**EPA
United States
Environmental Protection
Agency
Third External Review Draft of
Air Quality Criteria for
Particulate Matter (April, 2002)
Volume II
-------
EPA/600/P-99/002aC
April 2002
Third External Review Draft
Air Quality Criteria for Participate Matter
Volume II
National Center for Environmental Assessment-RTP Office
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
-------
1 DISCLAIMER
2
3 This document is an external review draft for review purposes only and does not constitute
4 U.S. Environmental Protection Agency policy. Mention of trade names or commercial products
5 does not constitute endorsement or recommendation for use.
6
7
8 PREFACE
9
10 National Ambient Air Quality Standards (NAAQS) are promulgated by the United States
11 Environmental Protection Agency (EPA) to meet requirements set forth in Sections 108 and 109
12 of the U.S. Clean Air Act (CAA). Sections 108 and 109 require the EPA Administrator (1) to list
13 widespread air pollutants that reasonably may be expected to endanger public health or welfare;
14 (2) to issue air quality criteria for them that assess the latest available scientific information on
15 nature and effects of ambient exposure to them; (3) to set "primary" NAAQS to protect human
16 health with adequate margin of safety and to set "secondary" NAAQS to protect against welfare
17 effects (e.g., effects on vegetation, ecosystems, visibility, climate, manmade materials, etc.); and
18 (5) to periodically (every 5 years) review and revise, as appropriate, the criteria and NAAQS for
19 a given listed pollutant or class of pollutants.
20 The original NAAQS for particulate matter (PM), issued in 1971 as "total suspended
21 parti culate" (TSP) standards, were revised in 1987 to focus on protecting against human health
22 effects associated with exposure to ambient PM less than 10 microns (< 10 //m) that are capable
23 of being deposited in thoracic (tracheobronchial and alveolar) portions of the lower respiratory
24 tract. Later periodic reevaluation of newly available scientific information, as presented in the
25 last previous version of this "Air Quality Criteria for Parti culate Matter" document published in
26 1996, provided key scientific bases for PM NAAQS decisions published in July 1997. More
27 specifically, the PM10 NAAQS set in 1987 (150 //g/m3, 24-h; 50 //g/m3, annual average) were
28 retained in modified form and new standards (65 //g/m3, 24-h; 15 //g/m3, annual average) for
29 particles <2.5 //m (PM25) were promulgated in July 1997.
April 2002 Il-ii DRAFT-DO NOT QUOTE OR CITE
-------
1 This Third External Review Draft of revised Air Quality Criteria for Particulate Matter
2 assesses new scientific information that has become available mainly between early 1996 through
3 December 2001. The present draft is being released for public comment and review by the Clean
4 Air Scientific Advisory Committee (CASAC) to obtain comments on the organization and
5 structure of the document, the issues addressed, the approaches employed in assessing and
6 interpreting the newly available information on PM exposures and effects, and the key findings
7 and conclusions arrived at as a consequence of this assessment. Public comments and CASAC
8 review recommendations will be taken into account in making any appropriate further revisions
9 to this document for incorporation into a final draft. Evaluations contained in the present
10 document will be drawn on to provide inputs to associated PM Staff Paper analyses prepared by
11 EPA's Office of Air Quality Planning and Standards (OAQPS) to pose alternatives for
12 consideration by the EPA Administrator with regard to proposal and, ultimately, promulgation of
13 decisions on potential retention or revision of the current PM NAAQS.
14 Preparation of this document was coordinated by staff of EPA's National Center for
15 Environmental Assessment in Research Triangle Park (NCEA-RTP). NCEA-RTP scientific
16 staff, together with experts from other EPA/ORD laboratories and academia, contributed to
17 writing of document chapters; and earlier drafts of this document were reviewed by experts from
18 federal and state government agencies, academia, industry, and NGO's for use by EPA in support
19 of decision making on potential public health and environmental risks of ambient PM. The
20 document describes the nature, sources, distribution, measurement, and concentrations of PM in
21 outdoor (ambient) and indoor environments. It also evaluates the latest data on human exposures
22 to ambient PM and consequent health effects in exposed human populations (to support decision
23 making regarding primary, health-related PM NAAQS). The document also evaluates ambient
24 PM environmental effects on vegetation and ecosystems, visibility, and man-made materials, as
25 well as atmospheric PM effects on climate change processes associated with alterations in
26 atmospheric transmission of solar radiation or its reflectance from the Earth's surface or
27 atmosphere (to support decision making on secondary PM NAAQS).
28 The NCEA of EPA acknowledges the contributions provided by authors, contributors, and
29 reviewers and the diligence of its staff and contractors in the preparation of this document.
April 2002 II-iii DRAFT-DO NOT QUOTE OR CITE
-------
Air Quality Criteria for Particulate Matter
VOLUME I
EXECUTIVE SUMMARY E-l
1. INTRODUCTION 1-1
2. PHYSICS, CHEMISTRY, AND MEASUREMENT OF
PARTICULATE MATTER 2-1
3. CONCENTRATIONS, SOURCES, AND EMISSIGNS OF
ATMOSPHERIC PARTICULATE MATTER 3-1
Appendix 3 A: Spatial and Temporal Variability of the Nationwide
AIRS PM25 and PM10.25 Data Sets 3A-1
Appendix 3B: Aerosol Composition Data from the Speciation
Network 3B-1
Appendix 3C: Organic Composition of Particulate Matter 3C-1
Appendix 3D: Composition of Particulate Matter Source Emissions .... 3D-1
4. ENVIRONMENTAL EFFECTS OF PARTICULATE MATTER 4-1
Appendix 4A: Colloquial and Latin Names 4A-1
5. HUMAN EXPOSURE TO PARTICULATE MATTER AND ITS
CONSTITUENTS 5-1
April 2002 Il-iv DRAFT-DO NOT QUOTE OR CITE
-------
Air Quality Criteria for Particulate Matter
(cont'd)
VOLUME II
6. DOSIMETRY OF PARTICULATE MATTER 6-1
7. TOXICOLOGY OF PARTICULATE MATTER IN HUMANS AND
LABORATORY ANIMALS 7-1
8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM
AMBIENT PARTICULATE MATTER 8-1
Appendix 8A: Short-Term PM Exposure-Mortality Studies:
Summary Table 8A-1
Appendix 8B: Particulate Matter-Morbidity Studies:
Summary Tables 8B-1
9. INTEGRATIVE SYNTHESIS 9-1
Appendix 9A: Key Quantitative Estimates of Relative Risk for
Particulate Matter-Related Health Effects Based on
Epidemiologic Studies of U.S. and Canadian Cities
Assessed in the 1996 Particulate Matter Air Quality
Criteria Document 9A-1
April 2002 II-v DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
Page
List of Tables II-xiv
List of Figures II-xx
Authors, Contributors, and Reviewers D-xxvii
U.S. Environmental Protection Agency Project Team for Development of
Air Quality Criteria for Particulate Matter II-xxxvi
Abbreviations and Acronyms D-xxxix
6. DOSIMETRY OF PARTICULATE MATTER 6-1
6.1 INTRODUCTION 6-1
6.1.1 Size Characterization of Inhaled Particles 6-3
6.1.2 Structure of the Respiratory Tract 6-4
6.2 PARTICLE DEPOSITION 6-6
6.2.1 Mechanisms of Deposition 6-6
6.2.2 Deposition Patterns in the Human Respiratory Tract 6-8
6.2.2.1 Total Respiratory Tract Deposition 6-8
6.2.2.2 Deposition in the Extrathoracic Region 6-12
6.2.2.3 Deposition in the Tracheobronchial and Alveolar Regions 6-16
6.2.2.4 Local Distribution of Deposition 6-16
6.2.2.5 Deposition of Specific Size Modes of Ambient Aerosol 6-20
6.2.3 Biological Factors Modulating Deposition 6-23
6.2.3.1 Gender 6-23
6.2.3.2 Age 6-25
6.2.3.3 Respiratory Tract Disease 6-28
6.2.3.4 Anatomical Variability 6-30
6.2.4 Interspecies Patterns of Deposition 6-32
6.3 PARTICLE CLEARANCE AND TRANSLOCATION 6-39
6.3.1 Mechanisms and Pathways of Clearance 6-39
6.3.1.1 Extrathoracic Region 6-41
6.3.1.2 Tracheobronchial Region 6-42
6.3.1.3 Alveolar Region 6-42
6.3.2 Clearance Kinetics 6-44
6.3.2.1 Extrathoracic Region 6-44
6.3.2.2 Tracheobronchial Region 6-44
6.3.2.3 Alveolar Region 6-46
6.3.3 Interspecies Patterns of Clearance 6-50
6.3.4 Factors Modulating Clearance 6-51
6.3.4.1 Age 6-51
6.3.4.2 Gender 6-51
6.3.4.3 Physical Activity 6-51
6.3.4.4 Respiratory Tract Disease 6-52
6.4 PARTICLE OVERLOAD 6-53
April 2002 H-vi DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
6.5 COMPARISON OF DEPOSITION AND CLEARANCE PATTERNS
OF PARTICLES ADMINISTERED BY INHALATION AND
INTRATRACHEAL INSTILLATION 6-55
6.6 MODELING THE DISPOSITION OF PARTICLES IN THE
RESPIRATORY TRACT 6-57
6.6.1 Modeling Deposition, Clearance, and Retention 6-57
6.6.2 Models To Estimate Retained Dose 6-64
6.6.3 Fluid Dynamics Models for Deposition Calculations 6-67
6.7 SUMMARY AND CONCLUSIONS 6-73
REFERENCES 6-76
7. TOXICOLOGY OF PARTICIPATE MATTER IN HUMANS AND LABORATORY
ANIMALS 7-1
7.1 INTRODUCTION 7-1
.2 RESPIRATORY EFFECTS OF PARTICIPATE MATTER IN HEALTHY
HUMANS AND LABORATORY ANIMALS: IN VIVO EXPOSURES 7-2
7.2.1 Ambient Combustion-Related and Surrogate Particulate Matter 7-7
7.2.1.1 Ambient Particulate Matter 7-15
7.2.1.2 Diesel Particulate Matter 7-18
7.2.1.3 Complex Combustion-Related Particles 7-22
7.2.2 Acid Aerosols 7-25
7.2.3 Metal Particles, Fumes, and Smoke 7-27
7.2.4 Ambient Bioaerosols 7-32
7.3 CARDIOVASCULAR AND SYSTEMIC EFFECTS OF PARTICULATE
MATTER IN HUMANS AND LABORATORY ANIMALS: IN VIVO
EXPOSURES 7-34
7.4 SUSCEPTIBILITY TO THE EFFECTS OF PARTICULATE MATTER
EXPOSURE 7-45
7.4.1 Pulmonary Effects of Particulate Matter in Compromised Hosts 7-45
7.4.2 Genetic Susceptibility to Inhaled Particles and their Constituents 7-50
7.4.3 Effect of Particulate Matter on Allergic Hosts 7-52
7.4.4 Resistance to Infectious Disease 7-59
7.5 PARTICULATE MATTER TOXICITY AND PATHOPHYSIOLOGY:
IN VITRO EXPOSURES 7-60
7.5.1 Introduction 7-60
7.5.2 Experimental Exposure Data 7-61
7.5.2.1 Ambient Particles 7-62
7.5.2.2 Comparison of Ambient and Combustion-Related Surrogate
Particles 7-72
7.5.3 Potential Cellular and Molecular Mechanisms 7-75
7.5.3.1 Reactive Oxygen Species 7-75
7.5.3.2 Intracellular Signaling Mechanisms 7-80
April 2002 H-vii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
7.5.3.3 Other Potential Cellular and Molecular Mechanisms 7-84
7.5.4 Specific Particle Size and Surface Area Effects 7-86
7.5.5 Pathophysiological Mechanisms for the Effects of Low Concentrations
of Paniculate Air Pollution 7-90
7.5.5.1 Direct Pulmonary Effects 7-91
7.5.5.2 Systemic Effects Secondary to Lung Injury 7-92
7.5.5.3 Direct Effects on the Heart 7-94
7.6 RESPONSES TO PARTICULATE MATTER AND GASEOUS
POLLUTANT MIXTURES 7-95
7.7 SUMMARY 7-105
7.7.1 Biological Plausibility 7-105
7.7.1.1 Link Between Specific Particulate Matter Components
and Health Effects 7-105
7.7.1.2 Susceptibility 7-110
7.7.2 Mechanisms of Action 7-110
REFERENCES 7-111
EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS
FROM AMBIENT PARTICULATE MATTER 8-1
8.1 INTRODUCTION 8-1
8.1.1 Types of Epidemiology Studies Reviewed 8-2
8.1.2 Confounding and Effect Modification 8-4
8.1.3 Selection of Studies for Review and Ambient PM Increments Used
to Report Risk Estimates 8-10
8.2 MORTALITY EFFECTS OF PARTICULATE MATTER EXPOSURE 8-12
8.2.1 Introduction 8-12
8.2.2 Mortality Effects of Short-Term Particulate Matter Exposure 8-13
8.2.2.1 Summary of 1996 Particulate Matter Criteria Document
Findings and Key Issues 8-13
8.2.2.2 Introduction to Newly Available Information on Short-Term
Mortality Effects 8-17
8.2.2.3 New Multi-City Studies 8-25
8.2.2.4 The Role of Particulate Matter Components 8-38
8.2.2.5 New Assessments of Cause-Specific Mortality 8-61
8.2.2.6 Salient Points Derived from Summarization of Studies of
Short-Term Particulate Matter Exposure Effects on Mortality . . . 8-66
8.2.3 Mortality Effects of Long-Term Exposure to Ambient Particulate
Matter 8-69
8.2.3.1 Studies Published Prior to the 1996 Particulate Matter
Criteria Document 8-69
April 2002 H-viii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
8.2.3.2 Prospective Cohort Analyses of Chronic Particulate Matter
Exposure Mortality Effects Published Since the 1996
Particulate Matter Air Quality Criteria Document 8-73
8.2.3.3 Studies by Particulate Matter Size-Fraction and Composition ... 8-96
8.2.3.4 Population-Based Mortality Studies in Children 8-101
8.2.3.5 Salient Points Derived from Analyses of Chronic Particulate
Matter Exposure Mortality Effects 8-105
.3 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE 8-108
8.3.1 Cardiovascular Effects Associated with Acute Ambient Particulate
Matter Exposure 8-108
8.3.1.1 Introduction 8-108
8.3.1.2 Summary of Key Findings on Cardiovascular Morbidity from
the 1996 Particulate Matter Air quality Criteria Document 8-109
8.3.1.3 New Particulate Matter-Cardiovascular Morbidity Studies 8-110
8.3.1.4 Issues in the Interpretation of Acute Cardiovascular Effects
Studies 8-132
8.3.2 Effects of Short-Term Particulate Matter Exposure on the Incidence of
Respiratory Hospital Admissions and Medical Visits 8-134
8.3.2.1 Introduction 8-134
8.3.2.2 Summary of Key Respiratory Hospital Admissions
Findings from the 1996 Particulate Matter Air Quality
Criteria Document 8-135
8.3.2.3 New Respiratory-Related Hospital Admissions Studies 8-135
8.3.2.4 Key New Respiratory Medical Visits Studies 8-145
8.3.2.5 Identification of Potential Susceptible Subpopulations 8-147
8.3.2.6 Summary of Key Findings on Acute Particulate Matter
Exposure and Respiratory-Related Hospital Admissions
and Medical Visits 8-150
8.3.3 Effects of Particulate Matter Exposure on Lung Function and
Respiratory Symptoms 8-154
8.3.3.1 Effects of Short-Term Particulate Matter Exposure on
Lung Function and Respiratory Symptoms 8-155
8.3.3.2 Long-Term Particulate Matter Exposure Effects on Lung
Function and Respiratory Symptoms 8-166
.4 DISCUSSION OF EPIDEMIOLOGIC STUDIES ON HEALTH EFFECTS
OF AMBIENT PARTICULATE MATTER 8-173
8.4.1 Introduction 8-173
8.4.2 Assessment of Confounding by Co-Pollutants 8-177
8.4.2.1 Introduction 8-177
8.4.2.2 Issues 8-181
April 2002 H-ix DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
8.4.2.2 Assessments of Confounding Using Multi-Pollutant Models
with Observed Gases 8-197
8.4.2.3 Assessment of Confounding in Multi-City Studies:
Pooling Effects 8-201
8.4.2.4 Assessment of Confounding in Multi-City Studies:
Regression 8-203
8.4.2.5 Assessment of Confounding Based on Exposure 8-207
8.4.2.6 Assessment of Confounding by Factor Analysis 8-217
8.4.2.7 Simulation Analysis of Confounding 8-218
8.4.2.8 Discussion 8-219
8.4.3 Role of Particulate Matter Components 8-220
8.4.3.1 Fine- and Coarse-Particle Effects on Mortality 8-220
8.4.3.2 PM10, PM2 5 (Fine), and PM10_2 5 (Coarse) Particulate Matter
Effects on Morbidity 8-231
8.4.4 The Question of Lags 8-237
8.4.5 New Assessments of Mortality Displacement 8-243
8.4.6 Concentration-Response Relationships for Ambient PM 8-245
8.4.7 New Assessments of Consequences of Measurement Error 8-250
8.4.7.1 Theoretical Framework for Assessment of Measurement
Error 8-250
8.4.7.2 Spatial Measurement Error Issues That May Affect the
Interpretation of Multi-Pollutant Models with Gaseous
Co-Pollutants 8-256
8.4.7.3 Measurement Error and the Assessment of Confounding by
Co-Pollutants in Multi-Pollutant Models 8-263
8.4.7.5 Air Pollution Exposure Proxies in Long-Term Mortality
Studies 8-264
8.4.9 Heterogeneity of Particulate Matter Effects Estimates 8-270
8.4.9.1 Evaluation of Heterogeneity of Particulate Matter Mortality
Effect Estimates 8-271
8.4.9.2 Comparison of Spatial Relationships in the NMMAPS and
Cohort Reanalyses Studies 8-275
8.4.9.3 Epidemiologic Studies of Ambient Air Pollution
Interventions 8-277
8.5 KEY FINDINGS AND CONCLUSIONS DERIVED FROM
PARTICULATE MATTER EPIDEMIOLOGY STUDIES 8-283
REFERENCES 8-289
APPENDIX 8A: Short-Term PM Exposure-Mortality Studies: Summary Table .... 8A-1
APPENDIX 8B: Particulate Matter-Morbidity Studies: Summary Tables 8B-1
April 2002 H-x DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
9. INTEGRATIVE SYNTHESIS 9-1
9.1 INTRODUCTION 9-1
9.2 BACKGROUND 9-3
9.2.1 Basic Concepts 9-3
9.2.2 Particle Size Distributions 9-3
9.2.3 Definitions of Particle Size Fractions 9-4
9.3 CHARACTERIZATION OF EMISSION SOURCES 9-10
9.4 AMBIENT CONCENTRATIONS 9-16
9.4.1 Measurement of Particulate Matter 9-16
9.4.2 Mass Concentrations 9-17
9.4.3 Physical and Chemical Properties of Ambient PM 9-18
9.5 AIR QUALITY MODEL DEVELOPMENT AND TESTING 9-23
9.6 EXPOSURE TO PARTICULATE MATTER AND COPOLLUTANTS 9-27
9.6.1 Central Site to Outdoor 9-28
9.6.1.1 Exposure for Acute Epidemiology 9-28
9.6.1.2 Exposure for Chronic Epidemiology 9-28
9.6.2 Home Outdoor Concentrations Versus Ambient Concentrations
Indoors and the Ambient Contribution to Total Personal Exposure 9-30
9.6.2.1 Mass Balance Model 9-30
9.6.2.2 Separation of Total Personal Exposure into its Ambient
and Nonambient Components 9-31
9.6.3 Variability in the Relationship Between Concentrations and Personal
Exposures 9-35
9.6.4 Exposure Relations for Co-Pollutants 9-35
9.6.5 Summary 9-41
9.7 EXPOSURE TO BIOLOGICALLY IMPORTANT CHARACTERISTICS
OF PARTICULATE MATTER 9-41
9.7.1 Exposure Relationships for Susceptible Subpopulations 9-42
9.7.2 Toxicologically Important Components of PM 9-42
9.7.3 Exposure-Measurement Techniques 9-43
9.7.4 Comprehensive Studies to Determine Population Exposure 9-44
9.7.5 Air Pollutants Generated Indoors 9-48
9.8 DOSIMETRY: DEPOSITION AND FATE OF PARTICLES IN THE
RESPIRATORY TRACT 9-50
9.8.1 Particle Deposition in the Respiratory Tract 9-50
9.8.2 Particle Clearance and Translocation 9-55
9.8.3 Deposition and Clearance Patterns of Particles Administered by
Inhalation Versus Intratracheal Instillation 9-57
9.8.4 Inhaled Particles as Potential Carriers of Toxic Agents 9-57
9.8.5 Summary of Particle Dosimetry 9-58
April 2002 H-xi DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
9.9 ASSESSMENT OF PARTICULATE-MATTER PROPERTIES LINKED
TO HEALTH EFFECTS 9-60
9.9.1 Introduction 9-60
9.9.2 Specific Properties of Ambient PM Linked to Health Effects 9-62
9.9.2.1 Physical Properties 9-62
9.9.2.2 Chemical Properties 9-64
9.9.2.3 Summary 9-67
9.9.3 Chemical Components and Source Categories Associated with Health
Effects in Epidemiologic Studies 9-67
9.9.3.1 Individual Chemical Species 9-68
9.9.3.2 Source Category Factors 9-68
9.10 SUSCEPTIBLE SUBPOPULATIONS 9-70
9.10.1 Introduction 9-71
9.10.2 Preexisting Disease as a Risk Factor for Particulate Matter Health
Effects 9-71
9.10.2.3 Ambient PM Exacerbation of Respiratory Disease
Conditions 9-73
9.10.2.4 Ambient PM Exacerbation of Cardiovascular Disease
Conditions 9-74
9.10.3 Age-Related At-Risk Population Groups: The Elderly and Children 9-77
9.11 MECHANISMS OF INJURY 9-80
9.12 HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER
OBSERVED IN POPULATION STUDIES 9-85
9.12.1 Introduction 9-85
9.12.2 Community-Health Epidemiologic Evidence for Ambient Particulate
Matter Effects 9-87
9.12.2.1 Short-Term Particulate Matter Exposure Effects on
Mortality 9-90
9.12.2.2 Relationships of Ambient Particulate Matter Concentrations
to Morbidity Outcomes 9-116
9.12.2.3 Methodological Issues 9-129
9.12.3 Coherence of Reported Epidemiologic Findings 9-136
9.13 EVALUATION OF STATISTICAL AND MEASUREMENT
ERROR ISSUES 9-138
9.13.1 Errors Related to Concentration, Exposure, and Dose 9-138
9.13.1.1 Opportunities for Error in the Use of Ambient PM
Concentration as a Surrogate for PM Dose in Epidemiologic
Studies 9-139
9.13.2 Possible Errors Related to Health and Epidemiology 9-146
9.13.3 Apportioning Health Effects to PM (by size, chemical component,
or source category) and Gaseous Co-Pollutants 9-148
April 2002 U-xii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
9.14 IMPLICATIONS OF HEALTH EFFECTS OF LONG-TERM
EXPOSURES TO PARTICIPATE MATTER 9-153
9.14.1 Methodological Issues 9-153
9.14.2 Overall Survival and Life Expectancy 9-154
9.14.3 Verification and Sensitivity Analyses 9-155
9.14.4 Impact on Life-Expectancy 9-155
9.14.5 Specific Causes of Death 9-156
REFERENCES 9-157
APPENDIX 9A: Key Quantitative Estimates of Relative Risk for Particulate
Matter-Related Health Effects Based on Epidemiologic Studies
of U.S. and Canadian Cities Assessed in the 1996 Parti culate
Matter Air Quality Criteria Document 9A-1
REFERENCES 9A-6
April 2002 H-xiii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
Number Page
6-1 Overview of Respiratory Tract Particle Clearance and Translocation
Mechanisms 6-40
7-1 Types of Particulate Matter Used in Toxicological Studies 7-4
7-2 Respiratory Effects of Ambient Parti culate Matter 7-8
7-3 Respiratory Effects of Complex Combustion-Related Parti culate Matter 7-9
7-4 Respiratory Effects of Surrogate Parti culate Matter 7-14
7-5 Respiratory Effects of Acid Aerosols in Humans and Laboratory Animals 7-26
7-6 Respiratory Effects of Metal Particles, Fumes, and Smoke in Humans and
Laboratory Animals 7-29
7-7 Respiratory Effects of Ambient Bioaerosols 7-33
7-8 Cardiovascular and Systemic Effects of Ambient and Combustion-Related
Paniculate Matter 7-35
7-9 Physicochemical Properties of Parti culate Matter 7-61
7-10 In Vitro Effects of Parti culate Matter and Parti culate Matter Constituents 7-63
7-11 Numbers and Surface Areas of Monodisperse Particles of Unit Density of
Different Sizes at a Mass Concentration of 10 //g/m3 7-87
7-12 Respiratory and Cardiovascular Effects of Mixtures 7-97
8-1 Recent U.S. and Canadian Time-Series Studies of PM-Related Daily Mortality ... 8-18
8-2 Synopsis of Short-Term Mortality Studies That Examined Relative Importance
of PM25 and PM10.25 8-40
8-3 Excess Total Mortality Risks Estimated to be Associated with Various
Ambient Particle Size-Related Indices 8-50
8-4 Summary of Parti culate Matter Chemical Components Analyzed in
Recent Studies 8-52
April 2002 II-xiv DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number
8-5 Summary of Source-Oriented Evaluations of Particulate Matter Components
in Recent Studies 8-58
8-6 Comparison of Six Cities and American Cancer Society Study Findings
from Original Investigators and Health Effects Institute Reanalysis 8-75
8-7 Summary of Results from the Extended ACS Study 8-79
8-8 Relative Risk of Mortality from All Nonexternal Causes, by Sex and Air
Pollutant, for an Alternative Covariate Model in the ASHMOG Study 8-87
8-9 Relative Risk of Mortality from Cardiopulmonary Causes, by Sex and Air
Pollutant, for an Alternative Covariate Model 8-88
8-10 Relative Risk of Mortality from Lung Cancer by Air Pollutant and by Gender
for an Alternative Covariate Model 8-88
8-11 Comparison of Excess Relative Risks for Three Particle Metrics in the Male
Subset of the AHSMOG Study 8-90
8-12 Comparison of Excess Relative Risks of Long-Term Mortality in the Harvard
Six Cities, ACS, AHSMOG, and VA Studies 8-94
8-13 Comparison of Estimated Relative Risks for All-Cause Mortality in Six
U.S. Cities Associated with the Reported Inter-City Range of Concentrations
of Various Parti culate Matter Metrics 8-97
8-14 Comparison of Reported SO4= and PM25 Relative Risks for Various Mortality
Causes in the American Cancer Society Study 8-97
8-15 Comparison of Total Mortality Relative Risk Estimates and T-statistics for
Particulate Matter Components in Three Prospective Cohort Studies 8-98
8-16 Comparison of Cardiopulmonary Mortality Relative Risk Estimates and
T-statistics for Particulate Matter Components in Three Prospective Cohort
Studies 8-99
8-17 Summary of Studies of PM10 or PM25 and Total CVD Hospital Visits 8-111
8-18 Percent Increase in Hospital Admissions per lO-^g/m3 Increase in PM10 in
14 U.S. Cities 8-137
April 2002 II-xv DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number
8-19 Summary of United States PM10 Respiratory Hospital Admission Studies 8-151
8-20 Summary of United States PM25 Respiratory Hospital Admission Studies 8-152
8-21 Summary of United States PM10_25 Respiratory Hospital Admission Studies .... 8-152
8-22 Summary of United States PM10, PM25, and PM10_25 Asthma Medical Visit
Studies 8-153
8-23 Summary of Asthma PM10 PFT Studies 8-157
8-24 Summary of PM25 PFT Asthma Studies 8-158
8-25 Summary of Asthma PM10 Cough Studies 8-162
8-26 Summary of Asthma PM10 Phlegm Studies 8-163
8-27 Summary of Asthma PM10 Lower Respiratory Illness Studies 8-163
8-28 Summary of Asthma PM10 Bronchodilator Use Studies 8-164
8-29 Summary of Asthma PM25 Respiratory Symptom Studies 8-166
8-30 Summary of Non-Asthma PM10 PFT Studies 8-167
8-31 Summary of Non-Asthma PM10 Respiratory Symptom Studies 8-168
8-32 Summary of Non-Asthma PM2 5 Respiratory Outcome Studies 8-169
8-33 Summary of Non-Asthma Coarse Fraction Studies of Respiratory Endpoints .... 8-170
8-34 Characterization of Co-Pollutant Effects on the Stability and Variance
Inflation or Deflation of PM Effect Size Estimate (in terms of excess RR) 8-193
8-35 Some New Daily Time Series Studies for Mortality or Morbidity with
Co-Pollutant Models and Gravimetric PM Indices 8-199
8-36 Single-Day Lags Used in Co-Pollutant Models in Lippmann et al., 2000,
Tables 13-14 8-201
April 2002 II-xvi DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
8-37 Number of Participants, N, in Each Block for the Exposure Study in
Sarnat et al. (2000) 8-209
8-38 Correlations Among Ambient Pollutants in Baltimore 8-212
8-39 Summary of Past Ecologic and Case-Control Epidemiologic Studies of
Outdoor Air and Lung Cancer 8-227
8-40 Comparison of PM10 Effect Sizes Estimated by NMMAPS Analyses for 0, 1,
and 2 Day Lags for the 20 Largest U.S. Cities 8-239
8-41 Maximum, Median, and Minimum 90th Percentile of Absolute Values of
Differences Between Fine Particle Concentrations at Pairs of Monitoring
Sites in 27 Metropolitan Areas in Order of Decreasing Maximum Difference . . . 8-259
8-42 Summary of Within-City Heterogeneity by Region 8-260
8-43 Summary of ACS Pollution Indices: Units, Primary Sources, Number of
Cities and Subjects Available for Analysis, and the Mean Levels 8-269
8A-1 Short-Term Particulate Matter Exposure Mortality Effects Studies 8A-2
8B-1 Acute Parti culate Matter Exposure and Cardiovascular Hospital Admissions 8B-3
8B-2 Acute Particulate Matter Exposure and Respiratory Hospital Admissions
Studies 8B-18
8B-3 Acute Particulate Matter Exposure and Respiratory Hospital Admissions
Studies 8B-40
8B-4 Short-Term Particulate Matter Exposure Effects on Pulmonary Function
Tests in Studies of Asthmatics 8B-52
8B-5 Short-Term Particulate Matter Exposure Effects on Symptoms in Studies
of Asthmatics 8B-57
8B-6 Short-Term Particulate Matter Exposure Effects on Pulmonary Function
Tests in Studies of Nonasthmatics 8B-62
April 2002 II-xvii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number
8B-7 Short-Term Particulate Matter Exposure Effects on Symptoms in Studies
ofNonasthmatics 8B-69
8B-8 Long-Term Particulate Matter Exposure Respiratory Health Indicators:
Respiratory Symptom, Lung Function 8B-74
9-1 Constituents of Atmospheric Particles and Their Major Sources 9-11
9-2 Emissions of Primary PM25 by Various Sources in 1999 9-14
9-3 Emissions of Precursors to Secondary PM25 Formation by Various Sources
in 1999 9-15
9-4 Comparison of Ambient Particles, Fine-Mode (nuclei mode plus accumulation
mode) and Coarse-Mode 9-21
9-5 Concentrations of PM2 5, PM10_25 and Selected Elements in the PM2 5 and
PM10.25 Size Range ' 9-22
9-6 Qualitative Estimates of Exposure Variables 9-38
9-7 Particulate Matter Characteristics Potentially Relevant to Health 9-43
9-8 Volume Mean Diameter of Indoor Particle Sources 9-49
9-9 Concentration Differences Between Constituents of Nonambient
(Indoor-Generated) and Ambient PM 9-50
9-10 Chemical Species Associated with Mortality in Epidemiologic Studies 9-68
9-11 Source Categories Associated with Mortality in Epidemiologic Studies 9-69
9-12 Incidence of Selected Cardiorespiratory Disorders by Age and by Geographic
Region, 1996 9-72
9-13 Number of Acute Respiratory Conditions per 100 persons per Year, by Age:
United States, 1996 9-78
9-14 Effect Estimates per Variable Increments in 24-Hour Concentrations of Fine
Particle Indicators (PM25, SOJ, H+) from U.S. and Canadian Studies 9-96
April 2002 II-xviii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number
9-15 Effect Estimates Per Variable Increments in 24-Hour Concentrations of
Coarse-Fraction Particles (PM10_25) from U.S. and Canadian Studies 9-102
9-16 Summary of Source-Oriented Evaluations of Particulate Matter Components
in Recent Studies 9-108
9-17 Effect Estimates per Increments in Long-Term Mean Levels of Fine and
Inhalable Particle Indicators from U.S. and Canadian Studies 9-112
9-18 Percent Increase in Hospital Admissions per 10-//g/m3 Increase in 24-Hour
PM10 in 14 U.S. Cities 9-120
9-19 Percent Increase in Mortality per 10 //g/m3 PM10 in Seven U.S. Regions 9-133
9-20 Percent Excess Risk (t-statistic) per 10 //g/m3 Increase in PM for the
Relationship of Various Indicators of PM with Various Types of Mortality
(CV = cardiovascular) in Several Different Locations 9-143
9-21 Examples of How % Excess Risk per 10 //g/m3 Increase in PM Indicator
Increases for Specific Chemical Components of PM 9-143
9-22 Percent Excess Risk (t-statistic) per Interquartile Increase in PM Indicator
for the Relationship of Various Indicators of PM with Cardiovascular
Mortality for Phoenix 9-143
9A-1 Effect Estimates per 50-//g/m3 Increase in 24-hour PM10 Concentrations from
U.S. and Canadian Studies 9A-2
9A-2 Effect Estimates per Variable Increments in 24-hour Concentrations of
Fine Particle Indicators (PM25, SOJ, H+) from U.S. and Canadian Studies 9A-4
9A-3 Effect Estimates per Increments in Annual Mean Levels of Fine Particle
Indicators From U.S. and Canadian Studies 9A-5
April 2002 H-xix DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
Number Page
6-1 Diagrammatic representation of respiratory tract regions in humans 6-5
6-2 Total respiratory tract deposition (as percentage deposition of amount inhaled)
in humans as a function of particle size 6-9
6-3 Total deposition fraction as a function of particle size in 22 healthy men and
women under six different breathing patterns 6-11
6-4 Extrathoracic deposition (as percentage deposition of the amount inhaled)
in humans as a function of particle size 6-13
6-5 Tracheobronchial deposition (as percentage deposition of the amount inhaled)
in humans as a function of particle size 6-17
6-6 Alveolar deposition (as percentage deposition of the amount inhaled) in humans
as a function of particle size 6-17
6-7 Lung deposition fractions in the tracheobronchial (TB) and alveolar (A) regions
obtained by the bolus technique 6-18
6-8 Lung deposition fractions in ten volumetric regions for particle sizes ranging
from ultrafine particle diameter (dp) of 0.04 to 0.01 //m (Panel A) to fine
(dp =1.0 //m) (Panel B) and coarse (dp = 3 and 5 //m) (Panels C and D) 6-21
6-9 Regional deposition fraction in laboratory animals as a function of particle size . . 6-34
6-10 Particle deposition efficiency in rats and humans as a function of particle size
for the (A) total respiratory tract, (B) thoracic region, (C) tracheobronchial
region, and (D) alveolar region 6-36
6-11 Major clearance pathways for particles deposited in the extrathoracic region
and tracheobronchial tree 6-40
6-12 Diagram of known and suspected clearance pathways for poorly soluble
particles depositing in the alveolar region 6-41
8-la Strong within-city association between PM and mortality, but no second-stage
association 8-7
8-lb Within-city association between PM and mortality ranges from negative to
positive with mean across cities approximately zero, but with strong positive
second-stage association 8-7
April 2002 H-xx DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number
8-2 (a) Graphical depiction of confounding; (b) Graphical depiction of effect
modification; (c) Graphical depiction of a causal agent with a secondary
confounder; (d) Graphical depiction of a causal agent and two potential
confounders
8-3 Estimated excess risks for PM mortality (1 day lag) for the 90 largest U.S.
cities as shown in the original NMAPS report 8-27
8-4 Map of the United States showing the 90 cities (the 20 cities are circled) and
the seven regions considered in the NMMAPS geographic analyses 8-28
8-5 Percent excess mortality risk (lagged 0, 1, or 2 days) estimated in the NMMAPS
90-City Study to be associated with 10-//g/m3 increases in PM10 concentrations
in cities aggregated within U.S. regions shown in Figure 8-2 8-29
8-6 Percent excess risks estimated per 25 //g/m3 increase in PM2 5 or PM10_2 5 from
new studies evaluating both PM2 5 and PM10_2 5 data for multiple years, based
on single pollutant (PM only) models 8-43
8-7 Excess risks estimated for sulfate per 5 //g/m3 increase from the studies in
which both PM2 5 and PM10_2 5 data were available 8-56
8-8 Natural logarithm of relative risk for total and cause-specific mortality per
10 //g/m3 PM25 (approximately the excess relative risk as a fraction), with
smoothed concentration-response functions 8-80
8-9 Relative risk of total and cause-specific mortality at 10 //g/m3 PM25 (mean
of 1979-1983) of alternative statistical models 8-81
8-10 Relative risk of total and cause-specific mortality for particle metrics and
gaseous pollutants over different averaging periods 8-82
8-11 Univariate relation between percentage of homes with central AC and
regression coefficients for (A) CVD, for cities nonwinter peaking PM10
concentrations (solid line) and winter peaking PM10 concentrations (dashed
line) and (B) univariate relation between percentage of PM10 from highway
vehicles and regression coefficients for CVD 8-115
8-12 Acute cardiovascular hospitalizations and particulate matter exposure excess
risk estimates derived from selected U.S. PM10 studies 8-124
April 2002 H-xxi DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number
8-13 Maximum excess risk of respiratory-related hospital admissions and visits per
50-//g/m3 PM10 increment in selected studies of U.S. cities 8-154
8-14 Selected acute pulmonary function change studies of asthmatic children 8-160
8-15 Odds ratios with 95% confidence interval for cough per 50-//g/m3 increase in
PM10 for selected asthmatic children studies at lag 0 8-165
8-16 Graphical depiction of actual confounding of the effects of ambient A and
ambient B 8-182
8-17 Graphical depiction of under-fitting of A and B 8-183
8-18 Only A is causal, B is not related to the outcome, but both regressors are
included in the model, a likely cause of variance inflation 8-184
8-19 Graphical depiction of over-fitting of A and B 8-184
8-20 Graphical depiction of mis-fitting of the effects of A and B 8-184
8-21 Effects of PM10 on total mortality in 20 large U.S. cities, as a function of
co-pollutant models 8-188
8-22 Effects of particles and gases on total mortality in eight Canadian cities 8-189
8-23 Effects of PM10 or PM25 on circulatory mortality in three U.S. cities as a
function of lag days 8-190
8-24 Total mortality from particles and gases in Santa Clara County, CA 8-191
8-25 Cause-specific fine or coarse particle mortality in Detroit, MI 8-191
8-26 Effects of fine particles on total mortality in Mexico City 8-192
8-27 Concentration of PM10 and NO2 versus distance 8-216
8-28 Marginal posterior distribution for effects of PM10 on all cause mortality at
lag 0, 1, and 2 for the 90 cities 8-238
April 2002 H-xxii DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number
8-29 Particulate matter <10 //m in aerodynamic diameter (PM10)-total mortality
dose-response curve for the mean lag PM10 and 95% credible regions (solid
lines), 20 largest U.S. cities, 1987-1994 8-247
8-30 The EPA-derived plot showing relationship of PM10 total mortality effects
estimates and 95% confidence intervals for all cities in the Samet et al.
(2000a,b) NMMAPS 90-cities analyses in relation to study size (i.e., the
natural logarithm or numbers of deaths times days of PM observations) 8-273
8-31 The EPA-derived plots showing relationships of PM10-mortality (total,
nonaccidental) effects estimates and 95% confidence intervals to study size
(defined as Figure 8-10) for cities broken out by regions as per the NMMAPS
regional analyses of Samet et al. (2000a,b) 8-274
9-1 A general framework for integrating particulate-matter research 9-2
9-2 Particle size distributions: (a) number of particles as a function of particle
diameter: number concentrations are shown on a logarithmic scale to display
the wide range by site and size and (b) particle volume as a function of particle
diameter: for the averaged urban and freeway-influenced urban number
distributions shown in Figure 2-1 of Chapter 2 9-5
9-3 Volume size distribution, measured in traffic, showing fine-mode and
coarse-mode particles and the nuclei and accumulation modes within the
fine-particle mode 9-6
9-4 Specified particle penetration (size-cut curves) through an ideal
(no-particle-loss) inlet for five different size-selective sampling criteria 9-8
9-5 An idealized distribution of ambient particulate matter showing fine-mode
particles and coarse-mode particles and the fractions collected by size-selective
samplers 9-9
9-6 Philadelphia, PA-NJ MSA 9-19
9-7 Occurrence of differences between pairs of sites in three MS As 9-20
9-8 Major chemical components of PM2 5 as determined in the pilot study for
EPA's national speciation network 9-23
April 2002 H-xxiii DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number
9-9 Main components of a comprehensive atmospheric chemistry modeling
system, such as Models 3 9-24
9-10 Correlograms showing the variation in site-to-site correlation coefficient for
PM25 as a function of distance between sites for several cities 9-29
9-11 Regression analysis of daytime total personal exposures to PM10 versus
ambient PM10 concentrations using data from the PTEAM study 9-32
9-12 Comparison of correlation coefficients for longitudinal analyses of personal
exposure for individual subjects versus ambient concentrations of PM25 and
sulfate 9-33
9-13 Regression analysis of daytime exposures to the ambient component of personal
exposure to PM10 (ambient exposure) versus ambient PM10 concentrations 9-34
9-14 Regression analysis of daytime exposures to the nonambient component of
personal exposure to PM10 (nonambient exposure) versus ambient PM10
concentrations 9-34
9-15 Distribution of individual, daily values of the infiltration factor, F^ =
C(AI)/C and the attenuation factor, a = A/C, estimated using data from the
PTEAM study 9-36
9-16 Percentage of homes with air conditioning versus the regression coefficient
for the relationship of cardiovascular-related hospital emissions to ambient
PM10 concentrations 9-37
9-17 Spatial variation of PM25, PM10, and PM10_25 as shown by site-to-site correlation
coefficients as a function of distance between sites for summer 1992 and 1993 in
Philadelphia, PA 9-45
9-18 Comparison of site-to-site correlation coefficients for PM25 and PM10_2 5 for
several cities 9-45
9-19 Site-to-site correlation coefficients for PM2 5 mass and some chemical
components of PM25 in 1994 in Philadelphia, PA 9-46
9-20 Site-to-site correlation coefficients for PM2 5 mass and several source category
factors in 1986 in the South Coast Basin (Los Angeles area) 9-46
April 2002 II-xxiv DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number
9-21 Values of geometric mean infiltration factor, F^ = A/C, as a function of
particle diameter for hourly nighttime data (assuming no indoor sources)
for summer and fall seasons 9-47
9-22 Values of penetration efficiency and deposition rate as a function of particle
diameter estimated from model of average nighttime indoor-outdoor
concentration data 9-48
9-23 Inhalation rates on a per body-weight basis for males (•) and females (±)
by age (Layton, 1993) 9-79
9-24 Schematic representation of potential pathophysiological pathways and
mechanisms by which ambient PM may increase risk of cardiovascular
morbidity and/or mortality 9-81
9-25 Percent excess risks estimated per 25-//g/m3 increase in PM2 5 or PM10_2 5
from new studies evaluating both PM2 5 and PM10_25 data for multiple years 9-105
9-26 Relative risks estimated per 5-//g/m3 increase in sulfate from U.S. and Canadian
studies in which both PM2 5 and PM10_25 data were available 9-107
9-27 Acute cardiovascular hospitalizations and PM exposure excess risk estimates
derived from selected U.S. PM10 studies 9-118
9-28 Maximum excess risk in selected studies of U.S. cities relating PM10 estimate
of exposure (50 //g/m3) to respiratory-related hospital admissions and visits .... 9-124
9-29 Selected acute pulmonary function change studies of asthmatic children 9-127
9-30 Odds ratios for cough for a 50-//g/m3 increase in PM10 for selected asthmatic
children studies, with lag 0 with 95% CI 9-127
9-31 Marginal posterior distributions for effect of PM10 on total mortality at lag 1,
with and without control for other pollutants, for the 90 cities 9-131
9-32 An expanded version of the Risk Assessment Framework: (a) PM sources
to PM exposure, (b) PM exposure to PM dose 9-140
9-33 Schematic showing major nonvolatile and semivolatile components of PM2 5 ... 9-141
April 2002 II-xxv DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number
9-34 The percent deposition of inhaled particles in the tracheobronchial and
alveolar regions of the lung as a function of particle size 9-147
9-35 Diagram showing relationships (correlations) between A and B and between
various concentration, exposure, and outcome measures 9-148
9-36 Diagram showing concentrations—exposure—outcome relationships
(correlations for CO or NO2, PM2 5, and source category factors for
vehicular traffic related PM and regional sulfate) 9-152
April 2002 II-xxvi DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
CHAPTER 6. DOSIMETRY OF PARTICULATE MATTER
Principal Authors
Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Ramesh Sarangapani—IFC Consulting, Research Triangle Park, NC 27711
Dr. Richard Schlesinger—New York University School of Medicine, Department of
Environmental Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. James McGrath—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. James Raub—National Center for Environmental Assessment (MD-52), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. Dan Costa—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Robert Devlin—National Health and Environmental Effects Research Laboratory
(MD58),U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Kevin Dreher—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Andrew Ohio—National Health and Environmental Effects Research Laboratory (MD-58D),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Judith Graham—National Exposure Research Laboratory (MD-75),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Chong Kim—National Health and Environmental Effects Research Laboratory (MD-58B),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Hillel Koren—National Health and Environmental Effects Research Laboratory (MD-58A),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
April 2002 II-xxvii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
Dr. Ted Martonen—National Health and Environmental Effects Research Laboratory (MD-74),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Jim Samet—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Ravi Subramaniam—National Center for Environmental Assessment (8623D),
U.S. Environmental Protection Agency, Washington, DC 20460
Dr. William Watkinson—National Health and Environmental Effects Research Laboratory
(MD-82), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Bennett—University of North Carolina at Chapel Hill, Campus Box 7310,
Chapel Hill, NC 37599
Dr. Mark Frampton—University of Rochester, 601 Elmwood Avenue, Box 692, Rochester, NY
14642
Dr. John Godleski—421 ConantRoad, Weston, MA 02493
Dr. Gunter Oberdorster—University of Rochester, Department of Environmental Medicine,
Rochester, NY 14642
Dr. Kent Pinkerton—University of California, ITEH, One Shields Avenue, Davis, CA 95616
Dr. Peter J.A. Rombout—National Institute of Public Health and Environmental Hygiene,
Department of Inhalation Toxicology, P.O. Box 1, NL-3720 BA Bilthoven, The Netherlands
Dr. Vanessa Vu—Office of Research and Development, U.S. Environmental Protection Agency
(8601), Waterside Mall, 401 M St. S.W., Washington, DC 20460
April 2002 II-xxviii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
CHAPTER 7. TOXICOLOGY OF PARTICULATE MATTER IN HUMANS AND
LABORA TOR Y ANIMALS
Principal Authors
Dr. Lung Chi Chen—New York University School of Medicine, Nelson Institute of
Environmental Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Terry Gordon—New York University Medical Center, Department of Environmental
Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. James McGrath—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Christine Nadziejko—Department of Environmental Medicine, New York University School
of Medicine, Tuxedo, NY
Mr. James Raub—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. Susanne Becker—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Dan Costa—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Robert Devlin—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Kevin Dreher—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Janice Dye—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
April 2002 II-xxix DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
Dr. Andrew Ohio—National Health and Environmental Effects Research Laboratory (MD-58D),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Ian Gilmour—National Health and Environmental Effects Research Laboratory (MD-82),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Tony Huang—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Judith Graham—National Exposure Research Laboratory (MD-75),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Chong Kim—National Health and Environmental Effects Research Laboratory (MD-58B),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Urmila Kodavanti—National Health and Environmental Effects Research Laboratory
(MD-82), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Hillel Koren—National Health and Environmental Effects Research Laboratory (MD-58A),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Michael Madden—National Health and Environmental Effects Research Laboratory
(MD-58B), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Ted Martonen—National Health and Environmental Effects Research Laboratory (MD-74),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Jim Samet—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Watkinson—National Health and Environmental Effects Research Laboratory
(MD-82), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Bennett—University of North Carolina at Chapel Hill, Campus Box 7310,
Chapel Hill, NC 37599
Dr. MarkFrampton—University of Rochester, 601 Elmwood Avenue, Box 692, Rochester, NY
14642
April 2002 II-xxx DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
Dr. John Godleski—421 ConantRoad, Weston, MA 02493
Dr. Gunter Oberdorster—University of Rochester, Department of Environmental Medicine,
Rochester, NY 14642
Dr. Kent Pinkerton—University of California, ITEH, One Shields Avenue, Davis, CA 95616
Dr. Peter J.A. Rombout—National Institute of Public Health and Environmental Hygiene,
Department of Inhalation Toxicology, P.O. Box 1, NL-3720 BA Bilthoven, The Netherlands
CHAPTER 8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM
AMBIENT PARTICULA TE MA TTER
Principal Authors
Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Vic Hasselblad—29 Autumn Woods Drive, Durham, NC 27713
Dr. Kazuhiko Ito—New York University Medical Center, Institute of Environmental Medicine,
Long Meadow Road, Tuxedo, NY 10987
Dr. Patrick Kinney, Columbia University, 60 Haven Avenue, B-l, Room 119,
New York, NY 10032
Dr. Dennis J. Kotchmar—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Allan Marcus—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. George Thurston—New York University Medical Center, Institute of Environmental
Medicine, Long Meadow Road, Tuxedo, NY 10987
April 2002 II-xxxi DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
Dr. Burt Brunekreef—Agricultural University, Environmental and Occupational Health,
P.O. Box 238, NL 6700 AE, Wageningen, The Netherlands
Dr. Richard Burnett—Health Canada, 200 Environmental Health Centre, Tunney's Pasture,
Ottawa, Canada KlA OL2
Dr. Raymond Carroll—Texas A & M University, Department of Statistics, College Station, TX
77843-3143
Dr. Robert Chapman—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Steven Colome—Integrated Environmental Services, 5319 University Drive, #430,
Irvine, CA 92612
Dr. Ralph Delfmo—University of California at Irvine, Epidemiology Division, Department of
Medicine, University of California at Irvine, Irvine, CA 92717
Dr. Douglas Dockery—Harvard School of Public Health, 665 Huntington Avenue, 1-1414,
Boston, MA 02115
Dr. Peter Guttorp—University of Washington, Department of Statistics, Box 354322
Seattle, WA 98195
Dr. Scott R. Kegler—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Fred Lipfert—23 Carll Court, Northport, NY 11768
Dr. Lee-Jane Sally Liu—University of Washington, Department of Environmental Health,
Box 357234, Seattle, WA 98195
Dr. Suresh Moolgavakar—Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue,
N-MP 665, Seattle, WA 98109
Dr. Robert D. Morris—Tufts University, 136 Harrison Avenue, Boston, MA 02111
Dr. Lucas Neas—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
April 2002 II-xxxii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
Dr. James Robins—Harvard School of Public Health, Department of Epidemiology,
Boston, MA 02115
Dr. Isabelle Romieu—Centers for Disease Control (CDC), 4770 Bufford Hwy, NE,
Atlanta, GA 30341
Dr. Mary Ross—Office of Air Quality Planning and Standards (MD-15),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lianne Sheppard—University of Washington, Box 357232, Seattle, WA 98195-7232
Dr. Richard L. Smith—University of North Carolina, Department of Statistics, Box 3260
Chapel Hill, NC 27599
Dr. Leonard Stefanski—North Carolina State University, Department of Statistics, Box 8203,
Raleigh, NC 27695
Dr. Duncan Thomas—University of Southern California, Preventative Medicine Department,
1540 Alcazar Street, CH-220, Los Angeles, CA 90033-9987
Dr. Clarice Weinberg—National Institute of Environmental Health Sciences, P.O. Box 12233,
Research Triangle Park, NC 27709
Dr. William Wilson—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
CHAPTER 9. INTEGRA TIVE SYNTHESIS: PARTICULA TE MATTER
ATMOSPHERIC SCIENCE, AIR QUALITY, HUMAN EXPOSURE,
DOSIMETRY, AND HEALTH RISKS
Principal Authors
Dr. William E. Wilson—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
April 2002 II-xxxiii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Principal Authors
(cont'd)
Dr. Dennis J. Kotchmar—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Allan Marcus—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Joseph P. Pinto—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. James Raub—National Center for Environmental Assessment (MD-52), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. Robert Chapman—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. William Ewald—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
April 2002 H-xxxiv DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
Executive Director
Dr. Lester D. Grant—Director, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Scientific Staff
Dr. William E. Wilson—Air Quality Coordinator, Physical Scientist, National Center for
Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle
Park,NC 27711
Dr. Lawrence J. Folinsbee—Health Coordinator, Chief, Environmental Media Assessment
Group, National Center for Environmental Assessment (MD-52), U.S. Environmental Protection
Agency, Research Triangle Park, NC 27711 (now deceased)
Dr. Dennis J. Kotchmar—Project Manager, Medical Officer, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Dr. Robert Chapman—Medical Officer, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Beverly Comfort—Health Scientist, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. William Ewald—Health Scientist, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. J.H.B. Garner—Ecological Scientist, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. David Mage—Physical Scientist, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Allan Marcus—Statistician, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. James McGrath—Visiting Senior Health Scientist, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
April 2002 H-xxxv DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Scientific Staff
(cont'd)
Dr. Joseph P. Pinto—Physical Scientist, National Center for Environmental Assessment,
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. James A. Raub—Health Scientist, National Center for Environmental Assessment (MD-52),
U. S. Environmental Protection Agency, Research Triangle Park, NC 27711
Technical Support Staff
Mr. Randy Brady—Deputy Director, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. Douglas B. Fennell—Technical Information Specialist, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Ms. Emily R. Lee—Management Analyst, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Diane H. Ray—Program Specialist, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Eleanor Speh—Office Manager, Environmental Media Assessment Branch, National Center
for Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research
Triangle Park, NC 27711
Ms. Donna Wicker—Administrative Officer, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. Richard Wilson—Clerk, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
April 2002 U-xxxvi DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Document Production Staff
Dr. Carol A. Seagle—Technical Editor, Computer Sciences Corporation,
2803 Slater Road, Suite 220, Morrisville, NC 27560
Ms. Diane G. Caudill—Graphic Artist, Computer Sciences Corporation,
2803 Slater Road, Suite 220, Morrisville, NC 27560
Ms. Carolyn T. Perry—Word Processor, Computer Sciences Corporation,
2803 Slater Road, Suite 220, Morrisville, NC 27560
Ms. Kelly Quifiones—Word Processor, InfoPro, Inc., 8405 Colesville Road, 2nd Floor,
Silver Spring, MD 20910
Technical Reference Staff
Mr. John A. Bennett—Technical Information Specialist, SANAD Support Technologies, Inc.,
11820 Parklawn Drive, Suite 400, Rockville, MD 20852
Ms. Sandra L. Hughey—Technical Information Specialist, SANAD Support Technologies, Inc.,
11820 Parklawn Drive, Suite 400, Rockville, MD 20852
Ms. Beth Olen—Records Management Technician, Reference Retrieval and Database Entry
Clerk, InfoPro, Inc., 8405 Colesville Road, 2nd Floor, Silver Spring, MD 20910
April 2002 U-xxxvii DRAFT-DO NOT QUOTE OR CITE
-------
Abbreviations and Acronyms
oabs light-absorption coefficient
oag light-absorption coefficient of gases
oap light-absorption coefficient of particles
°~ext
light-extinction coefficient
og geometric standard deviation
"scat
light-scattering coefficient
osg light-scattering coefficient of gases
osp light-scattering coefficient of particles
4-POBN
A
AAS
ACGffl
a-(4-pyridyl- 1 -oxide)-N-tert-butylnitrone
alveolar
atomic absorption spectrophotometry
American Conference of Governmental Industrial Hygienists
AD
ADS
AES
AIRS
AM
AQCD
AQI
AQRV
ARIES
ASOS
ATOM
ATOFMS
annular denuder system
atomic emission spectroscopy
Aerometric Information Retrieval System
alveolar macrophages
Air Quality Criteria Document
Air Quality Index
Air Quality Related Values
Aerosol Research and Inhalation Epidemiology Study
Automated Surface Observing System
aerosol and toxic deposition model
time-of-flight mass spectrometer
b
Ba
absorption coefficient
April 2002
D-xxxviii
DRAFT-DO NOT QUOTE OR CITE
-------
BAD
brachial artery diameter
BAL
bronchoalveolar lavage
BALF
bronchoalveolar lavage fluid
BAUS
brachial artery ultrasonography
BC
black carbon (see also CB)
BW
bronchial wash
BYU
Bringham Young University
C
apparent contrast
Ca+
calcium
CAA
Clean Air Act
CAAM
continuous ambient mass monitor
CAMNET
CAPs
concentrated ambient particles
CARS
California Air Resources Board
CASAC
Clean Air Scientific Advisory Committee
CASTNet
Clean Air Status and Trends Network
CAT
computer-aided tomography
CB
carbon black
base cation
CC
carbonate carbon
CC14
carbon tetrachloride
CCPM
continuous coarse particle monitor
CCSEM
computer-controlled scanning electron microscopy
CEN
European Standardization Committee
CF
Cystic Fibrosis
CFA
coal fly ash
CFCs
chlorofluorocarbons
CFD
computational fluid dynamics
April 2002
II-xxxix
DRAFT-DO NOT QUOTE OR CITE
-------
CFR
CH2O
GIF
CL
CMAQ
CMB
CMD
CMP
CMSA
C0
CO
CO CD
COPD
CPC
CPZ
CR
CRP
Code of Federal Regulations
formaldehyde
charcoal-impregnated cellulose fiber
chemiluminescence
Community Multi-Scale Air Quality
chemical mass balance
count mean diameter
copper smelter dust
Consolidated Metropolitan Statistical Area
initial contrast
carbon monoxide
Air Quality Criteria Document for Carbon Monoxide
chronic obstructive pulmonary disease
condensation particle counter
capsazepine
concentration-response
Coordinated Research Program
CSIRO
CSMCS
CTM
CV
Carbonaceous Species Methods Comparison Study
chemistry-transport model
coefficient of variation
D5o
Da
DAQM
DCFH
DE
DE
DEF
Denver Air Quality Model
dichlorofluorescin
deposition efficiencies
diesel exhaust
Deferoxamine
April 2002
n-xi
DRAFT-DO NOT QUOTE OR CITE
-------
DEP
diesel exhaust particles
DHR
dihydrorhodamine-123
DMS
dimethyl sulfide
DMTU
dimethylthiourea
DOFA
domestic oil fly ash
DPM
diesel particulate matter
DRG
dorsal root ganglia
dv
deciview index
BAD
electrical aerosol detector
EC
elemental carbon
ECAO
Environmental Criteria and Assessment Office
ECG
electrocardiogram
EDXRF
energy dispersive X-ray fluorescence
EGA
evolved gas analysis
EGF
epidermal growth factor
ELSIE
Elastic Light Scattering and Interactive Efficiency
ERK
extracellular receptor kinase
ESP
electrostatic precipitator
ESR
electron spin resonance
ET
extrathoacic
ETS
environmental tobacco smoke
EU
endotoxin units
EXPOLIS
flux
FEF
forced expiratory flow
FEVj
forced expiratory volume in 1 second
FID
flame ionization detection
FMD
flow-mediated dilation
April 2002
n-xii
DRAFT-DO NOT QUOTE OR CITE
-------
FPD
flame photometric detector
FRM
Federal Reference Method
gS02
gaseous sulfur dioxide
GC
gas chromatography
GCMs
General Circulation Models
GCVTC
Grand Canyon Visibility Transport Commission
GG/MSD
gas chromatography/mass-selective detection
GHG
greenhouse gases
GMCSF
granulocyte macrophage colony stimulating factor
GMPD
geometric mean particle diameter
GSD
geometric standard deviation (see also o )
GSH
glutathione
H2SO4
sulfuric acid
HAAQS
HDM
house dust mite
HDS
honeycomb denuder/filter pack sampler
HEADS
Harvard-EPA Annular Denuder Sampler
HEI
Health Effects Institute
hivol
High blume sampler
HNO3
nitric acid
HR
heart rate
HTGC-MS
high temperature gas chromotography-mass spectrometry
radiance
inhibitory kappa B alpha
apparent radiance of the background
transmitted radiance of the background
1C
ion chromatography
ICAM-1
intercellular adhesion molecule-1
April 2002
H-xlii
DRAFT-DO NOT QUOTE OR CITE
-------
ICP
inductively coupled plasma
ICRP
International Commission on Radiological Protection
le
equilibrium radiance or source function
IPS
Integrated Forest Study
IgE
immunoglobin E
IgG
immunoglobin G
IL
interleukin
IMPROVE
Interagency Monitoring of Protected Visual Environments
INAA
instrumental neutron activation analysis
IOVPS
integrated organic vapor/particle sampler
intraperitoneal
path radiance
IPCC
Intergovernmental Panel on Climate Change
IPM
inhalable paniculate matter
IPN
Inhalable Paniculate Network
ISO
International Standards Organization
transmitted radiance
INK
c-jun N-terminal kinase
'scp
light scattering by coarse particles
Jsfp
light scattering by fine particles
Jspd
light scattering coefficient of particles under dry conditions
'spw
light scattering coefficient of particles under humid conditions
K
Koschmieder constant
K+
potassium ion
KOH
potassium hydroxide
LAI
leaf area indices
LFA-1
leukocyte function-associated antigen-1
LN
lymph nodes
April 2002
H-xliii
DRAFT-DO NOT QUOTE OR CITE
-------
LoS
1pm, Lpm, L/min
LPS
LWCA
MAA
MAACS
MADPro
MAPK
MAQSIP
MCM
MCT
MEK
MIP
Mm
MMAD
MMD
MMPs
MOUDI
MPL
MPO
MS
MSA
MSAs
MSH
low sulfur
liters per minute
lipopolysaccharide
liquid water content analyzer
mineral acid anion
Metropolitan Acid Aerosol Characterization Study
Mountain Acid Deposition Program
mitogen-activated protein kinase
page 3-83
mass concentrations monitor
monocrotaline
mitogen-activated protein kinase
macrophage inflammatory protein
megameters
mean median aerodynamic diameter (see og)
mass median diameter
matrix metalloproteinases
micro-orifice uniform deposit impactor
multipath lung
myeloperoxidase
mass spectroscopy
methane sulfonic acid
metropolitan statistical areas
Mount St. Helens
MSP
NAC
NAL
NAMS
N-acetylcysteine (antioxidant)
nasal lavage fluid
National Ambient Monitoring Stations
April 2002
DRAFT-DO NOT QUOTE OR CITE
-------
NaN3
sodium azide
NAPAP
National Acid Precipitation Assessment Program
NAPRMN
NARSTO
NAST
National Assessment Synthesis Team
NCRPM
National Council on Radiation Protection and Measurements
ND
NIST diesel (also, not determined)
NDDN
National Dry Deposition Network
NDIR
nondispersive infrared spectrophotometry
NESCAUM
Northeast States for Coordinated Air Use Management
NF
nuclear factor
NF-KB
nuclear factor kappa B
NFRAQS
North Frontal Range Air Quality Study
NH3
ammonia
NIL
ammonium
(NH4)2 S04
ammonium sulfate
NILH,S(X
ammonium acid sulfate
NHBE
normal human bronchial epithelial
NIOSH
NIR
NIST
National Institute of Standards and Technology
NMD
nitroglycerine-mediated dilation
NMD
number mean diameter
NMRI
Naval Medical Research Institute
NO
nitrogen oxide
NO,
nitrogen dioxide
NO3-
nitrate
NOPL
naso-oro-pharyngo-laryngeal
April 2002
II-xlv
DRAFT-DO NOT QUOTE OR CITE
-------
NOX
NPP
NRC
NuCM
03
OAA
OAQPS
OAR
OC
OFA
CH-
ORD
OVA
P
P so42-
PAH
PAHs
PAN
PAR
PB
PEL
nitrogen oxides
net primary production
National Research Council
nutrient cycling model
ozone
Ottowa ambient air
Office of Air Quality Planning and Standards
Office of Air and Radiation
organic carbon
oil fly ask
hydroxyl ion
Office of Research and Development
ovalbumin
partial pressure
particulate sulfate
polynuclear aromatic hydrocarbon
polycyclic aromatic hydrocarbons
peroxyacetyl nitrate
photosynthetically active radiation
polymyxin-B
planetary boundary layer
PBY
PC
PC
PC-BOSS
PCA
PCBs
pyrolitic carbon
particle concentrator
Particulate Concentrator-Brigham Young University Organic
Sampling System
principal component analysis
polychloronated biphenyls
April 2002
DRAFT-DO NOT QUOTE OR CITE
-------
PCDD
polychlorinateddibenzo-^-dioxins
PCDF
polychlorinated dibenzofurans
PCM
particle composition monitor
pdf
probability density functions
PDGF
platelet-derived growth factor
PEM
Personal Environmental Monitor
PESA
proton elastic scattering analysis
PFA
PIXE
proton induced X-ray emission
PM
particulate matter
PM AQCD
PM Air Quality Criteria Document
PM(10.25)
coarse particulate matter
PM
-2.5
fine particulate matter
PMF
positive matrix factorization
PMN
polymorphonuclear leukocytes
equilibrium vapor pressure
poly I:C
polyionosinic-polycytidilic acid
POP
persistent organic pollutant
PROBDET
Probability of Detection Algorithm
PTEAMS
PTEP
PM10 Technical Enhancement Program
PTFE
polytetrafluoroethylene
PTFE
polytetrafluoroethylene
PUF
polyurethane foam
Q
respiratory flow rates
Qabs
efficiency of absorption
Qext
efficiency of extinction
Qscat
efficiency of scattering
April 2002
DRAFT-DO NOT QUOTE OR CITE
-------
aerodynamic resistance
RAAS
RADM
Regional Acid Deposition Model
RAMS
Real-Time Air Monitoring System
RAMS
Regional Air Monitoring Study
RAPS
Regional Air Pollution Study
boundary layer resistance
canopy resistance
REMSAD
Regulatory Modeling System for Aerosols and Deposition
RFC
residual fuels oils
RH
relative humidity
ROFA
residual oil fly ash
ROFA
residual oil fly ash
ROME
Reactive and Optics Model Emissions
ROS
reactive oxygen species
RPM
respirable particulate matter
RPM
Regional Particulate Model
RTE
rat tracheal epithelial
RTF
Research Triangle Park
SASS
sec
saturation ratio
SA
Sierra Anderson
SAD
small airway disease
SCAQS
Southern California Air Quality Study
scos
Southern California Ozone Study
sd
standard deviation
SEM
scanning electron microscopy
April 2002
DRAFT-DO NOT QUOTE OR CITE
-------
SES
sample equilibration system
SEV
Sensor Equivalent Visibility
SH
spontaneously hypertensive
SIP
State Implementation Plans
SIXE
synchrotron induced X-ray emission
SL
stochastic lung
SLAMS/NAMS
SLAMS
State and Local Air Monitoring Stations
SLE
St. Louis encephalitis
SMPS
scanning mobility particle sizer
SO,
sulfur dioxide
scx2-
sulfate
SOA
SOC
semivolatile organic compounds
SoCAB
South Coast Air Basin
SOD
superoxide dismutase
SOPM
secondary organic particulate matter
SP
Staff Paper
SPM
synthetic polymer monomers
SRI
SRM
standard reference method
SSM
solid sampler module
Stk
Stokes number
SUVB
solar ultraviolet B radiation
svoc
semivolatile organic compounds
SWMMC
Southwest Metropolitan Mexico City
T(CO)
core temperature
TB
tracheabronchial
April 2002
H-xlix
DRAFT-DO NOT QUOTE OR CITE
-------
TDF
total deposition fraction
TDMA
Tandem Differential Mobility Analyzer
TEOM
tapered element oscillating microbalance
TEOMs
TIMP
tissue inhibitor of metaloproteinase
TLN
TNF
tumor necrosis factor
TOFMS
aerosol time-of-flight mass spectroscopy
TOR
thermal/optical reflectance
TOT
thermal/optical transmission
TPM
thoracic paniculate matter
TRXRF
total reflection X-ray fluorescence
TSI
TSP
total suspended particulates
UAM-V
Urban Airshed Model Version V
UCM
unresolved complex mixture
ufCB
ultrafine carbon black
UFP
ultrafine fluorospheres
UNEP
United Nations Environment Programme
URG
University Research Glassware
USGCRP
U.S. Global Change Research Program
UVD
Utah Valley dust
VAPS
Versatile Air Pollution Samplers
VASM
Visibility Assessment Scoping Model
VBE
Japanese B encephalitis
VCAM-1
vascular cell adhesion molecule-1
deposition velocity
VDI
April 2002
II-l
DRAFT-DO NOT QUOTE OR CITE
-------
voc
vs
v,
v,
we
WEE
WINS
WIS
WKY
WMO
Wo
WRAC
X-XRF
XAD
XRF
volatile organic compounds
sedimentation velocity
turbulent diffusion velocity
tidal volume
tungsten carbide
western equine encephalitis
Well Impactor Ninety- Six
Wistar
Wi star-Kyoto
World Meteorological Organization
single scattering albedo
Wide Range Aerosol Classifier
synchrotron induced X-ray fluorescence
polystyrene-divinyl benzene
X-ray fluorescence
V*
April 2002
Il-li
DRAFT-DO NOT QUOTE OR CITE
-------
i 6. DOSIMETRY OF PARTICULATE MATTER
2
3
4 6.1 INTRODUCTION
5 A basic principle in health effects evaluation is that the dose delivered to the target site of
6 concern, rather than the external exposure, is the proximal cause of any biological response.
7 Characterization of the exposure-dose-response continuum for particulate matter (PM), a
8 fundamental objective of any dose-response assessment for evaluation of health effects, requires
9 the elucidation and understanding of the mechanistic determinants of inhaled particle dose.
10 Furthermore, dosimetric information is critical to an effective extrapolation to humans of health
11 effects demonstrated by toxicological studies using experimental animals and for comparison of
12 results from controlled clinical studies involving different types of human subjects, e.g., those
13 with preexisting respiratory disease and normals. Dosimetry provides a critical link in evaluating
14 the relevance of health effects found in animal models of susceptible humans because it allows
15 for discrimination between actual susceptibility differences from those due to differences in sites
16 of particle action.
17 Dose to target tissue is dependent initially on the deposition of particles within the
18 respiratory tract. Particle deposition refers to the removal of particles from their airborne state
19 because of their aerodynamic, thermodynamic, and/or electrostatic behavior. Once particles have
20 deposited onto the surfaces of the respiratory tract, they are subsequently subjected to either
21 absorptive or nonabsorptive particulate removal processes. This may result in their removal from
22 airway surfaces, as well as their removal, to varying degrees, from the respiratory tract itself.
23 The deposited PM thus cleared from initial deposition sites is said to have undergone
24 translocation. Clearance of deposited particles depends upon the initial site of deposition and
25 upon the physicochemical properties of the particles, both of which impact upon specific
26 translocation pathways. Retained particle burdens are determined by the dynamic relationship
27 between deposition and clearance rates.
28 This chapter is concerned with particle dosimetry, the study of the deposition, translocation,
29 clearance, and retention of particles within the respiratory tract and extrapulmonary tissues.
30 It summarizes basic concepts as presented in the 1996 EPA document, Air Quality Criteria for
April 2002 6-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 Particulate Matter or "PM AQCD" (U.S. Environmental Protection Agency, 1996), specifically
2 in Chapter 10; and it updates the state of the science based upon new literature appearing since
3 publication of the 1996 PM AQCD. Although our understanding of the basic mechanisms
4 governing deposition and clearance of inhaled particles has not changed, there has been
5 significant additional information on the role of certain biological determinants of the
6 deposition/clearance processes, such as gender and age. Also, the understanding of regional
7 dosimetry and the particle size range over which this has been evaluated has been expanded.
8 The dose from inhaled particles deposited and retained in the respiratory tract is governed
9 by a number of factors. These include exposure concentration and exposure duration, respiratory
10 tract anatomy and ventilatory parameters, and by physicochemical properties of the particles
11 themselves (e.g., particle size, hygroscopicity, solubility). The basic characteristics of particles
12 as they relate to deposition and retention, as well as anatomical and physiological factors
13 influencing particle deposition and retention, were discussed in depth in the 1996 PM AQCD.
14 Thus, in this current chapter, only an overview of basic information related to one critical factor
15 in deposition, namely particle size, is provided (Section 6.1.1), so as to allow the reader to
16 understand the different terms used in the remainder of this chapter and in subsequent ones
17 dealing with health effects. This is followed, in Section 6.1.2, by a basic overview of respiratory
18 tract structure as it relates to deposition evaluation. The ensuing major sections of this chapter
19 provide updated information on particle deposition, clearance, and retention in the respiratory
20 tract of humans, as well as laboratory animals, which are useful in the evaluation of PM health
21 effects. Issues related to the phenomenon of particle overload as it may apply to human exposure
22 and the use of instillation as an exposure technique to evaluate PM health effects also are
23 discussed. The final sections of the chapter deal with mathematical models of particle
24 disposition in the respiratory tract.
25 It must be emphasized that any dissection into discrete topics of factors that control dose
26 from inhaled particles tends to mask the dynamic and interdependent nature of the intact
27 respiratory system. For example, although deposition is discussed separately from clearance
28 mechanisms, retention (i.e., the actual amount of particles found in the respiratory tract at any
29 point in time) is, as noted previously, determined by the relative rates of both deposition and
30 clearance. Thus, assessment of overall dosimetry requires integration of these various
31 components of the overall process. In summarizing the literature on particle dosimetry, when
April 2002 6-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 applicable, changes from control are described if they were statistically significant at a
2 probability (p) value less than 0.05 (i.e., p < 0.05). When trends are described, an attempt will be
3 made to provide the actual p values given in the published reports.
4
5 6.1.1 Size Characterization of Inhaled Particles
6 Information about particle size distribution is important in the evaluation of effective
7 inhaled dose. This section summarizes particle attributes requiring characterization and provides
8 some general definitions important in understanding particle fate within the respiratory tract.
9 Particles exist in the atmosphere as components of aerosols, which are airborne suspensions
10 of finely dispersed solid or liquid particles. Because aerosols can consist of almost any material,
11 their description in simple geometric terms can be misleading unless important factors relating to
12 constituent particle size, shape, and density are considered. Although the size of particles within
13 aerosols can be described based on actual physical measurements (such as those obtained with a
14 microscope), in many cases it is better to use some equivalent diameter in place of the physical
15 diameter. The most commonly used metric is aerodynamic equivalent diameter (AED), whereby
16 particles of differing geometric size, shape, and density are compared in terms of aerodynamic
17 behavior (i.e., terminal setting velocity) to particles that are unit density (1 gm/cm3) spheres. The
18 aerodynamic behavior of unit density spherical particles constitutes a useful standard by which
19 many types of particles can be compared in terms of certain deposition mechanisms. (See
20 Chapter 2 for a more complete discussion.)
21 It is important to note that most aerosols present in natural and work environments are
22 polydisperse. This means that the constituent particles within an aerosol have a range of sizes
23 and are more appropriately described in terms of a size distribution parameter. The lognormal
24 distribution (i.e., the situation in which the logarithms of particle diameter are distributed
25 normally) can be used for describing size distributions of most aerosols. In linear form, the
26 logarithmic mean is the median of the distribution, and the metric of variability around this
27 central tendency is the geometric standard deviation (og). The og, a dimensionless term, is the
28 ratio of the 84th (or 16th) % particle size to the 50th % size. Thus, the only two parameters
29 needed to describe a log normal distribution of particle sizes for a specific aerosol are the median
30 diameter and the geometric standard deviation. However, the actual size distribution may be
31 obtained in various ways. For example, when a distribution is described by counting particles,
April 2002 6-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 the median is called the count median diameter (CMD). On the other hand, the median of a
2 distribution based on particle mass in an aerosol is the mass median diameter (MMD). When
3 using aerodynamic diameters, a term that is encountered frequently is mass median aerodynamic
4 diameter (MMAD), which refers to the median of the distribution of mass with respect to
5 aerodynamic equivalent diameter. Most of the present discussion will focus on MMAD because
6 it is the most commonly used measure of aerosol distribution. However, alternative distributions
7 should be used for particles with actual physical sizes below about 0.5 //m because, for these,
8 aerodynamic properties become less important. One such metric is thermodynamic-equivalent
9 size, which is the diameter of a spherical particle that has the same diffusion coefficient in air as
10 the particle of interest.
11
12 6.1.2 Structure of the Respiratory Tract
13 A detailed discussion of respiratory tract structure was provided in the 1996 PM AQCD
14 (U.S. Environmental Protection Agency, 1996), and only a brief synopsis is presented here.
15 For dosimetry purposes, the respiratory tract can be divided into three regions (Figure 6-1):
16 (1) extrathoracic (ET), (2) tracheobronchial (TB), and (3) alveolar (A). The ET region consists
17 of head airways (i.e., nasal and oral passages) through the larynx and represents the areas through
18 which inhaled air first passes. In humans, inhalation can occur through the nose or mouth (or
19 both, known as oronasal breathing). However, most laboratory animals commonly used in
20 respiratory toxicological studies are obligate nose breathers.
21 From the ET region, inspired air enters the TB region at the trachea. From the level of the
22 trachea, the conducting airways then undergo branching for a number of generations. The
23 terminal bronchiole is the most peripheral of the distal conducting airways and these lead,
24 in humans, to the respiratory bronchioles, alveolar ducts, alveolar sacs, and alveoli (all of which
25 comprise the A region). All of the conducting airways, except the trachea and portions of the
26 mainstem bronchi, are surrounded by parenchymal tissue. This is composed primarily of the
27 alveolated structures of the A region and associated blood and lymphatic vessels. It should be
28 noted that the respiratory tract regions are comprised of various cell types and that there are
29 distinct differences in the cells of airway surfaces in the ET, TB, and A regions. Although a
30 discussion of cellular structure of the respiratory tract is beyond the scope of this section, details
31 may be found in a number of sources (e.g., Crystal et al., 1997).
April 2002 6-4 DRAFT-DO NOT QUOTE OR CITE
-------
S
s
ol
4*
o
'S.
0 Co
- 0\
2
•_
w
*
-------
1 6.2 PARTICLE DEPOSITION
2 This section discusses the deposition of particles in the respiratory tract. It begins with an
3 overview of the basic physical mechanisms that govern deposition. This is followed by an
4 update on both total respiratory tract and regional deposition patterns in humans. Some critical
5 biological factors that may modulate deposition are then presented. The section ends with a
6 discussion of issues related to interspecies patterns of particle deposition.
7
8 6.2.1 Mechanisms of Deposition
9 Particles may deposit within the respiratory tract by five mechanisms: (1) inertial
10 impaction, (2) sedimentation, (3) diffusion, (4) electrostatic precipitation, and (5) interception.
11 Sudden changes in airstream direction and velocity cause particles to fail to follow the
12 streamlines of airflow. As a consequence, the particles contact, or impact, onto airway surfaces.
13 The ET and upper TB airways are characterized by high air velocities and sharp directional
14 changes and, thus, dominate as sites of inertial impaction. Impaction is a significant deposition
15 mechanism for particles larger than 1 //m AED.
16 All aerosol particles are continuously influenced by gravity, but particles with an
17 AED >1 //m are affected to the greatest extent. A particle will acquire a terminal settling
18 velocity when a balance is achieved between the acceleration of gravity acting on the particle and
19 the viscous resistance of the air, and it is this settling out of the airstream that takes it into contact
20 with airway surfaces. Both sedimentation and inertial impaction can influence the deposition of
21 particles within the same size range. These deposition processes act together in the ET and TB
22 regions, with inertial impaction dominating in the upper airways and gravitational settling
23 becoming increasingly dominant in the smaller conducting airways.
24 Particles having actual physical diameters <1 //m are subjected increasingly to diffusive
25 deposition because of random bombardment by air molecules, which results in contact with
26 airway surfaces. The root mean square displacement that a particle experiences in a unit of time
27 along a given cartesian coordinate is a measure of its diffusivity. The density of a particle is
28 unimportant in determining particle diffusivity. Thus, instead of having an aerodynamic
29 equivalent size, diffusive particles of different shapes can be related to the diffusivity of a
30 thermodynamic equivalent size based on spherical particles.
April 2002 6-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 The particle size region around 0.2 to 1.0 //m frequently is described as consisting of
2 particles that are small enough to be minimally influenced by impaction or sedimentation and
3 large enough to be minimally influenced by diffusion. Such particles are the most persistent in
4 inhaled air and undergo the lowest extent of deposition in the respiratory tract.
5 Interception is deposition by physical contact with airway surfaces. The interception
6 potential of any particle depends on its physical size, and fibers are the chief concern in relation
7 to the interception process. Their aerodynamic size is determined predominantly by their
8 diameter, but their length is the factor that influences probability of interception deposition.
9 Electrostatic precipitation is deposition related to particle charge. The minimum charge an
10 aerosol particle can have is zero. This condition rarely is achieved because of the random
11 charging of aerosol particles by air ions. Aerosol particles will acquire charges from these ions
12 by collisions with them because of their random thermal motion. Furthermore, many laboratory-
13 generated aerosols are charged. Such aerosols will generally lose their charge as they attract
14 oppositely charged ions, and an equilibrium state of these competing processes eventually is
15 achieved. This Boltzmann equilibrium represents the charge distribution of an aerosol in charge
16 equilibrium with bipolar ions. The minimum amount of charge is very small, with a statistical
17 probability that some particles within the aerosol will have no charge and others will have one or
18 more positive and negative charges.
19 The electrical charge on some particles will result in an enhanced deposition over what
20 would be expected from size alone. This results from image charges induced on the surface of
21 the airway by these particles or to space-charge effects, whereby repulsion of particles containing
22 like charges results in increased migration toward the airway wall. The effect of charge on
23 deposition is inversely proportional to particle size and airflow rate. This type of deposition is
24 often small compared to the effects of turbulence and other deposition mechanisms, and it
25 generally has been considered to be a minor contributor to overall particle deposition. However,
26 a study by Cohen et al. (1998), employing hollow airway casts of the human tracheobronchial
27 tree to assess deposition of ultrafine (0.02 //m) and fine (0.125 //m), particles found the
28 deposition of singly charged particles to be 5 to 6 times that of particles having no charge and
29 2 to 3 times that of particles at Boltzmann equilibrium. This suggests that electrostatic
30 precipitation may, in fact, be a significant deposition mechanism for ultrafine, and some fine,
31 particles within the TB region.
April 2002 6-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.2.2 Deposition Patterns in the Human Respiratory Tract
2 Knowledge of sites where particles of different sizes deposit in the respiratory tract and the
3 amount of deposition therein is necessary for understanding and interpreting the health effects
4 associated with exposure to particles. Particles deposited in the various respiratory tract regions
5 are subjected to large differences in clearance mechanisms and pathways and, consequently,
6 retention times. This section summarizes concepts of particle deposition in humans and
7 laboratory animals as reported in the 1996 PM AQCD (U.S. Environmental Protection Agency,
8 1996) and provides additional information based on studies published since that earlier
9 document.
10 Ambient air often contains particles too massive to be inhaled. The descriptor
11 "inhalability" is used to denote the overall spectrum of particle sizes that are potentially capable
12 of entering the respiratory tract. Inhalability is defined as the ratio of the number concentration
13 of particles of a certain aerodynamic diameter that are inspired through the nose or mouth to the
14 number concentration of the same diameter particle present in ambient air (International
15 Commission on Radiological Protection, 1994). In general, for humans, unit density particles
16 >100 (j,m diameter have a low probability of entering the mouth or nose in still air, but there is no
17 sharp cutoff to zero probability. Also, there is no lower limit to inhalability, so long as the
18 particle exceeds a critical size where the aggregation of atomic or molecular units is stable
19 enough to endow it with "particulate" properties, in contrast to those of free ions or gas
20 molecules.
21
22 6.2.2.1 Total Respiratory Tract Deposition
23 Total human respiratory tract deposition, as a function of particle size, is depicted in
24 Figure 6-2. These data were obtained by various investigators using different sizes of spherical
25 test particles in healthy male adults under different ventilation conditions; the large standard
26 deviations reflect interindividual and breathing pattern-related variability of deposition
27 efficiencies. Deposition in the ET region with nose breathing is generally higher than that with
28 mouth breathing because of the superior filtration capabilities of the nasal passages, resulting in
29 somewhat higher total deposition with mouth breathing for particles > l//m. For particles with
30 aerodynamic diameters greater than 1 //m, deposition is governed by impaction and
31 sedimentation, and it increases with increasing AED. When AED is >10 //m, almost all inhaled
April 2002 6-8 DRAFT-DO NOT QUOTE OR CITE
-------
100
90 -
80 -
70 -
60 -
50 -
o
8" 40 -
Q
30 -
20 -
10 -
c
o
0
Human (oral inhalation)
Human (nasal inhalation)
0.01
0.1 1.0
Particle Diameter
10
Figure 6-2. Total respiratory tract deposition (as percentage deposition of amount
inhaled) in humans as a function of particle size. All values are means with
standard deviations when available. Particle diameters are aerodynamic
(MMAD) for those >0.5
Source: Modified from Schlesinger (1989).
1 particles are deposited. As the particle size decreases from «0.5 //m, diffusional deposition
2 becomes dominant and total deposition depends more on the actual physical diameter of the
3 particle, with decreasing particle diameter leading to an increase in total deposition. Total
4 deposition shows a minimum for particle diameters in the range of 0.2 to 1.0 //m where, as noted
5 above, neither sedimentation, impaction, or diffusion deposition are very effective.
April 2002
6-9
DRAFT-DO NOT QUOTE OR CITE
-------
1 Besides particle size, breathing pattern is the most important factor affecting lung
2 deposition. Kim (2000) reported total lung deposition values in healthy adults for a wide range
3 of breathing patterns, tidal volumes (375 to 1500 mL), flow rates (150 to 1000 mL/s), and
4 respiratory times (2 to 12 s). Total lung deposition increased with increasing tidal volume at a
5 given flow rate and with increasing flow rate at a given respiratory time. Various deposition
6 values were correlated with a single composite parameter consisting of particle size, flow rate,
7 and tidal volume.
8 One of the specific size modes of the ambient aerosol that is being evaluated in terms of
9 potential toxicity is the ultrafme mode (i.e., particles having diameters <0.1 //m). There is,
10 however, little information on total respiratory tract deposition of such particles. Frampton et al.
11 (2000) exposed healthy adult human males and females, via mouthpiece, to 0.0267 //m diameter
12 carbon particles (at 10 //g/m3) for 2 h at rest. The inspired and expired particle number
13 concentration and size distributions were evaluated. Total respiratory tract deposition fraction
14 was determined for six particle size fractions, ranging from 0.0075 to 0.1334 //m. They found an
15 overall total lung deposition fraction of 0.66 (by particle number) or 0.58 (by particle mass),
16 indicating that exhaled mean particle diameter was slightly larger than inhaled diameter. There
17 was no gender difference. The deposition fraction decreased with increasing particle size within
18 the ultrafme range, from 0.76 at the smallest size to 0.47 at the largest.
19 Jaques and Kim (2000) measured total deposition fraction (TDF) of ultrafme particles
20 [number median diameter (NMD) = 0.04-0.1 //m and og = 1.3] in 22 healthy adults (men and
21 women in equal number) under a variety of breathing conditions. The study was designed to
22 obtain a rigorous data set for ultrafme particles that could be applied to health risk assessment.
23 TDF was measured for six different breathing patterns: tidal volume (Vt) of 500 mL at
24 respiratory flow rates (Q) of 150 and 250 mL/s; V, = 750 mL at Q of 250 and 375 mL/s; V, = 1 L
25 at Q of 250 and 500 mL/s. Aerosols were monitored continuously by a modified condensation
26 nuclei counter during mouthpiece inhalation with the prescribed breathing patterns. For a given
27 breathing pattern, TDF increased as particle size decreased, regardless of the breathing pattern
28 used. For example, at V, = 500 mL and Q = 250 mL/s, TDF was 0.26, 0.30, 0.35, and 0.44 for
29 NMD = 0.10, 0.08, 0.06, and 0.04 //m, respectively (see Figure 6-3). For a given particle size,
30 TDF increased with an increase in V, and a decrease in Q, indicating an importance of breathing
31 pattern in assessing respiratory dose. The study also found that TDF was greater for women than
April 2002 6-10 DRAFT-DO NOT QUOTE OR CITE
-------
\.u
o
03
^ OR
g
"H— ' n c;
w
o
°- HA
-------
1 A property of some ambient particulate species that affects deposition is hygroscopicity, the
2 propensity of a material for taking up and retaining moisture under certain conditions of humidity
3 and temperature. Such particles can increase in size in the humid air within the respiratory tract
4 and, when inhaled, will deposit according to their hydrated size rather than their initial size. The
5 implications of hygroscopic growth on deposition have been reviewed extensively by Morrow
6 (1986) and Hiller (1991); whereas the complications of studying lung deposition of hygroscopic
7 aerosols have been reviewed recently by Kim (2000). In general, compared to nonhygroscopic
8 particles of the same initial size, the deposition of hygroscopic aerosols in different regions of the
9 lung may be higher or lower, depending on the initial size. Thus, for particles with initial sizes
10 larger than «0.5 //m, the influence of hygroscopicity would be to increase total deposition with a
11 shift from peripheral to central or extrathoracic regions; whereas for smaller ones total deposition
12 would tend to be decreased.
13
14 6.2.2.2 Deposition in the Extrathoracic Region
15 The fraction of inhaled particles depositing in the ET region is quite variable, depending on
16 particle size, flow rate, breathing frequency and whether breathing is through the nose or the
17 mouth (Figure 6-4). Mouth breathing bypasses much of the filtration capabilities of the nasal
18 airways, leading to increased deposition in the lungs (TB and A regions). The ET region is
19 clearly the site of first contact with particles in the inhaled air and essentially acts as a "prefilter"
20 for the lungs.
21 Since release of the 1996 PM AQCD, a number of studies have explored ET deposition
22 with in vivo studies, as well as in both physical and mathematical model systems. In one study,
23 the relative distribution of particle deposition between the oral and nasal passages was assessed
24 during "inhalation" by use of a physical model (silicone rubber) of the human upper respiratory
25 system, extending from the nostrils and mouth through the main bronchi (Lennon et al., 1998).
26 Monodisperse particles ranging in size from 0.3 to 2.5 //m were used at various flow rates
27 ranging from 15 to 50 L/min. Total deposition in the model, as was regional deposition in the
28 oral passages, lower oropharynx-trachea, nasal passages, and nasopharynx-trachea, were
29 assessed. Deposition within the nasal passages was found to agree with available data obtained
30 from a human inhalation study (Heyder and Rudolf, 1977), being proportional to particle size,
31 density, and inspiratory flow rate. It also was found that for oral inhalation, the relative
April 2002 6-12 DRAFT-DO NOT QUOTE OR CITE
-------
100
90 -
80 -
70 -
^ 60 H
c
o
S 50 H
o
§• 40 H
Q
30 H
20 -
10 -
0
Human (oral inhalation)
Human (nasal inhalation)
1
'X T
A*
I I I I III
0.01
0.1 1.0
Particle Diameter (pm)
10
Figure 6-4. Extrathoracic deposition (as percentage deposition of the amount inhaled)
in humans as a function of particle size. All values are means with standard
deviations, when available. Particle diameters are aerodynamic (MMAD)
for those >0.5 (j,m and geometric (or diffusion equivalent) for those < 0.5 (j,m.
Source: Modified from Schlesinger (1989).
1 distribution between the oral cavity and the oropharynx-trachea was similar; whereas for nasal
2 inhalation, the nasal passages contained most of the particles deposited in the model, with only
3 about 10% depositing in the nasopharynx-trachea region. Furthermore, the deposition efficiency
4 of the nasopharynx-trachea region was greater than that of the oropharynx-trachea region.
5 For simulated oronasal breathing, deposition in the ET region depended primarily on particle size
6 rather than flow rate. For all flows and for all breathing modes, total deposition in the ET region
April 2002
6-13
DRAFT-DO NOT QUOTE OR CITE
-------
1 increased as particle diameter increased. Such information on deposition patterns in the ET
2 region is useful in refining empirical deposition models.
3 Deposition within the nasal passages was further evaluated by Kesavanathan and Swift
4 (1998), who examined the deposition of 1- to 10-//m particles in the nasal passages of normal
5 adults under an inhalation regime in which the particles were drawn through the nose and out
6 through the mouth at flow rates ranging from 15 to 35 L/min. At any particle size, deposition
7 increased with increasing flow rate; whereas at any flow rate, deposition increased with
8 increasing particle size. In addition, as was shown experimentally by Lennon et al. (1998) under
9 oronasal breathing conditions, deposition of 0.3- to 2.5-//m particles within the nasal passages
10 was significantly greater than within the oral passages, and nasal inhalation resulted in greater
11 total deposition in the model than did oral inhalation. These results are consistent with other
12 studies discussed in the 1996 PM AQCD and with the known dominance of impaction deposition
13 within the ET region.
14 Rasmussen et al. (2000) measured deposition in the nasal cavity of normal adult humans of
15 0.7 //m particles consisting of sodium chloride and radioactively-labeled DTPA. Inspiration
16 occurred under different levels of flow rate ranging from 10-30 L/min. They found that the
17 deposition fraction in the nasal passages increased as flow rate increased and that an estimate of
18 maximum linear air velocity was the best single predictor of nasal deposition fraction.
19 For ultrafine particles (dp < 0.1 //m), deposition in the ET region is controlled by diffusion,
20 which depends only on the particle's geometric diameter. Prior to 1996, ET deposition for this
21 particle size range had not been studied extensively in humans, and this remains the case. In the
22 earlier 1996 PM AQCD, the only data available for ET deposition of ultrafine particles were
23 from cast studies. More recently, deposition in the ET region was examined using mathematical
24 modeling. Three dimensional numerical simulations of flow and particle diffusion in the human
25 upper respiratory tract, which included the nasal region, oral region, larynx, and first two
26 generations of bronchi, were performed by Yu et al. (1998). Deposition of particles of 0.001 and
27 0.1 //m in these different regions was calculated under inspiratory and expiratory flow conditions.
28 Deposition efficiencies in the total model were lower on expiration than inspiration although
29 values for the former were quite high. Nasal deposition of ultrafine particles can also be quite
30 high. For example, nasal deposition accounted for up to 54% of total deposition in the model
31 system for 0.001-//m particles. The total deposition efficiency in the model was 75% (of the
April 2002 6-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 amount entering) for this size particle. With oral breathing, deposition efficiency was estimated
2 at 48% (of amount entering) (Yu et al., 1998).
3 Swift and Strong (1996) examined the deposition of ultrafine particles, ranging in size from
4 0.053 to 0.062 //m, in the nasal passages of normal adults during constant inspiratory flows of
5 6 to 22 L/min. The results are consistent with results noted in studies above, namely that the
6 nasal passages are highly efficient collectors for ultrafine particles. In this case, fractional
7 deposition ranged from 94 to 99% (of amount inhaled). There was found to be only a weak
8 dependence of deposition on flow rate, which contrasts with results noted above (i.e., Lennon
9 et al., 1998) for particles >0.3 //m, but is consistent with diffusion as the main deposition
10 mechanism.
11 Cheng et al. (1997) examined oral airway deposition in a replicate cast of the human nasal
12 cavity, oral cavity, and laryngeal-tracheal sections. Particle sizes ranged from 0.005 to 0.150 //m,
13 and constant inspiratory and expiratory flow rates of 7.5 to 30 L/min were used. They noted that
14 the deposition fractions within the oral cavity were essentially the same as that in the
15 laryngeal-tracheal sections for all particle sizes and flow rates. They ascribed this to the balance
16 between flow turbulence and residence time in these two regions. Svartengren et al. (1995)
17 examined the effect of changes in external resistance on oropharyngeal deposition of 3.6-//m
18 particles in asthmatics. Under controlled mouthpiece breathing conditions (flow rate 0.5 L/s), the
19 median deposition as a percentage of inhaled particles in the mouth and throat was 20%
20 (mean = 33%; range 12 to 84%). Although the mean deposition fell to 22% with added
21 resistance, the median value remained at 20% (range 13 to 47%). Fiberoptic examination of the
22 larynx revealed that there was a trend for increased mouth and throat deposition associated with
23 laryngeal narrowing. Katz et al. (1999) indicate, on the basis of mathematical model
24 calculations, that turbulence plays a key role in enhancing particle deposition in the larynx and
25 trachea.
26 The results of all of the above studies support the previously known ability of the ET
27 region, and especially the nasal passages, to act as an efficient filter for nanoparticles (<0.1 //m)
28 as well as for larger ones (>5//m), potentially reducing the amount of particles within a wide size
29 range that are available for deposition in the TB and A regions.
30
31
April 2002 6-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.2.2.3 Deposition in the Tracheobronchial and Alveolar Regions
2 Particles that do not deposit in the ET region of the respiratory tract enter the lungs;
3 however, their regional deposition within the lungs cannot be precisely measured. Much of the
4 available deposition data for the TB and A regions have been obtained from experiments with
5 radioactively labeled, poorly soluble particles (Figures 6-5 and 6-6, respectively). These have
6 been described previously (U.S. Environmental Protection Agency, 1996). Although there are no
7 new regional data obtained by means of the radioactive aerosol method since the publication of
8 that document, a novel serial bolus delivery method has been introduced. Using this bolus
9 technique, regional deposition has been measured for fine and coarse aerosols (Kim et al., 1996;
10 Kim and Hu, 1998) and for ultrafme aerosols (Kim and Jacques, 2000). The serial bolus method
11 uses nonradioactive aerosols and can measure regional deposition in a virtually unlimited number
12 of lung compartments. Because of experimental limitations of the technique, the investigators
13 measured regional lung deposition in ten serial, 50 mL increments from the mouth to the end of a
14 typical 500 mL tidal volume. Deposition measurements in the TB and A regions were obtained
15 for both men and women for particles ranging from 0.04 to 5.0 //m in diameter. It should be
16 noted that particle deposition in the TB and A regions was based on volumetric compartments of
17 50 to 150 mL and >150 mL, respectively. Deposition in the ET region was based on the 0 to
18 50 mL compartment. Lung deposition fractions in the TB and A regions obtained by the bolus
19 technique are shown in Figure 6-7. Of total particle deposition in the lung, 23 to 32% was
20 deposited in the TB region and 68 to 77% was deposited in the A region. Deposition in women
21 was consistently greater in the TB region by 21 to 47%, but was comparable or slightly smaller in
22 the A region when compared to men. As a result, total lung deposition was slightly greater in
23 women than men (~5 to 15%).
24
25 6.2.2.4 Local Distribution of Deposition
26 Airway structure and its associated air flow patterns are exceedingly complex, and
27 ventilation distribution of air in different parts of the lung is uneven. Thus, it is expected that
28 particle deposition patterns within the ET, TB, and A regions would be highly nonuniform, with
29 some sites exhibiting deposition that is much greater than average levels within these regions.
30 This was discussed in detail previously in the 1996 PM AQCD. Basically, using deposition data
31 from living subjects as well as from mathematical and physical models, enhanced deposition has
April 2002 6-16 DRAFT-DO NOT QUOTE OR CITE
-------
60
50 -
^ 40 H
o
:-E so
w
o
CL
Q 20 -
10 -
0.01
A Human (oral inhalation)
'\
0.1
1.0
10
Particle Diameter (|jm)
Figure 6-5. Tracheobronchial deposition (as percentage deposition of the amount
inhaled) in humans as a function of particle size. All values are means
with standard deviations, when available. Particle diameters are
aerodynamic (MMAD) for those >0.05 ^,m and geometric (or diffusion
equivalent) for those < 0.5 (j,m.
Source: Modified from Schlesinger (1989).
70
60 -
50 -
c 40 H
o 30 H
01
Q
20 -
0.01
Human (oral inhalation)
Human (nasal inhalation)
1
0.1 1.0
Particle Diameter (pm)
10
Figure 6-6. Alveolar deposition (as percentage deposition of the amount inhaled) in
humans as a function of particle size. All values are means with standard
deviations, when available. Particle diameters are aerodynamic (MMAD)
for those >0.05 ^m and geometric (or diffusion equivalent) for those
< 0.5
Source: Modified from Schlesinger (1989).
April 2002
6-17
DRAFT-DO NOT QUOTE OR CITE
-------
o
t3
ro
o
Q.
0>
Q
Male Vt = 500ml_
Female Q = 250 mL/s
Male
Female
Vt = 500m L
Q = 250 mL/s
0.04 0.06 0.08 0.10
Particle Diameter (|jm)
1 3 5
Particle Diameter (|jm)
Figure 6-7. Lung deposition fractions in the tracheobronchial (TB) and alveolar (A)
regions obtained by the bolus technique. Using a breathing pattern of 500 mL
at 15 breaths per min, TB deposition was 1.5,10.6, and 26.1% and
A deposition was 7.7, 39.4, and 39.8% for particles of 1, 3, and 5 ^m in
diameter, respectively, for men. In comparison to men, TB deposition in
women was 26 to 53% greater, whereas A deposition was comparable.
For ultrafine particles of 0.04 to 0.1 (j,m diameter, TB and A deposition ranged
from 5.7 to 15.6% and 18.2 to 33.1%, respectively. Both TB and A deposition
decreased with increasing particle size within the ultrafine range, which is
consistent with deposition theory.
Source: Kim and Hu (1998); Kim and Jaques (2000).
1
2
been shown to occur in the nasal passages and trachea and at branching points in the TB and
A regions (see Chapter 10 of U.S. Environmental Protection Agency, 1996). Churg and Vedal
April 2002
6-18
DRAFT-DO NOT QUOTE OR CITE
-------
1 (1996) examined retention of particles on carinal ridges and tubular sections of airways from
2 lungs obtained at necropsy. Results indicated significant enhancement of particle retention on
3 carinal ridges through the segmental bronchi; the ratios were similar in all airway generations
4 examined.
5 Kim and Fisher (1999) studied local deposition efficiencies and deposition patterns of
6 aerosol particles (2.9 to 6.7 //m) in sequential double bifurcation tube models with two different
7 branching geometries: one with in-plane (A) and another with out of plane (B) bifurcation. The
8 deposition efficiencies (DE) in each bifurcation increased with increasing Stokes number (Stk).
9 With symmetric flow conditions, DE was somewhat smaller in the second than the first
10 bifurcation in both models. DE was greater in the second bifurcation in model B than in model
11 A. With asymmetric flows, DE was greater in the low-flow side compared to the high-flow side;
12 and this was consistent in both models. Deposition pattern analysis showed highly localized
13 deposition on and in the immediate vicinity of each bifurcation ridge, regardless of branching and
14 flow patterns.
15 Comer et al. (2000) used a three-dimensional computer simulation technique to investigate
16 local deposition patterns in sequentially bifurcating airway models that were previously used in
17 experiments by Kim and Fisher (1999). The simulation was for 3-, 5-, and 7-//m particles and
18 assumed steady, laminar, constant air flow with symmetry about the first bifurcation. The overall
19 trend of the particle deposition efficiency, i.e., an exponential increase with Stokes number, was
20 similar for all bifurcations, and deposition efficiencies in the bifurcation regions agreed very well
21 with experimental data. Local deposition patterns consistently showed that the majority of the
22 deposition occurred within the carinal region.
23 Deposition "hot spots" at airway bifurcations have undergone additional analyses using
24 mathematical modeling techniques. Using calculated deposition sites, a strong correlation has
25 been demonstrated between secondary flow patterns and deposition sites and density both for
26 large (10 //m) particles and for ultrafme particles (0.01 //m) (Heistracher and Hofmann, 1997;
27 Hofmann et al., 1996). This supports experimental work, noted in U.S. Environmental
28 Protection Agency (1996), indicating that, like larger particles, ultrafme particles also show
29 enhanced deposition at airway branch points — even in the upper tracheobronchial tree.
30 The pattern of particle distribution on a more regional scale was evaluated by Kim et al.
31 (1996) and Kim and Hu (1998). Deposition patterns were measured in situ in nonsmoking
April 2002 6-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 healthy young adult males using, an aerosol bolus technique that delivered 1-, 3-, or 5-fj.m
2 particles into specific volumetric depths within the lungs. The distribution of particle deposition
3 was uneven; and it was noted that sites of peak deposition shifted from distal to proximal regions
4 of the lungs with increasing particle size (Figure 6-8). Furthermore, the surface dose was found
5 to be greater in the conducting airways than in the alveolar region for all of the particle sizes
6 evaluated. Within the conducting airways, the largest airway regions (i.e., 50 to 100 mL volume
7 distal to the larynx) received the greatest surface doses.
8 Kim and Jaques (2000) used the respiratory bolus technique to measure the deposition
9 distribution of ultrafme particles (0.04, 0.06, 0.08, and 0.1 //m) in young adults. Under normal
10 breathing conditions (tidal volume 500 mL, respiratory flow rate 250 mL/s), bolus aerosols were
11 delivered sequentially to a lung depth ranging from 50 to 500 mL in 50-mL increments. The
12 results indicate that regional deposition varies widely along the depth of the lung, regardless of
13 particle size (Figure 6-8). The deposition patterns for ultrafme particles, especially for very small
14 ultrafme particles, were similar to those for coarse particles. Peak deposition occurred in the
15 lung regions situated between 150 and 200 mL from the mouth, and sites of peak deposition
16 shifted proximally with a decrease in particle size. Deposition dose per unit average surface area
17 was greatest in the proximal lung regions and decreased rapidly with increased lung depth. Peak
18 surface dose was 5 to 7 times greater than average lung dose. These results indicate that local
19 enhancement of dose occurs in healthy lungs, which could be an important factor in eliciting
20 pathophysiological effects.
21
22 6.2.2.5 Deposition of Specific Size Modes of Ambient Aerosol
23 The studies described in previous sections generally evaluated deposition using individual
24 particle sizes within certain ranges without consideration of specific relevant ambient size ranges.
25 Some recent modeling studies, however, have considered the deposition profiles of particle
26 modes that exist in ambient air, so as to provide estimates on dosimetry of these "real world"
27 particle size fractions. One such study using a lung-anatomical model (Venkataraman and Kao,
28 1999) examined the contribution of two specific size modes of the PM10 ambient aerosol, namely
29 the fine mode (defined as particles with diameters up to 2.5 //m) and the thoracic fraction of the
30 coarse mode (defined as particles with diameters 2.5 to 10 //m), to total lung and regional lung
31 doses (i.e., a daily dose expressed as //g/day, and a surface dose expressed a //g/cm2/day)
April 2002 6-20 DRAFT-DO NOT QUOTE OR CITE
-------
c
.0
"o
CO
c
o
'(7)
o
Q.
0
Q
"CD
o
o
0.1-
0.0
Dp = 3|jm
100
200
300
400
500
Volumetric Lung Region (ml_)
Figure 6-8. Lung deposition fractions in ten volumetric regions for particle sizes ranging
from ultrafine particle diameter (dp) of 0.04 to 0.01 (j,m (Panel A) to fine
(dp = 1.0 jum) (Panel B) and coarse (dp = 3 and 5 ^m) (Panels C and D).
Healthy young adults inhaled a small bolus of monodisperse aerosols under
a range of normal breathing conditions (ie., tidal volume of 500 mL at
breathing frequencies of 9,15, and 30 breaths per niin.).
Source: Kim et al. (1996); Kim and Hu (1998); Kim and Jacques (2000).
April 2002
6-21
DRAFT-DO NOT QUOTE OR CITE
-------
1 resulting from a 24-h exposure to a particle concentration of 150 //g/m3. The study also
2 evaluated deposition in terms of two metrics, namely mass dose and number dose. Deposition
3 was calculated using a mathematical model for a healthy human lung under both simulated
4 moderate exertion (1 L at 15 breaths/min) and vigorous exertion (1.5 L at 15 breaths/min), and
5 for a compromised lung (0.5 L at 30 breaths/min). Regional deposition values were obtained for
6 the ET, TB, and A regions. Because the exposure scenario used is quite unrealistic, only general
7 trends should be inferred from this study rather than actual deposition values.
8 Daily mass dose peaked in the A airways for all breathing patterns; whereas that for the
9 coarse fractions was comparable in the TB and A regions. The mass per unit surface area of
10 various airways from the fine and coarse fractions was larger in the trachea and first few
11 generations of bronchi. It was suggested that these large surface doses may be related to
12 aggravation of upper respiratory tract illness in geographical areas where coarse particles are
13 present.
14 The daily number dose was different for fine and coarse fractions in all lung airways, with
15 the dose from the fine fraction higher by about 100 times in the ET and about 10s times in
16 internal lung airways. The surface number dose (particles/cm2/day) was 103 to 10s times higher
17 for fine than for coarse particles in all lung airways, indicating the larger number of fine particles
18 depositing. Particle number doses did not follow trends in mass doses and are much higher for
19 fine than coarse particles and are higher for different breathing patterns. It also was concluded
20 that the fine fraction contributes 10,000 times greater particle number per alveolar macrophage
21 than the coarse fraction particles. As noted, these results must be viewed with caution because
22 they were obtained using a pure mathematical model that must be validated in terms of realistic
23 physiologic conditions.
24 Another evaluation of deposition that included consideration of size mode of the ambient
25 aerosol was that of Broday and Georgopoulos (2001). In this case, a mathematical model was
26 used to account for particle hygroscopic growth, transport, and deposition in tracking the changes
27 in the size distribution of inhaled aerosols. It was concluded that different rates of particle
28 growth in the inspired air resulted in a change in the aerosol size distribution, such that increased
29 mass and number fractions of inspired ultrafine particles (< 0.1 //m) were found in the size range
30 between 0.1 to 1 //m and, therefore, deposited to a lesser extent due to a decrease in diffusion
31 deposition. On the other hand, particles that were originally in the 0.1 to 1 //m size range when
April 2002 6-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 inhaled will undergo enhanced deposition because of their increase in size resulting from
2 hygroscopic growth. Hence, the initial size distribution of the inhaled polydisperse aerosol
3 affects the evolution of size distribution once inhaled and, thus, its deposition profile in the
4 respiratory tract. Hygroscopicity of respirable particles must be considered for accurate
5 predictions of deposition. Because different size fractions likely have different chemical
6 composition, such changes in deposition patterns will affect biological responses.
7
8 6.2.3 Biological Factors Modulating Deposition
9 Experimental deposition data in humans are commonly derived using healthy adult
10 Caucasian males. Various factors can act to alter deposition patterns from those obtained in this
11 group. Evaluation of these factors is important to help understand potentially susceptible
12 subpopulations because differences in biological response following pollutant exposure may be
13 caused by dosimetry differences as well as by differences in innate sensitivity. The effects of
14 different biological factors on deposition were discussed in the 1996 PM AQCD (U.S.
15 Environmental Protection Agency, 1996) and are summarized below together with additional
16 information obtained from more recent studies.
17
18 6.2.3.1 Gender
19 Males and females have different body size and ventilatory parameter distributions;
20 therefore, it is expected that there would be gender differences in deposition. In some of the
21 controlled studies, however, men and women are breathing at the same tidal volume and
22 frequency. If the women are generally smaller than the men, the increased minute ventilation
23 compared to their normal ventilation would cause different changes in deposition patterns.
24 In these cases, it would be better for the investigators to have used size-adjusted tidal volumes.
25 This may help to explain some of the differing results discussed below.
26 Using particles in the 2.5- to 7.5-//m size range, Pritchard et al. (1986) indicated that, for
27 comparable particle sizes and inspiratory flow rates, females had higher ET and TB deposition
28 and smaller A deposition than did males. The ratio of A deposition to total thoracic deposition in
29 females also was found to be smaller. These differences were attributed to gender differences in
30 airway size.
April 2002 6-23 DRAFT-DO NOT QUOTE OR CITE
-------
1 In another study (Bennett et al., 1996), the total respiratory tract deposition of 2-//m
2 particles was examined in adult males and females aged 18 to 80 years who breathed with a
3 normal resting pattern. Deposition was assessed in terms of a deposition fraction, which was the
4 difference between the amount of particles inhaled and exhaled during oral breathing. Although
5 there was a tendency for a greater deposition fraction in females compared to males, and because
6 males had greater minute ventilation, the deposition rate (i.e., deposition per unit time) was
7 greater in males than in females.
8 Kim and Hu (1998) assessed regional deposition patterns in healthy adult males and
9 females using particles with median aerodynamic sizes of 1, 3, and 5 //m and a bolus delivery
10 technique that involved controlled breathing. The total deposition in the lungs was similar for
11 both genders with the smallest particle, but was greater in women for the 3- and 5-//m particles,
12 regardless of the inhalation flow rate used; this difference ranged from 9 to 31%, with higher
13 values associated with higher flow rates. The pattern of deposition was similar for both genders
14 although females showed enhanced deposition peaks for all three particle sizes. The volumetric
15 depth location of these peaks was found to shift from peripheral (i.e., increased volumetric depth)
16 to proximal (i.e., shallow volumetric depth) regions of the lung with increasing particle size, but
17 the shift was greater in females than in males. Thus, deposition appeared to be more localized in
18 the lungs of females compared to those of males. These differences were attributed to a smaller
19 size of the upper airways in females than in males, particularly of the laryngeal structure. Local
20 deposition of l-//m particles was somewhat flow dependent but, for larger (5-//m) particles, was
21 largely independent of flow (flows did not include those that would be typical of exercise).
22 In a related study, Kim et al. (2000) evaluated differences in deposition between males and
23 females in terms of exercise levels of ventilation and breathing patterns. Using particles at the
24 same size noted above and a number of breathing conditions, total lung deposition was
25 comparable between men and women for l-//m particles, but was slightly greater in women than
26 men for 3- and 5-//m particles with all breathing patterns. The gender difference was about 15%
27 at rest, and variable during exercise, depending on particle size. However, total lung deposition
28 rate (i.e., deposition per unit time) was found to be 3 to 4 times greater during moderate exercise
29 than during rest for all particle sizes. Thus, it was concluded that exercise may increase the
30 health risk from particles because of increased large airway deposition and that women may be
31 more susceptible to this exercise-induced change.
April 2002 6-24 DRAFT-DO NOT QUOTE OR CITE
-------
1 Jaques and Kim (2000) and Kim and Jaques (2000) expanded the evaluation of deposition
2 in males and females to particles <1 //m. They measured total lung deposition in healthy adults
3 using sizes in the ultrafme mode (0.04 to 0.1 //m), in addition to those having diameters of 1 and
4 5 //m. Total lung deposition was greater in females than in males for 0.04- and 0.06-//m
5 particles. The difference was negligible for 0.08-and 0.1-//m particles. Therefore, the gender
6 effect was particle-size dependent, showing a greater deposition in females for very small
7 ultrafme and large coarse particles, but not for particles ranging from 0.08 to 1 //m. A local
8 deposition fraction was determined in each volumetric compartment of the lung to which
9 particles are injected based on the inhalation procedure (Kim and Jaques, 2000). The deposition
10 fraction was found to increase with increasing lung depth from the mouth, reach a peak value,
11 and then decrease with further increase in lung volumetric depth. The height of the peak and its
12 depth did vary with particle size and breathing pattern. Peak deposition for the 5-//m particles
13 was more proximal than that for the l-//m particles; whereas that for the ultrafme particles
14 occurred between these two peaks. For the ultrafme particles, the peak deposition became more
15 proximal as particle size decreased. Although this pattern of deposition distribution was similar
16 for both men and women, the region of peak deposition was shifted closer to the mouth and peak
17 height was slightly greater for women than for men for all exposure conditions.
18
19 6.2.3.2 Age
20 Airway structure and respiratory conditions vary with age, and these variations may alter
21 the deposition pattern of inhaled particles. The limited experimental studies reported in the 1996
22 PM AQCD (U. S. Environmental Protection Agency, 1996) indicated results ranging from no
23 clear dependence of total deposition on age to slightly higher deposition in children than adults.
24 However, children have a different resting ventilation than do adults. The experimental studies
25 must adjust for the higher minute ventilation per unit body weight in children when comparing
26 deposition results to those obtained in adults.
27 Potential regional deposition differences between children and adults have been assessed to
28 a greater extent using mathematical models. These indicated that, if the entire respiratory tract
29 and a complete breathing cycle at normal rate are considered, then ET deposition in children
30 would be generally higher than that in adults, but TB and A regional deposition in children may
31 be either higher or lower than that in adults, depending on particle size (Xu and Yu, 1986).
April 2002 6-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 Enhanced deposition in the TB region would occur for particles <5 //m in children (Xu and Yu,
2 1986; Hofmann et al., 1989a).
3 An age dependent theoretical model to predict regional particle deposition in childrens'
4 lungs that incorporates breathing parameters and morphology of the growing lung was developed
5 by Musante and Martonen (1999). The model was used to compare deposition of monodisperse
6 aerosols, ranging from 0.25 to 5 //m, in the lungs of children (aged 7, 22, 48, and 98 mo) at rest
7 to that in adults (aged 30 years) at rest. Compared to adults, A deposition was highest in the
8 48- and 98-mo subjects for all particle sizes; TB deposition was found to be a monotonically
9 decreasing function of age for all sizes; and total lung deposition (i.e., TB+A) was generally
10 higher in children than adults, with children of all ages showing similar total deposition fractions.
11 This model was used by Musante and Martonen (2000a) to evaluate the deposition of a
12 polydisperse aerosol that has been extensively used in toxicological studies, namely residual oil
13 fly ash (ROFA) having an MMAD of 1.95 //m, a geometric standard deviation of 2.19, and a
14 CMD of 0.53 (assuming a particle density of 0.34 g/cm2). Deposition was evaluated under
15 resting breathing conditions. The mass based deposition fraction of the particles was found to
16 decrease with age from 7 mo to adulthood, but the mass deposition per unit surface area in the
17 lungs of children could be significantly greater than that in the adult.
18 Phalen and Oldham (2001) calculated the respiratory deposition of particles with sizes
19 ranging from 0.1 to 10 //m in diameter for 20 year-old adults and 2 year-old children. Total lung
20 deposition was comparable between adults and children for all particle sizes tested; however, TB
21 deposition was much greater in children than in adults (from 13 to 81%, depending on particle
22 size). Particle deposition in the A region was significantly reduced in children.
23 Cheng et al. (1995) examined deposition of ultrafme particles in replica casts of the nasal
24 airways of children aged 1.5 to 4 years. Particle sizes ranged from 0.0046 to 0.2 //m, and both
25 inspiratory and expiratory flow rates were used (3 to 16 L/min). Deposition efficiency was found
26 to decrease with increasing age for a given particle size and flow rate.
27 Oldham et al. (1997) examined the deposition of monodisperse particles having diameters
28 of 1, 5, 10, and 15 //m in hollow airway models that were designed to represent the trachea and
29 the first few bronchial airway generations of an adult, a 7-year-old child, and a 4-year-old child.
30 They noted that, in most cases, the total deposition efficiency was greater in the child-size
31 models than in the adult model.
April 2002 6-26 DRAFT-DO NOT QUOTE OR CITE
-------
1 Bennett et al. (1997a) analyzed the regional deposition of poorly soluble 4.5 //m particles
2 inhaled via mouthpiece. The subjects were children and adults with mild cystic fibrosis (CF), but
3 who likely had normal upper airway anatomy such that intra- and extrathoracic deposition would
4 be similar to that in healthy people. The mean age of the children was 13.8 years and for the
5 adults was 29.1 years. Extrathoracic deposition, as a percentage of total respiratory tract
6 deposition, was higher by about 50% in children compared to adults (30.7%, 20.1%, and 16.0%,
7 respectively). There was an age dependence of ET deposition in the children, in that the
8 percentage ET deposition tended to be higher at a younger age (p = 0.08); the younger group
9 (<14 years) (p = 0.05) had almost twice the percentage ET deposition of the older group
10 (>14 years). Additional analyses showed an inverse correlation of extrathoracic deposition with
11 body height. There was no significant difference in lung or total respiratory tract deposition
12 between the children and adults. Because ET deposition was age dependent, and total deposition
13 was not, this suggests that the ET region does a more effective job in children of filtering out
14 particles that would otherwise reach the TB region. However, because the lungs of children are
15 smaller than are those of adults, children may still have comparable deposition per unit surface
16 area as adults.
17 Bennett and Zeman (1998) measured the deposition of monodisperse 2 //m (MMAD)
18 particles in children (aged 7 to 14 years) and adolescents (aged 14 to 18 years) for comparison to
19 that in adults (19 to 35 years). Each subject inhaled the particles by following their previously
20 determined individual spontaneous resting breathing pattern. Deposition was assessed by
21 measuring the amount of particles inhaled and exhaled. There was no age-related difference in
22 deposition within the children group. There was also no significant difference in deposition
23 between the children and adolescents, between the children and adults, or between the
24 adolescents and adults. However, the investigators noted that, because the children had smaller
25 lungs and higher minute volumes relative to lung size, they likely would receive greater doses of
26 particles per lung surface area compared to adults. Furthermore, breath-to-breath fractional
27 deposition in children did vary with tidal volume, increasing with increasing volume. The rate of
28 deposition normalized to lung surface area tended (p = 0.07) to be greater (35%) in children
29 when compared to the combined group of adolescents and adults. These additional studies still
30 do not provide unequivocal evidence for significant differences in deposition between adults and
31 children, even when considering differences in lung surface area. However, it should be noted
April 2002 6-27 DRAFT-DO NOT QUOTE OR CITE
-------
1 that differences in levels of activity between adults and children are likely to play a fairly large
2 role in age-related differences in deposition patterns of ambient particles. Children generally
3 have higher activity levels during the day and higher associated minute ventilation per lung size,
4 which can contribute to a greater size-specific dose of particles. Activity levels in relationship to
5 exposure are discussed more fully in Chapter 5.
6 Another subpopulation of potential concern related to susceptibility to inhaled particles is
7 the elderly. In the study of Bennett et al. (1996), in which the total respiratory tract deposition of
8 2-//m particles was examined in people aged 18 to 80 years, the deposition fraction in the lungs
9 of people with normal lung function was found to be independent of age, depending solely on
10 breathing pattern and airway resistance.
11
12 6.2.3.3 Respiratory Tract Disease
13 The presence of respiratory tract disease can affect airway structure and ventilatory
14 parameters, thus altering deposition compared to that occurring in healthy individuals. The effect
15 of airway diseases on deposition has been studied extensively, as described in the 1996 PM
16 AQCD (U.S. Environmental Protection Agency, 1996). Studies described therein had shown that
17 people with chronic obstructive pulmonary disease (COPD) had very heterogeneous deposition
18 patterns, and differences in regional deposition compared to normals. People with asthma and
19 obstructive pulmonary disease tended to have greater TB deposition than did healthy people.
20 Furthermore, there tended to be an inverse relationship between bronchoconstriction and the
21 extent of deposition in the A region; whereas total respiratory tract deposition generally increased
22 with increasing degrees of airway obstruction. The described studies were performed during
23 controlled breathing; i.e., all subjects breathed with the same tidal volume and respiratory rate.
24 However, although resting tidal volume is similar or elevated in people with COPD compared to
25 normal, healthy individuals the former tend to breathe at a faster rate, resulting in higher than
26 normal tidal peak flow and resting minute ventilation. Thus, some of the reported differences in
27 the deposition of particles could have been caused by increased fractional deposition with each
28 breath. Although the extent to which lung deposition may change with respect to particle size,
29 breathing pattern, and disease status in people with COPD is still unclear, some recent studies
30 have attempted to provide additional insight into this issue.
April 2002 6-28 DRAFT-DO NOT QUOTE OR CITE
-------
1 Bennett et al. (1997b) measured the fractional deposition of insoluble 2-//m particles in
2 people with severe to moderate COPD (mix of emphysema and chronic bronchitis, mean age
3 62 years) and compared this to healthy older adults (mean age 67 years) under conditions where
4 the subjects breathed using their individual resting breathing pattern, as well as a controlled
5 breathing pattern. People with COPD tended to breathe with elevated tidal volume and at a
6 faster rate than people with healthy lungs, resulting in about 50% higher resting minute
7 ventilation. Total respiratory tract deposition was assessed in terms of deposition fraction, a
8 measure of the amount deposited based on measures of amount of aerosol inhaled and exhaled,
9 and deposition rate, the particles deposited per unit time. Under typical breathing conditions,
10 people with COPD had about 50% greater deposition fraction than did age-matched healthy
11 adults. Because of the elevation in minute ventilation, people with COPD had average
12 deposition rates about 2.5 times that of healthy adults. Similar to previously reviewed studies
13 (U.S. Environmental Protection Agency, 1996), these investigators observed an increase in
14 deposition with an increase in airway resistance, suggesting that, at rest, COPD resulted in
15 increased deposition of fine particles in proportion to the severity of airway disease. The
16 investigators also reported a decrease in deposition with increasing mean effective airspace
17 diameter; this suggested that the enhanced deposition was associated more with the chronic
18 bronchitic component of COPD than with the emphysematous component. Greater deposition
19 was noted with natural breathing compared to the fixed pattern.
20 Kim and Kang (1997) measured lung deposition of l-//m particles inhaled via the mouth by
21 healthy adults (mean age 27 years) and by those with various degrees of airway obstruction,
22 namely smokers (mean age 27 years), smokers with small airway disease (SAD; mean age
23 37 years), asthmatics (mean age 48 years), and patients with COPD (mean age 61 years)
24 breathing under the same controlled pattern. Deposition fraction was obtained by measuring the
25 number of particles inhaled and exhaled, breath by breath. There was a marked increase in
26 deposition in people with COPD. Deposition was 16%, 49%, 59%, and 103% greater in
27 smokers, smokers with SAD, asthmatics and people with COPD, respectively, than in healthy
28 adults. Deposition in COPD patients was significantly greater than that associated with either
29 SAD or asthma; there was no significant difference in deposition between people with SAD and
30 asthma. Deposition fraction was found to be correlated with percent predicted forced expiratory
31 volume (FEVj) and forced expiratory flow (FEF25-75%). Airway resistance was not correlated
April 2002 6-29 DRAFT-DO NOT QUOTE OR CITE
-------
1 strongly with total lung deposition. Kohlhaufl et al. (1999) showed increased deposition of fine
2 particles (0.9 //m) in women with bronchial hyperresponsiveness.
3 Segal et al. (2000a) developed a mathematical model for airflow and particle motion in the
4 lung that was used to evaluate how lung cancer affects deposition patterns in the lungs of
5 children. It was noted that the presence of airway tumors could affect deposition by increasing
6 probability of inertial deposition and diffusion. The former would occur on upstream surfaces of
7 tumors and the latter on downstream surfaces. It was concluded that particle deposition is
8 affected by the presence of airway disease, that effects may be systematic and could be predicted,
9 and that, therefore, they could be incorporated into dosimetry models.
10 Brown et al. (2001) examined the relationship between regional lung deposition for coarse
11 particles (5 //m) and ventilation patterns in healthy adults and in patients with cystic fibrosis
12 (CF). They found that deposition in the TB region was positively associated with regional
13 ventilation in healthy subjects, but negatively associated in CF patients. The relationships were
14 reversed for deposition in the A region. These data suggest that significant coarse particle
15 deposition may occur in the TB region of poorly ventilated lungs, as occurs in CF; whereas TB
16 deposition follows ventilation in healthy subjects.
17 Thus, the database related to particle deposition and lung disease suggests that total lung
18 deposition generally is increased with obstructed airways, regardless of deposition distribution
19 between the TB and A regions. Airflow distribution is very uneven in diseased lungs because of
20 the irregular pattern of obstruction, and there can be closure of small airways. In this situation, a
21 part of the lung is inaccessible, and particles can penetrate deeper into other, better ventilated
22 regions. Thus, deposition can be enhanced locally in regions of active ventilation, particularly in
23 the A region. The relationships between lung deposition and airway obstruction or ventilation
24 distribution were previously studied in vivo in animal models (Kim, 1989; Kim et al., 1989).
25
26 6.2.3.4 Anatomical Variability
27 As indicated above, variations in anatomical parameters between genders, and between
28 healthy people and those with obstructive lung disease, can affect deposition patterns. However,
29 previous analyses generally have overlooked the effect on deposition of normal interindividual
30 variability in airway structure in healthy individuals. This is an important consideration in
31 dosimetry modeling, which often is based on a single idealized structure. Studies that have
April 2002 6-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 become available since the 1996 PM AQCD have attempted to assess the influence of such
2 variation in respiratory tract structure on deposition patterns.
3 The ET region is the first to contact inhaled particles and, therefore, deposition within this
4 region would reduce the amount of particles available for deposition in the lungs. Variations in
5 relative deposition within the ET region will, therefore, propagate through the rest of the
6 respiratory tract, creating differences in calculated doses from individual to individual.
7 A number of studies have examined the influence of variations in airway geometry on deposition
8 in the ET region.
9 Cheng et al. (1996) examined nasal airway deposition in healthy adults using particles
10 ranging in size from 0.004 to 0.15 //m and at two constant inspiratory flow rates, 167 and
11 33 mL/s. Deposition was evaluated in relation to measures of nasal geometry as determined by
12 magnetic resonance imaging and acoustic rhinometry. They noted that interindividual variability
13 in deposition was correlated with the wide variation of nasal dimensions, in that greater surface
14 area, smaller cross-sectional area, and increasing complexity of airway shape were all associated
15 with enhanced deposition.
16 Using a regression analysis of data on nasal airway deposition derived from Cheng et al.
17 (1996), Guilmette et al. (1997) noted that the deposition efficiency within this region was highly
18 correlated with both nasal airway surface area and volume; this indicated that airway size and
19 shape factors were important in explaining intraindividual variability noted in experimental
20 studies of human nasal airway aerosol deposition. Thus, much of the variability in measured
21 deposition among people resulted from differences in the size and shape of specific airway
22 regions.
23 Kesavanathan and Swift (1998) also evaluated the influence of geometry in affecting
24 deposition in the nasal passages of normal adults from two ethnic groups. Mathematical
25 modeling of the results indicated that the shape of the nostril affected particle deposition in the
26 nasal passages, but that there still remained large intersubject variations in deposition when this
27 was accounted for, and which was likely caused by geometric variability in the mid and posterior
28 regions of the nasal passages.
29 Bennett et al. (1998) studied the role of anatomic dead space (ADS) in particle deposition
30 and retention in bronchial airways, using an aerosol bolus technique. They found that the
31 fractional deposition was dependant on the subject's ADS and that a significant number of
April 2002 6-31 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles was retained beyond 24 h. This finding of prolonged retention of insoluble particles in
2 the airways is consistent with the findings of Scheuch et al. (1995) and Stahlhofen et al. (1986a)
3 and with the predictions of asymmetric stochastic human lung models (Asgharian et al., 2001).
4 Bennett et al. (1999) also found a lung volume-dependent asymmetric distribution of particles
5 between the left and right lung; the leftright ratio was increased at increased percentage of total
6 lung capacity (e.g., at 70% TLC, L:R was 1.60).
7 From the analysis of detailed deposition patterns measured by a serial bolus mouth delivery
8 method, Kim and Hu (1998) and Kim and Jaques (2000) found a marked enhancement in
9 deposition in the very shallow region (lung penetration depth <150 mL) of the lungs in females.
10 The enhanced local deposition for both ultrafme and coarse particles was attributed to a smaller
11 size of the upper airways, particularly of the laryngeal structure.
12 Hofmann et al. (2000) examined the role of heterogeneity of airway structure in the rat
13 acinar region in affecting deposition patterns within this area of the lungs. By the use of different
14 morphometric models, they showed that substantial variability in predicted particle deposition
15 would result.
16
17 6.2.4 Interspecies Patterns of Deposition
18 The primary purpose of this document is to assess the health effects of particles in humans.
19 As such, human dosimetry studies have been stressed. Such studies avoid uncertainties
20 associated with extrapolation of dosimetry from laboratory animals to humans. Nevertheless,
21 animal models have been and are currently being used in evaluations of health effects from
22 particulate matter because there are ethical limits to the types of studies that can be performed on
23 human subjects. Because of this, there is considerable need to understand dosimetry in animals
24 and to understand dosimetric differences between animals and humans. In this regard, there are a
25 number of newly published studies that were designed to assess particle dosimetry in commonly
26 used animals and to relate this to dosimetry in humans.
27 The various species used in inhalation toxicology studies that serve as the basis for
28 dose-response assessment may not receive identical doses in a comparable respiratory tract
29 region (i.e., ET, TB, or A) when exposed to the same aerosol at the same inhaled concentration.
30 Such interspecies differences are important because any toxic effect is often related to the
31 quantitative pattern of deposition within the respiratory tract as well as to the exposure
April 2002 6-32 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentration; this pattern determines not only the initial respiratory tract tissue dose, but also the
2 specific pathways by which deposited material is cleared and redistributed (Schlesinger, 1985).
3 Differences in patterns of deposition between humans and animals were summarized previously
4 in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996) and by others
5 (Schlesinger et al., 1997). Such differences in initial deposition must be considered when
6 relating biological responses obtained in laboratory animal studies to effects in humans.
7 It is difficult to compare systematically interspecies deposition patterns obtained from
8 various reported studies because of variations in experimental protocols, measurement
9 techniques, definitions of specific respiratory tract regions, and so on. For example, tests with
10 humans are generally conducted under protocols that standardize the breathing pattern; whereas
11 those using laboratory animals involve a wider variation in respiratory exposure conditions (e.g.,
12 spontaneous breathing versus ventilated breathing or varying degrees of sedation). Much of the
13 variability in the reported data for individual species may be due to the lack of normalization for
14 specific respiratory parameters during exposure. In addition, the various studies have used
15 different exposure techniques, such as nasal mask, oral mask, oral tube, or tracheal intubation.
16 Regional deposition is affected by the exposure route and delivery technique employed.
17 Figure 6-9 shows the regional deposition data versus particle diameter in commonly used
18 laboratory animals obtained by various investigators, as compiled by Schlesinger (1988; 1989).
19 The results are described in detail in the 1996 PM AQCD (U.S. Environmental Protection
20 Agency, 1996). In general, there is much variability in the data; however, it is possible to make
21 some generalizations concerning comparative deposition patterns. The relationship between total
22 respiratory tract deposition and particle size is approximately the same in humans and most of
23 these animals; deposition increases on both sides of a minimum that occurs for particles of 0.2 to
24 1 //m. Interspecies differences in regional deposition occur due to anatomical and physiological
25 factors. In most laboratory animal species, deposition in the ET region is near 100 percent for a
26 particle diameter (dp) greater than 5 //m (Raabe et al., 1988), indicating greater efficiency than
27 that seen in humans. In the TB region, there is a relatively constant, but lower, deposition
28 fraction for dp greater than 1 //m in all species compared to humans. Finally, in the A region,
29 deposition fraction peaks at a lower particle size (dp about 1 //m) in laboratory animals than in
30 humans.
31
April 2002 6-33 DRAFT-DO NOT QUOTE OR CITE
-------
IUU
80
60
40
20
n
I I I IIW^
O Rat II F
n Hamster T
'k ^' $
fjf 1 V
u ™, . :
I | I T ^BAAsjari
0.01 0.1 1.0 1C
Particle Diameter (|jm)
Figure 6-9. Regional deposition fraction in laboratory animals as a function of particle
size. Particle diameters are aerodynamic (MMAD) for those > 0.5 ^m and
geometric (or diffusion equivalent) for those < 0.5 (j,m.
Source: Schlesinger (1988).
April 2002
6-34
DRAFT-DO NOT QUOTE OR CITE
-------
1 One of the issues that must be considered in interspecies comparisons of hazards from
2 inhaled particles is inhalability of the aerosol in the atmosphere of concern. Although this may
3 not be an issue for humans per se as far as exposure to ambient particles is concerned, it can be
4 an important issue when attempting to extrapolate to humans the results of studies using animal
5 species commonly employed in inhalation toxicological studies (Miller et al., 1995).
6 For example, differences between rat and human become very pronounced for particles >5 //m,
7 and some differences are also evident for particles as small as 1 //m (Figure 6-10).
8 A number of studies have addressed various aspects of interspecies differences in
9 respiratory tract deposition using mathematical modeling approaches. Hofmann et al. (1996)
10 compared deposition between rat and human lungs, using three-dimensional asymmetric
11 bifurcation models and mathematical procedures for obtaining air flow and particle trajectories.
12 Deposition in segmental bronchi and terminal bronchioles was evaluated under both inspiration
13 and expiration at particle sizes of 0.01, 1.0, and 10 //m, which covers the range of deposition
14 mechanisms from diffusion to impaction. Total deposition efficiencies of all particles in the
15 upper and lower airway bifurcations were comparable in magnitude for both rat and human.
16 However, the investigators noted that penetration probabilities from preceding airways must be
17 considered. When considering the higher penetration probability in the human lung, the resulting
18 bronchial deposition fractions were generally higher in human than in rat. For all particle sizes,
19 deposition at rat bronchial bifurcations was less enhanced on the carinas compared to that found
20 in human airways.
21 Hofmann et al. (1996) attempted to account for interspecies differences in branching
22 patterns in deposition analyses. Numerical simulations of three-dimensional particle deposition
23 patterns within selected (species-specific) bronchial bifurcations indicated that morphologic
24 asymmetry was a major determinant of the heterogeneity of local deposition patterns. They noted
25 that many interspecies deposition calculations used morphometry that was described by
26 deterministic lung models (i.e., the number of airways in each airway generation is constant, and
27 all airways in a given generation have identical lengths and diameters). Such models cannot
28 account for variability and branching asymmetry of airways in the lungs. Thus, their study
29 employed computations that used stochastic morphometric models of human and rat lungs
30 (Koblinger and Hofmann, 1985, 1988; Hofmann et al., 1989b) and evaluated regional and local
31 particle deposition. Stochastic models of lung structure describe, in mathematical terms, the
April 2002 6-3 5 DRAFT-DO NOT QUOTE OR CITE
-------
100
Total Respiratory Tract
Human
- Oral Breathing
0
100
Human
- Nasal Breathing
100 r Rat
0.01 0.1 1.0
Particle Diameter (|jm)
Tracheobronchial Region
Human
Oral Breathing
10
01
Q
100 r
Rat
0.01 0.1 1.0
Particle Diameter (|jm)
10
B
Extrathoracic Region
100r Human
. Oral Breathing
.-. 100
CD
Q
Human
- Nasal Breathing
100 r Rat
.01 0.1 1.0
Particle Diameter (|jm)
D
Alveolar Region
Human
. Oral Breathing
0
100
CD
Q
Human
- Nasal Breathing
100 r Rat
0.01 0.1 1.0
Particle Diameter (^m)
Figure 6-10. Particle deposition efficiency in rats and humans as a function of particle
size for the (A) total respiratory tract, (B) thoracic region,
(C) tracheobronchial region, and (D) alveolar region. Each curve
represents an eye fit through mean values (or centers of ranges) for the
data compiled by Schlesinger (1985). Particle diameters are aerodynamic
(MMAD) for those >0.5 ^,m and geometric (or diffusion equivalent) for
those < 0.5 (j,m.
Source: Modified from Schlesinger (1989).
April 2002
6-36
DRAFT-DO NOT QUOTE OR CITE
-------
1 inherent asymmetry and variability of the airway system, including diameter, length, and angle.
2 They are based on statistical analyses of actual morphometric analyses of lungs. The model also
3 incorporated breathing patterns for humans and rats. In a later analysis (Hoffmann and
4 Bergmann, 1998), the dependence of deposition on particle size was found to be qualitatively
5 similar in both rats and humans, with deposition minima in the size range of 0.1 to 1 //m for total
6 deposition as well as deposition within the TB and A regions. In addition, a deposition
7 maximum occurred at about 0.02 to 0.03 //m and between 3 and 5 //m in both species. The
8 deposition decrease in the A region at the smallest and largest sizes resulted from the filtering
9 efficiency of upstream airways. Although deposition patterns were qualitatively similar in rat
10 and human, deposition in the human lung appeared to be consistently higher than in the rat in all
11 regions of the lung (TB and A) over the entire size range. Both species showed a similar pattern
12 of dependence of deposition on flow rate.
13 The above model also assessed local deposition. In both human and rat, deposition of
14 0.001-//m particles was highest in the upper bronchial airways; whereas 0.1- and l-//m particles
15 showed higher deposition in more peripheral airways, namely the bronchiolar airways in rat and
16 the respiratory bronchioles in humans. Deposition was variable within any branching generation
17 because of differences in airway dimensions, and regional and total deposition also exhibited
18 intrasubject variations. Airway geometric differences between rats and humans were reflected in
19 deposition. Because of the greater branching asymmetry in rats, prior to about generation 12,
20 each generation showed deposition maxima at two particle sizes, reflecting deposition in major
21 and minor daughters. These geometric differences became reduced with depth into the lung;
22 beyond generation 12, these two maxima were no longer seen.
23 Another comparison of deposition in lungs of humans and rats was performed by Musante
24 and Martonen (2000b). An interspecies mathematical dosimetry model was used to determine
25 the deposition of ROFA in the lungs under sedentary and light activity breathing patterns. This
26 latter condition was mimicked in the rat by increasing the CO2 level in the exposure system. The
27 MMAD of the particle size distribution was 1.95 //m with a geometric standard deviation of 2.19.
28 They noted that physiologically comparable respiratory intensity levels did not necessarily
29 correspond to comparable dose distribution in the lungs. Because of this, the investigators
30 speculate that the resting rat may not be a good model for the resting human. The ratio of aerosol
31 mass deposited in the TB region to that in the A region for the human at rest was 0.961,
April 2002 6-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 indicating fairly uniform deposition throughout the lungs. On the other hand, in the resting rat,
2 the ratio was 2.24, indicating greater deposition in the TB region than in the A region. However,
3 by mimicking light activity in the rat, the ratio was reduced to 0.97, similar to the human. These
4 data suggest that ventilatory characteristics in animal models may have to be adjusted to provide
5 for comparable regional deposition to that in humans.
6 The relative distribution of particles deposited within the bronchial and alveolar regions of
7 the airways may differ in the lungs of animals and humans for the same total amount of deposited
8 matter because of structural differences. The effect of such structural differences between rat and
9 human airways on particle deposition patterns was examined by Hofmann et al. (1999; 2000) in
10 an attempt to find the most appropriate morphometric parameter to characterize local particle
11 deposition for extrapolation modeling purposes. Particle deposition patterns were evaluated as
12 functions of three morphometric parameters, namely (1) airway generation, (2) airway diameter,
13 and (3) cumulative path length. It was noted that airway diameter was a more appropriate
14 morphometric parameter for comparison of particle deposition patterns in human and rat lungs
15 than was airway generation.
16 The manner in which particle dose is expressed, that is, the specific dose metric, may affect
17 relative differences in deposition between humans and other animal species. For example,
18 although deposition when expressed on a mass per unit alveolar surface area basis may not be
19 different between rats and humans, dose metrics based on particle number per various anatomical
20 parameters (e.g., per alveolus or alveolar macrophage) can differ between rats and humans,
21 especially for particles around 0.1 to 0.3 //m (Miller et al., 1995). Furthermore, in humans with
22 lung disease (such as asthma or COPD), differences between rat and human can be even more
23 pronounced.
24 The probability of any biological effect occurring in humans or animals depends on
25 deposition and retention of particles, as well as the underlying tissue sensitivity. Interspecies
26 dosimetric extrapolation must consider these differences in evaluating dose-response
27 relationships. Thus, even similar deposition patterns may not result in similar effects in different
28 species, because dose also is affected by clearance mechanisms. In addition, the total number of
29 particles deposited in the lung may not be the most relevant dose metric for interspecies
30 comparisons. For example, it may be the number of deposited particles per unit surface area or
31 dose to a specific cell (e.g., alveolar macrophage) that determines response for specific regions.
April 2002 6-3 8 DRAFT-DO NOT QUOTE OR CITE
-------
1 More specifically, even if deposition is similar in rat and human, there would be a higher
2 deposition density in the rat because of the smaller surface area of rat lung. Thus, species-
3 specific differences in deposition density should be considered when health effects observed in
4 laboratory animals are being evaluated for potential effects occurring in humans.
5
6
7 6.3 PARTICLE CLEARANCE AND TRANSLOCATION
8 This section discusses the clearance and translocation of particles that have deposited in the
9 respiratory tract. First, a basic overview of biological mechanisms and pathways of clearance in
10 the various region of the respiratory tract is presented. This is then followed by an update on
11 regional kinetics of particle clearance. Interspecies patterns of clearance are then addressed,
12 followed by new information on biological factors that may modulate clearance.
13
14 6.3.1 Mechanisms and Pathways of Clearance
15 Particles that deposit on airway surfaces may be cleared from the respiratory tract
16 completely or may be translocated to other sites within this system by various regionally distinct
17 processes. These clearance mechanisms, which are outlined in Table 6-1, can be categorized as
18 either absorptive (i.e., dissolution) or nonabsorptive (i.e., transport of intact particles) and may
19 occur simultaneously or with temporal variations. It should be mentioned that particle solubility
20 in terms of clearance refers to solubility within the respiratory tract fluids and cells. Thus, a
21 poorly soluble particle is considered to be one whose rate of clearance by dissolution is
22 insignificant compared to its rate of clearance as an intact particle. All deposited particles,
23 therefore, are subject to clearance by the same basic mechanisms, with their ultimate fate a
24 function of deposition site, physicochemical properties (including solubility and any toxicity),
25 and sometimes deposited mass or number concentration. Clearance routes from the various
26 regions of the respiratory tract have been discussed previously in detail (U.S. Environmental
27 Protection Agency, 1996; Schlesinger et al., 1997). They are schematically shown in Figure 6-11
28 (for extrathoracic and tracheobronchial regions) and in Figure 6-12 (for poorly soluble particle
29 clearance from the alveolar region) and are reviewed only briefly below.
30
April 2002 6-39 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-1. OVERVIEW OF RESPIRATORY TRACT PARTICLE CLEARANCE
AND TRANSLOCATION MECHANISMS
Extrathoracic region (ET)
Mucociliary transport
Sneezing
Nose wiping and blowing
Dissolution and absorption into blood
Tracheobronchial region (TB)
Mucociliary transport
Endocytosis by macrophages/epithelial cells
Coughing
Dissolution and absorption into blood/lymph
Alveolar region (A)
Macrophages, epithelial cells
Interstitial
Dissolution and absorption into blood/lymph
Source: Schlesinger (1995).
( Nasal Passages j^
) Dissolution /*
*—c
Mucociliary
Transport
) Dissolution
-* ( Tracheobronchial Tree
Figure 6-11. Major clearance pathways for particles deposited in the extrathoracic region
and tracheobronchial tree.
Source: Adapted from Schlesinger et al. (1997).
April 2002
6-40
DRAFT-DO NOT QUOTE OR CITE
-------
uefjusn&u i-cHiiut;
P
i
Phagocytosis by "\
Alveolar Macrophages J
1 ^
Movement within . niaSS,a~
Alveolar Lumen " Alveola
|
Bronchiolar/ Bronchial < Inte
Endocyt
Epithelij
je Through
r Epithelium
i
1
rstitium "^
osis I
Iveol
alCe
1 ^ ^_
~ Lymphatic Channels "^
Mucociliary Blanket i
-
^y
ar
^ t
Passage through
Pulmonary Capillary
Endothelium
— /" Phagocytosis byA
- 1 Interstitial 1
V^ Macrophages J
Gl Tract
Figure 6-12. Diagram of known and suspected clearance pathways for poorly
soluble particles depositing in the alveolar region. (The magnitude of
various pathways may depend upon size of deposited particle.)
Source: Modified from Schlesinger et al. (1997).
1 6.3.1.1 Extrathoracic Region
2 The clearance of poorly soluble particles deposited in the posterior portions of the nasal
3 passages occurs via mucociliary transport, with the general flow of mucus being towards the
4 nasopharynx. Mucus flow in the most anterior portion of the nasal passages is forward, clearing
5 deposited particles to the vestibular region, where removal occurs by sneezing, wiping, or
6 blowing. Soluble material deposited on the nasal epithelium is accessible to underlying cells via
7 diffusion through the mucus. Dissolved substances may be translocated subsequently into the
8 bloodstream. The nasal passages have a rich vasculature, and uptake into the blood from this
9 region may occur rapidly.
10 Clearance of poorly soluble particles deposited in the oral passages is by coughing and
11 expectoration or by swallowing into the gastrointestinal tract. Soluble particles are likely to be
12 rapidly absorbed after deposition, but it depends on the rate of dissolution of the particle and the
13 molecular size of the solute.
April 2002
6-41
DRAFT-DO NOT QUOTE OR CITE
-------
1 6.3.1.2 Tracheobronchial Region
2 Poorly soluble particles deposited within the TB region are cleared by mucociliary transport
3 towards the oropharynx, followed by swallowing. Poorly soluble particles also may traverse the
4 epithelium by endocytotic processes, entering the peribronchial region, be engulfed via
5 phagocytosis by airway macrophages (which can then move cephalad on the mucociliary
6 blanket), or enter the airway lumen from the bronchial or bronchiolar mucosa. Soluble particles
7 may be absorbed through the epithelium into the blood. It has been shown that blood flow
8 affects translocation from the TB region, in that decreased bronchial blood flow is associated
9 with increased airway retention of soluble particles (Wagner and Foster, 2001). There is,
10 however, evidence that even soluble particles may be cleared by mucociliary transport (Bennett
11 and Howite, 1989; Matsui et al., 1998; Wagner and Foster, 2001).
12
13 6.3.1.3 Alveolar Region
14 Clearance from the A region occurs via a number of mechanisms and pathways. Particle
15 removal by macrophages comprises the main nonabsorptive clearance process in this region.
16 These cells, which reside on the epithelium, phagocytize and transport deposited material that
17 they contact by random motion or via directed migration under the influence of chemotactic
18 factors.
19 Although alveolar macrophages normally comprise up to about 5% of the total alveolar
20 cells in healthy, nonsmoking humans and other mammals, the actual cell count may be altered by
21 particle loading. The magnitude of any increase in cell number is related to the number of
22 deposited particles rather than to total deposition by weight. Thus, equivalent masses of an
23 identically deposited substance would not produce the same response if particle sizes differed,
24 and the deposition of smaller particles would tend to result in a greater elevation in macrophage
25 number than would deposition of larger particles.
26 Particle-laden macrophages may be cleared from the A region along a number of pathways.
27 As noted in Figure 6-11, this includes cephalad transport via the mucociliary system after the
28 cells reach the distal terminus of the mucus blanket; movement within the interstitium to a
29 lymphatic channel; or perhaps traversing of the alveolar-capillary endothelium, directly entering
30 the bloodstream. Particles within the lymphatic system may be translocated to tracheobronchial
31 lymph nodes, which can become reservoirs of retained material. Particles subsequently reaching
April 2002 6-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 the postnodal lymphatic circulation will enter the blood. Once in the systemic circulation, these
2 particles, or transmigrated macrophages, can travel to extrapulmonary organs. Deposited
3 particles that are not ingested by alveolar macrophages may enter the interstitium, where they are
4 subject to phagocytosis by resident interstitial macrophages, and may travel to perivenous,
5 peribronchiolar or subpleural sites, where they become trapped, increasing particle burden. The
6 migration and grouping of particles and macrophages within the lungs can lead to the
7 redistribution of initially diffuse deposits into focal aggregates. Some particles or components
8 can bind to epithelial cell membranes or macromolecules, or to other cell components, delaying
9 clearance from the lungs.
10 Churg and Brauer (1997) examined lung autopsy tissue from 10 never-smokers from
11 Vancouver, Canada. They noted that the geometric mean particle diameter (GMPD) in lung
12 parenchymal tissue was 0.38 //m (og = 2.4). Ultrafme particles accounted for less than 5% of the
13 total retained particulate matter. Metal particles had a GMPD of 0.17 //m, and silicates 0.49 //m.
14 Ninety-six percent of retained PM was less than 2.5 //m. A subsequent study considered
15 retention of actual ambient particles in the lungs, which is related to deposition. Brauer et al.
16 (2001) showed that small particles could undergo significant steady-state retention within the
17 lungs. Using lungs obtained at autopsy from long-term, nonsmoking residents of an area having
18 high levels of ambient PM (Mexico City, Mexico) and those from an area with relatively low PM
19 levels (Vancouver, Canada), the investigators measured the particle concentration per gram of
20 lung within the parenchyma. They found that living in the high PM region resulted in
21 significantly greater retention of both fine and ultrafine particles within the lungs; levels in the
22 lungs from Mexico City contained over 7.4 times the concentration of these particles as did the
23 lungs from residents of Vancouver. These results indicate a clear relationship between ambient
24 exposure concentration and retention in the A region.
25 Clearance by the absorptive mechanism involves dissolution in the alveolar surface fluid,
26 followed by transport through the epithelium and into the interstitium, and then diffusion into the
27 lymph or blood. Solubility is influenced by the particle's surface to volume ratio and other
28 properties, such as hydrophilicity and lipophilicity (Mercer, 1967; Morrow, 1973; Patton, 1996).
29
30
31
April 2002 6-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.3.2 Clearance Kinetics
2 The kinetics of clearance have been reviewed in U.S. Environmental Protection Agency
3 (1996) and in a number of monographs (e.g., Schlesinger et al., 1997) and are discussed only
4 briefly here. The actual time frame over which clearance occurs affects the cumulative dose
5 delivered to the respiratory tract, as well as the dose delivered to extrapulmonary organs.
6
7 6.3.2.1 Extrathoracic Region
8 Mucus flow rates in the posterior nasal passages are highly nonuniform, but the median rate
9 in a healthy adult human is about 5 mm/min, resulting in a mean anterior to posterior transport
10 time of about 10 to 20 min for poorly soluble particles (Rutland and Cole, 1981; Stanley et al.,
11 1985). Particles deposited in the anterior portion of the nasal passages are cleared more slowly
12 by mucus transport and are usually more effectively removed by sneezing, wiping, or nose
13 blowing (Fry and Black, 1973; Morrow, 1977).
14
15 6.3.2.2 Tracheobronchial Region
16 Mucus transport in the tracheobronchial tree occurs at different rates in different local
17 regions; the velocity of movement is fastest in the trachea, and it becomes progressively slower
18 in more distal airways. In healthy nonsmoking humans, using noninvasive procedures and no
19 anesthesia, average tracheal mucus transport rates have been measured at 4.3 to 5.7 mm/min
20 (Yeates et al., 1975, 1981; Foster et al., 1980; Leikauf et al., 1981, 1984); whereas that in the
21 main bronchi has been measured at -2.4 mm/min (Foster et al., 1980). Estimates for human
22 medium bronchi range between 0.2 to 1.3 mm/min; whereas those in the most distal ciliated
23 airways range down to 0.001 mm/min (Morrow et al., 1967; Cuddihy and Yeh, 1988; Yeates and
24 Aspin, 1978).
25 The total duration of bronchial clearance or some other time parameter often is used as an
26 index of mucociliary kinetics. Although clearance from the TB region is generally rapid, there is
27 experimental evidence, discussed in U.S. Environmental Protection Agency (1996), that a
28 fraction of material deposited in the TB region is retained much longer than the 24 h commonly
29 used as the outer range of clearance time for particles within this region (Stahlhofen et al.,
30 1986a,b; Scheuch and Stahlhofen, 1988; Smaldone et al., 1988). A study by Asgharian et al.
31 (2001) showed that it is not necessary to invoke a slow- and fast-phase for TB clearance to have
April 2002 6-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles retained longer than 24 h. Based upon asymmetric stochastic human lung modeling
2 data, intersubject variability in retained mass arising from the periphery of the TB can explain the
3 experimental observations while still fitting a single compartment clearance model. Other
4 studies described below, however, do support the concept that TB regional clearance consists of
5 both a fast and a slow component.
6 Falk et al. (1997) studied clearance in healthy adults using monodisperse Teflon particles
7 (6.2 //m) inhaled at two flow rates. A considerable fraction (about 50%) of particles deposited in
8 small airways had not cleared within 24 h following exposure. These particles cleared with a
9 half time of 50 days. Although the deposition sites of the particles were not confirmed
10 experimentally, calculations suggested these to be in the smaller ciliated airways. Camner et al.
11 (1997) also noted that clearance from the TB region was incomplete by 24 h postexposure and
12 suggested that this may be caused by incomplete clearance from bronchioles. Healthy adults
13 inhaled teflon particles (6, 8, and 10 //m) under low flow rates to maximize deposition in the
14 small ciliated airways. The investigators noted a decrease in 24-h retention with increasing
15 particle size, indicating a shift toward either a smaller retained fraction, deposition more
16 proximally in the respiratory tract, or both. They calculated that a large fraction, perhaps as high
17 as 75% of particles depositing in generations 12 through 16, was still retained at 24 h
18 postexposure.
19 In a study to examine retention kinetics in the tracheobronchial tree (Falk et al., 1999),
20 nonsmoking healthy adults inhaled radioactively tagged 6. l-//m particles at both a normal flow
21 rate and a slow flow rate designed to deposit particles preferentially in small ciliated airways.
22 Lung retention was measured from 24 h to 6 mo after exposure. Following normal flow rate
23 inhalation, 14% of the particles retained at 24 h cleared with a half time of 3.7 days and 86%
24 with a half time of 217 days. Following slow flow rate inhalation, 35% of the particles retained
25 at 24 h cleared with a half time of 3.6 days and 65% with a half time of 170 days. Estimates
26 using a number of mathematical models indicated higher deposition in the bronchiolar region
27 (generations 9 through 15) with the slow rate inhalation compared to the normal rate. The
28 experimental data and predictions of the deposition modeling indicated that 40% of the particles
29 deposited in the conducting airways during the slow inhalation were retained after 24 h. The
30 particles that cleared with the shorter half time were mainly deposited in the bronchi olar region,
April 2002 6-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 but only about 25% of the particles deposited in this region cleared in this phase. This study
2 provided additional confirmation for a phase of slow clearance from the bronchial tree.
3 The underlying sites and mechanisms of long-term TB retention in the smaller airways are
4 not known. Some proposals were presented in the earlier 1996 PM AQCD (U.S. Environmental
5 Protection Agency, 1996). This slow clearing tracheobronchial compartment likely is associated
6 with bronchioles <1 mm in diameter (Lay et al., 1995; Kreyling et al., 1999; Falk et al., 1999).
7 Based on a study in which an adrenergic agonist was used to stimulate mucus flow, so as to
8 examine the role of mucociliary transport in the bronchioles, it was found that clearance from the
9 smaller airways was not influenced by the drug, suggesting to the investigators that mucociliary
10 transport was not as an effective clearance mechanism from this region as it is in larger airways
11 (Svartengren et al., 1998, 1999). Although slower or less effective mucus transport may result in
12 longer retention times in small airways, other factors may account for long-term TB retention.
13 One such proposal is the movement of particles into the gel phase because of surface tension
14 forces in the liquid lining of the small airways (Gehr et al., 1990, 1991). The issue of particle
15 retention in the tracheobronchial tree certainly is not resolved.
16 Long-term TB retention patterns are not uniform. There is an enhancement at bifurcation
17 regions (Radford and Martell, 1977; Henshaw and Fews, 1984; Cohen et al., 1988), the likely
18 result of both greater deposition and less effective mucus clearance within these areas. Thus,
19 doses calculated based on uniform surface retention density may be misleading, especially if the
20 material is lexicologically slow acting.
21
22 6.3.2.3 Alveolar Region
23 Particles deposited in the A region generally are retained longer than are those deposited in
24 airways cleared by mucociliary transport. There are limited data on alveolar clearance rates in
25 humans. Within any species, reported clearance rates vary widely because, in part, of different
26 properties of the particles used in the various studies. Furthermore, some chronic experimental
27 studies have employed high concentrations of poorly soluble particles that may have interfered
28 with normal clearance mechanisms, resulting in clearance rates different from those that would
29 typically occur at lower exposure levels. Prolonged exposure to high particle concentrations is
30 associated with what is termed particle "overload." This is discussed in greater detail in
31 Section 6.4.
April 2002 6-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 There are numerous pathways of A region clearance, and the utilization of these may
2 depend on the nature of the particles being cleared. Little is known concerning relative rates
3 along specific pathways. Thus, generalizations about clearance kinetics are difficult to make.
4 Nevertheless, A region clearance is usually described as a multiphasic process, each phase
5 considered to represent removal by a different mechanism or pathway and often characterized by
6 increased retention half times following toxicant exposure.
7 The initial uptake of deposited particles by alveolar macrophages is very rapid and
8 generally occurs within 24 h of deposition (Lehnert and Morrow, 1985; Naumann and
9 Schlesinger, 1986; Lay et al., 1998). The time for clearance of particle-laden alveolar
10 macrophages via the mucociliary system depends on the site of uptake relative to the distal
11 terminus of the mucus blanket at the bronchiolar level. Furthermore, clearance pathways and
12 subsequent kinetics may depend to some extent on particle size. For example, some smaller
13 ultrafine particles (< 0.02 //m) may be less effectively phagocytosed than larger ones
14 (Oberdorster, 1993).
15 Uningested particles may penetrate into the interstitium within a few hours following
16 deposition. This transepithelial passage seems to increase as particle loading increases,
17 especially to that level above which macrophage numbers increase (Ferin, 1977; Ferin et al.,
18 1992; Adamson and Bowden, 1981). It also may be particle size dependent, because insoluble
19 ultrafine particles (<0.1 //m diameter) of low intrinsic toxicity show increased access to the
20 interstitum and greater lymphatic uptake than do larger particles of the same material
21 (Oberdorster et al., 1992; Ferin et al., 1992). However, ultrafine particles of different materials
22 may not enter the interstitium to the same extent. Similarly, a depression of phagocytic activity,
23 a reduction in macrophage ability to migrate to sites of deposition (Madl et al., 1998), or the
24 deposition of large numbers of ultrafine particles may increase the number of free particles in the
25 alveoli, perhaps enhancing removal by other routes. In any case, free particles may reach the
26 lymph nodes perhaps within a few days after deposition (Lehnert et al., 1988; Harmsen et al.,
27 1985) although this route is not definitive and may be species dependent.
28 The extent of lymphatic uptake of particles may depend on the effectiveness of other
29 clearance pathways, in that lymphatic translocation likely increases when phagocytic activity of
30 alveolar macrophages decreases. This may be a factor in lung overload. However, it seems that
31 the deposited mass or number of particles must exceed some threshold below which increases in
April 2002 6-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 loading do not affect translocation rate to the lymph nodes (Ferin and Feldstein, 1978; LaBelle
2 and Brieger, 1961). In addition, the rate of translocation to the lymphatic system may be
3 somewhat particle size dependent. Although no human data are available, translocation of latex
4 particles to the lymph nodes of rats was greater for 0.5- to 2-//m particles than for 5- and 9-//m
5 particles (Takahashi et al., 1992), and particles within the 3- to 15-//m size range were found to
6 be translocated at faster rates than were larger sizes (Snipes and Clem, 1981). On the other hand,
7 translocation to the lymph nodes was similar for both 0.4-//m barium sulfate or 0.02-//m gold
8 colloid particles (Takahashi et al., 1987). It seems that particles <2 //m clear to the lymphatic
9 system at a rate independent of size; and it is particles of this size, rather than those >5 //m, that
10 would have significant deposition within the A region following inhalation. In any case, the
11 normal rate of translocation to the lymphatic system is quite slow; and elimination from the
12 lymph nodes is even slower, with half times estimated in tens of years (Roy, 1989).
13 Soluble particles depositing in the A region may be cleared rapidly via absorption through
14 the epithelial surface into the blood. Actual rates depend on the size of the particle (i.e., solute
15 size), with smaller molecular weight solutes clearing faster than larger ones. Absorption may be
16 considered as a two stage process, with the first stage being dissociation of the deposited
17 particles into material that can be absorbed into the circulation (i.e., dissolution) and the second
18 stage being uptake of this material. Each of these stages may be time dependent. The rate of
19 dissolution depends on a number of factors, including particle surface area and chemical
20 structure. A portion of the dissolved material may be absorbed more slowly because of binding
21 to respiratory tract components. Accordingly, there is a very wide range for absorption rates,
22 depending on the physicochemical properties of the material deposited.
23 As indicated in both the toxicology and epidemiology chapters of this document (Chapters
24 7 and 8), one of the health outcome of concern relates to ambient PM effects on the
25 cardiovascular system. Thus, an important dosimetric issue involves the pathways by which
26 inhaled and deposited particles in the lungs could impact upon extrapulmonary systems.
27 Clearance and translocation pathways by which this may occur have been recently described.
28 Nemmar et al. (2001) instilled hamsters with radioactively-labeled colloidal albumin particles
29 (diameter < 0.080 //m) as a model for ambient ultrafine particles and measured the label
30 appearing in systemic blood and various extrapulmonary organs up to 1 h postexposure. They
31 found label in blood within 5 minutes after instillation. In their subsequent studies in which
April 2002 6-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 healthy volunteers were challenged with inhalation of 99mTechnitum-labeled ultrafine carbon
2 particles (Nemmar et al., 2002), the radioactivity was detected in blood as early as 1 min,
3 reaching a maximum between 10 and 20 min after inhalation of the aerosol. While label was
4 also noted in the other extrapulmonary organs examined (namely liver, heart, spleen, kidneys,
5 and brain), the liver had the highest levels and these increased with increasing time postexposure,
6 while the second highest levels were noted in the heart or kidney, depending upon the instilled
7 concentration. This suggests that ultrafine particles can rapidly diffuse from the lungs into the
8 systemic circulation, thus providing a pathway by which ambient PM may rapidly affect
9 extrapulmonary organs.
10 In another study, Takenaka et al. (2001) exposed rats by inhalation to 0.015 //m particles of
11 elemental silver and found evaluated levels of these particles in various extrapulmonary organs
12 up to 7 days postexposure. They found that the amount of particles in the lungs decreased
13 rapidly with time and, by day 7, only about 4% of the initial lung burden remained. At day 0,
14 particles were already found in the blood. The particles were found to be distributed in the liver,
15 kidney, heart, and brain by 1 day postexposure. The particle concentration was highest in the
16 kidney, followed by the liver, and then the heart. This study also indicates that inhaled ultrafine
17 particles were rapidly cleared from the lungs into the systemic circulation. However, a similar
18 cleance pattern was found after intratracheal instillation of AgNO3 solution. Therefore, the
19 investigators postulated that the rapid clearance of elemental silver particles was due to a fast
20 dissolution of ultrafine silver particles into the lung fluid and subsequent diffusion into the blood
21 stream, although a possibility of direct translocation of solid particles into the blood stream was
22 not excluded. The investigators also instilled an aqueous suspension of elemental silver into
23 some animals; in this case, there was more retention in the lungs, which was ascribed to
24 phagocytic accumulation of agglomerated particles in alveolar macrophages and slow dissolution
25 of particles in cells. Thus, this study also suggested that particle size and the tendency of
26 particles to aggregate can affect the translocation pathway from the lungs. Earlier studies
27 (Huchon et al., 1987; Peterson et al., 1989; Morrison et al., 1998) investigated lung clearance of
28 labeled macromolecule solutes with widely varying molecular weight and labeled albumin, as
29 well as albumin ultrafine aggregates. Clearance rates found from these earlier studies were much
30 slower than recent studies described above, suggesting that the possibility of a fast clearing
31 pathway of solid ultrafine particles may need further study.
April 2002 6-49 DRAFT-DO NOT QUOTE OR CITE
-------
l 6.3.3 Interspecies Patterns of Clearance
2 The inability to study the retention of certain materials in humans for direct risk assessment
3 requires use of laboratory animals. Because dosimetry depends on clearance rates and routes,
4 adequate toxicologic assessment necessitates that clearance kinetics in such animals be related to
5 those in humans. The basic mechanisms and overall patterns of clearance from the respiratory
6 tract are similar in humans and most other mammals. However, regional clearance rates can
7 show substantial variation between species, even for similar particles deposited under
8 comparable exposure conditions, as extensively reviewed elsewhere (U.S. Environmental
9 Protection Agency, 1996; Schlesinger et al., 1997; Snipes et al., 1989).
10 In general, there are species-dependent rate constants for various clearance pathways.
11 Differences in regional and total clearance rates between some species are a reflection of
12 differences in mechanical clearance processes. For example, the relative proportion of particles
13 cleared from the A region in the short- and longer-term phases differs between laboratory rodents
14 and larger mammals, with a greater percentage cleared in the faster phase in rodents. A recent
15 study (Oberdorster et al., 1997) showed interstrain differences in mice and rats in the handling of
16 particles by alveolar macrophages. Macrophages of B6C3F1 mice could not phagocytize 10-//m
17 particles, but those of C57 black/61 mice did. In addition, the nonphagocytized 10-//m particles
18 were efficiently eliminated from the alveolar region; whereas previous work in rats found that
19 these large particles, after uptake by macrophages, were retained persistently (Snipes and Clem,
20 1981; Oberdorster et al., 1992). The ultimate implication of interspecies differences in clearance
21 needing to be considered in assessing particle dosimetry is that the retention of deposited
22 particles can differ between species and may result in differences in response to similar PM
23 exposure atmospheres.
24 Hsieh and Yu (1998) summarized the existing data on pulmonary clearance of inhaled,
25 poorly soluble particles in the rat, mouse, guinea pig, dog, monkey, and human. Clearance at
26 different initial lung burdens, ranging from 0.001 to 10 mg particles/g lung, was analyzed using a
27 two-phase exponential decay function. Two clearance phases in the alveolar region, namely fast
28 and slow, were associated with mechanical clearance along two pathways, the former with the
29 mucociliary system and the latter with the lymph nodes. Rats and mice were noted to be fast
30 clearers in comparison to the other species. Increasing the initial lung burden resulted in an
31 increasing mass fraction of particles cleared by the slower phase. As lung burden increased
April 2002 6-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 beyond 1 mg particles/g lung, the fraction cleared by the slow phase increased to almost 100%
2 for all species. However, the rate for the fast phase was similar in all species and did not change
3 with increasing lung burden of particles; whereas the rate for the slow phase decreased with
4 increasing lung burden. At elevated burdens, the effect on clearance rate was greater in rats than
5 in humans, an observation consistent with previous findings (Snipes, 1989).
6
7 6.3.4 Factors Modulating Clearance
8 A number of factors have previously been assessed in terms of modulation of normal
9 clearance patterns, including: age, gender, workload, disease, and irritant inhalation. Such
10 factors have been discussed in detail previously (U.S. Environmental Protection Agency, 1996).
11
12 6.3.4.1 Age
13 Studies previously described in the 1996 PM AQCD (U.S. Environmental Protection
14 Agency, 1996) indicated that there appeared to be no clear evidence for any age-related
15 differences in clearance from the lung or total respiratory tract, either from child to adult, or
16 young adult to elderly. Studies of mucociliary function have shown either no changes or some
17 slowing in mucous clearance function with age after maturity, but at a rate that would be unlikely
18 to significantly affect overall clearance kinetics.
19
20 6.3.4.2 Gender
21 Previously reviewed studies (U.S. Environmental Protection Agency, 1996) indicated no
22 gender-related differences in nasal mucociliary clearance rates in children (Passali and Bianchini
23 Ciampoli, 1985) nor in tracheal transport rates in adults (Yeates et al., 1975).
24
25 6.3.4.3 Physical Activity
26 The effect of increased physical activity on mucociliary clearance is unresolved, with
27 previously discussed studies (U.S. Environmental Protection Agency, 1996) indicating either no
28 effect or an increased clearance rate with exercise. There are no data concerning changes in
29 A region clearance with increased activity levels. Breathing with an increased tidal volume was
30 noted to increase the rate of particle clearance from the A region, and this was suggested to result
31 from distension-related evacuation of surfactant into proximal airways, resulting in a facilitated
April 2002 6-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 movement of particle-laden macrophages or uningested particles because of the accelerated
2 motion of the alveolar fluid film (John et al., 1994).
3
4 6.3.4.4 Respiratory Tract Disease
5 Various respiratory tract diseases are associated with clearance alterations. Evaluation of
6 clearance in individuals with lung disease requires careful interpretation of results, because
7 differences in deposition of particles used to assess clearance function may occur between
8 normal individuals and those with disease; this would impact directly on the measured clearance
9 rates, especially in the tracheobronchial tree. Earlier studies reported in the 1996 PM AQCD
10 (U.S. Environmental Protection Agency, 1996) noted findings of (a) slower nasal mucociliary
11 clearance in humans with chronic sinusitis, bronchiectasis, rhinitis, or cystic fibrosis and (b)
12 slowed bronchial mucus transport associated with bronchial carcinoma, chronic bronchitis,
13 asthma, and various acute respiratory infections. However, a recent study by Svartengren et al.
14 (1996a) concluded, based on deposition and clearance patterns, that particles cleared equally
15 effectively from the small ciliated airways of healthy humans and those with mild to moderate
16 asthma; but, this similarity was ascribed to effective therapy for the asthmatics.
17 In another study, Svartengren et al. (1996b) examined clearance from the TB region in
18 adults with chronic bronchitis who inhaled 6-//m Teflon particles. Based on calculations,
19 particle deposition was assumed to be in small ciliated airways at low flow and in larger airways
20 at higher flow. The results were compared to those obtained in healthy subjects from other
21 studies. At low flow, a larger fraction of particles was retained over 72 h in people with chronic
22 bronchitis compared to healthy subjects, indicating that clearance resulting from spontaneous
23 cough could not fully compensate for impaired mucociliary transport in small airways. For larger
24 airways, patients with chronic bronchitis cleared a larger fraction of the deposited particles over
25 72 h than did healthy subjects, but this was reportedly because of differences in deposition
26 resulting from airway obstruction.
27 An important mechanism of clearance from the tracheobronchial region, under some
28 circumstances, is cough. Although cough can be a reaction to an inhaled stimulus, in most
29 individuals with respiratory infections and disease, spontaneous coughing also serves to clear the
30 upper bronchial airways by dislodging mucus from the airway surface. Recent studies confirm
31 that this mechanism likely plays a significant role in clearance for people with mucus
April 2002 6-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 hypersecretion, at least for the upper bronchial tree, and for a wide range of deposited particle
2 sizes (0.5 to 5 //m) (Toms et al., 1997; Groth et al., 1997). There appears to be a general trend
3 towards an association between the extent (i.e., number) of spontaneous coughs and the rate of
4 particle clearance, with faster clearance being associated with a greater number of coughs (Groth
5 et al., 1997). Thus, recent evidence continues to support cough as an adjunct to mucociliary
6 movement in the removal of particles from the lungs of individuals with COPD. However, some
7 recent evidence suggests that, like mucociliary function, cough-induced clearance may become
8 depressed with worsening airway disease. Noone et al. (1999) found that the efficacy of
9 clearance via cough in patients with primary ciliary dyskinesia (who rely on coughing for
10 clearance because of immotile cilia) correlated with lung function (FEV1), in that decreased
11 cough clearance was associated with decreased percentage of predicted FEV1.
12 Earlier reported studies (U.S. Environmental Protection Agency, 1996) indicated that rates
13 of A region particle clearance were reduced in humans with chronic obstructive lung disease and
14 in laboratory animals with viral infections; whereas the viability and functional activity of
15 macrophages were impaired in human asthmatics and in animals with viral-induced lung
16 infections. However, any modification of functional properties of macrophages appears to be
17 injury-specific, in that they reflect the nature and anatomic pattern of disease.
18 One factor that may affect clearance of particles is the integrity of the epithelial surface
19 lining of the lungs. Damage or injury to the epithelium may result from disease or from the
20 inhalation of chemical irritants. Earlier studies performed with particle instillation had shown
21 that alveolar epithelial damage in mice at the time of deposition resulted in increased
22 translocation of inert carbon to pulmonary interstitial macrophages (Adamson and Hedgecock,
23 1995). A similar response was observed in a more recent assessment (Adamson and Prieditis,
24 1998), whereby silica (<0.3 //m) was instilled into a lung having alveolar epithelial damage (as
25 evidenced by increased permeability) and particles were noted to reach the interstitium and
26 lymph nodes.
27
28
29 6.4 PARTICLE OVERLOAD
30 Experimental studies using some laboratory rodents have employed high exposure
31 concentrations of relatively nontoxic, poorly soluble particles. These particle loads interfered
April 2002 6-53 DRAFT-DO NOT QUOTE OR CITE
-------
1 with normal clearance mechanisms, producing clearance rates different from those that would
2 occur at lower exposure levels. Prolonged exposure to high particle concentrations is associated
3 with a phenomenon that has been termed particle "overload", defined as the overwhelming of
4 macrophage-mediated clearance by the deposition of particles at a rate that exceeds the capacity
5 of that clearance pathway. It has been suggested that, in the rat, overload is more dependent
6 upon the volume rather than the mass of particles (Tran et al., 2000) and that volumetric
7 overloading will begin when particle retention approaches 1 mg particles/g lung tissue (Morrow,
8 1988). Overload is a nonspecific effect noted in experimental studies using many different kinds
9 of poorly soluble particles and results in A region clearance slowing or stasis, with an associated
10 chronic inflammation and aggregation of macrophages in the lungs and increased translocation of
11 particles into the interstitium.
12 The relevance of lung overload to humans exposed to poorly soluble, nonfibrous particles
13 remains unclear. Although it is likely to be of little relevance for most "real world" ambient
14 exposures, it may be of concern in interpreting some long-term experimental exposure data and,
15 perhaps, also for occupational exposures. For example, it has been suggested that a condition
16 called progressive massive fibrosis, which is unique to humans, has features indicating that dust
17 overload is a factor in its pathogenesis (Green, 2000). This condition is associated with
18 cumulative dust exposure and impaired clearance and can occur following high exposure
19 concentrations associated with occupational situations. In addition, any relevance to humans is
20 clouded by the suggestion that macrophage-mediated clearance is normally slower, and perhaps
21 of less relative importance in overall clearance, in humans than in rats (Morrow, 1994), and that
22 there can be significant differences in macrophage loading between species. On the other hand,
23 overload may be a factor in individuals with compromised lungs even under normal exposure
24 conditions. Thus, it has been hypothesized (Miller et al., 1995) that localized overload of particle
25 clearance mechanisms in people with compromised lung status may occur, whereby clearance is
26 overwhelmed and results in morbidity or mortality from particle exposure.
27
28
29
April 2002 6-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.5 COMPARISON OF DEPOSITION AND CLEARANCE PATTERNS
2 OF PARTICLES ADMINISTERED BY INHALATION AND
3 INTRATRACHEAL INSTILLATION
4 The most relevant exposure route by which to evaluate the toxicity of paniculate matter is
5 inhalation. However, many toxicological studies deliver particles by intratracheal instillation.
6 This latter technique has been used because it is easy to perform, requires significantly less effort,
7 cost, and amount of test material than does inhalation, and can deliver a known, exact dose of a
8 toxicant to the lungs. Because particle disposition is a determinant of dose, it is important to
9 compare deposition and clearance of particles delivered by these two routes in order to evaluate
10 the relevance of studies using instillation. However, in most instillation studies, the effect of this
11 route of administration on particle deposition and clearance per se was not examined. Although
12 these parameters were evaluated in some studies, it has been very difficult to compare particle
13 deposition/clearance between different inhalation and instillation studies because of differences
14 in experimental procedures and in the manner by which particle deposition/clearance was
15 quantitated. A recent paper provides a detailed evaluation of the role of instillation in respiratory
16 tract dosimetry and toxicology studies (Driscoll et al., 2000); and a short summary derived from
17 this paper is provided below in this section.
18 The pattern of initial regional deposition is strongly influenced by the exposure technique
19 used. Furthermore, the patterns within specific respiratory tract regions also are influenced in
20 this regard. Depending on particle size, inhalation results in varying degrees of deposition within
21 the ET airways, a region that is completely bypassed by instillation. Thus, differences in amount
22 of particles deposited in the lower airways will occur between the two procedures, especially for
23 those particles in the coarse mode. This is important if inhaled particles in ambient air affect the
24 upper respiratory tract and such responses are then involved in the evaluation of health outcomes.
25 Exposure technique also influences the intrapulmonary distribution of particles, which
26 potentially would affect routes and rates of ultimate clearance from the lungs and dose delivered
27 to specific sites within the respiratory tract or to extrapulmonary organs. Intratracheal instillation
28 tends to disperse particles fairly evenly within the TB region but can result in heterogeneous
29 distribution in the A region; whereas inhalation tends to produce a more homogeneous
30 distribution throughout the major conducting airways as well as the A region for the same
31 particles. Thus, inhalation results in a randomized distribution of particles within the lungs;
April 2002 6-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 whereas intratracheal instillation produces an heterogeneous distribution, in that the periphery of
2 the lung receives little particle load and most of the instilled particles are found in regions that
3 have a short path length from the major airways. Furthermore, inhalation results in greater
4 deposition in apical areas of the lungs and less in basal areas; whereas intratracheal instillation
5 results in less apical than basal deposition. Thus, toxicological effects from instilled materials
6 may not represent those which would occur following inhalation, due to differences in sites of
7 initial deposition following exposure. In addition, instillation studies generally deliver high
8 doses to the lungs, much higher than those which would occur with realistic inhalation exposure.
9 This would also clearly affect the initial dose delivered to target tissue and its relevance to
10 ambient exposure.
11 Comparison of the kinetics of clearance of particles administered by instillation or
12 inhalation have shown similarities, as well as differences, in rates for different clearance phases,
13 depending on the exposure technique used (Oberdorster et al., 1997). However, some of the
14 differences in kinetics may be explained by differences in the initial sites of deposition. One of
15 the maj or pathways of clearance involves particle uptake and removal via pulmonary
16 macrophages. Dorries and Valberg (1992) noted that inhalation resulted in a lower percentage of
17 particles recovered in lavaged cells and a more even distribution of particles among
18 macrophages. More individual cells received measurable amounts of particles via inhalation than
19 via intratracheal instillation; whereas with the latter, many cells received little or no particles and
20 others received very high burdens. Furthermore, with intratracheal instillation, macrophages at
21 the lung periphery contained few, if any, particles; whereas cells in the regions of highest
22 deposition were overloaded, reflecting the heterogeneity of particle distribution when particles
23 are administered via instillation. Also, both the relative number of particles phagocytized by
24 macrophages as well as the percentage of these cells involved in phagocytosis is affected by the
25 burden of administered particles, which is clearly different in instillation and inhalation (Suarez
26 et al., 2001). Thus, when guinea pigs were administered latex microspheres (1.52-3.97 //m
27 MMAD) by inhalation or instillation, the percentage of cells involved in phagocytosis, as well as
28 the amount of particles per cell, were both significantly higher with the latter route. The route of
29 exposure, therefore, influences particle distribution in the macrophage population and could, by
30 assumption, influence clearance pathways and clearance kinetics.
April 2002 6-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 In summary, inhalation may result in deposition within the ET region, the extent of which
2 depends on the size of the particles used. Of course, intratracheal instillation bypasses this
3 portion of the respiratory tract and delivers particles directly to the tracheobronchial tree.
4 Although some studies indicate that short (0 to 2 days) and long (100 to 300 days postexposure)
5 phases of clearance of insoluble particles delivered either by inhalation or intratracheal
6 instillation are similar, other studies indicate that the percentage retention of particles delivered
7 by instillation is greater than that for inhalation at least up to 30 days postexposure. Thus, there
8 is some inconsistency in this regard.
9 Perhaps the most consistent conclusion regarding differences between inhalation and
10 intratracheal instillation is related to the intrapulmonary distribution of particles. Inhalation
11 generally results in a fairly homogeneous distribution of particles throughout the lungs. On the
12 other hand, instillation results in a heterogeneous distribution, especially within the alveolar
13 region, and focally high concentrations of particles. The bulk of instilled material penetrates
14 beyond the major tracheobronchial airways, but the lung periphery is often virtually devoid of
15 particles. This difference is reflected in particle burdens within macrophages, with those from
16 animals inhaling particles having more homogeneous burdens and those from animals with
17 instilled particles showing groups of cells with no particles and others with heavy burdens. This
18 difference impacts on clearance pathways, dose to cells and tissues, and systemic absorption.
19 Exposure method, thus, clearly influences dose distribution.
20
21
22 6.6 MODELING THE DISPOSITION OF PARTICLES IN THE
23 RESPIRATORY TRACT
24 6.6.1 Modeling Deposition, Clearance, and Retention
25 Over the years, mathematical models for predicting deposition, clearance and, ultimately,
26 retention of particles in the respiratory tract have been developed. Such models help interpret
27 experimental data and can be used to make dosimetry predictions for cases where data are not
28 available. In fact, model predictions described below are estimates based on the best available
29 models at the time of publication and, except where noted, have not been verified by
30 experimental data.
April 2002 6-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 A review of various mathematical deposition models was given by Morrow and Yu (1993)
2 and in U.S. Environmental Protection Agency (1996). There are three major elements involved
3 in mathematical modeling. First, a structural model of the airways must be specified in
4 mathematical terms. Second, deposition efficiency in each airway must be derived for each of
5 the various deposition mechanisms. Finally, a computational procedure must be developed to
6 account for the transport and deposition of the particles in the airways. As noted earlier, most
7 models are deterministic, in that particle deposition probabilities are calculated using anatomical
8 and airflow information on an airway generation by airway generation basis. Other models are
9 stochastic, whereby modeling is performed using individual particle trajectories and finite
10 element simulations of airflow.
11 Recent reports involve modeling the deposition of ultrafine particles and deposition at
12 airway bifurcations. Zhang and Martonen (1997) used a mathematical model to simulate
13 diffusion deposition of ultrafine particles in the human upper tracheobronchial tree and compared
14 the results to those in a hollow cast obtained by Cohen et al. (1990). The model results were in
15 good agreement with experimental data. Zhang and Martonen (1997) studied the inertial
16 deposition of particles in symmetric three-dimensional models of airway bifurcations,
17 mathematically examining effects of geometry and flow. They developed equations for use in
18 predicting deposition based on Stokes numbers, Reynolds numbers, and bifurcation angles for
19 specific inflows.
20 Models for deposition, clearance, and dosimetry of the respiratory tract of humans have
21 been available for the past four decades. For example, the International Commission on
22 Radiological Protection (ICRP) has recommended three different mathematical models during
23 this time period (International Commission on Radiological Protection, 1960, 1979, 1994).
24 These models make it possible to calculate the mass deposition and retention in different parts of
25 the respiratory tract and provide, if needed, mathematical descriptions of the translocation of
26 portions of the deposited material to other organs and tissues beyond the respiratory tract.
27 A somewhat simplified variation of the 1994 ICRP dosimetry model was used by Snipes et al.
28 (1997) to predict average particle deposition in the ET, T and A regions and retention patterns in
29 the A region, under a repeated exposure situation for two characterized environmental aerosols
30 obtained from Philadelphia, PA and Phoenix, AZ. Both of these aerosols had both fine and
31 coarse particles. They found similar retention for the fine particles in both aerosols, but
April 2002 6-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 significantly different retention for the coarse-mode particles. Because the latter type dominated
2 the aerosol in the Phoenix sample, this type of evaluation can be used to improve understanding
3 of the relationship between exposures to ambient PM and retention patterns that affect health
4 endpoints in residents of areas in which the particle distributions and, therefore, the particle
5 chemistry may differ.
6 A morphological model based on laboratory data from planar gamma camera and single-
7 photon emission tomography images has been developed (Martonen et al., 2000). This model
8 defines the parenchymal wall in mathematical terms, divides the lung into distinct left and right
9 components, derives a set of branching angles from experimental measurements, and confines
10 the branching network within the left and right components (so there is no overlapping of
11 airways). The authors conclude that this more physiologically realistic model can be used to
12 calculate PM deposition patterns for risk assessment.
13 Musante and Martonen (2000c) developed an age-dependent theoretical model to predict
14 dosimetry in the lungs of children. The model comprises dimensions of individual airways and
15 geometry of branching airway networks within developing lungs and breathing parameters as a
16 function of age. The model suggests that particle size, age, and activity level markedly affect
17 deposition patterns of inhaled particles. Simulations thus far predict a lung deposition fraction of
18 38% in an adult and 73% (nearly twice as high) in a 7-mo-old for 2 //m particles inhaled during
19 heavy breathing. The authors conclude that this model will be useful for estimating dose
20 delivered to sensitive subpopulations, such as children.
21 Segal et al. (2000a) developed a computer model, noted earlier, for airflow and particle
22 motion in the lungs of children to study how airway disease, specifically cancer, affects inhaled
23 PM deposition. The model considers how tumor characteristics (size and location) and
24 ventilatory parameters (breathing rates and tidal volumes) influence particle trajectories and
25 deposition patterns. The findings indicate that PM may be deposited on the upstream surfaces of
26 tumors because of enhanced efficiency of inertial impaction. Also, submicron particles and
27 larger particles, respectively, may be deposited on the downstream surfaces of tumors because of
28 enhanced efficiency of diffusion and sedimentation. The mechanisms of diffusion and
29 sedimentation are functions of the particle residence times in airways. Eddies downstream of
30 tumors would trap particles and allow more time for deposition to occur by diffusion and
April 2002 6-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 sedimentation. The authors conclude that particle deposition is complicated by the presence of
2 airway disease, but that the effects are systematic and predictable.
3 Segal et al. (2000b) have used a traditional mathematical model based on Weibel's lung
4 morphology and calculated total lung deposition fraction of 1 to 5 //m diameter particles in
5 healthy adults. Airway dimensions were scaled by individual lung volume. Deposition
6 predictions were made with both plug flow and parabolic flow profiles in the airways. The
7 individualized airway dimension improved the accuracy of the predicted values when compared
8 with experimental data. There were significant differences, however, between the model
9 predictions and experimental data depending on the flow profiles used, indicating that use of
10 more realistic parameters is essential to improving the accuracy of model predictions.
11 Broday and Georgopoulos (2001) presented a model that solves a variant of the general
12 dynamic equation for size evolution of respirable particles within human tracheobronchial
13 airways. The model considers polydisperse aerosols with respect to size but heterosperse with
14 respect to thermodynamic state and chemical composition. The aerosols have an initial bimodal
15 lognormal size distribution that evolves with time in response to condensation-evaporation and
16 deposition processes. Simulations reveal that submicron size particles grow rapidly and cause
17 increased number and mass fractions of the particle population to be found in the intermediate
18 size range. Because deposition by diffusion decreases with increasing size, hygroscopic fine
19 particles may persist longer in the inspired air than nonhygroscopic particles of comparable initial
20 size distribution. In contrast, the enhanced deposition probability of hygroscopic particles
21 initially from the intermediate size range increases their fraction deposited in the airways. The
22 model demonstrates that the combined effect of growth and deposition tends to decrease the
23 nonuniformity of the persistent aerosol, forming an aerosol which is characterized by size
24 distribution of smaller variance. These factors also alter the deposition profile along airways.
25 Lazaridis et al. (2001) developed a deposition model for humans that was designed to better
26 describe the dynamics of respirable particles within the airways. The model took into account
27 alterations in aerosol particle size and mass distribution that may result from processes such as
28 nucleation, condensation, coagulation, and gas phase chemical reactions. The airway geometry
29 used was the regular dichotomous model of Weibel, and it incorporated the influences of airway
30 boundary layers on particle dynamics, although simplified velocity profiles were used so as to
31 maintain a fairly uncomplicated description of respiratory physiology. Thus, this model was
April 2002 6-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 considered to be an improvement over previous models which did not consider either the effects
2 of boundary layers on both the airborne and deposited particles or the effects of gas-phase
3 transport processes, because it can account for the polydispersity, multimodality, and
4 heterogeneous composition of common ambient aerosols. The authors indicate that the model
5 predictions were both qualitatively and quantitatively consistent with experimental data for
6 particle deposition within the TB and A regions.
7 Another respiratory tract dosimetry model was developed concurrently with the ICRP
8 model by the National Council on Radiation Protection and Measurements (NCRP, 1997).
9 As with the ICRP model (International Commission on Radiological Protection, 1994), the
10 NCRP model addresses inhalability of particles, revised subregions of the respiratory tract,
11 dissolution-absorption as an important aspect of the model, and body size and age. The NCRP
12 model defines the respiratory tract in terms of a naso-oro-pharyngo-laryngeal (NOPL) region, a
13 tracheobronchial (TB) region, a pulmonary (P) region, and lung-associated lymph nodes (LN).
14 Deposition and clearance are calculated separately for each of these regions. As with the 1994
15 ICRP model, inhalability of aerosol particles is considered, and deposition in the various regions
16 of the respiratory tract is modeled using methods that relate to mechanisms of inertial impaction,
17 sedimentation, and diffusion.
18 Fractional deposition in the NOPL region was developed from empirical relationships
19 between particle diameter and air flow rate. Deposition in the TB and P regions were projected
20 from model calculations, based on geometric or aerodynamic particle diameter and physical
21 deposition mechanisms such as impaction, sedimentation, diffusion, and interception.
22 Deposition in the TB and P regions used the lung model of Yeh and Schum (1980) with a method
23 of calculation similar to that of Findeisen (1935) and Landahl (1950). This method was modified
24 to accomodate an adjustment of lung volume and substitution of realistic deposition equations.
25 These calculations were based on air flow information and idealized morphometry and used a
26 typical pathway model. Comparison of regional deposition fraction predictions between the
27 NCRP and ICRP models was provided in U.S. Environmental Protection Agency (1996). The
28 definition of inhalability was that of the American Conference of Governmental Industrial
29 Hygenists (1985). Breathing frequency, tidal volume, and functional residual capacity were the
30 ventilatory factors used to model deposition. These were related to body weight and to three
31 levels of physical activity, namely low activity, light exertion, and heavy exertion.
April 2002 6-61 DRAFT-DO NOT QUOTE OR CITE
-------
1 Clearance from all regions of the respiratory tract was considered to result from
2 competitive mechanical and absorptive mechanisms. Mechanical clearance in the NOPL and TB
3 regions was considered to result from mucociliary transport. This was represented in the model
4 as a series of escalators moving towards the glottis and where each airway had an effective
5 clearance velocity. Clearance from the P region was represented by fractional daily clearance
6 rates to the TB region, the pulmonary LN region, and the blood. A fundamental assumption in
7 the model was that the rates for absorption into blood were the same in all regions of the
8 respiratory tract; the rates of dissolution-absorption of particles and their constituents were
9 derived from clearance data primarily from laboratory animals. The effect of body growth on
10 particle deposition also was considered in the model, but particle clearance rates were assumed to
11 be independent of age. Some consideration for compromised individuals was incorporated into
12 the model by altering normal rates for the NOPL and TB regions.
13 Mathematical deposition models for a number of nonhuman species have been developed;
14 these were discussed previously in the 1996 PM AQCD (U.S. Environmental Protection Agency,
15 1996). Despite difficulties, modeling studies in laboratory animals remain a useful step in
16 extrapolating exposure-dose-response relationships from laboratory animals to humans.
17 Respiratory-tract clearance begins immediately upon deposition of inhaled particles. Given
18 sufficient time, the deposited particles may be removed completely by these clearance processes.
19 However, single inhalation exposures may be the exception rather than the rule. It generally is
20 accepted that repeated or chronic exposures are common for environmental aerosols. As a result
21 of such exposures, accumulation of particles may occur. Chronic exposures produce respiratory
22 tract burdens of inhaled particles that continue to increase with time until the rate of deposition is
23 balanced by the rate of clearance. This is defined as the "equilibrium respiratory tract burden".
24 It is important to evaluate these accumulation patterns, especially when assessing ambient
25 chronic exposures, because they dictate what the equilibrium respiratory tract burdens of inhaled
26 particles will be for a specified exposure atmosphere. Equivalent concentrations can be defined
27 as "species-dependent concentrations of airborne particles which, when chronically inhaled,
28 produce equal lung deposits of inhaled particles per gram of lung during a specified exposure
29 period" (Schlesinger et al., 1997). Available data and approaches to evaluate exposure
30 atmospheres that produce similar respiratory tract burdens in laboratory animals and humans
31 were discussed in detail in the 1996 PM AQCD.
April 2002 6-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 Several laboratory animal models have been developed to help interpret results from
2 specific studies that involved chronic inhalation exposures to nonradioactive particles (Wolff
3 et al., 1987; Strom et al., 1988; Stober et al., 1994). These models were adapted to data from
4 studies involving high level chronic inhalation exposures in which massive lung burdens of low
5 toxicity, poorly soluble particles were accumulated. Koch and Stober (2001) further adapted
6 clearance models for more relevant particle deposition in the pulmonary region. They published
7 a pulmonary retention model that accounts for dissolution and macrophage-mediated removal of
8 deposited polydisperse aerosol particles. The model provides a mathematical solution for the
9 size distribution of particles in the surfactant layer of the alveolar surface and in the cell plasma
10 of alveolar macrophages and accounts for the different kinetics and biological effects in the two
11 compartments. It does not, however, account for particle penetration to the lung interstitium and
12 particle clearance by the lymph system.
13 The multiple-path models of Anjilvel and Asgharian (1995) for rat lung and its extension
14 by Subramaniam et al. (1999) for human lungs describe a method for calculating a deposited
15 fraction for a specific size distribution based on a summary of published data on regional
16 deposition of different size particles. The method is based on constructing nomograms that are
17 used to estimate alveolar deposition fractions for three species (human, monkey, and rat). The
18 data are then incorporated into a regression model that calculates more exact deposition fractions
19 in the whole lung for monodisperse and polydisperse aerosols for ultrafme through coarse
20 particle sizes. The model is somewhat constrained at present because of limitations in the
21 underlying deposition database.
22 Tran et al. (1999) used a mathematical model of clearance and retention in the A region of
23 rats lungs to determine the extent to which a sequence of clearance mechanisms and pathways
24 could explain experimental data obtained from inhalation studies using relatively insoluble
25 particles. These pathways were phagocytosis by macrophages with subsequent clearance,
26 transfer of particles into the interstitium and to lymph nodes, and overloading of defense
27 mechanisms. The model comprised a description of the complete defense system in this region,
28 using both clearance and transfer processes represented by sets of equations. The authors suggest
29 that the model could be used to examine the consistency of various hypotheses concerning the
30 fate of inhaled particles and could be used for species other than the rat with appropriate scaling.
April 2002 6-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hofmann et al. (2000) used three different morphometric models of the rat lung to compute
2 particle deposition in the acinar (alveolar) airways: the multipath lung model (MPL) with a fixed
3 airway geometry; the stochastic lung (SL) model with a randomly selected branching structure;
4 and a hybrid of the MPL and SL models. They calculated total and regional deposition for a
5 range of particle sizes during quiet and heavy breathing. Although the total bronchial and acinar
6 deposition fractions were similar for the three models, the SL and the hybrid models predicted a
7 substantial variation in particle deposition among different acini. Acinar deposition variances in
8 the MPL model were consistently smaller than in the SL and the hybrid lung models. The
9 authors conclude that the similarity of acinar deposition variations in the latter two models and
10 their independence of the breathing pattern suggest that the heterogeneity of the acinar airway
11 structure is primarily responsible for the heterogeneity of acinar particle deposition.
12 The combination of MPL and SL models developed for the human lung takes into
13 consideration both intra- and inter-human variability in airway structure. The models also have
14 been developed to approximately the same level of complexity for laboratory animals and,
15 therefore, can be readily used for interspecies extrapolation (Asgharian et al., 1999). A variation
16 of these models will soon be developed for inclusion of the airway geometry of children. By the
17 incorporation of particle clearance in the TB region (Asgharian et al., 2001) and hopefully in the
18 alveolar region (Koch and Stober, 2001), this suite of models should prove to be very useful in
19 better predicting PM dosimetry in humans.
20
21 6.6.2 Models To Estimate Retained Dose
22 Models have been used routinely to express retained dose in terms of temporal patterns for
23 A region retention of acutely inhaled materials. Available information for a variety of
24 mammalian species, including humans, can be used to predict deposition patterns in the
25 respiratory tract for inhalable aerosols with reasonable degrees of accuracy. Additionally,
26 alveolar clearance data for non-human mammalian species commonly used in inhalation studies
27 are available from numerous experiments that involved inhaled radioactive particles.
28 An important factor in using models to predict retention patterns in laboratory animals or
29 humans is the dissolution-absorption rate of the inhaled material. Factors that affect the
30 dissolution of materials or the leaching of their constituents in physiological fluids and the
31 subsequent absorption of these constituents are not fully understood. Solubility is known to be
April 2002 6-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 influenced by the surface-to-volume ratio and other surface properties of particles (Mercer, 1967;
2 Morrow, 1973). The rates at which dissolution and absorption processes occur are influenced by
3 factors that include the chemical composition of the material. Temperature history of materials is
4 also an important consideration for some metal oxides. For example, in controlled laboratory
5 environments, the solubility of oxides usually decreases when the oxides are produced at high
6 temperatures, which generally results in compact particles having small surface-to-volume ratios.
7 It is sometimes possible to accurately predict dissolution-absorption characteristics of materials
8 based on physical/chemical considerations, but predictions for in-vivo dissolution-absorption
9 rates for most materials, especially if they contain multivalent cations or anions, should be
10 confirmed experimentally.
11 Phagocytic cells, primarily macrophages, clearly play a role in dissolution-absorption of
12 particles retained in the respiratory tract (Kreyling, 1992). Some particles dissolve within the
13 phagosomes because of the acidic milieu in those organelles (Lundborg et al., 1984, 1985), but
14 the dissolved material may remain associated with the phagosomes or other organelles in the
15 macrophage rather than diffuse out of the macrophage to be absorbed and transported elsewhere
16 (Cuddihy, 1984). This same phenomenon has been reported for organic materials. For example,
17 covalent binding of benzo[a]pyrene or metabolites to cellular macromolecules resulted in an
18 increased alveolar retention time for that compound after inhalation exposures of rats (Medinsky
19 andKampcik, 1985). Understanding these phenomena and recognizing species similarities and
20 differences are important for evaluating alveolar retention and clearance processes and for
21 interpreting the results of inhalation studies.
22 Dissolution-absorption of materials in the respiratory tract is clearly dependent on the
23 chemical and physical attributes of the material. Although it is possible to predict rates of
24 dissolution-absorption, it is prudent to determine this important clearance parameter
25 experimentally. It is important to understand the impact of this clearance process for the lungs,
26 tracheobronchial lymph nodes, and other body organs that might receive particles or their
27 constituents that enter the circulatory system from the lung.
28 Insufficient data were available to adequately model long-term retention of particles
29 deposited in the conducting airways of any mammalian species at the time of the 1996 PM
30 AQCD, and this still remains the case. Additional research must be done to provide the
31 information needed to properly evaluate retention of particles in conducting airways. However,
April 2002 6-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 a number of earlier studies, discussed in the 1996 document and in Section 6.2.2.2 herein, noted
2 that some particles were retained for relatively long times in the tracheobronchial regions,
3 effectively contradicting the general conclusion that almost all inhaled particles that deposit in
4 the TB region clear within hours or days. These studies have demonstrated that variable portions
5 of the particles that deposit in, or are cleared through, the TB region are retained with half times
6 on the order of weeks or months. Long-term retention and clearance patterns for particles that
7 deposit in the ET and TB regions must continue to be thoroughly evaluated because of the
8 implications of this information for respiratory tract dosimetry and risk assessment.
9 Model projections are possible for the A region using the cumulative information in the
10 scientific literature relevant to deposition, retention, and clearance of inhaled particles.
11 Clearance parameters for six laboratory animal species were summarized in U.S. Environmental
12 Protection Agency (1996). Nikula et al. (1997) evaluated results in rats and monkeys exposed to
13 high levels of either diesel soot or coal dust. Although the total amount of retained material was
14 similar in both species, the rats retained a greater portion in the lumens of the alveolar ducts and
15 alveoli than did monkeys; whereas the monkeys retained a greater portion of the material in the
16 interstitium. The investigators concluded that intrapulmonary retention patterns in one species
17 may not be predictive of those in another species at high levels of exposure, but this may not be
18 the case at lower levels of exposure.
19 The influence of exposure concentration on the pattern of particle retention in rats (exposed
20 to diesel soot) and humans (exposed to coal dust) was examined by Nikula et al. (2000) using
21 histological lung sections obtained from both species. The exposure concentrations for diesel
22 soot were 0.35, 3.5, or 7.0 //g/m3, and exposure duration was 7 h/day, 5 days/week for 24 mo.
23 The human lung sections were obtained from nonsmoking nonminers, nonsmoking coal miners
24 exposed to levels <2 //g dust/m3 for 3 to 20 years, or nonsmoking miners exposed to <10 //g/m3
25 for 33 to 50 years. In both species, the amount of retained material (using morphometric
26 techniques based on the volume density of deposition) increased with increasing dose (which is
27 related to exposure duration and concentration). In rats, the diesel exhaust particles were found
28 to be primarily in the lumens of the alveolar duct and alveoli; whereas in humans, retained dust
29 was found primarily in the interstitial tissue within the respiratory acini.
30
31
April 2002 6-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.6.3 Fluid Dynamics Models for Deposition Calculations
2 The available models developed to simulate particulate deposition in the lung are based on
3 simplifying assumptions about the morphometry of the lung and the fluid dynamics of inspired
4 air through a branching airway system. All of the approaches, whether analytic, symmetric, or
5 multiple-path models, simulate particle behavior in an "idealized" respiratory system with
6 homogeneous geometry and flow profile and can only predict average regional and total
7 dosimetry in the lung. As new models are developed, they will better predict particle deposition
8 patterns in a more realistic airway geometry under realistic flow conditions that can result in local
9 inhomogeneities of particle deposition and the formation of hot-spots. One example is the model
10 of ventilation distribution in the human lung developed by Chang and Yu (1999). This model
11 was designed as an improvement over those that assumed uniform ventilation in the lungs,
12 because it better simulated the effect of airway dynamics on the distribution of ventilation under
13 different conditions which may occur in the various lobes of the lungs and under various
14 inspiratory flow rates. The authors indicate that the results of the model compared favorably
15 with experimental data and that the model will be incorporated into a particle deposition model
16 that will allow for the evaluation of the nonuniformity of deposition within the lungs resulting
17 from the physiological situation of nonuniform distribution of ventilation. Computational fluid
18 dynamics (CFD) modeling adds another step to better model development by providing increased
19 ability to predict local airflow and particle deposition patterns and provide a better representation
20 of extrathoracic deposition in the human respiratory tract. The CFD models developed to date,
21 however, also are limited in scope because they are unable to simulate flow in the more complex
22 gas exchange regions. Due to a lack of more realistic simulations for the lower airways, they
23 impose another "idealized" boundary condition at the distal end of the human respiratory tract.
24 Airflow patterns within the lung are determined by the interplay of structural and
25 ventilatory conditions. These flow patterns govern the deposition kinetics of entrained particles
26 in the inspired air. A number of CFD software programs are available to simulate airflow
27 patterns in the lung by numerically solving the Navier-Stokes equations (White, 1974). The CFD
28 modeling requires a computer reconstruction of the appropriate lung region and the application of
29 boundary conditions. The flow field resulting from the CFD modeling is represented by velocity
30 vectors in the grid points of a two- or three-dimensional mesh. Numerical models of particle
31 deposition patterns are computed by simulating the trajectories of particles introduced into these
April 2002 6-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 flow streams after solving for the particles' equation of motion. Such CFD models have been
2 developed for different regions of the respiratory tract, including the nasal cavity (Yu et al., 1998;
3 Sarangapani and Wexler, 2000); larynx (Martonen et al. 1993; Katz et al., 1997; Katz, 2001);
4 major airway bifurcations (Gradon and Orlicki, 1990; Balashazy and Hofmann, 1993a,b, 1995,
5 2001; Heistracher and Hofmann, 1995; Lee et al., 1996; Zhang et al., 1997, 2000, 2001, 2002;
6 Comer et al., 2000, 2001a,b); and alveoli (Tsuda et al., 1994a,b; Chantal, 2001).
7 Kimbell (2001) has recently reviewed the literature on CFD models of the upper respiratory
8 tract (URT). Most of these models have focused on characterizing the airflow patterns in the
9 URT and have not included simulation of particulate dosimetry. Keyhani et al. (1995) were the
10 first to use computer-aided tomography (CAT) scans of the human nasal cavity to construct an
11 anatomically accurate three-dimensional airflow model of the human nose. Subramaniam et al.
12 (1998) used MRI scan data to extend these CFD studies to include the nasopharynx. However,
13 neither of these studies investigated particle deposition in the upper respiratory tract.
14 Yu et al. (1998) have developed a three-dimensional CFD model of the entire human upper
15 respiratory tract, including the nasal airway, oral airway, laryngeal airway, and the first two
16 generations of the tracheobronchial airway. They have used this CFD model to investigate the
17 effect of breathing pattern, i.e., nasal breathing, oral breathing, and simultaneous nasal and oral
18 breathing, on airflow and ultrafme particle deposition. They concluded that the ultrafme particle
19 deposition simulated using the CFD model was in reasonable agreement with the corresponding
20 experimental measurements. In a study led by Sarangapani and Wexler (2000), an upper
21 respiratory tract CFD model that included the nasal cavity, nasopharynx, pharynx, and larynx was
22 developed to study the deposition efficiency of hygroscopic and non-hygroscopic particles in this
23 region. They used the CFD model to simulate the temperature and water vapor conditions in the
24 upper airways and predicted high relative humidity conditions in this region. They also
25 simulated particle trajectories for 0.5 //m, 1 //m, and 5 //m particles under physiologically
26 realistic flow rates. The predictions of the CFD model indicated that high relative humidity
27 conditions contribute to rapid growth of hygroscopic particles and would dramatically alter the
28 deposition characteristics of ambient hygroscopic aerosols.
29 Stapleton et al. (2000) investigated deposition of a polydisperse aerosol (MMD = 4.8 //m
30 and GSD = 1.65) in a replica of a human mouth and throat, using both experimental results and
31 3-D CFD simulation. They found that CFD results were comparable with experimental results
April 2002 6-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 for a laminar flow case but were more than 200% greater for a turbulent flow case. The results
2 suggest that accurate predictions of particle deposition in a complex airway geometry requires a
3 careful evaluation of geometric and fluid dynamic factors in developing CFD models.
4 Due to the complex structural features and physiological conditions of the human laryngeal
5 region, only a limited number of modeling studies have been conducted to evaluate laryngeal
6 fluid dynamics and particle deposition. A high degree of inter-subject variability, a compliant
7 wall that presents challenges in setting appropriate boundary conditions, and a complex turbulent
8 flow field are some of the difficulties encountered in developing CFD models of the laryngeal
9 airways. Martonen et al. (1993) investigated laryngeal airflow using a two-dimensional CFD
10 model and concluded that laryngeal morphology exerts a pronounced influence on regional flow,
11 as well as fluid motion in the trachea and the main bronchi. In this study, the glottal aperture
12 (defined by the geometry of the vocal folds) was allowed to change in a prescribed manner with
13 the volume of inspiratory flow (Martonen and Lowe, 1983), and three flow rates corresponding
14 to different human activity were examined.
15 In a subsequent CFD analysis, a three-dimensional model of the larynx based on
16 measurements of human replica laryngeal casts (Martonen and Lowe, 1983; Katz and Martonen,
17 1996; Katz et al., 1997) simulated the flow field in the larynx and trachea under steady
18 inspiratory flow conditions at three flow rates. They observed that the complex geometry
19 produces jets, recirculation zones, and circumferential flow that may directly influence particle
20 deposition at select sites within the larynx and tracheobronchial airways. The primary
21 characteristics of the simulated flow field were a central jet penetrating into the trachea created
22 by the ventricular and vocal folds, a recirculating zone downstream of the vocal folds, and a
23 circumferential secondary flow. Recently, a computational model for fluid dynamics and particle
24 motion for inspiratory flow through the human larynx and trachea has been described (Katz,
25 2001). This model calculates the trajectory of single particles introduced at the entrance to the
26 larynx using a stochastic model for turbulent fluctuations incorporated into the particles'
27 equation of motion and time-averaged flow fields in the larynx and trachea. The effects of flow
28 rate and initial particle location on overall deposition were presented in the form of probability
29 density histograms of final particle deposition sites. At present, however, there are no
30 experimental data to validate results of such modeling.
April 2002 6-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 A number of CFD models have been developed to study fluid flow and particle deposition
2 patterns in airway bifurcations. The bifurcation geometries that have been modeled include:
3 two-dimensional (Li and Ahmadi, 1995); idealized three-dimensional using circular airways
4 (Kinsara et al., 1993) or square channels (Asgharian and Anjilvel, 1994); symmetric bifurcations
5 (Balashazy and Hofmann, 1993a,b); or physiologically realistic asymmetric single (Balashazy
6 and Hofmann, 1995; Heistracher and Hofmann, 1995) and multiple bifurcation models (Lee
7 et al., 1995; Heistracher and Hofmann, 1997; Comer et al., 2000, 2001; Zhang et al., 2000, 2001,
8 2002), with anatomical irregularities such as cartilaginous rings (Martonen et al., 1994a) and
9 carinal ridge (Martonen et al., 1994b; Comer et al., 2001a) shapes incorporated. The CFD flow
10 simulations in the bifurcating geometry models show distinct asymmetry in the axial (primary)
11 and radial (secondary) velocity profile in the daughter and parent airway during inspiration and
12 expiration, respectively. In a systematic investigation of flow patterns in airway bifurcations,
13 numerical simulations were performed to study primary flow (Martonen et al., 200la), secondary
14 currents (Martonen et al., 200Ib), and localized flow conditions (Martonen et al., 200Ic) for
15 different initial flow rates. The effects of inlet conditions, Reynolds numbers, ratio of airway
16 diameters, and branching angles with respect to intensity of primary flow, vortex patterns of the
17 secondary currents, and reverse flow in the parent-daughter transition region were investigated.
18 These simulated flow patterns match experimentally-observed flow profiles in airway
19 bifurcations (Schroter and Sudlow, 1969).
20 Gradon and Orlicki (1990) computed the local deposition flux of submicron size particles
21 in a three-dimensional bifurcation model for both inhalation and exhalation; and they found
22 enhanced deposition in the carinal ridge region during inspiration and in the central zone of the
23 parent airway during expiration. Numerical models of particle deposition in symmetric three-
24 dimensional bifurcations were developed by Balashazy and Hofmann (1993a,b), and these were
25 subsequently extended to incorporate effects of asymmetry in airway branching (Balashazy and
26 Hofmann, 1995) and physiologically realistic shapes of the bifurcation transition zone and the
27 carinal ridge (Heistracher and Hofmann, 1995; Balashazy and Hofmann, 2001). In these
28 numerical models, three-dimensional airflow patterns were computed by finite difference or
29 finite volume methods, and the trajectories of particles entrained in the airstream were simulated
30 using Monte Carlo techniques considering the simultaneous effects of gravitational settling,
31 inertial impaction, Brownian motion, and interception. The spatial deposition pattern of inhaled
April 2002 6-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles was examined for a range of particle sizes (0.01-10 //m) and flow rates (16-32 L/min
2 minute volume) by determining the intersection of particle trajectories with the surrounding
3 surfaces. The overall deposition rates derived using the CFD models correspond reasonably with
4 experimental data (Kim and Iglesias, 1989). These simulations predict deposition hot spots at the
5 inner side of the daughter airway downstream of the carinal ridge during inspiration,
6 corresponding to the secondary fluid motion of the inhaled air stream. During exhalation, the
7 CFD models predict enhanced deposition at the top and bottom parts of the parent airway,
8 consistent with secondary motion in the exhaled air stream. These studies indicate that
9 secondary flow patterns within the bifurcating geometry play a dominant role in determining
10 highly non-uniform local particle deposition patterns.
11 Zhang et al. (1997) numerically simulated particle deposition in three-dimensional
12 bifurcating airways (having varying bifurcation angles) due to inertial impaction during
13 inspiration for a wide range of Reynolds numbers (100-1000). Inlet velocity profile, flow
14 Reynolds number, and bifurcation angle had a substantial effect on particle deposition efficiency.
15 Based on the simulated results, equations were derived for particle deposition efficiency as a
16 function of nondimensional parameters, such as Stokes number, Reynolds number, and
17 bifurcation angle, and were shown to compare favorably with available experimental results.
18 More recently, Comer et al. (2000) have estimated the deposition efficiency of 3, 5, and 7 //m
19 particles in a three-dimensional double bifurcating airway model for both in-plane and out-of-
20 plane configurations for a wide range of Reynolds numbers (500-2000). They demonstrated
21 deposition in the first bifurcation to be higher than in the second bifurcation, with deposition
22 mostly concentrated near the carinal region. The non-uniform flow generated by the first
23 bifurcation had a dramatic effect on the deposition pattern in the second bifurcation. Based on
24 these results, they concluded that use of single bifurcation models are inadequate to capture the
25 complex fluid-particle interactions that occur in multigeneration airway systems.
26 Comer et al. (2001a) further investigated detailed characteristics of the axial and secondary
27 flow in a double bifurcation airway model using 3-D CFD simulation. Effects of carina shape
28 (sharp vs. rounded) and bifurcation plane (planar vs. non-planar) were examined. Particle
29 trajectories and deposition patterns were subsequently investigated in the same airway model
30 (Comer et al, 2001b). There was a highly localized deposition at and near the carina both in the
31 first and second bifurcation, and deposition efficiency was much lower in the second bifurcation
April 2002 6-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 than in the first bifurcation as demonstrated in the earlier study (Comer et al, 2000). They found
2 that deposition patterns were not much different between the sharp vs. rounded carina shape at
3 Stokes numbers of 0.04 and 0.12. However, deposition patterns were altered significantly for
4 these particles when the bifurcation plane was rotated, suggesting that a careful consideration of
5 realistic airway morphology is important for accurate prediction of particle deposition by CFD
6 modeling.
7 Zhang et al (2000, 2001) extended the studies of Comer et al. described above and
8 investigated effects of angled inlet tube as well as asymmetric flow distribution between daughter
9 branches. The flow asymmetry caused uneven deposition between downstream daughter
10 branches. Also noted was that the absolute deposition amount was higher, but deposition
11 efficiency per se was lower in the high flow branch than in the low flow branch. The intriguing
12 relationship between flow asymmetry and deposition was in fact consistent with experimental
13 data of Kim et al. (1999), indicating that the CFD model could correctly simulate complicated
14 airflow and particle dynamics that may occur in the respiratory airways.
15 Most CFD models use constant inspiratory or expiratory flows for simplicity and practical
16 reasons. However, the respiratory airflow is cyclic, and such flow characteristics cannot be fully
17 described by constant flows. Recent studies of Zhang et al. (2002) investigated particle
18 deposition in a triple bifurcation airway model under cyclic flow conditions mimicking resting
19 and light activity breathing. Deposition dose was obtained for every mm square area. They
20 found that deposition patterns were similar to those obtained with constant flows. However,
21 deposition efficiencies were greater with the cyclic flows than constant flows, and the difference
22 could be as high as 50% for 0.02 < mean Stk < 0.12 during normal breathing. The CFD results
23 are qualitatively comparable to experimental data (Kim et al, 1991) that showed about 25%
24 increase in deposition with cyclic flows. With further improvement of airway morphology and
25 computational scheme, CFD modeling could be a valuable tool for exploring the microdosimetry
26 in the airway structure.
27 Current CFD models of the acinar region are limited due to the complex and dynamic
28 nature of the gas exchange region. Flow simulation in a linearly increasing volume of a spherical
29 truncated two-dimensional alveolus model show distinct velocity maxima in the alveolar ducts
30 close to the entrance and exit points of the alveolus and a radial velocity profile in the interior
31 space of the alveolus (Tsuda et al., 1996). This is in contrast to simulations based on a rigid
April 2002 6-72 DRAFT-DO NOT QUOTE OR CITE
-------
1 alveolus (Tsuda, 1994a,b) and suggests that a realistic simulation of the flow pattern in the acinar
2 region should involve application of time-dependent methods with moving boundary conditions.
3 Nonuniform deposition patterns, with higher deposition near the alveolar entrance ring, have
4 been predicted using numerical models (Tsuda, 1994a,b, 1996).
5 Recent studies of Chantal (2001) examined aerosol transport and deposition in 6-generation
6 alveolated ducts using 2-D computer simulation. Particle trajectories and deposition patterns
7 were obtained for one complete breathing cycle (2 s inspiration and 2 s expiration). There were
8 large non-uniformities in deposition between generations, between ducts of a given generation,
9 and within each alveolated duct, suggesting that local deposition dose can be much greater than
10 the mean acinar dose.
11
12
13 6.7 SUMMARY AND CONCLUSIONS
14 An understanding of biological effects of inhaled particulate matter and underlying
15 mechanisms of action requires knowledge of the dosimetry of such material. This is because the
16 dose of particles delivered to a target site or sites of concern, rather than the actual exposure
17 concentration, is the proximal cause of the biological response. Such information is also critical
18 for extrapolation of effects found in controlled exposure studies of animals to those observed in
19 human clinical studies and, also, for relating effects in potentially susceptible persons to those in
20 normal, healthy persons. Dosimetry involves delineation of the processes of particle deposition,
21 translocation, and clearance. While the current understanding of basic mechanisms of particle
22 dosimetry, clearance, and retention has not changed since the 1996 PM ACQD (U.S.
23 Environmental Protection Agency, 1996), additional information has become available on the
24 role of certain biological determinants of these processes, such as gender and age; and there has
25 been an expansion of previous knowledge about the relationship between regional deposition and
26 translocation in regard to specific particle size ranges of significance to ambient particulate
27 exposure scenarios. There also has been significant improvement in the mathematical and CFD
28 modeling of particle dosimetry in the respiratory tract of humans. Although the models have
29 become more sophisticated and versatile, validation of the models is still needed.
30 One of the areas that has improved since the 1996 PM ACQD is consideration of specific
31 and relevant ambient size particle ranges in deposition studies. One such size mode is the
April 2002 6-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 ultrafine. While further information on respiratory deposition for this size mode is still needed,
2 there has been an improvement in the understanding of total deposition as a function of particle
3 size and breathing pattern and of certain aspects of regional deposition of ultrafme particles.
4 This new information indicates that the ET region, especially the nasal passages, is a very
5 efficient "filter" for these particles, reducing the amount which would be available for deposition
6 in the TB and A regions of the respiratory tract. Within the thoracic region, the deposition
7 distribution of ultrafme particles is highly skewed towards the proximal airway regions and
8 resembles that of coarse particles. In other words, deposition patterns of ultrafme particles are
9 very much like those of coarse particles. Another example involves studies which attempt to
10 evaluate the contribution of fine- and coarse-mode particles to deposition in various parts of the
11 respiratory tract, although there have been only a few of these.
12 It always has been clear that certain host factors affect deposition, and there has been
13 improvement since the 1996 PM AQCD in the understanding of some of these factors,
14 specifically gender and age. Recent information suggests that there are significant gender
15 differences in the homogeneity of deposition as well as the deposition rate, and this could affect
16 susceptibility. In regard to age, recent evaluations employed both mathematical models as well
17 as experimental studies, and most involved comparison of deposition in children compared to
18 adults. These studies generally indicate that children would receive greater doses of particles per
19 lung surface area than would adults. Unfortunately, deposition studies in another potentially
20 susceptible population, namely the elderly, are still lacking although there have been a number of
21 studies examining effects of chronic pulmonary disease on deposition. These studies confirmed
22 that significant increases in deposition in obstructed lungs could occur.
23 Once deposited on airway surfaces, particles are subjected to translocation and clearance.
24 While the general pathways of clearance have been known for years, recent information has
25 improved the understanding of translocation of particles within size ranges which may be of
26 specific concern for ambient exposures. One such size mode, as noted above, is the ultrafme;
27 and recent studies indicate that ultrafme particles can be rapidly cleared from the lungs into the
28 systemic circulation and reach extrapulmonary organs. This provides a mechanism whereby
29 inhaled particles may affect cardiovascular function, as noted in various epidemiological studies
30 (see Chapter 8).
April 2002 6-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 As with experimental studies, the major improvements in mathematical modeling of
2 dosimetry involve evaluation of realistic size modes for ambient conditions, as well as
3 improvements in the precision of these models for more realistic depictions of respiratory tract
4 airflow patterns and detailed airway structures that may result in deposition "hot spots". These
5 improvements include more detailed evaluations of enhanced deposition at airway bifurcations,
6 use of parameters that allow determination of age differences in dosimetry, and improvement in
7 the modeling of clearance mechanisms.
8 Thus, in general, while our understanding of specific aspects of particle dosimetry has
9 improved since the 1996 PM AQCD, there are still areas in need of further evaluation. These
10 areas of research include dosimetry in susceptible humans, better models for extrapolation
11 between animals used in inhalation studies and humans, and better understanding of differences
12 in the manner in which particles of different and relevant ambient size modes are handled
13 following deposition. This latter research need is important for determining the potential of
14 various particle types to exert effects systemically, rather than just locally within the respiratory
15 tract.
April 2002 6-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Adamson, I. Y. R.; Bowden, D. H. (1981) Dose response of the pulmonary macrophagic system to various
3 particulates and its relationship to transepithelial passage of free particles. Exp. Lung Res. 2: 165-175.
4 Adamson, I. Y. R.; Hedgecock, C. (1995) Patterns of particle deposition and retention after instillation to mouse
5 lung during acute injury and fibrotic repair. Exp. Lung Res. 21: 695-709.
6 Adamson, I. Y. R.; Prieditis, H. (1998) Silica deposition in the lung during epithelial injury potentiates fibrosis and
7 increases particle translocation to lymph nodes. Exp. Lung Res. 24: 293-306.
8 American Conference of Governmental Industrial Hygienists (ACGIH). (1985) Particle size-selective sampling in
9 the workplace: report of the ACGIH technical committee on air sampling procedures. Cincinnati, OH:
10 American Conference of Governmental Industrial Hygienists.
11 Anjilvel, S.; Asgharian, B. (1995) A multiple-path model of particle deposition in the rat lung. Fundam. Appl.
12 Toxicol. 28: 41-50.
13 Asgharian, B.; Anjilvel, S. (1994) Inertia! and gravitational deposition of particles in a square cross section
14 bifurcating airway. Aerosol Sci. Technol. 20: 177-193.
15 Asgharian, B.; Wood, R.; Schlesinger, R. B. (1995) Empirical modeling of particle deposition in the alveolar region
16 of the lungs: a basis for interspecies extrapolation. Fundam. Appl. Toxicol. 27: 232-238.
17 Asgharian, B.; Miller, F. J.; Subramaniam, R. P. (1999) Dosimetry software to predict particle deposition in humans
18 and rats. CUT Activities 19(3). Available: http://www.ciit.org/ACT99/ACTIVITIESMAR99/mar99.html
19 [19 December, 2001].
20 Asgharian, B.; Hofmann, W.; Miller, F. J. (2001) Mucociliary clearance of insoluble particles from the
21 tracheobronchial airways of the human lung. Aerosol Sci. 32: 817-832.
22 Balashazy, I.; Hofmann, W. (1993a) Particle deposition in airway bifurcations-I: inspiratory flow. J. Aerosol Sci.
23 24:745-772.
24 Balashazy, I.; Hofmann, W. (1993b) Particle deposition in airway bifurcations-II: expiratory flow. J. Aerosol Sci.
25 24:773-786.
26 Balashazy, I.; Hofmann, W. (1995) Deposition of aerosols in asymmetric airway bifurcations. J. Aerosol Sci.
27 26:273-292.
28 Balashazy, I.; Hofmann, W. (2001) Fluid dynamics and related particle deposition patterns in human airway
29 bifurcations. In: Martonen, T. B., ed., Medical Applications of Computer Modeling: The Respiratory System.
30 WIT Press, Southampton, Boston, p. 84-108.
31 Bennett, W. D.; Ilowite, J. S. (1989) Dual pathway clearance of 99mTc-DTPA from the bronchial mucosa. Am. Rev.
32 Respir. Dis. 139: 1132-1138.
33 Bennett, W. D.; Zeman, K. L. (1998) Deposition of fine particles in children spontaneously breathing at rest.
34 Inhalation Toxicol. 10: 831-842.
35 Bennett, W. D.; Zeman, K. L.; Kim, C. (1996) Variability of fine particle deposition in healthy adults: effect of age
36 and gender. Am. J. Respir. Crit. Care Med. 153: 1641-1647.
37 Bennett, W. D.; Zeman, K. L.; Kang, C. W.; Schechter, M. S. (1997) Extrathoracic deposition of inhaled, coarse
3 8 particles (4.5 ,am) in children vs. adults. In: Cherry, N.; Ogden, T., eds. Inhaled particles VIII: proceedings of
39 an international symposium on inhaled particles organised by the British Occupational Hygiene Society;
40 August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 497-502.
41 Bennett, W. D.; Zeman, K. L.; Kim, C.; Mascarella, J. (1997b) Enhanced deposition of fine particles in COPD
42 patients spontaneously breathing at rest. Inhalation Toxicol. 9: 1-14.
43 Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1998) Bronchial
44 airway deposition and retention of particles in inhaled boluses: effect of anatomic dead space. J. Appl.
45 Physiol. 85: 685-694.
46 Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1999) Regional
47 deposition and retention of particles in shallow, inhaled boluses: effect if lung volume. J. Appl. Physiol.
48 86: 168-173.
49 Brauer, M.; Avila-Casado, C.; Fortoul, T. I.; Vedal, S.; Stevens, B.; Churg, A. (2001) Air pollution and retained
50 particles in the lung. Environ. Health Perspect. 109: 1039-1043.
51 Broday, D. M.; Georgopoulos, P. G. (2001) Growth and deposition of hygroscopic paniculate matter in the human
52 lungs. Aerosol Sci. Technol. 34: 144-159.
53 Brown, J. S.; Kirby, Z. L.; Bennett, W. D. (2001) Regional deposition of coarse particles and ventilation distribution
54 in healthy subjects and patients with cystic fibrosis. J. Aerosol Med. 14: 443-454.
April 2002 6-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 Camner, P.; Anderson, M; Philipson, K.; Bailey, A.; Hashish, A.; Jarvis, N.; Bailey, M.; Svartengren, M. (1997)
2 Human bronchiolar deposition and retention of 6-, 8-, and 10-/u particles. Exp. Lung Res. 23: 517-535.
3 Chang, Y. H.; Yu, C. P. (1999) A model of ventilation distribution in the human lung. Aerosol Sci. Technol.
4 30:309-319.
5 Cheng, Y.-S.; Smith, S. M.; Yeh, H.-C.; Kim, D.-B.; Cheng, K.-H.; Swift, D. L. (1995) Deposition of ultrafine
6 aerosols and thoron progeny in replicas of nasal airways of young children. Aerosol. Sci. Technol.
7 23:541-552.
8 Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Guilmette, R. A.; Simpson, S. Q.; Yang, Y.-H.; Swift, D. L. (1996) In vivo
9 measurements of nasal airway dimensions and ultrafine aerosol deposition in the human nasal and oral
10 airways. J. Aerosol Sci. 27: 785-801.
11 Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Swift, D. L. (1997) An experimental method for measuring aerosol
12 deposition efficiency in the human oral airway. Am. Ind. Hyg. Assoc. J. 58: 207-213.
13 Churg, A.; Brauer, M. (1997) Human lung parenchyma retains PM2 5. Am. J. Respir. Crit. Care Med.
14 155:2109-2111.
15 Churg, A.; Vedal, S. (1996) Carinal and tubular airway particle concentrations in the large airways of non-smokers
16 in the general population: evidence for high particle concentration at airway carinas. Occup. Environ. Med.
17 53:553-558.
18 Cohen, B. S.; Harley, N. H.; Schlesinger, R. B.; Lippmann, M. (1988) Nonuniform particle deposition on
19 tracheobronchial airways: implications for lung dosimetry. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.;
20 Fisher, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung
21 dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32 (suppl. 1): 1045-1053.
22 Cohen, B. S.; Sussman, R. G.; Lippmann, M. (1990) Ultrafine particle deposition in a human tracheobronchial cast.
23 Aerosol Sci. Technol. 12: 1082-1091.
24 Cohen, B. S.; Xiong, J. Q.; Fang, C.-P.; Li, W. (1998) Deposition of charged particles on lung airways. Health
25 Phys. 74: 554-560.
26 Comer, J. K.; Kleinstreuer, C.; Hyun, S.; Kim, C. S. (2000) Aerosol transport and deposition in sequentially
27 bifurcating airways. J. Biomech. Eng. 122: 152-158.
28 Comer, J. K.; Kleinstreuer, C.; Zhang, Z. (2001a) Flow structures and particle deposition patterns in double
29 bifurcation airway models. Part 1. Air flow fields. J. Fluid Mech. 435: 24-54.
30 Comer, J. K.; Kleinstreuer, C.; Kim, C. S. (200Ib) Flow structures and particle deposition patterns in double
31 bifurcation airway models. Part 2. Aerosol transport and deposition. J. Fluid Mech. 435: 55-80.
32 Crystal, R. G., West, J. B.; Barnes, P. J.; Weibel, E. R., eds. (1997) The lung: scientific foundations. Volume 1.
33 Section III: major components. 2nd ed. Philadelphia, PA: Lippincott-Raven; chapters 30-67, pp. 445-991.
34 Cuddihy, R. G. (1984) Mathematical models for predicting clearance of inhaled radioactive materials. In: Smith, H.;
3 5 Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a meeting jointly
3 6 organized by the Commission of the European Communities and the National Radiological Protection Board;
37 March; Oxford, United Kingdom. Luxembourg: Commission of the European Communities; pp. 167-179;
38 report no. EUR 9384 EN.
39 Cuddihy, R. G.; Yeh, H. C. (1988) Respiratory tract clearance of particles and substances dissociated from particles.
40 In: Mohr, U.; Dungworth, D.; Kimmerle, G.; Lewkowski, J.; McClellan, R.; Stober, W., eds. Inhalation
41 toxicology: the design and interpretation of inhalation studies and their use in risk assessment. New York,
42 NY: Springer-Verlag; pp. 169-193.
43 Darquenne, C. (2001) A realistic two-dimensional model of aerosol transport and deposition in the alveolar zone of
44 the human lung. J. Aerosol Sci. 32: 1161-1174.
45 Dorries, A. M.; Valberg, P. A. (1992) Heterogeneity of phagocytosis for inhaled versus instilled material. Am. Rev.
46 Respir. Dis. 146:831-837.
47 Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
48 Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
49 limitations. Toxicol. Sci. 55: 24-35.
50 Falk, R.; Philipson, K.; Svartengren, M.; Jarvis, N.; Bailey, M.; Camner, P. (1997) Clearance of particles from small
51 ciliated airways. Exp. Lung Res. 23: 495-515.
52 Falk, R.; Philipson, K.; Svartengren, M.; Bergmann, R.; Hofmann, W.; Jarvis, N.; Bailey, M.; Camner, P. (1999)
5 3 Assessment of long-term bronchiolar clearance of particles from measurements of lung retention and
54 theoretical estimates of regional deposition. Exp. Lung Res. 25: 495-516.
April 2002 6-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ferin, J. (1977) Effect of particle content of lung on clearance pathways. In: Sanders, C. L.; Schneider, R. P.; Dagle,
2 G. E.; Ragan, H. A., eds. Pulmonary macrophages and epithelial cells: proceedings of the sixteenth annual
3 Hanford biology symposium; September 1976; Richland, WA. Oak Ridge, TN: Energy Research and
4 Development Administration; pp. 414-423. Available from: NTIS, Springfield, VA; CONF-760927.
5 (ERDA symposium series 43).
6 Ferin, J.; Feldstein, M. L. (1978) Pulmonary clearance and hilar lymph node content in rats after particle exposure.
7 Environ. Res. 16: 342-352.
8 Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
9 Respir. Cell Mol. Biol. 6: 535-542.
10 Findeisen, W. (1935) Uber das Absetzen kleiner, in der Luft suspendierter Teilchen in der menschlichen Lunge bei
11 der Atmung [The deposition of small airborne particles in the human lung during respiration]. Pfluegers
12 Arch. Gesamte Physiol. Menschen Tiere 236: 367-379.
13 Foster, W. M.; Langenback, E.; Bergofsky, E. H. (1980) Measurement of trachea! and bronchial mucus velocities in
14 man: relation to lung clearance. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 48: 965-971.
15 Frampton, M. W.; Chalupa, D.; Morrow, P. E.; Gibb, F. R.; Oberdorster, G.; Speers, D. M.; Zareba, W.; Utell, M. J.
16 (2000) Deposition and effects of inhaled ultrafine carbon particles in healthy subjects at rest. Presented at:
17 PM2000: paniculate matter and health—the scientific basis for regulatory decision-making, specialty
18 conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
19 Fry, F. A.; Black, A. (1973) Regional deposition and clearance of particles in the human nose. J. Aerosol Sci.
20 4: 113-124.
21 Gehr, P.; Schurch, S.; Berthaiume, Y.; ImHof, V.; Geiser, M. (1990) Particle retention in airways by surfactant.
22 J. Aerosol Med. 3:27-43.
23 Gehr, P.; ImHof, V.; Geiser, M.; Schurch, S. (1991) The fate of particles deposited in the intrapulmonary
24 conducting airways. J. Aerosol Med. 4: 349-362.
25 Gradofi, L.; Orlicki, D. (1990) Deposition and inhaled aerosol particles in a generation of the tracheobronchial tree.
26 J. Aerosol Sci. 21:3-19.
27 Green, F. H. Y. (2000) Pulmonary responses to inhaled poorly soluble paniculate in the human. In: Gardner, D. E.,
28 ed. ILSI Risk Science Institute Workshop: The Relevance of the Rat Lung Response to Particle Overload for
29 Human Risk Assessment; March, 1998. Inhalation Toxicol. 12: 59-95.
30 Groth, M. L.; Macri, K.; Foster, W. M. (1997) Cough and mucociliary transport of airway paniculate in chronic
31 obstructive lung disease. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an
32 international symposium on inhaled particles organised by the British Occupational Hygiene Society;
33 August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 515-521.
34 Guilmette, R. A.; Cheng, Y. S.; Griffith, W. C. (1997) Characterising the variability in adult human nasal airway
35 dimensions. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international
36 symposium on inhaled particles organised by the British Occupational Hygiene Society; August 1996;
37 Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 491-496.
38 Harmsen, A. G.; Muggenburg, B. A.; Snipes, M. B.; Bice, D. E. (1985) The role of macrophages in particle
39 translocation from lungs to lymph nodes. Science (Washington, DC) 230: 1277-1280.
40 Heistracher, T.; Hofmann, W. (1995) Physiologically realistic models of bronchial airway bifurcation. J. Aerosol
41 Sci. 26: 497-509.
42 Heistracher, T.; Hofmann, W. (1997) Flow and deposition patterns in successive airway bifurcations. In: Cherry,
43 N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international symposium on inhaled particles
44 organised by the British Occupational Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg.
45 41(suppl.): 537-542.
46 Henshaw, D. L.; Fews, A. P. (1984) The microdistribution of alpha emitting particles in the human lung.
47 In: Smith, H.; Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a
48 meeting jointly organized by the Commission of the European Communities and the National Radiological
49 Protection Board; March; Oxford, United Kingdom. Luxembourg: Commission of the European
50 Communities; pp. 199-218; report no. EUR 9384 EN.
51 Heyder, J.; Rudolf, G. (1977) Deposition of aerosol particles in the human nose. In: Walton, W. H.; McGovern, B.,
52 eds. Inhaled particles IV: proceedings of an international symposium, part 1; September 1975; Edinburgh,
53 United Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 107-126.
54 Killer, F. C. (1991) Health implications of hygroscopic particle growth in the human respiratory tract. J. Aerosol
55 Med. 4: 1-23.
April 2002 6-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hofmann, W.; Bergmann, R. (1998) Predictions of particle deposition patterns in human and rat airways. Inhalation
2 Toxicol. 10: 557-583.
3 Hofmann, W.; Martonen, T. B.; Graham, R. C. (1989a) Predicted deposition of nonhygroscopic aerosols in the
4 human lung as a function of subject age. J. Aerosol Med. 2: 49-68.
5 Hofmann, W.; Koblinger, L.; Martonen, T. B. (1989b) Structural differences between human and rat lungs:
6 implications for Monte Carlo modeling of aerosol deposition. In: Mahaffey, J. A., ed. 26th Hanford life
7 sciences symposium, modeling for scaling to man: biology, dosimetry, and response. Health Phys.
8 57(suppl. 1): 41-47.
9 Hofmann, W.; Balashazy, L; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
10 bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
11 Hofmann, W.; Bergmann, R.; Koblinger, L. (1999) Characterization of local particle deposition patterns in human
12 and rat lungs by different morphometric parameters. J. Aerosol Sci. 30: 651-667.
13 Hofmann, W.; Asgharian, B.; Bergmann, R.; Anjilvel, S.; Miller, F. J. (2000) The effect of heterogeneity of lung
14 structure on particle deposition in the rat lung. Toxicol. Sci. 53: 430-437.
15 Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
16 Inhalation Toxicol. 10: 121-130.
17 Huchon, G. J.; Montgomery, A. B.; Lipavsky, A.; Hoeffel, J. M.; Murray, J. F. (1987) Respiratory clearance of
18 aerosolized radioactive solutes of varying molecular weight. J. Nucl. Med. 28: 894-902.
19 International Commission on Radiological Protection. (1960) Report of Committee II on permissible dose for
20 internal radiation (1959). Health Phys. 3.
21 International Commission on Radiological Protection. (1979) Limits for intakes of radionuclides by workers.
22 Oxford, United Kingdom: Pergamon Press; ICRP publication 30, part 1.
23 International Commission on Radiological Protection. (1994) Human respiratory tract model for radiological
24 protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
25 United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP: v. 24, nos. 1-3).
26 Jaques, P. A.; Kim, C. S. (2000) Measurement of total lung deposition of inhaled ultrafine particles in healthy men
27 and women. Inhalation Toxicol. 12: 715-731.
28 John, J.; Wollmer, P.; Dahlback, M.; Luts, A.; Jonson, B. (1994) Tidal volume and alveolar clearance of insoluble
29 particles. J. Appl. Physiol. 76: 584-588.
3 0 Katz, I. M. (2001) Computer modeling of fluid dynamics and particle motion in the larynx and trachea. In:
31 Martonen, T. B., ed. Medical Applications of Computer Modeling: The Respiratory System. Boston, MA:
32 WIT Press, pp. 47-63.
33 Katz, I. M.; Martonen, T. B. (1996) Three-dimensional fluid particle trajectories in the human larynx and trachea.
34 J. Aerosol Med. 9:513-20.
35 Katz, I. M., Martonen, T. B., and Flaa, W. (1997) Three-dimensional computational study of inspiratory aerosol
36 flow through the larynx: the effect of glottal aperture modulation. J. Aerosol Sci. 28: 1073-83.
37 Katz, I. M.; Davis, B. M.; Martonen, T. B. (1999) A numerical study of particle motion within the human larynx
38 and trachea. J. Aerosol Sci. 30: 173-183.
39 Kesavanathan, J.; Swift, D. L. (1998) Human nasal passage particle deposition: the effect of particle size, flow rate,
40 and anatomical factors. Aerosol Sci. Technol. 28: 457-463.
41 Keyhani, K.; Scherer, P. W.; Mozell, M. M. (1995) Numerical simulation of airflow in the human nasal cavity.
42 J. Biomech. Eng. 117: 429-41.
43 Kim, C. S. (1989) Aerosol deposition in the lung with obstructed airways. J. Aerosol Med. 2: 111-120.
44 Kim, C. S. (2000) Methods of calculating lung delivery and deposition of aerosol particles. Respir. Care
45 45:695-711.
46 Kim, C. S.; Fisher, D. M. (1999) Deposition characteristics of aerosol particles in sequentially bifurcating airway
47 models. Aerosol Sci. Technol. 31:198-220.
48 Kim, C. S.; Garcia, L. (1991) Particle deposition in cyclic bifurcating tube flow. Aerosol Sci. Technol. 14: 302-315.
49 Kim, C. S.; Hu, S. C. (1998) Regional deposition of inhaled particles in human lungs: comparison between men and
50 women. J. Appl. Physiol. 84: 1834-1844.
51 Kim, C. S.; Iglesias, A. J. (1989) Deposition of inhaled particles in bifurcating airway models: I. inspiratory
52 deposition. J. Aerosol Med. 2: 1-14.
53 Kim, C. S.; Jaques, P. A. (2000) Respiratory dose of inhaled ultrafine particles in healthy adults. Phil. Trans. Roy.
54 Soc. London A 358: 2693-2705.
April 2002 6-79 DRAFT-DO NOT QUOTE OR CITE
-------
1 Kim, C. S.; Kang, T. C. (1997) Comparative measurement of lung deposition of inhaled fine particles in normal
2 subjects and patients with obstructive airway disease. Am. J. Respir. Crit. Care Med. 155: 899-905.
3 Kim, C. S.; Iglesias, A. J.; Sackner, M. A. (1987) Mucus clearance by two-phase gas-liquid flow mechanism:
4 Asymmetric periodic flow model. J. Appl. Physiol. 62: 959-971.
5 Kim, C. S.; Eldridge, M. A.; Garcia, L.; Wanner A. (1989) Aerosol deposition in the lung with asymmetric airways
6 obstruction: in vivo observation. J. Appl. Physiol. 67: 2579-2585.
7 Kim, C. S.; Hu, S. C.; DeWitt, P.; Gerrity, T. R. (1996) Assessment of regional deposition of inhaled particles in
8 human lungs by serial bolus delivery method. J. Appl. Physiol. 81: 2203-2213.
9 Kim, C. S.; Hu, S. C.; Jaques, P.; Ding, J.; DeWitt, P. (2000) Measurement of regional deposition dose of inhaled
10 particles in the size range of 0.04-5 /j. in human lungs. Presented at: PM2000: paniculate matter and
11 health~the scientific basis for regulatory decision-making, specialty conference & exhibition; January;
12 Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
13 Kimbell, J. S. (2001) Computational fluid dynamics of the extrathoracic airways. In: Martonen, T. B., ed. Medical
14 Applications of Computer Modeling: The Respiratory System. Boston, MA: WIT Press; pp. 11-45.
15 Kinsara, A. A.; Tompson, R. V.; Loyalka, S. K. (1993) Computational flow and aerosol concentration profiles in
16 lung bifurcations. Health Phys. 64: 13-22.
17 Koblinger, L.; Hofmann, W. (1985) Analysis of human lung morphometric data for stochastic aerosol deposition
18 calculations. Phys. Med. Biol. 30: 541-556.
19 Koblinger, L.; Hofmann, W. (1988) Stochastic morphological model of the rat lung. Anat. Rec. 221: 533-539.
20 Koch, W.; Stober, W. (2001) A simple pulmonary retention model accounting for dissolution and
21 macrophage-mediated removal of deposited polydisperse particles. Inhalation Toxicol. 13: 129-148.
22 Kohlhaufl, M.; Brand, P.; Scheuch, G.; Meyer, T. S.; Schulz, H.; Haussinger, K.; Heyder, J. (1999) Increased fine
23 particle deposition in women with asymptomatic nonspecific airway hyperresponsiveness. Am. J. Respir.
24 Crit. Care Med. 159: 902-906.
25 Kreyling, W. G. (1992) Intracellular particle dissolution in alveolar macrophages. Environ. Health Perspect.
26 97:121-126.
27 Kreyling, W. G.; Blanchard, J. D.; Godleski, J. J.; Haeussermann, S.; Heyder, J.; Hutzler, P.; Schulz, H.; Sweeney,
28 T. D.; Takenaka, S.; Ziesenis, A. (1999) Anatomic localization of 24- and 96-h particle retention in canine
29 airways. J. Appl. Physiol. 87: 269-284.
30 LaBelle, C. W.; Brieger, H. (1961) Patterns and mechanisms in the elimination of dust from the lung. In: Davies,
31 C. N., ed. Inhaled particles and vapours: proceedings of an international symposium; March-April 1960;
32 Oxford, United Kingdom. New York, NY: Pergamon Press; pp. 356-368.
33 Landahl, H. D. (1950) On the removal of air-borne droplets by the human respiratory tract: I. the lung. Bull. Math.
34 Biophys. 12: 43-56.
35 Lay, J. C.; Berry, C. R.; Kim, C. S.; Bennett, W. D. (1995) Retention of insoluble particles after local intrabronchial
36 deposition in dogs. J. Appl. Physiol. 79: 1921-1929.
37 Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular
3 8 distribution of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol.
39 18:687-695.
40 Lazaridis, M.; Broday, D. M.; Hov, 0.; Georgopoulos, P. G. (2001) Integrated exposure and dose modeling and
41 analysis system. 3. Deposition of inhaled particles in the human respiratory tract. Environ. Sci. Technol.
42 35: 3727-3734.
43 Lee, J. W.; Goo, J. H.; Chung, M. K. (1996) Characteristics of inertial deposition in a double bifurcation. J. Aerosol
44 Sci. 27: 119-138.
45 Lehnert, B. E.; Morrow, P. E. (1985) Association of 59iron oxide with alveolar macrophages during alveolar
46 clearance. Exp. Lung Res. 9: 1-16.
47 Lehnert, B. E.; Valdez, Y. E.; Bomalaski, S. H. (1988) Analyses of particles in the lung free cell, tracheobronchial
48 lymph nodal, and pleural space compartments following their deposition in the lung as related to lung
49 clearance mechanisms. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fisher, D. R., eds. Inhaled particles
50 VI: proceedings of an international symposium and workshop on lung dosimetry; September 1985;
51 Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 125-140.
52 Leikauf, G.; Yeates, D. B.; Wales, K. A.; Spektor, D.; Albert, R. E.; Lippmann, M. (1981) Effects of sulfuric acid
53 aerosol on respiratory mechanics and mucociliary particle clearance in healthy nonsmoking adults. Am. Ind.
54 Hyg. Assoc. J. 42: 273-282.
April 2002 6-80 DRAFT-DO NOT QUOTE OR CITE
-------
1 Leikauf, G. D.; Spektor, D. M; Albert, R. E.; Lippmann, M. (1984) Dose-dependent effects of submicrometer
2 sulfuric acid aerosol on particle clearance from ciliated human lung airways. Am. Ind. Hyg. Assoc. J.
3 45:285-292.
4 Lennon, S.; Zhang, Z.; Lessmann, R.; Webster, S. (1998) Experiments on particle deposition in the human upper
5 respiratory system. Aerosol Sci. Technol. 28: 464-474.
6 Li, A.; Ahmadi, G. (1995) Computer simulation of particle deposition in the upper tracheobronchial tree. Aerosol
7 Sci. Tech. 23:201-223.
8 Li, W.; Perzl, M.; Heyder, J.; Langer, R.; Brain, J. D.; Englmeier, K. H.; Niven, R. W.; Edwards, D. A. (1996)
9 Aerodynamic and aerosol particle deaggregation phenomena in model oral-pharyngeal cavities. J. Aerosol
10 Sci. 27: 1269-1286.
11 Lundborg, M.; Lind, B.; Camner, P. (1984) Ability of rabbit alveolar macrophages to dissolve metals. Exp. Lung
12 Res. 7:11-22.
13 Lundborg, M.; Eklund, A.; Lind, B.; Camner, P. (1985) Dissolution of metals by human and rabbit alveolar
14 macrophages. Br. J. Ind. Med. 42: 642-645.
15 Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
16 macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
17 Pharmacol. 153:28-38.
18 Martonen, T. B.; Lowe, J. (1983) Assessment of aerosol deposition patterns in human respiratory tract casts.
19 In: Marple, V. A.; Liu, B. Y. H., eds. Aerosols in the Mining and Industrial Work Environments Vol. 1
20 Fundamentals and Status, p. 151-164. Ann Arbor, MI: Ann Arbor Science Publishers.
21 Martonen, T. B.; Zhang, Z.; Lessmann, R. (1993) Fluid dynamics of the human larynx and upper tracheobronchial
22 airways. Aerosol Sci. Technol. 19: 133-156.
23 Martonen, T. B.; Yang, Y.; Xue, Z. Q. (1994a) Influences of cartilagious rings on tracheobronchial fluid dynamics.
24 Inhalation Toxicol. 6: 185-203.
25 Martonen, T. B.; Yang, Y.; Xue, Z. Q. (1994b) Effects of carinal ridge shapes on lung airstreams. Aerosol Sci.
26 Technol. 21: 119-136.
27 Martonen, T. B.; Schroeter, J. D.; Hwang, D.; Fleming, J. S.; Conway, J. H. (2000) Human lung morphology
28 models for particle deposition studies. In: Grant, L. D., ed. PM2000: paniculate matter and health.
29 Inhalation Toxicol. 12(suppl. 4): 109-121.
30 Martonen, T. B.; Guan, X.; Schreck, R. M. (2001a) Fluid dynamics in airway bifurcations: I. Primary flows.
31 Inhalation Toxicol. 13: 261-79.
32 Martonen, T. B.; Guan, X.; Schreck, R. M. (200Ib) Fluid dynamics in airway bifurcations: II. Secondary currents.
33 Inhalation Toxicol. 13: 281-9.
34 Martonen, T. B.; Guan, X.; Schreck, R. M. (2001c) Fluid dynamics in airway bifurcations: III. Localized flow
35 conditions. Inhalation Toxicol. 13: 291-305.
36 Matsui, H.; Randell, S. H.; Peretti, S. W.; Davis, C. W.; Boucher, R. C. (1998) Coordinated clearance of periciliary
37 liquid and mucus from airway surfaces. J. Clin. Invest. 102: 1125-1131.
3 8 Medinsky, M. A.; Kampcik, S. J. (1985) Pulmonary retention of [14C]benzo/o/pyrene in rats as influenced by the
39 amount instilled. Toxicology 35: 327-336.
40 Mercer, T. T. (1967) On the role of particle size in the dissolution of lung burdens. Health Phys. 13: 1211-1221.
41 Miller, F. J.; Anjilvel, S.; Menache, M. G.; Asgharian, B.; Gerrity, T. R.. (1995) Dosimetric issues relating to
42 paniculate toxicity. Inhalation Toxicol. 7: 615-632.
43 Morrison, D.; Skwarski, D.; Millar, A. M.; Adams, W.; MacNee, W. (1998) A comparison of three methods of
44 measuring 99mTc-DTPA lung clearance and their repeatability. Eur. Respir. J. 11: 1141-1146.
45 Morrow, P. E. (1973) Alveolar clearance of aerosols. Arch. Intern. Med. 131: 101-108.
46 Morrow, P. E. (1977) Clearance kinetics of inhaled particles. In: Brain, J. D.; Proctor, D. F.; Reid, L. M., eds.
47 Respiratory defense mechanisms (in two parts), part II. New York, NY: Marcel Dekker, Inc.; pp. 491-543.
48 (Lenfant, C., ed. Lung biology in health and disease: v. 5).
49 Morrow, P. E. (1986) Factors determining hygroscopic aerosol deposition in airways. Physiol. Rev. 66: 330-376.
50 Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
51 10:369-384.
52 Morrow, P. E. (1994) Mechanisms and significance of "particle overload." In: Mohr, U.; Dungworth, D. L.;
53 Mauderly, J. L.; Oberdorster, G., eds. Toxic and carcinogenic effects of solid particles in the respiratory tract:
54 [proceedings of the 4th international inhalation symposium]; March 1993; Hannover, Germany. Washington,
55 DC: International Life Sciences Institute Press; pp. 17-25.
April 2002 6-81 DRAFT-DO NOT QUOTE OR CITE
-------
1 Morrow, P. E.; Yu, C. P. (1993) Models of aerosol behavior in airways and alveoli. In: Moren, F.; Dolovich, M. B.;
2 Newhouse, M. T.; Newman, S. P., eds. Aerosols in medicine: principles, diagnosis and therapy. 2nd rev. ed.
3 Amsterdam, The Netherlands: Elsevier; pp. 157-193.
4 Morrow, P. E.; Gibb, F. R.; Gazioglu, K. M. (1967) A study of paniculate clearance from the human lungs.
5 Am. Rev. Respir. Dis. 96: 1209-1221.
6 Musante, C. J.; Martonen, T. B. (1999) Predicted deposition patterns of ambient paniculate air pollutants in
7 children's lungs under resting conditions. In: Proceedings of the third colloquium on paniculate air pollution
8 and human health; June; Durham, NC. Irvine, CA: University of California, Air Pollution Health Effects
9 Laboratory, p. 7-15 - 7-20.
10 Musante, C. J.; Martonen, T. B. (2000a) Computer simulations of particle deposition in the developing human lung.
11 J. Air Waste Manage. Assoc. 50: 1426-1432.
12 Musante, C. J.; Martonen, T. B. (2000b) An extrapolation model to aid in toxicological studies of paniculate air
13 pollutants. Presented at: PM2000: paniculate matter and health—the scientific basis for regulatory
14 decision-making, specialty conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste
15 Management Association.
16 Musante, C. J.; Martonen, T. B. (2000c) Paniculate matter deposition in the lungs of children and adults:
17 predictions of an age-dependent computer model. Presented at: PM2000: paniculate matter and health—the
18 scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
19 Pittsburgh, PA: Air & Waste Management Association.
20 National Council on Radiation Protection and Measurements. (1997) Deposition, retention and dosimetry of inhaled
21 radioactive substances. Bethesda, MD: National Council on Radiation Protection and Measurements; report
22 no. 125.
23 Naumann, B. D.; Schlesinger, R. B. (1986) Assessment of early alveolar particle clearance and macrophage
24 function following an acute inhalation of sulfuric acid mist. Exp. Lung Res. 11: 13-33.
25 Nemmar, A.; Vanbilloen, H.; Hoylaerts, M. F.; Hoet, P. H. M.; Verbruggen, A.; Nemery, B. (2001) Passage of
26 intratracheally instilled ultrafine particles from the lung into the systemic circulation in hamster. Am. J.
27 Respir. Crit. Care Med. 164: 1665-1668.
28 Nemmar, A.; Hoet, H. M.; Vanquickenborne, B.; Dinsdale, D.; Thomeer, M.; Hoylaerts, M. F.; Vanbilloen, H.;
29 Mortelmans, L.; Nemery, B. (2002) Passage of inhaled particles into the blood circulation in humans.
30 Circulation 105: 411-414.
31 Nikula, K. J.; Avila, K. J.; Griffith, W. C.; Mauderly, J. L. (1997) Lung tissue responses and sites of particle
32 retention differ between rats and cynomolgus monkeys exposed chronically to diesel exhaust and coal dust.
33 Fundam. Appl. Toxicol. 37: 37-53.
34 Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2001) Influence of exposure concentration or dose on
3 5 the distribution of paniculate material in rat and human lungs. Environ. Health Perspect. 109: 311-318.
36 Noone, P. G.; Bennett, W. D.; Regnis, J. A.; Zeman, K. L.; Carson, J. L.; King, M.; Boucher, R. C.; Knowles, M. R.
37 (1999) Effect of aerosolized uridine-5'-triphosphate on airway clearance with cough in patients with primary
38 ciliary dyskinesia. Am. J. Respir. Crit. Care Med.: 160: 144-149.
39 Oberdorster, G. (1993) Lung dosimetry: pulmonary clearance of inhaled particles. Aerosol Sci. Technol.
40 18:279-289.
41 Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
42 lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
43 Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer
44 particles for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
45 Oldham, M. J.; Mannix, R. C.; Phalen, R. F. (1997) Deposition of monodisperse particles in hollow models
46 representing adult and child-size tracheobronchial airways. Health Phys. 72: 827-834.
47 Passali, D.; Bianchini Ciampoli, M. (1985) Normal values of mucociliary transport time in young subjects. Int. J.
48 Pediatr. Otorhinolaryngol. 9: 151-156.
49 Fatten, J. S. (1996) Mechanisms of macromolecule absorption by the lungs. Adv. Drug Delivery Rev. 19: 3-36.
50 Peterson, B. T.; Dickerson, K. D.; James, H. L.; Miller, E. J.; McLarty, J. W.; Holiday, D. B. (1989) Comparison of
51 three tracers for detecting lung epithelial injury in anesthetized sheep. J. Appl. Physiol. 66: 2374-2383.
52 Phalen, R. F.; Oldham, M. J. (2001) Methods for modeling particle deposition as a function of age. Respir. Physiol.
53 128: 119-130.
54 Pritchard, J. N.; Jefferies, S. J.; Black, A. (1986) Sex differences in the regional deposition of inhaled particles in
55 the 2.5-7.5 ^m size range. J. Aerosol Sci. 17: 385-389.
April 2002 6-82 DRAFT-DO NOT QUOTE OR CITE
-------
1 Raabe, O. G.; Al-Bayati, M. A.; Teague, S. V.; Rasolt, A. (1988) Regional deposition of inhaled monodisperse,
2 coarse, and fine aerosol particles in small laboratory animals. In: Dodgson, J.; McCallum, R. I.; Bailey,
3 M. R.; Fischer, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop
4 on lung dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 53-63.
5 Radford, E. P.; Martell, E. A. (1977) Polonium-210: lead-210 ratios as an index of residence time of insoluble
6 particles from cigarette smoke in bronchial epithelium. In: Walton, W. H.; McGovern, B., eds. Inhaled
7 particles IV: proceedings of an international symposium, part 2; September 1975; Edinburgh, United
8 Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 567-581.
9 Rasmussen, T. R.; Andersen, A.; Pedersen, O. F. (2000) Particle deposition in the nose related to nasal cavity
10 geometry. Rhinology 38: 102-107.
11 Roy, M. (1989) Lung clearance modeling on the basis of physiological and biological parameters. Health Phys.
12 57(suppl. 1): 255-262.
13 Rutland, J.; Cole, P. J. (1981) Nasal mucociliary clearance and ciliary beat frequency in cystic fibrosis compared
14 with sinusitis and bronchiectasis. Thorax 36: 654-658.
15 Salma, I.; Balashazy, I.; Winkler-Heil, R.; Hofmann, W.; Zaray, G. (2002) Effect of particle mass size distribution
16 on the deposition of aerosols in the human respiratory system. J. Aerosol Sci. 33: 119-132.
17 Sarangapani, R.; Wexler, A. S. (2000) Modeling particle deposition in extrathoracic airways. Aerosol Sci. Technol.
18 32:72-89.
19 Scheuch, G.; Stahlhofen, W. (1988) Particle deposition of inhaled aerosol boluses in the upper human airways.
20 J. Aerosol Med. 1: 29-36.
21 Scheuch, G.; Philipson, K.; Falk, R.; Svartengren, M.; Stahlhofen, W.; Camner, P. (1995) Retention of particles
22 inhaled inboli with and without induced bronchoconstriction. Exp. Lung Res. 21: 901-916.
23 Schlesinger, R. B. (1985) Clearance from the respiratory tract. Fundam. Appl. Toxicol. 5: 435-450.
24 Schlesinger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
25 emissions. In: Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
26 health. Washington, DC: National Academy Press; pp. 239-298.
27 Schlesinger, R. B. (1989) Deposition and clearance of inhaled particles. In: McClellan, R. O.; Henderson, R. F.,
28 eds. Concepts in inhalation toxicology. New York, NY: Hemisphere Publishing Corp.; pp. 163-192.
29 Schlesinger, R. B. (1995) Deposition and clearance of inhaled particles. In: McClellan, R. O.; Henderson, R. F.,
30 eds. Concepts in inhalation toxicology. 2nd ed. Washington, DC: Taylor & Francis; pp. 191-224.
31 Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
32 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
33 Schroter, R. C.; Sudlow, M. F. (1969) Flow patterns in models of the human bronchial airways. Respir. Physiol.
34 7:341-55.
35 Segal, R. A.; Guan, X.; Shearer, M; Martonen, T. B. (2000a) Mathematical model of airflow in the lungs of children
36 I: effects of tumor sizes and locations. J. Theor. Med. 2:199-213.
37 Segal, R. A.; Martonen, T. B.; Kim, C. S. (2000b) Comparison of computer simulations to total lung deposition to
38 human subject data in healthy test subjects. J. Air Waste Manage. Assoc. 50: 1262-1268.
39 Smaldone, G. C.; Perry, R. J.; Bennett, W. D.; Messina, M. S.; Zwang, J.; Ilowite, J. (1988) Interpretation of
40 "24 hour lung retention" in studies of mucociliary clearance. J. Aerosol Med. 1: 11-20.
41 Snipes, M. B. (1989) Long-term retention and clearance of particles inhaled by mammalian species. CRC Crit. Rev.
42 Toxicol. 20: 175-211.
43 Snipes, M. B.; Clem, M. F. (1981) Retention of microspheres in the rat lung after intratracheal instillation.
44 Environ. Res. 24: 33-41.
45 Snipes, M. B.; McClellan, R. O.; Mauderly, J. L.; Wolff, R. K. (1989) Retention patterns for inhaled particles in the
46 lung: comparisons between laboratory animals and humans for chronic exposures. Health Phys.
47 57(suppl. 1): 69-78.
48 Snipes, M. B.; James, A. C.; Jarabek, A. M. (1997) The 1994 ICRP66 human respiratory tract dosimetry model as a
49 tool for predicting lung burdens from exposures to environmental aerosols. Appl. Occup. Environ. Hyg. 12:
50 547-554.
51 Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G.; Philipson, K. (1986a) Clearance from the human airways of
52 particles of different sizes deposited from inhaled aerosol boli. In: Aerosols: formation and reactivity,
53 proceedings of the second international aerosol conference; September; Berlin, Federal Republic of
54 Germany. Oxford, United Kingdom: Pergamon Press; pp. 192-196.
April 2002 6-83 DRAFT-DO NOT QUOTE OR CITE
-------
1 Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G. (1986b) Measurement of lung clearance with pulses of
2 radioactively-labelled aerosols. J. Aerosol Sci. 17: 333-336.
3 Stanley, P. J.; Wilson, R.; Greenstone, M. A.; Mackay, I. S.; Cole, P. J. (1985) Abnormal nasal mucociliary
4 clearance in patients with rhinitis and its relationship to concomitant chest disease. Br. J. Dis. Chest
5 79:77-82.
6 Stapleton, K. W.; Guentsch, E.; Hoskinson, M. K.; Finlay, W. H. (2000) On the suitability of K-e turbulence
7 modeling for aerosol deposition in the mouth and throat: a comparison with experiment. J. Aerosol Sci.
8 31:739-749.
9 Stober, W.; Morrow, P. E.; Koch, W.; Morawietz, G. (1994) Alveolar clearance and retention of inhaled insoluble
10 particles in rats simulated by a model inferring macrophage particle load distributions. J. Aerosol Sci.
11 25:975-1002.
12 Strom, K. A.; Chan, T. L.; Johnson, J. T. (1988) Pulmonary retention of inhaled submicron particles in rats: diesel
13 exhaust exposures and lung retention model. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.; Fischer, D. R.,
14 eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung dosimetry;
15 September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 645-657.
16 Suarez, S.; Kazantseva, M.; Bhat, M.; Costa, D.; Hickey, A. J. (2001) The influence of suspension nebulization or
17 instillation on particle uptake by guinez pig alveolar macrophages. Inhalation Toxicol. 13: 773-788.
18 Subramaniam, R. P.; Richardson, R. B.; Morgan, K. T.; Guilmette, R. A.; Kimbell, J. S. (1998) Computational fluid
19 dynamics simulations of inspiratory airflow in the human nose and nasopharynx. Inhalation Toxicol.
20 10:91-120.
21 Subramaniam, R. P.; Asgharian, B.; Leendert, V.; Cassee, F. R.; Freijer, J. L; Miller, F. J.; Rombout, P. J. A. (1999)
22 A software package for multiple-path modeling of paniculate matter deposition in human and rat lungs.
23 In: Phalen, R.; Bell, Y., eds. Proceedings of Third Colloquium on Paniculate Air Pollution and Human
24 Health. APHEL report no. 99-01, pp. 7.21-7.30. Irvine, CA: Air Pollution Health Effects Laboratory,
25 University of California.
26 Svartengren, K.; Lindestad, P.; Svartengren, M.; Philipson, K.; Bylin, G.; Camner, P. (1995) Added external
27 resistance reduces oropharyngeal deposition and increases lung deposition of aerosol particles in asthmatics.
28 Am. J. Respir. Crit. Care Med. 152: 32-37.
29 Svartengren, K.; Philipson, K.; Svartengren, M.; Anderson, M.; Camner, P. (1996a) Tracheobronchial deposition
30 and clearance in small airways in asthmatic subjects. Eur. Respir. J. 9: 1123-1129.
31 Svartengren, K.; Ericsson, C. H.; Svartengren, M.; Mossberg, B.; Philipson, K.; Camner, P. (1996b) Deposition and
32 clearance in large and small airways in chronic bronchitis. Exp. Lung Res. 22: 555-576.
33 Svartengren, K.; Philipson, K.; Svartengren, M.; Camner, P. (1998) Effect of adrenergic stimulation on clearance
34 from small ciliated airways in healthy subjects. Exp. Lung Res. 24: 149-158.
35 Svartengren, M.; Svartengren, K.; Aghaie, F.; Philipson, K.; Camner, P. (1999) Lung deposition and extremely slow
36 inhalations of particles. Limited effect of induced airway obstruction. Exp. Lung Res. 25: 353-366.
37 Swift, D. L.; Strong, J. C. (1996) Nasal deposition of ultrafine 218Po aerosols in human subjects. J. Aerosol Sci.
38 27: 1125-1132.
39 Takahashi, S.; Asaho, S.; Kubota, Y.; Sato, H.; Matsuoka, O. (1987) Distribution of 198Au and 133Ba in thoracic and
40 cervical lymph nodes of the rat following the intratracheal instillation of 198Au-colloid and 133BaSO4.
41 J. Radial Res. 28:227-231.
42 Takahashi, S.; Kubota, Y.; Hatsuno, H. (1992) Effect of size on the movement of latex particles in the respiratory
43 tract following local administration. Inhalation Toxicol. 4: 113-123.
44 Takenaka, S.; Karg, E.; Roth, C.; Schulz, H.; Ziesenis, A.; Heinzmann, U.; Schramel, P.; Heyder, J. (2001)
45 Pulmonary and systemic distribution of inhaled ultrafine silver particles in rats. Environ. Health Perspect.
46 109(suppl. 4): 547-551.
47 Toms, N.; Hasani, A.; Pavia, D.; Clarke, S. W.; Agnew, J. E. (1997) Effect of mucus hypersecretion on initial time
48 course of inert particle clearance from the lung. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII:
49 proceedings of an international symposium on inhaled particles organised by the British Occupational
50 Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 509-514.
51 Tran, C. L.; Jones, A. D.; Cullen, R. T.; Donaldson, K. (1999) Exploration of the mechanisms of retention and
52 clearance of low-toxicity particles in the rat lung using a mathematical model. Inhalation Toxicol.
53 11: 1077-1108.
April 2002 6-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 Tran, C. L.; Buchanan, D.; Cullen, R. T.; Searl, A.; Jones, A. D.; Donaldson, K. (2000) Inhalation of poorly soluble
2 particles. II. Influence of particle surface area on inflammation and clearance. Inhalation Toxicol. 12: 1113-1126.
3 Tsuda, A.; Butler, J. P.; Fredberg, J. J. (1994a) Effects of alveolated duct structure on aerosol kinetics. I.
4 Diffusional deposition in the absence of gravity. J. Appl. Physiol. 76: 2497-2509.
5 Tsuda, A.; Butler, J. P.; Fredberg, J. J. (1994b) Effects of alveolated duct structure on aerosol kinetics. II.
6 Gravitational sedimentation and inertial impaction. J. Appl. Physiol. 76: 2510-2516.
7 Tsuda, A.; Henry, F. S.; Otani, Y.; Haber, S.; Butler, J. P. (1996) Aerosol transport and deposition in the
8 rhythmically expanding pulmonary acinus. J. Aerosol. Med. 9: 389-408.
9 U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
10 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/00laF-cF. 3v.
11 Venkataraman, C.; Kao, A. S. (1999) Comparison of particle lung doses from the fine and coarse fractions of urban
12 PM-10 aerosols. Inhalation Toxicol. 11: 151-169.
13 Wagner, E. M; Foster, W. M. (2001) The role of the bronchial vasculature in soluble particle clearance. Environ.
14 Health Perspect. 109(suppl. 4): 563-565.
15 White, F. M. (1974) Viscous fluid flow. New York, NY: McGraw-Hill, pp. 65-72.
16 Wolff, R. K.; Henderson, R. F.; Snipes, M. B.; Griffith, W. C.; Mauderly, J. L.; Cuddihy, R. G.; McClellan, R. O.
17 (1987) Alterations in particle accumulation and clearance in lungs of rats chronically exposed to diesel
18 exhaust. Fundam. Appl. Toxicol. 9: 154-166.
19 Xu, G. B.; Yu, C. P. (1986) Effects of age on deposition of inhaled aerosols in the human lung. Aerosol Sci.
20 Technol. 5: 349-357.
21 Yeates, D. B.; Aspin, M. (1978) A mathematical description of the airways of the human lungs. Respir. Physiol.
22 32:91-104.
23 Yeates, D. B.; Aspin, N.; Levison, H.; Jones, M. T.; Bryan, A. C. (1975) Mucociliary trachea! transport rates in
24 man. J. Appl. Physiol. 19: 487-495.
25 Yeates, D. B.; Pitt, B. R.; Spektor, D. M.; Karron, G. A.; Albert, R. E. (1981) Coordination of mucociliary transport
26 in human trachea and intrapulmonary airways. J. Appl. Physiol.: Respir. Environ. Exercise Physiol.
27 51: 1057-1064.
28 Yeh, H.-C.; Schum, G. M. (1980) Models of human lung airways and their application to inhaled particle
29 deposition. Bull. Math. Biol. 42: 461-480.
30 Yu, G.; Zhang, Z.; Lessmann, R. (1996) Computer simulation of the flow field and particle deposition by diffusion
31 in a 3-D human airway bifurcation. Aerosol Sci. Technol. 25: 338-352.
32 Yu, G.; Zhang, Z.; Lessmann, R. (1998) Fluid flow and particle diffusion in the human upper respiratory system.
33 Aerosol Sci. Technol. 28: 146-158.
34 Zhang, Z.; Martonen, T. (1997) Deposition of ultrafine aerosols in human tracheobronchial airways. Inhalation
35 Toxicol. 9:99-110.
36 Zhang, L.; Asgharian, B.; Anjilvel, S. (1997) Inertial deposition of particles in the human upper airway bifurcations.
37 Aerosol Sci. Tech. 26: 97-110.
3 8 Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2000) Effects of asymmetric branch flow rates on aerosol deposition in
39 bifurcating airways. J. Med. Eng. Technol. 24: 192-202.
40 Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2001) Effects of curved inlet tube on air flow and particle deposition in
41 bifurcating lung models. J. Biomechanics 34: 659-669.
42 Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2002) Cyclic micron-size particle inhalation and deposition in a triple
43 bifurcation lung airway model. J. Aerosol Sci. 33: 257-281.
April 2002 6-85 DRAFT-DO NOT QUOTE OR CITE
-------
i 7. TOXICOLOGY OF PARTICULATE MATTER IN
2 HUMANS AND LABORATORY ANIMALS
3
4
5 7.1 INTRODUCTION
6 Toxicological research on ambient particulate matter (PM) is used to address several
7 related questions, including (1) does exposure to PM at relevant ambient concentrations cause
8 toxicological effects, (2) what mechanisms may be involved in the toxicological response to PM
9 exposure, (3) what factors affect individual or subpopulation susceptibility to the effects of PM
10 exposure, (4) what characteristics of PM (e.g., size, composition) contribute to the observed
11 toxicity, and (5) what are the combined effects of PM and gaseous co-pollutants in producing
12 toxic responses? A variety of research approaches are used to address these questions, including
13 studies of human volunteers exposed to PM under controlled conditions; in vivo studies of
14 laboratory animals such as nonhuman primates, dogs and rodent species; and in vitro studies of
15 tissue, cellular, genetic, and biochemical systems. Similarly, a variety of exposure conditions are
16 employed, including whole body and nose-only inhalation exposures to laboratory generated PM
17 or concentrated ambient PM, tracheal or pulmonary instillation, nasal or nasopharyngeal
18 instillation, and in vitro exposure to test materials in solution or suspension. The various
19 research approaches are targeted to test hypotheses and, ultimately, provide a scientific basis for
20 an improved understanding of the role of PM in producing the health effects identified by
21 epidemiological studies.
22 Because of the sparsity of toxicological data on ambient PM at the time the previous PM
23 Air Quality Criteria Document or "PM AQCD" (U.S. Environmental Protection Agency, 1996a)
24 was completed, the discussion of respiratory effects of PM was organized into specific chemical
25 components of ambient PM or model "surrogate" particles (e.g., acid aerosols, metals, ultrafme
26 particles, bioaerosols, "other particle matter"). In this chapter, the conclusions of the 1996 PM
27 AQCD are summarized for each of these components. Since completion of the previous
28 document, there are many new studies demonstrating the potentially toxic effects of combustion-
29 related particles. The main reason for this increased interest in combustion particles is that these
April 2002 7-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles, along with the secondary aerosols that they form, are typically the dominant sources
2 represented in the fine fraction of ambient PM.
3 This chapter is organized as follows. The respiratory effects of specific components of
4 ambient PM or surrogate particles delivered by in vivo exposures of both humans and laboratory
5 animals are discussed first (Section 7.2), followed by cardiovascular and systemic effects of
6 particles (Section 7.3) and effects in laboratory animal models that mimic human disease
7 (Section 7.4). The in vitro exposure studies are discussed next (Section 7.5) because they
8 provide valuable information on potential hazardous constituents and mechanisms of PM injury.
9 The remaining section on exposure studies examines the health effects of mixtures of ambient
10 PM or PM surrogates with gaseous pollutants (Section 7.6). This organization provides the
11 underlying data for evaluation in the final section of this chapter (Section 7.7), but it may fail to
12 adequately convey the extensive and intricate linkages among the pulmonary, cardiac, and
13 nervous systems, all of which may be involved individually and in concert to represent the effects
14 of exposure to PM.
15
16
17 7.2 RESPIRATORY EFFECTS OF PARTICULATE MATTER IN
18 HEALTHY HUMANS AND LABORATORY ANIMALS: IN VIVO
19 EXPOSURES
20 The following sections assess the respiratory effects of controlled human exposure to
21 various types of PM and also review and evaluate controlled laboratory animal toxicology
22 studies. A discussion of related in vitro studies using animal or human respiratory cells can be
23 found in Section 7.5. The discussion focuses on studies published since completion of the
24 previous 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a).
25 The biological responses occurring in the respiratory tract following controlled PM
26 inhalation include changes in pulmonary inflammation and systemic effects resulting from direct
27 effects on lung tissue. The observed responses may be dependent on the physicochemical
28 characteristics of the PM, the exposure, and the health status of the host. Many of the responses
29 are usually seen only at the higher concentrations characteristic of occupational and laboratory
30 animal exposures and not at (typically much lower) ambient particle concentrations. Moreover,
31 there are substantial differences in the inhalability and deposition profiles of PM in humans and
April 2002 7-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 rodents (see Chapter 6 for details). Observed responses and dose-response relationships also are
2 very dependent on the specific biological response being measured.
3 Paniculate matter is a broad term that encompasses a myriad of physical and chemical
4 species, some of which have been investigated in controlled laboratory animal or human studies
5 (see Table 7-1). However, a full discussion of all types of particles that have been studied is
6 beyond the scope of this chapter. Thus, specific criteria were used to select topics for
7 presentation. High priority was placed on studies that may (1) elucidate health effects of major
8 common constituents of ambient PM or (2) contribute to enhanced understanding of the
9 epidemiological studies. Most studies have been designed to address the question of biologic
10 plausibility, rather than providing dose-response or risk assessment quantification.
11 Diesel particulate matter (DPM) generally fits the criteria; however, because it is described
12 in other documents in great detail (U. S. Environmental Protection Agency, 1999; Health Effects
13 Institute, 1995), it is not covered extensively in this chapter. Particles with high inherent toxicity,
14 such as silica, that are of concern primarily because of occupational exposure, are excluded from
15 this chapter and are discussed in detail in other documents and reports (U.S. Environmental
16 Protection Agency, 1996b; Gift and Faust, 1997; Lippmann, 2000).
17 Most of the laboratory animal studies summarized here have used high particulate mass
18 concentrations administered by inhalation or by intratracheal instillation. The studies have used
19 doses that are generally quite high when compared to ambient levels, even when laboratory
20 animal-to-human dosimetric differences are considered. These high doses are necessary,
21 however, in laboratory animal studies that must explore potentially toxic effects using numbers
22 of subjects (animals) that are magnitudes fewer than those used in epidemiology studies. More
23 research on particle dose extrapolation is needed, therefore, to determine species differences and
24 the importance of exercise and other factors influencing particle deposition in humans that
25 together can account for a 50-fold or more difference in dose.
26 As mentioned earlier, the data available in the previous 1996 PM AQCD were from studies
27 that investigated the respiratory effects of specific components of ambient PM or surrogate
28 particles such as sulfuric acid droplets. More recently, pulmonary effects of controlled exposures
29 to ambient PM have been investigated by the use of particles collected from emission bag room
30 or ambient samplers (e.g., impactors; diffusion denuders) and by the use of aerosol concentrators
31 (e.g., Sioutas et al., 1995a,b, 2000; Gordon et al., 1998; Chang et al., 2000, Kim et al., 2000a,b).
April 2002 7-3 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-1. TYPES OF PARTICIPATE MATTER USED IN
TOXICOLOGICAL STUDIES
Source
category
Particle"
Source'
Label/Date1 Description
Referenced
Concentrated Ambient Particles
ambient
ambient
ambient
ambient
ambient
CAPs
CAPs
CAPs
CAPs
CAPs
New York, NY
Boston, MA
Boston, MA
Chapel Hill, NC;
Research Triangle
Park, NC
Los Angeles, CA
Gerber concentrator; 0.2 to 1.2 ,um
MMAD;
og = 0.2to3.9
Harvard concentrator;
0.2 to 0.3 ^m MMAD;
og = 2.9
1997 Harvard concentrator;
1998 0.23 to 0.34 //m MMAD;
og = 0.2to2.9
1997 Harvard concentrator; 0.65 ,um
1998 MMAD;
og = 2.35
Harvard concentrator; PM2 5
Gordon et al.
(1998; 2000)
Goldsmith et al.
(1998);
Clarke etal.
(1999; 2000a,b)
Godleski et al.
(2000)
Ohio et al.
(2000a);
Kodavanti et al.
(2000a)
Gong et al.,
(2000)
Ambient Particulate Matter Extracts
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
ambient
(aqueous
extracts)
urban PM (StL)
urban PM (Ott)
urban PM (Dus)
urban PM
urban PM (WDC)
urban PM
urban PM
urban CB & UCB
urban dust
St. Louis, MO
Ottawa, ONT
Dusseldorf,
Germany
Terni, Italy
Washington, DC
Provo, UT
Provo, UT
Edinburgh, UK
NIST;
Gaithersburg, MD
SRM 1648
EHC-93; videlon filter samples, mechanically
1993 sieved (36, 56, 80, 100, 300 ^m), and
stored at -80 °C
SRM 1649
1981,1982 TSP collected on glass-fiber filters,
suspended in aqueous medium,
centrifuged, lyophilized, and
resuspended in saline.
1986, 1987, TSP and PM10 collected on glass hi-
1988 vol filters, suspended, centrifuged,
lyophilized, and resuspended in
saline.
Dong et al.,
(1996); Becker
and Soukup
(1998)
Vincent et al.
(1997; 2001)
Costa and Dreher
(1997)
Fabiani et al.
(1997)
Becker and
Soukup (1998)
Kennedy et al.
(1998); Ohio
etal. (1999a,b)
Dye et al.,
(2001); Ohio and
Devlin, (2001)
Li etal. (1996;
1997)
April 2002
7-4
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-1 (cont'd). TYPES OF PARTICIPATE MATTER USED IN
TOXICOLOGICAL STUDIES
Source
category
Particle'
Source'
Label/Date1
Description
Referenced
Complex Combustion-Related Particulate Matter
stationary coal fly ash (CFA)
stationary oil fly ash (OFA)
stationary residual oil fly ash
(ROFA)
residential domestic oil fly ash
(DOFA)
residential wood stove
mobile diesel exhaust particles
(DEP)
mobile diesel particulate matter
(DPM)
U.S. power plants
Niagra power
plant
variable
Watkinson et al.
(1998; 2000a,b);
Campen et al.
(2000);
Muggenburg
et al. (2000)
home oil-burning
furnace
Durham, NC
Laboratory-Derived Surrogate Particulate Matter
ambient
(simulated)
ambient
(simulated)
stationary
(simulated)
mobile
natural
natural
inorganic
organic
acid aerosols
(e.g., H2S04)
bioaerosols (e.g.,
lipopolysaccharide, LPS)
inorganic metal oxides
(CdO, Fe2O, MnO2,
NiSO4, TiO2, V2O5, ZnO)
diesel soot NIST; SRM 1650
Gaithersburg, MD
Mt. St. Helens ash Ritzville, WA
(MSH)
coal dust
carbon black (CB)
synthetic polymer
microspheres (SPM)
See Table 7-2
See Table 7-7
See Table 7-3
"Particle Notes:
1. See Tables 7-4, 7-5, 7-6 and 7-8 for description and additional information on studies using ambient PM and PM substitutes.
2. See Table 7-2 for description and additional information on studies using acid aerosols.
3. See Table 7-3 for description and additional information on studies using metal oxides.
4. See Table 7-7 for description and additional information on studies using ambient bioaerosols.
5. For additional information on Diesel PM (DPM) or Diesel exhaust particles (DEP), see U.S. Environmental Protection Agency
(2000) and Health Effects Institute (1995).
6. UCB = fine or ultrafine urban carbon black particles.
b Source Notes:
1. Aerosol concentrators (e.g., Harvard; Gerber) were used to generate CAPs.
2. Particle samplers (e.g., impactors, diffusion denuders) were used to collect ambient PM.
3. NIST = National Institute of Standards and Technology.
'Label/Date Notes:
• SRM = standard reference material.
• EHC = Environmental Health Center in Ottawa, Canada.
Date of particle collection, when available.
dReference: Not an exhaustive list; see text for details.
April 2002
7-5
DRAFT-DO NOT QUOTE OR CITE
-------
1 Some ambient PM has been standardized as a reference material and compared to existing dust
2 and soot standards [e.g., National Institutes of Standards and Technology (NIST)]. Both ambient
3 PM and CAPs have been used to investigate effects in normal and compromised laboratory
4 animals and humans.
5 Particles from ambient air samplers are collected on filters or other media, stored, and
6 resuspended in an aqueous medium for use in experimental, tracheal installation, or in vitro
7 studies. The in vivo and in vitro studies discussed in this chapter have almost exclusively used
8 PM10 or PM2 5 as particle size cutoffs for studying the adverse effects of ambient PM. Studying
9 only particles less than a certain size is justified based upon earlier interests in setting standards
10 for PM10 and PM25. In addition, the collection of these size fractions is made easier by the
11 widespread availability of ambient sampling equipment for PM10 and PM2 5. Unfortunately, the
12 study of other important size fractions, such as the coarse fraction (PM10_2 5) and PMl 0 has been
13 largely ignored and little toxicology data are available to specifically address these potentially
14 important particle sizes. Similarly, organic compounds make up 20 to 60% of the dry fine
15 particle mass of ambient PM (Chapter 3, Section 3.2), yet very little research has addressed the
16 mechanisms by which this organic fraction contributes to the adverse effects associated with
17 acute exposure to PM. The potential contribution of organics in mutagenesis and carcinogenesis
18 has been studied, but these findings are not discussed within the context of this chapter which is
19 focused on understanding the epidemiologic evidence of increased cardiopulmonary morbidity
20 and mortality associated with acute exposure to ambient PM.
21 Particle concentrators provide a technique for exposing animals or humans by inhalation to
22 concentrated ambient particles (CAPs) that are 5 to 10-fold higher than typical ambient PM
23 levels. The development of particle concentrators has permitted the study of true ambient PM
24 under controlled conditions. This strength is somewhat weakened by the inability of CAPs
25 studies to precisely control the mass concentration and day-to-day variability in ambient particle
26 composition. Nonetheless, these studies are invaluable in the attempt to understand the
27 biological mechanisms responsible for the cardiopulmonary response to inhaled PM. Because
28 the composition of concentrated ambient PM varies in both time and location, a thorough
29 physical-chemical characterization is necessary to compare results among studies or even among
30 exposures within studies or to link particle composition to effect.
31
April 2002 7-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.2.1 Ambient Combustion-Related and Surrogate Particulate Matter
2 In vivo toxicology studies utilizing inhalation exposure as a technique for measuring the
3 respiratory effects of ambient particles in humans and laboratory animals have been conducted
4 with CAPs (see Table 1) and with DPM. The majority of the in vivo exposures have utilized
5 intratracheal instillation techniques. Discussions on the pros and cons of this technique in
6 comparison to inhalation are covered in Chapter 6 (Section 6.5), and these issues have also been
7 reviewed elsewhere (Driscoll et al., 2000; Oberdorster et al., 1997; Osier and Oberdorster, 1997).
8 In most of the studies, PM samples were collected on filters, resuspended in a vehicle (usually
9 saline), and a small volume of the suspension was instilled intratracheally into the animals. The
10 physiochemical characteristics of PM may be altered by deposition on a filter and resuspension in
11 an aqueous medium. In addition, the doses used in these instillation studies are generally high
12 compared to ambient concentrations, even when laboratory animal-to-human dosimetric
13 differences are considered. Therefore, in terms of direct extrapolation to humans in ambient
14 exposure scenarios, greater importance should be placed on inhalation studies. However,
15 delivery of PM by instillation has the advantages that much less material is needed and that the
16 delivered dose can be determined directly without extrapolating from estimates of lung
17 deposition. Instillation studies have proven valuable in comparing the effects of different types
18 of PM and for investigating some of the mechanisms by which particles may cause lung injury
19 and inflammation. Tables 7-2, 7-3, and 7-4 outline studies in which various biological endpoints
20 were measured following exposures to CAPs, ambient PM extracts, complex combustion-related
21 PM, or laboratory-derived surrogate PM, respectively.
22 There were only limited data available from human studies or laboratory animal studies on
23 ultrafine particles and even less on coarse particles at the time of the release of the previous
24 criteria document (U.S. Environmental Protection Agency, 1996a). In vitro studies have shown
25 that ultrafine particles have the capacity to cause injury to cells of the respiratory tract. High
26 levels of ultrafine particles, as metal or polymer "fume," are associated with toxic respiratory
27 responses in humans and other mammals. Such exposures are associated with cough, dyspnea,
28 pulmonary edema, and acute inflammation. At concentrations less than 50 //g/m3, freshly
29 generated insoluble ultrafine PTFE fume particles can be severely toxic to the lung. However, it
30 was not clear what role in the observed effects was played by fume gases which adhered to the
31 particles. Newer data from controlled exposures have demonstrated that particle composition, in
April 2002 7-7 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-2. RESPIRATORY EFFECTS OF AMBIENT PARTICIPATE MATTER
13.
to
o
o
to
1
oo
o
W
p>
H
6
O
H
O
Cj
O
w
L-LJ
o
o
1 — I
H
W
Species, Gender,
Strain, Age, etc.
Rats, male S-D
200-225 g,
control and
SO2-treated
Rats, male S-D
60 days
Humans, healthy
nonsmokers;
21 M, 3F;
26.4±2.2yrold
Rats, S-D
60 days
Humans, healthy
nonsmokers;
18 to 40 yr old
Mongrel dogs,
some with balloon
occluded LAD
coronary artery
n= 14
Humans, healthy;
n=4, 19-41 yr old
Rats, male F 344
Hamsters, male,
8-mo-old Bi TO-2
Rats, male, 90 to
100-day-old S-D,
with or without
SO2-induced
bronchitis
Rats, Wis
(HAN strain)
Particle"
Concentrated
ambient particles
(CAPs)
(Boston)
Provo, UT,
TSP filters
(10 years old)
Provo, UT,
PM10 filters
(10 years old)
Provo, UT,
TSP filters
(10 years old),
soluble and
insoluble extracts
CAPs
(Chapel Hill)
CAPs
(Boston)
CAPs
(LA)
CAPs
(NY)
CAPs
(RTF)
Ambient PM
Edinburgh, CB,
CB Ultrafine
(UCB)
Exposure
Technique
Inhalation;
Harvard/EPA
fine particle
concentrator;
animals restrained
in chamber
Intratracheal
instillation
Intrabronchial
instillation
Intratracheal
instillation
Inhalation
Inhalation via
tracheostomy
Inhalation
Inhalation
Inhalation
Intratracheal
instillation
Concentration
206,733, 607 //g/m3 for
Days 1-3; 29 °C,
59% RH
0.25, 1.0, 2.5, 5.0 mg of
PM extract in 0.3 mL
saline
500 Mg of PM extract in
10 mL saline
100-1000//gofPM
extract in 0.5 mL saline
23.1 to 311.1 //g/m3
69-828 //g/m3
148-246 Mg/m3
132 to 919 //g/m3
650 Mg/m3
50-125 //gin 0.2 mL
Particle Size Exposure Duration Effect of Particles
0.18//m 5 h/day for 3 days PEF and TV increased in CAPS exposed animals.
og = 2.9 Increased protein and % neutrophils and
lymphocytes in lavage fluid after CAPS exposure.
Responses were greater in SO2-bronchitis animals.
No changes in LDH. No deaths occurred.
N/A 24 h Inflammation (PMN) and pulmonary injury was
produced by particles collected while the steel
mill was in operation
N/A 24 h BAL Inflammation (PMN) and pulmonary injury was
produced by particles collected while the steel
mill was in operation
N/A 24 h Inflammation (PMN) and lavage fluid protein was
greater with the soluble fraction containing more
metal (Zn, Fe, Cu).
0.65 /mi 2 h; analysis at 18 h Increased BAL neutrophils in both bronchial and
og = 2.35 alveolar fractions
0.23 to 0.34 /mi 6 h/day x 3 days Decreased respiratory rate and increased lavage fluid
og = 0.2 to 2.9 neutrophils in normal dogs.
PM2 5 2 h No significant changes in lung function, symptoms,
SaO2, or Holter ECGs were observed.
0.2 to 1.2 /mi Ix3hor3x6h No inflammatory responses, no cell damage
og = 0.2 to 3.9 responses, no PFT changes.
6 h/day x 2-3 days No significant changes in healthy rats; increased
BALF protein and neutrophil influx in bronchitic
rats; responses were variable between exposure
regimens.
PM10 Sacrificed at 6 h Increased PMN, protein, and LDH following PM10;
CB = (200-500 nm) greater response with ultrafme CB but not CB;
UCB = 20 nm decreased GSH level in BAL; free radical activity
(deplete supercoil DNA); leukocytes from treated
animals produced greater NO and TNF.
Reference
Clarke et al.
(1999)
Dye etal. (2001)
Ohio and Devlin
(2001)
Ohio et al.
(1999a)
Ohio et al.
(2000a)
Godleski et al.
(2000)
Gong et al.
(2000)
Gordon et al.
(2000)
Kodavanti et al.
(2000a)
Li etal. (1996,
1997)
"See Table 1 for details
-------
>
to
o
o
to
TABLE 7-3. RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICIPATE MATTER
Species, Gender, Strain,
Age, etc.
Particle"
Exposure
Technique
Concentration
Particle Size
Exposure
Duration
Effect of Particles
Reference
Hamsters, Syrian golden,
male, 90-125g
Kuwaiti oil fire Intratracheal
particles; instillation
urban particles
from St. Louis,
MO
0.15, 0.75, and 3.75 Oil fire particles: Sacrificed 1 and
mg/lOOg -c 3.5 ,um, 10 days of 7 days post
24-h samples instillation
Increases in PMN, AM, albumin, LDH,
myeloperoxidase, and
p-N-acetylglucosaminidase;
acute toxicity of the particles found in the
smoke from the Kuwaiti oil fires is
comparable to that of urban particles.
Brain etal. (1998)
Mice, female, NMRI,
28-32g
Rats, male, S-D,
60 days old
CFA Intratracheal
CMP instillation
we
Emission Intratracheal
source PM instillation
(ROFA,
DOFA, CFA)
Ambient
airshed PM
ROFA
CMP: 20 Mg N/A
arsenic/kg, or CMP:
100 mg
particles/kg,
WC alone
(100 mg/kg), CFA
alone (100 mg/kg
[i.e., 20 //g
arsenic/kg]), CMP
mixed with WC
(CMP, 13.6 mg/kg
[(i.e., 20 //g
arsenic/kg]), WC
(86. 4 mg/kg) and
Ca3(AsO4)2 mixed
with WC (20 ,ug
arsenic/kg), WC
(100 mg/kg)
Total mass: Emission PM:
2.5mg/rat 1. 78-4.17 ^m
Total transition Ambient PM:
metal: 46 Mg/rat 3. 27-4.09 ^m
1, 5, 30 days
posttreatment,
lavage for total
protein content,
inflammatory
cell number and
type, and TNF-
a production
particle
retention
Analysis at 24
and 96 h
following
instillation
Rats, male WISTAR
Bor: WISW strain
Coal oil fly ash
Inhalation
(chamber)
0, 11, 32, and
103 mg/m3
1.9-2.6 Mm
og= 1.6-1.8
6 h/day,
5days/week,
4 weeks
Mild inflammation for WC; Ca3(AsO4)2 Broeckaert et al.
caused significant inflammation; (1997)
CMP caused severe but transient
inflammation; CFA caused persistent
alveolitis; cytokine production was
upregulated in WC- and Ca3(AsO4) treated
animals after 6 and 30 days, respectively;
a 90% inhibition of TNF-a production still
was still observed at Day 30 after
administration of CMP and CFA;
a significant fraction persisted (10-15% of
the arsenic administered) in the lung of
CMP- and CFA-treated mice at Day 30.
Suppression of TNF-a production is
dependent on the slow elimination of the
particles and their metal content from the
lung
Increases in PMNs, albumin, LDH, PMN, Costa and Dreher
and eosinophils following exposure to (1997)
emission and ambient particles;
induction of injury by emission and
ambient PM samples is determined
primarily by constituent metals and their
bioavailability.
At the highest concentration, type II cell Dormans et al.
proliferation and mild fibrosis occurred and (1999)
increased perivascular lymphocytes were
seen. The main changes at the lowest
concentration were particle accumulation in
AM and mediastinal lymph nodes.
Lymphoid hyperplasia observed at all
concentrations. Effects increased with
exposure duration.
-------
TABLE 7-3 (cont'd). RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICULATE MATTER
^
to
o
o
to
1
o
o
J>
"Tl
H
6
o
o
^^
H
O
CH
O
, 1
a
L-LJ
o
7^
o
H
w
Species, Gender, Strain,
Age, etc.
Rats, male, S-D,
60 days old
Rats, S-D, 5-day-old
Rats, male, S-D rats
60 days old
Rats, male, S-D,
5-day-old
Rats, male, S-D,
60 days old
Mice, female,
Balb/cJ7-15 weeks
Mice, female,
7-week-old Balb/cJ
(16-21 g)
Rats, male, S-D
Mice, normal and Hp,
105 days old
Mice, BALB/C,
2-day-old, sensitized to
ovalbumin (OVA)
Particle"
ROFA
ROFA
#6 ROFA,
volcanic ash
LowS
#6 ROFA,
volcanic ash
saline
Two ROFA
samples
Rl had 2x
saline-
leachable
sulfate, Ni, and
Vand40xFe
as R2; R2 had
3 1 x higher Zn
ROFA
ROFA lo-S
residual oil
ROFA
ROFA
Aerosolized
ROFA leachate
Exposure
Technique
Intratracheal
instillation
Intratracheal
Instillation
Intratracheal
Instillation
Intratracheal
Instillation
Intratracheal
instillation
Intratracheal
instillation
Inhalation and
intratracheal
instillation
challenge with
OVA
Intratracheal
instillation
Intratracheal
instillation
Nose-only
inhalation
Concentration Particle Size
8.33 mg/mL 1 .95 Mm MMAD
0.3 mL/rat
500 Mg/rat 1 .95 Mm MMAD
0.3,1.7 1.95 Mm
8.3 mg/mL og = 2.19
8.3 mg/mL 1.4 Mm
0.3,1.7, 1.95 Mm
8.3mg/kgBW og=1.95
in saline 1 .4 Mm
8.3 mg/kg BW
1 mL/kg BW
2.5mgin0.3mL Rl: 1.88 Mm,
MMAD
R2: 2.03 Mm,
MMAD
60 Mg in 50 ML < 2.5
(dose 3mg/kg)
158±3mg/m3 PM25 sample
500 Mg/animal 3. 6 Mm
50 Mg 1.95 Mm
50 mg/mL N/A
Exposure
Duration
Analysis at
24 and 96 h
24h
24 h
24 h
Analysis at
4 days
Analysis at
1-15 days after
instillation
1, 3, 8, 15 days
after
instillation
Analyzed
4 and 96 h
postexposure
Analysis at
24 h
30min
Effect of Particles
Increased PMNs, protein, LDH at both time points;
bioavailable metals were causal constituents of
pulmonary injury.
Increased neutrophilic inflammation was inhibited
by DMTU treatment, indicating role of ROS.
Plasma fibrinogen elevated after ROFA instillation
but not volcanic ash
Increased WBC count in ROFA-exposed rats
plasma fibrinogen increased 86% in ROFA rats at
highest concentration.
Four of the 24 animals treated with R2 or R2s
(supernatant) died; none in Rl s treated animals;
more AM, PMN, eosinophils protein, and LDH in
R2 and R2s animals; more focal alveolar lesions,
thickened alveolar septae, hyperplasia of type II
cells, alveolar fibrosis in R2 and R2s animals;
baseline pulmonary function and airway
hyperreactivity were worse in R2 and R2s groups.
ROFA caused increases in eosinophils, IL-4 and
IL-5 and airway responsiveness in ovalbumin-
sensitized and challenged mice.
Increased BAL protein and LDH at 1 and 3 days
but not at 15 days postexposure. Combined OVA
and ROFA challenge increased all damage markers
and enhanced allergen sensitization. Increased
methacholine response after ROFA.
Ferritin and transferrin were elevated; greatest
increase in ferritin, lactoferrin, transferrin occurred
24 h postexposure.
Diminished lung injury (e.g., decreased lavage
fluid ascorbate, protein, lactate dehydrogenase,
inflammatory cells, cytokines) in Hp mice lacking
transferrin; associated with increased metal storage
and transport proteins.
Increased airway response to methylcholine and
to OVA in ROFA exposed mice; increased airway
inflammation also.
Reference
Dreher et al.
(1997)
Dye etal. (1997)
Gardner et al.
(2000)
Gardner et al.
(2000)
Gavett et al.
(1997)
Gavett et al.
(1999)
Gavett et al.
(1999)
Ohio et al.
(1998b)
Ohio et al.
(2000b)
Hamada et al.
(1999)
-------
>
to
o
o
to
TABLE 7-3 (cont'd). RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICIPATE MATTER
Species, Gender, Strain,
Age, etc. Particle*
Rats, male, S-D, ROFA
60 days old
Rats, S-D, 250 g FOFA
MCT
Exposure
Technique
Intratracheal
instillation
Inhalation
Concentration
l.OmginO.5 mL
saline
580 ± 110Mg/m3
Particle Size
1.95 Mm
2.06 Mm MMAD
og=1.57
Exposure
Duration
Analysis at
24 h
6 h/day for
3 days
Effect of Particles
Increased PMNs, protein.
Death occurred only in MCT rats exposed to
ROFA. Neutrophils in lavage fluid were
increased significantly in MCT rats exposed to
ROFA versus filtered air. MIP-2 mRNA
expression in lavage cells was induced in normal
animals exposed to fly ash.
Reference
Kadiiska et al.
(1997)
Killingsworth
etal. (1997)
Rats, male, S-D and ROFA Intratracheal
F-344 (60 days old) instillation
Rats, male, S-D, WIS, ROFA Intratracheal
and F-344 (60 days old) instillation
1.95 Mm Sacrificed at Increase in neutrophils in both S-D and F-344
og = 2.14 24 h rats; a time-dependent increase in eosinophils
occurred in S-D rats but not in F-344 rats.
1.95 Mm Sacrificed at 6, Inflammatory cell infiltration, as well as alveolar,
og = 2.14 24, 48, and airway, and interstitial thickening in all three rat
72 h; 1, 3, and strains; a sporadic incidence of focal alveolar
12 weeks fibrosis in S-D rats, but not in WIS and F-344
rats; cellular fibronectin (cFn)
mRNA isoforms EIIIA(+) were up-regulated in S-
D and WIS rats but not in F-344 rats. Fn mRNA
expression by macrophage, alveolar and airway
epithelium, and within fibrotic areas in S-D rats;
increased presence of Fn EIIIA(+) protein in the
areas of fibrotic injury and basally to the airway
epithelium.
Kodavanti et al.
(1996)
Kodavanti et al.
(1997a)
Rats, male, S-D,
60 days old
Rats, male, S-D,
60 days old
ROFA Intratracheal
instillation
Fe2(S04)3,
VS04,
NiSO4
10 Intratracheal
compositionally instillation
different ROFA
particles from a
Boston power
plant
8.33 mg/kg 1.95 Mm
og = 2.14
ROFA-equivalent
dose of metals
0.833,3.33,8.3 1.99-2.59 Mm
mg/kg MMAD
Analysis at 3,
24, and 96 h,
postinstillation
Sacrificed at
24 h
ROFA-induced pathology lesions were as severe
as those caused by Ni. Metal mixture caused less
injury than ROFA or Ni alone; Fe was less
pathogenic. Cytokine and adhesion molecule
gene expression occurred as early as 3 h after
exposure. V-induced gene expression was
transient but Ni caused persistent expression and
injury.
ROFA-induced increases in BAL protein and
LDFI, but not PMN, were associated with water-
leachable total metal, Ni, Fe, and S; BALF
neutrophilic inflammation was correlated with V
but not Ni or S. Chemiluminescence signals in
vitro (AM) were greatest with ROFA containing
soluble V and less with Ni plus V.
Kodavanti et al.
(1997b)
Kodavanti et al.
(1998a)
-------
TABLE 7-3 (cont'd). RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICIPATE MATTER
to
o
o
to
to
H
6
o
o
H
O
o
HH
H
W
Species, Gender, Strain,
Age, etc. Particle"
Rats, male, S-D ROFA
60-day-old treated with
MCT (60 mg/kg)
Rats, male, WKY and ROFA
SH, 11-13 weeks old
Exposure
Technique
Intratracheal
instillation;
Nose-only
inhalation
Nose-only
Inhalation
Concentration
0, 0.83,
3.3 mg/kg
15 mg/m3
15 mg/m3
Exposure
Particle Size Duration
1.95,um 24-96 h
og = 2.14
6 h/day for
3 days analysis
at 0 or 18 h
1 .95 ^im 6 h/day x 3 day,
og = 2. 14 analysis at 0 or
18h
Effect of Particles
Both IT and IN rats showed inflammatory
responses (IL-6, MIP-2 genes upregulated).
58% of IT rats exposed to ROFA died within 96 h.
No mortality occurred by inhalation. ROFA
exacerbated lung lesions (edema, inflammation,
alveolar thickening) and gene expression in MCT
rats.
More pulmonary injury in SH rats. Increased
RBCs in BALF of SH rats. ROFA increased
airway reactivity to Ach in both SH and WKY
Reference
Kodavanti et al.
(1999)
Kodavanti et al.
(2000b)
Rats, male, WKY and
SH, 11-13 weeks old
rats. Increased protein, albumin, and LDH in
BALF after ROFA exposure (SH>WKY).
Increased oxidative stress in SH rats. SH rats
failed to increase glutathione. Inflammatory
cytokine gene expression increased in both SH and
WKY rats.
ROFA Intratracheal 3.33 1.95,um 1 and 4 days; Increased BALF protein and LDH alveotis with
Instillation mg/mL/kg og = 2.14 post instillation macrophage accumulation in alveoli; increased
analysis at 6 or neutrophils in BAL. Increased pulmonary protein
VSO4, 1.5|/molkg 24 h leakage and inflammation in SH rats. Effects of
NiSO4, or metal constituents of ROFA were strain specific;
saline vanadium caused pulmonary injury only in WKY
rats; nickel was toxic in both SH and WKY rats.
Kodavanti et al.
(2001)
Rats, Brown Norway
Rats, male, S-D,
60-day-old
Rats, male, S-D,
60-day-old
Rats, male, S-D;
60 days old
Rats, male, S-D,
60-day-old
ROFA
#6 ROFA from
Florida
NC ROFA;
Domestic oil
fly ash
#6 ROFA
(Florida)
NiSO4
VS04
ROFA
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
200 Mg N/A
100 Mg
1000 Mg in 0.5 ml 1.95 ±0. 18 ^m
1000 Mg in
0.5 mL saline
3.3 mg/ml/kg; 1 .9 ,um
ROFA equivalent og = 2. 14
dose of metals
400-1000 Mg/mL N/A
N/A ROFA enhanced the response to house dust mite
(HDM) antigen challenge. Eosinophil numbers,
LDH, BAL protein, and IL-10 were increased with
ROFA + HDM versus HDM alone.
15 min to 24 h Production of acetaldehyde increased at 2 h
postinstillation.
15 min to 24 h ROFA induced production of acetaldehyde with a
peak at about 2 h. No acetaldehyde was seen in
plasma at any time. DOFA increased
acetaldehyde, as did V and Fe.
3 or 24 h Inflammatory and stress responses were
upregulated; the numbers of genes upregulated
were correlated with metal type and ROFA
12 h post-IT ROFA increased PGE2 via cycloxygenase
expression.
Lambert et al.
(1999)
Madden et al.
(1999)
Madden et al.
(1999)
Nadadur et al.
(2000); Nadadur
and Kodavanti
(2002)
Samet et al.
(2000)
-------
TABLE 7-3 (cont'd). RESPIRATORY EFFECTS OF COMPLEX COMBUSTION-RELATED PARTICIPATE MATTER
^
to
o
o
to
Species, Gender, Strain,
Age, etc.
Rats, male, S-D,
60-day-old
Rats, male, S-D;
60-day-old; WKY and
SH; cold-stressed SH,
ozone-exposed SH, and
MCT-treated SH
Particle"
LoS,
#6 ROFA
Ottawa dust,
ROFA, and
volcanic ash
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation, nose-
only inhalation
Concentration
500 Mg in 0.5 ml
saline
Dose: IT 0, 0.25,
1.0, and
2.5 mg/rat; INH
15 mg/m3
Exposure
Particle Size Duration
3.6 /j,m 1, 4, or24h
1.95 tan 6h/dayfor
3 -day
inhalation;
instillation -
96 h post-IT
Effect of Particles
Mild and variable inflammation at 4 h;
no pronounced inflammation until 24 h when there
were marked increases in P-Tyr and P-MARKS.
IT ROFA caused acute and dose-related increase
in pulmonary inflammation; no effect of volcanic
ash.
Reference
Silbajoris et al.
(2000)
Watkinson et al.
(2000a,b)
H
6
o
o
H
O
"See Table 1 for details (CFA = Coal fly ash; CMP = Copper smelter dust; WC = Tungsten carbide; MCT = Monocrotaline; DOFA = Fly ash from a domestic oil-burning furnace; Fe2(SO4) = Iron
sulfate; V SO4 = Vanadium sulfate; NiSO4 = Nickel sulfate; LoS = low sulfur)
o
HH
H
W
-------
>
to
o
o
to
TABLE 7-4. RESPIRATORY EFFECTS OF SURROGATE PARTICIPATE MATTER
Species, Gender, Strain,
Age, etc.
Hamsters, Syrian golden
900 male, 900 female,
4-wks-old
Mice, C57B1/6J
Rats, male, F-344
200-230 g
Mice, male, C57BL/6J,
8 weeks and 8-mo-old
Rats, male, S-D,
MCT-treated
Rats, male, S-D
(200g)
Mice, male, Swiss-Webster,
6-8 weeks old (A/J, AKR/J,
B6C3F1/J, BALB/cJ,
C3H/HeJ-C3, C3HeOuJ,
CSTBL/6J-B6, SJL/J,
SWR/J, 129/J) strains
raised in a pathogen free
laboratory
Particle"
Toner
(carbon)
TiO2
Silica
PTFE
TiO2
PTFE Fumes
PTFE Fumes
Fluorescent
microspheres
Diesel,
SiO2,
carbon black
Carbon black
Regal 660
Carbon-
associated
so4-
Exposure
Technique
Nose-only
inhalation
Inhalation
Whole body
inhalation
Whole body
inhalation
Inhalation
Intratracheal
instillation
Nose only
inhalation
Concentration
1.5, 6.0, or
24 mg/m3
40 mg/m3
3 mg/m3
PTFE:
1.25, 2.5, or
5xl05
particles/cc
TiO2-F: 10 mg/m3
NiO: 5 mg/m3
Ni3S2: 0.5 mg/m3
1, 2.5, or 5 x 105
particles/cm3
1, 2.5, or 5 x 105
particles/cm3
3. 85 ±0.81
mg/m3
1 mg in 0.4 mL.
10 mg/m3
285 Mg/m3
Particle Size
4.0 Mm
1.1 A^m
1.4 ,um
PTFE: 18nm
TiO2-F: 200 nm
TiO2-D: 10 nm
18 nm
18 nm
1.38 ±0.10 Ann
og = 1.8 ±0.28
DEP Collected as
TSP - disaggregated
in solution by
sonication (20 nm);
SiO2 (7 nm);
carbon black
0.29 Aim
± 2.7 Aim
Exposure
Duration
3, 9, 15 mo
6 h/day
5days/week
30 min or
6 h/day,
5days/week,
6 mo
1 5 min,
analysis 4 h
postexposure
30-min
exposure,
analysis 6 h
following
exposure
3 h/day
x 3 days
Necropsy at 2,
7, 21, 42, and
84 days
postinstillation
4h
Effect of Particles
Retention increased with increased exposure.
Clearance halftimes retarded (males)
Effects on the epithelium caused by direct
interactions with particles, not a result of
macrophage-derived mediators, and suggest
a more significant role in the overall pulmonary
response than previously suspected; type II cell
growth factor production may be significant in
the pathogenesis of pulmonary fibrosis.
Increased PMN, mRNA of MnSOD and MT,
IL-la, IL-lp, IL-6, MIP-2, TNF-a mRNA of
MT and IL-6 expressed around all airways and
interstitial regions; PMN expressed IL-6, MT,
and TNF-a; AM and epithelial cells were
actively involved.
Increased PMN, lymphocytes, and protein
levels in old mice over young mice; increased
TNF-a mRNA in old mice over young mice;
no difference in LDH and p-Glucuronidase.
Monocrotaline-treated animals contained fewer
microspheres in their macrophages, probably
because of impaired chemotaxis.
Amorphous SiO2 increased permeability, and
neutrophillic inflammation. Carbon black
and DEP translocated to interstitum and lymph
nodes by 12 weeks.
Differences in inflammatory responses
(PMN) across strains. Appears to be genetic
component to the susceptibility.
Reference
Creutzenberg et al.
(1998)
Finkelstein et al.
(1997)
Johnston et al.
(1996)
Johnston et al.
(1998)
Madletal. (1998)
Murphy etal. (1998)
Ohtsuka et al.
2000a,b
aSee Table 1 for details (PTFE = polytetrafluoroethylene; TiO2 = titanium oxide; SiO2 = silicon dioxide)
-------
1 addition to particle size, may be responsible for the adverse health effects associated with
2 ambient PM exposures.
3 Toxicologic studies of other particulate matter species also were discussed in the previous
4 criteria document (U.S. Environmental Protection Agency, 1996a). These studies included
5 exposures to fly ash, volcanic ash, coal dust, carbon black, and miscellaneous other particles,
6 either alone or in mixture. Some of the particles discussed were considered to be models of
7 "nuisance" or "inert" dusts (i.e., those having low intrinsic toxicity) and were used in instillation
8 studies to delineate nonspecific particle effects from effects of known toxicants. A number of
9 studies on "other PM" examined effects of up to 50,000 //g/m3 of respirable particles with
10 inherently low toxicity. Although there was no mortality, some mild pulmonary function
11 changes after exposure to 5,000 to 10,000 //g/m3 of inert particles were observed in rats and
12 guinea pigs. Lung morphology studies revealed focal inflammatory responses, some epithelial
13 hyperplasia, and fibrotic responses after exposure to >5,000 //g/m3. Changes in macrophage
14 clearance after exposure to >10,000 //g/m3 were equivocal (no host defense effects). In studies of
15 mixtures of particles and other pollutants, effects were variable depending on the toxicity of the
16 associated pollutant. In humans, co-exposure to carbon particles appeared to increase responses
17 to formaldehyde but not to acid aerosol. None of the "other" particles mentioned above are
18 present in ambient air in more than trace quantities. Thus, it was concluded that the relevance of
19 any of these studies to standard setting for ambient PM may be extremely limited (see Chapter 6,
20 Section 4, Particle Overload).
21
22 7.2.1.1 Ambient Particulate Matter
23 Studies that examined the acute effects of intratracheal instillation of ambient PM obtained
24 from specific ambient sources have shown clearly that PM can cause lung inflammation and
25 injury. Costa and Dreher (1997) showed that instillation of relatively high concentrations of PM
26 samples from three emission sources (two oil and one coal fly ash) and four ambient airsheds (St.
27 Louis, MO; Washington, DC; Dusseldorf, Germany; and Ottawa, Canada) resulted in increases in
28 lung polymorphonuclear leucocytes (PMNs) and eosinophils in rats 24 h after instillation.
29 Biomarkers of permeability (total protein and albumin) and cellular injury (LDH) also were
30 increased. This study demonstrated that the lung dose of bioavailable transition metal, not
31 instilled PM mass, was the primary determinant of the acute inflammatory response. Kennedy et
April 2002 7-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 al. (1998) reported a similar dose-dependent inflammation (i.e., increase in protein and PMN in
2 lavage fluid, proliferation of bronchiolar epithelium, and intraalveolar hemorrhage) in rats
3 instilled with water-extracted particles (TSP) collected in Provo, UT. This study also indicated
4 that the metal constituent, in this case PM-associated Cu, was a plausible cause of the outcome.
5 Likewise, instillation of ambient PM10 collected in Edinburgh, Scotland, also caused pulmonary
6 injury and inflammation in rats (Li et al., 1996, 1997). Brain et al. (1998) examined the effects
7 of instillation of particles that resulted from the Kuwaiti oil fires in 1991 compared to urban
8 particulate matter collected in St. Louis (NIST SRM 1648, collected in a bag house in early
9 1980s) and showed that on an equal mass basis, the acute toxicity of the Kuwaiti oil fire particles
10 was similar to that of urban particles collected in the United States.
11 Toxicological studies of ambient PM collected around Provo, UT (Utah Valley) in the late
12 1980s are particularly interesting (Ohio and Devlin, 2001; Dye et al., 2001; Wu et al., 2001;
13 Soukup et al., 2000; Frampton et al., 1999). Epidemiological studies by Pope et al. (1989, 1991)
14 and reported in the previous PM AQCD (U.S. Environmental Protection Agency, 1996a) showed
15 that closure of an open-hearth steel mill over the winter of 1987 was associated with reductions
16 in hospital admissions for respiratory diseases (see Chapter 8 for details of the epidemiology
17 studies). Ambient PM was collected near the steel mill during the winter of 1986 (before
18 closure), 1987 (during closure), and again in 1988 (after plant reopening). The glass hi-vol filters
19 were stored, folded PM-side inward, in plastic sleeves at room temperature and humidity (Dye et
20 al., 2001). A description of the in vivo studies follows; the in vitro studies are discussed in
21 Section 7.5.2.1.
22 Ohio and Devlin (2001) investigated the biologic effect of PM from the Utah Valley to
23 determine if the biological responses mirrored the epidemiological findings, with greater injury
24 occurring after exposure to an equal mass of particles from those years in which the mill was in
25 operation. Aqueous extracts of the filters collected prior to closure of the steel mill, during the
26 closure and after its reopening, were instilled through a bronchoscope into the lungs of
27 nonsmoking volunteers. Twenty-four hours later, the same subsegment was lavaged. Exposure
28 to aqueous extracts of PM collected before closure and after reopening of the steel mill provoked
29 a greater inflammatory response than PM extract acquired during the plant shutdown. These
30 results indicate that the pulmonary effects observed after experimental exposure of humans to the
April 2002 7-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 Utah Valley PM can be correlated with health outcomes observed in epidemiologic studies of the
2 same material under normal exposure conditions.
3 Dye et al (2001) examined the relationship between Utah Valley ambient PM and
4 respiratory health effects. Sprague-Dawley rats were intratracheally instilled with equivalent
5 masses of aqueous extracts from filters originally collected during the winter before, during, and
6 after closure of the steel mill. Twenty-four hours after instillation, rats exposed to extracts of
7 particles collected when the plant was open developed significant pulmonary injury and
8 neutrophilic inflammation. Additionally, 50% of rats exposed to these extracts had increased
9 airway responsiveness to acetylcholine, compared to 17 and 25% of rats exposed to saline or the
10 extracts of particles collected when the plant was closed. By 96 hr, these effects were largely
11 resolved except for increases in lung lavage fluid neutrophils and lymphocytes in rats exposed to
12 PM extracts from prior to the plant closing. Analogous effects were observed with lung
13 histologic assessment. Extract analysis demonstrated that nearly 70% of the mass in all three
14 extracts appeared to be sodium-based salts derived from the glass filter matrix. Extracts of
15 particles collected when the plant was open contained more sulfate, cationic salts (i.e., calcium,
16 potassium, magnesium), and certain metals (i.e., copper, zinc, iron, lead, strontium, arsenic,
17 manganese, nickel). Although total metal content was ~ 1% of the extracts by mass, the greater
18 quantity detected in the extracts of particles collected when the plant was open suggests that
19 metals may be important determinants of the observed pulmonary toxicity. The authors conclude
20 that the pulmonary effects induced in rats by exposure to aqueous extracts of local ambient PM
21 filters were in good accord with the epidemiologic reports of adverse respiratory health effects in
22 Utah Valley residents.
23 The fact that instillation of ambient PM collected from different geographical areas and
24 from a variety of emission sources consistently caused pulmonary inflammation and injury tends
25 to corroborate epidemiological studies that report increased PM-associated respiratory effects in
26 populations living in many different geographical areas and climates. However, high-dose
27 instillation studies may produce different effects on the lung than inhalation exposures done at
28 more relevant concentrations. This concern is somewhat diminished by the results of inhalation
29 studies of concentrated PM in healthy nonsmokers.
30 Ohio et al. (2000a) exposed 38 healthy volunteers exercising intermittently at moderate
31 levels of exertion for 2 h to either filtered air or particles concentrated from the air in Chapel
April 2002 7-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hill, NC (23 to 311 //g/m2). Analysis of cells and fluid obtained 18 h after exposure showed a
2 mild increase in neutrophils in the bronchial and alveolar fractions of bronchoalveolar lavage
3 (BAL) in subjects exposed to the highest quartile concentration of concentrated PM (mean of
4 206.7 //g/m3). Lavage protein did not increase, and there were no other indicators of pulmonary
5 injury. No respiratory symptoms or decrements in pulmonary function were found after exposure
6 to CAPs.
7 The 38 human volunteers reported in Ohio (2000a) were also examined for changes in host
8 defense and immune parameters in BAL and blood (Harder et al., 2001). There were no changes
9 in the number of lymphocytes or macrophages, subcategories of lymphocytes (according to
10 surface marker analysis by flow cytometry), cytokines IL-6 and IL-8, or macrophage
11 phagocytosis. Similarly, there was no effect of concentrated ambient PM exposure on
12 lymphocyte subsets in blood. Thus, a mild inflammatory response to concentrated ambient PM
13 was not accompanied by an affect on immune defenses as determined by lymphocyte or
14 macrophage effects.
15 Other human inhalation studies with CAPs are limited by the small numbers of subjects
16 studied. Petrovic et al., 1999 exposed four healthy volunteers (aged 18 to 40) under resting
17 conditions to filtered air and 3 concentrations of concentrated ambient PM (23 to 124 //g/m3) for
18 2 hours using a face mask. The exposure was followed by 30 minutes of exercise. No cellular
19 signs of inflammation were observed in induced sputum samples collected at 2 or 24 hours after
20 exposure. There was a trend toward an increase in nasal lavage neutrophils although no
21 statistical significance was presented. The only statistically significant change in pulmonary
22 function was a 6.4% decrease in thoracic gas volume after exposure to 124 //g/m3 PM versus a
23 5.6% increase after air. A similar, small pilot study has been reported (Gong et al., 2000) in
24 which no changes in pulmonary function or symptoms were observed in four subjects aged 19 to
25 41 after a 2 hour exposure to air or mean concentrations of 148 to 246 //g/m3 concentrated
26 ambient PM in Los Angeles, CA. Both of these laboratories are currently expanding on these
27 preliminary findings, but no data are available at this time.
28
29 7.2.1.2 Diesel Particulate Matter
30 Other controlled human exposures of ambient PM that may be relevant to this discussion
31 were the DPM studies previously examined in detail in separate assessment documents (U.S.
April 2002 7-18 DRAFT-DO NOT QUOTE OR CITE
-------
1 Environmental Protection Agency, 2000; Health Effects Institute, 1995). Briefly, the data from
2 work shift studies suggest that the principle noncancer human hazard from exposure to diesel
3 exhaust (DE) includes increased acute sensory and respiratory symptoms (e.g., cough, phlegm,
4 chest tightness, wheezing) that are more sensitive indicators of possible health risks from
5 exposure to diesel exhaust than pulmonary function decrements. Immunological changes also
6 have been demonstrated under short-term exposure scenarios to either diesel exhaust or DPM,
7 and the evidence indicates that these immunological effects are caused by both the non-
8 extractable carbon core and the adsorbed organic fraction of the diesel particle. While noncancer
9 effects from long-term exposure to DPM of several laboratory animal species include pulmonary
10 histopathology and chronic inflammation, noncancer effects in humans from long-term chronic
11 exposure to DPM are not evident. The mode of action of DPM is not completely understood but
12 the effects on the upper respiratory tract, observed in acute studies, suggest an irritant mechanism
13 while the effects on the lung, observed in chronic studies, indicate an underlying inflammatory
14 response. Currently available data suggest that the carbonaceous core of the diesel particle, or
15 metabolites of metal components of the particle, are possible causative agents for the noncancer
16 lung effects which are mediated, at least in part, by a progressive impairment of alveolar
17 macrophage function. The noncancer lung effects occur in response to DPM in several species
18 and occur in rats at doses lower than those inducing particle overload.
19 Diesel particulate matter, therefore, can be relevant to the urban environment, particularly
20 in urban micro-environments with heavy diesel engine traffic. The findings of controlled studies
21 on DPM are included here and in Section 7.4.3 (allergic hosts/immunology).
22 Pulmonary function and inflammatory markers (as assayed in induced sputum samples or
23 BAL) have been studied in human subjects exposed to either resuspended or freshly generated
24 and diluted DPM. In a controlled human study, Sandstrom and colleagues (Rudell et al., 1994)
25 exposed eight healthy subjects in an exposure chamber to diluted exhaust from a diesel engine
26 for 1 h with intermittent exercise. Dilution of the diesel exhaust was controlled to provide a
27 median NO2 level of approximately 1.6 ppm. Median particle number was 4.3 x 106 /cm3, and
28 median levels of NO and CO were 3.7 and 27 ppm, respectively (particle size and mass
29 concentration were not provided). There were no effects on spirometry or on airway closing
30 volume. Five of eight subjects experienced unpleasant smell, eye irritation, and nasal irritation
31 during exposure. BAL was performed 18 hours after exposure and was compared with a control
April 2002 7-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 BAL performed 3 weeks prior to exposure. There was no control air exposure. Small yet
2 statistically significant reductions were seen in BAL mast cells, AM phagocytic function, and
3 lymphocyte CD4 to CD8+ cell ratios. A small increase in neutrophils was also observed. These
4 findings suggest that diesel exhaust may induce mild airway inflammation in the absence of
5 spirometric changes. Although this early study provided important information on the effect of
6 diesel exhaust exposure in humans, only one exposure level was used, the number of subjects
7 was low, and a limited range of endpoints was reported. A number of follow-up studies have
8 been done by the same and other investigators.
9 Rudell et al. (1996) later exposed 12 healthy volunteers to diesel exhaust for 1 h in an
10 exposure chamber. Light work on a bicycle ergometer was performed during exposure.
11 Random, double-blinded exposures included air, diesel exhaust, or diesel exhaust with particle
12 numbers reduced 46% by a particle trap. The engine used was a new Volvo model 1990, a six-
13 cylinder direct-injection turbocharged diesel with an intercooler, which was run at a steady speed
14 of 900 rpm during the exposures. Comparison of this study with others is difficult because
15 neither exhaust dilution ratios nor particle concentrations were reported. Carbon monoxide
16 concentrations of 27-30 ppm and NO of 2.6-2.7 ppm, however, suggested DPM concentrations
17 may have equaled several mg/m3. The most prominent symptoms during exposure were
18 irritation of the eyes and nose accompanied by an unpleasant smell. Both airway resistance and
19 specific airway resistance increased significantly during the exposures. Despite the 46%
20 reduction in particle numbers by the trap, effects on symptoms and lung function were not
21 significantly attenuated.
22 A follow-up study on the usefulness of a particle trap confirmed the lack of effect of the
23 filter on diesel exhaust-induced symptoms (Rudell et al., 1999). In this study, 10 healthy
24 volunteers also underwent BAL 24 hours after exposure. Exposure to diesel exhaust produced
25 inflammatory changes in BAL as evidenced by increases in neutrophils and decreases in
26 macrophage phagocytic function in vitro. A 50% reduction in the particle number concentration
27 by the particle trap did not alter these cellular changes in BAL. Salvi et al. (1999) exposed
28 healthy human subjects to diluted diesel exhaust (DPM = 300 //g/m3 ) for 1 h with intermittent
29 exercise. As reported in the studies by Rudell and Sandstrom, significant increases in neutrophils
30 and B lymphocytes, as well as histamine and fibronectin in airway lavage fluid, were not
31 accompanied by decrements in pulmonary function. Bronchial biopsies obtained 6 h after diesel
April 2002 7-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 exhaust exposure showed a significant increase in neutrophils, mast cells, and CD4+ and CD8+
2 T lymphocytes, along with upregulation of the endothelial adhesion molecules ICAM-1 and
3 vascular cell adhesion molecule-1 (VCAM-1) and increases in the number of leukoxyte function-
4 associated antogen-1 (LFA-1+) in the bronchial tissue. Importantly, extra-pulmonary effects
5 were observed in these subjects. Significant increases in neutrophils and platelets were observed
6 in peripheral blood following exposure to diesel exhaust.
7 In a follow-up investigation of potential mechanisms underlying the DE-induced airway
8 leukocyte infiltration, Salvi et al. (2000) exposed healthy human volunteers to diluted DE on two
9 separate occasions for 1 h each, in an exposure chamber. Fiber-optic bronchoscopy was
10 performed 6 h after each exposure to obtain endobronchial biopsies and bronchial wash (BW)
11 cells. These workers observed that diesel exhaust (DE) exposure enhanced gene transcription of
12 interleukin-8 (IL-8) in the bronchial tissue and BW cells and increased growth-regulated
13 oncogene-a protein expression and IL-8 in the bronchial epithelium; there was also a trend
14 toward an increase in interleukin-5 (IL-5) mRNA gene transcripts in the bronchial tissue.
15 Nightingale et al. (2000) have reported inflammatory changes in healthy volunteers
16 exposed to 200 //g/m3 resuspended DPM under resting conditions in a double-blinded study.
17 Small but statistically significant increases in neutrophils and myeloperoxidase (an index of
18 neutrophil activation) were observed in sputum samples induced 4 hours after exposure to DPM
19 in comparison to air. Exhaled carbon monoxide was measured as an index of oxidative stress
20 and was found to increase maximally at 1 hour after exposure. These biochemical and cellular
21 changes occurred in the absence of any decrements in pulmonary function, thus suggesting that
22 markers of inflammation are more sensitive than pulmonary function measurements.
23 Because of the considerable concern regarding the inhalation of ambient particles by
24 sensitive subpopulations, Sandstrom's laboratory also studied the effect of a 1 hour exposure to
25 300 //g/m3 DPM on 14 atopic asthmatics with stable disease on inhaled corticosteroid treatment
26 (Nordenhall et al., 2001). At 6 hours after exposure, there was a significant increase in IL-6 in
27 induced sputum. At 24 hours after exposure, there was a significant increase in the nonspecific
28 airway responsiveness to inhaled methacholine. Although the exposure level was high relative to
29 ambient PM levels, these findings are important in terms of their relation to the epidemiology
30 evidence of an increase in asthma morbidity associated with episodic exposure to ambient PM.
April 2002 7-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 The role of antioxidant defenses in protecting against acute diesel exhaust exposure has
2 been studied. Blomberg et al. (1998) investigated changes in the antioxidant defense network
3 within the respiratory tract lining fluids of human subjects following diesel exhaust exposure.
4 Fifteen healthy, nonsmoking, asymptomatic subjects were exposed to filtered air or diesel
5 exhaust (DPM 300 mg/m3) for 1 h on two separate occasions at least 3 weeks apart. Nasal lavage
6 fluid and blood samples were collected prior to, immediately after, and 5.5 h post-exposure.
7 Bronchoscopy was performed 6 h after the end of diesel exhaust exposure. Nasal lavage ascorbic
8 acid concentration increased tenfold during diesel exhaust exposure, but returned to basal levels
9 5.5 h post-exposure. Diesel exhaust had no significant effects on nasal lavage uric acid or GSH
10 concentrations and did not affect plasma, bronchial wash, or bronchoalveolar lavage antioxidant
11 concentrations or malondialdehyde or protein carbonyl concentrations. The authors concluded
12 that the acute increase in ascorbic acid in the nasal cavity induced by diesel exhaust may prevent
13 further oxidant stress in the respiratory tract of healthy individuals.
14
15 7.2.1.3 Complex Combustion-Related Particles
16 Because emission sources contribute to the overall ambient air particulate burden (Spengler
17 and Thurston, 1983), many of the studies investigating the response of laboratory animals to
18 particle exposures have used complex combustion-related particles for exposure (see Table 7-3).
19 For example, the residual oil fly ash (ROFA) samples used in toxicological studies have been
20 collected from a variety of sources such as boilers, bag houses used to control emissions from
21 power plants, and from the particles that are emitted downstream of the collection devices (see
22 Table 1).
23 ROFA has a high content of water soluble sulfate and metals, accounting for 82 to 92% of
24 water-soluble mass, while the water-soluble mass fraction in ambient air varies from low teens to
25 more than 60% (Costa and Dreher, 1997; Prahalad et al., 1999). More than 90% of the metals in
26 ROFA are transition metals; whereas these metals are only a small subfraction of the total
27 ambient PM mass. Thus, the dose of bioavailable metal that is delivered to the lung when ROFA
28 is instilled into a laboratory animal can be orders of magnitude greater than an ambient PM dose,
29 even under a worst-case scenario.
30 Intratracheal instillation of various doses of ROFA suspension has been shown to produce
31 severe inflammation, an indicator of pulmonary injury that includes recruitment of neutrophils,
April 2002 7-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 eosinophils, and monocytes into the airway. The biological effects of ROFA in rats have been
2 shown to depend on aqueous teachable chemical constituents of the particles (Dreher et al., 1997;
3 Kodavanti et al., 1997b). A leachate prepared from ROFA, containing predominantly Fe, Ni, V,
4 Ca, Mg, and sulfate, produced similar lung injury to that induced by the complete ROFA
5 suspension (Dreher et al., 1997). Depletion of Fe, Ni, and V from the ROFA leachate eliminated
6 its pulmonary toxicity. Correspondingly, minimal lung injury was observed in animals exposed
7 to saline-washed ROFA particles. A surrogate transition metal sulfate solution containing Fe, V,
8 and Ni largely reproduced the lung injury induced by ROFA. Interestingly, ferric sulfate and
9 vanadium sulfate antagonized the pulmonary toxicity of nickel sulfate. Interactions between
10 different metals and the acidity of PM were found to influence the severity and kinetics of lung
11 injury induced by ROFA and its soluble transition metals.
12 To further investigate the response to ROFA with differing metal and sulfate composition,
13 male Sprague-Dawley rats (60 days old) were exposed to ROFA or metal sulfates (iron,
14 vanadium, and nickel, individually or in combination) (Kodavanti et al., 1997b). Transition
15 metal sulfate mixtures caused less injury than ROFA or Ni alone, suggesting metal interactions.
16 In addition, this study showed that V-induced effects were less severe than that of Ni and were
17 transient. Ferric sulfate was least pathogenic. Cytokine gene expression was induced prior to the
18 pathology changes in the lung, and the kinetics of gene expression suggested persistent injury by
19 nickel sulfate. Another study by the same investigators was performed using 10 different ROFA
20 samples collected at various sites within a power plant burning residual oil (Kodavanti et al.,
21 1998a). Animals received intratracheal instillations of either saline (control), or a saline
22 suspension of whole ROFA (<3.0 //m MMAD) at three concentrations (0.833, 3.33, or
23 8.33 mg/kg). This study showed that ROFA-induced PMN influx was associated with its water-
24 teachable V content; however, protein leakage was associated with water-leachable Ni content.
25 ROFA-induced in vitro activation of alveolar macrophages (AMs) was highest with ROFA
26 containing teachable V but not with Ni plus V, suggesting that the potency and the mechanism of
27 pulmonary injury may differ between emissions containing bioavailable V and Ni.
28 Other studies have shown that soluble metal components play an important role in the
29 toxicity of emission source particles. Gavett et al. (1997) investigated the effects of two ROFA
30 samples of equivalent diameters, but having different metal and sulfate content, on pulmonary
31 responses in Sprague-Dawley rats. ROFA sample 1 (Rl) (the same emission particles used by
April 2002 7-23 DRAFT-DO NOT QUOTE OR CITE
-------
1 Dreher et al. [1997]) had approximately twice as much saline-leachable sulfate, nickel, and
2 vanadium, and 40 times as much iron as ROFA sample 2 (R2); whereas R2 had a 31-fold higher
3 zinc content. Rats were instilled with suspensions of 2.5 mg R2 in 0.3 mL saline, the supernatant
4 of R2 (R2s), the supernatant of 2.5 mg Rl (Rls), or saline only. By 4 days after instillation, 4 of
5 24 rats treated with R2s or R2 had died. None of those treated with Rls or saline died.
6 Pathological indices, such as alveolitis, early fibrotic changes, and perivascular edema, were
7 greater in both R2 groups. In surviving rats, baseline pulmonary function parameters and airway
8 hyperreactivity to acetylcholine were significantly worse in the R2 and R2s groups than in the
9 Rls groups. Other than BAL neutrophils, which were significantly higher in the R2 and R2s
10 groups, no other inflammatory cells (macrophages, eosinophils, or lymphocytes) or biochemical
11 parameters of lung injury were significantly different between the R2 and R2s groups and the
12 Rls group. Although soluble forms of zinc had been found in guinea pigs to produce a greater
13 pulmonary response than other sulfated metals (Amdur et al., 1978), and, although the level of
14 zinc was 30-fold greater in R2 than Rl, the precise mechanisms by which zinc may induce such
15 responses are unknown. Nevertheless, these results show that the composition of soluble metals
16 and sulfate is critical in the development of airway hyperractivity and lung injury produced by
17 ROFA, albeit at high concentrations.
18 Reactive oxygen species may play an important role in the in vivo toxicity of ROFA. Dye
19 et al. (1997) pretreated rats with an intraperitoneal injection of saline or dimethylthiourea
20 (DMTU) (500 mg/kg), followed 30 min later by intratracheal instillation of either acidic saline
21 (pH = 3.3) or an acidified suspension of ROFA (500 //g/rat). The systemic administration of
22 DMTU impeded development of the cellular inflammatory response to ROFA but did not
23 ameliorate biochemical alterations in BAL fluid. In a subsequent study, these investigators
24 determined that oxidant generation, possibly induced by soluble vanadium compounds in ROFA,
25 is responsible for the subsequent rat tracheal epithelial cells gene expression, inflammatory
26 cytokine production (MTP-2 and IL-6), and cytotoxicity (Dye et al., 1999).
27 In addition to transition metals, other components in fly ash also may cause lung injury.
28 The effects of arsenic compounds in coal fly ash or copper smelter dust on the lung integrity and
29 on the ex vivo release of TNFa by alveolar phagocytes were investigated by Broeckaert et al.
30 (1997). Female Naval Medical Research Institute (NMRI) mice were instilled with different
31 particles normalized for the arsenic content (20 //g/kg body weight [i.e., 600 ng arsenic/mouse])
April 2002 7-24 DRAFT-DO NOT QUOTE OR CITE
-------
1 and the particle load (100 mg/kg body weight [i.e., 3 mg/mouse]). Mice received tungsten
2 carbide (WC) alone, coal fly ash (CFA) alone, copper smelter dust (CMP) mixed with WC, and
3 Ca3(AsO4)2 mixed with WC (see Table 7-2 for concentration details). Copper smelter dust
4 caused a severe but transient inflammatory reaction; whereas a persisting alveolitis (30 days
5 postexposure) was observed after treatment with coal fly ash. In addition, TNFa production in
6 response to lipopolysaccharide (LPS) by alveolar phagocytes were significantly inhibited at Day
7 1 but was still observed at 30 days after administration of CMP and CFA. Although arsenic was
8 cleared from the lung tissue 6 days after Ca3(AsO4)2 administration, a significant fraction
9 persisted (10 to 15% of the arsenic administered) in the lung of CMP- and CFA-treated mice at
10 Day 30. It is possible that suppression of TNF-a production is dependent upon the slow
11 elimination of the particles and their metal content from the lung.
12 In summary, intratracheally instilled ROFA produced acute lung injury and inflammation.
13 The water soluble metals in ROFA appear to play a key role in the acute effects of instilled
14 ROFA. Although studies done with ROFA clearly show that combustion generated particles
15 with a high metal content can cause substantial lung injury, there are still insufficient data to
16 extrapolate the high dose effects to the low levels of particle associated transition metals in
17 ambient PM.
18
19 7.2.2 Acid Aerosols
20 There have been extensive studies of the effects of controlled exposures to aqueous acid
21 aerosols on various aspects of lung function in humans and laboratory animals. Many of these
22 studies were reviewed in the previous criteria document (U.S. Environmental Protection Agency
23 1996a) and in the Acid Aerosol Issue Paper (U.S. Environmental Protection Agency, 1989);
24 some of the more recent studies are summarized in this document (Table 7-5). Methodology and
25 measurement methods for controlled human exposure studies have been reviewed elsewhere
26 (Folinsbeeetal., 1997).
27 The studies summarized in the previous document illustrate that aqueous acidic aerosols
28 have minimal effects on symptoms and mechanical lung function in young healthy adult
29 volunteers at concentrations as high as 1000 //g/m3. However, at concentrations as low as
30 100 //g/m3, acid aerosols can alter mucociliary clearance. Brief exposures (< 1 h) to low
31 concentrations («100 //g/m3) may accelerate clearance while longer (multihour) exposures to
April 2002 7-25 DRAFT-DO NOT QUOTE OR CITE
-------
>
to
o
o
to
TABLE 7-5. RESPIRATORY EFFECTS OF ACID AEROSOLS IN HUMANS AND LABORATORY ANIMALS
to
H
6
o
o
H
O
Species, Gender, Strain
Age, etc.
Dogs, beagle, healthy;
n= 16
Humans, asthmatic;
13 M, 11 F
Rats, female, Fischer
344; Guinea Pigs,
female, Hartley
Humans, healthy
nonsmokers; 10 M,
2 F; 21-37 years old
Particle
Neutral sulfite
aerosol
Acidic sulfate
aerosol
H2SO4 aerosol
NH+4/SCr4
aerosol
H2SO4 aerosol
H2SO4 aerosol
Exposure
Technique
Inhalation
Inhalation
Inhalation by
face mask
Inhalation
Inhalation
Concentration Particle Size
1.5 mg/m3 1.0 Mm MMAD
og = 2.2
5. 7 mg/m3 1.1 Mm MMAD
og = 2.0
500 Mg/m3 9 Mm MMAD
7 Mm MMAD
94 mg/m3 0.80±1.89og
43 mg/m3 0.93 ±2. Hog
1,000 Mg/m3 0.8-0.9 Mm
MMAD
Exposure
Duration
16.5 h/day
for 13 mo
6 h/day for
13 mo
Ih
4h
3h
Effects of Particles
Long-term exposure to particle-associated sulfur and
hydrogen ions at concentrations close to ambient
levels caused only subtle respiratory responses and no
change in lung pathology.
Exposure to simulated natural acid fog did not induce
bronchoconstriction nor change bronchial
responsiveness in asthmatics.
Acid aerosol increased surfactant film compressibility
in guinea pigs.
No inflammatory responses; LDH activity in BAL was
elevated. Effect on bacterial killing by macrophages
was inconclusive; latex particle phagocytosis was
reduced 28%.
Reference
Heyderetal. (1999)
Leducetal. (1995)
Lee etal. (1999)
Zelikoffetal. (1997)
H2SO4 = Sulfuric acid
BAL = Bronchoalveolar lavage
LDH = Lactate dehydrogenase
MMAD = Mass median aerodynamic diameter
MMD = Mass median diameter
og = Geometric standard deviation
o
HH
H
W
-------
1 higher concentrations (>100 //g/m3) can depress clearance. Asthmatic subjects appear to be more
2 100 //g/m3, acid aerosols can alter mucociliary clearance. Brief exposures (< 1 h) to low
3 concentrations («100 //g/m3) may accelerate clearance while longer (multihour) exposures to
4 higher concentrations (>100 //g/m3) can depress clearance. Asthmatic subjects appear to be more
5 sensitive to the effects of acidic aerosols on mechanical lung function. Responses have been
6 reported in adolescent asthmatics at concentrations as low as 68 //g/m3, and modest
7 bronchoconstriction has been seen in adult asthmatics exposed to concentrations >400 //g/m3, but
8 the available data are not consistent.
9 A previously described, acid aerosol exposure in humans (1000 //g/m3 H2SO4) did not
10 result in airway inflammation (Frampton et al., 1992), and there was no evidence of altered
11 macrophage host defenses. A more recent study by Zelikoff et al. (1997) compared the responses
12 of rabbits and humans exposed to similar concentrations of H2SO4 aerosol. For both rabbits and
13 humans, there was no evidence of PMN infiltration into the lung and no change in BAL fluid
14 protein level, although there was an increase in LDH in rabbits but not in humans. Macrophages
15 showed less antimicrobial activity in rabbits; insufficient data were available for humans.
16 Macrophage phagocytic activity was slightly reduced in rabbits but not in humans. Superoxide
17 production by macrophages was somewhat depressed in both species. No respiratory effects of
18 long-term exposure to acid aerosol were found in dogs (Heyder et al., 1999). Thus, recent studies
19 have not provided any additional evidence to unequivocally demonstrate that relevant
20 concentrations of aqueous acid aerosols contribute to the acute cardiopulmonary effects of
21 ambient PM.
22
23 7.2.3 Metal Particles, Fumes, and Smoke
24 Data from occupational and laboratory animal studies reviewed in the previous criteria
25 document (U. S. Environmental Protection Agency, 1996a) indicated that acute exposures to very
26 high levels (hundreds of//g/m3 or more) or chronic exposures to lower levels (up to 15 //g/m3) of
27 metallic particles could have an effect on the respiratory tract. Therefore, it was concluded on
28 the basis of data available at that time that the metals at typical concentrations present in the
29 ambient atmosphere (1 to 14 //g/m3) were not likely to have a significant acute effect in healthy
30 individuals. The metals include arsenic, cadmium, copper, nickel, vanadium, iron, and zinc.
31 Other metals found at concentrations less than 0.5 //g/m3 were not reviewed in the previous
April 2002 7-27 DRAFT-DO NOT QUOTE OR CITE
-------
1 criteria document. However, more recently published data from high-dose laboratory animal
2 studies added to the existing PM data base indicate that particle-associated metals are among the
3 potential causal components of PM.
4 Since completion of the previous criteria document, only limited controlled human
5 exposure studies have been performed with particles other than acid aerosols (see Table 7-6).
6 Controlled inhalation exposure studies to high concentrations of two different fume particles,
7 MgO and ZnO, demonstrate the differences in response based on particle metal composition
8 (Kuschner et al., 1997). Up to 6400 mg/m3/min cumulative dose of MgO had no effect on lung
9 function (spirometry, DLCO), symptoms of metal fume fever, or changes in inflammatory
10 mediators or cells recovered by BAL. However, lower concentrations of ZnO fume (165 to
11 1110 mg/m3/min) induced a neutrophilic inflammatory response in the airways 20 h
12 postexposure. Lavage fluid PMNs, TNF-a, and IL-8 were increased by ZnO exposure. Although
13 the concentrations used in these exposure studies exceed ambient levels by more than 1000-fold,
14 the absence of a response to an almost 10-fold higher concentration of MgO compared with ZnO
15 indicates that metal composition, in addition to particle size (ultrafme/fine), is an important
16 determinant of the observed health responses to inhaled PM.
17 Several metals, including zinc, chromium, cobalt, copper, and vanadium, have been shown
18 to stimulate cytokine release in cultured human pulmonary cells. Boiler makers, exposed
19 occupationally to approximately 400 to 500 //g/m3 of fuel oil ash, which contains high levels of
20 soluble metals, showed acute nasal inflammatory responses characterized by increased
21 myeloperoxidase (MPO) and IL-8 levels; these changes were associated with increased vanadium
22 levels in the upper airway (Woodin et al., 1998). Irsigler et al. (1999) reported that V2O5 can
23 induce asthma and bronchial hyperreactivity in exposed workers.
24 Autopsy data suggest that chronic exposure to urban air pollution leads to an increased
25 retention of metals in human tissues. A comparison of autopsy cases in Mexico City from the
26 1950s with the 1980s indicated substantially higher (5- to 20-fold) levels of Cd, Co, Cu, Ni, and
27 Pb in lung tissue from the 1980s (Fortoul et al., 1996). Similar studies have examined metal
28 content in human blood and lung tissue (Tsuchiyama et al., 1997; Osman et al., 1998) with
29 similar results.
30
April 2002 7-28 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-6. RESPIRATORY EFFECTS OF METAL PARTICLES, FUMES, AND SMOKE IN HUMANS AND
LABORATORY ANIMALS
to
o
o
to
1
to
**Q
O
l>
H
M
\^
o
0
H
/O
r*^x
o
H
W
O
^
o
HH
H
W
Species, Gender,
Strain, Age, etc.
Mice, Swiss
Rats, SD; 60
days old
Humans, healthy
nonsmokers;
12 M, 4 F;
18-35 years old
Humans,
vanadium plant
workers; 40 M;
19-60 years old
Humans, healthy
nonsmokers;
4 M, 2 F;
21-43 years old
Humans, healthy
nonsmokers;
27 M, 7 F;
20-36 years old
Rats, Fischer
344. (250 g)
Humans, healthy
nonsmokers;
8 M, 8 F;
18-34 years old
Mice, NMRI;
Mouse
peritoneal
macrophage
Particle
EHC-93
soluble
metal
salts
VS04
NiS04
Colloidal
iron oxide
VA
MgO
ZnO
FeA
FeA
FeA
Mn02
Exposure
Technique
Intratracheal
instillation
Inhalation
Bronchial
instillation
Inhalation
Inhalation
Intrapulmonary
instillation
Intratracheal
instillation
Inhalation
Intratracheal
instillation;
in vitro
Exposure
Concentration Particle Size Duration
ImginO.lml 0.8±0.4,um 3 days
0.3 - 2.4 mg/m3 N/A 6h/day x 4 days
5 mg in 10 mL 2.6 //m 1, 2, and 4 days
after instillation
0.05-1.53 N/A Variable
mg/m3
5.8-230 mg/m3 99% < 1.8 fj,m 15-45 min
29% < 0.1 fj.m
3 x 108 2.6 urn N/A
microspheres in
10 mL saline.
7.7 x 107 2.6 fj.m N/A
microspheres in
5 mL saline
12.7 mg/m3 1.5 ^m 30 min
ag = 2.1
0.037, 0. 12, 0.75, surface area of Sacrificed at
2.5 mg/animal 0.16, 0.5; 17, 5 days
62 m2/g
Effect of Particles
Solution containing all metal salts (Al, Cu, Fe, Pb,
Mg, Ni, Zn) or ZnCl alone increased BAL
inflammatory cells and protein.
V did not induce any significant changes in BAL or
HR; Ni caused delayed bradycardia, hypothermia, and
arrhythmogenesis at> 1.2 mg/m3; possible synergistic
effects were found.
L-ferritin increased after iron oxide particle exposure;
transferrin was decreased. Both lactoferrin and
transferrin receptors were increased.
12/40 workers had bronchial hyperreactivity that
persisted in some for up to 23 mo.
No significant differences in BAL inflammatory cell
concentrations, BAL interleukins (IL-1, IL-6, IL-8),
tumor necrosis factor, pulmonary function, or
peripheral blood neutrophils.
Transient inflammation induced initially (neutrophils,
protein, LDH, IL-8) was resolved by 4 days
postinstillation.
Transient inflammation at 1 day postinstillation.
No significant difference in 98mTc-DTPA clearance
half-times, DLCO, or spirometry
LDH, protein and cellular recruitment increased with
increasing surface area; freshly ground particles had
enhanced cytotoxicity.
Reference
Adamson et al.
(2000)
Campen et al.
(2001)
Ohio etal.
(1998a)
Irsigler et al.
(1999)
Kuschner et al.
(1997)
Lay etal. (1998)
Lay etal. (1998)
Lay etal. (2001)
Li son et al.
(1997)
-------
>
to
o
o
to
TABLE 7-6 (cont'd). RESPIRATORY EFFECTS OF METAL PARTICLES, FUMES, AND SMOKE IN HUMANS AND
LABORATORY ANIMALS
Species, Gender,
Strain, Age, etc. Particle
Rats, WISTAR CdO Fume
Furth;
7-week-old,
Mice, C57BL6
and DBA3NCR
Rats, M, F344, TiO2
175-225 g
Rats, M. F344, TiO2
175-225 g
Rats NaVO3
VOSO4
V205
Humans, ROFA
boilermakers
(18 M), 26-61
years old, and
utility worker
controls (11 M),
30-55 years old
Exposure
Technique
Nose-only
Inhalation
Intratracheal
inhalation and
Intratracheal
instillation
Intratracheal
inhalation and
Intratracheal
instillation
Intratracheal
instillation
Inhalation of
fuel-oil ash
Concentration Particle Size
1.04 mg/m3 CMD = 0.008
Rats dose = ,um ag = 1.1
18.72,ug
Mouse dose =
4.59 Mg
Inhalation at Fine: 250 nm
125 mg/m3 for Ultrafine:
2 h; Instillation at 21 nm
500 //g for fine,
750 //g for
ultrafine
Inhalation at Fine: 250 nm
125 mg/m3 for Ultrafine:
2 h; Instillation at 21 nm
500 //g for fine,
750 //g for
ultrafine
21 or 210 Mg N/A
V/kg (NaVO3,
VOSO4 soluble)
42 or 420 //g
V/kg (V20S) less
soluble
0.4-0.47 mg/m3 10 ,um
0.1-0. 13 mg/m3
Exposure
Duration
1 x 3h
Inhalation
exposure, 2 h;
sacrificed at 0,
1, 3, and 7 days
postexposure for
both techniques
Inhalation
exposure, 2 h;
sacrificed at 0,
1, 3, and 7 days
postexposure for
both techniques
1 h or 10 days
following
instillation
6 weeks
Effect of Particles
Mice created more metallothionein than rats, which
may be protective of tumor formation.
Inflammation produced by intratracheal inhalation
(both severity and persistence) was less than that
produced by instillation; ultrafine particles produced
greater inflammatory response than fine particles for
both dosing methods.
MIP-2 increased in lavage cells but not in supernatant
in those groups with increased PMN (more in
instillation than in inhalation; more in ultrafine than in
fine); TNF-a levels had no correlation with either
particle size or dosing methods.
PMN influx was greatest following VOSO4, lowest for
V2O5; VOSO4 induced inflammation persisted longest;
MIP-2 and KC (CXC chemokines) were rapidly
induced as early as 1 h postinstillation and persisted
for 48 h; Soluble V induced greater chemokine mRNA
expression than insoluble V; AMs have the highest
expression level.
Exposure to fuel-oil ash resulted in acute upper airway
inflammation, possibly mediated by increased IL-8
and PMNs.
Reference
McKenna et al.
(1998)
Osier and
Oberdorster
(1997)
Osier et al.
(1997)
Pierce et al.
(1996)
Woodin et al.
(1998)
CdO = Cadmium oxide
Fe2O3 = Iron oxide
MgO = Magnesium oxide
MnO2 = Manganese oxide
NaVO3 =
TiO2 = Titanium oxide
VOSO4 = Vanadium sulfate
V2O5 = Vanadium oxide
ZnO = Zinc oxide
BAL = Bronchoalveolar lavage
CMD = Count median diameter
IL = Interleukin
LDH = Lactate dehydrogenase
MIP-2 = Macrophage inflammatory protein-2
mRNA = Messenger RNA (ribonucleic acid)
N/A = Data not available
-------
1 Iron is the most abundant of the elements that are capable of catalyzing oxidant generation
2 and also is present in ambient urban particles. Lay et al. (1998) and Ohio et al. (1998a) tested the
3 hypothesis that the human respiratory tract will attempt to diminish the added, iron-generated
4 oxidative stress. They examined the cellular and biochemical response of human subjects,
5 instilled via the intrapulmonary route, with a combination of iron oxyhydroxides that introduced
6 an oxidative stress to the lungs. Saline alone and iron-containing particles suspended in saline
7 were instilled into separate lung segments of human subjects. Subjects underwent
8 bronchoalveolar lavage at 1 to 91 days after instillation of 2.6-//m diameter iron oxide
9 agglomerates. Lay and colleagues found iron-oxide-induced inflammatory responses in both the
10 alveolar fraction and the bronchial fraction of the lavage fluid at 1 day postinstillation. Lung
11 lavage 24 h after instillation revealed decreased transferrin concentrations and increased ferritin
12 and lactoferrin concentrations, consistent with a host-generated response to decrease the
13 availability of catalytically reactive iron (Ohio et al., 1998a). Normal iron homeostasis returned
14 within 4 days of the iron particle instillation. The same iron oxide preparation, which contained
15 a small amount of soluble iron, produced similar pulmonary inflammation in rats. In contrast,
16 instillation of rats with two iron oxide preparations that contained no soluble iron failed to
17 produce injury or inflammation, thus suggesting that soluble iron was responsible for the
18 observed intrapulmonary changes. Although the total dose of iron oxide delivered acutely to the
19 lung segments (approximately 5 mg or 2.1 x 108 particles) is considerably higher than would be
20 deposited in the lung at the concentrations of iron present in ambient urban air (generally less
21 than 1 //g/m3), only a small amount of the iron instilled in human subjects was "active."
22 Therefore, it is still not clear how the amount of active iron in the PM extract compares with the
23 iron found in ambient air particles.
24 In a subsequent inhalation study, Lay et al. (2001) studied the effect of iron oxide particles
25 on lung epithelial cell permeability. Healthy, nonsmoking human subjects inhaled 12.7 mg/m3
26 low- and high-solubility iron oxide particles (MMAD = 1.5 //m and og = 2.1) for 30 minutes.
27 Neither pulmonary function nor alveolar epithelial permeability, as assessed by pulmonary
28 clearance of technetium-labeled DPT A, was changed at 0.5 or 24 hours after exposure to either
29 type of iron oxide particle. Because the exposure concentration was so high, the data suggest that
30 metals may play little role in the adverse effects of ambient, urban PM. Ohio et al. (2001) have
31 reported a case study, however, in which acute exposure to oil fly ash from a domestic oil-
April 2002 7-31 DRAFT-DO NOT QUOTE OR CITE
-------
1 burning stove produced diffuse alveolar damage, difficulty in breathing, and symptoms of angina.
2 While steroid treatment led to rapid improvement in symptoms and objective measurements, this
3 report suggests that the high metal content of oil fly ash can alter the epithelial cell barrier in the
4 alveolar region.
5
6 7.2.4 Ambient Bioaerosols
7 Ambient bioaerosols include fungal spores, pollen, bacteria, viruses, endotoxins, and plant
8 and animal debris. Such biological aerosols can produce various health effects including
9 irritation, infection, hypersensitivity, and toxic response. Bioaerosols present in the ambient
10 environment have the potential to cause disease in humans under certain conditions. However, it
11 was concluded in the previous criteria document (U.S. Environmental Protection Agency, 1996a)
12 that bioaerosols, at the concentrations present in the ambient environment, would not contribute
13 to the observed effects of paniculate matter on human mortality and morbidity reported in PM
14 epidemiological studies. Moreover, bioaerosols generally represent a rather small fraction of the
15 measured urban ambient PM mass and are typically present even at lower concentrations during
16 the winter months when notable ambient PM effects have been demonstrated. Bioaerosols tend
17 to be in the coarse fraction of PM, but some bioaerosols including nonagglomerated bacteria and
18 fragmented pollens, are found in the fine fraction.
19 More recent inhalation studies on ambient bioaerosols are summarized in Table 7-7.
20 In vitro studies on particle-associated endotoxin are discussed in Section 7.5.2.2. Endotoxin,
21 a cell wall component of gram negative bacteria, is ubiquitous in the environment. Although
22 there is strong evidence that inhaled endotoxin plays a major role in the toxic effects of
23 bioaerosols encountered in the work place (Vogelzang et al., 1998; Castellan et al., 1984, 1987),
24 it is not clear whether ambient concentrations of endotoxin are sufficient to produce toxic
25 pulmonary or systemic effects in healthy or sick individuals.
26 Michel et al. (1997) examined the dose-response relationship to inhaled lipopolysaccharide
27 (LPS: the purified derivative of endotoxin) in normal healthy volunteers exposed to 0, 0.5, 5, and
28 50 //g of LPS. Inhalation of 5 or 50 //g of LPS resulted in increased PMNs in blood and sputum
29 samples. At the higher concentration, a slight (3%) but not significant decrease in FEVj was
30 observed. Cormier et al. (1998) reported an approximate 10% decline in FEVj and an increase in
31 methacholine airway responsiveness after a 5-h exposure inside a swine containment building.
April 2002 7-32 DRAFT-DO NOT QUOTE OR CITE
-------
>
to
o
o
to
TABLE 7-7. RESPIRATORY EFFECTS OF AMBIENT BIOAEROSOLS
Species, Gender,
Strain, Age, etc.
Rats, Fischer 344,
8 weeks to
20 months old,
N = 3/group
Humans, healthy;
5 M, 4 F, 24 to 50
years of age
Humans, healthy;
32 M, 32 F, 16 to
50 years of age
Humans, pig
farmers,
82 symptomatic
and
89 asymptomatic
n= 171
Humans, potato
plant workers, low
exposures (37 M),
high exposures
(20 M)
Particle
LPS
(endotoxin)
UF carbon
ozone
LPS
(endotoxin)
Indoor pool
water spray
Dust
Endotoxin
Endotoxin
Exposure
Technique Concentration Particle Size
Inhalation 70 EU 0.72 ij,m
og= 1.6
100 //g/m3 25 nm
og= 1.6
1 ppm
Inhalation 0.5 //g 1-4 ,um
5.0,ug MMAD
50 Mg
Inhalation N/A 0.1-7.5^m
Inhalation 2.63 mg/m3 N/A
og= 1.3
105 ng/m3
og=1.5
Inhalation 21.2EU/m3low N/A
og = 1.6
55.7EU/m3
high
og = 2.1
Exposure
Duration Effect of Particles
12 min Significant interaction of LPS and O3 on inflammatory
responses in young rats. O3 and UF-C interacted with
"priming" by LPS to produce greater PMN response.
6 h LPS has a priming effect on lung inflammatory response
to O3 and UF-C.
30 min Significant decrease in PMN luminol-enhanced
chemiluminescence with 0.5 /^g LPS; increase in blood
CRP and PMNs, and increase in sputum PMNs,
monocytes, and MPO with 5.0 ,ug LPS; increase in
temperature, blood PMNs, blood and urine CRP, sputum
PMNs, monocytes, lymphocytes, TNFa, and ECP with
50 Mg LPS.
N/A Recurring outbreaks of pool-associated granulomatous
pneumonitis (n = 33); case patients had higher cumulative
work hours. Analysis indicated increased levels of
endotoxin in pool air and water.
5 h/day average Large decline in FEV! (73 ml/year) and FVC (55 ml/year)
lifetime exposure associated with long-term average exposure to endotoxin.
8 h Decreased FEV1: FVC, and MMEF over the work shift
that was concentration related; endotoxin effects on lung
function can be expected above 53 EU/m3 (=4.5 ng/m3)
over 8 h.
Reference
Elder et al.
(2000a,b)
Michel et al.
(1997)
Rose et al.
(1998)
Vogelzang et al.
(1998)
Zock et al.
(1998)
-------
1 This exposure induced significant neutrophilic inflammation in both the nose and the lung.
2 Although these exposures are massive compared to endotoxin levels in ambient PM in U.S.
3 cities, these studies serve to illustrate the effects of endotoxin and associated bioaerosol material
4 in healthy nonsensitized individuals.
5 Some health effects have been observed after occupational exposure to complex aerosols
6 containing endotoxin at concentrations relevant to ambient levels. Zock et al. (1998) reported a
7 decline in FEVj («3%) across a shift in a potato processing plant with up to 56 endotoxin units
8 (EU)/m3 in the air. Rose et al. (1998) reported a high incidence (65%) of BAL lymphocytes in
9 lifeguards working at a swimming pool where endotoxin levels in the air were on the order of
10 28 EU/m3. Although these latter two studies may point towards pulmonary changes at low
11 concentrations of airborne endotoxin, it is not possible to rule out the contribution of other agents
12 in these complex organic aerosols. The contribution of endotoxin to the toxicity of ambient PM
13 has been studied in vitro, and these studies provide preliminary evidence that endotoxin
14 contamination of ambient PM may play a role in the observed in vitro effects (discussed in
15 Section 7.5).
16
17
18 7.3 CARDIOVASCULAR AND SYSTEMIC EFFECTS OF PARTICULATE
19 MATTER IN HUMANS AND LABORATORY ANIMALS: IN VIVO
20 EXPOSURES
21 A growing number of epidemiology studies have demonstrated that increases in cardiac-
22 related deaths are associated with exposure to PM (U.S. Environmental Protection Agency,
23 1996a) and that PM-related cardiac deaths appear to be as great or greater than those attributed to
24 respiratory causes (see Chapter 8). The toxicological consequences of inhaled particles on the
25 cardiovascular system had not been extensively investigated prior to 1996. Since then (see
26 Table 7-8), Costa and colleagues (e.g., Costa and Dreher, 1997) have demonstrated that
27 intratracheal instillation of high levels of ambient particles can increase or accelerate death in an
28 animal model of cardiorespiratory disease related to monocrotaline administration in rats. These
29 deaths did not occur with all types of ambient particles tested. Some dusts, such as volcanic ash
30 from Mount Saint Helens, were relatively inert; whereas other ambient dusts, including those
31 from urban sites, were toxic. These early observations suggested that particle composition plays
April 2002 7-34 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-8. CARDIOVASCULAR AND SYSTEMIC EFFECTS OF AMBIENT AND COMBUSTION-RELATED
PARTICULATE MATTER
to
o
o
to
^,
OJ
^r<
O
c
"Tl
H
6
o
h-^
-Z
H
O
d
o
3
o
^d
O
i — 1
— 1
w
Species, Gender,
Strain Age, or Body
Weight
Rats, male, F-344;
200-250 g
Rats, male, S-D,
60 days old,
MCT-treated and
healthy, n = 64
Dogs, female
mongrel,
14 to 17 kg
Rats, male, S-D,
60 days old,
MCT-treated,
and healthy
Rats, male, S-D;
60 days old
Humans, healthy
nonsmokers,
18 to 40 years old
Dogs, mongrel,
some with balloon
occluded LAD
coronary artery,
n= 14
Rats
Rats, male, F-344,
MCT-treated
Hamsters, 6-8 mo
old; Bio TO-2
Rats, S-D,
MCT-treated, 250 g
Particle
OTT
ROFA
CAPs
Emission source
PM
Ambient airshed
PM
ROFA
ROFA
CAPs
CAPs
CAPs
CAPs
FOFA
Exposure
Technique
Nose-only
Inhalation
Instillation
Inhalation via
tracheostomy
Instillation
Instillation
Inhalation
Inhalation via
tracheostomy
Nose-only
inhalation
Inhalation
Inhalation
Mass
Concentration
40 mg/m3
0.0,0.25, 1.0,
and 2.5 mg/rat
3-360 Mg/m3
Total mass:
2.5 mg/rat
Total transition
metal: 46 Mg/rat
0.3, 1.7, or
8.3 mg/kg
23.1 to
311.1 Mg/m3
69-828 Mg/m3
110-350 Mg/m3
132-919 Mg/m3
580 ± 110 Mg/m3
Particle
Size
4 to 5 Mm MMAD
1.95 Mm
0.2 to 0.3 Mm
Emission PM:
1.78-4. 17 Mm
Ambient PM:
3. 27-4.09 Mm
1.95 Mm
og = 2.19
0.65 Mm
og = 2.35
0.23 to 0.34 Mm
og = 0.2to2.9
N/A
0.2-1. 2 Mm
og = 0.2-3.9
2.06 Mm MMAD
og= 1.57
Exposure
Duration
4h
Analysis at
96 h
6 h/day for
3 days
Analysis at
24 and 96 h
following
instillation
Analysis at
24 h
2 h, analysis
at 18 h
6 h/day for
3 days
3h
3 h, evaluated
at 3 and 24 h
6 h/day for
3 days
Cardiovascular Effects
Increased plasma levels of endothelin-1.
No acute lung injury; however, lung NO production
decreased and macrophage inflammatory protein-2 from
lung lavage cells increased after exposure.
Dose-related hypothermia and bradycardia in healthy rats,
potentiated by compromised models.
Peripheral blood parameters were related to specific
particle constituents. Factor analysis from paired and
crossover experiments showed that hematologic changes
were not associated with increases in total CAP mass
concentration.
ROFA alone induced some mild arrhythumas;
MCT-ROFA showed enhanced neutrophilic inflammation;
MCT-ROFA animals showed more numerous and
severe arrhythmias including S-T segment inversions
and A-V block.
Increased plasma fibrinogen at 8.3 mg/kg only.
Increased blood fibrinogen.
Decreased time to ST segment elevation and increased
magnitude in compromised dogs. Decreased heart and
respiratory rate and increased lavage fluid neutrophils in
normal dogs.
Small but consistent increase in HR; no pulmonary injury
was found; increased peripheral blood neutrophils and
decreased lymphocytes.
No increase in cardiac arrhythmias; PM associated
increases in HR and blood cell differential counts, and
atrial conduction time of rats were inconsistent. No
adverse cardiac or pulmonary effects in hamsters.
Increased expression of the proinflammatory chemokine
MP-2 in the lung and heart of MCT-treated rats; less in
healthy rats. Significant mortality only in MCT-treated
rats.
Reference
Bouthillier et al.
(1998)
Campen et al.
(2000)
Clarke et al.
(2000a)
Costa and
Dreher (1997)
Gardner et al.
(2000)
Ohio et al.
(2000a)
Godleski et al.
(2000)
Gordon et al.
(1998)
Gordon et al.
(2000)
Killingsworth
etal. (1997)
-------
TABLE 7-8 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF AMBIENT AND COMBUSTION-RELATED
PARTICULATE MATTER
to
O
O
to
7"1
LtJ
ON
M
\^
C
^Tl
H
6
O
2,
O
H
/— s
Species, Gender,
Strain Age, or Body
Weight
Rats, male WKY and
SH, 12 to 13-week-
old
Rats, male SH and
WKY; 12-13 weeks
old
Dogs, beagles,
10. 5 -year- old,
healthy, n = 4
Rabbits, female,
New Zealand White,
1.8 to 2.4 kg
Rats, Wistar
Rats, male, S-D,
MCT-treated
Particle
ROFA
ROFA from a
precipitator of
an oil-burning
power plant
ROFA
Colloidal carbon
Ottawa ambient
(EHC-93)
(ECH-93L)
Diesel soot
(DPM)
Carbon black
(CB)
ROFA
Exposure
Technique
Nose-only
inhalation
Inhalation;
and
intratracheal
instillation
Oral
inhalation
Instillation
Inhalation
(nose only)
Instillation
Mass Particle
Concentration Size
15 mg/m3 N/A
15 mg/m3 1.5 fj.m
1 and 5 mg/kg og = 1.5
3 mg/m3 2.22 ^m MMAD
og = 2.71
2mLof 1% <1 //m
colloidal carbon
(20 mg)
48 mg/m3 36, 56, 80, 100,
49 mg/m3 and 300 ,um
5 mg/m3
5mg/m3
0.25, 1.0, or 1.95,umMMAD
2.5mgin0.3mL og = 2.19
saline
Exposure
Duration
6 h/day for
3 days
6 h/days, 3
days per week
forl, 2, or 4
weeks
3 h/day for
3 days
Examined for
24 to 192 h
after
instillation
4h
Monitored for
96 h after
instillation of
ROFA
particles
Cardiovascular Effects
Cardiomyopathy and monocytic cell infiltration, along with
increased cytokine expression, was found in left ventricle
of SH rats because of underlying cardiovascular disease.
ECG showed exacerbated ST segment depression caused
by ROFA.
IT exposure increased plasma fibrinogen and decreased
peripheral lymphocytes in both SH and WKY rats. Acute
IH exposure increased plasma fibrinogen in SH rats only;
longer exposure caused pulmonary injury but no changes in
fibrinogen.
No consistent changes in ST segment, the form or
amplitude of the T wave, or arrhythmias; slight bradycardia
during exposure.
Colloidal carbon stimulated the release of BRDU-labeled
PMNs from the bone marrow. The supernatant of alveolar
macrophages treated with colloidal carbon in vitro also
stimulated the release of PMNs from bone marrow, likely
via cytokines.
EHC-93 elevated blood pressure and ET-1 and ET-3 levels
EHC-93 L No effect on blood pressure, transient effect on
ET-1, -2, -3 levels
DPM no effect on blood pressure, but elevated ET-3 levels
CB no effect
Dose-related increases in the incidence and duration of
serious arrhythmic events in normal rats. Incidence and
severity of arrythmias were increased greatly in the MCT
rats. Deaths were seen at each instillation level in MCT
rats only (6/12 died after MCT + ROFA).
Reference
Kodavanti et al.
(2000b)
Kodavanti et al.
(2002)
Muggenburg
et al. (2000)
Terashima et al.
(1997)
Vincent et al.
(2001)
Watkinson et al.
(1998)
O
HH
H
W
-------
TABLE 7-8 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF AMBIENT AND COMBUSTION-RELATED
PARTICULATE MATTER
to
o
o
to
Species, Gender,
Strain Age, or Body
Weight
Particle
Exposure
Technique
Mass
Concentration
Particle
Size
Exposure
Duration
Cardiovascular Effects
Reference
(l)Rats, S-D healthy
and cold-stressed,
ozone-treated, and
MCT-treated
(2) Rats, S-D, SH
rats, WKY rats,
healthy and
MCT-treated,
(3) Rats, SH,
15-mo-old
(4) Rats, S-D
MCT-treated
ROFA
ROFA
OTT
ROFA
MSH
Fe2(S04)3
VS04
NiSO,
Intratracheal
instillation
Inhalation
Intratracheal
instillation
Intratracheal
instillation
0.0,0.25,1.0,
or 2.5 mg/rat
15 mg/m3
2.5 mg
0.5 mg
2.5 mg
105 /^g
245 ,ug
262.5 //g
1.95 urn
1.95 urn
1.95 urn
Monitored for (1) Healthy rats exposed IT to ROFA demonstrated dose-
96 h after related hypothermia, bradycardia, and increased
instillation arrhythmias. Compromised rats demonstrated exaggerated
hypothermia and cardiac responses to IT ROFA. Mortality
was seen only in the MCT-treated rats exposed to ROFA by
6 h/day for IT. (2) Pulmonary hypertensive (MCT-treated S-D) and
3 days systemically hypertensive (SH) rats exposed to ROFA by
inhalation demonstrated similar effects, but of diminished
amplitude. There were no lethalities by the inhalation
Monitored for route. (3) Older rats exposed IT to OTT showed a
96 h after pronounced biphasic hypothermia and a severe drop in HR
instillation accompanied by increased arrhythmias; exposure to ROFA
caused less pronounced, but similar effects. No cardiac
Monitored for effects were seen with exposure to MSH. (4) Ni and V
96 h after showed the greatest toxicity; Fe-exposed rats did not differ
instillation from controls.
Watkinson et al.
(2000a,b)
H
6
o
o
H
O
o
HH
H
W
-------
1 an important role in the adverse health effects associated with episodic exposure to ambient PM,
2 despite the "general particle" effect attributed to the epidemiological associations of ambient PM
3 exposure and increased mortality in many regions of the United States (i.e., regions with varying
4 particle composition). Work that examines the role of inherent susceptibility to the adverse
5 effects of PM in compromised animal models of human pathophysiology provides a potentially
6 important link to epidemiological observations and is discussed below.
7 To date, studies examining the systemic and cardiovascular effects of particles have used a
8 number of compromised animal models, largely rodent models. Two studies in normal or
9 compromised dogs (Godleski et al., 2000; Muggenburg et al., 2000) also have been published as
10 well as the preliminary results from studies in which human subjects were exposed to
11 concentrated ambient PM (see Section 7.4.1). Although the majority of animal studies
12 examining the systemic effects of PM have used metal-laden ROFA as a source particle, a
13 growing number of studies have used collected and stored ambient PM or real-time generated
14 concentrated ambient particles. The following discussion of the systemic effects of PM first
15 describes the ROFA studies and then compares these findings with the ambient PM studies.
16 Killingsworth and colleagues (1997) used a fuel oil fly ash to examine the adverse effects
17 of a model urban particle in an animal model (monocrotaline, MCT) of cardiorespiratory disease;
18 MCT causes progressive lung injury and vascular inflammation in rats. The lung injury induced
19 with MCT can lead, within two weeks of treatment, to pulmonary hypertension and right heart
20 enlargement, common features of chronic obstructive pulmonary disease in humans. They
21 observed 42% mortality in MCT rats exposed to approximately 580 //g/m3 fly ash for 6 h/day for
22 3 consecutive days. Deaths did not occur in MCT rats exposed to filtered air or in saline-treated
23 rats exposed to fly ash. The increase in deaths in the MCT/fly ash group was accompanied by an
24 increase in neutrophils in lavage fluid and an increased immunostaining of MIP-2 in the heart
25 and lungs of the MCT/fly ash animals. Cardiac immunohistochemical analysis indicated
26 increased MIP-2 in cardiac macrophages. The fly ash-induced deaths did not result from a
27 change in pulmonary arterial pressure and the cause of death was not identified.
28 In a similar experimental model, Watkinson et al. (1998) examined the effects of
29 intratracheally instilled ROFA (0.0, 0.25, 1.0, 2.5 mg in 3 mL saline) on ECG measurements in
30 control and MCT rats. They observed a dose-related increase in the incidence and duration of
31 serious arrhythmic events in control animals exposed to ROFA particles, and these effects were
April 2002 7-3 8 DRAFT-DO NOT QUOTE OR CITE
-------
1 clearly exacerbated in the MCT animals. Similar to the results of Killingsworth et al. (1997),
2 healthy animals treated with ROFA suffered no deaths, but there were 1, 2, and 3 deaths in the
3 low-, medium-, and high-dose MCT groups, respectively. Thus, ROFA PM was linked to the
4 conductive and hypoxemic arrhythmias associated with cardiac-related deaths in the MCT
5 animals.
6 To examine the biological relevance of intratracheal instillation of ROFA particles,
7 Kodavanti et al. (1999) exposed MCT rats to ROFA by either instillation (0.83 or 3.33 mg/kg) or
8 nose-only inhalation (15 mg/m3, 6 h/day for 3 consecutive days). Similar to Watkinson et al.
9 (1998), intratracheal instillation of ROFA in MCT rats resulted in -50% mortality. Notably, no
10 mortality occurred in MCT rats exposed to ROFA by the inhalation route despite the high
11 exposure concentration (15 mg/m3). In addition, no mortality occurred in healthy rats exposed to
12 ROFA or in MCT rats exposed to clean air. Despite the fact that mortality was not associated
13 with ROFA inhalation exposure of MCT rats, exacerbation of lung lesions and pulmonary
14 inflammatory cytokine gene expression, as well as ECG abnormalities, clearly were evident.
15 Watkinson and colleagues further examined the effect of instilled ROFA in rodents
16 previously exposed to ozone or housed in the cold (Watkinson et al., 2000a,b; Campen et al.,
17 2000). The effect of ozone-induced pulmonary inflammation (preexposure to 1 ppm ozone for
18 6 h) or housing in the cold (10 °C) on the response to instilled ROFA in rats was similar to that
19 produced with MCT. Bradycardia, arrhythmias, and hypothermic changes were consistently
20 observed in the ozone exposed and hypothermic animals treated with ROFA, although, unlike in
21 the MCT animals, no deaths occurred. Thus, in rodents with cardiopulmonary disease/stress,
22 instillation of 0.25 mg or more of ROFA can produce systemic changes that may be used to study
23 potential mechanisms of toxicity that are consistent with the epidemiology and panel studies
24 showing cardiopulmonary effects in humans.
25 While studies of instilled residual oil fly ash demonstrated immediate and delayed
26 responses, consisting of bradycardia, hypothermia, and arrhythmogenesis in conscious,
27 unrestrained rats (Watkinson et al., 1998; Campen et al., 2000), further study of instilled ROFA-
28 associated transition metals showed that vanadium induced the immediate responses, while
29 nickel was responsible for the delayed effects (Campen et al., 2002a). Moreover, Ni, when
30 administered concomitantly, potentiated the immediate effects caused by V.
April 2002 7-39 DRAFT-DO NOT QUOTE OR CITE
-------
1 In another study, Campen et al. (2001) examined the responses to these metals in conscious
2 rats by whole-body inhalation exposure. The authors attempted to ensure valid dosimetric
3 comparisons with the instillation studies, by using concentrations of V and Ni ranging from 0.3-
4 2.4 mg/m3. The concentrations used in this study incorporated estimates of total inhalation dose
5 derived using different ventilatory parameters. Heart rate (HR), core temperature (T[CO]), and
6 electrocardiographic (ECG) data were measured continuously throughout the exposure. Animals
7 were exposed to aerosolized Ni, V, or Ni + V for 6 h per day for 4 days, after which serum and
8 bronchoalveolar lavage samples were taken. While Ni caused delayed bradycardia, hypothermia,
9 and arrhythmogenesis at concentrations > 1.2 mg/m3, V failed to induce any significant change in
10 HR or T (CO), even at the highest concentration. When combined, Ni and V produced
11 observable delayed effects at 0.5 mg/m3 and potentiated responses at 1.3 mg/m3, greater than
12 were produced by the highest concentration of Ni (2.1 mg/m3) alone. Although these studies
13 were performed at metal concentrations that were orders of magnitude greater than ambient
14 concentrations, the results indicate a possible synergistic relationship between inhaled Ni and V.
15 Watkinson and colleagues (2000a,b) also sought to examine the relative toxicity of
16 different particles on the cardiovascular system of spontaneously hypertensive rats. They
17 instilled 2.5 mg of representative particles from ambient (Ottawa) or natural (Mount Saint Helens
18 volcanic ash) sources and compared the response to 0.5 mg ROFA. Instilled particles were either
19 mass equivalent dose or adjusted to produce equivalent metal dose. They observed adverse
20 changes in ECG, heart rate, and arrhythmia incidence that were much greater in the Ottawa- and
21 ROFA-treated rats than in the Mount Saint Helens-treated rats. The cardiovascular changes
22 observed with the Ottawa particles were actually greater than with the ROFA particles. These
23 series of experiments by Watkinson and colleagues clearly demonstrate that instillation of
24 ambient air particles, albeit at a very high concentration, can produce cardiovascular effects.
25 They also demonstrate that PM exposures of equal mass dose did not produce the same
26 cardiovascular effects, suggesting that PM composition was responsible for the observed effects.
27 Because of concerns regarding the relevance of particles administered by intratracheal
28 instillation, investigators also have examined the cardiovascular effects of ROFA particles using
29 more realistic inhalation exposure protocols. Kodavanti et al. (2000b) found that exposure to a
30 high concentration of ROFA (15 mg/m3 for 6 h/day for 3 days) produced alterations in the ECG
31 waveform of spontaneously hypertensive (SH) but not normotensive rats. Although the ST
April 2002 7-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 segment area of the ECG was depressed in the SH rats exposed to air, further depressions in the
2 ST segment were observed at the end of the 6-h exposure to ROFA on Days 1 and 2. The
3 enhanced ST segment depression was not observed on the third day of exposure, suggesting that
4 adaptation to the response had occurred. Thus, exposure to a very high concentration of ROFA
5 exacerbated a defect in the electroconductivity pattern of the heart in an animal model of
6 hypertension. This ROFA-induced alteration in the ECG waveform was not accompanied by an
7 enhancement in the monocytic cell infiltration and cardiomyopathy that also develop in SH rats.
8 Further work is necessary to determine the relevance of this ROFA study to PM at concentrations
9 relevant to ambient exposures.
10 Godleski and colleagues (2000) have performed a series of experiments examining the
11 cardiopulmonary effects of inhaled concentrated ambient PM on normal mongrel dogs and on
12 dogs with coronary artery occlusion. Dogs were exposed by inhalation via a tracheostomy tube
13 to concentrated ambient PM for 6 h/day for 3 consecutive days. The investigators found little
14 biologically-relevant evidence of pulmonary inflammation or injury in normal dogs exposed to
15 PM (daily range of mean concentrations was approximately 100 to 1000 //g/m3). The only
16 statistically significant effect observed was a doubling of the percentage of neutrophils in lung
17 lavage. Despite the absence of major pulmonary effects, a significant increase in heart rate
18 variability (an index of cardiac autonomic activity), a decrease in heart rate, and an increase in T
19 alternans (an index of vulnerability to ventricular fibrillation) were observed. Exposure
20 assessment of particle composition produced no specific components of the particles that were
21 correlated with the day-to-day variability in response. The significance of these effects is not yet
22 clear because the effects did not occur on all exposure days. For example, the change in heart
23 rate variability was observed on only 10 of the 23 exposure days. Although the heart rate
24 variability change and the increase in T alternans suggest a possible proarrhythmic response to
25 inhaled concentrated ambient PM, the clinical significance of this effect is currently unknown.
26 The most important finding in the experiments of Godleski et al. (2000) was the
27 observation of a potential increase in ischemic stress of the cardiac tissue from repeated exposure
28 to concentrated ambient PM. During coronary occlusion in four dogs exposed to PM, they
29 observed a significantly more rapid development of ST elevation of the ECG waveform.
30 In addition, the peak ST-segment elevation was greater after PM exposure. Together, these
31 changes suggest that concentrated ambient PM can augment the ischemia associated with
April 2002 7-41 DRAFT-DO NOT QUOTE OR CITE
-------
1 coronary artery occlusion in this dog model. Additional work in more dogs as well as other
2 species is necessary to determine the significance of these findings to the human response to
3 ambient PM.
4 Muggenburg and colleagues (2000) reported that inhalation exposure to high
5 concentrations of ROFA produces no consistent changes in amplitude of the ST-segment, form
6 of the T wave, or arrhythmias in dogs. In their studies, four beagle dogs were exposed to
7 3 mg/m3 ROFA particles for 3 h/day for 3 consecutive days. They noted a slight but variable
8 decrease in heart rate, but the changes were not statistically or biologically significant. The
9 transition metal content of the ROFA used by Muggenburg was approximately 15% by mass,
10 a value that is on the order of a magnitude higher than that found in ambient urban PM samples.
11 Although the study did not specifically address the effect of metals, it suggests that inhalation of
12 high concentrations of metals may have little effect on the cardiovascular system of a healthy
13 individual.
14 In a series of studies, Gordon, Nadziejko, and colleagues examined the response of the
15 rodent cardiovascular system to concentrated ambient PM derived from New York City air
16 (Gordon et al., 2000). Particles of 0.2 to 2.5 //m in diameter were concentrated up to 10 times
17 their levels in ambient air (~ 150 to 900 //g/m3) to maximize possible differences in effects
18 between normal and cardiopulmonary-compromised laboratory animals. ECG changes were not
19 detected in normal Fischer 344 rats or hamsters exposed by inhalation to concentrated ambient
20 PM for 1 to 3 days. Similarly, no deaths or ECG changes were observed in MCT rats or
21 cardiomyopathic hamsters exposed to PM. Contrary to the decrease in heart rate observed in
22 dogs exposed to concentrated ambient PM (Godleski et al., 2000), heart rate was increased in
23 both normal and MCT rats exposed to PM. The increase was approximately 5% and was not
24 observed on all exposure days. Thus, extrapolation of the heart rate changes in these animal
25 studies to human health effects is difficult, although the increase in heart rate in rats is similar to
26 that observed in some human population studies.
27 Gordon and colleagues (1998) have reported other cardiovascular effects in animals
28 exposed to inhaled CAP. Increases in peripheral blood platelets and neutrophils were observed
29 in control and MCT rats at 3 h, but not 24 h, after exposure to 150 to 400 //g/m3 concentrated
30 ambient PM (CAP). This neutrophil effect did not appear to be dose related and did not occur on
31 all exposure days, suggesting that day-to-day changes in particle composition may play an
April 2002 7-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 important role in the systemic effects of inhaled particles. The number of studies reported was
2 small and; therefore, it is not possible to statistically determine if the day-to-day variability was
3 truly due to differences in particle composition or even to determine the size of this effect.
4 Terashima et al. (1997) also examined the effect of particles on circulating neutrophils. They
5 instilled rabbits with 20 mg colloidal carbon, a relatively inert particle (<1 //m), and observed a
6 stimulation of the release of 5'-bromo-2'deoxyuridine (BrdU)-labeled PMNs from the bone
7 marrow at 2 to 3 days after instillation. Because the instilled supernatant from rabbit AMs
8 treated in vitro with colloidal carbon also stimulated the release of PMNs from the bone marrow,
9 the authors hypothesized that cytokines released from activated macrophages could be
10 responsible for this systemic effect. The same research group (Tan et al., 2000) looked for
11 increased white blood cell counts as a marker for bone marrow PMN precursor release in humans
12 exposed to high levels of carbon from biomass burning during the 1997 Southeast Asian smoke-
13 haze episodes. They found a significant association between PM10 (1-day lag) and elevated band
14 neutrophil counts expressed as a percentage of total PMNs. The biological relevance of this
15 latter study to urban PM-induced systemic effects in unclear, however, because of the high dose
16 of carbon particles.
17 The results of epidemiology studies have suggested that homeostatic changes in the
18 vascular system can occur after episodic exposure to ambient PM. Studies by Vincent et al.
19 (2001) indicate that urban particles inhaled by laboratory rats can affect blood levels of
20 endothelin and cause a vasopressor response without causing acute lung injury. Moreover, the
21 potency to influence hemodynamic changes can be modified by removing the polar organic
22 compounds and soluble elements from the particles. In the study described previously
23 (Section 7.2.3), Ohio et al. (2000a) also have shown that inhalation of concentrated PM in
24 healthy nonsmokers causes increased levels of blood fibrinogen. They exposed 38 volunteers
25 exercising intermittently at moderate levels of exertion for 2 h to either filtered air or particles
26 concentrated from the air in Chapel Hill, NC (23 to 311 //g/m2). Blood obtained 18 h after
27 exposure contained significantly more fibrinogen than blood obtained before exposure. The
28 observed effects in blood may be associated with the mild pulmonary inflammation also found
29 18 h after exposure to CAP (see Section 7.2.3).
30 Gardner et al. (2000) examined whether the instillation of particles would alter blood
31 coagulability factors in laboratory animals. Sprague-Dawley rats were instilled with 0.3,1.7, or
April 2002 7-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3 mg/kg of ROFA or 8.3 mg/kg Mount Saint Helens volcanic ash. They observed an increase
2 in plasma fibrinogen in healthy rats. Because fibrinogen is a known risk factor for ischemic heart
3 disease and stroke, the authors suggested that this alteration in the coagulation pathway could
4 take part in the triggering of cardiovascular events in susceptible individuals. Elevations in
5 plasma fibrinogen, however, were observed in healthy rats only at the highest treatment dose; and
6 no other changes in clotting function were noted. Because the lower treatment doses are known
7 to cause pulmonary injury and inflammation, albeit to a lower extent, the absence of plasma
8 fibrinogen changes at these lower doses suggests that only high levels of pulmonary injury are
9 able to produce an effect in healthy test animals.
10 To establish the temporal relationship between pulmonary injury, increased plasma
11 fibrinogen, and changes in peripheral lymphocytes, Kodavanti et al., (2002) exposed
12 Spontaneously Hypertensive (SH) and Wistar-Kyoto (WKY) rats to ROFA using both
13 intratracheal and inhalation exposure (acute and long-term) scenarios. Increases in plasma
14 fibrinogen and decreases in circulating white blood cells were found during the acute phase
15 responses to ROFA exposure and were temporally associated with acute, but not long-term, lung
16 injury. A bolus intratracheal instillation of ROFA increased plasma fibrinogen in both SH and
17 WKY rats; whereas the increase was evident only in SH rats after acute ROFA inhalation. The
18 increased fibrinogen in SH rats corresponded to an inability to find increased pulmonary
19 glutathione and greater pulmonary injury and inflammation than was found in the WKY rats.
20 In summary, controlled animal studies have provided initial evidence that only high
21 concentrations of inhaled or instilled particles can have systemic, especially cardiovascular,
22 effects. In the case of MCT rats, these effects can be lethal. Controlled human exposure studies
23 also have shown ambient levels of inhaled PM can produce some biochemical and cellular
24 changes in the blood. Although some of these biochemical changes have been used as clinical
25 "markers" for cardiovascular diseases, the causal relationship between these changes and the
26 potential life-threatening diseases remains to be established. Understanding the pathways by
27 which very small concentrations of inhaled ambient PM can produce systemic, life-threatening
28 changes also is far from clear. Among the hypotheses that have been proposed to account for the
29 nonpulmonary effects of PM are activation of neural reflexes, cytokine effects on heart tissue
30 (Killingsworth et al., 1997), alterations in coagulability (Seaton et al., 1995; Sjogren, 1997),
31 perturbations in both conductive and hypoxemic arrythmogenic mechanisms (Watkinson et al.,
April 2002 7-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 1998; Campen et al., 2000), and altered endothelin levels (Vincent et al., 2001). A great deal of
2 research using controlled exposures of laboratory animals and human subjects to PM will be
3 necessary to test mechanistic hypotheses generated to date, as well as those that are likely to be
4 proposed in the future (see Section 7.5).
5
6
7 7.4 SUSCEPTIBILITY TO THE EFFECTS OF PARTICULATE
8 MATTER EXPOSURE
9 Susceptibility of an individual to adverse health effects of PM can vary depending on a
10 variety of host factors such as age, physiological activity profile, genetic predisposition, or
11 preexistent disease. The potential for preexistent disease to alter adverse response to toxicant
12 exposure is widely acknowledged but poorly understood. Because of inherent variability
13 (necessitating large numbers of subjects) and ethical concerns associated with using diseased
14 subjects in clinical research studies, a solid database on human susceptibilities is lacking. For
15 more control over both host and environmental variables, animal models often are used.
16 However, care must be taken in extrapolation from animal models of human disease to humans.
17 Rodent models of human disease, their use in toxicology, and the criteria for judging their
18 appropriateness as well as their limitations must be considered (Kodavanti et al., 1998b;
19 Kodavanti and Costa, 1999).
20
21 7.4.1 Pulmonary Effects of Particulate Matter in Compromised Hosts
22 Epidemiological studies suggest there may be subsegments of the population that are
23 especially susceptible to effects from inhaled particles (see Chapter 8). The elderly with chronic
24 cardiopulmonary disease, those with pneumonia and possibly other lung infections, and those
25 with asthma (at any age) appear to be at higher risk than healthy people of similar age.
26 Unfortunately, most toxicology studies have used healthy adult animals. An increasing number
27 of newer studies have examined effects of ambient particles in compromised host models. Costa
28 and Dreher (1997) used a rat model of cardiopulmonary disease to explore the question of
29 susceptibility and the possible mechanisms by which PM effects are potentiated. Rats with
30 advanced monocrotaline (MCT)-induced pulmonary vasculitis/hypertension were given
31 intratracheal instillations of ROFA (0, 0.25, 1.0, and 2.5 mg/rat). The MCT animals had a
April 2002 7-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 marked neutrophilic inflammation. In the context of this inflammation, ROFA induced a four- to
2 fivefold increase in BAL PMNs. There was increased mortality at 96 h that was ROFA-dose
3 dependent. The results of this study indicate that particles, albeit at a high concentration,
4 enhanced mortality in MCT animals but not in healthy animals.
5 As discussed previously, Kodavanti et al. (1999) also studied PM effects in the MCT rat
6 model of pulmonary disease. Rats treated with 60 mg/kg MCT were exposed to 0, 0.83. or 3.3
7 mg/kg ROFA by intratracheal instillation and to 15 mg/m3 ROFA by inhalation. Both methods
8 of exposure caused inflammatory lung responses; and ROFA exacerbated the lung lesions, as
9 shown by increased lung edema, inflammatory cells, and alveolar thickening.
10 The manner in which MCT can alter the response of rats to inhaled particles was examined
11 by Madl and colleagues (1998). Rats were exposed to fluorescent colored microspheres (1 //m)
12 2 weeks after treatment with MCT. In vivo phagocytosis of the microspheres was altered in the
13 MCT rats in comparison with control animals. Fewer microspheres were phagocytized in vivo
14 by alveolar macrophages, and there was a concomitant increase in free microspheres overlaying
15 the epithelium at airway bifurcations. The decrease in in vivo phagocytosis was not accompanied
16 by a similar decrease in vitro. Macrophage chemotaxis, however, was impaired significantly in
17 MCT rats compared with control rats. Thus, MCT appeared to impair particle clearance from the
18 lungs via inhibition of macrophage chemotaxis.
19 The sulfur dioxide (SO2)-induced model of chronic bronchitis has also been used to
20 examine the potential interaction of PM with preexisting lung injury. Clarke and colleagues
21 pretreated Sprague-Dawley rats for 6 weeks with air or 170 ppm SO2 for 5 h/day and 5 days/week
22 (Clarke et al., 1999). Exposure to concentrated ambient air particles for 5 h/day for 3 days at an
23 average concentration of 515 //g/m3 produced significant changes in both cellular and
24 biochemical markers in lavage fluid. In comparison to control animal values, protein was
25 increased approximately threefold in SO2-pretreated animals exposed to concentrated ambient
26 PM. Lavage fluid neutrophils and lymphocytes were increased significantly in both groups of
27 rats exposed to concentrated ambient PM, with greater increases in both cell types in the
28 SO2-pretreated rats. Thus, exposure to concentrated ambient PM produced adverse changes in
29 the respiratory system, but no deaths, in both normal rats and in a rat model of chronic bronchitis.
30 Clarke et al. (2000b) next examined the effect of concentrated ambient PM from Boston,
31 MA, in normal rats of different ages. Unlike the earlier study that used Sprague-Dawley rats,
April 2002 7-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 4- and 20-mo-old Fischer 344 rats were examined after exposure to concentrated ambient PM for
2 5 h/day for 3 consecutive days. They found that exposure to the daily mean concentrations of 80,
3 170, and 50 //g/m3 PM, respectively, produced statistically significant increases in total
4 neutrophil counts (over 10-fold) in lavage fluid of the young, but not the old, rats. Thus, repeated
5 exposure to relatively low concentrations of ambient PM produced an inflammatory response,
6 although the actual percent neutrophils in the concentrated ambient PM-exposed young adult rats
7 was low (approximately 3%). On the other hand, Gordon et al. (2000) found no evidence of
8 neutrophil influx in the lungs of normal and monocrotaline-treated Fischer 344 rats exposed in
9 nine separate experiments to concentrated ambient PM from New York, NY, as high as
10 400 //g/m3 for a 6-h exposure or 192 //g/m3 for three daily 6-h exposures. Similarly, normal and
11 cardiomyopathic hamsters showed no evidence of pulmonary inflammation or injury after a
12 single exposure to the same levels of concentrated ambient PM. Gordon and colleagues did
13 report a statistically significant doubling in protein concentration in lavage fluid in
14 monocrotaline-treated rats exposed for 6 h to 400 //g/m3 concentrated ambient PM. Because of
15 the disparity in findings in the response of normal Fischer 344 rats to concentrated ambient PM
16 between these two labs, it is important that the reproducibility of these experiments be examined.
17 Kodavanti and colleagues (1998b) also have examined the effect of concentrated ambient
18 PM in normal rats and rats with sulfur dioxide-induced chronic bronchitis. Among the four
19 separate exposures to PM, there was a significant increase in lavage fluid protein in bronchitic
20 rats from only one exposure protocol in which the rats were exposed to 444 and 843 //g/m3 PM
21 on 2 consecutive days (6 h/day). Neutrophil counts were increased in bronchitic rats exposed to
22 concentrated ambient PM in three of the four exposure protocols, but was decreased in the fourth
23 protocol. No other changes in normal or bronchitic rats were observed, even in the exposure
24 protocols with higher PM concentrations. Thus, rodent studies have demonstrated that
25 inflammatory changes can be produced in normal and compromised animals exposed to
26 concentrated ambient PM. These findings are important because only a limited number of
27 studies have used real-time inhalation exposures to actual ambient urban PM.
28 Pulmonary function measurements are often less invasive than other means to assess the
29 effects of inhaled air pollutants on the mammalian lung. After publication of the 1996 PM
30 AQCD, a number of investigators examined the response of rodents and dogs to inhaled ambient
31 particles. In general, these investigators have demonstrated that ambient PM has minimal effects
April 2002 7-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 on pulmonary function tests. Gordon et al. (2000) exposed normal and monocrotaline-treated
2 rats to filtered air or 181 //g/m3 concentrated ambient PM for 3 h. For both normal and
3 monocrotaline-treated rats, no differences in lung volumes or diffusion capacities for carbon
4 monoxide were observed between the air or PM exposed animals at 3 or 24 h after exposure.
5 Similarly, in cardiomyopathic hamsters, concentrated ambient PM had no effect on these same
6 pulmonary function measurements.
7 Other pulmonary function endpoints have been studied in animals exposed to concentrated
8 ambient PM. Clarke et al. (1999) observed that tidal volume was increased slightly in both
9 control rats and rats with sulfur dioxide-induced chronic bronchitis exposed to 206 to 733 //g/m3
10 PM on 3 consecutive days. No changes in peak expiratory flow, respiratory frequency, or minute
11 volume were observed after exposure to concentrated ambient PM. In the series of dog studies
12 by Godleski et al. (2000) (also see Section 7.3), no signficant changes in pulmonary function
13 were observed in normal mongrel dogs exposed to concentrated ambient PM, although a 20%
14 decrease in respiratory frequency was observed in dogs that underwent coronary artery occlusion
15 and were exposed to PM. Thus, studies using normal and compromised animal models exposed
16 to concentrated ambient PM have found minimal biological effects of ambient PM on pulmonary
17 function.
18 Johnston et al. (1998) exposed 8-week-old mice (young) and 18-mo-old mice (old) to
19 polytetrafluoroethylene (PTFE) fumes (0, 10, 25, and 50 //g/m3) for 30 min. Lung lavage
20 endpoints (PMN, protein, LDH, and p-glucuronidase) as well as lung tissue mRNA levels for
21 various cytokines, metallothionein and for Mn superoxide dismutase were measured 6 h
22 following exposure. Protein, lymphocyte, PMN, and TNF-a mRNA levels were increased in
23 older mice when compared to younger mice. These findings suggest that the inflammatory
24 response to PTFE fumes is altered with age, being greater in the older animals. Although
25 ultrafme PTFE fumes are not a valid surrogate for ambient ultrafme particles (Oberdorster et al.,
26 1992), this study did provide evidence to support the hypothesis that particle-induced pulmonary
27 inflammation is different between young and old mice. Further studies on age-related PM effects
28 are described in Section 7.6 (Responses to PM and Gaseous Pollutant Mixtures).
29 Kodavanti et al. (2000b; 2001) used genetically predisposed spontaneously hypertensive
30 (SH) rats as a model of cardiovascular disease to study PM-related susceptibility. The SH rats
31 were found to be more susceptible to acute pulmonary injury from intratracheal ROFA exposure
April 2002 7-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 than normotensive control Wistar Kyoto (WKY) rats (Kodavanti et al., 2001). The primary
2 metal constituents of ROFA, V and Ni, caused differential species-specific effects. Vanadium,
3 which was less toxic than Ni in both strains, caused inflammatory responses only in WKY rats;
4 whereas Ni was injurious to both WKY and SH rats (SH > WKY). This differential
5 responsiveness of V and Ni was correlated with their specificity for airway and parenchymal
6 injury, discussed in another study (Kodavanti et al., 1998b). When exposed to the same ROFA
7 by inhalation, SH rats were more sensitive than WKY rats in regards to vascular leakage
8 (Kodavanti et al., 2000b). The SH rats exhibited a hemorrhagic response to ROFA. Oxidative
9 stress was much higher in ROFA exposed SH rats than matching WKY rats. Also, SH rats,
10 unlike WKY rats, showed a compromised ability to increase BALF glutathione in response to
11 ROFA, suggesting a potential link to increased susceptibility. Cardiovascular effects were
12 characterized by ST-segment area depression of the ECG in ROFA-exposed SH but not WKY
13 rats. When the same rats were exposed to ROFA by inhalation (Kodavanti et al., 2002),
14 differences in effects were found depending on the length of exposure. After acute exposure,
15 increased plasma fibrinogen was associated with lung injury; longer-term, episodic ROFA
16 exposure resulted in progressive protein leakage and inflammation that was significantly worse in
17 SH rats when compared to WKY rats. These studies demonstrate the potential utility of
18 cardiovascular disease models for the study of PM health effects and show that genetic
19 predisposition to oxidative stress and cardiovascular disease may play a role in sensitivity to
20 increased PM-related cardiopulmonary injury.
21 On the basis of in vitro studies, Sun et al. (2001) predicted that the antioxidant and lipid
22 levels in the lung lining fluid may determine susceptibility to inhaled PM. In a subsequent study
23 from the same laboratory, Norwood et al. (2001) conducted inhalation studies on guinea pigs to
24 test this hypothesis. The guinea pigs were divided on the basis of dietary supplementation or
25 depletion of ascorbic acid (C) and glutathione (GSH) into four groups: (+C+GSH), (+C-GSH),
26 (-C+GSH), and (-C-GSH). All groups were exposed, nose-only, to clean air or 19-25 mg/m3
27 ROFA (< 2.5 jim) for 2 h. Nasal lavage and BAL fluid and cells were examined at 0 h and 24 h
28 postexposure. Exposure to ROFA increased lung injury in the (-C-GSH) group only, as shown
29 by increased BAL fluid protein, LDH, and PMNs and decreased BAL macrophages, and resulted
30 in lower antioxidant concentrations in BAL fluid than were found with single deficiencies.
April 2002 7-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 In summary, although more of these studies are just emerging and are only now being
2 replicated or followed more thoroughly to investigate the mechanisms, they do provide evidence
3 of enhanced susceptibility to inhaled PM in "compromised" hosts.
4
5 7.4.2 Genetic Susceptibility to Inhaled Particles and their Constituents
6 A key question in understanding the adverse health effects of inhaled PM is which
7 individuals are susceptible to PM. Although factors such as age and health status have been
8 studied in both epidemiology and toxicology studies, a number of investigators have begun to
9 examine the importance of genetic susceptibility in the response to inhaled particles because of
10 considerable evidence that genetic factors play a role in the response to inhaled pollutant gases.
11 To accomplish this goal, investigators typically have studied the interstrain response to particles
12 in rodents. The response to ROFA instillation in different strains of rats has been investigated by
13 Kodavanti et al. (1996, 1997a). In the first study, male Sprague-Dawley (SD) and Fischer-344
14 (F-344) rats were instilled intratracheally with saline or ROFA particles. ROFA instillation
15 produced an increase in lavage fluid neutrophils in both SD and F-344 rats; whereas a time-
16 dependent increase in eosinophils occurred only in SD rats. In the subsequent study (Kodavanti
17 et al., 1997a), SD, Wistar (WIS), and F-344 rats (60 days old) were exposed to saline or ROFA
18 (8.3 mg/kg) by intratracheal instillation and examined for up to 12 weeks. Histology indicated
19 focal areas of lung damage showing inflammatory cell infiltration as well as alveolar, airway, and
20 interstitial thickening in all three rat strains during the week following exposure. Trichrome
21 staining for fibrotic changes indicated a sporadic incidence of focal alveolar fibrosis at 1, 3, and
22 12 weeks in SD rats; whereas WIS and F-344 rats showed only a modest increase in trichrome
23 staining in the septal areas. One of the isoforms of fibronectin mRNA was upregulated in
24 ROFA-exposed SD and WIS rats, but not in F-344 rats. Thus, in rats there appears to be a
25 genetic based difference in susceptibility to lung injury induced by instilled ROFA.
26 Differences in the degree of pulmonary inflammation have been described in rodent strains
27 exposed to airborne pollutants. To understand the underlying causes, signs of airway
28 inflammation (i.e., airway hyper-responsiveness, inflammatory cell influx) were established in
29 responsive (BABL/c) and non-responsive (C57BL/6) mouse strains exposed to ROFA (Veronesi
30 et al., 2000). Neurons taken from the ganglia (i.e., dorsal root ganglia) that innervate the nasal
31 and upper airways were cultured from each mouse strain and exposed to ROFA. The difference
April 2002 7-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 in inflammatory response noted in these mouse strains in vivo was retained in culture, with
2 C57BL/6 neurons showing significantly lower signs of biological activation (i.e., increased
3 intracellular calcium levels) and cytokine (i.e., IL-6, IL-8) release relative to BALB/c mice.
4 RT-PCR and immunocytochemistry indicated that the BALB/c mouse strain had a significantly
5 higher number of neuropeptide and acid-sensitive (i.e., NK1, VR1) sensory receptors on their
6 sensory ganglia relative to the C57BL/6 mice. Such data indicate that genetically-determined
7 differences in sensory inflammatory receptors can influence the degree of PM-induced airway
8 inflammation.
9 Kleeberger and colleagues have examined the role that genetic susceptibility plays in the
10 effect of inhaled acid-coated particles on macrophage function. Nine inbred strains of mice were
11 exposed nose-only to carbon particles coated with acid (10 mg/m3 carbon with 285 //g/m3 sulfate)
12 for 4 h (Ohtsuka et al., 2000a). Significant inter-strain differences in Fc-receptor-mediated
13 macrophage phagocytosis were observed, with C57BL/6J mice being the most sensitive.
14 Although neutrophil counts were increased more in C3H/HeOuJ and C3H/HeJ strains of mice
15 than in the other strains, the overall magnitude of change was small and not correlated with the
16 changes in macrophage phagocytosis. In follow-up studies using the same type particle, Ohtsuka
17 et al. (2000a,b) performed a genome-wide scan with an intercross cohort derived from C57BL/6J
18 and C3H/HeJ mice. Analyses of macrophage dysfunction phenotypes of segregant and
19 nonsegregant populations derived from these two strains indicate that two unlinked genes control
20 susceptibility. They identified a 3-centiMorgan segment on mouse chromosome 17 that contains
21 an acid-coated particle susceptibility locus. Interestingly, this quantitative trait locus overlaps
22 with those described for ozone-induced inflammation (Kleeberger et al., 1997) and acute lung
23 injury (Prows et al., 1997) and contains several promising candidate genes that may be
24 responsible for the observed genetic susceptibility for macrophage dysfunction in mice exposed
25 to acid-coated particles.
26 Leikauf and colleagues (Leikauf et al., 2000; Wesselkamper et al., 2000; McDowell et al.,
27 2000; Prows and Leikauf, 2001; Leikauf et al., 2001) have identified a genetic susceptibility in
28 mice that is associated with mortality following exposures to high concentrations (from 15 to 150
29 |ig/m3) of a MSO4 aerosol (0.22 |im MMAD) for up to 96 h. These studies also have
30 preliminarily identified the chromosomal locations of a small number of genes that may be
31 responsible for this genetic susceptibility. This finding is particularly significant in light of the
April 2002 7-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 toxicology studies demonstrating that bioavailable, first-row transition metals participate in the
2 acute lung injury following exposure to emission and ambient air particles. Similar genes may be
3 involved in human responses to particle-associated metals. However, additional studies will be
4 required to determine whether the identified metal susceptibility genes are involved in human
5 responses to ambient levels of particulate-associated metals.
6 One study has examined the interstrain susceptibility to ambient particles. C57BL/6J and
7 C3H/HeJ mice were exposed to 250 //g/m3 concentrated ambient PM2 5 for 6 h and examined at
8 0 and 24 h after exposure for changes in lavage fluid parameters and cytokine mRNA expression
9 in lung tissue (Shukla et al., 2000). No interstrain differences in response were observed.
10 Surprisingly, although no indices of pulmonary inflammation or injury were increased over
11 control values in the lavage fluid, increases in cytokine mRNA expression were observed in both
12 murine strains exposed to PM2 5. Although the increase in cytokine mRNA expression was
13 generally small (approximately twofold), the effects on IL-6, TNF-a, TGF-P2, and y-interferon
14 were consistent.
15 Thus, a handful of studies have begun to demonstrate that genetic susceptibility can play a
16 role in the response to inhaled particles. However, the doses of PM administered in these
17 studies, whether by inhalation or instillation, were extremely high when compared to ambient
18 PM levels. Similar strain differences in response to inhaled metal particles have been observed
19 by other investigators (McKenna et al., 1998; Wesselkamper et al., 2000), although the
20 concentration of metals used in these studies were also more relevant to occupational rather than
21 environmental exposure levels. It remains to be determined whether genetic susceptibility plays
22 as significant a role in the adverse effects of ambient PM as does age or health status.
23
24 7.4.3 Effect of Particulate Matter on Allergic Hosts
25 Relatively little is known about the effects of inhaled particles on humoral (antibody) or
26 cell-mediated immunity. Alterations in the response to a specific antigenic challenge have been
27 observed in animal models at high concentrations of acid sulfate aerosols (above 1,000 //g/m3)
28 (Pinto et al., 1979; Kitabatake et al., 1979; Fujimaki et al., 1992). Several studies have reported
29 an enhanced response to nonspecific bronchoprovocation agents, such as acetylcholine and
30 histamine, after exposure to inhaled particles. This nonspecific airway hyperresponsiveness,
31 a central feature of asthma, occurs in animals and human subjects exposed to sulfuric acid under
April 2002 7-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 controlled conditions (Gearhart and Schlesinger, 1986; Utell et al., 1983). Although, its
2 relevance to specific allergic responses in the airways of atopic individuals is unclear, it
3 demonstrates that the airways of asthmatics may become sensitized to either specific or
4 nonspecific triggers that could result in increases in asthma severity and asthma-related hospital
5 admissions (Peters et al., 1997; Jacobs et al., 1997; Lipsett et al., 1997). Combustion particles
6 also may serve as carrier particles for allergens (Knox et al., 1997).
7 A number of in vivo and in vitro studies have demonstrated that DPM can alter the immune
8 response to challenge with specific antigens and suggest that DPM may act as an adjuvant.
9 These studies have shown that treatment with DPM enhances the secretion of antigen-specific
10 IgE in mice (Takano et al., 1997) and in the nasal cavity of human subjects (Diaz-Sanchez et al.,
11 1996, 1997; Ohtoshi et al., 1998). Because IgE levels play a major role in allergic asthma
12 (Wheatley and Platts-Mills, 1996), upregulation of its production could lead to an increased
13 response to inhaled antigen in particle-exposed individuals.
14 Van Zijverden et al. (2000a,b) used mouse models to assess the potency of particles to
15 adjuvate an immune response to a protein antigen. All particles exert an adjuvant effect on the
16 immune response to co-administered antigen, apparently stimulated by the particle core rather
17 than the attached chemical factors. Different particles, however, stimulate distinct types of
18 immune responses. In one model (Van Zijverden et al., 2001), BALB/c mice were intranasally
19 treated with a mixture of antigen (model antigen TNP-Ovalbumin, TNP-OVA) and particles on
20 three consecutive days. On day 10 after sensitization mice were challenged with the antigen
21 TNP-OVA alone and five days later the immune response was assessed. DPM, as well as carbon
22 black particles (CB), were capable of adjuvating the immune response to TNP-OVA as
23 evidenced by an increase of TNP-specific antibody (IgGl and IgE) secreting B cells antibodies in
24 the lung-draining lymph nodes. Increased antigen-specific IgGl, IgG2a, and IgE isotypes were
25 measured in the serum, indicating that the response resulted in systemic sensitization.
26 Importantly, an increase of eosinophils in the bronchio-alveolar lavage was observed with CB.
27 Companion studies with the intranasal exposure model showed that the adjuvant effect of
28 particles (CB) was even more pronounced when the particles were given during both the
29 sensitization and challenge phases; whereas administration during the challenge phase caused
30 only marginal changes on the immune response. These data show that particulate matter can
31 increase both the sensitization and challenge responses to a protein antigen, and the immune
April 2002 7-53 DRAFT-DO NOT QUOTE OR CITE
-------
1 stimulating activity of particles appears to be a time-dependent process, suggesting that an
2 inflammatory microenvironment, such as may be created by the particles, is crucial for enhancing
3 sensitization by particles.
4 Only a small number of studies have examined the mechanisms underlying the
5 enhancement of allergic asthma by ambient urban particles. Ohtoshi et al. (1998) reported that a
6 coarse size-fraction of resuspended ambient PM, collected in Tokyo, induced the production of
7 granulocyte macrophage colony stimulating factor (GMCSF), an upregulator of dendritic cell
8 maturation and lymphocyte function, in human airway epithelial cells in vitro. In addition to
9 increased GMCSF, epithelial cell supernatants contained increased IL-8 levels when incubated
10 with DPM, a principal component of ambient particles collected in Tokyo. Although the sizes of
11 the two types of particles used in this study were not comparable, the results suggest that ambient
12 PM, or at least the DPM component of ambient PM, may be able to upregulate the immune
13 response to inhaled antigen through GMCSF production. Similarly, Takano et al. (1998) has
14 reported airway inflammation, airway hyperresponsiveness, and increased GMGSF and IL-5 in
15 mice exposed to diesel exhaust.
16 In a study by Walters et al. (2001), PM10 was found to induce airway hyperresponsiveness,
17 suggesting that PM exposure may be an important factor in increases in asthma prevalence.
18 Naive mice were exposed to a single dose (0.5 mg/ mouse) of ambient PM, coal fly ash, or diesel
19 PM. Exposure to PM10 induced increases in airway responsiveness and BAL cellularity; whereas
20 diesel PM induced significant increases in BAL cellularity, but not airway responsiveness.
21 On the other hand, coal fly ash exposure did not elicit significant changes in either of these
22 parameters. Ambient PM-induced airway hyperresponsiveness was sustained over 7 days. The
23 increase in airway responsiveness was preceded by increases in BAL eosinophils; whereas a
24 decline in airway responsiveness was associated with increases in macrophages. Thus, ambient
25 PM can induce asthma-like parameters in naive mice.
26 In an examination of the effect of concentrated ambient PM on airway responsiveness in
27 mice, Goldsmith and colleagues (1999) exposed control and ovalbumin-sensitized mice to an
28 average concentration of 787 //g/m3 PM for 6 h/day for 3 days. Although ovalbumin
29 sensitization itself produced an increase in the nonspecific airway responsiveness to inhaled
30 methylcholine, concentrated ambient PM did not change the response to methylcholine in
31 ovalbumin-sensitized or control mice. For comparison, these investigators examined the effect
April 2002 7-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 of inhalation of an aerosol of the active soluble fraction of ROFA on control and ovalbumin-
2 sensitized mice and demonstrated that ROFA could produce nonspecific airway
3 hyperresponsiveness to methylcholine in both control and ovalbumin-sensitized mice. Similar
4 increases in airway responsiveness have been observed after exposure to ROFA in normal and
5 ovalbumin-sensitized rodents (Gavett et al., 1997, 1999; Hamada et al., 1999, 2000).
6 Gavett et al. (1999) have investigated the effects of ROFA (intratracheal instillation) in
7 ovalbumin (OVA) sensitized and challenged mice. Instillation of 3 mg/kg (approximately 60 //g)
8 ROFA induced inflammatory and physiological responses in the OVA mice that were related to
9 increases in Th2 cytokines (IL-4, IL-5). Compared to OVA sensitization alone, ROFA induced
10 greater than additive increases in eosinophil numbers and in airway responsiveness to
11 methylcholine.
12 Hamada et al. (1999, 2000) have examined the effect of a ROFA leachate aerosol in a
13 neonatal mouse model of allergic asthma. In the first study, neonatal mice sensitized by
14 intraperitoneal (ip) injection with OVA developed airway hyperresponsiveness, eosinophilia, and
15 elevated serum anti-ovalbumin IgE after a challenge with inhaled OVA. Exposure to the ROFA
16 leachate aerosol had no marked effect on the airway responsiveness to inhaled methacholine in
17 nonsensitized mice, but did enhance the airway hyperresponsiveness to methylcholine produced
18 in OVA-sensitized mice. No other interactive effects of ROFA exposure with OVA were
19 observed. In a subsequent study, Hamada et al. clearly demonstrated that, whereas inhaled OVA
20 alone was not sufficient to sensitize mice to a subsequent inhaled OVA challenge, pretreatment
21 with a ROFA leachate aerosol prior to the initial exposure to aerosolized OVA resulted in an
22 allergic response to the inhaled OVA challenge. Thus, exposure to a ROFA leachate aerosol can
23 alter the immune response to inhaled OVA both at the sensitization stage at an early age and at
24 the challenge stage.
25 Lambert et al. (1999) also examined the effect of ROFA on a rodent model of pulmonary
26 allergy. Rats were instilled intratracheally with 200 or 1,000 //g ROFA 3 days prior to
27 sensitization with house dust mite (HDM) antigen. HDM sensitization after 1000 //g ROFA
28 produced increased eosinophils, LDH, BAL protein, and IL-10 relative to HDM alone. The
29 immediate bronchoconstrictive and associated antigen-specific IgE response to a subsequent
30 antigen challenge was increased in the ROFA-treated group in comparison with the control
31 group. Together, these studies suggest the components of ROFA can augment the immune
April 2002 7-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 response to antigen. Evidence that metals are responsible for the ROFA-enhancement of an
2 allergic sensitization was demonstrated by Lambert et al. (2000). In this follow-up study, Brown
3 Norway rats were instilled with 1 mg ROFA or the three main metal components of ROFA (iron,
4 vanadium, or nickel) prior to sensitization with instilled house dust mite. The three individual
5 metals were found to augment different aspects of the immune response to house dust mite.
6 Nickel and vanadium produced an enhanced immune response to the antigen as seen by higher
7 house dust mite-specific IgE serum levels after an antigen challenge at 14 days after sensitization.
8 Nickel and vanadium also produced an increase in the lymphocyte proliferative response to
9 antigen in vitro. In addition, the antigen-induced bronchoconstrictive response was greater only
10 in nickel-treated rats. Thus, instillation of metals at concentrations equivalent to those present in
11 the ROFA leachate mimicked the response to ROFA, suggesting that the metal components of
12 ROFA are responsible for the increased allergic sensitization observed in ROFA-treated animals.
13 Although these studies demonstrate that inhalation or instillation of ROFA augments the
14 immune response in allergic hosts, the applicability of these findings to ambient PM is an
15 important consideration. Goldsmith et al. (1999) have compared the effect of inhalation of
16 concentrated ambient PM for 6 h/day for 3 days versus the effect of a single exposure to a ROFA
17 leachate aerosol on the airway responsiveness to methylcholine in OVA-sensitized mice.
18 Exposure to ROFA leachate aerosols significantly enhanced the airway hyperresponsiveness in
19 OVA-sensitized mice; whereas, exposure to concentrated ambient PM (average concentration of
20 787 //g/m3) had no effect on airway responsiveness in six separate experiments. Thus, the effect
21 of the ROFA leachate aerosols on the induction of airway hyperresponsiveness in allergic mice
22 was significantly different than that of a high concentration of concentrated ambient PM.
23 Although airway responsiveness was examined at only one post-exposure time point, these
24 findings do suggest that a great deal of caution should be used in interpreting the results of
25 studies using ROFA particles or leachates in the attempt to investigate the biologic plausibility of
26 the adverse health effects of PM.
27 Several other studies have examined in greater detail the contribution of the particle
28 component and the organic fraction of DPM to allergic asthma. Tsien et al. (1997) treated
29 transformed IgE-producing human B lymphocytes in vitro with the organic extract of DPM. The
30 organic phase extraction had no effect on cytokine production but did increase IgE production.
31 In these in vitro experiments, DPM appeared to be acting on cells already committed to IgE
April 2002 7-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 production, thus suggesting a mechanism by which the organic fraction of combustion particles
2 can directly affect B cells and influence human allergic asthma.
3 Cultured epithelial cells from atopic asthmatics show a greater response to DPM exposure
4 when compared with cells from nonatopic nonasthmatics. IL-8, GM-CSF, and soluble ICAM-1
5 increased in response to DPM at a concentration of 10 //g/mL DPM (Bayram et al., 1998a,b).
6 This study suggests that particles could modulate airway disease through their actions on airway
7 epithelial cells. This study also suggests that bronchial epithelial cells from asthmatics are
8 different from those of nonasthmatics in regard to their mediator release in response to DPM.
9 Sagai and colleagues (1996) repeatedly instilled mice with DPM for up to 16 weeks and
10 found increased numbers of eosinophils, goblet cell hyperplasia, and nonspecific airway
11 hyperresponsiveness, changes which are central features of chronic asthma (National Institutes of
12 Health, 1997). Takano et al. (1997) extended this line of research and examined the effect of
13 repeated instillation of DPM on the antibody response to antigen (OVA) in mice. They observed
14 that antigen-specific IgE and IgG levels were significantly greater in mice repeatedly instilled
15 with both DPM and OVA. Because this upregulation in antigen-specific immunoglobulin
16 production was not accompanied by an increase in inflammatory cells or cytokines in lavage
17 fluid, it would suggest that, in vivo, DPM may act directly on immune system cells, as described
18 in the work by Tsien et al. (1997). Animal studies have confirmed that the adjuvant activity of
19 DPM also applies to the sensitization of Brown-Norway rats to timothy grass pollen (Steerenberg
20 etal., 1999).
21 Diaz-Sanchez and colleagues (1996) have continued to study the mechanism of DPM-
22 induced upregulation of allergic response in the nasal cavity of human subjects. In one study, a
23 200 //L aerosol bolus containing 0.15 mg of DPM was delivered into each naris of subjects with
24 or without seasonal allergies. In addition to increases in IgE in nasal lavage fluid (NAL), they
25 found an enhanced production of IL-4, IL-6, and IL-13, cytokines known to be B cell
26 proliferation factors. The levels of several other cytokines also were increased, suggesting a
27 general inflammatory response to a nasal challenge with DPM. In a following study, these
28 investigators delivered ragweed antigen, alone or in combination with DPM, on two occasions, to
29 human subjects with both allergic rhinitis and positive skin tests to ragweed (Diaz-Sanchez et al.
30 1997). They found that the combined challenge with ragweed antigen and DPM produced
31 significantly greater antigen-specific IgE and IgG4 in NAL. A peak response was seen at 96 h
April 2002 7-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 postexposure. The combined treatment also induced expression of IL-4, IL-5, IL-10, and IL-13,
2 with a concomitant decrease in expression of Thl-type cytokines. Although the treatments were
3 not randomized (antigen alone was given first to each subject), the investigators reported that
4 pilot work showed no interactive effect of repeated antigen challenge on cellular and biochemical
5 markers in NAL. DPM also resulted in the nasal influx of eosinophils, granulocytes, monocytes,
6 and lymphocytes, as well as the production of various inflammatory mediators. The combined
7 DPM plus ragweed exposure did not increase the rhinitis symptoms beyond those of ragweed
8 alone. Thus, diesel exhaust (particles and gases) can produce an enhanced response to antigenic
9 material in the nasal cavity.
10 Extrapolation of these findings of enhanced allergic response in the nose to the human lung
11 would suggest that ambient combustion particles containing DPM may have significant effects
12 on allergic asthma. A study by Nordenhall et al. (2001) has addressed the effects of diesel PM on
13 airway hyperresponsiveness, lung function and airway inflammation in a group of atopic
14 asthmatics with stable disease. All were hyperresponsive to methacholine. Each subject was
15 exposed to DPM (300 //g/m3) and air for 1 h on two separate occasions. Lung function was
16 measured before and immediately after the exposures. Sputum induction was performed 6 h, and
17 methacholine inhalation test 24 h, after each exposure. Exposure to DE was associated with a
18 significant increase in the degree of hyperresponsiveness, as compared to after air, a significant
19 increase in airway resistance and in sputum levels of interleukin (IL)-6 (p=0.048). No changes
20 were detected in sputum levels of methyl-histamine, eosinophil cationic protein,
21 myeloperoxidase, and IL-8.
22 These studies provide biological plausibility for the exacerbation of allergic asthma
23 associated with episodic exposure to PM. Although DPM may make up only a fraction of the
24 mass of urban PM, because of their small size, DPM may represent a significant fraction of the
25 ultrafme particle mode in urban air, especially in cities and countries that rely heavily on diesel-
26 powered vehicles. It must be noted that the potential contribution of DPM to the rising
27 prevalence in asthma is complicated by the fact that DPM levels have been decreasing over the
28 last decade (CALEPA report). The reported decrease in DPM levels is a result of the increased
29 combustion efficiency of diesel engines. This improvement in diesel engine design also has
30 brought about a significant decrease in the particle size of diesel emissions. Thus, the balance
April 2002 7-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 between a decrease in diesel emissions (in terms of mass) versus the production of a smaller and
2 potentially more toxic particle size needs further exploration.
3
4 7.4.4 Resistance to Infectious Disease
5 The development of an infectious disease requires both the presence of the appropriate
6 pathogen, as well as host susceptibility to the pathogen. There are numerous specific and
7 nonspecific host defenses against microbes, and the ability of inhaled particles to modify
8 resistance to bacterial infection could result from a decreased ability to clear or kill microbes.
9 Rodent infectivity models frequently have been used to examine the effect of inhaled particles on
10 host defense and infectivity. Mice or rats are challenged with a bacterial or viral load either
11 before or after exposure to the particles (or gas) of interest; mortality rate, survival time, or
12 bacterial clearance are then examined. A number of studies that have used the infectivity model
13 to assess the effect of inhaled PM were discussed previously (U.S. Environmental Protection
14 Agency, 1982, 1989, 1996a). In general, acute exposure to sulfuric acid aerosols at
15 concentrations up to 5,000 //g/m3 were not very effective in enhancing mortality in a bacterially
16 mediated murine model. In rabbits, however, sulfuric acid aerosols altered anti-microbial
17 defenses after exposure for 2 h/day for 4 days to 750 //g/m3 (Zelikoff et al., 1994). Acute or
18 short-term repeated exposures to high concentrations of relatively inert particles have produced
19 conflicting results. Carbon black (10,000 //g/m3) was found to have no effect on susceptibility to
20 bacterial infection (Jakab, 1993); whereas a very high concentration (20,000 //g/m3) of TiO2
21 decreased the clearance of microbes and the bacterial response of lymphocytes isolated from
22 mediastinal lymph nodes (Gilmour et al., 1989a,b). In addition, exposure to DPM (2 mg/m3,
23 7h/d, 5d/wk for 3 and 6 mo) has been shown to enhance the susceptibility of mice to the lethal
24 effects of some, but not all, microbial agents (Hahon et al., 1985). Thus, the pulmonary response
25 to microbial agents has been shown to be altered at relatively high particle concentrations in
26 animal models. Moreover, these effects appear to be highly dependent on the microbial
27 challenge and the test animal studied. Pritchard et al. (1996) observed in rats exposed to particles
28 with a high concentration of metals (e.g., ROFA), that the increased mortality rate after
29 streptococcus infection was associated with the amount of metal in the PM.
30 Despite the reported association between ambient PM and deaths caused by pneumonia
31 (Schwartz, 1994), there are few recent studies that have examined the mechanisms that may be
April 2002 7-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 responsible for the effect of PM on infectivity. In one study, Cohen and colleagues (1997)
2 examined the effect of inhaled vanadium (V) on immunocompetence. Healthy rats were
3 repeatedly exposed to 2 mg/m3 V, as ammonium metavanadate, and then instilled with
4 polyinosinic-polycytidilic acid (poly I:C), a double-stranded polyribonucleotide that acts as a
5 potent immunomodulator. Induction of increases in lavage fluid protein and neutrophils was
6 greater in animals preexposed to V. Similarly, IL-6 and interferon-gamma were increased in
7 V-exposed animals. Alveolar macrophage function, as determined by zymosan-stimulated
8 superoxide anion production and by phagocytosis of latex particles, was depressed to a greater
9 degree after poly I:C instillation in V-exposed rats as compared to filtered air-exposed rats.
10 These findings provide evidence that inhaled V, a trace metal found in combustion particles and
11 shown to be toxic in vivo in studies using instilled or inhaled ROFA (Dreher et al., 1997;
12 Kodavanti et al., 1997b, 1999), has the potential to inhibit the pulmonary response to microbial
13 agents. It must be taken into consideration that these effects were found at very high exposure
14 concentrations of V, and as with many studies, care must be taken in extrapolating the results to
15 the ambient exposure of healthy individuals or those with preexisting cardiopulmonary disease to
16 trace concentrations (approximately 3 orders of magnitude lower concentration) of metals in
17 ambient PM.
18
19
20 7.5 PARTICULATE MATTER TOXICITY AND PATHOPHYSIOLOGY:
21 IN VITRO EXPOSURES
22 7.5.1 Introduction
23 Toxicological studies play an integral role in determining the biological plausibility for the
24 health effects associated with ambient PM exposure. At the time of completion of the previous
25 PM AQCD (U.S. Environmental Protection Agency, 1996a) very little was known about the
26 potential mechanisms that could explain the morbidity and mortality observed in populations
27 exposed to PM. One of the difficulties in trying to sort out possible mechanisms is the nature of
28 particles themselves. Ambient PM has diverse physicochemical properties (Table 7-9) ranging
29 from the physical characteristics of the particle to the chemical components in or on the surface
30 of the particle. Any one of these properties could change at any time in the ambient exposure
31 atmosphere, making it hard to replicate the actual properties in a controlled experiment. As a
April 2002 7-60 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-9. PHYSICOCHEMICAL PROPERTIES OF PARTICIPATE MATTER
Physical Characteristics
Chemical Components
particle mass (size, shape, density)
particle number
surface area
surface chemistry
surface charge
acidity
• elemental and organic carbon
• volatile organics
• metals (Fe, Cd, Co, Cu, Mn, Ni, Pb, Ti, V, Zn)
• biologicals (e.g., pollen, microbes)
• sulfates
• nitrates
* pesticides
1 result, controlled exposure studies as yet have not been able to unequivocally determine the
2 particle properties and the specific mechanisms by which ambient PM may affect biological
3 systems.
4 Despite these underlying difficulties, a larger number of toxicological studies have become
5 available since 1996 to help explain how ambient particles may exert toxic effects on the
6 cardiovascular and respiratory systems. The following section discusses the more recently
7 published studies that provide an approach toward identifying potential mechanisms by which
8 PM mediates health effects. The remaining sections discuss potential mechanisms in relation to
9 PM characteristics based on these available data.
10
11 7.5.2 Experimental Exposure Data
12 In vitro exposure is a useful technique to provide information on potential hazardous PM
13 constituents and mechanisms of PM injury, especially when only limited quantities of the test
14 material are available. In addition, in vitro exposure allows the examination of the response to
15 particles in only one or two cell types. Respiratory epithelial cells that line the airway lumen,
16 constitute the initial targets of airborne pollutants. These cells have been featured in numerous
17 studies involving airborne pollutants and show inflammatory responses similar to that of human
18 primary epithelial cultures. Limitations of in vitro studies include difficulty in extrapolating
19 dose-response relationships and from in vitro to in vivo biological response and mechanistic
20 extrapolations. In addition to alterations in physiochemcial characteristics of PM because of the
21 collection and resuspension processes, these exposure conditions do not simulate the air-cell
April 2002
7-61
DRAFT-DO NOT QUOTE OR CITE
-------
1 interface that actually exists within the lungs, and, thus, the exact dosage delivered to target cells
2 in vivo is not known. Furthermore, unless an in vitro exposure system that is capable of
3 delivering particles uniformly to monolayers of airway epithelial cells cultured in an air-liquid
4 interface system is used (Chen et al., 1993), the conventional incubation system alters the
5 microenvironment surrounding the cells and may alter the mechanisms of cellular injury induced
6 by these agents.
7 Even with these limitations, in vitro studies do provide an approach to identify potential
8 cellular and molecular mechanisms by which PM mediates health effects. These mechanisms
9 can then be evaluated in vivo. In vitro studies are summarized in Table 7-10.
10
11 7.5.2.1 Ambient Particles
12 Several studies have exposed airway epithelial cells, alveolar macrophages, or blood
13 monocytes and erythrocytes to aqueous extracts of ambient PM to investigate cellular processes
14 such as oxidant generation and cytokine production that may contribute to the pathophysiological
15 response seen in vivo. Among the ambient PM being examined were samples collected from
16 Boston, MA, (Goldsmith et al., 1998); North Provo, UT (Ohio et al., 1999a,b); St. Louis, MO
17 (SRM 1648, Dong et al., 1996; Becker and Soukup, 1998); Washington, DC (SRM 1649, Becker
18 and Soukup, 1998); Ottawa, Canada (EHC-93, Becker and Soukup, 1998); Dusseldorf and
19 Duisburg, Germany (Hitzfeld et al., 1997), Mexico City (Bonner et al., 1998), Terni, Italy
20 (Fabiani et al., 1997); and Rome, Italy (Diociaiuti et al., 2001). In any in vitro studies, however,
21 there is a potential for contamination of ambient PM by biologic material during collection on
22 filters. Endotoxin contamination, in particular, can occur at any time in the manufacture of the
23 filter media or during handling of the filter samples before, during, and after the particle
24 collection process. This potential inadvertent contamination of filter samples can make
25 extrapolation of the study results difficult, although careful handling, characterization, and
26 controls can eliminate these concerns.
27 Because soluble metals of ambient surrogates like ROFA have been associated with
28 biological effect and toxicity, several studies have investigated whether the soluble components
29 of ambient PM may have the same biological activities. Extracts of ambient PM samples
30 collected from North Provo, UT, (during 1981 and 1982) were used to test whether the soluble
31 components or ionizable metals, which accounted for approximately 0.1% of the mass, are
April 2002 7-62 DRAFT-DO NOT QUOTE OR CITE
-------
>
to
o
o
to
TABLE 7-10. IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE MATTER CONSTITUENTS
Species, Cell type,
etc.
Human bronchial
epithelial cells,
asthmatic (ASTH)
nonasthmatic (NONA)
Human bronchial
epithelial cells
(smokers)
Human and
rat AM
Human AM and
blood monocytes
Rat AM
NHBE cells
Particle or Exposure
Constituent Technique
DPM In vitro
DPM In vitro
Four Urban air In vitro exposure,
particles: 2 x 105 cells
ROFA exposed for 2 h
DPM
Volcanic ash
Silica
Urban air In vitro
particles;
St. Louis SRM
1648;
Washington,
DC, SRM 1649;
Ottawa, Canada,
EHC-93
PM10 In vitro
Mexico City
1993; volcanic
ash (MSHA)
ROFA In vitro
Concentration Particle Size Exposure Duration
10-100 //g/mL 0.4 //m 2, 4, 6, 24 h
10-100 Mg/mL 0.4 i/m 24 h
Urban and DPM: Urban particles: 2 h for cytotoxicity,
12,27,111,333, 0.3-0.4 Mm 16-18 h for cytokine
or 1000 //g/niL DPM: 0.3 /an assay;
SiO2 and TiO2: ROFA: 0.5 /an chemiluminescence at
4, 12, 35, or Volcanic ash: 30 minutes
167//g/mL 1.8//m
Fe2O3: 1:1, 3:1; Silica- 05-10 /an
10:1 particles/cell TiO2: <5 /an
ratio Latex: 3.8 /an
33 or 100 Mg/mL 0.2 to 0.7 /an 3, 6, or 18-20 h
1-100 //g/mL
-------
H
6
o
o
H
O
o
HH
H
W
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
to
o
o
to
Species, Cell type,
etc.
Human
erythrocytes;
RAW 264.7 cells
Particle or
Constituent
PM10.2.5; PM25 from
Rome, Italy
Exposure
Technique
In vitro
Concentration
50 ± 45 Mg/m3
31±24,ug/m3
19 ± 20 |/g/m3
Particle Size
PM10
PM25
PM10.2.5
Exposure
Duration
Ih
24 h
Effect of Particles
Oxidative stress on cell membranes is related to PM
surface per volume unit of suspension; small particles
are more effective at decreasing viability and increasing
markers of inflammation.
Reference
Diociaiuti
etal. (2001)
Supercoiled
DNA
PM10 from
Edinburgh, Scotland
Rat AM
UAP
DPM
In vitro 996.2± 181.8 PM10 8h PM10 caused damage to DNA; mediated by hydroxyl
l/g/filter in radicals (inhibited by mannitol) and iron (inhibited by
100 ,uL DEF). Clear supernatant has all of the suspension
activity. Free radical activity is derived either from a
fraction that is not centrifugeable on a bench centrifuge
or that the radical generating system is released into
solution.
In vitro 50 to DPM: 1.1 - 1.3 ^m 2 h exposure; Dose dependent increase in TNF-a, IL-6, CINC, MIP-2
200 ,ug/mL UAP: St Louis, supernatant gene expression by urban particles but not with DPM;
between 1974 and collected 18 h cytokine production were not related to ROS; cytokine
1976 in a baghouse, postexposure production can be inhibited by polymyxin B; LPS was
sieved through detected on UAP but not DPM; endotoxin is
200-mesh (125 ,um) responsible for the cytokine gene expression induced by
UAP in AM..
Donaldson
etal. (1997)
Dong et al.
(1996)
Primary cultures
ofRTE
ROFA
In vitro
1.95 ,um MMAD Analysis at 6 and Particle induced epithelial-cell detachment and lytic
24 h cell injury; alterations in the permeability of the
cultured RTE cell layer; increase in LDH, G-6-PDH,
gluathione reductase, glutathione S-transferase;
mechanism of ROFA-induced RTE cytotoxicity and
pulmonary cellular inflammation involves
the development of an oxidative burden.
Dye et al.
(1997)
Primary cultures
of RTE
Peripheral blood
monocytes
BEAS-2B
ROFA; metal
solutions
Organic extract of
TSP, Italy
Provo PM10 extract
In vitro 5, 10, or
20 ,ug/m2
In vitro 42.5 ^g
extract/m3
(acetone)
In vitro 125, 250,
1 .95 ij,m MMAD Analysis at 6 and
24 h
N/A, collected from 2 h
high-volume
sampler (60 m3/h)
PM10 2 and 24 h
Over 24 h ROFA, V, or Ni + V, but not Fe or Ni,
increased epithelial permeability, decreased cellular
glutathione, cell detachment, and lytic cell injury;
treatment with DMTU inhibited expression of MIP-2
and IL-6 genes.
Superoxide anion generation was inhibited at
a particulate concentration of 0. 17 mg/mL when
stimulated with PMA; 50% increase in LDH;
disintegration of plasma membrane.
Dose-dependent increase in IL-6 and IL-8 produced by
particles collected while the steel mill was in operation;
particles collected during plant closure had the lowest
concentrations of soluble Fe, Cu, And Zn
Dye et al.
(1999)
Fabiani et al.
(1997)
Frampton et al.
(1999)
-------
H
6
o
o
H
O
o
HH
H
W
Species, Cell
type, etc.
Rat AM
NHBE
BEAS-2B
BEAS-2B
respiratory
epithelial cells
BEAS-2B
0X174 RF1
DNA
Hamster AM
Hamster AM
AMs from
female CD rats
Exposure
Particle or Constituent Technique Concentration
ROFA, iron sulfate, In vitro 0.01-1.0 mg/mL
nickel sulfate, vanadyl (0.7 x 106
sulfate cells/mL)
Latex particles with
metal complexed on
the surface
ROFA In vitro 5-200 ^g/mL
ROFA In vitro 100 ,ug/mL
Provo In vitro 500 ,ug/mL
TSP soluble and
insoluble extract
PM10 from Edinburgh, In vitro 3.7 or 7.5 i/g/mL
Scotland
ROFA or CAPs In vitro 0, 25, 50, 100, or
200 //g/mL
CAPs, ROFA, and their In vitro 0-200 mg/mL
water-soluble and
particulate fractions
Vanadyl chloride In vitro 10-1000 //m
sodium metavanadate metavanadate
Exposure
Particle Size Duration Effect of Particles
3.6^mMMAD Up to 400 min Increase chemiluminescence, inhibited by DEF and
hydroxyl radical scavengers; solutions of metal
sulfates and metal-complexed latex particles
similarly elevated chemiluminescence in a dose-
and time-dependent manner.
3.6,um 2 and 24 h mRNA for ferritin did not change; ferritin protein
increase; mRNA for transferrin receptor decreased,
mRNA for lactoferrin increased; transferrin
decreased whereas lactoferrin increased;
deferoxamine alone increased lactoferrin mRNA.
N/A ~ 1 h Lactoferrin binding with PM metal occurred
within 5 min. V and Fe (m), but not Ni, increased
the concentration of lactoferrin receptor.
TSP 24 h Water soluble fraction caused greater release of
IL-than insoluble fraction. The effect was
blocked by deferoxamine and presumably
because of metals (Fe, Cu, Zn, Pb).
PM10 8 h Significant free radical activity on degrading
supercoiled DNA; mainly because of hydroxyl
radicals (inhibited by mannitol); Fe involvement
(DEF-B conferred protection); more Fe3+ was
released compared to Fe2+, especially at pH 4.6
than at 7.2.
CAPs: 30 min Dose-dependent increase in AM oxidant stress with
0.1-2.5 //m incubation, both ROFA and CAP. Increase in particle uptake;
(from Harvard analysis Mac-type SR mediate a substantial proportion of
concentrator) immediately AM binding; particle-associated components (e.g.,
TiO2: 1 fjm following transition metals) are likely to mediate intracellular
oxidant stress and proinflammatory activation.
CAPs = 0. 125 i^m 30 min ROFA and CAPs (water soluble components)
ROFA =1.0 fjm caused increases in DCFH oxidation; CAPs
samples and components showed substantial day-
to-day variability in their oxidant effects; ROFA
increased MIP-2 and TNF-a production in AM
and can be inhibitable by NAC.
N/A 30 min Metavanadate caused increased production of
ROS. The LOEL was 50 ^M.
Reference
Ohio et al.
(1997a)
Ohio et al.
(1998c)
Ohio et al.
(1999b)
Ohio et al.
(1999a)
Gilmour et al.
(1996)
Goldsmith
etal. (1997)
Goldsmith
etal. (1998)
Grabowski
etal. (1999)
-------
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
to
o
o
to
-;J
ON
ON
O
c
•T1
H
6
o
l>
-^H
n
s.**
H
O
w
o
H H
H
W
Species, Cell type,
etc.
Human PMN
Human AM
Primary GPTE
cells
Human Bronchial
Epithelial
(BEAS-2B) cells
Rat AM
Human lung
mucoepidermoid
carcinoma cell line,
NCI-H292
BEAS-2B, airway
epithelial cells
Particle or
Constituent
Aqueous and
organic extracts of
TSP in Dusseldorf
and Duisburg,
Germany
UAP
(#1648, 1649)
Volcanic ash
ROFA
ROFA
DOFA
STL
WDC
OT
MSH
TSP collected in
Provo
ROFA, 10 samples
with differing
metal composition
ROFA
ROFA
Exposure
Technique Concentration
In vitro 0.42-0.78 mg
dust/mL
In vitro 0, 25, 100, or
200 Mg/mL
In vitro 6-25, 12.5, 25,
and 50 Mg/cm3
In vitro TSP filter
samples
(36.5 mg/mL)
agitated in
deionized H2O2
for 96 h,
centrifuged at
1200g for 30 min,
lyophylized and
resuspended in
deionized H2O2 or
saline
In vitro 0 or 50 Mg/mL
In vitro 30 Mg/ml
In vitro 0, 0.5, or 2.0 mg
in 10 mL
Particle Size
Collected by high
volume sampler, 90%
<5 Mm, 50%
-------
to
o
o
to
H
6
o
o
H
O
o
HH
H
W
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
Species, Cell type,
etc.
Male (Wistar) rat
lung macrophages
Human blood
monocytes and
neutrophils (PMN)
Human airway
epithelium-derived
cell lines BEAS-2B
(S6-subclone)
Human airway
epithelium-derived
cell line BEAS 2B
Human airway
epithelium-derived
cell line BEAS
Human airway
epithelium-derived
Particle or
Constituent
Urban dust SRM
1649, TiO2,
quartz
Ambient air
particles, carbon
black, oil fly ash,
coal fly ash
ROFA
ROFA
ROFA
Synthetic ROFA
(soluble Ni, Fe,
and V)
Particle
components As,
Exposure
Technique Concentration Particle Size
In vitro 0-100 //g in 1 mL N/A
In vitro 100 //g in N/A
0.2 mL
In vitro 0, 6, 12, 25, or 1.96 fj.m
50 ,ug/mL
In vitro 2, 20, or 60 //g/cm2 1 .96 ^m
In vitro ROFA: 0-200 ROFA: 1. 96 ^m
Mg/mL Synthetic ROFA: N/A
Synthetic ROFA (soluble)
(lOO^g/mL):
Ni, 64 i/M
Fe, 63 ,uM
V, 370 mM
In vitro 500 ^M of As, F, N/A (soluble)
Cr (III), Cu, V, Zn
Exposure
Duration
18h
40 min.
1 and 24 h
24-h
exposure
Up to 24 h
20 min and
6 and 24 h
Effect of Particles
Cytotoxicity ranking was quartz > SRM 1649 > TiO2,
based on cellular ATP decrease and LDH, acid
phosphatase, and p-glucuronidase release.
ROS generation, measured by LCL increased in PMN,
was correlated with Si, Fe, Mn, Ti, and Co content but
not V, Cr, Ni, and Cu. Deferoxamine, a metal ion-
chelator, and did not affect LCL in PMN, suggesting that
metal ions are not related to the induction of LCL.
Activation of IL-6 gene by NF-KB activation and
binding to specific sequences in promoter of IL-6 gene;
inhibition of NF-KB activation by DEF and NAC;
increase in PGE2, IL-6, TNF, and IL-8; activation
NF-B may be a critical first step in the inflammatory
cascade following exposure to ROFA particles.
Epithelial cells exposed to ROFA for 24 h secreted
substantially increased amounts of the PHS products
prostaglandins E2 and F2a; ROFA-induced increase in
prostaglandin synthesis was correlated with a marked
increase in PHS activity.
Tyrosine phosphatase activity, which was known to
be inhibited by vanadium ions, was markedly diminished
after ROFA treatment; ROFA exposure induces
vanadium ion-mediated inhibition of tyrosine
phosphatase activity, leading to accumulation of protein
phosphotyrosines in BEAS cells.
Noncytotoxic concentrations of As, V, and Zn induced
a rapid phosphorylation of MAPK in BEAS cells;
Reference
Nadeau et al.
(1996)
Prahalad et al.
(1999)
Quay etal. (1998)
Samet et al.
(1996)
Samet et al.
(1997)
Samet et al.
(1998)
cell lines BEAS-2B Cr, Cu, Fe, Ni, V,
and Zn
A549
OX174RFIDNA
Urban particles:
SRM 1648,
St. Louis
SRM 1649,
Washington, DC
In vitro Img/mLforFe SRM 1648:
mobilization assay 50% < 10 ,um
SRM 1649:
30% < 10//m
Up to 25 h
activity assays confirmed marked activation of ERK,
JNK, and P38 in BEAS cells exposed to As, V, and
Zn. Cr and Cu exposure resulted in a relatively small
activation of MAPK, whereas Fe and Ni did not activate
MAPK under these conditions; the transcription factors
c-Jun and ATF-2, substrates of JNK and P38,
respectively, were markedly phosphorylated in BEAS
cells treated with As, Cr, Cu, V, and Zn; acute exposure
to As, V, or Zn that activated MAPK was sufficient to
induce a subsequent increase in IL-8 protein expression
in BEAS cells.
Single-strand breaks in DNA were induced by PM
only in the presence of ascorbate, and correlated with
amount of Fe that can be mobilized; ferritin in A549
cells was increased with treatment of PM suggesting
mobilization of Fe in the cultured cells.
Smith and Aust
(1997)
-------
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
to
o
o
to
^1
ON
oo
o
§>
-n
H
6
O
H
O
c|
o
3
o
^
o
H
W
Species, Cell type,
etc.*
Human AMs
Human AMs
Rat (Wistar) AM
RAM cells
(a rat AM cell line)
A549
A549
RLE-6TN cells
(type II like cell
line)
Rat, Long Evans
epithelial cells
BEAS-2B human
bronchial epithelial
cells
NHBE
BEAS-2B
Tell tunes- RTF = R
Particle or
Constituent*
Provo PM10
extract
Chapel Hill PM
extract; both H20
soluble(s) and
insoluble(is)
TiO2
ROFA, a-quartz,
TiO2
TiO2, Fe2O3,
CAP, and the
fibrogenic
particle a-quartz
PM2 5, Burlington,
VT;'
Fine/ultrafme
TiO2
CFA
PFA
a-quartz.
ROFA
Birmingham, AL.
188mg/gofVO
Provo PM10
extract
attrar.liHal Hnitlinlial r.H
Exposure
Technique Concentration Particle Size
In vitro 500 Mg PM10
In vitro 100 Mg/mL PM25
PM10.2.5
In vitro 20, 50, or N/A
80 Mg/mL
In vitro 1 mg/mL N/A
In vitro TiO2 [40 Mg/mL], N/A
Fe2O3 [100
Mg/mL], a-quartz
[200 Mg/mL], or
CAP [40 Mg/mL]
In vitro 1, 2.5, 5, or PM25: 39 nm
10 Mg/mL Fine TiO2: 159 nm
UF TiO2: 37 nm
2.6 Mm
17.7 Mm
2.5 Mm
In vitro 100 Mg/mL N/A
In vitro 50, 100, PM10
200 Mg/mL
!!«• ("rPTF = ("riiinHa mo trar.liRal Hnitlinlial r.nlls
Exposure
Duration Effect of Particles
24 h AM phagocytosis of (FITC)-labeled Saccharomyces
cerevisiae inhibited 30% by particles collected
before steel mill closure.
24 h Increased cytokine production (IL-6, TNFa,
MCP-1); is PM10 > , PM10 > is PM25; , PM25 was
inactive; endotoxin was partially responsible.
4 h Opsonization of TiO2 with surfactant components
resulted in a modest increase in AM uptake
compared with that of unopsonized TiO2; surfactant
components increase AM phagocytosis of particles.
60 min Exposure of A549 cells to ROFA, a-quartz, but not
TiO2, caused increased IL-8 production in TNF-a
primed cells in a concentration-dependent manner.
24 h TiO2 > Fe2O3 > a-quartz > CAP in particle binding;
binding of particle was found to be calcium-
dependent for TiO2 and Fe2O3, while a-quartz
binding was calcium-independent; scavenger
receptor, mediate particulate binding; a-quartz, but
not TiO2 or CAP, caused a dose-dependent
production of IL-8.
24 and 48 h Increases in c-Jun kinase activity, levels of
exposure phosphorylated c-Jun immunoreactive protein, and
transcriptional activation of activator protein-
1 -dependent gene expression; elevation in number
of cells incorporating 5 '-bromodeoxyuridine.
3 h CFA produced highest level of hydroxyl radicals;
iron content is more important than quartz content.
2-6 h ROFA caused increased intracellular Ca++, IL-6,
IL-and TNF-a through activation of capsicin-
and pH-sensitive receptors.
24 h Dose-dependent increase in expression of IL-8
produced by particles collected when the steel mill
was in operation.
Reference
Soukup et al.
(2000)
Soukup and
Becker (2001)
Stringer and
Kobzik
(1996)
Stringer and
Kobzik (1998)
Stringer et al.
(1996)
Timblin et al.
(1998)
Van Maanen
etal. (1999)
Veronesi et al.
(1999b)
Wu etal. (2001)
"•Particles: See Table 7-1
-------
1 responsible for the biological activity of the extracted PM components. The oxidant generation
2 (thiobarbituric acid reactive products), release of IL-8 from BEAS-2B cells, and PMN influx in
3 rats exposed to these samples correlated with sulfate content and the ionizable concentrations of
4 metals in these PM extracts (Ohio et al., 1999a,b). In addition, these extracts stimulated IL-6 and
5 IL-8 production as well as increased IL-8 mRNA and enhanced expression of intercellular
6 adhesion molecule-1 (ICAM-1) in BEAS-2B cells (Kennedy et al., 1998). Cytokine secretion
7 was preceded by activation of nuclear factor kappa B (NF-KB) and was reduced by treatment
8 with superoxide dismutase (SOD), Deferoxamine (DBF), or N-acetylcysteine. The addition of
9 similar quantities of Cu+2 as found in the Provo extract replicated the biological effects observed
10 with particles alone. When normal constituents of airway lining fluid (mucin or ceruloplasmin)
11 were added to BEAS cells, particulate-induced secretion of IL-8 was modified. Mucin reduced
12 IL-8 secretion; whereas ceruloplasmin significantly increased IL-8 secretion and activation of
13 NF-KB. The authors suggest that copper ions may cause some of the biologic effects of inhaled
14 PM in the Provo region and may provide an explanation for the sensitivity of asthmatics to Provo
15 PM seen in epidemiologic studies.
16 Frampton et al. (1999) examined the effects of the same ambient PM samples collected
17 from Utah Valley in the late 1980s (see Section 7.2.1). Aqueous extracts of the filters were
18 analyzed for metal and oxidant production and added to cultures of human respiratory epithelial
19 cells (BEAS-2B) for 2 or 24 h. Particles collected in 1987, when the steel mill was closed had
20 the lowest concentrations of soluble iron, copper, and zinc and showed the least oxidant
21 generation. Ambient PM collected before and after plant closing induced expression of IL-6 and
22 IL-8 in a dose-response relationship (125, 250, and 500 //g/mL). Ambient PM collected after
23 reopening of the steel mill also caused cytotoxicity, as demonstrated by microscopy and LDH
24 release at the highest concentration used (500 //g/mL).
25 Soukup et al. (2000) used similar ambient PM extracts as Frampton et al. (1999) to
26 examine effects on human alveolar macrophages. The phagocytic activity and oxidative response
27 of AMs was measured after segmental instillation of aqueous extracts from the Utah Valley or
28 after overnight in vitro cell culture. Ambient PM collected before closure of the steel mill
29 inhibited AM phagocytosis of (FITC)-labeled Saccharomyces cerevisiae by 30%; no significant
30 effect on phagocytosis was seen with the other two extracts. Furthermore, although extracts of
31 ambient PM collected before and after plant closure inhibited oxidant activity of AMs when
April 2002 7-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 incubated overnight in cell culture, only the former particles caused an immediate oxidative
2 response in AMs. Host defense effects were attributed to apoptosis which was most evident in
3 particles collected before plant closure. Interpretation of loss of these effects by chelation
4 removal of the metals was complicated by the observed differences in apoptosis despite similar
5 metal contents of ambient PM collected during the steel mill operation.
6 Wu et al (2001) investigated the intracellular signaling mechanisms for the pulmonary
7 responses to Utah Valley PM extracts. Human primary airway epithelial cells were exposed to
8 aqueous extracts of PM collected from the year before, during, and after the steel mill closure in
9 Utah Valley. Transfection with kinase-deficient extracellular signal-regulated kinase (ERK)
10 constructs partially blocked the PM-induced interleukin (IL)-8 promoter reporter activity. The
11 mitogen-activated protein kinase/ERK kinase (MEK) activity inhibitor PD-98059 significantly
12 abolished IL-8 released in response to the PM, as did the epidermal growth factor (EOF) receptor
13 kinase inhibitor AG-1478. Western blotting showed that the PM-induced phosphorylation of
14 EOF receptor tyrosine, MEK1/2, and ERK1/2 could be ablated with AG-1478 or PD-98059. The
15 results indicate that the potency of Utah Valley PM collected during plant closure was lower than
16 that collected while the steel mill was in operation and imply that Utah Valley PM can induce IL-
17 8 expression partially through the activation of the EGF receptor signaling.
18 There are regional as well as daily variations in the composition of ambient PM and, hence,
19 its biological activities. For example, concentrated ambient PM (CAP, from Boston urban air)
20 has substantial day-to-day variability in its composition and oxidant effects (Goldsmith et al.,
21 1998). Similar to Utah PM, the water-soluble component of Boston CAPs significantly
22 increased AM oxidant production and inflammatory cytokine (MIP2 and TNFa) production over
23 negative control values. These effects can be blocked by metal chelators or antioxidants. The
24 regional difference in biological activity of ambient PM has been shown by Becker and Soukup
25 (1998). The oxidant generation, phagocytosis, as well as the expressions of receptors important
26 for phagocytosis in human alveolar macrophage and blood monocyte were reduced significantly
27 by PM exposure.
28 Becker and Soukup (1998) and others (Dong et al., 1996, Becker et al., 1996) have
29 suggested that the biological activity of the ambient PM may result from the presence of
30 endotoxin on the particles rather than metal-associated oxidant generation. Using the same urban
31 particles (SRM 1648), cytokine production (TNF-a, IL-1,11-6, CINC, and MIP-2) was increased
April 2002 7-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 in macrophages following treatment with 50 to 200 //g/mL of urban PM (Dong et al., 1996). The
2 urban particle-induced TNF-a secretion was abrogated completely by treatment with polymyxin
3 B, an antibiotic that blocks LPS-associated activities, but not with antioxidants.
4 The involvement of endotoxin, at least partially, in PM induced biological effects was
5 supported more recently by Bonner et al. (1998) and Soukup and Becker (2001). Urban PM10
6 collected from north, south, and central regions of Mexico City was used with SD rat AM to
7 examine PM effects on platelet-derived growth factor (PDGF) receptors on lung myofibroblasts
8 (Bonner et al., 1998). Mexico City PM10 (but not volcanic ash) stimulated secretion of
9 upregulatory factors for the PDGF a receptor, possibly via IL-1 p. In the presence of an
10 endotoxin-neutralizing protein, the Mexico City PM10 effect on PDGF was blocked partially,
11 suggesting that LPS was responsible partially for the effect of the PM10 on macrophages.
12 In addition, both LPS and vanadium (both present in the PM10) acted directly on lung
13 myofibroblasts. However, the V levels in Mexico City PM10 were probably not high enough to
14 exert an independent effect. The authors concluded that PM10 exposure could lead to airway
15 remodeling by enhancing myofibroblast replication and chemotaxis.
16 Soukup and Becker (2001) collected fresh PM25 and PM10_2 5 from the ambient air of
17 Chapel Hill, NC, and compared the activity of these two particle size fractions. Both water
18 soluble and insoluble components were assessed for cytokine production, inhibition of
19 phagocytosis, and induction of apoptosis. The most potent fraction was the insoluble PM10_2 5.
20 Endotoxin was responsible for much of the cytokine production, while inhibition of phagocytosis
21 was induced by other moieties in the coarse material. None of the activities were inhibited by the
22 metal chelator deferoxamine.
23 The effects of water soluble as well as organic components (extracted in dichloromethane)
24 of ambient PM were investigated by exposing human PMN to PM extracts (Hitzfeld et al., 1997).
25 PM was collected with high-volume samplers in two German cities, Dusseldorf and Duisburg;
26 these sites have high traffic and high industrial emissions, respectively. Organic, but not
27 aqueous, extracts of PM alone significantly stimulated the production and release of ROS in
28 resting human PMN. The effects of the PM extracts were inhibited by SOD, catalase, and
29 sodium azide (NaN3). Similarly, the organic fraction (extractable by acetone) of ambient PM
30 from Terni, Italy, had been shown to produce cytotoxicity, superoxide release in response to
31 PMA and zymosan in peripheral monocytes (Fabiani et al., 1997).
April 2002 7-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 Diociaiuti et al. (2001) compared the in vitro toxicity of coarse (PM10_25) and fine (PM25)
2 particulate matter, collected in an urban area of Rome. The in vitro toxicity assays used included
3 human red blood cell hemolysis, cell viability, and nitric oxide (NO) release in the RAW 264.7
4 macrophage cell line. There was a dose-dependent hemolysis in human erythrocytes when they
5 were incubated with fine and coarse particles. The hemolytic potential was greater for the fine
6 particles than for the coarse particles in equal mass concentration. However, when data were
7 expressed in terms of PM surface area per volume of suspension, the hemolytic activity of the
8 fine fraction was equal to the coarse fraction. This result suggested that the oxidative stress
9 induced by PM on the cell membranes could be due mainly to the interaction between the particle
10 surfaces and the cell membranes. Although RAW 264.7 cells challenged with fine and coarse
11 particles showed decreased viability and an increased release of NO, a key inflammatory
12 mediator, both effects were not dose-dependent in the tested concentration range. The fine
13 particles were the most effective in inducing these effects when the data were expressed as mass
14 concentration or as surface area per unit volume. The authors concluded that these differences in
15 biological activity were due to the different physicochemical natures of the particles.
16
17 7.5.2.2 Comparison of Ambient and Combustion-Related Surrogate Particles
18 In vitro toxicology studies utilizing alveolar macrophages as target cells (Imrich et al.,
19 2000; Long et al., 2001; Ning et al., 2000; Mukae et al., 2000, 2001; van Eeden et al., 2001) have
20 found that urban air particles are much more potent for inducing cellular responses than
21 individual surrogate combustion particles such as diesel and ROFA. Similar to the results
22 described above in Section 7.5.2.1, these studies also show that when cytokine responses are
23 measured, LPS/endotoxin is found to be responsible for most of the activity. Metals, on the other
24 hand, do not seem to affect cytokine production, as confirmed by studies showing that ROFA
25 does not induce macrophage cytokine production. These results are important because LPS is an
26 important component associated with both coarse and fine mode particles (Menetrez et al., 2001).
27 In fact, in one study (Long et al., 2001), cytokine responses in the alveolar macrophages were
28 correlated with LPS content and more LPS was found associated with indoor PM2 5 than outdoor
29 PM25.
30 Imrich et al., (2000) found that when mice alveolar macrophages were stimulated with
31 CAPs (PM2 5), the resulting TNF responses could be inhibited by the use of an endotoxin
April 2002 7-72 DRAFT-DO NOT QUOTE OR CITE
-------
1 neutralizing agent [e.g., polymyxin-B (PB)]. Because the MIP-2 response (IL-8) was only partly
2 inhibited by PB; however, the authors concluded that endotoxin primed cells to respond to other
3 particle components. In a related study (Ning et al., 2000), the use of PB showed that particle-
4 absorbed endotoxin in CAPs suspensions caused activation of normal (control) AMs, while other
5 (nonendotoxin) components were predominantly responsible for the enhanced cytokine release
6 observed by primed AMs incubated with CAPs. The non-LPS component was not identified in
7 this study, however, the AM biological response did not correlate with any of a panel of elements
8 quantified within the insoluble CAPs samples (e.g., Al, Cd, Cr, Cu, Fe, Mg, Mn, Ni, S, Ti, V).
9 Van Eeden et al. (2001) compared ROFA, the atmospheric dust sample EHC-93, and
10 different size latex particles for cytokine induction on human alveolar macrophages. The
11 EHC-93 particles produced greater than 8-fold induction of various cytokines, including IL-1,
12 TNF, GMCSF; the other particles induced these cytokines approximately 2-fold. Using the same
13 EHC-93 particles, Mukae et al. (2000, 2001) found that inhalation exposure stimulated bone
14 marrow band cell-granulocyte precursor production. They also found that the magnitude of the
15 response was correlated with the amount of phagocytosis of the particles by alveolar
16 macrophages. These results may indicate that macrophages produce factors which stimulate
17 bone marrow, including IL-6 and GMCSF. In fact, alveolar macrophages exposed in vitro to
18 these particles released cytokines; and when the supernatant of PM-stimulated macrophages was
19 instilled into rabbits, the bone marrow was stimulated.
20 In a series of studies using the same ROFA samples, several in vitro experiments have
21 investigated the biochemical and molecular mechanisms involved in ROFA induced cellular
22 injury. Prostaglandin metabolism in cultured human airway epithelial cells (BEAS-2B and
23 NHBE) exposed to ROFA was investigated by Samet et al. (1996). Epithelial cells exposed to
24 ROFA for 24 h secreted substantially increased amounts of prostaglandins E2 and F2 a. The
25 ROFA-induced increase in prostaglandin synthesis was correlated with a marked increase in
26 activity of the PHS-2 form of prostaglandin H synthase as well as mRNA coded for this enzyme.
27 In contrast, expression of the PHS1 form of the enzyme was not affected by ROFA treatment of
28 airway epithelial cells. These investigators further demonstrated that noncytotoxic levels of
29 ROFA induced a significant dose- and time-dependent increase in protein tyrosine phosphate, an
30 important index of signal transduction activation leading to a broad spectrum of cellular
31 responses. ROFA-induced increases in protein phosphotyrosines were associated with its soluble
April 2002 7-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 fraction and were mimicked by V-containing solutions but not iron or nickel solutions (Samet
2 etal., 1997).
3 ROFA also stimulates respiratory cells to secrete inflammatory cytokines such as IL-6,
4 IL-8, and TNF. Normal human bronchial epithelial (NHBE) cells exposed to ROFA produced
5 significant amounts of IL-8, IL-6, and TNF, as well as mRNAs coding for these cytokines (Carter
6 etal., 1997). Increases in cytokine production were dose-dependent. The cytokine production
7 was inhibited by the addition of metal chelator, DBF, or the free radical scavenger
8 dimethylthiourea (DMTU). Similar to the data of Samet et al. (1997), V but not Fe or Ni
9 compounds were responsible for these effects. Cytotoxicity, decreased cellular glutathione levels
10 in primary cultures of rat tracheal epithelial (RTE) cells exposed to suspensions of ROFA
11 indicated that respiratory cells exposed to ROFA were under oxidative stress. Treatment with
12 buthionine sulfoxamine (an inhibitor of y-glutamyl cysteine synthetase) augmented ROFA-
13 induced cytotoxicity; whereas treatment with DMTU inhibited ROFA-induced cytoxicity further
14 suggested that ROFA-induced cell injury may be mediated by hydroxyl-radical-like reactive
15 oxygen species (ROS) (Dye et al., 1997). Using BEAS-2B cells, a time- and dose-dependent
16 increase in IL-6 mRNA induced by ROFA was shown to be preceded by the activation of nuclear
17 proteins, for example, nuclear factor-KB (NF-icB) (Quay et al., 1998). Taken together, ROFA
18 exposure increases oxidative stress, perturbs protein tyrosine phosphate homeostasis, activates
19 NF-KB, and up-regulates inflammatory cytokine and prostaglandin synthesis and secretion to
20 produce lung injury.
21 Stringer and Kobzik (1998) observed that "primed" lung epithelial cells exhibited enhanced
22 cytokine responses to PM. Compared to normal cells, exposure of tumor necrosis factor (TNF)-
23 a-primed A549 cells to ROFA or a -quartz caused increased IL-8 production in a concentration-
24 dependent manner for particle concentrations ranging from 0-200 //g/mL. Addition of the
25 antioxidant N-acetylcysteine (NAC) (1.0 mM) decreased ROFA and a -quartz-mediated IL-8
26 production by approximately 50% in both normal and TNF-a-primed A549 cells. Exposure of
27 A549 cells to ROFA caused an increase in oxidant levels that could be inhibited by NAC. These
28 data suggest that (1) lung epithelial cells primed by inflammatory mediators show increased
29 cytokine production after exposure to PM and (2) oxidant stress is an important mechanism for
30 this response.
April 2002 7-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 In summary, exposure of lung epithelial cells to ambient PM or ROFA leads to increased
2 production of cytokines and the effects may be mediated, at least in part, through production of
3 ROS. Day-to-day variations in the components of PM, such as soluble transition metals, which
4 may be critical to eliciting the response, are suggested. The involvement of organic components
5 in ambient PM also was suggested in some studies.
6
7 7.5.3 Potential Cellular and Molecular Mechanisms
8 7.5.3.1 Reactive Oxygen Species
9 Ambient particulate matter contains transition metals, such as iron (most abundant),
10 copper, nickel, vanadium, and cobalt. These metals are capable of catalyzing the one-electron
11 reductions of molecular oxygen necessary to generate reactive oxygen species (ROS). These
12 reactions can be demonstrated by the iron-catalyzed Haber-Weiss reactions that follow.
13
14 Reductant11 + Fe(III) -> Reductantn+1 + Fe(II) (1)
15 Fe(II) + O2 -» Fe(III) + O2 (2)
16 HO2+O2+H+^ O2+H2O2 (3)
17 Fe(II) + H2O2 -» Fe(III)+*OH + HO"(FentonReaction) (4)
18
19 Iron will continue to participate in the redox cycle in the above reactions as long as there is
20 sufficient O2 or H2O2 and reductants.
21 Soluble metals from inhaled PM dissolved into the fluid lining of the airway lumen can
22 react directly with biological molecules (acting as a reductant in the above reactions) to produce
23 ROS. For example, ascorbic acid in the human lung epithelial lining fluid can react with Fe(in)
24 from inhaled PM to cause single strand breaks in supercoiled plasmid DNA, cj)Xl74 RFI (Smith
25 and Aust, 1997). The DNA damage caused by a PM10 suspension can be inhibited by mannitol,
26 an hydroxyl radical scavenger, further confirming the involvement of free radicals in these
27 reactions (Gilmour et al., 1996; Donaldson et al., 1997; Li et al., 1997). Because the clear
28 supernatant of the centrifuged PM10 suspension contained all of the suspension activity, the free
29 radical activity is derived either from a fraction that is not centrifugable (10 min at 13,000 rpm
April 2002 7-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 on a bench centrifuge) or the radical generating system is released into solution (Gilmour et al.,
2 1996; Donaldson et al., 1997; Li et al., 1997).
3 In addition to measuring the interactions of ROS and biomolecules directly, the role of
4 ROS in PM-induced lung injury also can be assessed by measuring the electron spin resonance
5 (ESR) spectrum of radical adducts or fluorescent intensity of dichlorofluorescin (DCFH), an
6 intracellular dye that fluoresces on oxidation by ROS. Alternatively, ROS can be inhibited using
7 free radical scavengers, such as dimethylthiourea (DMTU); antioxidants, such as glutathione or
8 N-acetylcysteine (NAC); or antioxidant enzymes, such as superoxide dismutase (SOD). The
9 diminished response to PM after treatment with these antioxidants may indicate the involvement
10 of ROS; however, some antioxidants (e.g., thiol-containing) can interact with metal ions directly.
11 As described earlier, Kadiiska et al. (1997) used the ESR spectra of 4-POBN [a-(4-pyridyl
12 l-oxide)-N-tert-butylnitrone] adducts to measure ROS in rats instilled with ROFA and
13 demonstrated the association between ROS production within the lung and soluble metals in
14 ROFA. Using DMTU to inhibit ROS production, Dye et al. (1997) had shown that systemic
15 administration of DMTU impeded development of the cellular inflammatory response to ROFA,
16 but did not ameliorate biochemical alterations in BAL fluid. Goldsmith et al. (1998), as
17 described earlier, showed that ROFA and CAPs caused increases in ROS production in AMs.
18 The water-soluble component of both CAPs and ROFA significantly increased AM oxidant
19 production over negative control values. In addition, increased PM-induced cytokine production
20 was inhibited by NAC. Li et al. (1996, 1997) instilled rats with PM10 particles (collected on
21 filters from an Edinburgh, Scotland, monitoring station). Six hours after intratracheal instillation
22 of PM10, they observed a decrease in glutathione (GSH) levels in the BAL fluid. Although this
23 study does not describe the composition of the PM10, the authors suggest that changes in GSH, an
24 important lung antioxidant, support the contention that the free radical activity of PM10 is
25 responsible for its biological activity in vivo.
26 In addition to ROS generated directly by PM, resident or newly recruited AMs or PMNs
27 also are capable of producing these reactive species on stimulation. The ROS produced during
28 the oxidative burst can be measured using a chemiluminescence (CL) assay. With this assay,
29 AM CL signals in vitro have been shown to be greatest with ROFA containing primarily soluble
30 V and were less with ROFA containing Ni plus V (Kodavanti et al., 1998a). As described
31 earlier, exposures to Dusseldorf and Duisburg PM increased the resting ROS production in
April 2002 7-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 PMNs, which could be inhibited by SOD, catalase, and sodium azide (Hitzfeld et al., 1997).
2 Stringer and Kobzik (1998) showed that addition of NAC (1.0 mM) decreased ROFA-mediated
3 IL-8 production by approximately 50% in normal and TNF-a-primed A549 cells. In addition,
4 exposures of A549 cells to ROFA caused a substantial (and NAC inhibitable) increase in oxidant
5 levels as measured by DCFH oxidation. In human AMs, Becker et al. (1996) found a CL
6 response for ROFA, but not urban air particles (Ottawa and Dusseldorf) or volcanic ash.
7 Metal compounds of PM are the most probable species capable of catalyzing ROS
8 generation on exposure to PM. To determine elemental content and solubility in relation to their
9 ability to generate ROS, PMN or monocytes were exposed to a wide range of ambient air
10 particles from divergent sources (one natural dust, two types of oil fly ash, two types of coal fly
11 ash, five different ambient air samples, and one carbon black sample) (Prahalad et al., 1999), and
12 CL production was measured over a 20-min period postexposure. Percent of sample mass
13 accounted for by XRF detectable elements was 1.2% (carbon black); 22 to 29% (natural dust and
14 ambient air particles); 13 to 22% (oil fly ash particles); and 28 to 49% (coal fly ash particles).
15 The maj or proportion of elements in most of these particles were aluminosilicates and insoluble
16 iron, except oil derived fly ash particles in which soluble vanadium and nickel were in highest
17 concentration, consistent with particle acidity as measured in the supernatants. All particles
18 induced CL response in cells, except carbon black. The CL response of PMNs in general
19 increased with all washed particles, with oil fly ash and one urban air particle showing statistical
20 differences between deionized water washed and unwashed particles. These CL activities were
21 significantly correlated with the insoluble Si, Fe, Mn, Ti, and Co content of the particles.
22 No relationship was found between CL and soluble transition metals such as V, Cr, Ni, and Cu.
23 Pretreatment of the particles with a metal ion chelator, deferoxamine, did not affect CL activities.
24 Particle sulfate content and acidity of the particle suspension did not correlate with CL activity.
25 Soluble metals can be mobilized into the epithelial cells or AMs to produce ROS
26 intracellularly. Size fractionated coal fly ash particles (2.5, 2.5 to 10, and <10 //m) of bituminous
27 b (Utah coal), c (Illinois coal), and lignite (Dakota coal) were used to compare the amount of iron
28 mobilization in A549 cells and by citrate (1 mM) in cell-free suspensions (Smith et al., 1998).
29 Iron was mobilized by citrate from all three size fractions of all three coal types. More iron, in
30 Fe(ni) form, was mobilized by citrate from the <2.5-//m fraction than from the >2.5-//m
31 fractions. In addition, the amount of iron mobilized was dependent on the type of coal used to
April 2002 7-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 generate the fly ash (Utah coal > Illinois coal = Dakota coal) but not related to the total amount
2 of iron present in the particles. Ferritin (an iron storage protein) levels in A549 cells increased by
3 as much as 12-fold in cells treated with coal fly ash (Utah coal > Illinois coal > Dakota coal).
4 More ferritin was induced in cells treated with the <2.5-//m fraction than with the >2.5-//m
5 fractions. Mossbauer spectroscopy of a fly ash sample showed that the bioavailable iron was
6 assocated with the glassy aluminosilicate fraction of the particles (Ball et al., 2000). As with the
7 bioavailability of iron, there was an inverse correlation between the production of IL-8 and fly
8 ash particle size with the Utah coal fly ash being the most potent.
9 Using ROFA and colloidal iron oxide, Ohio et al. (1997b; 1998a,b,c; 1999c; 2000b) have
10 shown that exposures to these particles disrupted iron homeostasis and induced the production of
11 ROS in vivo and in vitro. Treatment of animals or cells with metal-chelating agents such as DEF
12 with an associated decrease in response has been used to infer the involvement of metal in PM-
13 induced lung injury. Metal chelation by DEF (1 mM) caused significant inhibition of particulate-
14 induced AM oxidant production, as measured using DCFH (Goldsmith et al., 1998). DEF
15 treatment also reduced NF-KB activation and cytokine secretion in a human bronchial epithelial
16 cell line (BEAS-2B cells) exposed to Provo PM (Kennedy et al., 1998). However, treatment of
17 ROFA suspension with DEF was not effective in blocking teachable metal induced acute lung
18 injury (Dreher et al., 1997). Dreher et al. (1997) indicated that DEF could chelate Fe(HI) and
19 V(n), but not Ni(n), suggesting that metal interactions played a significant role in ROFA-induced
20 lung injury.
21 Other than Fe, several V compounds have been shown to increase mRNA levels for
22 selected cytokines in BAL cells and also to induce pulmonary inflammation (Pierce et al., 1996).
23 NaVO3 and VOSO4, highly soluble forms of V, tended to induce pulmonary inflammation and
24 inflammatory cytokine mRNA expression more rapidly and more intensely than the less soluble
25 form, V2O5, in rats. Neutrophil influx was greatest following exposure to VOSO4 and lowest
26 following exposure to V2O5. However, metal components of fly ash have not been shown to
27 consistently increase ROS production from bovine AM treated with combustion particles
28 (Schluter et al., 1995). For example, As(ni), Ni(n), and Ce(in), which are major components of
29 fly ash, had been shown to inhibit the secretion of superoxide anions (O2") and hydrogen
30 peroxide. In the same study, O2" were lowered by Mn(II) and Fe(n); whereas V(IV) increased O2"
31 and H2O2. In contrast, Fe(in) increase O2" productions, demonstrating that the oxidation state of
April 2002 7-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 metal may influence its oxidant generating properties. Other components of fly ash, such as
2 Cd(II), Cr(HI), and V(V), had no effects on ROS.
3 It is likely that a combination of several metals rather than a single metal in PM is
4 responsible for the PM induced cellular response. For example, V and Ni+V but not Fe or Ni
5 alone (in saline with the final pH at 3.0) resulted in increased epithelial permeability, decreased
6 cellular glutathione, cell detachment, and lytic cell injury in rat tracheal epithelial cells exposed
7 to soluble salts of these metals at equivalent concentrations found in ROFA (Dye et al., 1999).
8 Treatment of V-exposed cells with buthionine sulfoximine further increased cytotoxicity.
9 Conversely, treatment with radical scavenger dimethyl thiourea inhibited the effects in a dose-
10 dependent manner. These results suggest that soluble metal or combinations of several metals in
11 ROFA may be responsible for these effects.
12 Similar to combustion particles such as ROFA, the biological response to exposure to
13 ambient PM also may be influenced by the metal content of the particles. Human subjects were
14 instilled with 500 //g (in 20 mL sterile saline) of Utah Valley dust (UVD1, 2, 3, collected during
15 3 successive years) on the left segmental bronchus and on the right side with sterile saline as
16 control. Twenty-four-hours post-instillation, a second bronchoscopy was performed and
17 phagocytic cells were obtained on both sides of the segmental bronchus. AM from subjects
18 instilled with UVD, obtained by bronchoaveolar lavage 24 h post-instillation, were incubated
19 with fluoresceinated yeast (Saccharomyces cerevisiae) to assess their phagocytic ability.
20 Although the same proportion of AMs were exposed to UVD phagocytized yeast, AMs exposed
21 to UVD1, which were collected while a local steel mill was open, took up significantly less
22 particles than AMs exposed to other extracts (UVD2 when the steel mill was closed and UVD3
23 when the plant reopened). AMs exposed to UVD1 also exhibited a small decrease in oxidant
24 activity (using dihydrorhodamine-123, DHR). AMs from healthy volunteers were incubated in
25 vitro with the various UVD extracts to assess whether similar effects on human AMs function
26 could be observed to those seen following in vivo exposure. The percentage of AMs that
27 engulfed yeast particles was significantly decreased by exposure to UVD1 at 100 //g/mL, but not
28 at 25 //g/mL. However, the amount of particles engulfed was the same following exposure to all
29 three UVD extracts. AMs also demonstrated increased oxidant stress (using chemiluminescence)
30 after in vitro exposure to UVD1, and this effect was not abolished with pretreatment of the
31 extract with the metal chelator deferoxamine. As with the AMs exposed to UVD in vivo, AM
April 2002 7-79 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposed to UVD in vitro had a decreased oxidant activity (DHR assay). UVD1 contains 61 times
2 and 2 times the amount of Zn compared to UVD 2 and UVD3, respectively; whereas UVD3
3 contained 5 times more Fe than UVD1. Ni and V were present only in trace amounts. Using
4 similarly extracted samples, Frampton et al. (1999) exposed BEAS-2B cells for 2 and 24 h.
5 Similar results were observed for oxidant generation in these cells (i.e., UVD 2, which contains
6 the lowest concentrations of soluble iron, copper, and zinc, produced the least response). Only
7 UVD 3 produced cytotoxicity at a dose of 500 //g/mL. UVD 1 and 3, but not 2, induced
8 expression of IL-6 and 8 in a dose-dependent fashion. Taken together, these data showed that the
9 biological response to ambient particle extracts is heavily dependent on the source and; hence,
10 the chemical composition of PM.
11
12 7.5.3.2 Intracellular Signaling Mechanisms
13 In has been shown that the intracellular redox state of the cell modulates the activity of
14 several transcription factors, including NF-KB, a critical step in the induction of a variety of
15 proinflammatory cytokine and adhesion-molecule genes. NF-KB is a heterodimeric protein
16 complex that in most cells resides in an inactive state in the cell cytoplasm by binding to
17 inhibitory kappa B alpha (IicBa). On appropriate stimulation by cytokines or ROS, IicBa is
18 phosphorylated and subsequently degraded by proteolysis. The dissociation of IicBa from NF-KB
19 allows the latter to translocate into the nucleus and bind to appropriate sites in the DNA to
20 initiate transcription of various genes. Two studies in vitro have shown the involvement of
21 NF-KB in particulate-induced cytokine and intercellular adhesion molecule-1 (ICAM-1)
22 production in human airway epithelial cells (BEAS-2B) (Quay et al., 1998; Kennedy et al.,
23 1998). Cytokine secretion was preceded by activation of NF-KB and was reduced by treatment
24 with antioxidants or metal chelators. These results suggest that metal-induced oxidative stress
25 may play a significant role in the initiation phase of the inflammatory cascade following
26 particulate exposure.
27 A second well-characterized human transcription factor, AP-1, also responds to the
28 intracellular ROS concentration. AP-1 exists in two forms, either in a homodimer of c-jun
29 protein or a heterodimer consisting of c-jun and c-fos. Small amounts of AP-1 already exist in
30 the cytoplasm in an inactive form, mainly as phosphorylated c-jun homodimer. Many different
31 oxidative stress-inducing stimuli, such as UV light and IL-1, can activate AP-1. Exposure of rat
April 2002 7-80 DRAFT-DO NOT QUOTE OR CITE
-------
1 lung epithelial cells to ambient PM in vitro resulted in increases in c-jun kinase activity, levels of
2 phosphorylated c-jun immunoreactive protein, and transcriptional activation of AP-1-dependent
3 gene expression (Timblin et al., 1998). This study demonstrated that interaction of ambient
4 particles with lung epithelial cells initiates a cell signaling cascade related to aberrant cell
5 proliferation.
6 Early response gene transactivation has been linked to the development of apoptosis, a
7 unique type of programmed cell injury and a potential mechanism to account for PM-induced
8 changes in cellular response. Apoptosis of human AMs exposed to ROFA (25 //g/mL) or urban
9 PM was observed by Holian et al. (1998). In addition, both ROFA and urban PM upregulated the
10 expression of the RFD1+ AM phenotype; whereas only ROFA decreased the RFDl+7+ phenotype.
11 It has been suggested that an increase in the AM phenotype ratio of RFDl+/RFDl+7+ may be
12 related to disease progression in patients with inflammatory diseases. These data showed that
13 ROFA and urban PM can induce apoptosis of human AMs and increase the ratio of AM
14 phenotypes toward a higher immune active state and may contribute to or exacerbate lung
15 inflammation.
16 Somatosensory neurons located in the dorsal root ganglia (DRG), innervate the upper
17 thoracic region of the airways and extend their terminals under and between the epithelial lining
18 of the lumen. Given this anatomical proximity, the sensory fibers and their tracheal epithelial
19 targets are the first resident cells to encounter inhaled pollutants, such as PM. The differential
20 response of these cell types to PM derived from various sources (i.e., industrial, residential,
21 volcanic) was examined with biophysical and immunological endpoints (Veronesi et al., 2002a).
22 Although the majority of PM tested stimulated IL-6 release in both BEAS-2B epithelial cells and
23 DRG neurons in a receptor-mediated fashion, the degree of these responses was markedly higher
24 in sensory neurons. Epithelial cells are damaged or denuded in many common health disorders
25 (e.g., asthma, viral infections), allowing PM particles to directly encounter the sensory terminals
26 and their acid sensitive receptors. This differential sensitivity of target cells to PM suggests that
27 non-genetic factors (i.e., cell-cell interactions) may also affect the inflammatory response to PM
28 in individuals whose epithelial lining is damaged.
29 Another intracellular signaling pathway that can lead to diverse cellular responses such as
30 cell growth, differentiation, proliferation, apoptosis, and stress responses to environmental
31 stimuli, is the phosphorylation-dependent, mitogen-activated protein kinase (MAPK).
April 2002 7-81 DRAFT-DO NOT QUOTE OR CITE
-------
1 Noncytotoxic levels of ROFA have been shown to induce significant dose- and time-dependent
2 increases in protein tyrosine phosphate levels in BEAS cells (Samet et al., 1997). In a
3 subsequent study, the effects of As, Cr, Cu, Fe, Ni, V, and Zn on the MAPK, extracellular
4 receptor kinase (ERK), c-jun N-terminal kinase (INK), and P38 in BEAS cells were investigated
5 (Samet et al., 1998). Noncytotoxic concentrations of As, V, and Zn induced a rapid
6 phosphorylation of MAPK in BEAS cells. Activity assays confirmed marked activation of ERK,
7 INK, and P38 in BEAS cells exposed to As, V, and Zn. Cr and Cu exposure resulted in a
8 relatively small activation of MAPK; whereas Fe and Ni did not activate MAPK. Similarly, the
9 transcription factors c-Jun and ATF-2, substrates of INK and P38, respectively, were markedly
10 phosphorylated in BEAS cells treated with As, Cr, Cu, V, and Zn. The same acute exposure to
11 As, V, or Zn that activated MAPK was sufficient to induce a subsequent increase in IL-8 protein
12 expression in BEAS cells. These data suggest that MAPK may mediate metal-induced
13 expression of inflammatory proteins in human bronchial epithelial cells. The ability of ROFA to
14 induce activation of MAPKs in vivo was demonstrated by Silbajoris et al. (2000) (see Table 7-3).
15 In addition, Gercken et al. (1996) showed that the ROS production induced by PM was markedly
16 decreased by the inhibition of protein kinase C as well as phospholipase A2.
17 The major cellular response downstream of ROS and the cell signaling pathways described
18 above is the production of inflammatory cytokines or other reactive mediators. In an effort to
19 determine the contribution of cyclooxygenase to the pulmonary responses to ROFA exposure
20 in vivo, Samet et al. (2000) intratracheally instilled Sprague-Dawley rats with ROFA (200 or
21 500 ^g in 0.5 mL saline). These animals were pretreated ip with 1 mg/kg NS398, a specific
22 prostaglandin H synthase 2 (COX2) inhibitor, 30 min prior to intratracheal exposure. At 12 h
23 after intratracheal instillations, ip injections (1 mL of NS398 in 20% ethanol in saline) were
24 repeated. ROFA treatment induced a marked increase in the level of PGE2 recovered in the BAL
25 fluid, which was effectively decreased by pretreating the animals with the COX2 inhibitor.
26 Immunohistochemical analyses of rat airway showed concomitant expression of COX2 in the
27 proximal airway epithelium of rats treated with soluble fraction of ROFA. This study further
28 showed that, although COX2 products participated in ROFA induced lung inflammation, the
29 COX metabolites are not involved in IL-6 expression nor the influx of PMN influx into the
30 airway. However, the rationale for the use of intraperitoneal challenge was not elaborated.
April 2002 7-82 DRAFT-DO NOT QUOTE OR CITE
-------
1 The production of cytokines and mediators also has been shown to depend on the type of
2 PM used in the experiments. A549 cells (a human airway epithelial cell line) were exposed to
3 several PM, carbon black (CB, Elftex-12, Cabot Corp.), diesel soot (ND from NIST, LD
4 produced from General Motors LH 6.2 V8 engine at light duty cycle), ROFA (from the heat
5 exchange section of the Boston Edison), OAA (Ottowa ambient air PM, EHC-93), SiO2, and
6 Ni3S2 at Img/cm2 (Seagrave and Nikula, 2000). Results indicated that (1) SiO2 and Ni3S2 caused
7 dose dependent acute toxicity and apototic changes; (2) ROFA and ND were significant only at
8 the highest concentrations; (3) SiO2 and Ni3S2 increased IL-8 (three and eight times over the
9 control, respectively) at low concentrations but suppressed IL-8 at high concentrations; (4) OAA
10 and ROFA also induced IL-8 but lower than SiO2 and Ni3S2; and (5) both diesel soots suppressed
11 IL-8 production. The order of potency in alkaline phasphatase production is OAA > LD =
12 ND > ROFA » SiO2 = Ni3S2. These results demonstrated that the type of particle used has a
13 strong influence on the biological response.
14 Expression of MIP-2 and IL-6 genes was significantly upregulated as early as 6 h
15 post-ROFA-exposure in rat tracheal epithelial cells; whereas gene expression of iNOS was
16 maximally increased 24 h postexposure. V but not Ni appeared to be mediating the effects of
17 ROFA on gene expression. Treatment with dimethylthiourea inhibited both ROFA and V
18 induced gene expression in a dose-dependent manner (Dye et al., 1999).
19 It appears that many biological responses are produced by PM whether it is composed of a
20 single component or a complex mixture. A technical approach is to use the newly developed
21 gene array to monitor the expressions of many mediator genes, which regulate complex and
22 coordinated cellular events involved in tissue injury and repair, in a single assay. Using an array
23 consisting of 84 rat genes representing inflammatory and anti-inflammatory cytokines, growth
24 factors, adhesion molecules, stress proteins, transcription factors, and antioxidant enzymes,
25 Nadadur et al. (2000) and Nadadur and Kodavanti (2002) measured the pulmonary expressions of
26 these genes in rats intratracheally instilled with ROFA (3.3 mg/kg), NiSO4 (1.3 //mol/kg), and
27 VSO4 (2.2 //mol/kg). Their data revealed a twofold induction of IL-6 and TEVIP-1 at 24 h post-
28 ROFA or Ni exposure. The expression of cellular fibronectin (cFn-EIIIA), ICAM-1, IL-lb, and
29 iNOS gene also were increased 24 h post-ROFA, V, or Ni exposure. This study demonstrated
30 that gene array may provide a tool for screening the expression profile of tissue specific markers
31 following exposure to PM.
April 2002 7-83 DRAFT-DO NOT QUOTE OR CITE
-------
1 To investigate the interaction between respiratory cells and PM, Kobzik (1995) showed that
2 scavenger receptors are responsible for AM binding of unopsonized PM and that different
3 mechanisms mediate binding of carbonaceous dusts such as DPM. In addition, surfactant
4 components can increase AM phagocytosis of environmental particulates in vitro, but only
5 slightly relative to the already avid AM uptake of unopsonized particles (Stringer and Kobzik,
6 1996). Respiratory tract epithelial cells are also capable of binding with PM to secrete cytokine
7 IL-8. Using a respiratory epithelial cell line (A549), Stringer et al. (1996) found that binding of
8 particles to epithelial cells was calcium-dependent for TiO2 and Fe2O3, while a-quartz binding
9 was not calcium dependent. In addition, as observed in AMs, PM binding by A549 cells also
10 was mediated by scavenger receptors, albeit those distinct from the heparin-insensitive
11 acetylated-LDL receptor. Furthermore, a-quartz, but not TiO2 or CAPs, caused a dose-dependent
12 production of IL-8 (range 1 to 6 ng/mL), demonstrating a particle-specific spectrum of epithelial
13 cell cytokine (IL-8) response.
14
15 7.5.3.3 Other Potential Cellular and Molecular Mechanisms
16 A potential mechanism involving in the alteration of surface tension may be related to
17 changes in the expression of matrix metalloproteinases (MMPs), such as pulmonary matrilysin
18 and gelatinase A and B, and tissue inhibitor of metalloproteinase (TEVIP) (Su et al., 2000a,b).
19 Sprague-Dawley rats exposed to ROFA by intratracheal injection (2.5 mg/rat) had increased
20 mRNA levels of matrilysin, gelatinase A, and TEVIP-1. Gelatinase B, not expressed in control
21 animals, was increased significantly from 6 to 24 h following ROFA exposure. Alveolar
22 macrophages, epithelial cells, and inflammatory cells were major cellular sources for the
23 pulmonary MMP expression. The expression of Gelatinase B in rats exposed to the same dose of
24 ambient PM (<1.7 //m and 1.7 to 3.7 //m) collected from Washington, DC, was significantly
25 increased as compared to saline control; whereas the expression of TIMP-2 was suppressed.
26 Ambient PM between 3.7 and 20 //m also increased the Gelatinase B expression. Increases in
27 MMPs, which degrade most of the extracellular matrix, suggest that ROFA and ambient PM can
28 similarly increase the total pool of proteolytic activity to the lung and contribute in the
29 pathogenesis of PM-induced lung injury.
30 The role of sensory nerve receptors in the initiation of PM inflammation has been described
31 in a series of recent studies. Neuropeptide and acid-sensitive sensory irritant (i.e., capsaicin,
April 2002 7-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 VR1) receptors were first identified on human bronchial epithelial cells (i.e., BEAS-2B).
2 To address whether PM could initiate airway inflammation through these acid sensitive sensory
3 receptors, BEAS-2B cells were exposed to ROFA and responded with an immediate increase in
4 [Ca+2]; followed by a concentration-dependent release of inflammatory cytokine (i.e., IL-6, IL-8,
5 TNFa) and their transcripts (Veronesi et al., 1999a). To test the relevance of neuropeptide or
6 capsaicin VR1 receptors to these changes, BEAS-2B cells were pretreated with neuropeptide
7 receptor antagonists or capsazepine (CPZ), the antagonist for the capsaicin (i.e., VR1) receptor.
8 The neuropeptide receptor antagonists reduced ROFA-stimulated cytokine release by 25%-50%.
9 However, pretreatment of cells with CPZ inhibited the immediate increases in [Ca+2];, diminished
10 transcript (i.e., IL-6, IL-8, TNFa) levels and reduced IL-6 cytokine release to control levels
11 (Veronesi et al., 1999b). The above studies suggested that ROFA inflammation was mediated by
12 acid sensitive VR1 receptors located on the sensory nerve fibers that innervate the airway and on
13 epithelial target cells.
14 Colloidal particles (like ROFA and other PM) carry an inherently negative surface charge
15 (i.e., zeta potential) that attracts protons from their vaporous milieu. These protons form a
16 neutralizing, positive ionic cloud around the individual particle (Hunter, 1981). Since VR1
17 irritant receptors respond to acidity (i.e., protonic charge), experiments were designed to
18 determine if the surface charge carried by ROFA and other PM particles could biologically
19 activate cells and stimulate inflammatory cytokine release. The mobility of ROFA particles was
20 measured in an electrically charged field (i.e., micro-electrophoresis) microscopically and their
21 zeta potential calculated. Next, synthetic polymer microspheres (SPM) (i.e., polymethacrylic
22 acid nitrophenylacrylate microspheres) were prepared with attached carboxyl groups to yield
23 SPM particles of the same size and with zeta potentials similar to ROFA (-29 + 0.9 mV)
24 particles. These SPM acted as ROFA surrogates with respect to their size and surface charge, but
25 lacked all other contaminants that were thought to be responsible for its toxicity (e.g., transition
26 metals, sulfates, volatile organics and biologicals). Similar concentrations of SPM and ROFA
27 particles were used to test BEAS-2B cells and mouse dorsal root ganglia (DRG) sensory neurons,
28 both targets of inhaled PM. Equivalent degrees of biological activation (i.e., increase in
29 intracellular calcium, [Ca+2];, IL-6 release) occurred in both cell types in response to either ROFA
30 or SPM and both responses could be reduced by antagonists to VR1 receptors or acid-sensitive
31 pathways. Neutrally charged SPM (i.e., zeta potential of 0 mV), however, failed to stimulated
April 2002 7-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 increases in [Ca+2]; or IL-6 release (Oortgiesen et al., 2000). To expand on these data, a larger set
2 of PM was obtained from urban (St. Louis, Ottawa), residential (wood stove), volcanic (Mt. St.
3 Helen), and industrial (oil fly ash, coal fly ash) sources. Each PM sample was described
4 physicochemically (i.e., size and number of visible particles, acidity, zeta potential) and used to
5 test BEAS-2B epithelial cells. The resulting biological effect (i.e., increases in [Ca+2];, IL-6
6 release) was related to their physicochemical descriptions. When examined by linear regression
7 analysis, the only measured physicochemical property that correlated with increases in [Ca+2]; and
8 IL-6 release was the zeta potential of the visible particles (r2 > 0.97) (Veronesi et al., 2002b).
9 Together, these studies have demonstrated a neurogenic basis for PM inflammation by which the
10 proton cloud associated with negatively-charged colloidal PM particles can activate acid-
11 sensitive VR1 receptors found on human airway epithelial cells and sensory terminals. This
12 activation results in an immediate influx of calcium and the release of inflammatory
13 neuropeptides and cytokines which proceed to initiate and sustain inflammatory events in the
14 airways through the pathophysiology of neurogenic inflammation (Veronesi and Oortgiesen,
15 2001).
16
17 7.5.4 Specific Particle Size and Surface Area Effects
18 Most particles used in laboratory animal toxicology and occupational studies are greater
19 than 0.1 //m in size. However, the enormous number and huge surface area of the ultrafine
20 particles demonstrate the importance of considering the size of the particle in assessing response.
21 Ultrafine particles with a diameter of 20 nm when inhaled at the same mass concentration have a
22 number concentration that is approximately 6 orders of magnitude higher than for a 2.5-//m
23 diameter particle; particle surface area is also greatly increased (Table 7-11).
24 Many studies summarized in U.S. Environmental Protection Agency (1996a), as well as in
25 this document, suggest that the surface of particles or substances that are released from the
26 surface (e.g., transition metals) interact with the biological system, and that surface-associated
27 free radicals or free radical-generating systems may be responsible for toxicity. Thus, if ultrafine
28 particles were to cause toxicity by a transition metal-mediated mechanism, for example, then the
29 relatively large surface area for a given mass of ultrafine particles would mean high
30 concentrations of transition metals being available to cause oxidative stress to cells.
31
April 2002 7-86 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-11. NUMBERS AND SURFACE AREAS OF MONODISPERSE
PARTICLES OF UNIT DENSITY OF DIFFERENT SIZES AT A MASS
CONCENTRATION OF 10
Particle Diameter
Cwm)
Source:
0.02
0.1
0.5
1.0
2.5
Oberdorster (1996).
Particle Number
(per cm3 air)
2,400,000
19,100
153
19
1.2
Particle Surface Area
(,wm2 per cm3 air)
3,016
600
120
60
24
1 Two groups have examined the toxic differences between fine and ultrafme particles, with
2 the general finding that the ultrafme particles show a significantly greater response at similar
3 mass doses (Oberdorster et al., 1992; Li et al., 1996, 1997, 1999). However, only a few studies
4 have investigated the ability of ultrafme particles to generate a greater oxidative stress when
5 compared to fine particles of the same material. Studies by Gilmour et al. (1996) have shown
6 that at equal mass, ultrafme TiO2 caused more plasmid DNA strand breaks than fine TiO2. This
7 effect could be inhibited with mannitol. Osier and Oberdorster (1997) compared the response of
8 rats (F344) exposed by intratracheal inhalation to "fine" (approximately 250 nm) and "ultrafme"
9 (approximately 21 nm) TiO2 particles with rats exposed to similar doses by intratracheal
10 instillation. Animals receiving particles through inhalation showed a smaller pulmonary
11 response, measured by BAL parameters, in both severity and persistence, when compared with
12 those animals receiving particles through instillation. These results demonstrate a difference in
13 pulmonary response to an inhaled versus an instilled dose, which may result from differences in
14 dose rate, particle distribution, or altered clearance between the two methods. Consistent with
15 these in vivo studies, Finkelstein et al. (1997) has shown that exposing primary cultures of rat
16 Type n cells to 10 //g/mL ultrafme TiO2 (20 nm) causes increased TNF and IL-1 release
17 throughout the entire 48-h incubation period. In contrast, fine TiO2 (200 nm) had no effect.
18 In addition, ultrafme polystyrene carboxylate-modified microspheres (UFP, fluorospheres,
19 molecular probes 44 ± 5 nm) have been shown to induce a significant enhancement of both
20 substance P and histamine release after administration of capsaicin (10"4 M), to stimulate C-fiber,
21 and carbachol (10"4 M), a cholinergic agonist in rabbit intratracheally instilled with UFP
April 2002 7-87 DRAFT-DO NOT QUOTE OR CITE
-------
1 (Nemmar et al., 1999). A significant increase in histamine release also was recorded in the
2 UFP-instilled group following the administration of both Substance P (10"6 M) plus thiorpan
3 (10"5 M) and compound 48/80 (C48/80, 10"3 M) to stimulate mast cells. BAL analysis showed an
4 influx of PMN, an increase in total protein concentration, and an increase in lung wet weight/dry
5 weight ratio. Electron microscopy showed that both epithelial and endothelial injuries were
6 observed. The pretreatment of rabbits in vivo with a mixture of either SR 140333 and SR 48368,
7 a tachykinin NKl and NK2 receptor antagonist, or a mixture of terfenadine and cimetidine,
8 a histamine Hx and H2 receptor antagonist, prevented UFP-induced PMN influx and increased
9 protein and lung WW/DW ratio.
10 Given the assumption that the chemical composition of ultrafme particles is the same as
11 larger particles, it is believed that ultrafme particles cause greater cellular injury because of the
12 relatively large surface area for a given mass. However, in a study that compared the response to
13 carbon black particles of two different sizes, Li et al. (1999) demonstrated that in the instillation
14 model, a localized dose of particle over a certain level causes the particle mass to dominate the
15 response, rather than the surface area. Ultrafme carbon black (ufCB, Printex 90), 14 nm in
16 diameter, and fine carbon black (CB, Huber 990), 260 nm in diameter, were instilled
17 intratracheally in rats and BAL profile at 6 h was assessed. At mass of 125 //g or below, ufCB
18 generated a greater response (increase LDH, epithelial permeability, decrease in GSH, TNF, and
19 NO production) than fine CB at various times postexposure. However, higher dose of CB caused
20 more PMN influx than the ufCB. In contrast to the effect of CB, which showed dose-related
21 increasing inflammatory response, ufCB at the highest dose caused less of a neutrophil influx
22 than at the lower dose, confirming earlier work reported by Oberdorster et al. (1992). Moreover,
23 when the PMN influx was expressed as a function of surface area, CB produced greater response
24 than ufCB at all doses used in this study. Although particle insterstitialization with a consequent
25 change in the chemotatic gradient for PMN was offered as an explanation, these results need
26 further scrutiny.
27 Oberdorster et al. (2000) recently completed a series of studies in rats and mice using
28 ultrafme particles of various chemical compositions (PTFE, TiO2, C, Fe, Fe2O3, Pt, V, and V2O5).
29 In old rats sensitized with endotoxin and exposed to ozone plus ultrafme carbon particles, they
30 found a ninefold greater release of reactive oxygen species in old rats than in similarly treated
April 2002 7-88 DRAFT-DO NOT QUOTE OR CITE
-------
1 young rats. Exposure to ultrafine PM alone in sensitized old rats also caused an inflammatory
2 response.
3 Although the exact mechanism of ultrafme-induced lung injury remains unclear, it is likely
4 that ultrafine particles, because of their small size, can easily penetrate the airway epithelium and
5 cause cellular damage. Using electron microscopy to examine rat tracheal explants treated with
6 fine (0.12 //m) and ultrafine (0.021 //m) TiO2 particles for 3 or 7 days, Churg et al. (1998) found
7 both size particles in the epithelium at both time points, but in the subepithelial tissues, they were
8 found only at Day 7. The volume proportion (the volume of TiO2 over the entire volume of
9 epithelium or subepithelium area) of both fine and ultrafine particles in the epithelium increased
10 from 3 to 7 days. It was greater for ultrafine at 3 days but was greater for fine at 7 days. The
11 volume proportion of particles in the subepithelium at day 7 was equal for both particles, but the
12 ratio of epithelial to subepithelial volume proportion was 2:1 for fine and 1:1 for ultrafine.
13 Ultrafine particles persisted in the tissue as relatively large aggregates; whereas the size of fine
14 particle aggregates became smaller over time. Ultrafine particles appeared to enter the
15 epithelium faster and, once in the epithelium, a greater proportion of them were translocated to
16 the subepithelial space compared to fine particles. However, the authors assumed that the
17 volume proportion is representative of particle number and the number of particles reaching the
18 interstitial space is directly proportional to the number applied (i.e., there is no preferential
19 transport from lumen to interstitium by size). These data are in contrast to the results of
20 instillation or inhalation of fine and ultrafine TIO2 particles reported earlier (Ferin et al., 1990,
21 1992). However, the explant and intratracheal instillation test systems differ in many aspects
22 making direct comparisons difficult. Limitations of the explant test system include traumatizing
23 the explanted tissue, introducing potential artifacts through the use of liquid suspension for
24 exposure, the absence of inflammatory cells, and possible overloading of the explants with dust.
25 Only two studies examined the influence of specific surface area on biological activity
26 (Lison et al., 1997; Oettinger et al., 1999). The biological responses to various MnO2 dusts with
27 different specific surface area (0.16, 0.5, 17, and 62 m2/g) were compared in vitro and in vivo
28 (Lison et al., 1997). In both systems, the results show that the amplitude of the response is
29 dependent on the total surface area that is in contact with the biological system, indicating that
30 surface chemistry phenomena are involved in the biological reactivity. Freshly ground particles
31 with a specific surface area of 5 m2/g also were examined in vitro. These particles exhibited an
April 2002 7-89 DRAFT-DO NOT QUOTE OR CITE
-------
1 enhanced cytotoxic activity that was almost equivalent to that of particles with a specific surface
2 area of 62 m2/g, indicating that undefined reactive sites produced at the particle surface by
3 mechanical cleavage also may contribute to the toxicity of insoluble particles. In another study,
4 two types of carbon black particles, Printex 90 (P90, Degussa, Germany, formed by controlled
5 combustion, consists of defined granules with specific surface area of 300 m2/g and particle size
6 of 14 nm) and FR 101 (Degussa, Germany, with specific surface area of 20 m2/g and particle size
7 of <95 nm, has a coarse structure, and the ability to adsorb polycyclic and other carbons) were
8 used in the study (Oettinger et al., 1999). Exposure of AMs to 100 //g/106 cells of FR 101 and
9 P90 resulted in a 1.4- and 2.1-fold increase in ROS release. These exposures also caused a
10 fourfold up-regulation of NF-KB gene expression. These studies indicated that PM of single
11 component with larger surface area produce greater biological response than similar particles
12 with smaller surface area. By exposing bovine AMs to metal oxide coated silica particles,
13 Schluter et al. (1995) showed that most of the metal coatings (As, Ce, Fe, Mn, Ni, Pb, and V) had
14 no effect on ROS production by these cells. However, coating with CuO markedly lowered the
15 O2" and H2O2, whereas V(IV) increases both ROI. This study demonstrated that, in addition to
16 specific area, chemical composition of the particle surface also influence its cellular response.
17 Thus, ultrafine particles have the potential to significantly contribute to the adverse effects of
18 PM. These studies, however, have overlooked the portion of ambient ultrafine particles that are
19 not solid in form. Droplets (e.g., sulfuric acid droplets) and organic based ultrafine particles do
20 exist in the ambient environment, but their role in the adverse effects of ultrafine particles has
21 been ignored. Moreover, the ability of these droplet ultrafine particles to spread, disperse, or
22 dissolve after contact with liquid surface layers must be considered.
23
24 7.5.5 Pathophysiological Mechanisms for the Effects of Low Concentrations
25 of Particulate Air Pollution
26 The pathophysiological mechanisms involved in PM-associated cardiovascular and
27 respiratory health effects still are not elucidated fully, but progress has been made since the 1996
28 PM AQCD (U. S. Environmental Protection Agency, 1996a) was prepared. This section
29 summarizes several current hypotheses and reviews the toxicological evidence for these potential
30 pathophysiological mechanisms.
31
April 2002 7-90 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.5.5.1 Direct Pulmonary Effects
2 When the 1996 PM AQCD (U. S. Environmental Protection Agency, 1996a) was written,
3 the lung was thought to be the primary organ that was affected by particulate air pollution.
4 Although the lung still is a primary organ affected by PM inhalation, there is growing
5 toxicological and epidemiological evidence that the cardiovascular system is affected, as well,
6 and may be a co-primary organ system related to certain health endpoints such as mortality.
7 Nonetheless, understanding how particulate air pollution causes or exacerbates respiratory
8 disease remains an important goal. There is some toxicological evidence for the following three
9 mechanisms for direct pulmonary effects.
10
11 Particulate Air Pollution Causes Lung Injury and Inflammation
12 In the last few years, numerous studies have shown that instilled and inhaled ROFA, a
13 product of fossil fuel combustion, can cause substantial lung injury and inflammation. The toxic
14 effects of ROFA are largely caused by its high content of soluble metals, and some of the
15 pulmonary effects of ROFA can be reproduced by equivalent exposures to soluble metal salts.
16 In contrast, controlled exposures of animals to sulfuric acid aerosols, acid-coated carbon, and
17 sulfate salts cause little lung injury or inflammation, even at high concentrations. Inhalation of
18 concentrated ambient PM (which contains only small amounts of metals) by laboratory animals
19 at concentrations in the range of 100 to 1000 //g/m3 have been shown in some (but not all)
20 studies to cause mild pulmonary injury and inflammation. Rats with SO2-induced bronchitis and
21 monocrotaline-treated rats have been reported to have a greater inflammatory response to
22 concentrated ambient PM than normal rats. These studies suggest that exacerbation of
23 respiratory disease by ambient PM may be caused in part by lung injury and inflammation.
24
25 Particulate Air Pollution Causes Increased Susceptibility to Respiratory Infections
26 At this time there are no newly published studies on the effects of inhaled concentrated
27 ambient PM on host susceptibility to infectious agents. Ohtsuka et al. (2000a,b) have shown that
28 in vivo exposure of mice to acid-coated carbon particles at a mass concentration of 10,000 //g/m3
29 carbon black causes decreased phagocytic activity of alveolar macrophages, even in the absence
30 of lung injury.
31
April 2002 7-91 DRAFT-DO NOT QUOTE OR CITE
-------
1 Particulate Air Pollution Increases Airway Reactivity and Exacerbates Asthma
2 The strongest evidence supporting this hypothesis is from studies on diesel particulate
3 matter (DPM). DPM has been shown to increase production of antigen-specific IgE in mice and
4 humans (summarized in Section 7.2.1.2). In vitro studies have suggested that the organic
5 fraction of DPM is involved in the increased IgE production. ROFA leachate also has been
6 shown to enhance antigen-specific airway reactivity in mice (Goldsmith et al., 1999), indicating
7 that soluble metals can also enhance an allergic response. However, in this same study, exposure
8 of mice to concentrated ambient PM did not affect antigen-specific airway reactivity. It is
9 premature to conclude from this one experiment that concentrated ambient PM does not
10 exacerbate allergic airways disease because the chemical composition of the PM (as indicated by
11 studies with DPM and ROFA) may be more important than the mass concentration.
12
13 7.5.5.2 Systemic Effects Secondary to Lung Injury
14 When the 1996 PM AQCD was written, it was thought that cardiovascular-related
15 morbidity and mortality most likely would be secondary to impairment of oxygenation or some
16 other consequence of lung injury and inflammation. Newly available toxicologic studies provide
17 some additional evidence regarding such possibilities.
18
19 Lung Injury from Inhaled Particulate Matter Causes Impairment of Oxygenation and
20 Increased Work of Breathing That Adversely Affects the Heart
71
22 Instillation of ROFA has been shown to cause a 50% mortality rate in monocrotaline-
23 treated rats (Watkinson et al., 2000a,b). Although blood oxygen levels were not measured in this
24 study, there were ECG abnormalities consistent with severe hypoxemia in about half of the rats
25 that subsequently died. Given the severe inflammatory effects of instilled ROFA and the fact
26 that monocrotaline-treated rats have increased lung permeability as well as pulmonary
27 hypertension, it is plausible that instilled ROFA can cause severe hypoxemia leading to death in
28 this rat model. Results from studies in which animals (normal and compromised) were exposed
29 to concentrated ambient PM (at concentrations many times higher than would be encountered in
30 the United States) indicate that ambient PM is unlikely to cause severe disturbances in
31 oxygenation or pulmonary function. However, even a modest decrease in oxygenation can have
32 serious consequences in individuals with ischemic heart disease. Kleinman et al. (1998) has
April 2002 7-92 DRAFT-DO NOT QUOTE OR CITE
-------
1 shown that a reduction in arterial blood saturation from 98 to 94% by either mild hypoxia or by
2 exposure to 100 ppm CO significantly reduced the time to onset of angina in exercising
3 volunteers. Thus, information is needed on the effects of PM on arterial blood gases and
4 pulmonary function to fully address the above hypothesis.
5
6 Lung Inflammation and Cytokine Production Cause Adverse Systemic Hemodynamic Effects
1 It has been suggested that systemic effects of particulate air pollution may result from
8 activation of cytokine production in the lung (Li et al., 1997). In support of this idea,
9 monocrotaline-treated rats exposed to inhaled ROFA (15,000 //g/m3, 6 h/day for 3 days) showed
10 increased pulmonary cytokine gene expression, bradycardia, hypothermia, and increased
11 arrhythmias (Watkinson et al., 2000a,b). However, spontaneously hypertensive rats had a similar
12 cardiovascular response to inhaled ROFA (except that they also developed ST segment
13 depression) with no increase in pulmonary cytokine gene expression. Studies in dogs exposed to
14 concentrated ambient PM showed minimal pulmonary inflammation and no positive staining for
15 IL-8, IL-1, or TNF in airway biopsies. However, there was a significant decrease in the time of
16 onset of ischemic ECG changes following coronary artery occlusion in PM-exposed dogs
17 compared to controls (Godleski et al., 2000). Thus, the link between changes in the production
18 of cytokines in the lung and cardiovascular function is not clear-cut, and basic information on the
19 effects of mild pulmonary injury on cardiovascular function is needed to understand the
20 mechanisms by which inhaled PM affects the heart.
21
22 Lung Inflammation from Inhaled Particulate Matter Causes Increased Blood Coagulability
23 That Increases the Risk of Heart Attacks and Strokes
24
25 There is abundant evidence linking risk of heart attacks and strokes to small prothrombotic
26 changes in the blood coagulation system. However, the published toxicological evidence that
27 moderate lung inflammation causes increased blood coagulability is inconsistent. Ohio et al.
28 (2000a) have shown that inhalation of concentrated ambient PM in healthy nonsmokers causes
29 increased levels of blood fibrinogen. Gardner et al. (2000) have shown that a high dose
30 (8,300 //g/kg) of instilled ROFA in rats causes increased levels of fibrinogen, but no effect was
31 seen at lower doses. Exposure of dogs to concentrated ambient PM had no effect on fibrinogen
32 levels (Godleski et al., 2000). The coagulation system is as multifaceted and complex as the
April 2002 7-93 DRAFT-DO NOT QUOTE OR CITE
-------
1 immune system, and there are many other sensitive and clinically significant parameters that
2 should be examined in addition to fibrinogen. Thus, it is premature to draw any conclusions on
3 the relationship between PM and blood coagulation.
4
5 Interaction of Particulate Matter with the Lung Affects Hematopoiesis
6 Terashima et al. (1997) found that instillation of fine carbon particles (20,000 //g/rabbit)
7 stimulated release of PMNs from the bone marrow. In further support of this hypothesis, Gordon
8 and colleagues reported that the percentage of PMNs in the peripheral blood increased in rats
9 exposed to ambient PM in some but not all exposures. On the other hand, Godleski et al. (2000)
10 found no changes in peripheral blood counts of dogs exposed to concentrated ambient PM.
11 Thus, direct evidence that PM ambient concentrations can affect hematopoiesis remains to be
12 demonstrated.
13
14 7.5.5.3 Direct Effects on the Heart
15 Changes in heart rate, heart rate variability, and conductance associated with ambient PM
16 exposure have been reported in animal studies (Godleski et al., 2000; Gordon et al., 2000;
17 Watkinson et al., 2000a,b; Campen et al., 2000), in several human panel studies (described in
18 Chapter 8), and in a reanalysis of data from the MONICA study (Peters et al., 1997). Some of
19 these studies included endpoints related to respiratory effects but few significant adverse
20 respiratory changes were detected. This raises the possibility that ambient PM may have effects
21 on the heart that are independent of adverse changes in the lung. There is certainly precedent for
22 this idea. For example, tobacco smoke (which is a mixture of combustion-generated gases and
23 PM) causes cardiovascular disease by mechanisms that are independent of its effect on the lung.
24 Two types of hypothesized direct effects of PM on the heart are noted below.
25
26 Inhaled Particulate Matter Affects the Heart by Uptake of Particles into the Circulation
27 or Release of a Soluble Substances into the Circulation,
29 Drugs can be rapidly and efficiently delivered to the systemic circulation by inhalation.
30 This implies that the pulmonary vasculature absorbs inhaled materials, including charged
31 substances such as small proteins and peptides. Cigarettes are a widely used method for
32 delivering nicotine to the blood stream. It is likely that soluble materials absorbed onto airborne
April 2002 7-94 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles find their way into the blood stream, but it is not clear whether the particles themselves
2 enter the blood. It is anticipated that more information will be available on this important
3 question in the next few years.
4
5 Inhaled Particulate Matter Affects Autonomic Control of the Heart and
6 Cardiovascular System
8 There is growing evidence for this idea as described above. This raises the question of how
9 inhaled particles could affect the autonomic nervous system. Activation of neural receptors in
10 the lung is a logical area to investigate. Studies in conscious rats have shown that inhalation of
11 wood smoke causes marked changes in sympathetic and parasympathetic input to the
12 cardiovascular system that are mediated by neural reflexes (Nakamura and Hayashida, 1992).
13 Although research on airway neural receptors and neural-mediated reflexes is a well established
14 discipline, the cardiovascular effects of stimulating airway receptors continue to receive less
15 attention than the pulmonary effects. Previous studies of airway reflex-mediated cardiac effects
16 usually employed very high doses of chemical irritants, and the results may not be applicable to
17 air pollutants. There is a need for basic physiological studies to examine effects on
18 cardiovascular system when airway and alveolar neural receptors are stimulated in a manner
19 relevant to air pollutants.
20
21
22 7.6 RESPONSES TO PARTICULATE MATTER AND GASEOUS
23 POLLUTANT MIXTURES
24 Ambient PM itself is a mixture of particles of varying size and composition. The following
25 discussion examines effects of mixtures of ambient PM, or PM surrogates, with gaseous
26 pollutants. Ambient PM co-exists in indoor and outdoor air with a number of co-pollutant gases,
27 including ozone, sulfur dioxide, oxides of nitrogen, and carbon monoxide. Toxicological
28 interactions between PM and gaseous co-pollutants may be antagonistic, additive, or synergistic
29 (Mauderly, 1993). The presence and nature of any interaction appears to depend on the chemcial
30 composition, size, concentration and ratios of pollutants in the mixture, exposure duration, and
31 the endpoint being examined. It may be difficult to predict a priori from the presence of certain
April 2002 7-95 DRAFT-DO NOT QUOTE OR CITE
-------
1 pollutants whether any interaction will occur and, if there is interaction, whether it will be
2 synergistic, additive, or antagonistic (Table 7-12).
3 Mechanisms responsible for the various forms of interaction are speculative. In terms of
4 potential health effects, the greatest hazard from pollutant interaction is the possibility of synergy
5 between particles and gases, especially if effects occur at concentrations at which no effects
6 occur when individual constituents are inhaled. Various physical and chemical mechanisms may
7 underlie synergism. For example, physical adsorption or absorption of some material on a
8 particle could result in transport to more sensitive sites, or sites where this material would not
9 normally be deposited in toxic amounts. This physical process may explain the interaction found
10 in studies of mixtures of carbon black and formaldehyde or of carbon black and acrolein (Jakab,
11 1992,1993).
12 Chemical interactions between PM and gases can occur on particle surfaces, thus, forming
13 secondary products that may be more active lexicologically than the primary materials and that
14 can then be carried to a sensitive site. The hypothesis of such chemical interactions has been
15 examined in the gas and particle exposure studies by Amdur and colleagues (Amdur and Chen,
16 1989; Chen et al., 1992) and Jakab and colleagues (Jakab and Hemenway, 1993; Jakab et al.,
17 1996). These investigators have suggested that synergism occurs as secondary chemical species
18 are produced, especially under conditions of increased temperature and relative humidity.
19 Another potential mechanism of gas-particle interaction may involve a pollutant-induced
20 change in the local microenvironment of the lung, enhancing the effects of the co-pollutant.
21 For example, Last et al. (1984) suggested that the observed synergism between ozone (O3) and
22 acid sulfates in rats was due to a decrease in the local microenvironmental pH of the lung
23 following deposition of acid, enhancing the effects of O3 by producing a change in the reactivity
24 or residence time of reactants, such as radicals, involved in O3-induced tissue injury. Likewise,
25 Pinkerton et al (1989) showed increased retention of the mass and number of asbestos fibers in
26 rats exposed to O3, suggesting an increase in lung fiber burden due to exposure to this gaseous
27 pollutant.
28 As noted in U.S. Environmental Protection Agency (1996a), the toxicology database for
29 mixtures containing PM other than acid sulfates was and is still quite sparse. Vincent et al.
30 (1997) exposed rats to 0.8 ppm O3 in combination with 5 or 50 mg/m3 of resuspended urban
31 particles for 4 h. Although PM alone caused no change in cell proliferation (3H-thymidine
April 2002 7-96 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-12. RESPIRATORY AND CARDIOVASCULAR EFFECTS OF MIXTURES
to
o
o
to
H
6
o
o
H
O
o
HH
H
W
Species, Gender,
Strain Age, or
Body Weight
Rats, Fischer
NNia, male,
22 to 24 mo old
Rats
Humans; healthy
15 M, 10 F,
34.9+10 years of
age
Humans; healthy
children
Humans; 59
healthy childern
in Mexico City;
19 controls in
Gulf port town
Humans; 15
healthy children
in Mexico City;
1 1 children in
Veracruz; 4-15
years of age
Humans; 83
healthy children
in Mexico City;
24 children in
Isla Mujeres;
6- 12 years of
age
Gases and PM
Carbon,
ammonium
bisulfate,
and O3
O3 and Ottawa
urban dust
CAPs
Ambient gases
and particles
Ambient gases
and particles
Ambient gases
and particles
Ambient gases
and particles
Particle
Exposure Technique Mass Concentration Size
Inhalation 50 ^g/m3 carbon + 0.4/^mMMAD
70 fj.g/m1 ammonium og = 2.0
bisulfate + 0.2 ppm
O3 or 100 Mg/m3
carbon +140 i/g/m3
ammonium bisulfate
+ 0.2 ppm O3
Inhalation 40,000 ,ug/m3 and 4.5 ,um
0.8 ppm O3 MM AD
Inhalation ISO^g/m3 PM25
120 ppb O3
Natural 24 h
exposure in
Southwest
Metropolitan
Mexico City
(SWMMC)
Natural 24 h
exposure in
SWMMC compared
to low pollution
Gulf of Mexico
Natural 24 h
exposure in
SWMMC compared
to low pollution
Gulf Coast
Natural 24 h
exposure in
SWMMC compared
to low pollution
Caribbean
Exposure Cardiopulmonary Effects of Inhaled
Duration PM and Gases
4 h/day, No changes in protein concentration in lavage
3 days/week for fluid or in prolyl 4-hydroxylase activity in
4 weeks blood. Slight, but statistically significant
decreases in plasma fibronectin in animals
exposed to the combined atmospheres
compared to animals exposed to O3 alone.
Single 4-h Co-exposure to particles potentiated O3-induced
exposure followed septal cellurity. Enhanced septal thickening
by 20 h clean air associated with elevated production of
macrophage inflammatory protein-2 and
endothelin 1 by lung lavage cells.
2 h Acute brachial artery vasoconstriction as
determined by vascular ultrasonography
performed before and 10 min after exposure.
Radiological evidence of lung hyperinflation
from chest X-rays.
Increased upper and lower respiratory
symptoms; bilateral symmetric mild lung
hyperinflation from chest X-rays.
Nasal biopsies revealed increased basal,
ciliated, goblet, and squamous metaplastic
and intermediate cells; cellular abnormalities
and possible dyskinesia were noted.
Nasal biopsies revealed p53 accumulation by
immunochemistry; increased upper and lower
respiratory symptoms.
Reference
Bolarin et al.
(1997)
Bouthillier et al.
(1998)
Brook et al. (2002)
Calderon-Garciduenas
et al. (2000a)
Calderon-Garciduenas
et al. (2000b)
Calderon-Garciduenas
etal. (2001a)
Calderon-Garciduenas
etal. (2001b)
-------
-------
TABLE 7-12 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF MIXTURES
13.
to
o
o
to
Species, Gender,
Strain Age, or
Body Weight
Rats
Humans,
children, healthy
and asthmatic
Gases and PM
H2SO4 and O3
H2S04,
SO2, and O3
Exposure Technique
Inhalation,
whole body
Inhalation
Mass Concentration
20 to 150 Mg/m3
H2SO4and0.12or
0.2 ppm O3
60 to 140 Mg/m3
H2SO4, 0.1 ppm
SO2, and 0.1 ppm O3
Particle
Size
0.4 to 0.8 Mm
0.6 Mm H2SO4
Exposure
Duration
Intermittent
(12 h/day) or
continuous
exposure for up
to 90 days
Single 4-h
exposure with
intermittent
Cardiopulmonary Effects of Inhaled
PM and Gases
No interactive effect of H2SO4 and O3 on
biochemical and morphometric endpoints.
A positive association between acid concentration
and symptoms, but not spirometry, in asthmatic
children. No changes in healthy children.
Reference
Last and
Pinkerton (1997)
Linn et al.
(1997)
-------
1 labeling), co-exposure to either concentration of resuspended PM with O3 greatly potentiated the
2 proliferative effects of exposure to O3 alone. These interactive changes occurred in epithelial
3 cells of the terminal bronchioles and the alveolar ducts. These findings using resuspended dusts,
4 although at high concentrations, are consistent with studies demonstrating interaction between
5 sulfuric acid (H2SO4) aerosols and O3. Kimmel and colleagues (1997) examined the effect of
6 acute co-exposure to O3 and fine or ultrafine H2SO4 aerosols on rat lung morphology. They
7 determined morphometrically that alveolar septal volume was increased in animals co-exposed to
8 O3 and ultrafine, but not fine, H2SO4. Interestingly, cell labeling, an index of proliferative cell
9 changes, was increased only in animals co-exposed to fine H2SO4 and O3, as compared to animals
10 exposed to O3 alone. Importantly, Last and Pinkerton (1997) extended their previous work and
11 found that subchronic exposure to acid aerosols (20 to 150 //g/m3 H2SO4) had no interactive
12 effect on the biochemical and morphometric changes produced by either intermittent or
13 continuous O3 exposure (0.12 to 0.2 ppm). Thus, the interactive effects of O3 and acid aerosol
14 co-exposure in the lung disappeared during the long-term exposure. Sindhu et al. (1998)
15 observed an increase in rat lung putrescine levels after repeated, combined exposures to O3 and a
16 nitric acid vapor.
17 Kleinman et al. (1999) examined the effects of O3 plus fine, H2SO4-coated, carbon particles
18 (MMAD = 0.26 //m) for 1 or 5 days. They found the inflammatory response with the O3-particle
19 mixture was greater after 5 days (4 h/day) than after Day 1. This contrasted with O3 exposure
20 alone (0.4 ppm), which caused marked inflammation on acute exposure, but no inflammation
21 after 5 consecutive days of exposure.
22 Kleinman et al. (2000) examined the effects of a mixture of elemental carbon particles, O3,
23 and ammonium bisulfate on rat lung collagen content and macrophage activity. Decreases in
24 lung collagen, and increases in macrophage respiratory burst and phagocytosis were observed
25 relative to other pollutant combinations. Mautz et al. (2001) used a similar mixture (i.e.,
26 elemental carbon particles, O3, ammonium bisulfate, but with NO2 also) and exposure regimen as
27 Kleinman (2000). There were decreases in pulmonary macrophage Fc-receptor binding and
28 phagocytosis and increases in acid phosphatase staining. Bronchoalveolar epithelial permeability
29 cell proliferation were increased. Altered breathing-patterns were also observed, with some
30 adaptations occurring.
April 2002 7-100 DRAFT-DO NOT QUOTE OR CITE
-------
1 Studies have examined interactions between carbon particles and gaseous co-pollutants.
2 Jakab et al. (1996) challenged mice with a single 4-h exposure to a high concentration of carbon
3 particles (10 mg/m3) in the presence of SO2 at low and high relative humidities. Macrophage
4 phagocytosis was depressed significantly only in mice exposed to the combined pollutants under
5 high relative humidity conditions. This study suggests that fine carbon particles can serve as an
6 effective carrier for acidic sulfates where chemical conversion of adsorbed SO2 to acid sulfate
7 species occurred. Interestingly, the depression in macrophage function was present as late as
8 7 days postexposure. Bolarin et al. (1997) exposed rats to only 50 or 100 //g/m3 carbon particles
9 in combination with ammonium bisulfate and O3. Despite 4 weeks of exposure, they observed
10 no changes in protein concentration in lavage fluid or blood prolyl 4-hydroxylase, an enzyme
11 involved in collagen metabolism. Slight decreases in plasma fibronectin were present in animals
12 exposed to the combined pollutants versus O3 alone. Thus as, previously noted, the potential for
13 adverse effects in the lungs of animals challenged with a combined exposure to particles and
14 gaseous pollutants is dependent on numerous factors, including the gaseous co-pollutant,
15 concentration, and time.
16 In a complex series of exposures, Oberdorster and colleagues examined the interaction of
17 ultrafine carbon particles (100 //g/m3) and O3 (1 ppm) in young and old Fischer 344 rats that were
18 pretreated with aerosolized endotoxin (Elder et al., 2000a,b). In old rats, exposure to carbon and
19 O3 produced an interaction that resulted in a greater influx in neutrophils than that produced by
20 either agent alone. This interaction was not seen in young rats. Oxidant release from lavage
21 fluid cells was also assessed and the combination of endotoxin, carbon particles, and O3
22 produced an increase in oxidant release in old rats. This combination produced the opposite
23 response in the cells recovered from the lungs of the young rats, indicating that the lungs of the
24 aged animals underwent greater oxidative stress in response to this complex pollutant mix of
25 particles, O3, and a biogenic agent.
26 Wagner et al. (2001) examined the synergistic effect of co-exposure to O3 and endotoxin on
27 the transition and respiratory epithelium of rats that also was mediated, in part, by neutrophils.
28 Fisher 344 rats (10 to 12 week old) exposed to 0.5 ppm O3, 8 h per day, for 3 days, developed
29 mucous cell metaplasia in the nasal transitional epithelium, an area normally devoid of mucous
30 cells; whereas, intratracheal instillation of endotoxin (20 //g) caused mucous cell metaplasia
31 rapidly in the respiratory epithelium of the conducting airways. A synergistic increase of
April 2002 7-101 DRAFT-DO NOT QUOTE OR CITE
-------
1 intraepithelial mucosubstances and morphological evidence of mucous cell metaplasia were
2 found in rat maxilloturbinates upon exposure to both ozone and endotoxin, compared to each
3 pollutant alone.
4 The effects of O3 modifying the biological potency of PM (diesel PM and carbon black)
5 was examined by Madden et al. (2000). Reaction of NIST Standard Reference Material # 2975
6 diesel PM with 0.1 ppm O3 for 48 hr increased the potency (compared to unexposed or
7 air-exposed diesel PM) to induce neutrophil influx, total protein, and LDH in lung lavage fluid in
8 response to intratracheal instillation. Exposure of the diesel PM to high, non-ambient O3
9 concentration (1.0 ppm) attenuated the increased potency, suggestion destruction of the bioactive
10 reaction products. Unlike the diesel particles, carbon black particles exposed to 0.1 ppm O3 did
11 not exhibit an increase in biological potency, which suggested that the reaction of organic
12 components of the diesel PM with O3 were responsible for the increased potency. Reaction of
13 particle components with O3 was ascertained by chemical determination of specific classes of
14 organic compounds.
15 The interaction of PM and O3 was further examined in a murine model of ovalbumin
16 (OVA)-induced asthma. Kobzik et al. (2001) investigated whether coexposure to inhaled,
17 concentrated PM from Boston, MA and to O3 could exacerbate asthma-like symptoms. On days
18 7 and 14 of life, half of the BALB/c mice used in this study were sensitized by ip injection of
19 OVA and then exposed to OVA aerosol on three successive days to create the asthma phenotype.
20 The other half received the ip OVA, but were exposed to a phosphate-buffered saline aerosol
21 (controls). The mice were further subdivided (n >6I/group) and exposed for 5 h to CAPs,
22 ranging from 63 to 1,569 //g/m3, 0.3 ppm O3, CAPs + O3, or to filtered air. Pulmonary resistance
23 and airway responsiveness to an aerosolized MCh challenge were measured after exposures. A
24 small, statistically significant increase in pulmonary resistance and airway responsiveness,
25 respectively, was found in both normal and asthmatic mice immediately after exposure to CAPs
26 alone and to CAPs + O3, but not to O3 alone or to filtered air. By 24 h after exposure, the
27 responses returned to baseline levels. There were no significant increases in airway
28 inflammation after any of the pollutant exposures. In this well-designed study of a small-animal
29 model of asthma, O3 and CAPs did not appear to be synergistic. In further analysis of the data
30 using specific elemental groupings of the CAPs, the acutely increased pulmonary resistance was
31 found to be associated withe the AISi fraction of PM. Thus, some components of concentrated
April 2002 7-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM2 5 may affect airway caliber in sensitized animals, but the results are difficult to extrapolate to
2 people with asthma.
3 Linn and colleagues (1997) examined the effect of a single exposure to 60 to 140 //g/m3
4 H2SO4, 0.1 ppm SO2, and 0.1 ppm O3 in healthy and asthmatic children. The children performed
5 intermittent exercise during the 4-h exposure to increase the inhaled dose of the pollutants. An
6 overall effect on the combined group of healthy and asthmatic children was not observed. A
7 positive association between acid concentration and symptoms was seen, however, in the
8 subgroup of asthmatic children. The combined pollutant exposure had no effect on spirometry in
9 asthmatic children, and no changes in symptoms or spirometry were observed in healthy children.
10 Thus, the effect of combined exposure to PM and gaseous co-pollutants appeared to have less
11 effect on asthmatic children exposed under controlled laboratory conditions in comparison with
12 field studies of children attending summer camp (Thurston et al., 1997). However, prior
13 exposure to H2SO4 aerosol may enhance the subsequent response to O3 exposure (Linn et al.,
14 1994; Frampton et al., 1995); and the timing and sequence of the exposures may be important.
15 Six unique animal studies have examined the adverse cardiopulmonary effects of complex
16 mixtures in urban and rural environments of Italy (Gulisano et al., 1997), Spain (Lorz and Lopez,
17 1997), and Mexico (Vanda et al., 1998; Calderon-Garciduefias et al., 2001c,d; Moss et al., 2001).
18 Five of these studies, identified in Table 7-12, have taken advantage of the differences in
19 pollutant mixtures of urban and rural environments to report primarily morphological changes in
20 the nasopharynx and lower respiratory tract (Gulisano et al., 1997; Lorz and Lopez, 1997;
21 Calderon-Garciduefias et al., 2001c) and in the heart (Calderon-Garciduefias et al., 2001d) of
22 lambs, pigeons, and dogs, respectively, after natural, continuous exposures to ambient pollution.
23 Each study has provided evidence that animals living in urban air pollutants have greater
24 pulmonary and cardiac changes than would occur in a rural and presumably cleaner,
25 environment. The study by Moss et al. (2001) examined the nasal and lung tissue of rats exposed
26 (23 h/day) to Mexico City air for up to 7 weeks and compared them to controls similarly exposed
27 to filtered air. No inflammatory or epithelial lesions were found using quantitative
28 morphological techniques; however, the concentrations of pollutants were low (see Table 7-12).
29 Extrapolation of these results to humans is restricted, however, by uncontrolled exposure
30 conditions, small sample sizes, and other unknown exposure and nutritional factors in the studies
31 in mammals and birds, and the negative studies in rodents. They also bring up the issue of which
April 2002 7-103 DRAFT-DO NOT QUOTE OR CITE
-------
1 species of "sentinel" animals is more useful for predicting urban pollutant effects in humans.
2 Thus, in these field studies, it is difficult to assign a specific role to PM (or to any other
3 component of the mixture) in the significant cardiopulmonary effects reported.
4 Similar morphological changes (Calderon-Garciduefias et al., 2000a; 2001a,b) and chest
5 X-ray evidence of mild lung hyperinflation (Calderon-Garciduefias et al., 2000b) have been
6 reported in children residing in urban and rural areas of Mexico City. The ambient air in urban
7 areas, particularly in Southwest Metropolitan Mexico City (SWMMC), is a complex mixture of
8 particles and gases, including high concentrations of O3 and aldehydes that previously have been
9 shown to cause airway inflammation and epithelial lesions in humans (e.g., Calderon-
10 Garciduefias et al., 1992, 1994, 1996) and laboratory animals (Morgan et al., 1986; Heck et al.,
11 1990; Harkema et al., 1994, 1997a,b). The described effects demonstrate a persistent, ongoing
12 upper and lower airway inflammatory process and chest X-ray abnormalities in children residing
13 predominantly in SWMMC. Again, extrapolation of these results to urban populations of the
14 United States is difficult because of the unique complex of urban air in Mexico City,
15 uncontrolled exposure conditions, and other unknown exposure and nutritional factors.
16 Only one controlled study has examined the effect of a combined inhalation exposure to
17 CAPs and O3 in human subjects. In a randomized, double-blind crossover study, Brook et al.
18 (2002) exposed 25 healthy male and female subjects, 34.9 ±10 (SD) years of age, to filtered
19 ambient air containing 1.6 //g/m3 PM2 5 and 9 ppb O3 (control) or to unfiltered air containing
20 150 //g/m3 CAPs and 120 ppb O3 while at rest for 2 h. Blood pressure was measured and high-
21 resolution brachial artery ultrasonography (BAUS) was performed prior to and 10 min after
22 exposure. The BAUS technique was used to measure brachial artery diameter (BAD),
23 endothelium-dependent flow-mediated dilation (FMD), and endothelial-independent
24 nitroglycerine-mediated dilation (NMD). Although no changes in blood pressure or endothelial-
25 dependent or independent dilatation were observed, a small (2.6%) but statistically significant
26 (p = 0.007) decrease in BAD was observed in CAPs plus O3 exposures (-0.09 mm) when
27 compared to filtered air exposures (+0.01 mm). Pre-exposure BAD showed no significant day-
28 to-day variation (0.03 mm), and no significant exposure differences were found for other gaseous
29 pollutants (CO, NOX, SO2) in the ambient air. This finding suggests that combined exposure to a
30 mixture of CAPs and O3 produces vasoconstriction, potentially via autonomic reflexes or as a
31 result of an increase in circulating endothelin, as has been described in rats exposed to urban PM
April 2002 7-104 DRAFT-DO NOT QUOTE OR CITE
-------
1 (Vincent et al., 2001). It is not known, however, whether this effect is caused by CAPS or O3
2 alone, or if vasoactive responses would be found at lower PM25 and O3 concentrations typically
3 found in most urban locations in North America.
4 The effects of gaseous pollutants on PM-mediated responses also have been examined by in
5 vitro studies, though to a limited extent. Churg et al. (1996) demonstrated increased uptake of
6 asbestos or TiO2 into rat tracheal explant cultures in response to 10 min O3 (up to 1.0 ppm) pre-
7 exposure. These data suggest that low concentrations O3 may increase the penetration of some
8 types of PM into epithelial cells. Additionally, Madden et al. (2000) demonstrated a greater
9 potency for ozonized diesel PM to induce prostaglandin E2 production from human epithelial cell
10 cultures, suggesting that O3 can modify the biological activity of PM derived from diesel exhaust.
11 No effect of NO2 exposure on PM-induced interleukin-8 production by A549 epithelial cell
12 line was found (Dick et al., 2001). The PM10 used in this study was collected from gas stoves.
13
14
15 7.7 SUMMARY
16 7.7.1 Biological Plausibility
17 Toxicological studies can play an integral role in answering the following two key
18 questions regarding biological plausibility of PM health effects.
19 (1) What component (or components) of ambient PM cause health effects?
20 (2) Are the statistical associations between PM and health effects biologically plausible?
21 This summary focuses on the progress that toxicological studies have made towards answering
22 these questions.
23
24 7.7.1.1 Link Between Specific Particulate Matter Components and Health Effects
25 Key to the validity of the biological plausibility is the need to understand the linkage
26 between the components of airborne PM responsible for the adverse effects and the individuals at
27 risk. The plausibility of the association between PM and increases in morbidity and mortality has
28 been questioned because the adverse cardiopulmonary effects have been observed at very low
29 PM concentrations, often below the current NAAQS for PM10. To date, toxicology studies on
30 PM have provided only very limited evidence for specific PM components being responsible for
April 2002 7-105 DRAFT-DO NOT QUOTE OR CITE
-------
1 observed cardiopulmonary effects of ambient PM. Studies have shown that some components of
2 particles are more toxic than others. For example, high concentrations of ROFA and associated
3 soluble metals have produced clinically significant effects (including death) in compromised
4 animals. The relevance of these findings to understanding the adverse effects of PM components
5 is tempered, however, by the large difference between metal concentrations delivered to the test
6 animals and metal concentrations present in the ambient urban environment. Such comparisons
7 must be applied to the interpretation of all studies that examine the individual components of
8 ambient urban PM. A summary of potential contributions of individual physical/chemical factors
9 of particles to cardiopulmonary effects is given below.
10
11 Acid Aerosols
12 There is relatively little new information on the effects of acid aerosols, and the conclusions
13 of the 1996 PM AQCD are unchanged. It was previously concluded that acid aerosols cause
14 little or no change in pulmonary function in healthy subjects, but asthmatics may develop small
15 changes in pulmonary function. This conclusion is supported by the recent study of Linn and
16 colleagues (1997) in which children (26 children with allergy or asthma and 15 healthy children)
17 were exposed to sulfuric acid aerosol (100 //g/m3) for 4 h. There were no significant effects on
18 symptoms or pulmonary function when data from the entire group was analyzed, but the allergy
19 group had a significant increase in symptoms after the acid aerosol exposure.
20 Although pulmonary effects of acid aerosols have been the subject of extensive research in
21 past decades, the cardiovascular effects of acid aerosols have received little attention. Zhang
22 et al. (1997) reported that inhalation of acetic acid fumes caused reflex-mediated increases in
23 blood pressure in normal and spontaneously hypertensive rats. Thus, acid components should
24 not be ruled out as possible mediators of PM health effects. In particular, the cardiovascular
25 effects of acid aerosols at realistic concentrations need further investigation.
26
27 Metals
28 The previous PM AQCD (U.S. Environmental Protection Agency, 1996a) mainly relied on
29 data related to occupational exposures to evaluate the potential toxicity of metals in particulate
30 air pollution. Since that time, in vivo and in vitro studies using ROFA or soluble transition
31 metals have contributed substantial new information on the health effects of particle-associated
April 2002 7-106 DRAFT-DO NOT QUOTE OR CITE
-------
1 soluble metals. Although there are some uncertainties about differential effects of one transition
2 metal versus another, water soluble metals leached from ROFA have been shown consistently
3 (albeit at high concentrations) to cause cell injury and inflammatory changes in vitro and in vivo.
4 Even though it is clear that combustion particles that have a high content of soluble metals
5 can cause lung injury and even death in compromised animals, it has not been established that the
6 small quantities of metals associated with ambient PM are sufficient to cause health effects.
7 Moreover, it cannot be assumed that metals are the primary toxic component of ambient PM.
8 In studies in which various ambient and emission source particulates were instilled into rats, the
9 soluble metal content did appear to be the primary determinant of lung injury (Costa and Dreher,
10 1997). However, one published study has compared the effects of inhaled ROFA (at 1 mg/m3) to
11 concentrated ambient PM (four experiments, at mean concentrations of 475 to 900 //g/m3) in
12 normal and SO2-induced bronchitic rats. A statistically significant increase in at least one lung
13 injury marker was seen in bronchitic rats with only one out of four of the concentrated ambient
14 exposures; whereas inhaled ROFA had no effect even though the content of soluble iron,
15 vanadium, and nickel was much higher in the ROFA sample than in the concentrated ambient
16 PM.
17
18 Ultraftne Particles
19 When this subject was reviewed in the 1996 PM AQCD (U. S. Environmental Protection
20 Agency, 1996a), it was not known whether the pulmonary toxicity of freshly generated ultrafme
21 teflon particles was due to particle size or a result of absorbed fumes. Subsequent studies with
22 other types of ultrafme particles have shown that the chemical constituents of ultrafmes
23 substantially modulate their toxicity. For example, Kuschner et al. (1997) have established that
24 inhalation of MgO particles produces far fewer respiratory effects than does ZnO. Also,
25 inhalation exposure of normal rats to ultrafme carbon particles generated by electric arc discharge
26 (100 //g/m3 for 6 h) caused minimal lung inflammation (Elder et al., 2000a,b), compared to
27 ultrafme Teflon or metal particles. On the other hand, instillation of 125 //g of ultrafme carbon
28 black (20 nm) caused substantially more inflammation than did the same dose of fine particles of
29 carbon black (200 to 250 nm), suggesting that ultrafme particles may cause more inflammation
30 than larger particles (Li et al., 1997). However, the chemical constituents of the two sizes of
31 carbon black used in this study were not analyzed, and it cannot be assumed that the chemical
April 2002 7-107 DRAFT-DO NOT QUOTE OR CITE
-------
1 composition was the same for the two sizes. Thus, there is still insufficient toxicological
2 evidence to conclude that ambient concentrations of ultrafme particles contribute to the health
3 effects of paniculate air pollution. With acid aerosols, studies of low concentrations of sulfuric
4 acid ultrafme metal oxide particles have demonstrated effects in the lung. However, it is possible
5 that inhaled ultrafme particles may have systemic effects that are independent of effects on the
6 lung.
7
8 Bioaerosols
9 Recent studies support the conclusion of the 1996 PM AQCD (U. S. Environmental
10 Protection Agency, 1996a), which stated that bioaerosols, at concentrations present in the
11 ambient environment, would not account for the reported health effects of ambient PM.
12 Dose-response studies in healthy volunteers exposed to 0.55 and 50 //g endotoxin, by the
13 inhalation route, showed a threshold for pulmonary and systemic effects for endotoxin between
14 0.5 and 5.0 //g (Michel et al., 1997). Monn and Becker (1999) examined effects of size
15 fractionated outdoor PM on human monocytes and found cytokine induction characteristic of
16 endotoxin activity in the coarse-size fraction but not in the fine fraction. Available information
17 suggests that ambient concentrations of endotoxin are very low and do not exceed 0.5 ng/m3.
18
19 Diesel Exhaust Particles
20 As described in Section 7.2.1.2, there is growing toxicological evidence that diesel PM
21 exacerbates the allergic response to inhaled antigens. The organic fraction of diesel exhaust has
22 been linked to eosinophil degranulation and induction of cytokine production, suggesting that the
23 organic constituents of diesel PM are the responsible part for the immune effects. It is not known
24 whether the adjuvant-like activity of diesel PM is unique or whether other combustion particles
25 have similar effects. It is important to compare the immune effects of other source-specific
26 emissions, as well as concentrated ambient PM, to diesel PM to determine the extent to which
27 exposure to diesel exhaust may contribute to the incidence and severity of allergic rhinitis and
28 asthma.
April 2002 7-108 DRAFT-DO NOT QUOTE OR CITE
-------
1 Organic Compounds
2 Published research on the acute effects of particle-associated organic carbon constituents is
3 conspicuous by its relative absence, except for diesel exhaust particles. Like metals, organics are
4 common constituents of combustion-generated particles and have been found in ambient PM
5 samples over a wide geographical range. Organic carbon constituents comprise a substantial
6 portion of the mass of ambient PM (10 to 60% of the total dry mass [Turpin, 1999]). The
7 organic fraction of ambient PM has been evaluated for its mutagenic effects. Although the
8 organic fraction of ambient PM is a poorly characterized heterogeneous mixture of an unknown
9 number of different compounds, strategies have been proposed for examining the health effects
10 of this potentially important constituent (Turpin, 1999).
11
12 Ambient Particle Studies
13 Ambient particle studies should be the most relevant in understanding the susceptibility of
14 individuals to PM and the underlying mechanisms. Studies have used collected urban PM for
15 intratracheal administration to healthy and compromised animals. Despite the difficulties in
16 extrapolating from the bolus delivery used in such studies, they have provided strong evidence
17 that the chemical composition of ambient particles can have a major influence on toxicity. More
18 recent work with inhaled concentrated ambient PM has observed cardiopulmonary changes in
19 rodents and dogs at high concentrations of fine PM. No comparative studies to examine the
20 effects of ultrafine and coarse ambient PM have been done, although a new ambient particle
21 concentrator developed by Sioutas and colleagues should permit the direct toxicological
22 comparison of various ambient particle sizes. Importantly, it has become evident that, although
23 the concentrated ambient PM studies can provide important dose-response information, identify
24 susceptibility factors in animal models, and permit examination of mechanisms related to PM
25 toxicity, they are not particularly well suited for the identification of toxic components in urban
26 PM. Because only a limited number of exposures using concentrated ambient PM can be
27 reasonably conducted by a given laboratory in a particular urban environment, there may be
28 insufficient information to conduct a factor analysis on an exposure/response matrix. This may
29 also hinder principal component analysis techniques that are useful in identifying particle
30 components responsible for adverse outcomes.
31
April 2002 7-109 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.7.1.2 Susceptibility
2 Progress has been made in understanding the role of individual susceptibility to ambient
3 PM effects. Studies have consistently shown that older animals or animals with certain types of
4 compromised health, either genetic or induced, are more susceptible to instilled or inhaled
5 particles, although the increased animal-to-animal variability in these models has created
6 problems. Moreover, because PM seems to affect broad categories of disease states, ranging
7 from cardiac arrhythmias to pulmonary infection, it can be difficult to know what disease models
8 to use in understanding the biological plausibility of the adverse health effects of PM. Thus, the
9 identification of susceptible animal models has been somewhat slow, but overall it represents
10 solid progress when one considers that data from millions of people are necessary in
11 epidemiology studies to develop the statistical power to detect small increases in PM-related
12 morbidity and mortality.
13
14 7.7.2 Mechanisms of Action
15 The mechanisms that underlie the biological responses to ambient PM are not clear.
16 Various toxicologic studies using particulate matter having diverse physicochemical
17 characteristics have shown that these characteristics have a great impact on the specific response
18 that is observed. Thus, there are multiple biological mechanisms that may be responsible for
19 observed morbidity/mortality due to exposure to ambient PM, and these mechanisms may be
20 highly dependent on the type of particle in the exposure atmosphere. However, it should be
21 noted that many controlled exposure studies used particle concentrations much higher than those
22 typically occurring in ambient air. Thus, some of the mechanisms elicited may not occur with
23 exposure to lower levels. Clearly, controlled exposure studies have not as yet been able to
24 unequivocally determine the particle characteristics and the toxicological mechanisms by which
25 ambient PM may affect biological systems.
26
April 2002 7-110 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Adamson, I. Y. R.; Prieditis, H.; Hedgecock, C.; Vincent. R. (2000) Zinc is the toxic factor in the lung response to
3 an atmospheric paniculate sample. Toxicol. Appl. Pharmacol. 166: 111-119.
4 Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects.
5 In: Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC. Environ.
6 Health Perspect. 79: 147-150.
7 Amdur, M. O.; Dubriel, M.; Creasia, D. A. (1978) Respiratory response of guinea pigs to low levels of sulfuric acid.
8 Environ. Res. 15: 418-423.
9 Ball, B. R.; Smith, K. R.; Veranth, J. M.; Aust, A. E. (2000) Bioavailability of iron from coal fly ash: mechanisms
10 of mobilization and of biological effects. In: Grant, L. D., ed. PM2000: paniculate matter and health.
11 Inhalation Toxicol. 12(suppl. 4): 209-225.
12 Bayram, H.; Devalia, J. L.; Khair, O. A.; Abdelaziz, M. M.; Sapsford, R. J.; Sagai, M.; Davies, R. J. (1998a)
13 Comparison of ciliary activity and inflammatory mediator release from bronchial epithelial cells of nonatopic
14 nonasthmatic subjects and atopic asthmatic patients and the effect of diesel exhaust particles in vitro.
15 J. Allergy Clin. Immunol. 102: 771-782.
16 Bayram, H.; Devalia, J. L.; Sapsford, R. J.; Ohtoshi, T.; Miyabara, Y.; Sagai, M.; Davies, R. J. (1998b) The effect
17 of diesel exhaust particles on cell function and release of inflammatory mediators from human bronchial
18 epithelial cells in vitro. Am. J. Respir. Cell Mol. Biol. 18: 441-448.
19 Becker, S.; Soukup, J. M. (1998) Decreased CD 1 IB expression, phagocytosis, and oxidative burst in urban
20 paniculate pollution-exposed human monocytes and alveolar macrophages. J. Toxicol. Environ. Health Part
21 A 55: 455-477.
22 Becker, S.; Soukup, J. M.; Gilmour, M. L; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
23 by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
24 Pharmacol. 141:637-648.
25 Blomberg, A.; Sainsbury, C.; Rudell, B.; Frew, A. J.; Holgate, S. T.; Sandstrom, T.; Kelly, F. J. (1998) Nasal cavity
26 lining fluid ascorbic acid concentration increases in healthy human volunteers following short term exposure
27 to diesel exhaust. Free Radical Res. 28: 59-67.
28 Bolarin, D. M.; Bhalla, D. K.; Kleinman, M. T. (1997) Effects of repeated exposures of geriatric rats to ozone and
29 particle-containing atmospheres: an analysis of bronchoalveolar lavage and plasma proteins. Inhalation
30 Toxicol. 9: 423-434.
31 Bonner, J. C.; Rice, A. B.; Lindroos, P. M.; O'Brien, P. O.; Dreher, K. L.; Rosas, L; Alfaro-Moreno, E.;
32 Osornio-Vargas, A. R. (1998) Induction of the lung myofibroblast PDGF receptor system by urban ambient
33 particles from Mexico City. Am. J. Respir. Cell Mol. Biol. 19: 672-680.
34 Bouthillier, L.; Vincent, R.; Goegan, P.; Adamson, I. Y. R.; Bjarnason, S.; Stewart, M.; Guenette, J.; Potvin, M.;
3 5 Kumarathasan, P. (1998) Acute effects of inhaled urban particles and ozone: lung morphology, macrophage
36 activity, and plasma endothelin-1. Am. J. Pathol. 153: 1873-1884.
37 Brain, J. D.; Long, N. C.; Wolfthal, S. F.; Dumyahn, T.; Dockery, D. W. (1998) Pulmonary toxicity in hamsters of
38 smoke particles from Kuwaiti oil fires. Environ. Health Perspect. 106: 141-146.
39 Broeckaert, F.; Buchet, J. P.; Huaux, F.; Lardot, C.; Lison, D.; Yager, J. W. (1997) Reduction of the ex vivo
40 production of tumor necrosis factor alpha by alveolar phagocytes after administration of coal fly ash and
41 copper smelter dust. J. Toxicol. Environ. Health 51: 189-202.
42 Brook, R. D.; Brook, J. R.; Urch, B.; Vincent, R.; Rajagopalan, S.; Silverman, F. (2002) Inhalation of fine
43 paniculate air pollution and ozone causes acute arterial vasoconstriction in health adults. Circulation 105:
44 1534-1536.
45 Calderon-Garciduenas, L.; Osorno-Velaquez, A.; Bravo-Alvarez, H.; Delgado-Chavez, R.; Barrios-Marquez, R.
46 (1992) Histopathologic changes of the nasal mucosa in southwest metropolitan Mexico City inhabitants. Am.
47 J. Pathol. 140: 225-232.
48 Calderon-Garciduenas, L.; Rodriguez-Alcaraz, A.; Garcia, R.; Sanchez, G.; Barragan, G.; Camacho, R.; Ramirez, L.
49 (1994) Human nasal mucosal changes after exposure to urban pollution. Environ. Health Perspect. 102:
50 1074-1080.
51 Calderon-Garciduenas, L.; Rodriguez-Alcaraz, A.; Garcia, R.; Ramirez, L.; Barragan, G. (1995) Nasal inflammatory
52 responses in children exposed to a polluted urban atmosphere. J. Toxicol. Environ. Health 45: 427-437.
April 2002 7-111 DRAFT-DO NOT QUOTE OR CITE
-------
1 Calderon-Garciduenas, L.; Osnaya-Brizuela, N.; Ramirez-Martinez, L.; Villarreal-Calderon, A. (1996) DNA strand
2 breaks in human nasal respiratory epithelium are induced upon exposure to urban pollution. Environ. Health
3 Perspect. 104: 160-168.
4 Calderon-Garciduenas, L.; Devlin, R. B.; Miller, F. J. (2000a) Respiratory tract pathology and cytokine imbalance
5 in clinically healthy children chronically and sequentially exposed to air pollutants. Med. Hypotheses 55:
6 373-378.
7 Calderon-Garciduenas, L.; Mora-Tascareno, A.; Chung, C. J.; Valencia, G.; Fordham, L. A.; Garcia, R.; Osnaya,
8 N.; Romero, L.; Acuna, H.; Villarreal-Calderon, A.; Devlin, R. B.; Koren, H. S. (2000b) Exposure to air
9 pollution is associated with lung hyperinflation in healthy children and adolescents in southwest Mexico
10 City: a pilot study. Inhalation Toxicol. 12: 537-561.
11 Calderon-Garciduenas, L.; Valencia-Salazar, G.; Rodriguez-Alcaraz, A.; Gambling, T. M.; Garcia, R.; Osnaya, N.;
12 Villarreal-Calderon, A.; Devlin, R. B.; Carson, J. L. (2001a) Ultrastructural nasal pathology in children
13 chronically and sequentially exposed to air pollutants. Am. J. Respir. Cell Mol. Biol. 24: 132-138.
14 Calderon-Garciduenas, L.; Rodriguez-Alcaraz, A.; Valencia-Salazar, G.; Mora-Tascareno, A.; Garcia, R.; Osnaya,
15 N.; Villarreal-Calderon, A.; Devlin, R. B.; Van Dyke, T. L. (200Ib) Nasal biopsies of children exposed to air
16 pollutants. Toxicol. Pathol. 29: 558-564.
17 Calderon-Garciduenas, L.; Mora-Tiscareno, A.; Fordham, L. A.; Chung, C. J.; Garcia, R.; Osnaya, N.; Hernandez,
18 J.; Acuna, H.; Gambling, T. M.; Villarreal-Calderon, A.; Carson, J.; Koren, H. S.; Devlin, R. B. (200Ic)
19 Canines as sentinel species for assessing chronic exposures to air pollutants: part 1. respiratory pathology.
20 Toxicol. Sci. 61:342-355.
21 Calderon-Garciduenas, L.; Gambling, T. M.; Acuna, H.; Garcia, R.; Osnaya, N.; Monroy, S.; Villarreal-Calderon,
22 A.; Carson, J.; Koren, H. S.; Devlin, R. B. (2001d) Canines as sentinel species for assessing chronic
23 exposures to air pollutants: part 2. cardiac pathology. Toxicol. Sci. 61: 356-367.
24 Campen, M. J.; Costa, D. L.; Watkinson, W. P. (2000) Cardiac and thermoregulatory toxicity of residual oil fly ash
25 in cardiopurmonary-compromised rats. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the third
26 colloquium on paniculate air pollution and human health (second special issue); June, 1999; Durham, NC.
27 Inhalation Toxicol. 12(suppl. 2): 7-22.
28 Campen, M. J.; Nolan, J. P.; Schladweiler, M. C. J.; Kodavanti, U. P.; Evansky, P. A.; Costa, D. L.; Watkinson, W.
29 P. (2001) Cardiovascular and thermoregulatory effects of inhaled PM-associated transition metals: a potential
30 interaction between nickel and vanadium surf ate. Toxicol. Sci. 64: 243-252.
31 Campen, M. J.; Nolan, J. P.; Schladweiler, M. C. J.; Kodavanti, U. P.; Costa, D. L.; Watkinson, W. P. (2002)
32 Cardiac and thermoregulatory effects of instilled paniculate matter-associated transition metals in healthy and
33 cardiopurmonary-compromised rats. J. Toxicol. Environ. Health (special issue): in press.
34 Carter, J. D.; Ohio, A. J.; Samet, J. M.; Devlin, R. B. (1997) Cytokine production by human airway epithelial cells
3 5 after exposure to an air pollution particle is metal-dependent. Toxicol. Appl. Pharmacol. 146: 180-188.
36 Castellan, R. M.; Olenchock, S. A.; Hankinson, J. L.; Millner, P. D.; Cocke, J. B.; Bragg, C. K.; Perkins, H. H., Jr.;
37 Jacobs, R. R. (1984) Acute bronchoconstriction induced by cotton dust: dose-related responses to endotoxin
38 and other dust factors. Ann. Intern. Med. 102: 157-163.
39 Castellan, R. M.; Olenchock, S. A.; Kinsley, K. B.; Hankinson, J. L. (1987) Inhaled endotoxin and decreased
40 spirometric values: an exposure-response relation for cotton dust. N. Engl. J. Med. 317: 605-610.
41 Chang, M. C.; Sioutas, C.; Kim, S.; Gong, H., Jr.; Linn, W. S. (2000) Reduction of nitrate losses from filter and
42 impactor samplers by means of concentration enrichment. Atmos. Environ. 34: 85-98.
43 Chen, L. C.; Fine, J. M.; Qu, Q.-S.; Amdur, M. O.; Gordon, T. (1992) Effects of fine and ultrafine sulfuric acid
44 aerosols in guinea pigs: alterations in alveolar macrophage function and intracellular pH. Toxicol. Appl.
45 Pharmacol. 113: 109-117.
46 Chen, L. C.; Fang, C. P.; Qu, Q. S.; Fine, J. M.; Schlesinger, R. B. (1993) A novel system for the in vitro exposure
47 of pulmonary cells to acid sulfate aerosols. Fundam. Appl. Toxicol. 20: 170-176.
48 Churg, A.; Brauer, M.; Keeling, B. (1996) Ozone enhances the uptake of mineral particles by tracheobronchial
49 epithelial cells in organ culture. Am. J. Respir. Crit. Care Med. 153: 1230-1233.
50 Churg, A.; Stevens, B.; Wright, J. L. (1998) Comparison of the uptake of fine and ultrafine TiO2 in a tracheal
51 explant system. Am. J. Physiol. 274: L81-L86.
52 Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.; Sioutas, C.; Paulauskis, J.; Coull, B.;
53 Ferguson, S.; Godleski, J. J. (1999) Urban air paniculate inhalation alters pulmonary function and induces
54 pulmonary inflammation in a rodent model of chronic bronchitis. Inhalation Toxicol. 11: 637-656.
April 2002 7-112 DRAFT-DO NOT QUOTE OR CITE
-------
1 Clarke, R. W.; Coull, B.; Reinisch, U.; Catalano, P.; Killingsworth, C. R.; Koutrakis, P.; Kavouras, I.; Murthy, G.
2 G. K.; Lawrence, J.; Lovett, E.; Wolf son, J. M; Verrier, R. L.; Godleski, J. J. (2000a) Inhaled concentrated
3 ambient particles are associated with hematologic and bronchoalveolar lavage changes in canines. Environ.
4 Health Perspect. 108: 1179-1187.
5 Clarke, R. W.; Catalano, P.; Coull, B.; Koutrakis, P.; Krishna Murthy, G. G.; Rice, T.; Godleski, J. J. (2000)
6 Age-related responses in rats to concentrated urban air particles (CAPs). In: Phalen, R. F., ed. Inhalation
7 Toxicology: proceedings of the third colloquium on paniculate air pollution and human health (first special
8 issue); June, 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 73-84.
9 Cohen, M. D.; Becker, S.; Devlin, R.; Schlesinger, R. B.; Zelikoff, J. T. (1997) Effects of vanadium upon
10 polyl:C-induced responses in rat lung and alveolar macrophages. J. Toxicol. Environ. Health 51: 591-608.
11 Cormier, Y.; Laviolette, M.; Bedard, G.; Dosman, J.; Israel-Assayag, E. (1998) Effect of route of breathing on
12 response to exposure ina swine confinement building. Am. J. Respir. Crit. Care Med. 157: 1512-1521.
13 Costa, D. L.; Dreher, K. L. (1997) Bioavailable transition metals in paniculate matter mediate cardiopulmonary
14 injury in healthy and compromised animal models. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of
15 the sixth international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
16 September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1053-1060.
17 Creutzenberg, O.; Bellmann, B.; Muhle, H.; Dasenbrock, C.; Morrow, P.; Mermelstein, R. (1998) Lung clearance
18 and retention of toner, TiO2, and crystalline silica, utilizing a tracer technique during chronic inhalation
19 exposure in Syrian Golden hamsters. Inhalation Toxicol. 10:731-751.
20 Diaz-Sanchez, D.; Tsien, A.; Casillas, A.; Dotson, A. R.; Saxon, A. (1996) Enhanced nasal cytokine production in
21 human beings after in vivo challenge with diesel exhaust particles. J. Allergy Clin. Immunol. 98: 114-123.
22 Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997) Combined diesel exhaust paniculate and ragweed
23 allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
24 production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
25 Dick, C. A. J.; Dennekamp, M.; Howarth, S.; Cherrie, J. W.; Seaton, A.; Donaldson, K.; Stone, V.; (2001)
26 Stimulation of IL-8 release from epithelial cells by gas cooker PM10: a pilot study. Occup. Environ. Med. 58:
27 208-210.
28 Diociaiuti, M.; Balduzzi, M.; De Berardis, B.; Cattani, G.; Stacchini, G.; Ziemacki, G.; Marconi, A.; Paoletti, L.
29 (2001) The two PM2 5 (fine) and PM2 5-10 (coarse) fractions: evidence of different biological activity. Environ.
30 Res. A 86: 254-262.
31 Donaldson, K.; Brown, D. M.; Mitchell, C.; Dineva, M.; Beswick, P. H.; Gilmour, P.; MacNee, W. (1997) Free
32 radical activity of PM10: iron-mediated generation of hydroxyl radicals. In: Driscoll, K. E.; Oberdorster, G.,
3 3 eds. Proceedings of the sixth international meeting on the toxicology of natural and man-made fibrous and
34 non-fibrous particles; September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5):
35 1285-1289.
36 Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor aexpression from alveolar
37 macrophages treated with urban air particles. Exp. Lung Res. 22: 577-592.
38 Dormans, J. A. M. A.; Steerenberg, P. A.; Arts, J. H. E.; VanBree, L.; De Klerk, A.; Verlaan, A. P. J.; Bruijntjes,
39 J. P.; Beekhof, P.; Van Soolingen, D.; Van Loveren, H. (1999) Pathological and immunological effects of
40 respirable coal fly ash in male wistar rats. Inhalation Toxicol. 11: 51-69.
41 Dreher, K. L.; Jaskot, R. H.; Lehmann, J. R.; Richards, J. H.; McGee, J. K.; Ohio, A. J.; Costa, D. L. (1997)
42 Soluable transition metals mediate residual oil fly ash induced acute lung injury. J. Toxicol. Environ. Health
43 50:285-305.
44 Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
45 Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
46 limitations. Toxicol. Sci. 55: 24-35.
47 Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1997) Epithelial injury induced by exposure to residual oil
48 fly-ash particles: role of reactive oxygen species? Am. J. Respir. Cell Mol. Biol. 17: 625-633.
49 Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1999) Role of soluble metals in oil fly ash-induced airway
50 epithelial injury and cytokine gene expression. Am. J. Physiol. 277: L498-L510.
51 Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. J.; Costa, D. L.
52 (2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
53 studies in Utah Valley residents. Environ. Health Perspect. 109(suppl. 3): 395-403.
54 Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000a) Endotoxin priming affects the lung
55 response to ultrafine particles and ozone in young and old rats. In: Phalen, R. F., ed. Inhalation Toxicology:
April 2002 7-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 proceedings of the third colloquium on paniculate air pollution and human health (first special issue); June,
2 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 85-98.
3 Elder, A. C. P.; Gelein, R.; Finkelstein, J. N.; Cox, C.; Oberdorster, G. (2000b) Pulmonary inflammatory response
4 to inhaled ultrafine particles is modified by age, ozone exposure, and bacterial toxin. In: Grant, L. D., ed.
5 PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 227-246.
6 Fabiani, R.; Pampanella, L.; Minelli, A.; Mezzasoma, L; Vecchiarelli, A.; Morozzi, G. (1997) Effect of airborne
7 paniculate extracts onmonocyte oxidative metabolism. J. Environ. Pathol. Toxicol. Oncol. 16: 195-199.
8 Ferin, J.; Oberdorster, G.; Penney, D. P.; Soderholm, S. C.; Gelein, R.; Piper, H. C. (1990) Increased pulmonary
9 toxicity of ultrafine particles? I. Particle clearance, translocation, morphology. J. Aerosol Sci. 21: 381-384.
10 Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
11 Respir. Cell Mol. Biol. 6: 535-542.
12 Finkelstein, J. N.; Johnston, C.; Barrett, T.; Oberdorster, G. (1997) Particulate-cell interactions and pulmonary
13 cytokine expression. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
14 on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
15 NY. Environ. Health Perspect. Suppl. 105(5): 1179-1182.
16 Folinsbee, L. J.; Kim, C. S.; Kehrl, H. R.; Prah, J. D.; Devlin, R. B. (1997) Methods in human inhalation toxicology.
17 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press LLC; pp. 607-670.
18 Fortoul, T. L; Osorio, L. S.; Tovar, A. T.; Salazar, D.; Castilla, M. E.; Olaiz-Fernandez, G. (1996) Metals in lung
19 tissue from autopsy cases in Mexico City residents: comparison of cases from the 1950s and the 1980s.
20 Environ. Health Perspect. 104: 630-632.
21 Frampton, M. W.; Voter, K. Z.; Morrow, P. E.; Roberts, N. J., Jr.; Culp, D. J.; Cox, C.; Utell, M. J. (1992) Sulfuric
22 acid aerosol exposure in humans assessed by bronchoalveolar lavage. Am. Rev. Respir. Dis. 146: 626-632.
23 Frampton, M. W.; Morrow, P. E.; Cox, C.; Levy, P. C.; Condemi, J. J.; Speers, D.; Gibb, F. R.; Utell, M. J. (1995)
24 Sulfuric acid aerosol followed by ozone exposure in healthy and asthmatic subjects. Environ. Res. 69: 1-14.
25 Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
26 extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol.
27 277: L960-L967.
28 Fujimaki, H.; Katayama, N.; Wakamori, K. (1992) Enhanced histamine release from lung mast cells of guinea pigs
29 exposed to sulfuric acid aerosols. Environ. Res. 58: 117-123.
30 Gardner, S. Y.; Lehmann, J. R.; Costa, D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in
31 rats. Toxicol. Sci. 56: 175-180.
32 Gavett, S. H.; Madison, S. L.; Dreher, K. L.; Winsett, D. W.; McGee, J. K.; Costa, D. L. (1997) Metal and sulfate
33 composition of residual oil fly ash determines airway hyperreactivity and lung injury in rats. Environ. Res.
34 72: 162-172.
35 Gavett, S. H.; Madison, S. L.; Stevens, M. A.; Costa, D. L. (1999) Residual oil fly ash amplifies allergic cytokines,
36 airway responsiveness, and inflammation in mice. Am. J. Respir. Crit. Care Med. 160: 1897-1904.
37 Gearhart, J. M.; Schlesinger, R. B. (1986) Sulfuric acid-induced airway hyperresponsiveness. Fundam. Appl.
38 Toxicol. 7: 681-689.
39 Gercken, G.; Berg, I.; Dorger, M.; Schluter, T. (1996) Mechanisms of particle-induced activation of alveolar
40 macrophages. Toxicol. Lett. 88: 121-129.
41 Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles. Am.
42 J. Respir. Crit. Care Med. 164: 704-708.
43 Ohio, A. J.; Meng, Z. H.; Hatch, G. E.; Costa, D. L. (1997a) Luminol-enhanced chemiluminescence after in vitro
44 exposures of rat alveolar macrophages to oil fly ash is metal dependent. Inhalation Toxicol. 9: 255-271.
45 Ohio, A. J.; Piantadosi, C. A.; Crumbliss, A. L. (1997b) Hypothesis: iron chelation plays a vital role in neutrophilic
46 inflammation. BioMetals 10: 135-142.
47 Ohio, A. J.; Carter, J. D.; Richards, J. H.; Brighton, L. E.; Lay, J. C.; Devlin, R. B. (1998a) Disruption of normal
48 iron homeostasis after bronchial instillation of an iron-containing particle. Am. J. Physiol. 274: L396-L403.
49 Ohio, A. J.; Richards, J. H.; Dittrich, K. L.; Samet, J. M. (1998b) Metal storage and transport proteins increase after
50 exposure of the rat lung to an air pollution particle. Toxicol. Pathol. 26: 388-394.
51 Ohio, A. J.; Carter, J. D.; Samet, J. M.; Reed, W.; Quay, J.; Dailey, L. A.; Richards, J. H.; Devlin, R. B. (1998c)
52 Metal-dependent expression of ferritin and lactoferrin by respiratory epithelial cells. Am. J. Physiol.
53 274: L728-L736.
April 2002 7-114 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ohio, A. J.; Stonehuerner, J.; Dailey, L. A.; Carter, J. D. (1999a) Metals associated with both the water-soluble and
2 insoluble fractions of an ambient air pollution particle catalyze an oxidative stress. Inhalation Toxicol. 11: 37-49.
3 Ohio, A. J.; Carter, J. D.; Dailey, L. A.; Devlin, R. B.; Samet, J. M. (1999b) Respiratory epithelial cells demonstrate
4 lactoferrin receptors that increase after metal exposure. Am. J. Physiol. 276: L933-L940.
5 Ohio, A. J.; Stoneheurner, J.; McGee, J. K.; Kinsey, J. S. (1999c) Sulfate content correlates with iron concentrations
6 in ambient air pollution particles. Inhalation Toxicol. 11: 293-307.
7 Ohio, A. J.; Kim, C.; Devlin, R. B. (2000a) Concentrated ambient air particles induce mild pulmonary inflammation
8 in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
9 Ohio, A. J.; Carter, J. D.; Richards, J. H.; Crissman, K. M.; Bobb, H. H.; Yang, F. (2000b) Diminished injury in
10 hypotransferrinemic mice after exposure to a metal-rich particle. Am. J. Physiol. 278: L1051-L1061.
11 Ohio, A. J.; Gilbey, J. G.; Roggli, V. L.; Richards, J. H.; McGee, J. K.; Carson, J. L.; Devlin, R. B.; Cascio, W. E.
12 (2001) Diffuse alveolar damage after exposure to an oil fly ash. Am. J. Respir. Crit. Care Med. 164:
13 1514-1518.
14 Gift, J. S.; Faust, R. A. (1997) Noncancer inhalation toxicology of crystalline silica: exposure-response assessment.
15 J. Exposure Anal. Environ. Epidemiol. 7: 345-358.
16 Gilmour, M. L; Taylor, F. G. R.; Baskerville, A.; Wathes, C. M. (1989a) The effect of titanium dioxide inhalation
17 on the pulmonary clearance of Pasteurella haemolytica in the mouse. Environ. Res. 50: 157-172.
18 Gilmour, M. L; Taylor, F. G. R.; Wathes, C. M. (1989b) Pulmonary clearance of Pasteurella haemolytica and
19 immune responses in mice following exposure to titanium dioxide. Environ. Res. 50: 184-194.
20 Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
21 effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
22 53:817-822.
23 Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms of morbidity and mortality from
24 exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
25 Goldsmith, C.-A.; Frevert, C.; Imrich, A.; Sioutas, C.; Kobzik, L. (1997) Alveolar macrophage interaction with air
26 pollution particulates. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
27 on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
28 NY. Environ. Health Perspect. Suppl. 105(5): 1191-1195.
29 Goldsmith, C.-A. W.; Imrich, A.; Danaee, H.; Ning, Y. Y.; Kobzik, L. (1998) Analysis of air pollution
30 particulate-mediated oxidant stress in alveolar macrophages. J. Toxicol. Environ. Health Part A 54: 529-545.
31 Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
32 (1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
33 Inhalation Toxicol. 11: 981-998.
34 Gong, H., Jr.; Sioutas, C.; Linn, W. S.; Clark, K. W.; Terrell, S. L.; Terrell, L. L.; Anderson, K. R.; Kim, S.; Chang,
3 5 M.-C. (2000) Controlled human exposures to concentrated ambient fine particles in metropolitan Los
36 Angeles: methodology and preliminary health-effect findings. In: Phalen, R. F., ed. Inhalation Toxicology:
3 7 proceedings of the third colloquium on paniculate air pollution and human health (first special issue); June,
38 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 107-119.
39 Gordon, T.; Nadziejko, C.; Schlesinger, R.; Chen, L. C. (1998) Pulmonary and cardiovascular effects of acute
40 exposure to concentrated ambient paniculate matter in rats. Toxicol. Lett. 96-97: 285-288.
41 Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
42 hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
43 Grabowski, G. M.; Paulauskis, J. D.; Godleski, J. J. (1999) Mediating phosphorylation events in the
44 vanadium-induced respiratory burst of alveolar macrophages. Toxicol. Appl. Pharmacol. 156: 170-178.
45 Gulisano, M.; Marceddu, S.; Barbara, A.; Pacini, A.; Buiatti, E.; Martini, A.; Pacini, P. (1997) Damage to the
46 nasopharyngeal mucosa induced by current levels of urban air pollution: a field study in lambs. Eur. Respir.
47 J. 10: 567-572.
48 Hahon, N.; Booth, J. A.; Green, F.; Lewis, T. R. (1985) Influenza virus infection in mice after exposure to coal dust
49 and diesel engine emissions. Environ. Res. 37: 44-60.
50 Hamada, K.; Goldsmith, C.-A.; Kobzik, L. (1999) Increased airway hyperresponsiveness and inflammation in a
51 juvenile mouse model of asthma exposed to air-pollutant aerosol. J. Toxicol. Environ. Health 58: 129-143.
52 Hamada, K.; Goldsmith, C.-A.; Goldman, A.; Kobzik, L. (2000) Resistance of very young mice to inhaled allergen
53 sensitization Is overcome by coexposure to an air-pollutant aerosol. Am. J. Respir. Crit. Care Med.
54 161: 1285-1293.
April 2002 7-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 Harder, S. D.; Soukup, J. M; Ohio, A. J.; Devlin, R. B.; Becker, S. (2001) Inhalation of PM25 does not modulate
2 host defense or immune parameters in blood or lung of normal human subjects. Environ. Health Perspect.
3 109(suppl. 4): 599-604.
4 Harkema, J. R.; Morgan, K. T.; Gross, E. A.; Catalano, P. J.; Griffith, W. C. (1994) Consequences of prolonged
5 inhalation of ozone on F344/N rats: collaborative studies. Part VII: effects on the nasal mucociliary
6 apparatus. Cambridge, MA: Health Effects Institute; research report no. 65.
7 Harkema, J. R.; Barr, E. B.; Hotchkiss, J. A. (1997a) Responses of rat nasal epithelium to short- and long-term
8 exposures of ozone: image analysis of epithelial injury, adaptation and repair. Microsc. Res. Tech. 36:
9 276-286.
10 Harkema, J. R.; Hotchkiss, J. A.; Griffith, W. C. (1997b) Mucous cell metaplasia in rat nasal epithelium after a
11 20-month exposure to ozone: a morphometric study of epithelial differentiation. Am. J. Respir. Cell Mol.
12 Biol. 16: 521-530.
13 Health Effects Institute. (1995) Diesel exhaust: a critical analysis of emissions, exposure, and health effects:
14 a special report of the Institute's Diesel Working Group. Cambridge, MA: Health Effects Institute.
15 Heck, H. d'A.; Casanova, M.; Starr, T. B. (1990) Formaldehyde toxicity~new understanding. Crit. Rev. Toxicol. 20:
16 397-426.
17 Heyder, J.; Beck-Speier, I.; Busch, B.; Dirscherl, P.; Heilmann, P.; Perron, G. A.; Josten, M.; Karg, E.; Kreyling,
18 W. G.; Lenz, A.-G.; Maier, K. L.; Miaskowski, U.; Platz, S.; Reitmeir, P.; Schulz, H.; Takenaka, S.;
19 Ziesenis, A. (1999) Health effects of sulfur-related environmental air pollution. I. Executive summary.
20 Inhalation Toxicol. 11: 343-359.
21 Hitzfeld, B.; Friedrichs, K. H.; Ring, J.; Behrendt, H. (1997) Airborne paniculate matter modulates the production
22 of reactive oxygen species in human polymorphonuclear granulocytes. Toxicology 120: 185-195.
23 Holian, A.; Hamilton, R. F., Jr.; Morandi, M. T.; Brown, S. D.; Li, L. (1998) Urban particle-induced apoptosis and
24 phenotype shifts in human alveolar macrophages. Environ. Health Perspect. 106: 127-132.
25 Hunter, R. J. (1981) Zeta potential in colloid science: principles and applications. Orlando, FL: Academic Press, pp.
26 1-124. [Ottewill, R. H.; Rowell, R. L., eds. Colloid Science].
27 Imrich, A.; Ning, Y. Y.; Kobzik, L. (2000) Insoluble components of concentrated air particles mediate alveolar
28 macrophage responses in vitro. Toxicol. Appl. Pharmacol. 167: 140-150.
29 Irsigler, G. B.; Visser, P. J.; Spangenberg, P. A. L. (1999) Asthma and chemical bronchitis in vanadium plant
30 workers. Am. J. Ind. Med. 35: 366-374.
31 Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
32 1983-1992. Environ. Health Perspect. 105: 980-985.
33 Jakab, G. J. (1992) Relationship between carbon black particulate-bound formaldehyde, pulmonary antibacterial
34 defenses, and alveolar macrophage phagocytosis. Inhalation Toxicol. 4: 325-342.
3 5 Jakab, G. J. (1993) The toxicologic interactions resulting from inhalation of carbon black and acrolein on
36 pulmonary antibacterial and antiviral defenses. Toxicol. Appl. Pharmacol. 121: 167-175.
37 Jakab, G. J.; Hemenway, D. R. (1993) Inhalation coexposure to carbon black and acrolein suppresses alveolar
3 8 macrophage phagocytosis and TNF-a release and modulates peritoneal macrophage phagocytosis. Inhalation
39 Toxicol. 5: 275-289.
40 Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
41 acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
42 Jiang, N.; Dreher, K. L.; Dye, J. A.; Richards, J. H.; Martin, L. D.; Adler, K. B. (2000) Residual oil fly ash induces
43 cytotoxicity and mucin secretion by guinea pig tracheal epithelial cells via an oxidant-mediated mechanism.
44 Toxicol. Appl. Pharmacol. 163: 221-230.
45 Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Baggs, R.; Oberdorster, G. (1996) Characterization of the early
46 pulmonary inflammatory response associated with PTFE fume exposure. Toxicol. Appl. Pharmacol.
47 140: 154-163.
48 Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Oberdorster, G. (1998) Pulmonary inflammatory responses and
49 cytokine and antioxidant mRNA levels in the lungs of young and old C57BL/6 mice after exposure to Teflon
50 fumes. Inhalation Toxicol. 10: 931-953.
51 Kadiiska, M. B.; Mason, R. P.; Dreher, K. L.; Costa, D. L.; Ohio, A. J. (1997) In vivo evidence of free radical
52 formation in the rat lung after exposure to an emission source air pollution particle. Chem. Res. Toxicol.
53 10: 1104-1108.
April 2002 7-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 Kennedy, T.; Ohio, A. I; Reed, W.; Samet, I; Zagorski, I; Quay, I; Carter, I; Dailey, L.; Hoidal, J. R.; Devlin,
2 R. B. (1998) Copper-dependent inflammation and nuclear factor-KB activation by paniculate air pollution.
3 Am. J. Respir. Cell Mol. Biol. 19: 366-378.
4 Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
5 (1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly
6 ash inhalation. Inhalation Toxicol. 9: 541-565.
7 Kim, S.; Chang, M.-C.; Kim, D.; Sioutas, C. (2000a) A new generation of portable coarse, fine, and ultrafine
8 particle concentrators for use in inhalation toxicology. In: Phalen, R. F., ed. Inhalation Toxicology:
9 proceedings of the third colloquium on paniculate air pollution and human health (first special issue); June,
10 1999; Durham, NC. Inhalation Toxicol. 12 (suppl. 1): 121-137.
11 Kim, S.; Sioutas, C.; Chang, M.-C.; Gong, H., Jr. (2000b) Factors affecting the stability of the performance of
12 ambient fine-particle concentrators. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
13 Toxicol. 12(suppl. 4): 281-298.
14 Kimmel, T. A.; Chen, L. C.; Bosland, M. C.; Nadziejko, C. (1997) Influence of acid aerosol droplet size on
15 structural changes in the rat lung caused by acute exposure to sulfuric acid and ozone. Toxicol. Appl.
16 Pharmacol. 144: 348-355.
17 Kitabatake, M.; Imai, M.; Kasama, K.; Kobayashi, I.; Tomita, Y.; Yoshida, K. (1979) Effects of air pollutants on the
18 experimental induction of asthma attacks in guinea pigs: sulfuric acid mist and mixture of the mist and sulfur
19 dioxide. Mie Med. J. 29: 29-36.
20 Kleeberger, S. R.; Levitt, R. C.; Zhang, L.-Y.; Longphre, M.; Harkema, J.; Jedlicka, A.; Eleff, S. M.;
21 DiSilvestre, D.; Holroyd, K. J. (1997) Linkage analysis of susceptibility to ozone-induced lung inflammation
22 in inbred mice. Nat. Genet. 17: 475-478.
23 Kleinman, M. T.; Leaf, D. A.; Kelly, E.; Caiozzo, V.; Osann, K.; O'Niell, T. (1998) Urban angina in the mountains:
24 effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
25 Health 53: 388-397.
26 Kleinman, M. T.; Mautz, W. J.; Bjarnason, S. (1999) Adaptive and non-adaptive responses in rats exposed to ozone,
27 alone and in mixtures, with acidic aerosols. Inhalation Toxicol. 11: 249-264.
28 Kleinman, M. T.; Bufalino, C.; Rasmussen, R.; Hyde, D.; Bhalla, D. K.; Mautz, W. J. (2000) Toxicity of chemical
29 components of ambient fine paniculate matter (PM2 5) inhaled by aged rats. J. Appl. Toxicol. 20: 357-364.
30 Knox, R. B.; Suphioglu, C.; Taylor, P.; Desai, R.; Watson, H. C.; Peng, J. L.; Bursill, L. A. (1997) Major grass
31 pollen allergen Lol p 1 binds to diesel exhaust particles: implications for asthma and air pollution. Clin. Exp.
32 Allergy 27: 246-251.
33 Kobzik, L. (1995) Lung macrophage uptake of unopsonized environmental particulates: role of scavenger-type
34 receptors. J. Immunol. 155: 367-376.
35 Kobzik, L.; Goldsmith, C.-A. W.; Ning, Y. Y.; Qin, G.; Morgan, B.; Imrich, A.; Lawrence J.; Murthy, G. G. K.;
36 Catalano, P. J. (2001) Effects of combined ozone and air pollution particle exposure in mice. Boston, MA:
37 Health Effects Institute; research report no. 106. Available: http://www.healtheffects.org/pubs-recent.htm [11
38 April 2002].
39 Kodavanti, U. P.; Costa, D. L. (1999) Animal models to study for pollutant effects. In: Holgate, S. T.; Samet, J. M.;
40 Koren, H. S.; Maynard, R. L., eds. Air pollution and health. New York, NY: Academic Press; pp. 165-197.
41 Kodavanti, U. P.; Jaskot, R. H.; Bonner, J.; Badgett, A.; Dreher, K. L. (1996) Eosinophilic lung inflammation in
42 particulate-induced lung injury: technical consideration in isolating RNA for gene expression studies.
43 Exp. Lung Res. 22: 541-554.
44 Kodavanti, U. P.; Jaskot, R. H.; Su, W. Y.; Costa, D. L.; Ohio, A. J.; Dreher, K. L. (1997a) Genetic variability in
45 combustion particle-induced chronic lung injury. Am. J. Physiol. 272: L521-L532.
46 Kodavanti, U. P.; Jaskot, R. H.; Costa, D. L.; Dreher, K. L. (1997b) Pulmonary proinflammatory gene induction
47 following acute exposure to residual oil fly ash: roles of particle-associated metals. Inhalation Toxicol.
48 9:679-701.
49 Kodavanti, U. P.; Hauser, R.; Christian!, D. C.; Meng, Z. H.; McGee, J.; Ledbetter, A.; Richards, J.; Costa, D. L.
50 (1998a) Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals.
51 Toxicol. Sci. 43:204-212.
52 Kodavanti, U. P.; Costa, D. L.; Bromberg, P. A. (1998b) Rodent models of cardiopulmonary disease: their potential
53 applicability in studies of air pollutant susceptibility. Environ. Health Perspect. 106(suppl. 1): 111-130.
54 Kodavanti, U. P.; Jackson, M. C.; Ledbetter, A. D.; Richards, J. R.; Gardner, S. Y.; Watkinson, W. P.; Campen,
55 M. J.; Costa, D. L. (1999) Lung injury from intratracheal and inhalation exposures to residual oil fly ash in a
April 2002 7-117 DRAFT-DO NOT QUOTE OR CITE
-------
1 rat model of monocrotaline-induced pulmonary hypertension. J. Toxicol. Environ. Health Part A
2 57: 101-121.
3 Kodavanti, U. P.; Mebane, R.; Ledbetter, A.; Krantz, T.; McGee, J.; Jackson, M. C.; Walsh, L.; Hilliard, H.; Chen,
4 B. Y.; Richards, J.; Costa, D. L. (2000a) Variable pulmonary responses from exposure to concentrated
5 ambient air particles in a rat model of bronchitis. Toxicol. Sci. 54: 441-451.
6 Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Watkinson, W. P.; Campen, M. J.; Winsett, D. W.;
7 Richards, J. R.; Crissman, K. M.; Hatch, G. E.; Costa, D. L. (2000b) The spontaneously hypertensive rat as a
8 model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress
9 from inhaled emission paniculate matter. Toxicol. Appl. Pharmacol. 164: 250-263.
10 Kodavanti, U. P.; Schladweiler, M. C. J.; Richards, J. R.; Costa, D. L. (2001) Acute lung injury from intratracheal
11 exposure to fugitive residual oil fly ash and its constituent metals in normo- and spontaneously hypertensive
12 rats. Inhalation Toxicol. 13: 37-54.
13 Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Hauser, R.; Christiani, D. C.; McGee, J.; Richards, J. R.;
14 Costa, D. L. (2002) Temporal association between pulmonary and systemic effects of paniculate matter in
15 healthy and cardiovascular compromised rats. J. Toxicol. Environ. Health: in press.
16 Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
17 experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
18 Perspect. 105: 1234-1237.
19 Lambert, A. L.; Dong, W.; Winsett, D. W.; Selgrade, M. K.; Gilmour, M. I. (1999) Residual oil fly ash exposure
20 enhances allergic sensitization to house dust mite. Toxicol. Appl. Pharmacol. 158: 269-277'.
21 Lambert, A. L.; Dong, W.; Selgrade, M. K.; Gilmour, M. I. (2000) Enhanced allergic sensitization by residual oil fly
22 ash particles is mediated by soluble metal constituents. Toxicol. Appl. Pharmacol. 165: 84-93.
23 Last, J. A.; Pinkerton, K. E. (1997) Chronic exposure of rats to ozone and sulfuric acid aerosol: biochemical and
24 structural responses. Toxicology 116: 133-146.
25 Last, J. A.; Hyde, D. M.; Chang, D. P. Y. (1984) A mechanism of synergistic lung damage by ozone and a
26 respirable aerosol. Exp. Lung Res. 7: 223-235.
27 Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular
28 distribution of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol.
29 18:687-695.
30 Lay, J. C.; Zeman, K. L.; Ohio, A. J.; Bennett, W. D. (2001) Effects of inhaled iron oxide particles on alveolar
31 epithelial permeability in normal subjects. Inhalation Toxicol. 13: 1065-1078.
32 Leduc, D.; Fally, S.; De Vuyst, P.; Wollast, R.; Yernault, J.-C. (1995) Acute exposure to realistic acid fog: effects
33 on respiratory function and airway responsiveness in asthmatics. Environ. Res. 71: 89-98.
34 Lee, M. M.; Green, F. H. Y.; Roth, S. H.; Karkhanis, A.; Bjarnason, S. G.; Schurch, S. (1999) Sulfuric acid aerosol
3 5 induces changes in alveolar surface tension in the guinea pig but not in the rat. Exp. Lung Res. 25: 229-244.
36 Leikauf, G. D.; McDowell, S. A.; Gammon, K.; Wesselkamper, S. C.; Bachurski, C. J.; Puga, A.; Wiest, J. S.;
37 Leikauf, J. E.; Prows, D. R. (2000) Functional genomics of particle-induced lung injury. Inhalation Toxicol.
38 12:59-73.
39 Leikauf, G.; McDowell, S. A.; Wesselkamper, S. C.; Miller, C. R.; Hardie, W. D.; Gammon, K.; Biswas, P. P.;
40 Korfhagen, T. R.; Bachurski, C. J.; Wiest, J. S.; Willeke, K.; Bingham, E.; Leikauf, J. E.; Aronow, B. J.;
41 Prows, D. R. (2001) Pathogenomic mechanisms for paniculate matter induction of acute lung injury and
42 inflammation in mice. Boston, MA: Health Effects Institute; research report no. 105. Available:
43 http://www.healtheffects.org/pubs-recent.htm [11 April 2002].
44 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1996) Free radical activity and pro-inflammatory effects of
45 particulate air pollution (PM10) in vivo and in vitro. Thorax 51: 1216-1222.
46 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
47 particulate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
48 international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
49 September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
50 Li, X. Y.; Brown, D.; Smith, S.; MacNee, W.; Donaldson, K. (1999) Short-term inflammatory responses following
51 intratracheal instillation of fine and ultrafine carbon black in rats. Inhalation Toxicol. 11: 709-731.
52 Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D. (1994) Effects of prolonged,
53 repeated exposure to ozone, sulfuric acid, and their combination in healthy and asthmatic volunteers. Am. J.
54 Respir. Crit. Care Med. 150: 431-440.
April 2002 7-118 DRAFT-DO NOT QUOTE OR CITE
-------
1 Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
2 mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
3 Lippmann, M. (2000) Environmental toxicants: human exposures and their health effects. 2nd ed. New York, NY:
4 Wiley-Interscience.
5 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
6 California. Environ. Health Perspect. 105: 216-222.
7 Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
8 insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
9 Long, C. M.; Suh, H. H.; Kobzik, L.; Catalano, P. I; Ning, Y. Y.; Koutrakis, P. (2001) A pilot investigation of the
10 relative toxicity of indoor and outdoor fine particles: in vitro effects of endotoxin and other paniculate
11 properties. Environ. Health Perspect. 109: 1019-1026.
12 Longphre, M.; Li, D.; Matovinovic, E.; Gallup, M.; Samet, J. M.; Basbaum, C. B. (2000) Lung mucin production is
13 stimulated by the air pollutant residual oil fly ash. Toxicol. Appl. Pharmacol. 162: 86-92.
14 Lorz, C.; Lopez, J. (1997) Incidence of air pollution in the pulmonary surfactant system of the pigeon (Columbia
15 livid). Anat. Rec. 249: 206-212.
16 Madden, M. C.; Thomas, M. J.; Ohio, A. J. (1999) Acetaldehyde (CH3CHO) production in rodent lung after
17 exposure to metal-rich particles. Free Radical Biol. Med. 26: 1569-1577.
18 Madden, M. C.; Richards, J. H.; Dailey, L. A.; Hatch, G. E.; Ohio, A. J. (2000) Effect of ozone on diesel exhaust
19 particle toxicity in rat lung. Toxicol. Appl. Pharmacol. 168: 140-148.
20 Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
21 macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
22 Pharmacol. 153:28-38.
23 Mauderly, J. L. (1993) Toxicological approaches to complex mixtures. Environ. Health Perspect. Suppl.
24 101(4): 155-165.
25 Mautz, W. J.; Kleinman, M. T.; Bhalla, D. K.; Phalen, R. F. (2001) Respiratory tract responses to repeated
26 inhalation of an oxidant and acid gas-particle air pollutant mixture. Toxicol. Sci. 61: 331-341.
27 McDowell, S. A.; Gammon, K.; Bachurski, C. J.; Weist, J. S.; Leikauf, J. E.; Prows, D. R. Leikauf, G. D. (2000)
28 Differential gene expression in the initiation and progression of nickel-induced acute lung injury. Am. J.
29 Respir. Cell Mol. Biol. 23: 466-474.
30 McKenna, I. M.; Gordon, T.; Chen, L. C.; Anver, M. R.; Waalkes, M. P. (1998) Expression of metallothionein
31 protein in the lungs of Wistar rats and C57 and DBA mice exposed to cadmium oxide fumes. Toxicol. Appl.
32 Pharmacol. 153: 169-178.
33 Menetrez, M. Y.; Foarde, K. K.; Ensor, D. S. (2001) An analytical method for the measurement of nonviable
34 bioaerosols. J. Air Waste Manage. Assoc. 51: 1436-1442.
35 Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
36 relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
37 Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
3 8 exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
39 155:245-252.
40 Morgan, K. T.; Patterson, D. L.; Gross, E. A. (1986) Responses of the nasal mucociliary apparatus of F-344 rats to
41 formaldehyde gas. Toxicol. Appl. Pharmacol. 82: 1-13.
42 Moss, O. R.; Gross, E. A.; James, R. A.; Janszen, D. B.; Ross, P. W.; Roberts, K. C.; Howard, A. M.; Harkema, J.
43 R.; Calderon-Garciduenas, L.; Morgan, K. T. (2001) Respiratory tract toxicity in rats exposed to Mexico City
44 air. Cambridge, MA: Health Effects Institute; research report no. 100. Available:
45 http://www.healtheffects.org/ [31 July, 2001].
46 Muggenburg, B. A.; Barr, E. B.; Cheng, Y. S.; Seagrave, JC.; Tilley, L. P.; Mauderly, J. L. (2000) Effect of inhaled
47 residual oil fly ash on the electrocardiogram of dogs. In: Grant, L. D., ed. PM2000: paniculate matter and
48 health. Inhalation Toxicol. 12(suppl. 4): 189-208.
49 Mukae, H.; Hogg, J. C.; English, D.; Vincent, R.; VanEeden, S. F. (2000) Phagocytosis of paniculate air pollutants
50 by human alveolar macrophages stimulates the bone marrow. Am. J. Physiol. 279: L924-L931.
51 Mukae, H.; Vincent, R.; Quinlan, K.; English, D.; Hards, J.; Hogg, J. C.; Van Eeden, S. F. (2001) The effect of
52 repeated exposure to paniculate air pollution (PM10) on the bone marrow. Am. J. Respir. Crit. Care Med.
53 163:201-209.
54 Murphy, S. A.; BeruBe, K. A.; Pooley, F. D.; Richards, R. J. (1998) The response of lung epithelium to well
55 characterised fine particles. Life Sci. 62: 1789-1799.
April 2002 7-119 DRAFT-DO NOT QUOTE OR CITE
-------
1 Nadadur, S. S.; Kodavanti, U. P. (2002) Altered gene expression profiles of rat lung in response to an emission
2 paniculate and its metal constituents. J. Toxicol. Environ. Health: in press.
3 Nadadur, S. S.; Schladweiler, M. C. J.; Kodavanti, U. P. (2000) A pulmonary rat gene array for screening altered
4 expression profiles in air pollutant-induced lung injury. Inhalation Toxicol. 12: 1239-1254.
5 Nadeau, D.; Vincent, R.; Kumarathasan, P.; Brook, J.; Dufresne, A. (1996) Cytotoxicity of ambient air particles to
6 rat lung macrophages: comparison of cellular and functional assays. Toxicol. In Vitro 10: 161-172.
7 Nakamura, T.; Hayashida, Y. (1992) Autonomic cardiovascular responses to smoke exposure in conscious rats. Am.
8 J. Physiol. 262: R73 8-745.
9 National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
10 Bethesda, MD: U. S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
11 publication no. 97-4051.
12 Nemmar, A.; Delaunois, A.; Nemery, B.; Dessy-Doize, C.; Beckers, J.-F.; Sulon, J.; Gustin, P. (1999) Inflammatory
13 effect of intratracheal instillation of ultrafine particles in the rabbit: role of C-fiber and mast cells.
14 Toxicol. Appl. Pharmacol. 160: 250-261.
15 Nightingale, J. A.; Maggs, R.; Cullinan, P.; Donnelly, L. E.; Rogers, D. F.; Kinnersley, R.; Chung, K. F.; Barnes, P.
16 J.; Ashmore, M.; Newman-Taylor, A. (2000) Airway inflammation after controlled exposure to diesel
17 exhaust particulates. Am. J. Respir. Crit. Care Med. 162: 161-166.
18 Ning, Y.; Imrich, A.; Goldsmith, C. A.; Qin, G.; Kobzik, L. (2000) Alveolar macrophage cytokine production in
19 response to air particles in vitro: role of endotoxin. J. Toxicol. Environ. Health A 59: 165-180.
20 Nordenhall, C.; Pourazar, J.; Ledin, M. C.; Levin, J. O.; Sandstrom, T.; Adelroth, E. (2001) Diesel exhaust
21 enhances airway responsiveness in asthmatic subjects. Eur. Respir. J. 17: 909-915.
22 Norwood, J., Jr.; Ledbetter, A. D.; Doerfler, D. L.; Hatch, G. E. (2001) Residual oil fly ash inhalation in guinea
23 pigs: influence of absorbate and glutathione depletion. Toxicol. Sci. 61: 144-153.
24 Oberdorster, G. (1996) Effects of ultrafine particles in the lung and potential relevance to environmental particles.
25 In: Marijnissen, J. C. M.; Gradon, L., eds. Proceedings of Aerosol inhalation, lung transport, deposition and
26 the relation to the environment: recent research frontiers, an international workshop; September, 1995;
27 Warsaw, Poland. London, United Kingdom: Kluwer Academic; pp. 165-174.
28 Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
29 lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
30 Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer
31 particles for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
32 Oberdorster, G.; Finkelstein, J. N.; Johnston, C.; Gelein, R.; Cox, C.; Baggs, R.; Elder, A. C. P. (2000) Acute
33 pulmonary effects of ultrafine particles in rats and mice. Cambridge, MA: Health Effects Institute; research
34 report no. 96.
35 Oettinger, R.; Drumm, K.; Knorst, M.; Krinyak, P.; Smolarski, R.; Kienast, K. (1999) Production of reactive oxygen
36 intermediates by human macrophages exposed to soot particles and asbestos fibers and increase in NF-kappa
37 B p50/p!05 mRNA. Lung 177: 343-354.
38 Ohtoshi, T.; Takizawa, H.; Okazaki, H.; Kawasaki, S.; Takeuchi, N.; Ohta, K.; Ito, K. (1998) Diesel exhaust
39 particles stimulate human airway epithelial cells to produce cytokines relevant to airway inflammation in
40 vitro. J. Allergy Clin. Immunol. 101: 778-785.
41 Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Branson, K.; Jakab, G. J.; Kleeberger, S. R. (2000a) Interstrain variation
42 in murine susceptibility to inhaled acid-coated particles. Am. J. Physiol. L469-L476.
43 Ohtsuka, Y.; Branson, K. J.; Jedlicka, A. E.; Mitzner, W.; Clarke, R. W.; Zhang, L.-Y.; Eleff, S. M.; Kleeberger,
44 S. R. (2000b) Genetic linkage analysis of susceptibility to particle exposure in mice. Am. J. Respir. Cell Mol.
45 Biol. 22: 574-581.
46 Oortgiesen, M.; Veronesi, B.; Eichenbaum, G.; Kiser, P. F.; Simon, S. A. (2000) Residual oil fly ash and charged
47 polymers activate epithelial cells and nociceptive sensory neurons. Am. J. Physiol. 278: L683-L695.
48 Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
49 Fundam. Appl. Toxicol. 40: 220-227.
50 Osier, M.; Baggs, R. B.; Oberdorster, G. (1997) Intratracheal instillation versus intratracheal inhalation: influence of
51 cytokines on inflammatory response. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
52 international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
53 September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1265-1271.
54 Osman, K.; Zejda, J. E.; Schtitz, A.; Mielzynska, D.; Blinder, C. G.; Vahter, M. (1998) Exposure to lead and other
55 metals in children from Katowice district, Poland. Int. Arch. Occup. Environ. Health71: 180-186.
April 2002 7-120 DRAFT-DO NOT QUOTE OR CITE
-------
1 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
2 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
3 Petrovic, S.; Urch, B.; Brook, J.; Datema, J.; Purdham, J.; Liu, L.; Lukic, Z.; Zimmerman, B.; Tofler, G.; Downar,
4 E.; Corey, P.; Tarlo, S.; Broder, L; Dales, R.; Silverman, F. (1999) Cardiorespiratory effects of concentrated
5 ambient PM2.5: a pilot study using controlled human exposures. In: Phalen, R. F., ed. Inhalation Toxicology:
6 proceedings of the third colloquium on paniculate air pollution and human health (first special issue); June,
7 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 173-188.
8 Pierce, L. M.; Alessandrini, F.; Godleski, J. J.; Paulauskis, J. D. (1996) Vanadium-induced chemokine mRNA
9 expression and pulmonary inflammation. Toxicol. Appl. Pharmacol. 138: 1-11.
10 Pinkerton, K. E.; Brody, A. R.; Miller, F. J.; Crapo, J. D. (1989) Exposure to low levels of ozone results in
11 enhanced pulmonary retention of inhaled asbestos fibers. Am. Rev. Respir. Dis. 140: 1075-1081.
12 Pinto, M.; Birnbaum, S. C.; Kadar, T.; Goldberg, G. M. (1979) Lung injury in mice induced by factors acting
13 synergistically with inhaled paniculate antigen. Clin. Immunol. Immunopathol. 13: 361-368.
14 Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
15 Am. J. Public Health 79: 623-628.
16 Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and
17 Cache Valleys. Arch. Environ. Health 46: 90-97.
18 Prahalad, A. K.; Soukup, J. M.; Inmon, J.; Willis, R.; Ohio, A. J.; Becker, S.; Gallagher, J. E. (1999) Ambient air
19 particles: effects on cellular oxidant radical generation in relation to paniculate elemental chemistry. Toxicol.
20 Appl. Pharmacol. 158: 81-91.
21 Pritchard, R. J.; Ohio, A. J.; Lehmann, J. R.; Winsett, D. W.; Tepper, J. S.; Park, P.; Gilmour, M. I.; Dreher, K. L.;
22 Costa, D. L. (1996) Oxidant generation and lung injury after paniculate air pollutant exposure increase with
23 the concentrations of associated metals. Inhalation Toxicol. 8: 457-477.
24 Prows, D. R.; Shertzer, H. G.; Daly, M. J.; Sidman, C. L.; Leikauf, G. D. (1997) Genetic analysis of ozone-induced
25 acute lung injury in sensitive and resistant strains of mice. Nat. Genet. 17: 471-474.
26 Prows, D. R.; Leikauf, G. D. (2001) Quantitative trait analysis of nickel-induced acute lung injury in mice. Am. J.
27 Respir. Cell Mol. Biol. 24: 740-746.
28 Quay, J. L.; Reed, W.; Samet, J.; Devlin, R. B. (1998) Air pollution particles induce IL-6 gene expression in human
29 airway epithelial cells via NF-KB activation. Am. J. Respir. Cell Mol. Biol. 19: 98-106.
30 Rose, C. S.; Martyny, J. W.; Newman, L. S.; Milton, D. K.; King, T. E., Jr.; Beebe, J. L.; McCammon, J. B.;
31 Hoffman, R. E.; Kreiss, K. (1998) "Lifeguard lung": endemic granulomatous pneumonitis in an indoor
32 swimming pool. Am. J. Public Health 88: 1795-1800.
33 Rudell, B.; Sandstrom, T.; Hammarstrom, U.; Ledin, M.-L.; Horstedt, P.; Stjernberg, N. (1994) Evaluation of an
34 exposure setup for studying effects of diesel exhaust in humans. Int. Arch. Occup. Environ. Health 66: 77-83.
35 Rudell, B.; Ledin, M.-C.; Hammarstrom, U.; Stjernberg, N.; Lundback, B.; Sandstrom, T. (1996) Effects on
36 symptoms and lung function in humans experimentally exposed to diesel exhaust. Occup. Environ. Med. 53:
37 658-662.
38 Rudell, B.; Blomberg, A.; Helleday, R.; Ledin, M.-C.; Lundback, B.; Stjernberg, N.; Horstedt, P.; Sandstrom, T.
39 (1999) Bronchoalveolar inflammation after exposure to diesel exhaust: comparison between unfiltered and
40 particle trap filtered exhaust. Occup. Environ. Med. 56: 527-534.
41 Sagai, M.; Furuyama, A.; Ichinose, T. (1996) Biological effects of diesel exhaust particles (DEP). III. Pathogenesis
42 of asthma like symptoms in mice. Free Radical Biol. Med. 21: 199-209.
43 Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (1999) Acute inflammatory
44 responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
45 volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.
46 Salvi, S. S.; Nordenhall, C.; Blomberg, A.; Rudell, B.; Pourazar, J.; Kelly, F. J.; Wilson, S.; Sandstrom, T.; Holgate,
47 S. T.; Frew, A. J. (2000) Acute exposure to diesel exhaust increases IL-8 and GRO-aproduction in healthy
48 human airways. Am. J. Respir. Crit. Care Med. 161: 550-557.
49 Samet, J. M.; Reed, W.; Ohio, A. J.; Devlin, R. B.; Carter, J. D.; Dailey, L. A.; Bromberg, P. A.; Madden, M. C.
50 (1996) Induction of prostaglandin H synthase 2 in human airway epithelial cells exposed to residual oil fly
51 ash. Toxicol. Appl. Pharmacol. 141: 159-168.
52 Samet, J. M.; Stonehuerner, J.; Reed, W.; Devlin, R. B.; Dailey, L. A.; Kennedy, T. P.; Bromberg, P. A.; Ohio, A. J.
53 (1997) Disruption of protein tyrosine phosphate homeostasis in bronchial epithelial cells exposed to oil fly
54 ash. Am. J. Physiol. 272: L426-L432.
April 2002 7-121 DRAFT-DO NOT QUOTE OR CITE
-------
1 Samet, J. M; Graves, L. M; Quay, I; Dailey, L. A.; Devlin, R. B.; Ohio, A. I; Wu, W.; Bromberg, P. A.; Reed, W.
2 (1998) Activation of MAPKs in human bronchial epithelial cells exposed to metals. Am. J. Physiol.
3 275:L551-L558.
4 Samet, J. M.; Ohio, A. J.; Costa, D. L.; Madden, M. C. (2000) Increased expression of cyclooxygenase 2 mediates
5 oil fly ash-induced lung injury. Exp. Lung Res. 26: 57-69.
6 Schluter, T.; Berg, L; Dorger, M.; Gercken, G. (1995) Effect of heavy metal ions on the release of reactive oxygen
7 intermediates by bovine alveolar macrophages. Toxicology 98: 47-55.
8 Schwartz, J. (1994) What are people dying of on high air pollution days? Environ. Res. 64: 26-35.
9 Seagrave, J. C.; Nikula, K. J. (2000) Multiple modes of responses to air pollution paniculate materials in
10 A549 alveolar type II cells. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
11 12(suppl. 4): 247-260.
12 Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Paniculate air pollution and acute health effects.
13 Lancet (8943): 176-178.
14 Shukla, A.; Timblin, C.; BeruBe, K.; Gordon, T.; McKinney, W.; Driscoll, K.; Vacek, P.; Mossman, B. T. (2000)
15 Inhaled paniculate matter causes expression of nuclear factor (NF)-icB-related genes and oxidant-dependent
16 NF-KB activation in vitro. Am. J. Respir. Cell Mol. Biol. 23: 182-187.
17 Silbajoris, R.; Ohio, A. J.; Samet, J. M.; Jaskot, R.; Dreher, K. L.; Brighton, L. E. (2000) In vivo and in vitro
18 correlation of pulmonary map kinase activation following metallic exposure. Inhalation Toxicol. 12: 453-468.
19 Sindhu, R. K.; Mautz, W. J.; Kikkawa, Y. (1998) Chronic exposure to ozone and nitric acid vapor results in
20 increased levels of rat pulmonary putrescine. Arch. Toxicol. 72: 445-449.
21 Sioutas, C.; Koutrakis, P.; Burton, R. M. (1995a) A technique to expose animals to concentrated fine ambient
22 aerosols. Environ. Health Perspect. 103: 172-177.
23 Sioutas, D.; Koutrakis, P.; Ferguson, S. T.; Burton, R. M. (1995b) Development and evaluation of a prototype
24 ambient particle concentrator for inhalation exposure studies. Inhalation Toxicol. 7: 633-644.
25 Sioutas, C.; Kim, S.; Chang, M.; Terrell, L. L.; Gong, H., Jr. (2000) Field evaluation of a modified DataRAM MIE
26 scattering monitor for real-time PM2 5 mass concentration measurements. Atmos. Environ. 34: 4829-4838.
27 Sjogren, B. (1997) Occupational exposure to dust: inflammation and ischaemic heart disease. Occup. Environ. Med.
28 54:466-469.
29 Smith, K. R.; Aust, A. E. (1997) Mobilization of iron from urban particulates leads to generation of reactive oxygen
30 species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chem. Res. Toxicol.
31 10:828-834.
32 Smith, K. R.; Veranth, J. M.; Lighty, J. S.; Aust, A. E. (1998) Mobilization of iron from coal fly ash was dependent
33 upon the particle size and the source of coal. Chem. Res. Toxicol. 11: 1494-1500.
34 Soukup, J. M.; Ohio, A. J.; Becker, S. (2000) Soluble components of Utah Valley paniculate pollution alter alveolar
35 macrophage function in vivo and in vitro. Inhalation Toxicol. 12: 401-414.
36 Soukup, J. M.; Becker, S. (2001) Human alveolar macrophage responses to air pollution particulates are associated
37 with insoluble components of coarse material, including paniculate endotoxin. Toxicol. Appl. Pharmacol.
38 171:20-26.
39 Spengler, J. D.; Thurston, G. D. (1983) Mass and elemental composition of fine and coarse particles in six U.S.
40 cities. J. AirPollut. Control Assoc. 33: 1162-1171.
41 Steerenberg, P. A.; Dormans, J. A. M. A.; Van Doom, C. C. M.; Middendorp, S.; Vos, J. G.; VanLoveren, H.
42 (1999) A pollen model in the rat for testing adjuvant activity of air pollution components. Inhalation Toxicol.
43 11: 1109-1122.
44 Stringer, B.; Kobzik, L. (1996) Alveolar macrophage uptake of the environmental paniculate titanium dioxide:
45 role of surfactant components. Am. J. Respir. Cell Mol. Biol. 14: 155-160.
46 Stringer, B.; Kobzik, L. (1998) Environmental particulate-mediated cytokine production in lung epithelial cells
47 (A549): role of preexisting inflammation and oxidant stress. J. Toxicol. Environ. Health Part A 55: 31-44.
48 Stringer, B.; Imrich, A.; Kobzik, L. (1996) Lung epithelial cell (A549) interaction with unopsonized environmental
49 particulates: quantitation of particle-specific binding and IL-8 production. Exp. Lung Res. 22: 495-508.
50 Su, W.-Y.; Jaskot, R. H.; Richards, J.; Abramson, S. R.; Woessner, J. F., Jr.; Yu, W.-H.; Dreher, K. L. (2000)
51 Induction of pulmonary matrilysin expression by combustion and ambient air particles. Am. J. Physiol. 279:
52 L152-L160.
53 Su, W.-Y.; Jaskot, R. H.; Dreher, K. L. (2000b) Paniculate matter induction of pulmonary gelatinase A, gelatinase
54 B, tissue inhibitor of metalloproteinase expression. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings
April 2002 7-122 DRAFT-DO NOT QUOTE OR CITE
-------
1 of the third colloquium on paniculate air pollution and human health (second special issue); June, 1999;
2 Durham, NC. Inhalation Toxicol. 12(suppl. 2): 105-119.
3 Sun, G.; Crissman, K.; Norwood, J.; Richards, J.; Slade, R.; Hatch, G. E. (2001) Oxidative interactions of synthetic
4 lung epithelial lining fluid with metal-containing paniculate matter. Am. J. Physiol. 281: L807-L815.
5 Takano, H.; Yoshikawa, T.; Ichinose, T.; Miyabara, Y.; Imaoka, K.; Sagai, M. (1997) Diesel exhaust particles
6 enhance antigen-induced airway inflammation and local cytokine expression in mice. Am. J. Respir. Crit.
7 Care Med. 156: 36-42.
8 Takano, H.; Ichinose, T.; Miyabara, Y.; Shibuya, T.; Lim, H.-B.; Yoshikawa, T. Sagai, M. (1998) Inhalation of
9 diesel exhaust enhances allergen-related eosinophil recruitment and airway hyperresponsiveness in mice.
10 Toxicol. Appl. Pharmacol. 150: 328-337.
11 Tan, W. C; Qiu, D.; Liam, B. L.; Ng, T. P.; Lee, S. H.; Van Eeden, S. F.; D'Yachkova, Y.; Hogg, J. C. (2000) The
12 human bone marrow response to acute air pollution caused by forest fires. Am. J. Respir. Crit. Care Med.
13 161: 1213-1217.
14 Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; Van Eeden, S. F. (1997) Phagocytosis of small carbon particles
15 (PM10) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
16 Am. J. Respir. Crit. Care Med. 155: 1441-1447.
17 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
18 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
19 Timblin, C.; BeruBe, K.; Chrug, A.; Driscoll, K.; Gordon, T.; Hemenway, D.; Walsh, E.; Cummins, A. B.;
20 Vacek, P.; Mossman, B. (1998) Ambient paniculate matter causes activation of the c-jun
21 kinase/stress-activated protein kinase cascade and DNA synthesis in lung epithelial cells. Cancer Res.
22 58:4543-4547.
23 Tsien, A.; Diaz-Sanchez, D.; Ma, J.; Saxon, A. (1997) The organic component of diesel exhaust particles and
24 phenanthrene, a major polyaromatic hydrocarbon constituent, enhances IgE production by IgE-secreting
25 EBV-transformed human B cells in vitro. Toxicol. Appl. Pharmacol. 142: 256-263.
26 Tsuchiyama, F.; Hisanaga, N.; Shibata, E.; Aoki, T.; Takagi, H.; Ando, T.; Takeuchi, Y. (1997) Pulmonary metal
27 distribution in urban dwellers. Int. Arch. Occup. Environ. Health 70: 77-84.
28 Turpin, B. J. (1999) Options for characterizing organic paniculate matter. Environ. Sci. Technol. 33: 76A-79A.
29 U.S. Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
3 0 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
31 Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
32 U.S. Environmental Protection Agency. (1989) An acid aerosols issue paper: health effects and aerometrics.
33 Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
34 Assessment Office; EPA report no. EPA-600/8-88-005F. Available from: NTIS, Springfield, VA;
35 PB91-125864.
36 U.S. Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park,
37 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/00laF-cF. 3v.
38 U.S. Environmental Protection Agency. (1996b) Ambient levels and noncancer health effects of inhaled crystalline
39 and amorphous silica: health issue assessment [draft]. Research Triangle Park, NC: National Center for
40 Environmental Assessment; EPA report no. EPA/600/R-95/115.
41 U.S. Environmental Protection Agency. (2000) Health assessment document for diesel exhaust [SAB review draft].
42 Washington, DC: Office of Research and Development, National Center for Environmental
43 Assessment-Washington Office; report no. EPA/600/8-90/057E. Available:
44 http://cfpub.epa.gov/ncea/cfm/nceahome.cfm [29 March, 2002].
45 Utell, M. J.; Morrow, P. E.; Hyde, R. W. (1983) Latent development of airway hyperreactivity in human subjects
46 after sulfuric acid aerosol exposure. J. Aerosol Sci. 14: 202-205.
47 Van Eeden, S. F.; Tan, W. C.; Suwa, T.; Mukae, H.; Terashima, T.; Fujii, T.; Qui, D.; Vincent, R.; Hogg, J. C.
48 (2001) Cytokines involved in the systemic inflammatory response induced by exposure to paniculate matter
49 air pollutants (PM10). Am. J. Respir. Crit. Care Med. 164: 826-830.
50 VanMaanen, J. M.; Borm, P. J.; Knaapen, A.; Van Herwijnen, M.; Schilderman, P. A.; Smith, K. R.; Aust, A. E.;
51 Tomatis, M.; Fubini, B. (1999) In vitro effects of coal fly ashes: hydroxyl radical generation, iron release,
52 and DNA damage and toxicity in rat lung epithelial cells. Inhalation Toxicol. 11: 1123-1141.
53 Van Zijverden, M.; Granum, B. (2000a) Adjuvant activity of paniculate pollutants in different mouse models.
54 Toxicology 152: 69-77.
April 2002 7-123 DRAFT-DO NOT QUOTE OR CITE
-------
1 VanZijverden, M; Van der Fiji, A.; Bol, M; VanPinxteren, F. A.; De Haar, C.; Penninks, A. H.; Van Loveren, H.;
2 Pieters, R. (2000b) Diesel exhaust, carbon black, and silica particles display distinct Thl/Th2 modulating
3 activity. Toxicol. Appl. Pharmacol. 168: 131-139.
4 Van Zijverden, M.; De Haar, C.; Van Beelen, A.; Van Loveren, H.; Penninks, A.; Pieters, R. (2001)
5 Co-administration of antigen and particles optimally stimulates the immune response in an intranasal
6 administration model in mice. Toxicol. Appl. Pharmacol. 177: 174-178.
7 Vanda, B.; de Buen, N.; Jasso, R.; Valero, G.; Vargas, M. H.; Olmos, R.; Arreola, J. L.; Santillan, P.; Alonso, P.
8 (1998) Inflammatory cells and ferruginous bodies in bronchoalveolar lavage in urban dogs. Acta Cytol.
9 42: 939-944.
10 Veronesi, B.; Oortgiesen, M. (2001) Neurogenic inflammation and paniculate matter (PM) air pollutants.
11 Neurotoxicology 22: 795-810.
12 Veronesi, B.; Carter, J. D.; Devlin, R. B.; Simon, S. A.; Oortgiesen, M. (1999a) Neuropeptides and capsaicin
13 stimulate the release of inflammatory cytokines in a human bronchial epithelial cell line. Neuropeptides 33:
14 447-456.
15 Veronesi, B.; Oortgiesen, M.; Carter, J. D.; Devlin, R. B. (1999b) Paniculate matter initiates inflammatory cytokine
16 release by activation of capsaicin and acid receptors in a human bronchial epithelial cell line. Toxicol. Appl.
17 Pharmacol. 154: 106-115.
18 Veronesi, B.; Oortgiesen, M.; Roy, J.; Carter, J. D.; Simon, S. A.; Gavett, S. H. (2000) Vanilloid (capsaicin)
19 receptors influence inflammatory sensitivity in response to paniculate matter. Toxicol. Appl. Pharmacol. 169:
20 66-76.
21 Veronesi, B.; De Haar, C.; Roy, J.; Oortgiesen, M. (2002a) Paniculate matter inflammation and receptor sensitivity
22 are target cell specific. Inhalation Toxicol. 14: 101-125.
23 Veronesi, B.; De Haar, C.; Lee, L.; Oortgiesen, M. (2002b) The surface charge of paniculate matter related to
24 cytokine release in human bronchial epithelial cells (BEAS-2B). Toxicol. Appl. Pharmacol. 179: XX
25 Vincent, R.; Bjarnason, S. G.; Adamson, I. Y. R.; Hedgecock, C.; Kumarathasan, P.; Guenette, J.; Potvin, M.;
26 Goegan, P.; Bouthillier, L. (1997) Acute pulmonary toxicity of urban paniculate matter and ozone. Am. J.
27 Pathol. 151: 1563-1570.
28 Vincent, R.; Kumarathasan, P.; Goegan, P.; Bjarnason, S. G.; Guenette, J.; Berube, D.; Adamson, I. Y.; Desjardins,
29 S.; Burnett, R. T.; Miller, F. J.; Battistini, B. (2001) Inhalation toxicology of urban ambient paniculate
30 matter: acute cardiovascular effect in rats. Boston, MA: Health Effects Institute; research report no. 104.
31 Available: http://www.healtheffects.org/Pubs/Vincent.pdf [29 January 02].
32 Vogelzang, P. F. J.; Van Der Gulden, J. W. J.; Folgering, H.; Kolk, J. J.; Heederik, D.; Preller, L.; Tielen, M. J. M.;
33 Van Schayck, C. P. (1998) Endotoxin exposure as a major determinant of lung function decline in pig
34 farmers. Am. J. Respir. Crit. Care Med. 157: 15-18.
35 Wagner, J. G.; Hotchkiss, J. A.; Harkema, J. R. (2001) Effects of ozone and endotoxin coexposure on rat airway
36 epithelium: potentiation of toxicant-induced alterations. Environ. Health Perspect. 109(suppl. 4): 591-598.
37 Walters, D. M.; Breysse, P. N.; Wills-Karp, M. (2001) Ambient urban Baltimore particulate-induced airway
38 hyperresponsiveness and inflammation in mice. Am. J. Respir. Crit. Care Med. 164: 1438-1443.
39 Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
40 fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
41 Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Kodavanti, U. P.; Dreher, K. L.; Su, W.-Y.; Highfill, J. W.; Costa,
42 D. L. (2000a) Cardiovascular effects following exposure to paniculate matter in healthy and
43 cardiopulmonary-compromised rats. In: Heinrich, U.; Mohr, U., eds. Relationships between acute and
44 chronic effects of air pollution. Washington, DC: ISLI Press, pp. 447-463.
45 Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U.P.; Highfill, J. W.; Costa, D. L. (2000b)
46 Thermoregulatory effects following exposure to paniculate matter in healthy and
47 cardiopulmonary-compromised rats. J. Therm. Biol. 25: 131-137.
48 Wesselkamper, S. C.; Prows, D. R.; Biswas, P.; Willeke, K.; Bingham, E.; Leikauf, G. D. (2000) Genetic
49 susceptibility to irritant-induced acute lung injury in mice. Am. J. Physiol. 279: L575-L582.
50 Wheatley, L. M.; Platts-Mills, T. A. E. (1996) Perennial allergens and the asthma epidemic. Sci. Med. 3: 6-13.
51 Woodin, M. A.; Hauser, R.; Liu, Y.; Smith, T. J.; Siegel, P. D.; Lewis, D. M.; Tollerud, D. J.; Christiani, D. C.
52 (1998) Molecular markers of acute upper airway inflammation in workers exposed to fuel-oil ash. Am. J.
53 Respir. Crit. Care Med. 158: 182-187.
54 Wu, W.; Samet, J. M.; Ohio, A. J.; Devlin, R. B. (2001) Activation of the EOF receptor signaling pathway in human
55 airway epithelial cells exposed to Utah Valley PM. Am. J. Physiol. 281: L483-L489.
April 2002 7-124 DRAFT-DO NOT QUOTE OR CITE
-------
1 Zelikoff, J. T.; Sisco, M; Cohen, M. D.; Frampton, M. W.; Utell, M. J; Schlesinger, R. B. (1994) Interspecies
2 comparison of immunotoxicity of inhaled sulfuric acid. II. New Zealand white rabbits. In: 1994 international
3 conference sponsored by the American Lung Association and the American Thoracic Society; May; Boston,
4 MA. Am. J. Respir. Crit. Care Med. 149: A621.
5 Zelikoff, J. T.; Frampton, M. W.; Cohen, M. D.; Morrow, P. E.; Sisco, M.; Tsai, Y.; Utell, M. J.; Schlesinger, R. B.
6 (1997) Effects of inhaled sulfuric acid aerosols on pulmonary immunocompetence: a comparative study in
7 humans and animals. Inhalation Toxicol. 9: 731-752.
8 Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
9 normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
10 Zock, J.-P.; Hollander, A.; Heederik, D.; Douwes, J. (1998) Acute lung function changes and low endotoxin
11 exposures in the potato processing industry. Am. J. Ind. Med. 33: 384-391.
April 2002 7-125 DRAFT-DO NOT QUOTE OR CITE
-------
i 8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS
2 FROM AMBIENT PARTICULATE MATTER
3
4
5 8.1 INTRODUCTION
6 Epidemiology studies linking community ambient PM concentrations to adverse health
7 effects played an important role in the 1996 PM Air Quality Criteria Document (PM AQCD), and
8 continue to play an important role. Those studies are indicative of measurable excesses in
9 pulmonary function decrements, respiratory symptoms, hospital and emergency department
10 admissions, and mortality in human populations being associated with ambient levels of PM25,
11 PM10_2 5, PM10, and other indicators of PM exposure. The numerous more recent epidemiologic
12 studies reviewed in this chapter generally identify more cities where ambient PM-relationships
13 with morbidity and mortality have been found and, thereby, both extend the earlier findings and
14 provide an expanded evidence base that substantiates health effects being associated with
15 exposures to PM at concentrations currently encountered in the United States.
16 The epidemiology studies presented here should be considered in combination with the
17 ambient concentration information presented in Chapter 3, the studies of human PM exposure in
18 Chapter 5, and the discussions of PM dosimetry and toxicology in Chapters 6 and 7. The
19 contribution of the epidemiology studies is to evaluate associations between health effects and
20 exposures of human populations to ambient PM and to help identify susceptible subgroups and
21 associated risk factors. Chapter 9 provides a concise interpretive synthesis of the information.
22 This chapter opens with a brief overview of key general features of the several types of
23 epidemiologic studies assessed in the chapter and a discussion of important general
24 methodological issues that must be considered in their critical assessment. After this brief
25 introduction, Section 8.2 assesses studies of PM effects on mortality. Section 8.3 evaluates
26 studies of morbidity as a health endpoint. Section 8.4 then provides an interpretive assessment of
27 the overall PM epidemiologic data base in relation to a variety of key issues and potential
28 inferences associated with studies reviewed in Sections 8.2 and 8.3. The overall key findings and
29 conclusions for this chapter are then summarized in Section 8.5.
30
April 2002 8-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.1.1 Types of Epidemiology Studies Reviewed
2 Definitions of various types of epidemiology studies used here were provided in the 1996
3 PM AQCD (U.S. Environmental Protection Agency, 1996a) and are briefly summarized here.
4 Briefly, the epidemiology studies are divided into mortality studies and morbidity studies.
5 Mortality studies evaluating PM effects on total (non-accidental) mortality and cause-specific
6 mortality have provided the most unambiguous evidence of a clearly adverse endpoint. The
7 morbidity studies further substantiate PM effects on a wide range of health endpoints, such as:
8 cardiovascular and respiratory-related hospital admissions, medical visits, reports of respiratory
9 symptoms, self-medication in asthmatics, changes in pulmonary function tests (PFT), low
10 birthweight infants, etc.
11 The epidemiology strategies most commonly used in PM health studies are of four types:
12 (1) ecologic studies; (2) time-series semi-ecologic studies; (3) longitudinal panel and prospective
13 cohort studies; and (4) case-control and crossover studies. All of these are observational studies
14 rather than experimental studies, since participants are not assigned at random to air pollution
15 exposures. In general, the exposure of the participant is not directly observed, and the
16 concentration of airborne particles and other air pollutants at one or more stationary air monitors
17 is used as a proxy for individual exposure to ambient air pollution.
18 In ecologic studies., the responses are at a community level (for example, annual mortality
19 rates), as are the exposure indices (for example, annual average particulate matter concentrations)
20 and covariates (for example, the percentage of the population greater than 65 years of age).
21 No individual data is used in the analysis, therefore the relation between health effect and
22 exposure calculated across different communities may not reflect individual-level associations
23 between health outcome and exposure. The use of proxy measures for individual exposure and
24 covariates or effects modifiers may also bias the results, and within-city or within-unit
25 confounding may be overlooked.
26 Time series studies are more informative because they allow study of associations between
27 changes in outcomes and changes in exposure indicators preceding or simultaneous with the
28 outcome. The temporal relationship supports a conclusion of a causal relation, even when both
29 the outcome (for example, the number of non-accidental deaths in a city during a day) and the
30 exposure (for example, daily air pollution concentration) are community indices.
April 2002 8-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 Prospective cohort (or panel) studies use data from individuals, including health status
2 (where available), individual exposure (not usually available), and individual covariates or risk
3 factors, observed over time. The participants in a prospective cohort study are ideally recruited,
4 using a simple or stratified random sample so as to represent a target population for which
5 individual or community exposure of the participants is known before and during the interval up
6 to the time the health endpoint occurs. The use of individual-level data is believed to give
7 prospective cohort studies greater inferential strength than other epidemiology strategies, but the
8 use of community-level or estimated exposure data may weaken this advantage, as in time-series
9 studies.
10 Case-control studies are retrospective studies in that exposure is determined after the health
11 endpoint occurs (this is common in occupational health studies). As Rothman and Greenland
12 (1998) describe it, "Case-control studies are best understood by defining a source population,
13 which represents a hypothetical study population in which a cohort study might have been
14 conducted ... In a case-control study, the cases are identified and their exposure status is
15 determined just as in a cohort study . . . [and] a control group of study subjects is sampled from
16 the entire source population that gives rise to the cases ... the cardinal requirement of control
17 selection is that the controls must be sampled independently of their exposure status."
18 The case-crossover design is suited to the study of a transient effect of an intermittent
19 exposure on the subsequent risk of a rare acute-onset disease hypothesized to occur a short time
20 after exposure. In the original development of the method, effect estimates were based on
21 within-subject comparisons of exposures associated with incident disease events with exposures
22 at times before the occurrence of disease, using matched case-control methods or methods for
23 stratified follow-up studies with spare data within each stratum. The principle of the analysis is
24 that the exposures of cases just before the event are compared with the distribution of exposure
25 estimated from some separate time period. This distribution is assumed to be representative of
26 the distribution of exposures for those individuals while they are at risk of developing the
27 outcome of interest.
28 When measurements of exposure or potential effect modifiers are available on an
29 individual level, it is possible to incorporate this information into a case-crossover study unlike a
30 time-series analysis. A disadvantage of the case-crossover design, however, is the potential for
31 bias due to time trends in the exposure time-series. Since case-crossover comparisons are made
April 2002 8-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 between different points in time, the case-crossover analysis implicitly depends on an assumption
2 that the exposure distribution is stable over time (stationary). If the exposure time-series is
3 non-stationary and case exposures are compared with referent exposures systematically selected
4 from a different period in time, a bias may be introduced into estimates of the measure of
5 association for the exposure and disease. These biases are particularly important when
6 examining the small associations that appear to exist between PM and health outcomes.
7
8 8.1.2 Confounding and Effect Modification
9 A pervasive problem in the analysis of epidemiology data, no matter what design or
10 strategy, is the unique attribution of the health outcome to the nominal causal agent (i.e., airborne
11 particles) in this document. The health outcomes attributed to particles are not specific (for
12 example, mortality in a broad range of ICD-9 categories) and may also be attributable to high or
13 low temperatures, influenza and other diseases, and/or exposure to gaseous criteria air pollutants.
14 Many of the other factors can be measured, directly or by proxies. Some of these co-variables
15 are confounders, others are effect modifiers. The distinctions are important.
16 Confounding \$ "... a confusion of effects. Specifically, the apparent effect of the
17 exposure of interest is distorted because the effect of an extraneous factor is mistaken for or
18 mixed with the actual exposure effect (which may be null)." (Rothman and Greenland, 1998,
19 p. 120). These authors list three criteria for a confounding factor:
20 (1) A confounding factor must be a risk factor for the disease (health effect).
21 (2) A confounding factor must be associated with the exposure under study in the source
22 population (the population at risk from which the cases are derived).
23 (3) A confounding factor must not be affected by the exposure or the disease, i.e., it cannot
24 be an intermediate step in the causal path between the exposure and the disease.
25 A causal pathway is one in which members of the population are exposed to putative causal
26 agents that can actually produce the observed health effect. The primary cause may be mediated
27 by secondary causes (possibly proximal to exposure) and may have either a direct effect on
28 exposure or an indirect effect through the secondary causes, or both, as illustrated below.
29 A non-causal pathway may involve factors that are not associated with the health effect or for
30 which there is no population exposure, so that the factors are not potential confounders.
April 2002 8-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 The determination of whether a potential confounder is an actual confounder depends on
2 biological or physical knowledge about its exposure and health effects. Patterns of association in
3 epidemiology may be helpful in suggesting where to look for this knowledge, but do not replace
4 it. Gaseous criteria pollutants (CO, NO2, SO2, O3) are candidates for confounders since: (1) all
5 of these have adverse health effects, with CO more often identified with cardiovascular effects
6 and the others with respiratory effects (including symptoms and hospital admissions), as part of
7 the wide spectrum of cardiopulmonary disease also associated with particles; (2) the gaseous
8 criteria pollutants may be associated with particles for several reasons, including (a) common
9 sources, (b) correlated changes in response to wind and weather, and (c) SO2 and NO2 may be
10 precursors to sulfate and nitrate components of ambient particle mixes, while NO2 contributes to
11 the formation of organic aerosols during photochemical transformations.
12 A common source, such as combustion of gasoline in motor vehicles emitting CO, NO2,
13 and primary particles, may play an important role in confounding among these pollutants, as does
14 weather and seasonal effects. Even though O3 is a secondary pollutant also associated with
15 emission of NO2, it is often less highly associated with particles. Levels of SO2 in the western
16 U.S. are often quite low, so that secondary formation of particle sulfates plays a much smaller
17 role there, resulting in usually relatively little confounding of SO2 with PM mass concentration in
18 the west. On the other hand, in the industrial midwest and northeastern states, SO2 and sulfate
19 levels during many of the epidemiology studies were relatively high and highly correlated with
20 fine particle mass concentrations, so that criterion 3 (no causal path leading from confounder to
21 exposure, or exposure to confounder to health effect) may not be strictly true for SO2 vs sulfate
22 or overall fine particle mass. If the correlation with PM and SO2 is not too high, it may be
23 possible to estimate some part of their independent effects. If there is a causal pathway, then it is
24 not clear whether the observed relation of exposure to health effect is a direct effect of the
25 exposure, an indirect effect mediated by the confounder, or a mixture of these.
26 Most extraneous variables fall into the category of effect modifiers. "Effect-measure
27 modification differs from confounding in several ways. The main difference is that, whereas
28 confounding is a bias that the investigator hopes to prevent or remove from the effect estimate,
29 effect-measure modification is a property of the effect under study ... In epidemiologic analysis
30 one tries to eliminate confounding but one tries to detect and estimate effect-measure
31 modification." (Rothman and Greenland, 1998, p. 254). Examples of effect modifiers in some
April 2002 8-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 of the studies evaluated in this chapter include environmental variables (such as temperature or
2 humidity in time-series studies), individual risk factors (such as education, cigarette smoking
3 status, age in a prospective cohort study), and community factors (such as percent of population
4 > 65 years old). It is often possible to stratify the relationship between health outcome and
5 exposure by one or more of these risk factor variables.
6 Effect modifiers may be encountered within a single-city time series studies, or across cities
7 in a two-stage hierarchical model or meta-analysis. We will use the latter case to illustrate some
8 of the possibilities using a hypothetical case with four cities in which a co-pollutant of the PM
9 index is to be evaluated as a possible effect modifier. In the examples in Figure 8-1, we assume
10 that the co-pollutant has a relatively high positive correlation with the PM index. It is also
11 assumed that the excess relative risk for PM is calculated in a model in which PM is the only air
12 pollutant. For any given co-pollutant concentration within each city, there is likely to be only a
13 modest range of values of the PM index and the associated excess relative risk, as is suggested by
14 the elliptical figures. The relationship between mortality and PM in Figure 8-la is assumed to be
15 the same and positive in all four cities; thus, with increasing co-pollutant concentration within
16 each city, the excess relative risk increases because the co-pollutant is strongly correlated with
17 the PM index. However, in the hypothetical 8-la, the co-pollutant is not an effect modifier for
18 PM, as can be shown by a regression of the estimated mean PM effect on the mean co-pollutant
19 concentration across the four cities.
20 The relationship between PM and mortality in Figure 8-lb is assumed to differ across the
21 four cities, ranging from strongly negative in City 1 to strongly positive in City 4. Thus, with
22 increasing co-pollutant concentration within each city, the excess relative risk decreases in City 1
23 and City 2 but increases in City 3 and City 4, because the co-pollutant is strongly correlated with
24 the PM index. In the hypothetical Figure 8-lb, the co-pollutant is an effect modifier for PM, as
25 can be shown by a regression of the estimated mean PM effect on the mean co-pollutant
26 concentration across the four cities, even though the simple mean of the excess relative risks
27 across the four cities is nearly zero. A relationship would be found if all within-city effects were
28 positive, or if the across-city ecological regression were negative. Stratification by levels of the
29 putative effect modifier is also often useful.
30 Potential confounding (Figure 8-2a) is more difficult to identify and several statistical
31 methods are available, none of them being completely satisfactory. The ususal methods are:
April 2002 8-6 DRAFT-DO NOT QUOTE OR CITE
-------
Q_
.0
0>
0)
LJJ
City 1
City 2
City 3
City 4
Co-Pollutant Concentration
Figure 8-la. Strong within-city association between PM and mortality, but no
second-stage association.
City 3
City 4
Co-Pollutant Concentration
Co-Pollutant Concentration
Figure 8-lb. Within-city association between PM and mortality ranges from negative
to positive with mean across cities approximately zero, but with strong
positive second-stage association.
April 2002
8-7
DRAFT-DO NOT QUOTE OR CITE
-------
Confounder
Figure 8-2a
, Outcomes
ft
Exposure
Outcomes
Modifier
Figure 8-2b
Primary Cause (s
Secondary Cause(s)
Figure 8-2c
Exposure
Outcomes
Exposure
Outcomes
Secondary Cause
Primary Cause(s)
Secondary Cause
: Exposure
Figure 8-2d
Figure 8-2. (a) Graphical depiction of confounding; (b) Graphical depiction of effect
modification; (c) Graphical depiction of a causal agent with a secondary
confounder; (d) Graphical depiction of a causal agent and two potential
confounders.
1
2
3
4
5
6
7
Within a city:
(A) Fit both a single-pollutant model and then several multi-pollutants models, and
determine if including the co-pollutants greatly changes the estimated effect and
inflates its estimated standard error;
(B) If the PM index and its co-pollutants are nearly multi-collinear, carry out a factor
analysis, and determine which gaseous pollutants are most closely associated with
PM in one or more common factors.
April 2002
DRAFT-DO NOT QUOTE OR CITE
-------
1 Using data from several cities:
2 (C) Proceed as in Method A and pool the effect size estimates across cities for single-
3 and multi-pollutant models;
4 (D) Carry out a hierarchical regression of the PM effects vs. the mean co-pollutant
5 concentration and determine if there is a significant relationship;
6 (E) First carry out a regression of PM vs. the co-pollutant concentration within each city
7 and the regression coefficient of mortality vs. PM for each city. Then fit a second-
8 stage model regressing the mortality-PM coefficient vs. the PM-co-pollutant
9 coefficient, concluding that the co-pollutant is a confounder if there is a significant
10 regression coefficient at the second stage (See Figure 8-2c).
11 The disadvantages of the methods are discussed in detail in Section 8.4. Briefly, the multi-
12 pollutant regression coefficients in method A may be unstable and have greatly inflated standard
13 errors, weakening their interpretation. In method B, the factors may be sensitive to the choice of
14 co-pollutants and the analysis method, and may be difficult to relate to real-world entities.
15 In method C, as with any meta-analysis, it is necessary to consider the heterogeneity of the
16 within-city effects before pooling them. Several large multi-city studies have revealed
17 unexpected heterogeneity, not fully explained at present.
18 While method D is sometimes interpreted as showing confounding if the regression
19 coefficient is non-zero, this is an argument for effect modification, not confounding.
20 Method E is sensitive to the assumptions being made. For example, if PM is the primary
21 cause in Figure 8-2c and the co-pollutant the secondary cause, then the two-stage approach may
22 be valid. However, if the model is mis-specified and there are two or more secondary causes,
23 some of which may not be identified, then the method may give misleading results.
24 An additional issue of great relevance is whether or not the population in a community time
25 series study or the participants in a prospective cohort study are exposed to measurable levels of
26 the potential confounder, particularly the ambient gaseous co-pollutants. If there is no exposure,
27 then the potential confounder does not satisfy the requirement that it is related to both exposure
28 and outcomes. This is discussed in Section 8.4 in connection with the role of exposure
29 measurement errors in air pollution epidemiology.
30
31
April 2002 8-9 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.1.3 Selection of Studies for Review and Ambient PM Increments Used to
2 Report Risk Estimates
3 Numerous PM epidemiology papers have been published since the 1996 PM AQCD.
4 An ongoing medline search has been and is continuing to be conducted in conjunction with other
5 strategies to identify PM literature pertinent to developing criteria for PM NAAQS. Those
6 epidemiologic studies that relate measures of ambient PM to human health outcomes are
7 assessed in this chapter, but occupational exposures studies are not. Some of the criteria used for
8 selecting relevant literature for consideration here include whether a given study presents:
9 (1) pertinent ambient PM indices: e.g., PM10, PM25, PM10_25, etc.; (2) analyses of health effects
10 of specific PM chemical or physical constituents (e.g., metals, sulfates, nitrates or ultrafme
11 particles, etc.); (3) health endpoints not previously extensively researched; (4) multiple pollutant
12 analyses; and/or (5) for long-term effects, mortality displacement information. The publication
13 of pertinent new studies has been and is proceeding at a prodigious rate; and the review and
14 evaluation of pertinent literature in this PM AQCD development process is an ongoing process
15 which continues to obtain and assess new evidence.
16 The literature review method is similar to those used by others (e.g., Basu and Samet,
17 2000): (a) Establish a publication base using Medline and other data bases using a set of key
18 words (particles, air pollution, mortality, morbidity, cause of death, PM, and others); (b) add
19 papers to the publication base by staff review of Current Contents and tables of contents of
20 journals in which relevant papers are published; and (c) staff requests to scientists known to be
21 active in this field for papers recently accepted for publication. Efforts have been made to assess
22 here pertinent new studies published mainly through December, 2001, as well as some studies
23 published in early 2002 as acquired (if, in the opinion of staff, such recent new papers provide
24 important inputs towards resolving critical scientific uncertainties).
25 The effect of mortality from exposure to PM or other pollutants in this document is usually
26 expressed as a relative risk or risk rate (RR) relative to a baseline mortality or morbidity rate.
27 The crude mortality rates in 88 cities in 48 contiguous states in the NMMAPS study ranged from
28 about 8 deaths per day per million population in Denver, CO to about 40 per day per million in
29 St. Petersburg, FL. It is likely that age-adjusted rates such as those used in the APHEA 2 study
30 (Katsouyanni et al., 2001) would have shown a smaller range. As reported in Samet et al.
31 (2000a), there was little association between PM10 effect size and crude mortality rate in the
April 2002 8-10 DRAFT-DO NOT QUOTE OR CITE
-------
1 continental U.S. cites; however, Katsouyanni et al. (2001) found a negative relation between
2 PM10-equivalent effect size estimates and age-adjusted mortality rate in 29 European cities.
3 We plotted the relationship between increased or decreased mortality rate in NMMAPS for
4 ranges between the 25th and 75 percentiles (results not shown), but there was little apparent new
5 information in those plots other than the RR.
6 The PM increments used in this document to convert regression coefficients into
7 meaningful increments of excess risk are based on data from the U.S. fine particle monitoring
8 network for 1999 and 2000, the most recent years available. The difference between the annual
9 mean and the annual 95th percentile was used to characterize annual variation within each site;
10 and the average across all sites was used to select an appropriate increment for short-term
11 studies, about 50 //g/m3 for PM10 and 25 //g/m3 for PM2 5 and PM10_2 5, after rounding for ease of
12 calculation. As there is little experimental evidence about differences in effects of fine (PM25)
13 and coarse (PM10_25) particles, common increments are used for both. The difference between the
14 average of annual mean PM concentrations across all sites and the average of the annual 95th
15 percentiles across all sites was about 20 //g/m3 for PM10 and 10 //g/m3 for PM2 5 and PM10_25,
16 which are values used here for PM increments in long-term studies.
17 Thus, the pollutant increments utilized here to report Relative Risks (RR's) or Odds Ratio
18 for various health effects are: for PM10, 50 //g/m3; for PM25, 25 //g/m3; for SO4=, 155 nmoles/m3
19 (15 //g/m3); and, for H+, 75 nmoles/m3 (3.6 //g/m3, if as H2SO4) for short-term (<24 h) exposure
20 studies. The increments for short-term studies are the same as used in the 1996 PM AQCD,
21 a choice now driven by current data. In the 1996 PM AQCD, the same increments were used for
22 the long- and short-term exposure studies. However, 20 //g/m3 is the increment used here for
23 PM10 and 10 //g/m3 for PM2 5 and PM10_25 for long-term exposure studies. These estimates
24 derived from new 1999-2000 data are smaller than these used for long-term studies in the 1996
25 PM AQCD.
26 Greater emphasis is placed in text discussions on integrating and interpreting findings from
27 the body of evidence provided by the newer studies (as well as relating them to those reviewed in
28 the 1999 PM AQCD), rather than detailed evaluation of each of the numerous newly available
29 studies. Particular emphasis is focused in the text on those studies and analyses thought to
30 provide the most pertinent information for U.S. standard setting purposes. For example, North
31 American studies conducted in the U.S. or Canada are generally accorded more text discussion
April 2002 8-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 than those from other geographic regions; and analyses using gravimetric (mass) measurements
2 are generally accorded more text attention than those using non-gravimetric ambient PM
3 measures, e.g., black smoke (BS) or coefficient of haze (COH). Also, more emphasis is placed
4 on text discussion of new multi-city studies that employ standardized methodological analyses
5 for evaluating PM effects across several or numerous cities and often provide overall effects
6 estimates based on combined analyses of information pooled across multiple cities.
7 In the sections that follow on PM mortality and morbidity effects, key points derived from
8 the 1996 PM AQCD assessment of then-available information are first concisely highlighted.
9 Succinct summary tables are included and key information is discussed below in the main text
10 with regard to the most important numerous new studies that have become available since that
11 prior PM AQCD. More detailed information for these and other newly available studies is
12 summarized in tabular form in Appendices 8A and 8B, in which important methodological
13 features and results are presented. The Appendix tables have a uniform general organization
14 with divisions that include: (1) information about study location and ambient PM levels,
15 (2) study description of methods employed, (3) results and comments and (4) quantitative
16 outcomes for PM measures.
17
18
19 8.2 MORTALITY EFFECTS OF PARTICIPATE MATTER EXPOSURE
20 8.2.1 Introduction
21 The relationship of PM and other air pollutants to excess mortality has been intensively
22 studied and has been an important issue addressed in previous PM criteria assessments (U.S.
23 Environmental Protection Agency, 1986, 1996a). Mortality is the most severe adverse health
24 endpoint and, in some ways, the easiest to study. Excellent death records are maintained at every
25 level of government in most all nations and are typically made available to researchers. Also,
26 from a narrowly technical point of view, individual deaths are more amenable to statistical
27 analyses, since individual deaths from natural causes (typically respiratory and cardiovascular
28 diagnoses) are statistically independent, except in rare extremely infectious instances. Individual
29 deaths are also non-recurring events, unlike hospital admissions or respiratory symptoms.
30
April 2002 8-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 Recent findings are evaluated here for the two most important epidemiology designs by
2 which mortality is studied: time-series mortality studies (Section 8.2.2) and prospective cohort
3 studies (Section 8.2.3). The time-series studies mostly assess acute responses to short-term PM
4 exposure, although some recent work suggests that time-series data sets are also useful to
5 examine responses to exposures over a longer time scale. Time-series studies use community-
6 level air pollution measurements to index exposure and community-level response (i.e., the total
7 number of deaths each day by age and/or by cause of death). Prospective cohort studies usefully
8 complement time-series studies; they use individual health records, with survival lifetimes or
9 hazard rates adjusted for individual risk factors, and typically evaluate human health impacts of
10 long-term PM exposures indexed by community-level measurements.
11
12 8.2.2 Mortality Effects of Short-Term Particulate Matter Exposure
13 8.2.2.1 Summary of 1996 Particulate Matter Criteria Document Findings and Key Issues
14 The time-series mortality studies reviewed in the 1996 and other past PM AQCD's
15 provided much evidence that ambient PM air pollution is associated with increases in daily
16 mortality. The 1996 PM AQCD assessed about 35 PM-mortality time-series studies published
17 between 1988 and 1996. Information derived from those studies was consistent with the
18 hypothesis that PM is a causal agent in the short-term mortality impacts of air pollution.
19 The PM10 relative risk estimates derived from short-term PM10 exposure studies reviewed
20 in the 1996 PM AQCD suggested that an increase of 50 //g/m3 in the 24-h average of PM10 is
21 most clearly associated with an increased risk of premature total nonaccidental mortality (total
22 deaths minus those from accident/injury) on the order of relative risk (RR) = 1.025 to 1.05 in the
23 general population or, in other words, 2.5 to 5.0% excess deaths per 50 //g/m3 PM10 increase.
24 Higher relative risks were indicated for the elderly and for those with pre-existing
25 cardiopulmonary conditions. Also, based on the then recently published Schwartz et al. (1996a)
26 analysis of Harvard Six City data, the 1996 PM AQCD found the RR for excess total mortality in
27 relation to 24-h fine particle concentrations to be in the range of RR = 1.026 to 1.055 per
28 25 //g/m3 PM2 5 (i.e., 2.6 to 5.5% excess risk per 25 //g/m3 PM2 5 increment).
29 While numerous studies reported PM-mortality associations, important issues needed to be
30 addressed in interpreting their findings. The 1996 PM AQCD extensively discussed most critical
31 issues, including: (1) seasonal confounding and effect modification; (2) confounding by weather;
April 2002 8-13 DRAFT-DO NOT QUOTE OR CITE
-------
1 (3) confounding by co-pollutants; (4) measurement error; (5) functional form and threshold;
2 (6) harvesting and life shortening; and (7) the role of PM components. As important issues
3 related to model specification became further clarified, more studies began to address the most
4 critical issues, with some having been at least partially resolved, whereas others required still
5 further investigation. The next several paragraphs summarize the status of these issues at the
6 1996 PM AQCD publication time.
7 One of the most important components in time-series model specification is adjustment for
8 seasonal cycles and other longer-term temporal trends. Residual over-dispersion and
9 autocorrelation result from inadequate control for these temporal trends, and not adequately
10 adjusting for them could result in biased RRs. Modern smoothing methods allow efficient fits of
11 temporal trends and minimize such statistical problems. Thus, most recent studies controlled for
12 seasonal and other temporal trends, and it was unlikely that inadequate control for such trends
13 seriously biased estimated PM coefficients. Effect modification by season was examined in
14 several studies. Season-specific analyses are often not feasible in small-sized studies (due to
15 marginally significant PM effect size), but some studies (e.g., Samet et al., 1996; Moolgavkar
16 and Luebeck, 1996) suggested that estimated PM coefficients varied from season to season.
17 It was not fully resolved, however, if these results represent real seasonal effect modifications or
18 may be due to varying extent of correlation between PM and co-pollutants or weather variables
19 by season.
20 While most available studies included control for weather variables, some reported
21 sensitivity of PM coefficients to weather model specification, leading some investigators to
22 speculate that inadequate weather model specifications may still have erroneously ascribed
23 residual weather effects to PM. Two PM studies (Samet et al., 1996, 1998; Pope and Kalkstein,
24 1996) involved collaboration with a meteorologist and utilized more elaborate weather modeling,
25 e.g., use of synoptic weather categories. These studies found that estimated PM effects were
26 essentially unaffected by the synoptic weather variables and also indicated that the synoptic
27 weather model did not provide better model fits in predicting mortality when compared to other
28 weather model specifications used in previous PM-mortality studies. Thus, these results
29 suggested that the reported PM effects were not explained by weather effects.
30 Many earlier PM studies considered at least one co-pollutant in the mortality regression,
31 and some also examined several co-pollutants. In most cases, when PM indices were significant
April 2002 8-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 in single pollutant models, addition of a co-pollutant diminished the PM effect size somewhat,
2 but did not eliminate the PM associations. When multiple pollutant models were performed by
3 season, the PM coefficients became less stable, again, possibly due to PM's varying correlation
4 with co-pollutants among season and/or smaller sample sizes. However, in many studies, PM
5 indices showed the highest significance (versus gaseous co-pollutants) in single and multiple
6 pollutant models. Thus, it was concluded that PM-mortality associations were not seriously
7 distorted by co-pollutants, but interpretation of the relative significance of each pollutant in
8 mortality regression as relative causal strength was difficult because of limited quantitative
9 information on relative exposure measurement/characterization errors among air pollutants.
10 Measurement error can influence the size and significance of air pollution coefficients in
11 time-series regression analyses and is also important in assessing confounding among multiple
12 pollutants, as varying the extent of such error among the pollutants could also influence the
13 corresponding relative significance. The 1996 PM AQCD discussed several types of such
14 exposure measurement or characterization errors, including site-to-site variability and site-to-
15 person variability—errors thought to bias the estimated PM coefficients downward in most cases.
16 However, there was not sufficient quantitative information available to estimate such bias.
17 The 1996 PM AQCD also reviewed evidence for threshold and various other functional
18 forms of short-term PM mortality associations. Several studies indicated that associations were
19 seen monotonically below the existing PM standards. It was considered difficult, however, to
20 statistically identify a threshold from available data because of low data density at lower ambient
21 PM concentrations, potential influence of measurement error, and adjustments for other
22 covariates. Thus, the use of relative risk (rate ratio) derived from the log-linear Poisson models
23 was considered adequate and appropriate.
24 The extent of prematurity of death (i.e., mortality displacement, or harvesting) in observed
25 PM-mortality associations has important public health policy implications. At the time of the
26 1996 PM AQCD review, only a few studies had investigated this issue. While one of the studies
27 suggested that the extent of such prematurity might be only a few days, this may not be
28 generalizable because this estimate was obtained for identifiable PM episodes. There was not
29 sufficient evidence to suggest the extent of prematurity for non-episodic periods, from which
30 most of the recent PM relative risks were derived. The 1996 PM AQCD concluded:
April 2002 8-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 "In summary, most available epidemiologic evidence suggests that increased mortality
2 results from both short-term and long-term ambient PM exposure. Limitations of available
3 evidence prevent quantification of years of life lost to such mortality in the population. Life
4 shortening, lag time, and latent period of PM-mediated mortality are almost certainly
5 distributed over long time periods, although these temporal distributions have not been
6 characterized." (p. 13-45)
7 Only a limited number of PM-mortality studies analyzed fine particles and chemically
8 specific components of PM. The Harvard Six Cities Study (Schwartz et al., 1996a) analyzed
9 size-fractionated PM (PM2 5, PM10/15, and PM10/15.2 5) and PM chemical components (sulfates and
10 H+). The results suggested that PM25 was most significantly associated with mortality among the
11 components of PM. While H+ was not significantly associated with mortality in this and an
12 earlier analysis (Dockery et al., 1992), the smaller sample size for H+ than for other PM
13 components made a direct comparison difficult. The 1996 PM AQCD also noted that mortality
14 associations with BS or COH reported in earlier studies in Europe and the U.S. during the 1950s
15 to 1970s most likely reflected contributions from fine particles, as those PM indices had low 50%
16 cut-off diameters (« 4.5yam). Furthermore, certain respiratory morbidity studies showed
17 associations between hospital admissions/visits with components of PM in the fine particle
18 range. Thus, the U.S. EPA 1996 PM AQCD concluded that there was adequate evidence to
19 suggest that fine particles play especially important roles in observed PM mortality effects.
20 Overall, then, the status of key issues raised in the 1996 PM AQCD can be summarized as
21 follows: (1) the observed PM effects are unlikely to be seriously biased by inadequate statistical
22 modeling (e.g., control for seasonality); (2) the observed PM effects are unlikely to be
23 significantly confounded by weather; (3) the observed PM effects may be to some extent
24 confounded or modified by co-pollutants, and such extent may vary from season to season;
25 (4) determining the extent of confounding and effect modification by co-pollutants requires
26 knowledge of relative exposure measurement characterization error among pollutants (there was
27 not sufficient information on this); (5) no clear evidence for any threshold for PM-mortality
28 associations was reported (statistically identifying a threshold from existing data was also
29 considered difficult, if not impossible); (6) some limited evidence for harvesting, a few days of
30 life-shortening, was reported for episodic periods (no study was conducted to investigate
31 harvesting in non-episodic U.S. data); (7) only a relatively limited number of studies suggested a
April 2002 8-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 causal role of fine particles in PM-mortality associations, but in the light of historical data,
2 biological plausibility, and the results from morbidity studies, a greater role for fine particles than
3 coarse particles was suggested in the 1996 PM AQCD as being likely. The AQCD concluded:
4 "The evidence for PM-related effects from epidemiologic studies is fairly strong, with most
5 studies showing increases in mortality, hospital admissions, respiratory symptoms, and
6 pulmonary function decrements associated with several PM indices. These epidemiologic
7 findings cannot be wholly attributed to inappropriate or incorrect statistical methods,
8 misspecification of concentration-effect models, biases in study design or implementation,
9 measurement of errors in health endpoint, pollution exposure, weather, or other variables,
10 nor confounding of PM effects with effects of other factors. While the results of the
11 epidemiology studies should be interpreted cautiously, they nonetheless provide ample
12 reason to be concerned that there are detectable human health effects attributable to PM at
13 levels below the current NAAQS." (p. 13-92)
14
15 8.2.2.2 Introduction to Newly Available Information on Short-Term Mortality Effects
16 Since the 1996 PM AQCD, numerous new studies have examined short-term associations
17 between PM indices and mortality. Newly available U.S. and Canadian studies on relationships
18 between short-term PM exposure and daily mortality are summarized in Table 8-1. More
19 detailed summaries of these and of other short-term exposure PM-mortality studies from other
20 geographic areas (e.g., Europe, Asia, etc) are described in Appendix Table 8A-1. Information on
21 study location, study period, levels of PM, outcomes, methods, results, and reported risk
22 estimates and lags is provided in Table 8A-1. In addition to these summary tables, discussion in
23 the text below highlights findings from several multi-city studies. Discussion of implications of
24 new study results for types of issues identified in foregoing text is mainly deferred to Section 8.4.
25 The summarization of studies in Table 8-1 and 8A-1 (and in other tables) is not meant to
26 imply that all listed studies should be accorded equal weight in the overall interpretive
27 assessment of evidence regarding PM-associated health effects. In general, increasing scientific
28 weight should be accorded to those studies (i.e., those not clearly flawed and which have
29 adequate control for confounding) in proportion to the precision of their estimate of a health
30 effect. Small studies and studies with an inadequate exposure gradient generally produce less
31 precise estimates than large studies with an adequate exposure gradient. Therefore, the range of
32 exposures (e.g., as indicated by the IQR), the size of the study as indexed by the total number of
April 2002 8-17 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-1. RECENT U.S. AND CANADIAN TIME-SERIES STUDIES OF PM-RELATED DAILY MORTALITY*
to
o
o
to
Reference
Location(s)
Pollutants in Models
Comments
Multi-City Mortality Studies in the U.S. and Canada
oo
oo
fe
H
6
o
o
H
O
O
H
W
O
PM10 studies using NMMAPS data
Samet et al. (2000a,b,c);
Dominici et al. (2000a,b);
Samet (2000)
Daniels et al. (2000)
Dominici et al. (2002)
88 cities in the 48 contiguous U.S. states
plus AK and HI, 1987-1994; mainly 20
largest.
20 cities in the 48 contiguous U.S. states,
1987-1994
88 cities in the 48 contiguous U.S. states,
1987-1994
PM10, O3, CO, NO2, SO2
PM10 only
PM10 only
Braga et al. (2000)
Five large U.S. cities: Chicago, IL;
Detroit, MI; Pittsburgh PA,
Minneapolis-St. Paul, MN; Seattle, WA
PM,n only
Numerous models; range of PM10 values
depending on city, region, co-pollutants.
Pooled estimates for 88 cities, individual
estimates for 20 largest with co-pollutant
models
Smooth non-parametric spline model for
concentration-response functions. Average
response curve nearly linear.
Smooth non-parametric spline models for PM10
concentration-response functions. Average
response curves are nearly linear in the
industrial Midwest and Northeast regions, and
overall, but non-linear (usually concave) in the
other regions. Possible thresholds in
Southwest, Southeast.
Pooled estimate across cities adjusted for
influenza epidemics.
Brief summary of new time-series studies on daily mortality since the 1996 Air Quality Criteria Document for Paniculate Matter (U.S. Environmental
Protection Agency, 1969a). More complete descriptive summaries are provided in Appendix Table 8A-1. The endpoint is total daily non-trauma mortality
unless noted otherwise. Due to the large number of models reported for sensitivity analyses for some of these papers, some evaluating various lags and
co-pollutant models, some for individual cities and others for estimates pooled across cities, quantitative risk estimates are not presented in this table.
Specific mortality risk estimates for fine and coarse particle models are shown in Table 8-2. Multiple-pollutant models are discussed in Section 8.4.2.2.
O
-------
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY
to
o
o
to
Reference
Location(s)
Pollutants in Models
Comments
Multi-City Mortality Studies in the U.S. and Canada (cont'd)
oo
VO
fe
H
6
o
o
H
O
O
H
W
O
Studies using everyday PM10 data
Schwartz (2000b)
Schwartz and Zanobetti
(2000)
Zanobetti and Schwartz
(2000)
Moolgavkar (2000a)
Popeetal. (1999a)
Laden et al. (2000)
Same ten U.S. cities as in
(Schwartz, 2000a)
Same ten U.S. cities as in
(Schwartz, 2000a)
Four large U.S. cities: Chicago, IL;
Detroit, MI, Minneapolis-St. Paul, MN;
Pittsburgh, PA
Three large U.S. counties (cities):
Cook City (Chicago), IL; Los Angeles,
CA; Maricopa Cty. (Phoenix), AZ.
Ogden, Provo-Orem, and Salt Lake City,
UT.
Same six cities as in Harvard Six city
study, with Harvard air monitors and
community daily mortality time series:
Boston (Watertown), MA, Harriman-
Kingston, TN; Portage-Madison, WI;
St. Louis, MO; Steubenville, OH;
Topeka, KS.
PM10 only.
PM10 only.
PM,n only.
PM10 in all three; PM2 5 in
Los Angeles. O3, CO, NO2,
and SO, in some models.
PM10 only in all three.
Chemically speciated PM2 5,
and factors aligned with
putative sources for each
city identified by specific
chemical elements as tracers.
Several pooled estimates across cities evaluated
for single day, moving average, and distributed
lags.
Pooled estimates of concentration-response
functions across cities using smooth semi-
parametric functions of PM10 with the same
span of 0.
Pooled estimate of effect size across cities was
modified somewhat by race and gender.
The results showed little consistency for
different time lags and cities, the PM10 or PM2 5
effects on CVD mortality were greatly
attenuated by including one or more gaseous
co-pollutants
Positive, significant and similar effects for
PM10 on total, CVD, and respiratory mortality
Different coefficients in different cities,
depending on source type, chemical indicators,
and principal factor method. The motor vehicle
combustion component was significant, other
factors occasionally, but not the crustal element
component.
O
-------
to
o
o
to
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY
Reference
Location(s)
Pollutants in Models
Comments
Multi-City Mortality Studies in the U.S. and Canada
oo
to
o
Tsaietal. (1999, 2000)
Clyde et al. (2000)
Burnett et al. (2000)
Burnett etal. (1998a)
Camden, Elizabeth, and Newark, NJ.
Phoenix, AZ, May, 1995-March, 1998.
Seattle, WA, 1990-1995.
PM2 5, PM15, sulfates.
PM2 5, PM10_2 5 in Phoenix.
PM10, PM2 5, nephelometer,
SO, in Seattle.
Eight Canadian cities: Montreal, Ottawa, PM10, PM2 5, PM10_2 5, SO4,
Toronto, Windsor, Calgary, Edmonton, O3, CO, NO2, SO2
Winnipeg, Vancouver
Eleven Canadian cities. 1980-1991.
Main emphasis on O3, CO,
NO2, SO2. PM2 5, PM10.2 5,
SO4 on varying schedules.
Significant effects of PM25, PM10, and sulfates
in Newark, Camden at most lags, but not
Elizabeth.
PM10_2 5 significant in most of the 25 "best"
models for Phoenix, PM2 5 in almost none.
PM2 5 and PM10 in some models for Seattle,
none in the 5 best.
Significant effects of PM2 5 and PM10, less so
for PM10_2 5; particle effects stable, co-pollutant
effects decreased by particles
Qualitative indication of effect modification
of gaseous pollutant effects by particles.
fe
H
6
o
o
H
O
O
H
W
O
Klemm et al. (2000)
Schwartz et al. (2002)
Same six cities as (Laden et al., 2000)
1979-1988.
Same six cities as (Laden et al., 2000)
1979-1988
PM10, PM2.5, PM10.2.5, S04
PM25, PM10.25, 15 elements in
PM2 5, O3, CO, NO2, SO2
Replicated Schwartz et al. (1996a) with
additional sensitivity analyses.
Five source factors identified, as in (Laden
et al., 2000). Meta-smoothing of
non-parametric concentration-mortality curves
for PM2 5 and for five source factors. Total and
"traffic" source PM2 5 significantly associated
with mortality, nearly linear for PM2 5, steeper
slope at low concentrations of traffic particles.
No apparent threshold.
O
-------
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY
to
o
o
to
Reference
Location(s)
Pollutants in Models
Comments
oo
to
fe
H
6
o
o
H
O
O
H
W
O
Single-City Mortality Studies in the U.S. and Canada
Ostro et al. (1999a, 2000) Coachella Valley (Palm Springs), CA
Fairley (1999)
Schwartz etal. (1999)
Schwartz and Zanobetti
(2000)
Lippmann et al. (2000)
Chock et al. (2000)
Santa Clara County (San Jose), CA
Spokane, WA
Chicago, IL
Detroit, MI
Pittsburgh, PA
PM10 in earlier study, PM2 5
and PM10_2 5 in later study; O3,
CO,NO2
PM10, PM2 5, PM10.2 5, sulfates,
nitrates, O3, CO, NO2.
PM10 only
PM,n only
PM10, PM2 5, PM10.2.5,
sulfates, acidity, TSP, O3,
CO, NO2, SO2
PM10,PM25,PM10.25,03,
CO, NO2, SO2
PM2 5 effects significant, PM10 and PM
effects non-significant for total mortality; for
cardiovascular mortality, PM10 and PM10_2 5
significant, PM2 5 not
All significant in one-pollutant models, nitrates
significant in all multi-pollutant models, PM2 5
significant except with particle nitrates.
No association between mortality and high
PM10 concentrations on dust storm days with
high crustal particles.
Larger effects with longer-term PM10 and
mortality moving averages for total, in-hospital,
and out-of-hospital mortality.
Positive but non-significant effects on mortality
for the 1992-1994 data, but significant effects
for respiratory mortality vs. PM10 or TSP in
1985-1990 data.
Fine and coarse particle data on about 1/3 of
days with PM10. Data split into ages < 75 and
75+, and seasons. Significant effects for PM10,
not for size fractions. Regional sulfate, traffic-
related PM, and biogenic combustion factors
have maximum associations on different lag
days.
O
-------
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY
to
o
o
to
Reference
Location(s)
Pollutants in Models
Comments
oo
to
to
H
6
o
*
o
H
O
O
H
W
O
Single-City Mortality Studies in the U.S. and Canada
Klemm and Mason (2000) Atlanta, GA
Gwynn et al. (2000)
Schwartz (2000c)
Lipfert et al. (2000a)
Levy (1998)
Mar etal. (2000)
Buffalo, NY
Boston, MA
Philadelphia, PA-Camden, NJ seven-
county area
King County (Seattle), WA
Phoenix, AZ, near the EPA platform
monitor.
PM2 5, PM10_2 5, nitrate,
oxygenated hydrocarbons
(HC), elemental carbon (EC),
organic carbon (OC)
PM10, CoH, H+, SO4, O3, CO,
NO2, SO2.
PM,
PM10, PM2 5, PM10.2 5, sulfates,
acids, metals, O3, CO, NO2,
S02.
(nephelometer), PM10,
CO, SO,.
PM10, PM2 5, PM10.2 5, fine
particle elements, estimated
soil and non-soil PM, EC,
OC, O3, CO, NO2, SO2;
sources by factor scores.
No significant effects due to short time series,
ca. one year. Larger effect and shorter
confidence interval for PM2 5 than for PM10_2 5.
All PM components significantly associated
with total mortality in single-pollutant models,
not gaseous pollutants.
Larger effects with longer-term PM2 5 and
mortality moving averages (span 15 to 60 days)
for total and cause-specific mortality.
Exploration of mortality in different areas
relative to air monitor location. Peak O3 very
significant, greatly reduces PM effects.
PMj associated only with out-of-hospital
ischemic heart disease deaths, total mortality
with neither PM10 nor PM[
Total mortality significantly associated with
NO2, CO, weakly with PM10, PM10.2 5, EC, SO2.
Cardiovascular mortality significantly
associated with PM10, PM2 5, PM10.2 5, EC, OC,
CO, NO2, SO2, source factors.
O
-------
to
o
o
to
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY
Reference
Location(s)
Pollutants in Models
Comments
Single-City Mortality Studies in the U.S. and Canada
Clyde et al. (2000) Phoenix, AZ
Smith et al. (2000)
Phoenix, AZ (within city and within
county), 1995-1997.
PM10, PM2 5
PM25,PM10.2
oo
to
Gamble (1998)
Ostro (1995)
Dallas, TX
1990-1994
San Bernar
CA
1980-1986.
PM10, O3, CO, NO2, SO2
San Bernardino and Riverside Counties, PM2 5 estimated from visual
range,
O3
Effect on elderly mortality consistently higher
for PM10_2 5 among 25 "best" models. Estimates
combined using Bayesian model averaging.
Significant linear relationship with PM10_2 5, not
PM2 5. Piecewise linear models with possible
PM10.2 5 threshold for elderly mortality 20-25
O3, CO, NO2 significantly associated with
mortality, PM10 and NO2 not associated
Positive, significant PM2 5 effect only in
summer
Kelsall et al. (1997)
Philadelphia, PA
1974-1988
TSP, SO2, NO2, O3, CO
TSP, O3, CO, NO2 significant alone, TSP effect
reduced when SO, included.
O
O
O
H
O
H
W
O
Moolgavkar and Luebeck Philadelphia, PA 1973 -1988
(1996)
Murray and Nelson (2000) Philadelphia, PA, 1973-1990
TSP, O3, NO2, SO2.
TSP only
NO2 most significant pollutant, TSP effects
stronger in summer and fall.
Kalman filtering used to estimate hazard
function in a state space model. Both TSP and
the product of TSP and average temperature are
significant, but not together. Includes estimate
of at-risk population.
O
-------
to
o
o
to
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY
Reference
Location(s)
Pollutants in Models
Comments
oo
to
fe
H
6
o
o
H
O
O
H
W
O
Single-City Mortality Studies in the U.S. and Canada
Neasetal. (1999)
Schwartz (2000d)
Burnett etal. (1998b)
Philadelphia, PA
1973-1980
Philadelphia, PA
1974-1988
Toronto, ON, Canada 1980-1994
Goldberg et al. (2001a,b,c,d) Montreal, PQ, Canada, 1984-1995
Ozkaynaketal. (1996)
Toronto, ON, Canada 1970-1991
TSP only
TSP, SO2, humidity-
corrected extinction
coefficient
TSP, CoH, SO4=, CO, NO2,
SO2, O3,PM10andPM25
estimated from every-sixth-
day data and observed daily
SO4=, TSP, and CoH
PM2 5 and PM10 every sixth
day until 1992, daily through
1993. CO, NO2, NO, O3,
SO2. Missing PM data
estimated from sulfates, CoH,
extinction coefficient.
TSP, CoH, O3, CO, NO2, SO2
Case-crossover study. Significant TSP effect.
No SO2 effect when TSP in model. TSP
significant unless extinction coefficient in
model.
Significant excess total mortality for PM2 5,
PM10, TSP
Excess total and cause-specific mortality with
most PM indices reported (estimated PM2 5,
sulfates, CoH). In the age 65+ age group, total
mortality significantly elevated in individuals
with prior cancer, acute lower resp. disease,
any cardiovascular disease, chronic coronary
artery disease, congestive heart failure.
Significant association with 0-day lag TSP.
Factor analysis identified a factor with high
loadings on CoH, CO, and NO2 (traffic
presumably) significantly associated with total
and most cause-specific deaths.
O
-------
1 observations (e.g., days) and total number of events (i.e., total deaths), and the inverse variance
2 for the principal effect estimate are all important indices useful in determining the likely
3 precision of health effects estimates and in according relative scientific weight to the findings of
4 a given study.
5 As can be seen in Tables 8-1 and 8A-1, with a few exceptions, nearly all of the newly
6 reported analyses continue to show statistically significant associations between short-term (24 h)
7 PM exposures indexed by a variety of ambient PM measurements and increases in daily mortality
8 in numerous U.S. and Canadian cities, as well as elsewhere around the world. Also, the effects
9 estimates from the newly reported studies are generally consistent with those derived from the
10 earlier 1996 PM AQCD assessment, with the newly reported PM risk estimates generally falling
11 within the range of ca. 1 to 8% increase in excess deaths per 50 //g/m3 PM10 and ca. 2 to 6%
12 increase per 25//g/m3 PM2 5. Several newly available PM epidemiology studies which
13 conducted time-series analyses in multiple cities are of particular interest, as discussed below.
14
15 8.2.2.3 New Multi-City Studies
16 The new multi-city studies are of particular interest here due to their evaluation of a wide
17 range of PM exposures and large numbers of observations holding promise of providing more
18 precise effects estimates than most smaller scale independent studies of single cities. Another
19 major advantage of the multi-city studies, over meta-analyses for multiple "independent" studies,
20 is the consistency in data handling and model specifications, which eliminates variation due to
21 study design. Further, unlike regular meta-analysis, they clearly do not suffer from potential
22 omission of negative studies due to "publication bias". Furthermore, geographic patterns of air
23 pollution effects can be systematically evaluated in multiple-city analyses. Thus, the results from
24 multi-city studies can provide especially valuable evidence regarding the consistency and/or
25 heterogeneity, if any, of PM-health effects relationships across geographic locations. Also, many
26 of the cities included in these multi-city studies were ones for which no time-series analyses had
27 been previously reported.
28
29
30
April 2002 8-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.2.3.1 U.S. Multi-City Studies
2 U.S. PM10 20-Cities and 90-Cities NMMAPS Analyses
3 The National Morbidity, Mortality, and Air Pollution Study (NMMAPS) focused on time-
4 series analyses of PM10 effects on mortality during 1987-1994 in the 90 largest U.S. cities (Samet
5 et al., 2000a,b), in the 20 largest U.S. cities in more detail (Dominici et al., 2000a), and PM10
6 effects on emergency hospital admissions in 14 U.S. cities (Samet et al., 2000a,b). These
7 NMMAPS analyses are marked by extremely sophisticated statistical approaches addressing
8 issues of measurement error biases, co-pollutant evaluations, regional spatial correlation, and
9 synthesis of results from multiple cities by hierarchical Bayesian meta-regressions and
10 meta-analyses. These analyses provide extensive new information of much importance in being
11 among that most highly relevant to the setting of U.S. PM standards, because no other study has
12 examined as many U.S. cities in such a consistent manner. NMMAPS used only one consistent
13 PM index (PM10) across all cities (noted PM10 samples were only collected every 6 days in most
14 of the 90 cities); death records were collected in a uniform manner; and demographic variables
15 were uniformly addressed. Both the 20 and 90 cities analyses studies employ multi-stage models
16 (see Table 8-1) in which heterogeneity in individual cities' coefficients in the first stage GAM
17 Poisson models were evaluated in the second stage models with city or region specific
18 explanatory variables.
19 In both the 20 and 90 cities studies, the combined estimates of PM10 coefficients were
20 positively associated with mortality at all the lags examined (0, 1, and 2 day lags), although the
21 1-day lag PM10 resulted in the largest overall combined estimate. Figure 8-3 shows the estimated
22 percent excess total deaths per 10 //g/m3 PM10 at lag 1 day in the 88 (90 minus Honolulu and
23 Anchorage) largest cities, as well as (weighted average) combined estimates for U.S. geographic
24 regions depicted in Figure 8-4. The majority of the coefficients were positive for the various
25 cities listed along the left axis of Figure 8-3. The estimates for the individual cities were first
26 made independently, without borrowing information from other cities. The cities were then
27 grouped into the 7 regions seen in Figure 8-4 (based on characteristics of the ambient PM mix
28 typical of each region, as delineated in the 1996 PM AQCD). The bolded segments represent the
29 posterior means and 95% posterior intervals of the pooled regional effects under the more
30 conservative prior A for the heterogeneity across both regions and cities within regions. The
31 solid circles and squares denote, respectively, the overall regional means without and with
April 2002 8-26 DRAFT-DO NOT QUOTE OR CITE
-------
% Increase in Mortality per 10 |jg/m3 Increase in PM10
seat
sanj
denv
stoc
mode
spok
overall.1
°VeDhoe
sana
elpa
aust
albu
corp
lubb
overall.1
la
s a
rive
?s
overall.1
minn
wich
overall.1
chic
det
clev
pitt
i? 5
cine
akr
tole
overall.1
phil
bait
h t
dc
prov
syra
overall.1
nous
miam
atla
stpe
tamp
memp
orla
jckv
birm
char
nash
tuls
no
batr
Itrk
gmb
knox
?Sf
hunt
overall.1
rail-north
overall
-4 -2 (
I I
•
° j
=^0
6
0 °
«
e —
fi —
—
9 * 5-
e
•
D 2 4
I I
0
> o
-e
^r-°
n
tlT
^=
n
0 °
° n
(?
/P
>"
0 o
-e 5-
0 °
— **-2 —
— ^ o
n °
e
5~e
^^^^^^^
^
'- — o
e
i
e
^
-------
Northwest
Southern
California
Upper
Midwest
Industrial
Midwest
Northeast
Southwest 1 Southeast
Figure 8-4. Map of the United States showing the 90 cities (the 20 cities are circled) and
the seven regions considered in the NMMAPS geographic analyses. Regions:
Northwestern; Southern California; Southwest; Upper Midwest; Industrial
Midwest; Northeast; Southeast.
1
2
3
4
5
9
10
11
borrowing information from other regions, ("overall 1" = the regional mean without other
regions, "overall 2" = with information from other regions). The triangles and bolded segments
at the bottom of Figure 8-3 display combined estimates of nationwide overall effects of PM10 for
all cities overall, and for all cities minus those in the Northeast (overall-north).
Note that there appears to be some regional-specific variation in the overall combined
estimates, shown as "overall 1" and "overall 2" for the two sets of modeling assumptions and
specifications used in analyses combining data from all the cities in a given region. This can be
discerned more readily in Figure 8-5 (which depicts overall region-specific excess risk estimates
for day 0 and 2 day lags, as well as for lag 1 day). For example, the coefficients for the Northeast
are generally higher than for other regions (the Northeast combined estimate, 4.5% excess total
deaths per 50 //g/m3 increase in PM10, was about twice that for the 90-cities overall). The overall
April 2002
8-28
DRAFT-DO NOT QUOTE OR CITE
-------
>
CO
E
o 2-
a=
-&0-
ro.£
Ho
•- 0
0) C
w
CO
£ -1-
o '
_c
s£
^ ^ o* ,X £
X X
^ 0^
i
!
tlf + t
i
i
T- CN O T- (M
O) O> O) D) D)
CO CO CO CO CO
en rn "fo "co Ja
0)
s
0)
§
Figure 8-5. Percent excess mortality risk (lagged 0, 1, or 2 days) estimated in the
NMMAPS 90-City Study to be associated with 10-jUg/m3 increases in PM10
concentrations in cities aggregated within U.S. regions shown in Figure 8-2.
1 national combined estimate (i.e., at lag 1 day, 2.3% excess total deaths per 50 //g/m3 increase in
1 PM10) for the 90 cities is consistent with the range of estimates reported in the 1996 PM AQCD.
3 In the 90 cities study, the weighted second-stage regression included five types of county-
4 specific variables: (1) mean weather and pollution variables; (2) mortality rate (crude mortality
5 rate); (3) sociodemographic variables (% not graduating from high school and median household
6 income); (4) urbanization (public transportation); (5) variables related to measurement error
7 (median of all pair-wise correlations between monitors). Some of these variables were
8 apparently correlated (e.g., mean PM10 and NO2, household income and education) so that the
9 sign of coefficients in the regression changed when correlated variables were included in the
10 model. Thus, while some of the county-specific variables were statistically significant (e.g.,
1 1 mean NO2 levels), interpreting the role of these county-specific variables may require caution.
12 Regarding the heterogeneity of PM10 coefficients, the investigators concluded that they "did not
13 identify any factor or factors that might explain these differences".
14 Another important finding from Samet and coworkers' analyses was the weak influence of
15 gaseous co-pollutants on the PM10 effect size estimates. In both the 20 and 90 cities analyses,
16 PM10 coefficients changed little when O3 was added to regression models. Additions of a third
17 pollutant (i.e., PM10 + O3 + another gaseous pollutant) did reduce PM10 coefficients somewhat
April 2002
8-29
DRAFT-DO NOT QUOTE OR CITE
-------
1 (e.g., from -2.2 to ~ 1.7 per 50 //g/m3 PM10 at lag 1 day in the combined 90 cities analysis), but
2 the PM10 coefficients remained statistically significant at p< 0.05. The gaseous pollutants
3 themselves in single-, two-, and three-pollutant models were less consistently associated with
4 mortality than PM10. Ozone was not associated with mortality using year-round data; but, in
5 season-specific analyses, it was associated with mortality negatively in winter and positively in
6 summer. SO2, NO2, and CO were weakly associated with mortality, but additions of PM10 and
7 other gaseous pollutants did not always reduce their coefficients, possibly suggesting their
8 independent effects. As noted in Section 8.1, CO and NO2 from motor vehicles are likely
9 confounders of PM25 and, thus, of PM10 when it is not dominated by the coarse particle fraction.
10 The investigators concluded that the PM10 effect on mortality "did not appear to be affected by
11 other pollutants in the model".
12
13 U.S. 10-Cities Studies
14 In another set of multi-city analyses, Schwartz (2000a,b), Schwartz and Zanobetti (2000),
15 Zanobetti and Schwartz (2000), Braga et al. (2000), and Braga et al. (2001) analyzed 1987-1995
16 air pollution and mortality data from ten U.S. cities (New Haven, CT; Pittsburgh, PA;
17 Birmingham, AL; Detroit, MI; Canton, OH; Chicago, IL; Minneapolis-St. Paul, MN; Colorado
18 Springs, CO; Spokane, WA; and Seattle, WA.) or subsets (4 or 5 cities) thereof. The selection of
19 these cities was based on the availability of daily (or near daily) PM10 data. The main results of
20 the study were presented in the Schwartz (2000a) paper and the other studies noted above
21 focused on each of several specific issues, including: potential confounding, effect modification,
22 distributed lag, and threshold. In this section, the results for the Schwartz (2000a) main analyses
23 and that of Braga et al. (2000) on confounding are discussed, and results for analyses of other
24 specific issues are discussed later in appropriate sections. For each of the 10 cities, daily total
25 (non-accidental) mortality was fitted using a GAM Poisson model adjusting for temperature,
26 dewpoint, barometric pressure, day-of-week, season, and time. Deaths stratified by location of
27 death (in or outside hospital) were also examined. The data were also analyzed by season
28 (November through April as heating season). In the second stage, the PM10 coefficients were
29 modeled as a function of city-dependent covariates including co-pollutant to PM10 regression
30 coefficient (to test potential confounding), education, unemployment rate, poverty level, and
31 percent non-white. Threshold effects were also examined. The inverse variance weighted
April 2002 8-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 averages of the ten cities' estimates were used to combine results. PM10 was significantly
2 associated with total deaths, and the effect size estimates were the same in summer and winter.
3 Adjusting for other pollutants did not substantially change the PM10 effect size estimates. The
4 socioeconomic variables did not modify the estimates. The effect size estimates for the deaths
5 outside hospital were substantially greater than for inside hospital. The combined percent excess
6 death estimate for total mortality was 3.4% (95% CI: 2.7-4.1) per 50 //g/m3 increase in PM10, but
7 was larger for days with PM10 < 50 //g/m3.
8 Braga et al. (2000) evaluated potential confounding of the reported PM-mortality
9 associations by effects of respiratory epidemics, using data from a subset of 5 of the 10 cities
10 evaluated by Schwartz (2000a). When adjustments were made for respiratory epidemics, small
11 decreases in PM10 effects were seen in the cities evaluated. The overall estimated percent excess
12 deaths per 50 //g/m3 PM10 for the five cities was 4.3% (CI 3.0, 5.6) without control for respiratory
13 epidemics, but slightly decreased to 4.0% (CI 2.6, 5.3) with control for epidemics.
14
15 U.S. 3-Cities Study
16 Moolgavkar (2000a) evaluated associations between short-term measures of major air
17 pollutants and daily deaths in three large U.S. metropolitan areas (Cook Co., IL, encompassing
18 Chicago; Los Angeles Co., CA; and Maricopa Co., AZ, encompassing Phoenix) during a 9-year
19 period (1987-1995). Generalized additive models (GAM) were used in a standard manner to
20 conduct time-series Poisson regression analyses independently for each of the three cities
21 (allowing comparison of results across them not due to methodological differences), but no
22 combined analyses were attempted to derive overall PM effects estimates. Total non-accidental
23 deaths and cause-specific deaths from cardiovascular disease (CVD), cerebrovascular disease
24 (CrD), and chronic obstructive lung disease (COPD), and associated conditions were analyzed in
25 relation to 24-h readings for PM, O3, CO, NO2, SO2 averaged over all monitors in a given county.
26 Daily readings were available for each of the gaseous pollutants in all three countries, as were
27 PM10 values for Cook County. However, PM10 values were only available every sixth day in
28 Maricopa and Los Angeles Counties; as were PM2 5 values in Los Angeles Co. PM values were
29 highest in the winter and fall in Los Angeles Co., in the fall in Maricopa Co., and in summer in
30 Cook Co., whereas the gases (except for O3) were highest in winter in all three counties (O3 was
31 highest in summer in all three). The PM indices were moderately correlated (r = 0.30 to 0.73)
April 2002 8-31 DRAFT-DO NOT QUOTE OR CITE
-------
1 with CO, NO2, and SO2 in Cook Co. and Los Angeles Co., but poorly correlated (r < 0.22) with
2 those gases in Maricopa Co. Ozone was very poorly (r < 0.20) or negatively correlated with PM
3 or the other gases in each location (except for Cook Co., r = 0.36 for O3 vs PM10). Total
4 non-accidental, CVD, and COPD deaths were all highest during winter in all three counties, but
5 CrD deaths were relatively constant from season to season (no season-specific analyses reported).
6 Controlling for temperature and relative humidity effects in separate analyses for each
7 mortality endpoint for each of the three countries, varying patterns of results were found from
8 one location to another, as noted in Table 8A-1. In general, although PM10 in each of the three
9 counties (and PM25 in Los Angeles) and each of the gaseous pollutants (except O3) were all
10 statistically significantly associated with total non-accidental mortality at one or more lag times
11 (0 to 5 days) in single pollutant models, the PM effect estimates tended to be reduced and non-
12 significant in many of the multi-pollutant (PM plus one other gas or PM plus all others) analyses.
13 In contrast, effect estimates for several of the gases (CO, SO2, and NO2) tended to be more robust
14 than those for PM in multi-pollutant models, with their estimates remaining statistically
15 significant (although usually somewhat attenuated) at one or more lag times when included in
16 multi-pollutant models with PM10 or PM2 5. Similarly, a somewhat analogous varying pattern of
17 results was observed for the cause-specific mortality analyses (discussed further below in Section
18 8.2.2.5). That is, although PM10 or PM25 were statistically significantly related to CVD and
19 COPD-related (and to CrD only in Maricopa Co., lag 5) mortality in single pollutant models,
20 their coefficients were typically markedly reduced and became non-significant in multi-pollutant
21 analyses with one or more of the gases included in the model. Moolgavkar (2000a) concluded
22 that, while direct effects of individual components of air pollution cannot be ruled out, individual
23 components can best be thought of as indices of the overall air pollution mix; and he noted
24 considerable heterogeneity of air pollution effects across the three geographic areas evaluated.
25 Moolgavkar (2000a) did not calculate any pooled effect estimates possibly because of the
26 heterogeneity seen among the cities studied.
27
28 8.2.2.3.2 Canadian Multi-City Study Analyses
29 Urban Air Pollution Mix and Daily Mortality in 11 Canadian Cities
30 The number of daily deaths for non-accidental causes during 1980-1991 were obtained for
31 11 Canadian cities and linked to concentrations of ambient gaseous air pollutants using relative
April 2002 8-32 DRAFT-DO NOT QUOTE OR CITE
-------
1 risk regression models for longitudinal count data (Burnett et al., 1998a). The GAM Poisson
2 models used evaluated daily mortality versus O3, NO2, SO2 and CO (including adjustments for
3 seasonal cycles, day-of-week effects, and weather effects), but no PM indices were included in
4 their analyses because daily PM measurements were not available. However, data were available
5 for fine and coarse PM mass from dichot samples, and sulfates, on variable schedules somewhat
6 more frequently than once per six days in Montreal, Toronto, and Windsor (with smaller
7 numbers in the other cities). This allowed an ecologic comparison of gaseous pollutant risks by
8 mean fine particle concentration (their Figure 1). These comparisons suggested a weak negative
9 confounding of NO2 and SO2 effects with fine particles, and a weak positive confounding of
10 particle effects with O3.
11
12 Eight Largest Canadian Cities Study
13 Burnett et al. (2000) analyzed various PM indices (PM10, PM25, PM10_2 5, sulfate, COH, and
14 47 elemental component concentrations for fine and coarse fractions) and gaseous air pollutants
15 (NO2, O3, SO2, and CO) for association with total mortality in the 8 largest Canadian cities:
16 Montreal, Ottawa-Hull, Toronto, Windsor, Winnipeg, Calgary, Edmonton, and Vancouver. This
17 study differs from (Burnett et al., 1998a), including fewer cities but more recent years of data
18 (1986-1996 vs. 1980-1991) and detailed analyses of particle mass components by size and
19 elemental composition. Each city's mortality, pollution, and weather variables were separately
20 filtered for seasonal trends and day-of-week patterns. The residual series from all cities were
21 then combined and analyzed in a GAM Poisson model. The weather model was selected from
22 spline-smoothed functions of temperature, relative humidity, and maximum change in barometric
23 pressure within a day and with 0 and 1 day lags, using forward stepwise procedures. Pollution
24 effects were examined at lags 0 through 5 days. To avoid unstable parameter estimates in multi-
25 pollutant models, principal components were also used as predictors in the regression models.
26 Ozone was weakly correlated with other pollutants, and other pollutants were "moderately"
27 correlated with each other (the highest was r = 0.65 for NO2 and CO). The strongest association
28 with mortality for all pollutants considered were for 0 or 1 day lags. PM2 5 was a stronger
29 predictor of mortality than PM10_25. The gaseous pollutant effects estimates were generally
30 reduced by inclusion of PM25 or PM10, but not PM10_2 5, where strength of prediction is measured
31 by the t value or statistical significance of the excess risk. In addition to the results implicating
April 2002 8-33 DRAFT-DO NOT QUOTE OR CITE
-------
1 the fine particle fraction (PM2 5) most clearly, other findings on fine particle components were
2 also of interest. Specifically, sulfate, Fe, Ni, and Zn were most strongly associated with
3 mortality. The total effect of these four components was greater than that for PM2 5 mass alone,
4 the authors suggesting that the characteristics of the complex chemical mixture in the fine
5 fraction may be a better predictor of mortality than the mass index alone.
6
7 8.2.2.3.3 European Multi-City APHEA Study Analyses
8 The Air Pollution and Health: a European Approach (APHEA) project is a multi-center
9 study of short-term effects of air pollution on mortality and hospital admissions with a wide
10 range of geographic, climatic, sociodemographic, and air quality patterns. The obvious strength
11 of this approach is to be able to evaluate potential effect modifiers in a consistent manner.
12 It should be noted that PM indices measured in those cities varied. In APHEAl, the PM indices
13 measured were mostly black smoke (BS), except for: Paris, Lyon (PM13); Bratislava, Cologne,
14 and Milan (TSP); and Barcelnoa (BS and TSP). In APHEA2, 10 out of the 29 cities used actual
15 PM10 measurements; in 11 additional cities, PM10 levels were estimated based on regression
16 models relating collocated PM10 measurements to BS or TSP. In the remaining 8 cities, only BS
17 measurements were available (14 cities had BS measurements). As discussed below, there have
18 been several papers published that present either a meta-analysis or pooled summary estimates of
19 these multi-city mortality results: (1) Katsouyanni et al. (1997) — SO2 and PM results from
20 12 cities; (2) Touloumi et al. (1997) — ambient oxidants (O3 and NO2) results from six cities;
21 (3) Zmirou et al. (1998) — cause-specific mortality results from 10 cities (see Section 8.2.2.5);
22 (4) Samoli et al. (2001) — a reanalysis of APHEAl using a different model specification to
23 control for long-term trends and seasonality; and (5) Katsouyanni et al. (2001) — APHEA2, with
24 emphasis on the examination of confounding and effect modification.
25
26 APHEAl Sulfur Dioxide and Particulate Matter Results for 12 Cities
27 The Katsouyanni et al. (1997) analyses evaluated data from the following cities: Athens,
28 Barcelona, Bratislava, Cracow, Cologne, Lodz, London, Lyons, Milan, Paris, Poznan, and
29 Wroclaw. In the western European cities, an increase of 50 //g/m3 in SO2 or BS was associated
30 with a 3% (95% CI = 2.0, 4.0) increase in daily mortality; and the corresponding figure was 2%
31 (95% CI = 1.0, 3.0) for estimated PM10 (they used conversion: PM10 = TSP*0.55). In the
April 2002 8-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 central/eastern European cities, the increase in mortality associated with a 50 //g/m3 change was
2 0.8% (CI = -0.1, 2.4) for SO2 and 0.6% (CI = 0.1, 1.1) per 50 //g/m3 change in BS. Estimates of
3 cumulative effects of prolonged (two to four days) exposure to air pollutants were comparable to
4 those for one day effects. The effects of both pollutants (BS, SO2) were stronger during the
5 summer and were mutually independent. Regarding the contrast between the western and
6 central/eastern Europe results, the authors speculated that this could be due to: difference in
7 exposure representativeness; difference in pollution toxicity or mix; difference in proportion of
8 sensitive sub-population; and model fit for seasonal control. Bobak and Roberts (1997)
9 commented that the heterogeneity between central/eastern and western Europe could be due to
10 the difference in mean temperature. However, Katsouyanni and Touloumi (1998) noted that,
11 having examined the source of heterogeneity, other factors could apparently explain the
12 difference in estimates as well as or better than temperature.
13
14 APHEA1 Ambient Oxidants (Ozone and Nitrogen Dioxide) Results for Six Cities
15 Touloumi et al. (1997) reported on additional APHEA data analyses, which evaluated
16 (a) short-term effects of ambient oxidants on daily deaths from all causes (excluding accidents),
17 and (b) impacts on effect estimates for NO2 and O3 of including a PM measure (BS) in
18 multi-pollutant models. Six cities in central and western Europe provided data on daily deaths
19 and NO2 and/or O3 levels. Poisson autoregressive models allowing for overdispersion were
20 fitted. Significant positive associations were found between daily deaths and both NO2 and O3.
21 Increases of 50 //g/m3 in NO2 (1-hour maximum) or O3 (1-hour maximum) were associated with
22 a 1.3% (95% CI 0.9-1.8) and 2.9% (95% CI 1.0-4.9) increase in the daily mortality, respectively.
23 There was a tendency for larger effects of NO2 in cities with higher levels of BS: when BS was
24 included in the model, the pooled estimate for the O3 effect was only slightly reduced, but the
25 coefficient for NO2 was reduced by half (but remained significant). The authors speculated that
26 the short-term effects of NO2 on mortality might be confounded by other vehicle-derived
27 pollutants (e.g., airborne ambient PM indexed by BS measurements). Thus, while this study
28 reports only relative risk levels for NO2 and O3 (but not for BS), it illustrates the importance of
29 confounding of NO2 and PM effects and the relative limited confounding of O3 and PM effects.
30
31
April 2002 8-35 DRAFT-DO NOT QUOTE OR CITE
-------
1 APHEA1: A Sensitivity Analysis for Controlling Long-Term Trends and Seasonally
2 In order to investigate further the source of the regional heterogeneity of PM effects and to
3 examine the sensitivity of the RRs, the APHEA1 data were reanalyzed by APHEA investigators
4 (Samoli et al., 2001). Unlike previous analysis (i.e., analysis by Katsouyanni et al., 1997) in
5 which sinusoidal terms for seasonal control and polynomial terms for weather were used, the
6 investigators this time used a GAM model with smoothing terms for seasonal trend and weather,
7 which is a more commonly used approach in recent years. Using this model, the estimated
8 relative risks for central-eastern cities were larger than those obtained in the previous analysis,
9 reducing the contrast of estimated PM effects between central-eastern and western European
10 cities. Also, restricting the analysis to days with concentration < 150 ug/m3 further reduced the
11 differences between the western and central-eastern European cities. The authors conclude that
12 part of the heterogeneity in the estimated air pollution effects between western and central -
13 eastern cities in previous publications was caused by the statistical approach and the data range.
14 These results indicate that the apparent regional heterogeneity could be somewhat sensitive to
15 model specification. Since the number of cities used in the APHEA1 study is relatively small
16 (eight western and five central-eastern cities), the apparent regional heterogeneity found in the
17 earlier publications could also be due to chance. Thus, such heterogeneity may be sensitive to
18 model specification and/or choice of data range. The combined estimate for 50 //g/m3 increase in
19 PM10 was reported to be 3.3% (95CI: 2.6, 4.1)
20
21 APHEA2: Confounding and Effect Modification Using Extended Data
22 The APHEA2 analyses (Katsouyanni et al. 2001) included more cities (29 cities) and a
23 more recent study period (variable years in 1990-1997, as compared to 1975-1992 in APHEA1).
24 As with the recent reanalysis of APHEA1 by Samoli et al. (2001), APHEA2 analyses used a
25 GAM Poisson model with a smoother to control for season and trends. The analyses put
26 emphasis on effect modification by city-specific factors. Thus, the city-specific coefficients from
27 the first stage of Poisson regressions were modeled in the second stage regression using city-
28 specific characteristics as explanatory variables. Inverse-variance weighted pooled estimates
29 (fixed-effects model) were obtained as part of this model. When substantial heterogeneity was
30 observed, the pooled estimates were obtained using random-effects models. These city-specific
31 variables included: (1) air pollution level and mix, such as average air pollution levels and
April 2002 8-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM/NO2 ratio (as an indicator of traffic-generated PM); (2) climatic variables, such as mean
2 temperature and relative humidity; (3) health status of the population, such as the age-adjusted
3 mortality rates, the percentage of persons over 65 years of age, and smoking prevalence;
4 (4) geographic area (three regions: central-eastern, southern, and north-western). The study also
5 addressed the issue of confounding by simultaneous inclusion of gaseous co-pollutants in city-
6 specific regressions, and obtaining the pooled PM estimates for each co-pollutant included.
7 Unlike APHEA1, in which the region (larger PM estimates in western Europe than in
8 central-eastern Europe) was highlighted as the important factor, APHEA2 found several effect
9 modifiers. NO2 (i.e., index of high pollution from traffic) was an important one. The cities with
10 higher NO2 levels showed larger PM effects. That is, the estimated PM10 risk was approximately
11 4-fold in cities with NO2 levels in the 75th percentile ("high"), as compared to cities with NO2
12 levels in the 25th percentile ("low") of the distribution. The cities with warmer climate showed
13 larger PM effects. The investigators noted that this might be due to the better estimation of
14 population exposures with outdoor community monitors (because of more open windows). Also,
15 the cities with low standardized mortality rate showed larger PM effects. The investigators
16 speculated that this may be because a smaller proportion of susceptible people (to air pollution)
17 are available in a population with a large age-standardized mortality rate. Interestingly, in the
18 pooled PM risk estimates from models with gaseous pollutants, it was also NO2 that affected
19 (reduced) PM risk estimates most. For example, in the fixed-effects models, approximately 50%
20 reductions in both PM10 and BS coefficients were observed when NO2 was included in the model.
21 SO2 only minimally reduced PM coefficients, whereas O3 actually increased PM coefficients.
22 Thus, in this analysis, NO2 was implicated both as a confounder and an effect modifier. The
23 overall combined estimate for total mortality for PM10 or BS was 3.0% (95CI: 2.0, 4.1).
24
25 8.2.2.3.4 An Examination of Effect Modification Using Past Results
26 Levy et al. (2000) sought to explain the apparent heterogeneity of PM effects found in past
27 studies. Their analysis is different from the other multi-city studies discussed above in that they
28 analyzed the PM coefficients from past studies, rather than obtaining city-specific coefficients in
29 a consistent time-series model specification. However, their results are worth mentioning here,
30 as they examined various city-specific covariates that are similar to those examined in other
31 multi-city studies, as well those that are unique to their study. They applied an empirical Bayes
April 2002 8-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 meta-analysis to 29 PM estimates from 21 published studies; and, in a second stage regression,
2 they considered city-specific variables such as mortality rate, gaseous pollutants' regression,
3 coefficients (that is, regressing a gaseous pollutant on PM), PM10 levels, PM2 5/PM10 ratio, central
4 air conditioning prevalence, and heating/cooling degree days. Among these variables,
5 PM2 5/PM10 ratio was a significant predictor (larger PM estimates for higher PM2 5/PM10 ratio)
6 in the 19 U.S. cities data subsets. While sulfate data were not available for all the 19 studies, the
7 investigators noted that the sulfate/PM10 ratio was highly correlated with both the mortality
8 (r = 0.84) and with the PM2 5/PM10 ratio in the limited subset of data, indicating that the
9 sulfate/PM10 ratio could be an even better predictor of regional heterogeneity of PM risk
10 estimates. It would be interesting to estimate PM25/PM10 ratios or sulfate/PM10 ratios for a larger
11 U.S. dataset (e.g., Samet et al.'s 90 cities study) and examine if Levy et al.'s finding holds for
12 larger geographic coverage. After adjusting for city-specific covariates, Levy et al.'s combined
13 total mortality excess death estimate for the 19 U.S. PM studies was 3.5% (95% CI: 2.7, 4.4) per
14 50 //g/m3 increase in PM10.
15
16 8.2.2.3.5 Comparison of Effects Estimates from Multi-City Studies
17 In summary, based on pooled analyses of data combined across multiple cities, the percent
18 excess (total, non-accidental) deaths estimated per 50 //g/m3 increase in PM10 in the above
19 multi-city studies were: (1) 2.3% in the 90 largest U.S. cities (4.5% in the Northeast region);
20 (2) 3.4% in 10 U.S. cities; (3) 3.5% in the 8 largest Canadian cities; and (4) 2.0% in western
21 European cities (using PM10 = TSP*0.55) in the original APHEA1; (5) 3.3% in the reanalysis of
22 APHEA1; (6) 3.0% in APHEA2; and (7) 3.5% in Levy et al.'s analysis of the 19 U.S. studies.
23 These combined estimates are all consistent with the range of PM10 estimates previously reported
24 in the 1996PM AQCD.
25
26 8.2.2.4 The Role of Particulate Matter Components
27 Delineation of the roles of specific ambient PM components in contributing to associations
28 between short-term PM exposures and mortality requires evaluation of several factors, e.g., size,
29 chemical composition, surface characteristics, and presence of gaseous co-pollutants. While
30 possible combinations of interactions among these factors can in theory be limitless, the actual
31 data tend to cover definable ranges of aerosol characteristics and co-pollutant environments due
April 2002 8-38 DRAFT-DO NOT QUOTE OR CITE
-------
1 to typical source characteristics (e.g., fine particles tend to be combustion products in most
2 cities). Newly available studies conducted in the last few years have begun to provide more
3 extensive information on the issue of PM component roles; their results are discussed below in
4 relation to three topics: (1) PM particle size (e.g., PM25 vs. PM10_25); (2) chemical components;
5 and (3) source oriented evaluations.
6
7 8.2.2.4.1 Particulate Matter Particle Size Evaluations
8 Numerous studies published since the 1996 PM AQCD substantiate associations between
9 PM25 and increased total mortality. Consistent with the 1996 PM AQCD findings, effect size
10 estimates from the new studies generally fall within the range of 2 to 6% excess total mortality
11 per 25 //g/m3 PM2 5, with many being statistically significant at p<0.05.
12 With regard to the relative importance of the fine and coarse fractions of inhalable PM10
13 particles capable of reaching thoracic regions of the respiratory tract, at the time of the 1996 PM
14 AQCD, there was only one acute mortality study (Schwartz et al., 1996a) that examined this
15 issue. That study suggested that fine particles (PM25), distinctly more so than coarse fraction
16 (PM10_25) particles, were associated with daily mortality. A recent study (Klemm et al., 2000), to
17 reconstruct the data and to replicate the original analyses, essentially reproduced the original
18 investigators' results.
19 Since the 1996 PM AQCD, several new studies have used size-fractionated PM data to
20 investigate the relative importance of fine (PM25) vs. coarse (PM10_25) fraction particles.
21 Table 8-2 provides synopses of those studies with regard to the relative importance of the two
22 size fractions, as well as some characteristics of the data. The average levels of PM25 ranged
23 from about 13 to 20 //g/m3 in the U.S. cities, but much higher average levels were measured in
24 Mexico City (27.4 //g/m3) and Santiago, Chile (64.0 //g/m3). As can be seen in Table 8-2, in the
25 northeastern U.S. cities (Pittsburgh, Philadelphia, and Detroit) and Atlanta, GA, there was more
26 PM25 mass than PM10_25 mass on the average, whereas in the western U.S. (Phoenix, AZ;
27 Coachella Valley, CA; Santa Clara County, CA) the average PM10_2 5 levels were higher than
28 PM2 5 levels. It should be noted that the three Phoenix studies in Table 8-2 use much the same
29 data set, using fine and coarse particle data from EPA's 1995-1997 platform study. Seasonal
30 differences in PM component levels should also be noted. For example, in Santa Clara County
31 and in Santiago, Chile, the winter PM2 5 levels averaged twice those during summer. The
April 2002 8-39 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-2. SYNOPSIS OF SHORT-TERM MORTALITY STUDIES THAT
EXAMINED RELATIVE IMPORTANCE OF PM2 5 AND PM10 2 5
Author, City
Means (^g/m3); ratio
ofPM25toPM10;and
correlation between
PM2 5 and PM10.2 5.
Results regarding relative importance of
PM2 5 vs. PM10_2 5 and comments.
Fairley (1999).
Santa Clara County,
CA
Ostro et al. (2000).
Coachella Valley,
CA
Clyde et al. (2000).
Phoenix, AZ
Mar etal. (2000).
Phoenix, AZ
1995-1997.
Smith et al. (2000).
Phoenix, AZ
Lippmann et al.
(2000). Detroit, MI
1992-1994.
Lipfert et al. (2000a).
Philadelphia, PA
1992-1995.
PM25 mean= 13;
PM25/PM10 = 0.38;
r=0.51.
PM2 5 (Palm Springs and
Indio, respectively)
mean= 12.7, 16.8;
PM25/PM10 = 0.43,0.35;
r= 0.46, 0.28.
PM25mean = 13.8;
PM25/PM10 = 0.30;
r=0.65.
PM25 (TEOM) mean= 13;
PM25/PM10 = 0.28;
r=0.42.
Not reported, but likely
same as Clyde's or Mar's
data from the same
location.
PM25 mean=18;
PM25/PM10 =0.58;
r=0.42.
PM25 mean=17.3;
PM2J/PM10 =0.72.
Of the various pollutants including PM10, PM25, PM10_25,
sulfates, nitrates, COH, CO, NO2, and O3, strongest
associations were found for ammonium nitrate and PM2 5.
PM2 5 was significantly associated with mortality, but PM10_2 5
was not, separately and together in the model. Sulfate was a
significant predictor of mortality in single pollutant model,
but not when PM2 5 was included simultaneously. Winter
PM2 5 level is more than twice that in summer.
Total mortality was more significantly associated with PM2 5
than with PM10_2 5. Cardiovascular mortality was associated
with PM10_2.5 more significantly than with PM2 5, but their
effect size estimates per IQR were similar.
Using Bayesian Model Averaging that incorporates model
selection uncertainty, with 29 covariates (lags 0- to 3-day),
effects of coarse particles (most consistent at lag 1 day) were
found to be stronger than that for fine particles. The
association was for mortality confined to the region where
fine particles (PM2 5) are expected to be uniform.
Total mortality was weakly (p < 0.10) associated with PM10_25.
It was less strongly (p > 0.10) associated with PM25.
Cardiovascular mortality was both significantly associated
with PM2 5 (lags 1, 3, 4) and PM10.2 5 (lag 0).
In linear PM effect model, a statistically significant mortality
association found with PM10_2 5, but not with PM2 5. In models
allowing for a threshold, evidence of a threshold for PM2 5
(in the range of 20-25 ,wg/m3) suggested, but not for PM10_2 5.
Seasonal interaction in the PM10_2 5 effect also reported: the
effect being highest in spring and summer when anthropogenic
concentration of PM10_25 is lowest.
Both PM2 5 and PM10_2 5 were positively associated with
mortality outcomes to a similar extent. Simultaneous inclusion
of PM2 5 and PM10_2 5 also resulted in comparable effect sizes.
Similar patterns were seen in hospital admission outcomes.
The authors conclude that no systematic differences were seen
according to particle size or chemistry. However, when PM2 5
and PM10_2.5 were compared, PM2 5 (at lag 1 or average of lag 0
and 1) was more significantly (with larger attributable risk
estimates) associated with cardiovascular mortality than
PM10.,5.
April 2002
8-40
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-2 (cont'd). SYNOPSIS OF SHORT-TERM MORTALITY STUDIES THAT
EXAMINED RELATIVE IMPORTANCE OF PM2 5 AND PM10 2 5
Author, City
Means Og/m3); ratio
ofPM25toPM10;and
correlation between
PM2 5 and PM10.2 5.
Results regarding relative importance of
PM2 5 vs. PM10_2 5 and comments.
Klemm and Mason
(2000). Atlanta, GA
Klemm et al. (2000)
Chock et al. (2000).
Pittsburgh, PA
Burnett et al. (2000)
8 Canadian cities
1986-1996
Castillejos et al.
(2000). Mexico City.
1992-1995
Cifuentes et al.
(2000).
Santiago, Chile
1988-1996.
Anderson et al.
(2001). The west
Midlands
conurbation, UK.
1994-1996.
PM25mean = 19.9;
PM25/PM10 =0.65.
Mean PM2 s ranges from
11.3 in Portage to 29.6 in
Steubenville. Mean
PM10_2 5 ranges from 6.6 in
Portage to 16.1 in
Steubenville. Mean
PM2 5/PM10 ranges from
50.1%inTopekato66%
in Kinston-Harriman.
Data distribution not
reported.
PM25/PM10 * 0.67.
PM25 mean=13.3;
PM25/PM10=0.51;
r=0.37.
PM25mean=27.4;
PM25/PM10=0.61;
r=0.52.
PM25 mean=64.0;
PM25/PM10=0.58;
r=0.52.
PM25mean=14.5;
PM2J/PM10 =0.62;
r=0.92.
No significant associations were found for any of the
pollutants examined, possibly due to a relatively short study
period (1-year). The coefficient and t-ratio were larger for
PM2 5 than for PM10.2 5.
Significant associations between total mortality and PM2 5 in 3
cities and in pooled effect. No significant association with
PM10_2.5 in the replications study for any city.
Seasonal dependence of correlation among pollutants, multi-
collinearity among pollutants, and instability of coefficients
were all emphasized in discussion and conclusion. These
considerations and small size of dataset stratified by age group
and season limit confidence in results finding no consistently
significant associations for any size fraction.
PM2 5 was a stronger predictor of mortality than PM10_2 5. For
chemical species, sulfate ion, nickel, and zinc from the fine
fraction were most strongly associated with mortality.
Both PM2 5 and PM10_2 5 were associated individually with
mortality, but the PM10_2 5 effect size was larger and more
significant. When both were included in the model, the effect
size of PM10_2 5 remained the same but that of PM2 5 was
virtually eliminated.
Results were different for warmer and colder months. PM2 5
was more important than PM10_2 5 in the whole year and in
winter, but not in summer. The mean of PM2 5 was more than
twice higher in winter (82.4 ^g/m3) than in summer (32.8),
whereas the mean of PM10_2 5 was more comparable for winter
(49.9 Mg/m3) and for summer (42.9).
No significant association seen between total mortality and
any of the PM indices in the all year analysis, but PM10 and
PM2 5 were significantly associated with total mortality in the
warm season (April-September). PM10_2 5 was generally more
weakly associated with mortality outcomes than PM10 or PM2 5
April 2002
8-41
DRAFT-DO NOT QUOTE OR CITE
-------
1 temporal correlation between PM25 and PM10_25 ranged between 0.30 and 0.65. Such differences
2 in ambient PM mix characteristics from season to season or from location to location
3 complicates assessment of the relative importance of PM25 and PM10_2 5.
4 To facilitate a quantitative overview of the effect size estimates and their corresponding
5 uncertainties from these studies, the percent excess risks are plotted in Figure 8-6. These
6 excluded the Clyde et al. study, in which the model specification did not obtain RRs for PM2 5
7 and PM10_2 5 separately, and the Smith et al. study, which did not present linear term RRs for
8 PM25 and PM10_2 5. Note that, in most of the original studies, the RRs were computed for
9 comparable distributional features (e.g., inter quartile range, mean, 5th-to-95th percentile, etc.).
10 However, the increments derived and their absolute values varied across studies; and therefore,
11 the RRs used in deriving the excess risk estimates delineated in Figure 8-6 were re-computed for
12 consistent increments of 25 //g/m3 for both PM2 5 and PM10_25. Note also that re-computing the
13 RRs per 25 //g/m3 in some cases changed the relative effect size between PM2 5 and PM10_25, but
14 it did not affect the relative significance.
15 All of the studies found positive associations between both the fine and coarse PM indices
16 and increased mortality risk, with most for PM2 5 and a few for PM10_25 being statistically
17 significant. However, most of the studies did not have large enough sample sizes to separate out
18 what often appear to be relatively small differences in effect size estimates; but several do show
19 statistical distinctly larger and significant mortality associations with PM2 5 than for non-
20 significant PM10_2 5 effects. For example, the Klemm et al. (2000) recomputation of the Harvard
21 Six Cities time-series study reconfirmed the original Schwartz et al. (1996a) finding of PM2 5
22 being significantly associated with excess mortality, whereas PM10_2 5 was not. Similar results
23 were obtained by the other multi-city study, i.e., the 8 largest Canadian cities study by Burnett
24 et al. (2000), and by the Atlanta (Klemm and Mason, 2000), Santa Clara (Fairley, 1999), and the
25 Coachella Valley (Ostro et al., 2000) studies. There were two studies in which the importance of
26 PM2 5 and PM10_2 5 were considered to be similar or, at least, not distinguishable: Philadelphia, PA
27 (Lipfert et al., 2000a) and Detroit, MI (Lippmann et al., 2000). The three Phoenix studies
28 obtained "mixed" results, in that the Smith et al. (2000) and Clyde et al. (2000) analyses (not
29 shown in Figure 8-6) found PM10_2 5 to appear to be more important in explaining mortality than
30 PM2 5, but Mar et al. (2000) found both to be significant (as depicted in Figure 8-6). Also, the
31 Mexico City analysis by Castillejos et al. (2000) implicated PM10_25 as the apparent more
April 2002 8-42 DRAFT-DO NOT QUOTE OR CITE
-------
>
to
o
o
to
• Q
~
Lag 2 day MA } ^ - } Winter
o
HH
H
W
Figure 8-6. Percent excess risks estimated per 25 Aig/m3 increase in PM2 5 or PM10_2 5 from new studies evaluating both PM2
-2.5
and PM10_2 5 data for multiple years, based on single pollutant (PM only) models. All lags = 1 day, unless indicated
otherwise (See Section 8.4.2 for same studies shown here that found different risk estimates in MP models with
both fine and coarse particles included).
-------
1 important fraction of PM10. However, the Santiago, Chile study (Cifuentes et al., 2000) found
2 significant associations with both fine and coarse fractions and interesting seasonal differences,
3 as well. In Chock et al.'s (2000) analysis of Pittsburgh, PA data, the authors emphasized the lack
4 of significant PM associations; and no specific comments were made regarding the relative
5 importance of PM25 versus PM10_25.
6 The Canadian 8-city study (Burnett et al., 2000) is noteworthy for a variety of reasons,
7 including the use of elemental composition and principal components analyses to provide
8 additional information about the relative importance of fine and coarse particles. The PM25
9 effect on mortality is greater than the PM10_2 5 effect for all gaseous-pollutant models in Table 5 of
10 Burnett et al. (2000) and in the principal component model 1 in their Table 8, where both PM
11 size fractions and the four gaseous co-pollutants are used simultaneously. PM component
12 models from this study are discussed further below, in Section 8.2.2.4.2.
13 The Lippmann et al. (2000) results for Detroit are also noteworthy in that additional PM
14 indices were evaluated besides those depicted in Figure 8-6 and the overall results obtained may
15 be helpful in comparing fine- versus coarse-mode PM effects. In analyses of 1985 to 1990 data,
16 PM-mortality relative risks and their statistical significance were generally in descending order:
17 PM10, TSP-SO4=, and TSP-PM10. For the 1992-1994 period, relative risks for equivalent
18 distributional increment (e.g., IQR) were comparable among PM10, PM25, and PM10_25 for both
19 mortality and hospital admissions categories; and SO4= was more strongly associated with most
20 outcomes than FT. Consideration of the overall pattern of results led the authors to state that the
21 mass of the smaller size index could explain a substantial portion of the variation in the larger
22 size indices. In these data, on average, PM25 accounted for 60% of PM10 (up to 80% on some
23 days) and PM10 for 66% of TSP mass. Also, the temporal correlation between TSP and PM2 5
24 was r = 0.63, and for PM25 vs. PM10 r = 0.90, suggesting that much of the apparent larger particle
25 effects may well be mainly driven by temporally covarying smaller PM2 5 particles. The stronger
26 associations for sulfates than FT, suggestive of non-acid fine particle effects, must be caveated by
27 noting the very low FT levels present (often circa non-detection limit).
28 Three research groups have examined the same Phoenix, AZ data set, using different
29 methods. While these groups used somewhat different approaches, there is some consistency
30 among their results in that PM10_2 5 appeared to emerge as possibly the more important predictor
31 of mortality versus PM2 5. In the Clyde et al. (2000) analysis, PM-mortality associations were
April 2002 8-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 found only for the geographic area where PM2 5 was considered uniformly distributed, but the
2 association was with PM10_2 5, not PM2 5. Based on the Bayes Information Criterion, the highly
3 ranked models consistently included 1-day lagged PM10_25. Smith et al.'s (2000) analyses found
4 that, based on a linear PM effect, PM10_2 5 was significantly associated with total mortality, but
5 PM25 was not. In the Mar et al. (2000) analysis, total mortality was significantly associated with
6 CO and NO2 and weakly (p < 0.1) associated with PM10 and PM10_2 5 (and PM2 5 was more weakly
7 associated); however, cardiovascular mortality (CVM) was significantly associated with both
8 PM25 and PM10_2 5 at p<0.05. CVM was also significantly associated with a motor vehicle source
9 category with loading of PM2 5, EC, OC, CO, NO2, and some trace metals, as shown by factor
10 analyses discussed below. The PM2 5 in Phoenix is mostly generated from motor vehicles,
11 whereas PM10_25 consists mainly of two types of particles: (a) crustal particles from natural (wind
12 blown dust) and anthropogenic (construction and road dust) processes, and (b) organic particles
13 from natural biogenic processes (endotoxin and molds) and anthropogenic (sewage aeration)
14 processes.
15 The Castillejos et al. (2000) and Cifuentes et al. (2000) analyses also appear to implicate
16 PM10_2 5, as well as PM25, as importantly contributing to mortality in two non-U.S. locations,
17 Mexico City and Santiago, Chile. The latter study also suggests possible seasonal differences in
18 Santiago, the PM effects in summer being more than double those in winter at that South
19 American location.
20
21 Crustal Particle Effects
22 Since the 1996 PM AQCD, several studies have yielded interesting new information
23 concerning possible roles of crustal wind-blown particles or crustal particles within the fine
24 particle fraction (i.e., PM25) in contributing to observed PM-mortality effects.
25 Schwartz et al. (1999), for example, investigated the association of coarse particle
26 concentrations with non-accidental deaths in Spokane, Washington, where dust storms elevate
27 coarse particle concentrations. During the 1990-1997 period, 17 dust storm days were identified.
28 The PM10 levels during those storms averaged 263 //g/m3, compared to 39 //g/m3 for the entire
29 period. The coarse particle domination of PM10 data on those dust storm days was confirmed by
30 a separate measurement of PM10 and PMX during a dust storm in August, 1996: the PM10 level
31 was 187 //g/m3, while PMX was only 9.5 //g/m3. The deaths on the day of a dust storm were
April 2002 8-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 contrasted with deaths on control days (n=95 days in the main analysis and 171 days in the
2 sensitivity analysis), which are defined as the same day of the year in other years when dust
3 storms did not occur. The relative risk for dust storm exposure was estimated using Poisson
4 regressions, adjusting for temperature, dewpoint, and day of the week. Various sensitivity
5 analyses considering different seasonal adjustment, year effects, and lags, were conducted. The
6 expected relative risk for these storm days with an increment of 221 //g/m3 would be about 1.04,
7 based on PM10 relative risk from past studies, but the estimated RR for high PM10 days was found
8 to be only 1.00 (95% CIO.95-1.05) per 50 //g/m3 PM10 change in this study. Schwartz et al.
9 concluded that there was no evidence to suggest that coarse (presumably crustal) particles were
10 associated with daily mortality.
11 Pope et al. (1999a) investigated PM10-mortality associations in three metropolitan areas
12 (Ogden, Salt Lake City, and Provo/Orem) in Utah's Wasatch Front mountain region during the
13 1985-1995 period. Although the three metropolitan areas shared common weather patterns,
14 pollution levels and patterns among the three areas were different due to different emission
15 sources. The authors utilized an index of air stagnation (the clearing index which the National
16 Weather Service computes from temperature, moisture and wind) to identify and screen obvious
17 windblown dust days, days clearly identified as with low stagnation index but high PM10. They
18 found that Salt Lake City experienced substantially more episodes of wind-blown dust. They
19 therefore conducted Poisson regression of mortality series using both unscreened and screened
20 PM10 data. The effects of screening were most apparent in Salt Lake City results. Before
21 screening, no significant relationships were observed; after screening, the RRs per 50 //g/m3
22 increase in PM10 for mortality in the three metropolitan areas were 1.12 (95% CI: 1.045 - 1.20),
23 1.023 (1.00 - 1.047), and 1.019 (0.979 - 1.06) for Ogden, Salt Lake City, and Provo/Orem,
24 respectively. These results suggest that the pollution episodes of wind-blown (crustal-derived)
25 dusts were less associated with mortality than were the episodes of (presumably) combustion-
26 related particles.
27 Ostro et al. (1999a) analyzed the Coachella Valley, CA data for 1989-1992. This desert
28 valley, where coarse particles of geologic origin comprise circa 50-60% of annual-average PM10
29 (> 90% during wind episodes throughout the year), includes the cities of Palm Springs and Indio,
30 CA. Total, respiratory, cardiovascular, non-cardiorespiratory and age-over-50 deaths were
31 analyzed. The correlation between gravimetric and beta-attenuation measurements, separated by
April 2002 8-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 25 miles, was high (r = 0.93); and the beta-attenuation data were used for analysis. GAM
2 Poisson models adjusting for temperature, humidity, day-of-week, season, and time were used.
3 Seasonally stratified analyses were also conducted. Lags 0 through 3 days (separately) of PM10,
4 along with moving averages of 3 and 5 days, were evaluated, as were O3, NO2, and CO.
5 Associations were found between 2- or 3-day lagged PM10 and all mortality categories examined,
6 except non-cardiorespiratory. Effect size estimates for total and cardiovascular deaths were
7 larger for warm season (May through October) than for all year, analogous to the Cifuentes et al.
8 (2000) findings for Santiago, Chile. NO2 and CO were statistically significant predictors of
9 mortality in single pollutant models; but in multi-pollutant models, all gaseous pollutants
10 coefficients were reduced and non-significant, whereas PM10 coefficients remained the same and
11 significant. Ostro et al. (2000) also conducted a follow-up study of the Coachella Valley data for
12 1989-1998, using actual PM2 5 and PM10_25 data for the last 2.5 years but PM2 5 and PM10_25
13 concentrations estimated for the other, earlier years. PM2 5, CO, and NO2 were significantly
14 associated with all-cause mortality; and PM10 and PM10_2 5 with cardiovascular mortality, but not
15 PM2 5 (possibly due to the low range of concentrations and reduced sample size for PM2 5 data
16 versus PM10 data). Thus, although the cardiovascular mortality results hint at crustal particle
17 effects possibly being important in this desert situation, the ability to discern more clearly the
18 role of fine particles would likely be improved by analyses of more years of actual data for PM2 5.
19 Laden et al. (2000) analyzed Harvard Six Cities study data and Mar et al. (2000) the
20 Phoenix data to investigate the role of crustal particles in PM2 5 samples on daily mortality.
21 These studies are discussed in more detail below in Section 8.2.2.4.3 on the source-oriented
22 evaluation of PM; and only the basic results regarding crustal particles are mentioned here. The
23 elemental abundance data (from X-ray fluorescence spectroscopy analysis of daily filters) were
24 analyzed to estimate the concentration of crustal particles in PM2 5 using factor analysis. Then
25 the association of mortality with fine crustal mass was estimated using Poisson regression
26 (regressing mortality on factor scores for "crustal factor"), adjusting for time trends and weather.
27 No positive association was found between fine crustal mass factor and mortality.
28 The above results, overall, mostly suggest that crustal particles (coarse or fine) per se are
29 not likely associated with daily mortality. However, as noted in the previous section, three
30 analyses of Phoenix, AZ data suggested that PM10_25 may be associated with mortality. The
31 results from one of the three studies (Smith et al., 2000) suggest that coarse particle mortality
April 2002 8-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 associations are stronger in spring and summer, when the anthropogenic portion of PM10_2 5 is
2 lowest as determined by factor analysis. However, during spring and summer, biogenic
3 processes (e.g., wind-blown endotoxins and molds) may contribute more to the PM10_2 5 fraction
4 in the Phoenix area, clouding any attribution of observed PM10_2 5 effects there to crustal particles,
5 per se. Disentangling potential contributions of biogenically-derived organic particle
6 components from those of crustal materials in the PM10_25 fraction in Mexico City and Santiago
7 poses further interesting challenges.
8
9 Ultrafine Particle Effects
10 The Wichmann et al. (2000) study evaluated the attribution of PM effects to specific size
11 fractions, including both the number concentration (NC) and mass concentration (MC) of
12 particles in a given size range. The study was carried out in the small German city of Erfurt
13 (pop. 200,000) in the former German Democratic Republic, by a team of scientists at the
14 Gessellschaft fur Strahlenforschung (GSF) and Ludwig Maximilian University in Germany.
15 Erfurt was heavily polluted by particles and SO2 in the 1980s, and excess mortality was attributed
16 to high levels of TSP by Spix et al. (1993). Concentrations of PM and SO2 have markedly
17 dropped since then. The present study provides a much more detailed look at the health effects
18 of ultrafine particles (diameter < 0.1 //m) than earlier studies, and allows examination of effects
19 related to number counts for fine and ultrafine particles, as well as to their mass.
20 The Mobile Aerosol Spectrometer (MAS), developed by GSF, produces number and mass
21 concentrations in three size classes of ultrafmes (0.01 to 0.1 //m) and three size classes of larger
22 fine particles (0.1 //m to 2.5 //m). The mass concentration MCO.01-2.5 is well correlated with
23 gravimetric PM2 5, and the number concentration NCO.01-2.5 is well correlated with total particle
24 counts from a condensation particle counter (CPC). Mortality data were coded by cause of death,
25 with some discrimination between underlying causes and prevalent conditions of the deceased.
26 Some analyses looked at cardiovascular causes without respiratory, respiratory without
27 cardiovascular, and both causes together as separate groups. Age was used as a modifying factor,
28 as was weekly data for all of Germany on influenza and similar diseases. Daily mortality data
29 were fitted using a Poisson Generalized Additive Model (GAM), with adjustments for weather
30 variables, time trends, day of week, and particle indices. Two types of models were fitted, one
April 2002 8-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 using the best single-day lag for air pollution and a second using the best polynomial distributed
2 lag (PDL) model for air pollution.
3 Winter PM generally had the most significant positive effects on mortality, and fall PM
4 effects were similar in magnitude, but less significant because of the smaller NC and MC in fall
5 than in winter. Summer PM effects were consistently lower and not significant. PDL models
6 generally had larger and more significant PM effects than single-day lag models. Log-
7 transformed pollution models occasionally provided better fits than untransformed pollutant
8 models, particularly for number concentration indices in single-day lag models. However, there
9 were some nonlinear relationships that could not be adequately described by either parametric
10 model, as shown by use of LOESS models. The results cited in Table 8-1 and Appendix
11 Table 8A-1 are all for linear PDL models, to facilitate comparison.
12 Mass concentration was most often significantly associated with excess mortality in
13 one-pollutant models, with excess risks for MAS MCO.1-2.5 being about 6.2% (CI1.4, 11.2) per
14 25 //g/m3. The non-significant estimate from filter PM25 was about 3% (CI -1.7, 7.9) per
15 25 //g/m3. Filter PM10 estimates were also significant predictors of mortality overall, about 6.6%
16 excess risk per 50 //g/m3 (CI 0.7 to 12.8) in PDL models.
17 Mass concentrations for smaller fine particles were also often significant, with excess risk
18 for MCO.01-1.0 being ca. 5.1% (CI 0.2, 10.2) per 25 //g/m3 in a linear PDL model. Smaller-size
19 components of MCO.01-1.0 were also significantly associated, or nearly so, with excess
20 mortality. The intermodal fraction MCI.0-2.5 was also significant in a PDL logarithmic model,
21 4.7% (CI 1.05, 8.5) per IQR in log concentration. No results were reported for the effects of
22 ultrafme mass concentrations in classes 0.01-0.03, 0.03-0.05, or 0.05-0.1 //g/m3.
23 Number concentrations of ultrafme particles were also associated with excess mortality,
24 significantly or nearly so in smaller size classes. The results for linear models are shown in
25 Table 8-3. The table also shows how much the estimated excess risks are reduced, sometimes
26 drastically, when co-pollutants (especially SO2 and NO2) are included in a two-pollutant model.
27 Number and mass concentrations of various ultrafme and fine particles in all size ranges are
28 rather well correlated with gaseous co-pollutants except for the intermodal size range MCI.0-2.5.
29 The correlations range from 0.44 to 0.62 with SO2, from 0.58 to 0.66 with NO2, and from 0.53 to
30 0.70 with CO. The mass correlations range from 0.53 to 0.62 with SO2, from 0.48 to 0.60 with
31 NO2, and from 0.56 to 0.62 with CO. The large decreases in excess risk for number
April 2002 8-49 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-3. EXCESS TOTAL MORTALITY RISKS ESTIMATED TO BE
ASSOCIATED WITH VARIOUS AMBIENT PARTICLE SIZE-RELATED INDICES
PM Index
NCO.01-0.03
NCO.03-0.05
NCO.05-0.1
NCO.01-2.5
NCO.01-0.1
MCO.01-2.5
Co-Pollutant
None
None
None
None
None
SO2
NO2
CO
MCO.01-2.5
None
SO2
Excess Risk, %
3.00a
3.80a
4.00a
6.891b
8.238b
4.758b
0.739b
3.594b
4.123b
6.194C
2.014C
Single-Pollutant Models
Lower 95% CL
-0.342
0.021
-0.307
0.662
0.252
-0.451
-3.951
-2.312
-1.437
1.409
-2.304
Upper 95% CL
6.455
7.722
8.493
13.504
16.86
10.239
5.658
9.856
9.996
11.205
6.523
"Risks estimates for mortality associated with number concentrations (NC) in specified ranges. At actual
interquartile range, respectively 8888, 2524, and 1525 particles/cm3.
bAt standard increment 25,000 particles/cm3; winter IQR is 22,211 particles/cm3, annual IQR is 12,690 particles/cm3.
cAt standard increment 25 ^g/m3.
Source: Based on Wichman et al. (2000), as calculated by U.S. EPA.
1
2
3
4
5
6
7
8
9
10
11
concentration, particularly when NO2 is a co-pollutant with NCO.01-0.1. clearly involves a more
complex structure than simple correlation. The large decrease in excess risk when SO2 is a
co-pollutant with MCO.01-2.5 is not readily explained, and it is discussed in some detail in
Wichmann et al. (2000).
SO2 is a strong predictor of excess mortality in this study; and its estimated effect is little
changed when different particle indicators are included in a two-pollutant model. The authors
noted: ". . .the [LOESS] smoothed dose response curve showed most of the association at the
left end, below 15 //g/m3, a level at which effects were considered biologically implausible. . ."
Replacement of sulfur-rich surface coal has reduced mean SO2 levels in Erfurt from 456 //g/m3
in 1988 to 16.8 //g/m3 during 1995 to 1998 and to 6 //g/m3 in 1998. The estimated concentration-
response functions for SO2 are very different in these time periods, comparing Spix et al. (1993)
April 2002
8-50
DRAFT-DO NOT QUOTE OR CITE
-------
1 with Wichmann et al. (2000) results. Wichmann et al. concluded "These inconsistent results for
2 SO2 strongly suggested that SO2 was not the causal agent but an indicator for something else."
3 The authors offered no specific suggestions as to what the "something else" might be, but they
4 did finally conclude that their studies from Germany strongly supported particulate air pollution
5 as more relevant than SO2 to observed mortality impacts.
6 The authors also found that ultrafme particles, NO2 and CO form a group of pollutants
7 strongly identified with motor vehicle traffic. Immediate and delayed effects seemed to be
8 independent in two-pollutant models, with single-day lags of 0 to 1 days and 4 to 5 days giving
9 'best fits' to data. The delayed effect of ultrafme particles was stronger than that for NO2 or CO.
10 Another finding of interest is that the excess risk in Erfurt is larger and more significantly
11 associated with ages < 70 years than with older ages. This is consistent for PDL models for
12 NCO. 01-0.1. MCO.01-2.5. andPM10. None of the single lag day models were significant.
13 Examination of prevalent disease categories found larger and more significant risks for
14 respiratory disease mortality than for cardiovascular mortality in almost all models. Combined
15 cardiovascular or combined respiratory diseases were generally the next highest category. Other
16 natural causes (i.e., neither respiratory nor cardiovascular) almost always had the lowest risk.
17
18 8.2.2.4.2 Chemical Components
19 Ten new studies from the U.S. and Canada examined specific chemical components of PM.
20 Table 8-4 shows the chemical components examined in these studies, the mean concentrations
21 for Coefficient of Haze (COH), sulfate, and H+, as well as the list of those that were found to be
22 associated with increased mortality. There are several chemical components of PM whose
23 associations with mortality can be compared across studies, including COH, sulfate, and H+.
24
25 Coefficient of Haze, Elemental Carbon, and Organic Carbon
26 COH is highly correlated with elemental carbon (EC) and is often considered as a good PM
27 index for motor vehicle sources (especially diesel), although other combustion processes such as
28 space heating likely also contribute to COH levels. Several studies (Table 8-4) examined COH;
29 and, in most cases, positive and significant associations with mortality outcomes were reported.
30 In terms of relative significance of COH in comparison to other PM components, COH was not
31 the clearly most significant PM component in any of these studies. The average level of COH in
April 2002 8-51 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-4. SUMMARY OF PARTICULATE MATTER CHEMICAL COMPONENTS
ANALYZED IN RECENT STUDIES
Mean COH Mean SCV
Author, City (1000ft) (Mg/m3)
Burnett et al. 0.42 9.2
(1998b)
Toronto, Canada.
1980-1994.
Burnett etal. 0.26 2.6
(2000).
8 largest Canadian
cities
1986-1996.
Fairley (1999). 0.5 1.8
Santa Clara County,
CA
Other PM
Mean H+ components
(nmol/m3) analyzed
TSP, estimated
PM10 and PM2 5
PM10, PM2 5,
PM10_2 5, and
47 trace elements
PM10, PM2 5,
PM10_2 5, and
nitrate
PM components
associated with mortality.
Comments.
TSP, COH, sulfate,
estimated PM10 and PM2 5.
However, CO together
with TSP explained most
of the association.
PM10, PM2 5, COH,
sulfate, Zn, Ni, and Fe
significantly associated
with total mortality.
COH, sulfate, nitrate,
PM10, and PM2 5 were
associated with mortality.
PM2 5 and nitrate most
significant.
Gwynn et al. 0.2
(2000).
Buffalo, NY
1988-1990
Lipfert et al. 0.28
(2000a).
Philadelphia, PA
1992-1995
Lippmann et al.
(2000). Detroit, MI
1992-1994
Klemm and Mason
(2000). Atlanta, GA
1998-1999
5.9 36.4 PM10
5.1 8.0 Nephelometry,
NH4+, TSP, PM10
PM2 5, and PM10.2 5
5.2 8.8 PM10PM25,and
PM10.2.5
5.2 0 Nitrate, EC, OC,
oxygenated HC,
PM10, PM25, and
PM10.2.5
Sulfate, H+, PM10, and
COH were associated with
total mortality. COH was
least significant predictor.
Essentially all PM
components were
associated with mortality.
PM10, PM2 5, and PM10.2 5
were more strongly
associated with mortality
outcomes than sulfate or
"Interim" results based on
one year of data.
No statistically significant
associations for any
pollutants. Those with
t-ratio of at least 1.0 were:
H+, PM10, and PM2 5,
April 2002
8-52
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-4 (cont'd). SUMMARY OF PARTICIPATE MATTER CHEMICAL
COMPONENTS ANALYZED IN RECENT STUDIES
Author, City
Mean COH Mean SO4~
(1000ft) (Mg/m3)
MeanH+
(nmol/m3)
Other PM
components
analyzed
PM components
associated with mortality.
Comments.
Mar et al. (2000).
Phoenix, AZ
1995-1997
Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ
1981-1983
12.7
Hoek et al. (2000).
The Netherlands
1986-1994
Goldberg et al.
(2000). Montreal,
Quebec, Canada.
1984-1993.
Anderson et al.
(2001). The west
Midlands
conurbation, UK.
1994-1996.
3.8
(median)
0.24
3.7
S, Zn, Pb,
soil-corrected K,
reconstructed soil,
EC, OC, TC,
PM10, PM2 5, and
PM,,
PM15, PM2 5,
sulfates
cyclohexane-
solubles (CX),
dichloromethane-
solubles (DCM),
and acetone-
solubles (ACE).
PM10, BS, and
nitrate
Predicted PM2 5,
and extinction
coefficient (visual-
range derived).
PM10, PM2 5,
PM10.25, andBS.
S, Pb, and soil were
negatively associated with
total mortality. PM10 and
PM10_2 5 were positively
associated with total
mortality. Soil-corrected
K, non-soil PM2 5, EC,
OC, TC, PM10, PM2 5, and
PM10_2 5 were associated
with cardiovascular
mortality.
PM15, PM2 5, sulfate, CX
and ACE were
significantly associated
with total and/or
cardiovascular mortality
in Newark and/or
Camden.
Sulfate, nitrate, and BS
were more consistently
associated with total
mortality than PM10.
COH, predicted PM2 5,
and sulfate were
associated with various
mortality outcomes
(mostly elderly and
stronger associations in
summer).
Significant associations
between all-cause
mortality with PM indices
(except PM10_2 5) were
seen only in warm season.
1 these studies ranged from 0.2 (Buffalo, NY) to 0.5 (Santa Clara County, CA) 1000 linear feet.
2 The correlations between COH and NO2 or CO in these studies (8 largest Canadian cities; Santa
3 Clara County, CA; and Buffalo, NY) were moderately high (r ~ 0.7 to 0.8), suggesting a likely
April 2002
8-53
DRAFT-DO NOT QUOTE OR CITE
-------
1 motor vehicle contribution. Some of the inconsistencies in the results across cities may be in
2 part, due to the differences in COH levels. For example, in Buffalo, NY (where COH was
3 lowest), no significant association was found for any pollutant, possibly due to small sample size
4 (« 1 year of data). However, both EC and OC were significant predictors of cardiovascular
5 mortality in the Phoenix study, with their effect sizes per IQR being comparable to those for
6 PM10, PM2 5, and PM10_2 5; there, EC and OC represented major mass fractions of PM25 (11% and
7 38%, respectively) and correlated highly with PM25 (r = 0.84 and 0.89, respectively). They were
8 also highly correlated with CO and NO2 (r ~ 0.8 to 0.9), indicating their associations with an
9 "automobile" factor. Thus, the COH and EC/OC results from the Mar et al. (2000) study suggest
10 that PM components from motor vehicle sources are likely associated with mortality.
11 In a recent study in Montreal, Quebec, by Goldberg et al. (2000), COH appeared to be
12 correlated with some of the mortality outcomes more strongly than other PM indices such as the
13 visual-range derived extinction coefficient (considered to be a good indicator of sulfate).
14 However, the main focus of the study was the role of cardio-respiratory risk factors for air
15 pollution, and the investigators warned against comparing the relative strength of associations
16 among PM indices, pointing out complications such as likely error involved in the visual range
17 measurements. Also, the estimated PM25 values were predicted from other PM indices,
18 including COH and extinction coefficient, making it difficult to compare straightforwardly the
19 relative importance of PM indices.
20
21 Sulfate and Hydrogen Ion
22 Sulfate and H+, markers of acidic components of PM, have been hypothesized to be
23 especially harmful components of PM (Lippmann and Thurston, 1996). The newly available
24 studies that examined sulfate are shown in Table 8-4; four of them also analyzed H+ data. The
25 sulfate concentrations ranged from 1.8 //g/m3 (Santa Clara County, CA) to 12.7 //g/m3 (three NJ
26 cities). Aside from the west versus east coast contrast, the higher levels observed in Toronto and
27 the three NJ cities are likely due to their study period coverage of the early 1980's, when sulfate
28 levels were higher. Sulfate explained 25 to 30% of PM25 mass in eastern U.S. and Canadian
29 cities, but it was only 14% of PM2 5 mass in Santa Clara County, CA. The mean H+level in the
30 Buffalo, NY study (36.4 nmol/m3) was much higher than the levels in Philadelphia, Detroit, or
31 Atlanta, in part because the Buffalo study covered the 1988 summer when summer-haze episodes
April 2002 8-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 occurred. The H+ levels measured in the other three cities were low, especially in Atlanta, GA
2 (where the mean concentration was reported to be 0.0 //g/m3). Even the mean H+ concentration
3 for Detroit, MI (the H+ was actually measured in Windsor, a Canadian city a few miles from
4 downtown Detroit), 8.8 nmol/m3, was low compared to the reported detection limit of
5 15.1 nmol/m3 (Brook et al., 1997) for the measurement system used in the study. Note that the
6 corresponding detection limit for sulfate was 3.6 nmol/m3 (or 0.34 //g/m3) and the mean sulfate
7 level for Detroit was 54 nmol/m3 (or 5.2 //g/m3), so that the signal-to-noise ratio is expected to be
8 higher for sulfate than for H+. Thus, the ambient levels and possible relative measurement errors
9 for these data should be considered in interpreting the results of the studies listed in Table 8-4.
10 Sulfate was a statistically significant (at p< 0.05) predictor of mortality, at least in single
11 pollutant models, in: Toronto, CN; the 8 largest Canadian cities; Santa Clara County, CA;
12 Buffalo, NY; Philadelphia, PA; Newark, NJ; Camden, NJ; and Montreal, Quebec, but not in
13 Detroit, MI, Elizabeth, NJ, or Atlanta, GA. However, it should be noted that the relative
14 significance across the cities is influenced by the sample size (both the daily mean death counts
15 and number of days available), as well as the range of sulfate levels, and therefore should be
16 interpreted with caution. Figure 8-7 shows the excess risks (± 95% CI) estimated per 5 //g/m3
17 increase in 24-h sulfate reported in these studies, compared to the earlier Six Cities Study result.
18 The largest estimate was seen for Santa Clara County, CA, but the wide confidence band
19 (possibly due to the small variance of the sulfate, since its levels were low) should be taken into
20 account. Also, in the Santa Clara County analysis, the sulfate effect was eliminated once PM2 5
21 was included in the model, perhaps being indicative of sulfate mainly serving as a surrogate for
22 fine particles in general there. In any case, more weight should be accorded to estimates from
23 other studies with narrower confidence bands. In the other studies, the effect size estimates
24 mostly ranged from about 1 to 4% per 5 //g/m3 increase in 24-h sulfate.
25 The relative significance of sulfate and FT compared to other PM components varied from
26 city to city, as seen in Table 8-4. Because each study included different combinations of
27 co-pollutants that had different extents of correlation with sulfate and because multiple mortality
28 outcomes were analyzed, it is difficult to assess the overall importance of sulfate across the
29 available studies. However, it can generally be seen that the associations were stronger in cities
30 where the sulfate and H+ levels were relatively high. For example, the Gwynn et al., 2000
31 finding for Buffalo, NY data that H+ and sulfate were most significantly associated with total
April 2002 8-55 DRAFT-DO NOT QUOTE OR CITE
-------
Percent (total mortality, unless otherwise noted)
per 5 |jg/m3 increase in sulfate
Schwartz et al. (1996)
Six Cities"
Burnett et al. (1998)
Toronto, Canada "
Burnett et al. (2000)
8 Largest -
Canadian Cities
Fairiey (1999) _
Santa Clara, Co. "
Gwynn et al. (2000)
Buffalo, NY "
Klemm et al. (2000) _
Atlanta, GA "
Lipfert et al. (2000a) _
Philadelphia, PA "
Lippman et al. (2000)
Detroit, Ml "
Tsai et a!. (2000)
3 NJ Cities
_2
I
0
I
2
i
4
i
6
i
8
i
10
i
-•—— Newwark
I Camden
Elizabeth
Figure 8-7. Excess risks estimated for sulfate per 5 Aig/m3 increase from the studies in
which both PM2 5 and PM10_2 5 data were available.
1 mortality may be in part due to the high acid aerosol levels in that data. Also, the fact that the
2 Lippmann et al. (2000) finding for Detroit, MI data on H+ and sulfate being less significantly
3 associated with mortality than the size-fractionated PM mass indices may be due to acidic
4 aerosols levels being mostly below the detection limit in that data. In this case, it appears that the
5 Detroit PM components show mortality effects even without much acidic input.
6 In summary, assessment of new study results for individual chemical components of PM
7 suggest that an array of PM components (mainly fine particle constituents) were associated with
8 mortality outcomes, including: COH, EC, OC, sulfate, H+, and nitrate. The discrepancies seen
9 with regard to the relative significance of these PM components across studies may be in part due
April 2002
8-56
DRAFT-DO NOT QUOTE OR CITE
-------
1 to the difference in their concentrations. This issue is further discussed below as part of the
2 assessment of new studies involving source-oriented evaluation of PM components.
3
4 8.2.2.4.3 Source-Oriented Evaluations
5 Several new studies have conducted source-oriented evaluation of PM components.
6 In these studies, daily concentrations of PM components (i.e., trace elements) and gaseous
7 co-pollutants were analyzed using factor analysis to estimate daily concentrations due to
8 underlying source types (e.g., motor vehicle emissions, soil, etc.), which are weighted linear
9 combinations of associated individual variables. The mortality outcomes were then regressed on
10 those factors (factor scores) to estimate the impact of source types, rather than just individual
11 variables. These studies differ in terms of: specific objectives/focus, the size fractions from
12 which trace elements were extracted, and the way factor analysis was used (e.g., rotation). The
13 main findings from these studies regarding the source-types identified (or suggested) and their
14 associations with mortality outcomes are summarized in Table 8-5.
15 The Laden et al. (2000) analysis of Harvard Six Cities data for 1979-1988 aimed to identify
16 distinct source-related fractions of PM25 and to examine each fraction's association with
17 mortality. Fifteen elements in the fine fraction samples were routinely found above their
18 detection limits and included in the data analyses. For each of the six cities, up to 5 common
19 factors were identified from among the 15 elements, using specific rotation factor analysis.
20 Using the Procrustes rotation (a type of oblique rotation), the projection of the single tracer for
21 each factor was maximized. This specification of the tracer element was based on:
22 (1) knowledge from previous source apportionment research; (2) the condition that regression of
23 total fine mass on that element must result in a positive coefficient; and (3) identifications of
24 additional local source factors that positively contributed to total fine mass regression. Three
25 source factors were identified in all six cities: (1) a soil and crustal material factor with Si as a
26 tracer; (2) a motor vehicle exhaust factor with Pb as a tracer; and, (3) a coal combustion factor
27 with Se as a tracer. City-specific analyses also identified a fuel combustion factor (V), a salt
28 factor (Cl), and selected metal factors (Ni, Zn, or Mn). For each city, a GAM Poisson regression
29 model, adjusting for trend/season, day-of-week, and smooth function of temperature/dewpoint,
30 was used to estimate impacts of each source type (using absolute factor scores) simultaneously.
31 Summary estimates across cities were obtained by combining the city-specific estimates, using
April 2002 8-57 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-5. SUMMARY OF SOURCE-ORIENTED EVALUATIONS OF
PARTICULATE MATTER COMPONENTS IN RECENT STUDIES
Author, City
Source types identified (or suggested) and associated
variables
Source types associated with mortality.
Comments.
Laden et. al., (2000)
Harvard Six Cities
1979-1988
Mar et al. (2000).
Phoenix, AZ
1995-1997
Soil and crustal material'. Si
Motor vehicle emissions'. Pb
Coal combustion'. Se
Fuel oil combustion'. V
Salt: Cl
Note: the trace elements are from PM2 5 samples
PM2 5 (fromDFPSS) trace elements:
Motor vehicle emissions and re-suspended road dust:
Mn, Fe, Zn, Pb, OC, EC, CO, and NO2
Soil: Al, Si, and Fe
Vegetative burning: OC, and Kg (soil-corrected
potassium)
LocalSO2 sources: SO2
Regional sulfate: S
Strongest increase in daily mortality associated
with mobile source factor. Coal combustion factor
was positively associated with mortality in all
metropolitan areas, with exception of Topeka.
Crustal factor from fine particles not associated
with mortality. Coal and mobile sources account
for majority of fine particles in each city.
PM, „ factors results: Soil factor and local SO2
factor were negatively associated with total
mortality. Regional sulfate was positively
associated with total mortality on the same day,
but negatively associated on the lag 3 day. Motor
vehicle factor, vegetative burning factor, and
regional sulfate factor were significantly positively
associated with cardiovascular mortality.
PMU_2S (from dichot) trace elements:
Soil: Al, Si, K, Ca, Mn, Fe, Sr, and Rb
A source of coarse fraction metals: Zn, Pb, and Cu
A marine influence: Cl
analyzed for associations with mortality because of
small sample size (every-3ri day samples from
June 1996).
Tsai et al. (2000).
Newark, Elizabeth, and
Camden, NJ.
1981-1983.
Motor vehicle emissions: Pb, CO
Geological (Soil): Mn, Fe
Oil burning: V, Ni
Industrial: Zn, Cu, Cd (separately)
Sulfate/'secondary aerosol: sulfate
Note: the trace elements are from PM15 samples
Oil burning, industry, secondary aerosol, and
motor vehicles factors were associated with
mortality.
Ozkaynak et al. (1996). Motor vehicle emissions: CO, COH, and NO2
Toronto, Canada.
Motor vehicle factor was a significant predictor
for total, cancer, cardiovascular, respiratory, and
pneumonia deaths.
April 2002
8-58
DRAFT-DO NOT QUOTE OR CITE
-------
1 inverse variance weights. The identified factors and their tracers are listed in Table 8-5. The
2 results from mortality regression analysis including these factors indicated that the strongest
3 increase in daily mortality was associated with the mobile source factor. Also, the coal
4 combustion factor was positively associated with mortality in all metropolitan areas, except for
5 Topeka. Lastly, S, Ni, and Pb were specific elements individually associated with mortality, but
6 the crustal factor from fine particles was not.
7 Mar et al. (2000) analyzed PM10, PM10_25, two measurements of PM25, and various
8 sub-components of PM25 for their associations with total (non-accidental) and cardiovascular
9 deaths in Phoenix, AZ during 1995-1997, using both individual PM components and factor
10 analysis-derived factor scores. GAM Poisson models were used, adjusting for season,
11 temperature, and relative humidity. The evaluated air pollution variables included: O3, SO2,
12 NO2, CO, TEOM PM10, TEOM PM25, TEOM PM10.25, DFPSS PM25, S, Zn, Pb, soil, soil-
13 corrected K (Ks), nonsoil PM, OC, EC, and TC. Lags 0 to 4 days were evaluated. As earlier
14 noted, individual PM component results indicated that PM10_2 5 was more significantly associated
15 with total mortality than PM2 5, although both TEOM PM2 5 and PM10_2 5 were significantly
16 associated with cardiovascular mortality. A factor analysis conducted on the chemical
17 components of DFPSS PM25 (Al, Si, S, Ca, Fe, Zn, Mn, Pb, Br, Ks, OC, and EC) identified
18 factors for: motor vehicle emissions/re-suspended road dust; soil; vegetative burning; local SO2
19 sources; and regional sulfate (see Table 8-5). The results of mortality regression with these
20 factors suggested that the soil factor and local SO2 factor were negatively associated with total
21 mortality. Regional sulfate was positively associated with total mortality on the same day, but
22 negatively associated on the lag 3 day. The motor vehicle factor, vegetative burning factor, and
23 regional sulfate factor were each significantly positively associated with cardiovascular mortality.
24 The authors also analyzed elements from dichot PM10_2 5 samples, and identified soil, a source of
25 coarse fraction metals (industry), and marine influence factors. However, these factors were not
26 analyzed for their associations with mortality outcomes due to the short measurement period
27 (starting in June 1996 with every-3rd-day sampling).
28 It should be noted here that the Smith et al. (2000) analysis of Phoenix data also included
29 factor analysis on the elements from the coarse fraction and identified essentially the same
30 factors ("a source of coarse fraction metals" factor in Mar et al.'s study was called "the
31 anthropogenic elements" in Smith et al.'s study). While Smith et al. did not relate these factors
April 2002 8-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 to mortality (due to a small sample size), they did show that the anthropogenic elements were
2 low in summer and spring, when the PM10_2 5 effect was largest. These results suggest that the
3 PM10_2 5 effects were not necessarily due to anthropogenic components of the coarse particles,
4 with biogenically-generated coarse particles perhaps being key during the warmer months (as
5 noted earlier above).
6 Tsai et al. (2000) conducted an exploratory analysis of mortality in relation to specific PM
7 source types for three New Jersey cities (Camden, Newark, and Elizabeth) using factor analysis -
8 Poisson regression techniques. During the three-year study period (1981-1983), extensive
9 chemical speciation data were available, including nine trace elements, sulfate, and particulate
10 organic matter. Total (excluding accidents and homicides), cardiovascular, and respiratory
11 mortality were analyzed. Tsai et al. first conducted a factor analysis of trace elements and
12 sulfate, identifying major source types: motor vehicle (Pb, CO); geological (Mn, Fe); oil burning
13 (V, Ni); industrial (Zn, Cu); and sulfate/secondary aerosols (sulfate). In addition to Poisson
14 regression of mortality on these factors, they also used an alternative approach in which the
15 inhalable particle mass (IPM, D50 < 15 //m) was first regressed on the factor scores of each of the
16 source types to apportion the PM mass; and then the estimated daily PM mass for each source
17 type was included in Poisson regression, so that RR could be calculated per mass concentration
18 basis for each PM source type. They found that oil burning (V, Ni), various industrial sources
19 (Zn, Cd), motor vehicle (Pb, CO), and the secondary aerosols, as well as the individual PM
20 indices IPM, FPM (D50 < 3.5 //m), and sulfates, were all associated with total and/or
21 cardiorespiratory mortality in Newark and Camden, but not in Elizabeth. In Camden, the RRs for
22 the source-oriented PM were higher (~ 1.10) than those for individual PM indices (« 1.02).
23 Ozkaynak et al. (1996) analyzed 21 years of mortality and air pollution data in Toronto,
24 Canada. In addition to the usual simultaneous inclusion of multiple pollutants in mortality
25 regressions, they also conducted a factor analysis of all the air pollution and weather variables,
26 including TSP, SO2, COH, NO2, O3, CO, relative humidity and temperature. The factor with the
27 largest variance contribution (~ 50%) had the highest factor loadings for CO, COH, and NO2,
28 which they considered to be representative of motor vehicle emissions, since this pollution
29 grouping was also consistent with the emission inventory information for that city. They then
30 regressed mortality on the factor scores (a linear combination of standardized scores for the
31 covariates), after filtering out seasonal cycles and adjusting for temperature and day-of-week
April 2002 8-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 effects. The estimated impacts on mortality from motor vehicle pollution ranged from 1 to 6%,
2 depending on the outcomes.
3 In summary, these studies suggest that a number of source-types are associated with
4 mortality, including motor vehicle emissions, coal combustion, oil burning, and vegetative
5 burning. The crustal factor from fine particles was not associated with mortality in the Harvard
6 Six Cities data. In Phoenix data, where coarse particles were reported to be associated with
7 mortality, the associations between the factors related to coarse particles (soil, marine influence,
8 and anthropogenic elements) and mortality could not be evaluated due to the small sample size.
9 However, the soil (i.e., crustal) factor from fine particles in the Phoenix data was negatively
10 associated with mortality. Thus, although some unresolved issues remain (mainly due to the lack
11 of sufficient data), the source-oriented evaluation approach, using factor analysis, thus far seems
12 to implicate fine particles of anthropogenic origin as being most important (versus crustal
13 particles of geologic origin) in contributing to observed increased mortality risks.
14
15 8.2.2.5 New Assessments of Cause-Specific Mortality
16 Consistent with similar findings described in the 1996 PM AQCD, most of the newly
17 available studies summarized in Tables 8-1 and 8A-1 that examined non-accidental total,
18 circulatory, and respiratory mortality categories (e.g., Samet et al., 2000a,b; Dominici et al.,
19 2000a; Moolgavkar, 2000a; Gwynn et al., 2000; Lippmann et al., 2000; Ostro et al., 1999a;
20 Schwartz, 2000c) found significant PM associations with both cardiovascular and/or respiratory-
21 cause mortality. Several (e.g., Ostro et al., 1998; Fairley, 1999; Gwynn et al., 2000; Borja-
22 Aburto et al., 1997; Wordley et al., 1997; Borja-Aburto et al., 1998; Prescott et al., 1998;)
23 reported estimated PM effects that were generally higher for respiratory deaths than for
24 circulatory or total deaths. Once again, the NMMAPS results for U.S. cities are among those of
25 particular note here due to the large study size and the combined, pooled estimates derived for
26 various U.S. regions.
27 The Samet et al. (2000a,b) NMMAPS 90-cities analyses not only examined all-cause
28 mortality (excluding accidents), but also evaluated cardiovascular, respiratory, and other
29 remaining causes of deaths. Results were presented for all-cause, cardio-respiratory, and "other"
30 mortality for lag 0, 1, and 2 days. The investigators commented that, compared to the result for
31 cardio-respiratory deaths showing 3.5% (CI 1.0, 5.9) increase per 50 //g/m3 PM10, there was less
April 2002 8-61 DRAFT-DO NOT QUOTE OR CITE
-------
1 evidence for non-cardio-respiratory deaths. However, the estimates for "other" mortality, though
2 half those for cardio-respiratory mortality, were nevertheless positive, with fairly high posterior
3 probability (e.g., 0.84 at lag 0 day) that the overall effects were greater than 0 (estimated percent
4 excess "other" deaths being ~ 1.3 per 50 //g/m3 PM10 at lag 0). Dominici et al. (2000a) evaluated
5 the 20 largest U. S. cities, a subset of the cities included in Samet et al.'s NMMAPS analyses.
6 The pattern of PM10 effects on cardiovascular and respiratory mortality was similar to that
7 discussed earlier for total mortality, with lag day 1 showing the largest estimates. In this case,
8 the PM10 effect in these analyses was smaller and weaker for "other" causes. Regional model
9 results suggested that PM10 effects in the western U.S. were larger than in the eastern or southern
10 U.S. The PM coefficients were little affected by including gaseous pollutants in the model.
11 The Lippmann et al. (2000) analyses of cause-specific mortality in Detroit also evaluated
12 such mortality at various lags (0-3 days) in relation to several PM indices (PM10, PM25, PM10_2 5,
13 sulfate, H+) and various gaseous pollutants (O3, SO2, NO2 and CO), with appropriate adjustment
14 for season, temperature, relative humidity, etc. Significant effects for both cardiovascular and
15 respiratory mortality were more consistently found for the first three PM indices than for H+ or
16 sulfate. Effect size estimates tended to be highest for lag 1 day. It is notable here that, in the
17 Lippmann et al. (2000) analysis of Detroit mortality data, the "other" mortality category also
18 showed statistically significant effect size estimates. The authors noted, however, that the
19 "other" (non-circulatory and non-respiratory) mortality showed seasonal cycles and apparent
20 influenza peaks, suggesting that this series may have also been influenced by respiratory
21 contributing causes.
22 Another U.S. study, that of Moolgavkar (2000a), evaluated possible PM effects on cause-
23 specific mortality across a broad range of lag (0-5 days) times. Moolgavkar reported that in
24 Poisson regression GAM analyses, controlling for temperature and relative humidity, varying
25 patterns of results were obtained for PM indices in evaluations of daily deaths related to
26 cardiovascular disease (CVD), cerebrovascular disease (CrD), and chronic obstructive lung
27 disease (COPD) in three large U.S. metropolitan areas. In Cook County (Chicago area), the
28 association of PM10 with CVD mortality was statistically significant at a lag of 3 days based on a
29 single-pollutant analysis and remained significantly associated with CVD deaths with a 3-day lag
30 in two pollutant models including one or another of CO, NO2, SO2, or O3. In joint analyses with
31 both O3 and SO2, however, the PM10 association became markedly reduced and non-significant.
April 2002 8-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 Also, in Los Angeles single-pollutant analyses, PM10 and PM2 5 were significantly associated with
2 CVD mortality with lags of 2 and 1 days, respectively; but their coefficients were not robust to
3 inclusion of one or more gaseous pollutants. In Maricopa Co., AZ, PM10 coefficients were large
4 for several lags and significantly associated with CVD mortality lagged 1 day, as were each of
5 the gaseous pollutants tested (except O3) at several different lag times; and PM10 coefficients
6 seemed to be robust in 2-pollutant models including PM10 and NO2. As for cerebrovascular
7 disease, Moolgavkar (2000) reported that there was little evidence of association for PM with
8 CrD deaths at any lag in any of the three counties analyzed. With regard to COPD deaths, PM10
9 was significantly associated with COPD mortality (lag 2 days) in Cook County.
10 Zmirou et al. (1998) presented cause-specific mortality analyses results for 10 of the
11 12 APHEA European cities (APHEA1). Using Poisson autoregressive models adjusting for
12 trend, season, influenza epidemics, and weather, each pollutant's relative risk was estimated for
13 each city and "meta-analyses" of city-specific estimates were conducted. The pooled excess risk
14 estimates for cardiovascular mortality were 1.0% (0.3, 1.7) per 25 //g/m3 increase in BS and 2.0%
15 (0.5, 3.0) per 50 //g/m3 increase in SO2 in western European cities. The pooled risk estimates for
16 respiratory mortality in the same cities were: 2.0% (0.8, 3.2) and 2.5% (1.5, 3.4) for BS and SO2,
17 respectively. Also of note, Wichmann et al. (2000) found significant associations of elevated
18 cardiovascular and respiratory disease mortality with various fine (and ultrafme) particle indices
19 evaluated in Erfurt, Germany. "Other" natural causes (neither cardio- or respiratory-related)
20 almost always had the lowest risk in those models evaluating cause-specific mortality.
21 Seeking unique cause-specificity of effects associated with various pollutants has been
22 difficult because the "cause specific" categories examined are typically rather broad (usually
23 cardiovascular and respiratory) and overlap; also cardiovascular and respiratory conditions tend
24 to occur together. Examinations of more specific cardiovascular and respiratory sub-categories
25 may be necessary to test hypotheses about any specific mechanisms, but smaller sample sizes for
26 more specific sub-categories may make a meaningful analysis difficult. The study by Rossi et al.
27 (1999), however, examined associations between TSP and detailed cardio-vascular and
28 respiratory cause-specific mortality in Milan, Italy for a 9-year period (1980-1989). They found
29 significant associations for respiratory infections (11% increase per 100 //g/m3 increase in TSP;
30 95%CI: 5, 17) and for heart failure (7%; 95%CI: 3, 11), both on the same day TSP. The
31 associations with myocardial infarction (10%; 95%CI: 3,18) and COPD (12%; 95%CI: 6, 17)
April 2002 8-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 were found for the average of 3 and 4 day TSP levels. They noted the difference in lags between
2 temperature effects (i.e., cold temp, at lag 1 day for respiratory infections; hot temp, at lag 0 for
3 heart failure and myocardial infarction) and air pollution (TSP) effects. The immediate hot
4 temperature effects and the lagged cold temperature effects for total and cardiovascular mortality
5 have been reported in past studies (e.g., Philadelphia, Chicago), but investigations of the
6 differences in lags of PM effects for specific cardiovascular or respiratory categories have rarely
7 been conducted in time-series mortality studies.
8 In the Hoek et al. (2001) study of the whole population of the Netherlands, the large sample
9 size (mean daily total deaths -330, or more than twice that of Los Angeles County) allowed
10 examination of specific cardiovascular cause of deaths. Deaths due to heart failure, arrhythmia,
11 cerebrovascular causes, and thrombocytic causes were more strongly (-2.5 to 4 times larger
12 relative risks) associated with air pollution than the overall cardiovascular deaths. The
13 investigators concluded that specific cardiovascular causes (such as heart failure) were more
14 strongly associated with air pollution than total cardiovascular mortality, but noted that the
15 largest contribution to the association between air pollution and cardiovascular mortality was
16 from ischemic heart disease (about half of all cardiovascular deaths).
17 An HEI report on an epidemiologic study conducted by Goldberg et al. (2000) in Montreal,
18 Canada also provides interesting new information regarding types of medical conditions putting
19 susceptible individuals at increased risk for PM-associated mortality effects; and it highlights the
20 potential importance of evaluating "contributing causes" in cause-specific mortality analyses.
21 First, the immediate causes of death, as listed on death certificates, were evaluated in relation to
22 various ambient PM indices (TSP, PM10, PM2 5, COH, sulfates, extinction coefficients) lagged for
23 0 to 4 days, with results reported emphasizing effects at 3 day lags for three main PM measures
24 (COH, sulfate, estimated PM2 5). Significant associations were observed between all three
25 measures and total nonaccidental deaths, respiratory diseases, and diabetes, with an approximate
26 2% increase in excess nonaccidental mortality being observed per 9.5 //g/m3 interquartile
27 increase in 3-day mean estimated PM25 exposure.
28 When underlying clinical conditions identified in decedents' medical records were then
29 evaluated in relation to ambient PM measures, all three measures (COH, sulfate, estimated PM2 5)
30 were associated with acute lower respiratory disease, congestive heart failure, and any
31 cardiovascular disease. Estimated PM2 5 and COH were also reported to be associated with
April 2002 8-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 chronic coronary artery disease, any coronary artery disease, and cancer; whereas, sulfate was
2 associated with acute and chronic upper respiratory disease. None of the three PM measures
3 were related to airways disease, acute coronary artery disease, or hypertension. These results
4 both tend to support previous findings identifying individuals with preexisting cardiopulmonary
5 diseases as being at increased risk for ambient PM effects and appear to implicate another risk
6 factor, diabetes (which typically also involves cardiovascular complications as it progresses), as a
7 possible susceptibility condition putting individuals at increased risk for ambient PM effects.
8 Two recent studies (Gouveia and Fletcher, 2000; Concei9ao et al., 2001), both using data
9 from Sao Paulo, Brazil, examined child mortality (age under 5 years). The study periods for
10 these studies did not overlap (1991-1993 for Concei9ao and Fletcher study; 1994-1997 for
11 Concei9ao). Although Gouveia and Fletcher found significant associations between air pollution
12 and elderly mortality, they did not find statistically significant associations between air pollution
13 and child respiratory mortality (PM10 coefficient was negative and not significant). In the
14 Concei9ao et al. (2001) analysis, significant associations were found between child respiratory
15 mortality and CO, SO2, and PM10 in single pollutant models, and coefficients for CO and SO2
16 remained significant in the multiple-pollutant (apparently all pollutants together) model. The
17 reported PM10 coefficient in the single pollutant model corresponds to percent excess respiratory
18 death of 7.1% (95% CI: 1.1, 13.7) per 50 Mg/m3 increase in PM10. However, it should be noted
19 that the average daily respiratory mortality counts for these studies were relatively small
20 (~2.4/day). With the modest length of observations (3 years for Gouveia and Fletcher study, and
21 4 years for Concei9ao et al.'s study), the statistical power of the data were likely less than
22 desirable. Thus, there have not been enough data to elucidate the range of short-term PM effects
23 on child (respiratory) mortality.
24 Overall, then, the above assessment of newly available information provides interesting
25 additional new information (beyond that in the 1996 PM AQCD) with regard to cause-specific
26 mortality related to ambient PM. That is, a growing number of studies continue to report
27 increased cardiovascular- and respiratory-related mortality risks as being significantly associated
28 with ambient PM measures at one or another varying lag times. When specific subcategory of
29 cardiovascular disease was examined in a large population (The Netherlands study by Hoek
30 et al.), some of the subcategories such as heart failure were more strongly associated with PM
31 and other pollutants than total cardiovascular mortality. Largest effects estimates are most
April 2002 8-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 usually reported for 0-1 day lags (with some studies also now noting a second peak at 3-4 day
2 lags). A few of the newer studies also report associations of PM metrics with "other" (i.e.,
3 non-cardiorespiratory) causes, as well. However, at least some of these "other" associations may
4 also be due to seasonal cycles that include relationships to peaks in influenza epidemics that may
5 imply respiratory complications as a "contributing" cause to the "other" deaths. Or, the "other"
6 category may include sufficient numbers of deaths due to diabetes or other diseases which may
7 also involve cardiovascular complications as contributing causes. Varying degrees of robustness
8 of PM effects are seen in the newer studies, as typified by estimates in multiple pollutant models
9 containing gaseous co-pollutants; many show little effect of gaseous pollutant inclusion on
10 estimated PM effect sizes, some show larger reductions in PM effects to non-significant levels
11 upon such inclusion, and a growing number also report significant associations of cardiovascular
12 and respiratory effects with one or more gaseous co-pollutants. Thus, the newer studies both
13 further substantiate PM effects on cardiovascular- and respiratory-related mortality, while also
14 pointing toward possible significant contributions of gaseous pollutants to such cause-specific
15 mortality, as well. The magnitudes of the PM effect size estimates are consistent with the range
16 of estimates derived from the few earlier available studies assessed in the 1996 PM AQCD.
17
18 8.2.2.6 Salient Points Derived from Summarization of Studies of Short-Term Particulate
19 Matter Exposure Effects on Mortality
20 The most salient key points to be extracted from the above discussion of newly available
21 information on short-term PM exposures relationships to mortality can be summarized as follow:
22
23 PM10 effects estimates. Since the 1996 PM AQCD, thus far, there have been more than 80 new
24 time-series PM-mortality analyses published. Estimated mortality relative risks in these studies
25 are generally positive, statistically significant, and consistent with the previously reported
26 PM-mortality associations. Of particular importance are several studies which evaluated
27 multiple cities using consistent data analytical approaches. The NMMAPS analyses for the
28 largest 90 U.S. cities (Samet et al., 2000a,b), which are thought to probably provide the most
29 precise estimates for PM10 effects applicable to the U.S., derived a combined nationwide excess
30 risk estimate of about 2.3% increase in total (non-accidental) mortality per 50 //g/m3 increase in
31 PM10. The other multi-city analyses, as well as various single city analyses, also obtained PM10
April 2002 8-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 effect sizes generally in the range of 1.5 to 8.5% per 50 //g/m3 increase in PM10, consistent with
2 the range of statistically significant estimates given in the 1996 PM AQCD. However, more
3 geographic heterogeneity is evident among the newer multi-city study results than was the case
4 among the fewer studies assessed in the 1996 PM AQCD. In particular, in the NMMAPS
5 analysis of the 90 largest U.S. cities data, the risk estimates varied by U.S. geographic region,
6 with the estimate for the Northeast being the largest (4.6% per 50 //g/m3 PM10 increase). The
7 observed heterogeneity in the estimated PM risks across cities/regions could not be explained
8 with the city-specific explanatory variables, such as the mean levels of pollution and weather,
9 mortality rate, sociodemographic variables (e.g., median household income), urbanization, or
10 variables related to measurement error. Notable apparent heterogeneity was also seen among
11 effects estimates for PM (and SO2) indices in the multi-city APHEA studies conducted in
12 European cities. In APHEA2, they found that several city-specific characteristics, such as NO2
13 levels and warm climate, were important effect modifiers. The issue of heterogeneity of effects
14 estimates is discussed further below in Section 8.4.
15
16 Confounding and effect modification by other pollutants. Numerous new short-term PM
17 exposure studies not only continue to report significant associations between various PM indices
18 and mortality, but also between gaseous pollutants (O3, SO2, NO2, and CO) and mortality as well.
19 In most of these studies, simultaneous inclusions of gaseous pollutants in the regression models
20 did not meaningfully affect the PM effect size estimates. This was the case for the NMMAPS 90
21 cities study with regard to the overall combined U.S. regional and nationwide risk estimates
22 derived for that study. The issue of confounding is discussed further in Section 8.4.
23
24 Fine and coarse particle effects. Newly available studies provide generally statistically
25 significant PM2 5 associations with mortality, with effect size estimates falling in the range
26 reported in the 1996 PM AQCD. New results from Germany appear to implicate both ultrafine
27 (nuclei-mode) and accumulation-mode fractions of urban ambient fine PM as being important
28 contributors to increased mortality risks. As to the relative importance of fine and coarse
29 particles, in the 1996 PM AQCD there was only one acute mortality study that examined this
30 issue. In that study, the authors suggested that fine particles (PM25), but not coarse particles
31 (PM10_2 5), were associated with daily mortality. Now, more than ten studies have analyzed both
April 2002 8-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM25 and PM10_2 5 for their associations with mortality. While the results from some of these new
2 studies (e.g., Santa Clara County, CA analysis [Fairley, 1999] and the largest 8 Canadian cities
3 analysis [Burnett et al., 2000]) did suggest that PM2 5 was more important than PM10_2 5 in
4 predicting mortality fluctuations, other studies (e.g., Phoenix, AZ analyses [Clyde et al., 2000;
5 Mar et al., 2000; Smith et al., 2000]; Mexico City and Santiago, Chile studies [Castillejos et al.,
6 2000; Cifuentes et al., 2000]) suggest that PM10_2 5 may also be important in at least some
7 locations. Seasonal dependence of size-related PM component effects observed in some of the
8 studies complicates interpretations.
9
10 Chemical components ofPM. Several new studies have examined the role of specific chemical
11 components of PM. The studies conducted in U.S. and Canadian cities showed mortality
12 associations with specific fine particle components of PM including H+, sulfate, nitrate, as well
13 as COH, but their relative importance varied from city to city, likely depending on their levels
14 (e.g., no clear associations in those cities where H+ and sulfate levels were very low, i.e., circa
15 non-detection limits). The results of several studies that investigated the role of crustal particles,
16 although somewhat mixed, do not appear overall to support associations between crustal particles
17 and mortality (see also the discussion of source-oriented evaluations presented below).
18
19 Source-oriented evaluations. Several studies conducted source-oriented evaluations of PM
20 components using factor analysis. The results from these studies generally indicate that several
21 combustion-related source-types are likely associated with mortality, including: motor vehicle
22 emissions; coal combustion; oil burning; and vegetative burning. The crustal factor from fine
23 particles was not associated with mortality in the Harvard Six Cities data, and the soil (i.e.,
24 crustal) factor from fine particles in the Phoenix data was negatively associated with mortality.
25 Thus, the source-oriented evaluations seem to implicate fine particles of anthropogenic origin as
26 being most important as contributing to increased mortality and generally do not support
27 increased mortality risks being related to short-term exposures to crustal materials in U.S.
28 ambient environments examined to date.
29
30 Cause-specific mortality. Findings for new results concerning cause-specific mortality comport
31 well with those for total (non-accidental) mortality, the former showing generally larger effect
April 2002 8-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 size estimates for cardiovascular, respiratory, and/or combined cardiorespiratory excess risks
2 than for total mortality risks. An analysis of specific cardiovascular causes in a large population
3 (The Netherlands) suggested the specific causes of deaths such as heart failure was more strongly
4 associated with PM (and other pollutants) than total cardiovascular mortality.
5
6 Lags. In general, maximum effect sizes for total mortality appear to be obtained with 0-1 day
7 lags, with some studies finding a second peak for 3-4 days lags. There is also some evidence
8 that, if effects distributed over multiple lag days are considered, the effect size may be larger than
9 for any single maximum effect size lag day. Lags are discussed further in Section 8.4.
10
11 Threshold. Few new short-term mortality studies explicitly address the issue of thresholds. One
12 study that analyzed Phoenix, AZ data (Smith et al., 2000) did report some limited evidence
13 suggestive of a possible threshold for PM25 there. However, several different analyses of larger
14 PM10 data sets across multiple cities (Dominici, et al., 2002; Daniels et al., 2000) generally
15 provide little or no support to indicate a threshold for PM10 mortality effects. Threshold issues
16 are discussed further in Section 8.4.
17
18 8.2.3 Mortality Effects of Long-Term Exposure to Ambient Particulate
19 Matter
20 8.2.3.1 Studies Published Prior to the 1996 Particulate Matter Criteria Document
21 8.2.3.1.1 Aggregate Population Cross-Sectional Chronic Exposure Studies
22 Mortality effects associated with chronic, long-term exposure to ambient PM have been
23 assessed in cross-sectional studies and, more recently, in prospective cohort studies. A number
24 of older cross-sectional studies from the 1970s provided indications of increased mortality
25 associated with chronic (annual average) exposures to ambient PM, especially with respect to
26 fine mass or sulfate (SO4=) concentrations. However, questions unresolved at that time regarding
27 the adequacy of statistical adjustments for other potentially important covariates (e.g., cigarette
28 smoking, economic status, etc.) across cities tended to limit the degree of confidence that was
29 placed by the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) on such purely
30 "ecological" studies or on quantitative estimates of PM effects derived from these studies.
31 Evidence comparing the toxicities of specific PM components was relatively limited. The sulfate
April 2002 8-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 and acid components had already been discussed in detail in the previous PM AQCD (U.S.
2 Environmental Protection Agency, 1986).
3
4 8.2.3.1.2 Semi-Individual (Prospective Cohort) Chronic Exposure Studies
5 Semi-individual cohort studies using subject-specific information about relevant covariates
6 (such as cigarette smoking, occupation, etc.) have provided more certain findings of long-term
7 PM exposure effects than past purely "ecological studies" (Kiinzli and Tager, 1997). At the same
8 time, these better designed cohort studies have largely confirmed the magnitude of PM effect
9 estimates from past cross-sectional study results.
10 Prospective cohort semi-individual studies of mortality associated with chronic exposures
11 to air pollution of outdoor origins have yielded especially valuable insights into the adverse
12 health effects of long-term PM exposures. The extensive Harvard Six-Cities Study (Dockery
13 et al., 1993) and the American Cancer Society (ACS) Study (Pope et al., 1995) agreed in their
14 findings of statistically significant positive associations between fine particles and excess
15 mortality, although the ACS study did not evaluate the possible contributions of other air
16 pollutants. Neither study considered multi-pollutant models, although the Six-City study did
17 examine various gaseous and particulate matter indices (including total particles, PM25, SO4=, H+,
18 SO2, and ozone), finding that sulfate and PM25 fine particles were best associated with mortality.
19 The excess RR estimates for total mortality in the Six-Cities study (and 95 percent confidence
20 intervals, CI) per increments in PM indicator levels were: Excess RR=15% (CI=6.1%, 32%) for
21 20 //g/m3 PM10; excess RR=11.4% (CI=4.3%, 23% )for 10 //g/m3 PM25; and excess RR=13.4%
22 (CI=5.1%, 29%) for 5 //g/m3 SO4=. The estimates for total mortality derived from the ACS study
23 were excess RR=6.5% (CI=3.5%, 9.7%) for 10 //g/m3 PM25 and excess RR 3.5% (CI=1.9%,
24 5.1%) for 5 //g/m3 SO4=. The ACS pollutant RR estimates were smaller than those from the
25 Six-Cities study, although their 95% confidence intervals overlap. In some cases in these studies,
26 the life-long cumulative exposure of the study cohorts included distinctly higher past PM
27 exposures, especially in cities with historically higher PM levels (e.g., Steubenville, OH); but
28 more current PM measurements were used to estimate the chronic PM exposures. In the ACS
29 study, the pollutant exposure estimates were based on concentrations at the start of the study
30 (during 1979-1983). Also, the average age of the ACS cohort was 56, which could overestimate
31 the pollutant RR estimates and perhaps underestimate the life-shortening associated with PM
April 2002 8-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 associated mortality. Still, although caution must be exercised regarding the use of the reported
2 quantitative risk estimates, the Six-Cities and ACS semi-individual studies provided consistent
3 evidence of a significant mortality association with long-term exposure to PM of ambient origins.
4 In contrast to the Six-Cities and ACS studies, early results reported by Abbey et al. (1991)
5 and Abbey et al. (1995a) from the Adventist Health Study on Smog (AHSMOG) found no
6 significant mortality effects of previous PM exposure in a relatively young cohort of California
7 nonsmokers. However, these analyses used TSP as the PM exposure metric, rather than more
8 health relevant PM metrics such as PM10 or PM25, included fewer subjects than the ACS study,
9 and considered a shorter follow-up time than the Six-Cities study (ten years vs. 15 years for the
10 Six-Cities study). Moreover, the AHSMOG study included only non-smokers, indicated by the
11 Six-Cities Study as having lower pollutant RR's than smokers, suggesting that a longer follow-up
12 time than considered in the past (10 years) might be required to have sufficient power to detect
13 significant pollution effects than is required in studies that include smokers (such as the
14 Six-Cities and ACS studies). Thus, greater emphasis has been placed thus far on the Six-Cities
15 and ACS studies.
16 Overall, the previously available chronic PM exposure studies collectively indicated that
17 increases in mortality are associated with long-term exposure to ambient airborne particles.
18 Also, effect size estimates for total mortality associated with chronic PM exposure indices
19 appeared to be much larger than those reported from daily mortality PM studies. This suggested
20 that a major fraction of the reported mortality relative risk estimates associated with chronic PM
21 exposure likely reflects cumulative PM impacts above and beyond those exerted by the sum of
22 acute exposure events (i.e., assuming that the latter are fully additive over time). The 1996 PM
23 AQCD (Chapter 12) reached several conclusions concerning four key questions about the
24 prospective cohort studies, as noted below:
25
26 (1) Have potentially important confounding variables been omitted?
27 "While it is not likely that the prospective cohort studies have overlooked plausible
28 confounding factors that can account for the large effects attributed to air pollution, there
29 may be some further adjustments in the estimated magnitude of these effects as individual
30 and community risk factors are included in the analyses." These include individual
31 variables such as education, occupational exposure to dust and fumes, and physical activity,
April 2002 8-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 as well as ecological (community) variables such as regional location, migration, and
2 income distribution. Further refinement of the effects of smoking status may also prove
3 useful."
4
5 (2) Can the most important pollutant species be identified?
6 "The issue of confounding with co-pollutants has not been resolved for the
7 prospective cohort studies . . . Analytical strategies that could have allowed greater
8 separation of air pollutant effects have not yet been applied to the prospective cohort
9 studies." The ability to separate the effects of different pollutants, each measured as a long-
10 term average on a community basis, was clearly most limited in the Six Cities study. The
11 ACS study offered a much larger number of cities, but did not examine differences
12 attributable to the spatial and temporal differences in the mix of particles and gaseous
13 pollutants across the cities. The AHSMOG study constructed time- and location-dependent
14 pollution metrics for most of its participants that might have allowed such analyses, but no
15 results were reported.
16
17 (3) Can the time scales for long-term exposure effects be evaluated?
18 "Careful review of the published studies indicated a lack of attention to this issue.
19 Long-term mortality studies have the potential to infer temporal relationships based on
20 characterization of changes in pollution levels over time. This potential was greater in the
21 Six Cities and AHSMOG studies because of the greater length of the historical air pollution
22 data for the cohort [and the availability of air pollution data throughout the study]. The
23 chronic exposure studies, taken together, suggest that there may be increases in mortality in
24 disease categories that are consistent with long-term exposure to airborne particles, and that
25 at least some fraction of these deaths are likely to occur between acute exposure episodes.
26 If this interpretation is correct, then at least some individuals may experience some years of
27 reduction of life as a consequence of PM exposure."
28
29 (4) Is it possible to identify pollutant thresholds that might be helpful in health
30 assessments?
April 2002 8-72 DRAFT-DO NOT QUOTE OR CITE
-------
1 "Model specification searches for thresholds have not been reported for prospective
2 cohort studies. . . . Measurement error in pollution variables also complicates the search
3 for potential threshold effects. . . . The problems that complicate threshold detection in the
4 population-based studies have a somewhat different character for the long-term studies."
5
6 8.2.3.2 Prospective Cohort Analyses of Chronic Particulate Matter Exposure Mortality
7 Effects Published Since the 1996 Particulate Matter Air Quality Criteria Document
8 Considerable progress has been made towards addressing further the above issues.
9 For example, extensive reanalyses (Krewski et al., 2000) of the Six-Cities and ACS Studies,
10 conducted under sponsorship by the Health Effects Institute (HEI), indicate that the published
11 findings of the original investigators (Dockery et al., 1993; Pope et al., 1995) are based on
12 substantially valid data sets and statistical analyses. The HEI reanalysis project has demonstrated
13 that small corrections in input data have very little effect on the findings and that alternative
14 model specifications further substantiate the robustness of the originally reported findings.
15 In addition, some of the above key questions have been further investigated by Krewski et al.
16 (2000) via sensitivity analyses (in effect, new analyses) for the Six City and ACS studies data
17 sets, including consideration of a much wider range of confounding variables. Newly published
18 analyses of ACS data for more extended time periods (Pope et al., 2002) further substantiate
19 original findings and, also, provide much clearer, stronger evidence for ambient PM exposure
20 relationships with increased lung cancer risk. Recently published analyses of AHSMOG data
21 (Abbey et al., 1999; Beeson et al., 1998) also extend the ASHMOG findings and show some
22 analytic outcomes different from earlier analyses reported out from the study. Results from the
23 Veterans' Administration- Washington University (hereafter called "VA") prospective cohort
24 study are now available (Lipfert et al., 2000). Still other, additional new studies suggestive of
25 possible effects of sub-chronic PM exposures on infant mortality (Woodruff et al., 1997; Bobak
26 and Leon, 1998; Lipfert, 2000; Chen et al., 2002) are also discussed below.
27
28 8.2.3.2.1 Health Effects Institute Reanalyses of the Six-Cities and ACS Studies
29 The overall objective of the HEI "Particle Epidemiology Reanalysis Project" was to
30 conduct a rigorous and independent assessment of the findings of the Six Cities (Dockery et al.,
31 1993) and ACS (Pope et al., 1995) Studies of air pollution and mortality. The following
April 2002 8-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 description of approach, key results, and conclusions is largely extracted from the Executive
2 Summary of the HEI final report (Krewski et al., 2000). The HEI-sponsored reanalysis effort
3 was approached in two steps:
4 •Parti: Replication and Validation. The Reanalysis Team sought to test: (a) if the
5 original studies could be replicated via a quality assurance audit of a sample of the
6 original data and; (b) if the original numeric results could be validated.
7 • Part II: Sensitivity Analyses. The Reanalysis Team tested the robustness of the original
8 analyses to alternate risk models and analytic approaches.
9 The Part I audit of the study population data for both the Six Cities and ACS Studies and of
10 the air quality data in the Six Cities Study revealed the data to be of generally high quality with
11 few exceptions. In both studies, a few errors were found in the data coding for and exclusion of
12 certain subjects; when those subjects were included in the analyses, they did not materially
13 change the results from those originally reported. Because the air quality data used in the ACS
14 Study could not be audited, a separate air quality database was constructed for the sensitivity
15 analyses in Part n.
16 The Reanalysis Team was able to replicate the original results for both studies using the
17 same data and statistical methods as used by the Original Investigators. The Reanalysis Team
18 confirmed the original point estimates, as shown in Table 8-6. For the Six Cities Study, they
19 reported the relative risk of mortality from all causes associated with an increase in fine particles
20 of 20.0 //g/m3 as 1.28, the same as the 1.28 per 20 //g/m3 reported by the Original Investigators.
21 For the ACS Study, the relative risk of all-cause mortality associated with a 20 //g/m3 increase in
22 fine particles was 1.19 in the reanalysis, close to the original 1.14 value.
23 The Part II sensitivity analysis applied an array of different models and variables to
24 determine whether the original results would remain robust to different analytic assumptions and
25 model specifications. The Reanalysis Team first applied the standard Cox model used by the
26 Original Investigators and included variables in the model for which data were available from
27 both original studies, but had not been used in the published analyses (e.g. physical activity, lung
28 function, marital status). The Reanalysis Team also designed models to include interactions
29 between variables. None of these alternative models produced results that materially altered the
30 original findings.
31
April 2002 8-74 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-6. COMPARISON OF SIX CITIES AND AMERICAN CANCER SOCIETY
(ACS) STUDY FINDINGS FROM ORIGINAL INVESTIGATORS AND HEALTH
EFFECTS INSTITUTE REANALYSIS
Type of Health
Effect & Location
Original Investigators'
Six City1'
Six City11
ACS Study
HEI reanalysis Phase I
Indicator
Findings
PM25
PM15/10
PM2S
: Replication
Mortality Risk per Increment in PMa
Total mortality
Excess Relative Risk (95% CI)
13% (4.4%, 23%)
18% (6%, 32%)
6.8% (3.4%, 10%)
Cardiopulmonary mortality
Excess Relative Risk (95% CI)
17% (5.8%, 42%)
e
11.8% (6.8%, 17%)
Six City Reanalysis4
PM15
ACS Study Reanalysis4 PM2S
PM15 (dichot)
PM15 (SSI)
11.3% (3%, 23%)
18% (6%, 34%)
9.1% (3.9%, 14.5%)
4% (1%, 7%)
2% (-1%, 4%)
18.7% (6.3%, 33%)
20% (2%, 41%)
15.3% (9.1%, 21%)
7% (2%, 12%)
6% (3%, 9%)
"Estimates calculated on the basis of differences between the most-polluted and least-polluted cities, scaled to
increments of 20 /^g/m3 increase for PM10, and 10 /ug/m3 increments for PM15 and PM2 s.
bDockery et al. (1993).
cPope et al. (1995).
dKrewski et al. (2000).
'Data presented only by smoking subgroup.
1 Next, for both the Six Cities and ACS Studies, the Reanalysis Team investigated the
2 possible effects of fine particles and sulfate on a range of potentially susceptible subgroups of the
3 population. These analyses did not find differences in PM-mortality associations among
4 subgroups based on various personal characteristics (e.g., including gender, smoking status,
5 exposure to occupational dusts and fumes, and marital status). However, estimated effects of
6 fine particles did vary with educational level; the association between an increase in fine particles
7 and mortality tended to be higher for individuals without a high school education than for those
8 with more education. The Reanalysis Team postulated that this finding could be attributable to
9 some unidentified socioeconomic effect modifier. The authors concluded, "The Reanalysis
10 Team found little evidence that questionnaire variables had led to confounding in either study,
11 thereby strengthening the conclusion that the observed association between fine particle air
12 pollution and mortality was not the result of a critical covariate that had been neglected by the
13 Original Investigators." (Krewski et al., 2000, pp. 219-220).
April 2002
8-75
DRAFT-DO NOT QUOTE OR CITE
-------
1 In the ACS study, the Reanalysis Team tested whether the relationship between ambient
2 concentrations and mortality was linear. They found some indications of both linear and
3 nonlinear relationships, depending upon the analytic technique used, suggesting that the shapes
4 of the concentration-response relationships warrant additional research in the future.
5 One of the criticisms of both original studies has been that neither analyzed the effects of
6 change in pollutant levels over time. In the Six Cities Study, for which such data were available,
7 the Reanalysis Team tested whether effect estimates changed when certain key risk factors
8 (smoking, body mass index, and air pollution) were allowed to vary over time. In general, the
9 reanalysis results did not change when smoking and body mass index were allowed to vary over
10 time. The Reanalysis Team did find for the Six Cities Study, however, that when the general
11 decline in fine particle levels over the monitoring period was included as a time-dependent
12 variable, the association between fine particles and all-cause mortality was reduced (Excess
13 RR = 10.4%, [1.5%, 20%]). This would be expected, since the most polluted cities would be
14 expected to have the greatest decline as pollution controls were applied. Despite this adjustment,
15 the PM25 effect estimate continued to be positive and statistically significant.
16 To test the validity of the original ACS air quality data, the Reanalysis Team constructed
17 and applied its own air quality dataset from available historical data. In particular, sulfate levels
18 with and without adjustment were found to differ by about 10% for the Six Cities Study. Both the
19 original ACS Study air quality data and the newly constructed dataset contained sulfate levels
20 inflated by approximately 50% due to artifactual sulfate. For the Six Cities Study, the relative
21 risks of mortality were essentially unchanged with adjusted or unadjusted sulfate. For the ACS
22 Study, adjusting for artifactual sulfate resulted in slightly higher relative risks of mortality from
23 all causes and cardiopulmonary disease compared with unadjusted data, while the relative risk of
24 mortality from lung cancer was lower after the data had been adjusted. Thus, the Reanalysis
25 Team found essentially the same results as the original Harvard Six-Cities and ACS studies, even
26 after using independently developed pollution datasets and after adjusting for sulfate artifact.
27 Because of the limited statistical power to conduct most model specification sensitivity
28 analyses for the Six Cities Study, the Reanalysis Team conducted the majority of its sensitivity
29 analyses using only the ACS Study dataset that considered 151 cities. When a range of city-level
30 (ecologic) variables (e.g., population change, measures of income, maximum temperature,
31 number of hospital beds, water hardness) were included in the analyses, the results generally did
April 2002 8-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 not change. The only exception was that associations with fine particles and sulfate were
2 reduced when city-level measures of population change or SO2 were included in the model.
3 A maj or product of the Reanalysis Proj ect is the determination that both pollutant variables
4 and mortality appear to be spatially correlated in the ACS Study dataset. If not identified and
5 modeled correctly, spatial correlation could cause substantial errors in both the regression
6 coefficients and their standard errors. The Reanalysis Team identified several methods for
7 addressing this, each of which resulted in some reduction in the estimated regression coefficients.
8 The full implications and interpretations of spatial correlations in these analyses have not been
9 resolved, and were noted to be an important subject for future research.
10 When the Reanalysis Team sought to take into account both the underlying variation from
11 city to city (random effects) and variation from the spatial correlation between cities, associations
12 were still found between mortality and sulfates or fine particles. Results of various models, using
13 alternative methods to address spatial autocorrelation and including different ecologic covariates,
14 found fine particle-mortality associations that ranged from 1.11 to 1.29 (RR reported by original
15 investigators was 1.17) per 24.5 //g/m3 increase in PM25. With the exception of SO2,
16 consideration of other pollutants in these models did not alter the associations found with
17 sulfates. The authors reported associations that were stronger for SO2 than for sulfate, which
18 may indicate that the sulfate with artifact was "picking up" some of the SO2 association, perhaps
19 because the artifact is in part proportional to the prevailing SO2 concentration (Coutant, 1977).
20 It should be recognized that the Reanalysis Team did not use data adjusted for artifactual sulfate
21 for most alternative analyses. When they did use adjusted sulfate data, relative risks of mortality
22 from all causes and cardiopulmonary disease increased. This result suggests that more analyses
23 with adjusted sulfate might result in somewhat higher relative risks associated with sulfate. The
24 Reanalysis Team concluded: "it suggests that uncontrolled spatial autocorrelation accounts for
25 24% to 64% of the observed relation. Nonetheless, all our models continued to show an
26 association between elevated risks of mortality and exposure to airborne sulfate" (Krewski et al.,
27 2000, p. 230).
28 In summary, the reanalyses generally confirmed the original investigator's findings of
29 associations between mortality and long-term exposure to PM, while recognizing that increased
30 mortality may be attributable to more than one ambient air pollution component. Regarding the
31 validity of the published Harvard Six-Cities and ACS Studies, the HEI Reanalysis Report
April 2002 8-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 concluded that: "Overall, the reanalyses assured the quality of the original data, replicated the
2 original results, and tested those results against alternative risk models and analytic approaches
3 without substantively altering the original findings of an association between indicators of
4 particulate matter air pollution and mortality."
5
6 8.2.3.2.2 The Extension of the ACS Study
1 A very recent publication by Pope et al. (2002) extends the analyses (Pope et al., 1995) and
8 reanalyses (Krewski et al., 2000) of the ACS CPS-II cohort to an additional eight years of foilow-
9 up. The new study has a number of advantages over the previous analyses, in that it: (a) doubles
10 the follow-up time from eight years to sixteen years, and triples the number of deaths;
11 (b) expands the ambient air pollution data substantially, including two recent years of fine
12 particle data, and adds data on gaseous co-pollutants; (c) improves statistical adjustments for
13 occupational exposure; (d) incorporates data on dietary covariates believed to be important
14 factors in mortality, including total fat consumption, and consumption of vegetables, citrus fruit,
15 and high-fiber grains; and (e) uses recent developments in non-parametric spatial smoothing and
16 random effects statistical models as input to the Cox proportional hazards model. Each
17 participant was identified with a specific metropolitan area, and mean pollutant concentrations
18 were calculated for all metropolitan areas with ambient air monitors in the one to two years prior
19 to enrollment. Ambient pollution during the follow-up period was extracted from the AIRS data
20 base. Averages of daily averages of the gaseous pollutants were used except for ozone, where
21 the average daily 1-hour maximum was calculated for the whole year and for the typical peak
22 ozone quarter (July, August, September). Mean sulfate concentrations for 1990 were calculated
23 from archived filters using quartz filters, virtually eliminating the historical sulfate artifact
24 leading to overestimation of sulfate concentrations.
25 The Krewski et al. (2000), Burnett et al. (200la), and Pope et al. (2002) studies were
26 concerned that survival times of participants in nearby locations might not be independent of
27 each other, due to missing, unmeasured or mis-measured risk factors or their surrogates that may
28 be spatially correlated with air pollution, thus violating an important assumption of the Cox
29 proportional hazards model. Model fitting proceeded in two stages, the first of which was an
30 adjusted relative risk model with a standard Cox proportional hazards model including
31 individual-specific covariates and indicator variables for each metropolitan area, but not air
April 2002 8-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 pollutants. In the second stage, the adjusted log(relative risks) were fitted to fine particle
2 concentrations or other air pollutants by a random effects linear regression model.
3 Models were estimated separately for each of four mortality (total, cardiopulmonary, lung
4 cancer, and causes other than cardiopulmonary or lung cancer deaths) endpoints for the entire
5 follow-up period, and for fine particles in three time periods (1979-1983, 1999-2000, and the
6 average of the mean concentrations in these two periods). The results are shown in Table 8-7.
7 Figures 8-8, 8-9, and 8-10 show the results displayed in Figures 2, 3, and 5 in Pope et al. (2002).
8 Figure 8-8 shows that a smooth non-parametric model can be reasonably approximated by a
9 linear model for all-cause mortality, cardiopulmonary mortality, and other mortality; but the
10 log(relative risk) model for lung cancer appears to be non-linear, with a steep linear slope up to
11 an annual mean concentration of about 13 //g/m3 and a flatter linear slope at fine particle
12 concentrations > 13 //g/m3.
13
14
TABLE 8-7. SUMMARY OF RESULTS FROM THE EXTENDED ACS STUDY*
PM2 5, average over PM2 5, average over PM2 5, average over all
Cause of death 1979-1983 1999-2000 seven years
All causes 4.1% (0.8,7.5%) 5.9% (2.0,9.9%) 6.2% (1.6, 11.0%)
Cardiopulmonary 5.9% (1.5, 10.5%) 7.9% (2.3, 14.0%) 9.3% (3.3, 15.8%)
Lung cancer 8.2% (1.1, 15.8%) 12.7% (4.1,21.9%) 13.5% (4.4,23.4%)
Other 0.8% (-3.0, 4.8%) 0.9% (-3.4, 5.5%) 0.5% (-4.8, 6.1%)
'Adjusted mortality excess risk ratios (95% confidence limits) per 10 ,wg/m3 PM2 5 by cause of death associated
with each of the multi-year averages of fine particle concentrations. The multi-year average concentrations are
used as predictors of cause-specific mortality for all of the 16 years (1982-1998) of the ACS follow-up study.
The excess risk ratios are obtained from the baseline random effects Cox proportional hazards models adjusted
for age, gender, race, smoking, education, marital status, BMI, alcohol consumption, occupational dust
exposure, and diet. Based on Table 2 in Pope et al. (2002) and more precise data from authors (G. Thurston,
personal communication, March 13. 2002).
1 Figure 4 in Pope et al. (2000) shows results of the stratified first-stage models: ages
2 < 60 and > 69 yr are marginally significant for total mortality; ages > 70 are significant for
3 cardiopulmonary mortality; and ages 60-69 for lung cancer mortality. Men are at significantly
April 2002 8-79 DRAFT-DO NOT QUOTE OR CITE
-------
o
0.2 -
0.1 -
-0.1 -
-0.3 -
-0.4 -
All Cause
0.2 -
^ 0.1 -
en
& nn
0, °-0 '
in -0.1 -
Q)
I**'
0
-1 -0.3 -
-0.4 -
^t** '
.. . -- ' ** ^-^--*"""""'"""\ — •
•^ .-••*•
y1"
x1"
(X*
Cardiopulmonary
10
15
20
10
15
20
0.2 :
^ 0.1 -
"
-------
Standard Cox Models
1.10-
1.15- £
3 «5 1
62
~ to
Random
Effects
(with all
covariates)
Random Effects (with
all covariates) plus
Spatial Smoothing
s£
o
CD
<> <
1 -L 1
1 ?
I i
f ]
> 4
5
5
.
<
>
[ ^
^
^
c
c
- £/
>
4
.
> 4
^ <
; 05 TO ^ ^
i §1 S?
3 -I > I 0.
> ^ >
_ _
, ^ 0
. .
— 1.05-
1.00 -
n 95 _
- M- f-f f f f :
>
4
Figure 8-9. Relative risk of total and cause-specific mortality at 10 Aig/m3 PM2 5 (mean
of 1979-1983) of alternative statistical models. The standard Cox models
are built up in a sequential stepwise manner from the baseline model
stratified by age, gender, and race by adding additional covariates. The
random effects model allows for additional city-to-city variation, and the
spatial smoothing models show the effects of increasingly aggressive
adjustment for spatial correlation. Based on Pope et al. (2002).
April 2002
8-81
DRAFT-DO NOT QUOTE OR CITE
-------
Fine Inhalable, coarse, and Sulfate
Particles total suspended Particlesparticles
Gaseous Pollutants
CO
T- CO 1.15-
3
ro 1.10-
O
-1 1.05-
^ H f\r\ _
I .UU "
0.95 -
0.90 -
£• 1.25-
C 1.20-
o
E 1.15-
3 1.10-
Q.
.° 1.05-
H -i nn _
i; 1 .UU
O 0.95 -
0.90 -
1.25 -
fe 120-
£ 1.15-
o 1.10-
g> 1.05-
D -i nn -
• 1 .UU
0.95 -
0.90 -
1.25 -
1.20 -
0) 1.15-
+= 1.10-
o
= 1.05-
**• 1.00-
0.95 -
0.90 -
0
c
r-
x».
^
^
Q
^
4
4
4
4
D C
1 C
^ 2
»" ^
. Q
^ <
4
> 4
^.....^
» -g
•> 6
^' -^.
r s
. c
J
4
>
4
^
4
>
^...^
J C
1 s
E" 2
D
^
^ 4
»
4
>
4
>
4
} OC
3 I--
-^ ---
E^ S
L Q
^
k
4
K
4
>
'.....^
>
3 Q, 00 OC
L "s S2- C
£ i Q. Q
: S w c/
. Q. 1- 1-
^
^ T ^
* 1 ]
^
j
^
-L -1 .
>
j
j
4
^
3 CO T-
co o c
^ ~^ s°, 5
3 CO O* C
- 1- CO 0
4
T i
> 1
[ I
T
0 I
. 1
-p
4. <
1
k
r • •
4
^ -
r i
) C
3 U
r
>
n
L <
k .
.
<
>•
4
>
L ]
00
? ^>
? CNI C
3 00 OC
)" O" C
3 CO 2
^ ^L
]
^
..
4
ii1
cc
c
r o
) CC
) C
: z
[ J
M
4
»
4
H
i
i I
r c
: c
r
' J
^ 1
4
>
».--:
oo ^
en T-
J- CN «
) 00 c
) O ^
) O C
^ 4
f ^
1
fr
4i
-N ^-
3 §
3 °
)" C
r 4,
[]
4
>
4
•<
>
u
3~ OC
^ c
g o
3 OC
.X S.
f c
4
>
4
>
4
>
.
>
j
3
5
J
3
*S
)"
^
^
>
>-
Figure 8-10. Relative risk of total and cause-specific mortality for particle metrics and
gaseous pollutants over different averaging periods (years 1979-2000 in
parentheses). Based on Pope et al. (2002).
April 2002
8-82
DRAFT-DO NOT QUOTE OR CITE
-------
1 smoothing can increase the magnitude of the RR and increase its significance by reducing the
2 width of the confidence intervals in the "50%-span" and "lowest variance" smoothing methods.
3 For lung cancer mortality, spatial smoothing very slightly decreases the magnitude of the RR but
4 also increases its significance by reducing the width of the confidence intervals in the "50%-
5 span" and "lowest variance" smoothing methods.
6 Figure 8-10 shows a statistically significant relationship between fine particles and total,
7 cardiopulmonary, and lung cancer mortality whatever averaging span was used for PM2 5, and
8 slightly larger for the average concentration of the 1979-1983 and 1999-2000 intervals. PM15 for
9 1979-1983 is significantly associated with cardiopulmonary mortality and marginally with total
10 mortality, whereas 1987-1996 PM15 is not quite significantly associated with cardiopulmonary
11 mortality only. Coarse particles and TSP are not significantly associated with any endpoint, but
12 are positively associated with cardiopulmonary mortality. Sulfate particles are very significantly
13 associated with all endpoints including mortality from all other causes, but only marginally for
14 lung cancer mortality using 1990 filters.
15 Figure 8-10 shows a highly significant relationship between SO2 and all endpoints
16 including mortality from other causes, although weaker for lung cancer mortality. Ozone (using
17 only the third quarter for 1982-1998) shows a marginally significant relationship with
18 cardiopulmonary mortality, but not the year-round average. The other criteria pollutants, CO and
19 NO2, are not significantly and positively related to any mortality endpoint, unlike the findings for
20 acute mortality studies.
21 This paper is noteworthy because it shows that the general pattern of findings in the first
22 eight years of the study (Pope et al., 1995; Krewski et al., 2000) could be reasonably extrapolated
23 to the patterns that remain present with twice the length of time on study and three times the
24 number of deaths. As shown later in Table 8-12 (Pg. 8-94), the excess relative risk estimate
25 (95% CI) per 10 //g/m3 PM25 for total mortality in the original ACS study (Pope et al., 1995) was
26 6.6% (3.6, 9.9%); in the ACS reanalysis (Krewski et al., 2000, Table 20, Full Model) it was 6.6%
27 (3.6, 9.9%); and, in the extended ACS data set (Pope et al., 2002), it was 4.1% (0.8, 7.5%) using
28 the 1979-1983 data and 6.2% (1.6, 11%) using the average of the 1979-1983 and 1999-2000 data.
29 The excess relative risk estimate (95% CI) per 10 //g/m3 PM25 for cardiopulmonary mortality in
30 the original ACS study (Pope et al., 1995) was 11.6% (6.6, 16.7%); in the ACS reanalysis
31 (Krewski et al., 2000, Table 20, Full Model), it was 10.6% (5.9, 15.4%); and, in the extended
April 2002 8-83 DRAFT-DO NOT QUOTE OR CITE
-------
1 ACS data set (Pope et al., 2002), it was 5.9% (1.5, 10.5%) using the 1979-1983 data and 9.3%
2 (3.3, 15.8%) using the average of the 1979-1983 and 1999-2000 data. Thus, the additional data
3 and statistical analyses in (Pope et al., 2000) yield somewhat smaller estimates than in the
4 original study (Pope et al., 1995), but similar estimates to the reanalysis of the original ACS data
5 set (Krewski et al., 2000).
6 The authors draw the following conclusions:
7 (1) The apparent association between fine particle pollution and mortality persists with longer
8 follow-up as the participants in the cohort grow older and more of them die.
9 (2) The estimated fine particle effect on cardiopulmonary mortality and cancer mortality was
10 relatively stable, even after adjustment for smoking status, although the estimated effect was
11 larger and more significant for never-smokers vs. former or current smokers. The estimates
12 were relatively robust against inclusion of many additional covariates: education, marital
13 status, BMI, alcohol consumption, occupational exposure, and dietary factors. However, as
14 the authors note, the data on individual risk factors was collected only at the time of
15 enrollment and has not been updated, so that changes in these factors since 1982 could
16 introduce risk factor exposure mis-classification, with a loss of precision in the estimates
17 and might limit the ability to characterize time dependency of effects.
18 (3) Additional assessments for potential spatial or regional differences not controlled in the
19 first-stage model were evaluated. If there are unmeasured or inadequately modeled risk
20 factors that are different across locations or spatially clustered, then PM risk estimates may
21 be biased. If the clustering is independent or random or independent across areas, then
22 adding a random-effects component to the Cox proportional hazards model can deal with the
23 problem. However, if location is associated with air pollution, then the spatial correlation
24 may be evaluated using non-parametric smoothing methods. No significant spatial auto-
25 correlation was found after controlling for fine particles. Even after adjusting for spatial
26 correlation, the estimated PM2 5 effects were significant and persisted for cardiopulmonary
27 mortality and lung cancer mortality and were borderline significant for total mortality, but
28 with much wider confidence intervals after spatial smoothing.
29 (4) Elevated total, cardiopulmonary, and lung cancer mortality risks were associated with fine
30 particles, but other mortality was not. PM10 for 1987-1996 and PM15 for 1979-1983 were
31 just significantly associated with cardiopulmonary mortality only, but PM10_2 5 and TSP were
April 2002 8-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 not associated with total or any cause-specific mortality. All endpoints were very
2 significantly associated with sulfates, except lung cancer with 1990 sulfate data. All
3 endpoints were very significantly associated with SO2 using 1980 data, with total and other
4 mortality using the 1982-1998 data, but cardiopulmonary and lung cancer mortality had only
5 a borderline significant association with the 1982-1998 SO2 data. None of the other gaseous
6 pollutants had a significant positive association with any endpoint, except for a borderline
7 association of third-quarter ozone to cardiopulmonary mortality. In summary, neither coarse
8 thoracic particles nor TSP were significantly associated with mortality, nor were NO2 and
9 CO on a long-term exposure basis. (It should be noted, however, that additional analyses
10 may yet be useful. The data would allow segmentation of mortality into smaller periods
11 rather than the whole 16 year duration of the mortality follow up, for example from 1982
12 through 1989 and from 1990 through 1998. In this way, it may be possible to evaluate any
13 changes in PM mortality rate over time.)
14 (5) The concentration-response curves estimated using non-parametric smoothers were all
15 monotonic and (except for lung cancer) nearly linear. However, the shape of the curve may
16 become non-linear at much higher concentrations.
17 (6) The excess risk from PM2 5 exposure is much smaller than that estimated for cigarette
18 smoking for current smokers in the same cohort (Pope et al., 1995), RR = 2.07 for total
19 mortality, RR = 2.28 for cardiopulmonary mortality, and RR = 9.73 for lung cancer
20 mortality. In the more polluted areas of the United States, the relative risk for substantial
21 obesity (a known risk factor for cardiopulmonary mortality) is larger than that for PM2 5, but
22 the relative risk from being moderately overweight is somewhat smaller.
23
24 8.2.3.2.3 AHSMOGAnalyses
25 The Adventist Health Study of Smog (AHSMOG) represents a third major U.S. prospective
26 cohort study of chronic PM exposure-mortality effects. In 1977, the study enrolled 6,338
27 non-smoking non-Hispanic white Seventh Day Adventist residents of California, ages 27 to
28 95 years. The participants had resided for at least 10 years within 5 miles (8 km) of their then-
29 current residence locations, either within the three major California air basins (San Diego,
30 Los Angeles, or San Francisco) or else were part of a random 10% sample of Adventist Health
31 Study participants residing elsewhere in California. The study has been extensively described
April 2002 8-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 and initial results reported elsewhere (Hodgkin et al., 1984; Abbey et al., 1991; Mills et al.,
2 1991). In the latest AHSMOG analyses (Abbey et al., 1999), mortality status of the subjects after
3 ca. 15-years of follow-up (1977-1992) was determined by various tracing methods, finding 1,628
4 deaths (989 female, 639 male) in the cohort. This is a 50% percent increase in the follow-up
5 period vs. previous AHSMOG reports, which increases the power of the latest analyses over past
6 published ones. Of 1,575 deaths from all natural (non-external) causes, 1,029 were
7 cardiopulmonary, 135 were non-malignant respiratory (ICD9 codes 460-529), and 30 were lung
8 cancer (ICD9 code 162) deaths. Abbey et al. (1999) also created another death category,
9 contributing respiratory causes (CRC). CRC included any mention of nonmalignant respiratory
10 disease as either an underlying or a "contributing cause" on the death certificate. Numerous
11 analyses were done for the CRC category, due to the large numbers and relative specificity of
12 respiratory causes as a factor in the deaths. Education was used to index socio-economic status,
13 rather than income. Physical activity and occupational exposure to dust were also used as
14 covariates.
15 Cox proportional hazard models adjusted for a variety of covariates, or stratified by sex,
16 were used. The "time" variable used in most of the models was survival time from date of
17 enrollment, except that age on study was used for lung cancer effects due to the expected lack of
18 short-term effects. A large number of covariate adjustments were evaluated, yielding results for
19 all non-external mortality as shown in Table 8-8 and described by Abbey et al. (1999).
20 Essentially no statistically significant PM related effects were observed for either males or
21 females, except RR = 1.08 for males in relation to 30 days per year with PM10 > 100 //g/m3.
22 An analogous pattern of results was found for cause-specific mortality analyses of the
23 AHSMOG data. That is, positive and statistically significant effects on cardiopulmonary deaths
24 were found in models that included both sexes and adjustment for age, pack-years of smoking,
25 and body-mass index (BMI) (RR = 1.14, 95% CI1.03-1.56 for 30 day/yr > 100 //g/m3 PM10).
26 Subsets of the cohort had elevated risks: (a) former smokers had higher RR's than never-
27 smokers (RR for PM10 exceedances for never-smokers was marginally significant by itself);
28 (b) subjects with low intake of anti-oxidant vitamins A, C, E had significantly elevated risk of
29 response to PM10, whereas those with adequate intake did not (suggesting that dietary factors or,
30 possibly, other socio-economic or life style factors for which they are a surrogate may be
31 important covariates); and (c) there also appeared to be a gradient of PM10 risk with respect to
April 2002 8-86 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-8. RELATIVE RISK OF MORTALITY FROM ALL NONEXTERNAL
CAUSES, BY SEX AND AIR POLLUTANT, FOR AN ALTERNATIVE
COVARIATE MODEL IN THE ASHMOG STUDY
Pollution Index
PM10>100, d/yr.
PM10 mean
SO4 mean
O3>100 ppb, h/yr.
SO2 mean
Pollution Incr.
30 days/yr.
20 Mg/m3
S^g/m3
551 h/yr. (IQR)
3.72 (IQR)
RR
0.958
0.950
0.901
0.90
1.00
Females
LCL
0.899
0.873
0.785
0.80
0.91
UCL
1.021
1.033
1.034
1.02
1.10
RR
1.082
1.091
1.086
1.140
1.05
Males
LCL
1.008
0.985
0.918
0.98
0.94
UCL
1.162
1.212
2.284
1.32
1.18
LCL = Lower 95% confidence limit
Source: Abbey etal. (1999).
UCL = Upper 95% confidence limit
1 time spent outdoors, with those who had spent at least 16 h/wk outside at greater risk from PM10
2 exceedances. The extent to which time spent outdoors is a surrogate for other variables or is a
3 modifying factor reflecting temporal variation in exposure to ambient air pollution is not certain.
4 For example, if the males spent much more time outdoors than females, outdoor exposure time
5 could be confounded with gender. When the cardiopulmonary analyses are broken down by
6 gender (Table 8-9), the RR's for female deaths were generally smaller than that for males,
7 although none of the risks for PM indices or gaseous pollutants were statistically significant.
8 The AHSMOG cancer analyses showed a confusing array of results for lung cancer mortality
9 (Table 8-10). For example, RR's for lung cancer deaths were statistically significant for males
10 for PM10 and O3 metrics, but not for females. In contrast, such cancer deaths were significant for
11 mean NO2 only for females (but not for males), but lung cancer metrics for mean SO2 were
12 significant for both males and females. This pattern is not readily interpretable, but is reasonably
13 attributable to the very small numbers of cancer-related deaths (18 for females and 12 for males),
14 resulting in wide RR confidence intervals and very imprecise effects estimates.
15 The analyses reported by Abbey et al. (1999) attempted to separate PM10 effects from those
16 of other pollutants by use of two-pollutant models, but no quantitative findings from such models
17 were reported. Abbey et al. mentioned that the PM10 coefficient for CRC remained stable or
April 2002
8-87
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-9. RELATIVE RISK OF MORTALITY FROM CARDIOPULMONARY
CAUSES, BY SEX AND AIR POLLUTANT, FOR AN ALTERNATIVE
COVARIATE MODEL
Pollution Index
PM10>100, d/yr.
PM10 mean
SO4 mean
O3>100 ppb, h/yr.
O3 mean
SO2 mean
Pollution Incr.
30 days/yr.
20 Mg/m3
5 A(g/m3
551 h/yr. (IQR)
10 ppb
3.72 (IQR)
RR
0.929
0.933
0.950
0.88
0.975
1.02
Females
LCL
0.857
0.836
0.793
0.76
0.865
0.90
UCL
1.007
1.042
1.138
1.02
1.099
1.15
RR
1.062
1.082
1.006
1.06
1.066
1.01
Males
LCL
0.971
0.943
0.926
0.87
0.920
0.86
UCL
1.162
1.212
1.086
1.29
1.236
1.18
LCL = Lower 95% confidence limit
Source: Abbey etal. (1999).
UCL = Upper 95% confidence limit
TABLE 8-10. RELATIVE RISK OF MORTALITY FROM LUNG CANCER BY AIR
POLLUTANT AND BY GENDER FOR AN ALTERNATIVE COVARIATE MODEL
Pollution
Index
PM10>100,d/yr.
PM10 mean
NO2 mean
O3>100 ppb,
h/yr
O3 mean
SO2 mean
Pollution
Incr.
30 days/yr.
20 ,wg/m3
19.78 (IQR)
551 h/yr
(IQR)
10 ppb
3.72 (IQR)
Smoking
Category
All1
All
All
All
never
smoker
past
smoker
All
All
never
smokers
RR
1.055
1.267
2.81
1.39
0.805
3.01
2.99
Females
LCL
0.657
0.652
1.15
0.53
0.436
1.88
1.66
UCL
1.695
2.463
6.89
3.67
1.486
4.84
5.40
RR
1.831
2.736
1.82
4.19
6.94
4.25
1.853
1.99
Males
LCL
1.281
1.455
0.93
1.81
1.12
1.50
0.994
1.24
UCL
2.617
5.147
3.57
9.69
43.08
12.07
3.453
3.20
'All = both never smokers and past smokers.
LCL = Lower 95% confidence limit
Source: Abbey etal. (1999).
UCL = Upper 95% confidence limit
April 2002
DRAFT-DO NOT QUOTE OR CITE
-------
1 increased when other pollutants were added to the model. Lung cancer mortality models for
2 males evaluated co-pollutant effects in detail and indicated that NO2 was non-significant in all
3 two-pollutant models but the other pollutant coefficients were stable. The PM10 and O3 effects
4 remained stable when SO2 was added, suggesting possible independent effects, but PM10 and O3
5 effects were hard to separate because these pollutants were highly correlated in this study.
6 Again, however, the very small number of lung cancer observations and likely great imprecision
7 of reported effects estimates markedly diminish the credibility of these results.
8 Other analyses, by Beeson et al. (1998), evaluated essentially the same data as in Abbey
9 et al. (1999), but focused on lung cancer incidence (1977-1992). There were only 20 female and
10 16 male lung cancer cases among the 6,338 subjects. Exposure metrics were constructed to be
11 specifically relevant to cancer, being the annual average of monthly exposure indices from
12 January, 1973 through the following months, but ending 3 years before date of diagnosis of the
13 case (i.e., representing a 3-year lag between exposure and diagnosis of lung cancer). The
14 covariates in the Cox proportional hazards model were pack-years of smoking and education, and
15 the time variable was attained age. Many additional covariates were evaluated for inclusion, but
16 only 'current use of alcohol' met criteria for inclusion in the final model. Pollutants evaluated
17 were PM10, SO2, NO2, and O3. No interaction terms with the pollutants proved to be significant,
18 including outdoor exposure times. The RR estimates for male lung cancer cases were:
19 (a) positive and statistically significant for all PM10 indicators; (b) positive and predominantly
20 significant for O3 indicators, except for mean O3, number of O3 exceedances > 60 ppb, and in
21 former smokers; (c) positive and significant for mean SO2, except when restricted to proximate
22 monitors; and (d) positive but not significant for mean NO2. When analyses are restricted to use
23 of air quality data within 32 km of the residences of subj ects, the RR over the IQR of 24 //g/m3 in
24 the full data set is 5.21 (or RR=1.989 for 10 //g/m3). The female RR's were all much smaller
25 than for males, not being statistically significant for any indicator of PM10 or O3, but being
26 significant for mean SO2.
27 The AHSMOG investigators also attempted to compare effects of fine vs. coarse particles
28 (McDonnell et al, 2000). For AHSMOG participants living near an airport (n=3,769), daily
29 PM2 5 concentrations were estimated from airport visibility using previously-described methods
30 (Abbey et al, 1995b). Table 8-11 shows the results of this analysis for the male subset near
31 airports (n=1266). Given the smaller numbers of subjects in these subset analyses, it is not
April 2002 8-89 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-11. COMPARISON OF EXCESS RELATIVE RISKS FOR THREE
PARTICLE METRICS IN THE MALE SUBSET OF THE AHSMOG STUDY
PM Metric
Underlying Cause of Mortality
All causes
Any contributing nonmalignant
respiratory cause
Lung cancer
PM25
8.4.% (-2.1,
—
9.3% (-3. 8,
22.6% (-2.9,
—
19.8% (-8.8,
39.1% (-21,
—
35.7% (-28,
21%)
24%)
55%)
58%)
146%)
157%)
PM10.2.5
—
5.2% (-8.2, 21%)
-1.0% (-16, 17%)
—
19.6% (-12, 64%)
6.2% (-27, 54%)
—
25. 9% (-3 8, 156%)
7.2% (-52, 137%)
PM10
9.9% (-4. 1,26%)
30.5% (-4.8, 140%)
5 1.2% (-30, 224%)
1 necessarily surprising that no pollutants are statistically significant in these regressions. It is
2 important, however, to caveat that the PM2 5 exposures were estimated from visibility
3 measurements (increasing exposure measurement error), and a very uneven and clustered
4 distribution of exposures was presented by the authors. Also, the PM10_2 5 values were calculated
5 from the differencing of PM10 and PM25, likely contributing to additional measurement error for
6 the coarse particle (PM10_2 5) variable used in the analyses.
7
8 8.2.3.2.4 The EPRI- Washington University Veterans' Cohort Mortality Study
9 Lipfert et al. (2000b) reported preliminary results from new large-scale mortality analyses
10 using a prospective cohort of up to 70,000 men assembled by the U.S. Veterans Administration
11 (VA) in the mid 1970s. While much smaller than the ACS cohort, this study group shares the
12 similarity that it was not originally formed to study air pollution, but was later linked to air
13 pollution data collected separately, much of it subsequent to the start of the study. The AHSMOG
14 and Six City studies were designed as prospective studies to evaluate long-term effects of air
15 pollution and had concurrent air pollution measurements. The ACS study was also a prospective
16 study, with air pollution data at about the approximate time of enrollment but not subsequently
April 2002
8-90
DRAFT-DO NOT QUOTE OR CITE
-------
1 (Pope et al., 1995). The extended ACS data incorporated much more air pollution data,
2 including TSP data back to the 1960s and more recent fine particle data. The PM25 data set was
3 smaller than the TSP data set and similar to the ACS data.
4 The study cohort was male, middle-aged (51 ± 12 years) and included a larger proportion of
5 African-Americans (35%) than the U.S. population as a whole and a large percentage of current
6 or former smokers (81%). The cohort was selected at the time of recruitment as being mildly to
7 moderately hypertensive, with screening diastolic blood pressure (DBF) in the range 90 to
8 114 mm Hg (mean 96, about 7 mm more than the U.S. population average) and average systolic
9 blood pressure (SBP) of 148 mm Hg. The subjects had all been healthy enough to be in the U.S.
10 armed forces at one time. A comparison of their pre-existing health status at time of study
11 recruitment vs. the initial health status of the other cohorts would be of interest. The study that
12 led to the development of this clinical cohort (Veterans Administration Cooperative Study Group
13 on Antihypertensive Agents, 1970; 1967) was a "landmark" VA cooperative study demonstrating
14 that anti-hypertensive treatment markedly decreased morbidity and mortality (Perry et al., 1982).
15 The clinical cohort itself involved actual clinical rather than research settings. Some differences
16 between the VA cohort and other prospective cohorts are noted below.
17 Pollutant levels of the county of residence at the time of entry into the study were used for
18 analyses versus levels at the VA hospital area. Contextual socioeconomic variables were also
19 assembled at the ZIP-code and county levels. The ZIP-code level variables were average
20 education, income, and racial mix. County-level variables included altitude, average annual
21 heating-degree days, percentage Hispanic, and socioeconomic indices. Census tract variables
22 included poverty rate and racial mix. County-wide air pollution variables included TSP, PM10,
23 PM25, PM15, PM15_25, SO4, O3, CO, and NO2 levels at each of the 32 VA clinics where veterans
24 were enrolled. In addition to considering average exposures over the entire period, three
25 sequential mortality follow-up periods (1976-81, 1982-88, 1989-96) were also considered
26 separately in statistical analyses, which evaluated relationships of mortality in each of those
27 periods to air pollution in different preceding, concurrent, or subsequent periods (i.e., up to 1975,
28 1975-81, 1982-88, and 1989-86, for TSP in the first three periods, PM10 for the last, and NO2,
29 95 percentile O3, and 95 percentile CO for all four periods). Mortality in the above-noted periods
30 was also evaluated in relation to SO4 in each of the same four periods noted for NO2, O3, and CO,
31 and to PM25, PM15, and PM15.25 in 1979-81 and 1982-84.
April 2002 8-91 DRAFT-DO NOT QUOTE OR CITE
-------
1 The use of diastolic and systolic blood pressure in the reported regression results may
2 require further evaluation. The VA Cohort participants were recruited on the basis of initial
3 diastolic blood pressure (DBF) of 90 to 114 mm Hg.
4 The participants in the VA Cohort clearly formed an "at-risk" population, and the results by
5 Vasan et al. (2001) make more plausible the hypothesis in (Lipfert et al., 2000b, p. 62) that
6 ". . .the relatively high fraction of mortality within this cohort may have depleted it of susceptible
7 individuals in the later periods of follow-up." The role of DBF and SBP as predictors in
8 regression models in the VA Cohort may be considered as closer to the endpoint (mortality) than
9 as a more distal behavioral, environmental, or contextual predictor of mortality such as air
10 pollution, temperature, smoking behavior, BMI, etc. The author (F. Lipfert, personal
11 communication, March 28, 2002) notes that personal-level variables tend to interact only with
12 each other, as do county-level variables with little correlation across spatial scales.
13 The estimated mean risk of cigarette smoking in this cohort (1.43 Relative Risk) is also
14 smaller than that of the Six City cohort (RR = 1.59) and the ACS cohort (RR = 2.07 for a current
15 smoker). Some possible differences include the higher proportion of former or current smokers
16 in this cohort (81%) vs. 51% in the ACS study and 42 to 53% in the Six City study. A possibly
17 more important factor may be the difference in education levels, with only 12% of the ACS
18 participants having less than a high school education vs 28% of the Six City cohort and not
19 reported for the VA Cohort (although the Armed Services do have enlistment standards). The
20 education differences may be associated with smoking behavior. Also, the large number of
21 interaction terms in the model may account for part of the difference.
22 The preliminary screening models used proportional hazards regression models (Miller
23 et al., 1994) to identify age, SBP, DBF, body mass index (BMI, nonlinear), age and race
24 interaction terms, and present or former smoking as baseline predictors, with one or two
25 pollution variables added. In the final model using 233 terms (of which 162 were interactions of
26 categorized SBP, DBF, and BMI variables with age), the most significant non-pollution variables
27 were SBP, DBF, BMI, and their interactions with age, smoking status, average ZIP education,
28 race, poverty, height, and a clinic-specific effect. Lipfert et al. (2000b) noted that the risk of
29 current cigarette smoking (1.43) that they found was lower than reported in other studies. The
30 most consistently positive effects were found for O3 and NO2 exposures in the immediately
31 preceding years. This study used peak O3 rather than mean O3 as in some other cohort studies.
April 2002 8-92 DRAFT-DO NOT QUOTE OR CITE
-------
1 This may account for the higher O3 and NO2 effects here. While the PM analyses considering
2 segmented (shorter) time periods gave differing results (including significantly negative mortality
3 coefficients for some PM metrics), when methods consistent with the past studies were used (i.e.,
4 many year average PM concentrations), similar results were reported, with the authors finding
5 that "(t)ne single-mortality-period responses without ecological variables are qualitatively similar
6 to what has been reported before (SO4= > PM25 > PM15)". With ecological variables included,
7 the only significant PM effect was that of TSP up to 1981 on 1976-81 mortality. It might be
8 instructive to evaluate more parsimonious regression models with fewer ecological covariates
9 and interaction terms. It is noteworthy that estimated PM effects appear to be smaller in the later
10 years of the study rather than in the earlier years. This may also be due to cohort depletion.
11
12 8.2.3.2.5 Relationship ofAHSMOG, Six Cities, ACS and VA Study Findings
13 The results of the more recent AHSMOG mortality analyses (Abbey et al., 1999; McDonnell
14 et al., 2000) are compared here with findings from the earlier Six Cities study (Dockery et al.,
15 1993), the ACS study (Pope et al., 1995), the HEI reanalyses of the latter two studies, the
16 extension of the ACS study (Pope et al., 2002), and the VA study (Lipfert et al., 2000).
17 Table 8-12 compares the estimated RR for total, cardiopulmonary, and cancer mortality among
18 the studies. The number of subjects in these studies varies greatly (8,111 subjects in the
19 Six-Cities Study; 295,223 subjects in the 50 fine particle (PM25) cities and 552,138 subjects in
20 the 151 sulfate cities of the ACS Study; 6,338 in the AHSMOG Study; 70,000 in the VA study);
21 and this may partially account for differences among their results.
22 As shown in Table 8-12, the Six Cities study found significant associations with all PM
23 indicators. In the Krewski et al. (2000) reanalysis of the ACS study data, larger associations
24 were found for both PM2 5 and PM15 (excess relative risks of 6.6% for 10 //g/m3 PM2 5 and 4% for
25 20 //g/m3 increments in annual PM15, respectively), although both associations were significant.
26 Most recently, McDonnell et al. (2000) reported evidence from the AHSMOG analyses
27 suggestive of somewhat stronger associations with fine particles than coarse particles, though the
28 associations were only reported for males and none reached statistical significance.
29 Overall, the results most recently reported for the AHSMOG study (Abbey et al., 1999;
30 McDonnell et al., 2000) do not find consistent, statistically significant associations between
31 mortality and long-term PM exposure, though the authors conclude that some evidence was
April 2002 8-93 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-12. COMPARISON OF EXCESS RELATIVE RISKS OF LONG-TERM
MORTALITY IN THE HARVARD SIX CITIES, ACS, AHSMOG, AND VA STUDIES
Study
Six City3
Six City
New4
ACS5
ACS6
New
ACS
New
ACS
New
ACS
New
ACS
Extend.7
ACS
Extend.
ACS
Extend.
AHSMOG8
AHSMOG9
AHSMOG10
VA12
PM1
PM25
PM25
PM25
PM25
PM15.,5
PM10/15
Dichot
PM10/15 SSI
PM25
1979-83
PM25
1999-000
PM2 5 Avg.
PM10/15
30+ days
PM10/15
>100
PM25
PM25
Total
Ex. RR2
13%
14%
6.6%
7.0%
0.3%
4%
2%
4.1%
5.9%
6.2%
2%
NA
9.3%u
-10.0%
Mortality
95% CI
(4.2, 23%)
(5.3, 23%)
(3.6, 9.9%)
(4.0, 10%)
(-0.9, 1.8%)
(1.0, 9%)
(-1.0,4%)
(0.8, 7.5%)
(2.0, 9.9%)
(1.6, 11%)
(-5, 9%)
NA
(-3.8,24%)
(-15,-4.6%)
Cardiopulmonary
Mortality
Ex.
RR
18%
19%
11.6%
12.0%
0.3%
7%
6%
5.9%
7.9%
9.3%
1%
14%
20%9
95% CI
(5.8, 32%)
(6.3, 33%)
(6.6, 17%)
(7.4, 17%)
(-1.5%, 2.4%)
(3, 12%)
(3, 9%)
(1.5, 10%
(2.3, 14%)
(3.3, 16%)
(-8, 10%)
(3, 26%)
(-9, 55%)
Lung Cancer Mortality
Ex. RR
18%
21%
1.2%
0.8%
-0.9%
0.4%
-0.8%
8.2%
12.7%
13.5%
174%9
NA
36%
95% CI
(-11,57%)
(-8.4,60%)
(-8,7, 12%)
(-8.7, 11%)
(-5.5%, 3.8%)
(-4.0, 5%)
(-4.4, 3%)
(1.1, 16%)
(4.1,22%)
(4.4, 23%)
(45, 415%)
NA
(-28, 157%)
'Increments are 10 ^g/m3 for PM2 5 and 20 ^g/m3 for PM10/15.
2Ex.RR (excess relative risk, percent) = 100 * (RR -1) where the RR has been converted from the
highest-to-lowest range to the standard increment A (10 or 20) by the equation.
RR = exp(log(RR for range) x A/range).
3From (Dockery et al., 1993; Krewski et al., 2000, Part II, Table 21a), original model.
4From (Krewski et al., 2000), Part II, Table 21c.
5From (Krewski et al., 2000), Part II, Table 25a.
Trom (Krewski et al., 2000), Part II, Table 25c.
7From (Pope et al., 2002).
8From (Abbey et al., 1999), pooled estimate for males and females.
9For males only; no significant excess risk for females with contributing respiratory causes.
10From (McDonnell et al., 2000), using two-pollutant (fine and coarse particle) models.
"Males only.
12Males only, exposure period 1979-81, mortality 1982-88 from Table 7 (Lipfert et al., 2000b).
April 2002
8-94
DRAFT-DO NOT QUOTE OR CITE
-------
1 suggestive of associations with fine particles. Also, the VA study (Lipfert et al., 2000) found no
2 association with PM2 5. Nevertheless, the lack of consistent findings in the AHSMOG study and
3 negative results of the VA study, do not cast doubt on the findings of the Six Cities and ACS
4 studies; both of the late studies had larger study populations, were based on measured PM data
5 (in contrast with AHSMOG PM estimates based on TSP or visibility measurements), and have
6 been validated through exhaustive reanalysis. When considering the results of these four studies,
7 including the reanalyses results for the Six Cities and ACS studies and the results of the ACS
8 study extension, it can be concluded that there is substantial evidence for a positive association
9 between long-term exposure to PM (especially fine particles) and mortality.
10 There is no obvious statistically significant relationship between PM effect sizes, gender,
11 and smoking status across these studies. The AHSMOG analyses show no significant
12 relationships between PM10 and total mortality or cardiovascular mortality for either sex, and
13 only for male lung cancer incidence and lung cancer deaths in a predominantly non-smoking
14 sample. The ACS results, in contrast, show similar and significant associations with total
15 mortality for both "never smokers" and "ever smokers", although the ACS cohort may include a
16 substantial number of long-term former smokers with much lower risk than current smokers.
17 The Six Cities study cohort shows the strongest evidence of a higher PM effect in current
18 smokers than in non-smokers, with female former smokers having a higher risk than male former
19 smokers. This study suggests that smoking status may be viewed as an "effect modifier" for
20 ambient PM, just as smoking may be a health effect modifier for ambient O3 (Cassino et al.,
21 1999).
22 When the ACS study results are compared with the AHSMOG study results for SO4=
23 (PM10_2 5 and PM10 were not considered in the ACS study, but were evaluated in ACS reanalyses
24 [Krewski et al., 2000; Pope et al, 2002]), the total mortality effect sizes per 15 //g/m3 SO4= for the
25 males in the AHSMOG population are seen to fall between the Six-Cities and the ACS effect
26 estimates for males: RR=1.28 for AHSMOG male participants; RR=1.61 for Six-Cities Study
27 male non-smokers; and RR=1.10 for never smoker males in the ACS study. The AHSMOG
28 study 95% confidence intervals encompass both of those other studies' sulfate RR's.
29
30
April 2002 8-95 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.3.3 Studies by Particulate Matter Size-Fraction and Composition
2 8.2.3.3.1 Six Cites, ACS, andAHSMOG Study Results
3 Ambient PM consists of a mixture that may vary in composition over time and from place
4 to place. This should logically affect the relative toxicity of PM indexed by mass at different
5 times or locations. Some semi-individual chronic exposure studies have investigated relative
6 roles of various PM components in contributing to observed air pollution associations with
7 mortality. However, only a limited number of the chronic exposure studies have included direct
8 measurements of chemical-specific constituents of the PM mixes indexed by mass measurements
9 used in their analyses.
10 As shown in Table 8-13, the Harvard Six-Cities study (Dockery et al., 1993) results
11 indicated that the PM2 5 and SO4= RR associations (as indicated by their respective 95% CFs and
12 t-statistics) were more consistent than those for the coarser mass components. However, the
13 effects of sulfate and non-sulfate PM2 5 are indicated to be quite similar. Acid aerosol (H+)
14 exposure was also considered by Dockery et al. (1993), but only less than one year of
15 measurements collected near the end of the follow-up period were available in most cities; so, the
16 Six-Cities results were much less conclusive for the acidic component of PM than for the other
17 PM metrics measured over many years during the study. The Six-Cities study also yielded total
18 mortality RR estimates for the reported range across those cities of PM2 5 and SO4= levels that,
19 although not statistically different, were roughly double the analogous RR's for the TSP-PM15
20 and PM15.2 5 mass components.
21 Table 8-14 presents comparative PM25 and SO4= results from the ACS study, indicating that
22 both had substantial, statistically significant effects on all-cause and cardiopulmonary mortality.
23 On the other hand, the RR for lung cancer was notably larger (and substantially more significant)
24 for SO4= than PM2 5 (not significant).
25 The most recent AHSMOG study analysis reported by Abbey et al. (1999) used PM10 as its
26 PM mass index, finding some significant associations with total and by-cause mortality, even
27 after controlling for potentially confounding factors (including other pollutants). This analysis
28 also considered SO4= as a PM index for all health outcomes studied except lung cancer, but SO4=
29 was not as strongly associated as PM10 with mortality and was not statistically significant for any
30 mortality category.
31
April 2002 8-96 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-13. COMPARISON OF ESTIMATED RELATIVE RISKS FOR
ALL-CAUSE MORTALITY IN SIX U.S. CITIES ASSOCIATED WITH
THE REPORTED INTER-CITY RANGE OF CONCENTRATIONS
OF VARIOUS PARTICULATE MATTER METRICS
PM Species
S04=
PM25 - SO4=
PM25
PM15.2.5
TSP-PM,,
Concentration Range
(^g/m3)
8.5
8.4
18.6
9.7
27.5
Relative Risk
Estimate
1.29
1.24
1.27
1.19
1.12
RR
95% CI
(1.06-1.56)
(1.16-1.32)
(1.06-1.51)
(0.91-1.55)
(0.88-1.43)
Relative Risk
t-Statistic
3.67
8.79
3.73
1.81
1.31
Source: Dockery et al. (1993); U.S. Environmental Protection Agency (1996a).
TABLE 8-14. COMPARISON OF REPORTED SO4= AND PM2 5 RELATIVE
RISKS FOR VARIOUS MORTALITY CAUSES IN THE AMERICAN
CANCER SOCIETY (ACS) STUDY
Mortality Cause
All Cause
Cardiopulmonary
Lung Cancer
SO4=
(Range = 19.9 //g/m3)
Relative
Risk
1.15
1.26
1.35
RR
95% CI
(1.09-1.22)
(1.15-1.37)
(1.11-1.66)
RR
t-Statistic
4.85
5.18
2.92
PM25
(Range = 24.5 ^g
Relative
Risk
1.17
1.31
1.03
RR
95% CI
(1.09-1.26)
(1.17-1.46)
(0.80-1.33)
/m3)
RR
t-Statistic
4.24
4.79
0.38
Source: Pope etal. (1995).
1 Also, very extensive results were reported in Lipfert et al. (2000b) for various components:
2 TSP, PM10, PM2 5, PM15_2 5, PM15, SO4=. There were no significant positive effects for any
3 exposure period concurrent or preceding the mortality period for any PM component, but there
4 was for O3.
5 Results from the Harvard Six Cities, the ACS, and the AHSMOG studies are compared in
6 Table 8-15 (for total mortality) and Table 8-16 (for cause-specific mortality). Results for the VA
7 study are not shown in Tables 8-15 and 8-16 for two reasons. First of all, the cohort is male and
April 2002 8-97 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-15. COMPARISON OF TOTAL MORTALITY RELATIVE RISK
ESTIMATES AND T-STATISTICS FOR PARTICULATE MATTER COMPONENTS
IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM10 (50 Mg/m3)
PM2 5 (25 Mg/m3)
S04= (15 Mg/m3)
Days/yr. with
PM10>100 Mg/m3
(30 days)
PMio-2.5 (25 Mg/m3)
Study
Six Cities
AHSMOG
Six Cities
ACS (50 cities)
Six Cities
ACS (151 cities)
AHSMOG
AHSMOG
Six Cities
Subgroup
All
Male Nonsmoker
Male Nonsmoker
All
Male Nonsmoker
All
Male Nonsmoker
All
Male Nonsmoker
All
Male Nonsmoker
Male Nonsmoker
Male Nonsmoker
All
Male Nonsmoker
Relative Risk
1.504a; 1.530b
1.280a
1.242
1.364a; 1.379b
1.207a
1.174
1.245
1.504a; 1.567b
1.359
1.111
1.104
1.279
1.082
1.814a; 1.560b
1.434a
t Statistic
2.94a; 3.27b
0.81a
1.616
2.94a; 3.73b
0.81a
4.35
1.96
2.94a; 3.67b
0.81a
5.107
1.586
0.960
2.183
2.94a'c 1.816b
0.81a
"Method 1 compares Portage vs. Steubenville (Table 3, Dockery et al., 1993).
bMethod 2 is based on ecologic regression models (Table 12-18, U.S. Environmental Protection Agency, 1996a).
"Method 1 not recommended for PM10_2 5 analysis, due to high concentration in Topeka.
1 largely current or former smokers (81%), thus not comparable to the total or male non-smoker
2 populations. Secondly, there is a wide variety of exposure periods and mortality periods.
3 Estimates for Six Cities parameters were calculated in two ways: (1) mortality RR for the
4 most versus least polluted city in Table 3 of Dockery et al. (1993) adjusted to standard
5 increments; and (2) ecological regression fits in Table 12-18 of U.S. Environmental Protection
6 Agency (1996a). The Six Cities study of eastern and mid-western U.S. cities suggests a strong
7 and highly significant relationship for fine particles and sulfates, a slightly weaker but still highly
April 2002
8-98
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-16. COMPARISON OF CARDIOPULMONARY MORTALITY RELATIVE
RISK ESTIMATES AND T-STATISTICS FOR PARTICULATE MATTER
COMPONENTS IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM10 (50 //g/m3)
PM25 (25 //g/m3)
S04= (15 ^g/m3)
Days/yr. with
PM10>100 (30 days)
PM10.2.5 (25 ^g/m3)
Study
Six Cities
AHSMOG
Six Cities
ACS (50 cities)
Six Cities
ACS (151 cities)
AHSMOG
AHSMOG
Six Cities
Subgroup
All
Male Nonsmoker
Male Non-CRCc
All
All
Male
Male Nonsmoker
All
All
Male
Male Nonsmoker
Male Nonsmoker
Male Non.-CRCc
Male Nonsmoker
Male Non.-CRCc
All
Relative Risk
1.744a
1.219
1.537
1.527a
1.317
1.245
1.245
1.743a
1.190
1.147
1.205
1.279
1.219
1.082
1.188
2.251s
t Statistic
2.94a
1.120
2.369
2.94a
4.699
3.061
1.466
2.94a
5.470
3.412
2.233
0.072
0.357
1.310
2.370
2.94a'b
"Method 1 compares Portage vs. Steubenville (Table 3, Dockery et al., 1993).
bMethod 1 not recommended for PM10.2.5 analysis due to high concentration in Topeka.
"Male non. - CRC = AHSMOG subjects who died of any contributing non-malignant respiratory cause.
1
2
3
4
5
significant relationship to PM10, and a marginal relationship to PM10_2 5. The ACS study looked at
a broader spatial representation of cities, and found a stronger statistically significant relationship
to PM2 5 than to sulfate (no other pollutants were examined). The AHSMOG study at California
sites (where sulfate levels are typically low) found significant effects in males for PM10
100 //g/m3 exceedances and a marginal effect of mean PM10, but no PM effects for females or
with sulfates. On balance, the overall results shown in Tables 8-15 and 8-16 suggest statistically
April 2002
8-99
DRAFT-DO NOT QUOTE OR CITE
-------
1 significant relationships between long-term exposures to PM10, PM2 5, and/or sulfates and excess
2 total and cause-specific cardiopulmonary mortality.
3 Overall, the semi-individual long-term PM exposure studies conducted to-date collectively
4 confirm earlier cross-sectional study indications that the fine mass component of PM10 (and
5 usually especially its sulfate constituent) are more strongly correlated with mortality than is the
6 coarse PM10_25 component. However, the greater precision of PM2 5 population exposure
7 measurement (both analytical and spatial) relative to PM10_2 5 makes conclusions regarding their
8 relative contributions to observed PM10-related associations less certain than if the effect of their
9 relative errors of measurement could be addressed.
10 Single-pollutant results about PM components are informative, as shown in Table 8-15 for
11 total mortality and in Table 8-16 for cardiopulmonary causes. The t-statistics are compared for
12 studies where appropriate: mean PM10, PM10_2 5, PM2 5, and sulfate for the Six Cities (Dockery
13 et al., 1993); mean PM2 5 and sulfate for ACS (Pope et al., 1995); mean PM10 and sulfate, and
14 PM10 exceedances of 100 //g/m3 for AHSMOG (Abbey et al., 1999).
15
16 8.2.3.3.2 Lipfert and Morris (2002): An Ecological Study
17 Although we have identified reasons for preferring to use prospective cohort studies to
18 assess the long-term exposure effects of particles and gases, additional useful information may
19 still be provided by ecological studies, particularly by repeated cross-sectional studies that may
20 provide another tool for examining changes in air-pollution-attributable mortality over time.
21 Lipfert and Morris (2002) carried out cross-sectional regressions for five time periods using
22 published data on mortality, air pollution, climate, and socio-demographic factors using county-
23 level data. Data were available for TSP and gaseous co-pollutants as far back as 1960 and for
24 PM25, PM15, and SO4= from the IPN. Attributable mortality at ages 45+ for 1979-1981 was
25 associated with TSP 1960-64, less strongly with TSP 1970-1974, but not with concurrent (1979-
26 1981) TSP. Attributable mortality for ages 45+ in 1979-1981 was associated with PM25 and
27 SO4= but not PM15 for 1979-1984. However, SO4= for most intervals 1960-64 up to 1979-1981
28 was associated with mortality for most ages . Concurrent SO2 (1979-1981) was associated with
29 mortality, but much less for earlier years.
30 Pollution-attributable mortality in 1989-91 was no longer significantly associated with TSP,
31 but remained significantly associated with PM2 5 and SO4= for ages 45+ for most time intervals:
April 2002 8-100 DRAFT-DO NOT QUOTE OR CITE
-------
1 1979-84, 1999, forPM25, and 1970-74, 1979-81, 1979-84 (fines), 1982-88 for SO4=. Pollution-
2 attributable mortality in 1995-1997 had little association with present or previous PM25 and
3 PM10, but a reasonably consistent and positive relationship to SO4=. There appeared to be a
4 systematic decrease in the TSP, IPN, PM2 5, and PM10 effects from the 1960s to the 1990s, and in
5 the AIRS and IPN SO4= effect over time, but an increase in the AIRS PM2 5 effect and in the NO2
6 and peak O3 effects.
7 One of the journal editors (Ayres, 2002) notes that this study uses some other ecological
8 variables that may improve the model. Two of the ecological variables, (vehicle miles of travel
9 per square mile per year by gasoline (VMTG) and diesel (VMTD) vehicles respectively in a
10 county, also used in Janssen et al., 2002) are likely to have important associations with air
11 pollution. As noted earlier, some ambient pollutants associated with fuel combustion have
12 higher concentrations near main roads, such as PM10_25 (EC if from diesel exhaust) , NO2, and
13 CO, whereas other pollutants such as O3 may have higher concentrations away from major
14 highways.
15
16 8.2.3.4 Population-Based Mortality Studies in Children
17 Older cross-sectional mortality studies suggest that the very young may represent an
18 especially susceptible sub-population for PM-related mortality. For example, Lave and Seskin
19 (1977) found mortality among those 0-14 years of age to be significantly associated with TSP.
20 More recently, Bobak and Leon (1992) studied neonatal (ages < 1 mo) and post-neonatal
21 mortality (ages 1-12 mo) in the Czech Republic and reported significant and robust associations
22 between post-neonatal mortality and PM10, even after considering other pollutants. Post-neonatal
23 respiratory mortality showed highly significant associations for all pollutants considered, but only
24 PM10 remained significant in simultaneous regressions. The exposure duration was longer than a
25 few days, but shorter than in the adult prospective cohort studies. Thus, the limited available
26 studies reviewed in the 1996 PM AQCD were highly suggestive of an association between
27 ambient PM concentrations and infant mortality, especially among post-neonatal infants.
28 More recent studies since the 1996 PM AQCD have focused specifically on ambient PM
29 relationships to (a) intrauterine mortality and morbidity and (b) early post neonatal mortality. In
30 a study by Pereira et al. (1998), of intrauterine (pre-natal) mortality during one year (1991-1992)
31 in Brazil, PM10 was not found to be a significant predictor, but involvement of CO was suggested
April 2002 8-101 DRAFT-DO NOT QUOTE OR CITE
-------
1 by an association between increased carboxyhemoglobin (COHb) in fetal blood and ambient CO
2 levels on the day of delivery measured in a separate study. Another study (Dejmek et al., 1999)
3 evaluated possible impacts of ambient PM10 and PM2 5 exposure (monitored by EPA-developed
4 VAPS methods) during pregnancy on intrauterine growth retardation (IUGR) risk in the highly
5 polluted Teplice District of Northern Bohemia in the Czech Republic during three years
6 (1993-1996). Mean levels of pollutants (PM, NO2, SO2) were calculated for each month of
7 gestation and three concentration intervals (low, medium, high) derived for each pollutant.
8 Preliminary analyses found significant associations of IUGR with SO2 and PM10 early in
9 pregnancy but not with NO2. Odds ratios for IUGR for PM10 and PM2 5 levels were determined
10 by logistic regressions for each month during gestation, after adjusting for potential confounding
11 factors (e.g., smoking, alcohol consumption during pregnancy, etc.). Definition of an IUGR birth
12 was any one for which the birth weight fell below the 10th percentile by gender and age for live
13 births in the Czech Republic (1992-93). The OR's for IUGR were significantly related to PM10
14 during the first month of gestation: that is, as compared to low PM10, the medium level PM10
15 OR = 1.47 (CI 0.99-2.16), and the high level PM10 OR = 1.85 (CI 1.29-2.66). PM25 levels were
16 highly correlated with PM10 (r = 0.98) and manifested similar patterns (OR = 1.16, CI 0.08-0.69
17 for medium PM25 level; OR = 1.68, CI 1.18-2.40 for high PM25 level). These results suggest
18 effects of PM exposures (probably including fine particles such as sulfates, acid aerosols, and
19 PAHs in the Teplice ambient mix) early in pregnancy (circa embryo implantation) on fetal
20 growth and development.
21 A recent study relating air pollution to birth weight in the metropolitan Reno, Nevada area
22 (Chen et al., 2002) examined the associations between air pollutant variables and birth weight
23 (BW) as a continuous variable and the prevalence of low birth weight (LEW, BW < 2500 gtn) as
24 a dichotomous variable. Mean daily concentrations of the pollution variables PM10, O3, and CO
25 were relatively low: 31.5 //g/m3 for PM10 (range 1 to 157), 27.2 ppb for O3 (range 2.8 to 62), and
26 1.0 ppm for CO (range 0.25 to 4.9). Ordinary least squares regression of BW on one, two, or
27 three air pollutants, and numerous covariates (e.g., age, race, education, prenatal care, maternal
28 behaviors) were included in the models. Third-trimester maternal exposure to PM10 was
29 significantly associated with an approximately 1 g reduction in BW per //g/m3 PM10, a finding
30 robust across different model specifications. Another finding was that the reduction in BW was
31 9 to 12 g for third-trimester exposures > 90th percentile PM10 (45 //g/m3). However, none of the
April 2002 8-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM odds ratios were significantly associated with increased risk of LEW. Neither ambient CO
2 nor O3 were associated were significantly associated with LEW, unlike findings for intrauterine
3 growth reduction (IUGR) in Los Angeles by Ritz and Yu (1999) and Ritz et al. (2002).
4 More consistent results indicating likely early post-natal PM exposure effects on neonatal
5 infant mortality have emerged from other new studies. Woodruff et al. (1997), for example, used
6 cross-sectional methods to evaluate possible association of post-neonatal mortality with ambient
7 PM10 pollution. This study involved an analysis of a cohort of circa 4 million infants born during
8 1989 - 1991 in 86 U.S. metropolitan statistical areas (MSAs). Data from the National Center for
9 Health Statistics-linked birth/infant death records were combined at the MSA level with PM10
10 data from EPA's Aerometric database. Infants were categorized as having high, medium, or low
11 exposures based on tertiles of PM10 averaged over the first 2 postnatal months. Relationships
12 between this early neonatal PM10 exposure and total and cause-specific post-neonatal mortality
13 rates (from 1 mo to 1 y of age) were examined using logistic regression analyses, adjusting for
14 demographic and environmental factors. Overall post-neonatal mortality rates per 1,000 live
15 births were 3.1 among infants in areas with low PM10 exposures, 3.5 among infants with medium
16 PM10 exposures, and 3.7 among highly PM exposed infants. After adjustment for covariates, the
17 odds ratio (OR) and 95% confidence intervals for total post-neonatal mortality for the high
18 versus the low exposure group was 1.10 (CI=1.04-1.16). In normal birth weight infants, high
19 PM10 exposure was associated with mortality for respiratory causes (OR = 1.40, CI=1.05-1.85)
20 and sudden infant death syndrome (OR = 1.26, CI=1.14-1.39). Among low birth weight babies,
21 high PM10 exposure was positively (but not significantly) associated with mortality from
22 respiratory causes (OR = 1.18, CI=0.86-1.61). However, other pollutants (e.g., CO) were not
23 considered as possible confounders. This study provides results consistent with some earlier
24 reports indicating that outdoor PM air pollution may be associated with increased risk of post-
25 neonatal mortality (e.g., Bobak and Leon, 1992), but lack of consideration of other air pollutants
26 as potential confounders in this new study reduces the certainty that PM is the specific causal
27 outdoor air pollutant in this case.
28 Lipfert et al. (2000c) have reported replicating the basic findings of Woodruff et al. (1997)
29 using a similar modeling approach but annual average PM10 air quality data for one year (1990)
30 instead of PM10 averaged over the first two post natal months during 1989-1991. The
31 quantitative relationship between the individual risk of infant mortality did not differ among
April 2002 8-103 DRAFT-DO NOT QUOTE OR CITE
-------
1 infant categories (by age, by birthweight, or by cause), but PM10 risks for SIDs deaths were
2 higher for babies of smoking mothers. SO4= was a strong negative predictor of SIDs mortality for
3 all age and birth weight categories. The authors (a) noted difficulties in ascribing the reported
4 PM10 and SO4= associations to effects of the PM pollutants per se versus the results possibly
5 reflecting interrelationships between the air pollution indices, a strong well-established
6 East-West gradient in U.S. SIDS cases, and/or underlying sociodemographic factors (e.g., the
7 socioeconomic or education level of parents) and (b) hypothesized that a parallel gradient in use
8 of wood burning in fireplaces or woodstoves and consequent indoor wood smoke exposure might
9 explain the observed cross-sectional study results.
10 The basic findings from Woodruff et al. (1997) also appear to be bolstered by a more recent
11 follow-up study by Bobak and Leon (1999), who conducted a matched population-based
12 case-control study covering all births registered in the Czech Republic from 1989 to 1991 that
13 were linked to death records. They used conditional logistic regression to estimate the effects of
14 suspended particles and nitrogen oxides on risk of death in the neonatal and early post-neonatal
15 period, controlling for maternal socioeconomic status and birth weight, birth length, and
16 gestational age. The effects of all pollutants were strongest in the post-neonatal period and
17 specific for respiratory causes. Only PM showed a consistent association when all pollutants
18 were entered in one model. Thus, in this study, it appears that long-term exposure to PM is the
19 air pollutant metric most strongly associated with excess post-neonatal deaths.
20 A study of changes in annual air pollution and infant mortality over time (rather than
21 spatially) in the U.S. was also recently conducted for the period 1981-1982 (Chay and
22 Greenstone, 2001a,b). These studies used sharp, differential air quality changes across sites
23 attributable to geographic variation in the effects of the 1981-1982 recession to estimate the
24 relationship between PM air pollution and infant mortality. During the narrow period of these
25 two years, there was substantial variation across counties in changes in particulate (TSP)
26 pollution and these differential pollution reductions appeared to be independent of changes in
27 numerous socioeconomic and health care factors that may be related to infant mortality. The
28 authors found that a 1 ug/m3 reduction in TSP resulted in about 4-8 fewer infant deaths per
29 100,000 live births at the county level (a 0.35-0.45 elasticity), the estimates being remarkably
30 stable across a variety of specifications. The estimated effects in this study were driven almost
31 entirely by fewer deaths occurring within one month and one day of birth (i.e., neonatal),
April 2002 8-104 DRAFT-DO NOT QUOTE OR CITE
-------
1 suggesting that fetal exposure to pollution (via the mother) may have adverse health
2 consequences. Findings of the population reductions in infant birth weight in this study provide
3 evidence consistent with the infant mortality effects found, suggestive of a causal relationship
4 between PM exposure and infant mortality.
5 The study by Loomis et al. (1999) of infant mortality in Mexico City during 1993-1995
6 adds additional interesting information pointing towards likely fine particle impacts on infant
7 mortality. That is, in Mexico City (where mean 24-h PM2 5 = 27.4 //g/m3), infant mortality was
8 found to be associated with PM2 5, NO2, and O3 in single pollutant GAM Poisson models, but
9 much less consistently with NO2 and O3 than PM2 5 in multipollutant models. The estimated
10 excess risk for PM25-related infant mortality lagged 3-5 days was 18.2% (95% CI 6.4, 30.7) per
11 25 //g/m3 PM2 5. It is not clear, however, the extent to which such a notable increased risk for
12 infant mortality might be extrapolated to U.S. situations, due to possible differences in prenatal
13 maternal or early postnatal infant nutritional status.
14
15 8.2.3.5 Salient Points Derived from Analyses of Chronic Particulate Matter Exposure
16 Mortality Effects
17 A review of the studies summarized in the previous PM AQCD (U.S. Environmental
18 Protection Agency, 1996a) indicates that past epidemiologic studies of chronic PM exposures
19 collectively indicate increases in mortality to be associated with long-term exposure to airborne
20 particles of ambient origins. The PM effect size estimates for total mortality from these studies
21 also indicate that a substantial portion of these deaths reflected cumulative PM impacts above
22 and beyond those exerted by acute exposure events.
23 The recent HEI-sponsored reanalyses of the ACS and Harvard Six-Cities studies (Krewski
24 et al., 2000) "replicated the original results, and tested those results against alternative risk
25 models and analytic approaches without substantively altering the original findings of an
26 association between indicators of paniculate matter air pollution and mortality." Several
27 questions, including the questions (1-4) posed at the outset of this Section (8.2.3) were
28 investigated by the Krewski et al. (2000) sensitivity analyses for the Six City and ACS studies
29 data sets. Key results emerging from the HEI reanalyses and other new chronic PM mortality
30 studies are as follow:
April 2002 8-105 DRAFT-DO NOT QUOTE OR CITE
-------
1 (1) A much larger number of confounding variables and effects modifiers were considered
2 in the Reanalysis Study than in the original Six City and ACS studies. The only significant air
3 pollutant other than PM2 5 and SO4 in the ACS study was SO2, which greatly decreased the PM2 5
4 and sulfate effects when included as a co-pollutant (Krewski et al., 2000, Part n, Tables 34-38).
5 A similar reduction in particle effects occurred in any multi-pollutant model with SO2. The most
6 important new effects modifier was education. The AHSMOG study suggested that other metrics
7 for air pollution, and other personal covariates such as time spent outdoors and consumption of
8 anti- oxidant vitamins, might be useful. Both individual- level covariates and ecological-level
9 covariates shown in (Krewski et al., 2000, Part n, Table 33) were evaluated.
10 (2) Specific attribution of excess long-term mortality to any specific particle component or
11 gaseous pollutant was refined in the reanalysis of the ACS study. Both PM25 and sulfate were
12 significantly associated with excess total mortality and cardiopulmonary mortality and to about
13 the same extent whether the air pollution data were mean or median long-term concentrations or
14 whether based on Original Investigator or Reanalysis Team data. The association of mortality
15 with PM15 was much smaller, though still significant, and the associations with the coarse
16 fraction (PM15.2 5) or TSP were even smaller and not significant. The lung cancer effect was
17 significant only for sulfate with the original investigator data or for new investigators with
18 regional sulfate artifact adjustment for the 1980-1981 data (Krewski et al., 2000, Part II,
19 Table 31). Associations of mortality with long-term mean concentrations of criteria gaseous
20 co-pollutants were generally non-significant except for SO2 (Krewski et al., 2000, Part n, Tables
21 32, 34-38) which was highly significant, and for cardiopulmonary disease with warm-season
22 ozone. However, the regional association of SO2 with SO4 and SO2 with PM2 5 was very high,
23 and the effects of the separate pollutants could not be distinguished. Krewski et al. (2000,
24 p. 234) concluded that, "Collectively, our reanalyses suggest that mortality may be associated
25 with more than one component of the complex mix of ambient air pollutants in urban areas of the
26 United States." In the most recent extension of the ACS study, Pope et al. (2002) confirmed the
27 strong association with SO2 but found little evidence of effects for long-term exposures to other
28 gaseous pollutants.
29 (3) The extensive temporal data on air pollution concentrations over time in the Six City
30 Study allowed the Reanalysis Team to evaluate time scales for mortality for long-term exposure
31 to a much greater extent than reported in Dockery et al. (1993). The first approach was to
April 2002 8-106 DRAFT-DO NOT QUOTE OR CITE
-------
1 estimate the log- hazard ratio as a function of follow up time using a flexible spline-function
2 model (Krewski et al., 2000, Part H, Figures 2 and 3). The results for both SO4 and PM25 suggest
3 very similar relationships, with larger risk after initial exposure decreasing to 0 after about 4 or
4 5 years, and a large increase in risk at about 10 years follow-up time.
5 The analyses of the ACS Study proceeded somewhat differently, with less temporal data
6 but many more cities. Flexible spline regression models for PM2 5 and sulfate as function of
7 estimated cumulative exposure (not defined) were very nonlinear and showed quite different
8 relationships (Krewski et al., 2000, Part II, Figures 10 and 11). The PM2 5 relationship shows the
9 mortality log-hazard ratio increasing up to about 15 //g/m3 and relatively flat above about
10 22 //g/m3, then increasing again. The sulfate relationship is almost piecewise linear, with a low
11 near- zero slope below about 11 //g/m3 and a steep increase above that concentration.
12 A third approach evaluated several time-dependent PM2 5 exposure indicators in the Six
13 City study. They are: (a) constant (at the mean) over the entire follow-up period; (b) annual
14 mean within each of the 13 years of the study; (c) city-specific mean concentration for the earliest
15 years of the study, i.e., very long-term effect; (d) exposure estimate in 2 years preceding death;
16 (e) exposure estimate in 3 to 5 years preceding death; (f) exposure estimate > 5 years preceding
17 death. The time-dependent estimates (a-e) for mortality risk are generally similar and statistically
18 significant (Krewski et al., 2000, Part II, Table 53), with RR of 1.14 to 1.19 per 24.5 //g/m3 being
19 much lower than the risk of 1.31 estimated for exposure at the constant mean for the period.
20 Thus, it is highly likely the duration and time patterns of long-term exposure affect the risk of
21 mortality, and further study of this question (along with that of mortality displacement from
22 short-term exposures) would improve estimates of life-years lost from PM exposure.
23 (4) The Reanalysis Study also advanced our understanding of the shape of the relationship
24 between mortality and PM. Again using flexible spline modeling, Krewski et al. (2000, Part n,
25 Figure 6) found a visually near-linear relationship between all-cause and cardiopulmonary
26 mortality residuals and mean sulfate concentrations, near-linear between cardiopulmonary
27 mortality and mean PM2 5, but a somewhat nonlinear relationship between all-cause mortality
28 residuals and mean PM2 5 concentrations that flattens above about 20 //g/m3. The confidence
29 bands around the fitted curves are very wide, however, neither requiring a linear relationship nor
30 precluding a nonlinear relationship if suggested by reanalyses. An investigation of the mortality
April 2002 8-107 DRAFT-DO NOT QUOTE OR CITE
-------
1 relationship for other indicators may be useful in identifying a threshold, if one exists, for chronic
2 PM exposures.
3 (5) With regard to the role of various PM constituents in the PM-mortality association, past
4 cross-sectional studies have generally found the fine particle component, as indicated either by
5 PM2 5 or sulfates, to be the PM constituent most consistently associated with mortality. While
6 relative measurement errors of various PM indicators must be further evaluated as a possible
7 source of bias in these estimate comparisons, the Six-Cities and AHSMOG prospective semi-
8 individual studies both indicate that the fine mass components of PM are more strongly
9 associated with mortality effects of chronic PM exposure than are coarse fraction indicators.
10
11
12 8.3 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE
13 This morbidity discussion is presented below in several subsections, dealing with: (a) acute
14 cardiovascular morbidity effects of ambient PM exposure; (b) effects of short-term PM exposure
15 on the incidence of respiratory and other medical visits and hospital admissions; and (c) short-
16 and long-term PM exposure effects on lung function and respiratory symptoms in asthmatics and
17 non-asthmatics.
18
19 8.3.1 Cardiovascular Effects Associated with Acute Ambient Particulate
20 Matter Exposure
21 8.3.1.1 Introduction
22 Very little information specifically addressing acute cardiovascular morbidity effects of PM
23 existed at the time of the 1996 PM AQCD. Since that time, a significantly expanded body of
24 literature has emerged, both on the ecologic relationship between ambient particles and
25 cardiovascular hospital admissions and on physiological and/or biochemical measures that have
26 been associated with PM exposures. The latter studies are particularly important in that they
27 suggest possible mechanisms.
28 This section begins with a brief summary of the conclusions that were reached in the 1996
29 PM AQCD regarding acute cardiovascular impacts of PM. Next, new studies are reviewed in the
30 two categories noted above, i.e., ecologic time series studies and individual-level studies of
31 physiological measures of cardiac function and/or biochemical measures in blood as they relate
April 2002 8-108 DRAFT-DO NOT QUOTE OR CITE
-------
1 to ambient pollution. This review is followed by discussion of several issues that are important
2 in interpreting the available data, including the identification of potentially susceptible sub-
3 populations, the roles of environmental co-factors such as weather and other air pollutants,
4 temporal lags in the relationship between exposure and outcome, and the relative importance of
5 various size-classified PM components (e.g., PM25, PM10, PM10_25).
6 Studies of cardiovascular PM effects presented in this section were identified by ongoing
7 Medline searches in conjunction with other search strategies. Specific studies were summarized
8 in text and/or tables based on criteria that include the following: (1) preference was given to
9 results reported for PM10, PM10_2 5, and PM2 5; (2) studies relating cardiovascular effects to levels
10 of ambient PM exposure in a quantitative manner are the focus of presentations; and (3) other
11 factors discussed earlier in Section 8.1 of this chapter.
12
13 8.3.1.2 Summary of Key Findings on Cardiovascular Morbidity from the 1996 Particulate
14 Matter Air quality Criteria Document
15 Just two studies were available for review in the 1996 PM AQCD that provided data on
16 acute cardiovascular morbidity outcomes (Schwartz and Morris, 1995; Burnett et al., 1995).
17 Both studies were of ecologic time series design, using standard statistical methods. Analyzing
18 four years of data on the > 65 year old Medicare population in Detroit, MI, Schwartz and Morris
19 (1995) reported significant associations between ischemic heart disease admissions and PM10,
20 controlling for environmental covariates. Based on an analysis of admissions data from
21 168 hospitals throughout Ontario, Canada, Burnett and colleagues (1995) reported significant
22 associations between fine particle sulfate concentrations, as well as other air pollutants, and daily
23 cardiovascular admissions. The relative risk due to sulfate particles was slightly larger for
24 respiratory than for cardiovascular hospital admissions. The 1996 PM AQCD concluded on the
25 basis of these studies that: "There is a suggestion of a relationship to heart disease, but the
26 results are based on only two studies, and the estimated effects are smaller than those for other
27 endpoints" (U.S. Environmental Protection Agency, 1996a p. 12-100). The PM AQCD went on
28 to state that acute impacts on CVD admissions had been demonstrated for elderly populations
29 (i.e., > 65), but that insufficient data existed to assess relative impacts on younger populations.
30 When viewed alongside the more extensive literature on acute CVD mortality that was
31 available at that time, the evidence from ecologic time series studies reviewed in the 1996 PM
April 2002 8-109 DRAFT-DO NOT QUOTE OR CITE
-------
1 AQCD was consistent with the notion that acute health risks of PM are larger for cardiovascular
2 and respiratory causes than for other causes. Given the tendency for end-stage disease states to
3 include both respiratory and cardiovascular impairment, and the associated diagnostic overlap
4 that often exists, it was not possible on the basis of these studies alone to determine which of the
5 two organ systems, if either, was more critically impacted.
6
7 8.3.1.3 New Particulate Matter-Cardiovascular Morbidity Studies
8 8.3.1.3.1 Acute Hospital Admission Studies in United States Cities
9 Numerous new studies have examined associations between daily measures of ambient PM
10 and daily hospital admissions for cardiovascular disease (see Table 8-17 and Table 8B-1 in
11 Appendix 8B). Of particular relevance are two new multi-city studies (Schwartz, 1999; Samet
12 et al., 2000a,b; Zanobetti et al., 2000a), which provide evidence substantiating significant PM
13 effects on cardiovascular-related hospital admissions and visits. Numerous other studies, carried
14 out by individual investigators in a variety of locales, present a more varied picture, especially
15 when gaseous co-pollutants have been analyzed on equal footing with PM.
16 For example, Schwartz (1999) extended the analytical approach he had used in Tucson
17 (described below) to eight more U.S. metropolitan areas, limiting analyses to a single county in
18 each location to enhance representativeness of the air pollution data. The locations analyzed
19 were: Chicago, IL; Colorado Springs, CO; New Haven, CT; Minneapolis, MN; St. Paul, MN;
20 Seattle, WA; Spokane, WA; and Tacoma, WA. Again, the analyses focused on total
21 cardiovascular (CVD) hospital admissions among persons >65 years old. In univariate
22 regressions, remarkably consistent PM10 associations with CVD admissions were found across
23 the eight locations, with a 50 //g/m3 increase in PM10 associated with 3.6 to 8.6% increases in
24 admissions. The univariate eight-county pooled PM10 effect was 5.0% (CI 3.7-6.4), similar to the
25 6.1 % effect per 50 //g/m3 observed in the previous Tucson analysis. In a bivariate model that
26 included CO, the pooled PM10 effect size diminished somewhat to 3.8% (CI 2.0-5.5) and the CO
27 association with CVD admissions was generally robust to inclusion of PM10 in the model.
28 Additional new results were based on analyses of daily CVD hospital admissions in persons
29 65 and older in relation to PM10 in 14 cities from the NMMAPS multi-city study (Samet et al.,
30 2000a,b). Cities included Birmingham, AL; Boulder, CO; Canton, OH; Chicago, IL; Colorado
31 Springs, CO; Detroit, MI; Minneapolis/ St. Paul, MN; Nashville, TN; New Haven, CT;
April 2002 8-110 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-17. SUMMARY OF STUDIES OF PM10 OR PM2 5 AND TOTAL CVD HOSPITAL VISITS
^
to
o
o
to
Reference citation,
location, etc.
Outcome Measure
Mean Paniculate
Levels (IQR) ^g/m3
Co-pollutants
Analyzed with PM
Lag Structure
Effect measures standardized to
50 Mg/m3 PM10 or 25 ^g/m3
PM * PM **
rlV12.5 •> rlV110-2.5
U.S. Results Without Co-pollutants
oo
^
O
^
-LJ
H
1
O
n
s.**
O
H
O
O
H
W
0
^\
Samet et al. (2000a,b)
14 Cities
Schwartz (1999)
8 Counties
Linn et al. (2000)
Los Angeles
Schwartz (1997)
Tucson, AZ
Gwynn et al. (2000)
Buffalo, NY
Moolgavkar (2000b)
Cook County, IL
Moolgavkar (2000b)
Los Angeles County, CA
Moolgavkar (2000b)
Maricopa County, AZ
Zanobetti et al., 2000a
Cook County, IL
Tolbertetal., (2000a)
Atlanta, GA 1993-1998
Tolbertetal., (2000a)
Atlanta, GA 1998-1999
Total CVD admiss.
> 65 yrs
Total CVD admiss. > 65 yrs
Total CVD admiss. > 30 yrs
Total CVD admiss. > 65 yrs
CVD HA
Total CVD admiss. > 65 yrs
Total CVD admiss. > 65 yrs
Total CVD admiss. > 65 yrs
Total CVD admiss. > 65 yrs
Total CVD emerg. dept.
visits, > 16 yrs
Total CVD emerg. dept.
visits, > 16 yrs
Mean 24.4-45.3
Median 23-37
45, 18
42, IQR 23
mn/max 24. 1/90.8
35, IQR 22
44, IQR 26
41, IQR 19
Median 3 3, IQR 23
30.1, 12.4
Period 1
29.1, 12.0
Period 2
none
none
none
none
none
none
none
none
none
none
none
Oday
Oday
Oday
Oday
3 day
Oday
Oday
Oday
0-1 day avg.
0-2 day avg.
0-2 day avg.
5.5% (4.7, 6.2)
5.0% (3.7, 6.4)
3. 25% (2.04, 4.47)
6.07% (1.12, 1.27)
5.7% (-3.3, 15.5)
4.2% (3.0, 5.5)
3.2% (1.2, 5.3)
4.3% (2.5, 6.1)*
-2.4% (-6.9, 2.3)
6.6% (4.9, 8.3)
-8.2%(p=0.002)
5.1% (-7.9, 19.9)
6.1% (-3.1, 16.2)*
17.6% (-4.6, 45.0)**
U.S. Results With Co-pollutants
Schwartz (1999)
8 Counties
Schwartz (1997)
Tucson, AZ
Total CVD admiss. > 65 yrs
Total CVD admiss. > 65 yrs
Median 23-37
42, IQR 23
CO
CO
Oday
Oday
3. 8% (2.0, 5.5)
5.22% (0.17, 10.54)
-------
TABLE 8-17 (cont'd). SUMMARY OF STUDIES OF PM10 OR PM2 5 AND TOTAL CVD HOSPITAL VISITS
to
o
o
to
oo
fe
H
6
o
o
H
O
O
H
W
O
O
Reference citation,
location, etc. Outcome Measure
Mean Paniculate
Levels (IQR)
Co-pollutants Lag Structure
Analyzed with PM
Effect measures standardized to
50 Mg/m3 PM10 or 25 ^g/m3
PM25*
U.S. Results With Co-pollutants (cont'd)
Moolgavkar (2000b) Total CVD admiss. > 65 yrs
Cook County, IL
Moolgavkar (2000b) Total CVD admiss. > 65 yrs
Los Angeles County, CA
35, IQR22
44, IQR 26
NO2 0 day
none 0 day
1.8% (0.4, 3.2)
-1.8% (-4.4, 0.9)
0.8% (-1.3, 2.9)*
Non-U.S. Results Without Co-pollutants
Burnett et al., (1997a) Total CVD admiss. all ages
Toronto, Canada
Stieb et al. (2000) Total CVD emerg. dept.
Saint John, Canada visits, all ages
Atkinson et al. (1999b) Total emerg. CVD admiss.
Greater London, England > 65 yrs
Prescott et al. (1998) Total CVD admiss. > 65 yrs
Edinburgh, Scotland
Wong et al. (1999) Total emerg. CVD admiss.
> 65 yrs
28, IQR 22
14.0, 9.0
28.5, 90-10 %tile
range: 30.7
20.7, 8.4
Median 45.0,
IQR 34.8
none 1-4 day avg.
none 1-3 day avg.
none 0 day
none 1-3 day avg.
none 0-2 day avg.
7.7% (0.9, 14.8)
5. 9% (1.8, 10.2) PM2 *
13.5% (5. 5, 22.0)**
29.3% (p=0.003)
14.4% (p = 0.055) PM2 5*
2.5% (-0.2, 5.3)
12.4% (4.6, 20.9)
4.1% (1.3, 6.9)
Non-U.S. Results With Co-pollutants
Burnett etal., (1997a)
Toronto, Canada
Stieb et al. (2000)
Saint John, Canada
Atkinson et al. (1999b)
Greater London, England
Prescott etal. (1998)
Edinburgh, Scotland
Wong etal. (1999)
Total CVD admiss. all ages 28, IQR 22
Total CVD emerg. dept.
visits, all ages
14.0, 9.0
Total emerg. CVD admiss. 28.5, 90-10 %tile
>65 yrs range: 30.7
Total CVD admiss. > 65 yrs 20.7, 8.4
Total emerg. CVD admiss. Median 45.0,
> 65 yrs IQR 34.8
O3, NO2, SO2, CO 1-4 day avg.
CO, H2S, NO2, O3, 1-3 day avg.
SO2, total reduced
sulfur
NO2, O3, SO2, CO 0 day
SO2, NO2, O3, CO 1-3 day avg.
NO2, O3, SO2
0-2 day avg.
-0.9% (-8.3, 7.1)
-1.1% (-7.8, 6.0)PM25*
8.1% (-1.3, 18.3)**
PM10 not significant; no
quantitative results presented
PM10 not significant; no
quantitative results presented
PM10 effect robust; no
quantitative results presented
PM10 effect robust; no
quantitative results presented
*PM2 5 entries. **PM10.2 s. All others relate to PM10.
-------
1 Pittsburgh, PA; Provo/Orem, UT; Seattle, WA; Spokane, WA; and Youngstown, OH. The range
2 of years studied encompassed 1985-1994, although this varied by city. Covariates included SO2,
3 NO2, O3, and CO; however these were not analyzed directly as regression covariates. Individual
4 cities were analyzed first by Poisson regression methods on PM10 for lags from 0 to 5 days.
5 An overall PM10 risk estimate was then computed by taking the inverse-variance weighted mean
6 of the city-specific risk estimates. The city-specific risk estimates for PM10 were also examined
7 for correlations with omitted covariates, including other pollutants. No relationship was
8 observed between city-specific risk estimates and measures of socioeconomic status, including
9 percent living in poverty, percent non-white, and percent with college educations. The overall
10 weighted mean risk estimate for PM10 was greatest for lag 0 and for the mean of lags 0-1.
11 For example, the mean risk estimate for the mean of lags 0-1 was a 6.0% increase in CVD
12 admissions per 50 //g/m3 PM10 (95% CI: 5.1 - 6.8). The mean risk was larger in a subgroup of
13 data where PM10 was less than 50 //g/m3, suggesting the lack of a threshold. A weakness of this
14 study was its failure to report multipollutant results. The authors argued that confounding by
15 co-pollutants was not present because the city-specific risk estimates did not correlate with city-
16 specific regressions of PM10 on co-pollutant levels. However, the validity of this method for
17 identifying meaningful confounding by co-pollutants at the daily time-series level has not been
18 demonstrated. Thus, it is not possible to conclude from these results alone that the observed
19 PM10 associations were independent of co-pollutants.
20 Janssen et al. (2002), in further analyses of the data set examined above by Samet et al.
21 (2000a,b), evaluated whether differences in prevalence in air conditioning (AC) and/or the
22 contribution of different sources to total PM10 emissions could partially explain the observed
23 variability in exposure-effect relations in the 14 cities. Cities were characterized and analyzed as
24 either winter or nonwinter peaking for the AC analyses. Data on the prevalence of AC from the
25 1993 American Housing Survey of the United States Census Bureau (1995) were used to
26 calculate the percentage of homes with central AC for each metropolitan area. Data on PM10
27 emissions by source category were obtained by county from the U.S. EPA emissions and air
28 quality data web site (2000). In an analysis of all 14 cities, central AC was not strongly
29 associated with PM10 coefficients. However, separate analysis for nonwinter peaking and winter
30 peaking PM10 cities yielded coefficients for CVD-related hospital admissions that decreased
31 significantly with increased percentage of central AC for both groups of cities, as shown in
April 2002 8-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 Figure 8-1 la. Another plot shown in Figure 8-1 Ib depicts the relationships between PM10
2 percent emissions from highways and CVD, showing significant positive relationships. For both
3 analyses, similar patterns were found for hospitalization for COPD and pneumonia. The authors
4 note that the stronger relationship for hospital admission rates for CVD over COPD and
5 pneumonia may relate to the 10 times higher CVD hospital admissions rate (which would result
6 in less error). However, no co-pollutant analyses were reported. The ecologic nature and limited
7 sample size also indicate the need for further study.
8 Zanobetti et al. (2000a) re-analyzed a subset of 10 cities from among the 14 evaluated by
9 Samet et al. (2000a,b). The same basic pattern of results obtained by Samet et al. (2000a,b) were
10 found, with strongest PM10 associations on lag 0 day, smaller effects on lag 1 and 2, and none at
11 longer lags. The cross-city weighted mean estimate at 0 day lag was excess risk = 5.6% (95%
12 CI 4.7, 6.4) per 50 //g/m3 PM10 increment. The 0-1 day lag average excess CVD risk = 6.2%
13 (95% CI 5.4, 7.0) per 50 //g/m3 PM10 increment. Effect size estimates increased when data were
14 restricted to days with PM10 < 50 //g/m3. As before, no evidence of gaseous (CO, O3, SO2)
15 co-pollutant modification of PM effects was seen in the second stage analyses. Again, however,
16 co-pollutants were not tested as independent explanatory variables in the regression analysis.
17 Turning to some examples of independent single-city analyses, PM10 associations with
18 CVD hospitalizations were also examined in a study by Schwartz (1997), which analyzed three
19 years of daily data for Tucson, AZ linking total CVD hospital admissions for persons > 65 years
20 old with PM10, CO, O3, and NO2. As was the above case in Chicago, only one site monitored
21 daily PM10, whereas multiple sites did so for gaseous pollutants (O3, NO2, CO). Both PM10 and
22 CO were independently (i.e., robustly) associated with CVD-related admissions, whereas O3 and
23 NO2 were not. The percent effect of a 50 //g/m3 increase in PM10 changed only slightly from 6.07
24 (CI 1.12-11.27) to 5.22 (CI 0.17 - 10.54) when CO was included in the model along with PM10.
25 Morris and Naumova (1998) reported results for PM10, as well as for O3, NO2, and SO2 in
26 an analysis of four years of congestive heart failure data among people > 65 years old in Chicago,
27 IL. As many as eight monitoring sites were available for calculating daily gaseous pollutant
28 concentrations; however, only one site in Chicago monitored daily PM10. Only same-day results
29 were presented, based on an initial exploratory analysis showing strongest effects for same-day
30 pollution exposure (i.e., lag 0). Associations between hospitalizations and PM10 were observed
31 in univariate regressions (3.9% [1.0, 6.9] per 50 //g/m3 PM10 increase), but these diminished
April 2002 8-114 DRAFT-DO NOT QUOTE OR CITE
-------
0.0025
0.0020 -
z: 0.0015 -
0.0010 -
0.0005 -
c
CD
'O
it
CD
O
O
Q
O
0.0000
New Haven, CT
* Ratio of Summer to
Winter PM10 levels.
Boulder, CO
o1.35
Pittsburgh,
.63
Colorado Springs, CO
01.75
- , »-74
Seattle WA^v Youngstown, OH
1'82 ^vs Spokane WA
"^ »1.29
Cahtoji, OH
o v
Provo UT
2.11
10 20 30 40 50
Central Air Conditioning (%)
Nashville.TN
2.09
60
70
80
0.0025
0.0020 -
i= 0.0015 H
Q)
i£
CD
Q
O
0.0010 -
0.0005 -
0.0000
Boulder, CO
Colorado Springs, CO
• Youngstown,
„,,-.,„..»,
Birmingham, AL ^ Seattle, WA
Nashville, TN
New Haven, Ct.
Chicago, IL
B
Detroit, Ml
123456
PM10 emission from highway vehicles (%)
Figure 8-11. Univariate relation between percentage of homes with central AC and
regression coefficients for (A) CVD, for cities nonwinter peaking PM10
concentrations (solid line) and winter peaking PM10 concentrations
(dashed line) and (B) univariate relation between percentage of PM10
from highway vehicles and regression coefficients for CVD. Circle area
is proportional to the inverse of the variance of the effect estimate. Lines
represent inverse variance regression equations (fixed-effects model).
Source: Adapted by EPA from Janssen et al. (2002).
April 2002
8-115
DRAFT-DO NOT QUOTE OR CITE
-------
1 somewhat in a multi-pollutant model (2.0%, [-1.4, 5.4]). Strong, robust associations were seen
2 between CO and congestive heart failure admissions. These results seem to suggest a more
3 robust association with CO than with PM10. However, the observed differences might also be
4 due in part to differential exposure misclassification for PM10 (monitored at one site) as
5 compared with CO (eight sites).
6 In one of two U.S. studies comparing multiple PM indices, Lippmann et al. (2000)
7 analyzed associations between PM10, PM2 5, or PM10_2 5 and various categories of CVD hospital
8 admissions among the elderly (65+ yr) in Detroit on 490 days in the period 1992-1994. The most
9 striking findings were notable percent excess risk for: (a) ischemic heart disease (IHD) in
10 relation to PM indices, i.e. 8.9% (0.5, 18.0) per 50 //g PM10; 10.5% (2.8, 18.9) per 25 //g/m3
11 PM10.25; and 4.3% (-1.4, 10.4) per 25 //g/m3 PM25 (all at lag 2d); and (b) heart failure, i.e. 9.7%
12 (0.2, 20.1) per 50 //g/m3 PM10; 5.2% (-3.3, 14.5) per 25 //g/m3 PM10.25; and 9.1% (2.4, 6.2) per
13 25 //g/m3 PM2 5 (the first two at lag 0 d and the latter at lag 1 d). The PM effects generally were
14 robust when co-pollutants were added to the model. As discussed earlier with regard to the
15 Lippmann et al. (2000) mortality findings, it is difficult to discern whether the observed
16 associations with coarse fraction particles (PM10_25) are independently due to such particles or
17 may possibly be attributed to the moderately correlated fine particle (PM25) fraction in Detroit.
18 Also, power was limited by the small sample size.
19 Tolbert et al. (2000a) reported very preliminary results on multiple PM indices as they
20 relate to daily hospital emergency department visits for dysrhythmias (DYS) and all CVD
21 categories for persons aged 16 yrs or older, based on analyses of data from 18 of 33 participating
22 hospitals in an ongoing study in Atlanta. During Period 1 of the study (1993-1998), PM10 from
23 the EPA AIRS database was reported to be negatively associated with CVD visits. In a
24 subsequent one-year period (Aug. 1998 - Aug. 1999), when data became available from the
25 Atlanta PM supersite, positive but non-significant associations were seen between CVD and
26 PM10 (RR of 5.1% per 50 //g/m3 PM10) and PM2.5 (RR of 6.1% per 25 //g/m3 PM2 5); and
27 significant positive associations were seen with certain fine particle components, i.e., elemental
28 carbon (p < 0.005) and organic carbon (p < 0.02), along with CO (p < 0.005). No multi-pollutant
29 results were reported. Study power was limited due to the short data record in Period 2.
30 In addition, caution applies to acceptance of the Tolbert et al. (2000a) findings until more
31 complete analyses from all participating hospitals are carried out and reported.
April 2002 8-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 In an analysis of 1992-1995 Los Angeles data, Linn et al. (2000) also found that PM10, CO,
2 and NO2 were all significantly associated with increased cardiovascular admission in single-
3 pollutant models among persons aged 30 yr and older. Associations generally appeared to be
4 stronger for CO than for PM10. No PM10 results were presented with co-pollutants in the model.
5 Lastly, Moolgavkar (2000b) analyzed PM10, CO, NO2, O3, and SO2 in relation to daily total
6 cardiovascular (CVD) and total cerebrovascular (CrD) admissions for persons aged >65 from
7 three urban counties (Cook, IL; Los Angeles, CA; Maricopa, AZ) in the period 1987-1995.
8 Consistent with most studies, in univariate regressions, PM10 (and PM2 5 in LA) was associated at
9 some lags with CVD admissions in Cook and LA counties, but not in Maricopa county.
10 However, in two-pollutant models in Cook and LA counties, the PM risk estimates diminished
11 substantially and/or were rendered non-significant, whereas co-pollutant (CO or NO2) risk
12 estimates were less affected. Results of this study suggest that gaseous pollutants, with the
13 exception of 03, were more strongly associated with CVD hospitalizations than was PM.
14 The above analyses of daily PM10 and CO in U.S. cities, overall, indicate that elevated
15 concentrations of both PM10 and CO may enhance risk of CVD-related morbidity leading to acute
16 hospitalizations. The Lippmann results appear to implicate PM25 and/or PM10_25 in increased
17 hospital admissions for some categories of CVD among the elderly.
18
19 8.3.1.3.2 Studies in Non-U.S. Cities
20 Four separate analyses of hospitalization data in Canada have been reported by Burnett and
21 coworkers since 1995 (Burnett et al., 1995, 1997a,b, 1999). A variety of locations, outcomes,
22 PM exposure metrics, and analytical approaches were used in these studies, which hinders
23 somewhat the ability to draw broad conclusions across the full group. The first (Burnett et al.,
24 1995), reviewed briefly in the 1996 PM AQCD, analyzed six years of data from 168 hospitals in
25 Ontario, CN. Cardiovascular (CVD) and respiratory hospital admissions were analyzed in
26 relation to sulfate and ozone concentrations. Sulfate lagged one day was associated with CVD
27 admissions, with a percent effect of 2.8 (CI 1.8-3.8) per 13 //g/m3 without O3 in the model and
28 3.3 (CI 1.7-4.8) with O3 included. When CVD admissions were split out into sub-categories,
29 larger associations were seen between sulfates and coronary artery disease and heart failure than
30 for cardiac dysrhythmias. Sulfate associations with total admissions were larger for the elderly
April 2002 8-117 DRAFT-DO NOT QUOTE OR CITE
-------
1 sub-population > 65 yr old (3.5% per 13 //g/m3) than for those <65 yr old (2.5% per 13 //g/m3).
2 There was little evidence for seasonal differences in sulfate associations.
3 Burnett et al. (1997c) analyzed daily congestive heart failure hospitalizations in relation to
4 carbon monoxide and other air pollutants (O3, NO2, SO2, CoH) in ten large Canadian cities as a
5 replication of an earlier U.S. study by Morris et al. (1995). The Burnett Canadian study
6 expanded upon the previous work both by its size (11 years of data for each of 10 large cities)
7 and also by including a measure of PM air pollution (coefficient of haze, CoH), whereas no PM
8 data were included in the earlier Morris et al. study. The Burnett study was restricted to the
9 population > 65 years old. The authors noted that all pollutants except O3 were correlated,
10 making it difficult to separate them statistically. CoH, CO, and NO2 measured on the same day
11 as admission (i.e., lag 0) were all strongly associated with congestive heart failure admissions in
12 univariate models. In multi-pollutant models, CO remained a strong predictor, whereas COH did
13 not (gravimetric PM measures were not evaluated).
14 The roles played by size-selected gravimetric and chemically speciated particle metrics as
15 predictors of CVD hospitalizations were explored in analysis of data from metropolitan Toronto
16 for the summers of 1992-1994 (Burnett et al., 1997a). The analysis used dichotomous sampler
17 (PM25, PM10, and PM10_25), hydrogen ion, and sulfate data collected at a central site as well as O3,
18 NO2, SO2, CO, and COH data collected at multiple sites in Toronto. Hospital admissions
19 categories included total cardiovascular (i.e., the sum of ischemic heart disease, cardiac
20 dysrhythmias, and heart failure) and total respiratory. Model specification with respect to
21 pollution lags was completely data-driven, with all lags and averaging times out to 4 days prior to
22 admission evaluated in exploratory analyses and "best" metrics chosen on the basis of maximal
23 t-statistics. The relative risks of CVD admissions were positive and generally statistically
24 significant for all pollutants analyzed in univariate regressions, but especially so for O3, NO2,
25 COH, and PM10_2 5 (i.e., regression t-statistics > 3). Associations for gaseous pollutants were
26 generally robust to inclusion of PM covariates, whereas the PM indices (aside from COH) were
27 not robust to inclusion of multiple gaseous pollutants. In particular, PM2 5 was not a robust
28 predictor of CVD admissions in multi-pollutant models: whereas an 25 //g/m3 increase in PM25
29 was associated with a 5.9% increase (t=l.8) in CVD admissions in a univariate model, the
30 percent effect was reduced to -1.1 (t=0.3) in a model that included O3, NO2, and SO2. COH, like
31 CO and NO2, is generally thought of as a measure of primary motor-vehicle emissions during the
April 2002 8-118 DRAFT-DO NOT QUOTE OR CITE
-------
1 non-heating season. The authors concluded that "particle mass and chemistry could not be
2 identified as an independent risk factor for exacerbation of cardiorespiratory diseases in this
3 study beyond that attributable to climate and gaseous air pollution."
4 Burnett et al. (1999) later reported results of a more extensive attempt to explore cause-
5 specific hospitalizations for persons of all ages in relation to a large suite of gaseous and PM air
6 pollutant measures, using 15 years of Toronto data. Cardiovascular admissions were split out
7 into separate categories for analysis: dysrhythmias, heart failure, and ischemic heart disease.
8 The analyses also examined several respiratory causes, as well as cerebrovascular and diseases of
9 the peripheral circulation (the latter categories being included because they should show PM
10 associations if one mechanism of PM action is related to increased plasma viscosity, as suggested
11 by Peters et al. (1997a). The PM metrics analyzed were PM2 5, PM10, and PM10_25 estimated from
12 daily TSP and TSP sulfate data, based on a regression analysis on dichotomous sampling data
13 that were available every sixth day during an eight-year subset of the full study period. This use
14 of estimated rather than measured PM components limits the interpretation of the PM results
15 reported here. In general, use of estimated PM exposure metrics will tend to increase exposure
16 measurement error and thereby tend to decrease effects estimates. Model specification for lags
17 was again data-driven, based on maximal t-statistics. Although some statistically significant
18 associations with one or another PM metric were found in univariate models, there were no
19 significant PM associations with any of the three CVD hospitalization outcomes in multi-
20 pollutant models. For example, whereas an 25 //g/m3 increase in estimated PM2 5 was associated
21 with a 8.05% increase (t-statistic = 6.08) in ischemic heart disease admissions in a univariate
22 analysis, the PM25 association was reduced to 2.25% (n.s.) when NO2 and SO2 were included in
23 the model. The gaseous pollutants dominated most regressions. There also were no associations
24 between PM and cerebral or peripheral vascular disease admissions.
25 The Burnett et al. studies provide some of the most extensive results for PM in conjunction
26 with multiple gaseous pollutants, but the inconsistent use of alternative PM metrics in the various
27 analyses confuses the picture somewhat. A general finding appears to be lack of robustness of
28 associations between cardiovascular outcomes and PM in multi-pollutant analyses. This was
29 seen for COH in the analysis of 10 Canadian cities (Burnett et al., 1997c), for PM2 5 and PM10 in
30 the analysis of summer data in Toronto (Burnett et al., 1997a), and for linear combinations of
31 TSP and sulfates (i.e., estimated PM25, PM10, and PM10_25) in the analysis of 15 years of data in
April 2002 8-119 DRAFT-DO NOT QUOTE OR CITE
-------
1 Toronto (Burnett et al., 1999). One exception was the association reported between CVD
2 admissions to 168 Ontario hospitals and sulfate concentrations (Burnett et al., 1995), where the
3 sulfate association was robust to the inclusion of O3. Also, although gravimetric PM variables
4 were not robust predictors in the Toronto summer analysis, COH was (Burnett et al., 1997a),
5 perhaps reflecting the impact of primary motor vehicle emissions. This contrasts, however, with
6 COH's lack of robustness in the 10-city analysis (Burnett et al., 1997c).
7 Stieb et al. studied all-age acute cardiac emergency room visits in relation to a rich set of
8 pollution covariates in Saint John, Canada for the period 1992-1996. Daily data were available
9 on PM2 5, PM10, fine fraction hydrogen and sulfate ions, COH, CO, H2S, NO2, O3, SO2, and total
10 reduced sulfur. In a multi-pollutant model, neither PM10 nor PM2 5 were significantly related to
11 total cardiac ED visits, though O3 and SO2 were.
12 Several additional non-U.S. studies, mainly in the U.K., have also been published since the
13 1996 PM AQCD. Most of these studies evaluated co-pollutant effects along with those of PM.
14 Interpretation is hindered somewhat, however, by the failure to report quantitative results for
15 PM10 in the presence of co-pollutants. In univariate models, Atkinson et al. (1999a) reported
16 significant associations of both ambient PM10 and black smoke (BS), as well as all other
17 co-pollutants, with daily admissions for total cardiovascular disease and ischemic heart disease
18 for 1992-1994 in London, UK, using standard time series regression Methods. Co-pollutants
19 included NO2, O3, SO2, and CO. PM associations were observed for persons aged < 65 yr and for
20 persons aged > 65 yr. In two-pollutant models, the associations with PM10, NO2, SO2, and CO
21 were moderated by the presence of BS in the model, but the BS association was robust to
22 co-pollutants. Interpretation is hampered somewhat by the lack of quantitative results for
23 two-pollutant models. In another U.K. study, associations with PM10, and to a lesser extent BS,
24 SO2, and CO, were reported for analyses of daily emergency hospital admissions for
25 cardiovascular diseases from 1992-1995 for Edinburgh, UK (Prescott et al., 1998).
26 No associations were observed for NO2 and ozone. Significant PM10 associations were present
27 only in persons 65 and older. The authors reported that the PM10 associations were unaffected by
28 inclusion of other pollutants; however, results were not shown. On the other hand, no
29 associations between PM10 and daily ischemic heart disease admissions were observed by
30 Wordley and colleagues (1997) in an analysis of two years of daily data from Birmingham, UK.
31 However, PM10 was associated with respiratory admissions and cardiovascular mortality during
April 2002 8-120 DRAFT-DO NOT QUOTE OR CITE
-------
1 the same study period. This inconsistency of results across causes and outcomes is difficult to
2 interpret, but may relate in part to the relatively short time series analyzed. The authors stated
3 that gaseous pollutants did not have significant associations with health outcomes independent of
4 PM, but no results were presented for models involving gaseous pollutants.
5 In eight European cities, the APHEAII (Le Tertre et al., 2002) project examined the
6 association between PM10 and hospital admissions for cardiac causes. They found a significant
7 effect of PM10 (0.5%; 0.2, 0.8) on admission for cardiac causes (all ages) and cardiac causes
8 (0.7%; 0.4, 1.0) and ischemic heart disease (0.8%; 0.3, 1.2) for people over 65 years with the
9 impact of PM10 per unit of pollution being half that found in the United States. PM10 did not
10 seem to be confounded by O3 and SO2. The PM10 effect was reduced when CO was incorporated
11 in the regression model and eliminated when controlling for NO2.
12 A study in Hong Kong by Wong et al (1999) found associations between CVD admissions
13 and PM10, SO2, NO2, and O3 in univariate models, but did not examine multi-pollutant models.
14 Ye and colleagues analyzed a 16 year record of daily emergency hospital visits for July and
15 August in Tokyo among persons age 65 and older (Ye et al., 2001). In addition to PM10, the
16 study included NO2, ozone, SO2, and CO. Models were built using an objective significance
17 criterion for variable inclusion. NO2 was the only pollutant significantly associated with angina,
18 cardiac insufficiency, and myocardial infarction hospital visits.
19
20 8.3.1.3.3 Summary and Conclusions
21 The ecologic time series studies reviewed here add substantially to the body of available
22 literature on acute CVD morbidity effects of PM and co-pollutants. Two U.S. multi-city studies
23 offer the strongest current evidence for effects of PM10 on acute CVD hospital admissions.
24 However, uncertainties regarding the possible role of co-pollutants in the larger of the two
25 studies hinders interpretation with respect to independent PM10 effects. Among single-city
26 studies carried out in the U.S. and elsewhere by a variety of investigators (see Summary
27 Table 8-17), less consistent evidence for PM effects is seen. Of particular importance is the
28 possible roles of co-pollutants (e.g., CO) as confounders of the PM effect. Among
29 13 independent studies that included gravimetrically-measured PM10 and co-pollutants, three
30 reported PM effects that appeared independent of co-pollutants (Schwartz, 1997; Lipmmann
31 et al., 2000; Prescott et al., 1998), eight reported no significant PM10 effects after inclusion of
April 2002 8-121 DRAFT-DO NOT QUOTE OR CITE
-------
1 co-pollutants (Morris and Naumova, 1998; Moolgavkar, 2000b; Tolbert et al., 2000a; Burnett
2 et al., 1997a; Steib et al., 2000; Atkinson et al., 1999b; Wordley et al. (1997); Morgan et al.,
3 1998; Ye et al., 2001), and two studies were unclear regarding independent PM effects (Linn
4 et al., 2000; Wong et al., 1999). In a recent quantitative review of published results from
5 12 studies on airborne particles and hospital admissions for cardiovascular disease, Morris
6 (2001) noted that adjustment for co-pollutants consistently reduced the PM10 effect, with
7 reductions ranging from 10 to 320% across studies. Thus, although several studies appear to
8 provide evidence for PM effects on CVD hospital admissions independent of co-pollutant
9 effects, still other studies examining co-pollutants yield results, showing PM effects in some
10 studies while not in others.
11 With respect to the question of particle size, only a handful of studies have examined the
12 relative impacts of different particle indicators (Lippmann et al., 2000, Burnett et al., 1997a,
13 Tolbert et al., 2000a, Steib et al., 2000, Moolgavkar, 2000b). Perhaps due to statistical power
14 issues, no clear picture has emerged as to the particle size fraction most associated with acute
15 CVD effects.
16 Because hospitalization can be viewed as a less severe manifestation of the same
17 pathophysiologic mechanism that may be responsible for acute mortality following PM exposure,
18 it is of interest to assess the coherence between the morbidity results reviewed here and the
19 mortality results reviewed in Section 8.2.2 (Borja-Aburto et al., 1997, 1998; Braga et al., 2001;
20 Goldberg et al., 2000; Gouveia and Fletcher, 2001; Hoek et al., 2001; Kwon et al., 2001;
21 Michelozzi et al., 1998; Morgan et al., 1998; Ponka et al., 1998; Schwartz et al., 1996a; Simpson
22 et al., 1997; Wordley et al., 1997; Zeghnoun et al., 2001; Zmirou et al., 1998). The mortality
23 studies reported significant associations between acute CVD mortality and measures of ambient
24 PM, though the PM metrics utilized and the relative risk estimates varied across studies. PM
25 measurement methods included gravimetrically analyzed filter samples (TSP, PM10, PM2 5,
26 PM10_25), beta gauge (particle attenuation of beta radiation), nephelometry (light scattering), and
27 black smoke (filter reflectance). Where tested, PM associations with acute CVD mortality
28 appeared to be generally more robust to inclusion of gaseous covariates than was the case for
29 acute hospitalization studies (Borja-Aburto et al., 1997, 1998; Morgan et al., 1998; Wordley
30 et al., 1997; Zmirou et al., 1998). One study (Goldberg et al., 2000) which examined multiple
31 alternative PM metrics reported strongest associations with PM2 5 and no associations for PM10_2 5
April 2002 8-122 DRAFT-DO NOT QUOTE OR CITE
-------
1 and hydrogen ion. Three studies (Braga et al., 2001; Goldberg et al., 2000; Hoek et al., 2001),
2 as noted in Section 8.2.2, provide data indicating that some specific cardiovascular causes of
3 mortality (such as heart failure) were more strongly associated with air pollution than total
4 cardiovascular mortality; but it was noted that ischemic heart disease (which contributes about
5 half of all CVD deaths) was the strongest contribution to the association between air pollution
6 and cardiovascular mortality. These results for acute cardiovascular mortality are qualitatively
7 consistent with those reviewed above for hospital admissions.
8 Figure 8-12 illustrates PM10 excess risk estimates for single-pollutant models derived from
9 selected U.S. studies of PM10 exposure and total cardiovascular disease (CVD) hospital
10 admissions, standardized to a 50 //g/m3 exposure to PM10. Results are shown both for studies
11 yielding pooled outcomes for multiple U.S. cities and for studies of single U.S. cities. The Samet
12 et al. (2000a) pooled cross-city results for 14 U.S. cities provides the most precise estimate for
13 relationships of U.S. ambient PM10 exposure to increased risk for CVD hospitalization. That
14 estimate, and those derived from most other studies depicted in Figure 8-6, generally appear to
15 confirm likely excess risk of CVD-related hospital admissions for U.S. cities in the range of
16 3-10% per 50 //g/m3 PM10, especially among the elderly (>65 yr). Also, other individual-city
17 results from Detroit are indicative of excess risk for ischemic heart disease and heart failure in
18 the range of approximately 4.0 to 10.0% per 25 //g/m3 of PM25 or PM10_25, as are preliminary
19 individual-city findings from Atlanta suggestive of 4.3% and 10.5% excess risk per 25 //g/m3 of
20 PM2 5 and PM10_2 5, respectively. However, the extent to which PM affects CVD hospitalization
21 risk independently of or together with other co-pollutants (such as CO), remains to be further
22 resolved.
23
24 8.3.1.3.4 Individual-Level Studies of Cardiovascular Physiology
25 New studies carried out by various groups have evaluated longitudinal associations
26 between ambient PM and physiologic measures of cardiovascular function or biochemical
27 changes in the blood that may be associated with cardiac risks. In contrast to the ecologic time-
28 series studies discussed above, these studies measure outcomes and most covariates at the
29 individual level, making it possible to draw conclusions regarding individual risks, as well as to
30 explore mechanistic hypotheses. Heterogeneity of responses across individuals, and across
31 subgroups defined on the basis of age, sex, pre-existing health status, etc., also can in principle
April 2002 8-123 DRAFT-DO NOT QUOTE OR CITE
-------
Samet et al. (2000)
14US Cities
Schwartz (1999)
8 US Cities
Moolgavkar (2000b)
Maricopa, AZ
Moolgavkar (2000b)
LA.CA
Moolgavkar (2000c)
Cook County
Linn et al. (2000)
LA.CA
Schwartz (1997)
Tucson,AZ
Tolbert et al. (2000a)
Atlanta
Morris and Naumova (1998)
Chicago
Lippmann etal.(2000) -
Total CVD
Period 1 (AIRS Data)
I « 1
CHF
HF
IHD
i » 1
Period 2 (Supersite Data)
I « 1
-15 -10 -5 0 5 10
Reconstructed Excess Risk Percentage
50 i^g/m3 Increase in
Figure 8-12. Acute cardiovascular hospitalizations and particulate matter exposure excess
risk estimates derived from selected U.S. PM10 studies. CVD =
cardiovascular disease. CHF = congestive heart failure.
1 be assessed. While exposure assessment remains largely ecologic (i.e., the entire population is
2 usually assigned the same exposure value on a given day), exposure is generally well
3 characterized in the small, spatially-clustered study populations. The recent studies fall into two
4 broad classes: those addressing cardiac rhythm or adverse events, and those addressing blood
5 characteristics. While significant uncertainty still exists regarding the interpretation of results
6 from these new studies, the varied responses that have been reported to be associated with
7 ambient PM and co-pollutants are of much interest in regard to mechanistic hypotheses
April 2002
8-124
DRAFT-DO NOT QUOTE OR CITE
-------
1 concerning pathophysiologic processes potentially underlying CVD-related mortality/morbidity
2 effects discussed in preceding sections.
3
4 Cardiac Physiology and Adverse Cardiac Events
5 Alterations in heart rate and/or rhythm have been hypothesized as possible mechanisms by
6 which ambient PM exposures may exert acute effects on human health. Decreased heart rate
7 variability, in particular, has been identified as a predictor of increased cardiovascular morbidity
8 and mortality. Several independent studies have recently reported temporal associations between
9 PM exposures and various measures of heart beat rhythm in panels of elderly subjects (Liao
10 et al., 1999; Pope et al., 1999a,b,c; Dockery et al., 1999; Peters et al., 1999a, 2000a; Gold et al.
11 2000; Creason et al., 2001). Changes in blood pressure may also reflect increases in risk (Linn
12 et al., 1999; Ibald-Mulli et al., 2001). Finally, one important new study has linked acute (2- and
13 24-h) ambient PM25 and PM10 concentrations with increased risk of myocardial infarction in
14 subsequent hours and days (Peters et al., 2001).
15 Liao and colleagues (1999) studied 26 elderly subjects (age 65-89 years; 73% female) over
16 three consecutive weeks at a retirement center in metropolitan Baltimore, 18 of whom were
17 classified as "compromised" based on previous cardiovascular conditions (e.g., hypertension).
18 Daily six-minute resting electrocardiogram (ECG) data were collected, and time intervals
19 between sequential R-R intervals recorded. A Fourier transform was applied to the R-R interval
20 data to separate its variance into two major components: low frequency (LF, 0.04-0.15 Hz) and
21 high frequency (FTP, 0.15-0.40 Hz). The standard deviation of all normal-to-normal (N-N; also
22 designated R-R) heartbeat intervals (SDNN) was computed for use as a time-domain outcome
23 variable. PM2 5 was monitored indoors by TEOM and outdoors by dichotomous sampler.
24 Outdoor PM25 levels ranged from 8.0 to 32.2 //g/m3 (mean = 16.1 //g/m3). Regression analyses
25 controlled for inter-subject differences in average variability, allowing each subject to serve as
26 his/her own control. Consistent associations were seen between decreases in all three outcome
27 variables (LF, FTP, SDNN) and increases in PM2 5 concentrations (both indoors and outdoors),
28 with associations being stronger for the 18 "compromised"subjects. No analyses of heart rate
29 were reported.
30 Creason and colleagues (2001) recently reported results of a subsequent study using similar
31 methods among 56 elderly residents of a retirement center in Baltimore County, MD. The
April 2002 8-125 DRAFT-DO NOT QUOTE OR CITE
-------
1 11 men and 45 women ranged in age from 72 to 97 years and were all Caucasian. Associations
2 between decreased HRV and ambient PM2 5 were again observed, though not significant at the
3 0.05 level and smaller in magnitude than in the previous Baltimore study. When two episodic
4 PM2 5 days with rainfall were excluded from the 24-day data set, the PM2 5 associations increased
5 in magnitude and became statistically significant. There was no evidence of larger effects among
6 subsets of subjects with compromised health status. No results were presented for other
7 pollutants besides PM2 5.
8 Pope and colleagues (1999c) reported similar findings in a panel of six elderly subjects
9 (69-89 years, 5/6 male) with histories of cardiopulmonary disease, and one 23-year old male
10 subject suffering from Crohn's disease and arrhythmias. Subjects carried Hotter monitors for up
11 to 48 hours during different weeks that varied in ambient PM10 concentrations. N-N heartbeat
12 intervals were recorded and used to calculate several measures of heart rate variability in the time
13 domain: the standard deviation of N-N intervals (SDNN), which is a broad measure of both high
14 and low frequency variations; the standard deviation of the averages of N-N intervals in all five
15 minute segments (SDANN), which is a measure of ultra-low frequency variations; and the root
16 mean squared differences between adjacent N-N intervals (r-MSSD), which is a measure of high
17 frequency variations. Daily gravimetric PM10 data obtained from three sites in the study area
18 ranged from circa 10 //g/m3 to 130 //g/m3 during the study. A simple step function in
19 concentration was observed with high levels occurring only during the first half of the 1.5 month
20 study period. Regression analysis with subject-specific intercepts was performed, with and
21 without control for daily barometric pressure and mean heart rate. Same-day, previous-day, and
22 the two-day mean of PM10 were considered. SDNN and SDANN were negatively associated with
23 both same-day and previous-day ambient PM10, and results were unaffected by inclusion of
24 covariates. Heart rate, as well as r-MSSD, were both positively, but less strongly, associated
25 with PM10. No co-pollutants were studied.
26 The Pope et al. (1999c) study discussed above was nested within a larger cohort of
27 90 subjects who participated in a study of heart rate and oxygen saturation in the Utah Valley
28 (Dockery et al., 1999; Pope et al., 1999b). The investigators hypothesized that decreases in
29 oxygen saturation might occur as a result of PM exposure, and that this could be a risk factor for
30 adverse cardiac outcomes. The study was carried out in winter months (mid November through
31 mid-March), when frequent inversions lead to fine particle episodes. PM10 levels at the three
April 2002 8-126 DRAFT-DO NOT QUOTE OR CITE
-------
1 nearest sites averaged from 35 to 43 //g/m3 during the study, with daily 24-h levels ranging from
2 5 to 147//g/m3. Two populations were studied: 52 retired Brigham Young University
3 faculty/staff and their spouses, and 38 retirement home residents. Oxygen saturation (SpO2) and
4 heart rate (HR) were measured once or twice daily by an optical sensor applied to a finger.
5 In regression analyses that controlled for inter-individual differences in mean levels, SpO2 was
6 not associated with PM10, but was highly associated with barometric pressure. In contrast, HR
7 significantly increased in association with PM10 and significantly decreased in association with
8 barometric pressure in joint regressions. Including CO in the regressions did not change these
9 basic findings. This was the first study of this type to examine the interrelationships among
10 physiologic measures (i.e., SpO2 and HR), barometric pressure, and PM10. The profound
11 physiological effects of barometric pressure noted here highlight the importance of carefully
12 controlling for barometric pressure effects in studies of cardiac physiology.
13 Gold and colleagues (2000) obtained somewhat different results in a study of heart rate
14 variability among 21 active elderly subjects, aged 53-87 yr, in a Boston residential community.
15 Resting, standing, exercising, and recovering ECG measurements were performed weekly using a
16 standardized protocol on each subject, which involved 25 min/week of continuous Hotter ECG
17 monitoring. Two time-domain measures were extracted: SDNN and r-MSSD (see above for
18 definitions). Heart rate also was analyzed as an outcome. Continuous PM10 and PM25
19 monitoring was conducted by TEOM at a site 6 km from the study site, with PM data corrected
20 for loss of semivolatile mass. Data on CO, O3, NO2, SO2, temperature and relative humidity
21 were available from nearby sites. Outcomes were regressed on PM2 5 levels in the 0-24 hour
22 period prior to ECG testing, with and without control for HR and temperature. As for the other
23 studies discussed above, declines in SDNN were associated with PM2 5 levels, in this case
24 averaged over 4 hours. These associations reached statistical significance at the 0.05 level only
25 when all testing periods (i.e., resting, standing, exercise) were combined. In contrast to the above
26 studies, both HR and r-MSSD here were negatively associated with PM2 5 levels (i.e., lower HR
27 and r-MSSD) when PM2 5 was elevated. These associations were statistically significant overall,
28 as well as for several of the individual testing periods, and were unaffected by covariate control.
29 Peters and colleagues (1999a) reported HR results from a retrospective analysis of data
30 collected as part of the MONICA study (monitoring of trends and determinants in cardiovascular
31 disease) carried out in Augsburg, Germany. Analyses focused on 2,681 men and women aged
April 2002 8-127 DRAFT-DO NOT QUOTE OR CITE
-------
1 25-64 years who had valid ECG measurements taken in winter 1984-1985 and again in winter
2 1987-1988. Ambient pollution variables included TSP, SO2, and CO. The earlier winter included
3 a 10-day episode with unusually high levels of SO2 and TSP, but not of CO. Pollution effects
4 were analyzed in two ways: dichotomously comparing the episode and non-episode periods, and
5 continuously using regression analysis. However, it is unclear from the report to what extent the
6 analyses reflect between-subject vs. within-subject effects. A statistically significant increase in
7 mean heart rate was observed during the episode period versus other periods, controlling for
8 cardiovascular risk factors and meteorology. Larger effects were observed in women. In single-
9 pollutant regression analyses, all three pollutants were associated with increased HR.
10 In another retrospective study, Peters and colleagues (2000a) examined incidence of cardiac
11 arrhythmias among 100 patients (mean age 62.2 yr; 79% male) with implanted cardioverter
12 defibrillators followed over a three year period. PM2 5 and PM10 were measured in South Boston
13 by the TEOM method, along with black carbon, O3, CO, temperature and relative humidity; SO2
14 and NO2 data were obtained from another site. The 5th percentile, mean, and 95th percentiles of
15 PM10 concentrations were 7.8, 19.3, and 37.0 //g/m3, respectively. The corresponding values for
16 PM25 were 4.6, 12.7, and 26.6 //g/m3. Logistic regression was used to analyze arrhythmia events
17 in relation to pollution variables, controlling for between-person differences, seasons, day-of-
18 week, and meteorology in two subgroups: 33 subjects with at least one arrhythmia event; and
19 6 subjects with 10 or more arrhythmia events. In the larger subgroup, only NO2 on the previous
20 day, and the mean NO2 over five days, were significantly associated with arrhythmia incidence.
21 In patients with 10 or more events, the NO2 associations were stronger. Also, some of the PM25
22 and CO lags became significant in this subgroup. These results should be interpreted cautiously
23 given the large number of statistical tests performed.
24 Linn and colleagues (1999) reported associations between both diastolic and systolic blood
25 pressure and PM10 in a panel study of 30 Los Angeles residents with severe COPD. Recently,
26 Ibald-Mulli et al. (2001) reported similar findings from a study of blood pressure among 2607
27 men and women aged 25-64 years who participated in the MONICA study in Augsburg,
28 Germany. Systolic blood pressure increased on average during an episode of elevated TSP and
29 SO2, but the effect disappeared after controlling for meteorological parameters including
30 temperature and barometric pressure. However, when TSP and SO2 were analyzed as continuous
31 variables, both were associated with elevated systolic blood pressure, controlling for
April 2002 8-128 DRAFT-DO NOT QUOTE OR CITE
-------
1 meteorological variables. In two-pollutant models, TSP was more robust than SO2. Further, the
2 TSP association was greater in the subgroups of subjects with elevated blood viscosity and heart
3 rates.
4 An exploratory study of a panel of COPD patients (Brauer et al., 2001) examined several
5 PM indicators in relation to CVD and respiratory health effects. The very low levels of ambient
6 particle (PM10 mean - 18[7] //g/m3) and low variability in these levels plus the sample size of 16
7 limit the conclusions that can be drawn. Nevertheless, for cardiovascular endpoints, single-
8 pollutant models indicated that both systolic and diastolic BP decreased with increasing
9 exposure, but this is not statistically significant. Also, the size of the ambient PM10 effect
10 estimate for AFEVj was larger than the effect estimate for ambient PM25 and personal PM2 5 but
11 not statistically significant. While the quantitative health relationships outcome results are
12 inconclusive, the results related to PM indicators is informative while requiring future research.
13 This initial effort indicates that ambient PM10 consistently had the largest effect estimates while
14 models using personal exposure measurements did not show larger or more consistently positive
15 effect estimates relative to those using ambient exposure metrics.
16 An important study by Peters et al. (2001) reported associations between onset of
17 myocardial infarction and ambient PM (either PM10 or PM2 5) in a cohort of 772 MI patients
18 studied in Boston, MA as part of the determinants of myocardial infarction onset study. Precise
19 information on the timing of the MI, obtained from patient interviews, was linked with
20 concurrent air quality data measured at a single Boston site. A case crossover design enabled
21 each Subject to serve as his/her own control. One strength of this study was its analysis of
22 multiple PM indices and co-pollutants, including real-time PM2 5, PM10, the PM10-PM2 5
23 difference, black carbon, Ozone, CO, NO2, and SO2. Only PM2 5 and PM10 were significantly
24 associated with MI risk in models adjusting for season, meteorological parameters, and day of
25 week. Both the mean PM2 5 concentration in the previous two hours and in the 24 hours lagged
26 one day were independently associated with MI, with odds ratios of 1.48 (1.09-2.02) for 25
27 ug/m3 and 1.62 (1.13-2.34) for 20 ug/m3, respectively. PM10 associations were similar. The
28 non-significant findings for other pollution metrics should be interpreted in the context of
29 potentially differing exposure misclassification errors associated with the single monitoring site.
30 The above studies present a range of intriguing findings suggesting possible effects of PM
31 on cardiac rhythm and adverse events. Four separate studies reported decreases in HR variability
April 2002 8-129 DRAFT-DO NOT QUOTE OR CITE
-------
1 associated with PM in elderly cohorts, although r-MSSD (a measure of high-frequency HR
2 variability) showed elevations with PM in one study (Pope et al., 1999a). Also, all of the studies
3 which examined HR found an association with PM; most reported positive associations, whereas
4 one (Gold et al., 2000) reported a negative relationship. However, variations in methods and
5 results across the studies argue for caution in drawing strong conclusions regarding PM effects
6 from them, especially in light of the complex intercorrelations which exist among measures of
7 cardiac physiology, meteorology, and air pollution (Dockery et al., 1999).
8
9 Viscosity and Other Blood Characteristics
10 Peters et al. (1997a) state that plasma viscosity is determined by fibrinogen and other large
11 asymmetrical plasma proteins such as immunoglobulin M and K2-macr°gl°bulin. They note that
12 in a cohort study of elderly men and women, fibrinogen concentrations were strongly related to
13 inflammatory markers such as neutrophil count and acute-phase proteins, (C-reactive protein and
14 Kj-antichymotrypsin) and to self-reported infections. Fibrinogen contributes to plasma viscosity,
15 which is a risk factor for ischemic heart disease.
16 Support for a mechanistic hypothesis, relating to enhanced blood viscosity, is suggested in
17 an analysis of plasma viscosity data collected in a population of 3256 German adults in the
18 MONICA study (Peters et al., 1997a). Each subject provided one blood sample during October
19 1984 to June 1985. An episode of unusually high air pollution concentrations occurred during a
20 13 day period while these measurements were being collected. The authors reported that, among
21 the 324 persons who provided blood during the episode, there was a statistically significant
22 elevation in plasma viscosity as compared with the 2932 persons studied at other times. The
23 odds ratio for plasma viscosity exceeding the 95th percentile was 3.6 (CI 1.6-8.1) among men
24 and 2.3 (CI 1.0-5.3) among women. Analysis of the distribution of blood viscosity data
25 suggested that these findings were driven by changes in the upper tail of the distribution rather
26 than by a general shift in mean viscosity, consistent with the likelihood of a susceptible
27 sub-population of individuals.
28 Peters et al. (2000b) reported on a prospective cohort study of a subset of male participants
29 from the above-described Augsburg, Germany MONICA study. Based on a survey conducted in
30 1984/85, a sample of 631 randomly selected men/aged 45-64 yr), free of cardiovascular disease at
31 entry, were evaluated in a 3-yr follow-up that examined relationships of air pollution to serum
April 2002 8-130 DRAFT-DO NOT QUOTE OR CITE
-------
1 C-reactive protein concentrations. C-reactive protein is a sensitive marker of inflammation,
2 tissue damage, and infections, with acute and chronic infections being related to coronary events,
3 as well as inflammation being related to systemic hypercoagulability and the onset of acute
4 ischemic syndromes. During the 1985 air pollution episode affecting Augsburg and other parts
5 of Germany, the odds of abnormal increases in serum C-reactive protein (i.e., >90th percentile of
6 pre-episode levels = 5.7 mg/L) tripled and associated increases in TSP levels of 26 //g/m3 (5-day
7 averages) were associated with an odds ratio of 1.37 (95% CI1.08-1.73) for C-reactive protein
8 levels exceeding the 90th percentile levels in two pollutant models also including SO2 levels.
9 The estimated odds ratio for a 30 //g/m3 increase in the 5-day mean for SO2 was 1.12 (95% CI
10 0.92 to 1.47; non-significant).
11 Two other recent studies also examined blood indices in relation to PM pollution (Seaton
12 et al., 1999; Prescott et al., 1999). Seaton and colleagues collected sequential blood samples (up
13 to 12) over an 18 month period in 112 subjects (all over age 60) in Belfast and Edinburgh, UK.
14 Blood samples were analyzed for hemoglobin, packed cell volumes, blood counts, fibrinogen,
15 factor Vn, interleuken 6, C-reactive protein. In a subset of 60 subjects, plasma albumin also was
16 measured. PM10 data monitored by TEOM were collected from ambient sites in each city.
17 Personal exposure estimates for the three days preceding each blood draw were derived from
18 ambient data adjusted by time-activity patterns and I/O penetration factors. No co-pollutants
19 were analyzed. Data were analyzed by analysis of covariance, controlling for city, seasons,
20 temperature, and between-subject differences. Significant changes in several of the blood indices
21 were observed in association with either ambient or estimated personal PM10 levels. All changes
22 were negative, except for C reactive protein in relation to ambient PM10, which was positive.
23 Prescott et al. (1999) also investigated factors that might increase susceptibility to adverse
24 cardiovascular events resulting from PM exposure. Using data from a cohort of 1592 subjects
25 aged 55-74 in Edinburgh, UK, baseline measurements of blood fibrinogen and blood and plasma
26 viscosity were examined as modifiers of the effects of PM (indexed by BS) on the incidence of
27 fatal and non-fatal myocardial infarction or stroke. All three blood indices were strong predictors
28 of increased cardiac event risk. However, there was no clear evidence of either a main effect of
29 BS, nor interactions between BS and blood indices.
30 Two other new studies examined air pollution associations with plasma fibrinogen.
31 Pekkanen and colleagues (2000) analyzed plasma fibrinogen data from a cross-sectional survey
April 2002 8-131 DRAFT-DO NOT QUOTE OR CITE
-------
1 of 4982 male and 2223 female office workers in relation to same-day and previous three-days
2 concentrations of PM10, black smoke, NO2, CO, SO2, and ozone. In the full analysis, NO2 and
3 CO were significantly associated with fibrinogen levels. When the analysis was restricted to the
4 summer season, NO2 and CO, as well as PM10 and black smoke, showed significant univariate
5 associations. Schwartz (2001) reported significant associations between plasma fibrinogen levels
6 and PM10 exposures in a subset of the NHANES in cohort. PM10 associations also were
7 observed for platelet and white cell counts. The PM10 associations were robust when ozone,
8 NO2, or SO2 was included. CO was not analyzed.
9 The above findings add support for some intriguing hypotheses regarding possible
10 mechanisms by which PM exposure may be linked with adverse cardiac outcomes. They are
11 especially interesting in terms of implicating both increased blood viscosity and C-reactive
12 protein, a biological marker of inflammatory responses thought to be predictive of increased risk
13 for serious cardiac events.
14
15 8.3.1.4 Issues in the Interpretation of Acute Cardiovascular Effects Studies
16 Susceptible subpopulations. Because they lack data on individual subject characteristics,
17 ecologic time series studies provide only limited information on susceptibility factors based on
18 stratified analyses. The relative impact of PM on cardiovascular (and respiratory) admissions
19 reported in ecologic time series studies are generally somewhat higher than those reported for
20 total admissions. This provides some limited support for hypothesizing that acute effects of PM
21 operate via cardiopulmonary pathways or that persons with pre-existing cardiopulmonary disease
22 have greater susceptibility to PM, or both. Although there is some data from the ecologic time
23 series studies showing larger relative impacts of PM on cardiovascular admissions in adults aged
24 >65 yr as compared with younger populations, the differences are neither striking nor consistent.
25 One recent study reported larger CVD hospitalization effects among persons with current
26 respiratory infections. The individual-level studies of cardiophysiologic function assessed above
27 generally do suggest that elderly persons with pre-existing cardiopulmonary disease are
28 susceptible to subtle changes in heart rate variability in association with PM exposures. Because
29 younger and healthier populations have not yet been assessed, it is not yet possible to say whether
30 the elderly clearly have especially increased susceptibility, but this does represent a reasonable
31 working hypothesis.
April 2002 8-132 DRAFT-DO NOT QUOTE OR CITE
-------
1 Role of other environmental factors. The ecologic time series studies published since 1996 all
2 have controlled adequately for weather influences. Thus, it is deemed unlikely that residual
3 confounding by weather accounts for the PM associations observed. With one possible
4 exception (Pope et al., 1999a), the roles of meteorological factors have not been analyzed
5 extensively as yet in the individual-level studies of cardiac function. Thus, the possibility of
6 confounding in such studies cannot yet be readily discounted. Co-pollutants have been analyzed
7 rather extensively in many of the recent time-series studies of hospital admissions and PM. In
8 some studies, PM clearly carries an independent association after controlling for gaseous co-
9 pollutants. In others, the "PM effects" are markedly reduced once co-pollutants are added to the
10 model; but this may in part be due to colinearity between PM10 and co-pollutants and/or the
11 gaseous pollutants such as CO having independent effects on cardiovascular function.
12
13 Temporal patterns of responses following PM exposure. The evidence from recent time series
14 studies of CVD admissions suggests rather strongly that PM effects tend to be maximal at lag 0,
15 with some carryover to lag 1, with little evidence for important effects beyond lag 1.
16
17 Relation of CVD effects to PM size and chemical composition attributes. Insufficient data
18 exist from the time series CVD admissions literature or from the emerging individual-level
19 studies to provide clear guidance as to which ambient PM components, defined either on the
20 basis of size or composition, determine ambient PM CVD effect potency. The epidemiologic
21 studies published to date have been constrained by the limited availability of multiple PM
22 metrics. Where multiple metrics exist, they often are highly correlated or of differential quality
23 due to differences in numbers of monitoring sites and in monitoring frequency.
24
25 PM effects on blood characteristics related to CVD events. Interesting, though limited, new
26 evidence has also been derived which is highly suggestive of associations between ambient PM
27 and increased blood viscosity, increased serum C-reactive protein, and fibrinogen (both related
28 to increased risks of serious cardiac events)
29
30
April 2002 8-133 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3.2 Effects of Short-Term Particulate Matter Exposure on the Incidence of
2 Respiratory Hospital Admissions and Medical Visits
3 8.3.2.1 Introduction
4 Among the most severe morbidity measures evaluated with regard to PM exposure are
5 hospital admissions. Hospital emergency department (ED) visits represent a related outcome that
6 has also been studied in relation to air pollution. Also doctors' visits represent a related health
7 measure that, although less studied, is relevant to those who also suffer severe health effects.
8 This latter category of pollution-affected persons can represent a large population, yet one largely
9 unevaluated due to the usual lack of centralized data regarding doctors' visits.
10 This section evaluates present knowledge regarding the epidemiologic associations of
11 ambient PM exposure with respiratory hospital admissions and medical visits. It intercompares
12 various studies examining each of the size-related PM mass exposure measures (e.g., for PM10)
13 and study results for various PM chemical components vis-a-vis their relative associations with
14 health effects, and their respective extents of coherence with PM associations exhibited across
15 related health effects measures. In the following discussion, the main focus for quantitative
16 intercomparisons is on studies and results considering PM metrics that quantitatively measure
17 mass or a specific mass constituent, i.e.,: PM10, PM10_25, PM25, sulfates (SO4=), or acidic aerosols
18 (H+). Study results for other related PM metrics (e.g., Black Smoke; BS) are also considered, but
19 only qualitatively, primarily with respect to their coherence or lack of coherence with studies
20 using mass or composition metrics measured in North America. In order to consider potentially
21 confounding effects of other co-existing pollutants, study results for various PM metrics are
22 presented both for: (1) when the PM metric is the only pollutant in the model; and, (2) the case
23 where a second pollutant (e.g., ozone) is also included. Results from models with more than two
24 pollutants included simultaneously are not used for quantitative estimates of coefficient size or
25 statistical strength, due to increased likelihood of bias and variance inflation due to multi-
26 collinearity of various pollutants (e.g., see Harris, 1975).
27 The approach taken in this section is: first, to summarize briefly results and implications of
28 the 1996 PM AQCD document regarding this topic; then the most important (pertinent for
29 present purposes) findings from newly available key studies published since the 1996 PM AQCD
30 are discussed in the text. More detailed descriptions of these and other new studies are provided
31 in tabular form in Appendix 8B.
April 2002 8-134 DRAFT-DO NOT QUOTE OR CITE
-------
1 Studies of respiratory hospital admissions and medical visits presented in this section were
2 identified by ongoing Medline searches in conjunction with other search strategies. Specific
3 studies were summarized in tables and/or text based on criteria that include the following:
4 (1) preference was given to results reported for PM10, PM10_2 5, and PM2 5 and/or smaller PM,
5 (2) studies relating respiratory hospital admissions and medical visits to levels of ambient PM
6 exposure in a quantitative manner are the focus of presentation, and (3) other factors discussed
7 earlier in Section 8.1.3 of this chapter.
8
9 8.3.2.2 Summary of Key Respiratory Hospital Admissions Findings from the 1996
10 Particulate Matter Air Quality Criteria Document
11 In the 1996 PM AQCD, it was found that both COPD and pneumonia hospitalization
12 studies showed moderate, but statistically significant, relative risks in the range of 1.06 to
13 1.25 (or 6 to 25% excess risk increment) per 50 //g/m3 PM10 increase or its equivalent. While a
14 substantial number of hospitalizations for respiratory illnesses occur in those >65 years of age,
15 there are also numerous hospitalizations for those under 65 years of age. Several of the
16 hospitalization studies restricted their analysis by age of the individuals, but did not explicitly
17 examine younger age groups. One exception noted was Pope (1991), who reported an increase in
18 hospitalization for Utah Valley children (aged 0 to 5) for monthly numbers of admissions in
19 relation to PM10 monthly averages, as opposed to daily admissions in relation to daily PM levels
20 used in other studies. Studies examining acute associations between indicators of components of
21 fine particles (e.g., BS; sulfates, SO4=; and acidic aerosols, H+) and hospital admissions were also
22 reported as finding significant relationships. While sulfates were especially predictive of
23 respiratory health effects, it was not clear whether the sulfate-related effects were attributable to
24 their acidity, to the broader effects of associated combustion-related fine particles, or to other
25 factors.
26
27 8.3.2.3 New Respiratory-Related Hospital Admissions Studies
28 New studies since 1996 have confirmed PM associations with respiratory hospital
29 admissions. These studies have examined various admissions categories, including: total
30 respiratory admissions for all ages and by age; asthma for all ages and by age; chronic obstructive
31 pulmonary disease (COPD) admissions (usually for patients > 64 yrs.), and pneumonia
April 2002 8-135 DRAFT-DO NOT QUOTE OR CITE
-------
1 admissions (for patients > 64 yrs.). Table 8B-2 in Appendix 8B summarizes important details
2 regarding the study area, study period, study population, PM indices considered and their
3 concentrations, the methods employed, study results and comments, and the "bottom-line" PM
4 index percent excess risks per standard PM increment (e.g., 50 //g/m3 for PM10) from studies
5 published since the 1996 PM AQCD.
6 The percent excess risk (ER) estimates presented in Table 8B-2 are based upon the relative
7 risks (RR's) provided by the authors, but converted into percent increments per standardized
8 increments used by the U.S. EPA to facilitate direct intercomparisons of results across studies, as
9 discussed in Section 8.1. The ER's shown in the table are for the most positively significant
10 pollutant coefficient. The maximum lag model is used here to provide an estimate of the
11 pollutant-health effects impact.
12 Among the numerous new epidemiological studies published on PM10 morbidity, many
13 evaluated effects of relatively high PM10 concentrations. However, a large number of studies did
14 evaluate associations at low PM10 concentration levels and associations have been reported by
15 several investigators between acute PM10 exposures and total respiratory-related hospital
16 admissions for numerous U.S. cities with annual mean ambient concentrations extending to
17 below 50//g/m3.
18 The NMMAPS multi-city study (Samet et al., 2000a,b) of PM10 concentrations and hospital
19 admissions by persons 65 and older in 14 U.S. cities is of particular interest. As noted in
20 Table 8-18, this study indicates PM10 effects similar to other cities, but with narrower confidence
21 bands, due to its greater power derived by combining multiple cities in the same analysis. This
22 allows significant associations to be identified, despite the fact that many of the cities considered
23 have relatively small populations and that each of the 14 cities had mean PM10 below 50 //g/m3.
24 The cities considered and their respective annual mean/daily maximum PM10 concentrations
25 (in //g/m3) are: Birmingham (34.8/124.8); Boulder (24.4/125.0); Canton (28.4/94.8); Chicago
26 (36.4/144.7); Colorado Springs (26.9/147.2); Detroit (36.8/133.6);Minneapolis/St Paul
27 (36.8/133.6); Nashville (31.6/128.0); New Haven (29.3/95.4); Pittsburgh (36.0/139.3);
28 Provo/Orem (38.9/241.0); Seattle (31.0/145.9); Spokane (45.3/605.8); and Youngstown
29 (33.1/104.0). As seen in Table 8-18, the PM10 association remained even when only those days
30 with PM10 less than 50 //g/m3 were considered. The city-specific value results ranged from -0.06
April 2002 8-136 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-18. PERCENT INCREASE IN HOSPITAL ADMISSIONS PER 10-^g/m3
INCREASE IN PM,n IN 14 U.S. CITIES
CVD COPD Pneumonia
% Increase (95% CI) %Increase (95% CI) % Increase (95% CI)
Constrained lag models (Fixed Effect
One day mean (lag 0)
Previous day mean
Two day mean
(for lag 0 and 1)
PM10 <50 Mg.m3
(two day mean)
Quadratic distributed lag
Unconstrained distributed
Fixed effects estimate
Random effects estimate
1.07
0.68
1.17
1.47
1.18
Lag
1.19
1.07
Estimates)
(0.93,
(0.54,
(1.01,
(1.18,
(0.96,
(0.97,
(0.67,
1.22)
0.81)
1.33)
1.76)
1.39)
1.41)
1.46)
1
1
1
2
2
2
2
.44
.46
.98
.63
.49
.45
.88
(1.00,
(1.03,
(1.49,
(1.71,
(1.78,
(1.75,
(0.19,
1.89)
1.88)
2.47)
3.55)
3.20)
3.17)
5.64)
1.
1.
1.
2.
1.
1
2.
57
31
98
84
68
.9
07
(1.27,
(1.03,
(1.65,
(2.21,
(1.25,
(1.46,
(0.94,
1.87)
1.58)
2.31)
3.48)
2.11)
2.34)
3.22)
Source: Samet et al. (2000a,b)
1 for Boulder to 6.43 for Detroit, with a combined result of 9.9 for fixed effects and 8.7 for random
2 effects models for 2 day mean PM10 for values less than 50 //g/m3 for CVD as an example.
3 Janssen et al. (2002) did further analyses for the Samet et al. (2000a,b) 14-city data set
4 examining the associations for the variable prevalence in AC and/or the contribution of different
5 sources to total PM10. For COPD and pneumonia, the associations were less significant, but the
6 pattern of association were similar to that for CVD as discussed in Section 8.3.1.
7 If day-to-day increases in air pollution cause rises in hospital admissions, as indicated by
8 time-series studies, then short-term removal of pollution should lower admissions. However, it
9 is rarely possible to test this hypothesis by examining a situation when pollution sources are
10 abruptly "turned off and then "turned on" again. One past case in point was a steel mill strike
11 and concomitant reductions in both PM and respiratory admissions that were experienced in Utah
12 Valley, but not in surrounding valleys with out the steel mill, as documented by Pope (1991).
13 A more broadly relevant case where this hypothesis was similarly tested was a study of air
14 quality improvements during the Atlanta Summer Olympics of 1996 (Friedman et al., 2001).
15 These improvements were compared to changes that occurred in children's hospital admissions,
16 while weather and other "natural" influences on admissions stayed unchanged from normal.
April 2002 8-137 DRAFT-DO NOT QUOTE OR CITE
-------
1 Compared to a baseline period, traffic related pollution declined, with PM10 levels declining by
2 16%, and ozone by 28% as a result of the alternative mass transportation strategy implemented to
3 reduce road traffic during the Games. At the same time, SO2, not related to traffic, actually
4 increased during the Games. Concentrations of both PM and ozone also rose noticeably after the
5 end of the Olympics. A significant reduction in asthma events was associated with ozone
6 concentration, but the PM10 association was not statistically significant. While the high
7 correlation between PM and ozone limit the ability to determine which pollutant may have
8 accounted for the reduction in asthma events, this study supports the hypothesis that reductions
9 of acute air pollution levels can provide immediate health improvements.
10 Other U.S. studies finding associations of respiratory-related hospital admissions or
11 medical visits with PM10 levels extending below 50 //g/m3 include: Schwartz (1995) in Tacoma;
12 Schwartz (1994) in Minneapolis; Schwartz et al. (1996b) in Cleveland; Sheppard et al. (1999) in
13 Seattle; Gwynn et al. (2000) in Buffalo, NY; Linn et al. (2000) in Los Angeles, Nauenberg and
14 Basu (1999) in Los Angeles; and Moolgavkar et al. (1997) in Minneapolis-St. Paul, MN, but not
15 in Birmingham, AL. The excess risk estimates appear to most consistently fall in the range of
16 5-25% per 50 //g/m3 PM10 increment, with those for asthma visits and hospital admissions
17 usually being higher than those for COPD and pneumonia hospital admissions.
18 Similar associations between increased respiratory related hospital admissions/medical
19 visits and relatively low short-term PM10 levels were also reported by various investigators for
20 several non-U.S. cities. Wordley et al. (1997), for example, reported positive and significant
21 associations between PM10 (mean = 25.6 //g/m3, max. = 131 //g/m3) and respiratory admissions in
22 Birmingham, UK; and Atkinson et al. (1999a) found significant increases in hospital admissions
23 for respiratory disease to be associated with PM10 (mean = 28.5 //g/m3) in London, UK. Hagen
24 et al. (2000) and Prescott et al. (1998) also found positive but non-significant PM10 associations
25 with hospital admissions in Drammen, Sweden (mean = 16.8 //g/m3) and Edinburgh, Scotland
26 (mean = 20.7 //g/m3), respectively. Admissions in Drammen considered relatively small
27 populations, limiting statistical power in this study. Petroeschevsky et al. (2001) examined
28 associations between outdoor air pollution and hospital admissions in Brisbane, Australia during
29 1987-1994 using a light scattering index (BSP) for fine PM. The levels of PM are quite low in
30 this city, relative to most U.S. cities. BSP was positively and significantly associated with total
31 respiratory admissions, but not for asthma.
April 2002 8-138 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3.2.3.1 Particulate Matter Mass Fractions and Composition Comparisons
2 While PM10 mass is the metric most often employed as the particle pollution index in the
3 U.S. and Canada, some new studies have begun to examine the relative roles of various PM10
4 mass fractions and chemical constituents (such as SO4=) in the PM-respiratory hospital
5 admissions association. Several new studies report significant associations of increased
6 respiratory-cause medical visits and/or hospital admissions with ambient PM2 5 and/or PM10_2 5
7 ranging to quite low concentrations. These include the Lippmann et al. (2000) study in Detroit,
8 where all PM metrics (PM10, PM2 5, PM10_2 5, H+) were positively related to pneumonia and COPD
9 admissions among the elderly (aged 65+ yr) in single pollutant models, with their RR values
10 generally remaining little changed (but with broader confidence intervals) in multipollutant
11 models including one or more gaseous pollutant (e.g., CO, O3, NO2, SO2). Excess risks for
12 pneumonia admissions in the one pollutant model were 13% (3.7, 22) and 12% (0.8, 24) per
13 25 //g/m3 of PM25 and PM10_25, respectively; those for COPD admissions were 5.5% (-4.7, 17)
14 and 9.3% (-4.4, 25) per 25 //g/m3 PM25 and PM10_25, respectively. Also of note, Moolgavkar
15 found ca. 5.0% excess risk for COPD hospital admissions among the elderly (64+ yr) in
16 Los Angeles to be significantly related to both PM2 5 and PM10_25 in one pollutant models; but the
17 magnitudes of the risk estimates dropped by more than half to non-statistically significant levels
18 in two-pollutant models including CO. In the same study, similar magnitudes of excess risk (i.e.,
19 in the range of ca. 4 to 7%) were found in one-pollutant models to be associated with PM2 5 or
20 PM10_25 for other age groups (0-19 yr; 20-64 yr) in Los Angeles, as well. Moolgavkar et al.
21 (2000) also found 5.6% (0.2, 11.3) excess risk for all-ages COPD hospital admissions per
22 25 //g/m3 PM2 5 increase in King County, WA.
23 Tolbert et al. (2000a) reported no significant associations of PM2 5 or PM10_2 5 with COPD
24 emergency department visits in Atlanta, based on data from less than half of all participating
25 hospitals and ca. 1 yr of supersite air quality data. However, more complete analyses from all
26 participating hospitals over a longer time period are required before this can be adequately
27 evaluated.
28 Gwynn et al. (2000) considered a 2.5 yr period (May 1988-Oct. 1990) in the Buffalo, NY
29 region in a time-series analysis of daily mortality and hospital admissions for total, respiratory,
30 and circulatory hospital admissions categories. Pollutants considered included: PM10, H+, SO4=
31 COH, O3, CO, SO2, and NO2. The H+ and SO4= concentrations were determined from daily PM2 5
April 2002 8-139 DRAFT-DO NOT QUOTE OR CITE
-------
1 samples not analyzed for mass (in order to avoid possible acid neutralization) using the
2 sequential acid aerosol system. Various modeling techniques were applied to control for
3 confounding of effect estimates due to seasonality, weather and day-of-week effects. They found
4 multiple significant pollutant-health effect associations, the most significant being between SO4=
5 and respiratory hospital admissions. When calculated in terms of increments employed across
6 analyses in this report, various PMRR's were: PM10RR=1.11, 95% C.I.=1.05-1.18(for
7 50 Aig/m3); H+ RR=1.06, 95% C.I.=1.03-1.09 (for 75 nmoles/m3= 3.6 Mg/m3, if as H2SO4); and
8 SO4= RR=1.08, 95% C.I.=1.04-1.12 (for 155 nmoles/m3=15 //g/m3). As in the Burnett et al.
9 (1997a) study described below, H+ yielded the highest RR per //g/m3 of concentration. These
10 various PM metric associations were not significantly affected by inclusion of gaseous
11 co-pollutants in the regression model. Thus, all PM components considered except COH were
12 found to be associated with increased hospital admissions, but H+, SO4= and O3 had the most
13 coherent associations with respiratory admissions.
14 Lumley and Heagerty (1999) illustrate the effect of reliable variance estimation on data
15 from hospital admissions for respiratory disease on King County, WA for eight years (1987-94),
16 together with air pollution and weather information. However, their weather controls were
17 relatively crude (i.e., seasonal dummy variables and linear temperature terms). This study is
18 notable for having compared sub-micron PM (PMLO) versus coarse PM10.LO and for finding
19 significant hospital admission associations only with PML0. This may suggest that the PM2 5 vs.
20 PM10 separation may not always be sufficient to differentiate submicron fine particle vs. coarse-
21 particle toxicities.
22 Asthma hospital admission studies conducted in various U.S. communities provide
23 additional important new data. Of particular note is a study by Sheppard et al. (1999) which
24 evaluated relationships between measured ambient pollutants (PM10, PM2 5, PM10_2 5, SO2, O3 and
25 CO) and non-elderly adult (<65 years of age) hospital admissions for asthma in Seattle, WA.
26 PM and CO were found to be jointly associated with asthma admissions. An estimated 4 to 5%
27 increase in the rate of asthma hospital admissions (lagged 1 day) was reported to be associated
28 with interquartile range changes in PM indices (19 //g/m3 for PM10, 11.8 //g/m3 for PM2 5, and
29 9.3 //g/m3 for PM10_25), equivalent to excess risk rates as follows: 13% (95% CI 05, 23) per
30 50 Aig/m3 for PM10; 9% (95% CI 3, 14) per 25 Aig/m3 PM2 5; 11% (95% CI 3, 20) per 25 Aig/m3
31 PM10_2 5. Also of note for the same region is the Norris et al. (1999) study showing associations
April 2002 8-140 DRAFT-DO NOT QUOTE OR CITE
-------
1 of low levels of PM2 5 (mean = 12 //g/m3) with markedly increased asthma ED, i.e., excess risk =
2 44.5% (CI 21.7, 71.4) per 25 //g/m3 PM25.
3 Burnett et al. (1997a) evaluated the role that the ambient air pollution mix, comprised of
4 gaseous pollutants and PM indexed by various physical and chemical measures, plays in
5 exacerbating daily admissions to hospitals for cardiac diseases and for respiratory diseases
6 (tracheobronchitis, chronic obstructive long disease, asthma, and pneumonia). They employed
7 daily measures of PM25 and PM10_25, aerosol chemistry (sulfates and H+), and gaseous pollutants
8 (ozone, nitrogen dioxide, sulfur dioxide, and carbon monoxide) collected in Toronto, Ontario,
9 Canada, during the summers of 1992, 1993, and 1994. Positive associations were observed for
10 all ambient air pollutants for both respiratory and cardiac diseases. Ozone was the most
11 consistently significant pollutant and least sensitive to adjustment for other gaseous and
12 particulate measures. The PM associations with the respiratory hospital admissions were
13 significant for: PM10 (RR=1.1 Ifor 50 //g/m3; CI=1.05-1.17); PM2 5 (fine) mass (RR=1.09 for
14 25 Aig/m3; CI=1.03-1.14); PM10.25 (coarse) mass (RR=1.13 for 25 //g/m3; CI=1.05-1.20); sulfate
15 levels (RR=1.11 for 155 nmoles/m3=15 //g/m3; CI=1.06-1.17); and H+ (RR=1.40 for
16 75 nmoles/m3= 3.6 //g/m3, as H2SO4; CI=1.15-1.70). After simultaneous inclusion of ozone in
17 the model, the associations with the respiratory hospital admissions remained significant for:
18 PM10(RR=1.10;CI=1.04-1.16);fmemass(RR=1.06;CI=1.01-1.12); coarse mass (RR= 1.11;
19 CI=1.04-1.19); sulfate levels (RR=1.06; CI=1.0-1.12); and H+(RR= 1.25; CI=1.03-1.53), using
20 the same increments. Of the PM metrics considered here, H+ yielded the highest RR estimate.
21 Regression models that included all recorded pollutants simultaneously (with high
22 intercorrelations among the pollutants) were also presented.
23 There have also been numerous new time-series studies examining associations between air
24 pollution and respiratory-related hospital admissions in Europe, as summarized in Appendix 8B,
25 Table 8B-2, but most of these studies relied primarily on black smoke (BS) as their PM metric.
26 BS is a particle reflectance measure that provides an indicator of particulate blackness and is
27 highly correlated with airborne carbonaceous particle concentrations (Bailey and Clayton, 1982).
28 In the U.S., Coefficient of Haze (COH) is a metric of particle transmittance that similarly most
29 directly represents a metric of particle blackness and ambient elemental carbon concentration
30 (Wolff et al., 1983) and has been found to be highly correlated with BS (r = 0.9) (Lee et al.,
31 1972). However, the relationship between airborne carbon and total mass of overall aerosol
April 2002 8-141 DRAFT-DO NOT QUOTE OR CITE
-------
1 (PM) composition varies over time and from locality to locality, so the BS-mass ratio is less
2 reliable than the BS-carbon relationship (Bailey and Clayton, 1982). This means that the BS-
3 mass relationship is likely to be very different between Europe and the U.S., largely due to
4 differences in local PM source characteristics (e.g., percentages of diesel powered motor
5 vehicles). Therefore, while these European BS-health effects studies are of qualitative use for
6 evaluating the PM-health effects associations, they are not as useful for quantitative assessment
7 of PM effects relevant to the U. S.
8 Hagan et al. (2000) compared the association of PM10 and co-pollutants with hospital
9 admissions for respiratory causes in Drammen, Norway during 1994-1997. Respiratory
10 admissions averaged only 2.2 per day; so, the power of this analysis is weaker than studies
11 looking at larger populations and longer time periods. The HEI IB Multi-city Report modeling
12 approach was employed. While a significant association was found for PM10 as a single
13 pollutant, it became non-significant in multiple pollutant models. In two pollutant models, the
14 associations and effect size of pollutants were generally diminished, and when all eight pollutants
15 were considered in the model, all pollutants became non-significant. These results are typical of
16 the problems of analyzing and interpreting the coefficients of multiple pollutant models when the
17 pollutants are even moderately inter-correlated over time. A unique aspect of this work was that
18 benzene was considered in this community strongly affected by traffic pollution. In two pollutant
19 models, benzene was most consistently still associated. The authors conclude that PM is mainly
20 an indicator of air pollution in this city and that emissions from vehicles seem most important for
21 health effects. Thompson et al. (2001) report a similar result in Belfast, Northern Ireland, where,
22 after adjusting for multiple pollutants, only the benzene level was independently associated with
23 asthma emergency department admissions.
24 The most recent European air pollution health effects analyses have mainly been conducted
25 as part of the APHEA study, which evaluated 15 European cities from 10 different countries with
26 a total population of over 25 million. All studies used a standardized data collection and analysis
27 approach, which included consideration of the same suite of air pollutants (BS, SO2, NO2, SO2,
28 and O3) and the use of time-series regression addressing: seasonal and other long-term patterns;
29 influenza epidemics; day of the week; holidays; weather; and autocorrelation (Katsouyanni et al.,
30 1996). The general coherence of the APHEA results with other results gained under different
31 conditions strengthens the argument for causality in the air pollution-health effects association.
April 2002 8-142 DRAFT-DO NOT QUOTE OR CITE
-------
1 In earlier studies, the general use of the less comparable suspended particle (SPM) measures and
2 BS as PM indicators in some of the APHEA locations and analyses lessens the quantitative
3 usefulness of such analyses in evaluating associations between PM and health effects most
4 pertinent to the U.S. situation. However, Atkinson et al. (2001) report results of PM10 analysis in
5 a study of eight APHEA cities.
6
7 8.3.2.3.2 Methodological Studies
8 One study by Lumley and Sheppard (2000) applied a simulation approach to examine the
9 effects of seasonal confounding and model selection on hospital admissions effect estimates in
10 Seattle, Washington. It was found that the bias introduced by model selection was small, but
11 could be on the same order as the estimated health impacts. This problem was the case when
12 seasonal adjustments were not accounted for in the model, and was larger when the maximum
13 lag of many was selected. However, the distributed lag nature of air pollution effects was not
14 simulated, making the tests of the maximum lag vs. real effect unrealistic. Also, in the now usual
15 case in which seasonal adjustments were included, any bias was consistently much smaller and
16 was non-significant in cases where there was no simulated PM effect. This suggests that model
17 selection bias is not a concern in the type of modeling routinely done today, and also points out
18 the need to consider statistical significance when evaluating and inter-comparing effect estimates.
19 Several studies looked at the potential influence of exposure error on pollutant impact
20 estimates. Lipfert (2000) surveyed the sources and magnitudes of such errors and concluded that
21 they can have "profound effects on the results of epidemiological studies", noting especially
22 comparisons between fine particles and less accurately measured coarse particle associations
23 with health. In a related paper, Lipfert and Wyzga (1999) consider this issue and argue against
24 the use of statistical significance for pollutant impact inter-comparisons because the distribution
25 characteristics of the variable can play a role in its strength of association. They recommend the
26 use of effect size to inter-compare pollutants, even though differing choices of a particular
27 increment for the pollutant effect estimation (e.g., IQR vs. mean vs. median vs. max-mean, etc.)
28 will usually give differing rankings across pollutants, as their relative sizes are influenced by
29 pollutant distribution, as well. The authors argue that, until uncertainties have been fully
30 explored, such as those introduced by exposure error, such epidemiological studies should only
31 be considered as suggestive of causality. Huang and Batterman (2000) also look at exposure
April 2002 8-143 DRAFT-DO NOT QUOTE OR CITE
-------
1 error, noting that most studies have not looked at the population exposure errors, and concluding
2 that, "Unless exposure levels among groups are verified, it cannot be determined whether
3 nonsignificant associations between exposure and health endpoints indicate a lack of measurable
4 health effects, or are merely a result of exposure misclassification". However, a recent study by
5 Sheppard and Damian (2000) using quasi-likelihood simulation techniques investigated the effect
6 of pure measurement error, but not spatial or within-day personal exposure variations,
7 concluding that adjustment for measurement error does not alter the conclusions from the time
8 series analyses typically reported in the literature.
9 Schwartz (2001) examined another relevant methodological and mechanistic aspect of the
10 PM-health association: the harvesting question (i.e., as to whether the associations between air
11 pollution and health effects are due to the moving up of an event [e.g., death] that would have
12 happened in a few days, anyway, or not)? Using a smoothing technique, he estimated the "net"
13 change in mortality and hospital admissions in Chicago associated with PM, after accounting for
14 any decline in events in the follow-up period, ranging from 15 to 60 days. Analyses indicated
15 that the health effect estimates stayed the same, or increased, when any harvesting effects were
16 adjusted for in the analysis. He concluded that the results are consistent with air pollution
17 increasing the size of the risk pool, and for most of the air pollution associated deaths being
18 advanced by months to years.
19 Dewanji and Moolgavkar (2000) implemented a flexible parametric model analysis in the
20 example of multiple hospital admissions for chronic respiratory disease in King County, WA,
21 that views the data on each subject as the realization of a point process, which allows
22 incorporation of subject specific covariate and the previous history of the process. In single
23 pollutant analyses, measures of PM (PM10 and PM25) and CO are associated with hospital
24 admissions. The effect of PM was stronger than that of CO in the multipollutant models. This
25 result is inconsistent with other analysis of the same data (Moolgavkar et al., 2000) which find
26 that the effect of PM becomes insignificant when CO is simultaneously considered in the
27 analysis.
28 Pollen is an atmospheric constituent that might potentially be a factor that may confound
29 PM-asthma admissions associations, if it is correlated with both PM and hospital admissions.
30 In a London study, airborne pollen did not confound the analysis of air pollution (including black
31 smoke) and daily admissions for asthma during the time period 1987-1992 (Anderson et al.,
April 2002 8-144 DRAFT-DO NOT QUOTE OR CITE
-------
1 1998). Moolgavkar et al. (2000) in a study in Seattle found that adding pollens to PM time-series
2 regressions of respiratory admissions diminished the PM effect estimates more for PM10 than that
3 for PM2 5. Confounding by pollens would require correlation between daily PM2 5 levels and
4 seasonal pollution events and weather-related specification events. Further, for a different
5 outcome measure, Delfino et al. (1996) found that pollen was not associated with asthma
6 symptoms in an asthma panel (see Section 8.3.3)
7
8 8.3.2.4 Key New Respiratory Medical Visits Studies
9 As discussed above, medical visits include both hospital emergency department (ED) visits
10 and doctors' office visits. As in the past PM AQCD's, most of the available morbidity studies
11 presented in Appendix 8B, Table 8B-3 are of ED visits and their associations with air pollution.
12 These studies collectively confirm the results provided in the previous AQCD, indicating a
13 positive and significant association between ambient PM levels and increased respiratory-related
14 hospital visits.
15 Of the medical visit and hospital admissions studies since the 1996 PM AQCD, the most
16 informative are those that evaluate health effects associations at levels below previously well-
17 implicated PM concentrations. In the case of medical visits, the Norris et al. (1999, 2000) studies
18 of asthma ED visits found significant PM-associated health effects among children in Seattle,
19 even at quite low average PM levels and even after incorporating the effects of other air
20 pollutants (study mean PM10 = 21.7 //g/m3; estimated mean PM2 5 =12 //g/m3). Tolbert et al.
21 (2000b) reported a significant PM10 association with pediatric ED visits in Atlanta where the
22 maximum PM10 concentration was 105 //g/m3. The Lipsett et al. (1997) study of winter air
23 pollution and asthma emergency visits in Santa Clara County, CA, may provide insight where
24 one of the principal sources of PM10 is residential wood combustion (RWC). Their results
25 demonstrate an association between PM and asthma concentrations. Also, Delfino et al. (1997)
26 found significant PM10 and PM2 5 associations for respiratory ED visits among older adults in
27 Montreal when mean PM10 =21.7 //g/m3 and mean PM2 5 = 12.2 Mg/m3. Medina et al. (1997)
28 reported significant associations between doctor's asthma house visits and PM13 (which would
29 have a slightly higher concentration value than PM10) in Paris when mean PM13 = 25 //g/m3 and
30 maximum daily PM13 = 95 //g/m3. Hajat et al. (1999) reported significant PM10 associations with
31 asthma doctor's visits for children and young adults in London when mean PM10 = 28.2 //g/m3
April 2002 8-145 DRAFT-DO NOT QUOTE OR CITE
-------
1 and the PM10 90th percentile was only 46.4 //g/m3. Overall, then, numerous new medical visits
2 studies indicate PM-health effects associations at lower PM2 5 and PM10 levels than demonstrated
3 previously for this health outcome.
4
5 8.3.2.4.1 Scope of Medical Visit Morbidity Effects
6 Several of these recent medical visit studies consider a new endpoint for comparison with
7 ED visits: visits in the primary care setting. In particular, key studies showing PM-health effects
8 associations for this health outcome include: the study by Medina et al. (1997) for Paris, France
9 which evaluated doctors' visits to patients in that city; the study by Hajat et al. (1999) that
10 evaluated the relationship between daily General Practice (GP) doctor consultations for asthma
11 and other lower respiratory disease (LRD) and air pollution in London, UK; the study by
12 Choudhury et al. (1997) of private asthma medical visits in Anchorage, Alaska; and the study by
13 Ostro et al. (1999b) of daily visits by young children to primary care health clinics in Santiago,
14 Chile for upper or lower respiratory symptoms.
15 While limited in number, the above studies collectively provide new insight into the fact
16 that there is a broader scope of severe morbidity associated with PM air pollution exposure than
17 previously documented. As the authors of the London study note: "There is less information
18 about the effects of air pollution in general practice consultations but, if they do exist, the public
19 health impact could be considerable because of their large numbers." Indeed, the studies of Paris
20 doctors' house calls and London doctors' GP office visits both indicate that the effects of air
21 pollution, including PM, can affect many more people than indicated by hospital admissions
22 alone.
23 These new studies also provide indications as to the quantitative nature of medical visits
24 effects, relative to those for hospital admissions. In the London case, comparing the number of
25 admissions from the authors' earlier study (Anderson et al., 1996) with those for GP visits in the
26 1999 study (Hajat et al., 1999) indicates that there are approximately 24 asthma GP visits for
27 every asthma hospital admission in that city. Also, comparing the PM10 coefficients indicates
28 that the all-ages asthma effect size for the GP visits (although not statistically different) was
29 about 30% larger than that for hospital admissions. Similarly, the number of doctors' house calls
30 for asthma approximated 45/day in Paris (based on an average 9 asthma house calls in the SOS-
31 Medocina data base, representing 20% of the total; Medina et al., [1997]), versus an average
April 2002 8-146 DRAFT-DO NOT QUOTE OR CITE
-------
1 14 asthma admissions/day (Dab et al., 1996), or a factor of 3 more doctors' house calls than
2 hospital admissions. Moreover, the RR for Paris asthma doctors' house calls was much higher
3 than asthma admissions (RR=1.18 for 25 //g/m3 BS for house calls vs. RR=1.01 per 25 //g/m3 BS
4 for hospital admissions). Thus, these new studies suggest that looking at only hospital
5 admissions and emergency hospital visit effects may greatly underestimate the overall numbers
6 of respiratory morbidity events in a population due to acute ambient PM exposure.
7
8 8.3.2.4.2 Evaluation of Factors Potentially Affecting Respiratory Medical Visit
9 Study Outcomes
10 Some newly available studies have examined certain factors that might extraneously affect
11 the outcomes of PM-medical visit studies. Stieb et al. (1998a) examined the occurrence of bias
12 and random variability in diagnostic classification of air pollution and daily cardiac respiratory
13 emergency department visits such as asthma, COPD, respiratory infection and cardiac. They
14 concluded that there was no evidence of diagnostic bias in relation to daily air pollution levels.
15 Also, Stieb et al. (1998b) reported that for a population of adults visiting an emergency
16 department with cardiac respiratory disease, fixed site sulfate monitors appear to accurately
17 reflect daily variability in average personal exposure to particulate sulfate, whereas paniculate
18 acid exposure was not as well represented by fixed site monitors. Another study investigated
19 possible confounding of respiratory visit effects due to pollens. In London, Atkinson et al.
20 (1999a) studied the association between the number of daily visits to emergency departments for
21 respiratory complaints and measures of outdoor air pollution for PM10, NO2, SO2 and CO. They
22 examined different age groups and reported the strongest association for children for visits for
23 asthma, but were unable to separate the effects of PM10 and SO2. Pollen levels did not influence
24 the results, similar to results from the asthma panel studies described below in Section 8.3.3.
25
26 8.3.2.5 Identification of Potential Susceptible Subpopulations
27 Associations between ambient PM measures and respiratory admissions have been found
28 for all age groups, but older adults and children have been indicated by a number of hospital
29 admissions studies to exhibit the most consistent PM-health effects associations in the literature.
30 As reported in this and previous PM AQCDs, numerous studies of older adults (e.g., those 65+
31 years of age) have related acute PM exposure with an increased incidence of hospital admissions
April 2002 8-147 DRAFT-DO NOT QUOTE OR CITE
-------
1 (e.g., see Anderson et al, 1998). However, only a limited number have specifically studied
2 children as a subgroup. Burnett et al. (1994) examined the differences in air pollution-hospital
3 admissions associations as a function of age in the province of Ontario, reporting that the largest
4 percentage increase in admissions was found among infants (neonatal and post-neonatal, one year
5 or less in age).
6 Considerable efforts have aimed at identifying and quantifying air pollution effects among
7 potentially especially susceptible sub-populations of the general public, especially among
8 children. Burnett et al. (200Ib) studied the association between air pollution and hospitalization
9 for acute respiratory diseases in children less than 2 years of age in Toronto, Canada during
10 1980-1994. In single pollutant analyses, PM25, PM10_25, ozone, NO2, and CO were all significant
11 predictors of young children's respiratory admissions, but only ozone and CO stayed significant
12 in 2 pollutant models, with ozone also having a robust effect estimate in co-pollutant models.
13 These effects were found to be bigger than those for older children or adults studied in a previous
14 publication (Burnett et al., 1994). Two other recent studies of children's morbidity support the
15 indication of air pollution effects among children. Pless-Mulloli et al. (2000) looked at
16 children's respiratory health and air pollution near opencast coal mining sites in a cohort of
17 nearly 5,000 children aged 1-11 in England. Mean levels were not high (mean less than
18 20 //g/m3 PM10), but statistically significant PM10 associations were found with respiratory
19 symptoms. A roughly 5 percent increase General Practitioner medical visits was also noted, but
20 the effect was not significant in this cohort. Dabaca et al. (1999) found an association between
21 levels of fine PM and emergency visits for pneumonia and other respiratory illnesses among
22 children less than 15 years of age living in the eastern part of Santiago, Chile, where the levels of
23 PM25 were very high (mean=71.3 //g/m3) during 1995-1996. The authors found it difficult to
24 separate out the effects of various pollutants, but concluded that PM (especially the fine
25 component) is associated with the risk of these respiratory illnesses. Overall, these new studies
26 support past assertions that children, and especially those less than 2 years of age, are especially
27 susceptible to the adverse health effects of air pollution.
28 Several new studies have further investigated the hypothesis that the elderly are especially
29 affected by air pollution. Zanobetti et al. (2000b) analyzed Medicare hospital admissions for
30 heart disease, COPD, and pneumonia in Chicago, IL between 1985 and 1994, finding that the
31 PM10 risk estimate was nearly doubled by the co-presence of respiratory infections, but that there
April 2002 8-148 DRAFT-DO NOT QUOTE OR CITE
-------
1 was no effect modification by sex or race. Zanobetti et al. (2000a) similarly examined PM10
2 associations with hospital admissions for heart and lung disease in ten U.S. cities, finding an
3 overall association for COPD, pneumonia, and CVD. They found that these results were not
4 significantly modified by poverty rate or minority status in this population of Medicare patients.
5 Ye et al. (2001) examined emergency transports to the hospital. Both PM10 and N02 levels were
6 significantly associated with daily hospital transports for angina, cardiac insufficiency,
7 myocardial infarction, acute and chronic bronchitis, and pneumonia. The pollutant effect sizes
8 were generally found to be greater in men than in women, except those for angina and acute
9 bronchitis, which were the same across genders. Thus, in these various studies, cardiopulmonary
10 hospital visits and admissions among the elderly were seen to be consistently associated with PM
11 levels across numerous locales in the U.S. and abroad, generally without regard to race or
12 income; but sex was sometimes an effect modifier.
13 Gwynn and Thurston (2001) examined race as a factor in the air pollution-hospital
14 admissions association. This study considered persons of all ages in New York City during
15 1988-1990, which provided a large and diverse population ideal for investigating this question.
16 Although not statistically different from each other, the various air pollutants' relative risk
17 estimates for the Hispanic non-White category in NYC were generally larger in magnitude than
18 those of the non-Hispanic White group. The greatest difference between the White and non-
19 White subgroups was observed for O3, but the same trend was found for PM10 and sulfates.
20 However, when insurance status was used as an indicator of socioeconomic/health coverage
21 status, higher RR's were indicated for the poor and working poor (i.e., those on Medicaid and the
22 uninsured) than for economically better off (i.e., the privately insured), even among the non-
23 Hispanic Whites. This result is consistent with the past analyses in California by Nauenberg and
24 Basu (1999). Thus, the within-race analyses by insurance coverage suggested that most of the
25 generally higher effects of air pollution found for minorities (i.e., Hispanics and non-Whites)
26 were actually caused by overall socioeconomic and/or health care disparities in these populations
27 vs. the generally wealthier non-Hispanic White population. This suggests that those living in
28 poverty may represent an especially affected sub-population.
29 The respiratory-related hospital admissions studies summarized in Appendix 8B, Table
30 8B-2 reveals that the PM RR's for all children (e.g., 0-18) are not usually noticeably larger than
31 those for adults, but such comparisons of RR's must adjust for differences in the baseline risks
April 2002 8-149 DRAFT-DO NOT QUOTE OR CITE
-------
1 for each group. For example, if hospital admissions per 100,000 per day for young children are
2 double the rate for adults, then they will have a pollution relative risk (RR) per //g/m3 that is half
3 that of the adults given the exact same impact on admissions/100,000///g/m3/day. Thus, it is
4 important to adjust RR's or Excess Risks (ER's) for each different age groups' baseline, but this
5 information is usually not available (especially regarding the population catchment for each age
6 group in each study).
7 One of the few indications that is notable when comparing children with other age group
8 effect estimates in Table 8B-2 is the higher excess risk estimate for infants (i.e., the group <1 yr.
9 of age) in the Gouveia and Fletcher (2000) study, an age group that has estimated risk estimate
10 roughly twice as large as for other children or adults. This is confirmatory of the excess risk
11 pattern previously found in the above-noted Burnett et al. (1994) study for respiratory-related
12 hospital admissions.
13
14 8.3.2.6 Summary of Key Findings on Acute Particulate Matter Exposure and
15 Respiratory-Related Hospital Admissions and Medical Visits
16 The results of new studies discussed above are generally consistent with and supportive of
17 findings presented in the previous PM AQCD (U.S. Environmental Protection Agency, 1996a),
18 with regard to ambient PM associations of short-term exposures with respiratory-related hospital
19 admissions/medical visits. Excess risk estimates for specific subcategories of respiratory-related
20 hospital admissions/medical visits for U.S. cities are summarized in Tables 8-19 to 8-22 and
21 graphically depicted in Figure 8-13. The excess risk estimates fall most consistently in the range
22 of 5 to 25% per 50 //g/m3 PM10 increments, with those for asthma visits and hospital admissions
23 tending to be somewhat higher than for COPD and pneumonia hospital admissions.
24 More limited new evidence substantiates increased risk of respiratory-related hospital
25 admissions due to ambient fine particles (PM2 5, PMLO, etc.) and also points towards such
26 admissions being associated with ambient coarse particles (PM10_2 5). Excess risk estimates tend
27 to fall in the range of ca. 5.0 to 15.0% per 25 //g/m3 PM25 or PM10_25 for overall respiratory
28 admissions or for COPD admissions, whereas larger estimates are found for asthma admissions
29 (ranging upwards to ca. 40 to 50% for children < 18 yr. old in one study).
30 Various new medical visits studies (including non-hospital physician visits) indicate that
31 the use of hospital admissions alone can greatly understate the total clinical morbidity effects of
April 2002 8-150 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-19. SUMMARY OF UNITED STATES PM10 RESPIRATORY HOSPITAL
ADMISSION STUDIES
Reference
Moolgavkar et al.
(1997)
Minneapolis, St. Paul
(MSP
Birmingham (BI)
BI
Gwynn et al. (2000)
Linn et al. (2000)
Schwartz et al.
(1996b)
Zanobetti and
Schwartz (2001)
Samet et al. (2000a,b)
Zanobetti et al.
(2000a)
Chen et al. (2000)
Zanobetti et al.
(2000b)
Lippman et al. (2000)
Moolgavkar et al.
(2000)
Moolgavkar (2000a)
Samet et al. (2000a,b)
Zanobetti et al.
(2000b)
Lippman et al. (2000)
Zanobetti et al.
(2000a)
Jacobs etal. (1997)
Sheppard et al. (1999)
Nauenberg and Basu
(1999)
Outcome
Measures
Respiratory
Respiratory
Respiratory
Respiratory
Respiratory
Respiratory
Respiratory
Respiratory
COPD
COPD
COPD
COPD
COPD
COPD
COPD
COPD (>64 yrs)
(median)
Pneumonia
Pneumonia
Pneumonia
Pneumonia
Asthma
Asthma
Asthma
Mean Levels
MSP PM10 34
MSP PM10 34
BIPM1043.4
BIPM1043.4
PM10
mn/max 24. 1/90.8
45.5
43
PM10 - 33 med
PM10 - 32.9
PM10 - 32.9
PM10-36.5
33.6
PM10-31
PM10 - 30.0
PM10 - 35, Chicago
PM10 -44, LA
PM10 - 41, Phoenix
PM10 - 44, LA
PM10 - 32.9
33.6
PM10-31
PM10 - 32.9
34.3
PM10-31.5
44.81
Co-Pollutants
Measured
03
03
gaseous
pollutants
CO, NO2, O3
SO3
—
SO2, O3, NO2,
CO
SO2, O3, CO
—
—
SO2, O3, NO2,
CO, H+
none
CO
CO
SO2, O3, NO2,
CO
—
SO2, O3, NO2,
CO, H+
SO2, O3, CO
O3, CO
CO, O3, SO2
03
Lag
1
1
0
0
0
0
—
—
0
1
0-1
—
0
3
3
2
2
0
2
0
2
0
1
0
1
1
0-1
—
1
0
Effect Estimate (95% CL)
8.7 (4.6, 13)
6.9(2.7, 11.3)
1.5 (-1.5, 4.6)
3.2 (-0.7, 7.2)
11% (4.0, 18)
3.3(1.7, 5)
5.8(0.5, 11.4)
w/ diabetes: 2.29 (-0.76, 5.44)
w/o diabetes: 1.50(0.42,2.6)
7.4(5.1,9.8)
7.5 (5.3, 9.8)
10.6(7.9, 13.4)
9.4(2.2, 17.1)
w/o prior RI: 8.8(3.3, 14.6)
w/ prior RI: 17.1 (-6.7, 46.9)
No Co Poll: 9.6 (-5. 1,27)
Co Poll: 1.0 (-15, 20)
No Co-Poll: 5.1(0, 10.4)
Co-Poll: 2.5 (-2.5, 7.8)
2% (-0.2, 4.3)
6.1(1.1, 11.3)
6.9 (-4.1, 19.3)
0.6 (-5. 16.7)
Two pollutant model
8.1(6.5,9.7)
6.7(5.3, 8.2)
w/o prior asthma: 11 (7.7, 14.3)
w/o prior asthma: 22.8 (5.1, 43.6)
No Co Poll: 22(8.3,36)
Co Poll: 24(8.2,43)
8.1(6.5,9.7)
6.11(NR)
13.7(5.5,22.6)
16.2(2.0,3.0)
April 2002
8-151
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-20. SUMMARY OF UNITED STATES PM2 5 RESPIRATORY HOSPITAL
ADMISSION STUDIES
Reference
Lumley and Heagerty
(1999)
Lippmann et al. (2000)
Moolgavkar et al.
(2000)
Outcome
Measures
Respiratory
COPD
COPD
Mean Levels
Mg/in3
PM1;NR
PM25-18
PM25-18.1
Two-Pollutants
Co-Pollutants
none
S02,03,N02,
CO,H+
none
CO
Lag
1
3
3
3
3
Effect Estimate
(95% CL)
5.9(1.1,11.0)
No Poll: 5.5 (-4.7, 17)
Co Poll: 2.8 (-9.2, 16)
6.4(0.9,12.1)
5.6(0.2,11.3)
Moolgavkar (2000a) COPD (>64 yrs) PM25 - 22, LA
(median)
PM2.5 - 22, LA
CO
2 5.1(0.9,9.4)
2 2.0 (-2.9, 7.1)
Two pollutant model
Lippmann et al. (2000)
Sheppard et al. (1999)
Freidman et al. (2001)
Pneumonia
Asthma
Asthma
PM2.5
PM2.5
PM25
(36.7
-18
-16.7
- 30.8 decrease)
S02,03,N02,
CO,H+
CO, O3, SO2
03
1
1
1
3d.
cu
m
No Poll: 13(3.7,22)
Co Poll: 12(1.7,23)
8.7(3.3,14.3)
1.4(0.80-2.48)
TABLE 8-21. SUMMARY OF UNITED STATES PM10 2 5 RESPIRATORY HOSPITAL
ADMISSION STUDIES
Reference
Moolgavkar (2000a)
Lippmann et al. (2000)
Sheppard et al. (1999)
Outcome
Measures
COPD
COPD
Pneumonia
Asthma
Mean
A*g
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
Levels
-12
-12
-16.2
Two-Pollutant
Co-Pollutants
—
SO2, O3, NO2, CO, H+
S02,03,N02,CO,H+
CO, O3, SO2
Lag
3
3
3
3
3
1
Effect Estimates
(95% CL)
5.1% (-0.4, 10.9)
No Poll: 9. 3 (-4.4, 25)
Co Poll: 0.3 (-14, 18)
No Poll: 9.3 (-4.4, 25)
Co Poll: 0.3 (-14, 18)
11.1 (2.8,20.1)
1 air pollution. Thus, these results support the hypothesis that considering only hospital
2 admissions and emergency hospital visit effects may greatly underestimate the numbers of
3 medical visits occurring in a population as a result of acute ambient PM exposure. Those groups
4 identified in these morbidity studies as most strongly affected by PM air pollution are older
5 adults and the very young.
April 2002 8-152 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-22. SUMMARY OF UNITED STATES PM1(1, PM? „ and PM1(1,. ASTHMA MEDICAL VISIT STUDIES
10, 2 5,
10 2 5
3.
to
o
to
H
O'
0
2;
O
H
0
0
H
W
O
O
H
W
Outcome Mean Levels Co-Pollutants
Reference Measures (/-tg/m3) Measured
PM10
Choudhury et al. (1997) Asthma PM10-41.5 Not considered
Lipsett et al. (1997) Asthma PM10 - 61.2 NO2, 03
Norris et al. (1999) Asthma PM10 - 21.7 CO, SO2, NO2
PM10
Norris et al. (2000) Asthma PM10 - Spokane 27.9 MP
PM10- Seattle 21. 5 MP
Tolbert et al. (2000b) Asthma PM10-38.9 O3
Tolbert et al. (2000a) Asthma PM10 -29.1 NO2, 03, CO, SO2
PM25
Norris et al. (1995) Asthma PM2 5 - 12.0 CO, SO2, NO2
Tolbert et al. (2000a) Asthma PM2 s - 19 A NO2, O3, CO, SO2
PM10.2.5
Tolbert et al. (2000a) Asthma PM10.2 5 - 9.39 NO2, O3, CO, SO2
*SP = single pollutant model; MP = multipollutant model.
Lag
0
2
1
1
3
3
1
1
0-2
1
1
0-2
0-2
Effect Estimate
(95% CL)
20.9(11.8,30.8)
34.7(16, 56.5) at 20 °C
SP 75. 9 (25.1, 147.4)
MP 75. 9 (16.3, 166)
2.4 (-10.9, 17.6)
56.2(10.4, 121.1)
SP 13.2 (1.2, 26.7)
MP 8.2 (-7.1, 26.1)
18.8 (-8.7, 54.4)
SP 44.5 (21.7, 71.4)
MP 51.2 (23.4, 85.2)
2.3 (-14.8, 22.7)
21.1 (-18.2, 79.3)
-------
Tolbert et al. (2000) Atlanta -
Morris et al. (2000) Seattle -
Morris et al. (2000) Spokane -
Morris etal. (1999) Seattle -
Choudhury et al. (1997) Anchorage -
Nauenberg and Basu (1999) LA.CA -
Sheppard etal. (1999) Seattle -
Zanobetti et al. (2000a) Chicago -
Samet et al. (2000a) 14 US Cities -
Moolgavkar (2000b) Phoenix -
Moolgavkar (2000b) LA,CA -
Moolgavkar (2000b) Chicago -
Moolgavkar et al. (2000) King C -
Moolgavkar et al. (1997) Minn-SP -
Moolgavkar et al. (1997) Birm. -
Chen et al. (2000) Reno.NV -
Zanobetti et al. (2000a) Chicago -
Samet etal. (2000a) 14 US Cities -
_
h
Asthma Visits
1 * 1
Asthma Hospital Admissions
^
m
i « i
+i COPD Hospital Admissions
l-»-H
«H
1 A 1
w Pneumonia Hospital Admissions
-25
25 50 75 100
Excess Risk, %
125
150
Figure 8-13. Maximum excess risk of respiratory-related hospital admissions and visits
per 50-jUg/m3 PM10 increment in selected studies of U.S. cities.
1 8.3.3 Effects of Particulate Matter Exposure on Lung Function and
2 Respiratory Symptoms
3 In the 1996 PM AQCD, the available respiratory disease studies used a wide variety of
4 designs examining pulmonary function and respiratory symptoms in relation to PM10. The
5 models for analysis varied and the populations included several different subgroups. Pulmonary
6 function studies were suggestive of short term effects resulting from ambient PM exposure. Peak
7 expiratory flow rates showed decreases in the range of 2 to 5 1/min resulting from an increase of
8 50 //g/m3 in 24-h PM10 or its equivalent, with somewhat larger effects in symptomatic groups
9 such as asthmatics. Studies using FEVj or FVC as endpoints showed less consistent effects. The
10 chronic pulmonary function studies were less numerous than the acute studies, and the results
11 were inconclusive.
April 2002
8-154
DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3.3.1 Effects of Short-Term Particulate Matter Exposure on Lung Function and
2 Respiratory Symptoms
3 The available acute respiratory symptom studies discussed in the 1996 PM AQCD included
4 several different endpoints, but typically presented results for: (1) upper respiratory symptoms,
5 (2) lower respiratory symptoms, or (3) cough. These respiratory symptom endpoints had similar
6 general patterns of results. The odds ratios were generally positive, the 95% confidence intervals
7 for about half of the studies being statistically significant (i.e., the lower bound exceeded 1.0).
8 The earlier studies of morbidity health outcomes of PM10 exposure on asthmatics were
9 limited in terms of conclusions that could be drawn because of the few available studies on
10 asthmatic subjects. Lebowitz et al. (1987) reported a relationship with TSP exposure and
11 productive cough in a panel of 22 asthmatics but not for peak flow or wheeze. Pope et al. (1991)
12 studied respiratory symptoms in two panels of asthmatics in the Utah Valley. The 34 asthmatic
13 school children panel yielded estimated odd ratios of 1.28 (1.06, 1.56) for lower respiratory
14 illness (LRI) and the second panel of 21 subjects aged 8 to 72 for LRI of 1.01 (0.81, 1.27) for
15 exposure to PM10. Ostro et al. (1991) reported no association for PM25 exposure in a panel of
16 207 adult asthmatics in Denver; but, for a panel of 83 asthmatic children age 7 to 12 in central
17 Los Angeles, reported a relationship of shortness of breath to O3 and PM10, but could not separate
18 effects of the two pollutants (Ostro et al., 1995). These few studies did not indicate a consistent
19 relationship for PM10 exposure and health outcome in asthmatics.
20 Numerous new studies of short-term PM exposure effects on lung function and respiratory
21 symptoms published since 1996 were identified by an ongoing medline search.. Most of these
22 followed a panel of subjects over one or more periods and evaluated daily lung function and/or
23 respiratory symptom associations with changes in ambient PM10, PM10_2 5, and/or PM2 5. Lung
24 function was usually measured daily with many studies including forced expiratory volume
25 (FEV), forced vital capacity (FVC) and peak expiratory flow rate (PEF). Most analyses included
26 both morning and afternoon measurements. A variety of respiratory symptoms were measured,
27 including cough, phlegm, difficulty breathing, wheeze, and bronchodilator use. Finally, several
28 measures of airborne particles were used, including: PM10, PM25, PM10_25, ultrafme PM, TSP,
29 BS, and sulfate fraction of ambient PM.
30 These various studies are summarized in several tables presented in Appendix 8B. Data on
31 physical and chemical aspects of ambient PM levels (especially for PM10, PM10_2 5. PM2 5, and
April 2002 8-155 DRAFT-DO NOT QUOTE OR CITE
-------
1 smaller size fractions) are of particular interest, as are new studies examining health outcome
2 effects and/or exposure measures not studied as much in the past. Each table is organized by
3 study location, PM measure, etc. Where possible, results are presented in terms of the units
4 described earlier. Specific studies were selected for summarization based on the following
5 criteria:
6 • Peak flow was used as the primary lung function measurement of interest.
7 • Cough, phlegm, difficulty breathing, wheeze, and bronchodilator use were summarized as
8 measures of respiratory symptoms when available.
9 • Quantitative relationships were estimated using PM10, PM2 5, PM10_2 5, and/or smaller PM
10 as independent variables.
11 • The analysis of the study was done such that each individual served as their own control.
12 Other factors are discussed earlier in Section 8.1.3 of this chapter selection of Studies for
13 Review, as well as in Chapter 1.
14
15 8.3.3.1.1 Lung Function and Respiratory Symptom Effects in Asthmatic Subjects
16 Tables 8B-4 and 8B-5 in Appendix B summarize short-term PM exposure effects on lung
17 function and respiratory symptoms, respectively, in asthmatic subjects. The peak flow analyses
18 results for asthmatics tend to show small decrements for PM10 and PM2 5 as shown in studies to
19 include Gielen et al. (1997), Peters et al. (1997b), Romieu et al. (1997), and Pekkanen et al.
20 (1997) listed in summary Table 8-23 for PM10, and Table 8-24 for PM25, and in more detail in
21 Appendix 8B, Table 8B-4. Pekkanen et al. (1997) reported similar changes in peak flow to be
22 related to several sizes of PM with PN 0.032-0.10 -0.970 (0.502) l(cm3) and PM10.32 -0.901
23 (0.536) and PM10 -1.13 (0.478) for morning PEF lag 2. Peters et al. (1997c) report the strongest
24 effects on peak flow were found with ultrafme particles. PMMC001.0 x: -1.21 (-2.13, -0.30);
25 PMMC001.25: -1.01 (-1.92, -0.11); and PM10, -1.30 (-2.36, -0.24).
26 Penttinen et al. (2001) using biweekly spirometry over 6 months on a group of 54 adult
27 asthmatics found that FVC, FEVj, and spirometric PEFR were inversely, but mostly
28 nonsignificantly-associated with ultra fine particle concentrations. Compared to the effect
29 estimates for self-monitored PEFR, the effect estimates for spirometric PEFR tended to be larger.
30 The strongest associations were observed in the size range of 0.1 to 1 //m.
31
April 2002 8-156 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-23. SUMMARY OF ASTHMA PM PFT STUDIES
10
3.
to
o
o
to
-------
TABLE 8-23 (cont'd). SUMMARY OF ASTHMA PM10 PFT STUDIES
3.
to
o
o
to
-------
1 In the Uniontown reanalysis, peak flow for PM21 for a 14 //g/m3 increment was -0.91 1/m
2 (-1.14, -1.68) and PM10.2A for 15 //g/m3 +1.04 1/m (-1.32, +3.4); for State College PM2A -0.56
3 (-1.13, +0.01) andPM10.21 -0.17 (-2.07, +1.72). The Schwartz andNeas (2000) reanalyses
4 allows comparison of fine and coarse effects using two pollutant models for fraction of PM.
5 Coull et al. (2001) reanalyzed data from the Pope et al. (1991) study of PM effects on
6 pulmonary function of children in the Utah Valley, using additive mixed models which allow for
7 assessment of heterogeneity of response or the source of heterogeneity. These additive models
8 describe complex covariate effects on each child's peak expiratory flow while allowing for
9 unexplained population heterogeneity and serial correlation among repeated measurements. The
10 analyses indicates that there is heterogeneity in that population with regard to PM10 (i.e.,
11 specifically that there are three subjects in the Utah Valley study who exhibited a particularly
12 acute response to PM10). However the limited demographic data available in the Utah Valley
13 Study does not explain the heterogeneity in PM sensitivity among the school children population.
14 The peak flow analyses results for asthmatics tend to show small decrements for both PM10
15 and PM2 5. For PM10, the available point estimates for morning PEF lagged one day showed
16 decreases, but the majority of the studies were not statistically significant, see Table 8-22 and as
17 shown in Figure 8-14 as an example of PEF outcomes. Lag 1 may be more relevant for morning
18 measurement of asthma outcome from the previous day. The figure presents studies which
19 provided such data. The results were consistent for both AM and PM peak flow analyses. The
20 effects using 2 to five-day lags averaged about the same as did the zero to one-day lags, but the
21 effects had wider confidence limits. Similar results were found for the PM2 5 studies, although
22 there were fewer studies. Several studies included PM2 5 and PM10 independently in their
23 analyses of peak flow. Of these, Naeher et al. (1999), Tiittanen et al. (1999), Pekkanen et al.
24 (1997), and Romieu et al. (1996) all found similar results for PM25 and PM10. The study of
25 Peters et al. (1997c) found slightly larger effects for PM25. The study of Schwartz and Neas
26 (2000) found larger effects for fine particle measures (PM2 5, sulfate, etc.) than for the coarse
27 fraction. Naeher et al. (1999) found that FT was significantly related to a decrease in morning
28 PEF. Overall, then, PM10 and PM2 5 both appear to affect lung function in asthmatics, but there is
29 only limited evidence for a stronger effect of fine versus coarse fraction particles. Also, of the
30 studies provided, few if any analyses were able to separate out the effects of PM10 and PM2 5 from
31 other pollutants.
April 2002 8-159 DRAFT-DO NOT QUOTE OR CITE
-------
Romleuetal. (1996)
* (Mexico)
Pekkanenetal. (1997)
(Finland)
Romleu et al. (1996)
(Mexico)
Gielenetal, (1937)
(Netherlands)
-10,0
-5.0 o.O 5-°
Change In Pulmonary Function, L/min
10.0
Figure 8-14. Selected acute pulmonary function change studies of asthmatic children.
Effect of 50 Aig/m3 PM10 on morning Peak flow lagged one-day.
1 The effects of PM on respiratory symptoms in asthmatics tended to be positive, although
2 they were much less consistent than the effects on lung function. Vedal et al. (1998) reported
3 that increases in PM10 were associated with increased reporting of cough, phlegm production, and
4 sore throat and that children with diagnosed asthma are more susceptible to the effects than are
5 other children. Similarly, in the Gielen et al. (1997) study of a panel of children, most of whom
6 had asthma, low levels of PM increased symptoms and medication use. Peters et al. (1997c)
7 study of asthmatics examined particle effects by size which indicated that fine particles were
8 associated with increases in cough, of which MC 0.01-2.5 was the best predictor.
9 Delfino et al. (1998) used an asthma symptom score to evaluate the effect of acute pollutant
10 exposures. PM10 1-and 8-hr maximum had larger effects than the 24-hr mean. Subgroup
11 analyses showed effects of current day PM maximums were strongest in 10 more frequently
12 symptomatic children; the odds ratios for adverse symptoms from 90th percentile increases were
13 2.24 (1.46, 3.46), for l-hrPM10; 1.82 (1.18, 2.8), for 8-hr PM10, and 1.50 (0.80-2.80) for 24-hr
14 PM10. Analyses suggested that effects of O3 and PM10 were largely independent.
15 Romieu et al. (1996) found children with mild asthma to be more strongly affected by high
16 ambient levels of PM observed in northern Mexico City than in a study (Romieu et al., 1997)
April 2002
8-160
DRAFT-DO NOT QUOTE OR CITE
-------
1 conducted in a nearby area with lower PM10 levels (mean PM10 = 166.8 //g/m3 versus 54.2
2 Mg/m3). Yu et al. (2000) reported estimates of odd ratios for asthma symptoms and 10 //g/m3
3 increments in PM10 and PM10 values of 1.18(1.05, 1.33) and 1.09(1.01, 1.18), respectively.
4 Multipollutant models with CO and SO2 yielded for PM10, 1.06 (0.95, 1.19) for PM10, and 1.11
5 (0.98, 1.26) for PMX 0, thus showing a lower value for PMX 0 and a lower value for PM10 with a
6 loss of significance. The correlation between CO and PMl 0 and PM10 was 0.82 and 0.86. Ostro
7 et al. (2001) studied a panel of inner-city African American children using a GEE model with
8 several measures of PM, including PM10 (both 24-hour average and 1-hour max.) and PM25,
9 demonstrating positive associations with daily probability of shortness of breath, wheeze, and
10 cough.
11 Most studies showed increases in cough, phlegm, difficulty breathing, and bronchodilator
12 use, although these increases were generally not statistically significant for PM10 (see
13 Tables 8-25, 8-26, 8-27, and 8-28; and, for cough as an example, see Figure 8-15). For PM25
14 results, see Table 8-29. Several studies included two indicators for PM; PM10_2 5 or PM10 and
15 PM25 in their analyses. The studies of Peters et al. (1997c) and Tiittanen et al. (1999) found
16 similar effects for the two PM measures, whereas the Romieu et al. (1996) study found slightly
17 1 arger effects for PM2 5.
18
19 8.3.3.1.2 Lung Function and Respiratory Symptom Effects in Nonasthmatic Subjects
20 Results of the PM10 peak flow analyses in non-asthmatic studies (see Appendix 8B, Table
21 8B-6) were inconsistent, with fewer studies reporting results in the same manner as for the
22 asthmatic studies (see Table 8-30). Many of the point estimates showed increases rather than
23 decreases. Similar results were found in the PM2 5 studies (see Summary Table 8-31). The
24 effects on respiratory symptoms in non-asthmatics (see Appendix 8B, Table 8B-7) were similar
25 to those in asthmatics (see Table 8-32). Most studies showed that PM10 increases cough, phlegm,
26 difficulty breathing, and bronchodilator use, although these increases were generally not
27 statistically significant. For PM2 5 see Tables 8-32 and for PM coarse studies see Table 8-33.
28 The Schwartz and Neas (2000) reanalyses allows comparison of fine and coarse particle effects
29 on healthy school children using two pollutant models of fine and coarse PM. CM was estimated
30 by subtracting PM21 from PM10 data. They report for cough for reanalysis of the Harvard Six
31 City Diary Study in the two PM pollutant model PM2 5 (increment 15 //g/m3) OR = 1.07 (0.90,
April 2002 8-161 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-25. SUMMARY OF ASTHMA PM1(1 COUGH STUDIES
10
3.
to
o
to
-------
TABLE 8-26. SUMMARY OF ASTHMA PM1(1 PHLEGM STUDIES
10
3.
to
o
o
to
Reference citation,
location, etc.
Vedal et al. (1998)
Peters etal. (1997b)
Romieuetal. (1997)
Romieuetal. (1996)
Vedal et al. (1998)
Romieuetal. (1997)
Romieuetal. (1996)
Peters etal. (1997b)
Outcome Measure
OR Phlegm
OR Phlegm
OR Phlegm
OR Phlegm
OR Phlegm
OR Phlegm
OR Phlegm
OR Phlegm
Mean Paniculate
Levels (Range) ^g/m3
19.1 (1, 159)
47 (29, 73)
(12, 126)
166.8 (29, 363)
19.1 (1, 159)
(12, 126)
166.8 (29, 363)
47 (29, 73)
Co-Pollutants
Measured
None
SO2, sulfate, H+
Ozone
Ozone
None
Ozone
Ozone
SO2, sulfate, H+
Lag
Structure
Oday
Oday
Oday
Oday
2 day
2 day
2 day
1-5 day
Effect measures
50 ,wg/n
1.28 (0.86, 1.89)
1.13(1.04, 1.23)
1.05 (0.83, 1.36)
1.21 (1.00, 1.48)
1.40(1.03, 1.90)
1.00(0.86, 1.16)
1.16(0.91, 1.49)
1.17(1.09, 1.27)
standardized to
13PM10
oo
TABLE 8-27. SUMMARY OF ASTHMA PM10 LOWER RESPIRATORY ILLNESS (LRI) STUDIES
00
o
§
H
6
0
o
H
0
0
H
W
O
HH
7°
o
H
W
Reference citation,
location, etc.
Vedal et al. (1998)
Gielen etal. (1997)
Romieuetal. (1997)
Romieuetal. (1996)
Vedal et al. (1998)
Gielen etal. (1997)
Segala etal. (1998)
Romieuetal. (1997)
Romieuetal. (1996)
Delfinoetal. (1998)
Outcome Measure
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
Mean Paniculate
Levels (Range)
19.1(1, 159)
30.5 (16, 60)
(12, 126)
166.8 (29, 363)
19.1(1, 159)
30.5 (16, 60)
34.2 (9, 95)
(12, 126)
166.8 (29, 363)
24 h 26 (6, 51)
8-h43 (23-73)
1-h 57 (30-108)
Co-pollutants
Measured
None
Ozone
Ozone
Ozone
None
Ozone
SO2, NO2
Ozone
Ozone
Ozone
Ozone
Ozone
Lag
Structure
Oday
Oday
Oday
Oday
2 day
2 day
2 day
2 day
2 day
Oday
Oday
Oday
Effect measures standardized to
50 ,wg/m3 PM10
1.10(0.82, 1.48)
1.26 (0.94, 1.68)
1.00 (0.95, 1.05)
1.21(1.10, 1.42)
1.16(1.00, 1.34)
1.05 (0.74, 1.48)
1.66 (0.84, 3.30)
1.00 (0.93, 1.08)
1.10(0.98, 1.24)
1.47(0.90-2.39)
2.17(1.33 -3.58)
1.78(1.25-2.53)
-------
TABLE 8-28. SUMMARY OF ASTHMA PM1(1 BRONCHODILATOR USE STUDIES
3.
to
o
o
to
10
Reference citation,
location, etc.
Outcome Measure
Mean Paniculate
Levels (Range)
Co-pollutants
Measured
Effect measures standardized to
Lag Structure 50 Atg/m3 PM10
Gielenetal. (1997)
Hiltermann et al. (1998)
Peters etal. (1997b)
Gielenetal. (1997)
Hiltermann et al. (1998)
Peters etal. (1997b)
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
30.5 (16, 60)
39.7(16,98)
47 (29, 73)
30.5 (16, 60)
39.7(16,98)
47 (29, 73)
Ozone 0 day
Ozone, NO2, SO2 0 day
SO2, sulfate, H+ 0 day
Ozone 2 day
Ozone, NO2, SO2 1-7 day
SO2, sulfate, H+ 1-5 day
0.94 (0.59, 1.50)
1.03 (0.93, 1.15)
1.06 (0.88, 1.27)
2.90(1.81,4.66)
1.12(1.00, 1.25)
1.23 (0.96, 1.58)
oo
H
6
o
o
H
O
O
H
W
O
O
H
W
-------
Glelenetal. (1997)
(Netherlands)
Romieu et al, (1997)
(Mexico)
Peters etal. (1997a)
(Czech Republic)
Vedal etal. (1998)
(Canada)
0.5
1.0 2.0
Odds Ratio for Cough
4.0
8.0
Figure 8-15. Odds ratios with 95% confidence interval for cough per 50-jUg/m3 increase in
PM10 for selected asthmatic children studies at lag 0.
1
2
3
4
5
6
1
8
9
10
11
12
13
14
15
1.26) and PM10_25 (increment 8 //g/m3) OR 1.18 (1.04, 1.34) in contrast to lower respiratory
symptom results of PM25 OR 1.29 (1.06, 1.57) andPM10.25 1.05 (0.9, 1.23).
Jalaludin et al. (2000) analyses using a multipollutant model evaluated O3, PM10, and NO2.
They found in metropolitan Sydney that ambient ozone and PM10 concentrations are poorly
correlated (0.13). For PEFR PM10 only was 0.0045 (0.0125) p-0.72, and with O3, 0.0051
(0.0124), p-0.68. Ozone was also unchanged in the one- and two-pollutant models. Gold et al.
(1999) attempted to study the interaction of PM2 5 and ozone on PEF. The authors found
independent effects of the two pollutants, but found that the joint effect was slightly less than the
sum of the independent effects.
Three authors, Schwartz and Neas (2000), Tiittanen et al. (1999) and Neas et al. (1999),
used PM10_2 5 as a coarse fraction particulate measure. Schwartz and Neas (2000) found that PM10
was significantly related to cough. Tiittanen found that one day lag of PM10_25 was related to
morning PEF, but there was no effect on evening PEF. Neas et al. found no effects of PM10_2 5 on
PEF.
April 2002
8-165
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-29. SUMMARY OF ASTHMA PM2< RESPIRATORY SYMPTOM STUDIES
Reference Mean Participate Effect measures
citation, Outcome Levels (Range) Co-pollutants Lag standardized to
location, etc. Measure Mg/ni3 Measured Structure 25 /-tg/m3 PM25
Peters et al.
(1997b)
Romieu et al.
(1996)
Tittanen et al.
(1999)
Romieu et al.
(1996)
Tittanen et al.
(1999)
Ostro et al.
(2001)
Peters et al.
(1997b)
Romieu et al.
(1996)
Romieu et al.
(1996)
Romieu et al.
(1996)
Romieu et al.
(1996)
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR Phlegm
OR Phlegm
ORLRI
ORLRI
50.8 (9, 347) SO2, sulfate, H+ 0 day
85.7 (23, 177)
15 (3, 55)
85.7 (23, 177)
85.7 (23, 177)
85.7 (23, 177)
85.7 (23, 177)
85.7 (23, 177)
Ozone
Oday
NO2, SO2, CO, 0 day
ozone
Ozone
2 day
15 (3, 55) NO2, SO2, CO, 2 day
ozone
40.8 (4, 208) Ozone, NO2
3 day
Ozone
Ozone
Ozone
Ozone
1.22(1.08, 1.38)
1.27(1.08, 1.42)
1.04 (0.86, 1.20)
1.16(0.98, 1.33)
1.24(1.02, 1.51)
1.02 (0.98, 1.06)
50.8(9,347) SO2, sulfate, H+ 1-5 day 1.02(0.90,1.17)
Oday 1.21(0.98,1.48)
2 day 1.16(0.99,1.39)
Oday 1.21(1.05,1.42)
2 day 1.16(1.05,1.42)
1 8.3.3.2 Long-Term Particulate Matter Exposure Effects on Lung Function and
2 Respiratory Symptoms
3 8.3.3.2.1 Summary of the 1996 Particulate Matter Air Quality Criteria Document Key
4 Findings
5 In the 1996 PM AQCD, the available long-term PM exposure-respiratory disease studies
6 were limited in terms of conclusions that could be drawn. At that time, three studies based on a
7 similar type of respiratory symptom questionnaire administered at three different times as part of
8 the Harvard Six-City and 24-City Studies provided data on the relationship of chronic respiratory
9 disease to PM. All three studies suggest a long-term PM exposure effect on chronic respiratory
10 disease. The analysis of chronic cough, chest illness and bronchitis tended to be significantly
11 positive for the earlier surveys described by Ware et al. (1986) and Dockery et al. (1989). Using
April 2002
8-166
DRAFT-DO NOT QUOTE OR CITE
-------
p>
us.
to
O
o
to
10% AM PEFR Deer.
OR >10% AM PEFR Deer.
OR >10% AM PEFR Deer.
Morning PEFR
% change in morning PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
OR >10% PM PEFR Deer.
OR >10%PM PEFR Deer.
OR >10%PM PEFR Deer.
OR >10%PM PEFR Deer.
OR >10%PM PEFR Deer.
OR >10%PM PEFR Deer.
% change in evening PEFR
Mean Particulate
Levels (Range) /-ig/m3
51 (23,878)
28 (5, 122)
32
28 (5, 122)
42 (5, 146)
42 (5, 146)
42 (5, 146)
32
(not given)
32
(not given)
(not given)
28 (5, 122)
28 (5, 122)
51(23,878)
32
42 (5, 146)
42 (5, 146)
42 (5, 146)
34 (?, 106)
34 (?, 106)
34 (?, 106)
(not given)
Co-pollutants
Measured
Ozone
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
NO2, SO2
NO2, SO2
NO2, SO2
Ozone
NO2, SO2, CO
Ozone
Sulfate fraction
Sulfate fraction
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Ozone
Ozone
NO2, SO2
N02, S02
N02, S02
NO2, SO2, sulfate
NO2, SO2, sulfate
NO2, SO2, sulfate
N02, S02, CO
Lag Structure
Iday
Oday
1-5 day
1-4 day
Iday
2 day
1-5 day
Oday
1 day
Oday
Oday
Oday
Oday
Oday
Oday
1-5 day
Oday
2 day
1-5 day
Oday
2 day
1-5 day
1 day
Effect measures standardized to
50 Mg/m3 PM10
-0.20 (-0.47, 0.07)
1.21 (-0.43,2.85)
2. 64 (-6.56, 11.83)
-1.26 (-5. 86, 3.33)
1.04(0.95, 1.13)
1.02(0.93, 1.11)
1.05(0.91, 1.21)
-8.16(-14.81,-1.55)
0.07 (-0.50, 0.63)
-1.44 (-7.33, 4.44)
-1.52 (-2. 80, -0.24)
-0.93 (-1.88, 0.01)
0.72 (-0.63, 1.26)
2.33 (-2.62, 7.28)
-0.14 (-0.45, 0.17)
1.47 (-7.31, 10.22)
1.17(1.08,1.28)
1.08(0.99,1.17)
1.16(1.02,1.33)
1.44(1.02,2.03)
1.14(0.83,1.58)
1.16(0.64,2.10)
-0.22 (-0.57, 0.16)
-------
TABLE 8-31. SUMMARY OF NON-ASTHMA PM1(1 RESPIRATORY SYMPTOM STUDIES
10
3.
to
o
o
to
oo
1
oo
O
H
1
O
0
o
H
0
0
H
W
O
O
H
w
Reference citation,
location, etc.
Schwartz & Neas (2000)
Boezenetal. (1998)
Van der Zee etal. (1999)
Urban areas
Tittanen etal. (1999)
Van der Zee etal. (1999)
Urban areas
Van der Zee etal. (1999)
Urban areas
Tittanen etal. (1999)
Boezenetal. (1998)
Tittanen etal. (1999)
Schwartz & Neas (2000)
Van der Zee etal. (1999)
Urban areas
Van der Zee etal. (1999)
Urban areas
Outcome Measure
OR cough-no
other symptoms
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR Phlegm
OR Phlegm
LRI
LRI
LRI
Mean Paniculate Levels
(Range) ^g/m3
(not given)
42 (5, 146)
34 (?, 106)
28 (5, 122)
34 (?, 106)
34 (?, 106)
28 (5, 122)
42 (5, 146)
28 (5, 122)
(not given)
34 (?, 106)
34 (?, 106)
Co-pollutants
Measured
Sulfate fraction
NO2, SO2
NO2, SO2, sulfate
NO2, SO2, CO, ozone
NO2, SO2, sulfate
NO2, SO2, sulfate
NO2, SO2, CO, ozone
NO2, SO2
NO2, SO2, CO, ozone
Sulfate fraction
NO2, SO2, sulfate
NO2, SO2, sulfate
Lag
Structure
Oday
Oday
Oday
Oday
2 day
1-5 day
1-4 day
Oday
2 day
Oday
Oday
2 day
Effect measures standardized to
50 mg/m3 PM10
1.20 (1.07, 1.35)
1.06(0.93, 1.21)
1.04(0.95, 1.14)
1.00(0.87, 1.16)
0.94 (0.89, 1.06)
0.95(0.80, 1.13)
1.58(0.87,2.83)
1.11(0.91, 1.36)
Positive but not significant
0.98 (0.89, 1.08)
1.01(0.93, 1.10)
-------
TABLE 8-32. SUMMARY OF NON-ASTHMA PM RESPIRATORY OUTCOME STUDIES
2 5
3.
to
o
o
to
-------
TABLE 8-33. SUMMARY OF NON-ASTHMA COARSE FRACTION STUDIES OF RESPIRATORY ENDPOINTS
3.
to
o
o
to
-------
1 a design similar to the earlier one, Dockery et al. (1996) expanded the analyses to include
2 24 communities in the United States and Canada. Bronchitis was found to be higher (odds ratio
3 =1.66) in the community with the highest particle strong acidity when compared with the least
4 polluted community. Fine particulate sulfate was also associated with higher reporting of
5 bronchitis (OR = 1.65, 95% CI1.12, 2.42).
6 Interpretation of such studies requires caution in light of the usual difficulties ascribed to
7 cross-sectional studies. That is, evaluation of PM effects is based on variations in exposure
8 determined by a different number of locations. In the first two studies, there were six locations
9 and, in the third, twenty-four. The results seen in all studies were consistent with a PM gradient,
10 but it was impossible to separate out effects of PM and any other factors or pollutants having the
11 same gradient.
12 Chronic pulmonary function studies by Ware et al. (1986), Dockery et al. (1989), and Neas
13 et al. (1994) had good monitoring data and well-conducted standardized pulmonary function
14 testing over many years, but showed no effect for children from airborne particle pollution
15 indexed by TSP, PM15, PM25 or sulfates. In contrast, the Raizenne et al. (1996) study of U.S. and
16 Canadian children found significant associations between FEVj and FVC and acidic particles
17 (FT). Overall, the available studies provided only limited evidence suggestive of pulmonary lung
18 function decrements being associated with chronic exposure to PM indexed by various measures
19 (TSP, PM10, sulfates, etc.). However, it was noted that cross-sectional studies require very large
20 sample sizes to detect differences because they cannot eliminate person to person variation,
21 which is much larger than the within person variation.
22
23 8.3.3.2.2 New Studies of Long- Term Particulate Matter Exposure Respiratory Effects
24 Several studies have been published since 1996 which evaluate effects of long-term PM
25 exposure on lung function and respiratory illness, as summarized in Appendix 8B, Table 8B-8.
26 The new studies examining PM10 and PM25 in the United States include McConnell et al. (1999),
27 Abbey et al. (1998), Berglund et al. (1999), Peters et al. (1999a,b), Gauderman et al. (2000), and
28 Avol et al. (2001), which all examined effects in California cohorts but produced inconsistent
29 results. McConnell et al. (1999) noted that as PM10 increased across communities, an increase in
30 bronchitis risk per interquartile range also occurred, results consistent with those reported by
31 Dockery et al. (1996), although the high correlation of PM10, acid, and NO2 precludes clear
April 2002 8-171 DRAFT-DO NOT QUOTE OR CITE
-------
1 attribution of the McConnell et al. bronchitis effects specifically to PM alone. Avol et al. (2001)
2 reported that, for 110 children that moved to other locations as a group, subjects who moved to
3 areas of lower PM10 showed increased growth in lung function and subjects who moved to
4 communities with higher PM10 showed slowed lung function growth.
5 For non-U.S. studies, particularly interesting results were obtained by Leonard! et al. (2000)
6 as part of the Central European Air Quality and Respiratory Health (CESAR) study. Blood and
7 serum samples were collected from school children aged 9-11 yrs. in each of 17 communities in
8 Central Europe (N = 10 to 61 per city). Numbers of lymphocytes increased as PM concentrations
9 increased across the cities. Regression slopes, adjusted for confounder effects, were largest and
10 statistically significant for PM2 5, but small and non-significant for PM10_25. A similar positive
11 relationship was found between IgG concentration in serum and PM2 5 gradient, but not for PM10
12 or PM10_2 5. These results tend to suggest a PM effect on immune function more strongly due to
13 ambient fine particle than coarse particle exposure.
14 Other non-U.S. studies examined PM measures such as TSP and BS in European countries.
15 In Germany, Heinrich et al. (2000) reported a cross-sectional survey of children, conducted twice
16 (with the same 971 children included in both surveys). TSP levels decreased between surveys as
17 did the prevalence of all respiratory symptoms (including bronchitis). Also, Kramer et al. (1999)
18 reported a study in six East and West Germany communities, which found yearly decreasing TSP
19 levels to be related to ever-diagnosed bronchitis from 1991-1995. Lastly, Jedrychowski et al.
20 (1999) reported an association between both BS and SO2 levels in various areas of Krakow,
21 Poland, and slowed lung function growth (FVC and FEVj).
22
23 8.3.3.2.3 Summary of Long- Term Particulate Matter Exposure Respiratory Effects
24 The methodology used in the long-term studies varies much more than the methodology in
25 the short-term studies. Some studies reported highly significant results (related to PM) while
26 others reported no significant results. The cross-sectional studies are often confounded, in part,
27 by unexplained differences between geographic regions. The studies that looked for a time trend
28 are also confounded by other conditions that were changing over time. Probably the most
29 credible cross-sectional study remains that described by Dockery et al. (1996) and Raizenne et al.
30 (1996). This study, reported in the previous 1996 PM AQCD, found differences in peak flow
31 and bronchitis rates associated with fine particle strong acidity.
April 2002 8-172 DRAFT-DO NOT QUOTE OR CITE
-------
1 Newly available studies since the 1996 PM AQCD, overall, provide evidence consistent
2 with the findings from the above 24-City Study. Most notably, several U.S. and European
3 studies report associations between PM measures and bronchitis rates and/or lung function
4 decrements or slowed lung function growth. One also provided evidence of PM effects on
5 immune function in school children, with stronger associations for fine particle indicators than
6 for ambient coarse particles.
7
8
9 8.4 DISCUSSION OF EPIDEMIOLOGIC STUDIES ON HEALTH
10 EFFECTS OF AMBIENT PARTICULATE MATTER
11 8.4.1 Introduction
12 Numerous PM epidemiology studies assessed in the 1996 PM AQCD implicated ambient
13 PM as a likely contributor to mortality and morbidity effects associated with ambient air
14 pollution exposures in epidemiology studies. Since preparation of the last previous PM AQCD
15 in 1996, the epidemiologic evidence concerning ambient PM-related health effects has expanded
16 greatly. Past regulatory decisions have played an important role in the selection of PM indices
17 and in the evolution of the PM epidemiologic literature base. The adoption of PM10 standards in
18 1987, and of PM25 standards in 1997, have generated ambient air concentration databases that
19 made it possible for research to address and resolve many of previously unresolved linkages
20 between airborne PM and human health; and the newly authorized network of speciation
21 samplers holds promise for further advances in the near future on the identification of the more
22 influential components of the ambient air pollution mixture. The most important types of
23 additions to the database beyond that assessed in the 1996 PM AQCD, as evaluated above in this
24 chapter, are:
25
26 (1) New multi-city studies on a variety of endpoints which provide more precise estimates of the
27 average PM effect sizes than most smaller-scale individual city studies, but also showing much
28 greater heterogeneity among studies than previously observed;
29
April 2002 8-173 DRAFT-DO NOT QUOTE OR CITE
-------
1 (2) More studies of various health endpoints using ambient PM10 and/or closely related mass
2 concentration indices (e.g., PM13 and PM7), which substantially lessen the need to rely on
3 non-gravimetric indices (e.g., BS or COH);
4
5 (3) New studies evaluating relationships of a variety of endpoints to the ambient PM coarse
6 fraction (PM10_2 5), the ambient fine-particle fraction (PM2 5), and even ambient ultrafme particles
7 measures (PM0 x and smaller) using direct mass measurements and/or estimated from site-specific
8 calibrations;
9
10 (4) A few new studies in which the relationship of some health endpoints to ambient particle
11 number concentrations were evaluated;
12
13 (5) Many new studies which evaluated the sensitivity of estimated PM effects to the inclusion of
14 gaseous co-pollutants in the model;
15
16 (6) Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures
17 including PM components, based on empirical combinations (e.g., factor analysis) or source
18 profiles;
19
20 (7) Numerous new studies of cardiovascular endpoints, with particular emphasis on assessment
21 of cardiovascular risk factors as well as symptoms;
22
23 (8) Additional new studies on asthma and other respiratory conditions potentially exacerbated by
24 PM exposure;
25
26 (9) New analyses of lung cancer associations with long-term exposures to ambient PM.
27
28 (10) New studies of infants and children as a potentially susceptible population.
29
30 As discussed in Sections 8.2 and 8.3, numerous new PM epidemiology studies, both of
31 short-term and long-term PM exposure, show statistically significant excess risk for various
April 2002 8-174 DRAFT-DO NOT QUOTE OR CITE
-------
1 mortality and/or morbidity endpoints in many U.S. cities and elsewhere to be associated with
2 ambient PM indexed by a variety of ambient community monitoring methods.
3 Still, several methodological issues discussed in the 1996 PM AQCD continue to be
4 important in assessing and interpreting the overall PM epidemiology database and its
5 implications for estimating risks associated with exposure to ambient PM concentrations in the
6 United States. The fundamental issue essentially subsuming all of the other modeling issues is
7 the selection of an appropriate statistical model. These critical methodological issues are:
8 (1) potential confounding of PM effects by co-pollutants (especially major gaseous pollutants
9 such as O3, CO, NO2, SO2); (2) the attribution of PM effects to specific PM components (e.g.,
10 PM10, PM10_25, PM25, ultrafines, sulfates, metals, etc.) or source-oriented indicators (motor
11 vehicle emissions, vegetative burning, etc.); (3) the temporal relationship between exposure and
12 effect (lags, mortality displacement, etc.); (4) the general shape of exposure-response
13 relationship(s) between PM and/or other pollutants and observed health effects (e.g., potential
14 indications of thresholds for PM effects); and (5) the consequences of measurement error.
15 Assessing the above issue(s) in relation to the PM epidemiology data base remains quite a
16 challenge. The basic issue is that there are an extremely large number of possible models, any of
17 which may turn out to give the best statistical "fit" of a given set of data, and only some of which
18 can be dismissed a priori as biologically or physically illogical or impossible, except that
19 putative cause clearly cannot follow effect in time. Most of the models for daily time series
20 studies are fitted by adjusting for changes over long time intervals and across season, by day of
21 week, weather, and climate. Many of the temporal and weather variable models have been fitted
22 to data using semi-parametric methods such as spline functions or local regression smoothers
23 (loess). The goodness of fit of these base models has been evaluated by criteria suitable for
24 generalized linear models with Poisson or hyper-Poisson responses (number of events) with a log
25 link function, particularly the Akaike Information Criterion (AIC) and the more conservative
26 Bayes or Schwarz information criterion (BIC), which adjust for the number of parameters
27 estimated from the data. The Poisson over-dispersion index and the auto-correlation of residuals
28 are also often used. It is often assumed, but rarely proven, that the best-fitting models with PM
29 would be models with the largest and most significant PM indices. Also, if high correlations
30 between PM and one or more gaseous pollutants emitted from a common source (e.g., motor
31 vehicles) exist in a given area, then disentangling their relative individual partial contributions to
April 2002 8-175 DRAFT-DO NOT QUOTE OR CITE
-------
1 observed health effects associations becomes very difficult. However, there have been very few
2 attempts at broad, systematic investigations of the model selection issue and little reporting of
3 goodness-of-fit criteria among competing models that provide a better basis by which to better
4 assess or compare models.
5 One systemic analysis of model choice was carried out by Clyde et al. (2000), using
6 Bayesian Model Averaging for the same Birmingham, AL, data as analyzed by Smith et al.
7 (2000). Several different calibrated information criterion priors were tried in which models with
8 large numbers of parameters are penalized to various degrees. After taking out a baseline trend
9 (estimated using a GLM estimate with a 30-knot thin-plate smoothing spline), 7,860 models were
10 selected for use in model averaging. These included lags 0-3 of a daily monitor PM10, an
11 area-wide average PM10 value with the same lags, temperature (daily extremes and average)
12 lagged 0-2 days, humidity (dewpoint, relative humidity min and max, average specific humidity)
13 lagged 0-2, and atmospheric pressure, lagged 0-2. The model choice is sensitive to the
14 specification of calibrated information criterion priors, in particular disagreeing as to whether
15 different PM10 variables should be included or not. For example, one or another PM10 variable is
16 included in all the top 25 Akaike Information Criterion (AIC) models, but only in about 1/3 of
17 the top Bayes Information Criterion (BIC) models. Both approaches give a relative risk estimate
18 of about 1.05 (to be compared to the Schwartzvalueofl.il fora 100 unit increase), with
19 credibility intervals of (0.94, 1.17) for the AIC prior and (0.99, 1.11) for the BIC prior.
20 A validation study in which randomly selected data were predicted using the different priors
21 favored Bayesian model averaging with BIC prior over model selection (picking the best model)
22 with BIC or any approach with AIC. This method could be useful in assessing multi-pollutant
23 models.
24 The possibility that an observed effect is "real" (i.e., likely to be found in an independent
25 replication of the study) or merely a statistical artifact is usually characterized by its confidence
26 interval or by its estimated significance level. In most of this document, confidence intervals, or
27 credible intervals for Bayesian analyses, are reported in order to emphasize that the effect size is
28 not known with certainty, but some values are more nearly consistent with the data than effect
29 size values outside the interval. P-values or t-values are implicitly associated with a null
30 hypothesis of no effect. A nominal significance level of 5% (i.e., a 95% confidence interval) is
31 usually used as a guide for the reader, but P-values should not be used as a rigid decision-making
April 2002 8-176 DRAFT-DO NOT QUOTE OR CITE
-------
1 tool. If the observed confidence intervals were arrived at by a number of prior model
2 specification searches, eliminating some worse fitting models, the true interval may well be
3 wider.
4 Given the now extremely large number of published epidemiologic studies of ambient PM
5 associations with health effects in human populations and the considerably wide diversity in
6 applications of even similar statistical approaches (e.g., "time-series analyses" for short-term PM
7 exposure effects), it is neither feasible nor useful here to try to evaluate the methodological
8 soundness of every individual study. Rather, two feasible approaches are likely to yield useful
9 evaluative information: (1) an overall characterization of evident general commonalities (and/or
10 notable marked differences) among findings from across the body of studies dealing with
11 particular PM exposure indices and types of health outcomes; and (2) more thorough, critical
12 assessment of key newly published multi-city analyses of PM effects, given that greater scientific
13 weight is likely ascribable to their results than those of smaller sized studies (often of individual
14 cities) yielding presumably less precise effects estimates. However, while pooling estimates
15 across cities may give more precise estimates of mean effect size, the uncertainty in the estimated
16 mean effect may also be inflated by differences in effect size among cities.
17 In the sections that follow, each of the five issues listed above (e.g., potential confounding
18 of PM effects by co-pollutants and so on) are critically discussed. In addition, given that the
19 newer multi-city study results, e.g, the NMMAPS analysis of the 90 largest U.S. cities (Samet
20 et al., 2000a,b) show evidence of more geographical heterogeneity in the estimated PM risks
21 across cities and regions than had been seen in the studies assessed in the 1996 PM AQCD, the
22 issue of geographical heterogeneity in PM effects estimates also warrants further evaluation here
23 (as is done in Section 8.4.9).
24
25 8.4.2 Assessment of Confounding by Co-Pollutants
26 8.4.2.1 Introduction
27 Airborne particles are found among a complex mixture of atmospheric pollutants, some of
28 which are well measured (such as gaseous criteria co-pollutants O3, CO, NO2, SO2) and others
29 which are not routinely measured. The basic question here is one of determining the extent to
30 which observed health effects can be attributed to airborne particles acting alone or in
31 combination with other air pollutants. Many of the pollutants are closely correlated due to
April 2002 8-177 DRAFT-DO NOT QUOTE OR CITE
-------
1 emissions by common sources and dispersion by common meteorological factors (so that it may
2 be difficult to disentangle their effects (as noted in Section 8.1.1), because some are in the
3 pathway of formation of other pollutants, e.g.: NO —* NO2 —¥ NO34 —> Particle Mass.
4 It is widely accepted that some PM metrics are associated with health effects, and that PM
5 has effects independent of the gaseous co-pollutants. The extent to which ambient gaseous
6 co-pollutants may have health effects independent of PM is less certain, but this is important in
7 considering the extent to which health effects attributed to PM may actually be due in part to
8 co-pollutants or to some other environmental factors, and conversely. EPA produces Air Quality
9 Criteria Documents for four gaseous pollutants: CO, NO2, SO2, and O3. The possible health
10 effects of the gaseous pollutants exerted independently from PM, and in some cases jointly with
11 PM, are discussed in those documents. They are also considered to some extent in this section
12 and elsewhere in this document because they affect quantitative assessments of the effects of
13 various PM metrics when these other pollutants are also present in the atmosphere. The gaseous
14 pollutants may also be of interest as PM effect modifiers, or through interactions with PM.
15 Co-pollutant models have received a great deal of attention in the last few years because
16 there now exist improved statistical methods for estimating PM effects by analyses of daily time
17 series of mortality (Schwartz and Marcus, 1990; Schwartz, 1991) or hospital admissions
18 (Schwartz, 1994) and/or in prospective cohort studies (Dockery et al., 1993). For example, in the
19 most recent AQCD for NO2 (U.S. EPA, 1993), there are only three epidemiology studies on
20 mortality, a daily time series study (Lebowitz, 1971) and two ecological analyses (Hickey et al.,
21 1970; Mendelsohn and Orcutt, 1979). The results of these earlier studies are described by U.S.
22 EPA (1993) as non-significant or inconclusive. By comparison, many of the studies using the
23 new methods have found significant positive relationships between mortality and one or more of
24 the four gaseous criteria pollutants in daily time series studies, and between SO2 and mortality in
25 the reanalyses of two large prospective cohort studies (Krewski et al., 2000). In the daily time
26 series studies, the estimated PM effect is relatively stable when the co-pollutant is included in the
27 model in some cities, whereas the estimated PM effect in other cities changes substantially when
28 certain co-pollutants are included. In the Krewski et al. (2000) analyses, the estimated effect of
29 SO4= is greatly decreased when SO2 is also included as a predictor in a proportional hazards
30 model. How should these findings be interpreted?
April 2002 8-178 DRAFT-DO NOT QUOTE OR CITE
-------
1 A number of the analyses presented below also discuss models in which multiple particle
2 metrics are present, either with or without the gaseous criteria pollutants. These mixtures are
3 encountered in urban air. Included among the studies evaluating both fine and coarse particles
4 simultaneously are: Burnett et al. (2000), Chock et al. (2000), Clyde et al. (2000), Fairley et al.
5 (1999), Lippmann et al. (2000), Mar et al.(2000); Cifuentes et al. (2001), and Castillejos et al.
6 (2000).
7 Gaseous co-pollutant levels may be correlated with total PM mass, but may be even more
8 strongly correlated with specific PM constituents due to their emission from a common source
9 (e.g., CO and NO2 from motor vehicle exhaust). The levels of a specific gaseous co-pollutant
10 may serve as an indicator of the day-to-day variation in the contribution of a distinct emission
11 source and to the varying composition of airborne PM. In a model with total PM mass, a gaseous
12 co-pollutant may serve as a surrogate for the source-apportioned contribution to ambient air PM.
13 It would be interesting to evaluate models with both source-relevant particle components (e.g.,
14 attributable to motor vehicles, coal combustion, oil combustion) and gaseous pollutants. The
15 closest approach is Model n in Burnett et al. (2000).
16 Carbon monoxide and NO2 may be acting as indicators of distinct emission sources
17 (primarily motor vehicles) and as indicators of PM from these sources (primary particles and
18 secondary nitrate particles). However, there are other sources of NO2, such as emissions from
19 coal- or oil-burning electric power plants.
20 The role of gaseous pollutants as surrogates for source-apportioned PM may be distinct
21 from confounding. The true health effect may be independently associated with a particular
22 ambient PM constituent that may be more or less toxic than the particle mix as a whole. Thus,
23 a gaseous co-pollutant may give rise to the appearance of confounding in a regression model.
24 By serving as an indicator of the more toxic particles, the gaseous co-pollutant could greatly
25 diminish the coefficient for total particle mass. In such a model, the coefficient for total particle
26 mass would most properly be interpreted an indicator of the other, less-toxic particles. The
27 conceptual issues in evaluating potential confounding are at least as complex as the technical
28 aspects discussed below. We restrict our discussion to daily time series studies.
29
30
31
April 2002 8-179 DRAFT-DO NOT QUOTE OR CITE
-------
1 The conceptual problems in answering the question about confounding are:
2
3 (a) Biological plausibility: Can some of the gaseous criteria pollutants cause increases in
4 mortality or hospital admissions rates in the (presumably most susceptible) sub-populations at
5 current levels of exposure to ambient concentrations? If so, are these increases in mortality or
6 hospital admissions likely to be associated with cardiovascular or respiratory causes?
7
8 (b) Exposure plausibility: Do some members of the population have personal exposure to both
9 the particle metrics of ambient origin and the gaseous pollutants of ambient origin? Also, do
10 susceptible subpopulations have greater or smaller personal exposure to ambient particles or
11 gases than the population as a whole?
12
13 The technical problems in answering these question(s) are:
14
15 (c) Is the model mis-specified (omission of predictive regressors, inclusion of correlated but non-
16 predictive regressors, non-linearity, lags, measurement error from use of proxy variables)?
17
18 (d) Is there a bias in effect size estimates as a result of model mis-specification?
19
20 (e) Are the estimates of effect size standard errors sensitive to model mis-specification?
21
22 (f) Do some of the mis-specification errors compound each other, e.g., non-linearity combined
23 with measurement error?
24
25 (g) Are effect size estimates and their standard errors really significantly different among
26 models?
27
28
29
April 2002 8-180 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.2.2 Issues
2 8.4.2.2.1 Conceptual Issues in Assessing Confounding
3 These concerns overlap two of Hill's (1965) suggested criteria for causal inference.
4 (a) Biological plausibility: It is generally accepted that O3, NO2, and SO2 are associated
5 with diminished pulmonary function and increased respiratory symptoms as well as more serious
6 consequences, and CO exposure has been associated with cardiovascular effects. While one may
7 question whether adverse health effects occur in most healthy people at current exposure to
8 ambient concentrations, there may be a susceptible sub-population for whom ambient gaseous
9 pollutants cause health effects. One should remember that less than 20 years ago, current levels
10 of exposure to ambient concentrations of PM were thought to be safe. It would be premature to
11 conclude that the gaseous co-pollutants at current ambient levels are not associated with
12 respiratory and cardiovascular health effects in susceptible subpopulations.
13 Ambient gaseous co-pollutants can be potential confounders of ambient PM if: (a) both the
14 gas and PM are able to cause the same health effects; (b) if personal exposure is correlated with
15 ambient concentrations for both particles and gases respectively; (c) if the personal exposure to
16 gases and to particles are correlated, and; (d) if the ambient concentrations of particles and gases
17 are correlated. If any of these conditions fail, then we may have any of the conditions called
18 "under-fitting", "over-fitting", or "mis-fitting" described in Section 8.4.2.2.2.
19 (b) Exposure plausibility: While most Americans spend most of their time in indoor
20 microenvironments, there is still sufficient personal exposure to O3 to cause frank respiratory
21 symptoms among sensitive children or adults exercising outdoors when ambient O3
22 concentrations are high (hence the declaration of "ozone alert" days). It is also likely that some
23 fraction of ambient CO also contributes to indoor air pollution and total personal CO exposure.
24 Nitrogen dioxide, while reactive, also penetrates indoors; and an ambient pollution component of
25 total personal exposure to NO2 can be identified among individuals without indoor NO2 sources
26 and close to strong outdoor sources such as highways. While there may be some, perhaps many,
27 individuals exposed to elevated concentrations of the gaseous criteria pollutants, in order to
28 contribute to the health effects associated with ambient concentrations of a co-pollutant (e.g,.
29 PM), the ambient gaseous pollutants must be significantly and positively correlated with the
30 exposure to the co-pollutant.
31
April 2002 8-181 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.2.2.2 Statistical Issues in the Use of Multi-Pollutant Models
2 Confounding describes a condition in which one observable potentially explanatory
3 variable in an epidemiological study can stand in for another one, leading to a confusion as to
4 which variable may be causing the outcome. In most PM epidemiology studies, the gaseous
5 pollutants can often stand in for the PM metric because there is frequently a high degree of
6 positive linear correlation among PM metrics and all criteria gaseous pollutants but ozone. This
7 condition, known as multi-collinearity, is necessary to establish confounding, but not sufficient.
8 We will demonstrate these important concepts graphically using causal pathway models.
9 Figure 8-16 shows a model with two pollutants (A and B) whose ambient concentrations and
10 personal exposure are correlated, and both are capable of producing the health outcome. If both
11 A and B exposure concentrations are used in a regression model for the health outcome, and both
12 are in fact causal, then the model is correctly specified. If the personal exposures are available,
13 then estimates of the health effects of both A and B as covariates or regressors will be unbiased,
14 but are likely to have large variances because exposures to A and B are correlated. If only
15 ambient concentrations of A and B are available, then the effect estimates will be biased as well
16 as having large variances, but may still be predictive of health effects for personal exposures to
17 A and B of ambient origin. Disentangling the effects of A and B may be difficult. This
18 corresponds to common notions of confounding.
19
20
Amhipnt A fr- Ambient A is used as a regressor
AmDemA + Personal exposure to A
Personal exposure to B
Ambient B is used as a regressor
Health Outcomes
Figure 8-16. Graphical depiction of actual confounding of the effects of ambient A and
ambient B.
1 In the figures below, a solid line with an arrow suggests a causal relationship, dot-dash
2 lines suggest a non-direct association, and dotted lines suggest the absence of a pathway, either
3 for exposure or outcome. In principle, it is possible to carry out additional studies to determine
April 2002 8-182 DRAFT-DO NOT QUOTE OR CITE
-------
1 whether A and B are both capable of causing independent health effects, although clinical trials
2 with a small number of participants may not have sufficient power to detect and cannot be used
3 to study highly adverse effects (death, hospital admissions) associated with particles or gases at
4 current ambient concentrations and personal exposure levels. Case-crossover study designs may
5 allow larger populations with adverse events to be studied, but are limited by the amount of
6 personal exposure data that can be attributed to the case prior to the occurrence of the adverse
7 outcome. There are a growing number of studies relating ambient concentrations and personal
8 exposures for particles and gases.
9 In Figure 8-16, we assume that both pathways occur in nature, hence a large increase in the
10 standard errors or variances of the effect size estimates in a multi-pollutant study a natural
11 description of confounding. However, variance inflation and effect size instability may also be
12 found in the absence of confounding, as shown in Figures 8-17 through 8-20. In summary,
13 multi-pollutant models may be useful tools for assessing whether the gaseous co-pollutants may
14 be potential confounders of PM effects, but cannot determine if in fact they are. Variance
15 inflation and effect size instability can occur in non-confounded models as well as in confounded
16 models. Our usual regression diagnostic tools can only determine whether there is a potential for
17 confounding. Therefore, although multi-colinearity leading to effect size estimate instability and
18 variance inflation are necessary conditions for confounding, they are not sufficient by themselves
19 to determine whether confounding exists.
20
21
Amhipnt A fc. Ambient A is used as a reqressor „
AmoentA +• Personal exposure to A 1
' !
I I Health Outcomes
J
fc. Personal exposure to B __
r Ambient B is not used as a regressor
Figure 8-17. Graphical depiction of under-fitting of A and B. Only ambient A is used as
a regressor (covariate) and B is omitted. The estimate of A's effect is biased
because it includes the causal effect of the omitted variable B which is
correlated with A.
April 2002 8-183 DRAFT-DO NOT QUOTE OR CITE
-------
Amhipnt A i»- Ambient A is used as a reqressor
AmmentA » Personal exposure to A
Health Outcomes
A
Personal exposure to B
Ambient B is used as a regressor
Figure 8-18. Only A is causal, B is not related to the outcome, but both regressors are
included in the model, a likely cause of variance inflation.
Amhipnt A h- Ambient A is used as a regressor
AmmentA > Personal exposure to A
Health Outcomes
t
** Personal exposure to B
>- Ambient B is useg as a regressor
Figure 8-19. Graphical depiction of over-fitting of A and B. Both A and B are causal, and
both ambient A and ambient B are used as regressors (covariates). However,
there is no relationship of ambient B to personal B. The estimate of A's
effect is biased because it includes the hypothetical causal effect of ambient
B, which is correlated with A, but for which there is no personal exposure.
Amhipnt A <»> Ambient A is not used as a reqressor_
Arm em A »• Personal exposure to A
Personal exposure to B
Ambient B is useg as a regressor
Health Outcomes
A
Figure 8-20. Graphical depiction of mis-fitting of the effects of A and B. Only A is causal,
B is not related to the outcome, but B is used as a regressor in the model and
the effects of A are transferred to B.
April 2002 8-184 DRAFT-DO NOT QUOTE OR CITE
-------
1 The most commonly used methods include multi-pollutant models in which both the
2 putative causal agent (PM) and one or more putative co-pollutants are used to estimate the health
3 effect of interest. If the effect size estimate for PM is "stable", then it is often assumed that the
4 effects of confounding are minimal. "Stable" is usually interpreted as meaning that the
5 magnitude of the estimated effect is similar in models with PM alone and in models with PM and
6 one or more co-pollutants, and the statistical significance or width of the confidence interval for
7 the PM effect is similar for all models, with or without co-pollutants. These (usually
8 unquantified) criteria diagnose confounding in a narrow sense, interpreted as synonymous with
9 multi-collinearity, not as a failure of the study design or other forms of model mis-specification.
10 (c) Model mis-specification assumes many forms. The omission of predictive regressors
11 ("underfitting", defined by Chen et al., 2000) may produce biased estimates of the effects of truly
12 predictive regressors that are included in the model. Inclusion of unnecessary or non-predictive
13 regressors along with all truly predictive regressors ("over-fitting") will produce unbiased
14 estimates of effect, but may increase the estimated standard error of the estimated effect if it is
15 correlated with other predictors. Omitting a truly predictive regressor while including a
16 correlated but non-causal variable ("mis-fitting") will attribute the effect of the causal regressor
17 to the non-causal regressor. Interaction terms are candidates for omitted regressor variables. It is
18 important to avoid the "mis-fitting" scenario. Assuming there is a linear relationship when the
19 true concentration-response function is non-linear will produce a biased estimate of the effect
20 size, high or low at different concentrations. One of the most common forms of model mis-
21 specification is to use the wrong set of multi-day lags, which could produce any of the
22 consequences described as "under-fitting" (e.g., using single-day lags when a multi-day or
23 distributed lag model is needed), "over-fitting" (e.g., including a longer span of days than is
24 needed), or "mis-fitting" (e.g., using a limited set of lags while the effects are in fact associated
25 with different set of lags). Different PM metrics and gaseous pollutants may have different lag
26 structures, so that in a multi-pollutant model, forcing both PM and gases to have the same lag
27 structure is likely to yield "mis-fitting". Finally, classical exposure measurement errors (from
28 use of proxy variables) attenuates (biases) effect size estimates under most assumptions about the
29 correlations among the regressors and among their measurement errors (Zeger et al., 2000).
30
31
April 2002 8-185 DRAFT-DO NOT QUOTE OR CITE
-------
1 (d) Bias: All of the mis-specifications listed in (c) can bias the effect size estimate except
2 for "over-fitting" and measurement error of Berkson type. The estimates of the standard error of
3 the effect size estimate under "over-fitting" or Berkson error cases are inflated, however.
4 (e) Estimates of effect size standard errors are usually sensitive to model mis-specification.
5 When all truly predictive regressors are added to an "underfit" model, the uncertainty will almost
6 always be reduced sufficiently that the standard errors of estimated effect size are reduced
7 ("variance deflation"). Adding correlated non-causal variables to "over-fitted" or "mis-fitted"
8 models will further increase the estimated standard errors ("variance inflation"). Variance
9 inflation can occur whenever a covariate is highly correlated with the regressor variable that is
10 presumably the surrogate for the exposure of interest. Confounding with the regressor variable
11 can occur only when the covariate is correlated (a) with the regressor variable proxy for the
12 exposure of interest and (b) with the outcome of interest in the absence of the exposure of
13 interest.
14 (f) Mis-specification errors may compound each other. If the concentration-response
15 function is nonlinear but there is measurement error in the exposures, then different sub-
16 populations will have greater or smaller risk than assigned by a linear model. Consider the
17 hypothetical case of a "hockey-stick" model with a threshold. If there were no exposure
18 measurement error, then the part of the population with measured concentrations above the
19 threshold would have excess risk, whereas those below would not. If exposures were measured
20 with error, even if the measured concentration were above the threshold, some people would
21 actually have exposures below the threshold and no excess risk. Conversely, if the measured
22 concentration was below the threshold, some people would actually have concentrations above
23 the threshold and would have excess risk. The flattening of a non-linear concentration-response
24 curve by measurement error is a well known phenomenon that may be detected by standard
25 methods (Cakmak et al., 1999).
26 (g) The question of whether effect size estimates and their standard errors are really
27 significantly different among models is usually not addressed quantitatively. Some authors
28 report various goodness-of-fit criteria such as AIC, BIC, deviance, or over-dispersion index, e.g.,
29 (Chock et al., 2000; Clyde et al., 2000), but the practice is not yet so wide-spread as to assist in
30 analyses of secondary data for use in this document. These models are not strictly nested.
31
April 2002 8-186 DRAFT-DO NOT QUOTE OR CITE
-------
1 Variance inflation may also happen with a correctly specified model when both pollutants
2 are causal and highly correlated, compared to a model in which only one pollutant is causal and
3 the non-causal pollutant is omitted. The situation where variance or the standard error decreases
4 when an additional variable is added (variance deflation) suggests that the model with the
5 covariate is more nearly correct and that the standard errors of all covariates may decrease.
6 Statistical significance is a concept of dubious usefulness in assessing or comparing results of
7 many models from the same data set. Still, it is a familiar criterion, and one we address by using
8 a nominal two-sided 5% significance level for all tests and 95% confidence intervals for all
9 estimates, acknowledging their limitations. There is at present no consensus on what clearly
10 constitutes "stability" of a model estimate effect size, e.g., effect sizes that differ by no more than
11 20% (or some other arbitrary number) from the single-pollutant models. Simple comparison of
12 the overlap of the confidence intervals of the models is not used because the model estimates use
13 the same data, and the confidence intervals for effect size in different models are more-or-less
14 correlated. In analyses with missing days of data for different pollutants, comparisons must also
15 incorporate differences in sample size or degrees of freedom. Some examples of (a) changes in
16 the statistical significance of PM effects in different models are evident from inspection of
17 Figures 8-21 to 8-25 and of (b) relative stability in significance of PM effects in Figure 8-26.
18 In any case, statistical comparisons cannot answer questions about either conceptual or
19 statistical issues in confounding with claims about statistical significance. If the model is
20 mis-specified in any of the numerous ways described above, then effect size estimates or their
21 estimated standard errors are biased. Statistical assessments alone can determine if the PM
22 metric is too closely correlated with the other pollutants to provide an accurate quantitative effect
23 size estimate, which is of course useful information even if we conclude that it is not feasible to
24 estimate the separate effects of PM and its gaseous co-pollutants. Confounding cannot occur if
25 the gaseous co-pollutants cannot produce the health outcome, or if there is no personal exposure
26 to the gaseous co-pollutants, or the personal exposure to them is not correlated with their ambient
27 concentrations.
28 We will start by considering what can be learned from the most commonly used approach
29 to diagnose potential confounding, fitting multi-pollutant models and evaluating the stability of
30 the estimated particle effect sizes against inclusion of co-pollutants.
31
April 2002 8-187 DRAFT-DO NOT QUOTE OR CITE
-------
San Diego
a
S.
Angeles
Cleveland
Dallas - Ft. Worth
Pittsburgh
San Bernadino
^^5^t°o'^0'1
V^V1 -g??
Philadeiphia
^V*
Seattle
\
San Antonio
Santa Ana -Anaheim
Minneapolis
Figure 8-21. Effects of PM10 on total mortality in 20 large U.S. cities, as a function of
co-pollutant models. EPA presentation of results from Samet et al. (2000b).
April 2002
8-188 DRAFT-DO NOT QUOTE OR CITE
-------
>VV"
Model Type for Sulfur Dioxide
Figure 8-22. Effects of particles and gases on total mortality in eight Canadian cities.
EPA presentation of results in Burnett et al. (2000).
1 The studies identified in Table 8-34 are too numerous to allow detailed narrative
2 description here. Rather, Table 8-34 provides a summary emphasizing the points of greatest
3 relevance in evaluating multi-pollutant models. The issue of the stability of the effect size
4 estimate in multi-pollutant models may perhaps be assessed best by reporting the range of effect
5 size estimates across different co-pollutant models, in the absence of quantitative goodness-of-fit
6 comparison criteria in almost all of the papers cited. Thus, in addition to identifying the study,
7 the endpoint (usually total mortality), the PM metric, and the lags used in the analyses, the
8 minimum and maximum effect size estimates and the co-pollutants (if any) for which the
9 estimates were calculated are included. It is not uncommon for the single-pollutant PM model to
10 have the maximum PM effect size, in which case the co-pollutant is listed as "nothing."
11 There is some agreement on what constitutes "variance deflation". If an additional
12 covariate is added to a baseline model (e.g., with PM alone) and the model predicts the outcome
April 2002
8-189
DRAFT-DO NOT QUOTE OR CITE
-------
Maricopa County, Az.
Cook County, II.
S 2
i i i
012345
PM,0Lag
Los Angeles
10
E
S
ir -10
5 -5
0123456
PM,0Lag
Los Angeles
0123456
PM2 5 Lag
Figure 8-23. Effects of PM10 or PM25 on circulatory mortality in three U.S. cities as a
function of lag days. Dark shading is a co-pollutant model, light shading is
a single-pollutant PM model, medium shading shows overlap between single-
pollutant and co-pollutant models. EPA presentation of results in
Moolgavkar (2000b).
1 better with the covariate, then the reduction in variance (or deviance for generalized linear or
2 additive models [GLM or GAM]) outweighs the loss of degrees of freedom for variability.
3 Although not always true, it is reasonable to expect a decrease in the estimated asymptotic
4 standard error of the effect size estimate, but improved goodness-of-fit may not reduce the
5 standard errors of all parameters in equal proportion because introducing the new covariate
6 modifies the covariate variance-covariance matrix. The weighted inverse covariance matrix
7 provides an exact estimate for standard errors in ordinary linear regression models, and
8 approximately so in GLM or GAM. The effects on other parameter estimates are rarely reported.
9 "Variance inflation" may occur under several circumstances, including "under-fitting" and
10 "mis-fitting" in which a truly predictive covariate is omitted or replaced by a correlated proxy,
April 2002
8-190
DRAFT-DO NOT QUOTE OR CITE
-------
Model Type for Fine Particles Model Type for Coarse Particles Model Type for Nitrate in PMie
Model Type for Ozone
Model Type for Carbon Monoxide Model Type for Nitrogen Dioxide
Figure 8-24. Total mortality from particles and gases in Santa Clara County, CA.
EPA presentation of results in Fairley (1999).
Total Mortality
Circulatory Mortality
Respiratory Mortality
PM:, Model Type
E
S, 5
8
B K
\ /
, , , Y ,
PMlc.-j Model Type
Figure 8-25. Cause-specific fine or coarse particle mortality in Detroit, MI.
EPA presentation of results in Lippmann et al. (2000).
April 2002
8-191 DRAFT-DO NOT QUOTE OR CITE
-------
i
I ,
Total Mortality
Elderly Mortality
PMF Model Type
Cardiovascular Respiratory Mortality Other Mortality
Mortality
PMF Model Type
^f ^°- ^-°- o.*P-
PMF Mode] Type
Figure 8-26. Effects of fine particles on total mortality in Mexico City. EPA presentation
of results in Borja-Aburo et al. (1998).
1 and "over-fitting" in which a non-predictive covariate correlated with the PM metric is also
2 included in the model. The potential for over-fitting can be diagnosed by evaluating the
3 eigenvalues of the correlation matrix of the predictors, with very small values identifying near-
4 collinearity. However, the complete covariate correlation matrix is almost never reported,
5 including all weather variables and nonlinear functions entered separately as covariates.
6 Nonetheless, even a correlation matrix among all pollutants would be informative. Furthermore,
7 composite correlation matrices in multi-city studies may conceal important differences among
8 the correlation matrices.
9 In the absence of any better criterion, we arbitrarily define "variance deflation" and
10 "variance inflation" as occurring when the estimated standard error of the effect size estimate
11 differs by more than 25% from the single-pollutant models. These are included in Table 8-34 for
12 models for total mortality. Figures 8-21 through 8-26 show results for two multi-city studies, one
13 in the U.S, (Samet et al., 2000b) and one in Canada (Burnett et al., 2000), as well as for some
14 single-city studies in the U.S. and Mexico (Lippmann et al., 2000; Fairley et al., 1999; Borja-
15 Abuto et al., 1999). We do not show other studies because of limitations in space. Readers may
16 form their own judgements about "stability" and "variance inflation/deflation". There are
17 examples of both parameter estimate stability and instability, and of variance inflation and
18 deflation, as noted in Table 8-34.
19
April 2002
8-192
DRAFT-DO NOT QUOTE OR CITE
-------
3.
TABLE 8-34. CHARACTERIZATION
INFLATION OR DEFLATION
OF CO-POLLUTANT EFFECTS ON THE STABILITY AND VARIANCE
OF PM EFFECT SIZE ESTIMATE (IN TERMS OF EXCESS RR)
l^
o
o
to
oo
VO
oo
o
^
H
6
0
o
H
0
0
H
W
O
O
H
w
Authors
Samet et al
Samet et al.
Samet et al.
Samet et al.
Samet et al.
Samet et al
Samet et al.
Samet et al.
Samet et al.
Samet et al.
Samet et al
Samet et al.
Samet et al.
Samet et al.
Samet et al.
Samet et al
Samet et al.
Samet et al.
Samet et al.
Samet et al.
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Year
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
Oty
Los Angeles
New York
Chicago
Dallas-FtW
Houston
San Diego
SantaAna-
Phoenix
Detroit
Miami
Philadelphia
Minneapolis
Seattle
San Jose
San Bernard
Cleveland
Pittsburgh
Oakland
Atlanta
San Antonio
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Los Angeles
Los Angeles
Endpoint
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
PM
Metric
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
Lag
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
1
2
3
4
5
0
1
PM
Alone Age or
Exc. RR Season
1.9
5.7
1.6
-2
1
5.6
3.5
3.3
2.4
3.5
3.9
2.4
1.4
1.6
1.3
-0.2
2
10.8
0.2
3.5
2.4
2.1
1.6
1.1
0.8
-0.5
0.4
0.9
Min.
ExcRR
-0.3
1.1
-0.2
-2
0.1
2.1
1.9
-0.4
2.1
3
3.5
NA
1.4
-1.4
1
-0.2
1.3
5.7
-5.4
3.5
1.9
1.9
1.6
1
0.8
-0.5
-2
-4
PM with
O3, NO2,
CO
O3, SO2
03
nothing
O3, NO2
O3, SO2
O3! NO2
O3, CO
O3, CO
03
03
NA
nothing
O3! NO2
O3, NO2
nothing
O3, NO2
O3, CO
O3, NO2
nothing
CO
CO
CO
CO
CO
nothing
CO
CO
Max.
ExcRR
2.1
6.1
1.6
3.4
1.1
5.8
5.3
10.2
3.3
3.5
6.5
NA
7.4
1.9
1.6
0.4
2.2
10.8
0.25
4
2.4
2.1
1.6
1.1
0.8
-0.4
0.4
0.9
PM
with
03
03
nothing
03
03
03
03
O3, NO2
O3, SO2
nothing
O3, NO2
NA
03
03
O3, SO2
O3, NO2,
CO
O3, SO2
nothing
nothing
O3, CO
nothing
nothing
either
nothing
either
CO
nothing
nothing
Variance
Inflation Inflation Inflation Inflation Deflation
O3, CO O3, NO2
O3! CO O3, NO2 O3, SO2
O3, SO2
O3, CO O3! NO2
O3! CO O3, NO2
O3, CO O3, NO2 O3, SO2
O3, CO O3, NO2 O3, SO2
O3! NO2 O3, SO2
O3 O3, CO
O3, CO O3! NO2
O3 O3, CO O3, NO2 O3, SO2
O3! CO O3, NO2 O3, SO2
O3, CO O3! NO2 O3, SO2
O3, CO O3, NO2
CO
CO
CO
CO
CO
CO
CO
-------
3.
TABLE 8-34 (cont'd). CHARACTERIZATION OF CO-POLLUTANT EFFECTS ON THE STABILITY AND VARIANCE
INFLATION OR DEFLATION OF PM EFFECT SIZE ESTIMATE (IN TERMS OF EXCESS RR)
l^
o
o
to
oo
^
VO
•^
O
^
H
1
O
o
2!
-*— I
O
H
O
0
H
W
O
H
W
Authors
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Moolgavkar
Fairley
Fairley
Fairley
Burnett et al.
Burnett et al.
Burnett et al.
Burnett et al.
Lippmann
etal.
Year
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
2000b
1999
1999
1999
2000
2000
2000
2000
2000
Oty
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
San Jose
San Jose
San Jose
Canada 8
Canada 8
Canada 8
Canada 8
Detroit
Endpoint
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
PM
Metric
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM2.5
PM2.5
PM2.5
PM2.5
PM2.5
PM2.5
PM25
PM10.2.5
PM10,
nitrate
PM2.5
PM25
PM10.2.5
PM10.2.5
PM25
Lag
2
3
4
5
0
1
9
3
4
5
0
1
2
3
4
5
0
0
0
0
1
0
1
3
PM
Alone Age or
Exc. RR Season
2.4
0.8
0.6
-2
2
3
4.5
-0.8
0.5
6.1
1.5
1.4
0.6
-0.6
0.2
-1.2
8.5
4.5
4.6
2.3
3
1.2
1.8
3.1
Min.
ExcRR
-1.9
-3.9
-1.8
-5.5
0.2
0.4
0.8
-0.8
-0.8
3
0.4
0.5
-1.2
-1.2
-1.1
-1.5
-0.1
-5.6
4.4
1.1
1.9
0.2
1
2.8
1.6
PM with
CO
CO
CO
CO
CO
CO
CO
nothing
CO
CO
CO
CO
CO
CO
CO
CO
PM10,
nitrate
PMF
NO2
03
four
gases
03
CO,NO2,
SO2
03
PM10.2.5
Max.
ExcRR
2.4
0.8
0.6
-2
2
3
4.5
-0.8
0.5
6.1
1.5
1.4
0.6
-0.6
0.2
-1.2
10.8
4.5
5.6
2.6
3
2.6
1.8
3.9
PM Variance
with Inflation Inflation Inflation Inflation Deflation
nothing
nothing
nothing
nothing CO
nothing CO
nothing CO
nothing CO
CO CO
nothing CO
nothing CO
nothing CO
nothing CO
nothing CO
nothing CO
nothing CO
nothing CO
NO2 O3, CO, nitrate
NO2
nothing
PM25 NO2 PM25
CO NO2
nothing CO NO2 O3, CO, NO2 Model I
SO2
nothing
SO2
PM10.2.5
-------
3.
TABLE 8-34 (cont'd). CHARACTERIZATION OF CO-POLLUTANT EFFECTS ON THE STABILITY AND VARIANCE
INFLATION OR DEFLATION OF PM EFFECT SIZE ESTIMATE (IN TERMS OF EXCESS RR)
l^
o
o
to
oo
VO
01
o
£j
H
6
0
o
H
0
0
w
o
o
H
w
Authors
Lippmann
etal.
Chock et al.
Chock et al.
Chock et al.
Chock et al.
Chock et al.
Chock et al.
Chock et al.
Chock et al.
Goldberg
etal.
Goldberg
etal.
Borja-Abuto
Year City
2000 Detroit
2000 Pittsburgh
2000 Pittsburgh
2000 Pittsburgh
2000 Pittsburgh
2000 Pittsburgh
2000 Pittsburgh
2000 Pittsburgh
2000 Pittsburgh
2000a Montreal
2000a Montreal
previous
previous
previous
previous
any previous
any previous
age previous
65+
1999 Mexico City
Endpoint
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
Total Mort.
cancer
acute LRI
airways
congestive
coronary
cardio-
airways
Total Mort.
PM
Metric
PM10.2.5
PM10
PM10
PM10
PM10
PM2.5
PM2.5
PM10.2.5
PM10.2.5
PM2.5 (est)
PM2.5 (est)
disease
heart
artery
vascular
disease
PM25
PM
Alone
Lag Exc. RR
3
Tbl.30
Tbl. 90
Tbl.60
Tbl. 90
Tbl. 90
Tbl. 90
Tbl. 90
Tbl. 90
MeanO,
1,2
Mean 0,
failure
disease
disease
no
meds.
4
4
3.1
1.3
2
2.2
2.6
1.5
0.7
1.3
5.8
1,2
4.9
12.9
3.5
10.9
4.9
7.4
11.2
3.4
Age or Min.
Season Exc RR
3.1
2.8
age < 75 2.6
age < 75 1.3
age 75+ 2
age 75+ 2.2
age < 75 2.6
age 75+ 1
age < 75 0.7
age 75+ 1.3
5.1
with previous
4.3
12.8
1.6
9.4
4.4
7.3
10.5
3.4
PM with
03
PM2,
CO
nothing
nothing
nothing
nothing
all 4
gases
nothing
nothing
CO
illness
03
03
03
03
03
03
03
NO2
Max.
ExcRR
4.1
4.3
2
3.7
2.7
3.3
1.5
0.8
1.4
5.9
4.9
13
3.6
10.9
4.9
7.4
11.7
4.2
PM Variance
with Inflation Inflation Inflation Inflation Deflation
SO2
PM25
NO2 NO2 CO, SO2 all 4
gases
all 4
gases
CO NO2 CO, NO2 all 4
gases
all 4 all 4
gases gases
all 4
gases
nothing
all
4 gases
all 4
gases
SO2 CO
(slight)
nothing
SO2
SO2
nothing
nothing
nothing O3 SO2
SO2
O3, NO2 NO2 O3, NO2
-------
3.
TABLE 8-34 (cont'd). CHARACTERIZATION OF CO-POLLUTANT EFFECTS ON THE STABILITY AND VARIANCE
INFLATION OR DEFLATION OF PM EFFECT SIZE ESTIMATE (IN TERMS OF EXCESS RR)
l^J
o
o
to
oo
1 .
VO
Oi
H
1
O
0
o
H
0
0
H
W
O
O
H
w
Authors
Castillej os
etal.
Castillej os
etal.
Castillej os
etal.
Ci&entes
etal.
Ci&entes
etal.
Ci&entes
etal.
Ci&entes
etal.
Ci&entes
etal.
Ci&entes
etal.
Atkinson
etal.
Katsouyanni
etal.
Year City
2000 Mexico City
2000 Mexico City
2000 Mexico City
2000 Santiago
2000 Santiago
2000 Santiago
2000 Santiago
2000 Santiago
2000 Santiago
2000 8 European
2001 29
European
PM
Endpoint Metric
Total Mort. PM10
Total Mort. PM2 5
Total Mort. PM10.2 5
Total Mort. PM2 5
Total Mort. PM2 5
Total Mort. PM2 5
Total Mort. PM10.2 5
Total Mort. PM10.2 5
Total Mort. PM10.2 5
Resp. Mort. PM10
Total Mort. PM10
PM
Alone Age or Min. Max. PM Variance
Lag Exc. RR Season Exc RR PM with Exc RR with Inflation Inflation Inflation Inflation Deflation
Avg. 9.5 9.5 nothing 13 NO2 O3, NO2
1,2
Avg. 3.7 0.4 PM10.25 3.7 nothing O3, NO2 PM10.25
1,2
Avg. 10.5 10.3 PM25 11 O3 O3, NO2 PM25
1,2
Avg. 4.5 summer 2.8 O3, 4.5 nothing
1, 2 PM10.2.5
Avg. 1.6 winter 0.8 NO2 1.8 PM10.25 PM10.25 CO SO2
1,2
Avg. 1.8 all year 0.8 NO2 1.8 nothing PM10.25 CO
1,2
Avg. 5.5 summer 3.7 PM10_25 5.5 nothing
1,2
Avg. 1.8 winter -0.5 PM10.25 1.8 nothing PM10.25 CO NO2
1,2
Avg. 2.3 all year 0.3 PM10.25 2.3 nothing PM10.25 CO NO2
1,2
4.6 3.6 SO2 5.6 NO2 O3 SO2
Avg. 3.4 1.8 NO2 3 SO2 O3 NO2
0, 1
Note: Stadies with ozone as the only co-pollutant are omitted.
-------
1 8.4.2.2 Assessments of Confounding Using Multi-Pollutant Models with Observed Gases
2 The most common approach to evaluation of confounding of PM effects by gaseous
3 co-pollutants is to compare the estimates of the PM effect size in models with and without the
4 gases. The single pollutant model is, in general, of the form
5
6 RR = exp(pPM x PM + other covariates) (8-1)
7
8 and the corresponding multi-pollutant model is of the form
9
10 RR = exp(pPM x PM + pgasl x [gasl] + Pgas2 x [gas2] + other covariates) (8-2)
11
12 If the estimates of PPM in model specifications in Equations 8-1 and 8-2 are very different, or if its
13 estimated standard error is much larger in Equation 8-2 than in Equation 8-1, then one may
14 conclude that PM is confounded with the gaseous co-pollutants, particularly if PM and the co-
15 pollutants are highly inter-correlated, as they often are. Variance inflation (large standard error)
16 of the estimated PPM in the multi-pollutant model suggests that the pollutants are collinear.
17 A large change in estimated PPM without much variance inflation suggests that some of the total
18 effect might be shared among PM and the gaseous pollutants, whereas a large change in the PM
19 coefficient along with a large increase in the estimated variance of the PM regression coefficient
20 suggests only that the PM coefficient is unstable.
21 The results in Table 8-34 show numerous cases of variance inflation, an expected
22 consequence of the multi-colinearity of PM and gaseous pollutants in most cities where
23 combustion products from motor vehicles, power plants, home heating, and industrial processes
24 dominate the urban air mix. We have not tabulated findings for which the only co-pollutant is
25 ozone (Dominici et al., 2000a; Lipfert et al., 2000; Samet et al., 2000a,c), as these results appear
26 to add little to the findings in (Samet et al., 2000b). Samet et al. (2000b) found that including
27 ozone along with PM10 tends to slightly increase the PM10 effect, but increasing the variance
28 substantially only in Cleveland and Seattle. Moolgavkar (2000b) finds that adding CO to a PM10
29 model substantially increases its variability at most single-lags in Chicago and Phoenix, but less
30 so in Los Angeles. Adding CO to the PM2 5 model in Los Angeles results in a substantial
31 reduction in the uncertainty of the PM2 5 effect, one of the few cases of variance deflation,
April 2002 8-197 DRAFT-DO NOT QUOTE OR CITE
-------
1 implying a better-fitting model. This suggests that it may be easier to separate the effects of CO
2 from the effects of PM2 5 than from the effects of PM10 in Los Angeles.
3 Some of the studies in which either PM10, PM25, or PM10_2 5 coefficients are evaluated in
4 multi-pollutant models are discussed below. A partial list of recent studies for which such
5 assessments can be done is given in Table 8-35. Only a subset of these studies are discussed as
6 examples of what can be learned from multi-pollutant analyses. Table 8-34 presents a summary
7 of the results, ordered starting with studies having clearer indications of PM effect size instability
8 against co-pollutants through to studies having clear indications of PM effect size stability.
9
10 Samet et al. (2000b) mortality in 20 U.S. cities
11 Most cities in Samet et al. (2000b) show a considerable reduction in effect size with
12 inclusion of ozone and another pollutant in the model, compared to the model with PM10 alone.
13 The maximum PM10 effect size across co-pollutant models is rarely much larger than the single-
14 pollutant PM10 effect, except in Dallas-Fort Worth, Phoenix, and Seattle. The overall impression
15 in Figure 8-21 is that the co-pollutant models are neither consistently stable nor unstable, so that
16 the sensitivity of the model to co-pollutants may vary substantially from city to city. The results
17 shown are all for a single lag day 1. Results in Moolgavkar (2000b) suggest that different single-
18 day lags in different cities may affect the apparent stability and variance inflation or deflation of
19 the estimated PM effects.
20
21 Moolgavkar (2000b) total and cardiovascular mortality in 3 U.S. cities.
22 Results for total mortality are shown in Figures 1, 2, and 3 in the Moolgavkar (2000b)
23 paper, using CO as the only co-pollutant. The assessment of stability and variance
24 inflation/deflation against co-pollutants and lags is shown in Table 8-34 for total mortality where
25 CO is the only co-pollutant, for consistency with Samet et al. (2000b). The results for
26 cardiovascular mortality are shown in Figure 8-23. It is clear that the results depend on both city
27 and lag. The PM10 models for total mortality in Los Angeles and Phoenix show systematic
28 attenuation of effect when CO is added, whatever the lag. The PM10 effect size in Chicago is
29 only somewhat attenuated by CO. Both Chicago and Phoenix show variance inflation by CO at
30 all lags. In Los Angeles, however, the PM10 effect shows little variance inflation by CO except at
31 lags 0 ands 5, even though the PM10 effect is strongly attenuated. The PM2 5 effect size on total
April 2002 8-198 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-35. SOME NEW DAILY TIME SERIES STUDIES FOR
MORTALITY OR MORBIDITY WITH CO-POLLUTANT MODELS AND
GRAVIMETRIC PM INDICES
Study
City or Cities
PM Indices
Co-Pollutants
Studies in the U.S. and Canada
Burnett et al. (2000)
Chock et al. (2000)
Dominici et al. (2000a)
Fairleyetal. (1999)
Goldberg et al. (2000,
2001abcd)
Gwynn et al. (2000);
Gwynn and Thurston (2001)
Lipfertetal. (2001)
Lippmann et al. (2000)
Moolgavkar (2000a)
Moolgavkar (2000b)
Samet et al. (2000abc)
8 Canadian Cities
Pittsburgh, PA
19 U.S. cities with
ozone data
Santa Clara County,
CA
Montreal, PQ, Canada
Buffalo, NY
Philadelphia, PA -
Camden, NJ
Detroit, MI
Los Angeles, CA
Chicago, IL
Phoenix, AZ
Los Angeles, CA
Chicago, IL
Phoenix, AZ
19 U.S. cities with
co-pollutant models
PM2 5, PM10.2.5, PM10
PM2 5, PM10.2.5, PM10
PM10
PM25,PM10.25,PM10,
COH, NO3', SO4=
Estimated PM2 5,
Sutton sulfate, COH
PM25,PM10, COH,
SO;, H+
PM2.5, PM10.2.5, PM10
PM2 5, PM10.2.5, PM10
PM2 5, PM10.2.5, PM10
PM10
PM10
PM2 5, PM10.2 .5, PM10
PM10
PM10
PM10
O3, CO, NO2, SO2
O3, CO, NO2, SO2
03
O3, CO, NO2
O3, CO, NO2, SO2, NO
O3, CO, NO2, SO2
O3 (only particle
co-pollutant)
O3, CO, NO2, SO2
O3, CO, NO2, SO2
O3, CO, NO2, SO2
O3, CO, NO2, SO2
O3, CO, NO2, SO2
O3, CO, NO2, SO2
O3, CO, NO2, SO2
O3, CO, NO2, SO2
Studies in Latin America
Borja-Abuto et al. (1998)
Castillejos et al. (2000)
Cifuentes et al. (2000)
Loomisetal. (1999)
Mexico City, D.F.,
Mexico
Mexico City, D.F.,
Mexico
Santiago, Chile
Mexico City, D.F.,
Mexico
PM25
PM2 5, PM10.2 .5, PM10
PM2.5, PM10.2.5, PM10
PM25
O3, NO2
03, N02
O3, CO, NO2, SO2
Studies in Europe
Atkinson etal. (2001)
Katsouyanni et al. (2001)
Sunyer and Basagna (2001)
8 European cities in
APHEA2
29 European cities
Barcelona, Spain
Several, converted
to PM10
PM10, Black Smoke (BS)
PM10
O3, CO, NO2, SO2
O3, NO2, SO2
O3, CO, NO2
Studies in Asia
Kwon etal. (2001)
Seoul, South Korea
PM10
O3, CO, NO2, SO2
April 2002
8-199
DRAFT-DO NOT QUOTE OR CITE
-------
1 mortality in Los Angeles is also greatly attenuated, but there is substantial variance deflation.
2 Perhaps CO is either a surrogate or a potential confounder of PM10 in Los Angeles.
3
4 Fairley et al. (1999) total mortality in Santa Clara County (San Jose), CA.
5 This study is noteworthy because it is, as far as we are aware, the only study using particle
6 nitrates as an exposure index. The PM10-nitrate component is very stable and shows variance
7 inflation from NO2 and PM2 5, as might be expected. The extent to which nitrate is a component
8 of PM25 rather than PM10 in this study is unknown. In many western cities, PM25 is much more
9 alkaline than in the eastern U.S., so that nitrates are less likely to be displaced to the coarse PM10
10 fraction. In the eastern U.S., where particle acidity is greater, there may be a greater
11 displacement of nitrates from fine particles where sulfates are a much larger fraction of particle
12 mass than in the west, and consequently nitrates are more likely to reside in the coarse fraction in
13 the east. It is therefore uncertain that the nitrate component of the atmosphere can account for
14 the large adverse health effects of PM10 observed in many cities in the northeast and industrial
15 midwest. Clearly, it would be desirable to have more epidemiology studies with the nitrate
16 component (size-stratified and measured in a manner to avoid evaporated losses) used as a PM
17 component in models, notwithstanding technical difficulties that might be encountered in
18 measuring the samples. As shown in Table 8-34, the PM2 5 effect size estimate is almost
19 eliminated by including PM10-nitrates , and the estimated PM25 effect size variance inflated by
20 including the criteria gaseous pollutants. As shown in Figure 8-24, the PM10-nitrate and PM2 5
21 effect size estimates are stable against gaseous pollutants. It is unlikely that the gaseous
22 pollutants are confounders of PM10-nitrate.
23
24 Other studies
25 The results from other studies are also summarized in Table 8-34 and Figures 8-19 to 8-26.
26 There are numerous examples of effect size instability and variance inflation. The only other
27 cases of substantial variance deflation (i.e., better prediction of PM effect size by inclusion of
28 co-pollutants) are in Burnett et al. (2000), Lippmann et al. (2000), and Goldberg et al. (2000a).
29 In Burnett et al. (2000), the model with sulfates as a surrogate for PM2 5, three gases, and four
30 metals in PM2 5 give a better estimate of the sulfate effect than do the models with the gravimetric
31 PM2 5 index. In Lippmann et al. (2000) the models for total mortality with fine or coarse particles
April 2002 8-200 DRAFT-DO NOT QUOTE OR CITE
-------
1 give better predictions of the particle effect when the other size component is included than do
2 any of the gaseous co-pollutant models, regardless of the single day lags used for the various PM
3 or gaseous co-pollutants (see Table 8-36).
4
5
TABLE 8-36. SINGLE-DAY LAGS USED IN CO-POLLUTANT MODELS IN
(Lippmann et al., 2000, Tables 13-14)
Pollutant
Endpoint
Total Mortality
Circulatory Mortality
Respiratory Mortality
Pneumonia Admissions
COPD Admissions
Ischemic Heart Disease (IHD)
Admissions
Dysrhythmia Admissions
Heart Failure Admissions
Stroke Admissions
PM10
1
1
0
1
3
2
1
0
1
PM25
3
1
0
1
3
2
1
1
0
PM10_2.5
1
1
2
1
3
2
0
0
1
03
0
0
0
3
3
3
3
3
3
CO
1
1
1
3
3
3
3
3
3
NO2
1
1
1
3
3
3
3
3
3
SO2
3
3
3
3
3
3
3
3
3
1 8.4.2.3 Assessment of Confounding in Multi-City Studies: Pooling Effects
2 One approach to evaluating confounding used in a number of multi-city studies is to pool or
3 combine the results of individual within-city studies using either standard analytic techniques
4 such as inverse variance averaging (Atkinson et al., 2001; Katsouyanni et al., 2001) or Bayesian
5 second-stage meta-analysis methods (Samet et al., 2000a) which may be thought of as another
6 kind of averaging. The argument is that if the pooled or combined PM effect size estimates for
7 the single-pollutant models across a number of cities differing greatly in PM and co-pollutant
8 distributions and correlations are similar to those obtained from PM effect size estimates of
9 multi-pollutant models across the same cities, then it is unlikely that the co-pollutants are
10 confounding the PM effect.
April 2002
8-201
DRAFT-DO NOT QUOTE OR CITE
-------
1 The basis of this argument is not self-evident. Examination of the results of Samet et al.
2 (2000a) for the 20 largest U.S. cities discussed in Section 8.2.2.2 shows that there are a variety of
3 different patterns of change in PM10 effect size associated with including gaseous co-pollutants in
4 a co-pollutant model. Results for multi-pollutant models in the NMMAPS Part n study (Samet
5 et al., 2000a) for the 20 largest U.S. cities are shown in Section 8.4.2.2, Figure 8-21. The
6 Bayesian posterior distribution of the estimates was shown earlier in Figure 8-3. It should be
7 noted that the posterior distribution for the mean PM10 effect has about the same standard
8 deviation for all of the co-pollutant models, as might be expected by examining Figure 8-21. The
9 posterior distribution for the mean PM10 effect size estimate remains relatively unchanged from
10 the single-pollutant model when O3 is included as a co-pollutant, and tends to decrease
11 substantially (by about 30%) when either CO or NO2 are added as co-pollutants in addition to O3.
12 Adding SO2 causes a smaller reduction the estimated PM10 effect. The estimated PM10 effect
13 follows a similar pattern of association with the gaseous criteria pollutants in many large U.S.
14 cities, but shows a very different pattern in other cities, either being stable with respect to co-
15 pollutants or showing increasing effects of PM10 with the inclusion of CO, NO2, or SO2 in a
16 multi-pollutant model for apparently dissimilar cities including Seattle, WA, Phoenix, AZ,
17 Dallas-Form Worth, TX, and Philadelphia, PA. One can argue that CO and NO2 are often
18 closely and positively associated with PM10, especially if PM10 is dominated by the fine particle
19 fraction, predominantly from combustion, thus reducing the magnitude and significance of the
20 estimated PM10 effect. This explanation is less convincing in cities such as Phoenix, AZ, where
21 excess mortality is better associated with coarse particles than with fine particles (Mar et al.,
22 2000; Smith et al., 2000; Clyde et al., 2000).
23 Moolgavkar (2000b) finds different effects of PM10 on circulatory mortality in Phoenix,
24 where the single-pollutant and multi-pollutant models largely agree, disagreeing slightly more in
25 magnitude and significance at lag days 0 and 1, 4 and 5, than at lag days 2 and 3. In general,
26 Moolgavkar finds different temporal patterns of the effect of PM10 for single-pollutant and
27 co-pollutant models among Phoenix, AZ, Chicago, IL, and Los Angeles, CA. There are many
28 possible reasons for city-to-city variations in these relationships. One possibility is differences in
29 the mix of fine and coarse particles (which may be quite different in those cities). The combined
30 estimates across cities using the one-day lags for all cities or all regions may overlook different
April 2002 8-202 DRAFT-DO NOT QUOTE OR CITE
-------
1 delays associated with different causes for total or cause-specific mortality in different cities,
2 suggesting that caution be used when pooling data from different places.
3 Analogous differences in the stability of the estimated PM coefficients have also been
4 noted in other city-specific studies discussed in Section 8.4.2.2. Does the occasional instability
5 of PM coefficients in co-pollutant models across different cities reflect real differences, or is it
6 merely another kind of statistical variability that might be explained in a second-stage
7 regression? Second-stage regression approaches are discussed next in Section 8.4.2.4.
8
9 8.4.2.4 Assessment of Confounding in Multi-City Studies: Regression
10 8.4.2.4.1 Second-Stage Regression and Identification of Effects Modifiers
11 The approach used by Atkinson et al. (2001); Janssen et al. (2002); Katsouyanni et al.
12 (2001); Levy et al. (2000); and Samet et al. (2000b) is to accept the estimated PM effect-size
13 estimates as samples from a distribution of possible effect sizes in different cities (a "meta-
14 population") and to fit a weighted regression model of the estimated effect sizes on various
15 community-wide indices. The community-wide indices for these studies have included: (a) the
16 mean or median levels of co-pollutants; (b) the median or trimmed mean of the correlations of
17 the PM10 concentrations at different sites across the city (as an index of spatial measurement
18 error); (c) characteristics of particles such as the observed or estimated PM25/PM10 ratio;
19 (d) some characteristics of the distribution of meteorological variables (e.g., mean maximum
20 annual temperature); (e) some community-wide surrogates for possible sources (e.g., density of
21 vehicle miles traveled, percent of population using public transportation, or PM25/NO2 ratio as
22 indicators of motor vehicle use); (f) factors possibly affecting exposure (e.g., percentage of
23 residences with air conditioning); and (g) sociodemographic characteristics possibly affecting
24 exposure and susceptibility to particles (e.g., average education level, percentage of residents
25 >64 years of age, measures of immigration or emigration from the community).
26 Samet et al. (2000b) found that estimated PM10 effects on total mortality (a) increased with
27 increasing mean O3 (not significant in any model), increasing mean NO2 (significant in three- and
28 four-variable models, only marginally significant in a five-variable model), and increasing
29 percentage without a high school diploma, but (b) decreased with increasing mean PM10
30 (significant only in the best five-pollutant model) and the median PM10 cross correlation.
April 2002 8-203 DRAFT-DO NOT QUOTE OR CITE
-------
1 Janssen et al. (2002) found variations among PM10 effects on hospital admissions in single-
2 pollutant models to be associated with differences in: the community-wide prevalence of air
3 conditioners; sources of PM10, population density; and density of vehicle traffic (mean daily
4 vehicle miles of urban travel per square mile). The same 14 cities used in the hospital
5 admissions studies in Samet et al. (2000b) were divided into two groups, five in which PM10
6 concentrations peaked in the winter (Boulder and Colorado Springs, CO; Provo-Orem, UT;
7 Seattle and Spokane, WA) and nine others where PM10 concentrations peaked in the summer
8 (Birmingham, AL; Canton and Youngstown, OH; Chicago, IL; Detroit, MI; Minneapolis, MN;
9 Nashville, TN; New Haven, CT; Pittsburgh, PA). There was a statistically significant negative
10 relationship between the percentage of homes with central air conditioning and the regression
11 coefficient (excess relative risk) for cardiovascular disease, this being lower in winter-peaking
12 cities than in summer-peaking cities. The relationship between exposure and ventilation rate is
13 discussed further in Section 8.4.2.5. Ventilation rate also effects personal exposure to gaseous
14 co-pollutants. Additional studies similar to Jannsen et al. (2002) would likely help clarify further
15 the associations between particles and gases and may offer a useful alternative method for
16 assessing multi-pollutant models across different cities.
17 Katsouyanni et al. (2001) found that PM10 effects on mortality increased with mean NO2
18 level, mean temperature, and percentage of population with age > 65 years. The estimated PM10
19 effects decreased with increasing PM10/NO2 ratio, increasing relative humidity, and increasing
20 age-adjusted mean mortality rate in the 29 cities in the APHEAII study in Europe.
21 Atkinson et al. (2001) studied respiratory mortality in eight European cities and found a
22 significant positive relationship between asthma mortality at ages 0 to 14 years and the percent of
23 population > age 65, a significant negative relationship with community smoking prevalence, and
24 a negative relationship with relative humidity. In the analyses for age 65+ total respiratory
25 mortality, and age 65+ mortality from asthma and COPD, effect size increased significantly with
26 larger mean O3.
27 These studies, while quite informative, do not address the core issue in assessment of
28 potential confounders: in any single city, it is often difficult to disentangle the components of the
29 mixture of air pollutants that actually exists. The problem derives from the typically co-linear
30 relationship of particles and gaseous co-pollutants, where co-pollutants may have relatively high
31 linear correlation coefficients among themselves. The assessment of confounding depends on the
April 2002 8-204 DRAFT-DO NOT QUOTE OR CITE
-------
1 correlations among the pollutants, not on their absolute levels. It would be possible for
2 co-pollutants to have exactly the same correlations with PM even if the absolute concentrations
3 of the co-pollutants differed greatly from one city to the next. For this reason, even if the
4 co-pollutant concentrations are very low, co-linearity could still be manifested in a
5 multi-pollutant model because the co-pollutant would increase and decrease in step with the PM
6 index due to common meteorological conditions. For example, in a recent study of respiratory
7 symptoms in Port Alberni, BC, Canada, the levels of SO2 and other pollutants are low, and are
8 less likely to be the cause of the observed effects (so not a confounder), but may complicate a
9 multi-pollutant model because of their co-linearity with PM.
10 Another issue is whether the mean or median PM or co-pollutant concentration is the best
11 covariate in a second-stage model. An indicator of high-level concentrations might be
12 informative, e.g., the mean or median of 95th percentiles for each year (approximately the 3rd
13 largest of 61 annual every-6th-day observations for PM and the 18th largest for 365 daily measures
14 of the co-pollutants).
15
16 8.4.2.4.2 Regression of Effect Size Coefficients on Co-Pollutant vs. PM Coefficients in a
17 Multi-City Study
18 Several authors (Samet et al., 2000b; Schwartz, 1999, 2000a) have applied two-stage
19 regression techniques in an effort to identify the extent to which a pollutant is directly associated
20 with human health effects, as opposed to acting indirectly through its association with other
21 pollutants. Noting that relationships between co-pollutants differ by city, it has been proposed
22 that such differences can facilitate the separation of direct and indirect effects through use of a
23 second-stage meta-regression. The second-stage meta-regression approach regresses the city-
24 specific PM regression coefficients (which may have been adjusted for individual-level or time-
25 varying covariates in a previous stage) against the city-specific correlations between PM and a
26 selected co-pollutant. Typically the second-stage regression model is limited to a single
27 independent variable, and the regression results are presented graphically along with the data
28 points. For the results of the second-stage regression, a non-zero slope is taken as evidence of
29 confounding by the co-pollutant, while a non-zero intercept is taken as evidence of a true,
30 unconfounded association with PM.
31 Marcus and Kegler (2001) demonstrate that the use of the intercept as an indicator an
32 unconfounded association will only work if the second-stage model has been correctly specified.
April 2002 8-205 DRAFT-DO NOT QUOTE OR CITE
-------
1 Their counter-example is simply the case where the PM association is independently confounded
2 by two co-pollutants and only one of these confounders is included in the second-stage regression
3 model. In this case, the PM association would still be confounded by the omitted co-pollutant
4 and a non-zero intercept from the mis-specified model would not be valid evidence of an
5 unconfounded association with PM.
6 The counter-example is an illustration of residual confounding which is already well-
7 understood by epidemiologists. Residual confounding may arise (1) when a regression model
8 fails to include all of the potential confounders or (2) when a regression model includes a poorly
9 measured covariate that captures only a portion of the confounding by that covariate. The
10 approach to the evaluation of confounding proposed by Schwartz and Samet et al. could be
11 improved by the simple expedient of use of a multi-variate second-stage regression, in which the
12 model includes several gaseous co-pollutants. Even a multi-variate model would not fully
13 resolve the issue of residual confounding in the case of (a) poorly measured covariates or (b)
14 other omitted covariates. In either of these cases, the use of a non-zero intercept as evidence of a
15 true, unconfounded association with PM would be incorrect.
16 However, the most important aspect of the approach proposed by Schwartz and Samet et al.
17 was not their use of the intercept, but rather their use of the second-stage slope as an indication of
18 both the presence and magnitude of confounding by the modeled co-pollutant. The simulations
19 illustrated in the article by Marcus and Kegler clearly demonstrate that a confounding
20 co-pollutant would produce a non-zero slope, even in the presence of omitted confounders.
21 In addition to confounding, a non-zero slope might, under certain circumstances, also be
22 evidence of effect modification of a true PM association by the modeled co-pollutant.
23 Conversely, the absence of a slope in the second-stage regression could be evidence that the
24 particulate matter association is neither confounded nor modified by the modeled co-pollutant, or
25 that positive slopes could be obscured by negative slopes in relationships among co-pollutants,
26 suggested by the observation that the longitudinal correlations between PM10 and O3 or between
27 NO2 and O3 are often negative.
28
29
30
April 2002 8-206 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.2.5 Assessment of Confounding Based on Exposure
2 8.4.2.5.1 Review of Sarnat et al. (2001): Is there significant personal exposure to gaseous
3 co-pollutants?
4 A direct method for evaluating whether a putative causal factor is confounded by another
5 factor is based on the requirements that a confounder be (1) associated with the health outcome
6 or disease, and (2) associated with exposure to the putative causal factor. If individuals are not
7 exposed to a potential confounder, then it cannot be a confounder of another agent to which the
8 individual is exposed, although there is no guarantee that the putative causal factor causes the
9 outcome unless there is evidence of biological effects at the levels to which the individual is
10 exposed. The most likely potential confounders in air pollution epidemiology studies are the
11 gaseous criteria pollutants, specific size components or chemical components of particles, and
12 meteorological variables associated with exposure to PM or other pollutants. Many of these
13 potential confounding factors have been shown to cause adverse cardiovascular or respiratory
14 effects after exposure to elevated levels in laboratory animal studies, in in vitro experimental
15 studies, or as small physiological or functional changes in human adult volunteers. There is also
16 evidence for toxic effects from either short-term or long-term exposures to other ambient air
17 pollutants with which the criteria pollutants are associated. Finally, while extremes of
18 temperature and humidity are known to be independently associated with increases in mortality
19 and morbidity, they are also associated with concentrations and possibly even exposures (e.g., by
20 closing the windows and using air conditioners) to these pollutants, as well as ambient particles.
21 Thus, the gaseous co-pollutants and other environmental variables cannot be totally precluded as
22 confounders on the basis of lack of effects independent of PM.
23 The question raised in two important papers by Sarnat et al. (2000, 2001) addresses the
24 exposure aspect of confounding, i.e., are the gaseous pollutants confounders or surrogates of PM
25 effects? The first study (Sarnat et al., 2000) enrolled 15 non-smoking elderly participants (age
26 65+, average 75 years) in Baltimore, MD, who wore multi-pollutant personal samplers during the
27 summer of 1998 and the winter of 1999. Selection of participants was non-random, as they all
28 were healthy (i.e., asymptomatic) non-smokers, living in private residences, all with central air
29 conditioning but one (denoted SA4). The participants came from a range of socio-economic
30 backgrounds and locations within Baltimore (details not reported). The ambient pollutant
31 concentrations were measured at seven state or federal monitoring sites (shown in Chang et al.,
32 2000, Figure 2), five in the city of Baltimore and two in suburban counties, all within nine miles
April 2002 8-207 DRAFT-DO NOT QUOTE OR CITE
-------
1 of the central business district site denoted CBD1. Longitudinal within-subject correlations of
2 personal and ambient PM2 5 were high in the summer (median Spearman r = 0.74) and low in the
3 winter (median r = 0.25), and residential ventilation was an important determinant of this
4 association (high in well-ventilated environments, low in poorly-ventilated environments). The
5 highest association, on average, was between personal and ambient SO4=, given that SO4= has few
6 indoor sources. There were successively smaller mean correlations between personal and
7 ambient concentrations for PM2 5, PM10, O3, NO2, and PM10_2 5 in the summer, and PM10, PM2 5,
8 PM10_2 5, O3, NO2, SO2 in the winter, as shown in Table 3-6 of Sarnat et al. (2000). The lower
9 correlations between personal and ambient concentrations of the gaseous pollutants may reflect
10 both the much greater spatial heterogeneity of the gaseous pollutants (except possibly for ozone),
11 and the fact that many gaseous pollutant concentrations were below the seasonal detection limit
12 in Baltimore, producing negative median concentrations for some participants.
13 The study design should be considered in some detail. Each study period was divided into
14 three 12-day segments (denoted A, B, C) , with participants providing n = 9 to 12 days of
15 personal exposure data from which the longitudinal correlation coefficients for each participant
16 were calculated. The participants in each successive 12-day block or group were different
17 individuals with different residence locations, possible patterns of behavior, and household
18 characteristics that may have affected exposure. The number of participants (N) in each block is
19 shown in Table 8-37. One might hypothesize no difference among blocks in this small sample.
20 The results presented in Sarnat et al. (2000) have been aggregated over these three waves or
21 blocks. In spite of the admittedly very small numbers, one might ask whether the aggregation or
22 pooling across blocks is appropriate, given that the participants in each block are independent of
23 each other. The evaluate this, we transformed the correlation data to more nearly normally
24 distributed observations with constant variance, as if the Spearman correlation coefficient r was a
25 Pearson coefficient, using the Fisher Z-transformation, Z = 0.5 (ln(l + r ) - ln(l - r )). We then
26 tested the hypothesis that the mean values of the transformed personal to ambient correlations
27 were the same in blocks A, B, C on average, using a standard analysis of variance test.
28 There were no significant differences among the correlation between summertime personal
29 exposure and ambient air pollution among blocks A, B, C for pollutants that are believed to have
30 a reasonably uniform spatial distribution in summer, including SO4=, PM25, PM10, and O3. The
31 PM10_2 5 difference is also very non-significant if SC5 is included in the data, but becomes
April 2002 8-208 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-37. NUMBER OF PARTICIPANTS, N, IN EACH BLOCK FOR THE
EXPOSURE STUDY IN SARNAT ET AL. (2000)
Summer 1998
Block
A
B
C
Total
Approximate Dates
June 3 0 - July 12
July 14 - July 25
July 27 - August 7
June 30 - August 7
Number N
O
6
5
14*
Winter 1999
A
B
C
Total
February 2 - February 13
February 16 - February 27
March 2 -March 13
February 2 - March 13
4
4
6
14*
One of the 15 participants was excluded because of high exposure to environmental tobacco smoke outside
the residence.
1 statistically significant when PM10_2 5 from this participant is excluded. Differences among the
2 correlation between summertime personal exposure and ambient NO2 among blocks A, B, C is
3 nearly significant, even though there is a larger within-block variance for the other pollutants,
4 consistent with NO2 having a somewhat non-uniform spatial distribution.
5 We note significant between-block differences among the correlation between wintertime
6 personal exposure and ambient air pollution among blocks A, B, C for pollutants that had a
7 reasonably uniform spatial distribution in summer, including SO4= and PM10 in spite of the small
8 numbers and large within-block variance of personal correlations. There was also greater
9 evidence for between-block differences in the wintertime correlation between total personal
10 exposure and ambient PM2 5 concentration (P = 0.11 for Z) than for the summertime correlation
11 (P = 0.24). There was little indication of wintertime temporal variability among the personal -
12 ambient correlations for NO2, SO2, and O3, the ambient concentrations of which were often
13 below the wintertime detection limit. The PM10_2 5 interblock difference in person al vs. ambient
14 correlation is also very non-significant. We believe it is useful to recognize that large differences
April 2002 8-209 DRAFT-DO NOT QUOTE OR CITE
-------
1 among people and temporal blocks of personal-ambient correlation coefficients for gases and
2 particles may require personal exposure studies with a sufficient number of participants and of
3 longer duration to establish those relationships with a high degree of statistical certainty. This
4 analysis would be meaningful only if one block was high and significant and another was low
5 and insignificant.
6 The correlation of personal PM25 exposure to ambient concentrations of other pollutants
7 was also reported in Table 7 of Sarnat et al. (2000). The median correlation of personal PM25
8 exposure with ambient PM25 was higher than the correlation of personal exposure to PM2 5 with
9 any of the ambient concentrations of PM10_25, O3, NO2 in summer, but actually smaller in winter
10 with PM10_25 and NO2. The correlation of winter PM25 exposure and ambient O3 is very negative.
11 It would also have been useful to have ambient measurements at locations near the
12 participants' residences, in order to determine whether the difference in strength of association is
13 related to the heterogeneity in the spatial and temporal distribution of co-pollutant gases and
14 coarse particles relative to central site monitors versus the comparative (but not absolute)
15 homogeneity of PM25 measurements, a "measurement error" problem reviewed in
16 Sections 8.4.5.3 and 8.4.7.2. Finally, it would have also been desirable to have reported the
17 correlations between personal exposure to each of the gaseous co-pollutants versus the ambient
18 concentration of PM25 for each participant, analogous to Table 6. This would have allowed a
19 more direct comparison of the hypothesis in Sarnat et al. (2001) reviewed next in
20 Section 8.4.2.5.2.
21
22 8.4.2.5.2 Review of Sarnat et al. (2001): Are gaseous pollutants confounders or surrogates?
23 The Sarnat et al. (2001) paper extends the results in Sarnat et al. (2000) to additional
24 cohorts and participants: (a) 21 healthy children, ages 9 to 13 years; (b) 15 individuals with
25 COPD, average age 65 years; and (c) a total of 20 older healthy adults of average age 75 years,
26 6 more than in the earlier paper. All participants were non-smokers who lived nonsmoking
27 private residences. Fourteen of the healthy adults participated in both summer and winter
28 sampling campaigns described above. The COPD cohort consisted of individuals with
29 physician-diagnosed COPD, with an average age younger than the healthy adult cohort. The
30 sampling plan is shown in Table 1 of Sarnat et al. (2001). Although the participants lived in
April 2002 8-210 DRAFT-DO NOT QUOTE OR CITE
-------
1 various parts of Baltimore city and County and had a range of socio-economic backgrounds, they
2 were not selected as a representative sample of susceptible sub-populations.
3 The Sarnat et al. publication reports on personal monitoring of 56 subjects for fine PM, O3,
4 NO2, and SO2 in comparison with ambient concentrations of these substances. For fine PM, the
5 personal measurements are associated with the ambient fine PM central-site measurements, O3
6 (negative in winter), NO2, CO (winter only), and SO2 (winter only and negative). For the gaseous
7 co-pollutants, the various personal measurements are not positively associated with the ambient
8 central-site measurements of the same gas. The authors conclude that the ambient, central-site
9 measurements of the gaseous co-pollutants may be surrogates for specific constituents of fine PM
10 rather than confounders.
11 Among the combined sample of 56 participants, the highest median correlation between
12 personal exposure and ambient concentration was for particle sulfates, a component of
13 predominately ambient origin. Sulfates had a summertime median Spearman correlation of 0.88,
14 13 correlations being significant out of 14 for older healthy adults, and a wintertime median
15 correlation of 0.71, 16 out of 29 being significant correlations including 14 healthy adults. The
16 median Spearman correlation between total personal exposure and ambient concentration was
17 also high for PM25, with a median Spearman correlation of 0.65, 13 significant correlations
18 among 24 healthy older adults and children combined, and a wintertime median correlation of
19 0.22, with 10 of 44 significant correlations using data combined from the three cohorts. Among
20 the gaseous co-pollutants, the personal-ambient correlation was highest for NO2, with 7 out of
21 44 significant correlations using data combined from the three cohorts.
22 The ambient pollutants are correlated as expected, with high positive summertime
23 correlations seen between the regionally more correlated pollutants PM2 5 and O3, between two
24 combustion products NO2 and CO, and a positive significant correlation between ambient PM2 5
25 and NO2 shown in Table 8-38. There are high negative wintertime correlations between the
26 regionally correlated pollutant O3 and the combustion products PM2 5, NO2, or CO, and high
27 positive correlations among the combustion products PM2 5, NO2, and CO.
28 Total personal exposure to PM2 5 and exposure to estimated ambient PM2 5 is not
29 significantly correlated with personal exposure to gases, except for NO2 exposure in summer.
30 Sarnat et al. (2001) expressed this as a linear regression, with personal PM2 5 the dependent
31 variable and personal exposure to NO2, O3, or SO2 as the independent variable, finding that:
April 2002 8-211 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-38. CORRELATIONS AMONG AMBIENT POLLUTANTS IN
BALTIMORE. SUMMERTIME CORRELATIONS IN UPPER RIGHT,
WINTERTIME IN LOWER LEFT. STATISTICALLY SIGNIFICANT SPEARMAN'S
CORRELATIONS ARE SHOWN AS UNDERLINED BOLD VALUES.
Pollutant
PM25
03
NO2
CO
SO2
PM25
1
-0.72
0.75
0.69
-0.17
03
0.67
1
-0.71
-0.67
0.41
NO2
0.37
0.02
1
0.76
-0.17
CO SO2
0.15 —
-0.06 —
0.75 —
1
-0.12 1
Source: Based on Table 3 in Sarnat et al. (2001).
1 (Personal exposure to PM2 5) = 18.65 + 0.18 (Personal exposure to NO2) (8-4)
2
3 in summer, with both slope and intercept terms statistically significant and NO2 measured in
4 units of parts per billion (ppb). It is likely that the relative measurement error of PM25 exposure
5 is smaller—possibly much smaller - than that of personal exposure to the gaseous pollutants, as
6 shown from Tables 1, 4, and 5 in Sarnat et al. (2000). Among the 14 healthy older adults in the
7 earlier study, only 3 had mean summer exposure concentrations for O3 greater than the
8 summertime limit of detection (LOD) of 6.6 ppb (SC2, SC4, SC5, all in block C), only 3/14 of
9 the mean NO2 personal exposures below the LOD (SA1, SA2, SB3) of 5.5 ppb, but all of the
10 mean PM25 exposure concentrations were 5 to 12 times larger than the summertime LOD of 2.6.
11 None of the wintertime mean O3 exposure concentrations is larger than the LOD, and 5 of the
12 6 mean O3 exposures in block C are negative. Only 2 of 14 wintertime mean NO2 exposures is
13 smaller than the LOD (WAI, WB1) of 11.7 ppb, whereas all of the PM25 personal exposures
14 means are above the LOD, some by a factor of 12 to 13.
15 Figure 2 in Sarnat et al. (2001) shows box plots of the distribution of Spearman correlations
16 between personal and ambient concentrations for individual participants. In the summer, the
17 median correlation between personal O3 and ambient O3 is quite low (left-hand box) and only one
18 correlation is statistically significant, much lower than the median correlation between personal
April 2002 8-212 DRAFT-DO NOT QUOTE OR CITE
-------
1 O3 and ambient PM2 5 (right-hand box) where five correlations are statistically significant.
2 However, while the summertime median correlation between personal NO2 and ambient NO2 is
3 also quite low (left-hand bar) and only three correlations are statistically significant, it is not
4 much lower than the median correlation between personal NO2 and ambient PM2 5 for which four
5 correlations are significant. The median correlations are, of course, much higher for ambient
6 PM2 5 vs. ambient O3 or NO2. Thus, ambient PM2 5 may be a better proxy for personal exposure
7 to O3 than is ambient O3. However, personal exposure to NO2 is almost as well correlated with
8 ambient NO2 as with ambient PM25. If it is believed that exposure to NO2 also causes adverse
9 health effects, along with exposure to PM2 5, then it is not clear that ambient PM2 5 is merely a
10 proxy for ambient NO2.
11 In the winter, the median correlation between personal O3 and ambient O3 is quite low
12 (left-hand box) and no correlations are statistically significant, whereas the wintertime median
13 correlation between personal O3 and ambient PM25 (right-hand box) where seven correlations are
14 statistically significant and negative. While the wintertime median correlation between personal
15 NO2 and ambient NO2 is also quite low (left-hand bar), six correlations are positive and
16 statistically significant, the median personal-ambient NO2 correlation is not much lower than the
17 median correlation between personal NO2 and ambient PM2 5 for which four correlations are
18 significant. The median correlations are, of course, much higher for ambient PM2 5 vs. ambient
19 O3 or NO2. Thus, ambient PM2 5 may be a better proxy for personal exposure to O3 than is
20 ambient O3. However, personal exposure to NO2 is about as well correlated with ambient NO2 as
21 with ambient PM25. If it is believed that exposure to NO2 also causes adverse health effects,
22 along with exposure to PM2 5, then it is not clear that ambient PM2 5 is merely a proxy for ambient
23 NO2. Wintertime personal exposure to SO2 tends to be negatively associated with both ambient
24 SO2 and ambient PM2 5, with similar median correlations, but with one significantly negative
25 correlation against ambient SO2 vs. four significantly negative correlations against ambient
26 PM2 5. Thus, ambient PM2 5 may be a surrogate for personal SO2 exposure.
27 CO personal exposures have little correlation with ambient PM2 5 in summer, with only two
28 statistically significant correlations, but may be more strongly associated with ambient PM2 5 in
29 winter, showing five positive and one negative significant correlation and a larger positive
30 median correlation. However, there is no distribution of correlations of personal vs. ambient CO
31 with which to compare these findings.
April 2002 8-213 DRAFT-DO NOT QUOTE OR CITE
-------
1 Personal exposures to O3 are more positively correlated with personal exposures to PM2 5 in
2 summertime, and more negatively correlated in wintertime, than are personal O3 exposures with
3 ambient O3, and the associations of personal O3 with ambient PM25 even more so, suggesting that
4 ambient PM2 5 may be a good surrogate for personal O3. On the other hand, summertime
5 personal NO2 is no more closely associated with personal PM2 5 than with ambient NO2, and only
6 slightly less so than with ambient PM2 5, so that even if personal PM2 5 measurements were
7 available, one would expect them to be no more informative about NO2 personal exposures than
8 would ambient NO2. Wintertime personal NO2 is not associated with personal PM2 5, and about
9 equally associated with ambient NO2 and with ambient PM2 5, so that even if personal PM25
10 measurements were available, one would expect them to be completely uninformative about NO2
11 personal exposures compared to ambient NO2.
12 Table 9 in Sarnat et al. (2001) contains extensive results on the wintertime linear
13 relationships of personal total PM25 exposure, exposure to estimated PM25 of ambient origin,
14 personal SO4= exposure, and personal elemental carbon (EC) exposure, versus ambient
15 concentrations of the gaseous co-pollutants. Many of these are statistically significant: negative
16 relationships of exposure to all of these particle components vs. ambient O3, positive
17 relationships of personal exposure to EC and to estimated PM25 of ambient origin versus NO2 or
18 CO, and negative relationships of exposure to SO4= and estimated PM2 5 of ambient origin versus
19 ambient SO2. However, as noted above, and as done by the authors in Figure 2, comparison of
20 personal exposures of the gaseous co-pollutants to these particle components might be more
21 useful in evaluating the question proposed by the authors: are gases confounders or surrogates of
22 fine particles or fine particle components?
23
24 8.4.2.5.3 Confounding of co-pollutant effects arising from the spatial distribution of particles
25 and gases.
26 We focus here mainly on the question of co-pollutant exposures as a potential confounder
27 of PM exposure, identified in the previous sections. There is a component of total exposure to
28 NO2, CO, and other pollutants derived from ambient air most easily detected in the vicinity of
29 strong sources, such as the large number of automobiles and heavy-duty vehicles on major
30 highways or trunk roads near the exposed populations. The link to human health effects, if any,
31 would require satisfying three steps: (a) co-pollutant concentrations are high near certain line or
32 point sources, and decrease with increasing distance from the source much more rapidly than
April 2002 8-214 DRAFT-DO NOT QUOTE OR CITE
-------
1 does the PM10 or PM2 5 concentration decrease with increasing distance; (b) humans residing near
2 strong sources are more highly exposed to co-pollutants than those living farther away; and
3 (c) increased risks of adverse health effects attributable to the co-pollutants occur in proximity to
4 stronger sources of the co-pollutants. For this reason, one may attribute some the difficulties in
5 interpreting the findings of multi-pollutant PM epidemiology models including co-pollutants as
6 spatial "measurement errors" associated with the non-uniform distribution of the co-pollutants in
7 an urban area. Thus, there may be a reduced likelihood that central site co-pollutant monitors
8 will accurately characterize population exposure compared to the ability of central site PM
9 monitors to characterize PM population exposure.
10 Recent reviews about traffic-oriented concentration gradients and exposures to other
11 pollutants have been published by van Wijnen and van der Zee (1998) and by Monn (2001).
12 Location effects are particularly noticeable at the neighborhood level if there is a strong line
13 source (arterial street or freeway, for example) or point source (fossil-fuel-burning power plant,
14 for example) near the micro-environment. Some studies provide quantitative relationships
15 between distance from a heavily traveled roadway and concentrations of various pollutants (van
16 Wijnen and van der Zee, 1998).
17 The spatial correlations for PM2 5 were generally higher than for those of PM10 in the
18 PTEAM Riverside study (Clayton et al., 1993; Wallace, 1996) and in Philadelphia (Burton et al.,
19 1996; Wilson and Suh, 1997). However, larger spatial variations may occur for particles with
20 important local sources, such as highways carrying a large number of diesel trucks. Where PM10
21 is dominated by coarse particles, substantial variations (±20%) occurred between pairs of
22 monitors within 4 to 14 km in California's San Joaquin Valley (Blanchard et al., 1999). Several
23 European studies have found modest variations in ambient PM10 concentrations for residences
24 and housing close to roadways (Kingham et al., 2000, for Huddersfield, U.K.; Monn et al., 1997,
25 for Zurich), but large differences in NO2 concentrations occurred within a few meters of a Swiss
26 street during summer (Monn et al., 1997). The Monn results are shown as Figure 8-27. The
27 location and behavior of participants in a personal exposure study, as well as the spatial aspects
28 of socioeconomic differences in exposure (Rotko et al., 2000), may be important in defining
29 differences in exposure among sub-populations in an epidemiology study.
30 Numerous studies on personal exposure to airborne particles are discussed in detail in
31 Chapter 5. There is little doubt that elevated ambient concentrations of sulfates and fine particles
April 2002 8-215 DRAFT-DO NOT QUOTE OR CITE
-------
50
40
o
to
§ 30
O
20
. N02 Summer
PM10 Summer
20
40
J_
60 80
Figure 8-27. Concentration of PM10 and NO2 versus distance.
Source: Monn et al. (2000).
1 are closely related to elevated personal exposures to fine particles of ambient origin and to
2 elevated personal exposure to sulfates. In general, concentrations of ambient fine particles and
3 sulfates are more closely correlated to distance from a major highway or other sources than are
4 PM10_2 5, NO2, and CO, whose concentrations decrease with increasing distance. Rotko et al. also
5 reported that PM2 5 was more uniformly distributed in Helsinki than was NO2.
6 Janssen et al. (2001) evaluated personal indoor and outdoor NO2 and PM2 5 concentrations
7 at 24 schools located within 400 m of 22 different stretches of freeway in the Netherlands.
8 Indoor PM2 5 exposure was correlated with the distance from the school to the freeway and was
9 moderately correlated with the truck traffic volume, but not with the total or car traffic volume.
10 Indoor NO2 concentration was significantly associated with car traffic volume and with percent
11 of time downwind, but not with distance from the freeway or with truck traffic volume. Outdoor
12 NO2 concentration was significantly correlated only with percent of time downwind from the
13 freeway. PM25 concentrations indoors and outdoors were both significantly correlated with truck
14 traffic and distance from the freeway, but not with car traffic (at P < 0.05) or downwind
15 percentage.
April 2002
8-216
DRAFT-DO NOT QUOTE OR CITE
-------
1 Traffic-oriented health effects were reviewed by Wjst. et al. (1993) and by van Wijnen and
2 van der Zee (1998). New studies have appeared since then, including those by Venn et al. (2001)
3 and Roemer and van Wijnen (2001). The study by Roemer and van Wijnen is notable for three
4 reasons: (i) the endpoint is total mortality, as a more serious outcome than respiratory symptoms
5 in children; (ii) the populations in the study were divided into a "traffic" population living along
6 roads with traffic volume greater than 10,000 vehicles per day (about 10 percent of the total
7 population of Amsterdam) and a background population; (iii) measured air pollution
8 concentrations of BS, PM10, NO2, NO, CO, SO2, and O3 (8-hour mean) were available for the
9 background populations, and BS, NO2, NO, CO, SO2, and O3 for the "traffic" populations. All of
10 the pollutant concentrations except for O3 were higher for the "traffic" sites than for the
11 background sites. The excess risk rates in Table 3 in Roemer and van Wijnen (2001) for Black
12 Smoke (lags 1 and 2) were much higher and, for NO2 (lag 1) somewhat higher, than in the traffic
13 population. However, the statistically significant risk rates for the total population using the
14 "traffic" air pollution sites was lower than those using the background sites. No results were
15 reported for risk rates for the "traffic" population using traffic monitor sites, but the background
16 monitor concentrations were moderately correlated with those at the traffic monitoring sites.
17
18 8.4.2.6 Assessment of Confounding by Factor Analysis
19 How can one assess confounding in a single-city study if PM and its gaseous co-pollutants
20 are inextricably mixed in the urban atmosphere? One possibility is to make use of the correlation
21 structure of the data by extracting its principal components, or factors if the principal components
22 are rotated to provide a clearer picture of the main components. The principal components (p.c.)
23 or factors are linear combinations of the pollutant concentrations and are exactly independent for
24 p.c. and nearly so for rotated factors. An important advantage of this method is that there is no
25 problem of instability when all or most p.c. or factors are used in a multi-p.c. model. Several
26 variations of this approach have been used in PM epidemiology studies. Ozkaynak et al. (1996)
27 studied the relationship between mortality, air pollution, and weather using factor analysis
28 methods, where the factors were constructed from a particle index CoH, CO, and weather
29 variables. The analyses in Laden et al. (2000) and Tsai et al. (1999, 2000) used factors based on
30 chemical elements in fine particles for each day for which there was a fine particle filter sample.
31 Laden et al. (2000) identified up to seven sources in six cities, with some differences in the target
April 2002 8-217 DRAFT-DO NOT QUOTE OR CITE
-------
1 elements for sources across the cities. Three sources were present in all cities (mobile sources,
2 coal combustion, crustal particles). Poisson regression models were fitted to mortality data with
3 all source factors included simultaneously. Positive and statistically significant effects were
4 found for the motor vehicle and coal combustion sources in most cities, but not the crustal
5 source. These studies did not include gaseous pollutants.
6 Mar et al. (2000) developed a factor-analytic model for Phoenix, AZ, based on
7 12 components of particles and three gaseous pollutants (CO, NO2, SO2) measured on the same
8 day. They reported relative risks of total and cardiovascular mortality for five factors,
9 representing: motor vehicle exhaust and resuspended road dust; soil; vegetative burning; local
10 sources of SO2; and regional sulfate. The results reported in Tables 9 of Mar et al. (2000) are for
11 single-factor models. The authors state that regression analysis with all of the factors included in
12 a multisource model produced similar results.
13 This is a promising approach to analyzing multi-pollutant models and, in principle, could
14 be used in other studies with particles and gaseous criteria pollutants, even if little or no chemical
15 composition data were available. At the very least, evaluating the principal components of a
16 particle-gas mixture might help to identify which combinations of particles and gases are most
17 difficult to separate statistically in a regression analysis.
18
19 8.4.2.7 Simulation Analysis of Confounding
20 Since no single model specification can a priori be designated as "correct" in addressing
21 confounding effects of co-pollutants, discrepancies in results among studies, even for the same
22 dataset, are to be expected. While any assessment of relative "adequacy" of these alternative
23 model specifications is difficult with observational data, the implication of "inadequate" model
24 specifications may be studied through simulations using synthetic data in which the "correct"
25 model is known. Chen et al. (1999) conducted such simulations using a synthetic data set in
26 which the causal variables are known, and the effects of model misspecification were studied in
27 the presence of two variables (xx and x2), with varying levels of correlation, in a Poisson model.
28 They considered three situations: (1) model under/it, in which mortality was generated with both
29 Xj and x2, but regressed only on xt; (2) model overfit, in which mortality was generated with only
30 xl3 but regressed on both x1 and x2; and (3) model misfit., in which mortality was generated with
31 either xt or x2 but regressed on the other variable. They observed that the confounding of
April 2002 8-218 DRAFT-DO NOT QUOTE OR CITE
-------
1 covariates in an overfltted model does not bias the estimated coefficients but does reduce their
2 significance; and that the effect of model underfit or misfit leads not only to erroneous estimated
3 coefficients but also to erroneous significance. Based on these observations, Chen et al.
4 suggested that "models which use only one or two air quality variables (such as PM10 and SO2)
5 are probably unreliable, and that models containing several correlated and toxic or potentially
6 toxic air quality variables should also be investigated...". While conceptually useful, this
7 simulation study ignored one factor that is crucial in evaluating the implication of confounding,
8 the relative error. For example, including several correlated pollutants in a regression model may
9 lead to erroneous inferences, unless one considers the relative error associated with each of the
10 pollutants.
11
12 8.4.2.8 Discussion
13 In the preceding several section, a number of methods for evaluating the potential for
14 gaseous co-pollutants to confound particle effects were discussed. Multi-pollutant models may
15 be sensitive to multi-colinearity (high correlations among particle and gaseous pollutant
16 concentrations), and to so-called "measurement errors", possibly associated with spatial
17 variability. Combining multi-pollutant models across several cities may not improve the
18 precision of the mean PM effect size estimate combined, if the differences among the cities is as
19 large or larger in the multi-pollutant models as in the single-pollutant PM model. Second-stage
20 regressions have been useful in identifying effect modifiers in the NMMAPS and APHEA 2
21 studies, but may not, in general, provide a solution to the problem that confounding of effects is a
22 within-city phenomenon. Furthermore, the correlations among pollutants may change from
23 season to season and from place to place, suggesting that confounding as indicated by co-
24 linearity is not always the same.
25 Two promising approaches are also discussed, the first based on personal exposures to
26 particles and gases of three panels of participants in Baltimore, MD (Sarnat et al., 2000, 2001).
27 This directly addresses the premise that if individuals are not exposed to a potential confounder,
28 then it cannot really be a confounder of the presumed causal effect. While the results in this
29 paper support the conclusion that personal exposure to sulfates, fine particles, and PM10 are well
30 correlated with the corresponding fixed site ambient concentrations, the correlations are much
31 lower for PM10_2 5, O3, and NO2. There is however a great deal of variation from one of three
April 2002 8-219 DRAFT-DO NOT QUOTE OR CITE
-------
1 two-week panels from one season to the next. The sample size is small (N = 56), but did detect
2 marginally significant associations between personal and ambient NO2 for the personal-ambient
3 correlation, although much lower than for particles. There were, however, a number of
4 residences in which personal and ambient NO2 were highly correlated. This has been known to
5 happen in other studies when the residences are close to a major road, which describes several
6 members in the three cohorts (health elderly adults, adults with COPD, children 9-13 years.)
7 The other promising approach is to use principal component or factor analysis to determine
8 which combinations of gaseous criteria pollutants and PM size fractions or chemical constituents
9 together cannot be easily disentangled, and which pollutants are substantially independent of the
10 linear combinations of the others. For example, Mar et al. (2000) shows independent effects of
11 regional sulfate, motor vehicle-related particles, particles from vegetive burning, and PM10_25 for
12 cardiovascular mortality in Phoenix.
13
14 8.4.3 Role of Particulate Matter Components
15 In the 1996 PM AQCD, extensive epidemiologic evidence substantiated very well positive
16 associations between ambient PM10 concentrations and various health indicators, e.g., mortality,
17 hospital admissions, respiratory symptoms, pulmonary function decrements, etc.. A somewhat
18 more limited number of studies were then available which substantiated mortality and morbidity
19 associations with various fine particle indicators (e.g., PM2 5, sulfate, FT, etc.); and only one, the
20 Harvard Six Cities analysis by Schwartz et al. (1996a), evaluated relative contributions of the
21 fine (PM2 5) versus the coarse (PM10_2 5) fraction of PM10, with PM25 appearing to be associated
22 more strongly with mortality effects than PM10_2 5. Lastly, only a very few studies seemed to be
23 indicative of possible coarse particle effects, e.g., increased asthma risks associated with quite
24 high PM10 concentrations in a few locations where coarse particles strongly dominated the
25 ambient PM10 mix.
26
27 8.4.3.1 Fine- and Coarse-Particle Effects on Mortality
28 A greatly enlarged and still rapidly growing number of new studies published since the
29 1996 PM AQCD provide much new evidence further substantiating ambient PM associations
30 with increased human mortality and morbidity. As indicated in Table 8-1, most newly reported
31 analyses, with few exceptions, continue to show statistically significant associations between
April 2002 8-220 DRAFT-DO NOT QUOTE OR CITE
-------
1 short-term (24-h) PM concentrations and increases in daily mortality in many U.S. and Canadian
2 cities (as well as elsewhere). Also, the reanalyses of Harvard Six City and ACS study data
3 substantiate the original investigator's findings of long-term PM exposure associations with
4 increased mortality as well.
5
6 8.4.3.1.1 Effects on Total Mortality
1 The effects estimates from the newly reported studies are generally consistent with those
8 derived from the earlier 1996 PM AQCD assessment, which reported risk estimates for excess
9 total (nonaccidental) deaths associated with short-term PM exposures as generally falling within
10 the range of ca. 1.5 to 8.5% per 50 //g/m3 PM10 (24-h) increment and ca. 2.5 to 5.5% increase per
11 25 Mg/m3 PM2 5 (24-h) increment.
12 Several new PM epidemiology studies which conducted time-series analyses in multiple
13 cities were noted to be of particular interest, in that they provide evidence of effects across
14 various geographic locations (using standardized methodologies) and more precise pooled effect
15 size estimates with narrow confidence bounds, reflecting the typically much stronger power of
16 such multi-city studies over individual-city analyses to estimate a mean effect. Based on pooled
17 analyses across multiple cities, the percent total (non-accidental) excess deaths per 50 //g/m3
18 PM10 increment were estimated in different multi-city analyses to be: (a) 2.3% in the 90 largest
19 U.S. cities; (b) 3.4% in 10 large U.S. cities; (c) 3.5% in the 8 largest Canadian cities; and
20 (d) 2.0% in European cities.
21 Many new individual-city studies found positive associations (most statistically significant
22 at p < 0.05) for the PM2 5 fraction, with effect size estimates typically ranging from ca. 2.0 to ca.
23 8.5% per 25 //g/m3 PM25 for U.S. and Canadian cities. Of the 10 or so new analyses that not
24 only evaluated PM10 effects but also made an effort to compare fine versus coarse fraction
25 contributions to total mortality, only two are multi-city analyses yielding pooled effects
26 estimates: (a) the Klemm and Mason (2000) recomputation of Harvard Six Cities data,
27 confirming the original published findings by Schwartz et al. (1996a); and (b) the Burnett et al.
28 (2000) study of the 8 largest Canadian cities. Both of these studies found roughly comparable,
29 statistically significant excess risk estimates for PM2 5, i.e., approximately 3% increased total
30 mortality risk per 25 //g/m3 PM2 5 increment.
April 2002 8-221 DRAFT-DO NOT QUOTE OR CITE
-------
1 With regard to possible coarse particle short-term exposure effects on mortality, in those
2 new studies which evaluated PM10_2 5 effects as well as PM2 5 effects, the coarse particle (PM10_2 5)
3 fraction was also consistently positively associated with increased total mortality, albeit the
4 coarse fraction effect size estimates were generally less precise than those for PM2 5 and
5 statistically significant at p < 0.05 in only a few studies (Figure 8-6). Still, the overall picture
6 tends to suggest that excess total mortality risks may well reflect actual coarse fraction particle
7 effects, in at least some locations. This may be most consistently the case in arid areas, e.g., in
8 Mexico City, Santiago, Chile, or in the Phoenix area (as shown in Mar et al., 2000). On the other
9 hand, significant (or nearly significant) elevations in coarse PM-related total mortality risks have
10 also been detected for Steubenville, PH (an eastern U.S. urban area in the Harvard Six City
11 Study), as shown by Schwartz et al. (1996a). These results may reflect contamination of later-
12 resuspended coarse PM by metals in fine PM emitted from smelters (Phoenix) or steel mills
13 (Steubenville) that was earlier deposited on nearby soils. Excess total mortality risks associated
14 with short-term (24-h) exposures to coarse fraction particles capable of depositing in the lower
15 respiratory tract generally fall in the range of 0.5 to 6.0% per 25 //g/m3 PM10_25 increment for U.S.
16 and Canadian cities.
17 Three new papers provide particularly interesting new information on relationships between
18 short-term coarse particle exposures and total elderly mortality (age 65 and older), using
19 exposure TEOM data from the EPA ORD NERL monitoring site in Phoenix, AZ. Each used
20 quite different models but each reported statistically significant relationships between mortality
21 and coarse PM, specifically PM10_2 5, an indicator for the thoracic fraction of coarse-mode PM.
22 Smith et al. (2000), using a three-day running average as the exposure metric, performed
23 linear regression of the square root of daily mortality on the long-term trend, meteorological and
24 PM-based variables. Two mortality variables were used, total (non-accidental) deaths for the city
25 of Phoenix and the same for a larger, regional area. Using a linear analysis, effects based on
26 coarse PM were statistically significant for both regions, whereas effects based on fine PM
27 (PM2 5) were not. However, when the possibility of a nonlinear response was taken into account,
28 no evidence was found for a nonlinear effect for coarse PM, but fine PM was found to have a
29 statistically significant effect for concentration thresholds of 20 and 25 //g/m3. There was no
30 evidence of confounding between fine and coarse PM, suggesting that fine and coarse PM are
31 "essentially separate pollutants having distinct effects". Smith et al. (2000) also observed a
April 2002 8-222 DRAFT-DO NOT QUOTE OR CITE
-------
1 seasonal effect for coarse PM, the effect being statistically significant only during spring and
2 summer. Based on a principal component analysis of elemental concentrations, crustal elements
3 are highest in spring and summer and anthropogenic elements lowest, but Smith et al. (2000) felt
4 that the implication that crustal, rather than anthropogenic elements, were responsible for the PM
5 mortality was counterintuitive.
6 Clyde et al. (2000) used a more conventional model, a Poisson regression of log deaths on
7 linear PM variables; but they employed Bayesian model averaging to consider a wide variety of
8 variations in the basic model. They considered three regions: the Phoenix metropolitan area;
9 a small subset of zip code to give a region presumably with uniform PM25; and a still smaller zip
10 code region surrounding the monitoring site (thought to be uniform as to PM10 concentrations).
11 The models considered lags of 0, 1, 2, or 3 days but only for single day PM variables (no running
12 averages as used by Smith et al., 2000). A PM effect with a reasonable probability was found
13 only in the uniform PM2 5 region and only for coarse PM.
14 Mar et al. (2000) used conventional Poisson regression methods and limited their analyses
15 to the smallest area (called Uniform PM10 by Clyde et al.). They reported modeling data for lag
16 days 0 to 4. Coarse fraction PM was marginally significant on lag day 0. No direct fine particle
17 measures were statistically significant on day 0. A regional sulfate factor determined from
18 source apportionment, however, was statistically significant. No correlations were reported for
19 the source apportionment factors, but the correlation coefficient between sulfur (S) in PM25 (as
20 measured by XRF) with coarse fraction PM was only 0.13, suggesting separate and distinct
21 effects for regional sulfate and coarse fraction PM.
22 The above three studies of PM- total mortality relationships in Phoenix tend to suggest a
23 statistical association of coarse fraction PM with total elderly mortality in addition to and
24 different from any relationship with fine PM, fine PM components, or source factors for fine PM.
25 With regard to long-term PM exposure effects on total (non-accidental) mortality, the
26 newly available evidence from the HEI Reanalyses of Harvard Six Cities and ACS data (and
27 extensions, thereof), substantiate well associations attributable to chronic exposures to inhalable
28 thoracic particles (indexed by PM15 or PM10) and the fine fraction of such particles (indexed by
29 PM2 5 and/or sulfates). Statistically significant excess risk for total mortality was shown by the
30 reanalyses to fall in the range of 4-18% per 20 //g/m3 PM15/10 increment and 14-28% per
31 20 //g/m3 PM2 5 increase, thus suggesting likely stronger associations with fine versus coarse
April 2002 8-223 DRAFT-DO NOT QUOTE OR CITE
-------
1 fraction particles. Significant fine PM associations with total mortality were also found in the
2 latest reported AHSMOG results for males, but not in females.
3 Other recent studies on the relation of mortality to particle composition and source (Laden
4 et al., 2000; Mar et al., 2000; Ozkaynak et al., 1996; Tsai et al., 2000) suggest that particles from
5 certain sources may have much higher potential for adverse health effects than others, as
6 delineated by source-oriented evaluations involving factor analyses. Laden et al. (2000)
7 conducted factor analyses of the elemental composition of PM25 for Harvard Six Cities study
8 data for 1979-1988. In the analysis for all six cities combined, the excess risk for daily mortality
9 was estimated to be 3.4% (CI, 1.7 to 5.2) per 10 //g/m3 increment in a mobile source factor; 1.1%
10 (CI, 0.3 to 2.0) per 10 //g/m3 for a coal source factor, and -2.3% (CI, -5.8 to 1.2) per 10 //g/m3
11 for a crustal factor. There was large variation among the cities and some suggestion of an
12 association with a fuel oil factor identified by V or Mn, but it was not statistically significant.
13 Mar et al. (2000) applied factor analysis to evaluate mortality in relation to 1995-1997 fine
14 particle elemental components and gaseous pollutants (CO, NO2, SO2) in an area of Phoenix, AZ,
15 close to the air pollution monitors. The PM25 constituents included sulfur, Zn, Pb, soil-corrected
16 potassium, organic and elemental carbon, and a soil component estimated from oxides of Al, Si,
17 Ca, Fe, and It. Based on models fitted using one pollutant at a time, statistically significant
18 associations were found between total mortality and PM10, CO (lags 0 and 1), NO2 (lags 0, 1, 3,
19 4), S (negative), and soil (negative). Statistically significant associations were also found
20 between cardiovascular mortality and CO (lags 0 to 4), NO2 (lags 1 and 4), SO2 (lags 3 and 4),
21 PM25 (lags 1,3,4), PM10 (lag 0), PM10_2 5 (lag 0), and elemental, organic, or total carbon.
22 Cardiovascular mortality was significantly related to a vegetative burning factor (high loadings
23 on organic carbon and soil-corrected potassium), motor vehicle exhaust/resuspended road dust
24 factor (with high loadings on Mn, Fe, Zn, Pb, OC, EC, CO, and NO2), and a regional sulfate
25 factor (with a high loading on S). However, total mortality was negatively associated with a soil
26 factor (high loadings on Al, Fe, Si) and a local SO2 source factor, but was positively associated
27 with the regional sulfate factor.
28 Tsai et al. (2000) analyzed daily time series of total and cardiorespiratory deaths, using
29 short periods of 1981-1983 data for Newark, Elizabeth, and Camden, NJ. In addition to
30 inhalable particle mass (PM15) and fine particle mass (PM2 5), the study evaluated data for metals
31 (Pb, Mn, Fe, Cd, V, Ni, Zn, Cu) and for three fractions of extractable organic matter. Factor
April 2002 8-224 DRAFT-DO NOT QUOTE OR CITE
-------
1 analyses were carried out using the metals, CO, and sulfates. The most significant sources or
2 factors identified as predictors of daily mortality were oil burning (targets V, Ni), Zn and Cd
3 processing, and sulfates. Other factors (dust, motor vehicles targeted by Pb and CO, industrial
4 Cu or Fe processing) were not significant predictors. In Newark, oil burning sources and sulfates
5 were positive predictors, and Zn/Cd a negative predictor for total mortality. In Camden oil
6 burning and motor vehicle emissions predicted total mortality, but copper showed a marginal
7 negative association. Oil burning, motor vehicle emissions, and sulfates were predictors of
8 cardiorespiratory mortality in Camden. In Elizabeth, resuspended dust indexed by Fe and Mn
9 showed marginal negative associations with mortality, as did industrial sources traced by Cu.
10 The set of results from the above factor analyses studies do not yet allow one to identify
11 with great certainty a clear set of specific high-risk chemical components of PM. Nevertheless,
12 some commonalities across the studies seem to highlight the likely importance of mobile source
13 and other fuel combustion emissions (and apparent lesser importance of crustal particles) as
14 contributing to increased total or cardiorespiratory mortality.
15
16 8.4.3.1.2 Effects on Cause-Specific Mortality
17 Cardiovascular- and Respiratory-Related Mortality
18 Numerous new studies have evaluated PM-related effects on cause-specific mortality.
19 Most all report positive, often statistically significant (at p < 0.05), short-term (24-h) PM
20 exposure associations with cardiovascular (CVD)- and respiratory-related deaths. Cause-specific
21 effects estimates appear to mainly fall in the range of 3.0 to 7.0% per 25 //g/m3 24-h PM25 for
22 cardiovascular or combined cardiorespiratory mortality and 2.0 to 7.0% per 25 //g/m3 24-h PM2 5
23 for respiratory mortality in U.S. cities. Effect size estimates for the coarse fraction (PM10_25) for
24 cause-specific mortality generally fall in the range of ca. 3.0 to 8.0% for cardiovascular and ca.
25 3.0 to 16.0% for respiratory causes per 25 //g/m3 increase in PM10_25.
26 Also of particular interest, the above noted study by Mar et al. examined the associations of
27 a variety of PM indicators with cardiovascular mortality (for age >65), again in the zip code area
28 near the Phoenix monitoring site. For this end point, coarse PM was statistically significant on
29 lag day 0 but not on subsequent lag days. PM25 and a number of fine PM indicators were
30 statistically significant on lag day 1 but not on lag day 0. This suggests a distinct and separate
31 relationship of PM2 5 and PM10_2 5. As in the case of total mortality, the only fine PM indicator
April 2002 8-225 DRAFT-DO NOT QUOTE OR CITE
-------
1 found to be statistically significant on lag day 0 was regional sulfate. However, the low
2 correlation coefficient between S in PM2 5 and PM10_2 5 (r = 0.13) suggests that the two
3 relationships represent different sets of deaths. Thus, there is some evidence suggesting that the
4 risk of cardiovascular mortality , as well as that of total mortality, may be statistically associated
5 with PM10_2 5 and that this relationship may be independent of any relationships with fine particle
6 indicators.
7
8 Long- Term PM Exposure and Lung Cancer
9 Of particular interest with regard to PM-related effects on cause-specific mortality is a
10 growing body of evidence linking long-term PM exposure with increased risk of lung cancer.
11 Historical evidence has included studies of lung cancer trends, studies of occupational groups,
12 comparisons of urban and rural populations, and case-control and cohort studies using diverse
13 exposure metrics (Cohen and Pope, 1995). Table 8-39 (derived from Cohen, 2000) indicates
14 that, despite possible problems with respect to potential errors in exposure and other risk factor
15 measurement errors, numerous past ecological and case-control studies of PM and lung cancer
16 have generally indicated a lung cancer RR greater than 1.0 to be associated with living in areas
17 indicated as having higher PM exposures.
18 Prospective cohort studies offer a potentially more powerful approach to evaluate the
19 apparent association between PM exposures and the development of lung cancer. The 1996 PM
20 AQCD (U.S. Environmental Protection Agency, 1996a) summarized three of these more
21 elaborate studies that carefully evaluated the effects of PM air pollution exposure on lung cancer
22 using the prospective cohort design. In the Adventist Health Smog Study (AHSMOG), Abbey
23 et al. (1991) followed a cohort of Seventh Day Adventists, whose extremely low prevalence of
24 smoking and uniform, relatively healthy dietary patterns reduce the potential for confounding by
25 these factors. Excess lung cancer incidence was observed in females in relation to both particle
26 (TSP) and ozone exposure after 6 years follow up time. Dockery et al. (1993) reported the
27 results ofa 14-to 16-year prospective follow-up of 8,111 adults living in six U.S. cities that
28 evaluated associations between air pollution and mortality. After controlling for individual
29 differences in age, sex, cigarette smoking, BMI, education, and occupational exposure, Dockery
30 et al. (1993) found an elevated but non-significant risk for lung cancer (RR = 1.37; 95%CI = 0.81
31 to 2.31) for a difference in PM2 5 pollution equal to that of the most polluted versus the least
April 2002 8-226 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-39. SUMMARY OF PAST ECOLOGIC AND CASE-CONTROL
EPIDEMIOLOGIC STUDIES OF OUTDOOR AIR AND LUNG CANCER
Study Type Authors
Ecologic Henderson et al. ,
1975
Buffleretal.,
1988
Archer, 1990
Case-Control Pike et al., 1979
Vena, 1982
Jedrychowski,
etal., 1990
Katsouyanni,
etal., 1990
Barbone etal.,
1995
Nyberg et al.,
2000
Locale
Los Angeles, CA
Houston, TX
Utah
Los Angeles
Buffalo, NY
Cracow, Poland
Athens, Greece
Trieste, Italy
Stockholm,
Sweden
Exposure Classification
High PAH Areas
TSP by Census Tract
TSP by county
BAP Geo. Areas
TSP Geo. Areas
TSP and SO2
Geo. Areas
Soot Concentration
Geo. Areas
High Particle
Deposition Areas
High NO2 Areas
Rate Ratio (95% CI)
1.3@96-116ug/m3TSP
(CI: N/A)
1.9@16ug/m3TSP
(CI: N/A)
1.6@85ug/m3TSP
(CI: N/A)
1.3 @96-116ug/m3TSP
1. 7 @ 80-200 ug/m3 TSP
(CI: 1.0-2.9)
1.1@TSP> 150 ug/m3
(CI: N/A)
I.l@sootupto400
ug/m3
(CI: N/A)
1.4@>0.3g/m2/day
(CI: 1.1-1.8)
1.3
(CI: 0.9-1.9)
Source: Cohen (2000).
1 polluted city. Pope et al. (1995) similarly analyzed PM2 5 and sulfate (SO4) air pollution as
2 predictors of mortality in a prospective study of 7-year survival data (1982 to 1989) for about
3 550,000 adult volunteers obtained by the American Cancer Society (ACS). Both the ACS and
4 Harvard studies have been subjected to much scrutiny, including an extensive independent audit
5 and re-analysis of the original data (Krewski et al., 2000) that confirmed the originally published
6 results. The ACS study controlled for individual differences in age, sex, race, cigarette smoking,
7 pipe and cigar smoking, exposure to passive cigarette smoke, occupational exposure, education,
8 BMI, and alcohol use. Lung cancer mortality was significantly associated with particulate air
9 pollution when SO4= was used as the index,, but not when PM2 5 mass was used as the index for a
10 smaller subset of the study population that resided in metropolitan areas where PM2 5 data were
11 available from the Inhalable Particle (IP) Network. Thus, while these prospective cohort studies
April 2002
8-227
DRAFT-DO NOT QUOTE OR CITE
-------
1 have also indicated that long-term PM exposure is associated with an increased cancer risk, the
2 effect estimates were generally not statistically significant, quite possibly due to inadequate
3 statistical power by these studies at that time (e.g., due to inadequate population size and/or
4 follow-up time for long-latency cancers).
5 The AHSMOG investigators have re-examined the association between long-term PM
6 exposure and increased risk of both lung cancer incidence and lung cancer mortality in
7 nonsmokers using longer-term follow-up of this cohort and improved analystical approaches.
8 Beeson et al. (1998) considered this cohort of some 6,338 nonsmoking, non-Hispanic, white
9 Californian adults, ages 27-95, that was followed from 1977 to 1992 for newly diagnosed
10 cancers. Incident lung cancer in males was positively and significantly associated with IQR
11 increases for mean concentrations of PM10 (RR = 5.21; 95% CI =1.94-13.99). For females in the
12 cohort, incident lung cancer was positively associated with Inter-Quartile Range (IQR) increases
13 for SO2(RR = 2.14; CI, 1.36-3.37) and IQR increases for PM10 exceedance frequencies of 50
14 ug/m3 (RR =1.21; 95% CI = 0.55-2.66) and 60 ug/m3 (RR = 1.25; 95% CI = 0.57-2.71). Thus,
15 increased risks of incident lung cancer were deemed by the authors to be associated with elevated
16 long-term ambient concentrations of PM10 and SO2 in both genders. The higher PM10 risk effect
17 estimate for cancer in males appeared to be partially due to gender differences in long-term air
18 pollution exposures. Abbey et al. (1999) also related long-term ambient concentrations of PM10,
19 SO4=, SO2, O3, and NO2 to 1977-1992 mortality in the AHSMOG cohort. After adjusting for a
20 wide range of potentially confounding factors, including occupational and indoor sources of air
21 pollutants, PM10 showed a strong association with lung cancer deaths in males (PM10 IQR
22 RR=2.38; 95% CI: 1.42 - 3.97). In this cohort, males spent more time outdoors than females,
23 thus having higher estimated air pollution exposures than the cohort females. Ozone showed an
24 even stronger association with lung cancer mortality for males, and SO2 showed strong
25 associations with lung cancer mortality for both sexes. The authors reported that other pollutants
26 showed weak or no association with mortality. Therefore, increases in both lung cancer
27 incidence and lung cancer mortality in the extended follow-up analysis of the AHSMOG study
28 were found to be most consistently associated with elevated long-term ambient concentrations of
29 PM10 and SO2, especially among males.
30 A recent follow-up analysis of the major ACS study by Pope et al. (2002) responds to a
31 number of criticisms previously noted for the earlier ACS analysis (Pope et al., 1995) in the 1996
April 2002 8-228 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM AQCD (U.S. Environmental Protection Agency, 1996a), most notably by including
2 examinations of other pollutants, better occupational indices, and diet information, while also
3 addressing possible spatial auto-correlations due to regional location. The recent extension of the
4 ACS study includes approximately 500,000 adult men and women drawn from ACS-CPS-II
5 enrollment and follow-up during 1982-1998. This new analysis of the ACS cohort substantially
6 expands the prior analysis, including: (1) a more than doubling of the follow-up time to 16 years
7 (and a more than tripling of the number of deaths in the analysis); (2) substantially expanded
8 exposure data, including gaseous co-pollutant data and new PM2 5 data collected in 1999-2001;
9 (3) improved control of occupational exposures; (4) incorporation of dietary variables that
10 account for total fat consumption, as well as consumption of vegetables, citrus and high-fiber
11 grains; and (5) utilization of recent advances in statistical modeling, including the incorporation
12 of random effects and non-parametric spatial smoothing components in the Cox proportional
13 hazards model.
14 In this extended ACS analysis, it was found that long-term exposure to air pollution, and
15 especially to PM25, is associated with increased annual risk of mortality. With the longer 15-year
16 follow-up period and with improved metrics of PM25 exposures, this study for the first time
17 detected a statistically significant association between living in a city with higher PM25 and
18 increased risk of dying of lung cancer. Each 10 ug/m3 elevation in annual average fine PM was
19 associated with a 13 percent (95% CI=4%-23%) increase in lung cancer mortality. Coarse
20 particles and gaseous pollutants were generally not significantly associated with excess lung
21 cancer mortality. SO4= was significantly associated with mortality and lung cancer deaths in this
22 extended data set, yielding RR's consistent with (i.e., not significantly different from) the SO4=
23 RR's reported in the previously published 7-year follow-up (Pope et al, 1995). However, while
24 PM2 5 was specific to the causes most biologically plausible to be influenced by air pollution in
25 this analysis (i.e., cardio-pulmonary and cancer), SO4= was significantly associated with every
26 mortality category in this new analysis, including that for "all-other causes", This suggests that
27 the PM2 5 associations found are more biologically plausible than the less specific SO4=
28 associations found. The PM2 5 cancer risk appears greatest for non-smokers and among those
29 with lower socio-economic status (as indicated by lower educational attainment).
30 Overall, these new cohort studies confirm and strengthen the published older ecological and
31 case-control evidence indicating that living in an area that has experienced higher PM exposures
April 2002 8-229 DRAFT-DO NOT QUOTE OR CITE
-------
1 can cause a significant increase in the RR of lung cancer incidence and associated mortality.
2 In particular, the new ACS cohort analysis more clearly indicates that living in a city with higher
3 PM25 levels is associated with an elevated risk of lung cancer amounting to an increase of some
4 10 to 15% above the lung cancer risk in a cleaner city.
5 With regard to specific ambient fine particle constituents that may significantly contribute
6 to the observed ambient PM-related increases in lung cancer, PM components of diesel engine
7 exhaust represent one class of likely important contributors. Diesel emission PM typically
8 comprises a noticeable fraction of ambient fine particles in many urban areas, having been
9 estimated to comprise from approximately 5 to 35% of ambient PM25 in some U.S. urban areas
10 (see Chapter 3). Also, as discussed in a separate Health Effects Assessment of Diesel Engine
11 Exhaust (U.S. Environmental Protection Agency, 2002), extensive epidemiologic and toxicologic
12 evidence links diesel emissions (including fine PM components) to increased risk of lung cancer.
13
14 8.4.3.1.3 Shortening-of-Life Associated With Long-Term Ambient Particulate
15 Matter Exposure
16 The public health burden of mortality associated with exposure to ambient PM depends not
17 only on the increased risk of death, but also on the length of life shortening that is attributable to
18 those deaths. However, the 1996 PM AQCD concluded that confident quantitive determination
19 of years of life lost to ambient PM exposure was not yet possible; life shortening may range from
20 days to years (U.S. Environmental Protection Agency, 1996a). Now, some newly available
21 analyses provide further interesting insights with regard to potential life-shortening associated
22 with chronic PM exposures.
23
24 8.4.3.1.3.1 Life-Shortening Estimates Based on Semi-Individual Cohort Study Results
25 Brunekreef (1997) reviewed the available evidence of the mortality effects of long-term
26 exposure to PM air pollution and, using life table methods, derived an estimate of the reduction
27 in life expectancy implied by those effect estimates. Based on the results of Pope et al. (1995)
28 and Dockery et al. (1993), a relative risk of 1.1 per 10 //g/m3 exposure over 15 years was
29 assumed for the effect of PM air pollution on men 25-75 years of age. A 1992 life table for men
30 in the Netherlands was developed for 10 successive five-year categories that make up the
31 25-75 year old age range. Life expectancy of a 25 year old was then calculated for this base case
32 and compared with the calculated life expectancy for the PM-exposed case, where the death rates
April 2002 8-230 DRAFT-DO NOT QUOTE OR CITE
-------
1 were increased in each age group by a factor of 1.1. A difference of 1.11 years was found
2 between the "exposed" and "clean air" cohorts' overall life expectancy at age 25. Looked at
3 another way, this implies that the expectation of the lifespan for persons who actually died from
4 air pollution was reduced by more than 10 years, since they represent a small percentage of the
5 entire cohort population. A similar calculation by the authors for the 1969-71 life table for U.S.
6 white males yielded an even larger reduction of 1.31 years for the entire population's life
7 expectancy at age 25. Thus, these calculations imply that relatively small differences in long-
8 term exposure to ambient PM can have substantial effects on life expectancy.
9
10 8.4.3.1.3.2 Potential Effects of Infant Mortality on Life-Shortening Estimates
11 Deaths among children can logically have the greatest influence on a population's overall
12 life expectancy, but the Brunekreef (1997) life table calculations did not consider any possible
13 long-term air pollution exposure effects on the population aged <25 years. As discussed above,
14 some of the older cross-sectional studies and the more recent studies by Bobak and Leon (1992),
15 Woodruff et al. (1997), Bobak and Leon (1999), and Loomis et al. (1999) suggest that infants
16 may be among sub-populations notably affected by long-term PM exposure. Thus, although it is
17 difficult to quantify, any premature mortality that does occur among children due to long-term
18 PM exposure (as suggested by these new studies) would significantly increase the overall
19 population life shortening over and above that estimated by Brunekreef (1997) for long-term PM
20 exposure of adults aged 25 years and older.
21
22 8.4.3.2 PM10, PM2 5 (Fine), and PM10 2 5 (Coarse) Particulate Matter Effects on Morbidity
23 At the time of the 1996 PM AQCD, fine particle morbidity studies were mostly limited to
24 Schwartz et al. (1994) , Neas et al. (1994, 1995); Koenig et al. (1993); Dockery et al. (1996); and
25 Raizenne et al. (1996); and discussion of coarse particles morbidity effects was also limited to
26 only a few studies (Gordian et al., 1996; Hefflin et al., 1994) which implicated PM10_25 as a
27 possible important fraction of PM10. Since the 1996 PM AQCD, several new studies have been
28 published in which newly available size-fractionated PM data allowed investigation of the effects
29 of both fine (PM25) and coarse fraction (PM10_25) particles. Fine (FP) and coarse fraction (CP)
30 particle results are noted below for studies by morbidity outcome areas, as follows:
April 2002 8-231 DRAFT-DO NOT QUOTE OR CITE
-------
1 cardiovascular disease (CVD) hospital admissions (HA's); respiratory medical visits and hospital
2 admissions; and respiratory symptoms and pulmonary function changes.
3 As discussed in Section 8.3.1 (on cardiovascular effects associated with acute ambient PM
4 exposure), an extensive new body of evidence has emerged since the 1996 PM AQCD that
5 evaluates PM10 effects on cardiovascular-related hospital admissions and visits. Especially
6 notable new evidence has been provided by several new multi-city studies (Schwartz, 1999;
7 Samet et al., 2000a,b) that yield pooled estimates of PM-CVD effects across numerous U.S.
8 cities and regions. These studies found not only significant PM associations, but also
9 associations with other gaseous pollutants as well, thus hinting at likely independent effects of
10 certain gases (O3, CO, NO2, SO2) and/or interactive effects with PM. These and other individual-
11 city studies generally appear to confirm likely excess risk of CVD-related hospital admission for
12 U.S. cities in the range of 3-10% per 50 //g/m3 PM10, especially among the elderly (> 65 yr).
13 In addition to the PM10 studies, several new U.S. and Canadian studies evaluated fine-mode
14 PM effects on cardiovascular outcomes. Moolgavkar (2000a) reported PM2 5 to be significantly
15 associated with CVD HA for lag 0 and 1 in Los Angeles. Burnett et al. (1997a) reported that fine
16 particles were significantly associated with CVD HA in a single pollutant model, but not when
17 gases were included in multipollutant models for the 8 largest Canadian city data. Stieb et al.
18 (2000) reported both PM10 and PM25 to be associated with CVD emergency department (ED)
19 visits in single pollutant, but not multipollutant models. Similarly, Morgan et al. (1998) reported
20 that PM2 5 measured by nepholonetry was associated with CVD HA for all ages and 65+ yr, but
21 not in the multipollutant model. Tolbert et al. (2000a) reported that coarse particles were
22 significantly associated with dysrhythmias, whereas PM2 5 was not. Other studies (e.g., Liao
23 et al., 1999; Creason et al., 2001; Pope et al., 1999b,c) reported associations between increases in
24 PM25 and several measures of decreased heart rate variability, but Gold et al. (2000) reported a
25 negative association of PM25 with heart rate and decreased variability in r-MSSD (one heart rate
26 variability measure). A recent study by Peters and colleagues (2001) reported significant
27 temporal associations between acute (2-h or 24-h) measures of PM25 and myocardial infarction.
28 Overall, these new studies collectively appear to implicate fine particles, as well as possibly some
29 gaseous co-pollutants, in cardiovascular morbidity, but the relative contributions of fine particles
30 acting alone or in combination with gases such as O3, CO, NO2 or SO2 remain to be more clearly
31 delineated and quantified. The most difficult issue relates to interpretation of reduced PM effect
April 2002 8-232 DRAFT-DO NOT QUOTE OR CITE
-------
1 size and /or statistical significance when co-pollutants derived from the same source(s) as PM are
2 included in multipollutant models.
3 Section 8.3.1 also discussed U.S. and Canadian studies that present analyses of coarse
4 fraction particles (CP) relationships to CVD outcomes. Lippmann et al. (2000) found significant
5 positive associations of PM10_2 5 with ischemic heart disease hospital admissions in Detroit
6 (RR =1.10, CI 1.026, 1.18). Tolbert et al. (2000a) reported significant positive associations of
7 heart dysrhythmias with CP (p = 0.04) as well as for elemental carbon (p = 0.004), but these
8 preliminary results must be interpreted with caution until more complete analyses are carried out
9 and reported. Burnett et al. (1997b) noted that CP was the most robust of the particle metrics
10 examined to inclusion of gaseous covariates for cardiovascular hospitalization, but concluded
11 that particle mass and chemistry could not be identified as an independent risk factor for
12 exacerbation of cardiorespiratory disease in this study. Based on another Canadian study,
13 Burnett et al. (1999), reported statistically significant associations for CP in univariate models
14 but not in multipollutant models; but the use of estimated rather than measured PM exposures
15 indices limits the interpretation of the PM results reported.
16 The collective evidence reviewed above, in general, appears to suggest excess risks for
17 CVD-related hospital admissions of approximately 4.0 to 10% per 25 //g/m3 PM2 5 or PM10_2 5
18 increment.
19 Section 8.3.2 also discussed new studies of effects of short-term PM exposure on the
20 incidence of respiratory hospital admissions and medical visits. Several new U.S. and Canadian
21 studies have yielded particularly interesting results suggestive of roles of both fine and coarse
22 particles in respiratory-related hospital admissions. In an analysis of Detroit data, Lippmann
23 et al. (2000) found comparable effect size estimates for PM2 5 and PM10_2 5. That is, the excess
24 risk for pneumonia hospital admissions (in no co-pollutant model) was 13% (CI 3.7, 22) per
25 25 Mg/m3 PM25 and 12% (CI 0.8, 24) per 25 //g/m3 PM10_25. Because PM25 and PM10_25 were not
26 highly correlated, the observed association between coarse particles and health outcomes were
27 possibly not confounded by smaller particles. Despite the greater measurement error associated
28 with PM10_25 than with either PM2 5 and PM10, this indicator of the coarse particles within the
29 thoracic fraction was associated with some of the outcome measures. The interesting result is
30 that PM10_2 5 appeared to be a separate factor from other PM metrics, especially given the effect
31 estimates of PM10_25 with pneumonia hospital admissions (lag 1; RR= 1.11, 95% CI: 1.006,
April 2002 8-233 DRAFT-DO NOT QUOTE OR CITE
-------
1 1.233). Burnett et al. (1997b) also reported PM (PM10, PM2 5, and PM10.2 5) associations with
2 respiratory hospital admissions, even with O3 in the model. Notably, the PM10_25 association was
3 significant (RR =1.13 for 25 //g/m3; CI = 1.05 - 1.20); and inclusion of ozone still yielded a
4 significant coarse mass RR =1.11 (CI = 1.04 - 1.19). Moolgavkar et al. (2000) showed the most
5 consistent association for PM10 across lags (0-4d), while PM2 5 yielded the strongest positive PM
6 metric association at lag 3 days. Also, Moolgavkar (2000a) reported that, in Los Angeles, both
7 PM10 and PM2 5 yielded both positive and negative associations at different lags for single
8 pollutant models but not in two pollutant models. Delfmo et al. (1997) reported that both PM2 5
9 and PM10 are positively associated with ED visits for respiratory disease. Morgan et al. (1998)
10 reported that PM25 estimated from nephelometry yielded a PM25 association with COPD hospital
11 admissions for 1-hr max PM that was more positive than 24-h average PM2 5.
12 Some new studies appear to substantiate PM associations with asthma-related hospital
13 admissions. For example, Norris et al (1999) reported associations of emergency department
14 visits for asthma in children with both PM25 and PM10_25. Two other studies presented uniquely
15 different analyses of hospital admissions in the Seattle, Washington area. Sheppard et al. (1999)
16 studied relationships between PM metrics that included PM10_2 5 and non-elderly adult hospital
17 admissions for asthma in the greater Seattle area and reported significant relative rates for PM10,
18 PM25 and PM10.25 (lagged 1 day). For PM10.25, the relative risk was 1.04 (95% CI 1.01, 1.07).
19 In a different analysis, Lumley and Heagerty (1999) examined PMX and PM104 in the King
20 County, WA (Seattle) area during the same time period but for hospital admissions for overall
21 respiratory disease. Since only a significant hospital admission association was found with PMl 0
22 and not PM1(M, a dominant role by sub-micron particles in PM2 5 - asthma HA association was
23 suggested, but this may not be an appropriate conclusion based on several differences between
24 the study analysis methods and differences between asthma versus respiratory outcome measures
25 used in the two Seattle studies. For a 16% decrease in PM10 levels, Friedman et al. (2001)
26 reported decreased hospital admissions for asthmatics during the Olympics in Atlanta.
27 Several other studies (Chen et al. 2000; Choudhury et al., 1997; Moolgavlar 2000a;
28 Lippsett et al., 1997) report results for areas (e.g., Reno-Sparks, NV; Anchorage, AK; Phoenix,
29 AZ; Santa Clara, CA) where coarse fraction particles tend to constitute a large fraction of PM10
30 but no measures of PM10_2 5 were available. These studies showing significant PM10 effects on
31 respiratory hospital admissions provide additional data suggestive of likely coarse fraction
April 2002 8-234 DRAFT-DO NOT QUOTE OR CITE
-------
1 particle effects on respiratory morbidity. It is possible that vegetative burning (e.g., wood) in
2 these western cities may produce coarse particles whose toxicity may differ from that of coarse
3 crustal fraction particles.
4 Thus, although PM10 mass has most often been implicated as the PM pollution index
5 affecting respiratory hospital admissions, the overall collection of new studies reviewed in
6 Section 8.3.2 appear to suggest relative roles for both fine and coarse PM mass fractions, such as
7 PM2 5 and PM10.2 5.
8 Section 8.3.3 assessed relationships between PM exposure on lung function and respiratory
9 symptoms. While most data examine PM10 effects, several studies also examined fine and coarse
10 fraction particle effects. Schwartz and Neas (2000) report that cough was the only response in
11 which coarse fraction particles appeared to provide an independent contribution to explaining the
12 increased incidence. The correlation between CM and PM25 was moderate (0.41). Coarse
13 fraction particles had little association with evening peak flow. Tiittanen et al. (1999) also
14 reported a significant effect of PM10_25 for cough. Thus, cough may be an appropriate outcome
15 related to coarse fraction particle effects. However, the limited data base suggests that further
16 study is appropriate. The report by Zhang, et al. (2000) of an association between coarse fraction
17 particles and the indicator "runny nose" is noted also.
18 Published epidemiological studies have collectively indicated that exposure to PM air
19 pollution can be associated with adverse human health effects, and that asthmatics represent a
20 population that can be especially affected by acute exposures to air pollution (e.g., see Koren and
21 Utell, 1997). In particular, prospective epidemiologic studies of panels of individuals confirm
22 the air pollution-asthma exacerbation association.
23 For respiratory symptoms and PFT changes, several new asthma studies report relationships
24 with ambient PM measures. The peak flow analyses results for asthmatics tend to show small
25 decrements for both PM10 and PM2 5. Several studies included PM2 5 and PM10 independently in
26 their analyses of peak flow. Of these, Naeher et al. (1999), Tiittanen et al. (1999), Pekkanen et
27 al. (1997), and Romieu et al. (1996) all found comparable results for PM2 5 and PM10. The study
28 of Peters et al. (1997c) found slightly larger effects for PM25. The study of Schwartz and Neas
29 (2000) found larger effects for PM2 5 than for coarse fraction particles. Three studies included
30 both PM10 and PM25 in their analyses of respiratory symptoms. The studies of Peters et al.
31 (1997c) and Tiittanen et al. (1999) found similar effects for the two PM measures. Only the
April 2002 8-23 5 DRAFT-DO NOT QUOTE OR CITE
-------
1 Romieu et al. (1996) study found slightly larger effects for PM25. While the PM associations
2 with adverse health effects among asthmatics and others are well documented, the type/source(s)
3 of those particles most associated with adverse health effects among asthmatics are not known at
4 this time. Indeed, the makeup of PM varies greatly from place to place and over time, depending
5 upon factors such as the sources that contribute to the pollution and the prevailing atmospheric
6 conditions, affecting particle formation, coagulation, transformation, and transport. One
7 suspected causal PM agent is the fine particle component of diesel combustion exhaust.
8 Two studies (Delfino et al., 1998; Ostro et al., 2001) examined PM effects on asthmatics
9 using one hour maximum exposure measures by TEOM, and both studies indicate a relationship
10 with measures of respiratory symptoms. Further research is needed at these shorter exposure
11 times for different PM size fractions.
12 For non-asthmatics, several studies evaluated PM2 5 effects. Naeher et al. (1999) reported
13 similar AM PEF decrements for both PM2 5 and PM10. Neas et al. (1996) reported a
14 nonsignificant negative association for PEF and PM2 b and Neas et al. (1999) also reported
15 negative but nonsignificant PEF results. Schwartz and Neas (2000) reported a significantly PM
16 PEF association with PM2 5, and Tiittanen et al. (1999) also reported negative but nonsignificant
17 association for PEF andPM25. Gold et al. (1999) reported significantly PEF results. Schwartz
18 and Neas (2000) reported significant PM2 5 effects relative to lower respiratory symptoms.
19 Tiittanen et al. (1999) showed significant effects for cough and PM25 for a 4-day average.
20 Another study conducted by Peters et al. (1997c) in Erfurt, Germany in 1992 is unique for
21 two reasons: (1) they studied the size distribution in the range 0.01 to 2.5 //m and (2) examined
22 the number of particles. They report that the health effects of 5 day means of the number count
23 (NC) for ultrafme particles were larger than those related to the mass of the fine particles. For
24 NC 0.01 -0.1, cough was significant for the same day and the five day mean.
25 In a chronic respiratory disease study of 22-24 North American communities evaluated in
26 the 1996 PM AQCD, Raizenne et al. (1996) found PM2A to be related to a statistically significant
27 FVC deficit of -3.21% (-4.98, -1.41). Dockery et al. (1996) also reported PM2A associations
28 with increased bronchitis; odds ratio =1.50 (95% CI = 0.91, 2.47).
29 The above new studies offer much more information than was available in 1996. Effects
30 were noted for several morbidity endpoints: cardiovascular hospital admissions, respiratory
31 hospital admissions and cough. Still insufficient data exists from these relatively limited studies
April 2002 8-236 DRAFT-DO NOT QUOTE OR CITE
-------
1 to allow strong conclusions at this time as to which size-related ambient PM components may be
2 most strongly related to one or another morbidity endpoints. Very preliminarily, however, fine
3 particles appear to be more strongly implicated in cardiovascular outcomes than are coarse
4 fraction particles, whereas both seem to impact respiratory endpoints.
5
6 8.4.4 The Question of Lags
7 The effect of selecting lags on the resulting model for PM health effects is one of the main
8 issues in model selection. Using simulated data with parameters similar to a Seattle PM10_25 data
9 series, Lumley and Sheppard (2000) showed that the bias resulting from the selection is shown to
10 be similar in size to the relative risk estimates from the measured data. More precisely, the log
11 relative risk from the measured Seattle data is about twice the mean bias in the simulated control
12 data, and the published estimate of relative risk is only at the 90th percentile of the bias
13 distribution in these control analysis. The selection rule used was to choose the lag (between
14 0 and 6 day) with the largest estimated relative risk. In comparisons to real data from Seattle for
15 other years and from Portland, OR (with similar weather patterns to Seattle), similar bias issues
16 became evident.
17 In most of the past air pollution health effects time-series studies, after the basic model (the
18 best model with weather and seasonal cycles as covariates) was developed, several pollution lags
19 (usually 0 to 3 or 4 days) were individually introduced and the most significant lag(s) chosen for
20 the RR calculation. While this practice may bias the chance of finding a significant association,
21 without a firm biological reason to establish a fixed pre-determined lag, it appears reasonable.
22 Due to likely individual variability in response to air pollution, the apparent lags of effects
23 observed for aggregated population counts are expected to be "distributed" (i.e., symmetric or
24 skewed bell-shape). The "most significant lag" in such distributed lags is also expected to
25 fluctuate statistically. The "vote-counting" of the most significant lags reported in the past
26 PM-mortality studies shows that 0 and 1 day lags are, in that order, the most frequently reported
27 "optimal" lags, but such estimates may be biased because these lags are also likely the most
28 frequently examined ones. Thus, a more systematic approach across different data sets was
29 needed to investigate this issue.
30 The Samet et al. (2000b) analysis of the 90 largest U.S. cities provides particularly useful
31 information on this matter. Figure 8-28 depicts the Samet et al. (2000b) overall pooled results,
April 2002 8-237 DRAFT-DO NOT QUOTE OR CITE
-------
I
-0.2
I
0.0
I
0.2
I
0.4
0.6
I
0.8
% Change in Mortality per 10 jjg/m3 Increase in PM
i
1.0
10
Figure 8-28. Marginal posterior distribution for effects of PM10 on all cause mortality at
lag 0,1, and 2 for the 90 cities. From Samet et al. (2000a,b). The numbers in
the upper right legend are posterior probabilities that overall effects are
greater than 0.
Source: Samet et al. (2000b).
1 showing the posterior distribution of PM10 effects for the 90 cities for lag 0, 1, and 2 days. It can
2 be seen that the effect size estimate for lag Iday is about twice that for lag 0 or lag 2 days,
3 although their distributions overlap. However, a careful examination of Figures 6 and 7 in the
4 NMMAPS I Report suggests that the maximum PM10 effect may occur in different cities with
5 somewhat different lag relationships. In terms of the magnitude of the estimated PM10 effects,
6 Table 8-40, based on NMMAPS I Figure 7 (posterior bivariate distribution for each county; PM10
7 effect adjusted for O3), suggests that somewhat different patterns may apply in different
8 locations. These data suggest that while lag 1 effects are typically the largest, there may be some
9 situations in which lag 0 or lag 2 effects are larger.
10 The NMMAPS mortality and morbidity analyses and another HEI-sponsored study on PM
11 components (Lippmann et al., 2000) illustrate three different ways to deal with temporal
12 structure: (1) assume all sites have the same lag, e.g., 1 day, for a given effect; (2) use the lag or
April 2002
8-238
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-40. COMPARISON
ANALYSES FOR 0,1, AND
OF PM10 EFFECT SIZES ESTIMATED BY NMMAPS
2 DAY LAGS FOR THE 20 LARGEST U.S. CITIES
County
Ordered PM,n effect sizes
10
Los Angeles
New York
Chicago
Dallas/Fort Worth
Houston
San Diego
Santa Ana /Anaheim
Phoenix
Detroit
Miami
Philadelphia
Seattle
San Jose
Cleveland
San Bernardino
Pittsburgh
Oakland
San Antonio
Riverside
Lag 0 < lag 1 « lag 2
Lag 0 = lag 1 » lag 2
Unreadable
Lag 0> lag 1, lag Klag2
Lag 0< lag 1, lag 1 > lag 2
Lag 0 = lag 1 > lag 2
Lag 0 > lag 1 > lag 2
Lag 0 = lag 1< lag 2
Lag 0 < lag 1, lag 1 > lag 2
Lag 0 < lag 1 = lag 2
Lag 0< lag 1, lag 1 > lag 2
Lag 0< lag 1, lag 1 > lag 2
Lag 0 > lag 1 = lag 2
Lag 0> lag 1, lag Klag2
Lag 0 > lag 1 = lag 2
Lag 0< lag 1, lag 1 > lag 2
Lag 0 < lag 1 = lag 2
Lag 0 = lag 1< lag 2
Lag 0< lag 1, lag 1 > lag 2
1 moving average giving the largest or most significant effect and for each pollutant and endpoint;
2 and (3) use a flexible distributed lag model, with parameters adjusted to each site
3 The NMMAPS mortality analyses used the first approach. This approach introduces a
4 consistent response model across all locations. However, since the cardiovascular, respiratory, or
5 other causes of acute mortality usually associated with PM are not at all specific, there is little
6 a priori reason to believe that they must have the same relation to current or previous PM
7 exposures at different sites. The imposed consistency in lag that maximizes the aggregate effect
April 2002
8-239
DRAFT-DO NOT QUOTE OR CITE
-------
1 of lag 1 across all cities, in Figure 15-18 and 24 of NMMAPS II, may obscure important regional
2 or local differences for lags other than 1 day. Moolgavkar (2000a,b) illustrates this point for
3 three large U.S. cities where strong PM effects on cardiovascular mortality occur at lags 4-5 and
4 1-2 days in Maricopa County, lag 3 in Cook County, and lag 0 in Los Angeles County. These
5 may correspond to the onset or exacerbation of different illnesses leading to cardiovascular
6 mortality.
7 The NMMAPS morbidity studies evaluate 0- and 1-day lags, the moving average of 0 and
8 1-day lags, polynomial distributed lag models, and unrestricted distributed lag models. The
9 first-stage models for each city in the study were fitted for each city, with no restriction as to a
10 consistent model across all cities, and combined across all 14 cities in the second stage as shown
11 in Table 14 and Figure 23 of NMMAPS H. A comparison of the data tabulated in the NMMAPS
12 Report Appendices shows large differences across cities in the apparent magnitude of the PM10
13 effect, depending on how the PM concentration data over the preceding few days are used.
14 The approach used in Lippmann et al. (2000) and many other studies is to use the model
15 that maximizes some global model goodness-of-fit criterion. This leads to selection of different
16 models at different sites, as might be expected. However, the best-fitting model (for lags, for
17 example) is often the model with the largest or most significant PM10 coefficient. All models for
18 the pollutant(s) of interest are usually compared among themselves only after a preliminary
19 baseline model has been fitted. The baseline model takes into account most of the other
20 variables with which PM10 could be plausibly associated, so that the remaining variation in
21 morbidity or mortality that can be explained by including PM10 indicators with different temporal
22 structures is nearly "orthogonal" or independent of the baseline model. The restriction to the
23 same lag day at all sites certainly increases the precision of that estimate, but possibly at the cost
24 of obscuring different relationships between time of exposure and health effect at other sites.
25 An additional complication in assessing the shape of a distributed lag is that the apparent
26 spread of the distributed lag may depend on the pattern of persistence of air pollution (i.e.,
27 episodes may persist for a few days), which may vary from city to city and from pollutant to
28 pollutant. If this is the case, fixing the lag across cities or across pollutants may not be ideal, and
29 may tend to obscure important nuances of lag structures that may provide important clues to
30 possible different lags between PM exposures and different cause-specific effects.
April 2002 8-240 DRAFT-DO NOT QUOTE OR CITE
-------
1 Thus, it is possible that the extent of lag and its spread may vary depending on the cause of
2 death. For example, Rossi et al. (1999) report that, in their analysis of TSP-cause specific
3 mortality in Milan, Italy, the lags varied for different cause of death (i.e., same day for respiratory
4 infections and heart failure; 3-4 days for myocardial infarction and COPD). Thus, the lag for
5 total mortality may exhibit mixed lags (weighted by the frequency of deaths in each cause).
6 Another example was reported for a recent Mexico City study (Borja-Aburto et al., 1998), in
7 which they found significant PM2 5-total mortality associations for same day and 4-day lag, but
8 not for the intervening 2 to 3 days (percent increases per 25 //g/m3 were 3.38, -4.00, 1.03,
9 1.08,3.43, 2.49, for 0 through 5 day lags, respectively). The authors state: "This phenomenon is
10 consistent with both a harvesting of highly susceptible persons on the day of exposure to high
11 pollution levels and a lagged increase in mortality due to delayed effects of reduction of
12 pulmonary defenses, cardiovascular complications, or other homeostatic changes among
13 less-compromised individuals". It is interesting to note that Wichmann et al. (2000) also
14 reported that the most predictive single day effects on mortality for mass concentrations of
15 0.01-2.5 (j, particles were either immediate (0-1 d lag) or delayed (4-5 d lag) for their data from
16 Erfurt, Germany.
17 It should also be noted that if one chooses the most significant single lag day only, and if
18 more than one lag day shows positive (significant or otherwise) associations with mortality, then
19 reporting a RR for only one lag would also underestimate the pollution effects. Schwartz
20 (2000b) investigated this issue, using the 10 U.S. cities data where daily PM10 values were
21 available for 1986-1993. Daily total (non-accidental) deaths of persons 65 years of age and older
22 were analyzed. For each city, a GAM Poisson model adjusting for temperature, dewpoint,
23 barometric pressure, day-of-week, season, and time was fitted. Effects of distributed lag were
24 examined using four models: 1-day mean at lag 0 day; 2-day mean at lag 0 and 1 day; second-
25 degree distributed lag model using lags 0 through 5 days; unconstrained distributed lag model
26 using lags 0 through 5 days. The inverse variance weighted averages of the ten cities' estimates
27 were used to combine results. The results indicated that the effect size estimates for the
28 quadratic distributed model and unconstrained distributed lag model were similar. Both
29 distributed lag models resulted in substantially larger effect size estimates (7.25% and 6.62%,
30 respectively, as percent excess total death per 50 //g/m3 increase in PM10) than the single day lag
31 (3.29%) and moderately larger effect size estimates than the two-day average models (5.36%).
April 2002 8-241 DRAFT-DO NOT QUOTE OR CITE
-------
1 Samet et al. (2000a,b) also applied 7- and 14-day unconstrained distributed lag models to
2 Chicago, Minneapolis/St. Paul, and Pittsburgh data, and reported that the sum of the 7-day
3 distributed lag coefficients was greater than the estimates based on a single day's value, but the
4 14-day estimate was substantially lower than the 7-day estimate in Chicago and Minneapolis/
5 St. Paul. Thus, it is possible that the usual RR estimate using one lag day may underestimate PM
6 effects.
7 Mis-specification of the lag structure may cause important modeling biases. Most of the
8 published literature for the U.S. evaluates only single-day models, a choice dictated by the every-
9 sixth-day sampling schedule used for PM10 in many U.S. cities. When this occurs, it is not
10 possible to evaluate multi-day models with greater biological plausibility, such as moving
11 average models and distributed lag models. Only three of the 20 largest U.S. cities used in the
12 NMMAPS mortality study (Chicago, Minneapolis-St. Paul, Pittsburgh) had daily data (Samet
13 et al., 2000a,b,c). The 14 cities used in the NMMAPS hospital admissions study had daily PM10
14 data, but some of these cities were too small to be included among the 90 largest cities in the
15 mortality study (Canton and Youngstown, OH, Boulder and Colorado Springs, CO). An every-
16 other-day sampling schedule was used in the Harvard Six City Study, for which the PM data on a
17 given day has been used as though it were a two-day moving, alternately concurrent with
18 mortality on half the days and lagging mortality by one day on the other days. While the most
19 commonly used lags in PM time series models are zero or one day, some studies have found PM
20 effects with longer lags (Loomis et al., 1999, in Mexico City; Ponka et al., 1999, for Helsinki),
21 and other studies have found effects at both short and long time lags in some cities (Moolgavkar,
22 2000a,b). It is therefore plausible that mortality or hospital admissions from PM may arise from
23 different responses or PM-associated diseases with different characteristic lags, for example, that
24 cardiovascular responses may arise almost immediately after exposure, within zero or one days
25 or even within two hours (Peter et al., 2002, for myocardial infarction).
26 One would then expect to see different best-fitting lags for different cause-specific
27 mortality or hospital admissions. This idea was fully demonstrated in Lippmann et al. (2000)
28 where different single-day lag models for different health endpoints, PM metrics, and gaseous
29 pollutants were included in the model. The best-fitting PM models had lag 0 to 3 days,
30 depending on the endpoint. This problem is not solved by use of distributed lag models.
31 Schwartz and Zanobetti (2002) found it necessary to use different distributed lag models in each
April 2002 8-242 DRAFT-DO NOT QUOTE OR CITE
-------
1 of the 10 cities whose concentration-response functions were combined by meta-smoothing. One
2 wonders if the meta-smoothing results by region (Dominici et al., 2002) might have been
3 changed if the concentration-response function and optimal lag for each city had been used.
4 In this case, model mis-specification may involve a combination of two potential biases.
5
6 8.4.5 New Assessments of Mortality Displacement
7 There have been a few studies that investigated the question of "harvesting", a phenomenon
8 in which a deficit in mortality occurs following days with (pollution-caused) elevated mortality,
9 due to depletion of the susceptible population pool. This issue is very important in interpreting
10 the public health implication of the reported short-term PM mortality effects. The 1996 PM
11 AQCD discussed suggestive evidence observed by Spix et al. (1993) during a period when air
12 pollution levels were relatively high. Recent studies, however, generally typically used data from
13 areas with lower, non-episodic pollution levels.
14 Schwartz (2000c) separated time-series air pollution, weather, and mortality data from
15 Boston, MA, into three components: (1) seasonal and longer fluctuations; (2) "intermediate"
16 fluctuations; (3) "short-term" fluctuations. By varying the cut-off between the intermediate and
17 short term, evidence of harvesting was sought. The idea is, for example, if the extent of
18 harvesting were a matter of a few days, associations between weekly average values of mortality
19 and air pollution (controlling for seasonal cycles) would not be seen. For COPD, Schwartz
20 (2000c) reported evidence indicating that most of the mortality was only displaced by a few
21 weeks; for pneumonia, heart attacks, and all cause mortality, the effect size increased as longer
22 time scales were included. The percent increase in deaths associated with a 25 //g/m3 increase in
23 PM25 increased from 5.3% (95%CI: 6.8, 9.0) to 9.64% (95%CI: 8.2, 11.1).
24 Schwartz and Zanobetti (2000) used the same approach described above to analyze a larger
25 data set from Chicago, IL for 1988-1993. Total (non-accidental), in-hospital, out-of-hospital
26 deaths, as well as heart disease, COPD, and pneumonia elderly hospital admissions were
27 analyzed to investigate possible PM10 "harvesting" effects. GAM Poisson models adjusting for
28 temperature, relative humidity, day-of-week, and season were applied in baseline models using
29 the average of the same day and previous day's PM10. Seasonal and trend decomposition
30 techniques called STL were applied to the health outcome and exposure data to decompose them
31 into different time-scales (i.e., short-term to long-term), excluding long seasonal cycles (120 day
April 2002 8-243 DRAFT-DO NOT QUOTE OR CITE
-------
1 window). The associations were examined with smoothing windows of 15, 30, 45, and 60 days.
2 The effect size estimate for deaths outside hospital was larger than for deaths inside hospital.
3 All cause mortality showed an increase in effect size at longer time scales. The effect size for
4 deaths outside hospital increases more steeply with increasing time scale than that for inside
5 hospital deaths.
6 Zanobetti et al. (2000b) used GAM distributed lag models to help quantify mortality
7 displacement in Milan, Italy, 1980-1989. Non-accidental total deaths were regressed on smooth
8 functions of TSP distributed over the same day and the previous 45 days using penalized splines
9 for the smooth terms and seasonal cycles, temperature, humidity, day-of-week, holidays, and
10 influenza epidemics. The mortality displacement was modeled as the initial positive increase,
11 negative rebound (due to depletion), followed by another positive coefficients period, and the
12 sum of the three phases were considered as the total cumulative effect. TSP was positively
13 associated with mortality up to 13 days, followed by nearly zero coefficients between 14 and
14 20 days, and then followed by smaller but positive coefficients up to the 45th day (maximum
15 examined). The sum of these coefficients was over three times larger than that for the single-day
16 estimate.
17 Zeger et al. (1999) first illustrated, through simulation, the implication of harvesting for PM
18 regression coefficients (i.e., mortality relative risk) as observed in frequency domain. Three
19 levels of harvesting, 3 days, 30 days, and 300 days were simulated. As expected, the shorter the
20 harvesting, the larger the PM coefficient in the higher frequency range. However, in the real data
21 from Philadelphia, the regression coefficients increased toward the lower frequency range,
22 suggesting that the extent of harvesting, if it exists, is not in the short-term range. Zeger
23 suggested that "harvesting-resistant" regression coefficients could be obtained by excluding the
24 coefficients in the very high frequency range (to eliminate short-term harvesting) and in the very
25 low frequency range (to eliminate seasonal confounding). Since the observed frequency domain
26 coefficients in the very high frequency range were smaller than those in the mid frequency range,
27 eliminating the "short-term harvesting" effects would only increase the average of those
28 coefficients in the rest of the frequency range.
29 Frequency domain analyses are rarely performed in air pollution health effects studies,
30 except perhaps the spectral analysis (variance decomposition by frequency) to identify seasonal
31 cycles. Examinations of the correlation by frequency (coherence) and the regression coefficients
April 2002 8-244 DRAFT-DO NOT QUOTE OR CITE
-------
1 by frequency (gain) may be useful in evaluating the potentially frequency-dependent
2 relationships among multiple time series. A few past examples in air pollution health effects
3 studies include: (1) Shumway et al.'s (1983) analysis of London mortality analysis, in which
4 they observed that significant coherence occurred beyond two week periodicity (they interpreted
5 this as "pollution has to persist to affect mortality"); (2) Shumway et al.'s (1988) analysis of
6 Los Angeles mortality data, in which they also found larger coherence in the lower frequency;
7 (3) Ito's (1990) analysis of London mortality data in which he observed relatively constant gain
8 (regression coefficient) for pollutants across the frequency range, except the annual cycle. These
9 results also suggest that associations and effect size, at least, are not concentrated in the very high
10 frequency range.
11 Schwartz (2000c), Zanobetti et al. (2000b), and Zeger et al.'s (1999) results all suggest that
12 the extent of harvesting, if any, is not a matter of only a few days. Other past studies that used
13 frequency domain analyses are also at least qualitatively in agreement with the evidence against
14 the short-term only harvesting. Since very long wave cycles (> 6 months) need to be controlled
15 in time-series analyses to avoid seasonal confounding, the extent of harvesting beyond 6 months
16 periodicity is not possible in time-series study design. While these studies suggest that observed
17 short-term associations are not simply due to short-term harvesting, more data are needed to
18 obtain quantitative estimates of the extent of prematurity of deaths.
19
20 8.4.6 Concentration-Response Relationships for Ambient PM
21 In the 1996 PM AQCD, the limitations of identifying 'threshold' in the concentration-
22 response relationships in observational studies were discussed including the low data density in
23 the lower PM concentration range, the small number of quantile indicators often used, and the
24 possible influence of measurement error. Also, a threshold for a population, as opposed to a
25 threshold for an individual, has some conceptual issues that need to be noted. For example,
26 Schwartz (1999) discussed that, since individual thresholds would vary from person to person
27 due to individual differences in genetic level susceptibility and pre-existing disease conditions, it
28 would be almost mathematically impossible for a threshold to exist in the population. This
29 argument holds only if the most sensitive members of a population are sensitive to very low
30 concentrations, which may not be the case. The person-to-person difference in the relationship
31 between personal exposure and the concentration observed at a monitor would also add to the
April 2002 8-245 DRAFT-DO NOT QUOTE OR CITE
-------
1 variability. Because one cannot directly measure but can only compute or estimate a population
2 threshold, it would be difficult to interpret an observed threshold, if any, biologically. Despite
3 these issues, several studies have attempted to address the question of threshold by analyzing
4 large databases, or by conducting simulations.
5 Daniels et al. (2000) examined the presence of threshold using the largest 20 U.S. cities for
6 1987-1994. The authors compared three log-linear GAM regression models: (1) using a linear
7 PM10 term; (2) using a cubic spline of PM10 with knots at 30 and 60 //g/m3 (corresponding
8 approximately to 25 and 75 percentile of the distribution); and, (3) using a threshold model with
9 a grid search in the range between 5 and 200 //g/m3 with 5 //g/m3 increment. The covariates
10 included in these models are similar to those used by the same research group previously (Kelsall
11 et al., 1997; Samet et al., 2000a,b), including the smoothing function of time, temperature and
12 dewpoint, and day-of-week indicators. Total, cardiorespiratory, and other mortality series were
13 analyzed. These models were fit for each city separately, and for model (1) and (2), the
14 combined estimates across cities were obtained by using inverse variance weighting if there was
15 no heterogeneity across cities, or by using a two-level hierarchical model if there was
16 heterogeneity. The best fit among the models, within each city and over all cities, were also
17 determined using the Akaike's Information Criterion (AIC). The results using the spline model
18 showed that, for total and cardiorespiratory mortality, the spline curves were roughly linear,
19 consistent with the lack of a threshold. For mortality from other causes, however, the curve did
20 not increase until PM10 concentrations exceeded 50 //g/m3. While the test of heterogeneity
21 indicated that there was considerable heterogeneity in these curves across cities (see Figure
22 8-29), the shapes of the curves were similar across cities, with no indication of one city unduly
23 influencing the overall estimate of the curves. The hypothesis of linearity was examined by
24 comparing the AIC values across models. The results suggested that the linear model was
25 preferred over the spline and the threshold models. Thus, these results suggest that linear models
26 without a threshold may well be appropriate for estimating the effects of PM10 on the types of
27 mortality of main interest.
28 Thus, while these studies do not refute the usual assumption of a linear no-threshold
29 concentration-response function, neither do they provide unqualified support for that assumption.
30 Sensitivity analyses for individual city studies' concentration-response function would be
31 helpful. Schwartz and Zanobetti (2000) investigated the presence of threshold by simulation and
April 2002 8-246 DRAFT-DO NOT QUOTE OR CITE
-------
Total
o.os
-c
o
-------
1 span (0.7) was used in each of the cities. The predicted values of the log relative risks were
2 computed for 2 //g/m3 increments between 5.5 //g/m3 and 69.5 //g/m3 of PM10 levels. Then, the
3 predicted values were combined across cities using inverse-variance weighting. The simulation
4 results indicated that the "meta-smoothing" approach did not bias the underlying relationships for
5 the linear and threshold models, but did result in a slight downward bias for the logarithmic
6 model. Measurement error (additive or multiplicative) in the simulations did not cause upward
7 bias in the relationship below threshold. The threshold detection in the simulation was not very
8 sensitive to the choice of span in smoothing. In the analysis of real data from 10 cities, the
9 combined curve did not show evidence of a threshold in the PM10-mortality associations.
10 Cakmak et al. (1999) investigated methods to detect and estimate threshold levels in time
11 series studies. Based on the realistic range of error observed from actual Toronto pollution data
12 (average site-to-site correlation: 0.90 for O3; 0.76 for COH; 0.69 for TSP; 0.59 for SO2; 0.58 for
13 NO2; and 0.44 for CO), pollution levels were generated with multiplicative error for six levels of
14 exposure error (1.0, 0.9, 0.8, 0.72, 0.6, 0.4, site-to-site correlation). Mortality series were
15 generated with three PM10 threshold levels (12.8 Mg/m3, 24.6 Mg/m3, and 34.4 Mg/m3). LOESS
16 with a 60% span was used to observe the exposure-response curves for these 18 combinations of
17 exposure-response relationships with error. A parameter threshold model was also fit using non-
18 linear least squares. Graphical presentations indicate that LOESS adequately detects threshold
19 under no error, but the thresholds were "smoothed out" under the extreme error scenario. Use of
20 a parametric threshold model was adequate to give "nearly unbiased" estimates of threshold
21 concentrations even under the conditions of extreme measurement error, but the uncertainty in
22 the threshold estimates increased with the degree of error. They concluded, "if threshold exists,
23 it is highly likely that standard statistical analysis can detect it".
24 The Smith et al. (2000) study of associations between daily total mortality and PM25 and
25 PM10_2 5 in Phoenix, AZ (during 1995-1997) also investigated the possibility of a threshold.
26 In the linear model, the authors found that mortality was significantly associated with PM10_2 5,
27 but not with PM2 5. In modeling possible thresholds, they applied: (1) a piecewise linear model
28 in which several possible thresholds were specified; and (2) a B-spline (spline with cubic
29 polynomials) model with 4 knots. Using the piecewise model, there was no indication that there
30 was a threshold for PM10_2 5. However, for PM2 5, the piecewise model resulted in suggestive
31 evidence for a threshold, around 20 to 25 //g/m3. The B-spline results also showed no evidence
April 2002 8-248 DRAFT-DO NOT QUOTE OR CITE
-------
1 of threshold for PM10_2 5, but for PM2 5, a non-linear curve showed a change in the slope around
2 20 //g/m3. A further Bayesian analysis for threshold selection suggested a clear peak in the
3 posterior density of PM2 5 effects around 22 //g/m3. These results, if they in fact reflect reality,
4 make it difficult to evaluate the relative roles of different PM components (in this case, PM2 5
5 versus PM10_2 5). However, the concentration-response curve for PM2 5 presented in this
6 publication suggests more of a U- or V-shaped relationship than the usual "hockey stick"
7 relationship. Such a relationship is, unlike the temperature-mortality relationship, difficult to
8 interpret biologically. Because the sample size of this data (~3 years) is relatively small, further
9 investigation of this issue using similar methods but a larger data set is warranted. Other studies
10 evaluate non-linear relationships using a multi-city meta-smoothing approach based on non- or
11 semi-parametric smoothers rather than on linear parametric models.
12 Many ad hoc decisions go into model selection in air pollution health effects studies. The
13 effect of some of these decisions on relative risk estimates for Birmingham, AL, PM10 data,
14 previously analyzed by Schwartz (1993) and others, is illustrated by Smith et al. (2000). The
15 response variable is non-accidental mortality. Specifically, the selection of meteorological
16 variables, the selection of an exposure variable (as a weighted average of lagged PM values), and
17 the possibility of nonlinear effects, such as threshold effects, are investigated. The results are
18 sensitive to the inclusion of humidity in addition to temperature. This inclusion decreases the
19 resulting PM10 coefficient. The model is highly sensitive to the definition of an exposure
20 measure. For example, when lags 0-4 were averaged, there was no significant effect. In an
21 attempt to account for a nonlinear PM-mortality effect, there appeared to be little effect of
22 exposure below 80 //g/m3, and a threshold analysis (as well as a generalized additive models
23 approach) supported the conclusion that the main effect is at higher values of PM. Although this
24 paper was based on an intensive analysis of a single data set (in contrast to other studies, such as
25 NMMAPS analysis, which combined data form many cities), it demonstrated the very wide range
26 of interpretations that are possible using alternative, but statistically valid, analyses of the same
27 data.
28
29
30
April 2002 8-249 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.7 New Assessments of Consequences of Measurement Error
2 8.4.7.1 Theoretical Framework for Assessment of Measurement Error
3 Since the 1996 PM AQCD, there have been some advances in conceptual framework
4 development to investigate the effects of measurement error on PM health effects estimated in
5 time-series studies. Several new studies evaluated the extent of bias caused by measurement
6 errors under a number of scenarios with varying extent of error variance and covariance structure
7 between co-pollutants.
8 Zidek et al. (1996) investigated, through simulation, the joint effects of multi-collinearity
9 and measurement error in Poisson regression model, with two covariates with varying extent of
10 relative errors and correlation. Their error model was of classical error form (W=X+U, where W
11 and X are surrogate and true measurements, respectively, and the error U is normally distributed).
12 The results illustrated the transfer of effects from the "causal" variable to the confounder.
13 However, for the confounder to have larger coefficients than the true predictor, the correlation
14 between the two covariates had to be large (r = 0.9), with moderate error (a > 0.5) for the true
15 predictor, and no error for the confounder in their scenarios. The transfer-of-causality effect was
16 mitigated when the confounder also became subject to error. Another interesting finding that
17 Zidek et al. reported is the behavior of the standard errors of these coefficients: when the
18 correlation between the covariates was high (r = 0.9) and both covariates had no error, the
19 standard errors for both coefficients were inflated by factor of 2; however, this phenomenon
20 disappeared when the confounder had error. Thus, multi-collinearity influences the significance
21 of the coefficient of the causal variable only when the confounder is accurately measured.
22 Zeger et al. (2000) also conducted a mathematical analysis of PM mortality effects in
23 ordinary least square model (OLS) with the classical error model, under varying extent of error
24 variance and correlation between two predictor variables. The error described here was
25 analytical error (e.g., discrepancy between the co-located monitors). In general, they found that
26 positive regression coefficients are only attenuated, but null predictors (zero coefficient) or weak
27 predictors are only able to appear stronger than true positive predictors under unusual conditions:
28 (1) true predictors must have very large positive or negative correlation (i.e., |r| > 0.9);
29 (2) measurement error must be substantial (i.e., error variance ~ signal variance); and
30 (3) measurement errors must have a large negative correlation. They concluded that estimated
April 2002 8-250 DRAFT-DO NOT QUOTE OR CITE
-------
1 FP health effects are likely underestimated, although the magnitude of bias due to the analytical
2 measurement error is not very large.
3 Zeger et al. (1999) illustrated the implication of the classical error model and the Berkson
4 error model (i.e., X = W + U) in the context of time-series study design. Their simulation of the
5 classical error model with two predictors, with various combinations of error variance and
6 correlation between the predictors/error terms, showed results similar to those reported by Zidek
7 et al. (1996). Most notably, for the transfer of the effects of one variable to the other (i.e., error-
8 induced confounding) to be large, the two predictors or their errors need to be substantially
9 correlated. Also, for the spurious association of a null predictor to be more significant than the
10 true predictor, their measurement errors have to be extremely negatively correlated—a condition
11 not yet demonstrated as occurring in actual air pollution data sets.
12 Zeger et al. also laid out a comprehensive framework for evaluating the effects of exposure
13 measurement error on estimates of air pollution mortality relative risks in time-series studies.
14 The error, the difference between personal exposure and the central station's measurement of
15 ambient concentration was decomposed into three components: (1) the error due to having
16 aggregate rather than individual exposure; (2) the difference between the average personal
17 exposure and the true ambient concentration level; and, (3) the difference between the true and
18 measured ambient concentration level. By aggregating individual risks to obtain expected
19 number of deaths, they showed that the first component of error (the aggregate rather than
20 individual) is a Berkson error, and, therefore is not a significant contributor to bias in the
21 estimated risk. The second error component is a classical error and can introduce bias if there are
22 short-term associations between indoor source contributions and ambient concentration levels.
23 Recent analysis, however, both using experimental data (Mage et al., 1999; Wilson et al., 2000)
24 and theoretical interpretations and models (Ott et al., 2000) indicate that there is no relationship
25 between the ambient concentration and the nonambient components of personal exposure to PM.
26 However, a bias can arise due to the difference between the personal exposure to ambient PM
27 (indoors plus outdoors) and the ambient concentration. The third error component is the
28 difference between the true and the measured ambient concentration. According to Zeger et al.
29 the final term is largely of the Berkson type if the average of the available monitors is an
30 unbiased estimate of the true spatially averaged ambient level.
April 2002 8-251 DRAFT-DO NOT QUOTE OR CITE
-------
1 Using this framework, Zeger et al. (2000) then used PTEAM Riverside, CA data to
2 estimate the second error component and its influence on estimated risks. The correlation
3 coefficient between the error (the average population PM10 total exposure minus the ambient
4 PM10 concentration) and the ambient PM10 concentration was estimated to be -0.63. Since this
^,
5 correlation is negative, the fiz (the estimated value of the pollution-mortality relative risk in the
6 regression of mortality onzp the daily ambient concentration) will tend to underestimate the
yv
7 coefficient fix that would be obtained in the regression of mortality on xt, the daily average total
8 personal exposure, in a single-pollutant analysis. Zeger et al. (2000) then proceed to assess the
9 size of the bias that will result from this exposure misclassification, using daily ambient
10 concentration, zt. As shown in Equation 9, the daily average total personal exposure, xt, can be
11 separated into a variable component, Ql zp dependent on the daily ambient concentration, zp and
12 a constant component, 60, independent of the ambient concentration.
13
14 xt = OQ + Qlzt + £t (8-5)
15 where £t is an error term.
16 If the nonambient component of the total personal exposure is independent of the ambient
17 concentration, as appears to be the case, Equation 9 from Zeger et al. (2000) becomes the
18 regression analysis equation familiar to exposure analysts (Dockery and Spengler, 1981; Ott
19 et al., 2000; Wilson et al., 2000). In this case, 60 gives the average nonambient component of the
20 total personal exposure and Ql gives the ratio of the ambient component of personal exposure to
21 the ambient concentration. (The ambient component of personal exposure includes exposure to
22 ambient PM while outdoors and, while indoors, exposure to ambient PM that has infiltrated
23 indoors.) In this well-known approach to adjust for exposure measurement error, called
~ ~ yv
24 regression calibration (Carroll et al., 1995), the estimate of f)x has the simple form ftx = ftz/9l.
25 Thus, for the regression calibration, the value of (3X (based on the total personal exposure) does
26 not depend on the total personal exposure but is given by [)z, based on the ambient concentration,
27 times 0l3 the ratio of the ambient component of personal exposure to the ambient concentration.
28 A regression analysis of the PTEAM data gave an estimate Ql = 0.60.
29 Zeger et al. (2000) use Equation 9, with 9o = 59.95 and 6j = 0.60, estimated from the
30 PTEAM data, to simulate values of daily average personal exposure, x*t, from the ambient
April 2002 8-252 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentrations, zt, for PM10 in Riverside, CA, 1987-1994. They then compare the mean of the
/\
2 simulated jGf. s, obtained by the series of log-linear regressions of mortality on the simulated x*t,
/^
3 with the normal approximation of the likelihood function for the coefficient Pz from the
yv. yv.
4 log-linear regression of mortality directly on zt. The resulting /3Z / fix = 0.59, is very close to
yv
5 0j = 0.60. Dominici et al. (2000) provide a more complete analysis of the bias in f5z as an
6 estimate of fix using the PTEAM Study and four other data sets and a more complete statistical
^, ^,
7 model. Their findings were qualitatively similar in that fix was close to j3z /Oj. Thus, it appears
8 that the bias is very close to Ql which depends not on the total personal exposure but only on the
9 ratio of the ambient component of personal exposure to the ambient concentration.
10 Zeger et al. (2000), in the analyses described above, also suggested that the error due to the
11 difference between the average personal exposure and the ambient level (the second error type
12 described above) is likely the largest source of bias in estimated relative risk. This suggestion at
13 least partly comes from the comparison of PTEAM data and site-to-site correlation (the third type
14 of error described above) for PM10 and O3 in 8 US cities. While PM10 and O3 both showed
15 relatively high site-to-site correlation («0.6-0.9), a similar extent of site-to-site correlation for
16 other pollutants is not necessarily expected. Ito et al. (1998) estimated site-to-site correlations
17 (after adjusted for seasonal cycles) for PM10, O3, SO2, NO2, CO, temperature, dewpoint
18 temperature, and relative humidity, using multiple stations' data from seven central and eastern
19 states (IL, IN, MI, OH, PA, WV, WI), and found that, in a geographic scale of less 100 miles,
20 these variables could be categorized into three groups in terms of the extent of correlation:
21 weather variables (r > 0.9); O3, PM10, NO2 (r: 0.6 - 0.8); CO and SO2 (r < 0.5). These results
22 suggest that the contribution from the third component of error, as described in Zeger et al.
23 (2000), would vary among pollution and weather variables. Furthermore, the contribution from
24 the second component of error would also vary among pollutants; i.e., the ratio of ambient
25 exposure to ambient concentration, called the attenuation coefficient, is expected to be different
26 for each pollutant. Some of the ongoing studies are expected to shed some light on this issue.
27 However, more information is needed on attenuation coefficients for a variety of pollutants.
28 With regard to the PM exposure, longitudinal studies (Wallace, 2000; Mage et al., 1999),
29 show reasonably good correlation (r = 0.6 to 0.9) between ambient PM concentrations and
30 average population PM exposure, lending support for the use of ambient data as a surrogate for
31 personal exposure to ambient PM in time-series mortality or morbidity studies. Furthermore,
April 2002 8-253 DRAFT-DO NOT QUOTE OR CITE
-------
1 fine particles are expected to show even better site-to-site correlation than PM10. Wilson and Suh
2 (1997) examined site-to-site correlation of PM10, PM25, and PM10_25 in Philadelphia and
3 St. Louis, and found that site-to-site correlations were high (r ~ 0.9) for PM2 5 but low for PM10_2 5
4 (r ~ 0.4), indicating that fine particles have smaller errors in representing community-wide
5 exposures. This finding supports Lipfert and Wyzga's (1997) speculation that the stronger
6 mortality associations for fine particles than coarse particles found in the Schwartz et al. (1996a)
7 study may be due in part to larger measurement error for coarse particles.
8 However, as Lipfert and Wyzga (1997) suggested, the issue is not whether the fine particle
9 association with mortality is a "false positive", but rather, whether the weaker mortality
10 association with coarse particles is a "false negative". Carrothers and Evans (2000) also
11 investigated the joint effects of correlation and relative error, but they specifically addressed the
12 issue of fine (FP) vs. coarse particle (CP) effect, by assuming three levels of relative toxicity of
13 fine versus coarse particles (Ppp / PCP = 1, 3, and 10) and, then, evaluating the bias, (B = |E[PF]/
14 E[PC,]} / (PF / PC), as a function of FP-CP correlation and relative error associated with FP and
15 CP. Their results indicate: (1) if the FP and CP have the same toxicity, there is no bias (i.e.,
16 B=l) as long as FP and CP are measured with equal precision, but, if, for example, FP is
17 measured more precisely than CP, then FP will appear to be more toxic than CP (i.e., B > 1);
18 (2) when FP is more toxic than CP (i.e., Ppp / PCP = 3 and 10), however, the equal precision of FP
19 and CP results in downward bias of FP (B < 1), implying a relative overestimation of the less
20 toxic CP. That is, to achieve non-bias, FP must be measured more precisely than CP, even more
21 so as the correlation between FP and CP increases. They also applied this model to real data
22 from the Harvard Six Cities Study, in particular, the data from Boston and Knoxville. Estimation
23 of spatial variability for Boston was based on external data and a range of spatial variability for
24 Knoxville (since there was no spatial data available for this city). For Boston, where the
25 estimated FP-CP correlation was low (r = 0.28), estimated error was smaller for FP than for CP
26 (0.85 vs. 0.65, as correlation between true vs. error-added series), and the observed FP to CP
27 coefficient ratio was high (11), the calculated FP to CP coefficient ratio was even larger (26)-thus
28 providing evidence against the hypothesis that FP is absorbing some of the coefficient of CP.
29 For Knoxville, where FP-CP correlation was moderate (0.54), the error for FP was smaller than
30 for CP (0.9 vs. 0.75), and the observed FP to CP coefficient ratio was 1.4, the calculated true FP
31 to CP coefficient ratio was smaller (0.9) than the observed value, indicating that the coefficient
April 2002 8-254 DRAFT-DO NOT QUOTE OR CITE
-------
1 was overestimated for the better-measured FP, while the coefficient was underestimated for the
2 worse-measured CP. Since the amount (and the direction) of bias depended on several variables
3 (i.e., correlation between FP and CP; the relative error for FP and CP; and, the underlying true
4 ratio of the FP toxicity to CP toxicity), the authors concluded "...for instance, it is inadequate to
5 state that differences in measurement error among fine and coarse particles will lead to false
6 negative findings for coarse particles".
7 Fung and Krewski (1999) conducted a simulation study of measurement error adjustment
8 methods for Poisson models, using scenarios similar to those used in the simulation studies that
9 investigated implication of joint effects of correlated covariates with measurement error. The
10 measurement error adjustment methods employed were the Regression Calibration (RCAL)
11 method (Carroll et al., 1995) and the Simulation Extrapolation (SEVIEX) method (Cook and
12 Stefanski, 1994). Briefly, RCAL algorithm consists of: (1) estimation of the regression of X on
13 W (observed version of X, with error) and Z (covariate without error); (2) replacement of X by
14 its estimate from (1), and conducting the standard analysis (i.e., regression); and (3) adjustment
15 of the resulting standard error of coefficient to account for the calibration modeling. SIMEX
16 algorithm consists of: (1) addition of successively larger amount of error to the original data;
17 (2) obtaining naive regression coefficients for each of the error added data sets; and, (3) back
18 extrapolation of the obtained coefficients to the error-free case using a quadratic or other
19 function. Fung and Krewski examined the cases for: (1) Px = 0.25; Pz = 0.25; (2) Px = 0.0;
20 pz = 0.25; (3) Px = 0.25; Pz = O.O., all with varying level of correlation (-0.8 to 0.8) with and
21 without classical additive error, and also considering Berkson type error. The behaviors of naive
22 estimates were essentially similar to other simulation studies. In most cases with the classical
23 error, RCAL performed better than SIMEX (which performed comparably when X-Z correlation
24 was small), recovering underlying coefficients. In the presence of Berkson type error, however,
25 even RCAL did not recover the underlying coefficients when X-Z correlation was large ( > 0.5).
26 This is the first study to examine the performance of available error adjustment methods that can
27 be applied to time-series Poisson regression. The authors recommend RCAL over SEVIEX.
28 Possible reasons why RCAL performed better than SEVIEX in these scenarios were not discussed,
29 nor are they clear from the information given in the publication. There has not been a study to
30 apply these error adjustment methods in real time-series health effects studies. These
31 methodologies require either replicate measurements or some knowledge on the nature of error
April 2002 8-255 DRAFT-DO NOT QUOTE OR CITE
-------
1 (i.e., distributional properties, correlation, etc.). Since the information regarding the nature of
2 error is still being collected at this time, it may take some time before applications of these
3 methods become practical.
4 Another issue that measurement error may affect is the detection of threshold in time-series
5 studies. Lipfert and Wyzga (1996) suggested that measurement error may obscure the true shape
6 of the exposure-response curve, and that such error could make the exposure-response curve to
7 appear linear even when a threshold may exist. However, based on a simulation with realistic
8 range of exposure error (due to site-to-site correlation), Cakmak et al. (1999) illustrated that the
9 modern smoothing approach, LOESS, can adequately detect threshold levels (12.8 //g/m3,
10 24.6 //g/m3, and 34.4 //g/m3) even with the presence of exposure error (see also Section 8.4.6
11 above).
12 Other issues related to exposure error that have not been investigated include potential
13 differential error among subpopulations. If the exposure errors are different between susceptible
14 population groups (e.g., people with COPD) and the rest of the population, the estimation of bias
15 may need to take such differences into account. Also, the exposure errors may vary from season
16 to season, due to seasonal differences in the use of indoor emission sources and air exchange
17 rates due to air conditioning and heating. This may possibly explain reported season-specific
18 effects of PM and other pollutants. Such season-specific contributions of errors from indoor and
19 outdoor sources are also expected to be different from pollutant to pollutant.
20 In summary, the studies that examined joint effects of correlation and error suggest that PM
21 effects are likely underestimated, and that spurious PM effects (i.e., qualitative bias such as
22 change in the sign of coefficient) due to transferring of effects from other covariates require
23 extreme conditions and are, therefore, unlikely. Also, one simulation study suggests that, under
24 the likely range of error for PM, it is unlikely that a threshold is ignored by common smoothing
25 methods. More data are needed to examine the exposure errors for other pollutants, since their
26 relative error contributions will influence their relative significance in relative risk estimates.
27
28 8.4.7.2 Spatial Measurement Error Issues That May Affect the Interpretation of
29 Multi-Pollutant Models with Gaseous Co-Pollutants
30 The measurement error framework put forth in Dominici et al. (2000) and Zeger et al.
31 (2000) explicitly assumes that one of the error components has a Berkson error structure.
32 As summarized in (Zeger et al., 2000, p. 421): "This Berkson model is appropriate when z
April 2002 8-256 DRAFT-DO NOT QUOTE OR CITE
-------
1 represents a measurable factor [e.g. measured PM or another pollutant] that is shared by a group
2 of participants whose individual [true] exposures x might vary because of time-activity patterns.
3 For example, z might be the spatially averaged ambient level of a pollutant without major indoor
4 sources and x might be the personal exposures that, when averaged across people, match the
5 ambient level." This assumption is likely accurate for sulfates, less so for fine particles and for
6 PM10, and almost certainly incorrect for gases such as CO and NO2 that may vary substantially on
7 an intra-urban spatial scale with widely distributed local sources.
8 The usual characterization of longitudinal or temporal pollutant correlation may not
9 adequately characterize the spatial variation that is the more important aspect of association in
10 evaluating possible Berkson errors. Temporal correlation coefficients, even across large
11 distances (e.g. Ito et al., 2000) may be a consequence of large-scale weather patterns affecting the
12 concentrations of many pollutants. Local concentrations for some pollutants with strong local
13 sources and low regional dispersion (especially for CO and NO2, and PM10_2 5 to a lesser extent)
14 may have somewhat smaller temporal correlations and much greater relative spatial variations
15 than PM. Thus, individuals in a large metropolitan area may have roughly similar levels of PM
16 exposure x on any given day for which the ambient average PM concentration z is an adequate
17 surrogate, whatever their space-time activity patterns, residence, or non-residential micro-
18 environments, while the same individuals may be exposed to systematically higher or lower
19 concentrations of a co-pollutant than the spatial average of the co-pollutant. This violates the
20 basic assumption of the Berkson error model that within each stratum of the measured (spatially
21 averaged) level z, the average value of the true concentration x is equal to z, i.e.,
22
23 E{ x z } = z, (8-6)
24
25 where E{.} is the average or expected value over the population.
26 There are empirical reasons to believe that if the strata are chosen to be locations within a
27 metropolitan area, some individuals far from local sources have consistently less exposure than
28 the average ambient concentration (denoted p) for co-pollutants with local sources such as CO
29 and NO2, and PM25, whose true exposure (denoted q) depends on the location of the person's
30 residence or other micro-environment where most exposure occurs. For this group,
31
April 2002 8-257 DRAFT-DO NOT QUOTE OR CITE
-------
1 E{q p}p. (8-8)
7
8 There is a substantial and growing body of evidence that adverse health effects are
9 associated with proximity to a major road or highway (Wijst et al., 1994; Monn et al., 2000;
10 Roemer and Hoek, 2001). As shown below, there is good reason to believe that the intra-city
11 variation even in PM25 is substantial within some U.S. cities. If we assume for the sake of
12 argument that concentrations of PM10 or PM2 5 are relatively uniformly distributed, then
13 associations of adverse health effects with proximity to a source cannot be attributed to a
14 pollutant such as PM with a uniform spatial distribution. NO2 is a pollutant often used to
15 illustrate the spatial non-uniformity of the gaseous co-pollutants. Figure 8- from Monn et al.
16 (1997) compares the concentrations of NO2 and PM10 as a function of curbside distance in a
17 moderately busy urban street in Zurich. The PM10 concentrations decrease only slightly with
18 increasing distance, with the decrease more likely due to decreasing coarse particle levels than to
19 decreasing fine particle concentrations. The NO2 concentrations show a much stronger seasonal
20 dependence, decreasing rapidly with increasing distance in the summer and showing little
21 decrease with distance in the winter. However, the belief that PM2 5 is spatially uniform should
22 also not be accepted uncritically, as recent analyses for 27 U.S. cities shown in Chapter 3 and
23 Appendix 3 A of this document demonstrate.
24 The 90th Percentile differences (P90) between a pair of sites may provide a useful guide to
25 the differences between monitor pairs (and by implication, personal exposure to fine particles)
26 that might be reasonably expected within a metropolitan area. Shown below in Table 8-41 are the
27 maximum, median, and minimum differences between monitor pairs, the monitor pairs at which
28 the largest 90th percentile difference occurs (by reference to the tables in Appendix 3 A). Based
29 on these differences, we have shown in Table 8-42 a characterization of cities as "relatively
30 homogeneous" with P90 < 10 //g/m3 and "relatively heterogeneous" if P90 > 10 //g/m3. The
31 results in Appendix 3 A and Table 8-42 show a variety of spatial patterns of association of PM2 5
April 2002 8-258 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-41. MAXIMUM, MEDIAN, AND MINIMUM 90th PERCENTILE OF
ABSOLUTE VALUES OF DIFFERENCES BETWEEN FINE PARTICLE
CONCENTRATIONS AT PAIRS OF MONITORING SITES IN 27 METROPOLITAN
AREAS IN ORDER OF DECREASING MAXIMUM DIFFERENCE
(based on Chapter 3 and Appendix 3A).
City
Los Angeles
Pittsburgh
Riverside-
San Bernardino
Birmingham
Seattle
Gary
Cleveland
Atlanta
Detroit
Salt Lake C.
St. Louis
San Diego
Louisville
Chicago
Washington DC
Steubenville
Boise
Kansas City
Philadelphia
Portland OR
Grand Rapids
Dallas
Milwaukee
Columbia
Tampa
Norfolk
Baton Rouge
N Sites
5
4 (w/o E) *
4
5
5
5
4 (w/o A) *
4
7
7
6 (w/o G) *
5
6
4
4
4
11
6
5 (w/o F)
5
4
6
5
4
4
7
8
4
4
5
3
Maximum
31.0
20.2
21.3
20.2
15.4
15.3
8.4
14.9
14.9
14.0
10.6
13.3
12.6
12.5
11.9
11.2
10.5
10.1
7.7
9.9
8.9
7.4
6.9
6.5
6.1
5.6
5.5
5.3
5.0
4.7
3.2
Pair
CE
AD
BD
BC
AE
AE
CE
BD
BG
EG
CF
CD
AC
AD
CD
AD
EK
AF
AD
AE
BD
CF
BC
AB
BC
AE
FH
AB
BD
AC
AC
Median
13.8
13.7
10.8
12.6
10.1
8.2
7.6
8.2
7.1
9.4
8.3
8.6
7.6
9.5
10.6
8.7
6.2
7.4
6.25
8.45
5.2
4.1
5.2
4.45
4.8
3.3
3.65
3.95
4.45
3.55
2.9
Minimum
11.8
11.8
4.1
7.0
7.5
3.8
3.8
5.9
3.8
6.5
6.5
4.9
3.9
6.0
7.4
6.3
3.5
4.2
4.2
2.5**
3.8
1.9
3.3
4.0
2.8
2.0
2.9
2.7
3.6
2.6
2.5
* Without one site > 100 km from the others.
** Collocated monitors at sites D and E.
April 2002
8-259
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-42. SUMMARY OF WITHIN-CITY HETEROGENEITY BY REGION
Relative Heterogeneity Among Pairs of Monitors
Relatively Heterogenous
Relatively Homogeneous
West
Los Angeles, CA
Riverside, CA
Salt Lake City, UT
San Diego, CA
East
Atlanta, GA
Birmingham, AL
Chicago, IL
Cleveland, OH
Detroit, MI
Gary, IN
Louisville, KY
Pittsburgh, PA
St. Louis, MO
Washington, DC (with F) Seattle, WA (with A)
East West
Baton Rouge, LA Boise, ID
Columbia, SC Portland, OR
Dallas, TX
Grand Rapids, MI
Kansas City, KS-MO
Milwaukee, WI
Norfolk, VA
Philadelphia, PA
Steubenville, OH
Tampa, FL
Washington, DC (w/o F) Seattle, WA (w/o A)
1 within a Metroplitan Statistical Area (MSA). There may be some discernable regional
2 differences; but, because many major population centers are not represented in Appendix 3A,
3 further investigation is likely warranted.
4 The results shown here provide clear evidence that fine particle concentrations may be less
5 homogenous in at least some MSAs than has been previously assumed. This provides support
6 for earlier studies using TSP and PM10 cited below. As noted in Chapter 3, these differences may
7 not be strictly related to the distance between monitors, especially where topography plays a role.
8 In many eastern sites, however, particle distribution may be more substantially governed by
9 regional particle concentrations than by local concentrations.
10 A number of recent studies have examined the role of spatial siting of monitors on the
11 estimation of PM effects. Ito et al. (1995) examined the ability of single-site vs. multi-site
12 averages to best estimate total mortality vs. PM10 in Cook County (Chicago), IL and Los Angeles
13 County, CA. In order to have a sufficiently large sample size to detect effects, Ito et al. used six
14 PM10 sites in Cook County (Chicago), IL and four sites in Los Angeles County, CA.
April 2002
8-260
DRAFT-DO NOT QUOTE OR CITE
-------
1 A sinusoidal model was used to account for temporal components, although spline or LOESS
2 methods would now be used. Only one Cook County site had every-day PM samples, and the
3 others as well as the Los Angeles sites had a one-in-six-day sampling schedule. The monitor
4 sites were located in urban and suburban settings, according to the State's objectives. Three of
5 the Los Angeles sites were residential and one was commercial use. One of the Cook County
6 sites was residential, two were commercial, and three were industrial. One of the Chicago sites
7 was intended to monitor population exposure, three to monitor maximum concentrations, and
8 two to monitor both maximum concentrations and personal exposure. There was considerable
9 variation among the distribution of PM10 in Cook County (Chicago), IL sites, and among
10 Los Angeles County, CA sites, especially at the upper end of the distribution. The sites were
11 temporally correlated, 0.83 to 0.63 in Cook County, 0.9 to 0.7 in Los Angeles (except for one site
12 pair), across distances of 4 to 26 miles. The Cook County mortality estimates were better
13 estimated by some single-site estimates (Site 2 with everyday data, N = 1251) than by an average
14 using all available data with missing values estimated from non-missing data (N = 1357). The
15 every-six-day subsamples from Site 1 (N = 281) and Site 2 (lag 0, N = 246) were better
16 predictors, and from Site 4 (N = 243) and Site 6 (N = 292) about as good predictors of mortality
17 as the corresponding every-six-day averages (N = 351). In Los Angeles, only Site 4 (N = 349)
18 was about as predictive as the spatial averages (N = 405).
19 Lipfert et al. (2000) examined the relationship between the area in which mortality occurred
20 among residents and the locations of monitoring sites or averages over monitoring sites for
21 several particle size components and particle metrics. The mortality data were located for
22 Philadelphia, PA, for three aditional suburban Philadelphia counties, for Camden, NJ and other
23 New Jersey counties in the Philadelphia - Camden MSA. A single site was used for fine and
24 coarse particles from the Harvard School of Public Health monitors. Additional PA and NJ
25 thoracic particle data were available for 2 to 4 stations and results averaged for at least two
26 stations reporting data. The authors conclude that mortality in any part of the region may be
27 associated with air pollution concentrations or average concentrations in any other part of the
28 region, whether particles or gases. The authors suggest two interpretations: (a) the associations
29 of mortality with pollution were random (from carrying out multiple significance tests) and not
30 causal, or (b) both particles and gaseous pollutants have a broad regional distribution. The
31 authors note that interpretation (b) may lead to large uncertainties in identifying which pollutant
April 2002 8-261 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposures for the population are primarily responsible for the observed effects. These data could
2 be studied further to evaluate smaller-scale spatial relationships among health effects and gases.
3 Lippmann et al. (2000) evaluated the effects of monitor siting choice using 14 TSP
4 monitoring stations in Detroit, MI, and nearby Windsor, ON, Canada. The stations operated
5 from 1981-1987 with almost complete data. When a standard log-linear link Poisson regression
6 model for mortality was fitted to TSP data for each of the 14 sites, the relative risk estimates
7 were similar for within-site increments of 5th to 95th percentiles, generally highest and positive at
8 lag day 1, but not statistically significant except for site "w" (site 12, south of the urban center of
9 Wayne County) and nearly significant at sites "f" (west of the city of Detroit), "g" (south of the
10 city) and "v" (suburban site in northwestern Wayne County, MI, generally "upwind" of the
11 urban center). However, as the authors note, all of the reported relative risks are for site-specific
12 increments, which vary by a factor of about 2.5 over the Wayne County - Windsor area. When
13 converted to a common increment of 100 //g/m3 TSP, the largest excess risks are found when the
14 monitor used in the model is "f' (4.5%), "v" (4.2%), or "w" (3.8%), which also show the most
15 significant effects among the 14 monitors. As the authors note, "... the distributional increments
16 [used] to calculate relative risk tend to standardize the scale of relative risks. This actually makes
17 sense in that if there is a concentration gradient of TSP within a city, and if the various TSP
18 concentrations fluctuate together, then using a site with a low mean TSP for time-series analysis
19 would result in a larger coefficient. This result does warn against extrapolating the effects from
20 one city to an other using a raw regression coefficient [excess relative risk]"
21 Other recent studies also point out other aspects of intra-urban spatial variation in PM
22 concentrations. Kinney et al. (2000) note that in a personal and ambient PM2 5 and diesel exhaust
23 particle (DEP) exposure study in a dense urban area of New York City, PM25 concentrations
24 showed only a moderate site-to-site variation (37 to 47 //g/m3), probably due to broader regional
25 sources of PM2 5, whereas elemental carbon concentrations (EC) showed a four-fold range of site-
26 to-site variations, reflecting the greater local variation in EC from DEP.
27 Several PM health studies for the city of Seattle (King County), WA (e.g. Levy et al., 2001,
28 for out-of-hospital primary cardiac arrests) have found few statistically significant relationships,
29 attributed by the authors in part to the fact that Seattle has a topographically diverse terrain with
30 local "hot spots" of residential wood burning, especially in winter. Sheppard et al. (2001) have
31 explored reasons for these findings, particularly focusing on adjustments for location by use of a
April 2002 8-262 DRAFT-DO NOT QUOTE OR CITE
-------
1 "topographic index" that includes the "downstream" normal flow of wood smoke from higher
2 elevations, and the trapping of wood smoke in topographic bowls or basins even at higher
3 elevations. They also adjusted for weather using a "stagnation index" (the average number of
4 hours per day with wind speed less than the 25th percentile of wind speeds), and temperature, as
5 well as interaction terms for stagnation on hilltop sites and temperature at suburban wood-
6 smoke-exposed valley sites. The adjustments for exposure measurement error based on methods
7 developed in (Sheppard and Damian, 2000; Sheppard et al., 2001) had little effect on effect size
8 estimates for the case-crossover study (Levy et al., 2001), but may be useful in other studies
9 where localized effects are believed to be important, particularly for the gaseous co-pollutants.
10 Daniels et al. (2001) evaluated the relative sources of variability or heterogeneity in
11 monitoring PM10 in Pittsburgh, PA in 1996. The site is data-rich, having 25 monitors in a
12 rectangle approximately 40 by 80 km. The authors found no isotropic spatial dependence after
13 accounting for other sources of variability, but an indication of heterogeneity in the variability of
14 the small-scale processes over time and space, and heterogeneity in the mean values and
15 covariate effects across sites. Important covariates included temperature, precipitation, wind
16 speed and direction. The authors concluded that significant unmeasured processes might be in
17 operation. These methods should also be useful in evaluating the spatial and temporal variations
18 in gaseous co-pollutants, where small-scale processes are clearly important.
19
20 8.4.7.3 Measurement Error and the Assessment of Confounding by Co-Pollutants in
21 Multi-Pollutant Models.
22 The discussion in Zeger et al. (2000) may be interpreted as addressing the question of
23 whether the apparent lack of a PM10_2 5 effect in models with both fine and coarse particles
24 demonstrates a "false negative" due to the larger measurement error of coarse particle
25 concentrations. However, the more important question may involve the relative attenuation of
26 estimated effects of PM25 and gaseous co-pollutants, especially those such as CO that are known
27 to be highly correlated with PM2 5. Tables 1 and 2 in (Zeger et al., 2000) may be particularly
28 relevant here. The evidence discussed in this chapter supports the hypothesis that PM has
29 adverse health effects, but leaves open the question as to whether the co-pollutants have effects
30 as well when their exposure is measured much less accurately than that of the PM metric. If both
31 the PM metric and the co-pollutant have effects, Table 1 shows that the co-pollutant effect size
32 estimate may be greatly attenuated and the PM effect size estimate much less so, depending on
April 2002 8-263 DRAFT-DO NOT QUOTE OR CITE
-------
1 the magnitude of the correlation between the true PM and gaseous pollutant exposures, and the
2 correlation between their measurement errors. One would expect that PM2 5, CO, and NO2 would
3 often have a high positive correlation, and their "exposure measurement errors" would also be
4 positively correlated if PM and the gaseous pollutants were positively correlated due to common
5 activity patterns, weather, and source emissions. Thus, the line with corr(x1,x2) = 0.5, var^) =
6 0.5, var(62) = 2, corr^, 52) = 0.7 seems appropriate. This implies that the estimated effect of
7 the more accurately measured pollutant is 64% of the true value, and that of the less accurately
8 measured pollutant is 14% of the true value. In view of the substantially greater spatial
9 heterogeneity of traffic-generated ambient pollutants such as CO and NO2, and the relative
10 (though not absolute) regional spatial uniformity of ambient PM2 5 in some cities, but not in
11 others, it is likely that effect size estimates in multi-pollutant models are attenuated downward to
12 a much greater extent for the gaseous co-pollutants than for the PM metric in some cities, but not
13 in others. This may explain part of the heterogeneity of findings for multi-pollutant models in
14 different cities discussed in Section 8.4.2.2.3. Low effect size estimates for the gaseous co-
15 pollutants in a multi-pollutant model should be interpreted cautiously, as noted in Section
16 8.4.2.2.3. The representativeness of the monitoring sites for population exposure of both the
17 particle metrics and gaseous pollutants should be evaluated as part of the interpretation of the
18 analysis. Indices such as the maximum 90th percentile of the absolute difference in
19 concentrations between pairs of sites as well as the median cross-correlation across sites may be
20 useful for characterizing for spatially heterogeneity of gaseous co-pollutants as well as for fine
21 particles.
22
23 8.4.7.5 Air Pollution Exposure Proxies in Long-Term Mortality Studies
24 The AHSMOG Study of mortality (Abbey et al., 1999; McDonnell et al., 2000), the
25 Harvard 6-Cities Study of mortality (Dockery et al, 1993), the ACS Study (Pope et al., 1995), and
26 the VA/Washington Univ. Study (Lipfert et al., 2000b) together provided a major step forward in
27 the assessment of the long-term effects of air pollution. These cohort studies responded to many
28 of the major criticisms of the prior cross-sectional mortality studies, while largely confirming the
29 results of those prior studies. In particular, unlike the ecological cross-sectional studies, these
30 new cohort studies had individual-level information about the members of the study cohort,
April 2002 8-264 DRAFT-DO NOT QUOTE OR CITE
-------
1 allowing the analysis to more properly control for other major factors in mortality, such as
2 smoking and socio-economic factors.
3 While several of these studies made use of newly available fine particle mass (PM25) data
4 to derive useful estimates of health effects of PM25 well before it was routinely measured, these
5 studies utilized air pollution exposure information in a manner similar to that used in the past
6 studies. These studies used central site metropolitan area (MA) spatial and time averages of air
7 pollution exposures, rather than exposure information at the individual level. For this reason, the
8 AHSMOG, Harvard Six-Cities, ACS, and VA/Washington Univ. studies have been term
9 "semi-individual" cohort studies of air pollution.
10
11 The AHSMOG Study
12 Although this study covers a large number of years (1977-1992 in Abbey et al., 1999), it is
13 considerably more limited in the availability of particle metrics that were actually observed rather
14 than estimated. Prior to 1987, PM10 could only be estimated from TSP, not observed. Likewise,
15 for the more recent years, McDonnell et al. (2000) used participants who lived near an airport so
16 that PM2 5, and PM10_2 5 as the difference of PM10 and PM25, could be estimated from airport
17 visibility data using the method described in an earlier publication (Abbey et al., 1995b). All of
18 these issues add potential measurement error to the exposure estimates.
19
20 The Veterans' Administration/Washington University Study
21 The air pollution concentrations for the participants' counties of residence at the time of
22 enrollment were used in the analyses, rather concentrations at the 32 VA hospitals in the final
23 study. County-wide pollution variables for five particle metrics and three gaseous pollutants
24 were used in the study, although TSP was most often the particle metric observed for the earlier
25 years of the study (before 1975 up to!988), which are important in assessing pollution effects for
26 many years of exposure. However, IPMN data for fine particles and sulfates were available for
27 ca. 1979-1983, as in the ACS study. Effects on average mortality for the intervals 1976-1981,
28 1982-1988, and 1989-1996 were related to multi-year particle exposures for four long intervals:
29 < 1975, 1975-1981, 1982-1988, and 1989-1996. TSP was used in the first three exposure
30 intervals, PM10 in the most recent. This study examined "concurrent" exposures (same interval
31 as average mortality), "causal" prior exposures (exposure interval precedes mortality interval),
April 2002 8-265 DRAFT-DO NOT QUOTE OR CITE
-------
1 and "non-causal" PM vs. mortality associations. The mortality associations were also examined
2 for PM2 5, PM15, and PM15.25 for 1979-1981 and 1982-1984. This study has a considerable
3 amount of air pollution data and should be as adequate as other studies for characterizing fixed-
4 site air pollution concentrations in the place of residence at the time of enrollment. However, if
5 any participants moved away from the county where air pollution is measured, but were retained
6 in the study because they continue to participate in follow-ups at the same clinic, then the use of
7 the initial residence location may not be an adequate proxy for actual exposure after initial
8 enrollment.
9
10 Harvard Six-Cities Air Pollution Exposure Data
11 In the case of the Harvard Six Cities Study, ambient concentrations of fine particles (PM2 5),
12 total suspended particles (TSP), sulfur dioxide (SO2), ozone (O3), nitrogen dioxide (NO2), and
13 sulfate (SO4=) were measured at a centrally located air monitoring station established within each
14 of the six communities. Long-term mean concentrations for each pollutant were calculated for
15 periods that were consistent among the six cities, but not across pollutants. The original
16 epidemiologic analysis characterized ambient air quality as long-term mean concentrations of
17 total particles (TSP) (1977-1985), inhalable and fine particles (1979-1985), sulfate particles
18 (1979-1984), aerosol acidity (H+ ) (1985-1988), sulfur dioxide (1977-1985), nitrogen dioxide
19 (1977-1985), and ozone (1977-1985), as follows:
20 Gases: The gases (SO2, NO2, and O3) were monitored hourly by conventional continuous
21 instrumentation in parts per billion.
22
23 Particles: Mean PM concentrations were reported for four classifications of particles in each of
24 the six cities: TSP (particles with aerodynamic diameters up to 50 //m), inhalable particles, fine
25 particles, and sulfate particles. Values of mass for TSP and sulfate particles were determined
26 from 24-h high-volume samplers. Inhalable particle mass was calculated from coarse and fine
27 particle mass, which had been determined from 24-h sample pairs collected by dichotomous
28 samplers. In these, the fine particle channel collected particles smaller than about 2.5 //m and the
29 measurement was recorded directly as fine particle (FP) mass. The coarse particle channel
30 collected particles between 2.5 //m and 10 or 15 //m in aerodynamic diameter (the upper bound
31 measurement depended on the inlet size used at the time).
April 2002 8-266 DRAFT-DO NOT QUOTE OR CITE
-------
1 Acidity: Aerosol acidity (H+) was measured for about one year in each city. However,
2 measurements were conducted in only two cities at a time. Thus, it was not possible to compare
3 acidity for a common time period. Furthermore, the acidity data were not linked with particle
4 data in the same city. Thus, intercity and inter-pollutant comparisons of FT in this study were
5 confounded by inter-annual variability.
6
7 ACS Study Air Pollution Exposure Data
8 In the ACS Study (Pope et al., 1995), two measures of particulate air pollution, were
9 considered: fine particles and sulfate. No gaseous pollutants were considered. The mean
10 concentration of sulfate air pollution by metropolitan area (MA) during 1980 was estimated using
11 data from the EPA Aerometric Information Retrieval System (AIRS) database. These means
12 were calculated as the averages of annual arithmetic mean 24-h sulfate values for all monitoring
13 sites in the 151 MA's considered. The median concentration of fine particles between 1979 and
14 1983 was estimated from the EPA's dichotomous sampler network. These estimates of fine
15 particle levels had been used previously in a population-based cross-sectional mortality study of
16 50 MA's. Gaseous co-pollutants were not considered in Pope et al's original ACS analysis.
17
18 Six-City Study and ACS Exposure Data Strengths and Weaknesses
19 In each of these studies, there was a single mean pollution concentration assigned for each
20 city for each pollutant for the entire follow-up period considered. Concentrations were not
21 broken into each year or sub-groups of years (e.g., 5 year averages), largely because data were not
22 available in this form. This may represent a significant weakness, as a single number could not
23 accurately account for the different exposures in different years of follow-up. However, it is
24 possible that the simultaneous or immediately preceding years alone might not as well represent
25 the effects of long-term pollution exposure.
26 The ACS analysis also uses metropolitan area (MA) pollutant concentrations for air
27 pollution exposure estimates, rather than individual level measurements. Thus, spatial variability
28 in air pollution levels and potential effects of different housing infiltration rates were not
29 addressed as potential factors in exposure variability. However, individual exposure data would
30 be economically impractical for such large cohorts, and the use of more localized measurements
31 (e.g., by county) might well lead to more error, due to day-to day mobility between counties by
April 2002 8-267 DRAFT-DO NOT QUOTE OR CITE
-------
1 individuals (e.g., to work and back) and changes of specific residence within an MA over time.
2 Thus, the MA average may yet be the best metric that can be developed in the absence of
3 individual level exposure data.
4 Another notable weakness of the original ACS Study was that only two PM air pollution
5 metrics were considered. Thus, this study did not consider the potentially confounding
6 influences of gaseous air pollutants or other particle indicators.
7 These two studies' analyses assign the subjects' residence MA on the basis of where they
8 were enrolled, which can lead to exposure errors if the subjects moved to another MA during the
9 follow-up period. However, a recent reanalysis of the Six Cities Study cohort (Krewski et al.,
10 2000) indicates that mobility in these older populations is limited, with only 18.5% leaving the
11 original city of enrollment over subsequent decades.
12
13 The HEI Reanalysis of the ACS Study
14 The HEI Reanalysis of these two cohort studies (Krewski et al, 2000) confirmed the
15 databases used in these two studies, but also developed new exposure data for the ACS Study
16 cohort. In particular, data for the gaseous pollutants (for the year 1980) were added to the
17 analysis. Table 8-43 below displays summary data for the most recent data available for the
18 analysis of the ACS cohort (Pope et al., 2002). The variables noted with the data source "HEI"
19 were added to the analysis during the HEI reanalysis. These HEI results largely confirmed the
20 original ACS analysis results for PM, but also indicated that SO2 was also correlated with U.S.
21 mortality.
22
23 The 16-Year Follow-Up of the ACS Cohort
24 Also included in Table 8-43 are summaries of the pollutant data developed to provide
25 exposure estimates for the latest 16-year follow-up analysis of the ACS cohort (Pope et al, 2002).
26 These new data are similarly city-wide averages of all monitoring stations in the MA's
27 considered, but for the entire period of follow-up (1982-1998), when possible. In addition, this
28 new analysis has incorporated the new PM25 air monitoring data collected routinely from 1999
29 onward. As a result, this new analysis has increased the analysis power both by extending the
30 length of follow-up, and by adding significant new multiple and multi-year air pollution exposure
31 data to the analysis.
April 2002 8-268 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-43. SUMMARY OF ACS POLLUTION INDICES: UNITS, PRIMARY
SOURCES, NUMBER OF CITIES AND SUBJECTS AVAILABLE FOR ANALYSIS,
AND THE MEAN LEVELS (standard deviations)
Pollutant
(years of data)
PM25 (79-83)
PM2 5 (99-00)
PM2 5 (ave)
PM10 (82-98)
PM15 (79-83)
PM15.25 (79-83)
TSP (80-81)
TSP (79-83)
TSP (82-98)
SO4 (80-81)
SO4 (90)
SO2 (80)
SO2 (82-98)
NO2 (80)
NO2 (82-98)
CO (80)
CO (82-98)
03 (80)
O3 (82-98)
O3 (82-98 3rd Q.)
Units
Mg/rn3
Mg/nf
Mg/ni3
^tg/rn3
^tg/rn3
Mg/m3
Mg/m3
^g/m3
Mg/m3
Mg/m3
Mg/m3
ppb
ppb
ppb
ppb
ppm
ppm
ppb
ppb
ppb
Sources of Data*
IPMN (HEI)
AIRS (NYU)
Average of two above
AIRS (NYU)
IPMN (HEI)
IPMN (HEI)
NAD (HEI.)
IPMN (HEI)
AIRS (NYU)
IPMN and NAD,
artifact adjusted (HEI)
NYU compilation and
analysis of PM10 filters
AIRS (HEI)
AIRS (NYU)
AIRS (HEI)
AIRS (NYU)
AIRS (HEI)
AIRS (NYU)
AIRS (HEI)
AIRS (NYU)
AIRS (NYU)
No. of Metro
Areas
61
116
51
102
63
63
156
58
150
149
53
118
126
78
101
113
122
134
119
134
No. of Sub.
(1000s)
359
500
319
415
359
359
590
351
573
572
269
520
539
409
493
519
536
569
525
557
Mean (SD)
21.1 (4.6)
14.0 (3.0)
17.7(3.7)
28.8 (5.9)
40.3 (7.7)
19.2(6.1)
68.0(16.7)
73.7(14.3)
56.7(13.1)
6.5 (2.8)
6.2 (2.0)
9.7 (4.9)
6.7 (3.0)
27.9 (9.2)
21.4(7.1)
1.7 (0.7)
1.1(0.4)
47.9(11.0)
45.5 (7.3)
59.7(12.8)
Source: Pope et al. (2002)
1 Conclusions
2 The pollution exposure data used in these studies, while state-of-the-art when they were
3 conducted, have weaknesses, most notably that these studies, of necessity, have employed city-
4 wide estimates of air pollution exposure, rather than individual-level exposure data. In the case
5 of the mortality control variables (e.g., race and education), the use of individual-level data did
April 2002
8-269
DRAFT-DO NOT QUOTE OR CITE
-------
1 not significantly change the air pollution effect estimates from those given by prior "ecological"
2 cross-sectional mortality analyses using MA aggregate data (e.g., Ozkaynak and Thurston, 1987).
3 Future research into the human health effects of long-term air pollution exposures needs to
4 similarly assess whether the use of individual level exposure data would or would not
5 substantially change the pollution effect estimates.
6
7 8.4.9 Heterogeneity of Particulate Matter Effects Estimates
8 Approximately 35 then-available acute PM exposure community epidemiologic studies
9 were assessed in the 1996 PM AQCD as collectively demonstrating increased risks of mortality
10 being associated with short-term (24-h) PM exposures indexed by various ambient PM
11 measurement indices (e.g., PM10, PM2 5, BS, COH, sulfates, etc.) in many different cities in the
12 United States and internationally. Much homogeneity appeared to exist across various
13 geographic locations, with many studies suggesting, for example, increased relative risk (RR)
14 estimates for total nonaccidental mortality on the order of 1.025 to 1.05 (or 2.5 to 5.0% excess
15 deaths) per 50 //g/m3 increase in 24-h PM10, with statistically significant results extending more
16 broadly in the range of 1.5 to 8.0%. The elderly > 65 yrs. old and those with preexisting
17 cardiopulmonary conditions had somewhat higher excess risks. One study, the Harvard Six City
18 Study, also provided estimates of increased RR for total mortality falling in the range of 1.02 to
19 1.056 (2.0 to 5.6% excess deaths) per 25 //g/m3 24-h PM2 5 increment.
20 Now, more than 80 new time-series PM-mortality studies assessed earlier in this chapter
21 provide extensive additional evidence which, qualitatively, largely substantiates significant
22 ambient PM-mortality relationships, again based on 24-h exposures indexed by a wide variety of
23 PM metrics in many different cities of the United States, in Canada, in Mexico, and elsewhere (in
24 South America, Europe, Asia, etc.). The newly available effect size estimates from such studies
25 are reasonably consistent with the ranges derived from the earlier studies reviewed in the 1996
26 PM AQCD. For example, newly estimated PM10 effects generally fall in the range of 1.0 to 8.0%
27 excess deaths per 50 //g/m3 PM10 increment in 24-h concentration; whereas new PM25 excess
28 estimates for short-term exposures generally fall in the range of 2 to 8% per 25 //g/m3 increment
29 in 24-h PM2 5 concentration.
30 However, somewhat greater spatial heterogeneity appears to exist across newly reported
31 study results, both with regard to PM-mortality and morbidity effects. The newly apparent
April 2002 8-270 DRAFT-DO NOT QUOTE OR CITE
-------
1 heterogeneity of findings across locations is perhaps most notable in relation to reports based on
2 multiple-city studies in which investigators used the same analytical strategies and models
3 adjusted for the same or similar co-pollutants and meteorological conditions, raising the
4 possibility of different findings reflecting real location-specific differences in exposure-response
5 relationships rather than potential differences in models used, pollutants measured and included
6 in the models, etc. Some examples of newly reported and well-conducted multiple-city studies
7 include: the NMMAPS analyses of mortality and morbidity in 20 and 90 U.S. cities (Samet et al.,
8 2000a,b; Dominici et al., 2000a); the Schwartz (2000b,c) analyses of 10 U.S. cities; the study of
9 eight largest Canadian cities (Burnett et al., 2000); the study of hospital admissions in eight U.S.
10 counties (Schwartz, 1999); and the APHEA studies of mortality and morbidity in several
11 European cities (Katsouyanni et al., 1997; Zmirou et al., 1998). The recently completed large
12 NMMAPS studies of morbidity and mortality in U.S. cities add especially useful and important
13 information about potential U.S. within- and between-region heterogeneity.
14
15 8.4.9.1 Evaluation of Heterogeneity of Particulate Matter Mortality Effect Estimates
16 In all of the U.S. multi-city analyses, the heterogeneity in the PM estimates across cities
17 was not explained by city-specific characteristics in the 2nd stage model. The heterogeneity of
18 effects estimates across cities in the multi-city analyses may be due to chance alone, to mis-
19 specification of covariate effects in small cities, or to real differences from location to location in
20 effects of different location-specific ambient PM mixes, for which no mechanistic explanations
21 are yet known. Or, the apparent heterogeneity may simply reflect imprecise PM effect estimates
22 derived from smaller-sized analyses of less extensive available air pollution data or numbers of
23 deaths in some cities tending to obscure more precise effects estimates from larger-size analyses
24 for other locations, which tend to be consistently more positive and statistically significant.
25 Some of these possibilities can be evaluated by using data from the NMMAPS study
26 (Samet et al., 2000b). Data in Figure 8-3 were optically scanned and digitized, producing
27 reasonably accurate estimates by comparison with the 20 largest U.S. cities in their Table A-2.
28 The cities were divided among 7 regions, and excess risk with 95% confidence intervals plotted
29 against the total number of effective observations, measured by the number of days of PM10 data
30 times the mean number of daily deaths in the community. This provides a useful measure of the
31 weight that might be assigned to the results, since the uncertainty of the RR estimate based on a
April 2002 8-271 DRAFT-DO NOT QUOTE OR CITE
-------
1 Poisson mean is roughly inversely proportional to this product. That is, the expected pattern
2 typically shows less spread of estimated excess risk with increasing death-days of data. A more
3 refined weight index would also include the spread in the distribution of PM concentrations. The
4 results are plotted in Figure 8-30 for all cities and Figure 8-31 for each of the 7 regions.
5 Figure 8-30 for all cities suggests some relationship between precision of the effects
6 estimates and study weight, overall. That is, the more the mortality-days observations, the
7 narrower the 95% confidence intervals and the more precise the effects estimates (with nearly all
8 these for cities with > log 9 mortality-days being positive and many statistically significant at
9 p < 0.05).
10 The Figure 8-31 depiction for each of the 7 regions is also informative. In the Northeast,
11 there is considerable homogeneity (not heterogeneity) of effect size for larger study-size cities,
12 even with moderately wide confidence intervals for those with log mortality-days = 8 to 9, and all
13 clearly exceed the overall nationwide grand mean indicated by the dashed line. On the other
14 hand, the smaller study-size Northeast cities (with much wider confidence intervals at log
15 < 8) show much greater heterogeneity of effects estimates and less precision. Also, most of the
16 estimates for larger study-size (log > 9) cities in the industrial midwest are positive and several
17 statistically significant, so that an overall significant regional risk is plausible there as well. There
18 may even be some tendency for relatively large risks for some cities with small study sizes and
19 wide confidence intervals in the industrial midwest, and further investigation of that would be of
20 interest. The plot for Southern California in Figure 8-31 clearly shows a rather consistent
21 estimate of effect size and width of the confidence intervals across cities of varying study-size.
22 All risk estimates are positive and most are significant at p < 0.05 or nearly so for the Southern
23 California cities. For Northwestern cities plotted in Figure 8-31, the value for Oakland, CA (at
24 ca. log 9.5) is notable (it being very positive and significant), whereas many but not all of the
25 other cities have positive effect estimates not too far off the nationwide grand mean, but with
26 sufficiently wide confidence intervals so as not to be statistically significant at p < 0.05. The
27 Southwestern cities (except for 2 cities), too, mostly appear to have effect sizes near the
28 nationwide mean, but with confidence intervals too wide to be significant at p < 0.05. The
29 "Other" (non-industrial or "Upper", as per NMMAPS) Midwest cities and the Southeastern cities
30 in Figure 8-31 show more heterogeneity, although most of the larger study size cities (log > 9.0)
31 tend to be positive and not far off the nationwide mean (even though not significant at p < 0.05).
April 2002 8-272 DRAFT-DO NOT QUOTE OR CITE
-------
5.4
CO
1.8-
0.0
C/)
CO
0
O
X
LU
-5.4
I I I
All Cities
-1.8-
Natural Log of Mortality (Days)
Figure 8-30. The EPA-derived plot showing relationship of PM10 total mortality effects
estimates and 95% confidence intervals for all cities in the Samet et al.
(2000a,b) NMMAPS 90-cities analyses in relation to study size (i.e., the
natural logarithm or numbers of deaths times days of PM observations).
Note generally narrower confidence intervals for more homogeneously
positive effects estimates as study size increases beyond about In (mortality-
days) (i.e., beyond about 8,000 deaths-days of observation). The dashed line
depicts the overall nationwide effect estimate (grand mean) of approximately
0.5% per 10 fJ-g/m3 PM10 for models with no co-pollutants.
April 2002
8-273
DRAFT-DO NOT QUOTE OR CITE
-------
5
CL
CO
| 1.8
8.0.0
« -1.8
'(T
Ul
Ul
01
o
X
-
-
Northwest
[fiV-<
s
-
-
• J
Industrial Midwest
JU T T -T-
iff Ti I ^
1 -"-11
-
-
i
_
-
i i i
Northeast
tiff I"
hp-
i i i
7 8 9 10 11 7 8 9 10 11 12 13 6 7 8 9 10 11 12
Natural Log of Mortality (Days) Natural Log of Mortality (Days) Natural Log of Mortality (Days)
5.4
CL
CO
a o.o
tr
Ul
Ul
01
o
X
1 1 1
Southern California
"TI 1
11
-
ill
i *
-
•
-
- .
Other Midwest
i
i
• i
T
-f--'
j
• '
•
v
•
j
L_L
i
Southeast
h ^R
-L
8 9 10 11 12
Natural Log of Mortality (Days)
7 8 9 10 11 12
Natural Log of Mortality (Days)
8 9 10 11
Natural Log of Mortality (Days)
CL
CO
5.4
1.8
8.0.0
« -1'8
ui
ui
x
UJ r- ,
Southwest
'8.0 8.5 9.0 9.5 10.0
Natural Log of Mortality (Days)
Figure 8-31. The EPA-derived plots showing relationships of PM10-mortality (total,
nonaccidental) effects estimates and 95% confidence intervals to study size
(defined as Figure 8-10) for cities broken out by regions as per the NMMAPS
regional analyses of Samet et al. (2000a,b). Dashed line on each plate depicts
overall nationwide effect estimate (grand mean) of approximately 0.5% per
10 Aig/m3 PM10 for models with no co-pollutants.
April 2002
8-274
DRAFT-DO NOT QUOTE OR CITE
-------
1 Given the wide range of effects estimates and confidence intervals seen for Southeastern cities,
2 further splitting of the region might be informative.
3 In fact, closer reexamination of results for each of the regions may reveal interesting new
4 insights into what factors may account for any apparent disparities among the cities within a
5 given region or across regions. Several possibilities readily come to mind. First, cursory
6 inspection of the mean PM10 levels shown for each city in (Samet et al., 2000b; Appendix A)
7 suggests that many of the cities showing low effects estimates and wide confidence intervals tend
8 to be among those having the lowest mean PM10 levels and, therefore, likely the smallest range of
9 PM10 values across which to distinguish any PM-related effect, if present. It may also be possible
10 that those areas with higher PM25 proportions of PM10 mass (i.e., larger percentages of fine
11 particles) may show higher effects estimates (e.g., in Northeastern cities) than those with higher
12 coarse-mode fractions (e.g., as would be more typical of Southwestern cities). Also, more
13 industrialized cities with greater fine-particle emissions from coal combustion (e.g., in the
14 industrial Midwest) and/or those with high fine-particle emissions from heavy motor vehicle
15 emissions (e.g., typical of Southern California cities) may show larger PM10 effects estimates
16 than other cities. Lastly, the extent of air-conditioning use may also account for some of the
17 differences, with greater use in many Southeastern and Southwestern cities perhaps decreasing
18 actual human exposure to ambient particles present versus higher personal exposure to ambient
19 PM (including indoors) in those areas where less air-conditioning is used (e.g., the Northeast and
20 industrial Midwest). See, for example, Janssen et al. (2002) results reproduced as Figure 8-11.
21
22 8.4.9.2 Comparison of Spatial Relationships in the NMMAPS and Cohort Reanalyses
23 Studies
24 Both the NMMAPS and HEI Cohort Reanalyses studies had a sufficiently large number of
25 U.S. cities to allow considerable resolution of regional PM effects within the "lower 48" states,
26 but an attempt was made to take this approach to a much more detailed level in the Cohort
27 Reanalysis studies than in NMMAPS. There were: 88 cities with PM10 effect size estimates in
28 NMMAPS; 50 cities with PM25 and 151 cities with sulfates in the original Pope et al. (1995)
29 ACS analyses and in the HEI reanalyses using the original data; and 63 cities with PM2 5 data and
30 144 cities with sulfate data in the additional analyses done by the HEI Cohort Reanalysis team.
31 The relatively large number of data points utilized in the HEL reanalyses effort and additional
April 2002 8-275 DRAFT-DO NOT QUOTE OR CITE
-------
1 analyses allowed estimation of surfaces for elevated long-term concentrations of PM2 5, sulfates,
2 and SO2 with resolution on a scale of a few tens to hundreds of kilometers.
3 The patterns for PM2 5 and sulfates are similar, but not identical. In particular, the modeled
4 PM25 surface (Krewski et al., 2000; Figure 18) has peak levels around Chicago - Gary, in the
5 eastern Kentucky - Cleveland region, and around Birmingham AL, with elevated but lower PM2 5
6 almost everywhere east of the Mississippi, as well as southern California. This is similar to the
7 modeled sulfate surface (Krewski et al., 2000; Figure 16), with the absence of a peak in
8 Birmingham and an emerging sulfate peak in Atlanta. The only area with markedly elevated SO2
9 concentrations is the Cleveland - Pittsburgh region. A preliminary evaluation is that secondary
10 sulfates in particles derived from local SO2 are more likely to be important in the industrial
11 midwest, south from the Chicago - Gary region into Ohio, northeastern Kentucky, West Virginia,
12 and southwest Pennsylvania, possibly related to combustion of high-sulfur fuels.
13 The overlay of mortality with air pollution patterns is also of much interest. The spatial
14 overlay of long-term PM2 5 and mortality (Krewski et al., 2000; Figure 21) is highest from
15 southern Ohio to northeastern Kentucky /West Virginia, but also includes a significant association
16 over most of the industrial midwest from Illinois to the eastern non-coastal parts of North
17 Carolina, Virginia, Pennsylvania, and New York. This is reflected, in diminished form, by the
18 sulfates and SO2 maps (Krewski et al., 2000; Figures 19 and 20), where there appears to be a
19 somewhat tighter focus of elevated risk in the upper Ohio River Valley area. This suggests that,
20 while S02 may be an important precursor of sulfates in this region, there may also be some other
21 (non-sulfur) contributors to associations between PM2 5 and long-term mortality, embracing a
22 wide area of the North Central Midwest and non-coastal Mid-Atlantic region.
23 It should be noticed that, while a variety of spatial modeling approaches were discussed in
24 the NMMAPS methodology report (NMMAPS Part I, pp. 66-71 [Samet et al., 2000a]), the
25 primary spatial analyses in the 90-city study (NMMAPS, Part II [Samet et al., 2000b]) were
26 based on a simpler seven-region breakdown of the contiguous 48 states. The 20-city results
27 reported for the spatial model in NMMAPS I show a much smaller posterior probability of a
28 PM10 excess risk of short-term mortality, with a spatial posterior probability vs. a non-spatial
29 probability of a PM10 effect of 0.89 vs. 0.98 at lag 0, of 0.92 vs. 0.99 at lag 1, and of 0.85 vs. 0.97
30 at lag 2. The evidence that PM10 is associated with an excess short-term mortality risk is still
31 moderately strong with a spatial model, but less strong than with a non-spatial model.
April 2002 8-276 DRAFT-DO NOT QUOTE OR CITE
-------
1 The apparently substantial differences in PM10 and/or PM2 5 effect sizes across different
2 regions should not be attributed merely to possible variations in measurement error or other
3 statistical artifact(s). Some of these differences may reflect: real regional differences in particle
4 composition or co-pollutant mix; differences in relative human exposures to ambient particles or
5 other gaseous pollutants; sociodemographic differences (e.g., percent of infants or elderly in
6 regional population); or other important, as of yet unidentified PM effect modifiers.
7
8 8.4.9.3 Epidemiologic Studies of Ambient Air Pollution Interventions
9 To date, assessment of health risk in epidemiologic studies of ambient air pollutants,
10 including PM, has relied largely on studies that focus on increases in exposure, and that inquire
11 whether health risk changes in relation to such increases. Such studies are used to support
12 qualitative and quantitative inference as to whether decreases in exposure will bring about
13 reductions in health risk, or improvement in health status.
14 Ambient criteria air pollutants are rarely, if ever, the only etiology of the health disorders
15 with which exposure to these pollutants is associated. For example, numerous reports have
16 implicated ambient air pollution exposure with exacerbations of pre-existing asthma. These
17 reports justify the expectation that further reduction in ambient air pollution exposure would
18 reduce the public health burden of asthma exacerbations. However, many other factors,
19 including allergens, passive smoking, exercise, cold, and stress are also associated with such
20 exacerbations. Asthmatics would continue to be exposed to these factors even with further
21 reduction in ambient air pollution exposure.
22 Thus, reduction of ambient air pollution exposure, even to zero concentration, would not
23 bring about zero risk of the health disorders with which such exposure is associated. Also, it is
24 likely that at least some non-pollution risk factors would behave differently in the absence of
25 ambient air pollution exposure as in its presence. That is, in the real world, risk factors probably
26 do not behave in discrete, additive fashion.
27 Therefore, truly quantitative characterization of effects of reduction in air pollution
28 concentrations and exposures requires study of situations in which such reductions actually
29 occur. In such studies, it is important to measure both exposure and health status before and after
30 exposure is reduced. It is also highly desirable to identify risk factors other than ambient air
31 pollution, and to ascertain their effects before and after air pollution exposure reduction.
April 2002 8-277 DRAFT-DO NOT QUOTE OR CITE
-------
1 In his classic monograph (The Environment and Disease: Association or Causation?), Hill
2 (1965) addressed the topic of preventive action and its consequences under Aspect 8, stating:
3 "Experiment: Occasionally it is possible to appeal to experimental, or semi-experimental,
4 evidence. For example, because of an observed association some preventive action is taken. Does
5 it in fact prevent? The dust in the workshop is reduced, lubricating oils are changed, persons stop
6 smoking cigarettes. Is the frequency of the associated events affected? Here the strongest support
7 for the causation hypothesis may be revealed."
8 The available epidemiologic literature on ambient air pollution offers a limited evidence
9 related to this aspect. In these studies, air pollution concentrations have been temporarily or
10 permanently reduced through regulatory action, industrial shutdown, or other intervening
11 factor(s).
12 In the U.S., the most thoroughly studied example of such ambient air pollution reduction
13 occurred in the Utah Valley, UT, during the 1980s. The Valley's largest stationary source of PM,
14 a steel mill, was closed due a labor dispute for 13 months from autumn 1986 until autumn 1987.
15 This offered the opportunity to study health effects not only of the closure-related reduction in
16 ambient PM concentrations, but also of the increases in PM that occurred after the re-opening of
17 the mill. Pope et al. have reported extensively on such health effects. These reports were
18 addressed in more detail in the 1996 PM AQCD than in the present document. Briefly, these
19 investigators observed reduction in frequency of a variety of health disorders during the period in
20 which the mill was closed. These included daily mortality (Pope et al., 1992), respiratory
21 hospital admissions (Pope, 1989), bronchitis and asthma admissions for preschool children
22 (Pope, 1991), reductions in lung function (Pope et al., 1991), and elementary school absences
23 (Ransom and Pope, 1992). Changes in these endpoints were reflected by differing strength of
24 positive associations between measures of these health endpoints and PM mass measurements
25 from filters collected before, during, and after the steel mill shut down.
26 Five experimental studies investigated effects of aqueous extracts of ambient Utah Valley
27 particulate filters employing filter extracts from January through March 1986 (mill open), 1987
28 (mill closed), and 1988 (mill open) (Frampton et al., 1999; Dye et al., 2001; Soukup et al., 2000;
29 Wu et al., 2001; and Ohio and Devlin, 2001). In all of these studies, investigators observed less
30 intense in vivo or in vitro effects when treating with the 1987 extracts than when treating with
31 extracts from 1986 and/or 1987 (see Chapter 7 of this document).
April 2002 8-278 DRAFT-DO NOT QUOTE OR CITE
-------
1 Frampton et al. (1999) state that extracts were taken from filters collecting PM10, and that a
2 total of 36 filters was used, 12 per year. Soukup et al. (2000) state that PM10 filters were used,
3 and that 34 filters per year were used (total 102 filters). Dye et al. (2001) state that TSP filters
4 were extracted, and that 12 filters per year were used (total 36 filters). Wu et al. (2001) state that
5 PM10 filters were used, and a total of 102 filters was used. Ohio and Devlin (2001) state that
6 "filters containing PM10" were extracted, and that 34 filters each year were used (total 102
7 filters). Taken together, these descriptions raise the question whether the two studies that
8 employed 12 filters per year (Frampton et al. and Dye et al.) were using TSP filters exclusively,
9 whereas the other three studies, that employed 34 filters per year, employed a mixture of TSP
10 filters and PM10 filters. In any event, the degree of comparability of source filters among these
11 five studies is not entirely clear. Also, there is some uncertainty as to the within-study
12 comparability of filters from year to year, particularly in the studies that employed 34 filters per
13 year. Furthermore, a substantial proportion of the extracted material was probably derived from
14 filter matrix, not ambient PM, and about 10 years elapsed between collection and extraction of
15 the filter samples.
16 Even so, the combined results of these five experimental studies provide support and
17 corroboration for the epidemiologic observations of reduced frequency and severity of health
18 disorders during the period of steel mill closure. The experimental studies also provide
19 hypotheses regarding potential biological mechanisms underlying some of the observed effects.
20 Perhaps the strongest of these hypotheses is that PM-associated metals were etiologically related
21 to some of the observed disorders, and that reduction in ambient concentrations of these metals
22 was at least partially responsible for the health benefits observed during steel mill closure. In any
23 event, these experimental studies underscore the importance of particle composition in
24 production or promotion of harmful health effects (Beckett, 2001).
25 Avol and colleagues investigated effects of reductions and increases in ambient air
26 pollution concentrations on longitudinal lung function growth in a subsample of participants in
27 the Children's Health Study conducted by the University of Southern California (Avol et al.,
28 2001). Follow-up lung function tests were administered to 110 children who had moved away
29 from the study area after the baseline lung function test, which was administered while the
30 children lived within the study area. Lung function growth rates were analyzed against
31 differences between the children's original and new communities in annual average
April 2002 8-279 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentrations of PM10, NO2, and O3. Analytical models were adjusted for anthropometric
2 variables and other relevant covariates. No multi-pollutant analyses were reported. Moving to a
3 community with lower ambient PM10 concentration was associated with increased growth rates of
4 FVC, FEV1, MMEF and PEFR, and moving to a community with higher PM10 concentrations
5 was associated with decreased growth of these metrics. These associations were statistically
6 significant for MMEF and PEFR, and appear to have been marginally significant for FVC and
7 FEV1. Moving to a community with lower ambient NO2 or O3 concentration was generally
8 associated with increased lung function growth, and vice versa. However, associations of change
9 in lung function growth with change in community levels of NO2 and O3 were not statistically
10 significant. This study suggests that reduction in long-term ambient PM10 levels is indeed
11 associated with improvement of children's lung growth, and that increase in these levels is
12 associated with retardation of lung growth.
13 Friedman et al. (2001) investigated the influence of temporary changes in transportation
14 behaviors (instituted to reduce downtown traffic congestion during the 1996 Summer Olympic
15 Games in Atlanta, GA) on ambient air quality and acute care visits and hospitalizations for
16 asthma in children residing in Atlanta. Ambient air quality and childhood asthma during the
17 17 days of the Games were compared to those during a baseline period consisting of the four
18 weeks before and the four weeks after the Games. During the Games, concentrations of PM10
19 (24-h average), O3 (daily peak 1-h average), CO (8-h average), and NO2 (daily peak 1-h average)
20 were, respectively, 16.1%, 27.9%, 18.5%, and 6.8% lower than during the baseline period.
21 Twenty-four hour average concentrations of SO2 were 22.1% higher during the Games than
22 during the baseline period. Reductions in O3, PM10, and carbon monoxide were statistically
23 significant at alpha = 0.05 (p = 0.01, p < 0.001, and p = 0.02, respectively). Ambient mold
24 counts during the Games did not differ significantly from those during the baseline period. Four
25 sources of asthma frequency data were examined: (1) the Georgia Medicaid claims file; (2) files
26 of a health maintenance organization; (3) emergency department records for two of Atlanta's
27 three pediatric hospitals; and (4) the Georgia Hospital Discharge Database. For all four sources,
28 asthma-related unadjusted and adjusted relative risks during the Games were less than 1 (as
29 compared to RR = 1 during the baseline period). Relative risks from the Medicaid database were
30 statistically significant (p < 0.005), and those from the HMO approached significance (p < 0.10).
31 These findings suggest strongly that, in Atlanta in summer 1996, temporary improvement in
April 2002 8-280 DRAFT-DO NOT QUOTE OR CITE
-------
1 ambient air quality contributed to temporary reduction in severity of pre-existing asthma. This
2 reduction could not be attributed specifically to any individual air pollutant. In the opinion of
3 Friedman et al., reductions in morning rush-hour traffic played an important role in reduction of
4 asthma-related visits and hospitalizations.
5 Heinrich et al. (2000) studied effects of long-term air pollution reduction in the former East
6 Germany on prevalence of respiratory illnesses and symptoms in 5 to 14 year-old children.
7 Cross-sectional surveys were conducted in 1992-1993 and 1995-1996 in three areas, all of which
8 experienced reductions in annual mean ambient SO2 and TSP concentrations in the time interval
9 between the surveys. Percentage reductions in SO2 and TSP were substantial, ranging from
10 about 40%-60% and about 20%-35%, respectively, in the three areas. Longitudinal changes were
11 not measured for size-specific PM metrics. After adjustment for relevant covariates, statistically
12 significant temporal decreases in prevalences of bronchitis, otitis media, frequent colds, and
13 febrile infections were observed.
14 In Hong Kong, a regulation prohibiting the use of fuel oil containing more than 0.5% sulfur
15 by weight went into effect in July 1990. Investigators from the University of Hong Kong studied
16 respiratory health in children and non-smoking women before and after the regulation was
17 implemented. In a relatively polluted district (District A), the regulation resulted in rapid and
18 substantial reduction in the ambient concentration of sulfur dioxide, and in appreciable but less
19 marked reduction in the concentration of sulfate ion in "respirable suspended particulates" (RSP,
20 thought to be equivalent to PM10). Percentage reductions in these sulfur-containing pollutants
21 were considerably smaller in a less polluted district (District B). The regulation was not
22 accompanied by appreciable reductions in levels of PM metrics (TSP and RSP) in either district.
23 Tarn et al. (1994) reported that the prevalence of bronchial hyperreactivity (BHR) in
24 children (as defined by a > 20% drop in FEV1 in response to histamine challenge) was higher in
25 District A than in District B, even after exclusion of children with wheeze and asthma. Wong
26 et al. (1998) measured BHR prevalence rates in these districts in 1991 and 1992, and compared
27 these to rates before the regulation was implemented. In both districts, BHR prevalence was
28 statistically significantly lower in 1991 than before the intervention. In 1992, the pre- to post-
29 intervention decrease in BHR prevalence was significantly larger in District A than in
30 District B. Peters et al. reported that before the intervention, prevalences of children's respiratory
31 symptoms (e.g., cough, sore throat, wheeze) were statistically significantly higher in District A
April 2002 8-281 DRAFT-DO NOT QUOTE OR CITE
-------
1 than in District B. About one year after the intervention, there were greater pre- to post-
2 intervention declines in prevalences of cough or sore throat, phlegm, and wheezing in District A
3 than in District B. Wong et al. reported that before the intervention, the prevalence of poor
4 respiratory health in non-smoking women was significantly higher in District A than in District
5 B. Also, effects of passive smoking on the women's respiratory health were stronger in District
6 A than in District B, but not significantly so. About one year after the intervention, declines in
7 frequency of poor respiratory health were observed, but these declines did not differ significantly
8 between districts. Taken together, these Hong Kong studies suggest that reduction of sulfur in
9 fuel oil brought about appreciable improvement in children's respiratory health, and discernible
10 but lesser improvement in non-smoking women's respiratory health. These studies also suggest
11 that these benefits were associated with reduction in sulfur-containing ambient air pollutants, but
12 not necessarily with reduction in TSP or RSP per se.
13 Taken together, these epidemiologic intervention studies lend confidence that further
14 reduction of ambient air pollution exposures in the U. S. would benefit public health. It is likely
15 that such reduction would bring about both respiratory and cardiovascular health benefits.
16 Available studies also give reason to expect that further reductions in both particulate and
17 gaseous air pollutants would benefit health. On balance, these studies suggest that selective
18 reduction in ambient PM concentrations might well bring about greater benefit than would
19 selective reduction in concentrations of other ambient criteria air pollutants. Furthermore, the
20 experimental studies of Utah Valley filter extracts point to PM-associated metals as a likely
21 cause or promoter of at least some of the health disorders associated with ambient PM. Beyond
22 this, available epidemiologic intervention studies do not yet give direct, quantitative evidence as
23 to the relative health benefits that would result from selective reduction of specific PM size
24 fractions. Also, these studies do not yet provide firm grounds for quantitative prediction of the
25 relative health benefits of single-pollutant reduction strategies vs. multi-pollutant reduction
26 strategies. Even in an almost ideal "natural experiment" such as Utah Valley, potentially
27 confounding factors other than ambient PM concentrations also changed during the steel mill
28 closure. These included concentrations of other pollutants and possible changes in population
29 due to out- and in-migration influenced by the closing and re-opening of the steel mill. While
30 changes in ambient PM concentrations undoubtedly played a role, other factors may also have
31 modified the size of the changes in health effects.
April 2002 8-282 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.5 KEY FINDINGS AND CONCLUSIONS DERIVED FROM
2 PARTICULATE MATTER EPIDEMIOLOGY STUDIES
3 It is not possible to assign any absolute measure of certainty to conclusions based on the
4 findings of the epidemiology studies discussed in this chapter. However, these observational
5 study findings would be further enhanced by supportive findings of causal studies from other
6 scientific disciplines (dosimetry, toxicology, etc.), in which other factors could be eliminated or
7 controlled, as discussed in Chapters 6 and 7. The epidemiology studies discussed in this chapter
8 demonstrate biologically-plausible responses in humans exposed at ambient concentrations. The
9 most salient conclusions derived from the PM epidemiology studies include:
10
11 (1) A large and reasonably convincing body of epidemiology evidence confirms earlier
12 associations between short- and long-term ambient PM10 exposures (inferred from
13 stationary air monitor measures) and mortality/morbidity effects and suggest that PM10
14 (or one or more PM10 components) is a probable contributing cause of adverse human
15 health effects.
16 (2) It is likely that there is significant spatial heterogeneity in the city-specific excess risk
17 estimates for the relationships between short-term ambient PM10 concentrations and acute
18 health effects. The reasons for such variation in effects estimates are not well understood at
19 this time, but do not negate ambient PM's likely causative contribution to observed PM-
20 mortality and/or morbidity associations in many locations. Possible factors contributing to
21 the heterogeneity include geographic differences in air pollution mixtures, composition of
22 PM components, and personal and sociodemographic factors affecting PM exposure (such
23 as use of air conditioners, education, and so on).
24 (3) A growing body of epidemiology studies confirm associations between short- and long-
25 term ambient PM2 5 exposures (inferred from stationary air monitor measures) and adverse
26 health effects and suggest that PM2 5 (or one or more PM2 5 components) is a probable
27 contributing cause of observed PM-associated health effects. Some new epidemiology
28 findings also suggest that health effects are associated with mass or number concentrations
29 of ultrafme (nuclei-mode) particles, but not necessarily more so than for other ambient fine
30 PM components.
April 2002 8-283 DRAFT-DO NOT QUOTE OR CITE
-------
1 (4) A smaller body of evidence appears to support an association between short-term ambient
2 thoracic coarse fraction (PM10_2 5) exposures (inferred from stationary air monitor measures)
3 and short-term health effects in epidemiology studies. This suggests that PM10_2 5, or some
4 constituent component(s) of PM10_25, may be a contributory cause of adverse health effects
5 in some locations. Reasons for differences among findings on coarse-particle health effects
6 reported for different cities are still poorly understood, but several of the locations where
7 significant PM10_2 5 effects have been observed (e.g., Phoenix, Mexico City, Santiago) tend
8 to be in drier climates and may have contributions to observed effects due to higher levels
9 of organic particles from biogenic processes (endotoxins, molds, etc.) during warm months.
10 Other studies suggest that coarse thoracic fraction (PM10_2 5) particles of crustal origin are
11 generally unlikely to exert notable health effects under most ambient exposure conditions,
12 (however, see Item 14, below). Also, in some western U.S. cities where PM10_25 is a large
13 part of PM10, the relationship between hospital admissions and PM10 may be an indicator of
14 response to coarse thoracic particles from wood burning.
15 (5) Long-term PM exposure durations, on the order of months to years, as well as on the order
16 of a few days, are statistically associated with serious human health effects (indexed by
17 mortality, hospital admissions/medical visits, etc.). More chronic PM exposures, on the
18 order of years or decades, appear to be associated with life shortening well beyond that
19 accounted for by the simple accumulation of the more acute effects of short-term PM
20 exposures (on the order of a few days). Some uncertainties remain regarding the magnitude
21 of and mechanisms underlying chronic health effects of long-term PM exposures and the
22 relationship between chronic exposure and acute responses to short-term exposure.
23 (6) Recent investigations of the public health implications of such chronic PM exposure-
24 mortality effect estimates were also reviewed. Life table calculations by Brunekreef (1997)
25 found that relatively small differences in long-term exposure to airborne PM of ambient
26 origin can have substantial effects on life expectancy. For example, a calculation for the
27 1969-71 life table for U.S. white males indicated that a chronic exposure increase of
28 10 //g/m3 PM was associated with a reduction of 1.31 years for the entire population's life
29 expectancy at age 25. Also, new evidence of associations of PM exposure with infant
30 mortality (Bobak and Leon, 1992, 1999; Woodruff et al., 1997; Loomis et al., 1999) and/or
31 intrauterine growth retardation (Dejmek et al., 1999) and consequent increase risk for many
April 2002 8-284 DRAFT-DO NOT QUOTE OR CITE
-------
1 serious health conditions associated with low birth weight, if further substantiated, would
2 imply that life shortening in the entire population from long-term PM exposure could well
3 be significantly larger than that estimated by Brunekreef (1997).
4 (7) Considerable coherence exists among effect size estimates for ambient PM health effects.
5 For example, results derived from several multi-city studies, based on pooled analyses of
6 data combined across multiple cities (thought to yield the most precise estimates of mean
7 effect size), show the percent excess total (non-accidental) deaths estimated per 50 //g/m3
8 increase in 24-h PM10 to be: 2.3% in the 90 largest U.S. cities (4.5% in the Northeast U.S.
9 region); 3.4% in 10 large U.S. cities; 3.5% in the 8 largest Canadian cities; and 2.0% in
10 western European cities (using PM10 = TSP*0.55). These combined estimates are
11 consistent with the range of PM10 estimates previously reported in the 1996 PM AQCD.
12 These and excess risk estimates from many other individual-city studies, generally falling
13 in the range of ca. 1.5 to 8.0% per 50 //g/m3 24-h PM10 increment, also comport well with
14 numerous new studies confirming increased cause-specific cardiovascular- and respiratory-
15 related mortality. They are also coherent with larger effect sizes reported for cardiovascular
16 (in the range of ca. 3.0 to 10.0% per 50 //g/m3 24-h PM10 increment) and respiratory (in the
17 range of ca. 5 to 25% per 50 //g/m3 24-h PM10) hospital admissions and visits, as would be
18 expected for these morbidity endpoints versus those for PM10-related mortality.
19 (8) Several independent panel studies (but not all) that evaluated temporal associations
20 between PM exposures and measures of heart beat rhythm in elderly subjects provide
21 generally consistent indications of decreased heart rate variability (HRV) being associated
22 with ambient PM exposure (decreased HRV being an indicator of increased risk for serious
23 cardiovascular outcomes, e.g., heart attacks). Other studies point toward changes in blood
24 characteristics (e.g., C-reactive protein levels) related to increased risk of ischemic heart
25 disease also being associated with ambient PM exposures. However, these heart rhythm
26 and blood characteristics findings should currently be viewed as providing only limited or
27 preliminary support for PM-related cardiovascular effects.
28 (9) Notable new evidence now exists which substantiates positive associations between
29 ambient PM concentrations and increased respiratory-related hospital admissions,
30 emergency department, and other medical visits, particularly in relation to PM10 levels.
31 Of much interest are new findings tending to implicate not only fine particle components
April 2002 8-285 DRAFT-DO NOT QUOTE OR CITE
-------
1 but also coarse thoracic (e.g., PM10_2 5) particles as likely contributing to exacerbation of
2 asthma conditions. Also of much interest are emerging new findings indicative of likely
3 increased occurrence of chronic bronchitis in association with (especially chronic) PM
4 exposure. Also of particular interest are reanalyses or extensions of earlier prospective
5 cohort studies of long-term ambient PM exposure effects which demonstrate substantial
6 evidence for association of increased lung cancer risk with such PM exposures, especially
7 exposure to fine PM or its subcomponents.
8 (10) One major methodological issue affecting epidemiology studies of both short-term and
9 long-term PM exposure effects is that ambient PM of varying size ranges is typically found
10 in association with other air pollutants, including gaseous criteria pollutants (e.g, O3, NO2,
11 SO2, CO), air toxics, and/or bioaerosols. Available statistical methods for assessing
12 potential confounding arising from these associations may not yet be fully adequate. The
13 inclusion of multiple pollutants often produces statistically unstable estimates. Omission of
14 other pollutants may incorrectly attribute their independent effects to PM. Second-stage
15 regression methods may have certain pitfalls that have not yet been fully evaluated. Much
16 progress in sorting out relative contributions of ambient PM components versus other
17 co-pollutants is nevertheless being made and, overall, tends to substantiate that observed
18 PM effects are at least partly due to ambient PM acting alone or in the presence of other
19 covarying gaseous pollutants. However, the statistical association of health effects with
20 PM acting alone or with other pollutants should not be taken as an indicator of a lack of
21 effect of the other pollutants. Indeed, the effects of the other pollutants may at times be
22 greater or less than the effects attributed to PM and may vary from place to place or from
23 time to time.
24 (11) It is possible that differences in observed health effects will be found to depend on site-
25 specific differences in chemical and physical composition characteristics of ambient
26 particles and on factors affecting exposure (such as air conditioning) as well as on
27 differences in PM mass concentration. For example, the Utah Valley study (Dockery et al.,
28 1999; Pope et al., 1991, 1999b) showed that PM10 particles, known to be richer in metals
29 during exposure periods while the steel mill was operating, were more highly associated
30 with adverse health effects than was PM10 during the PM exposure reduction while the steel
31 mill was closed. In contrast, PM10 or PM2 5 was relatively higher in crustal particles during
April 2002 8-286 DRAFT-DO NOT QUOTE OR CITE
-------
1 windblown dust episodes in Spokane and in three central Utah sites than at other times, but
2 was not associated with higher total mortality. These differences require more research that
3 may become more feasible as the PM2 5 sampling network produces air quality data related
4 to speciated samples.
5 (12) The above reasons suggest it is inadvisable to pool epidemiology studies at different
6 locations, different time periods, with different population sub-groups, or different health
7 endpoints, without assessing potential causes and the consequences of these differences.
8 Published multi-city analyses using common data bases, measurement devices, analytical
9 strategies, and extensive independent external review, as carried out in the APHEA and
10 NMMAPS studies are likely to be useful. Pooled analyses of more diverse collections of
11 independent studies of different cities, using varying methodology and/or data quality or
12 representativeness, are likely less credible and should not, in general, be used without
13 careful assessment of their underlying scientific comparability.
14 (13) It may be possible that different PM size components or particles with different
15 composition or sources produce effects by different mechanisms manifested at different
16 lags, or that different preexisting conditions may lead to different delays between exposure
17 and effect. Thus, although maximum effect sizes for PM effects have often been reported
18 for 0-1 day lags, evidence is also beginning to suggest that more consideration should be
19 given to lags of several days. Also, if it is considered that all health effects occurring at
20 different lag days are all real effects, so that the risks for each lag day should be additive,
21 then higher overall risks may exist that are higher than implied by maximum estimates for
22 any particular single or two-day lags. In that case, multi-day averages or distributed lag
23 models should be used.
24 (14) Certain classes of ambient particles may be distinctly less toxic than others and may not
25 exert human health effects at typical ambient exposure concentrations or only under special
26 circumstances. Coarse thoracic particles of crustal origin, for example, may be relatively
27 non-toxic under most circumstances compared to those of combustion origin such as wood
28 burning. However, crustal particles may be sufficiently toxic to cause human health effects
29 under some conditions; resuspended crustal particles, for example, may carry toxic trace
30 elements and other components from previously deposited fine PM, e.g., metals from
31 smelters (Phoenix) or steel mills (Steubenville, Utah Valley), PAH's from automobile
April 2002 8-287 DRAFT-DO NOT QUOTE OR CITE
-------
1 exhaust, or pesticides from administration to agricultural lands. Likewise, fine particles
2 from different sources have different effect sizes. More research is needed to identify
3 conditions under which one or another class of particles may cause little or no adverse
4 health effects, as well as conditions under which particles may cause notable effects.
5 (15) Certain epidemiology evidence suggests that reducing ambient PM10 concentrations may
6 reduce a variety of health effects on a time scale from a few days to a few months. This has
7 been found in epidemiology studies of "natural experiments" such as in the Utah Valley,
8 and by supporting toxicology studies using the particles from ambient community sampling
9 filters from the Utah Valley. Recent studies in Germany and in the Czech Republic also
10 tend to support a hypothesis that reductions in air pollution are associated with reductions
11 in the incidence of adverse health effects, but these studies cannot unambiguously attribute
12 improved health to reduced PM alone.
13 (16) Adverse health effects in children are emerging as a more important area of concern than in
14 the 1996 PM AQCD. Unfortunately, relatively little is known about the relationship of PM
15 to the most serious health endpoints (low birth weight, preterm birth, neonatal and infant
16 mortality, emergency hospital admissions and mortality in older children).
17 (17) Little is yet known about involvement of PM exposure in the progression from less serious
18 childhood conditions, such as asthma and respiratory symptoms, to more serious disease
19 endpoints later in life. This is an important health issue because childhood illness or death
20 may cost a very large number of productive life-years. Lastly, new epidemiologic studies
21 of ambient PM associations with increased non-hospital medical visits (physician visits)
22 and asthma effects suggest likely much larger health impacts and costs to society due to
23 ambient PM than just those indexed by mortality and/or hospital admissions/visits.
24
25
26
April 2002 8-288 DRAFT-DO NOT QUOTE OR CITE
-------
REFERENCES
Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
Adventists. Environ. Health Perspect. 94: 43-50.
Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
7: 19-34.
Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
2.5 microns in aerodynamic diameter (PM2 5) from airport visibility data in California. J. Exposure Anal.
Environ. Epidemiol. 5: 161-180.
Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
158: 289-298.
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
Care Med. 159: 373-382.
Ackermann-Liebrich, U.; Leuenberger, P.; Schwartz, J.; Schindler, C.; Monn, C.; Bolognini, B.; Bongard, J. P.;
Brandli, O.; Domenighetti, G.; Elsasser, S.; Grize, L.; Karrer, W.; Keller, R.; Keller-Wossidlo, H.;
Ktinzli, N.; Martin, B. W.; Medici, T. C.; Perruchoud, A. P.; Schoni, M. H.; Tschopp, J. M.; Villiger, B.;
Wuthrich, B.; Zellweger, J. P.; Zemp, E. (1997) Lung function and long term exposure to air pollutants in
Switzerland. Am. J. Respir. Crit. Care Med. 155: 122-129.
Alberdi Odriozola, J. C.; Diaz Jimenez, J.; Montero Rubio, J. C.; Miron Perez, I. J.; Pajares Ortiz, M. S.; Ribera
Rodrigues, P. (1998) Air pollution and mortality in Madrid, Spain: a time-series analysis. Int. Arch. Occup.
Environ. Health 71: 543-549.
Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Strachan, D. P. (1996) Air pollution and daily
mortality in London: 1987-92. Br. Med. J. 312: 665-669.
Anderson, H. R.; Spix, C.; Medina, S.; Schouten, J. P.; Castellsague, J.; Rossi, G.; Zmirou, D.; Touloumi, G.;
Wojtyniak, B.; Ponka, A.; Bacharova, L.; Schwartz, J.; Katsouyanni, K. (1997) Air pollution and daily
admissions for chronic obstructive pulmonary disease in 6 European cities: results from the APHEA project.
Eur. Respir. J. 10: 1064-1071.
Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Emberlin, J.; Strachen, D. P. (1998) Air pollution,
pollens, and daily admissions for asthma in London 1987-92. Thorax 53: 842-848.
Anderson, H. R.; Bremner, S. A.; Atkinson, R. W.; Harrison, R. M.; Walters, S. (2001) Paniculate matter and daily
mortality and hospital admissions in the west midlands conurbation of the United Kingdom: associations with
fine and coarse particles, black smoke and sulphate. Occup. Environ. Med. 58: 504-510.
Archer, V. E. (1990) Air pollution and fatal lung disease in three Utah counties. Arch. Environ. Health 45: 325-334.
Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Ponce de Leon, A. (1999a)
Short-term associations between emergency hospital admissions for respiratory and cardiovascular disease
and outdoor air pollution in London. Arch. Environ. Health 54: 398-411.
Atkinson, R. W.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Bremner, S. A.; Ponce de Leon, A. (1999b)
Short-term associations between outdoor air pollution and visits to accident and emergency departments in
London for respiratory complaints. Eur. Respir. J. 13: 257-265.
Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2001) Acute effects of paniculate air pollution on
respiratory admissions: results from APHEA 2 project. Am. J. Respir. Crit. Care Med. 164: 1860-1866.
Avol, E. L.; Gauderman, W. J.; Tan, S. M.; London, S. J.; Peters, J. M. (2001) Respiratory effects of relocating to
areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
Awasthi, S.; Click, H. A.; Fletcher, R. H.; Ahmed, N. (1996) Ambient air pollution & respiratory symptoms
complex in preschool children. Indian J. Med. Res. 104: 257-262.
Bailey, D. L. R.; Clayton, P. (1982) The measurement of suspended particle and total carbon concentrations in the
atmosphere using standard smoke shade methods. Atmos. Environ. 16: 2683-2690.
April 2002 8-289 DRAFT-DO NOT QUOTE OR CITE
-------
Baldi, I.; Tessier, J. F.; Kauffmann, F.; Jacqmin-Gadda, H.; Nejjari, C.; Salamon, R. (1999) Prevalence of asthma
and mean levels of air pollution: results from the French PAARC survey. Eur. Respir. J. 14: 132-138.
Barbone, F.; Bovenzi, M; Cavalleri, F.; Stanta, G. (1995) Air pollution and lung cancer in Trieste, Italy. Am. J.
Epidemiol. 141: 1161-1169.
Basu, R.; Samet, J. M. (1999) A review of the epidemiological evidence on health effects of nitrogen dioxide
exposure from gas stoves. J. Environ. Med. 1: 173-187.
Bateson, T. F.; Schwartz, J. (1999) Control for seasonal variation and time trend in case-crossover studies of acute
effects of environmental exposures. Epidemiology 10: 539-544.
Beckett, W. S. (2001) The air pollution detectives. Am. J. Respir. Crit. Care Med. 164: 515-516.
Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
Beyer, U.; Franke, K.; Cyrys, J.; Peters, A.; Heinrich, J.; Wichmann, H. E.; Brunekreef, B. (1998) Air pollution and
respiratory health of children: the PEACE panel study in Hettstedt and Zerbst, Eastern Germany. Eur. Respir.
Rev. 8: 61-69.
Blanchard, C. L.; Lurmann, F. W.; Roth, P. M.; Jeffries, H. E.; Korc, M. (1999) The use of ambient data to
corroborate analyses of ozone control strategies. Atmos. Environ. 33: 369-381.
Bobak, M.; Leon, D. A. (1992) Air pollution and infant mortality in the Czech Republic, 1986-1988.
Lancet (8826): 1010-1014.
Bobak, M.; Leon, D. (1998) Air pollution and infant mortality: the effects are specific for respiratory causes in
postneonatal period. Epidemiology 9: S58.
Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
Bobak, M.; Roberts, A. (1997) Heterogeneity of air pollution effects is related to average temperature [letter].
Br. Med. J. 315: 1161.
Boezen, M.; Schouten, J.; Rijcken, B.; Vonk, J.; Gerritsen, J.; VanDerZee, S.; Hoek, G.; Brunekreef, B.;
Postma, D. (1998) Peak expiratory flow variability, bronchial responsiveness, and susceptibility to ambient
air pollution in adults. Am. J. Respir. Crit. Care Med. 158: 1848-1854.
Boezen, H. M.; VanDerZee, S. C.; Postma, D. S.; Vonk, J. M.; Gerritsen, J.; Hoek, G.; Brunekreef, B.;
Rijcken, B.; Schouten, J. P. (1999) Effects of ambient air pollution on upper and lower respiratory symptoms
and peak expiratory flow in children. Lancet 353: 874-878.
Borja-Aburto, V H.; Loomis, D. P.; Bangdiwala, S. I.; Shy, C. M.; Rascon-Pacheco, R. A. (1997) Ozone,
suspended particulates, and daily mortality in Mexico City. Am. J. Epidemiol. 145: 258-268.
Borja-Aburto, V H.; Castillejos, M.; Gold, D. R.; Bierzwinski, S.; Loomis, D. (1998) Mortality and ambient fine
particles in southwest Mexico City, 1993-1995. Environ. Health Perspect. 106: 849-855.
Braga, A. L. F.; Conceicao, G. M. S.; Pereira, L. A. A.; Kishi, H. S.; Pereira, J. C. R.; Andrade, M. F.; Goncalves,
F. L. T.; Saldiva, P. H. N.; Latorre, M. R. D. O. (1999) Air pollution and pediatric respiratory hospital
admissions in Sao Paulo, Brazil. J. Environ. Med. 1: 95-102.
Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2000) Do respiratory epidemics confound the association between air
pollution and daily deaths? Eur. Respir. J. 16: 723-728.
Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2001) The lag structure between paniculate air pollution and
respiratory and cardiovascular deaths in ten U.S. cities. J. Occup. Environ. Med. 43: 927-933.
Brauer, M.; Ebelt, S. T.; Fisher, T. V; Brumm, J.; Petkau, A. J.; Vedal, S. (2001) Exposure of chronic obstructive
pulmonary disease patients to particles: respiratory and cardiovascular health effects. J. Exposure Anal.
Environ. Epidemiol. 11: 490-500.
Braun-Fahrlander, C.; Vuille, J. C.; Sennhauser, F. H.; Neu, U.; Ktinzle, T.; Grize, L.; Gassner, M.; Minder, C.;
Schindler, C.; Varonier, H. S.; Wuthrich, B.; SCARPOL team. (1997) Respiratory health and long-term
exposure to air pollutants in Swiss schoolchildren. Am. J. Respir. Crit. Care Med. 155: 1042-1049.
Bremner, S. A.; Anderson, H. R.; Atkinson, R. W.; McMichael, A. J.; Strachan, D. P.; Bland, J. M.; Bower, J. S.
(1999) Short term associations between outdoor air pollution and mortality in London 1992-4. Occup.
Environ. Med. 56: 237-244.
Brook, J. R.; Wiebe, A. H.; Woodhouse, S. A.; Audette, C. V; Dann, T. F.; Callaghan, S.; Piechowski, M.;
Dabek-Zlotorzynska, E.; Dloughy, J. F. (1997) Temporal and spatial relationships in fine particle strong
acidity, sulphate, PM10, andPM25 across multiple Canadian locations. Atmos. Environ. 31: 4223-4236.
Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
April 2002 8-290 DRAFT-DO NOT QUOTE OR CITE
-------
Buffler, P. A.; Cooper, S. P.; Stinnett, S.; Contant, C.; Shirts, S.; Hardy, R. J.; Agu, V.; Gehan, B.; Burau, K. (1988)
Air pollution and lung cancer mortality in Harris County, Texas, 1979-1981. Am. J. Epidemiol.
128: 683-699.
Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M;
Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of
respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
142: 15-22.
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997a) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105: 614-620.
Burnett, R. T.; Brook, J. R.; Yung, W. T.; Dales, R. E.; Krewski, D. (1997b) Association between ozone and
hospitalization for respiratory diseases in 16 Canadian cities. Environ. Res. 72: 24-31.
Burnett, R. T.; Dales, R. E.; Brook, J. R.; Raizenne, M. E.; Krewski, D. (1997c) Association between ambient
carbon monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities.
Epidemiology 8: 162-167.
Burnett, R. T.; Cakmak, S.; Brook, J. R. (1998a) The effect of the urban ambient air pollution mix on daily mortality
rates in 11 Canadian cities. Can. J. Public Health 89: 152-156.
Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
Ozkaynak, H. (1998b) The association between ambient carbon monoxide levels and daily mortality in
Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54:130-139.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12(suppl. 4): 15-39.
Burnett, R.; Ma, R.; Jerrett, M.; Goldberg, M. S.; Cakmak, S.; Pope, C. A., Ill; Krewski, D. (2001a) The spatial
association between community air pollution and mortality: a new method of analyzing correlated geographic
cohort data. Environ. Health Perspect. 109(suppl. 3): 375-380.
Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Raizenne, M. E.; Brook, J. R.; Dales, R. E.; Leech, J. A.; Cakmak, S.;
Krewski, D. (200 Ib) Association between ozone and hospitalization for acute respiratory diseases in children
less than 2 years of age. Am. J. Epidemiol. 153: 444-452.
Burton, R. M.; Suh, H. H.; Koutrakis, P. (1996) Spatial variation in paniculate concentrations within metropolitan
Philadelphia. Environ. Sci. Technol. 30: 400-407.
Cakmak, S.; Burnett, R. T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
concentrations for air pollution-related mortality with exposure measurement error. Risk Anal. 19: 487-496.
Calderon-Garciduenas, L.; Mora-Tiscareno, A.; Chung, C. J.; Valencia, G.; Fordham, L. A.; Garcia, R.; Osnaya, N.;
Romero, L.; Acuna, H.; Villarreal-Calderon, A.; Devlin, R. B.; Koren, H. S. (2000) Exposure to air pollution
is associated with lung hyperinflation in healthy children and adolescents in southwest Mexico City: a pilot
study. Inhalation Toxicol. 12: 537-561.
Cassino, C.; Ito, K.; Bader, L; Ciotoli, C.; Thurston, G.; Reibman, J. (1999) Cigarette smoking and ozone-associated
emergency department use for asthma by adults in New York City. Am. J. Respir. Crit. Care Med.
159: 1773-1779.
Castillejos, M.; Borja-Aburto, V. H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
mortality. In: Inhalation Toxicology: proceedings of the third colloquium on paniculate air pollution and
human health; June, 1999; Durham, NC. Inhalation Toxicology 12(suppl. 1): 61-72.
Chang, L.-T.; Koutrakis, P.; Catalano, P. J.; Suh, H. H. (2000) Hourly personal exposures to fine particles and
gaseous pollutants—results from Baltimore, Maryland. J. Air Waste Manage. Assoc. 50: 1223-1235.
Chay, K. Y.; Greenstone, M. (2001) The impact of air pollution on infant mortality: evidence from geographic
variation in pollution shocks induced by a recession. Berkeley, CA: University of California, Berkeley,
Department of Economics. Available: http://elsa.berkeley.edu/users/kenchay/ [29 April, 2002].
Chen, C.; Chock, D. P.; Winkler, S. L. (1999) A simulation study of confounding in generalized linear models for
air pollution epidemiology. Environ. Health Perspect. 107: 217-222.
April 2002 8-291 DRAFT-DO NOT QUOTE OR CITE
-------
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
Chen, L.; Yang, W.; Jennison, B. L.; Goodrich, A.; Omaye, S. T. (2002) Air pollution and birth weight in northern
Nevada, 1991-1999. Inhalation Toxicol. 14: 141-157.
Chew, F. T.; Goh, D. Y. T.; Ooi, B. C.; Saharom, R.; Hui, J. K. S.; Lee, B. W. (1999) Association of ambient
air-pollution levels with acute asthma exacerbation among children in Singapore. Allergy (Copenhagen)
54: 320-329.
Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52: 113-117.
Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
50: 1287-1298.
Clayton, C. A.; Perritt, R. L.; Pellizzari, E. D.; Thomas, K. W.; Whitmore, R. W.; Wallace, L. A.; Ozkaynak, H.;
Spengler, J. D. (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. J. Exposure Anal. Environ. Epidemiol. 3: 227-250.
Clench-Aas, J.; Bartonova, A.; Skjensberg, O. H.; Leegaard, J.; Hagen, L. O.; Gisver, P.; Moseng, J.; Roemer, W.
(1998) Air pollution and respiratory health of children: the PEACE study in Oslo, Norway. Eur. Respir. Rev.
8: 36-43.
Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
Cohen, A. J. (2000) Outdoor air pollution and lung cancer. Environ. Health Perspect. 108(suppl. 4): 743-750.
Cohen, A. J.; Pope, C. A., III. (1995) Lung cancer and air pollution. Environ. Health Perspect.
103(suppl. 8): 219-224.
Conceicao, G. M. S.; Miraglia, S. G. E. K.; Kishi, H. S.; Saldiva, P. H. N.; Singer, J. M. (2001) Air pollution and
child mortality: a time-series study in Sao Paulo, Brazil. Environ. Health Perspect. 109(suppl. 3): 347-350.
Coull, B. A.; Schwartz, J.; Wand, M. P. (2001) Respiratory health and air pollution: additive mixed model analyses.
Biostatistics 2: 337-349.
Coutant, R. W. (1977) Effect of environmental variables on collection of atmospheric sulfate. Environ. Sci.
Technol. 11:873-878.
Creason, J.; Neas, L.; Walsh, D.; Williams, R.; Sheldon, L.; Liao, D.; Shy, C. (2001) Paniculate matter and heart
rate variability among elderly retirees: the Baltimore 1998 PM study. J. Exposure Anal. Environ. Epidemiol.
11: 116-122.
Cropper, M. L.; Simon, N. B.; Alberini, A.; Arora, S.; Sharma, P. K. (1997) The health benefits of air pollution
control in Delhi. Am. J. Agric. Econ. 79: 1625-1629.
Cuijpers, C. E. J.; Swaen, G. M. H.; Wesseling, G.; Wouters, E. F. M. (1994) Acute respiratory effects of summer
smog in primary school children. Toxicol. Lett. 72: 227-235.
Dab, W.; Medina, S.; Quenel, P.; Le Moullec, Y.; Le Tertre, A.; Thelot, B.; Monteil, C.; Lameloise, P.; Pirard, P.;
Momas, L; Ferry, R.; Festy, B. (1996) Short term respiratory health effects of ambient air pollution: results of
the APHEA project in Paris. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on
health: a European approach using epidemiological time series data. J. Epidemiol. Community Health
50 (suppl. 1): S42-S46.
Damia, A. D.; Fabregas, M. L.; Tordera, M. P.; Torrero, L. C. (1999) Effects of air pollution and weather conditions
on asthma exacerbation. Respiration 66: 52-58.
Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
152: 397-406.
Dejmek, J.; Selevan, S. G.; Benes, I.; Solansky, I.; Sram, R. J. (1999) Fetal growth and maternal exposure to
paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154: 633-641.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
April 2002 8-292 DRAFT-DO NOT QUOTE OR CITE
-------
Delfino, R. J.; Murphy-Moulton, A. M.; Becklake, M. R. (1998) Emergency room visits for respiratory illnesses
among the elderly in Montreal: association with low level ozone exposure. Environ. Res. 76: 61-11.
Dewanji, A.; Moolgavkar, S. H. (2000) A Poisson process approach for recurrent event data with environmental
covariates. Environmetrics 11: 665-673.
Diaz, J.; Garcia, R.; Ribera, P.; Alberdi, J. C.; Hernandez, E.; Pajares, M. S.; Otero, A. (1999) Modeling of air
pollution and its relationship with mortality and morbidity in Madrid, Spain. Int. Arch. Occup. Environ.
Health 72: 366-376.
Dockery, D. W.; Spengler, J. D.; Neas, L. M.; Speizer, F. E.; Ferris, B. G., Jr.; Ware, J. H.; Brunekreef, B. (1989)
An epidemiologic study of respiratory health status and indicators of indoor air pollution from combustion
sources. In: Harper, J. P., ed. Combustion processes and the quality of the indoor environment: transactions
of an international specialty conference; September 1988; Niagara Falls, NY. Pittsburgh, PA: Air & Waste
Management Association; pp. 262-271. (A&WMA transactions series: TR-15).
Dockery, D. W.; Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily mortality: associations with particulates
and acid aerosols. Environ. Res. 59: 362-373.
Dockery, D. W.; Pope, C. A., Ill; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
(1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
329: 1753-1759.
Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
symptoms. Environ. Health Perspect. 104: 500-505.
Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
Health Effects Institute; research report no. 83.
Dominici, F.; Samet, J. M.; Zeger, S. (2000a) Combining evidence on air pollution and daily mortality from the
largest 20 US cities: a hierarchical modeling strategy. J. R. Stat. Soc. A 163: 263-302.
Dominici, F.; Zeger, S. L.; Samet, J. (2000b) A measurement error model for time-series studies of air pollution and
mortality. Biostatistics 1: 157-175.
Dominici, F.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2002) Air pollution and mortality: estimating regional and
national dose-response relationships. J. Am. Stat. Assoc. 97: 100-111.
Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. L; Ohio, A. J.; Costa, D. L.
(2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
studies in Utah Valley residents. Environ. Health Perspect. 109(suppl. 3): 395-403.
Fairley, D. (1990) The relationship of daily mortality to suspended particulates in Santa Clara county, 1980-86.
Environ. Health Perspect. 89: 159-168.
Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
Perspect. 107: 637-641.
Forsberg, B.; Segerstedt, B.; Stjernberg, N.; Roemer, W. (1998) Air pollution and respiratory health of children:
the PEACE panel study in Umea, Sweden. Eur. Respir. Rev. 8: 12-19.
Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol.
277: L960-L967.
Friedman, M. S.; Powell, K. E.; Hutwagner, L.; Graham, L. M.; Teague, W. G. (2001) Impact of changes in
transportation and commuting behaviors during the 1996 summer Olympic games in Atlanta on air quality and
childhood asthma. JAMA J. Am. Med. Assoc. 285: 897-905.
Frischer, T.; Studnicka, M.; Gartner, C.; Tauber, E.; Horak, F.; Veiter, A.; Spengler, J.; Ktihr, J.; Urbanek, R.
(1999) Lung function growth and ambient ozone: a three-year population study in school children. Am. J.
Respir. Crit. Care Med. 160: 390-396.
Fusco, D.; Forastiere, F.; Michelozzi, P.; Spadea, T.; Ostro, B.; Area, M.; Perucci, C. A. (2001) Air pollution and
hospital admissions for respiratory conditions in Rome, Italy. Eur. Respir. J. 17: 1143-1150.
Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98-MP26.03.
Garcia-Aymerich, J.; Tobias, A.; Anto, J. M.; Sunyer, J. (2000) Air pollution and mortality in a cohort of patients
with chronic obstructive pulmonary disease: a time series analysis. J. Epidemiol. Community Health
54: 73-74.
April 2002 8-293 DRAFT-DO NOT QUOTE OR CITE
-------
Garty, B. Z.; Kosman, E.; Ganor, E.; Berger, V.; Garty, L.; Wietzen, T.; Waisman, Y.; Mimouni, M.; Waisel, Y.
(1998) Emergency room visits of asthmatic children, relation to air pollution, weather, and airborne allergens.
Ann. Allergy Asthma Immunol. 81: 563-570.
Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
Gauvin, S.; Zmirou, D.; Pin, L; Quentin, J.; Balducci, F.; Boudet, C.; Poizeau, D.; Brambilla, C. (1999) Short-term
effect of exposure to suspended paniculate matter (PM10) on the respiratory function of urban asthmatic and
control adults. J. Environ. Med. 1: 71-79.
Gherghinova, M.; Kostyanve, S.; Ivanova, M. (1989) Theoretic values of body plethysmography indices in healthy
children between 7 and 14 years of age. Pediatria 28: 52-58.
Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
Am. J. Respir. Crit. Care Med. 164: 704-708.
Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
155:2105-2108.
Gold, D. R.; Damokosh, A. L; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
Castillejos, M. (1999) Paniculate and ozone pollutant effects on the respiratory function of children in
southwest Mexico City. Epidemiology 10: 8-16.
Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm
[15 February, 2001].
Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Brook, J.; Bonvalot, Y.; Tamblyn, R.; Singh, R.; Valois, M.-F.
(200 la) The association between daily mortality and ambient air particle pollution in Montreal, Quebec. 1.
Nonaccidental mortality. Environ. Res. 86: 12-25.
Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Brook, J.; Bonvalot, Y.; Tamblyn, R.; Singh, R.; Valois, M.-F.;
Vincent, R. (200 Ib) The association between daily mortality and ambient air particle pollution in Montreal,
Quebec. 2. Cause-specific mortality. Environ. Res. 86: 26-36.
Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Tamblyn, R.; Ernst, P.; Flegel, K.; Brook, J.; Bonvalot, Y.;
Singh, R.; Valois, M.-F.; Vincent, R. (200Ic) Identification of persons with cardiorespiratory conditions who
are at risk of dying from the acute effects of ambient air particles. Environ. Health Perspect.
109(suppl. 4): 487-494.
Goldberg, M. S.; Burnett, R. T.; Brook, J.; Bailar, J. C., Ill; Valois, M.-F.; Vincent, R. (2001d) Associations
between daily cause-specific mortality and concentrations of ground-level ozone in Montreal, Quebec. Am. J.
Epidemiol. 154: 817-826.
Gordian, M. E.; Ozkaynak, H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Paniculate air pollution and respiratory
disease in Anchorage, Alaska. Environ. Health Perspect. 104: 290-297.
Gouveia, N.; Fletcher, T. (2000) Respiratory diseases in children and outdoor air pollution in Sao Paulo, Brazil:
a time series analysis. Occup. Environ. Med. 57: 477-483.
Grievink, L.; Van der Zee, S. C.; Hoek, G.; Boezen, H. M.; Van't Veer, P.; Brunekreef, B. (1999) Modulation of the
acute respiratory effects of winter air pollution by serum and dietary antioxidants: a panel study. Eur. Respir.
J. 13: 1439-1446.
Gtintzel, O.; Bollag, U.; Helfenstein, U. (1996) Asthma and exacerbation of chronic bronchitis: sentinel and
environmental data in a time series analysis. Zentralbl. Hyg. Umweltmed. 198: 383-393.
Guo, Y. L.; Lin, Y.-C; Sung, F.-C; Huang, S.-L.; Ko, Y.-C; Lai, J.-S.; Su, H.-J.; Shaw, C.-K.; Lin, R.-S.; Dockery,
D. W. (1999) Climate, traffic-related air pollutants, and asthma prevalence in middle-school children in
Taiwan. Environ. Health Perspect. 107: 1001-1006.
Gwynn, R. C.; Thurston, G. D. (2001) The burden of air pollution: impacts among racial minorities. Environ. Health
Perspect. 109(suppl. 4): 501-506.
Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
April 2002 8-294 DRAFT-DO NOT QUOTE OR CITE
-------
Hagen, J. A.; Nafstad, P.; Skrondal, A.; Bjerkly, S.; Magnus, P. (2000) Associations between outdoor air pollutants
and hospitalization for respiratory diseases. Epidemiology 11: 136-140.
Hajat, S.; Haines, A.; Goubet, S. A.; Atkinson, R. W.; Anderson, H. R. (1999) Association of air pollution with
daily GP consultations for asthma and other lower respiratory conditions in London. Thorax 54: 597-605.
Hajat, S.; Haines, A.; Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Emberlin, J. (2001) Association between
air pollution and daily consultations with general practitioners for allergic rhinitis in London, United
Kingdom. Am. J. Epidemiol. 153: 704-714.
Haluszka, J.; Pisiewicz, K.; Miczynski, J.; Roemer, W.; Tomalak, W. (1998) Air pollution and respiratory health in
children: the PEACE panel study in Krakow, Poland. Eur. Respir. Rev. 8: 94-100.
Harre, E. S. M; Price, P. D.; Ayrey, R. B.; Toop, L. J.; Martin, I. R.; Town, G. I. (1997) Respiratory effects of air
pollution in chronic obstructive pulmonary disease: a three month prospective study. Thorax 52: 1040-1044.
Harris, R. J. (1975) A primer of multivariate statistics. New York, NY: Academic Press; pp. 155-224.
Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
Heinrich, J.; Hoelscher, B.; Wjst, M.; Ritz, B.; Cyrys, J.; Wichmann, H.-E. (1999) Respiratory diseases and
allergies in two polluted areas in East Germany. Environ. Health Perspect. 107: 53-62.
Heinrich, J.; Hoelscher, B.; Wichmann, H. E. (2000) Decline of ambient air pollution and respiratory symptoms in
children. Am. J. Respir. Crit. Care Med. 161: 1930-1936.
Henderson, B. E.; Gordon, R. J.; Menck, H.; SooHoo, J.; Martin, S. P.; Pike, M. C. (1975) Lung cancer and air
pollution in southcentral Los Angeles County. Am. J. Epidemiol. 101: 477-488.
Hickey, R. J.; Boyce, D. E.; Harner, E. B.; Clelland, R. C. (1970) Ecological statistical studies on environmental
pollution and chronic disease in metropolitan areas of the United States. Philadelphia, PA: Regional Science
Research Institute. (RSRI discussion paper series no. 35).
Hill, A. B. (1965) The environment and disease: association or causation? Proc. R. Soc. Med. 58: 295-300.
Hiltermann, T. J. N.; de Bruijne, C. R.; Stolk, J.; Zwinderman, A. H.; Spieksma, F. Th. M.; Roemer, W.;
Steerenberg, P. A.; Fischer, P. H.; van Bree, L.; Hiemstra, P. S. (1997) Effects of photochemical air pollution
and allergen exposure on upper respiratory tract inflammation in asthmatics. Am. J. Respir. Crit. Care Med.
156: 1765-1772.
Hiltermann, T. J. N.; Stolk, J.; VanderZee, S. C.; Brunekreef, B.; De Bruijne, C. R.; Fischer, P. H.; Ameling,
C. B.; Sterk, P. J.; Hiemstra, P. S.; Van Bree, L. (1998) Asthma severity and susceptibility to air pollution.
Eur. Respir. J. 11: 686-693.
Hoek, G.; Schwartz, J. D.; Groot, B.; Eilers, P. (1997) Effects of ambient paniculate matter and ozone on daily
mortality in Rotterdam, the Netherlands. Arch. Environ. Health 52: 455-463.
Hoek, G.; Dockery, D. W.; Pope, A.; Neas, L.; Roemer, W.; Brunekreef, B. (1998) Association between PM10 and
decrements in peak expiratory flow rates in children: reanalysis of data from five panel studies. Eur. Respir.
J. 11: 1307-1311.
Hoek, G.; Brunekreef, B.; Verhoeff, A.; van, Wijnen, J.; Fischer, P. (2000) Daily mortality and air pollution in the
Netherlands. J. Air Waste Manage. Assoc. 50: 1380-1389.
Hoek, G.; Brunekreef, B.; Fischer, P.; Van Wijnen, J. (2001) The association between air pollution and heart
failure, arrhythmia, embolism, thrombosis, and other cardiovascular causes of death in a time series study.
Epidemiology 12: 355-357.
Hodgkin, J. E.; Abbey, D. E.; Euler, G. L.; Magie, A. R. (1984) COPD prevalence in nonsmokers in high and low
photochemical air pollution areas. Chest 86: 830-838.
Hong, Y.-C.; Leem, J.-H.; Ha, E.-H.; Christiani, D. C. (1999) PM10 exposure, gaseous pollutants, and daily
mortality in Inchon, South Korea. Environ. Health Perspect. 107: 873-878.
Howel, D.; Darnell, R.; Pless-Mulloli, T. (2001) Children's respiratory health and daily paniculate levels in
10 nonurban communities. Environ. Res. 87: 1-9.
Hruba, F.; Fabianova, E.; Koppova, K.; Vandenberg, J. J. (2001) Childhood respiratory symptoms, hospital
admissions and long-term exposure to airborne paniculate matter. J. Exposure Anal. Environ. Epidemiol.
11:33-40.
Huang, Y.-L.; Batterman, S. (2000) Residence location as a measure of environmental exposure: a review of air
pollution epidemiology studies. J. Exposure Anal. Environ. Epidemiol. 10: 66-85.
Ibald-Mulli, A.; Stieber, J.; Wichmann, H.-E.; Koenig, W.; Peters, A. (2001) Effects of air pollution on blood
pressure: a population-based approach. Am. J. Public Health 91: 571-577.
April 2002 8-295 DRAFT-DO NOT QUOTE OR CITE
-------
Ilabaca, M; Olaeta, I.; Campos, E.; Villaire, J.; Tellez-Rojo, M. M; Romieu, I. (1999) Association between levels
of fine paniculate and emergency visits for pneumonia and other respiratory illnesses among children in
Santiago, Chile. J. Air Waste Manage. Assoc. 49: 154-163.
Ito, K. (1990) An examination of the role of aerosol acidity in historical London, England daily mortality
[dissertation]. Syracuse, NY: New York University. Available from: University Microfilms International,
Ann Arbor, MI; AAD91-13012.
Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
1983-1992. Environ. HealthPerspect. 105: 980-985.
Jalaludin, B. B.; Chey, T.; O'Toole, B. I.; Smith, W. T.; Capon, A. G.; Leeder, S. R. (2000) Acute effects of low
levels of ambient ozone on peak expiratory flow rate in a cohort of Australian children. Int. J. Epidemiol.
29: 549-557.
Jamason, P. F.; Kalkstein, L. S.; Gergen, P. J. (1997) A synoptic evaluation of asthma hospital admissions in
New York City. Am. J. Respir. Crit. Care Med. 156: 1781-1788.
Janssen, N. A. H.; Van Vliet, P. H. N.; Aarts, F.; Harssema, H.; Brunekreef, B. (2001) Assessment of exposure to
traffic related air pollution of children attending schools near motorways. Atmos. Environ. 35: 3875-3884.
Janssen, N. A. H.; Schwartz, J.; Zanobetti, A.; Suh, H. H. (2002) Air conditioning and source-specific particles as
modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ. Health Perspect.
110:43-49.
Jedrychowski, W.; Becher, H.; Wahrendorf, J.; Basa-Cierpialek, Z. (1990) A case-control study of lung cancer with
special reference to the effect of air pollution in Poland. J. Epidemiol. Commun. Health 44: 114-120.
Jedrychowski, W.; Flak, E. (1998) Separate and combined effects of the outdoor and indoor air quality on chronic
respiratory symptoms adjusted for allergy among preadolescent children. Int. J. Occup. Med. Environ. Health
11: 19-35.
Jedrychowski, W.; Flak, E.; Mroz, E. (1999) The adverse effect of low levels of ambient air pollutants on lung
function growth in preadolescent children. Environ. Health Perspect. 107: 669-674.
Kalandidi, A.; Gratziou, C.; Katsouyanni, K.; Manalis, N.; Tzala, L.; Pantazopoulou, A.; Efthimiou, M.;
Roussos, C.; Roemer, W. (1998) Air pollution and respiratory health of children: the PEACE panel study in
Athens, Greece. Eur. Respir. Rev. 8: 117-124.
Katsouyanni, K.; Touloumi, G. (1998) Causes of regional differences in air pollution effects are being studied
further [letter]. Br. Med. J. 316: 1982.
Katsouyanni, K.; Karakatsani, A.; Messari, L; Touloumi, G.; Hatzakis, A.; Kalandidi, A.; Trichopoulos, D. (1990)
Air pollution and cause specific mortality in Athens. J. Epidemiol. Commun. Health 44: 321-324.
Katsouyanni, K.; Schwartz, J.; Spix, C.; Touloumi, G.; Zmirou, D.; Zanobetti, A.; Wojtyniak, B.; Vonk, J. M.;
Tobias, A.; Ponka, A.; Medina, S.; Bacharova, L.; Andersen, H. R. (1996) Short term effects of air pollution
on health: a European approach using epidemiology time series data: the APHEA protocol. In: St Leger, S.,
ed. The APHEA project. Short term effects of air pollution on health: a European approach using
epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S12-S18.
Katsouyanni, K.; Touloumi, G.; Spix, C.; Schwartz, J.; Balducci, F.; Medina, S.; Rossi, G.; Wojtyniak, B.;
Sunyer, J.; Bacharova, L.; Schouten, J. P.; Ponka, A.; Anderson, H. R. (1997) Short term effects of ambient
sulphur dioxide and paniculate matter on mortality in 12 European cities: results from time series data from
the APHEA project. Br. Med. J. 314: 1658-1663.
Katsouyanni, K.; Touloumi, G.; Samoli, E.; Gryparis, A.; Le Tertre, A.; Monopolis, Y.; Rossi, G.; Zmirou, D.;
Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
Schindler, C.; Schwartz, J. (2001) Confounding and effect modification in the short-term effects of ambient
particles on total mortality: results from 29 European cities within the APHEA2 project. Epidemiology
12:521-531.
Keatinge, W. R.; Donaldson, G. C. (2001) Mortality related to cold and air pollution in London after allowance for
effects of associated weather patterns. Environ. Res. A 86: 209-216.
Keles, N.; Ilicali, O. C.; Deger, K. (1999) Impact of air pollution on prevalence of rhinitis in Istanbul. Arch.
Environ. Health 54: 48-51.
Kelsall, J. E.; Samet, J. M.; Zeger, S. L.; Xu, J. (1997) Air pollution and mortality in Philadelphia, 1974-1988.
Am. J. Epidemiol. 146: 750-762.
Kingham, S.; Briggs, D.; Elliott, P.; Fischer, P.; Lebret, E. (2000) Spatial variations in the concentrations of
traffic-related pollutants in indoor and outdoor air in Huddersfield, England. Atmos. Environ. 34: 905-916.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
April 2002 8-296 DRAFT-DO NOT QUOTE OR CITE
-------
Klemm, R. I; Mason, R. M, Jr.; Heilig, C. M; Neas, L. M; Dockery, D. W. (2000) Is daily mortality associated
specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
50: 1215-1222.
Koenig, J. Q.; Larson, T. V.; Hanley, Q. S.; Rebolledo, V.; Dumler, K.; Checkoway, H.; Wang, S.-Z.; Lin, D.;
Pierson, W. E. (1993) Pulmonary function changes in children associated with fine paniculate matter.
Environ. Res. 63:26-38.
Kontos, A. S.; Fassois, S. D.; Deli, M. F. (1999) Short-term effects of air pollution on childhood respiratory illness
in Piraeus, Greece, 1987-1992: nonparametric stochastic dynamic analysis. Environ. Res. 81: 275-296.
Koren, H. S.; Utell, M. J. (1997) Asthma and the environment. Environ. Health Perspect. 105: 534-537.
Korrick, S. A.; Neas, L. M.; Dockery, D. W.; Gold, D. R.; Allen, G. A.; Hill, L. B.; Kimball, K. D.; Rosner, B. A.;
Speizer, F. E. (1998) Effects of ozone and other pollutants on the pulmonary function of adult hikers.
Environ. Health Perspect. 106: 93-99.
Kostianev, S.; Gerginova, M.; Ivanova, M. (1994) Reference values of lung function parameters in Bulgarian girls
aged 7 to 14 years. Pediatria 33: 30-33.
Kotesovec, F.; Vitnerova, N.; Leixner, M.; Benes, I.; Skorkovsky, J.; Roemer, W. (1998) Air pollution and
respiratory health of children: the PEACE panel study in Teplice, Czech Republic. Eur. Respir. Rev.
8: 70-77.
Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Peters, A.; Heinrich, J. (2000) Daily mortality and air pollution in
northern Bohemia; different effects for men and women. Cent. Eur. J. Public Health 8: 120-127.
Kousa, A.; Monn, C.; Rotko, T.; Aim, S.; Oblesby, L.; Jantunen, M. J. (2001) Personal exposures to NO2 in the
EXPOLIS-study: relation to residential indoor, outdoor and workplace concentrations in Basel, Helsinki and
Prague. Atmos. Environ. 35: 3405-3412.
Kramer, U.; Behrendt, H.; Dolgner, R.; Ranft, U.; Ring, J.; Wilier, H.; Schlipkoter, H.-W. (1999) Airway diseases
and allergies in East and West German children during the first 5 years after reunification: time trends and
the impact of sulphur dioxide and total suspended particles. Int. J. Epidemiol. 28: 865-873.
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
W. H. (2000) Reanallysis of the Harvard Six Cities study and the American Cancer Society study of
paniculate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
Project. Cambridge, MA: Health Effects Institute.
Ktinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared
to ecologic studies. Environ. Health Perspect. 105: 1078-1083.
Ktinzli, N.; Ackermann-Liebrich, U.; Brandli, O.; Tschopp, J. M.; Schindler, C.; Leuenberger, P.; SAPALDIA
Team. (2000) Clinically "small" effects of air pollution on FVC have a large public health impact. Eur.
Respir. J. 15: 131-136.
Kwon, H.-J.; Cho, S.-H.; Nyberg, F.; Pershagen, G. (2001) Effects of ambient air pollution on daily mortality in a
cohort of patients with congestive heart failure. Epidemiology 12: 413-419.
Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
Lave, L. B.; Seskin, E. P. (1977) Air pollution and human health. Baltimore, MD: The Johns Hopkins University
Press.
Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2002) Short term effects of paniculate
air pollution on cardiovascular diseases in eight European cities. J. Epidemiol. Community Health: in press.
Lebowitz, M. D. (1971) A stimulus response method for the analysis of environmental and health events related in
time. Environ. Lett. 2: 23-24.
Lebowitz, M. D.; Collins, L.; Holberg, C. J. (1987) Time series analyses of respiratory responses to indoor and
outdoor environmental phenomena. Environ. Res. 43: 332-341.
Lee, J.-T.; Schwartz, J. (1999) Reanalysis of the effects of air pollution on daily mortality in Seoul, Korea:
a case-crossover design. Environ. Health Perspect. 107: 633-636.
Lee, J.-T.; Shy, C. M. (1999) Respiratory function as measured by peak expiratory flow rate and PM10:
six communities study. J. Exposure Anal. Environ. Epidemiol. 9: 293-299.
Lee, R. E., Jr.; Caldwell, J. S.; Morgan, G. B. (1972) The evaluation of methods for measuring suspended
particulates in air. Atmos. Environ. 6: 593-622.
Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
Perspect. 107: 149-154.
April 2002 8-297 DRAFT-DO NOT QUOTE OR CITE
-------
Lee, J.-T.; Kim, H.; Hong, Y.-C.; Kwon, H.-J.; Schwartz, J.; Christiani, D. C. (2000) Air pollution and daily
mortality in seven major cities of Korea, 1991-1997. Environ. Res. 84: 247-254.
Leonardi, G. S.; Houthuijs, D.; Steerenberg, P. A.; Fletcher, T.; Armstrong, B.; Antova, T. (2000) Immune
biomarkers in relation to exposure to paniculate matter: a cross-sectional survey in 17 cities of central
Europe. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 1-14.
Levy, D. (1998) Fine paniculate air pollution and out-of-hospital mortality in King County, Washington. In: Vostal,
J. J., ed. Health effects of paniculate matter in ambient air. Proceedings of an international conference; 1997;
Prague, Czech Republic. Pittsburgh, PA: Air & Waste Management Association; pp. 262-271. (A&WMA
publication VIP-80).
Levy, J. I.; Hammitt, J. K.; Spengler, J. D. (2000) Estimating the mortality impacts of paniculate matter: what can
be learned frombetween-study variability? Environ. Health Perspect. 108: 109-117.
Lewis, P. R.; Hensley, M. J.; Wlodarczyk, J.; Toneguzzi, R. C.; Westley-Wise, V. J.; Dunn, T.; Calvert, D. (1998)
Outdoor air pollution and children's respiratory symptoms in the steel cities of New South Wales. Med. J.
Aust. 169: 459-463.
Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
Lin, C. A.; Martins, M. A.; Farhat, S. C. L.; Pope, C. A., Ill; Conceicao, G. M. S.; Anastacio, V. M.; Hatanaka, M.;
Andrade, W. C.; Hamaue, W. R.; Bohm, G. M.; Saldiva, P. H. N. (1999) Air pollution and respiratory illness
of children in Sao Paulo, Brazil. Paediatr. Perinat. Epidemiol. 13: 475-488.
Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D.; Gong, H., Jr. (1996)
Short-term air pollution exposures and responses in Los Angeles area schoolchildren. J. Exposure Anal.
Environ. Epidemiol. 6: 449-472.
Linn, W. S.; Gong, H., Jr.; Clark, K. W.; Anderson, K. R. (1999) Day-to-day paniculate exposures and health
changes in Los Angeles area residents with severe lung disease. J. Air Waste Manage. Assoc.
49:PM108-PM115.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lipfert, F. W. (2000) The use and misuse of surrogate variables in environmental epidemiology. J. Environ. Med.
1:267-278.
Lipfert, F. W.; Wyzga, R. E. (1999) Statistical considerations in determining the health significance of constituents
of airborne paniculate matter. J. Air Waste Manage. Assoc. 49: PM-182-191.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
Lipfert, F. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) The Washington
University-EPRI veterans' cohort mortality study: preliminary results. In: Grant, L. D., ed. PM2000:
paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 41-73.
Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000c) Infant mortality and air pollution: a comprehensive analysis of U.S.
data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
Lippmann, M.; Thurston, G. D. (1996) Sulfate concentrations as an indicator of ambient paniculate matter air
pollution for health risk evaluations. J. Exposure Anal. Environ. Epidemiol. 6: 123-146.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report no.
95.
Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
California. Environ. Health Perspect. 105: 216-222.
Long, W.; Tate, R. B.; Neuman, M.; Manfreda, J.; Becker, A. B.; Anthonisen, N. R. (1998) Respiratory symptoms
in a susceptible population due to burning of agricultural residue. Chest 113: 351-357.
Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
mortality in Mexico City. Epidemiology 10: 118-123.
Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
J. R. Stat. Soc. B 61(part 2): 459-477.
Lumley, T.; Sheppard, L. (2000) Assessing seasonal confounding and model selection bias in air pollution
epidemiology using positive and negative control analyses. Environmetrics 11: 705-717.
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
April 2002 8-298 DRAFT-DO NOT QUOTE OR CITE
-------
Marcus, A. H.; Kegler, S. R. (2001) Confounding in air pollution epidemiology: when does two-stage regression
identify the problem? Environ. Health Perspect. 109: 1193-1196.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
California children with asthma. Environ. Health Perspect. 107: 757-760.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Islam, T.; Gauderman, W. J.; Avol, E.; Margolis, H. G.;
Peters, J. M. (2002) Asthma in exercising children exposed to ozone: a cohort study. Lancet 359: 386-391.
McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of
mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers.
J. Exposure Anal. Environ. Epidemiol. 10: 427-436.
McGregor, G. R.; Walters, S.; Wordley, J. (1999) Daily hospital respiratory admissions and winter air mass types,
Birmingham, UK. Int. J. Biometeorol. 43: 21-30.
Medina, S.; Le Tertre, A.; Quenel, P.; Le Moullec, Y.; Lameloise, P.; Guzzo, J. C.; Festy, B.; Ferry, R.; Dab, W.
(1997) Air pollution and doctors' house calls: results from the ERPURS system for monitoring the effects of
air pollution on public health in greater Paris, France, 1991-1995. Environ. Res. 75: 73-84.
Mendelsohn, R.; Orcutt, G. (1979) An empirical analysis of air pollution dose-response curves. J. Environ. Econ.
Manage. 6: 85-106.
Michelozzi, P.; Forastiere, F.; Fusco, D.; Perucci, C. A.; Ostro, B.; Ancona, C.; Pallotti, G. (1998) Air pollution and
daily mortality in Rome, Italy. Occup. Environ. Med. 55: 605-610.
Mills, P. K.; Abbey, D.; Beeson, W. L.; Petersen, F. (1991) Ambient air pollution and cancer in California
Seventh-day Adventists. Arch. Environ. Health 46: 271-280.
Monn, C. (2001) Exposure assessment of air pollutants: a review on spatial heterogeneity and
indoor/outdoor/personal exposure to suspended paniculate matter, nitrogen dioxide and ozone. Atmos.
Environ. 35: 1-32.
Monn, Ch.; Carabias, V.; Junker, M.; Waeber, R.; Karrer, M.; Wanner, H. U. (1997) Small-scale spatial variability
of paniculate matter <10 um(PM10) and nitrogen dioxide. Atmos. Environ. 31: 2243-2247.
Moolgavkar, S. H. (2000a) Air Pollution and Mortality in Three U.S. Counties. Environ. Health Perspect.
108: 777-784.
Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health.
Inhalation Toxicol. 12(suppl. 4): 75-90.
Moolgavkar, S. H.; Luebeck, E. G. (1996) A critical review of the evidence on paniculate air pollution and
mortality. Epidemiology 7: 420-428.
Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and
admissions for chronic respiratory disease in King County, Washington. In: Inhalation toxicology:
proceedings of the third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC.
Inhalation Toxicology 12(suppl. 1): 157-171.
Morgan, G.; Corbett, S.; Wlodarczyk, J.; Lewis, P. (1998) Air pollution and daily mortality in Sydney, Australia,
1989 through 1993. Am. J. Public Health 88: 759-764.
Morris, R. D. (2001) Airborne particulates and hospital admissions for cardiovascular disease: a quantitative review
of the evidence. Environ. Health Perspect. 109(suppl. 4): 495-500.
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
Murray, C. J.; Nelson, C. R. (2000) State-space modeling of the relationship between air quality and mortality.
J. Air Waste Manage. Assoc. 50: 1075-1080.
Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
Healthy women's PEF variations with ambient summer concentrations of PM10, PM2 5, SO42", H+, and O3.
Am. J. Respir. Crit. Care Med. 160: 117-125.
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
and exposure to air pollution. Public Health Rep. 114: 135-148.
April 2002 8-299 DRAFT-DO NOT QUOTE OR CITE
-------
Neas, L. M; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
139: 1088-1099.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
141: 111-122.
Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and
other determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
Philadelphia. Environ. Health Perspect. 107: 629-631.
Neukirch, F.; Segala, C.; Le Moullec, Y.; Korobaeff, M.; Aubier, M. (1998) Short-term effects of low-level winter
pollution on respiratory health of asthmatic adults. Arch. Environ. Health 53: 320-328.
Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
Perspect. 107: 489-493.
Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
stagnant air. Thorax 55: 466-470.
Nyberg, F.; Gustavsson, P.; Jarup, L.; Bellander, T.; Berglind, N.; Jakobsson, R.; Pershagen, G. (2000) Urban air
pollution and lung cancer in Stockholm. Epidemiology 11: 487-495.
Oglesby, L.; Rotko, T.; Kriitli, P.; Boudet, C.; Kruize, H.; Jantunen, M. J.; Ktinzli, N. (2000) Personal exposure
assessment studies may suffer from exposure-relevant selection bias. J. Exposure Anal. Environ. Epidemiol.
10:251-266.
Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
70: 98-104.
Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
Am. J. Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
Ostro, B.; Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1998) The impact of fine paniculate matter in Bangkok,
Thailand. In: Chow, J.; Koutrakis, P., eds. PM2.5: a fine particle standard. Volume II: proceedings of an
international specialty conference; January; Long Beach, CA. Pittsburgh, PA: Air & Waste Management
Association; pp. 939-949. (A&WMA publication VIP-81).
Ostro, B. D.; Hurley, S.; Lipsett, M. J. (1999a) Air pollution and daily mortality in the Coachella Valley, California:
a study of PM10 dominated by coarse particles. Environ. Res. 81: 231-238.
Ostro, B. D.; Eskeland, G. S.; Sanchez, J. M.; Feyzioglu, T. (1999b) Air pollution and health effects: a study of
medical visits among children in Santiago, Chile. Environ. Health Perspect. 107: 69-73.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol.: 10: 412-419.
Ostro, B.; Lipsett, M.; Mann, J.; Braxtron-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
African-American children in Los Angeles. Epidemiology 12: 200-208.
Osunsanya, T.; Prescott, G.; Seaton, A. (2001) Acute respiratory effects of particles: mass or number? Occup.
Environ. Med. 58: 154-159.
Ozkaynak, H.; Thurston, G. D. (1987) Associations between 1980 U.S. mortality rates and alternative measures of
airborne particle concentration. Risk Anal. 7: 449-461.
Ozkaynak, H.; Schatz, A. D.; Thurston, G. D.; Isaacs, R. G.; Husar, R. B. (1985) Relationships between aerosol
extinction coefficients derived from airport visual range observations and alternative measures of airborne
particle mass. J. AirPollut. Control Assoc. 35: 1176-1185.
Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
Environmental Health; March 25.
Pantazopoulou, A.; Katsouyanni, K.; Kourea-Kremastinou, J.; Trichopoulos, D. (1995) Short-term effects of air
pollution on hospital emergency outpatient visits and admissions in the greater Athens, Greece area. Environ.
Res. 69:31-36.
April 2002 8-300 DRAFT-DO NOT QUOTE OR CITE
-------
Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
Pekkanen, J.; Brunner, E. J.; Anderson, H. R.; Tiittanen, P.; Atkinson, R. W. (2000) Daily concentrations of air
pollution and plasma fibrinogen in London. Occup. Environ. Med. 57: 818-822.
Penttinen, P.; Timonen, K. L.; Tiittanen, P.; Mirme, A.; Ruuskanen, J.; Pekkanen, J. (2001) Number concentration
and size of particles in urban air: effects on spirometric lung function in adult asthmatic subjects. Environ.
HealthPerspect. 109: 319-323.
Pereira, L. A. A.; Loomis, D.; Conceicao, G. M. S.; Braga, A. L. F.; Areas, R. M; Kishi, H. S.; Singer, J. M;
Bohm, G. M.; Saldiva, P. H. N. (1998) Association between air pollution and intrauterine mortality in
Sao Paulo, Brazil. Environ. Health Perspect. 106: 325-329.
Perry, H. M., Jr.; Schnaper, H. W.; Meyer, G.; Swatzell, R. (1982) Clinical program for screening and treatment of
hypertension in veterans. J. Natl. Med. Assoc. 74: 433-444.
Peters, A.; Goldstein, I. F.; Beyer, U.; Franke, K.; Heinrich, J.; Dockery, D. W.; Spengler, J. D.; Wichmann, H.-E.
(1996) Acute health effects of exposure to high levels of air pollution in eastern Europe. Am. J. Epidemiol.
144: 570-581.
Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
episode: a link to mortality? Lancet 349: 1582-1587.
Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997b) Short-term effects of paniculate air pollution on
respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997c) Respiratory effects are associated with the
number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
Peters, A.; Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Heinrich, J. (1999a) Akute Auswirkung der
Schwebstaubkonzentrationen in der Aulknluft auf die Mortalitat - Vergleichsstudie Nordost-Bayern /
Nordbohmen. AbschluBbericht [Acute effects of suspended particle concentrations in the atmosphere on
mortality - a study comparing northeast Bavaria and north Bohemia. Final report]. Bavaria, Federal Republic
of Germany: InstitutfurEpidemiologie; report no. GSF-EP S 1/99.
Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
Med. 159: 760-767.
Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999c) A study of twelve southern California communities with
differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
159: 768-775.
Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
Monahan, K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology
11: 11-17.
Peters, A.; Frohlich, M.; Doring, A.; Immervoll, T.; Wichmann, H.-E.; Hutchinson, W. L.; Pepys, M. B.;
Koenig, W. (2000b) Paniculate air pollution is associated with an acute phase response in men: results from
the MONICA-Augsburg Study. Eur. Heart J.: in press.
Peters, A.; Dockery, D. W.; Muller, J. E.; Mittleman, M. A. (2001) Increased paniculate air pollution and the
triggering of myocardial infarction. Circulation 103: 2810-2815.
Petroeschevsky, A.; Simpson, R. W.; Thalib, L.; Rutherford, S. (2001) Associations between outdoor air pollution
and hospital admissions in Brisbane, Australia. Arch. Environ. Health 56: 37-52.
Pike, M. C.; Jing, J. S.; Rosario, I. P.; Henderson, B. E.; Menck, H. R. (1979) Occupation: "explanation" of an
apparent air pollution related localized excess of lung cancer in Los Angeles County. In: Breslow, N. E.;
Whitemore, A. S., eds. Energy and Health: proceedings of a conference sponsored by Siam Institute for
Mathematics and Society; 1978; Alta, Utah. [SIAM-SIMS conference series 6].
Pless-Mulloli, T.; Howel, D.; King, A.; Stone, I.; Merefield, J.; Bessell, J.; Darnell, R. (2000) Living near opencast
coal mining sites and children's respiratory health. Occup. Environ. Med. 57: 145-151.
Ponce de Leon, A.; Anderson, H. R.; Bland, J. M.; Strachan, D. P.; Bower, J. (1996) Effects of air pollution on
daily hospital admissions for respiratory disease in London between 1987-88 and 1991-92. In: St Leger, S.,
ed. The APHEA project. Short term effects of air pollution on health: a European approach using
epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S63-S70.
Ponka, A.; Savela, M.; Virtanen, M. (1998) Mortality and air pollution in Helsinki. Arch. Environ. Health
53:281-286.
April 2002 8-301 DRAFT-DO NOT QUOTE OR CITE
-------
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and
Cache Valleys. Arch. Environ. Health 46: 90-97.
Pope, C. A., Ill; Kalkstein, L S. (1996) Synoptic weather modeling and estimates of the exposure-response
relationship between daily mortality and paniculate air pollution. Environ. Health Perspect. 104: 414-420.
Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM10 pollution:
a daily time series analysis. Am. Rev. Respir. Dis. 144: 668-674.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah valley.
Arch. Environ. Health 47: 211-217.
Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
Respir. Crit. Care Med. 151: 669-674.
Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
Front. Environ. Health Perspect. 107: 567-573.
Pope, C. A., Ill; Dockery, D. W.; Kanner, R. E.; Vollegas, G. M.; Schwartz, J. (1999b) Oxygen saturation, pulse
rate, and paniculate air pollution: a daily time-series panel study. Am. J. Respir. Crit. Care Med.
159: 365-372.
Pope, C. A., Ill; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
G. M.; Gold, D. R.; Dockery, D. W. (1999c) Heart rate variability associated with paniculate air pollution.
Am. Heart J. 138: 890-899.
Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
Assoc. 287: 1132-1141.
Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
Prescott, E.; Lange, P.; Vestbo, J.; Copenhagen City Heart Study Group. (1999) Socioeconomic status, lung
function and admission to hospital for COPD: results from the Copenhagen City Heart Study. Eur. Respir. J.
13: 1109-1114.
Qian, Z. Chapman, R. S.; Tian, Q.; Chen, Y.; Lioy, P. J.; Zhang, J. (2000) Effects of air pollution on children's
respiratory health in three Chinese cities. Arch. Environ. Health 55: 126-133.
Rahlenbeck, S. I.; Kahl, H. (1996) Air pollution and mortality in East Berlin during the winters of 1981-1989. Int. J.
Epidemiol. 25: 1220-1226.
Raizenne, M.; Neas, L. M.; Damokosh, A. L; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
Health Perspect. 104: 506-514.
Ransom, M. R.; Pope, C. A., III. (1992) Elementary school absences and PM10 pollution in Utah Valley. Environ.
Res. 58: 204-219.
Rijnders, E.; Janssen, N. A.; Van Vliet, P. H.; Brunekreef, B. (2001) Personal and outdoor nitrogen dioxide
concentrations in relation to degree of urbanization and traffic density. Environ. Health Perspect.
109(suppl. 3): 411-417.
Ritz, B.; Yu, F. (1999) The effect of ambient carbon monoxide on low birth weight among children born in southern
California between 1989 and 1993. Environ. Health Perspect. 107: 17-25.
Ritz, B.; Yu, F.; Fruin, S.; Chapa, G.; Shaw, G. M.; Harris, J. A. (2002) Ambient air pollution and risk of birth
defects in southern California. Am. J. Epidemiol. 155: 17-25.
Rodes, C. E.; Holland, D. M. (1981) Variations of NO, NO2 and O3 concentrations downwind of aLos Angeles
freeway. Atmos. Environ. 15: 243-250.
Roemer, W.; Hoek, G.; Brunekreef, B. (1993) Effect of ambient winter air pollution on respiratory health of
children with chronic respiratory symptoms. Am. Rev. Respir. Dis. 147: 118-124.
Roemer, W.; Hoek, G.; Brunekreef, B.; Haluszka, J.; Kalandidi, A.; Pekkanen, J. (1998) Daily variations in air
pollution and respiratory health in a multicentre study: the PEACE project. Eur. Respir. J. 12: 1354-1361.
Roemer, W.; Hoek, G.; Brunekreef, B.; Clench-Aas, J.; Forsberg, B.; Pekkanen, J.; Schutz, A. (2000) PM10
elemental composition and acute respiratory health effects in European children (PEACE project). Eur.
Respir. J. 15: 553-559.
Roemer, W. H.; Van Wijnen, J. H . (2001) Daily mortality and air pollution along busy streets in Amsterdam,
1987-1998. Epidemiology 12: 649-653.
April 2002 8-302 DRAFT-DO NOT QUOTE OR CITE
-------
Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
154: 300-307.
Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
Mexico City. Arch. Environ. Health 52: 368-376.
Rooney, C.; McMichael, A. J.; Kovats, R. S.; Coleman, M. P. (1998) Excess mortality in England and Wales, and in
greater London, during the 1995 heatwave. J. Epidemiol. Community Health 52: 482-486.
Roorda-Knape, M. C.; Janssen, N. A. H.; De Hartog, J. J.; Van Vliet, P. H. N.; Harssema, H.; Brunekreef, B. (1998)
Air pollution from traffic in city districts near major motorways. Atmos. Environ. 32: 1921-1930.
Rosas, I.; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, J.; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis
of the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
emergency admissions to a hospital in Mexico City. Allergy (Copenhagen) 53: 394-401.
Rossi, G.; Vigotti, M. A.; Zanobetti, A.; Repetto, F.; Gianelle, V.; and Schwartz, J. (1999) Air pollution and
cause-specific mortality in Milan, Italy, 1980-1989. Arch. Environ. Health 54: 158-164.
Rothman, K. J.; Greenland, S., eds. (1998) Modern epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven
Publishers.
Rotko, T.; Koistinen, K.; Hanninen, O.; Jantunen, M. (2000) Sociodemographic descriptors of personal exposure to
fine particles (PM25) inEXPOLISHelsinki. J. Exposure Anal. Environ. Epidemiol. 10: 385-393.
Rutherford, S.; Clark, E.; McTainsh, G.; Simpson, R.; Mitchell, C. (1999) Characteristics of rural dust events shown
to impact on asthma severity in Brisbane, Australia. Int. J. Biometeorol. 42: 217-225.
Samet, J. M. (2000) What properties of paniculate matter are responsible for health effects? In: Inhalation
Toxicology: proceedings of the third colloquium on paniculate air pollution and human health; June, 1999;
Durham, NC. Inhalation Toxicology 12(suppl. 1): 19-21.
Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1996) Weather, air pollution and mortality in
Philadelphia, 1973-1980, report to the Health Effects Institute on phase IB, Particle Epidemiology Evaluation
Project. Cambridge, MA: Health Effects Institute; review draft.
Samet, J.; Zeger, S.; Kelsall, J.; Xu, J.; Kalkstein, L. (1998) Does weather confound of modify the association of
paniculate air pollution with mortality? An analysis of the Philadelphia data, 1973-1980. Environ. Res.
77: 9-19.
Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
(2000a) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000b) National morbidity, mortality, and
air pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute;
research report no. 94.
Samet, J. M.; Dominici, F.; Curriero, F. C.; Coursac, L; Zeger, S. L. (2000c) Fine paniculate air pollution and
mortality in 20 U.S. cities, 1987-1994. N. Engl. J. Med. 343: 1742-1749.
Samoli, E.; Schwartz, J.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Balducci, F.; Medina, S.; Rossi, G.; Sunyer, J.;
Bacharova, L.; Anderson, H. R.; Katsouyanni, K. (2001) Investigating regional differences in short-term
effects of air pollution on daily mortality in the APHEA project: a sensitivity analysis for controlling
long-term trends and seasonality. Environ. Health Perspect. 109: 349-353.
Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
Sarnat, J. A.; Schwartz, J.; Catalano, P. J.; Suh, H. H. (2001) Gaseous pollutants in paniculate matter epidemiology:
confounders or surrogates? Environ. Health Perspect. 109: 1053-1061.
Scarlett, J. F.; Abbott, K. J.; Peacock, J. L.; Strachan, D. P.; Anderson, H. R. (1996) Acute effects of summer air
pollution on respiratory function in primary school children in southern England. Thorax 51:1109-1114.
Schouten, J. P.; Vonk, J. M.; de Graaf, A. (1996) Short term effects of air pollution on emergency hospital
admissions for respiratory disease: results of the APHEA project in two major cities in The Netherlands,
1977-89. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on health: a European
approach using epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S22-S29.
Schwartz, J. (1991) Paniculate air pollution and daily mortality in Detroit. Environ. Res. 56: 204-213.
Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137: 1136-1147.
Schwartz, J. (1994) PM10, ozone, and hospital admissions for the elderly in Minneapolis, MN. Arch. Environ.
Health 49: 366-374.
April 2002 8-303 DRAFT-DO NOT QUOTE OR CITE
-------
Schwartz, J. (1995) Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory
disease. Thorax 50: 531-538.
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. (2000a) Assessing confounding, effect modification, and thresholds in the association between ambient
particles and daily deaths. Environ. Health Perspect. 108: 563-568.
Schwartz, J. (2000b) The distributed lag between air pollution and daily deaths. Epidemiology 11: 320-326.
Schwartz, J. (2000c) Harvesting and long term exposure effects in the relation between air pollution and mortality.
Am. J. Epidemiol. 151: 440-448.
Schwartz, J. (2000d) Daily deaths are associated with combustion particles rather than SO2 in Philadelphia.
Occup. Environ. Med. 57: 692-697.
Schwartz, J. (2001) Air pollution and blood markers of cardiovascular risk. Environ. Health Perspect.
109(suppl. 3): 405-409.
Schwartz, J.; Dockery, D. W. (1992) Increased mortality in Philadelphia associated with daily air pollution
concentrations. Am. Rev. Respir. Dis. 145: 600-604.
Schwartz, J.; Marcus, A. (1990) Mortality and air pollution in London: a time series analysis. Am. J. Epidemiol.
131: 185-194.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
Michigan. Am. J. Epidemiol. 142: 23-35.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
with application to air pollution and daily death. Epidemiology 11: 666-672.
Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
Ferris, E.G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
Am. J. Respir. Crit. Care Med. 150: 1234-1242.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles?
J. Air Waste Manage. Assoc. 46: 927-939.
Schwartz, J.; Spix, C.; Touloumi, G.; Bacharova, L.; Barumamdzadeh, T.; le Tertre, A.; Piekarksi, T.; Ponce de
Leon, A.; Ponka, A.; Rossi, G.; Saez, M.; Schouten, J. P. (1996b) Methodological issues in studies of air
pollution and daily counts of deaths or hospital admissions. In: St Leger, S., ed. The APHEA project. Short
term effects of air pollution on health: a European approach using epidemiological time series data.
J. Epidemiol. Community Health 50(suppl. 1): S3-S11.
Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
Schwartz, J.; Laden, F.; Zanobetti, A. (2002) The concentration-response relation between PM25 and daily deaths.
Environ. Health Perspect.: in press.
Seaton, A.; Soutar, A.; Crawford, V.; Elton, R.; McNerlan, S.; Cherrie, J.; Watt, M.; Agius, R.; Stout, R. (1999)
Paniculate air pollution and the blood. Thorax 54: 1027-1032.
Segala, C.; Fauroux, B.; Just, J.; Pascual, L.; Grimfeld, A.; Neukirch, F. (1998) Short-term effect of winter air
pollution on respiratory health of asthmatic children in Paris. Eur. Respir. J. 11: 677-685.
Sheppard, L.; Damian, D. (2000) Estimating short-term PM effects accounting for surrogate exposure
measurements from ambient monitors. Environmetrics 11: 675-687.
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Shima, M.; Adachi, M. (1998) Indoor nitrogen dioxide in homes along trunk roads with heavy traffic. Occup.
Environ. Med. 55: 428-433.
Shumway, R. H.; Tai, R. Y.; Tai, L. P.; Pawitan, Y. (1983) Statistical analysis of daily London mortality and
associated weather and pollution effects. Sacramento, CA: California Air Resources Board; contract no.
Al-154-33.
Shumway, R. H.; Azari, A. S.; Pawitan, Y. (1988) Modeling mortality fluctuations in Los Angeles as functions of
pollution and weather effects. Environ. Res. 45: 224-241.
Simpson, R. W.; Williams, G.; Petroeschevsky, A.; Morgan, G.; Rutherford, S. (1997) Associations between
outdoor air pollution and daily mortality in Brisbane, Australia. Arch. Environ. Health 52: 442-454.
April 2002 8-304 DRAFT-DO NOT QUOTE OR CITE
-------
Smith, M. A.; Jalaludin, B.; Byles, J. E.; Lim, L.; Leader, S. R. (1996) Asthma presentations to emergency
departments in western Sydney during the January 2194 bushfires. Int. J. Epidemiol. 25: 1227-1236.
Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
Soukup, J. M.; Ohio, A. J.; Becker, S. (2000) Soluble components of Utah Valley paniculate pollution alter alveolar
macrophage function in vivo and in vitro. Inhalation Toxicol. 12: 401-414.
Spix, C.; Heinrich, J.; Dockery, D.; Schwartz, J.; Volksch, G.; Schwinkowski, K.; Collen, C.; Wichmann, H. E.
(1993) Air pollution and daily mortality in Erfurt, East Germany, 1980-1989. Environ. Health Perspect.
101: 518-526.
Spix, C.; Anderson, H. R.; Schwartz, J.; Vigotti, M. A.; LeTertre, A.; Vonk, J. M.; Touloumi, G.; Balducci, F.;
Piekarski, T.; Bacharova, L.; Tobias, A.; Ponka, A.; Katsouyanni, K. (1998) Short-term effects of air
pollution on hospital admissions of respiratory diseases in Europe: a quantitative summary of APHEA study
results. Arch. Environ. Health 53: 54-64.
Stieb, D. M.; Burnett, R. T.; Beveridge, R. C.; Brook, J. R. (1996) Association between ozone and asthma
emergency department visits in Saint John, New Brunswick, Canada. Environ. Health Perspect.
104: 1354-1360.
Stieb, D. M.; Beveridge, R. C.; Rowe, B. H.; Walter, S. D.; Judek, S. (1998a) Assessing diagnostic classification in
an emergency department: implications for daily time series studies of air pollution. Am. J. Epidemiol.
148: 666-670.
Stieb, D. M.; Brook, J. R.; Broder, L; Judek, S.; Burnett, R. T.; Beveridge, R. C. (1998b) Personal exposure of
adults with cardiorespiratory disease to paniculate acid and sulfate in Saint John, New Brunswick, Canada.
Appl. Occup. Environ. Hyg. 13: 461-468.
Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
John, Canada. J. Exposure Anal. Environ. Epidemiol.: 10: 461-477.
Sunyer, J.; Basagana, X. (2001) Particles, and not gases, are associated with the risk of death in patients with
chronic obstructive pulmonary disease. Int. J. Epidemiol. 30: 1138-1140.
Sunyer, J.; Spix, C.; Quenel, P.; Ponce-de-Leon, A.; Ponka, A.; Barumandzadeh, T.; Touloumi, G.; Bacharova, L.;
Wojtyniak, B.; Vonk, J.; Bisanti, L.; Schwartz, J.; Katsouyanni, K. (1997) Urban air pollution and emergency
admissions for asthma in four European cities: the APHEA project. Thorax 52: 760-765.
Sunyer, J.; Schwartz, J.; Tobias, A.; Macfarlane, D.; Garcia, J.; Anto, J. M. (2000) Patients with chronic obstructive
pulmonary disease are at increased risk of death associated with urban particle air pollution: a case-crossover
analysis. Am. J. Epidemiol. 151: 50-56.
Taggart, S. C. O.; Custovic, A.; Francis, H. C.; Faragher, E. B.; Yates, C. J.; Higgins, B. G.; Woodcock, A. (1996)
Asthmatic bronchial hyperresponsiveness varies with ambient levels of summertime air pollution. Eur.
Respir. J. 9: 1146-1154.
Tarn, A. Y. C.; Wong, C. M.; Lam, T. H.; Ong, S. G.; Peters, J.; Hedley, A. J. (1994) Bronchial responsiveness in
children exposed to atmospheric pollution in Hong Kong. Chest 106: 1056-1060.
Tan, W. C.; Qiu, D.; Liam, B. L.; Ng, T. P.; Lee, S. H.; Van Eeden, S. F.; D'Yachkova, Y.; Hogg, J. C. (2000) The
human bone marrow response to acute air pollution caused by forest fires. Am. J. Respir. Crit. Care Med.
161: 1213-1217.
Tanaka, H.; Honma, S.; Nishi, M.; Igarashi, T.; Teramoto, S.; Nishio, F.; Abe, S. (1998) Acid fog and hospital visits
for asthma: an epidemiological study. Eur. Respir. J. 11: 1301-1306.
Tellez-Rojo, M. M.; Romieu, I.; Ruiz-Velasco, S.; Lezana, M.-A.; Hernandez-Avila, M. M. (2000) Daily respiratory
mortality and PM10 pollution in Mexico City: importance of considering place of death. Eur. Respir. J.
16: 391-396.
Tenias, J. M.; Ballester, F.; Rivera, M. L. (1998) Association between hospital emergency visits for asthma and air
pollution in Valencia, Spain. Occup. Environ. Med. 55: 541-547.
Thompson, A. J.; Shields, M. D.; Patterson, C. C. (2001) Acute asthma exacerbations and air pollutants in children
living in Belfast, Northern Ireland. Arch. Environ. Health 56: 234-241.
Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
Timonen, K. L.; Pekkanen, J. (1997) Air pollution and respiratory health among children with asthmatic or cough
symptoms. Am. J. Respir. Crit. Care Med. 156: 546-552.
April 2002 8-305 DRAFT-DO NOT QUOTE OR CITE
-------
Tobias, A.; Campbell, M. J. (1999) Modelling influenza epidemics in the relation between black smoke and total
mortality. A sensitivity analysis. J. Epidemiol. Community Health 53: 583-584.
Tolbert, P. E.; Klein, M.; Metzger, K. B.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.; Frumkin, H.
(2000a) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure Anal.
Environ. Epidemiol. 10: 446-460.
Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000b) Air quality and
pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
Touloumi, G.; Katsouyanni, K.; Zmirou, D.; Schwartz, J.; Spix, C.; Ponce de Leon, A.; Tobias, A.; Quennel, P.;
Rabczenko, D.; Bacharova, L.; Bisanti, L.; Vonk, J. M.; Ponka, A. (1997) Short-term effects of ambient
oxidant exposure on mortality: a combined analysis within the APHEA project. Am. J. Epidemiol.
146: 177-185.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (1999) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Berkeley, CA: Lawrence Berkeley National Laboratory,
Environmental Energy Technologies Division; report no. LBNL-43583.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
Turnovska, T.; Kostianev, S. (1999) Effects of reduced air pollution on children's pulmonary function. Cent. Eur. J.
Public Health 7: 77-79.
U.S. Census Bureau. (1995) American housing survey for the United States in 1993. Washington, DC: U.S.
Department of Commerce; current housing reports H150/93. Available:
http://www.census.gov/hhes/www/housing/ahs/nationaldata.html [29 April, 2002].
U.S. Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
U.S. Environmental Protection Agency. (1986) Second addendum to air quality criteria for paniculate matter and
sulfur oxides (1982): assessment of newly available health effects information. Research Triangle Park, NC:
Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report
no. EPA-600/8-86-020F. Available from: NTIS, Springfield, VA; PB87-176574.
U.S. Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park,
NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525, and
PB95-124517.
U.S. Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/00laF-cF. 3v.
U.S. Environmental Protection Agency. (1996b) Air quality criteria for ozone and related photochemical oxidants.
Research Triangle Park, NC: Office of Research and Development; report nos. EPA/600/AP-93/004aF-cF.
3v. Available from: NTIS, Springfield, VA; PB96-185582, PB96-185590, andPB96-185608. Available
online at: www.epa.gov/ncea/ozone.htm.
U.S. Environmental Protection Agency. (2000) Air quality criteria for carbon monoxide. Research Triangle Park,
NC: National Center for Environmental Assessment; report no. EPA/600/P-99/001F. Available:
www.epa.gov/ncea/co/ [2000, October 6].
VanDerZee, S. C.; Hoek, G.; Boezen, H. M.; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (1999) Acute
effects of urban air pollution on respiratory health of children with and without chronic respiratory
symptoms. Occup. Environ. Med. 56: 802-813.
VanDerZee, S. C.; Hoek, G.; Boezen, M. H.; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (2000) Acute
effects of air pollution on respiratory health of 50-70 yr old adults. Eur. Respir. J. 15: 700-709.
Van Wijnen, J. H.; Van Der Zee, S. C. (1998) Traffic-related air pollutants: exposure of road users and populations
living near busy roads. Rev. Environ. Health 13: 1-25.
Vasan, R. S.; Larson, M. G.; Leip, E. P.; Kannel, W. B.; Levy, D. (2001) Assessment of frequency of progression to
hypertension in non-hypertensive participants in the Framingham Heart Study: a cohort study. Lancet
358: 1682-1686.
Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
Vena, J. E. (1982) Air pollution as a risk factor in lung cancer. Am. J. Epidemiol. 116: 42-56.
April 2002 8-306 DRAFT-DO NOT QUOTE OR CITE
-------
Venn, A. J.; Lewis, S. A.; Cooper, M; Hubbard, R.; Britton, J. (2001) Living near a main road and the risk of
wheezing illness in children. Am. J. Respir. Crit. Care Med. 164: 2177-2180.
Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1967) Effects of treatment on
morbidity in hypertension. Results in patients with diastolic blood pressures averaging 115 through 129 mm
Hg. JAMA J. Am. Med. Assoc. 202: 1028-1034.
Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1970) Effects of treatment on
morbidity in hypertension. II. Results in patients with diastolic blood pressure averaging 90 through 114 mm
Hg. JAMA J. Am. Med. Assoc. 213: 1143-1152.
Vigotti, M. A.; Rossi, G.; Bisanti, L.; Zanobetti, A.; Schwartz, J. (1996) Short term effects of urban air pollution on
respiratory health in Milan, Italy, 1980-89. In: Leger, S., ed. The APHEA project. Short term effects of air
pollution on health: a European approach using epidemiological time series data. J. Epidemiol. Community
Health 50(suppl. 1): S71-S75.
Wallace, L. (1996) Indoor particles: a review. J. Air Waste Manage. Assoc. 46: 98-126.
Wang, B.; Peng, Z.; Zhang, X.; Xu, Y.; Wang, H.; Allen, G.; Wang, L.; Xu, X. (1999) Paniculate matter, sulfur
dioxide, and pulmonary function in never-smoking adults in Chongqing, China. Int. J. Occup. Environ.
Healths: 14-19.
Ward, D. J.; Miller, M. R.; Walters, S.; Harrison, R. M.; Ayres, J. G. (2000) Impact of correcting peak flow for
nonlinear errors on air pollutant effect estimates from a panel study. Eur. Respir. J. 15: 137-140.
Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
Respir. Dis. 133:834-842.
Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
Cambridge, MA: Health Effects Institute; Research Report no. 98.
Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
Wjst, M.; Reitmeir, P.; Dold, S.; Wulff, A.; Nicolai, T.; Von Loeffelholz-Colberg, E. F.; Von Mutius, E. (1993)
Road traffic and adverse effects on respiratory health in children. Br. Med. J. 307: 596-600.
Wolff, G. T.; Stroup, C. M.; Stroup, D. P. (1983) The coefficient of haze as a measure of paniculate elemental
carbon. J. AirPollut. Control Assoc. 33: 746-750.
Wong, C. M.; Lam, T. H.; Peters, J.; Hedley, A. J.; Ong, S. G.; Tarn, A. Y. C.; Liu, J.; Spiegelhalter, D. J. (1998)
Comparison between two districts of the effects of an air pollution intervention on bronchial responsiveness
in primary school children in Hong Kong. J. Epidemiol. Community Health 52: 571-578.
Wong, C. M.; Hu, Z. G.; Lam, T. H.; Hedley, A. J.; Peters, J. (1999) Effects of ambient air pollution and
environmental tobacco smoke on respiratory health of non-smoking women in Hong Kong. Int. J. Epidemiol.
28: 859-864.
Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal
infant mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
Wordley, J.; Walters, S.; Ayres, J. G. (1997) Short term variations in hospital admissions and mortality and
paniculate air pollution. Occup. Environ. Med. 54: 108-116.
Wu, W.; Samet, J. M.; Ohio, A. J.; Devlin, R. B. (2001) Activation of the EOF receptor signaling pathway in human
airway epithelial cells exposed to Utah Valley PM. Am. J. Physiol. 281: L483-L489.
Xu, Z.; Yu, D.; Jing, L.; Xu, X. (2000) Air pollution and daily mortality in Shenyang, China. Arch. Environ. Health
55: 115-120.
Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
Inhalation Toxicol. 9: 15-29.
Ye, F.; Piver, W. T.; Ando, M.; Portier, C. J. (2001) Effects of temperature and air pollutants on cardiovascular and
respiratory diseases for males and females older than 65 years of age in Tokyo, July and August 1980-1995.
Environ. Health Perspect. 109: 355-359.
Yu, O.; Sheppard, L.; Lumley, T.; Koenig, J. Q.; Shapiro, G. G. (2000) Effects of ambient air pollution on
symptoms of asthma in Seattle-area children enrolled in the CAMP study. Environ. Health Perspect.
108: 1209-1214.
Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
J. Occup. Environ. Med. 42: 469-474.
Zanobetti, A.; Schwartz, J. (2001) Are diabetics more susceptible to the health effects of airborne particles? Am. J.
Respir. Crit. Care Med. 164: 831-833.
April 2002 8-307 DRAFT-DO NOT QUOTE OR CITE
-------
Zanobetti, A.; Schwartz, J.; Dockery, D. W. (2000a) Airborne particles are a risk factor for hospital admissions for
heart and lung disease. Environ. Health Perspect. 108: 1071-1077.
Zanobetti, A.; Wand, M. P.; Schwartz, J.; Ryan, L. M. (2000b) Generalized additive distributed lag models:
quantifying mortality displacement. Biostatistics 1: 279-292.
Zeger, S. L.; Dominici, F.; Samet, J. (1999) Harvesting-resistant estimates of air pollution effects on mortality.
Epidemiology 10: 171-175.
Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
Perspect. 108: 419-426.
Zeghnoun, A.; Beaudeau, P.; Carrat, F.; Delmas, V.; Boudhabhay, O.; Gayon, F.; Guincetre, D.; Czernichow, P.
(1999) Air pollution and respiratory drug sales in the city of Le Havre, France, 1993-1996. Environ. Res.
81:224-230.
Zeghnoun, A.; Czernichow, P.; Beaudeau, P.; Hautemaniere, A.; Froment, L.; Le Tertre, A.; Quenel, P. (2001)
Short-term effects of air pollution on mortality in the cities of Rouen and Le Havre, France, 1990-1995.
Arch. Environ. Health 56: 327-335.
Zemp, E.; Elsasser, S.; Schindler, C.; Ktinzli, N.; Perruchoud, A. P.; Domenighetti, G.; Medici, T.;
Ackermann-Liebrich, U.; Leuenberger, P.; Monn, C.; Bolognini, G.; Bongard, J.-P.; Brandli, O.; Karrer, W.;
Keller, R.; Schoni, M. H.; Tschopp, J.-M; Villiger, B.; Zellweger, J.-P.; SAPALDIA Team. (1999)
Long-term ambient air pollution and respiratory symptoms in adults (SAPALDIA study). Am. J. Respir. Crit.
Care Med. 159: 1257-1266.
Zhang, J.; Qian, Z.; Kong, L.; Zhou, L.; Yan, L.; Chapman, R. S. (1999) Effects of air pollution on respiratory
health of adults in three Chinese cities. Arch. Environ. Health 54: 373-381.
Zhang, H.; Triche, E.; Leaderer, B. (2000) Model forthe analysis of binary time series of respiratory symptoms.
Am. J. Epidemiol. 151: 1206-1215.
Zmirou, D.; Schwartz, J.; Saez, M.; Zanobetti, A.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Ponce de Leon, A.;
Le Moullec, Y.; Bacharova, L.; Schouten, J.; Ponka, A.; Katsouyanni, K. (1998) Time-series analysis of air
pollution and cause-specific mortality. Epidemiology 9: 495-503.
April 2002 8-308 DRAFT-DO NOT QUOTE OR CITE
-------
APPENDIX 8A
SHORT-TERM PM EXPOSURE-MORTALITY
STUDIES: SUMMARY TABLE
April 2002 8A-1 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8A-1. SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
IQRin//g/m3.
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
fe
H
6
o
o
H
O
O
H
W
O
O
United States
Samet et al. (2000a,b).
90 largest U.S. cities.
1987-1994.
PM10 mean ranged from
15.3 (Honolulu) to
52.0 (Riverside).
Dominici et al. (2000).
20 largest U.S. cities.
1987-1994. PM10 mean
ranged from 23.8 /j,g/w?
(San Antonio) to
52.0 Mg/m3 (Riverside).
Non-accidental total deaths and cause-specific (cardiac,
respiratory, and the other remaining) deaths, stratified in
three age groups (<65, 65-75, 75+), were examined for their
associations with PM10, O3, SO2, NO2, and CO (single, two,
and three pollutant models) at lags 0, 1, and 2 days. In the
first stage of the hierarchical model, RRs for the pollutants
for each city were obtained using GAM Poisson regression
models, adjusting for temperature and dewpoint (0-day and
average of 1-3 days for both variables), day-of-week,
seasonal cycles, intercept and seasonal cycles for three age
groups. In the second stage, between-city variation in RRs
were modeled within region. The third stage modeled
between-region variation (7 regions). Two alternative
assumptions were made regarding the prior distribution:
one with possibly substantial heterogeneity and the other
with less or no heterogeneity within region. The weighted
second-stage regression included five types of county-
specific variables: (1) mean weather and pollution
variables; (2) mortality rate; (3) socio-demographic
variables; (4) urbanization; (5) variables related to
measurement error.
Non-accidental total deaths (stratified in three age
groups: <65, 65-75, 75+) were examined for their
associations with PM10 and O3 (single, 2, and
3 pollutant models) at lags 0, 1, and 2 days. In the
first stage of the hierarchical model, RRs for PM10
and O3 for each city were obtained using GAM
Poisson regression models, adjusting for temperature
and dewpoint (0-day and average of 1-3 days for both
variables), day-of-week, seasonal cycles, intercept
and seasonal cycles for three age groups. In the
second stage, between-city variation in RRs were
modeled as a function of city-specific covariates
including mean PM10 and O3 levels, percent poverty,
and percent of population with age 65 and over. The
prior distribution assumed heterogeneity across cities.
To approximate the posterior distribution, a Markov
Chain Monte Carlo (MCMC) algorithm with a block
Gibbs sampler was implemented. The second stage
also considered spatial model, in which RRs in closer
cities were assumed to be more correlated.
The estimated city-specific coefficients were mostly positive at
lag 0, 1, and 2 days (estimated overall effect size was largest at
lag 1, with the estimated percent excess death rate per 10 ,ug/m3
PM10 being about 0.5%). The posterior probabilities that the
overall effects are greater than 0 at these lags were 0.99, 1.00,
and 0.98, respectively. None of the county-specific variables
(effect modifiers) in the second-stage regression significantly
explained the heterogeneity of PM10 effects across cities. In the
3-stage regression model with the index for 7 geographical
regions, the effect of PM10 varied somewhat across the 7 regions,
with the effect in the Northeast being the greatest. Adding O3
and other gaseous pollutants did not markedly change the
posterior distributions of PM10 effects. O3 effects, as examined
by season, were associated with mortality in summer (-0.5
percent per 10 ppb increase), but not in all season data (negative
in winter).
Lag 1 day PM10 concentration positively associated with
total mortality in most locations (only 2 out of
20 coefficients negative), though estimates ranged from
2.1% to -0.4% per 10 Mg/m3 PM10 increase. PM10
mortality associations changed little with the addition of
O3 to the model, or with the addition of a third pollutant
in the model. The pattern of PM10 effects with respiratory
and cardiovascular were similar to that of total mortality.
The PM10 effect was smaller (and weaker) with other
causes of deaths. The pooled analysis of 20 cities data
confirmed the overall effect on total and cardiorespiratory
mortality, with lag 1 day showing largest effect estimates.
The posterior distributions for PM10 were generally not
influenced by addition of other pollutants. In the data for
which the distributed lags could be examined (i.e., nearly
daily data), the sum of 7-day distributed lag coefficients
was greater than each of single day coefficients. City-
specific covariates did not predict the heterogeneity
across cities. Regional model results suggested that PM10
effects in West U.S. were larger than in East and South.
Posterior mean estimates and 95%
credible intervals for total mortality
excess deaths per 50 ,ug/m3 increase in
PM10 at lag 1 day: 2.3% (0.1, 4.5) for
"more heterogeneity" across-city
assumption; 2.2% (0.5, 4.0) for "less or
no heterogeneity" across cities
assumption. The largest PM10 effect
estimated for 7 U.S. regions was for the
Northeast: 4.6% (2.7, 6.5) excess
deaths per 50 //g/m3 PM10 increment.
Total mortality excess deaths per 50
Mg/m3 increase in PM10: 1.8 (-0.5,
4.1) for lag 0; 1.9 (-0.4, 4.3) for lag
1; 1.2 (-1.0, 3.4) for lag 2.
Cardiovascular disease excess
deaths per 50 /-ig/m3 PM10:
3.4(1.0,5.9).
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments. Design Issues, Uncertainties,
Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
United States (cont'd)
Braga et al. (2000). Five U.S.
cities: Pittsburgh, PA; Detroit,
MI; Chicago, IL; Minneapolis-
St. Paul, MN; Seattle, WA.
1986-1993. PM10 means were
35, 37, 37, 28, and 33 //g/m3,
respectively in these cities.
Braga etal. (2001).
Ten U.S. cities.
Same as Schwartz (2000b).
Potential confounding caused by respiratory epidemics on When respiratory epidemics were adjusted for, small decreases The overall estimated percent excess
PM-total mortality associations was investigated in a subset
of the 10 cities evaluated by Schwartz (2000a,b), as
summarized below. GAM Poisson models were used to
estimate city-specific PM10 effects, adjusting for
temperature, dewpoint, barometric pressure, time-trend and
day-of-week. A cubic polynomial was used to for each
epidemic period, and a dummy variable was used to control
for isolated epidemic days. Average of 0 and 1 day lags
were used.
The study examined the lag structure of PM10 effects on
respiratory and cardiovascular cause-specific mortality.
Using GAM Poisson model adjusting for temporal pattern
and weather, three types of lag structures were examined:
(1) 7-day unconstrained distributed lags; (2) 2-day average
(0- and 1-day lag); and (3) 0-day lag. The results were
combined across 10 cities.
in the PM10 effect were observed (9% in Chicago, 11% in
Detroit, 3% in Minneapolis, 5% in Pittsburgh, and 15% in
Seattle).
The authors reported that respiratory deaths were more affected
by air pollution levels on the previous days, whereas
cardiovascular deaths were more affected by same-day pollution.
Pneumonia, COPD, all cardiovascular disease, and myocardial
infarction were all associated with PM10 in the three types of lags
examined. The 7-day unconstrained lag model did not always
give larger effect size estimates compared others.
deaths per 50 ,ug/m3 increase in PM10
was 4.3% (3.0, 5.6) before controlling
for epidemics and 4.0% (2.6, 5.3) after.
Average of 0 and 1 day lags.
In the 7-day unconstrained distributed
lag model, the estimated percent excess
deaths per 10ug/m3 PM10 were 2.7%
(1.5,2.9), 1.7% (0.1, 3.3), 1.0% (0.6,
1.4), and 0.6% (0.0, 1.2) for
pneumonia, COPD, all cardiovascular,
and myocardial infarction mortality,
respectively.
fe
H
6
o
o
H
O
O
H
W
O
Schwartz (2000a).
Ten U.S. cities: New Haven,
CT; Pittsburgh, PA; Detroit,
MI; Birmingham, AL;
Canton, OH; Chicago, IL;
Minneapolis-St. Paul, MN;
Colorado Springs, CO;
Spokane, WA; and Seattle,
WA. 1986-1993. PM10 means
were 29, 35, 36, 37, 29, 37, 28,
27, 41, and 33, respectively in
these cities.
Daily total (non-accidental) deaths (20, 19, 63, 60, 10, 133,
32, 6, 9, and 29, respectively in these cities in the order
shown left). Deaths stratified by location of death (in or
outside hospital) were also examined. For each city, a GAM
Poisson model adjusting for temperature, dewpoint,
barometric pressure, day-of-week, season, and time was
fitted. The data were also analyzed by season (November
through April as heating season). In the second stage, the
PM10 coefficients were modeled as a function of city-
dependent covariates including copollutant to PM10
regression coefficient (to test confounding), unemployment
rate, education, poverty level, and percent non-white.
Threshold effects were also examined. The inverse variance
weighted averages of the ten cities' estimates were used to
combine results.
PM10 was significantly associated with total deaths, and the
effect size estimates were the same in summer and winter.
Adjusting for other pollutants did not substantially change PM10
effect size estimates. Also, socioeconomic variables did not
modify the estimates. The effect size estimate for the deaths that
occurred outside hospitals was substantially greater than that for
inside hospitals. The effect size estimate was larger for subset
with PM10 less than 50 /ig/m3.
The total mortality RR estimates
combined across cities per 50 //g/m3
increase of mean of lag 0- and 1-days
PM10: overall 3.4 (2.7, 4.1); summer
3.4 (2.4, 4.4); winter 3.3 (2.3, 4.4); in-
hospital 2.5 (1.5, 3.4); out-of-hospital
4.5 (3.4, 5.6); days < 50 ,ug/m3 4.4 (3.1,
5.7); with SO2 2.9 (1.2, 4.6); with CO
4.6 (3.2, 6.0); with O3 3.5 (1.6, 5.3).
o
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Schwartz (2000b).
Ten U.S. cities: New Haven,
CT; Pittsburgh, PA;
Birmingham, AL; Detroit, MI;
Canton, OH; Chicago, IL;
Minneapolis-St. Paul, MN;
Colorado Springs, CO;
Spokane, WA; and Seattle,
WA. 1986-1993. PM10 means
were 29, 35, 36, 37, 29, 37, 28,
27, 41, and 33, respectively in
these cities.
Schwartz and Zanobetti (2000).
Ten U.S. cities.
Same as above.
Zanobetti and Schwartz (2000).
Four U.S. cities: Chicago, IL;
Detroit, MI; Minneapolis-St.
Paul, MN; Pittsburgh, PA.
1986-1993. PM10 median = 33,
33, 25, and 31 respectively for
these cities.
The issue of distributed lag effects was the focus of this
study. Daily total (non-accidental) deaths of persons 65
years of age and older were analyzed. For each city, a GAM
Poisson model adjusting for temperature, dewpoint,
barometric pressure, day-of-week, season, and time was
fitted. Effects of distributed lag were examined using four
models: (1) 1-day mean at lag 0 day; (2) 2-day mean at lag 0
and 1 day; (3) second-degree distributed lag model using
lags 0 through 5 days; (4) unconstrained distributed lag
model using lags 0 through 5 days.
The inverse variance weighted averages of the ten cities'
estimates were used to combine results.
The issue of a threshold in PM-mortality exposure-response
curve was the focus of this study. First, a simulation was
conducted to show that the "meta-smoothing" could
produce unbiased exposure-response curves. Three
hypothetical curves (linear, piecewise linear, and
logarithmic curves) were used to generate mortality series in
10 cities, and GAM Poisson models were used to estimate
exposure response curve. Effects of measurement errors
were also simulated. In the analysis of actual 10 cities data,
GAM Poisson models were fitted, adjusting for temperature,
dewpoint, and barometric pressure, and day-of-week.
Smooth function of PM10 with the same span (0.7) in each
of the cities. The predicted values of the log relative risks
were computed for 2 ,ug/m3 increments between 5.5 ,ug/m3
and 69.5 ,ug/m3 of PM10 levels. Then, the predicted values
were combined across cities using inverse-variance
weighting.
Separate daily counts of total non-accidental deaths,
stratified by sex, race (black and white), and education
(education > 12yrs or not), were examined to test hypothesis
that people in each of these groups had higher risk of PM10.
GAM Poisson models adjusting for temperature, dewpoint,
barometric pressure, day-of-week, season, and time were
used. The mean of 0- and 1-day lag PM10 was used. The
inverse variance weighted averages of the four cities'
estimates were used to combine results.
The effect size estimates for the quadratic distributed model and
unconstrained distributed lag model were similar. Both
distributed lag models resulted in substantially larger effect size
estimates than the single day lag, and moderately larger effect
size estimates than the two-day average models.
The simulation results indicated that the "meta-smoothing"
approach did not bias the underlying relationships for the linear
and threshold models, but did result in a slight downward bias
for the logarithmic model. Measurement error (additive or
multiplicative) in the simulations did not cause upward bias in
the relationship below threshold. The threshold detection in the
simulation was not very sensitive to the choice of span in
smoothing. In the analysis of real data from 10 cities, the
combined curve did not show evidence of a threshold in the
PM10-mortality associations.
The differences in the effect size estimates among the various
strata were modest. The results suggest effect modification with
the slope in female deaths one third larger than in male deaths.
Potential interaction of these strata (e.g., black and female) were
not investigated.
Total mortality percent increase
estimates combined across cities per
50 /^g/m3 increase in PM10: 3.3 (2.5,
4.1) for 1-day mean at lag 0; 5.4 (4.4,
6.3) 2-day mean of lag 0 and 1; 7.3
(5.9, 8.6) for quadratic distributed lag;
and 6.6 (5.3, 8.0) for unconstrained
distributed lag.
The combined exposure-response curve
indicates that an increase of 50 ,ug/m3 is
associated with about a 4% increase in
daily deaths. Avg. of 0 and 1 day lags.
The total mortality RR estimates
combined across cities per 50 ,ug/m3
increase of mean of lag 0- and 1-days
PM10: white 5.0 (4.0, 6.0); black 3.9
(2.3, 5.4); male 3.8 (2.7, 4.9); female
5.5 (4.3, 6.7); education <12y 4.7 (3.3,
6.0); education> 12y3.6 (1.0, 6.3).
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Moolgavkar (2000a)
Cook County, Illinois
Los Angeles County, CA
Maricopa County, AZ
1987-1995
PM10, CO, O3, NO2, SO2 in
all three locations.
PM2 5 in Los Angeles County.
Cook Co:
PM10 Median = 47 //g/m3.
Maricopa Co:
PM10 Median = 41.
Los Angeles Co:
PM10 Median = 44;
PM,, Median = 22.
Associations between air pollution and time-series of daily
deaths evaluated for three U.S. metropolitan areas with
different pollutant mixes and climatic conditions. Daily
total non-accidental deaths and deaths from cardiovascular
disease (CVD), cerebrovascular (CrD), and chronic
obstructive lung disease and associated conditions (COPD)
were analyzed by generalized additive Poisson models in
relation to 24-h readings for each of the air pollutants
averaged over all monitors in each county. All models
included an intercept term for day-of-week and a spline
smoother for temporal trends. Effects of weather were first
evaluated by regressing daily deaths (for each mortality
endpoint) against temp and rel. humidity with lag times of 0
to 5 days. Then lags that minimized deviance for temp and
rel. humidity were kept fixed for subsequent pollutant effect
analyses. Each pollutant entered linearly into the regression
and lags of between 0 to 5 days examined. Effects of two or
more pollutants were then evaluated in multipollutant
models. Sensitivity analyses were used to evaluate effect of
degree of smoothing on results.
In general, the gases, especially CO (but not O3) were much
more strongly associated with mortality than PM. Specified
pattern of results found for each county were as follows. For
Cook Co., in single pollutant analyses PM10, CO, and O3 were all
associated (PM10 most strongly on lag 0-2 days) with total
mortality, as were SO2 and NO2 (strongest association on lag 1
day for the latter two). In joint analyses with one of gases, the
coefficients for both PM10 and the gas were somewhat
attenuated, but remained stat. sig. for some lags. With
3-pollutant models, PM10 coefficient became small and non-sig.
(except at lag 0), whereas the gases dominated. For Los
Angeles, PM10, PM2 5, CO, NO2, and SO2, (but not O3), were all
associated with total mortality. In joint analyses with CO or SO2
and either PM10 or PM2 5, PM metrics were markedly reduced
and non-sig., whereas estimates for gases remained robust. In
Maricopa Co. single-pollutant analyses, PM10 and each of the
gases, (except O3), were associated with total morality; in
2-pollutant models, coefficients for CO, NO2, SO2, were more
robust than for PM10. Analogous patterns of more robust
gaseous pollutant effects were generally found for cause-specific
(CVD, CrD, COPD) mortality analyses. Author concluded that
while direct effect of individual components of air pollution
cannot be ruled out, individual components best thought of as
indices of overall pollutant mix.
In single pollutant models, estimated
daily total mortality % excess deaths per
50 /2g/m3 PM10 was mainly in range of:
0.5-1.0% lags 0-2 Cook Co.; 0.25-1.0%
lags 0-2 LA; 2.0% lag 2 Maricopa.
Percent per 25 ^g/m3 PM25 0.5% lags 0,
1 for Los Angeles.
Maximum estimated COPD % excess
deaths (95% CI) per 50 ,ug/m3 PM10:
Cook Co. 5.4 (0.3,10.7), lag 2; with O3,
3.0 (-1.8, 8.1) lag 2; LA 5.9 (-1.6,
14.0) lag 1; Maricopa 8.2 (-4.2, 22.3)
lag 1; per 25 ,ug PM25 in LA 2.7 (-3.4,
9.1).
CVD % per 50 ,ug/m3 PM10:
Cook 2.2 (0.4, 4.1) lag 3; with O3, SO2
1.99 (-0.06, 4.1) lag 3; LA 4.5 (1.7,
7.4) lag 2; with CO -0.56 (-3.8, 2.8)
lag 2; Maricopa 8.9 (2.7, 15.4) lag 1;
with NO2 7.4 (-0.95, 16.3) lag 1.
Percent per 25 ,ug/m3 PM2 5, LA 2.6
(0.4, 4.9) lagl; with CO 0^60 (-2.1,
3.4).
CrD % per 50 ,ug/m3 PM10:
Cook 3.3 (-0.12, 6.8) lag 2; LA 2.9
(-2.3, 8.4) lag 3; Maricopa 11.1 (0.54,
22.8) lag 5. Percent per 25 Mg/m3
PM25, LA 3.6 (-0.6, 7.9) lag 3.
o
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
IQRin,ug/m3.
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Ostroetal. (1999a).
Coachella Valley, CA.
1989-1992. PM10
(beta-attenuation)
Mean = 56.8 ,ug/m3.
Ostro et al. (2000).
Coachella Valley, CA.
1989-1998.
PM25= 16.8;
PM10.25 = 25.8inIndio;
PM25 = 12.7;
PM10.25 = 17.9 in Palm Springs.
Study evaluated total, respiratory, cardiovascular, non-
cardiorespiratory and age >50 yr deaths (mean = 5.4, 0.6,
1.8, 3.0, and 4.8 per day, respectively). The valley is a
desert area where 50-60% of PM10 estimated to be coarse
particles. Correlation between gravimetric and beta-
attenuation, separated by 25 miles, was high (r = 0.93).
Beta-attenuation data were used for analysis. GAM Poisson
models adjusting for temperature, humidity, day-of-week,
season, and time were used. Seasonally stratified analyses
were also conducted. Lags 0-3 days (separately) of PM10
along with moving averages of 3 and 5 days examined, as
were O3, NO2, and CO.
A follow-up study of the Coachella Valley data, with PM2 5
and PM10.2 5 data in the last 2.5 years. Both PM2 5 and PM10.
2 5 were estimated for the remaining years to increase power
of analyses.
Associations were found between 2- or 3-day lagged PM10 and
all mortality categories examined, except non-cardiorespiratory
series. The effect size estimates for total and cardiovascular
deaths were larger for warm season (May through October) than
for all year period. NO2 and CO were significant predictor of
mortality in single pollutant models, but in multi-pollutant
models, none of the gaseous pollutants were significant
(coefficients reduced), whereas PM10 coefficients remained the
same and significant.
Several pollutants were associated with all-cause mortality,
including PM2 5, CO, and NO2. More consistent results were
found for cardiovascular mortality, for which significant
associations were found for PM10_2 5 and PM10, but not PM2 5
(possibly due to low range of PM2 5 concentrations and reduced
sample size for PM25 data).
Total mortality percent excess deaths
per 50 f2g/ m3 PM10 at 2-day lag= 4.6
(0.6, 8.8).
Cardiac deaths:
8.33(2.14, 14.9)
Respiratory deaths:
13.9(3.25,25.6)
Total percent excess deaths:
PM10 = 2.0 (-1.0, 5.1) per 50 //g/m3
PM25= 11.5(0.2, 24.1) per 25 |/g/m3
PMio-2.5 = 1.3 (-0.6, 3.5) per 25 ^g/m3
Cardio deaths:
PM10 = 6.1 (2.0, 10.3) per 50 ,ug/m3
PM25 = 8.6 (-6.4, 25.8) per 25 ^g
PM10.25 = 2.6 (0.7, 4.5) per 25
Respiratory deaths:
PM10= -2.0 (-11.4, 8.4) per 50
PM25= 13.3 (-43.1,32.1) per 25 Mg/
PM10.25 = -1.3 (-6.2, 4.0) per 25 ,ug/m3
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Fairley(1999).
Santa Clara County, CA
1989-1996.
PM25(13);PM10(34);
PM10.25(ll);COH(0.5unit);
NO3(3.0);SO4(1.8)
Schwartz etal. (1999).
Spokane, WA
1989-1995
PM10: "control" days:
42 Mg/m3;
dust-storm days: 263
Pope etal. (1999a).
Ogden, Salt Lake City, and
Provo/Orem, UT
1985-1995
PM10 (32 for Ogden;
41 for SLC; 38 for P/O)
Total, cardiovascular, and respiratory deaths were regressed
on PM10, PM2 5, PM10.2 5, COH, nitrate, sulfate, O3, CO, NO2,
adjusting for trend, season, and min and max temperature,
using Poisson GAM model. Season-specific analysis was
also conducted. The same approach was also used to re-
analyze 1980-1986 data (previously analyzed by Fairley,
1990).
Effects of high concentration of coarse crustal particles were
investigated by comparing death counts on 17 dust storm
episodes to those on non-episode days on the same day of
the years in other years, adjusting for temperature,
dewpoint, and day-of-week, using Poisson regression.
Associations between PM10 and total, cardiovascular, and
respiratory deaths studied in three urban areas in Utah's
Wasatch Front, using Poisson GAM model and adjusting for
seasonality, temperature, humidity, and barometric pressure.
Analysis was conducted with or without dust (crustal coarse
particles) storm episodes, as identified on the high "clearing
index" days, an index of air stagnation.
PM2 5 and nitrate were most significantly associated with
mortality, but all the pollutants (except PM10_2 5) were
significantly associated in single poll, models. In 2 and 4 poll.
models with PM2 5 or nitrate, other pollutants were not
significant. The RRs for respiratory deaths were always larger
than those for total or cardiovascular deaths. The difference in
risk between season was not significant for PM2 5. The 1980-
1986 results were similar, except that COH was very
significantly associated with mortality.
No association was found between the mortality and dust storm
days on the same day or the following day.
Salt Lake City (SLC), where past studies reported little PM10-
mortality associations, had substantially more dust storm
episodes. When the dust storm days were screened out from
analysis and PM10 data from multiple monitors were used,
comparable RRs were estimated for SLC and Provo/Orem (P/O).
Total mortality per 25 ,ug/m3 PM2 5 at 0
d lag: 8% in one pollutant model;
9-12
'. pollutant model; 12
4-pollutant model. Also, 8% per
50 f/g/m3 PM10 in one pollutant model
and 2% per 25 ,ug/m3 PM10.25.
Cardiovascular mortality:
PM10 = 9% per 50 ^g/m3
PM25= 13%per25,ug/m3
PM10.25 = 3% per 25 ^g/m3
Respiratory mortality:
PM10 = 1 1% per 50 ^g/m3
PMio-2.5 = 16% per 25 ,ug/m3
0% (-4.5, 4.7) for dust storm days at 0
day lag (50 ,ug/m3 PM10) (lagged days
also reported to have no associations).
Ogden PM10
Total (0 d) = 12.0% (4.5, 20.1)
CVD(0-4 d) = 1.4% (-8.3, 12.2)
Resp. (0-4 d) = 23.8 (2.8, 49.1)
SLC PM10
Total (0 d) = 2.3% (0.47)
CVD(0-4d) = 6.5%(2.2, 11.0)
Resp. (0-4 d) = 8.2 (2.4, 15.2)
Provo/Ovem PM10
Total (Od)= 1.9% (-2.1,6.0)
CVD (0-4 d) = 8.6% (2.4, 15.2)
Resp. (0-4 d) = 2.2% (-9.8, 15.9)
Note: Above % for PM2.5 and PM10.25
all per 25 ,ug/m3; all PM10 % per
50
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
OO
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Schwartz and Zanobetti (2000).
Chicago 1988-1993.
PM10. Median = 36 ,ug/m3.
Lippmann et al. (2000).
Detroit, MI. 1992-1994.
PM10 = 31;
PM25 = 18;
PM10.25=13.
For 1985-1990 period
TSP, PM10, TSP-PM10,
Sulfate from TSP (TSP-SO4')
Total (non-accidental), in-hospital, out-of-hospital deaths
(median = 132, 79, and 53 per day, respectively), as well as
heart disease, COPD, and pneumonia elderly hospital
admissions (115, 7, and 25 per day, respectively) were
analyzed to investigate possible "harvesting" effect of PM10.
GAM Poisson models adjusting for temperature, relative
humidity, day-of-week, and season were applied in baseline
models using the average of the same day and previous
day's PM10. The seasonal and trend decomposition
techniques called STL was applied to the health outcome
and exposure data to decompose them into different time-
scales (i.e, short-term to long-term), excluding the long,
seasonal cycles (120 day window). The associations were
examined with smoothing windows of 15, 30, 45, and 60
days.
For 1992-1994 study period, total (non-accidental),
cardiovascular, respiratory, and other deaths were analyzed
using GAM Poisson models, adjusting for season,
temperature, and relative humidity. The air pollution
variables analyzed were: PM10, PM2 5, PM10_2 5, sulfate, H+,
O3, SO2, NO2, and CO.
For earlier 1985-1990 study period, total non-accidental,
circulatory, respiratory, and "other" (non-circulatory or
respiratory non-accidental) mortality were evaluated versus
noted PM indices and gaseous pollutants.
The effect size estimate for deaths outside of the hospital is
larger than for deaths inside the hospital. All cause mortality
shows an increase in effect size at longer time scales. The effect
size for deaths outside of hospital increases more steeply with
increasing time scale than the effect size for deaths inside of
hospitals.
PM10, PM2 5, and PM10_2 5 were more significantly associated with
mortality outcomes than sulfate or H+. PM coefficients were
generally not sensitive to inclusion of gaseous pollutants. PM10,
PM2 5, and PM10_2 5 effect size estimates were comparable per
same distributional increment (5th to 95th percentile).
Both PM10 (lag 1 and 2 day) and TSP (lag 1 day) but not TSP-
PM10 or TSP-SO4" significantly associated with respiratory
mortality for 1985-1990 period. The simultaneous inclusions of
gaseous pollutants with PM10 or TSP reduced PM effect size by 0
to 34%. Effect size estimates for total, circulatory, and "other"
categories were smaller than for respiratory mortality.
Mortality RR estimates per 50 ,ug/m3
increase of mean of lag 0- and 1-days
PM10: total deaths 4.5 (3.1, 6.0);
in-hospital 3.9 (2.1, 5.8); out-of-
hospital 6.3 (4.1, 8.6). For total deaths,
the RR approximately doubles as the
time scale changes from 15 days to 60
days. For out-of-hospital deaths, it
triples from 15 days to 60 days time
scale.
Total mortality percent excess deaths:
PM10(1 d) = 4.4%(-1.0, 10.1)
PM25(Od) = 3.1%(-0.6, 7.0)
PM10.25 (1 d) = 4.0% (-1.2, 9.4)
PM10(1 d) = 6.9%(-1.3, 15.7)
PM25(1 d) = 3.2%(-2.3, 8.9)
PM10.25 (1 d) = 7.8% (0.0, 16.2)
Respiratory mortality:
PM10 (0 d) = 7.8% (-10.2, 29.5)
Circulatory mortality:
PM25(od) = 2.3%(-10.3, 16,6)
PM10.25 (2 d) = 7.4% (-9.1, 26.9)
Note: All above PM10 per 50 Mg/m3; all
PM25 and PM10.25 % per 25 //g/m3.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Chock et al. (2000).
1989-1991
Pittsburgh, PA
PM10 (daily)
PM2 5 (every 2 days)
Klemm and Mason (2000).
Atlanta, GA
1998-1999
PM25mean=19.9;
PM2yPM10 =0.65.
Nitrate, EC, OC, and
oxygenated HC.
Gwynn et al. (2000).
Buffalo, N.Y. 1988-1990.
PM10 (24); COH (0.2 /1000ft);
SO4= (62 nmoles/m3)
Schwartz (2000c).
Boston, MA.
1979-1986.
PM25mean= 15.6.
Study evaluated associations between daily mortality and
several air pollution variables (PM10, PM2 5, CO, O3, NO2,
SO2) in two age groups (<75 yr., > 75 yr.) in Pittsburgh, PA,
during 3-yr. period. Poisson regression used, including
filtering of data based on cubic B-spline basis functions,
with adjustments for seasonal trends, day-of-week effects,
temp., dew point. Single- and multi-pollutant models run
for 0, 1, 2, and 3 day lags. PM25/PM10 * 0.67.
Reported "interim" results for 1 yr period of observations
regarding total mortality in Atlanta, GA during 1998-1999.
Generalized additive model used to assess effects of PM25
vs PM10_2 5, and for nitrate, EC, OC and oxygenated HC
components.
Total, circulatory, and respiratory mortality and unscheduled
hospital admissions were analyzed for their associations
with H+, SO4=, PM10, COH, O3, CO, SO2, and NO2,
adjusting for seasonal cycles, day-of-week, temperature,
humidity, using. Poisson and negative binomial GAM
models.
Non-accidental total, pneumonia, COPD, and ischemic
heart disease mortality were examined for possible
"harvesting" effects of PM. The mortality, air pollution, and
weather time-series were separated into seasonal cycles
(longer than 2-month period), midscale, and short-term
fluctuations using STL algorithm. Four different midscale
components were used (15, 30, 45, and 60 days) to examine
the extent of harvesting. GAM Poisson regression analysis
was performed using deaths, pollution, and weather for each
of the four midscale periods.
Issues of seasonal dependence of correlation among pollutants,
multi-collinearity among pollutants, and instability of
coefficients emphasized. Single- and multi-pollutant non-
seasonal models show significant positive association between
PM10 and daily mortality, but seasonal models showed much
multi-collinearity, masking association of any pollutant with
mortality. Also, based on data set half the size for PM10, the
PM2 5 coefficients were highly unstable and, since no
consistently significant associations found in this small data set
stratified by age group and season, no conclusions drawn on
relative role of PM25 vs. PM10_25.
No significant associations were found for any of the pollutants
examined, possibly due to a relatively short study period (1-
year). The coefficient and t-ratio were larger for PM2 5 than for
PM10.2.5.
For total mortality, all the PM components were significantly
associated, with H+ being the most significant, and COH the
least significant predictors. The gaseous pollutants were mostly
weakly associated with total mortality.
For COPD deaths, the results suggest that most of the mortality
was displaced by only a few months. For pneumonia, ischemic
heart disease, and total mortality, the effect size increased with
longer time scales.
Total mortality percent increase per
25 ,ug/m3 for aged <75 yrs:
PM2 5 = 2.6% (2.0, 7.3)
PM10.25 = 0.7%(-1.7, 3.7)
Total mortality percent increase per 25
,ug/m3 for aged >75 yrs:
PM25= 1.5%(-3.0, 6.3)
PM10.25 = 1.3%(-1.3, 3.8)
Total mortality percent increase per 25
,ug/m3 for:
PM25 = 4.8%(-3.2, 13.4)
PM10.2.5 = 1.4% (-11.3, 15.9)
12% (2.6, 22.7) per 50 ,ug/m3 PM10 at 2-
day lag.
Total mortality percent increase per
25 ,ug/m3 increase in PM25:
5.3 (1.8, 9.0) for short-term
fluctuations; 9.6 (8.1, 11.1) for 1he 60
day window.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Lipfert et al. (2000a).
Philadelphia (7 county
Metropolitan area), 1992-1995.
Harvard PM measurements:
PM25(17.3);PM10(24.1);
PM10.2.5 (6.8),
sulfate (53.1 nmol/m3);
H+(8.0nmol/m3).
Laden et. al. (2000)
Six Cities (means):
Watertown, MA (16.5);
Kingston-Harriman, TN (21.1);
St. Louis, MO (19.2);
Steubenville, OH (30.5);
Portage, WI (11.3); Topeka, KS
(12.2). 1979-1988?. 15 trace
elements in the dichot PM2 5:
Si, S, Cl, K, Ca, V, Mn, Al, Ni,
Zn, Se, Br, Pb, Cu, and Fe.
Levy (1998).
King County, WA.
1990-1994.
PM10 Nephelometer (30);
(0.59bspunit)
12 mortality variables, as categorized by area, age, and
cause, were regressed on 29 pollution variables (PM
components, O3, SO2, NO2, CO, and by sub-areas), yielding
348 regression results. Both dependent and explanatory
variables were pre-filtered using the!9-day-weighted
average filter prior to OLS regression. Covariates were
selected from filtered temperature (several lagged and
averaged values), indicator variables for hot and cold days
and day-of-week using stepwise procedure. The average of
current and previous days' pollution levels were used.
Total (non-accidental), ischemic heart disease, pneumonia,
and COPD (mean daily total deaths for the six cities: 59,
12, 55, 3, 11, and 3, respectively in the order shown left). A
factor analysis was conducted on the 15 elements in the fine
fraction of dichot samplers to obtain five common factors;
factors were rotated to maximize the projection of the single
"tracer" element (as in part identified from the past studies
conducted on these data) for each factor; PM2 5 was
regressed on the identified factors scores so that the factor
scores could be expressed in the mass scale. Using GAM
Poisson models adjusting for temperature, humidity, day-of-
week, season, and time, mortality was regressed on the
factor scores in the mass scale. The mean of the same-day
and previous day (increasing the sample size from 6,211 to
9,108 days) mass values were used. The city-specific
regression coefficients were combined using inverse
variance weights.
Out-of-hospital deaths (total, respiratory, COPD, ischemic
heart disease, heart failure, sudden cardiac death screening
codes, and stroke) were related to PM10, nephelometer (0.2 -
1.0 /an fine particles), SO2, and CO, adjusting for day-of-
week, month of the year, temperature and dewpoint, using
Poisson regression.
Significant associations were found for a wide variety gaseous
and parti culate pollutants, especially for peak O3. No systematic
differences were seen according particle size or chemistry.
Mortality for one part of the metropolitan area could be
associated with air quality from another, not necessarily
neighboring part.
Three sources of fine particles were defined in all six cities with
a representative element for each source type: Si for soil and
crustal material; Pb for motor vehicle exhaust; and Se for coal
combustion sources. In city-specific analysis, additional sources
(V for fuel oil combustion, Cl for salt, etc.) were considered.
Five source factors were considered for each city, except Topeka
with the three sources. Coal and mobile sources account for the
majority of fine particles in each city. In all of the metropolitan
areas combined, 46% of the total fine particle mass was
attributed to coal combustion and 19% to mobile sources. The
strongest increase in daily mortality was associated with the
mobile source factor. The coal combustion factor was positively
associated with mortality in all metropolitan areas, with the
exception of Topeka. The crustal factor from the fine particles
was not associated with mortality.
Nephelometer data were not associated with mortality. Cause-
specific death analyses suggest PM associations with ischemic
heart disease deaths. Associations of mortality with SO2 and CO
not mentioned. Mean daily death counts were small (e.g., 7.7 for
total; 1.6 for ischemic heart disease). This is an apparently
preliminary analysis.
The fractional Philadelphia mortality
risk attributed to the pollutant levels:
"average risk" was 0.0423 for 25 //g/m3
PM25; 0.0517 for 25 ,ug/m3 PM10.25;
0.0609 for 50 //g/m3 PM10, using the
Harvard PM indices at avg. of 0 and 1 d
lags-
Total mortality percent excess overall:
4.0 (2.8, 5.3), 2.7 (0.6, 5.0) with each
25 /ig/m3 increase in the two-day mean
of coal combustion fine PM factor; 8.7
(4.2, 13.4) with each 25 //g/m3 increase
in the two-day mean of mobile source
fine PM factor; -5.7 (-13.7, 3.2) with
each 25 ,ug/m3 increase in the two-day
mean of the crustal source fine PM
factor.
Total mortality percent excess:
5.6% (-2.4, 1.43) per 50 //g/m3 PM10 at
avg. of 2 to 4 d lag; 7.2% (-6.3, 22.8)
with SO2 CO. 1.8% (-3.5, 7.3) per
25 Mg/m3 PMj; -1.0 (-8.7,. 7.7) with
SO, and CO.
o
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Mar et al. (2000).
Phoenix, AZ. 1995-1997.
PM10, PM25, andPM10.25
(TEOM), with means = 46.5,
13.0, and 33.5, respectively;
and PM2 5 (DFPSS),
mean= 12.0.
Clyde et al. (2000).
Phoenix, AZ. 1995-1998.
PM10, and PM25, (from
TEOM), with means = 45.4,
and 13.8. PM10_25 computed
as PM10-PM25.
Total (non-accidental) and cardiovascular deaths (mean =
8.6 and 3.9, respectively) for only those who resided in the
zip codes located near the air pollution monitor were
included. GAM Poisson models were used, adjusting for
season, temperature, and relative humidity. Air pollution
variables evaluated included: O3, SO2, NO2, CO, TEOM
PM10, TEOM PM25, TEOM PM10.25, DFPSS PM25, S, Zn,
Pb, soil, soil-corrected K (KS), nonsoil PM, OC, EC, and
TC. Lags 0 to 4 days evaluated. Factor analysis also
conducted on chemical components of DFPSS PM25 (Al, Si,
S, Ca, Fe, Zn, Mn, Pb, Br, KS, OC, and EC); and factor
scores included in mortality regression.
Elderly (age > 65 years) non-accidental mortality for three
regions of increasing size in Phoenix urban area analyzed to
evaluate influence of spatial uniformity of PM10 and PM25.
All-age accidental deaths for the metropolitan area also
examined as a "control". GAM Poisson models adjusting
for season (smoothing splines of days), temperature, specific
humidity, and lags 0- to 3-d of weather variables. PM
indices for lags 0-3 d considered. Bayesian Model
Averaging (BMA) produces posterior mean relative risks by
weighting each model (out of all possible model
specifications examined) based on support received from the
data.
Total mortality was significantly associated with CO and NO2
and weakly associated with SO2, PM10, PM10_2 5, and EC.
Cardiovascular mortality was significantly associated with CO,
NO2, SO2, PM25, PM10, PM10.25, OC and EC. Combustion-
related factors and secondary aerosol factors were also associated
with cardiovascular mortality. Soil-related factors, as well
as individual variables that are associated with soil were
negatively associated with total mortality.
The BMA results suggest that a weak association was found only
for the mortality variable defined over the region with uniform
PM2 5, with a 0.91 probability that RR is greater than 1. The
other elderly mortality variables, including the accidental deaths
("control"), had such probabilities in the range between 0.46 to
0.77. Within the results for the mortality defined over the region
with uniform PM2 5, the results suggested that effect was
primarily due to coarse particles rather than fine; only the lag 1
coarse PM was consistently included in the highly ranked
models.
Total mortality percent excess: 5.4 (0.1,
11.1) for PM10 (TEOM) 50 ,ug/m3 at lag
0 d; 3.0 (-0.5, 6.6) for PM10.25 (TEOM)
25 Mg/m3 at lag 0 d; 3.0 (-0.7, 6.9) for
PM2 5 (TEOM) 25 Mg/m3 at lag 0 d.
Cardiovascular mortality RRs: 9.9(1.9,
18.4) for PM10 (TEOM) 50 ^g/m3 at lag
0 d; 18.7 (5.7, 33.2) for PM25 (TEOM)
25 |/g/m3 at lag 1 d; and 6.4 (1.4, 11.7)
PM10 (TEOM) 25 |/g/m3 PM10.2 5 at lag 0
d.
Posterior mean RRs and 90%
probability intervals per changes of
25 |/g/m3 in all lags of fine and coarse
PM for elderly mortality for uniform
PM10 region: 1.06(1+, 1.11).
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
to
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Smith et al. (2000).
Phoenix, AZ.
1995-1997
Tsai et al. (2000).
Newark, Elizabeth, and
Camden, NJ. 1981-1983.
PM15: 55.5, 47.0, 47.5; and
PM25:42.1, 37.1, 39.9, for
Newark, Elizabeth, and
Camden, respectively.
Study evaluated effects of daily and 2- to 5-day average
coarse (PM10_2 5) and fine (PM2 5) particles from an
EPA-operated central monitoring site on nonaccidental
mortality among elderly (65+ years), using time-series
analyses for residents within city of Phoenix and, separately,
for region of circa 50 mi around Phoenix. Initial model
selected to represent long-term trends and weather variables
(e.g., ave. daily temp., max daily temp., daily mean specific
humidity, etc.); then PM variables added to model one at a
time to ascertain which had strongest effect. Piecewise
linear analysis and spline analysis used to evaluate possible
nonlinear PM-mortality relationship and to evaluate
threshold possibilities. Data analyzed most likely same as
Clyde's or Mar's Phoenix data.
Factor analysis-derived source type components were
examined for their associations with mortality in this study.
Non-accidental total deaths and cardiorespiratory deaths
were examined for their associations with PM15, PM2 5
sulfate, trace metals from PM15, three fractions of
extractable organic matter, and CO. Data were analyzed
with Poisson GEE regression models with autoregressive
correlation structure, adjusting for temperature, time-of-
week, and season indicator variables. Individual pollution
lag days from 0 to 3, as well as the average concentrations
of current and preceding 3 days were considered. Factor
analysis of the trace elements, sulfate, and CO data was
conducted, and mortality series were regressed on these
factor scores.
In linear PM effect model, a statistically significant mortality
association found with PM10_2 5, but not with PM2 5. In the model
allowing for a threshold, evidence suggestive of possible
threshold for PM2 5 (in the range of 20-25 ,ug/m3) found, but not
for PM10_25. A seasonal interaction in the PM10_25 effect was also
reported: the effect being highest in spring and summer when
anthropogenic concentration of PM10_25 is lowest.
Factor analysis identified several source types with tracer
elements. In Newark, oil burning factor, industrial source factor,
and sulfate factor were positively associated with total mortality;
and sulfate was associated with cardio-respiratory mortality. In
Camden, oil burning and motor vehicle factors were positively
associated with total mortality; and, oil burning, motor vehicles,
and sulfate were associated with cardio-respiratory mortality. In
Elizabeth, resuspended dust was not associated with total
mortality; and industrial source (traced by Cd) showed positive
associations with cardio-respiratory mortality. On the mass basis
(source-contributed mass), the RRs estimates per 10 i/g/m3 were
larger for specific sources (e.g., oil burning, industry, etc.) than
for total mass. The choice of lag/averaging reported to be not
important.
Percent excess deaths per 50 ,ug/m3
increase in current day PM15: in
Newark, 5.7 (4.6, 6.7) for total
mortality, 7.8 (3.6, 12.1) for cardioresp.
mortality; in Camden, 11.1 (0.7, 22.5)
and 15.0 (4.3, 26.9); and in Elizabeth,
-4.9 (-17.9, 10.9) and 3.0 (-11.0,
19.4), respectively. Percent excess
deaths per 25 ,ug/m3 PM25; in Newark,
4.3 (2.8, 5.9) for total and 5.1 (3.1, 7.2)
for cardiorespiratory mortality; in
Camden, 5.7(0.1, 11.5) and 6.2 (0.6,
12.1); in Elizabeth, 1.8 (-5.4, 9.5) and
2.3 (-5.0, 10.1), respectively.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Gamble (1998).
Dallas, TX. 1990-1994.
PM10(25)
Ostro (1995).
San Bernardino and Riverside
Counties, CA, 1980-1986.
PM2 5 (estimated from visual
range). Mean = 32.5.
Kelsalletal. (1997).
Philadelphia, PA
1974-1988.
TSP (67)
Moolgavkar and Luebeck
(1996). Philadelphia, PA.
1973-1988. TSP (68)
Relationships of total, respiratory, cardiovascular, cancer,
and remaining non-accidental deaths to PM10, O3, NO2, SO2,
and CO evaluated, adjusting for temperature, dewpoint, day-
of-week, and seasonal cycles (trigonometric terms) using
Poisson regression.
Study evaluated total, respiratory, cardiovascular, and age >
= 65 deaths (mean = 40.7, 3.8, 18.7, and 36.4 per day,
respectively). PM2 5 estimated based on airport visual range
and previously published empirical formula. Autoregressive
OLS (for total) and Poisson (for sub-categories) regressions
used, adjusting for season (sine/cosine with cycles from 1 yr
to 0.75 mo; prefiltering with 15-day moving ave.;
dichotomous variables for each year and month; smooth
function of day and temp.), day-of-week, temp, and
dewpoint. Evaluated lags 0, 1, and 2 of estimated PM25, as
well as moving averages of 2, 3, and 4 days and O3.
Total, cardiovascular, respiratory, and by-age mortality
regressed on TSP, SO2, NO2, O3, and CO, adjusting for
temporal trends and weather, using Poisson GAM model.
A critical review paper, with an analysis of total daily
mortality for its association with TSP, SO2, NO2, and O3,
adjusting for temporal trends, temperature, and also
conducting analysis by season, using Poisson GAM model.
O3 (avg. of 1-2 day lags), NO2 (avg.. 4 -5 day lags), and CO
(avg. of lags 5- 6 days) were significantly positively associated
with total mortality. PM10 and SO2 were not significantly
associated with any deaths.
The results were dependent on season. No PM2 5 - mortality
association found for the full year-round period. Associations
between estimated PM2 5 (same-day) and total and respiratory
deaths found during summer quarters (April - Sept.).
Correlation between the estimated PM2 5 and daily max temp.
was low (r = 0.08) during the summer quarters. Ozone was also
associated with mortality, but was also relatively highly
correlated with temp, r = 0.73). Moving averages of PM25 did
not improve the associations.
TSP, SO2, O3, and 1-day lagged CO individually showed
statistically significant associations with total mortality. No NO2
associations unless SO2 or TSP was also considered. The effects
of TSP and SO2 were diminished when both pollutants were
included.
RR results presented as figures, and seasonal difference noted.
TSP, SO2, O3 - mortality associations varied across season. TSP
associations were stronger in summer and fall. NO2 was the most
significant predictor.
-3.6% (-12.7, 6.6) per 50 //g/m3 PM10
at 0 lag (other lags also reported to have
no associations)
Percent excess deaths per 25 ,ug/m3 of
estimated PM25, lag 0: Full year: 0.3
(-0.6, 1.2) for total; 2.1 (-0.3, 4.5) for
respiratory; and 0.7 (-0.3, 1.7) for
circulatory. Summer quarters: 1.6
(0.03, 3.2) for total; 5.5 (1.1, 10.0) for
respiratory; and 0 (-1.0, 1.0) for
circulatory.
Total mortality excess risk: 3.2% (0,
6.1) per 100 ,ug/m3 TSP at 0 day lag.
Total mortality excess risk: ranged ~ 0
(winter) to =4% (summer) per
100 Mg/m3 TSP at 1 day lag.
o
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Neas et. al. (1999).
Philadelphia. 1973-1980.
TSP mean = 77.2.
Schwartz (2000d).
Philadelphia. 1974-1988.
TSP. Mean = 70//g/m3 for
warm season (April through
August) and 64 //g/m3 for
cold season.
Levy et al. (2000).
Years vary from study to study
ranging between 1973 to 1994.
21 published studies included
U.S., Canadian, Mexican,
European, Australian, and
Chilean cities. PM10 levels in
the 19 U.S. cities (in some
cases TSP were converted to
PM10 using factor of 0.55)
ranged from -20 to -60 ug/m3.
Total, age over 65, cancer, and cardiovascular deaths
analyzed for association with TSP. Conditional logistic
regression analysis with case-crossover design conducted.
Average values of current and previous days' TSP used.
Case period is the 48-hr period ending at midnight on day of
death. Control periods are 7, 14, and 21 days before and
after the case period. Other covariates included temperature
on the previous day, dewpoint on the same day, an indicator
for hot days (> SOT), an indicator for humid days
(dewpoint > 66°F), and interaction of same-day temp, and
winter season.
Total (non-accidental) deaths analyzed. GAM Poisson
models adjusting for temperature, dewpoint, day-of-week,
and season applied to each of 15 warm and cold seasons.
Humidity-corrected extinction coefficient, derived from
airport visual range, also considered as explanatory variable.
In the second stage, resulting 30 coefficients were regressed
on regression coefficients of TSP on SO2. Results of first
stage analysis combined using inverse variance weighting.
To determine whether across-study heterogeneity of PM
effects could be explained by regional parameters, Levy
et al. applied an empirical Bayes meta-analysis to 29 PM
estimates from 21 published studies. They considered such
city-specific variables as mortality rate, gaseous pollutants^
regression coefficients, PM10 levels, central air conditioning
prevalence, heating and cooling degree days.
In each set of the six control periods, TSP was associated with
total mortality. A model with four symmetric reference periods 7
and 14 days around the case period produced a similar result. A
model with only two symmetric reference periods of 7 days
around the case produced a larger estimate. A larger effect was
seen for deaths in persons > 65 years of age and for deaths due
to pneumonia and to cardiovascular disease. Cancer mortality
was not associated with TSP.
When TSP controlled for, no significant association between SO2
and daily deaths. SO2 had no association with daily mortality
when it was poorly correlated with TSP. In contrast, when SO2
was controlled for, TSP was more strongly associated with
mortality than when it was less correlated with SO2. However,
all of the association between TSP and mortality was explained
by its correlation with extinction coefficient.
Among the city-specific variables, PM2 5/PM10 ratio was a
significant predictor (larger PM estimates for higher PM2 5/PM10
ratios) in the 19 U.S. cities data subsets. While the sulfate data
were not available for all the 19 cities, the investigators noted
that, based on their analysis of the limited data with sulfate for
10 estimates, the sulfate/PMlO ratio was highly correlated with
both the mortality (r = 0.84) and with the PM25/PM10 ratio (r =
0.70). This indicates that the sulfate/PM10 ratio may be even
better predictor of regional heterogeneity of PM RR estimates.
Odds Ratio (OR) for all cause mortality
per 100 /ig/m3 increase in 48-hr mean
TSP was 1.056(1.027, 1.086). The
corresponding number for those aged 65
and over was 1.074(1.037, 1.111), and
1.063 (1.021, 1.107) for cardiovascular
disease.
Total mortality excess risk estimates
combined across seasons/years: 9.0
(5.7, 12.5) per 100//g/m3 TSP.
The pooled estimate froml9 U.S. cities
was 0.70% (0.54, 0.84) per 10 ug/m3
increase in PM10.
o
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
fe
H
6
o
o
H
O
O
H
W
O
Canada
Burnett etal. (1998a).
11 Canadian cities.
1980-1991.
No PM index data available
on consistent daily basis.
Burnett et al. (2000).
8 largest Canadian cities.
1986-1996. All city mean
PM1025.9;PM25 13.3;
PM10.25 12.6; sulfate 2.6.
Burnett etal. (1998b).
Toronto, 1980-1994.
TSP(60);COH(0.42);
SO4= (9.2 Mg/m3);
PM10 (30, estimated);
PM25(18, estimated)
Total non-accidental deaths were linked to gaseous air
pollutants (NO2, O3, SO2, and CO) using GAM Poisson
models adjusting for seasonal cycles, day-of-week, and
weather (selected from spline-smoothed functions of
temperature, dewpoint, relative humidity with 0, 1, and 2
day lags using forward stepwise procedure). Pollution
variables evaluated at 0, 1, 2, and up to 3-day lag averages
thereof. No PM index included in analyses because daily
PM measurements not available. City-specific models
containing all four gaseous pollutants examined. Overall
risks computed by averaging risks across cities.
Total non-accidental deaths linked to PM indices (PM10,
PM25, PM10_25, sulfate, 47 elemental component
concentrations for fine and coarse fractions) and gaseous air
pollutants (NO2, O3, SO2, and CO). Each city's mortality,
pollution, and weather variables separately filtered for
seasonal trends and day-of-week patterns. The residual
series from all the cities then analyzed in a GAM Poisson
model. The weather model was selected from spline-
smoothed functions of temperature, relative humidity, and
maximum change in barometric pressure within a day, with
0 and 1 day lags using forward stepwise procedure.
Pollution effects were examined at lags 0 through 5 days.
To avoid unstable parameter estimates in multi-pollutant
models, principal components were also used as predictors
in the regression models.
Total, cardiac, and other nonaccidental deaths (and by age
groups) were regressed on TSP, COH, SO4=, CO, NO2, SO2,
O3, estimated PM10 and PM2 5 (based on the relationship
between the existing every-6th-day data and SO4=, TSP and
COH), adjusting for seasonal cycles, day-of-week,
temperature, and dewpoint using Poisson GAM model.
NO2 had 4.1% increased risk per mean concentration; O3 had
1.8%; SO2 had 1.4%, and CO had 0.9% in multiple pollutant
regression models. A 0.4% reduction in excess mortality was
attributed to achieving a sulfur content of gasoline of 30 ppm in
five Canadian cities. Daily PM data for fine and coarse mass
and sulfates available on varying (not daily) schedules allowed
ecologic comparison of gaseous pollutant risks by mean fine
particle indicators mass concentrations.
O3 was weakly correlated with other pollutants and other
pollutants were "moderately" correlated with each other (the
highest was r = 0.65 for NO2 and CO). The strongest association
with mortality for all pollutants considered were for 0 or 1 day
lags. PM25 was a stronger predictor of mortality than PM10_25.
The estimated gaseous pollutant effects were generally reduced
by inclusion of PM2 5 or PM10, but not PM10_2 5. Sulfate, Fe, Ni,
and Zn were most strongly associated with mortality. Total
effect of these four components was greater than that for PM2 5
mass alone.
Essentially all pollutants were significant predictors of total
deaths in single pollutant models, but in two pollutant models
with CO, most pollutants' estimated RRs reduced (all PM
indices remained significant). Based on results from the co-
pollutant models and various stepwise regressions, authors noted
that effects of the complex mixture of air pollutants could be
almost completely explained by the levels of CO and TSP.
Found suggestion of weak negative
confounding of NO2 and SO2 effects
with fine particles and weak positive
confounding of particle effects with O3.
No quantitative RR or ER estimates
reported for PM indicators.
Percentage increase in daily filtered
non-accidental deaths associated with
increases of 50 ,ug/m3 PM10 and
25 |/g/m3 PM2 5 or PM10.2 5 at lag 1 day:
3.5 (1.0, 6.0) for PM10; 3^0 (1.1, 5.0) for
PM25; and 1.8 (-0.7, 4.4) for PM10.25.
In the multiple pollutant model with
PM2 5, PM10_2 5, and the 4 gaseous
pollutants, 1.9 (0.6, 3.2) for PM25; and
1.2 (-1.3, 3.8)forPM10.25.
Total mortality percent excess: 2.3%
(0.8, 3.8) per 100 ,ug/m3 TSP; 3.5%
(1..8, 5.3) per 50 ,ug/m3 PM10; 4.8%
(3.3, 6.4) per 25 ^g/m3 PM25. 0 day lag
for TSP and PM10; Avg. of 0 and 1 day
forPM,,.
o
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
fe
H
6
o
o
H
O
O
H
W
O
O
Canada (cont'd)
Goldberg et al. (2000)
Montreal, Quebec
1984-95 Mean
TSP=53.1
(14.6-211.1)Mg/m3
PM10 = 32.2
(6.5- 120.5),ug/m3
PM25 = 3.3 (0.0-30.0) ,ug/m3
Goldberg etal. (2001).
Montreal, Quebec.
1984-1993. Predicted PM2
mean = 17.6. CoH
(1000ft)) mean = 0.24,
sulfate mean =3.3.
Goldberg etal. (2001).
Data same as above.
Ozkaynaketal. (1996).
Toronto, 1970-1991.
TSP (80); COH (0.42
/1000ft).
Study aimed to shed light on population subgroups that my
be susceptible to PM effects. Linked data on daily deaths
with other health data (physician visits, pharmaceutical R,,,
etc.) to identify individuals with presenting health
conditions. PM10 and PM2 5 measured by dichotomous
sampler 1 in 6 days until 1992 (2 stations), then daily
through 1993. PM missing days interpolated from COH,
ext. coefficient, sulfates. Used quasi likelihood estimation
in GAM's to assess PM associations with total and cause-
specific mortality; and, also, in subgroups by age and/or
preexisting health conditions. Adjusted for CO, NO2, NO,
O3 and SO2 in 2-pollutant and all-pollutant models.
The investigators used the universal Quebec medicare
system to obtain disease conditions prior to deaths,
and the roles of these respiratory and cardiovascular
conditions in the PM-mortality associations were
examined. GAM Poisson model adjusting for
temporal pattern and weather was used.
Cause-specific mortality (non-accidental, neoplasm,
lung cancer, cardiovascular, coronary artery disease,
diabetes, renal disease, and respiratory) series were
examined for their associations with O3, using GAM
Poisson model adjusting for temporal pattern and
weather. Results were also reported for models with
adjustments for other pollutants (SO2, CO, NO2, CoH,
etc.).
Total, cardiovascular, COPD, pneumonia, respiratory,
cancer, and the remaining mortality series were
related to TSP, SO2, COH, NO2, O3, and CO,
adjusting for seasonal cycles (by high-pass filtering
each series) temperature, humidity, day-of-week,
using OLS regression. Factor analysis of multiple
pollutants was also conducted to extract automobile
related pollution, and mortality series were regressed
on the resulting automobile factor scores.
Significant associations found for all-cause (total non-
accidental) and cause-specific (cancer, CAD, respiratory disease,
diabetes) with PM measures. Results reported for PM25, COH
and sulfates. All three PM measures associated with increases in
total, resp., and "other nonaccidental", and diabetes-related
mortality. No PM associations found with digestive, accidental,
renal or neurologic causes of death. Also, mainly in 65+ yr
group, found consistent associations with increased total
mortality among persons who had cancer, acute lower resp.
diseases, any cardiovascular disease, chronic CAD and
congestive heart failure (CHF).
The PM-mortality associations were found for those who
had acute lower respiratory diseases, chronic coronary
diseases, and congestive heart failure. They did not find
PM-mortality associations for those chronic upper
respiratory diseases, airways disease, cerebrovascular
diseases, acute coronary artery diseases, and
hypertension. Adjusting for gaseous pollutants generally
attenuated PM RR estimates, but the general pattern
remained. Effects were larger in summer.
The effect of O3 was generally higher in the warm season
and among persons aged 65 years and over. O3 showed
positive and statistically significant associations with
non-accidental cause, neoplasms, cardiovascular disease,
and coronary artery disease. These associations were not
reduced when the model adjusted for SO2, CO, NO2,
CoH. simultaneously (or when CoH was replaced with
PM25 or total sulfates).
TSP (0 day lag) was significantly associated with total
and cardiovascular deaths. NO2 (0-day lag) was a
significant predictor for respiratory and COPD deaths. 2-
day lagged O3 was associated with total, respiratory, and
pneumonia deaths. Factor analysis showed factor with
high loadings for NO2, COH, and CO (apparently
representing automobile factor) as significant predictor
for total, cancer, cardiovascular, respiratory, and
pneumonia deaths.
Percent excess mortality per 25 //g/m3
estimated PM25:
Total deaths (3 d ave.) = 4.4% (2.5, 6.3)
CV deaths (3 d ave.) = 2.6% (-0.1, 5.5)
Resp deaths (3 d ave.) = 16.0% (9.7,
22.8)
Coronary artery (3 d ave.) = 3.4%
(-0.2,7.1)
Diabetes (3 d ave.) = 15.7% (4.8, 27.9)
Lower Resp Disease (3 d ave.) = 9.7%
(4.5, 15.1)
Airways disease (3 d ave.) = 2.7%
(-0.9, 6.4)
CHF (3 d ave.) = 8.2% (3.3, 13.4)
The percent excess deaths estimates
for non-accidental deaths per IQR
(average of 0-2 day lags) for CoH,
predicted PM25, and sulfate were:
1.98% (1.07, 2.90), 2.17% (1.26,
3.08), and 1.29% (0.68, 1.90),
respectively.
PM RRs not reported.
Total mortality excess risk: 2.8%
per 100 Mg/m3 TSP at 0 day lag.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
IQRin,ug/m3.
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Katsouyanni et al. (1997).
12 European (APHEA) cities.
1975-1992 (study years
different from city to city).
Median Black Smoke (BS)
levels ranged from 13 in
London to 73 in Athens
and Kracow.
Samolietal. (2001).
APHEA 1 cities (see
Katsouyanni (1997)). At least
five years between 1980-1992.
The PM levels are the same as
those in Katsouyanni et al.
(1997).
Katsouyanni et al. (2001).
1990-1997 (variable from city
to city). Median PM10 ranged
from 14 (Stockholm) to 66
(Prague). Median BS ranged
from 10 (Dublin) to 64
(Athens).
Touloumi et al. (1997).
6 European (APHEA) cities.
1977-1992 (study years
different from city to city).
Median Black Smoke (BS)
levels ranged from 14.6 in
London to 84.4 in Athens.
Total daily deaths regressed on BS or SO2 using Poisson
models, adjusting for seasonal cycles, day-of-week,
influenza epidemic, holidays, temp., humidity. Final
analysis done with autoregressive Poisson models to allow
for overdispersion and autocorrelation. Pollution effects
examined at 0 through 3 day lags and multi-day averages
thereof. When city-specific coefficients tested to be
homogeneous, overall estimates obtained by computing
variance-weighted means of city-specific estimates (fixed
effects model). When significant heterogeneity present,
source of heterogeneity sought by examining a predefined
list of city-specific variables, including annual and seasonal
means of pollution and weather variables, number of
monitoring sites, correlation between measurements from
different sites, age-standardized mortality, proportion of
elderly people, smoking prevalence, and geographic
difference (north-south, east-west). A random effects model
was fit when heterogeneity could not be explained.
In order to further investigate the source of the regional
heterogeneity of PM effects, and to examine the sensitivity
of the RRs, the APHEA data were reanalyzed by the
APHEA investigators themselves (Samoli et al., 2001).
Unlike previous model in which sinusoidal terms for
seasonal control and polynomial terms for weather, the
investigators this time used a GAM model with smoothing
terms for seasonal trend and weather, which is more
commonly used approach in recent years.
The 2nd phase of APHEA (APHEA 2) put emphasis on the
effect modification by city-specific factors. The first stage
of city specific regressions used GAM Poisson model. The
second stage regression analysis was conducted to explain
any heterogeneity of air pollution effects using city-specific
variables. These city-specific variables included average air
pollution levels, average temperature/humidity, age-
standardize mortality rate, region indicators, etc.
Results of the short-term effects of ambient NO2 and/or O3
on daily deaths from all causes (excluding accidents) were
discussed to provide a basis for comparison with estimated
SO2 or BS effects in APHEA cities. Poisson models,
lag/averaging of pollution, and the computation of
combined effects across the cities were done in the same
way as done by Katsouyanni et al. (1997), as above.
Substantial variation in pollution levels (winter mean SO2 ranged
from 30 to 330 ,ug/m3), climate, and seasonal patterns were
observed across cities. Significant heterogeneity was found for
the effects of BS and SO2, but only the separation between
western and central eastern European cities resulted in more
homogeneous subgroups. Significant heterogeneity for SO2
remained in western cities. Cumulative effects of prolonged
(two to four days) exposure to air pollutants resulted in estimates
comparable with the one day effects. The effects of both SO2
and BS were stronger during the summer and were independent.
T he estimated relative risks for central-eastern cities were larger
than those obtained from the previous model. Also, restricting
the analysis to days with concentration < 150ug/m3 further
reduced the differences between the western and central-eastern
European cities. The authors concluded that part of the
heterogeneity in the estimated air pollution effects between
western and central eastern cities in previous publications was
caused by the statistical approach and the data range.
The authors found several effect modifiers. The cities with
higher NO2 levels showed larger PM effects. The cities with
warmer climate showed larger PM effects. The cities with low
standardized mortality rate showed larger PM effects. The
combined estimate of mortality RRs per 10ug/m3 PM10 or BS
was: 0.6% (0.4, 0.8). The PM RR estimates for cities with low
vs. high NO2 levels were 0.19% (0, 0.41) and 0.80% (0.67,
0.93); 0.29% (0.16, 0.42) for cities with cold climate and 0.82%
(0.69, 0.96) for warm climate, respectively.
Significant positive associations found between daily deaths and
both NO2 and O3. Tendency for larger effects of NO2 in cities
with higher levels of BS. When BS included in the model,
pooled estimate for O3 effect only slightly reduced, but
coefficient for NO2 reduced by half. Authors speculated that
short-term effects of NO2 on mortality confounded by other
vehicle-derived pollutants.
Total mortality excess deaths per
25 |/g/m3 increase in single day BS for
western European cities: 1.4(1.0, 1.8);
and 2 (1, 3) per 50 ,ug/m3 PM10
increase. In central/eastern Europe
cities, corresponding figure was 0.3
(0.05, 0.5) per 25 ^g/m3 BS.
Total mortality RRs per 50ug/m3 BS for
all cities, western cities, and central-
eastern cities using the GAM approach
were: 2.2% (1.8, 2.6); 3.1% (2.4, 3.9);
and, 2.2% (1.4, 2.3), respectively. In
contrast, those with old method were:
1.3% (0.9, 1.7); 2.9% (2.1, 3.7); and,
0.6% (0.1, 1.1), respectively.
NO2 and/or O3 estimates only.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
OO
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Zmirouetal. (1998).
10 European (APHEA) cities.
1977-1992 (study years
different from city to city).
Median Black Smoke (BS)
levels ranged from 13 in
London to 73 in Kracow.
Bremneretal. (1999).
London, UK, 1992-1994.
BS (13), PM10 (29).
Prescottetal. (1998).
Edinburgh, UK, 1981-1995.
PM10 (21, by TEOM only for
1992-1995); BS (8.7).
Rooneyetal. (1998).
England and Wales, and
Greater London, UK
PM10 (56, during the worst heat
wave; 39, July-August mean)
Wordleyetal. (1997).
Birmingham, UK,
1992-1994.
PM10 (apparently
beta-attenuation, 26)
Cardiovascular, respiratory, and digestive mortality series
in 10 European cities analyzed to examine cause-specificity
of air pollution. The mortality series were analyzed for
associations with PM (BS, except TSP in Milan and
Bratislava; PM13 in Lyon), NO2, O3, and SO2. Poisson
models, lag/averaging of pollution, and computation of
combined effects across the cities done in the same way as
by Katsouyanni et al. (1997), above.
Total, cardiovascular, and respiratory (by age) mortality
series were regressed on PM10, BS, O3, NO2, CO, and SO2,
adjusting for seasonal cycles, day-of-week, influenza,
holidays, temperature, humidity, and autocorrelation using
Poisson model.
Both mortality (total, cardiovascular, and respiratory) and
emergency hospital admissions (cardiovascular and
respiratory), in two age groups (<65 and >= 65), were
analyzed for their associations with PM10, BS, SO2, NO2, O3,
and CO, using Poisson regression adjusting for seasonal
cycles, day-of-week, temperature, and wind speed.
Excess deaths, by age, sex, and cause, during the 1995 heat
wave were estimated by taking the difference between the
deaths during heat wave and the 31-day moving averages
(for 1995 and 1993-94 separately). The pollution effects,
additively for O3, PM10, and NO2, were estimated based on
the published season-specific coefficients from the 1987-
1992 study (Anderson et al., 1996).
Mortality data were analyzed for COPD, pneumonia, all
respiratory diseases, all circulatory diseases, and all causes.
Mortality associations with PM10, NO2, SO2, and O3 were
examined using OLS (with some health outcomes log- or
square-root transformed), adjusting for day-of-week, month,
linear trend, temperature and relative humidity. The study
also analyzed hospital admission data.
The cardiovascular and respiratory mortality series were
associated with BS and SO2 in western European cities, but not
in the five central European cities. NO2 did not show consistent
mortality associations. RRs for respiratory causes were at least
equal to, or greater than those for cardiovascular causes. No
pollutant exhibited any association with digestive mortality.
Pooled cardiovascular mortality percent
excess deaths per 25 ,ug/m3 increase in
BS for western European cities: 1.0
(0.3, 1.7); for respiratory mortality, it
was 2.0 (0.8, 3.2) in single lag day
models (the lags apparently varied
across cities).
All effect size estimates (except O3) were positive for total deaths 1.9% (0.0, 3.8) per 25 ^g/m3 BS at lag
(though not significant for single lag models). The effects of O3
found in 1987-1992 were not replicated, except in
cardiovascular deaths. Multiple day averaging (e.g., 0-1, 0-2
days) tend to give more significant effect size estimates. The
effect size for PM10 and BS were similar for the same
distributional increment.
Among all the pollutants, BS was most significantly associated
with all cause, cardiovascular, and respiratory mortality series.
In the subset in which PM10 data were available, the RR
estimates for BS and PM10 for all cause elderly mortality were
comparable. Other pollutants' mortality associations were
generally inconsistent.
Air pollution levels at all the locations rose during the heat wave.
8.9% and 16.1% excess deaths were estimated for England and
Wales, and Greater London, respectively. Of these excess
deaths, up to 62% and 38%, respectively for these locations, may
be attributable to combined pollution effects.
1 day; 1.3% (-1.0, 3.6) per 50 ^g/
PM10 at lag 1 d for total deaths. Resp.
deaths (3 d) = 4.9% (0.5, 9.4). CVD
deaths (1 d) = 3.0%(0.3, 5.7).
3.8 (1.3, 6.4) per 25 i/g/m3 increase in
BS for all cause mortality in age 65+
group, avg. of 1-3 day lags.
2.6% increase for PM10 in Greater
London during heat wave.
Total, circulatory, and COPD deaths were significantly
associated with 1-day lag PM10. The gaseous pollutants "did not
have significant associations independent from that of PM10",
and the results for gaseous pollutants were not presented. The
impact of reducing PM10 to below 70 ,ug/m3 was estimated to be
"small" (0.2% for total deaths), but the PM10 level above 70
,ug/m3 occurred only once during the study period.
5.6% (0.5, 11.0) per 50 ^g/m3 PM10 at 1
d lag for total deaths. COPD (1 d lag)
deaths = 27.6 (5.1,54.9).
Circulatory (1 d) deaths = 8.8 (1.9,
17.1)
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
vo
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Hoek et al. (2000).
The Netherlands, 1986-1994.
PM10 (median 34);
BS (median 10).
Hoek etal. (2001).
The Netherlands. 1986-1994.
PM10 (median 34);
BS (median 10).
Ponkaetal. (1998).
Helsinki, Finland, 1987-1993.
TSP (median 64);
PM10 (median 28)
Peters etal. (1999a).
A highly polluted coal basin
area in the Czech Republic and
a rural area in Germany,
northeast Bavaria districts.
1982-1994. TSP: mean = 121.1
and 51.6, respectively, for these
two regions. PM10 and PM2 5
were also measured in the coal
basin during 1993-1994 (mean
= 65.9 and 51.0, respectively).
Total, cardiovascular, COPD, and pneumonia mortality
series were regressed on PM10, BS, sulfate, nitrate, O3, SO2,
CO, adjusting for seasonal cycles, day-of-week, influenza,
temperature, and humidity using Poisson GAM model.
Deaths occurring inside and outside hospitals were also
examined.
This study of the whole population of the Netherlands, with
its large sample size (mean daily total deaths ~ 330, allowed
examination of specific cardiovascular cause of deaths.
GAM Poisson regression models, adjusting for seasonal
cycles, temperature, humidity, day-of-week was used.
Total and cardiovascular deaths, for age groups < 65 and 65
+, were related to PM10, TSP, SO2, NO2, and O3, using
Poisson model adjusting for temperature, relative humidity,
day-of-week, temporal patterns, holiday and influenza
epidemics.
Non-accidental total and cardiovascular deaths (mean =
18.2 and 12.0 per day, for the Czech and Bavaria areas,
respectively). The APHEA approach (Poisson model with
sine/cosine, temperature as a quadratic function, relative
humidity, influenza, day-of-week as covariates), as well as
GAM Poisson models were considered. Logarithm of TSP,
SO2, NO2, O3, and CO (and PM10 and PM25 for 1993-1994)
were examined at lags 0 through 3 days.
Particulate air pollution was not more consistently associated
with mortality than were the gaseous pollutants SO2 and NO2.
Sulfate, nitrate, and BS were more consistently associated with
total mortality than was PM10. The RRs for all pollutants were
larger in the summer months than in the winter months.
Deaths due to heart failure, arrhythmia, cerebrovascular causes,
and thrombocytic causes were more strongly (~ 2.5 to 4 times
larger relative risks) associated with air pollution than the overall
cardiovascular deaths (CVD) or myocardial infarction (MI) and
other ischemic heart disease (IHD).
No pollutant significantly associated with mortality from all
cardiovascular or CVD causes in 65+ year age group. Only in
age <65 year group, PM10 associated with total and CVD deaths
with 4 and 5 d lags, respectively. The "significant" lags were
rather "spiky". O3 was also associated with CVD mortality <65
yr. group with inconsistent signs and late and spiky lags (neg. on
d 5 and pos. on d 6).
In the coal basin (i.e., the Czech Republic polluted area), on the
average, 68% of the TSP was PM10, and most of PM10 was PM25
(75%). For the coal basin, associations were found between the
logarithm of TSP and all-cause mortality at lag 1 or 2 days. SO2
was also associated with all-cause mortality with slightly lower
significance. PM10 and PM25 were both associated with all-
cause mortality in 1993-1994 with a lag of 1-day. NO2, O3 and
CO were positively but more weakly associated with mortality
than PM indices or SO2. In the Bavarian region, neither TSP nor
SO2 was associated with mortality, but CO (at lag 1-day) and O3
(at lag 0-day) were associated with all-cause mortality.
0.9 (0.1, 1.7) per 50 ,ug/m3 PM10; 1.0
(0.5, 1.5) per 25 ^g/m3 BS; 3.2 (0.6,
5.9) per 25 ,ug/m3 sulfate; and 4.1 (1.4,
6.9) per 25 ^g/m3 nitrate, all at 1 day
lag.
For PM10 (7-day mean), RRs for total
CVD, MI/IHD, arrhythmia, heart
failure, cerebrovascular, and
thrombocytic mortality per 80 ug/m3
increase were: 1.2% (-1.6, 4.1), 0.5% (-
3.6, 4.8), 4.1% (-6.8, 16.3), 3.6% (-4.0,
11.8), 3.1%(-2.9, 9.4), and 1.0%(-
10.6, 14.3), respectively. The RRs for
BS were larger and more significant
than those for PM10.
18.8% (5.6, 33.2) per 50 ^g/m3 PM10 4
day lag (other lags negative or zero).
Total mortality excess deaths per 100
//g/m3 increase in TSP for the Czech
region: 3.8 (0.8, 6.9) at lag 2-day for
1982-1994 period. For period 1993-
1994, 9.5 (1.2, 18.5) per 100 ,ug/m3
increase in TSP at lag 1-day, and 4.8
(0.7, 9.0) per 50 ,ug/m3 increase in
PM10; and 1.4 (-0.5, 3.4) per 25 ,ug/m3
PM,,.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
to
o
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Hoeketal. (1997).
Rotterdam, the Netherlands,
1983-1991. TSP (median 42);
BS (median 13).
Kotesovec et al. (2000).
Northern Bohemia, Czech
Republic, 1982-1994.
TSP (121.3).
Zanobetti et al. (2000a).
Milan, Italy. 1980-1989.
TSP mean = 142.
Anderson et al. (1996).
London, UK, 1987-1992.
BS(15)
Total mortality (also by age group) was regressed on TSP,
Fe (from TSP filter), BS, O3, SO2, CO, adjusting for
seasonal cycles, day-of-week, influenza, temperature, and
humidity using Poisson GAM model.
Total (excluding accidents and children younger than 1 yr),
cause specific (cardiovascular and cancer), age (65 and less
vs. otherwise), and gender specific mortality series were
examined for their associations with TSP and SO2 using
logistic model, adjusting for seasonal cycles, influenza
epidemics, linear and quadratic temperature terms. Lags 0
through 6 days, as well as a 7 day mean values were
examined.
The focus of this study was to quantify mortality
displacement using GAM distributed lag models. Non-
accidental total deaths were regressed on smooth function of
TSP distributed over the same day and the previous 45 days
using penalized splines for the smooth terms and seasonal
cycles, temperature, humidity, day-of-week, holidays, and
influenza epidemics. The mortality displacement was
modeled as the initial positive increase, negative rebound
(due to depletion), followed by another positive coefficients
period, and the sum of the three phases were considered as
the total cumulative effect.
Total, cardiovascular, and respiratory mortality series were
regressed on BS, O3, NO2, and SO2, adjusting for seasonal
cycles, day-of-week, influenza, holidays, temperature,
humidity, and autocorrelation using Poisson model.
Daily deaths were most consistently associated with TSP. TSP
and O3 effects were "independent" of SO2 and CO. Total iron
(from TSP filter) was associated "less consistently" with
mortality than TSP was. The estimated RRs for PM indices were
higher in warm season than in cold season.
For the total mortality, TSP, but not SO2, was associated. There
were apparent differences in associations were found between
men and women. For example, for age below 65 cardiovascular
mortality was associated with TSP for men but not for women.
TSP was positively associated with mortality up to 13 days,
followed by nearly zero coefficients between 14 and 20 days,
and then followed by smaller but positive coefficients up to the
45th day (maximum examined). The sum of these coefficients
was over three times larger than that for the single-day estimate.
Both O3 (0 day lag) and BS (1 day lag) were significant
predictors of total deaths. O3 was also positively significantly
associated with respiratory and cardiovascular deaths. The effect
size estimates per the same distributional increment (10% to
90%) were larger for O3 than for BS. These effects were larger
in warm season. SO2 and NO2 were not consistently associated
with mortality.
5.5 (1.1, 9.9) per 100 //g/m3 TSP at 1
day lag.
Total mortality percent excess deaths
per 100 |/g/m3 increase in TSP at 2 day
lag was 3.4 (0.5, 6.4).
Total mortality percent increase
estimates per IQR increase in TSP: 2.2
(1.4, 3.1) for single-day model; 6.7 (3.8,
9.6) for distributed lag model.
2.8% (1.4, 4.3) per 25 ^g/m3
lag for total deaths.
CVD(1 d)= 1.0 (-1.1, 3.1).
Resp. (1 d)= 1.1 (-2.7, 5.0).
BS at 1-d
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Michelozzietal. (1998).
Rome, Italy, 1992-1995.
TSP ("PM13" beta attenuation,
84).
Garcia-Aymerich et al. (2000).
Barcelona, Spain. 1985-1989.
Black Smoke no data
distribution was reported).
Rahlenbeck and Kahl (1996).
East Berlin,
1981-1989.
"SP" (beta attenuation, 97)
Rossi etal. (1999).
Milan, Italy, 1980-1989
TSP ("PM13" beta attenuation,
142)
Sunyer et al. (2000).
Barcelona, Spain.
1990-1995.
BS means: 43.9 for case period,
and 43.1 for control period.
Total mortality was related to PM13, SO2, NO2, CO, and O3,
using Poisson GAM model, adjusting for seasonal cycles,
temperature, humidity, day-of-week, and holiday. Analysis
of mortality by place of residence, by season, age, place of
death (in or out of hospital), and cause was also conducted.
Daily total (mean = 1.8/day), respiratory, and cardiovascular
mortality counts of a cohort (9,987 people) with COPD or
asthma were associated with black smoke (24-hr), SO2 (24-
hr and 1-hr max), NO2 (24-hr and 1-hr max), O3 (1-hr max),
temperature, and relative humidity. Poisson regression
models using APHEA protocol were used. The resulting
RRs were compared with those of the general population.
Total mortality (as well as deviations from long-wave
cycles) was regressed on SP and SO2, adjusting for day-of-
week, month, year, temperature, and relative humidity,
using OLS, with options to log-transform pollution, and w/
and w/o days with pollution above 150 //g/m3.
Specific causes of death (respiratory, respiratory infections,
COPD, circulatory, cardiac, heart failure, and myocardial
infarction) were related to TSP, SO2, and NO2, adjusting for
seasonal cycles, temperature, and humidity, using Poisson
GAM model.
Those over age 35 who sought emergency room services for
COPD exacerbation during 1985-1989 and died during
1990-1995 were included in analysis. Total, respiratory,
and cardiovascular deaths were analyzed using a conditional
logistic regression analysis with a case-crossover design,
adjusting for temperature, relative humidity, and influenza
epidemics. Bi-directional control period at 7 days was used.
Average of the same and previous 2 days used for pollution
exposure period. Data also stratified by potential effect
modifiers (e.g., age, gender, severity and number of ER
visits, etc.).
PM13 and NO2 were most consistently associated with mortality.
CO and O3 coefficients were positive, SO2 coefficients negative.
RR estimates higher in the warmer season. RRs similar for in-
and out-of hospital deaths.
Daily mortality in COPD patients was associated with all six
pollution indices. This association was stronger than in the
general population only for daily 1-hr max of SO2, daily 1-hr
max and daily means of NO2. BS and daily means of SO2
showed similar or weaker associations for COPD patients than
for the general population.
Both SP and SO2 were significantly associated with total
mortality with 2 day lag in single pollutant model. When both
pollutants were included, their coefficients were reduced by 33%
and 46% for SP and SO2, respectively.
All three pollutants were associated with all cause mortality.
Cause-specific analysis was conducted for TSP only.
Respiratory infection and heart failure deaths were both
associated with TSP on the concurrent day, whereas the
associations for myocardial infarction and COPD deaths were
found for the average of 3 to 4 day prior TSP.
BS levels were associated with all cause deaths. The association
was stronger for respiratory causes. Older women, patients
admitted to intensive care units, and patients with a higher rate
of ER visits were at greater risk of deaths associated with BS.
1.9% (0.5, 3.4) per 50 //g/m3 PM13 atO
day lag.
Total mortality percent increase per
25 //g/m3 increase in avg. of 0-3 day
lags of BS: 2.76 (1.31, 4.23) in general
population, and 1.14 (-4.4, 6.98) in the
COPD cohort.
6.1% per 100 //g/m3 "SP" at 2 day lag.
3.3% (2.4, 4.3) per 100 ,ug/m3 TSP at 0
day lag.
Percent increase per 25 //g/m3 increase
in 3-day average BS: 14.2(1.6,28.4)
for all causes; 9.7 (-10.2, 34.1) for
cardiovascular deaths; 23.2 (3.0, 47.4)
for respiratory deaths.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
Europe (cont'd)
Sunyer and Basagana (2001).
Barcelona, Spain. 1990-1995.
See Sunyer et al. (2000) for
PM levels.
The analysis assessed any "independent" particle effects,
after controlling for gaseous pollutants, on a cohort of
patients with COPD (see the summary description for
Sunyer et al. (2000) for analytical approach). PM10, NO2,
O,, and CO were analyzed.
PM10, but not gaseous pollutants were associated with mortality Odds ratio for all cause mortality per
for all causes. In the two-pollutant models, the PM10-mortality
associations were not diminished, whereas those with gaseous
pollutants were.
IQR PM10 on the same-day (27 ug/m3)
was 11% (0, 24). In two pollutant
models, the PM10 RRs were 10.5%,
12.9%, and 10.8% with NO2, O3, and
CO, respectively.
OO
to
to
fe
H
6
o
o
H
O
O
H
W
O
Tobias and Campbell (1999).
Barcelona, Spain.
1991-1995.
Black Smoke (BS)
(no data distribution
was reported).
Alberdi Odriozola et al. (1998).
Madrid, Spain, 1986-1992.
"TSP" (beta attenuation,
47 for average of 2 stations)
Diaz etal. (1999).
Madrid, Spain. 1990-1992.
TSP (no data distribution
was reported).
Study examined the sensitivity of estimated total mortality
effects of BS to different approaches to modeling influenza
epidemics: (1) with a single dummy variable; (2) with three
dummy variables; (3) using daily number of cases of
influenza. Poisson regression used to model total daily
mortality, adjusting for weather, long-term trend, and
season, apparently following APHEA protocol.
Total, respiratory, and cardiovascular deaths were related to
TSP and SO2. Multivariate autoregressive integrated
moving average models used to adjust for season,
temperature, relative humidity, and influenza epidemics.
Non-accidental, respiratory, and cardiovascular deaths
(mean = 62.4, 6.3, and 23.8 per day, respectively). Auto-
regressive Integrated Moving Average (ARIMA) models fit
to both depend, and independ. variables first to remove
auto-correlation and seasonality (i.e., pre-whitening"),
followed by examining cross-correlation to find optimal
lags. Multivariate OLS models thus included ARIMA
components, seasonal cycles (sine/cosine), V-shaped temp.,
and optimal lags found for pollution and weather variables.
TSP, SO2, NO2, and O3 examined. Season-specific analyses
also conducted.
Using the reported daily number of influenza cases resulted in a
better fit (i.e., a lower AIC) than those using dummy variables.
In the "better" model, the black smoke coefficient was about
10% smaller than those in the models with dummy influenza
variables, but remained significant. Lags not reported.
TSP (1-day lag) and SO2 (3-day lagged) were independently
associated with mortality.
TSP was significantly associated with non-accidental mortality
at lag 0 for year around and winter, but with a 1-day lag in
summer. A similar pattern was seen for circulatory deaths. For
respiratory mortality, a significant association with TSP was
found only in summer (0-day lag). SO2, NOx, and NO2 showed
similar associations with non-accidental deaths at lag 0 day. O3'
associations with non-accidental mortality was U-shaped, with
inconsistent lags (1, 4, and 10).
Total mortality excess deaths per 25
Mg/m3 increase in BS: 1.37 (0.20, 2.56)
for model using the daily case of
influenza; 1.71 (0.53, 2.91) for model
with three influenza dummy variables.
4.8% (1.8, 7.7) per 100 ,ug/m3 TSP at
lag 1 day.
For non-accidental mortality, excess
deaths was 7.4% (confidence bands not
reported; p < 0.05) per 100 ,ug/m3 TSP
at 0 day lag.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Wichmann et al., (2000).
Erfurt, Germany.
1995-1998.
Number counts (NC) & mass
concentrations (MC) of
ultrafme particles in three size
classes, 0.01 to 0.1 fan, and
fine particles in three size
classes from 0.1 to 2.5 fan.
diameter, using Spectrometryll
Mobile Aerosol Spectrometry
(MAS).
MAS MC PM2.5-0.01 (mean
25.8, median 18.8, IQR 19.9).
Filter measurements of PM10
(mean 38.2, median 31.0, IQR
27.7) and PM25 (mean 26.3,
median 20.2, IQR 18.5). MAS
NC2.5-0.01 (mean 17,966 per
cu.cm, median 14,769, IQR
13,269).
Zeghnoun et al. (2001).
Rouen and Le Havre, France.
1990-1995. PM13 mean = 32.9
for Rouen, 36.4 for Le Havre.
BS mean = 18.7 for Rouen,
16.3 for Le Havre.
Roemeretal. (2001).
Amsterdam. 1987-1998.
BS and PM10 means in
"background" = 10 and 39;
BS mean in "traffic" area = 21.
(No PM10 measurements
available at traffic sites)
Total non-accidental, cardiovascular, and respiratory deaths
(mean 4.88, 2.87, 1.08 per day, respectively) were related to
particle mass concentration and number counts in each size
class, and to mass concentrations of gaseous co-pollutants
NO2, CO, SO2, using GAM regression models adjusted for
temporal trends, day of week, weekly national influenza
rates, temperature and relative humidity. Data analyzed by
season, age group, and cause of death separately. Single-
day lags and polynomial distributed lag models (PDL) used.
Particle indices and pollutants fitted using linear, log-
transformed, and LOESS transformations. Two-pollutant
models with a particle index and a gaseous pollutant were
fitted. The "best" model as used by Wichmann et al. (2000)
was that having the highest t-statistic, since other criteria
(e.g., log-likelihood for nested models) and AIC for non-
nested models could not be applied due to different numbers
of observations in each model. There should be little
difference between these approaches and resulting
differences in results should be small in practice.
Sensitivity analyses included stratifying data by season,
winter year, age, cause of death, or transformation of the
pollution variable (none, logarithmic, non-parametric
smooth).
Total, cardiovascular, and respiratory mortality series were
regressed on BS, PM13, SO2, NO2, and O3 in 1- and 2-
pollutant models using GAM Poisson models adjusting for
seasonal trends, day-of-week, and weather.
Daily deaths for those who lived along roads with more than
10,000 motor vehicle, as well as deaths for total population,
were analyzed using data from background and traffic
monitors. Poisson GAM model was used adjusting for
season, day-of-week, and weather. BS, PM10, SO2, NO2,
CO, and O3 were analyzed.
Loss of stat. power by using a small city with a small number of
deaths was offset by advantage of having good exposure
representation from single monitoring site. Since ultrafme
particles can coagulate into larger aggregates in a few hours,
ultrafme particle size and numbers can increase into the fine
particle category, resulting in some ambiguity. Significant
associations were found between mortality and ultrafme particle
number concentration (NC), ultrafme particle mass
concentration (MC), fine particle mass concentration, or SO2
concentration. The correlation between MCO.01-2.5 and
NCO.01-0.1 is only moderate, suggesting it may be possible to
partially separate effects of ultrafme and fine particles. The most
predictive single-day effects are either immediate (lag 0 or 1) or
delayed (lag 4 or 5 days), but cumulative effects characterized by
PDL are larger than single-day effects. The significance of SO2
is robust, but hard to explain as a true causal factor since its
concentrations are very low. Age is an important modifying
factor, with larger effects at ages < 70 than > 70 years.
Respiratory mortality has a higher RR than cardio- vascular
mortality. A large number of models were fitted, with some
significant findings of association between mortality and particle
mass or number indices.
In Rouen, O3, SO2, and NO2 were each significantly associated
with total, respiratory, and cardiovascular mortality, respectively.
In Le Havre, SO2 and PM13 were associated with cardiovascular
mortality. However, the lack of statistical significance reported
for most of these results may be in part due to the relatively
small population size of these cities (430,000 and 260,000,
respectively).
Correlations between the background monitors and traffic
monitors were moderate for BS (r = 0.55) but higher for NO2 (r =
0.79) and O3 (r = 0.80). BS and NO2 were associated with
mortality in both total and traffic population. Estimated RR for
traffic population using background sites was larger than the RR
for total population using background sites. The RR for total
pop. using traffic sites was smaller that RRs for total population
using background sites. This is not surprising since the mean BS
for traffic sites were larger that for background sites.
Total mortality excess deaths:
Filter PM10 (0-4 d lag) = 6.6 (0.7, 12.8)
per 50 Mg/m3. Filter PM25 (0-1 d) = 3.0
(-1.7,7.9). MCforPM001.256.2%(1.4,
11.2) for all year; by season,
Winter = 9.2% (3.0, 15.7)
Spring = 5.2% (-2.0, 12.8)
Summer = -4.7% (-18.7, 11.7)
Fall = 9.7% (1.9, 18.1)
For ultrafme PM, NC 0.01-0.1 (0-4 d
lag):
All Year =8.2% (0.3, 16.9)
Winter = 9.7% (0.3, 19.9)
Spring = 10.5% (-1.4, 23.9)
Summer = -13.9% (-29.8, 5.7)
Fall = 12.0% (2.1,22.7)
PM13 total mortality RRs per IQR were
0.5% (-1.1, 2.1) in Rouen (IQR=20.6,
1-day lag) and 1.9% (-0.8, 7.4) in Le
Havre (IQR=23.9, 1-day lag ). BS total
mortality RRs per IQR were 0.5% (-1.8,
2.9) in Rouen (IQR=14.2, 1-day lag)
and 0.3% (-1.6, 2.2) in Le Havre
(IQR=11.5, 0-1 day lag avg.).
The RRs per 100 ug/m3 BS (at lag
1-day) were 1.383 (1.153, 1.659),
1.887 (1.207, 2.949), and 1.122 (1.023,
1.231) for total population using
background sites, traffic population
using background sites, and total
population using traffic sites,
respectively. Results for traffic pop.
using traffic sites not reported)
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Anderson et al. (2001).
The west Midlands
conurbation, UK. 1994-1996.
PM means: PM10 = 23,
PM25 = 15, PM10.25 = 9,
BS = 13.2, sulfate = 3.7.
Keatinge and Donaldson
(2001). Greater London,
England, 1976-1995.
BS mean = 17.7.
Latin America
Cifuentes et al. (2000).
Santiago, Chile.
1988-1996.
PM25(64.0), andPM1025
(47.3).
Castillejosetal. (2000).
Mexico City.
1992-1995.
PM10 (44.6), PM25 (27.4),
andPM10.25(17.2).
Loomisetal. (1999).
Mexico-City, 1993-1995.
PM25 (mean: 27.4 ,ug/m3)
Non-accidental cause, cardiovascular, and respiratory
mortality (as well as hospital admissions) were analyzed for
their associations with PM indices and gaseous pollutants
using GAM Poisson models adjusting for seasonal cycles,
day-of-week, and weather.
The study examined potential confounding effects of
atypical cold weather on air pollution/mortality
relationships. First, air pollution variables (SO2, CO and
BS) were modeled as a function of lagged weather variables
These variables were deseasonalized by regressing on seine
and cosine variables. Mortality regression included various
lagged and averaged weather and pollution variables.
Analyses were conducted in the linear range of
mortality/temperature relationship (15 to 0 degrees C).
Non-accidental total deaths (56.6 per day) were examined
for associations with PM2 5, PM10.2 5, O3, CO, SO2, and NO2.
Data analyzed using GAM Poisson regression models,
adjusting for temperature, seasonal cycles. Single and two
pollutant models with lag days from 0 to 5, as well as the
2- to 5-day average concentrations evaluated.
Non-accidental total deaths, deaths for age 65 and over, and
cause-specific (cardiac, respiratory, and the other
remaining) deaths were examined for their associations with
PM10, PM2 5, PM10_2 5, O3, and NO2. Data were analyzed
using GAM Poisson regression models, adjusting for
temperature (average of 1-3 day lags) and seasonal cycles.
Individual pollution lag days from 0 to 5, and average
concentrations of previous 5 days were considered.
Infant mortality (avg. ~ 3/day) related to PM2 5, O3, and
NO2, adjusting for temperature and smoothed time, using
Poisson GAM model.
Daily non-accidental mortality was not associated with PM
indices or gaseous pollutants in the all-year analysis. However,
all the PM indices (except coarse particles) were positively and
significantly associated with non-accidental mortality (age over
65) in the warm season. Of gaseous pollutants, NO2 and O3
were positively and significantly associated with non-accidental
mortality in warm season. Two pollutant models were not
considered because "so few associations were found".
Polluted days were found to be colder and less windy and rainy
than usual. In the regression of mortality on the multiple-lagged
temperature, wind, rain, humidity, sumshine, SO2, CO, and BS,
cold temperature was associated with mortality increase, but not
SO2 or CO. BS suggestive evidence, though not statistically
significant, of association at 0- and 1-day lag.
Both PM size fractions associated with mortality, but different
effects found for warmer and colder months. PM2 5 and PM10_2 5
both important in whole year, winter, and summer. In summer,
PM10_25 had largest effect size estimate. NO2 and CO also
associated with mortality, as was O3 in warmer months. No
consistent SO2-mortality associations.
All three particle size fractions were associated individually with
mortality. The effect size estimate was largest for PM10_25. The
effect size estimate was stronger for respiratory causes than for
total, cardiovascular, or other causes of death. The results were
not sensitive to additions of O3 and NO2. In the model with
simultaneous inclusion of PM25 and PM10_25, the effect size for
PM10_2 5 remained about the same, but the effect size for PM2 5
became negligible.
Excess infant mortality associated with PM2 5, NO2, and O3 in the
same average/lags. NO2 and O3 associations less consistent in
multi-pollutant models.
Percent excess mortality for PM10,
PM25, and PM10.25 (avg. lag 0 and 1
days) were 0.2% (-1.8, 2.2) per 24.4
ug/m3 PM10, 0.6% (-1.5, 2.7) per 17.7
ug/m3 PM25, and -0.6% (-4.2, 2.3) per
11.3 ug/m3 PM10_25 in all-year analysis.
The results for season specific analysis
were given only as figures.
3% (95% CI not reported) increase in
daily mortality per 17.7 ug/m3 of BS
(lag 0 and 1).
Percent excess total deaths per 25 ,ug/m3
increase in the average of previous two
days for the whole year: 1.8(1.3, 2.4)
for PM2.5 and 2.3 (1.4, 3.2) for PM10.25
in single pollutant models.
Total mortality percent increase
estimates per increase for average of
previous 5 days: 9.5 (5.0, 14.2) for
50 |/g/m3 PM10; 3.7 (0, 7.6) for
25 Mg/m3 PM25; and 10.5 (6.4, 14.8) for
25,ug/m3PM10.25.
Infant mortality excess risk: 18.2% (6.4,
30.7) per 25 ,ug/m3 PM25 at avg. 3-5 lag
days.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
Latin America (cont'd)
Borja-Aburto et al. (1998).
Mexico-City,
1993-1995.
PM2.5(mean: 27)
Borja-Aburto et al. (1997).
Mexico-City,
1990-1992.
TSP (median: 204)
Tellez-Rojo et al. (2000).
Mexico City. 1994.
PM10 mean = 75.1.
Total, respiratory, cardiovascular, other deaths, and age-
specific (age >= 65) deaths were related to PM2 5, O3, and
NO2, adjusting for 3-day lagged temperature and periodic
cycles, using Poisson GAM model.
Total, respiratory, cardiovascular, and age-specific (age >=
65) deaths were related to O3, TSP, and CO, adjusting for
minimum temperature (temperature also fitted seasonal
cycles) using Poisson models. The final models were
estimated using the iteratively weighted and filtered least
squares method to account for overdispersion and
autocorrelation.
One year of daily total respiratory and COPD mortality
series were analyzed for their associations with PM10 and
O3 using Poisson model adjusting for cold or warm months,
and 1-day lagged minimum temperature. The data were
stratified by the place of deaths.
PM2 5, O3, and NO2 were associated with mortality with different
lag/averaging periods (1 and 4 day lags; 1-2 avg.; 1-5 avg.,
respectively). PM25 associations were most consistently
significant. SO2 was available, but not analyzed because of its
"low" levels.
O3, SO2, and TSP were all associated with total mortality in
separate models, but in multiple pollutant model, only TSP
remained associated with mortality. CO association weak.
The average number of daily respiratory deaths, as well as that of
COPD deaths, was similar for in and out of hospital. They found
that the estimated PM10 relative risks were consistently larger for
the deaths that occurred outside medical units. The results are
apparently consistent with the assumption that the extent of
exposure misclassification may be smaller for those who died
outside medical units.
For total excess deaths, 3.4% (0.4, 6.4)
per 25 ,ug/m3 PM2 5 for both 0 and 4 d
lags. For respiratory (4 d) = 6.4 (-2.6,
16.2); for
CVD(4d) = 5.6(-0.1, 11.5)
Total deaths:
6% (3.3, 8.3) per 100 ,ug/m3 TSP at 0 d
lag.
CVD deaths:
5.2% (0.9, 9.9).
Resp. deaths:
9.5% (1.3, 18.4).
Percent excess for total respiratory and
COPD mortality were 2.9% (0.9, 4.9)
and 4.1% (1.3, 6.9) per 10 ug/m3
increase in 3-day lag PM10,
fe
H
6
o
o
H
O
O
H
W
O
O
Pereiraetal. (1998).
Sao Paulo, Brazil, 1991-1992.
PM10 (beta-attenuation, 65)
Gouveia and Fletcher (2000).
Sao Paulo, Brazil. 1991-1993.
PM10mean = 64.3.
Conceicao et al. (2001).
Sao Paulo, Brazil. 1994-1997.
PM10mean = 66.2
Intrauterine mortality associations with PM10, NO2, SO2,
CO, and O3 investigated using Poisson regression adjusting
for season and weather. Ambient CO association with
blood carboxyhemoglobin sampled from umbilical cords of
non-smoking pregnant mothers studied in separate time
period.
All non-accidental causes, cardiovascular, and respiratory
mortality were analyzed for their associations with air
pollution (PM10, SO2, NO2, O3, and CO) using Poisson
model adjusting for trend, seasonal cycles, and weather.
Potential roles of age and socio-economic status were
examined by stratifying data by these factors.
Daily respiratory deaths for children under 5 years of age
were analyzed for their associations with air pollution
(PM10, SO2, O3, and CO) using GAM Poisson model
adjusting for seasonal cycles and weather.
NO2, SO2, and CO were all individually significant predictor of
the intrauterine mortality. NO2 was most significant in multi-
pollutant model. PM10 and O3 were not significantly associated
with the mortality. Ambient CO levels were associated with and
carboxyhemoglobin of blood sampled from the umbilical cords.
There was an apparent effect modification by age categories.
Estimated PM10 effects were higher for deaths above age 65
(highest for the age 85+ category), and no associations were
found in age group < 65 years. Respiratory excess deaths were
larger than those for cardiovascular or non-accidental deaths.
Other pollutants were also associated with the elderly mortality.
Significant mortality associations were found for CO, SO2, and
PM10 in single pollutant models. When all the pollutants were
included, PM10 coefficient became negative and non-significant.
Intrauterine mortality excess risk: 4.1%
(-1.8, 10.4) per 50 ,ug/m3 PM10 at 0 day
lag.
Percent excess for total non-accidental,
cardiovascular, and respiratory
mortality for those with age > 65 were
3.3% (0.6, 6.0), 3.8% (0.1, 7.6), and 6.0
(0.5, 11.8), respectively, per 64.2 ug/m3
increase in PM10 (0-, 0-, and 1-day lag,
respectively).
Percent excess for child (age < 5)
respiratory deaths: 9.7% (1.5, 18.6) per
66.2 ug/m3 PM10 (2-day lag) in single
pollutant model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
IQRin,ug/m3.
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
fe
H
6
o
o
H
O
O
H
W
O
Australia
Morgan etal. (1998).
Sydney, 1989-1993.
Nephelometer (0.30
bscat/104m).
Site-specific conversion:
PM25 = 9; PM10 = 18
Simpson etal. (1997).
Brisbane, 1987-1993.
PM10 (27, not used in analysis).
Nephelometer
(0.26 bscat/104m,
size range: 0.01-2 ,um).
Asia
Hong etal. (1999).
Inchon, South Korea,
1995-1996 (20 months).
PM10mean = 71.2.
Lee etal. (1999).
Seoul and Ulsan, Korea,
1991-1995. TSP(beta
attenuation, 93 for Seoul
and 72 for Ulsan)
Total, cardiovascular, and respiratory deaths were related to
PM (nephelometer), O3, and NO2, adjusting for seasonal
cycles, day-of-week, temperature, dewpoint, holidays, and
influenza, using Poisson GEE to adjust for autocorrelation.
Total, cardiovascular, and respiratory deaths (also by age
group) were related to PM (nephelometer), O3, SO2, and
NO2, adjusting for seasonal cycles, day-of-week,
temperature, dewpoint, holidays, and influenza, using
Poisson GEE to adjust for autocorrelation. Season-specific
(warm and cold) analyses were also conducted.
Non-accidental total deaths, cardiovascular, and respiratory
deaths were examined for their associations with PM10, O3,
SO2, CO, and NO2. Data were analyzed using GAM
Poisson regression models, adjusting for temperature,
relative humidity, and seasonal cycles. Individual pollution
lag days from 0 to 5, as well as the average concentrations
of previous 5 days were considered.
Total mortality series was examined for its association with
TSP, SO2, and O3, in Poisson GEE (exchangeable
correlation for days in the same year), adjusting for season,
temperature, and humidity.
PM, O3, and NO2 all showed significant associations with total
mortality in single pollutant models. In multiple pollutant
models, the PM and O3 effect estimates for total and
cardiovascular deaths were marginally reduced, but the PM
effect estimate for respiratory deaths was substantially reduced.
Same-day PM and O3 were associated most significantly with
total deaths. The O3 effect size estimates for cardiovascular and
respiratory deaths were consistently positive (though not
significant), and larger in summer. PM's effect size estimates
were comparable for warm and cold season for cardiovascular
deaths, but larger in warm season for respiratory deaths. NO2
and SO2 were not associated with mortality.
A greater association with mortality was seen with the 5-day
moving average and the previous day's exposure than other
lag/averaging time. In the models that included a 5-day moving
average of one or multiple pollutants, PM10 was a significant
predictor of total mortality, but gaseous pollutants were not
significant. PM10 was also a significant predictor of
cardiovascular and respiratory mortality.
All the pollutants were significant predictors of mortality in
single pollutant models. TSP was not significant in multiple
pollutant models, but SO2 and O3 remained significant.
4.7% (1.6, 8.0) per 25 ,ug/m3 estimated
PM25 or 50 |/g/m3 estimated PM10 at
avg. ofO and 1 day lags.
(Note: converted from nephelometry
data)
3.4% (0.4, 6.4) per 25 ,ug/m3 1-h PM25
increment at 0 d lag; and 7.8% (2.5,
13.2) per 25 ^g/m3 24-h PM25
increment.
Percent excess deaths (t-ratio) per 50
//g/m3 increase in the 5-day moving
average of PM10: 4.1 (0.1, 8.2) for total
deaths; 5.1 (0.1, 10.4) for
cardiovascular deaths; 14.4 (-3.2, 35.2)
for respiratory deaths.
5.1% (3.1, 7.2) for Seoul, and -0.1% (-
3.9, 3.9) for Ulsan, per lOO^g/m3 TSP
at avg. of 0, 1, and 2 day lags.
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
>
to
fe
H
6
o
o
H
O
O
H
W
O
Asia (cont'd)
Lee and Schwartz (1999).
Seoul, Korea. 1991-1995.
TSPmean = 925.
Xu et al. (2000).
Shenyang, China, 1992
TSP (430).
Ostroetal. (1998).
Bangkok, Thailand,
1992-1995
PM10 (beta attenuation, 65)
Total deaths were analyzed for their association with TSP,
SO2, and O3. A conditional logistic regression analysis with
a case-crossover design was conducted. Three-day moving
average values (current and two past days) of TSP and SO2,
and l-hrmaxO3 were analyzed separately. The control
periods are 7 and 14 days before and/or after the case
period. Both unidirectional and bi-directional controls (7 or
7 and 14 days) were examined, resulting in six sets
of control selection schemes. Other covariates included
temperature and relative humidity.
Total (non-accidental), CVD, COPD, cancer and other
deaths examined for their associations with TSP and
SO2,using Poisson (GAM, and Markov approach to adjust
for mortality serial dependence) models, adjusting for
seasonal cycles, Sunday indicator, quintiles of temp, and
humidity. Ave. pollution values of concurrent and
3 preceding days used.
Total (non-accidental), cardiovascular, respiratory deaths
examined for associations with PM10 (separate
measurements showed =50% of PM10 was PM25),using
Poisson GAM model adjusting for seasonal cycles, day-of-
week, temp., humidity.
Among the six control periods, the two unidirectional
retrospective control schemes resulted in odds ratios less than
1; the two unidirectional prospective control schemes resulted
in larger odds ratios (e.g., 1.4 for 50 ppb increase in SO2); and
bi-directional control schemes resulted in odds ratios between
those for uni-directional schemes. SO2 was more significantly
associated with mortality than TSP.
Total deaths were associated with TSP and SO2 in both single
and two pollutant models. TSP was significantly associated
with CVD deaths, but not with COPD. SO2 significantly
associated with COPD, but not with CVD deaths. Cancer
deaths not associated with TSP or SO,.
All the mortality series were associated with PM10 at various
lags. The effects appear across all age groups. No other
pollutants were examined.
OR for non-accidental mortality
per 100 ,ug/m3 increase in 3-day
average TSP was 1.010 (0.988,
1.032).
Percent total excess deaths per
100 /^g/m3 increase in 0-3 day
ave. of TSP = 1.75(0.65,2.85);
with SO2 = 1.31(0.14, 2.49)
COPD TSP = 2.6 (-0.58, 5.89);
with SO2 = 0.76 (-2.46, 4.10).
CVD TSP = 2.15 (0.56, 3.71);
with SO2 = 1.95 (1.19, 3.74).
Cancer TSP = 0.87 (-1.14,
2.53); with SO2 = 1.07 (-1.05,
3.23).
Other deaths TSP = 3.52 (0.82,
6.30); with SO2 = 2.40 (-0.51,
5.89).
Total mortality excess risk: 5.1%
(2.1, 8.3) per 50 ,ug/m3 PM10 at
3 d lag (0 and 2 d lags also
significant).
CVD (3d ave.) = 8.3 (3.1, 13.8)
Resp. (3d ave.) = 3.0 (-8.4,
15.9)
O
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
o
o
to
Reference, Location, Years,
PM Index, Mean or Median,
IQRin//g/m3.
Study Description: Outcomes, Mean outcome rate, and
ages. Concentration measures or estimates. Modeling
methods: lags, smoothing, and covariates.
Results and Comments.
Design Issues, Uncertainties, Quantitative Outcomes.
PM Index, lag, Excess Risk%
(95% LCL, UCL), Co-pollutants.
OO
to
OO
fe
H
6
o
o
H
O
O
H
W
O
Asia (cont'd)
Cropper etal. (1997).
Delhi, India, 1991-1994
TSP(375)
Kwonetal. (2001).
Seoul, South Korea,
1994-1998.
PM10 mean = 68.7.
Lee et al. (2000).
Seven major cities, Korea.
1991-1997.
TSP mean = 77.9.
Total (by age group), respiratory and CVD deaths related to
TSP, SO2, and NOx, using GEE Poisson model (to control
for autocorrelation), adjusting for seasonal cycles
(trigonometric terms), temperature, and humidity. 70%
deaths occur before age 65 (in U.S., 70% occur after age
65).
The study was planned to test the hypothesis that patients
with congestive heart failure are more susceptible to the
harmful effects of ambient air pollution than the general
population. GAM Poisson regression models, adjusting for
seasonal cycles, temperature, humidity, day-of-week, as
well as the case-crossover design, with 7 and 14 days before
and after the case period, were applied
All non-accidental deaths were analyzed for their
associations with TSP, SO2, NO2, O3, and CO using GAM
Poisson model adjusting for trend, seasonal cycles, and
weather. Pollution relative risk estimates were obtained for
each city, and then pooled.
TSP was significantly associated with all mortality series
except with the very young (age 0-4) and the "very old" (age
>=65). The results were reported to be unaffected by addition
of SO2 to the model. The authors note that, because those who
are affected are younger (than Western cities), more life-years
are likely to be lost per person from air pollution impacts.
The estimated effects were larger among the congestive heart
failure patients than among the general population (2.5 — 4.1
times larger depending on the pollutants). The case-crossover
analysis showed similar results. In two pollutant models, the
PM10 effects were much lower when CO, NO2, or SO2 were
included. O3 had little impact on the effects of the other
pollutants.
In the results of pooled estimates for multiple pollutant models,
the SO2 relative risks were not affected by addition of other
pollutants, whereas the relative risks for other pollutants,
including TSP, were. The SO2 levels in these Korean cities were
much higher than the levels observed in the current U.S. For
example, the 24-hr mean SO2 levels in the Korean cities ranged
from 12.1 to 31.4 ppb, whereas, in Samet et al.'s 20 largest U.S.
cities, the range of 24-hr mean SO2 levels were 0.7 to 12.8 ppb.
2.3% (significant at 0.05, but SE
of estimate not reported) per 100
Mg/m3 TSP at 2 day lag.
The RRs for PM10 (same-day)
using the GAM approach for the
general population and for the
cohort with congestive heart
failure were 1.4% (0.6, 2.2) and
5.8 (-1.1, 13.1), respectively, per
42.1ug/m3. Corresponding
ORs using the case-crossover
approach were 0.1% (-0.9, 1.2)
and 7.4% (-2.2, 17.9),
respectively.
Percent excess deaths for all non-
accidental deaths was 1.7% (0.8,
2.6) per 100 ug/m3 2-day moving
average TSP.
O
-------
APPENDIX 8B
PARTICULATE MATTER-MORBIDITY STUDIES:
SUMMARY TABLES
April 2002 8B-1 DRAFT-DO NOT QUOTE OR CITE
-------
Appendix 8B.1: PM-Cardiovascular Admissions Studies
April 2002 8B-2 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8B-1. ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes.
Mean outcome rate, sample or population size, ages.
Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
UnitedStates
Samet et al. (2000a,b)
14 US cities
1985-1994, but range of years varied by city
PM10 (,ug/m3) mean, median, IQR:
Birmingham, AL: 34.8, 30.6, 26.3
Boulder, CO: 24.4, 22.0, 14.0
Canton, OH: 28.4, 25.6, 15.3
Chicago, II: 36.4, 32.6, 22.4
Colorado Springs, CO: 26.9, 22.9, 11.9
Detroit, MI: 36.8, 32.0, 28.2
Minneapolis/St. Paul, MN: 27.4, 24.1, 17.9
Nashville, TN: 31.6, 29.2, 17.9
New Haven, CT: 29.3,26.0,20.2
Pittsburgh, PA: 36.0, 30.5, 27.4
Provo/Orem, UT: 38.9, 30.3, 22.8
Seattle, WA: 31.0,26.7,20.0
Spokane, WA: 45.3,36.2,33.5
Youngstown, OH: 33.1, 29.4, 18.6
Daily medicare hospital admissions for total
cardiovascular disease, CVD (ICD9 codes 390-429), in
persons 65 or greater. Mean CVD counts ranged from 3 to
102/day in the 14 cities. Covariates: SO2, NO2, O3, CO,
temperature, relative humidity, barometric pressure. Stats:
In first stage, performed city-specific, PM10-ONLY,
generalized additive robust Poisson regression with
seasonal, weather, and day of week controls. Repeated
analysis for days with PM10 less than 50 ,ug/m3 to test for
threshold. Lags of 0-5 considered, as well as the quadratic
function of lags 0-5. Individual cities analyzed first. The
14 risk estimates were then analyzed in several second
stage analyses: combining risks across cities using inverse
variance weights, and regressing risk estimates on
potential effect-modifiers and slopes of PM10 on co-
pollutants.
City-specific risk estimates for a 10 i/g/m3
increase in PM10 ranged from -1.2% in Canton to
2.2% in Colorado Springs. Across-city weighted
mean risk estimate was largest at lag 0,
diminishing rapidly at other lags. Only the mean
of lags 0 and 1 was significantly associated with
CVD. There was no evidence of statistical
heterogeniety in risk estimates across cities for
CVD. City-specific risk estimates were not
associated with the percent of the population that
was non-white, living in poverty, college
educated, nor unemployed. No evidence was
observed that PM10 effects were modified by
weather. No association was observed between
the city-specific PM10 risk estimates and the city-
specific correlation between PM10 and co-
pollutants. However, due to the absence of multi-
pollutant regression results, it is not clear
whether this study demonstrates an independent
effect of PM10.
Percent Excess CVD Risk (95% CI),
combined over cities per 50 ,ug/m3
change in PM10.
PM10: Odlag.
5.5% (4.7, 6.2)
PM10: 0-Id lag.
6.0% (5.1,6.8)
PM10 < 50 Mg/m3: 0-Id lag.
7.6% (6.0, 9.1)
fe
H
6
o
o
H
O
O
H
W
O
o
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Janessen et al. (2002)
14 U.S. cities studied in Samet et al.
(2000a,b) above
Mean
Summer/Winter Ratio
Birmingham
Boulder*
Canton
Chicago
Colorado Springs*
Detroit
Minneapolis
Nashville
New Haven
Pittsburgh
Seattle*
Spokane*
Provo-Urem*
Youngstown
40.0/27.4
26.8/36.3
36.6/25.8
42.5/30.4
21.3/37.3
42.8/32.8
30.5/23.0
40.1/31.9
30.3/31.6
46.6/29.4
23.8/43.3
32.7/42.2
31.4/66.3
40.7/30.1
Zanobetti et al. (2000b)
10 US cities
1986-1994
PM10 (Mg/m3) median, IQR:
Canton, OH: 26, 15
Birmingham, AL: 31,26
Chicago, II: 33,23
Colorado Springs, CO: 23, 13
Detroit, MI: 32, 28
Minneapolis/St. Paul, MN: 24, 18
New Haven, CT: 26,21
Pittsburgh, PA: 30, 28
Seattle, WA: 27,21
Spokane, WA: 36, 34
0.69
1.35
0.70
0.71
1.75
0.77
0.75
0.80
1.04
0.63
1.82
1.29
2.11
0.74
Examined same database as Samet et al. (2000a,b) to
evaluate whether differences in prevalence in air
conditioning (AC) and/or the contribution of different
sources to total PM10 emissions could partially explain the
observed variability in exposure effect relations. Variables
included 24-hr means of temperature. Cities were
characterized and analyzed as either winter or nonwinter
peaking. Rations between mean concentrations during
summer (June, July August) and winter (January,
February, March) were calculated. ('Winter peaking PM10
concentration.)
Derived from the Samet et al. (2000a,b) study, but for a
subset of 10 cities. Daily hospital admissions for total
cardiovascular disease, CVD (ICD9 codes 390-429), in
persons 65 or greater. Median CVD counts ranged from 3
to 103/dayinthe 10 cities. Covariates: SO2, O3, CO,
temperature, relative humidity, barommetric pressure.
Stats: In first stage, performed single-pollutant generalized
additive robust Poisson regression with seasonal, weather,
and day of week controls. Repeated analysis for days with
PM10 less than 50 ,ug/m3 to test for threshold. Lags of 0-5
considered, as well as the quadratic function of lags 0-5.
Individual cities analyzed first. The 10 risk estimates were
then analyzed in several second stage analyses: combining
risks across cities using inverse variance weights, and
regressing risk estimates on potential effect-modifiers and
pollutant confounders.
Analysis of city groups of winter peaking, PM10
and nonwinter peaking PM10 yielded coefficients
for CVD-related hospitalization admissions that
decreased significantly with increasing
percentage of central AC for both city groups.
Four source related variables coefficients for
hospital admissions for CVD increased
significantly with increasing percentage of PM10
from highway vehicles, highway diesels, oil
combustion, metal processing, increasing
population, and vehicle miles traveled (VMT)
per sq mg and with decreasing percentage of
PM10 from fugitive dust. For COPD and
pneumonia association were less significant but
the pattern of association were similar to that for
CVD.
Same basic pattern of results as in Samet et al.
(2000a,b). For distributed lag analysis, lag 0 had
largest effect, lags 1 and 2 smaller effects, and
none at larger lags. City-specific slopes were
independent of percent poverty and percent non-
white. Effect size increase when data were
restricted to days with PM10 less than 50 ,ug/m3.
No multi-pollutant models reported; however, no
evidence of effect modification by co-pollutants
in second stage analysis. As with Samet et al.
2000., it is not clear whether this study
demonstrates an independent effect of PM10.
Homes with AC
PCVD
% change (SE)
All cities
-15.2(14.8)
Nonwinter peak cities
-50.3" (17.4)
Winter peak cities
-51.7" (13.8)
Source PM10 from
highway vehicles
% change (SE) p CVD
58.0*(9.9)
[**p<0.05]
Percent Excess Risk (SE) combined
over cities:
Effects computed for 50 ,ug/m3 change
in PM10.
PM10: Od.
5.6 (4.7, 6.4)
PM10: 0-1 d.
6.2(5.4,7.0)
PM10 < 50//g/m3: 0-ld.
7.8 (6.2, 9.4)
-------
to
o
o
to
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
>
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Schwartz (1999)
8 US metropolitan counties
1988-1990
median, IQR for PM10 (//g/m3):
Chicago, IL: 35, 23
Colorado Springs, CO: 23, 14
Minneapolis, MN: 28, 15
New Haven, CT: 37,25
St. Paul, MN: 34, 23
Seattle, WA: 29,20
Spokane, WA: 37,33
Tacoma, WA: 37,27
Linn et al. (2000)
Los Angeles
1992-1995
mean, SD:
PM10est(,ug/m3):45, 18
Daily hospital admissions for total cardiovascular diseases
(ICD9 codes 390-429) among persons over 65 years.
Median daily hospitalizations: 110, 3, 14, 18, 9, 22, 6, 7,
alphabetically by city. Covariates: CO, temperature,
dewpoint temp. Stats: robust Poisson regression after
removing admission outliers; generalized additive models
with LOESS smooths for control of trends, seasons, and
weather. Day of week dummy variables. Lag 0 used for
all covariates.
Hospital admissions for total cardiovascular diseases
(CVD), congestive heart failure (CHE), myocardial
infarction (MI), cardiac arrhythmia (CA) among all
persons 30 years and older, and by sex, age, race, and
season. Mean hospital admissions for CVD: 428.
Covariates: CO, NO2, O3, temperature, rainfall. Daily
gravimetric PM10 estimated by regression of every sixth
day PM10 on daily real-time PM10 data collected by
TEOM. Poisson regression with controls for seasons and
day of week. Reported results for lag 0 only. Results
reported as Poisson regression coefficients and their
standard errors. The number of daily CVD admissions
associated with the mean PM10 concentration can be
computed by multiplying the PM10 coefficient by the PM10
mean and then exponentiating. Percent effects are
calculated by dividing this result by the mean daily
admission count for CVD.
In single-pollutant models, similar PM10 effect
sizes obtained for each county. Five of eight
county-specific effects were statistically
significant, as was the PM10 effect pooled across
locations. CO effects significant in six of eight
counties. The PM10 and CO effects were both
significant in a two pollutant model that was run
for five counties where the PM10/CO correlation
was less than 0.5. Results reinforce those of
Schwartz, 1997.
In year-round, single-pollutant models,
significant effects of CO, NO2, and PM10
on CVD were reported. PM10 effects appeared
larger in winter and fall than in spring and
summer. No consistent differences in PM10
effects across sex, age, and race. CO risk was
robust to including PM10 in the model; no results
presented on PM10 robustness to co-pollutants.
Percent Excess Risk (95% CI):
Effects computed for 50 ,ug/m3 change
in PM10.
PM10: Od.
Individual counties:
Chicago: 4.7(2.6,6.8)
COSpng: 5.6 (-6.8, 19.0)
Minneap: 4.1 (-3.6, 12.5)
NewHav: 5.8(2.1,9.7)
St. Paul: 8.6 (2.9, 14.5)
Seattle: 3.6 (-0.1, 7.4)
Spokane: 6.7 (0.9, 12.8)
Tacoma: 5.3(3.1,7.6)
Pooled: 5.0(3.7,6.4)
3.8(2.0, 5.5) w. CO
% increase with PM10 change of
50 ,ug/m3:
PM10est: Od.
CVD ages 30+
3.25% (2.04, 4.47)
MI ages 30+
3.04% (0.06, 6.12)
CHE ages 30+
2.02% (-0.94, 5.06)
CA ages 30+
1.01% (-1.93, 4.02)
o
-------
to
o
o
to
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
United States (cont'd)
Morris and Naumova (1998)
Chicago, IL
1986-1989
mean, median, IQR, 75th percentile:
PM10 Cug/m3): 41,38,23,51
Schwartz (1997)
Tucson, AZ
1988-1990
mean, median, IQR:
PM10 (,ug/m3): 42, 39, 23
Daily hospital admissions for congestive heart failure,
CHF (ICD9 428), among persons over 65 years. Mean
hospitalizations: 34/day. Covariates: O3, NO2, SO2,
CO, temperature, relative humidity. Gases measured at up
to eight sites; daily PM10 measured at one site. Stats:
GLM for time series data. Controlled for trends and cycles
using dummy variables for day of week, month, and year.
Residuals were modeled as negative binomial distribution.
Lags of 0-3 days examined.
Daily hospital admissions for total cardiovascular diseases
(ICD9 codes 390-429) among persons over 65 years.
Mean hospitalizations: 13.4/day. Covariates: O3, NO2,
CO, SO2, temperature, dewpoint temperature. Gases
measured at multiple sites; daily PM10 at one site. Stats:
robust Poisson regression; generalized additive model with
LOESS smooth for controlling trends and seasons, and
regression splines to control weather. Lags of 0-2 days
examined.
CO was only pollutant statistically significant in
both single- and multi-pollutant models.
Exposure misclassification may have been larger
for PM10 due to single site. Results suggest
effects of both CO and PM10 on congestive heart
failure hospitalizations among elderly, but CO
effects appear more robust.
Both PM10 (lag 0) and CO significantly and
independently associated with admissions,
whereas other gases were not. Sensitivity
analyses reinforced these basic results. Results
suggest independent effects of both PM10 and
CO for total cardiovascular hospitalizations
among the elderly.
Percent Excess Risk (95% CI)
per 50 ,ug/m3 change in PM10.
PM10: Od.
3.92% (1.02, 6.90)
1.96% (-1.4, 5.4) with
4 gaseous pollutants
Percent Excess Risk (95% CI)
per 50 Mg/m3 change in PM10.
PM10: Od.
6.07% (1.12, 1.27)
5.22% (0.17, 10.54) w. CO
fe
H
6
o
o
H
O
O
H
W
O
Gwynn et al (2000)
Buffalo, NY
mn/max
PM10 = 24.1/90.8,ug/m3
SO4- = 2.4/3.9
H+ = 36.4/38.2 nmol/m3
CoH = 0.2/0.9 10'3 ft
Air pollution health effects associations with total,
respiratory, and CVD hospital admissions (HA's)
examined using Poisson model controlling for weather,
seasonality, long-wave effects, day of week, holidays.
Positive, but non-significant assoc. found
between all PM indices and circulatory hospital
admissions. Addition of gaseous pollutants to
the model had minimal effects on the PM RR
estimates.
Percent excess CVD HA risks (95%
CI) per PM10 = 50 ,ug/m3; SO4 =
15,ug/m3;
H+ = 75 nmoles/m3; COH = 0.5
units/1,000 ft:
PM10 (lag 3) = 5.7% (-3.3, 15.5)
SO4 (lag 1) = 0.1% (-0.1, 0.4)
H+(lag 0) = 1.9% (-0.3, 4.2)
COH (lag 1) = 2.2% (-1.9, 6.3)
o
-------
to
o
o
to
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
Reference citation. Location, Duration
PM Index, Mean or Median, IQR ,ug/m3
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
United States (cont'd)
Lippmann et al. (2000)
Detroit, MI
1992-1994
mean, median, IQR:
PM25(Mg/m3): 18, 15, 11
PM10(,ug/m3):31, 28, 19
PM10.25(//g/m3):13, 12,9
Various cardiovascular (CVD)-related hospital admissions
(HA's) for persons 65+ yr. analyzed, using GAM Poisson
models, adjusting for season, day of week, temperature,
and relative humidity. The air pollution variables analyzed
were: PM10, PM2 5, PM10.2 5, sulfate, H+, O3, SO2, NO2, and
CO. However, this study site/period had very low acidic
aerosol levels. As noted by the authors 85% of H+ data
was below detection limit (8 nmol/m3).
For heart failure, all PM metrics yielded
significant associations. Associations for IHD,
dysrhythmia, and stroke were positive but
generally non-sig. with all PM indices. Adding
gaseous pollutants had negligible effects on
various PM metric RR estimates. The general
similarity of the PM25 and PM10_25 effects per
,ug/m3 in this study suggest similarity in human
toxicity of these two inhalable mass components
in study locales/periods where PM2 5 acidity not
usually present. However, small sample size
limits power to distinguish between pollutant-
specific effects.
Percent excess CVD HA risks (95%
CI) per 50 //g/m3 PM10, 25 ,ug/m3
PM25andPM10.25:
IHD:
PM25(lag2)4.3(-1.4, 10.4)
PM10 (lag 2) 8.9 (0.5, 18.0)
PM10.2.5 (lag 2) 10.5 (2.7, 18.9)
Dysrhythmia:
PM25 (lag 1)3.2 (-6.5, 14.0)
PM10 (lag 1)2.9 (-6.8, 13.7)
PM10.2.5 (lag 0)0.2 (-12.2, 14.4)
Heart Failure:
PM25 (lag 1)9.1(2.4, 16.2)
PM10 (lag 0)9.7 (0.2, 20.1)
PM10.2.5 (lag 0)5.2 (-3.3, 14.5)
Stroke:
PM25 (lag 0)1.8 (-5.3, 9.4)
PM10 (lag 1)4.8 (-5.5, 16.2)
PM10.25 (lag 1)4.9 (-4.7, 15.5)
fe
H
6
o
o
H
O
O
H
W
O
O
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
OO
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Moolgavkar (2000b)
Three urban counties: Cook, IL; Los
Angeles, CA; Maricopa, AZ.
1987-1995
Pollutant median, IQR:
Cook: PM10: 35, 22
LA: PM10: 44, 26
PM25: 22, 16
Maricopa: PM10: 41, 19
Analysis of daily hospital admissions for total
cardiovascular diseases, CVD, (ICD9 codes 390-429) and
cerebrovascular diseases, CRD, (ICD9 430-448) among
persons aged 65 and over. For Los Angeles, a second age
group, 20-64, was also analyzed. Median daily CVD
admissions were 110, 172, and 33 in Cook, LA, and
Maricopa counties, respectively. PM10 available only
every sixth day in LA and Maricopa counties. In LA,
every-sixth-day PM2 5 also was available. Covariates: CO,
NO2, O3, SO2, temperature, relative humidity. Stats:
generalized additive Poisson regression, with controls for
day of week and smooth temporal variability. Single-
pollutant models estimated for individual lags from 0 to 5.
Two-pollutant models also estimated, with both pollutants
at same lag.
In single-pollutant models in Cook and LA
counties, PM was significantly associated with
CVD admissions at lags 0, 1, and 2, with
diminishing effects over lags. PM2 5 also was
significant in LA for lags 0 and 1. For the 20-64
year old age group in LA, risk estimates were
similar to those for 65+. In Maricopa county, no
positive PM10 associations were observed at any
lag. In two-pollutant models in Cook and LA
counties, the PM10/PM25 risk estimates
diminished and/or were rendered non-
significant. Little evidence observed for
associations between CRD admissions and PM.
These results suggest that PM is not
independently associated with CVD or CRD
hospital admissions.
Percent Excess CVD Risk (95% CI)
Effects computed for 50 ,ug/m3 change
in PM10 and 25 ,ug/m3 change in
PM25.
Cook 65+:
PM10, 0 d.
4.2(3.0,5.5)
PM10, 0 d. w/NO2.
1.8(0.4,3.2)
LA 65+:
PM10, 0 d.
3.2(1.2,5.3)
PM10, 0 d. w/CO
-1.8 (-4.4, 0.9)
PM25, Od.
4.3(2.5,6.1)
PM25, Od. w/CO
0.8 (-1.3, 2.9)
LA 20-64 years old:
PM10, 0 d.
4.4 (2.2, 6.7)
PM10, 0 d. w/CO
1.4 (-1.3, 4.2)
PM25, Od.
3.5(1.8,5.3)
PM25, Od., w/CO)
2.3 (-0.2, 4.8)
Maricopa:
PM10, 0 d.
-2.4 (-6.9, 2.3)
O
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Zanobetti et al. (2000a)
Cook County, IL
1985-1994
Median, IQR:
PM10 Og/m3): 33, 23
Tolbert et al. (2000a)
Atlanta
Period 1: 1/1/93-7/31/98
Mean, median, SD:
PM10 Og/m3): 30.1,28.0, 12.4
Period 2: 8/1/98-8/31/99
Mean, median, SD:
PM10 Og/m3): 29.1,27.6, 12.0
PM25Og/m3): 19.4, 17.5,9.35
CP Og/m3): 9.39, 8.95, 4.52 10-100 nm PM
counts
(count/cm3): 15,200, 10,900, 26,600
10-100 nm PM surface area (umVcm3):
62.5,43.4, 116
PM25 soluble metals Og/m3): 0.0327,
0.0226, 0.0306
PM2.5 Sulfates Og/m3): 5.59, 4.67, 3.6
PM2 5 Acidity Og/m3): 0.0181,0.0112,
0.0219
PM25 organic PM Og/m3): 6.30, 5.90, 3.16
PM2 5 elemental carbon Og/m3): 2.25, 1.88,
1.74
Total cardivascular hospital admissions in persons 65 and
older (ICD 9 codes390-429) in relation to PM10. Data
were analyzed to examine effect modification by
concurrent or preexisting cardiac and/or respiratory
conditions, age, race, and sex. No co-pollutants included.
Preliminary analysis of daily emergency department (ED)
visits for dysrhythmias, DYS, (ICD 9 code 427) and all
cardiovascular diseases, CVD, (codes 402, 410-414, 427,
428, 433-437, 440, 444, 451-453) for persons aged 16 and
older in the period before (Period 1) and during (Period 2)
the Atlanta superstation study. ED data analyzed here
from just 18 of 33 participating hospitals; numbers of
participating hospitals increased during period 1. Mean
daily ED visits for dysrhythmias and all CVD in period 1
were 6.5 and 28.4, respectively. Mean daily ED visits for
dysrhythmias and all CVD in period 2 were 11.2 and 45.1,
respectively. Covariates: NO2, O3, SO2, CO temperature,
dewpoint, and, in period 2 only, VOCs. PM measured by
both TEOM and Federal Reference Method; unclear which
used in analyses. For epidemiologic analyses, the two time
periods were analyzed separately. Poisson regression
analyses were conducted with cubic splines for time,
temperature and dewpoint. Day of week and hospital
entry/exit indicators also included. Pollutants were treated
a-priori as three-day moving averages of lags 0, 1, and 2.
Only single-pollutant results reported.
Evidence seen for increased CVD effects among
persons with concurrent respiratory infections or
with previous admissions for conduction
disorders.
In period 1, significant negative association
(p=0.02) observed between CVD and 3-day
average PM10. There was ca. 2% drop in CVD
per 10 ,ug/m3 increase in PM10. CVD was
positively associated with NO2 (p=0.11) and
negatively associated with SO2 (p=0.10). No
association observed between dysrhythmias and
PM10 in period 1. However, dysrhythmias were
positively associated with NO2 (p=0.06). In
period 2, i.e., the first year of operation of the
superstation, no associations seen with PM10 or
PM25. However, significant positive
associations observed between CVD and
elemental carbon (p=0.005) and organic matter
(p=0.02), as well as with CO (p=0.001).
For dysrhythmias, significant positive
associations observed with elemental carbon
(p=0.004), CP (p=0.04), and CO (p=0.005).
These preliminary results should be interpreted
with caution given the incomplete and variable
nature of the databases analyzed.
Percent Excess CVD Risk (95% CI)
Effects computed for 50 ,ug/m3
PM10, 0-1 D. AVG.
CVD: 6.6 (4.9-8.3)
Percent Excess Risk (p-value):
Effects computed for 50 ,ug/m3 change
in PM10; 25 ^g/m3 for CP and PM2 5;
25,000 counts/cm3 for 10-100 nm
counts.
Period 1:
PM10, 0-2 d. avg.
CVD: -8.2(0.02)
DYS: 4.6 (0.58)
Period 2:
0-2 d. avg. in all cases
CVD % effect; DYS % effect:
PM10: 5.1 (-7.9, 19.9); 13.1 (-14.1,
50.0)
PM25: 6.1 (-3.1, 16.2); 6.1 (-12.6,
28.9)
CP: 17.6 (-4.6, 45.0); 53.2 (2.1,
129.6)
10-100 nm counts: -11.0
(0.17); 3.0 (0.87)
O
-------
to
o
o
to
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
Canada
Burnett etal. (1995)
Ontario, Canada
1983-1988
Sulfate
Mean: 4.37 ,ug/m3
Median: 3.07,ug/m3
95th percentile: 13 ^
Burnett etal. (1997a)
Canada's 10 largest cities
1981-1994
COH daily maximum
Mean: 0.7 103 In feet
Median: 0.6 103 In feet
95th percentile: 1.5103lnfeet
168 Ontario hospitals. Hospitalizations for coronary artery
disease, CAD (ICD9 codes 410,413), cardiac
dysrhythmias, DYS (code 427), heart failure, HF (code
428), and all three categories combined (total CVD).
Mean total CVD rate: 14.4/day. 1986 population of study
area: 8.7 million. All ages, <65, >=65. Both sexes, males,
females. Daily sulfates from nine monitoring stations.
Ozone from 22 stations. Log hospitalizations filtered with
19-day moving average prior to GEE analysis. Day of
week effects removed. 0-3 day lags examined.
Covariates: ozone, ozone2, temperature, temperature2.
Linear and quadratic sulfate terms included in model.
Daily hospitalizations for congestive heart failure (ICD9
code 427) for patients over 65 years at 134 hospitals.
Average hospitalizations: 39/day. 1986 population of
study area: 12.6 million. Regressions on air quality using
generalized estimating equations, controlling for long-term
trends, seasonality, day of week, and inter-hospital
differences. Models fit monthly and pooled over months.
Log hospitalizations filtered with 19-day moving average
prior to GEE analysis. 0-3 day lags examined. Covariates:
CO, SO2, NO2, O3, temperature, dewpoint temperature.
Sulfate lagged one day significantly assoc. with
total CVD admissions with and without ozone in
the model. Larger associations observed for
coronary artery disease and heart failure than for
cardiac dysrhythmias. Suggestion of larger
associations for males and the sub-population 65
years old and greater. Little evidence for
seasonal differences in sulfate effects after
controlling for covariates.
COH significant in single-pollutant models with
and without weather covariates. Only InCO and
In NO2 significant in multi-pollutant models.
COH highly colinear with CO and NO2.
Suggests no particle effect independent of gases.
However, no gravimetric PM data were included.
Effects computed for 95th percentile
change in SO4
SO4, Id, no covariates:
Total CVD: 2.8(1.8,3.8)
CAD: 2.3(0.7,3.8)
DYS: 1.3 (-2.0, 4.6)
HF: 3.0(0.6,5.3)
Males: 3.4(1.8, 5.0)
Females: 2.0(0.2,3.7)
<65: 2.5(0.5,4.5)
>=65: 3.5(1.9,5.0)
SO4, Id, w. temp and O3:
Total CVD: 3.3 (1.7,4.8)
Effects computed for 95% change in
COH:
0 d lag:
5.5% (2.5, 8.6)
0 d lag w/weather:
4.7% (1.3, 8.2)
0 d lag w/CO, NO2, SO2, O3:
-2.26 (-6.5, 2.2)
O
-------
to
o
o
to
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
Canada (cont'd)
Burnett etal. (1997b)
Metro-Toronto, Canada
1992-1994
Pollutant: mean, median, IQR:
COH (103 In ft): 0.8,0.8,0.6
H+ (nmol/m3): 5, 1, 6
SO4 (nmol/m3): 57, 33, 57
PM10 0/g/m3): 28,25,22
PM25 (,ug/m3): 17, 14, 15
PM10.25 Cug/m3): 12,10,7
Daily unscheduled cardiovascular hospitalizations (ICD9
codes 410-414,427, 428) for all ages. Average hospital
admissions: 42.6/day. Six cities of metro-Toronto
included Toronto, North York, East York, Etobicoke,
Scarborough, and York, with combined 1991 population
of 2.36 million. Used same stat model as in Burnett et al.,
1997c. 0- 4 day lags examined, as well as multi-day
averages. Covariates: O3, NO2, SO2, CO, temperature,
dewpoint temperature.
Relative risks > 1 for all pollutants in univariate
regressions including weather variables; all but
H+ and FP statistically significant. In
multivariate models, the gaseous pollutant
effects were generally more robust than were
particulate effects. However, in contrast to
Burnett et al. (1997A), COH remained
significant in multivariate models. Of the
remaining particle metrics, CP was the most
robust to the inclusion of gaseous covariates.
Results do not support independent effects of FP,
SO4, or H+ when gases are controlled.
Percent excess risk (95% CI) per
50 Mg/m3 PM10, 25 ,ug/m3 PM25 and
PM10_2 5, and IQR for other indicators.
COH: 0-4 d.
6.2(4.0, 8.4)
5.9(2.8, 9.1) w. gases
H+: 2-4 d.
2.4(0.4,4.5)
0.5 (-1.6, 2.7) w. gases
S04: 2-4 d.
1.7 (-0.4, 3.9)
-1.6 (-4.4, 1.3) w. gases
PM10: l-4d.
7.7(0.9, 14.8)
-0.9 (-8.3, 7.1) w. gases
PM25 :2-4d.
5.9(1.8, 10.2)
-1.1 (-7.8, 6.0) w. gases
PM10.25:0-4d.
13.5(5.5,22.0)
8.1 (-1.3, 18.3) w. gases
O
-------
to
o
o
to
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
to
fe
H
6
o
o
H
O
O
H
W
O
Canada (cont'd)
Burnett etal. (1999)
Metro-Toronto, Canada
1980-1994
Pollutant: mean, median, IQR:
FPe!t 0/g/m3): 18,16,10
CPe!t Cug/m3): 12, 10, 8
PM10 ea (Mg/m3): 30,27, 15
Daily hospitalizations for dysrhythmias, DYS (ICD9 code
427; mean 5/day); heart failure, HF (428; 9/d); ischemic
heart disease, IHD (410-414; 24/d); cerebral vascular
disease, CVD (430-438; 10/d); and diseases of the
peripheral circulation, DPC (440-459; 5/d) analyzed
separately in relation to environmental covariates. Same
geographic area as in Burnett et al., 1997b. Three size-
classified PM metrics were estimated, not measured, based
on a regression on TSP, SO4, and COH in a subset of every
6th-day data. Generalized additive models used and non-
parametric LOESS prefilter applied to both pollution and
hospitalization data. Day of week controls. Tested 1-3
day averages of air pollution ending on lags 0-2.
Covariates: O3, NO2, SO2, CO, temperature, dewpoint
temperature, relative humidity.
In univariate regressions, all three PM metrics
were associated with increases in cardiac
outcome (DVS, HF, IHD). No associations with
vascular outcomes, except for CPest with DPC.
In multi-pollutant models, PM effects estimates
reduced by variable amounts (often >50%) for
specific endpoints and no statistically significant
(at p<0.05) PM associations seen with any
cardiac or circulatory outcome (results not
shown). Use of estimated PM metrics limits
interpretation of pollutant-specific results.
However, results suggest that linear combination
of TSP, SO4, and COH does not have a strong
independent association with cardiovascular
admissions when full range of gaseous pollutants
also modeled.
Single pollutant models:
Percent excess risk (95% CI) per
50 ,ug/m3 PM10; 25 ,ug/m3 PM25; and
25 Mg/m3 PM10.2.5.
All cardiac HA (lags 2-5 d):
PM25 1-poll = 8.1(2.45, 14.1)
PM25 w/4 gases = -1.6 (-10.4, 8.2);
w/CO = 4.60 (-3.39, 13.26)
PM10 1-poll = 12.07 (1.43, 23.81)
w/4 gases = -1.40 (-12.53, 11.16)
w/CO= 10.93 (0.11,22.92)
PM10.25 1-poll = 20.46 (8.24, 34.06)
w/4 gases = 12.14 (-1.89, 28.2);
w/CO=19.85(7.19, 34.0)
DYS:
FPest(Od): 6.1 (1.9, 10.4)
CPest(Od): 5.2 (-0.21, 1.08)
PM10 est: (0 d): 8.41 (2.89, 14.2)
HF:
FPe!t(0-2d): 6.59(2.50, 10.8)
CPest (0-2 d): 7.9 (2.28, 13)
PM10 est (0-2 d): 9.7(4.2, 15.5)
IHD:
FPest(0-2d): 8.1 (5.4, 10.8)
CPest(Od): 3.7(1.3,6.3)
PM10est(0-ld): 8.4(5.3, 11.5)
O
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
Canada (cont'd)
Stieb et al. (2000)
Saint John, Canada
7/1/92-3/31/96
mean and S.D.:
PM10 (,ug/m3): 14.0,9.0
PM25(,ug/m3): 8.5, 5.9
HOSPITAL ADMISSIONS
H+(nmol/m3): 25.7,36.8
Sulfate (nmol/m3): 31.1,29.7
COHmean(103lnft): 0.2,0.2
COHmax(103lnft): 0.6,0.5
Study of daily emergency department (ED) visits for
angina/myocardial infarction (mean 1.8/day), congestive
heart failure (1.0/day), dysrhythmia/conduction
disturbance (0.8/day), and all cardiac conditions (3.5/day)
for persons of all ages. Covariates included CO, H2S,
NO2, O3, SO2, total reduced sulfur (TRS), a large number
of weather variables, and 12 molds and pollens. Stats:
generalized additive models with LOESS prefiltering of
both ED and pollutant variables, with variable window
lengths. Also controlled for day of week and LOESS-
smoothed functions of weather. Single-day, and five day
average, pollution lags tested out to lag 10. The strongest
lag, either positive or negative, was chosen for final
models. Both single and multi-pollutant models reported.
Full-year and May-Sep models reported.
In single-pollutant models, significant positive
associations observed between all cardiac ED
visits and PM10, PM25, H2S, O3, and SO2.
Significant negative associations observed with
H+, sulfate, and COH max. PM results were
similar when data were restricted to May-Sep. In
multi-pollutant models, no PM metrics were
significantly associated with all cardiac ED visits
in full year analyses, whereas both O3 and SO2
were. In the May-Sep subset, significant
negative association found for sulfate. No
quantitative results presented for non-significant
variables in these multi-pollutant regressions. In
cause-specific, single-pollutant models, PM
tended to be positively associated with
dysrhythmia/conductive disturbances but
negatively associated with congestive heart
failure (no quantitative results presented). The
objective decision rule used for selecting lags
reduced the risk of data mining; however, the
biological plausibility of lag effects beyond 3-5
days is open to question. Rich co-pollutant data
base. Results imply no effects of PM
independent of co-pollutants.
Percent Excess Risk (p-value)
computed for 50 ,ug/m3 PM10, 25
f/g/m3 PM2 5 and mean levels of
sulfate and COH.
Full year results for all cardiac
conditions, single pollutant models:
PM10: 3d.
29.3 (P=0.003)
PM25: 3d.
14.4(P=0.055)
H+: 4-9 d. avg.
-1.8(0.010)
Sulfate: 4d.
-6.0(0.001)
COH max: 7d.
-5.4(0.027)
Full year results for all cardiac
conditions, multi-pollutant models:
No significant PM associations.
O
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe
Le Tertre et al. (2002)
Eight-City - APHEA 2
Study mean (SD) PM10 ,ug/m3
Barcelona- 1/94-12/96
55.7(18.4)
Birmingham - 3/92-12/94
24.8(13.1)
London - 1/92-12/94
28.4(12.3)
Milan - No PM10
Netherlands - 1/92-9/95
39.5(19.9)
Paris - 1/92-9/96
PM13-22.7 (10.8)
Rome - No PM10
Stockholm - 3/94-12/96
15.5(7.2)
Atkinson etal. (1999a)
Greater London, England
1992-1994
Pollutant: mean, median, 90-10 percentile
range:
PM10 (Mg/m3): 28.5, 24.8, 30.7
Black Smoke (Mg/m3): 12.7,10.8,16.1
Examined the association between measures of PM to
include PM10 and hospital admissions for cardiac causes in
eight European cities with a combined population of
38 million. Examined age factors and ischemic heart
disease and studies also stratified by age using
autoregressive Poisson models controlled for long-term
trends, season, influenza, epidemics, and meteorology, as
well as confounding by other pollutants. In a second
regression examined, pooled city-specific results for
sources of heterogeneity.
Daily emergency hospital admissions for total
cardiovascular diseases, CVD (ICD9 codes 390-459), and
ischemic heart disease, IHD (ICD9 410-414), for all ages,
for persons less than 65, and for persons 65 and older.
Mean daily admissions for CVD: 172.5 all ages, 54.5 <65,
117.8 >65; for IHD: 24.5 <65, 37.6 >65. Covariates: NO2,
O3, SO2, CO, temperature, relative humidity. Poisson
regression using APHEA methodology; sine and cosine
functions for seasonal control; day of week dummy
variables. Lags of 0-3, as well as corresponding multi-day
averages ending on lag 0, were considered.
Pooled results were reported for the cardiac
admissions results in table format. City-specific
and pooled results were depicted in figures only.
Found a significant effect of PM10 and black
smoke on admissions for cardiac causes (all
ages) and cardiac causes and ischemic heart
disease for people over 65 years with the impact
of PM10 per unit of pollution being half that
found in the United States. PM10 did not seem to
be confounded by O3 or SO2. The effect was
reduced when CO was incorporated in the
regression model and eliminated when
controlling for NO2. There was little evidence of
an impact of particles on hospital admissions for
ischemic heart disease for people below 65 years
or stroke for people over 65 years. The authors
state results were consistent with a role for traffic
exhaust/diesel in Europe.
In single-pollutant models, both PM metrics
showed positive associations with both CVD and
IHD admissions across age groups. In Two-
pollutant models, the BS effect, but not the PM10
effect, was robust. No quantitative results
provided for two-pollutant models. Study does
not support a PM10 effect independent of co-
pollutants.
For a 10 ,ug/m3 increase in PM10
Cardiac admissions/all ages
0.5% (0.2, 0.8)
Cardiac admissions/over 65 years
0.7% (0.4, 1.0)
Ischemic heart disease/over 65 years
0.8% (0.3, 1.2)
For cardiac admissions for people
over 65 years: All the city-specific
estimates were positive with London,
Milan, and Stockholm significant at
1he 5% level.
Effects computed for 50 ,ug/m3 PM10
and 25 ^g/m3 BS
PM10 0 d.
All ages:
CVD: 3.2(0.9, 5.5)
0-64 yr:
CVD: 5.6(2.0,9.4)
IHD: 6.8(1.3, 12.7)
65+ yr:
CVD: 2.5 (-0.2, 5.3)
IHD: 5.0(0.8,9.3)
Black Smoke 0 d.
All ages:
CVD: 2.95(1.00,4.94)
0-64 yr:
CVD: 3.12(0.05,6.29)
IHD: 2.78 (-1.88, 7.63)
65+ yr:
CVD: 4.24(1.89,6.64)
IHD (lag 3): 4.57 (0.86, 8.42)
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Prescottetal. (1998)
Edinburgh, Scotland
1981-1995 (BS and SO2)
1992-1995 (PM10, NO2, O3, CO)
Means for long and short series:
BS: 12.3, 8.7
PM10: NA, 20.7
Wordleyetal. (1997)
Birmingham, UK
4/1/92-3/31/94
mean, min, max:
PM10 (,ug/m3): 26,3, 131
Diaz etal. (1999)
Madrid, Spain
1994-1996
TSP by beta attenuation
Summary statistics not given.
Daily emergency hospital admissions for cardiovascular
disease (ICD9 codes 410-414, 426-429, 434-440) for
persons less than 65 years and for persons 65 or older.
Separate analyses presented for long (1981-1995) and
short (1992-1995) series. Mean hospital admissions
for long and short series: <65, 3.5, 3.4; 65+, 8.0, 8.7.
Covariates: SO2, NO2, O3, CO, wind speed, temperature,
rainfall. PM10 measured by TEOM. Stats: Poisson
log-linear regression; trend and seasons controlled by
monthly dummy variables over entire series; day of week
dummy variables; min daily temperature modeled using
octile dummies. Pollutants expressed as cumulative lag 1-
3 day moving avg.
Daily hospital admissions for acute ischemic heart disease
(ICD9 codes 410-429) for all ages. Mean hospitalizations:
25.6/day. Covariates: temperature and relative humidity.
Stats: Linear regression with day of week and monthly
dummy variables, linear trend term. Lags of 0-3
considered, as well as the mean of lags 0-2.
Daily emergency hospital admissions for all cardiovascular
causes (ICD9 codes 390-459) for the Gregorio Maranon
University Teaching Hospital. Mean admissions: 9.8/day.
Covariates: SO2, NO2, O3, temperature, pressure, relative
humidity, excess sunlight. Stats: Box-Jenkins time-series
methods used to remove autocorrelations, followed by
cross-correlation analysis; sine and cosine terms for
seasonality; details unclear.
In long series, neither BS nor NO2 were
associated with CVD admissions in either age
group. In the short series, only 3-day moving
average PM10 was positively and significantly
associated with CVD admissions in single-
pollutant models, and only for persons 65 or
older. BS, SO2, and CO also snowed positive
associations in this subset, but were not
significant at the 0.05 level. The PM10 effect
remained largely unchanged when all other
pollutants were added to the model, however
quantitative results were not given. Results
appear to show an effect of PM10 independent of
co-pollutants.
No statistically significant effects observed for
PM10 on ischemic heart disease admissions for
any lag. Note that PM10 was associated with
respiratory admissions and with cardiovascular
mortality in the same study (results not shown
here).
No significant effects of TSP on CVD reported.
Percent Excess Risk (95% CI):
Effects computed for 50 //g/m3 change
in PM10 and 25 ,ug/m3 change in BS.
Long series:
BS, 1-3d. avg.
<65: -0.5 (-5.4, 4.6)
65+: -0.5 (-3.8, 2.9)
Short series:
BS, 1-3 d. avg.
<65: -9.5 (-24.6, 8.0)
65+: 5.8 (-4.9, 17.8)
PM10, 1-3d. avg.
<65: 2.0 (-12.5, 19.0)
65+: 12.4(4.6,20.9)
% change (95% CI) per
50 ,ug/m3 change PM10
IHD admissions:
PM10 0-dlag:
1.4% (-4.4, 7.2)
PM10 1-dlag:
-1.3% (-7.1, 4.4)
No quantitative results presented for
PM.
O
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
HOSPITAL ADMISSIONS
to
o
o
to
Reference citation. Location, Duration
PM Index, Mean or Median, IQR
Study Description: Health outcomes or codes,
Mean outcome rate, sample or population size,
ages. Concentration measures or estimates.
Modeling methods: lags, smoothing, co-pollutants,
covariates, concentration-response
Results and Comments. Design Issues,
Uncertainties, Quantitative Outcomes
PM Index, Lag, Excess
Risk % (95% LCL, UCL),
Co-Pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
Australia
Morgan etal. (1998)
Sydney, Australia
1990-1994
mean, median, IQR, 90-10 percentile range:
Daily avg. bscat/104m: 0.32, 0.26, 0.23,
0.48
Daily max l-hrbscat/104m: 0.76, 0.57, 60,
1.23
Asia
Wong etal. (1999)
Hong Kong
1994-1995
median, IQR for PM10 (|/g/m3): 45.0, 34.8
Daily hospital admissions for heart disease (ICD9 codes
410, 413, 427, 428) for all ages, and separately for persons
less than 65 and persons 65 or greater. Mean daily
admissions: all ages, 47.2; <65, 15.4; 65+, 31.8. PM
measured by nephelometry (i.e., light scattering), which is
closely associated with PM25. Authors give conversion for
Sydney as PM2 5 =30 x bscat. Covariates: O3, NO2,
temperature, dewpoint temperature. Stats: Poisson
regression; trend and seasons controlled with linear time
trend and monthly dummies; temperature and dewpoint
controlled with dummies for eight levels of each variable;
day of week and holiday dummies. Single and cumulative
lags from 0-2 considered. Both single and multi-pollutant
models were examined.
Daily emergency hospital admissions for cardiovascular
diseases, CVD (ICD9 codes 410-417, 420-438, 440-444),
heart failure, HF (ICD9 428), and ischemic heart disease,
IHD (ICD9 410-414) among all ages and in the age
categories 5-64, and 65+. Median daily CVD admissions
for all ages: 101. Covariates: NO2, O3, SO2, temperature,
relative humidity. PM10 measured by TEOM. Stats:
Poisson regression using the APHEA protocol; linear and
quadratic control of trends; sine and cosine control for
seasonality; holiday and day of week dummies;
autoregressive terms. Single and cumulative lags from 0-5
days considered.
In single-pollutant models, NO2 was strongly
associated with heart disease admissions in all
age groups. PM was more weakly, but still
significantly associated with admissions for all
ages and for persons 65+. The NO2 association
in the 65+ age group was unchanged in
the multi-pollutant model, whereas the PM effect
disappeared when NO2 and O3 were added to the
model.. These results suggest that PM is not
robustly associated with heart disease admissions
when NO2 is included, similar to the sensitivity
of PM to CO in other studies.
In single-pollutant models, PM10, NO2, SO2, and
O3 all significantly associated with CVD
admissions for all ages and for those 65+. No
multi-pollutant risk coefficients were presented;
however, the PM10 effect was larger when O3 was
elevated (i.e., above median). A much larger
PM10 effect was observed for HF than for CVD
or IHD. These results confirm the presence of
PM10 associations with cardiovascular
admissions in single-pollutant models, but do not
address the independent role of PM10.
Percent Excess Risk (95% CI):
Effects computed for 25 //g/m3 PM2
(converted from bscat).
24-hr avg. PM25Od.
<65: 1.8 (-2.9, 6.7)
65+: 4.9(1.6,8.4)
All: 3.9(1.1,6.8)
24-hr PM2 5, 0 d w. NO2 and O3.
65+: 0.12 (-1.3, 1.6)
l-hrPM25, Od.
<65: (X19(-1.6, 2.0)
65+: 1.8(0.5,3.2)
All: 1.3(0.3,2.3)
Percent Excess Risk (95% CI):
Effects computed for 50 ,ug/m3 change
in PM10.
PM10, 0-2 d. avg.
CVD:
5-64: 2.5 (-1.5, 6.7)
65+: 4.1(1.3,6.9)
All: 3.0(0.8,5.4)
HF (PM10, 0-3 d ave.):
All: 26.4(17.1,36.4)
IHD (PM10, 0-3 d ave.):
All: 3.5 (-0.5, 7.7)
O
-------
Appendix 8B.2. PM-Respiratory Hospitalization Studies
April 2002 8B-17 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
TABLE 8B-2. ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation.
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
oo
fe
H
6
o
o
H
O
O
H
W
O
United States
Samet et al, (2000a,b)
Study Period.: 84-95
14 U.S. Cities: Birmingham,
Boulder, Canton, Chicago, Col.
Springs, Detroit, Minn./St. Paul,
Nashville, New Haven, Pittsburgh,
Provo/Orem, Seattle, Spokane,
Youngstown. Mean pop. aged
65+yr per city =143,000
PM10 mean = 32.9 /j,g/m3
PM10IQR = NR
Zanobetti et al. (2000b)
10 U.S. Cities
Hospital admissions for adults 65+ yrs. for
CVD (mean=22.1/day/city), COPD
(mean=2.0/day/city), and Pneumonia
(mean=5.6/day/city) related to PM10, SO2,
O3, NO2, and CO. City-specific Poisson
models used with adjustment for season,
mean temperature (T) and relative humidity
(RH) (but not their interaction), as well as
barometric pressure (BP) using LOESS
smoothers (span usually 0.5). Indicators for
day-of-week and autoregressive terms also
included.
Derived from the Samet et al. (2000a,b)
study, but for a subset of 10 cities. Daily
hospital admissions for total cardiovascular
and respiratory disease in persons aged >65
yr. Covariates: SO2, O3, CO, temperature,
relative humidity, barometric pressure. In
first stage, performed single-pollutant
generalized additive robust Poisson
regression with seasonal, weather, and day
of week controls. Repeated analysis for
days with PM10 less than 50 /-ig/m3 to test
for threshold. Lags of 0-5 d considered, as
well as the quadratic function of lags 0-5.
Individual cities analyzed first. The 10 risk
estimates were then analyzed in several
second stage analyses: combining risks
across cities using inverse variance weights,
and regressing risk estimates on potential
effect-modifiers and pollutant confounders.
PM10 positively associated with all three
hospital admission categories, but city
specific results ranged widely, with less
variation for outcomes with higher daily
counts. PM10 effect estimates not found to
vary with co-pollutant correlation,
indicating that results appear quite stable
when controlling for confounding by
gaseous pollutants. Analyses found little
evidence that key socioeconomic factors
such as poverty or race are modifiers, but
it is noted that baseline risks may differ,
yielding differing impacts for a given RR.
Same basic pattern of results as in Samet
et al. (2000a,b). For distributed lag
analysis, lag 0 had largest effect, lags 1
and 2 smaller effects, and none at larger
lags. City-specific slopes were
independent of percent poverty and
percent non-white. Effect size increase
when data were restricted to days with
PM10 less than 50^g/m3. No multi-
pollutant models reported; however, no
evidence of effect modification by co-
pollutants in second stage analysis.
Suggests association between PM10 and
total respiratory hospital admissions
among the elderly.
COPD HA's for Adults 65+ yrs.
LagOER = 7.4%(CI:5.1,9.8)
Lag 1 ER = 7.5% (CI: 5.3, 9.8)
2 day mean (lagOJagl) ER = 10.3%
(CI: 7.7, 13)
Pneumonia HA's for Adults 65+ yrs.
Lag 0 ER =8.1% (CI: 6.5, 9.7)
LaglER = 6.7%(CI:5.3, 8.2)
2 day mean (lagO, lagl) = 10.3%
(CI: 8.5, 12.1)
Percent excess respiratory risk (95% CI) per
50 Mg/rn3 PM10 increase:
COPD (0-1 d lag) = 10.6 (7.9, 13.4)
COPD (unconstrained dist. lag) = 13.4 (9.4,
17.4)
Pneumonia (0-1 d lag) = 8.1 (6.5, 9.7)
Pneumonia (unconstrained dist. lag) =10.1
(7.7, 12.6)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Jamason et al. (1997)
New York City, NY (82 - 92)
Population = NR
PM10 mean = 38.6 ,ug/rn3
Chen et al. (2000)
Reno-Sparks, NV (90 - 94)
Population = 307,000
B-Gauge PM10 mean=36.5 /-
PM10 IQR = 18.3-44.9 Mg/m
PM10 maximum = 201.3 /-ig/
Gwynn et al. (2000)
Buffalo, NY (5/88-10/90)
PM10 mn./max. = 24.1/90.8 Mg/
PM10IQR = 14.8-29.2 Mg/m3
SO4= mn./max. = 2.4/3.9 Mg/m3
SO4=IQR = 23.5-7.5 Mg/m3
H+ mn/max = 36.4/382 nmol/m3
H+ IQR = 15.7-42.2 nmol/m3
CoH mn/max = 0.2/0.9 10~3 ft.
CoH IQR =0.1-0.3
Gwynn etal. (2001)
New York City, NY
1988,89,90
PM10 37.4 Mg/ni3 mean
Weather/asthma relationships examined
using a synoptic climatological multivariate
methodology. Procedure relates
homogenous air masses to daily counts of
overnight asthma hospital admission.
Log of COPD (mean=1.72/day) and
gastroenteritis (control) admissions from 3
hospitals analyzed using GAM regression,
adjusting for effects of day-of-week,
seasons, Weather effects (T, WS), and long-
wave effects. No co-pollutants considered.
Air pollutant-health effect associations
with total, respiratory, and circulatory
hospital admissions and mortality examined
using Poisson methods controlling for
weather, seasonality, long-wave effects, day
of week, holidays,
Respiratory hospital admissions, race
specific for PM10, H+, O3, SO4=. Regression
model used to model daily variation in
respiratory hospital admissions, day-week,
seasonal, and weather aspects addressed in
modeling.
Air pollution reported to have little role in
asthma variations during fall and winter.
During spring and summer, however, the
high risk categories are associated with
high concentration of various pollutants
(i.e., PM10, SO2,NO2, O3).
PM10 positively associated with COPD
admissions, but no association with
gastroenteritis (GE) diseases, indicating
biologically plausible specificity of the
PM10-health effects association.
Association remained even after excluding
days with PM10 above 150 ^g/m3.
Strongest associations found between
SO4= and respiratory hospital admissions,
while secondary aerosol H+ and SO4=
demonstrated the most coherent
associations across both respiratory
hospital admissions and mortality.
Addition of gaseous pollutants to the
model had minimal effects on the PM RR
estimates. CoH weakness in associations
may reflect higher toxicity by acidic sulfur
containing secondary particles versus
carbonaceous primary particles.
Greatest difference between the white and
non-white subgroups was observed for O3.
However, within race analyses by
insurable coverage suggested that most of
the higher effects of air pollution found
for minorities were related to
socio-economic studies.
NR
COPD All age Admissions
50 Mg/m3 IQR PM10 (single pollutant):
ER = 9.4% (CI: 2.2, 17.1)
Respiratory Hospital Admissions(all ages) PM
Index (using standardized cone, increment)
-Single Pollutant Models
H+ = 75nmoles/m3;COH = 0.5 units/1 000ft
PM10(lag 0) ER = 11% (CI: 4.0, 18)
S04=(lag 0) ER = 8.2% (CI: 4.1, 12.4)
H+(lag 0) ER = 6% (CI: 2.8, 9.3)
CoH(lagO) ER = 3% (CI: - 1.2, 7.4)
PM10 (max-min) increment
1 day lag
white 1.027 (0.971-1.074)
non-white (1.027 (0.988-1.069)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
United States (cont'd)
Jacobs etal. (1997)
Butte County, CA (83 - 92)
Population = 182,000
PM10 mean = 34.3 ,ug/rn3
PM10 mm/max = 6.6 / 636 /-ig/m3
CoH mean = 2.36 per 1000 lin. ft.
CoH mm/max = 0/16.5
Association between daily asthma HA's
(mean = 0.65/day) and rice burning using
Poisson model with a linear term for
temperature, and indicator variables for
season and yearly population.
Co-pollutants were O3 and CO. PM10
estimated for 5 of every 6 days from CoH.
Increases in rice straw bum acreage found
to correlate with asthma HA's over time.
All air quality parameters gave small
positive elevations in RR. PM10 showed
the largest increase in admission risk.
Asthma HA's (all ages)
For an increase of 50 ,ug/ni3 PM10:
ER = 6.11% (not statistically significant)
oo
td
to
o
fe
H
6
o
o
H
O
O
H
W
O
Linn et al. (2000)
Los Angeles, CA (92-95)
Population = NR
PM10 mean = 45.5 /^g/m3
PM10 Min/Max = 5/132 Mg/
Moolgavkar et al. (1997)
Minneapolis-St. Paul 86-91
Populations NR
Birmingham, AL '86-'91
Population. = NR
PM10 mean = 34 ,ug/m3 (M-SP)
PM10IQR =22-41 Mg/m3 (M-SP)
PM10 mean =43.4 ^g/m3(Birm)
PM10 IQR =26-56 Mg/m3(Birm)
Pulmonary hospital admissions (HA's)
(mean=74/day) related to CO, NO2, PM10,
and O3 in Los Angeles using Poisson model
with long-wave, day of week, holidays, and
weather controls.
Investigated associations between air
pollution (PM10, SO2, NO2 O3, and CO) and
hospital admissions for COPD
(mean/day=2.9 in M-SP; 2.3 in Birm) and
pneumonia (mean=7.6 in M-SP; 6.0 in
Birm) among older adults (>64 yrs.).
Poisson GAM's used, controlling for day-
of-week, season, LOESS of temperature
(but neither RH effects nor T-RH
interaction considered).
PM10 positively associated with
pulmonary admissions year-round,
especially in winter. No association with
cerebro-vascular or abdominal control
diseases. However, use of linear
temperature, and with no RH interaction,
may have biased effect estimates
downwards for pollutants here most
linearly related to temperature (i.e., O3
and PM10).
In the M-SP area, PM10 significantly and
positively associated with total daily
COPD and pneumonia admissions among
elderly, even after simultaneous inclusion
of O3. When four pollutants included in
the model (PM10, SO2, O3, NO2), all
pollutants remained positively associated.
In Birm., neither PM10 nor O3 showed
consistent associations across lags. The
lower power (fewer counts) and lack of T-
RH interaction weather modeling in this
Southern city vs. M-SP may have
contributed to the differences seen
between cities.
Pulmonary HA's (>29 yrs.)
PM10 = 50 Mg/m3
(Lag 0)ER = 3.3% (CI: 1.7, 5)
COPD + Pneumonia Admissions (>64yrs.)
In M-SP, For PM10 = 50 /j,g/m3 (max Ig)
ER(lgl) = 8.7%(CI:4.6, 13)
With O3 included simultaneously:
ER(lgl)= 6.9%(95 CI: 2.7, 11.3)
In Birm, For PM10=50 /-ig/m3 (max Ig.)
ER(lgO)=1.5%(CI: -1.5,4.6)
With O3 included simultaneously:
ER(lgO) = 3.2% (CI: -0.7, 7.2)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Nauenberg and Basu (1999)
Los Angeles (9 1-94)
Wet Season =11/1-3/1
Dry Season =5/1 -8/1 5
Population .= 2.36 Million
PM10SE= 17.23,ug/m3
Schwartz et al. (1996b)
Cleveland (Cayahoga County), Ohio
(88 - 90)
PM10 mean = 43 /j.g/m3
PM10IQR = 26-56 ,ug/m3
Zanobetti, et al. (2000a)
Study Period: 86 - 94
Chicago (Cook Count), IL
Population = 633,000 aged 65+
PM10 mean = 33.6 Mg/m3
PM10 range = 2.2, 157.3 Mg/m3
The effect of insurance status on the
association between asthma-related hospital
admissions and exposure to PM10 and O3
analyzed, using regression techniques with
same day and 8-day weighted moving
average levels, after removing trends using
Fourier series. Compared results during wet
season for all asthma HA's (mean = 8.7/d),
for the uninsured (mean=0.77/d), for
MediCal (poor) patients (mean = 4.36/d),
and for those with other private health or
government insurance (mean = 3.62/d).
Review paper including an example drawn
from respiratory hospital admissions of
adults aged 65 yr and older (mean = 22/day)
in Cleveland, OH. Categorical variables for
weather and sinusoidal terms for filtering
season employed.
Analyzed HA's for older adults (65 + yr)
for COPD (mean = 7.8/d), pneumonia
(mean = 25.5/d), and CVD, using Poisson
regression controlling for temperature, dew
point, barometric pressure, day of week,
long wave cycles and autocorrelation, to
evaluate whether previous admission or
secondary diagnosis for associated
conditions increased risk from air pollution.
Effect modification by race, age, and sex
also evaluated.
No associations found between asthma
admissions and O3. No O3 or PM10
associations found in dry season. PM10
averaged over eight days associated with
increase in asthma admissions, with even
stronger increase among MediCal asthma
admissions in wet season. The authors
conclude that low income is useful
predictor of increased asthma
exacerbations associated with air
pollution. Non-respiratory HA's showed
no such association with PM10.
Hospital admissions for respiratory illness
of persons aged 65 yr and over in
Cleveland strongly associated with PM10
and O3, and marginally associated with
SO2 after control for season, weather, and
day of the week effects.
Air pollution- associated CVD HA's were
nearly doubled for those with concurrent
respiratory infections (RI) vs. those
without concurrent RI. For COPD and
pneumonia admissions, diagnosis of
conduction disorders or dysrhythmias
(Dyshr.) increased PM10 RR estimate. The
PM10 RR effect size did not vary
significantly by sex, age, or race, but
baseline risks across these groups differ
markedly, making such sub-population RR
inter-comparisons difficult to interpret.
All Age Asthma HA's
PM10 = 50 Mg/m3, no co-pollutant, during wet
season (Jan. 1 - Mar. 1):
All Asthma Hospital Admissions
0-d lag PM10 ER = 16.2 (CI: 2.0, 30)
8-d avg. PM10 ER = 20.0 (CI: 5.3, 35)
MediCal Asthma Hospital Admissions
8-d avg. PM10 ER = 13.7 (3.9, 23.4)
Other Insurance Asthma HA's
8-davg. PM10ER=6.2(-3.6, 16.1)
Respiratory HA's for persons 65+ years
50 Mg/m3 PM10
ER=5.8%(CI:0.5, 11.4)
PM10 = 50 ^g/m3(average of lags 0,1)
COPD (adults 65+ vrs.)
W/o prior RI. ER = 8.8% (CI: 3.3, 14.6)
With prior RI ER = 17.1% (CI: -6.7, 46.9)
COPD (adults 65+ vrs.)
W/o concurrent Dys. ER = 7.2% (CI: 1.3, 13.5)
With concurrent Dys. ER = 16.5%(CI: 3.2,
31.5)
Pneumonia (adults 65+ vrs.)
W/o pr. Asthma ER =11% (CI: 7.7, 14.3)
Withpr. Asthma ER = 22.8% (CI: 5.1, 43.6)
Pneumonia (adults 65+ vrs.)
W/o pr. Dyshr. ER = 10.4% (CI: 6.9, 14)
With pr. Dyshr. ER = 18.8% (CI: 6.3, 32.7)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
to
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Lippmann et al. (2000)
Detroit, MI ('92-'94)
Population = 2.1 million
PM10 Mean = 31 /-ig/m3
(IQR=19, 38Mg/m3;
max=105,ug/m3)
PM25Mean= IS^g/m3
(IQR= 10, 21 Mg/m3; max=86
Mg/m3)
PM10.25 Mean = 12 /-ig/m3
(IQR= 8, 17 Mg/m3; max=50 M
SO4TVIean = 5 Mg/m3
(IQR=1.8, 6.3,ug/m3;
max=34.5 ^g/m3)
H+ Mean =8.8 nmol/m3 = 0.4
(IQR=0, 7nmol/m3;max=279)
Respiratory (COPD and Pneumonia) HA's
for persons 65 + yr. analyzed, using GAM
Poisson models, adjusting for season, day
of week, temperature, and relative humidity.
The air pollution variables analyzed were:
PM10, PM2 5, PM10.2 5, sulfate, H+, O3, SO2,
NO2, and CO. However, this study
site/period had very low acidic aerosol
levels. As noted by the authors 85% of H+
data was below detection limit (8 nmol/m3).
For respiratory HA's, all PM metrics
yielded RR's estimates >1, and all were
significantly associated in single pollutant
models for pneumonia. For COPD, all
PM metrics gave RR's >1, with H+ being
associated most significantly, even after
the addition of O3 to the regression.
Adding gaseous pollutants had negligible
effects on the various PM metric RR
estimates. The most consistent effect of
adding co-pollutants was to widen the
confidence bands on the PM metric RR
estimates: a common statistical artifact of
correlated predictors. Despite usually
non-detectable levels, H+ had strong
association with respiratory admissions on
the few days it was present. The general
similarity of the PM25 and PM10_25 effects
per Mg/ni3 in this study suggest similarity
in human toxicity of these two inhalable
mass components in study locales/periods
where PM2 5 acidity is usually not present.
Pneumonia HA's for 65+ yrs.
No co-pollutant:
PM10 (50 Mg/m3) Id lag
ER = 22%(CI: 8.3,36)
PM25(25Mg/m3)ldlag:
ER=13%(CI: 3.7,22)
PM25.10(25Mg/m3)ldlag:
ER=12%(CI:0.8,24)
H+ (75 nmol/m3) 3d lag:
ER=12%(CI:0.8,23)
O, co-pollutant (lag 3) also in model:
PM10 (50 Mg/m3) Id lag,
ER = 24% (CI: 8.2, 43)
PM25(25Mg/m3)ldlag:
ER=12%(CI: 1.7,23)
PM25.10(25Mg/m3)ldlag:
ER = 14% (CI: 0.0, 29)
H+ (75 nmol/m3) 3d lag:
ER=11%(CI: -0.9,24)
COPD Hospital Admissions for 65+ yrs.
No co-pollutant:
PM10 (50
3d lag
ER = 9.6%(CI: -5.1,27)
PM25(25,ug/m3)3dlag:
ER = 5.5%(CI: -4.7,17)
PM25.10(25Aig/m3)3dlag:
ER = 9.3%(CI: -4.4,25)
H+ (75 nmol/m3) 3d lag:
ER=13%(CI:0.0,28)
O, co-pollutant (lag 3) also in model:
PM10 (50 Mg/m3) 3d lag,
ER=1.0%(-15,20)
PM25(25Mg/m3)3dlag:
ER = 2.8%(CI: -9.2,16)
PM25.10(25Mg/m3)3dlag:
ER = 0.3%(CI: -14, 18)
H+ (75 nmol/m3) 3d lag:
ER=13%(CI: -0.6,28)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Lumley and Heagerty (1999)
Seattle (King Cty.), WA (87-94)
Population = NR
PM[ daily mean = NR
PM^K) daily mean = NR
From Sheppard et al, 1999:
PM10 mean = 31.5 /-ig/m3
PM10IQR=19-39Mg/m3
Moolgavkar et al. (2000)
King County, WA (87 - 95)
Population = NR
PM10 mean = 30.0 /-ig/m3
PM10IQR =18.9-37.3 Mg/m3
PM2 5 mean =18.1 ,ug/m3
PM2^IQR=10-23,ug/m3
Estimating equations based on marginal
generalized linear models applied to
respiratory HA's for persons <65 yrs. of age
(mean ~ 8/day) using class of variance
estimators based upon weighted empirical
variance of the estimating functions.
Poisson regression used to fit a marginal
model for the log of admissions with linear
temperature, day of week, time trend, and
dummy season variables. No co-pollutants
considered.
Association between air pollution and
hospital admissions (HA's) for COPD
(all age mean=7.75/day; 0-19 yrs.
mean=2.33/day) investigated using Poisson
GAM's controlling for day-of-week,
season, and LOESS of temperature. Co-
pollutants addressed: O3, SO2, CO, and
pollens. PM25 only had one monitoring site
versus multiple sites averaged for other
pollutants.
PM[ at lag 1 day associated with
respiratory HA's in children and younger
adults (<65), but not PM1(ri, suggesting a
dominant role by the submicron particles
in PM2 5-asthma HA associations reported
by Sheppard et al. (1999). 0-day lag PM;
and 0 and 1 day lag PM^^ had RR near 1
and clearly non-significant. Authors note
that model residuals correlated at r=0.2,
suggesting the need for further long-wave
controls in the model (e.g., inclusion of
the LOESS of HA's).
Of the PM metrics, PM10 showed the most
consistent associations across lags (0-4 d).
PM25 yielded the strongest positive PM
metric association at Iag3 days, but gave a
negative association at Iag4 days. That
PM2 5 only had one monitoring site may
have contributed to its effect estimate
variability. Residual autocorrelations (not
reported) may also be a factor. Adding
gaseous co-pollutants or pollens decreased
the PM2 5 effect estimate less than PM10.
Analyses indicated that asthma HA's
among the young were driving the overall
COPD-air pollution associations.
Respiratory HA's for persons <65 vrs. old
PM[ = 25 Mg/m3, no co-pollutant:
1-dlag ER= 5.9 (1.1, 11.0)
COPD HA's all ages (no co-pollutant)
PM10 (50 Mg/m3, lag 2)
ER = 5.1%(CI: 0,10.4)
PM25(25^g/m3,lag3)
ER = 6.4% (CI: 0.9, 12.1)
COPD HA's all ages (CO as co-pollutant)
PM10 (50 Mg/m3, lag 2)
ER = 2.5%(CI: -2.5,7.8)
PM25(25Mg/m3,lag3)
ER = 5.6% (CI: 0.2, 11.3)
o
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Moolgavkar (2000a)
Study Period: 1987-1995
Chicago (Cook County), IL
Population = NR
PM10 median = 35 /j.g/w?
PM10IQR = 25-47 ,ug/m3
Los Angeles (LA County), CA
Population = NR
PM10 median = 44 /-ig/m3
PM10 IQR = 33-59 Mg/m3
PM2 5 median = 22 /-ig/m3
PM2^5IQR=15-3lMg/m3
Phoenix (Maricopa County), AZ
Population = NR
PM10 median = 41 /j.g/w?
PM10 IQR = 32-5 l,ug/m3
Investigated associations between air
pollution (PM10, O3, SO2, NO2, and CO)
and COPD Hospital Admissions (HA's).
PM25 also analyzed in Los Angeles. HA's
for adults >65 yr.: median=12/day in
Chicago, =4/d in Phoenix; =20/d in LA.
In LA, analyses also conducted for children
0-19 yr. (med.=17/d) and adults 20-64
(med.=24/d). Poisson GAM's used
controlling for day-of-week, season, and
splines of temperature and RH (but not their
interaction) adjusted for overdispersion.
PM data available only every 6th day
(except for daily PM10 in Chicago), vs.
every day for gases. Power likely differs
across pollutants, but number of sites and
monitoring days not presented. Two
pollutant models forced to have same lag
for both pollutants. Autocorrelations or
intercorrelations of pollutant coefficients
not presented or discussed.
For >64 adults, CO, NO2 and O3 (in
summer) most consistently associated with
the HA's. PM effects more variable,
especially in Phoenix. Both positive and
negative significant associations for PM
and other pollutants at different lags
suggest possible unaddressed negative
autocorrelation. In LA, PM associated
with admissions in single pollutant
models, but not in two pollutant models.
The forcing of simultaneous pollutants to
have the same lag (rather than maximum
lag), which likely maximizes
intercorrelations between pollutant
coefficients, may have biased the two
pollutant coefficients, but information not
presented.. Analysis in 3 age groups in
LA yielded similar results. Author
concluded that "the gases, other than
ozone, were more strongly associated with
COPD admissions than PM, and that there
was considerable heterogeneity in the
effects of individual pollutants in different
geographic areas".
Most Significant Positive ER
Single Pollutant Models:
COPD HA's (>64vrs.) (50 Mg/m3 PM10):
Chicago: Lag 0 ER =2% (CI: -0.2, 4.3)
LA: Lag 2 ER = 6. 1%(CI: 1.1,11.3)
Phoenix: Lag 0 ER = 6.9% (CI: -4.1, 19.3)
LA COPD HA's
(50
PM10, 25
3 PM2.5 or PM1
(0-19 yrs.): PM10 lg2=10.7%(CI: 4.4, 17.3)
(0-19 yrs.): PM25 lgO=4.3%(CI: -0.1, 8.9)
(0-19 yrs.): PM10.25 lg2=17.1%(CI: 8.9, 25.8)
(20-64 yrs.): PM10 lg2=6.5%(CI: 1.7, 11.5)
(20-64 yrs.): PM25 lg2=5.6%(CI: 1.9, 9.4)
(20-64 yrs.): PM10.2.5 lg2=9%(CI: 3, 15.3)
(> 64 yrs): PM10 Ig2 = 6.1% (1.1, 11.3)
(> 64 yrs): PM2.5 Ig2 = 5.1% (0.9, 9.4)
(>64 yrs.): PM10.25 Ig3=5.1% (CI: -0.4, 10.9)
(>64 yr) 2 Poll. Models (CO = co-poll.)
PM10: Lag 2 ER= 0.6% (CI: -5.1,6.7)
PM25: Lag 2 ER = 2.0% (-2.9, 7.1)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Sheppard et al. (1999)
Seattle, WA, Pop. = NR
1987-1994
PM10 mean = 31.5 ,ug/rn3
PM10IQR= 19-39Aig/m3
PM25 mean =16.7 /j.g/m3
PM25.10 mean = 16.2 ,ug/rn3
PM2^.10IQR=9-21,ug/m3
Friedman etal. (2001)
Atlanta, GA
Summer 1996/control vs. Olympics
PM10 decrease for 36.7 ,ug/m3 to
30.8,ug/m3
Zanobetti and Schwartz (2001)
Cook County, Illinois
1988-1994
PM10: 33 Mg/ni3 median
Janssen et al. (2002)
14 U.S. cities
1985-1994
see Samet et al. (2000a,b)
Daily asthma hospital admissions (HA's)
for residents aged <65 (mean=2.7/day)
regressed on PM10, PM2 5, PM2 5.10, SO2, O3,
and CO in a Poisson regression model with
control for time trends, seasonal variations,
and temperature-related weather effects.
Appendicitis HA's analyzed as a control.
Except O3 in winter, missing pollutant
measures estimated in a multiple imputation
model. Pollutants varied in number of sites
available for analysis, CO the most (4) vs. 2
forPM.
Asthma events in children aged 1 to
16 years were related to pollutant levels
contrasting those during the Summer
Olympics games during a 17 day period to
control periods before and afer the
Olympics.
Respiratory admissions for lung disease in
persons with or without diabetes as a
co-morbidity related to PMIO measures.
Regression coefficients of the relation
between PM10 and hospital admissions for
respiratory disease from Samet et al.
(2000a,b) and prevalence of air
conditioning (AC).
Asthma HA's significantly associated with
PM10, PM25, and PM10_25 mass lagged 1
day, as well as CO. Authors found PM
and CO to be jointly associated with
asthma admissions. Highest increase in
risk in spring and fall. Results conflict
with hypothesis that wood smoke (highest
in early study years and winter) would be
most toxic. Associations of CO with
respiratory HA's taken by authors to be an
index of incomplete combustion, rather
than direct CO biological effect.
Asthma events were reduced during the
Olympic period. A significant reduction
in asthma events was associated with
ozone concentration. The high correlation
between ozone and PM limit the ability to
determine which pollutants may have
accounted for the reduction in asthma
events.
Weak evidence that diabetes modified the
risks of PM10 induced respiratory hospital
admissions while diabetes modified the
risk of PM10 induced COPD admissions in
older people. Found a significant
interaction with hospital admissions for
heart disease and PM with more than
twice the risk in diabetics as in persons
without diabetes.
Regression coefficients of the relation
between ambient PM10 and hospital
admissions for COPD decreased with
increasing percentage of homes with
central AC.
Asthma Admissions (ages 0-64)
PM10 (lag=lday); 50 Mg/m3
ER= 13.7%(CI: 5.5%, 22.6)
PM25 (lag=lday); 25 Mg/m3
ER=8.7%(CI: 3.3%, 14.3)
PM25.10 (lag=lday); 25 Mg/m3
ER= 11.1%(CI:2.8%,20.1)
3 day cumulative exposure PM10
per 10 Mg/m3
1.0(0.80-2.48)
COPD
PM10
10Aig/m3
with diabetes
2.29 (-0.76-5.44)
without diabetes
1.50(0.42-2.60)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
fe
H
6
o
o
H
O
O
H
W
O
Canada
Burnett etal. (1997b)
Toronto, Canada (1992-1994),
Pop. = 4 mill.
PM25 mean =16.8 Mg/ni3
PM2^5 IQR = 8-23 Mg/m3
PM10.2 5 mean =11.6 Mg/m3
PM1
5 IQR = 7-14
PM10 mean = 28.4 ,ug/m3
PM10IQR= 16-38 Mg/m3
CoH mean = 0.8 (per 103 lin. ft.)
CoHIQR = 0.5-1. l(per 103 lin ft)
SO4 mean = 57.1 nmole/m3
SO4 IQR =14-71 nmole/m3
H+ mean = 5 nmole/m3
H+ IQR = 0-6 nmole/m3
Burnett etal. (1999)
Metro-Toronto, Canada
1980-1994
Pollutant: mean, median, IQR:
FPest (Mg/m3): 18,16,10
CPest (Mg/m3): 12,10,8
PM10 est (Mg/m3): 30,27,15
Hospital admissions (HA's) for respiratory
diseases (tracheobronchitis, chronic
obstructive long disease, asthma,
pneumonia) analyzed using Poisson
regression (adjusting for long-term temporal
trends, seasonal variations, effects of short-
term epidemics, day-of-week, ambient
temperature and dew point). Daily particle
measures: PM25, coarse particulate
mass(PM10_25), PM10, SO4, H+, and gaseous
pollutants (O3, NO2, SO2, and CO)
evaluated.
Daily hospitalizations for asthma (493,
mean 1 I/day), obstructive lung disease
(490-492, 496, mean 5/day), respiratory
infection (464, 466, 480-487, 494, mean
13/day) analyzed separately in relation to
environmental covariates. Same geographic
area as in Burnett et al., 1997b. Three size-
classified PM metrics were estimated, not
measured, based on a regression on TSP,
SO4, and COH in a subset of every 6th-day
data. Generalized additive models. Non-
parametric LOESS prefilter applied to both
pollution and hospitalization data. Day of
week controls. Tested 1-3 day averages of
air pollution ending on lags 0-2. Covariates:
O3, NO2, SO2, CO, temperature, dewpoint
temperature, relative humidity.
Positive air pollution-HA associations
found, with ozone being pollutant least
sensitive to adjustment for co-pollutants.
However, even after the simultaneous
inclusion of O3 in the model, the
association with the respiratory hospital
admissions were still significant for PM10,
PM25, PM10.25, CoH,, S04, and H+.
In univariate regressions, all three PM
metrics were associated with increases in
respiratory outcome. In multi-pollutant
models, there were no significant PM
associations with any respiratory outcome
(results not shown). Use of estimated PM
metrics limits the interpretation of
pollutant-specific results reported.
However, results suggest that a linear
combination of TSP, SO4, and COH does
not have a strong independent association
with cardiovascular admissions when a
full range of gaseous pollutants are also
modeled.
Respiratory HA's all ages(no co-pollutant)
PM10 (50 Mg/m3, 4d avg. lag 0)
ER = 10.6% (CI: 4.5-17.1)
PM25 (25 Mg/m3, 4d avg. lag 1)
ER = 8.5%(CI: 3.4,13.8)
PM10.25 (25 Mg/m3, 5d avg. lag 0)
ER=12.5%(CI: 5.2,20.0)
Respiratory HA's all ages(O, co-pollutant)
PM10 (50 Mg/m3, 4d avg. lag 0)
ER = 9.6% (CI: 3.5, 15.9)
PM25 (25 Mg/m3, 4d avg., lag 1)
ER = 6.2% (1.0, 11.8)
PM10.25 (25 Mg/m3, 5d avg. lag 0)
ER = 10.8% (CI: 3.7, 18.1)
Percent excess risk (95% CI) per 50 ,
PM10; 25 Mg/m3 PM25 and PM10.25:
Asthma
PM25 (0-1-2 d): 6.4(2.5,10.6)
PM10 (0-1 d): 8.9 (3.7, 14.4)
PM10.2.5 (2-3-4 d): 11.1(5.8,16.6)
COPD
PM25: 4.8 (-0.2, 10.0)
PM10: 6.9(1.3,12.8)
PM10.25(2-3-4d): 12.8(4.9,21.3)
Resp. Infection:
PM25: 10.8(7.2, 14.5)
PM10: 14.2(9.3,19.3)
PM10.25 (0-1-2 d): 9.3 (4.6, 14.2)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
Canada (cont'd)
Burnett etal. (1997c)
16 Canadian CitiesO 81-91)
Population=12.6 MM
CoHmean=0.64(per 103 lin. ft)
CoH IQR=0.3-0.8(per 103 lin ft)
Burnett etal. (2001)
Toronto, Canada
1980-1994
PM25: 18Mg/m3
PM10.2.5: 16.2 Mg/m3
(both estimated values)
Air pollution data were compared to
respiratory hospital admissions
(mean=l .46/million people/day) for
16 cities across Canada. Used a random
effects regression model, controlling for
long-wave trends, day of week, weather,
and city-specific effects.
Respiratory admissions in children aged
<2 years relates to mean pollution levels.
O3, NO2, SO2, and CO
(ICD-9: 493 asthma; 466 acute bronchitis;
464.4 croup or pneumonia, 480-486).
Time-series analysis adjusted with LOESS.
The 1 day lag of 03 was positively
associated with respiratory admissions in
the April to December period, but not in
the winter months. Daily maximum 1-hr.
CoH from 11 cities and CO also positively
associated with HA's, even after
controlling for O3.
Summertime urban air pollution,
especially ozone, increases the risk that
children less than 2 years of age will be
hospitalized for respiratory disease.
Respiratory HA's all ages (with O,,CO)
CoHIQR = 0.5,lagO:
CoH ER = 3.1% (CI: 1.0-4.6%)
PM25lagO
15.8%(t=3.29)
PM25lagO
with O3 1.4% (0.24)
PM10.2.5 lag 1
18.3%(t=3.29)
with O3 4.5% (0.72)
fe
H
6
o
o
H
O
O
H
W
O
O
Europe
Atkinson et al. (1999b)
London (92 - 94)
Population = 7.2 MM
PM10Mean = 28.5
10th-90thIQR = 15.8-46.5
BS mean = 12.7 Mg/m3
10th-90thIQR = 5.5-21.6 M
All-age respiratory (mean=150.6/day), all-
age asthma (38.7/day), COPD plus asthma
in adults >64 yr. (22.9/day), and lower
respiratory (64.1/day) in adults >64 yr
(16.7/day) hospital admissions in London
hospitals considered. Counts for ages 0-14,
15-64, and >64 yr also examined. Poisson
regression used, controlling for season, day-
of-week, meteorology, autocorrelation,
overdispersion, and influenza epidemics.
Positive associations found between
respiratory-related emergency hospital
admissions and PM10 and SO2, but not for
O3orBS. When SO2 and PM10 included
simultaneously, size and significance of
each was reduced. Authors concluded that
SO2 and PM10 are both indicators of the
same pollutant mix in this city. SO2 and
PM10 analyses by temperature tertile
suggest that warm season effects
dominate. Overall, results consistent with
earlier analyses for London, and
comparable with those for North America
and Europe.
PM10 (50 Mg/ni3), no co-pollutant.
All Respiratory Admissions:
All age (lag Id) ER = 4.9% (CI: 1.8, 8.1)
0-14 y (lag Id) ER = 8.1% (CI: 3.5, 12.9)
15-64y (lag 2d) ER = 6.9% (CI: 2.1, 12.9)
65+ y (lag 3d) ER = 4.9% (CI: 0.8, 9.3)
Asthma Admissions:
All age (lag 3d) ER = 3.4% (CI: -1.8, 8.9)
0-14y(lag3d)ER = 5.4%(CI: -1.2, 12.5)
15-64 y(lag 3d) ER= 9.4% (CI: 1.1, 18.5)
65+y.(lag Od) ER = 12% (CI: -1.8,27.7)
COPD & Asthma Admissions (65+yrs.)
(lag 3d) ER = 8.6% (CI: 2.6, 15)
Lower Respiratory Admissions (65+ yrs.)
(lag 3d) ER = 7.6% (CI: 0.9, 14.8)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
Europe (cont'd)
oo
td
to
oo
fe
H
6
o
o
H
O
O
H
W
O
O
Wordleyetal. (1997)
Study Period: 4/92 -3/94
Birmingham, UK
Population = NR
PM10 daily values:
Mean = 25.6 /j.g/m3
range = 2.8, 130.9 ^g/m3
PM10 3 day running, mean:
Mean = 25.5 /-ig/m3
range = 7.3, 104.7 Mg/m3
Prescott et al. (1998)
Edinburgh (10/92-6/95)
Population = 0.45 MM
PM10 mean. =20.7 /j,g/m3
PM10 min/max=5/72 /-ig/m
PM10 90°% -10°% = 20
McGregor et al. (1999)
Birmingham, UK.
Population = NR
MeanPM,0=30.0ug/m3
Relation between PM10 and total HA's for
respiratory (mean = 21.8/d), asthma
(mn.=6.2/d), bronchitis (mn.=2.4/d),
pneumonia (mn.=3.4/d), and COPD
(mn.=3.2/d) analyzed, using linear
regression after adjusting for day of week,
month, linear trend, RH, and T (but not T-
RH interaction). RR's compared for
various thresholds vs. mean risk of HA.
Poisson log linear regression models used
to investigate relation of daily HA's with
NO2, O3, CO, and PM10. Adjustments made
for seasonal and weekday variation, daily T
(using 8 dummy variables), and wind speed.
Separate analyses for age<65 yr. (mean resp
HA = 3.4/day) and age >64 yr. (mean resp
HA = 8.7/day), and for subjects with
multiple HA's.
A synoptic climatological approach used to
investigate linkages between air mass types
(weather situations), PM10, and all
respiratory hospital admissions (mean=
19.2/day) for the Birmingham area.
PM10 positively associated with all HA's
for respiratory, asthma, bronchitis,
pneumonia, and COPD. Pneumonia, all
respiratory, and asthma HA's also
significantly positively associated with the
mean of PM10 over the past three days,
which gave 10 to 20% greater RR's per 10
Mg/m3, as expected given smaller day to
day deviations. Other air pollutants
examined but not presented, as "these did
not have a significant association with
health outcomes independent from that of
PM10".
The two strongest findings were for
cardiovascular HA's of people aged >64,
which showed a positive association with
PM10 as a mean of the 3 previous days.
PM10 was consistently positively
associated with Respiratory HA's in both
age groups, with the greatest effect size in
those >64, especially among those with
>4 HA's during '81-'95. Weak
significances likely contributed to by low
population size.
Study results show distinct differential
responses of respiratory admission rates to
the six winter air mass types. Two of
three types of air masses associated with
above- average admission rates also favor
high PM10 levels. This is suggestive of
possible linkage between weather, air
quality, and health.
50
in PM1
All Respiratory HA's (all ages)
(lagOd) ER = 12.6% (CI: 5.7, 20)
Asthma HA's (all ages)
(Iag2d) ER = 17.6% (CI: 3, 34.4)
Bronchitis HA's (all ages)
(lagOd) ER= 32.6% (CI: 4.4, 68.3)
Pneumonia HA's (all ages)
(lag3d)ER=31.9%(CI: 15,51.4)
COPD HA's (all ages)
(lagld)ER= 11.5%(CI: -3,28.2)
Single Pollutant Models
PM10 = 50 Mg/m3, mean of lags 1-3
Respiratory HA's (age<65)
ER= 1.25 (-12.8, 17.5)
Respiratory HA's (age>64)
ER= 5.33 (-9.3, 22.3)
Respiratory HA's (age>64, >4 HA's)
ER= 7.93 (-19.0,43.7)
NR
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
to
VO
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Hagen et al. (2000)
Drammen, Sweden( 11794-12/97)
Population = 110,000
PM10 mean = 16.8 Mg/m3
PM10IQR= 9.8-20.9 Mg/m3
Dab etal. (1996)
Paris, France (87 - 92)
Population =6.1 MM
PM13 mean = 50.8 ,ug/rn3
PM13 S^S111 range = 19.0-137.3
BS S^S111 Range =11. 0-123. 3
Anderson etal. (1997)
Amsterdam(77 - 89)
Barcelona ( 86- 92)
London (87-91)
Milan ( 80- 89)
Paris ( 87 - 92)
Rotterdam ( 77 - 89)
Populations .= 0.7(A), 1.7(B),
7.2(L),1.5(M),6.5(P),0.6(R)MM
BS Means = 6, 41, 13, -, 26, 22
TSP Means = 41,155, -, 105, -,41
Examined PM10, SO2, NO2, VOC's, and O3
associations with respiratory hospital
admissions from one hospital (mean =
2.2/day). Used Poisson GAM controlling
for temperature and RH (but not their
interaction), long-wave and seasonality,
day-of-week, holidays, and influenza
epidemics.
Daily mortality and general admissions to
Paris public hospitals for respiratory causes
were considered (means/day: all
resp.=79/d, asthma=14/d, COPD=12/d).
Time series analysis used linear regression
model followed by a Poisson regression.
Epidemics of influenza A and B,
temperature, RH, holidays, day of week,
trend, long-wave variability, and nurses'
strike variables included. No two pollutant
models considered.
All-age daily hospital admissions (HA's)
for COPD considered in 6 APHEA cities;
Mean/day = l.l(A), 11(B), 20(L), 5(M),
11(P), 1.1 (R). Poisson regression
controlling for day of week, holidays,
seasonal and other cycles, influenza
epidemics, temperature, RH, and
autocorrelation. Overall multi-city
estimates made using inverse variance wts.,
allowing for inter-city variance.
As a single pollutant, the PM10 effect was
of same order of magnitude as reported in
other studies. The PM10 association
decreased when other pollutants were
added to the model. However, the VOC's
showed the strongest associations.
For the all respiratory causes category, the
authors found "the strongest association
was observed with PM13" for both hospital
admissions and mortality, indicating a
coherence of association across outcomes.
Asthma was significantly correlated with
NO2 levels, but not PM13.
Ozone gave the most consistent
associations across models. Multi-city
meta-estimates also indicated associations
for BS and TSP. The warm/cold season
RR differences were important only for
ozone, having a much stronger effect in
the warm season. COPD effect sizes
found were much smaller than in U.S.
studies, possibly due to inclusion of non-
emergency admissions or use of less
health-relevant PM indices.
Respiratory Hospital Admissions(all ages)
For IQR=50 Mg/m3
-Single Pollutant Model:
PM10 (lag 0) ER = 18.3% (CI: -4.2, 46)
-Two Pollutant Model (with O3):
PM10 (lag 0) ER = 18.3% (CI: -4.2, 45.4)
-Two Pollutant Model (with Benzene):
PM10 (lag 0) ER = 6.5% (CL-14 ,31.8)
For PM,, = 50 Mg/m3; BS = 25 ug/m3;
Respiratory HA's (all ages):
PM13 Lag 0 ER = 2.2% (CI: 0.2, 4.3)
BSLagOER=1.0%(0.2, 1.8)
COPD HA's (all ages):
PM13 Lag 2 ER = ~2.3% (CI: -6.7, 2.2)
BS Lag 2 ER = -1.1% (-2.9, 0.6)
Asthma HA's (all ages):
PM13 Lg 2 ER = -1.3% (CI: -4.6, 2.2)
BS Lg 0 ER = 1.2% (-0.5, 2.9)
BS (25 Mg/m3) Id lag, no co-pollutant:
All Age COPD Hospital Admissions
ER= 1.7% (0.5, 2.97)
TSP (100 Mg/m3) Id lag, no co-pollutant:
All Age COPD Hospital Admissions
ER = 4.45%(CI: -0.53,9.67)
O
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Diaz etal. (1999)
Madrid (94 - 96)
Population = NR
TSP mean * 40 A/
Spixetal. (1998)
London (L) (87 -91)
Pop. =7.2 Million (MM)
BSMean= 13 A/g/m3
Amsterdam (A) (77 - 89)
Pop. =0.7 MM
BS Mean = 6 ^g/m3
TSP mean = 41 ,ug/m3
Rotterdam (R) (77 - 89)
Pop. =0.6MM
BS Mean = 22 Mg/m3
TSP mean = 41 ,ug/m3
Paris (P) (87 - 92),
Pop.= 6.14MM
BS Mean = 26 Mg/m3
Milano (M) (80 - 89)
Pop. = 1.5 MM
TSP Mean =120
Vigottietal. (1996)
Study Period.: 80-89
Milan, IT
Population =1.5 MM
TSP mean = 139.0,ug/m3
TSP IQR= 82.0, 175.7 Mg
ARIMA modeling used to analyze
emergency respiratory and circulatory
admissions (means/day=7.8,7.6) from one
teaching hospital. Annual, weekly, and 3
day periodicities controlled, but no time
trend included, and temperature crudely fit
with v-shaped linear relationship.
Respiratory (ICD9 460-519) HA's in age
groups 15-64 yr and 65 + yrs. related to
SO2, PM (BS or TSP), O3, and NO2 in the
APHEA study cities using standardized
Poisson models with confounder controls
for day of week, holidays, seasonal and
other cycles, temperature, RH, and
autocorrelation. PM lag considered ranged
from 0-3 day, but varied from city to city.
Quantitative pooling conducted by
calculating the weighted means of local
regression coefficients using a fixed-effects
model when no heterogeneity could be
detected; otherwise, a random-effects model
employed.
Association between adult respiratory HA's
(15-64 yr mean =11.3/day, and 65 + yr
mean =8.8/day) and air pollution evaluated,
using the APHEA protocol. Poisson
regression used with control for weather
and long term trend, year, influenza
epidemics, and season
Although TSP correlated at zero lag with
admissions in winter and year-round, TSP
was never significant in ARIMA models;
so effect estimates not reported for TSP.
Also, found biologically implausible u-
shaped relationship for O3, possibly
indicating unaddressed temperature
effects.
Pollutant associations noted to be stronger
in areas where more than one monitoring
station was used for assessment of daily
exposure. The most consistent finding
was an increase of daily HA's for
respiratory diseases (adults and elderly)
with O3. The SO2 daily mean was
available in all cities, but SO2 was not
associated consistently with adverse
effects. Some significant PM associations
were seen, although no conclusion related
to an overall particle effect could be
drawn. The effect of BS was significantly
stronger with high NO2 levels on the same
day, but NO2 itself was not associated with
HA's. Authors concluded that "there was
a tendency toward an association of
respiratory admissions with BS, but the
very limited number of cities prevented
final conclusions."
Increased risk of respiratory HA was
associated with both SO2 and TSP. The
relative risks were similar for both
pollutants. There was no modification of
the TSP effect by SO2 level. There was a
suggestion of a higher TSP effect on
hospital admissions in the cool months.
N/A
Respiratory Admissions (BS = 25 ug/
BS (L, A, R, P)
15-64 yrs: 1.4% (0.3, 2.5)
65+yrs: 1.0% (-0.2, 2.2)
TSP(A,R,M)(100Mg/m3)
15-64 yrs: 2.0 (-2.1, 6.3)
65+yrs: 3.2 (-1.2, 7.9)
Respiratory HA's
BS (L, A, R, P): Warm (25 Mg/m3)
15-64 yrs: -0.5% (-5.2, 4.4)
65+yrs: 3.4% (-0.1, 7.1)
BS (L, A, R, P): Cold (25 Mg/m3)
15-64 yrs: 2,0% (0.8, 3.2)
65+yrs: 0% (-2.2, 2.3)
TSP (A, R, M): Warm (100 Mg/m3)
15-64 yrs: 6.1% (0.1, 12.5)
65+yrs: 2.0% (-3.9, 8.3)
TSP (A, R, M): Cold (100 Mg/m3)
15-64 yrs: -5.9% (-14.2, 3.2)
65+yrs: 4.0% (-0.9, 9.2)
Young Adult (15-64 yrs.) Resp. HA's
100 Mg/ni3 increase in TSP
Lag 2 ER = 5% (CI: 0, 10)
Older Adult (65+ yrs.) Resp. HA's
100 Mg/ni3 increase in TSP
Lag 1 ER = 5% (CI: -1,10)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Anderson et al. (1998)
London (87 - 92)
Population = 7.2 MM
BS daily mean = 14.6 ,ug/rn3
BS25-75thIQR = 24-38
Kontosetal. (1999)
Piraeus, Athens GR (87 - 92)
Population = NR
BS mean =46.5 Mg/m3
BS max =200 ,ug/m3
Ponce de Leon et al. (1996)
London (4/87-2/92)
Population = 7.3 million
BS mean. =14.6 Mg/ni3
BS S^-gS111 %=6 -
Poisson regression used to estimate the RR
of London daily asthma hospital admissions
associated with changes in O3, SO2, NO2
and particles (BS) for all ages and for 0-14
yr. (mean=19.5/d), 15-64 yr. (mean=13.1/d)
and 65 + yr. (mean =2.6/d). Analysis
controlled for time trends, seasonal factors,
calendar effects, influenza epidemics, RH,
temperature, and auto-correlation.
Interactions with co-pollutants and
aeroallergens tested via 2 pollutant models
and models with pollen counts (grass, oak
and birch).
Relation of respiratory HA's for children
(0-14 yrs.) (mean = 4.3/day) to BS, SO2,
NO2, and O3 evaluated, using a
nonparametric stochastic dynamical system
approach and frequency domain analyses.
Long wave and effects of weather
considered, but non-linearity and
interactions of T and RH relation with HA's
not addressed.
Poisson regression analysis of daily counts
of HA's (means/day: all ages=125.7; Ages
0-14=45.4; Ages 15-64=33.6; Ages
65+=46.7). Effects of trend, season and
other cyclical factors, day of the week,
holidays, influenza epidemic, temperature,
humidity, and
autocorrelation addressed. However,
temperature modeled as linear, with no RH
interaction. Pollution variables were BS,
SO2, O3, andNO2, lagged 0-3 days.
Daily hospital admissions for asthma
found to have associations with O3, SO2,
NO2, and particles (BS), but there was
lack of consistency across the age groups
in the specific pollutant. BS association
was strongest in the 65 + group, especially
in winter. Pollens not consistently
associated with asthma HA's, sometimes
being positive, sometimes negative. Air
pollution associations with HA's not
explained by airborne pollens in
simultaneous regressions, and there was
no consistent pollen-pollutant interaction.
Pollution found to explain significant
portion of the HA variance. Of pollutants
considered, BS was consistently among
most strongly explanatory pollutants
across various reported analyses.
O3 associated with increase in daily HA's,
especially in the "warm" season.
However, u-shape of the O3 dose-response
suggests that linear temperature control
was not adequate. Few significant
associations with other pollutants, but
these tended to be positive (especially in
cold season, Oct-March, and for older
individuals for BS).
Asthma Admissions. BS=25 Mg/m3
BS Lag = 0-3 day average concentration
All age ER= 5.98% (0.4, 11.9)
<15yr. ER = 2.2% (-4.6, 9.5)
15-64yrER= 1.2% (-5.3, 8.1)
65+ yr. ER = 22.8% (6.1, 42.5)
BS=50 Mg/m3, 2d lag & co-pollutant:
Older Adult (>64 yrs.) Asthma Visits:
BS alone: ER = 14.6% (2.7, 27.8)
&O3: ER = 20.0% (3.0, 39.8)
&N02: ER= 7.4% (-8.7, 26.5)
SO2: ER= 11.8% (-2.2, 27.8)
NR
Respiratory HA's (all ages)
Single Pollutant Models
For Oct-Mar. BS = 25 ,ug/m3
Lag 1ER = 0.2% (-1.9, 2.3)
For Apr-Sep. BS = 25 ,ug/m3
Lag 1ER =-2.7% (-6.0, 0.8)
Respiratory HA's (>65)
Single Pollutant Models
For Oct-Mar. BS = 25 Mg/m3
Lag 2 ER= 1.2% (-2.1,4.5)
For Apr-Sep. BS = 25 Mg/m3
Lag 2 ER = 4.5% (-1.0, 10.4)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
OJ
to
Europe (cont'd)
Schouten et al. (1996)
Amsterdam/Rotterdam (77 - 89)
Amsterdam Pop. = 0.69 Million
Rotterdam Pop. = 0.58 Million
Amsterdam, NE
BSmean. =11 Mg/ni3
BS S^-gS^/o =1-37 ,ug/m3
Rotterdam, NE
BS mean. =26 Mg/m3
BS5th-95tll%=6-6lMg/m3
Daily emergency HA's for respiratory
diseases (ICD 460-519), COPD (490-492,
494, 496), and asthma (493). The mean
HA/d (range) for these were: 6.70 (0-23),
1.74 (0-9) and 1.13 (0-7) respectively in
Amsterdam and 4.79 (0-19), 1.57 (0-9),
and 0.53 (0-5) in Rotterdam. HA
associations with BS, O3, NO2, and SO2
analyzed, using autoregressive Poisson
regression allowing for overdispersion and
controlling for season, day of week,
meteorological factors, and influenza
epidemics.
BS did not show any consistent effects in
Amsterdam; but in Rotterdam BS was
positively related to HA's. Most
consistent BS associations in adults >64
yrs. in winter. Positive O3 association in
summer in people aged >64 in Amsterdam
and Rotterdam. SO2 and NO2 did not
show any clear effects. Results not
changed in pollutant interaction analyses.
The authors concluded short-term air
pollution-emergency HA's association is
not always consistent at these individual
cities' relatively low counts of daily HA's
and low levels of air pollution. Analyses
for all ages of all the Netherlands gave a
strong BS-HA association in winter.
Single Pollutant Models
For BS=25 Mg/m3, 2 day lag
For all of the Netherlands:
Respiratory HA's (all ages)
Winter:
ER = 2.0% (-1.5, 5.7)
Summer:
ER = 2.4% (0.6, 4.3)
fe
H
6
o
o
H
O
O
H
W
O
o
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Sunyeretal. (1997)
Barcelona (86 - 92)
Population = NR
BS Median: 40 Mg/m3
BS Range: 11-258 (B
Helsinki (86 - 92)
Population = NR
BS Median: -
BS Range: -
Paris (86 - 92)
Population = NR
BS Median: 28 Mg/m3
BS Range: 4-186Mg/m3
London (86 - 92)
Population = NR
BS Median: IS^g/m3
BS Range: 3-95 ,ug/m3
Teniasetal(1998)
Study Period.: 94-95
Valencia, Spain
Hosp. Cachment Pop. =200,000
BS mean = 57.7 Mg/m3
BSIQR = 25.6-47.7 Mg/m3
Evaluated relations of BS, SO2, NO2, and
O3 to daily counts of asthma HA's and ED
visits in adults [ages 15-64 years: mean/day
= 3.9 (B); 0.7 (H); 13.1 (H); 7.3 (P)] and
children [ages < 15 years: mean/day = 0.9
(H); 19.8 (L); 4.6 (P)]. Asthma
(ICD9=493) studied in each city, but the
outcome examined differed across cities:
ED visits in Barcelona; emergency hospital
asthma admissions in London and Helsinki,
and total asthma admissions in Paris.
Estimates from all cities obtained for entire
period and also by warm or cold seasons,
using Poisson time-series regression,
controlling for temperature and RH, viral
epidemics, day of week effects, and
seasonal and secular trends. Combined
associations were estimated using meta-
analysis.
Associations between adult (14+ yrs.)
emergency asthma ED visits to one city
hospital (mean =1.0/day) and BS, NO2, O3,
SO2 analyzed, using Poisson auto-
regressive modeling, controlling for
potential confounding weather and time
(e.g., seasonal) and trends using the
APHEA protocol.
Daily admissions for asthma in adults
increased significantly with increasing
ambient levels of NO2, and positively (but
non-significantly) withBS. The
association between asthma admissions
and pollution varied across cities, likely
due to differing asthma outcomes
considered. In children, daily admissions
increased significantly with SO2 and
positively (but non-significantly) with BS
and NO2, though the latter only in cold
seasons. No association observed in
children for O3. Authors concluded that
"In addition to particles, NO2 and SO2 (by
themselves or as a constituent of a
pollution mixture) may be important in
asthma exacerbations".
Association with asthma was positive and
more consistent for NO2 and O3 than for
BS or SO2. Suggests that secondary
oxidative-environment pollutants may be
more asthma relevant than primary
reduction-environment pollutants (e.g.,
carbonaceous particles). NO2 had greatest
effect on BS in co-pollutant models, but
BS became significant once 1993 was
added, showing power to be a limitation of
this study.
ER per 25 Mg/m3 BS (24 h Average)
Asthma Admissions/Visits:
<15 yrs.:
London ER=1.5% (IgOd)
Paris ER= 1.5%(lg2d)
Total ER=1.5%(-1.1,4.1)
15-64 yrs:
Barcelona ER = 1.8% (Ig 3d)
London ER = 1.7% (Ig Od)
Paris ER= 0.6% (IgOd)
Total ER= 1.0% (-0.8, 2.9)
Two Pollutant (per 25 ug/m3 BS)
Asthma Admissions (24 h Avg)
<15yrs, (BS&NO2):
London ER = 0.6% (Ig Od)
ParisER = 2.9%(lg2d)
Total ER= 1.8% (-0.6, 4.3)
<15 yrs, (BS & SO2):
London ER = -1.1 % (Ig Od)
ParisER=-1.4%(lg2d)
Total ER=-1.3 (-5.0, 2.5)
15-64 yrs, (BS&NO2):
Barcelona ER = 1.5% (Ig Od)
London ER = -4.7% (Ig Od)
ParisER=-0.7%(lgld)
Total ER = -0.5% (-5.1, 4.4)
Adult Asthma HA's, BS = 25 Mg/m3
For 1993-1995:
Lag OER= 10.6% (0.9, 21.1)
For 1994-1995:
Lag OER = 6.4% (-4.8, 18.8)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Anderson et al. (2001)
West Midland, England
(October 1994-December 1996)
Population = 2.3 million
PM10 mean = 23.3 Mg/m3
PM25 mean = 14.5 /j.g/m3
PM10.2.5 = 9.0 Mg/m3
(by subtraction)
Respiratory hospital admissions (mean =
66/day) related to PM10, PM25, PM10.25, BS,
SO4", NO2, O3, SO2, CO. Regression with
quasilikelihood approach controlling for
seasonal patterns, temp, humidity, influenza
episodes, day week. Adjusted for residual
serial correlation and over-dispersion.
Respiratory admissions (all ages) not
associated with any pollutant. Analyses
by age revealed some associations to PM10
and PM2 5 and respiratory admissions in
the 0-14 age group. There was a striking
seasonal interaction in the cool season
versus the warm season. PM10_25 effects
cannot be excluded. Two pollutant
models examined particulate measures.
PM2 5 effects reduced by inclusion of
black smoke.
Respiratory HA - lag 0+1 days
PM,n IncrementlO-90% (11.4-38.3 Mg
All ages: 1.5 (-0.7 to 3.6)
Ages 0-14: 3.9 (0.6 to 7.4)
Ages 15-64: 0.1 (-4.0 to 4.4)
Ages>65: -1.1 (-4.3 to 2.1)
PM, ,(6.0-25.8)
All ages: 1.2 (-0.9 to 3.4)
Ages 0-14: 3.4 (-0.1 to 7.0)
Ages 15-64: -2.1 (-6.4 to 2.4)
Ages>65: -1.3 (-4.7 to 2.2)
PM10.25(4.1tol5.2)
All ages: 0.2 (-2.5 to 3.0)
Ages 0-14: 4.4 (-0.3 to 9.4)
Ages 15-64: -4.9 (-9.9 to 0.4)
Ages>65: -1.9 (-6.0 to 2.5)
COPD (ICD-9 490-492, 494-496)
EMiou
Age>65: -1.8 (-6.9 to 3.5)
PM,.
Age>65: -3.9 (-9.0 to 1.6)
EMio-2,.
Age>65: -1.7 (-8.9 to 5.3)
Asthma (ICD- 9-493) (mean lag 0+1)
PM10
Ages 0-14: 8.3 (1.7 to 15.3)
Ages 15-64: -2.3 (-10.0 to 6.1)
EM.2.5.
Ages 0-14: 6.0 (-0.9 to 13.4)
Ages 15-64: -8.4 (-16.4 to 0.3)
PM10.25
Ages 0-14: 7.1 (-2.1 to 17.2)
Ages 15-64: -10.7 (-19.9 to-0.5)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Atkinson et al. (2001) Eight city
study: Median/range
Barcelona 1/94 - 12/96
PM10 53.3,ug/m3(17.1 - 131.7)
Birmingham 3/92-12/94
PM10 21.5Mg/m3(6.5 - 115)
London 1/92 - 12/94
PM10 24.9 ,ug/m3 (7.2 - 80.4)
Milan -No PM10
Netherlands 1/92 - 9/95
PM10 33.4 Mg/m3(l 1.3- 130.8)
Paris 1/92 - 9/96
PM1020.lMg/m3(5.8- 80.9)
Rome - No PM10
Stockholm 3/94 - 12/96
PM1013.6A/g/m3(4.3-43.3)
Thompson et al. (2001) Belfast,
Northern Ireland 1/1/93 - 12/31/95.
PM10 Mg/m3 mean (SD)
May-October24.9 (13.7)
November-April 31.9 (24.3)
Fusco et al. (2001) Rome, Italy
1995-1997
PM - suspended particles measured
As part of the APHEA 2 project,
association between PM10 and daily counts
of emergency hospital admissions for
Asthma (0-14 and 15-64 yrs), COPD and
all-respiratory disease (65+ yrs) controlling
for environmental factors and temporal
patterns were studied.
The rates of acute asthma admission to
children's emergency was studied in
relation to day-to-day fluctuation of PM10
and other pollutants using Poisson
regression.
Daily counts of hospital admissions for total
respiratory conditions, acute respiratory
infection including pneumonia, COPD, and
asthma was analyzed in relation to PM
measures and gaseous pollutants using
generalized additive models controlling for
mean temperature, influenza, epidemics,
and other factors.
This study reports that PM was associated
with daily admissions for respiratory
disease in a selection of European cities.
Average daily ozone levels explained a
large proportion of the between-city
variability in the size of the particle effect
estimates in the over 65 yr age group. In
children, the particle effects were
confounded with NO2 on a day-to-day
basis.
A weak, but significant association
between PM10 concentration and asthma
emergency-department admissions was
seen. After adjusting for multiple
pollutants only the benzene level was
independently associated with asthma
emergency department admission.
Benzene was highly correlated to PM10,
SO2 and NO2 levels.
No effect was found for PM. Total
respiratory admissions were significantly
associated with same-day level of NO2 and
CO. There was no indication that the
effects of air pollution were present at lags
>2 days. Among children, total
respiratory and asthma admissions were
strongly associated with NO2 and CO.
Multipollutant model analysis yielded
weaker and more unstable results.
For 10 Mg/rn3 increase
Asthma Admission Age 0-14 yrs:
PM10 for cities ranged from -0.9% (-2.1, 0.4) to
2.8% (0.8, 4.8) with an overall effect estimate
of 1.2% (0.2-2.3)
Asthma Admission Age 15-64 yrs:
Overall PM 1.1% (0.3- 1.8)
Admission of COPD and Asthma Age 65+
years:
Overall PM 1.0% (0.4 - 1.5)
Admission All Respiratory Disease Age 65+
years:
Overall PM 0.9% (0.6 -1.3)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Hrubaetal. (2001)
(Central Slovakia) (1996)
TSP 87 Mg/m3
Latin America
Bragaetal. (1999)
Sao Paulo, Brazil (92 - 93)
Population = NR
PM10 mean = 66.3 ,ug/rn3
PM10 Std. Deviation = 26.1
PM,0 Min./Max. = 26.7/165.4
Gouveia and Fletcher (2000)
Study Period. 92-94
Sao Paulo, Brazil
Population =9.5 MM x 66%
PM10 mean = 64.9 Mg/m3
PM10IQR = 42.9-75.5 Mg/m3
PM1010/90th%=98.lMg/m3
PM1095th%=131.6Mg/m3
Logistic regression modeled TSP exposure
and hospital admission for asthma,
bronchitis, or pneumonia in children, ages
7-11 years, N=667.
Pediatric (<13 yrs.) hospital admissions
(mean=67.6/day) to public hospitals serving
40% of the population were regressed
(using both Poisson and maximum
likelihood methods) on air pollutants,
controlling for month of the year, day-of-
week, weather, and the daily number of
non-respiratory admissions
(mean=120.7/day). Air pollutants
considered included PM10, O3, SO2, CO,
andNO2.
Daily public hospital respiratory disease
admissions for children (mean resp. < 5y =
56.1/d; mean pneumonia <5y =40.8/d;
mean asthma <5 y = 8.5/d; mean
pneum.
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Latin America (cont'd)
Rosas etal. (1998)
SW Mexico City (1991)
Population = NR
PM10 mean. =77 /j.g/m3
PM10 min/max= 25/183
Morgan etal. (1998)
Sydney, AU (90 - 94)
Population = NR
PM25 24 h mean = 9.6 A(g/m3
PM2^5 10th-90th% = 3.6-18 Mg/
PM25 max-1 h mean = 22.8 /-i
PM2^5 10th-90th% = 7.5-44.4 M
Log-regression analysis of relations
between emergency hospital admissions for
asthma for children <15 yrs
(mean=2.5/day), adults (mean=3.0/day),
and adults >59 yrs (mean=0.65/day) and lag
0-2 d pollen, fungal spores, air pollutants
(O3, NO2, SO2, and PM10) and weather
factors. Long wave controlled only by
separating the year into two seasons: "dry"
and "wet". Day-of-week not included in
models.
A Poisson analysis, controlled for
overdispersion and autocorrelation via
GEE, of asthma (means: 0-14
yrs.=15.5/day; 15-64=9/day)), COPD (mean
65+yrs =9.7/day), and heart disease HA's.
PM25 estimated from nephelometry.
Season and weather controlled using
dummy variables.
Few statistical associations were found
between asthma admissions and air
pollutants. Grass pollen was associated
with child and adult admissions, and
fungal spores with child admissions.
Authors conclude that aeroallergens may
be more strongly associated with asthma
than air pollutants, and may act as
confounding factors in epidemiologic
studies. Results are limited by low power
and the lack of long-wave auto-correlation
controls in the models.
Childhood asthma was primarily
associated with NO2, while COPD was
associated with both NO2 and PM. 1 -hr.
max PM25 more consistently positively
related to respiratory HA's than 24-h avg
PM25. Adding all other pollutants
lowered PM effect sizes, although
pollutant inter-correlations makes many
pollutant model interpretations difficult.
No association found between asthma and
O3 or PM. The authors cited the error
introduced by estimating PM2 5 and the
low PM levels as possible reasons for the
weak PM-respiratory HA associations.
NR
Asthma HA's
Single Pollutant Model:
For24hrPM25 = 25Mg/m3
1-14 yrs.(lagl) ER = -1.5% (CI: -7.8, 5.3)
15-64 yrs.(lagO) ER = 2.3% (CI: -4, 9)
ForlhPM25=25Mg/m3
1-14 yrs.(lagl)ER = + 0.5% (CI: -1.9, 3.0)
15-64 yrs.(lagO) ER = 1.5%(CI: -0.9,4)
Multiple Pollutant Model:
1-14 yrs.(lagl) ER = -0.6% (CI: -7.4, 6.7)
COPD (65+yrs.)
Single Pollutant Model:
(lag 0) ER =4.2% (CI: -1.5, 10.3)
(lag 0) ER = 2% (CI: -0.3, 4.4)
Multiple Pollutant Model:
(lag 0) ER = 1.5% (CI: -0.9, 4)
-------
to
o
o
to
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
OJ
oo
fe
H
6
o
o
H
O
O
H
W
O
Tanakaetal. (1998)
StdyPd.: 1/92-12/93
Kushiro, Japan
Pop. = 102 adult asthmatics
PM10 mean = 24.0 Mg/m3
PM10IQR = NR
Wong etal. (1999)
Study Period.: 94-95
Hong Kong
Population = NR
PM10 mean = 50.1 /-ig/
PM10 median = 45.0 Mg
PM10 IQR =30.7, 65.5
g/m3
Associations of HA's for asthma (in 44
non-atopic and 58 atopic patients) with
weather or air pollutants (NO, NO2,
SO2,PM10, O3, and acid fog) evaluated.
Odds ratios (OR) and 95% CFs calculated
between high and low days for each
environmental variable. Poisson regression
was performed for the same dichototomized
variables.
Poisson regression analyses were applied to
assess association of daily NO2, SO2, O3,
and PM10 with emergency HA's for all
respiratory (median = 13I/day) and COPD
(median = 101/day) causes. Effects by age
groups (0-4, 5-64, and 65+ yrs.) also
evaluated. Using the APHEA protocol,
models accounted for time trend, season
and other cyclical factors, T, RH,
autocorrelation and overdispersion. PM10
measured by TEOM, which likely
underestimates mass.
Only the presence of acid fog had a
significant OR >1.0 for both atopies and
non-atopies. PM10 associated with a
reduction in risk (OR<1.0) for both
atopies and non-atopies. Poisson
regression gave a non-significant effect by
PM10 on asthma HA's. However, no long- Poisson Coefficient for PM10 > 30
For same-day (lag=0) PM10
Adult Asthma HA's
OR for <30 vs. >30 ,ug/m3PM10:
Non-atopic OR = 0.77 (CI: 0.61, 0.98)
Atopic OR = 0.87 (CI: 0.75, 1.02)
wave or serial auto-correlation controls
applied, so the opposing seasonalities of
PM vs. HA's indicated in time series data
plots are likely confounding these results.
Positive associations were found for HA's
for all respiratory diseases and COPD with
all four pollutants. PM10 results for lags
0-3 cumulative. Admissions for asthma,
pneumonia, and influenza were associated
with NO2, O3, and PM10. Those aged > or
= 65 years were at higher risk, except for
PM10. No significant respiratory HA
interactions with PM10 effect were found
for high NO2, high O3, or cold season.
Non-atopic B = -0.01 (SE = 0.15)
Atopic B = -0.002 (SE = 0.09)
g/m3 (Lags = 0-3 days)
Respiratory HA's
Allage: ER = 8.3%(CI: 5.1, 11.5)
0-4yrs.: ER = 9.9% (CI: 5.4, 14.5)
5-64yrs.: ER = 8.8% (CI: 4.3, 13.4)
65+yrs.: ER= 9.3%(CI: 5.1, 13.7)
Asthma HA's (all ages)
ER= 7.7% (1.0, 14.9)
COPD HA's (all ages)
ER= 10.0% (5.6, 14.3)
Pneumonia and Influenza HA's (all ages)
ER= 13.1% (7.2, 19.4)
O
-------
Appendix 8B.3: PM-Respiratory Visits Studies
April 2002 8B-39 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
to
o
o
TABLE 8B-3. ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
United States
Choudhury et al. (1997)
Anchorage, Alaska (90 - 92)
Population = 240,000
PM10 mean = 41.5 ,ug/rn3
PM10(SD) = 40.87
PM10 maximum=565 /j.g/w?
Lipsettetal. (1997)
Santa Clara County, CA
Population = NR
(Winters 88 - 92)
PM10 mean = 61.2 /-ig/m3
PM10 Min/Max = 9/165 /j,g/m3
Norrisetal. (1999)
Seattle, WA (9/95-12/96)
Pop. Of Children <18= 107,816
PM10 mean. =21.7 Mg/m3
PM10IQR=11.6Mg/m3
asp mean = 0.4 nrl/10~4
(«12.0Mg/m3PM25)
(=9.5Mg/m3PM25)
Using insurance claims data for state
employees and dependents living in
Anchorage, Alaska, number of daily
medical visits determined for asthma (mean
= 2.42/day), bronchitis, and upper
respiratory infections. Used linear
regression, including a time-trend variable,
crude season indicator variables (i.e.,
spring, summer, fall, winter), and a variable
for the month following a volcanic eruption
in 1992.
Asthma emergency department (ER) visits
from 3 acute care hospitals (mean=7.6/day)
related to CoH, NO2, PM10, and O3 using
Poisson model with long-wave, day of
week, holiday, and weather controls
(analysis stratified by minimum T). Every
other day PM10 estimated from CoH.
Residential wood combustion (RWC)
reportedly a major source of winter PM.
Gastro-enteritis (G-E) ER admissions also
analyzed as a control disease.
The association between air pollution and
childhood (<18 yrs.) ED visits for asthma
from the inner city area with high asthma
hospitalization rates (0.8/day, 23/day/10K
persons) were compared with those from
lower hospital utilization areas(l. I/day,
8/day/10K persons). Daily ED counts were
regressed against PM10, light scattering
(asp), CO, SO2, andNO2 using a
semiparametric Poisson regression model
evaluated for over-dispersion and auto-
correlation.
Positive association observed between
asthma visits and PM10. Strongest
association with concurrent-day PM10
levels. No co-pollutants considered.
Temperature and RH did not predict visits,
but did interact with the PM10 association.
Morbidity relative risk higher with respect
to PM10 pollution during warmer days.
Consistent relationships found between
asthma ER visits and PM10, with greatest
effect at lower temperatures. Sensitivity
analyses supported these findings. NO2
also associated, but in simultaneous
regressions only PM10 stayed associated.
ER visits for gastroenteritis not
significantly associated with air pollution.
Results demonstrate an association
between wintertime ambient PM10 and
asthma exacerbations in an area where
RWC is a principal PM source.
Associations found between ED visits for
asthma in children and fine PM and
CO. CO and PM10 highly correlated with
each other (r=.74) and K, an indicator of
woodsmoke pollution. There was no
stronger association between ED visits for
asthma and air pollution in the higher
hospital utilization area than in the lower
utilization area in terms of RR's.
However, considering baseline risks/1 OK
population indicates a higher PM
attributable risk (AR) in the inner city.
Asthma Medical Visits (all ages):
For mean = 50 Mg/m3 PM10 (single poll.)
Lag = 0 days
ER = 20.9%(CI: 11.8,30.8)
Asthma ED Visits (all ages)
PM10 = 50 Mg/m3 (2 day lag):
At20°F,ER=34.7%(CI: 16,56.5)
At30°F,ER = 22%(CI: 11,34.2)
At 41 ° F, ER = 9.1% (CI: 2.7, 15.9)
Children's (<18 yrs.) Asthma ED Visits
Single Pollutant Models:
24h PM10 =50 Mg/m3
Lagl ER= 75.9% (25.1, 147.4)
For25Mg/m3PM25
Lagl ER = 44.5% (CI: 21.7, 71.4)
Multiple Pollutant Models:
24h PM10 =50 Mg/m3
Lagl ER= 75.9%(CI: 16.3, 166)
For25Mg/m3PM25
Lagl ER= 51.2% (CI: 23.4, 85.2)
-------
to
o
o
to
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
United States (cont'd)
Norris et al. (2000)
Spokane, WA (1/95 - 3/97)
Population = 300,000
PM10 mean. = 27.9 /j,g/m3
PM10 Mm/Max =4.7/186.4 Mg/m3
PM10IQR = 21.4Mg/m3
Seattle, WA (9/95 - 12/96)
Pop. Of Children <18 = 107,816
PM10 mean. = 21.5 ,ug/ni3
PM10 Min/Max = 8/69.3 Mg/m3
PM10IQR=11.7Mg/m3
Tolbert et al. (2000b)
Atlanta, GA (92 - 94 Summers)
Population = 80% of children in
total population of 3 million
PM10 mn. (SE) = 38.9 (15.5) Mg/
PM10 Range = 9, 105 ,ug/m3
Associations investigated between an
atmospheric stagnation index (# of hours
below median wind speed), a "surrogate
index of pollution", and asthma ED visits
for persons <65 yr. (mean=3.2/d) in
Spokane and for children <18 yr.
(mean=l .8/d) in Seattle. Poisson GAM
model applied, controlling for day of week,
long-wave effects, and temperature and dew
point (as non-linear smooths). Factor
Analysis (FA) applied to identify PM
components associated with asthma HA's.
Pediatric (<17 yrs. of age) ED visits (mean
= 467/day) related to air pollution (PM10,
O3, NO,,, pollen and mold) using GEE and
logistic regression and Bayesian models.
Autocorrelation, day of week, long-term
trend terms, and linear temperature controls
included.
Stagnation persistence index was strongly
associated with ED visits for asthma in
both cities. Factor analysis indicated that
products of incomplete combustion
(especially wood-smoke related K, OC,
EC, and CO) are the air pollutants driving
this association. Multi-pollutant models
run with "stagnation" as the "co-pollutant"
indicated importance of general air
pollution over any single air pollutant
index, but not of the importance of various
pollutants relative to each other.
Both PM10 and O3 positively associated
with asthma ED visits using all three
modeling approaches. In models with
both O3 and PM10, both pollutants become
non-significant because of high
collinearity of the variables (r=0.75).
Asthma ED Visits
Single Pollutant Models
Persons<65 years (Spokane)
For PM10IQR = 50 ,ug/m3
Lag 3 ER = 2.4% (CI: -10.9, 17.6)
Persons<18 years (Seattle)
For PM10 IQR = 50 Mg/m3
Lag 3 ER = 56.2% (95 CI: 10.4,121.1)
Pediatric (<17 yrs. of age) ED Visits
PM10 = 50 Mg/m3
Lag 1 day ER = 13.2% (CI: 1.2, 26.7)
With O3 8.2 (-7.1, 26.1)
fe
H
6
o
o
H
O
O
H
W
O
o
-------
to
o
o
to
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
-U
to
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Tolbert et al. (2000a)
Atlanta
Period 1: 1/1/93-7/31/98
Mean, median, SD:
PM10 Cug/m3): 30.1,28.0,12.4
Period 2: 8/1/98-8/31/99
Mean, median, SD:
PM10 (,ug/m3): 29.1,27.6,12.0
PM25 Og/m3): 19.4, 17.5, 9.35
CP (>g/m3): 9.39, 8.95, 4.52 10-
100 nmPM counts
(count/cm3): 15,200, 10,900,
26,600
10-100 nm PM surface area
(unrVcm3): 62.5,43.4,116
PM25 soluble metals (,ug/m3):
0.0327, 0.0226, 0.0306
PM25 Sulfates Og/m3): 5.59, 4.67,
3.6
PM2 5 Acidity Cwg/m3): 0.0181,
0.0112,0.0219
PM2 5 organic PMCwg/m3): 6.30,
5.90,3.16
PM25 elemental carbon (,ug/m3):
2.25,1.88,1.74
Preliminary analysis of daily emergency
department (ED) visits for asthma (493),
wheezing (786.09) COPD (491, 492, 4966)
LRI 466.1, 480, 481, 482, 483, 484, 485,
486), all resp disease (460-466, 477, 480-
486, 491, 492, 493, 496, 786.09) for
persons > 16 yr in the period before (Period
1) and during (Period 2) the Atlanta
superstation study. ED data analyzed here
from just 18 of 33 participating hospitals;
numbers of participating hospitals increased
during period 1. Mean daily ED visits for
dysrhythmias and all DVD in period 1 were
6.5 and 28.4, respectively. Covariates:
NO2, O3, SO2, CO temperature, dewpoint,
and, in period 2 only, VOCs. PM measured
by both TEOM and Federal Reference
Method; unclear which used in analyses.
For epidemiologic analyses, the two time
periods were analyzed separately. Poisson
regression analyses were conducted with
cubic splines for time, temperature and
dewpoint. Day-of-week and hospital
entry/exit indicators also included.
Pollutants treated a-priori as three-day
moving averages of lags 0, 1, and 2. Only
single-pollutant results reported.
In period 1, observed significant COPD
association with 3-day average PM10.
COPD was also positively associated with
NO2, O3, CO and SO2. No statistically
significant association observed between
asthma and PM10 in period 1. However,
asthma positively associated with ozone
(p=0.03). In period 2, i.e., the first year of
operation of the superstation, no
statistically significant associations
observed with PM10 or PM2 5. These
preliminary results should be interpreted
with caution given the incomplete and
variable nature of the databases analyzed.
Period 1:
PM10 (0-2 d):
asthma:
5.6% (-8.6, 22.1)
COPD:
19.9% (0.1, 43.7)
Period 2: (all 0-2 day lag)
PM10: asthma
18.8% (-8.7, 54.4)
COPD
-3.5% (29.9- 33.0)
PM25: asthma
2.3% (-14.8, 22.7)
COPD
12.4% (-7.9, 37.2)
PM10_25: asthma
21.1% (-18.2, 79.3)
COPD
-23.0% (50.7-20.1)
O
-------
to
o
o
to
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
United States (cont'd)
Yang etal( 1997)
Study Period: 92 - 94
Reno-Sparks, Nevada
Population = 298,000
PM10 mean = 33.6 Mg/m3
PM10 range = 2.2, 157.3
Association between asthma ER visits
(mean = 1.75/d, SD=1.53/d) and PM10, CO
and O3 assessed using linear WLS and
ARIMA regression, including adjustments
for day-of-week, season, and temperature
(but not RH or T-RH interaction). Season
adjusted only crudely, using month dummy
variable.
Only O3 showed significant associations
with asthma ER visits. However, the
crude season adjustment and linear model
(rather than Poisson) may have adversely
affected results. Also, Beta-gauge PM10
mass index used, rather than direct
gravimetric mass measurements.
NR
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Canada
Delfmoetal. (1997)
Montreal, Canada
Population= 3 million
6-9/92, 6-9/93
1993 Means (SD):
PM10=21.7Mg/m3(10.2)
PM25=12.2,ug/m3(7.1)
SO4== 34.8nmol/m3(33.1)
H+= 4nmol/m3(5.2)
Delfmoetal. (1998)
Montreal, Canada
6-8/89,6-8/90
MeanPM10= 18.6 ,ug/m3
(SD=9.3, 90th% = 30.0 Mg
Association of daily respiratory emergency
department (ED) visits (mean = 98/day
from 25 of 31 acute care hospitals) with O3,
PM10, PM25, SO4~, and H+ assessed using
linear regression with controls for temporal
trends, auto-correlation, and weather. Five
age sub-groups considered.
Examined the relationship of daily ED
visits for respiratory illnesses by age
(mean/day: <2yr.=8.9; 2-34yr.=20.1; 35-
64yr.=22.6; >64yr.=20.3) with O3 and
estimated PM25. Seasonal and day-of-week
trends, auto-correlation, relative humidity
and temperature were addressed in linear
time series regressions.
No associations with ED visits in '92, but
33% of the PM data missing then. In'93,
only H+ associated for children <2, despite
very low H+ levels. H+ effect stable in
multiple pollutant models and after
excluding highest values. No associations
for ED visits in persons aged 2-64 yrs.
For patients >64 yr, O3, PM10, PM2 5, and
SO4~ positively associated with visits
(p < 0.02), but PM effects smaller than for
03.
There was an association between PM2 5
and respiratory ED visits for older adults
(>64), but this was confounded by both
temperature and O3. The fact that PM2 5
was estimated, rather than measured, may
have weakened its relationship with ED
visits, relative to O3.
Respiratory ED Visits
Adults >64: (pollutant lags = 1 day)
50 Mg/m3 PM10ER = 36.6% (10.0, 63.2)
25 Mg/m3 PM2.5 ER = 23.9% (4.9, 42.8)
Older Adults(>64 yr) Respiratory ED Visits
Estimated PM25 = 25 /-ig/m3
Single Pollutant:
(lag 1 PM25) ER = 13.2 (-0.2, 26.6)
With Ozone (lag 1PM2 5):
Est. PM25 (lagl) ER = 0.8% (CI: -14.4, 15.8)
o
-------
to
o
o
to
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Canada (cont'd)
Stiebetal. (1996)
New Brunswick, Canada
Population = 75,000
May-Sept. 84 - 92
SO42'Mean =5.5 Mg/m3
Range: 1-23, 95th% =14 ,ug/m3
TSP Mean = 36.7 Mg/m3
Range:5-108, 95°% =70 Mg/m3
Stieb et al. (2000)
Saint John, Canada
7/1/92-3/31/96
mean and S.D.:
PM10(Mg/m3): 14.0,9.0
PM25(Aig/m3): 8.5,5.9
H+(nmol/m3): 25.7,36.8
Sulfate (nmol/m3): 31.1,29.7
COHmean(103lnft): 0.2,0.2
COHmax(103lnft): 0.6,0.5
Asthma ED visits (mean=l .6/day) related to
daily O3 and other air pollutants (SO2, NO2,
SO42", and TSP). PM measured only every
6th day. Weather variables included
temperature, humidex, dewpoint, and RH.
ED visit frequencies were filtered to remove
day of week and long wave trends. Filtered
values were regressed on pollution and
weather variables for the same day and the
3 previous days.
Study of daily emergency department (ED)
visits for asthma (mean 3.5/day), COPD
(mean 1.3/day), resp infections (mean
6.2/day), and all respiratory conditions
(mean 10.9/day) for persons of all ages.
Covariates included CO, H2S, NO2, O3,
SO2, total reduced sulfur (TRS), a large
number of weather variables, and 12 molds
and pollens. Stats: generalized additive
models with LOESS prefiltering of both ED
and pollutant variables, with variable
window lengths. Also controlled for day of
week and LOESS-smoothed functions of
weather. Single-day, and five day average,
pollution lags tested out to lag 10. The
strongest lag, either positive or negative,
was chosen for final models. Both single
and multi-pollutant models reported. Full-
year and May-Sep models reported.
Positive, statistically significant (p < 0.05)
association observed between O3 and
asthma ED visits 2 days later; strength of
the association greater in nonlinear
models. Ozone effect not significantly
influenced by addition of other pollutants.
However, given limited number of
sampling days for sulfate and TSP, it was
concluded that "a particulate effect could
not be ruled out".
In single-pollutant models, significant
positive associations were observed
between all respiratory ED visits and
PM10, PM25, H2S, O3, and SO2.
Significant negative associations were
observed with H+, and COH max. PM
results were similar when data were
restricted to May-Sep. In multi-pollutant
models, no PM metrics significantly
associated with all cardiac ED visits in full
year analyses, whereas both O3 and SO2
were. In the May-Sep subset, significant
negative association found for sulfate. No
quantitative results presented for non-
significant variables in these multi-
pollutant regressions.
Emergency Department Visits (all ages)
Single Pollutant Model
100 Mg/m3 TSP = 10.7% (-66.4, 87.8)
PM2.5, (lag 3) 15.1 (-0.2, 32.8)
PM10, (lag 3) 32.5 (10.2, 59.3)
o
-------
to
o
o
to
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe
Atkinson et al. (1999a)
London (92 - 94)
Population = NR
PM10Mean = 28.5,ug/m3
10th-90thIQR = 15.8-46.5 ,ug/m3
BSmean=12.7Mg/m3
10th-90thIQR = 5.5-21.6 Mg/m3
Hajatetal. (1999)
London, England (92 - 94)
Population = 282,000
PM10 mean = 28.2 /j,g/m3
PM1010'-90th%=16.3-46.4 Aig/
BSmean=10.lMg/m3
BS 10'-90th%=4.5-15.9Mg/m3
All-age Respiratory (mean=90/day),
Asthma (25.9/day), and Other Respiratory
(64.1/day) ED visits from 12 London
hospitals considered, but associated
population size not reported. Counts for
ages 0-14, 15-64, and>64 also examined.
Poisson regression used, controlling for
season, day of week, meteorology,
autocorrelation, overdispersion, and
influenza epidemics.
Examined associations of PM10, BS, NO2,
O3, SO2, and CO, with primary care general
practitioner asthma and "other LRD"
consultations. Asthma consultation means
per day = 35.3 (all ages); 14.(0-14 yrs,);
17.7 (15-64 yrs.); 3.6 (>64 yrs.). LRD
means = 155 (all ages); 39.7(0-14 yrs,);
73.8 (15-64 yrs.); 41.1 (>64 yrs.). Time-
series analyses of daily numbers of
consultations performed, controlling for
time trends, season factors, day of week,
influenza, weather, pollen levels, and serial
correlation.
PM10 positively associated, but not BS, for
all-age/all-respiratory category. PM10
results driven by significant children and
young adult associations, while older adult
visits had negative (but non-significant)
PM10-ED visit relationship. PM10
positively associated for all ages, children,
and young adults for asthma ED visits.
However, PM10-asthma relationship
couldn't be separated from SO2 in multi-
pollutant regressions. Older adult ED
visits most strongly associated with CO.
No O3-ED visits relationships found (but
no warm season analyses attempted).
Positive associations, weakly significant
and consistent across lags, observed
between asthma consultations and NO2
and CO in children, and with PM10 in
adults, and between other LRD
consultations and SO2 in children.
Authors concluded that there are
associations between air pollution and
daily concentrations for asthma and other
lower respiratory disease in London. In
adults, the authors concluded that the only
consistent association was with PM10.
Across all of the various age, cause, and
season categories considered, PM10 was
the pollutant most coherent in giving
positive pollutant RR estimates for both
asthma and other LRD (11 of 12
categories positive) in single pollutant
models considered.
PM10 (50 Mg/m3) No co-pollutant:
All Respiratory ED visits
All age(lag ld)ER = 4.9% (CI: 1.3, 8.6)
<15yrs(lag 2d)ER = 6.4% (CI: 1, 12.2)
15-64yr(lagld)ER= 8.6% (CI: 3.4, 14)
Asthma ED visits
All age (lag Id) ER = 8.9% (CI: 3, 15.2)
<15yrs (lag 2d) ER = 12.3% (CI: 3.4, 22)
15-64yr(lg ld)ER= 13% (CI: 4.6,22.1)
PM10 (50 Mg/m3) 2d lag & co-pollutant:
Children's (<15 yrs.) Asthma ED Visits:
PM alone: ER = 12.3% (CI: 3.4, 22)
&N02: ER= 7.8% (CI: -1.2, 17.6)
&O3: ER =10.5% (CI: 1.6,20.1)
&S02: ER = 8.1%(CI: -1.1,18.2)
&CO: ER = 12.1%(CI: 3.2,21.7)
Asthma Doctor's Visits:
50 Mg/m3 PM10
-Year-round, Single Pollutant:
All ages (Ig 2): ER = 5.4% (CI: -0.6, 11.7)
0-14 yrs.(lg 1): ER = 6.4% (-1.5, 14.6)
15-64 yrs.(lg 0): ER = 9.2% (CI: 2.8, 15.9)
>64yrs.(lg 2): ER= 11.7% (-1.8, 26.9)
-Year-round, 2 Pollutant, Children (0, 14):
(PM10 lag = 1 day) PM10 ER's:
W/N02: ER= 0.8% (CI: -8.7, 11.4)
W/O3: ER= 5.5% (-2.1, 13.8)
W/S02: ER= 3.2% (CI: -6.4, 13.7)
Other Lower Resp. Pis. Doctor's Visits:
50 Mg/m3 PM10
-Year-round, Single Pollutant:
All ages (Ig 2): ER = 3.5% (CI: 0, 7.1)
0-14yrs.(lg 1): ER = 4.2%(CI: -1.2, 9.9)
15-64 yrs.(lg 2): ER= 3.7% (CI: 0.0, 7.6)
>64yrs.(lg 2): ER = 6.2% (CI: 0.5, 12.9)
-------
to
o
o
to
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Europe (cont'd)
Hajatetal. (2001)
London (1992-1994)
44,406-49,596 registered patients
<1 to 14 years
PM10mean28.5(13.9)
Medina etal. (1997)
Greater Paris 91 -95
Populations 6.5 MM
Mean PM13 = 25 Mg/m
PM13 min/max = 6/95 /-
BS min/max = 3/130 ,ug/m3
Damiaetal. (1999)
Valencia, Spain (3/94-3/95)
Population = NR
BS mean = 101 Mg/m3
BS range = 34-213 ,ug/m3
Daily physician consultations (mean daily
4.8 for children; 15.3 for adults) for allergic
rhinitis (ICD-9, 477), SO2, O3, NO2, CO,
PM10, and pollen using generalized additive
models with nonparametric smoother.
Evaluated short-term relationships between
PM13 and BS concentrations and doctors'
house calls (mean=8/day; 20% of city total)
in Greater Paris. Poisson regression used,
with non-parametric smoothing functions
controlling for time trend, seasonal patterns,
pollen counts, influenza epidemics, day-of-
week, holidays, and weather.
Associations of BS and SO2 with weekly
total ED admissions for asthma patients
aged > 12 yrs (mean = 10/week) at one
hospital over one year assessed, using linear
stepwise regression. Season-specific
analyses done for each of 4 seasons, but no
other long-wave controls. Linear T, RH,
BP, rain, and wind speed included as crude
weather controls in ANOVA models.
SO2 and O3 show strong associations with
the number of consultations for allergic
rhinitis. Estimates largest for a lag of 3 or
4 days prior to consultations, with
cumulative measures stronger than single
day lags. Stronger effects were found for
children than adults. The two-pollutant
analysis of the children's model showed
that PM10 and NO2 associations
disappeared once either SO2 or O3 was
incorporated into the model.
A relationship between all age (0-64 yrs.)
asthma house calls and PM13, BS, SO2,
NO2, and O3 air pollution, especially for
children aged 0-14 (mean = 2/day).
In two-pollutant models including BS
with, successively, SO2, NO2, and O3, only
BS and O3 effects remained stable. These
results also indicate that air pollutant
associations noted for hospital ED visits
are also applicable to a wider population
that visits their doctor.
Both BS and SO2 correlated with ED
admissions for asthma (SO2: r=0.32; BS:
i=0.35), but only BS significant in
stepwise multiple regression. No linear
relationship found with weather variables.
Stratified ANOVA found strongest BS-ED
association in the autumn and during
above average temperatures. Uncontrolled
autocorrelation (e.g., within-season) and
weather effects likely remain in models.
PM10 - Increment (10-90%)
(15.8-46.5)
Age <1-14 years
lag 3: 10.4 (2.0 to 19.4)
Cum 0-3: 17.4 (6.8 to 29.0)
Ages 15-64 years
lag 2: 7.1(2.6 to 11.7)
Cum 0-6: 20.2 (14.1 to 26.6)
Doctor's Asthma House Visits:
50 Mg/m3 PM13
Year-round, Single Pollutant:
All ages (Ig 2): ER = 12.7% (CI: 4.1, 21.9)
0-14 yrs.(lg 0-3): ER = 41.5% (CI: 20, 66.8)
15-64 yrs.(lg 2): ER = 6.3% (CI: -4.6, 18.5)
Asthma ED Visits (all ages):
BS = 40 Mg/m3 (single pollutant)
BS as a lag 0 weekly average:
ER = 41.5% (CI = 39.1, 43.9)
O
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
o
o
to
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
Europe (cont'd)
Pantazopoulou et al. (1995)
Athens, GR( 1988)
Population = NR
Winter (1/88-3/88,9/88-12/88)
BS mean. =75 /-ig/m3
BS 5th-95th%=26 - 161 Mg/m3
Summer (3/22/88-3/88,9/21/88)
BS mean. =55 /-ig/m3
BS5th-95th%=19-90Mg/m3
Garty etal. (1998)
PM10 mean ~ 45 Mg/ni3
Tel Aviv, Israel (1993)
Examined effects of air pollution on daily
emergency outpatient visits and admissions
for cardiac and respiratory causes. Air
pollutants included: BS, CO, andNO2.
Multiple linear regression models used,
controlling for linear effects of temperature
and RH, day of week, holidays, and dummy
variables for month to crudely control for
season, separately for winter and summer.
Seven day running mean of asthma ED
visits by children (1-18 yrs.) to a pediatric
hospital modeled in relation to PM10 in Tel
Aviv, Israel.
Daily number of emergency visits related
positively with each air pollutant, but only
reached nominal level of statistical
significance for NO2 in winter. However,
the very limited time for each within-
season analysis (6 mo.) undoubtably
limited the power of this analysis to detect
significant effects. Also, possible lagged
pollution effects were apparently not
investigated, which may have reduced
effect estimates.
No PM10 associations found with ED
visits. The ER visits-pollutant correlation
increased significantly when the
September peak was excluded. Use of a
week-long average and associated
uncontrolled long-wave fluctuations (with
resultant autocorrelation) likely prevented
meaningful analyses of short -term PM
associations with ED visits.
Single Pollutant Models
For Winter (BS = 25 Mg/m3
Outpatient Hospital Visits
ER= 1.1% (-0.7, 2.3)
Respiratory HA's
ER = 4.3%(0.2, 8.3)
For Summer, BS = 25 Mg/
Outpatient Hospital Visits
ER= 0.6% (-4.7, 6.0))
Respiratory HA's
ER= 5.5% (-3.6, 14.7)
N/A
fe
H
6
o
o
H
O
O
H
W
O
O
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
o
o
to
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
-U
oo
Latin Amercia
Ilabacaetal. (1999)
Santiago, Chile
February 1995-August 1996
PM10: warm: 80.3 Mg/ni3
cold: 123.9,ug/m3
PM25: warm: 34.3 Mg/rn3
cold: 71.3 Mg/m3
Number of daily respiratory emergency
visits (REVs) related to PM by Poisson
model smooths for longer- and short-term
trends. SO2,NO2,O3.
Stronger coefficients for models including
PM2 5 than for models including PM10 or
PM10_2 5. Copollutant effects were
significantly associated with REVs. For
respiratory patients, the median number of
days between the onset of the first
symptoms and REV was two to three days.
For the majority of patients (70%) this
corresponded to the lag observed in this
study indicating that the timing of the
pollutant effect is consistent with the
temporal pattern of REV in this
population.
REV, lag 2
Cold
PM25,lag2
OR: 1.027 (1.01 to 1.04) for a45
increment
PM10, lag 2
OR: 1.02 (1.01 to 1.04) for a 76 M
increment
PM25,lag2
OR: 1.01 (1.00* to 1.03) for a 32
increment
Pneumonia, lag 2
PM10: 1.05 (1.00* to 1.10)
64 ,ug/rn3 increment
PM25: 1.04 (1.00* to 1.09)
45 Mg/nr3 increment
PM10.25: 10.5 (1.00* to 1.10)
32 ,ug/rn3 increment
'decimals < 1.00
fe
H
6
o
o
H
O
O
H
W
O
O
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
o
o
to
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
oo
td
-U
VO
fe
H
6
o
o
H
O
O
H
W
O
Latin Amercia (cont'd)
Lin etal. (1999)
Sao Paulo, BR (91-93)
Population=NR
PM10 mean =65 Mg/m3
PM10 SD=27 Mg/m3
PM10 range=15-193 Mg/
Ostroetal. (1999b)
Santiago, CI (7/92—12/93)
<2 yrs. Population « 20,800
3-14 yrs. Population ~ 128,000
PM10 mean. =108.6 Mg/m3
PM10 Min/Max=18.5/380 Mg/m3
PM10IQR = 70.3 - 135.5 Mg/m3
Respiratory ED visits by children (0-12
yrs.) To a major pediatric hospital
(mean=56/day) related to PM10, SO2, NO2,
CO, and O3 using Gaussian linear
regression modeling, Poisson modeling, and
a polynomial distributed lag model. Lower
respiratory (mean = 8/day) and upper
respiratory (mean = 9/day) all evaluated.
Analyses considered effects of season, day
of week, and extreme weather (using T, RH
dummy variables).
Analysis of daily visits to primary health
care clinics for upper (URS) or lower
respiratory symptoms (LRS) for children 2-
14 yr (mean LRS=111. I/day) and < age 2
(mean LRS=104.3/day). Daily PM10 and O3
and meteorological variables considered.
The multiple regression GAM included
controls for seasonality (LOESS smooth),
temperature, day of week, and month.
PM10 was found to be "the pollutant that
exhibited the most robust and stable
association with all categories of
respiratory disease". O3 was the only
other pollutant that remained associated
when other pollutants all simultaneously
added to the model. However, some
pollutant coefficients went negative in
multiple pollutant regressions, suggesting
coefficient intercorrelations in the multiple
pollutant models. More than 20%
increase in ED visits found on the most
polluted days, "indicating that air
pollution is a substantial pediatric health
concern".
Analyses indicated an association between Lower Resp. Symptoms Clinic Visits
50 Mg/nf PM10 (0-5-day lag mean)
Respiratory ED Visits (<13 yrs.)
Single pollutant model:
PM10ER=21.7%(CI: 18.2,25.2)
All pollutant models: PM10 ER=28.8%
(CI: 21.4,36.7)
Lower Respiratory ED Visits (<13 yrs.)
Single pollutant model:
PM10 ER=22.8% (CI: 12.7, 33.9)
All pollutant models: PM10 ER=46.9%
(CI: 27.9,68.8)
PM10 and medical visits for LRS in
children ages 2-14 and in children under
age 2 yr. PM10 was not related to non-
respiratory visits (mean =208/day).
Results unchanged by eliminating high
PM10 (>235 ,ug/m3) or coldest days
(<8°C). Adding O3 to the model had little
effect on PM,n-LRS associations.
PM10 = 50 Mg/m3
Single Pollutant Models:
-Children<2 years
Lag3ER = 2.5%(CI:0.2,4.8)
-Children 2-14 years
Lag 3 ER = 3.7% (CI: 0.8, 6.7%)
Two Pollutant Models (with O3):
-Children<2 years
Lag 3 ER = 2.2% (CI: 0, 4.4)
-Children 2-14 years
Lag 3 ER = 3.7% (CI: 0.9, 6.5)
O
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
o
o
to
Reference/Citation
Location, Duration
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
Australia
Smith etal. (1996)
StdyPd.: 12/92-1/93,12/93-1/94
West Sydney, AU
Population = 907,000
-Period 1 (12/92-1/93)
Bscalt median = 0.25 10~4/m
Bsc
= 0.18-0.39 10-4/m
Bscatt95th% = 0.86 10-4/m
-Period 2 (12/93-1/94)
Bscalt median = 0.19 10~4/m
Bsc
= 0.1-0.38 10-4/m
Bscatt 95th5% = 3.26 10-4/mPM10
median =18 Mg/m3
PM10 IQR =11.5-28.8 Mg/m3
PM1095th% = 92.5Mg/m3
Study evaluated whether asthma visits to
emergency departments (ED) in western
Sydney (mean-10/day) increased as result
of bushfire-generated PM ( Bscalt from
nephelometry) in Jan., 1994 (period 2). Air
pollution data included nephelometry
(Bscatt), PMio, SO2, andNO2. Data analyzed
using two methods: (1) calculation of the
difference in proportion of all asthma ED
visits between the time periods, and; (2)
Poisson regression analyses. Control
variables included T, RH, BP, WS, and
rainfall.
oo
td
No difference found in the proportion of
all asthma ED visits during a week of
bushfire-generated air pollution, compared
with the same week 12 months before,
after adjusting for baseline changes over
the 12-month period. The max. Bscatt
reading was not a significant predictor of
the daily asthma ED visits in Poisson
regressions. However, no long-wave
controls applied, other than indep. vars.,
and the power to detect differences was
weak (90% for a 50% difference). Thus,
the lack of a difference may be due to low
statistical strength or to lower toxicity of
particles from burning vegetation at
ambient conditions vs. fossil fuel
combustion.
ED Asthma Visits (all ages)
Percent change between bushfire and non
bushfire weeks:
PM10 = 50 Mg/m3
ER = 2.1%(CI: -0.2,4.5)
Asia
Ye etal. (2001)
Tokyo, Japan
Summer months
July-August, 1980-1995
PM10 46.0 mean
Hospital emergency transports for
respiratory disease for >65 years of age
were related to pollutant levels NO2, O3,
PM10, S02, and CO.
For chronic bronchitis PM10 with a lag
time of 2 days was the most statistically
significant model covariate.
Asthma (ICD-9-493)
Coefficienct estimate (SE)
0.003 (0.001)
fe
H
6
o
o
H
O
O
H
W
O
Chew etal. (1999)
Singapore (90 - 94)
Population = NR
TSPmean = 51.2,ug/m3
TSPSD = 20.3,ug/m3
TSP range = 13-184,ug/m3
Child (3-13 yrs.) ED visits (mean =
12.8/day) and HA's (mean = 12.2/day) for
asthma related to levels of SO2, NO2, TSP,
and O3 using linear regression with
weather, day-of-week controls. Auto-
correlation effects controlled by including
prior day response variable as a regression
variable. Separate analyses done for
adolescents (13-21 yrs.) (mean ED=12.2,
meanHA=3.0/day).
Positive associations found between TSP,
SO2, and NO2, and daily HA and ED visits
for asthma in children, but only with ED
visits among adolescents. Lack of power
(low counts) for adolescents' HA's
appears to have been a factor in the lack of
associations. When ED visits stratified by
year, SO2 and TSP remained associated in
every year, but not NO2. Analyses for
control diseases (appendicitis and urinary
tract infections) found no associations.
TSP(100 Mg/m3)No co-pollutant:
Child (3-13 vrs.Wsthma ED visits
Lag ldER = 541% (CI: 198.4, 1276.S
O
-------
Appendix 8B.4: Pulmonary Function Studies
April 2002 8B-51 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
TABLE 8B-4. SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
(^
to
fe
H
6
o
o
H
O
O
H
W
O
O
UnitedStates
Thurstonetal. (1997)
Summers 1991-1993.
O3, H+, sulfate
Canada
Vedaletal. (1998)
Port Alberni, BC
PM10 measurements were made using a Sierra-Anderson
dichotomous sampler. PM10 ranged from 1 to
159,ug/m3.
Europe
Gielenetal. (1997)
Amsterdam, NL
Mean PM10 level: 30.5 ,ug/m3 (16, 60.3).
Mean maximum 8 hr O,. 67 i/g/m3.
Hiltermannetal. (1998)
Leiden, NL
July-Oct, 1995
O3, NO2, SO2, BS, and PM10 ranged from 16.4 to
97.9 Mg/m3)
Three 5-day summer camps conducted in 1991,
1992, 1993. Study measured symptoms and
change in lung function (morning to evening).
Poisson regression for symptoms.
Study of 206 children aged 6 to 13 years living
in Port Alberni, British Columbia. 75 children
had physician-diagnosed asthma, 57 had an
exercised induced fall in FEV1, 18 children
with airway obstruction, and 56 children
without any symptoms. Respiratory symptom
data obtained from diaries. An autoregressive
model was fitted to the data, using GEE
methods. Covariates included temp., humidity,
and precipitation.
Study evaluated 61 children aged 7 to 13 years
living in Amsterdam, The Netherlands. 77
percent of the children were taking asthma
medication and the others were being
hospitalized for respiratory problems. Peak
flow measurements were taken twice daily.
Associations of air pollution were evaluated
using time series analyses. The analyses
adjusted for pollen counts, time trend, and day
of week.
270 adult asthmatic patients from an out-patient
clinic in Leiden, The Netherlands were studied
from July 3 to October 6, 1995. Peak flow
measured twice daily. An autoregressive model
was fitted to the data. Covariates included
temp, and day of week. Individual responses
not modeled.
The O3-APEFR relationship was seen as
the strongest.
In general, PM10 was associated with
changes in both peak flow and respiratory
symptoms. Ozone, SO2, and sulfate levels
were low because of low vehicle
admissions.
The strongest relationships were found
with ozone, although some significant
relationships found with PM10.
No relationship between ozone or PM10
and PET was found
Lag 0, PM10 average PEE- -0.27 (-0.54,
-0.01) per 10 ,ug/m3 increment
Lag 0, PM10:
Evening PEE = -0.08 (-2.49, 2.42)
Lag 1, PM10:
Morning PEE = 1.38 (-0.58, 3.35)
Lag 2, PM10:
Morning PEE = 0.34 (-1.78, 2.46)
Evening PEE = -1.46 (-3.23, 0.32)
Lag 0, PM10:
Average PEE = -0.80 (-3.84, 2.04)
7 day ave., PM10:
Average PEE = -1.10 (-5.22, 3.02)
-------
to
o
o
to
TABLE 8B-4 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY
FUNCTION TESTS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
Europe (cont'd)
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Peters etal. (1996)
Erfurt and Weimar, Germany
SO2, TSP, PM10, sulfate fraction, and PSA.
Mean PM10 level was 112 ,ug/m3.
PM was measured by a Marple-Harvard impactor.
Peters etal. (1997b)
Erfurt, Germany
PM fractions measured over range of sizes from
ultrafme to fine, including PM10.
Particles measured using size cuts of 0.01 to 0.1, 0.1 to
0.5, and 0.5 to 2.5 //m.
Mean PM10 level: 55 ,ug/m3 (max 71). Mean SO2:
100,ug/m3(max383).
PM was measured using a Harvard impactor. Particle
size distributions were estimated using a conduction
particle counter.
Peters etal. (1997c)
Sckolov, Czech Republic
Winter 1991-1992
PM10, SO2, TSP, sulfate, and particle strong acid.
Median PM10 level: 47 ^g/m3 (29, 73).
Median SO2: 46 ^g/m3 (22, 88).
PM was measured using a Harvard impactor. Particle
size distributions were estimated using a conduction
particle counter.
Panel of 155 asthmatic children in the cities of
Erfurt and Weimar, E. Germany studied. Each
panelist's mean PEE over the entire period
subtracted from the PEE value to obtain a
deviation. Mean deviation for all panelists on
given day was analyzed using an autoregressive
moving average. Regression analyses done
separately for adults and children in each city
and winter; then combined results calculated.
Study of 27 non-smoking adult asthmatics
living in Erfurt, Germany during winter season
of 1991-1992. Morning and evening peak flow
readings recorded. An auto-regressive model
was used to analyze deviations in individual
peak flow values, including terms for time
trend, temp., humidity, and wind speed and
direction.
Five day average SO2 was associated with
decreased PEE. Changes in PEE were not
associated with PM levels.
89 children with asthma in Sokolov, Czech
Republic studied. Subjects kept diaries and
measured peak flow for seven months during
winter of 1991-2. The analysis used linear
regression for PET. First order autocorrelations
were observed and corrected for using
polynomial distributed lag (PDL) structures.
Strongest effects on peak flow found with
ultrafme particles. The two smallest
fractions, 0.01 to 0.1 and 0.1 to 0.5 were
associated with a decrease of PEF.
Five day mean SO2, sulfates, and particle
strong acidity were also associated with
decreases in PM PFT as well as PM10.
Lag 0, PM10:
Evening PEF =-0.38 (-1.83, 1.08)
Lag 1, PM10:
Morning PEF = -1.30 (-2.36, 0.24)
5 Day Mean, PM10:
Morning PEF =-1.51 (-3.20,0.19)
Evening PEF = -2.31 (-4.54, -0.08)
LagO, PM25:
Evening PEF = -0.75 (-1.66, 0.17)
Lagl, PM25:
Morning PEF =-0.71 (-1.30,0.12)
5 Day Mean, PM25:
Morning PEF =-1.19 (-1.81, 0.57)
Evening PEF = -1.79 (-2.64, -0.95)
Lag 0, PM10:
Morning PEF = -0.71 (-2.14, 0.70)
Evening PEF = -0.92 (-1.96, 0.12)
5 Day mean PM10:
Evening PEF = -1.72 (-3.64, 0.19)
Morning PEF = -0.94 (-2.76, 0.91
o
-------
to
o
o
to
TABLE 8B-4 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY
FUNCTION TESTS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 fj.g/m1 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Timonen and Pekkanen (1997)
Kupio, Finland
PM10, BS, NO2, and SO2.
The intequartile range on PM10 was 8 to '<
Penttinen et al. (2001) studied adult asthmatics for
6 months in Helsinki, Findland. PM was measured
using a single-stage Harvard impactor. Particle number
concentrations were measured using an Electric Aerosol
Spectrometer. NO2PM10 ranged from 3.8 to 73.7 //g/m3.
PM25 ranged from 2.4 to 38.3 ,ug/m3.
Pekkanen etal. (1997)
Kuopio, Finland
PM fractions measured over range of sizes from
ultrafme to fine, including PM10.
Mean PM10 level: 18 ^g/m3 (10, 23).
Mean NO2 level: 28,ug/m3.
Segalaetal. (1998)
Paris, France
Nov. 1992-May 1993.
BS, SO2, NO2, PM13 (instead of PM10), measured.
Mean PM13 level: 34.2 ,ug/m3 (range 8.8, 95).
Mean SO2 level: 21.7 ^g/m3 (range 4.4, 83.8).
Mean NO2 level: 56.9 ^g/m3 (range 23.8, 121.9).
PM was measured by p-radiometry.
Gauvinetal. (1999)
Grenoble, France
Summer 1996, Winter 1997
Mean (SD) ^g/m3
PM10 Summer 23 (6.7)
PM10 Winter 38(17.3)
Sunday 15.55(5.12)
Weekday 24.03 (7.2)
Studied 74 asthmatic children (7 to 12 yr) in
Kuoio, Finland. Daily mean PEF deviation
calculated for each child. Values were
analyzed, then using linear first-order
autoregressive model. PM was measured using
single stage Harvard Impactors.
57 asthmatics were followed with daily PEF
measurements and symptom and medications
diaries from November 1996 to April 1997.
PEF deviations from averages were used as
dependent variables. Independent variables
included PM[, PM2 5, PM10, particle counts, CO,
NO, and
Studied 39 asthmatic children aged 7-12 years
living in Kuopio, Finland. Changes in peak flow
measurements were analyzed using a linear
first-order autoregressive model. PM was
measured using single stage Harvard impactors.
Study of 43 mildly asthmatic children aged 7-15
years living in Paris, France from Nov. 15, 1992
to May 9, 1993. Peak flow measured three
times a day. Covariates in the model included
temperature and humidity. An autoregressive
model was fitted to the data using GEE
methods.
Two panels: mild adult asthmatics, ages 20-60
years, (summer-18 asthmatics, 20 control
subjects; winter-19 asthmatics, 21 control
subjects) were examined daily for FEV! and
PEF. Bronchial reactivity was compared
Sunday vs. weekday. Temperature and RH
controlled.
Lagged concentrations of NO2 related to
declines in morning PEF as well as PM10
and BS.
The strongest relationships were found
between PEF deviations and PM particles
below 0.1 ,um. No associations were
found between particulate pollution and
respiratory symptoms.
Changes in peak flow found to be related
to all measures of PM, after adjusting for
minimum temperature. PNO.032-0.10
(I/cm3) and PN1.0-3.2 (I/cm3) were most
strongly associated with morning PEF
deviations.
Effects found related to PM10 were less
than those found related to the other
pollutants. The strongest effects were
found with SO,.
Respiratory function decreased among
asthmatic subjects a few days (lag
2/4 days) after daily PM10 levels had
increased. Bronchial reactivity was not
significantly different between the
weekdays and weekends. No copollutant
analysis conducted.
AM PEF = -.115 (-.448, .218) PM25 lag one
day
AM PEF = -.001 (-.334, .332) PM25 lag two
days
Lag 0, PM10:
Evening PEF = -0.35 (-1.14, 0.96)
Lag 1, PM10:
Morning PEF = -2.70 (-6.65, 1.23)
Lag 2, PM10:
Morning PEF = -4.35 (-8.02, -0.67)
Evening PEF = -1.10 (-4.70, 2.50)
Small sized particles had relationships similar
to those of PM10 for morning and evening
PEF.
Lag 4, PM13:
Morning PEF = -0.62 (-1.52, 0.28)
For a 10 ,ug/m3 increase in PM10
Summer
FEVj
-1.25%(-0.58to-1.92)
PEF
-0.87% (-0.1 to-1.63)
-------
to
o
o
to
TABLE 8B-4 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY
FUNCTION TESTS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Agocs etal. (1997)
Budapest, Hungary
SO2 and TSP were measured. TSP was measured by
beta reactive absorption methods.
Australia
Rutherford etal. (1999)
Brisbane, Australia
PM10, TSP, and particle diameter.
PM10 ranged form 11.4to 158.6 ,ug/m3. Particle sizing
was done by a Coulter Multisizer.
Latin America
Romieu etal. (1996)
Mexico City, Mexico
During study period, maximum daily 1-h O3 ranged
from 40 to 370 ppb (mean 190 ppb, SD = 80 ppb).
24 h ave, PM10 levels ranged from 29 to 363 ,ug/m3
(mean 166.8 Mg/m3, SD 72.8 ,ug/m3).
For 53 percent of study days, PM10 levels exceeded
150 /^g/m3. PM10 was measured by a Harvard impactor.
Romieu etal. (1997)
Mexico City, Mexico
During study period, maximum daily 1-h ozone ranged
from 40 to 390 ppb (mean 196 ppb SD = 78 ppb)
PM10 daily average ranged from 12 to 126 i/g/m3.
PM10 was measured by a Harvard impactor.
Panel of 60 asthmatic children studied for two
months in Budapest, Hungary. Mixed model
used relating TSP to morning and evening
PEER measurements, adjusting for SO2, time
trend, day of week, temp., humidity
Study examined effects of 11 dust events on
peak flow and symptoms of people with asthma
in Brisbane, Australia. PEE data for each
individual averaged for a period of 7 days prior
to the identified event. This mean was
compared to the average for several days of PEE
after the event, and the difference was tested
using a paired t-test.
Study of 71 children with mild asthma aged 5-7
years living in the northern area of Mexico City.
Morning and evening peak flow measurements
recorded by parents. Peak flow measurements
were standardized for each person and a model
was fitted using GEE methods. Model included
terms for minimum temperature.
Study of 65 children with mild asthma aged 5-
13 yr in southwest Mexico City. Morning and
evening peak flow measurements made by
parents. Peak flow measurements standardized
for each person and model was fitted using GEE
methods. Model included terms for minimum
temperature.
The paired t-tests were stat. significant for
some days, but not others. No general
conclusions could be drawn.
Ozone strongly related to changes in
morning PEE as well as PM10.
Strongest relationships were found
between ozone (lag 0 or 1) and both
morning and evening PET.
No significant TSP-PEFR relationships found.
Lag 0, PM10:
Evening PEE = -4.80 (-8.00, -1.70)
Lag 2, PM10:
Evening PEE = -3.65 (-7.20, 0.03)
LagO, PM25:
Evening PEE = -4.27 (-7.12, -0.85)
Lag 2, PM25:
Evening PEE = -2.55 (-7.84, 2.74)
Lag 1, PM10
Morning PEE = -4.70 (-7.65, -1.7)
Lag 2, PM10
Morning PEE = -4.90 (-8.4, -1.5)
Lag 0, PM10:
Evening PEE = -1.32 (-6.82, 4.17)
Lag 2, PM10:
Evening PEE = -0.04 (-4.29, 4.21)
Morning PEE = 2.47 (-1.75, 6.75)
Lag 0, PM10:
Morning PEE = 0.65 (-3.97, 5.32)
-------
Appendix 8B.5: Short-Term PM Exposure Effects
On Symptoms in Asthmatic Individuals
April 2002 8B-56 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
Reference citation, location, duration,
pollutants measured, summary of values
TABLE 8B-5. SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON
SYMPTOMS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Type of study, sample size, health outcomes
measured, analysis design, covariates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
United States
Delfmoetal. (1996)
San Diego, CA
Sept-Oct 1993
Ozone and PM2 5 measured. PM was measured by a
Harvard impactor. PM2 5 ranged from 6 to 66 ^g/m3 with
a mean of 25.
Delfmoetal. (1997)
San Diego County, CA
PM10 and ozone PM was measured using a tapered-
element oscillating microbalance. PM10 ranged from 6 to
51 ,ug/m3 with a mean of 26.
Delfmoetal. (1998)
So. California community
Aug. - Oct. 1995
Highest 24-hour PM10 mean: 54//g/m3.
PM10 and ozone PM was measured using a tapered-
element oscillating microbalance. PM10 ranged from 6 to
51 ,ug/m3 with a mean of 26.
Yu et al. (2000) study of a panel of 133 children aged
5-12 years in Seattle, WA. PM was measured by
gravimetric and nephelometry methods. PM[ 0 ranged
from 2 to 62 ,ug/m3 with a mean of 10.4. PM10 9 to
86 ^g/m3 mean 24.7.
Ostro et al. (2001) studied exacerbation of asthma in
African-American children in Los Angeles. PM was
measured by a beta-attenuated Andersen monitor. PM1(
ranged from 21 to 119 ,ug/m3 with a mean of 51.8.
Study of 12 asthmatic children with history of
bronchodilator use. A random effects model
was fitted for ordinal symptoms scores and
bronchodilator use in relation to 24-hr PM,,.
A panel of 9 adults and 13 children were
followed during late spring 1994 in semi-rural
area of San Diego County at the inversion
zone elevation of around 1,200 feet. A
random effects model was fitted to ordinal
symptom scores, bronchodilator use, and PEE
in relation to 24-hour PM10. Temp., relative
humidity, fugal spores, day of week and O3
evaluated
Relationship of asthma symptoms to O3 and
PM10 examined in a So. California
community with high O3 and low PM. Panel
of 25 asthmatics ages 9-17 followed daily,
Aug. - Oct., 1995. Longitudinal regression
analyses utilized GEE model controlling for
autocorrelation, day of week, outdoor fangi
and weather.
Daily diary records were collected from
November 1993 through August 1995 during
screening for the CAMP study. A repeated
measures logistic regression analysis was used
applied using GEE methods
138 children aged 8 to 13 years who had
physician diagnosed asthma were included.
A daily diary was used to record symptoms
and medication use. GEE methods were used
to estimate the effects of air pollution on
symptoms controlling for meteorological and
temporal variables.
Pollen not associated with asthma symptom
scores. 12-hr personal O3 but not ambient
O3 related to symptoms.
Although PM10 never exceeded 51 i/g/m3,
bronchodilator use was significantly
associated with PM10(0.76 [0.027, 0.27])
puffs per 50 //g/m3. Fungal spores were
associated with all respiratory outcomes.
Asthma symptoms scores significantly
associated with both outdoor O3 and PM10
in single pollutant and co-regressions. 1-hr
and 8-hr maxi PM10 had larger effects than
24-hr mean.
One day lag CO and PM10 levels and the
same day PM10 and S)2 levels had the
strongest effects on asthma symptoms after
controlling for subject specific variables
and time-dependent confounders.
Symptoms were generally related to PM10
and NO2, but not to ozone. Reported
associations were for pollutant variables
lagged 3 days. Results for other lag times
were not reported.
No significant relationships with PM10.
24-h-1.47 (0.90-2.39)
8-h-2.17 (1.33-3.58)
1-h-1.78 (1.25-2.53)
OR symptom = 1.18 (1.05, 1.33) (PM10 same
day)
OR symptom = 1.17 (1.04, 1.33) (PM10 one
day lag)
24-h
OR wheeze = 1.02 (0.99, 106) )PM10 lag 3
days)
OR cough = 1.06(1.02, 1.09) (PM10 lag 3
days)
OR shortness of breath = 1.08 (1.02, 1.13)
(PM10 lag 3 days)
1-h
OR cough = 1.05 (1.02, 1.18) lag 3 days
-------
to
o
o
to
TABLE 8B-5 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OFASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
(^
OO
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Thurstonetal. (1997)
Summers 1991-1993.
O3, H+, sulfate, pollen, daily max temp, measured.
Canada
Vedaletal. (1998)
PM10 measured by Sierra-Anderson dichotomous sampler
PM10 range: -1 to 159 ,ug/m3
Port Alberrni
British, Columbia
Europe
Gielenetal. (1997)
Amsterdam, NL
PM10 and ozone.
PM10 was measured using a Sierra-Anderson
dichotomous sampler. PM10 ranged from 15 to 60 ,ug/m3.
Hiltermannetal. (1998)
Leiden, NL
July-Oct 1995.
Ozone, PM10, NO2, SO2, BS
PM10 ranged from 16 to 98 ,ug/m3 with a mean of 40.
Hiltermannetal. (1997)
The Netherlands
Ozone and PM10
PM10 averaged 40 ,ug/m3,
Three 5-day summer camps conducted in
1991, 1992, 1993. Study measured
symptoms and change in lung function
(morning to evening). Poisson regression for
symptoms.
206 children aged 6 to 13 years, 75 with
physician's diagnosis of asthma. Respiratory
symptom data from diaries, GEE model.
Temp., humidity.
Study of 61 children aged 7 to 13 years living
in Amsterdam, NL. 77 percent were taking
asthma medication and the others were being
hospitalized for respiratory problems.
Respiratory symptoms recorded by parents in
diary. Associations of air pollution evaluated
using time series analyses, adjusted for pollen
counts, time trend, and day of week.
Study of 270 adult asthmatic patients from an
out-patient clinic in Leiden, NL from July 3,
to October 6, 1995. Respiratory symptom
data obtained from diaries. An autoregressive
model was fitted to the data. Covariates
included temperature and day of week.
Sixty outpatient asthmatics examined for
nasal inflammatory parameters in The
Netherlands from July 3 to October 6, 1995.
Associations of log transformed inflammatory
parameters to 24-h PM10 analyzed, using a
linear regression model. Mugwort-pollen and
O, were evaluated.
Ozone related to respiratory symptoms
No relationship between symptoms and
other pollutants.
PM10 associated with respiratory symptoms.
Strongest relationships found with O3,
although some significant relationships
found with PM10.
PM10, O3, and NO2 were associated with
changes in respiratory symptoms.
Inflammatory parameters in nasal lavage of
patients with intermittent to severe
persistent asthma were associated with
ambient O3 and allergen exposure, but not
with PM10 exposure.
LagO
Cough OR = 1.08(1.00, 1.16)perlO//g/m3
PM,n increments
Lag 0, Symptoms:
Cough OR = 2.19 (0.77, 6.20)
Branch. Dial. OR = 0.94 (0.59, 1.50)
Lag 2, Symptoms:
Cough OR = 2.19 (0.47, 10.24)
Branch. Dial. OR = 2.90 (1.80, 4.66)
Lag 0, Symptoms:
Cough OR = 0.93 (0.83, 1.04)
Short, breath OR = 1.17 (1.03, 1.34)
7 day average, Symptoms:
Cough OR = 0.94 (0.82, 1.08)
Short, breath OR = 1.01 (0.86, 1.20)
-------
to
o
o
to
TABLE 8B-5 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 //g/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Peters etal. (1997b)
Erfurt, Germany
PM fractions measured over range of sizes from ultrafme
to fine, including PM10.
Mean PM10 level: 55 ,ug/m3 (max 71).
Mean SO2: 100 ,ug/m3 (max 383).
PM was measured using a Harvard impactor.
Peters etal. (1997c)
Sokolov, Czech Republic
Winter 1991-1992
PM10, SO2, TSP, sulfate, and particle strong acid.
Median PM10: 47 ^g/m3 (29, 73).
Median SO2: 46 ^g/m3 (22, 88).
PM was measured using a Harvard impactor. Particle
size distributions were estimated using a conduction
particle counter.
Peters etal. (1997c)
Sokolov, Czech Republic
PM10 one central site. SO4 reported.
MeanPM10: 55 Mg/m3, max 177^g/m3.
SO4 - fine: mean 8.8 |/g/m3, max 23.8 ,ug/m3. PM was
measured using a Harvard impactor. Particle size
distributions were estimated using a conduction particle
counter.
Neukirch etal. (1998)
Paris, France
SO2, NO2, PM13 and BS.
PM was measured by radiometry.
PM13 ranged from 9 to 95 ,ug/m3 with a mean of 34.
Study of 27 non-smoking adult asthmatics
living in Erfurt, Germany during winter
season 1991-1992. Diary used to record
presence of cough. Symptom information
analyzed using multiple logistic regression
analysis.
Study of 89 children with asthma in Sokolov,
Czech Republic. Subjects kept diaries and
measured peak flow for seven months during
winter of 1991-2. Logistic regression for
binary outcomes used. First order
autocorrelations were observed and corrected
for using polynomial distributed lag
structures.
Role of medication use evaluated in panel
study of 82 children, mean ages 9.8 yr., with
mild asthma in Sokolov, Czech Republic
Nov. 1991 - Feb 1992. Linear and logistic
regression evaluated PM10, SO2, temp, RH
relationships to respiratory symptoms.
Panel of 40 nonsmoking adult asthmatics in
Paris studied. GEE models used to associate
health outcomes with air pollutants. Models
allowed for time-dependent covariates,
adjusting for time trends, day of week, temp.
and humidity.
Weak associations found with 5 day mean
sulfates and respiratory symptoms.
Significant relationships found between
TSP and sulfate with both phlegm and
runny nose.
Medicated children, as opposed to those not
using asthma medication, increased their
beta-agonist use in direct association with
increases in 5-day mean of SO4 particles
<2.5 ,um, but medication did not prevent
decrease in PEF and increase in prevalence
of cough attributable to PM air pollution.
Significant relationships found for
incidence of respiratory symptoms and
three or more day lags of SO2, and NO2.
Only selected results were given.
Lag 0, PM10:
Cough OR = 1.32(1.16, 1.50)
Feeling ill OR = 1.20(1.01, 1.44)
5 Day Mean, PM10:
Cough OR = 1.30(1.09, 1.55)
Feeling ill OR = 1.47 (1.16, 1.86)
LagO, PM25:
Cough OR = 1.19(1.07, 1.33)
Feeling ill OR = 1.24(1.09, 1.41)
5 Day Mean, PM25:
Cough OR = 1.02(0.91, 1.15)
Feeling ill OR = 1.21 (1.06, 1.38)
Lag 0, Symptoms:
Cough OR = 1.01 (0.97, 1.07)
Phlegm OR =1.13 (1.04, 1.23)
5 Day Mean, Symptoms:
Cough OR = 1.10(1.04, 1.17)
Phlegm OR =1.17 (1.09, 1.27)
Cough 1.16 (1.00, 1.34) 6.5 ^g/m3 increase
5-day mean SO4
5-d Mean SO4/increase of 6.5 ,ug/m3
Beta-Agonist Use 1.46 (1.08, 1.98)
Theophylline Use 0.99 (0.77, 1.26)
No PM10 analysis
Significant relationships found between
incidence of respiratory symptoms and three
or more day lags of PM13.
-------
to
o
o
to
TABLE 8B-5 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates
included, analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Segalaetal. (1998)
Paris, France
SO2, NO2, PM13 (instead of PM10), and BS.
PM was measured by p-radiometry.
Giintzeletal. (1996)
Switzerland
SO2, NO2, TSP
Taggartetal. (1996)
Northern England
SO2, NO2 and BS.
Latin America
Romieuetal. (1997)
Mexico City, Mexico
During study period, max daily 1-h O3 range: 40 to 390
ppb (mean 196 ppb SD = 78 ppb)
PM10 daily average range: 12 to 126 ,ug/m3.
PM was measured by a Harvard impactor.
Romieuetal. (1996)
During study period, max daily range: 40 to 370 ppb
(mean 190 ppb, SD = 80 ppb).
24 h ave. PM10 levels range: 29 to 363 ,ug/m3 (mean
166.8 Mg/m3, SD 72.8 ,ug/m3).
PM10 levels exceeded 150 ,ug/m3 for 53% of study days.
24-h ave. PM25 levels range 23-177 i/g/m3 (mean
85.7,ug/m3)
PM was measured by a Harvard impactor.
Study of 43 mildly asthmatic children aged
7-15 yr in Paris. Patients followed Nov. 15,
1992 to May 9, 1993. Respiratory symptoms
recorded daily in diary. An autoregressive
model fitted to data using GEE methods.
Covariates included temp, and humidity.
An asthma reporting system was used in
connection with pollutant monitoring in
Switzerland from fall of 1988 to fall 1990.
A Box-Jenkins ARIMA time series model was
used to relate asthma to TSP, O3, SO2, and
NO2 after adjusting for temperature.
Panel of 38 adult asthmatics studied July 17
to Sept. 22, 1993 in northern England. Used
generalized linear model to relate pollutants to
bronchial hyper-responsiveness, adjusting for
temperature.
Study of 65 children with mild asthma aged
5-13 yr living in southwest Mexico City.
Respiratory symptoms recorded by the
parents in daily diary. An autoregressive
logistic regression model used to analyze
presence of respiratory symptoms.
Study of 71 children with mild asthma aged
5-7 yr living in northern Mexico City.
Respiratory symptoms recorded by parents in
daily diary. An autoregressive logistic
regression model was used to analyze the
presence of respiratory symptoms.
Effects found related to PM13 were less than
those found related to the other pollutants.
No significant relationships found.
Small effects seen in relation to NO2 and
BS.
Strongest relationships found between O3
and respiratory symptoms.
Cough and LRI were associated with
increased O3 and PM10 levels.
Lag 2, Symptoms:
Short. Breath OR = 1.22 (0.83, 1.81)
Resp. Infect. OR = 1.66 (0.84, 3.30)
Lag 0, Symptoms:
Cough OR = 1.05(0.92, 1.18)
Phlegm OR =1.05 (0.83, 1.36)
Diff Breath OR = 1.13 (0.95, 1.33)
Lag 2, Symptoms:
Cough OR = 1.00(0.92, 1.10)
Phlegm OR = 1.00(0.86, 1.16)
Diff. Breath OR = 1.2(1.1, 1.36)
PM10 (lag 0) increase of 50 //g/m3 related to:
LRI= 1.21 (1.10, 1.42)
Cough = 1.27(1.16, 1.42)
Phlegm = 1.21 (1.00, 1.48)
PM25 (lag 0) increase of 25 ,ug/m3 related to:
LRI= 1.18(1.05, 1.36)
Cough = 1.21(1.05, 1.39)
Phlegm = 1.21 (1.03, 1.42)
-------
Appendix 8B.6: Short-Term PM Exposure Effects
On Pulmonary Function in Nonasthmatics
April 2002 8B-61 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
TABLE 8B-6. SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
ON
to
fe
H
6
o
o
H
O
O
H
W
O
UnitedStates
Hoeketal. (1998)
(summary paper)
Lee and Shy (1999)
North Carolina
Mean 24 h PM10 cone, over two years: 25.1 ,ug/m3.
Korricketal. (1998)
Mt. Washington, NH
O3 levels measured at 2 sites near top of the mountain.
PM2 5 measured near base of the mountain.
PM was measured by a Harvard impactor.
Naeheretal. (1999)
Virginia
PM10, PM25, sulfate fraction, H+, and ozone
Neas etal. (1996)
State College, PA
PM21. mean 23.5; max 85.8 ,ug/m3.
Results summarized from several other studies
reported in the literature. These included:
asymptomatic children in the Utah Valley (Pope
et al., 1991), children in Bennekom, NL (Roemer
et al., 1993), children in Uniontown, PA (Neas et
al., 1995), and children in State College, PA
(Neas et al., 1996). Analyses done using a first-
order autoregressive model with adjustments for
time trend and ambient temp.
Study of the respiratory health status of residents
whose households lived in six communities near
an incinerator in southwestern North Carolina.
Daily PEER measured in the afternoon was
regressed against 24 hour PM10 level lagged by
one day. Results were adjusted for gender, age,
height, and hypersensitivity.
Study of the effects of air pollution on adult
hikers on Mt. Washington, NH. Linear and non-
linear regressions used to evaluate effects of
pollution on lung function.
Daily change in PEE studied in 473 non-smoking
women in Virginia during summers 1995-1996.
Separate regression models run, using normalized
morning and evening PEE for each individual.
Study of 108 children in State College, PA, during
summer of 1991 for daily variations in symptoms
and PEFRs in relation to PM2 [ An autoregressive
linear regression model was used. The regression
was weighted by reciprocal number of children of
each reporting period. Fungus spore cone., temp.,
O, and SO2 were examined.
Other pollutants not considered.
PM10 was not related to variations in
respiratory health as measured by
PEER.
PM2 5 had no effect on the O3
regression coefficient.
Ozone was only pollutant related to
evening PEE.
Spore concentration associated with
deficient in morning PERF.
Significant decreases in peak flow found to be
related to PM10 increases.
Morning PEE decrements were associated with
PM10, PM2 5, and H+. Estimated effect from
PM2 5 and PM10 was similar. No PM effects
found for evening PEE.
PM2 j (25 //g/m3) related to RR of:
PM PFER (lag 0) = -0.05 (-1.73, 0.63)
PM PEER (lag 1)= -0.64 (-1.73, 0.44)
O
-------
to
o
o
to
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 //g/m3
PM10 (25 fj.g/m1 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
United States (cont'd)
Neasetal. (1999)
Philadelphia, PA
Median PM10 level: 31.6 in SW camps,
27.8 in NE camps (IQR ranges of about 18).
Median PM2 5 level: 22.2 in the SW camps,
20.7 in NE camps (IQR ranges about 16.2 and 12.9,
respectively).
Particle-strong acidity, fine sulfate particle, and O3 also
measured.
Schwartz and Neas (2000)
Eastern U.S.
PM2 5 and CM (PM10.2 5) measured.
Summary levels not given.
Linn etal. (1996)
So. California
NO2 ozone, and PM5 measured.
PM5 was measured using a Marple low volume sampler PM5
ranged from 1-145 //g/m3 with a mean of 24.
Europe
Boezen etal. (1999)
Netherlands
PM10, BS, SO2, and NO2 measured, but methods were not
given. PM10 ranged from 4.8 to 145//g/m3 with site means
ranging from 26 to 54 //g/m3.
Panel study of 156 normal children attending
YMCA and YWCA summer camps in greater
Philadelphia area in 1993. Children followed for
at most 54 days. Morning and evening deviations
of each child's PEE were analyzed using a mixed-
effects model adjusting for autocorrelation.
Covariates included time trend and temp. Lags
not used in the analysis.
Analyses for 1844 school children in grades 2-5
from six urban areas in eastern U.S. and from
separate studies from Uniontown and State
College, PA. Lower resp. symptoms, cough and
PEE used as endpoints. The authors replicated
models used in the original analyses. CM and
were used individually and jointly in the analyses.
Sulfate fractions also used in the analyses.
Details of models not given.
Study of 269 school children in Southern
California twice daily for one week in fall, winter
and spring for two years. A repeated measures
analysis of covariance was used to fit an
autoregressive model, adjusting for year, season,
day of week, and temperature.
Data collected from children during three winters
(1992-1995) in rural and urban areas of The
Netherlands. Study attempted to investigate
whether children with bronchial
hyperresponsiveness and high serum Ige levels
were more susceptible to air pollution. Prevalence
of a 10 percent PEE decrease was related to
pollutants for children with bronchial
hyperresponsiveness and high serum Ige levels.
Analyses that included sulfate fraction
and O3 separately also found
relationship to decreased flow. No
analyses reported for multiple
pollutant models.
Sulfate fraction was highly correlated
with PM25 (0.94), and, not
surprisingly, gave similar answers.
Morning FVC was significantly
decreased as a function of PM5 and
NO,
No consistent pattern of effects
observed with any of the pollutants
for 0, 1, and 2 day lags.
Lag 0, PM10:
Morning PEE = -8.16 (-14.81, -1.55)
Evening PEE = -1.44 (-7.33, 4.44)
5 day ave, PM10
Morning PEE = 2.64 (-6.56, 11.83)
Evening PEE = 1.47 (-7.31, 10.22)
LagO,PM25
Morning PEE = -3.28 (-6.64, 0.07)
Evening PEE = -0.91 (-4.04, 2.21)
5 day ave., PM2 5
Morning PEE = 3.18 (-2.64, 9.02)
Evening PEE = 0.95 (-4.69, 6.57)
Uniontown Lag 0,PM25 :
Evening PEE = -1.52 (-2.80, -0.24)
State College Lag 0, PM25:
Evening PEE = -0.93 (-1.88, 0.01)
Results presented for CM showed no effect.
Results for PM10 were not given.
-------
to
o
o
to
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Frischeretal. (1999)
Austria
PM10 measured gravimettrically for 14-d periods.
Annual mean PM10 levels range: 13.6 - 22.9 ,ug/m3.
O3 range: 39.1 ppb - 18.5 pbs between sites.
Grievinketal. (1999)
Netherlands
PM10 and BS.
PM10 ranged from 12 to 123 //g/m3 with a mean of 44.
Kiinzli et al. (2000)
Roemer et al. (2000)
PM10 means for 17 panels ranged 11.2 to 98.8 ,ug/m3.
SO2, NO2, and elemental content of PM also measured.
Measurement methods were not described.
Scarlett etal. (1996)
PM10, O3, and NO2 measured.
At nine sites in Austria during 1994, 1995, and
1996, a longitudinal study designed to evaluate O3
was conducted. During 1994 - 1996, children
were measured for FVC, FEVj and MEF50 six
times, twice a year in spring and fall. 1060
children provided valid function tests. Mean age
7.8±0.7yr. GEE models used. PM10, SO2, NO2,
and temp, evaluated.
A panel of adults with chronic respiratory
symptoms studied over two winters in The
Netherlands starting in 1993/1994. Logistic
regression analysis was used to model the
prevalence of large PEF decrements.
Individual linear regression analysis of PEF on
PM was calculated and adjusted for time trends,
influenza incidence, and meteorological variables.
Ackermann-Liebrich et al. (1997) data reanalyzed.
Authors showed that a small change in FVC
(-3.14 percent) can result in a 60% increase in
number of subjects with FVC less than 80 percent
of predicted.
Combined results from 1208 children divided
among 17 panels studied. Separate results
reported by endpoints included symptoms as
reported in a dairy and PEF. Individual panels
were analyzed using multiple linear regression
analysis on deviations from mean PEF adjusting
for auto-correlation. Parameter estimates were
combined using a fixed-effects model where
heterogeneity was not present and a random-
effects model where it was present.
In study of 154 school children, pulmonary
function was measured daily for 31 days.
Separate autoregressive models for each child
were pooled, adjusting for pollen, machine,
operator, time of day, and time trend.
Small but consistent lung function
decrements in cohort of school
children associated with ambient O3
exposure.
Subjects with low levels of serum
P-carotene more often had large PEF
decrements when PM10 levels were
higher, compared with subjects with
high serum P-carotene.
Results suggested serum P-carotene
may attenuate the PM effects on
decreased PEF.
The results were for two hypothetical
communities, A and B.
Daily concentrations of most elements
were not associated with the health
effects.
PM10 was related to changes in FEV
and FVC
PM10 showed little variation in exposure
between study site. For PM10, positive effect
seen for winter exposure but was completely
confounded by temperature.
PM10 Summertime
P = 0.003 SE 0.012 p=0.77
PM10 analyses not focus of this paper.
-------
to
o
o
to
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
van der Zee etal. (1999)
Netherlands
PM10 averages ranged 20 to 48 ,ug/m3.
BS, sulfate fraction, SO2, and NO2 also measured.
van der Zee et al. (2000)
Netherlands
PM10 averages ranged 24 to 53 /2g/m3.
BS, sulfate fraction, SO2, and NO2 also measured.
PM10 was measured using a Sierra Anderson 241
dichotomous sampler.
Tiittanen etal. (1999)
Kupio, Finland
Median PM10 level: 28 (25th, 75th percentiles = 12, 43).
Median PM25 level: 15 (25th, 75th percentiles = 9, 23).
Black carbon, CO, SO2, NO2, and O3 also measured.
PM was measured using single stage Harvard samplers.
Panel study of 795 children aged 7 to 11 years,
with and without chronic respiratory symptoms
living in urban and nonurban areas in the
Netherlands. Peak flow measured for three
winters starting in 1992/1993. Peak flow
dichotomized at 10 and 20% decrements below
the individual median. Number of subjects was
used as a weight. Minimum temperature day of
week, and time trend variables were used as
covariates. Lags of 0, 1 and 2 days were used, as
well as 5 day moving average.
Panel study of 489 adults aged 50-70 yr, with and
without chronic respiratory symptoms, living in
urban and nonurban areas in the Netherlands.
Resp. symptoms and peak flow measured for three
winters starting in 1992/1993. Symptom
variables analyzed as a panel instead of using
individual responses. The analysis was treated as
a time series, adjusting for first order
autocorrelation. Peak flow dichotomized at
10 and 20% decrements below the individual
median. The number of subjects used as a weight.
Minimum temp., day of week, and time trend
variables used as covariates. Lags of 0, 1 and
2 days used, as well as 5 day moving average.
Six-week panel study of 49 children with chronic
respiratory disease followed in the spring of 1995
in Kuopio, Finland. Morning and evening
deviations of each child's PEF analyzed, using a
general linear model estimated by PROC MIXED.
Covariates included a time trend, day of week,
temp., and humidity. Lags of 0 through 3 days
were used, as well as a 4-day moving average.
Various fine particles were examined.
In children with symptoms,
significant associations found
between PM10, BS and sulfate fraction
and the health endpoints. No multiple
pollutant models analyses reported.
BS tended to have the most consistent
relationship across endpoints. Sulfate
fraction also related to increased
respiratory effects. No analyses
reported for multiple pollutant
models. Relationship found between
PM10 and the presence of 20%
decrements in symptomatic subjects
from urban areas.
Ozone strengthened the observed
associations. Introducing either NO2
or SO2 in the model did not change
the results markedly. Effects varied
by lag. Separating effects by size was
difficult.
Lag 0, PM10, Urban areas
Evening PEF OR = 1.15 (1.02, 1.29)
Lag 2, PM10, Urban areas
Evening PEF OR = 1.07 (0.96, 1.19)
5 day ave, PM10, Urban areas
Evening PEF = 1.13 (0.96, 1.32)
Lag 0, PM10, Urban areas
Morning large decrements
OR= 1.44(1.02,2.03)
Lag 2, PM10, Urban areas
Morning large decrements
OR= 1.14(0.83, 1.58)
5 day ave, PM10, Urban areas
Morning large decrements
OR= 1.16(0.64,2.10)
Results should be viewed with caution because
of problems in analysis.
Lag 0, PM10:
Morning PEF = 1.21 (-0.43, 2.85)
Evening PEF = 0.72 (-0.63, 1.26)
4 day ave, PM10
Morning PEF= -1.26(-5.86, 3.33)
Evening PEF = 2.33 (-2.62, 7.28)
LagO,PM25
Morning PEF = 1.11 (-0.64, 2.86)
Evening PEF = 0.70 (-0.81, 2.20)
4 day ave., PM25
Morning PEF = -1.93 (-7.00, 3.15)
Evening PEF = 1.52 (-3.91, 6.94)
-------
to
o
o
to
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
Europe (cont'd)
oo
td
ON
Oi
fe
H
6
o
o
H
O
O
H
W
O
Ward et al. (2000)
West Midlands, UK
Daily measurements of PM10, PM2 5, SO2, CO, O3, and
oxides of nitrogen.
Details on PM monitoring were inomplete.
Osunsanya et al. (2001) studied 44 patients aged > 50 with
COPD in Aberdeen, UK. PM was measured using tapered
element oscillating microbalance. Particle sizes were
measured a TSI model 3934 scanning particle sizer. PM10
ranged from 6 to 34 /2g/m3 with a median of 13.
Cuijpers et al. (1994)
Maastricht, NL
SO2, NO2, BS, ozone, and H+ measured. PM measurements
were made with a modified Sierra Anderson sampler. PM10
ranged from 23 to 54 //g/m3.
Latin America
Gold etal. (1999)
Mexico City, Mexico
Mean 24 h O3 levels: 52ppb.
MeanPM25: 30 |/g/m3.
MeanPM10: 49,ug/m3.
Panel study of 9 yr old children in West Midlands,
UK for two 8-week periods representing winter
and summer conditions. Individual PEE values
converted to z-values. Mean of the z-values
analyzed in a linear regression model, including
terms for time trend, day of week, meteorological
variables, and pollen count. Lags up to four days
also used.
Symptom scores, bronchodilator use, and PEE
were recorded daily for three months. GEE
methods were used to analyze the dichotomous
outcome measures. PEE was converted to a
dichotomous measure by defining a 10 percent
decrement as the outcome of interest.
Summer episodes in Maastricht, The Netherlands
studied. Paired t tests used for pulmonary
function tests.
Peak flow studied in a panel of 40 school-aged
children living in southwest Mexico City. Daily
deviations from morning and afternoon PEFs
calculated for each subject. Changes in PEE
regressed on individual pollutants allowing for
autocorrelation and including terms for daily
temp., season, and time trend.
Results on effects of pollution on lung
function to be published elsewhere.
No associations were found between
actual PEE and PM10 or ultrafme
particles. A change of PM10 from 10 to
20 |/g/m3 was associated with a 14
percent decrease in the rate of high
scores of shortness of breath. A
similar change in PM10 was associated
with a rate of high scores of cough.
Small decreases in lung function
found related to pollutants.
O3 significantly contributed to
observed decreases in lung function,
but there was an independent PM
effect.
The endpoint was measured in terms of scores
rather than L/min.
Quantitative results not given.
Both PM2 5 and PM10 significantly related to
decreases in morning and afternoon peak flow.
Effects of the two pollutants similar in
magnitude when compared on percent change
basis.
O
-------
to
o
o
to
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3
PM10 (25 ,ug/m3 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
New Zealand
Harreetal. (1997)
Christchurch, NZ
SO2, NO2, PM10, and CO measured.
Details on monitoring methods and pollutant ranges were
not given.
Jalaludin et al. (2000) studied PEE in 148 children 6
primary schools in Sydney, Australia. PM was measured by
tapered element oscillating microbalance. Mean PM10 was
22.8+- 13.9 ,ug/m3.
Asia
Chen etal. (1999)
Taiwan
Beta-gauge PM10 ranged 44.5 to 189.0 ,ug/m3 for peak
concentrations.
Tan et al. (2000) Southeast Asian smoke-haze event
9/29-10/27 1997
PM10 mean daily was 125.4 ± 44.9 |/g/m3
ultra range of 47 to 216 ,ug/m3 in Singapore
Study of 40 subjects aged over 55 years with
COPD living in Christchurch, New Zealand
conducted during winter of 1994. Subjects
recorded their peak flow measurements. A log-
linear regression model with adjustment for first
order auto-correlation was used to analyze peak
flow data and a Poisson regression model was
used to analyze symptom data.
148 children in grades 3-5 were followed for 11
months, recording PEE twice daily. The
normalized change in PEE was analyzed using
GEE methods. PEE was related to SO3, PM10,
NO2, as well as meteorological variables.
In 3 Taiwan communities in 1995, PM10 by B-
gauge measured at selected primary schools in
each community. Spirometry tests (FVC, FEVLO,
FEF25.75%, PEE) obtained in period May 1995 to
Jan. 1996 using ATS protocol in study pop. aged
8 to 13 yr. 895 children were analyzed. Study
was designed to investigate short-term effect of
ambient air pollution in cross-sectional survey.
Multivariate linear model analysis used in both
one pollutant and multipollutant models, with 1-,
2-, and 7-day lags. SO2, CO, O3, NO2 and PM10
examined, as were meteorol. variables.
Examined the association between acute air
pollution caused by biomass burning and
peripheral UBC counts in human serial
measurement made during the event were
compared with a period after the haze cleared
(Nov. 21 -Dec. 5, 1997)
Few significant associations found
between the health endpoints and the
pollutants.
Daily mean deviations in PEE were
related to ozone, but no relationships
were found with PM10 or NO2.
Multiple pollutant models gave
similar results to those given by the
single pollutant models.
In the one-pollutant model, daytime
peak O3 cone, with a 1-day lag
significantly affected both FVC and
FEVj. NO2, SO2, CO affected FVC.
PM10 showed nonsignificant
decrement. No significant result
demonstrated in the model for the
exposure with 7 days lag. In the
multi-pollutant model, only peak O3
cone, with 1-day lag showed sig.
effect on FVC and FEVL0.
Indices of atmospheric pollution were
significantly associated in the
elevated band neutrophil counts
expressed as a percentage of total
polymonphonuclear leukocytes
(PMN). No statistically significant
difference in FEU! and FUC were
observed during and after haze
exposure.
Lag 0, PM10:
PEF=-0.£
(-2.33, 0.61)
Change from AM to PM PEE = 0.045 (-.205,
2.95) lag one day
One pollutant model daytime average
PM10-2daylag
FVC-0.37 se 0.39
-------
Appendix 8B.7: Short-Term PM Exposure Effects
On Symptoms in Nonasthmatics
April 2002 8B-68 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
Reference citation, location, duration,
pollutants measured, summary of values
TABLE 8B-7. SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 //g/m3
PM10 (25 fj.g/m1 PM2.5). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
ON
VO
fe
H
6
o
o
H
O
O
H
W
O
United States
Schwartz and Neas (2000)
Eastern U.S.
PM25 and CM (PM10_25 by substation)..
Summary levels not given
Zhang et al. (2000)
Vinton, Virginia
24- h PM10, PM25, sulfate and strong acid measured in
1995.
Canada
Long etal. (1998)
Winnepeg, CN
PM10, TSP, and VOC measured.
Methods for PM monitoring not given. Ranges of values
also not given.
Europe
Boezen etal. (1998)
Amsterdam, NL
PM10, SO2, and NO2 measured.
PM10 ranged from 7.9 to 242.2 //g/m3 with a median of 43.
Reported on analysis of 1844 school children in
grades 2-5 from six urban areas in the eastern
U.S., and from separate studies from Uniontown
and State College, PA. Lower respiratory
symptoms, and cough used as endpoints. The
authors replicated the models used in the original
analyses. CM and PM25 were used individually
and jointly in the analyses. Sulfates fractions
were also used in the analyses. Details of the
models were not given.
In southwestern Virginia, 673 mothers were
followed June 10 to Aug. 31, 1995 for the daily
reports of present or absence of runny or stuffy
nose. PM indicator, O3, NO2 temp., and random
sociodemographic characteristics considered.
Study of 428 participants with mild airway
obstruction conducted during a Winnepeg
pollution episode. Gender specific odds ratios of
symptoms were calculated for differing PM10
levels using the Breslow-Day test.
Study of 75 symptomatic and asymp. adults near
Amsterdam for three months during winter 1993-
1994. An autoregressive logistic model was used
to relate PM10 to respiratory symptoms, cough,
and phlegm, adjusting for daily min. temp., time
trend, day of week.
Sulfate fraction was highly correlated
with PM25 (0.94), and not
surprisingly gave similar answers.
Of all pollutants considered, only the
level of coarse particles as calculated
(PM10 - PM2 5) independently related
to incidence of new episode of runny
Cough, wheezing, chest tightness, and
shortness of breath were all increased
during the episode
No relationship found with pulmonary
function. Some significant
relationships with respiratory disease
found in subpopulations
PM2 5 was found to be significantly related to
lower respiratory symptoms even after
adjusting for CM, whereas the reverse was not
true. However, for cough, CM was found to be
significantly related to lower respiratory
symptoms even after adjusting for PM2 5,
whereas the reverse was not true.
O
-------
TABLE 8B-7 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3 PM10
(25 ,ug/m3 PM2.5). Negative coefficients for
lung function and ORs greater than 1 for other
endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
Europe (cont'd)
Howel et al. (2001) study of children's respiratory
health in 10 non-urban communities of northern
England. PM levels were measured using a single
continuous real-time monitor. PM10 levels ranged
from 5 to 54 ,ug/m3.
Roemeretal. (1998)
Mean PM10 levels measured at local sites ranged
11.2 to 98.8 ,ug/m3 over the 28 sites.
Roemer et al. (2000)
PM10 means for the 17 panels ranged 11.2 to
98.8 //g/m3.
SO2, NO2, and PM elemental content also measured.
Measurement methods were not described.
The study included 5 pairs of non-urban
communities near and not so near 5 coal mining
sites. 1405 children aged 1-11 years were
included. 275 of the children reported having
asthma. Diaries of respiratory symptoms were
collected over a 6 week period. PM10, measured
by a single continuous real-time monitor, ranged
from 5 to 54 ,ug/m3.
Pollution Effects on Asthmatic Children in
Europe (PEACE) study was a multi-center study
of PM10, BS, SO2, and NO2 on respiratory health
of children with chronic respiratory symptoms.
Results from individual centers were reported by
Kotesovec et al. (1998), Kalandidi et al. (1998),
Haluszka et al. (1998), Forsberg et al. (1998),
Clench-Aas et al. (1998), and Beyer et al. (1998).
Children with chronic respiratory symptoms
were selected into the panels. The symptom with
one of the larger selection percentages was dry
cough (range over sample of study communities
29 to 92% [22/75; 84/91] with most values over
50%). The group as a whole characterized as
those with chronic respiratory disease, especially
cough.
Combined results from 1208 children divided
among 17 panels studied. Endpoints included
symptoms as reported in a dairy and PEE.
Symptom variables analyzed as a panel instead of
using individual responses. The analysis was
treated as a time series, adjusting for first order
autocorrelation. Parameter estimates were
combined using a fixed-effects model where
heterogeneity was not present and a random-
effects model where it was present.
The associations found between daily
PM10 levels and respiratory symptoms
were frequently small and positive and
sometimes varied by community.
These studies modeled group rates and
are an example of the panel data
problem.
OR wheeze = 1.16(1.05, 1.28(PM10)
OR cough = 1.09(1.02, 1.16)(PM10)
OR reliever use = 1.00 (0.94, 1.06) (PM10)
Daily concentrations of most elements
were not associated with the health
effects.
The analysis of PM1(
paper.
was not a focus of this
-------
to
o
o
to
TABLE 8B-7 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3 PM10
(25 fj.g/m1 PM2.5). Negative coefficients for
lung function and ORs greater than 1 for other
endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
van der Zee etal. (1999)
Netherlands
PM10 averages ranged 20 to 48 ,ug/m3.
BS, sulfate fraction, SO2, and NO2 also measured.
van der Zee et al. (2000)
Netherlands
Daily measurements of PM10, BS, fine sulfate, nitrate,
ammonium and strong acidity.
PM10 was measured using a Sierra Anderson 241
dichotomous sampler.
Tiittanen etal. (1999)
Kupio, Finland
Median PM10 level: 28 (25th, 75th percentiles = 12,
43).
Median PM25: 15 (25th and 75th percentiles of 9 and
23). Black carbon, CO, SO2, NO2, and O3 also
measured. PM was measured using single stage
Harvard samplers.
Keles etal. (1999)
Istanbul, Turkey
Nov. 1996 to Jan. 1997.
TSP levels ranged from annual mean of 22 ,ug/m3 in
unpolluted area to 148.8 //g/m3 in polluted area.
A panel study of 795 children aged 7 to 11 yr,
with and without chronic respiratory symptoms,
living in urban and nonurban areas in the
Netherlands. Respiratory symptoms measured for
3 winters starting 1992/1993. Symptom variables
analyzed as a panel instead of using individual
responses. The analysis was treated as a time
series, adjusting for first order autocorrelation.
The number of subjects was used as a weight.
Minimum temp., day of week, and time trend
variables used as covariates. Lags of 0, 1 and 2
days used, as well as 5 day moving average.
Panel study of adults aged 50 to 70 yr during 3
consecutive winters starting in 1992/1993.
Symptom variables analyzed as a panel instead of
using individual responses. Analysis treated as a
time series, adjusting for first order
autocorrelation. Number of subjects used as a
weight. Min. temp., day of week, time trend
variables used as covariates. Lags 0, 1 and 2
days used, as well as 5 day moving average.
Six-week panel study of 49 children with chronic
respiratory disease followed in spring 1995 in
Kuopio, Finland. Cough, phlegm, URS, LRS and
medication use analyzed, using a random effects
logistic regression model (SAS macro
GLIMMFX). Covariates included a time trend,
day of week, temp., and humidity. Lags of 0 to 3
days used, as well as 4-day moving average.
Symptoms of rhinitis and atopic status were
evaluated in 386 students grades 9 and 10 using
statistical package for the social sciences, Fisher
tests, and multiple regression model as
Spearman's coefficient of correlation.
In children with symptoms, significant
associations found between PM10, BS
and sulfate fraction and the health
endpoints. No analyses reported with
multiple pollutant models.
BS was associated with upper
respiratory symptoms.
Ozone strengthened the observed
associations. Introducing either NO2 or
SO2 in the model did not change the
results markedly.
No difference found for atopic status in
children living in area with different air
pollution levels.
Lag 0, PM10, Urban areas
Cough OR = 1.04(0.95, 1.14)
Lag 2, PM10, Urban areas
Cough OR = 0.94 (0.89, 1.06)
5 day ave, PM10, Urban areas
Cough OR = 0.95 (0.80, 1.13)
Lag 0, Symptoms, Urban areas
LRS OR = 0.98 (0.89, 1.08)
URS OR = 1.04(0.96, 1.14)
Lag 2, Symptoms, Urban areas
LRS OR = 1.01(0.93, 1.10)
URS OR = 1.04(0.96, 1.13)
5 day ave, Symptoms, Urban areas
LRS OR = 0.95 (0.82, 1.11)
URS OR = 1.17(1.00, 1.37)
Lag 0, PM10:
Cough OR = 1.00 (0.87, 1.16)
4 day ave, PM10
Cough OR = 1.58(0.87,2.83)
Lag 0, PM2 5
Cough OR = 1.04(0.88, 1.23)
4 day ave., PM25
Cough OR = 2.01 (1.04,3.89)
-------
to
o
o
to
TABLE 8B-7 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON SYMPTOMS
IN STUDIES OF NONASTHMATICS
Effect measures standardized to 50 ,ug/m3 PM10
(25 fj.g/m1 PM2.5). Negative coefficients for
lung function and ORs greater than 1 for other
endpoints suggest PM effects
Reference citation, location, duration,
pollutants measured, summary of values
Type of study, sample size, health outcomes
measured, analysis design, covariates included,
analysis problems, etc.
Results and Comments
Effects of co-pollutants
OO
td
ij
to
New Zealand
Harreetal. (1997)
Christchurch, NZ
SO2, NO2, PM10, and CO measured.
Details on monitoring methods and pollutant ranges
were not given.
Asia
Awasthietal. (1996)
India
Suspended particulate matter, SO2, nitrates, coal,
wood, PM and kerosene measured. SPM was
measured using a high-volume sampler.
Study of 40 subjects aged 55 years with COPD
living in Christchurch, New Zealand during
winter 1994. Subjects recorded completed diaries
twice daily. Poisson regression model used to
analyze symptom data.
A cohort of 664 preschool children studied for
two weeks each in northern India. Ordinary least
squares was used to relate a respiratory symptom
complex pollutants.
NO2 was associated with increased
bronchodilator use.
PM10 was associated with increased nighttime
chest symptoms.
A significant regression coefficient
between PM and symptoms was found
fe
H
6
o
o
H
O
O
H
W
O
o
-------
Appendix 8B.8: Long-Term PM Exposure Effects On
Respiratory Health Indicators, Symptoms, and Lung Function
April 2002 8B-73 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
to
TABLE 8B-8. LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
United States
Abbey etal. (1998)
California Communities
20 year exposure to respirable particulates,
suspended sulfates, ozone, and PM10.
PM10 ranged from 1 to 145 ,ug/rn3 with a mean
value of 32.8.
Sex specific multiple linear regressions were
used to relate lung function measures to
various pollutants in long-running cohort study
of Seven Day Adventists (ASHMOG Study).
Sulfates were associated with
decreases in FEV.
Frequency of days where PM10 >
100 Aig/m3 associated with FEV
decrement in males whose parents had
asthma, bronchitis, emphysema, or hay
fever. No effects seen in other subgroups.
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Berglundetal.. (1999)
California communities
Peters etal. (1999a,b)
12 demographically similar communities in
So. California.
O3, PM acids, and NO2 evaluated.
PM was measured using a tapered element
oscillating microbalance instrument.
Avol etal. (2001)
Subjects living in Southern California in 1993
that moved to other western locations in 1998.
Pollutants O3, NO2, PM10 differences 15 to
66 ,ug/m3.
Cohort study of Seventh Day Adventists.
Multivariate logistic regression analysis of risk
factors (e.g., PM) for chronic airway disease in
elderly non-smokers, using pulmonary function
test and respiratory symptom data.
Stepwise logistic regression was used to relate
prevalence rates for symptoms to community-
specific ambient pollutants after adjustment for
race, sex, asthma, body mass, hay fever, and
membership in an insurance plan.
Studied 110 children who were 10 yrs of age at
enrollment and 15 at follow-up who had
moved from communities filled out health
questions and underwent spirometry. Linear
regression used to determine whether annual
average change in lung function correlated
with average changes in PM.
Significant risk factors identified:
childhood respiratory illness,
reported ETS exposure, age, sex and
parental history.
Wheeze prevalence was associated
with both acid and NO,.
As a group, subjects who moved to
areas of lower PM10 showed
increased growth in lung function
and subjects who moved to
communities with a higher PM10
showed decreased growth in lung
function.
For PM10 > 100Mg/m3, 42 d/yr:
RR = -1.09 CT (0.92, 1.30) for
obstructive disease determined by
pulmonary function tests.
No significant relationships were found
between PM10 and symptoms.
PM10 24 hr average
PERFml/sperlOMg/m3
mean = -34.9
95% CI
-59.8,-10.1
O
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
Gauderman et al. (2000)
12 So. California communities 1993 to 1997
Pollutants: O3, NO2, PM10, and PM25.
PM10 levels ranged from 16.1 to 67.6 Mg/m3
across the communities.
McConnell et al. (1999)
12 Southern California communities
1994 air monitoring data.
PM10 (mean 34.8; range 13.0 - 70.
PM2 5 (yearly mean 2 week averaged mean
15.3 /-ig/m3', range 6.7 - 31.5 ,ug/m3).
Studies of lung function growth of 3035
children in 12 communities within 200-mile
radius of Los Angeles during 1993 to 1997.
Cohorts of fourth, seventh, and tenth-graders
studied. By grade cohort, a sequence of linear
regression models were used to determine over
the 4yr of follow-up, if average lung function
growth rate of children was associated with
average pollutant levels. Adjustment were
made for height, weight, body mass index,
height by age interaction, report of asthma
activity or smoking. Two-pollutant models
also used.
Cross-sectional study of 3,676 school children
whose parents completed questionnaires in
1993 that characterized the children's history
of respiratory illness. Three groups examined:
(1) history of asthma; (2) wheezing but no
asthma; and (3) no history of asthma or
wheezing. Logistic regression model used to
analyze PM, O3, NO2, acid vapor effects. This
study also described in Peters et al. (1999b,c).
Lung growth rate for children in
most polluted community, as
compared to least polluted, was
estimated to result in cumulative
reduction of 3.4% in FEV! and
5.0% in MMEF over 4-yr study
period. Estimated deficits mostly
larger for children spending more
time outdoors. Due to the high
correlation in concentrations across
communities, not able to separate
effects of each pollutant. No sig.
associations seen with O3.
Positive association between air
pollution and bronchitis and phlegm
observed only among children with
asthma. As PM10 increased across
communities, a corresponding
increase in risk of bronchitis per
interquartile range occurred.
Strongest association with phlegm
was for NO2. Because of high
correlation of PM air pollution,
NO2, and acid, not possible to
distinguish clearly which most
likely responsible for effects.
From the lowest to highest observed
concentration of each pollutant, the
predicted differences in annual growth
rates were: -0.85% for PM10 (p = 0.026);
-0.64% for PM25 (p = 0.052); -0.90% for
PM10.2.5 (p = 0.030); -0.77% forNO2 (p =
0.019); and -0.73% for inorganic acid
vapor (p = 0.042).
PM10
Asthma
Bronchitis 1.4 CI( 1.1 - 1.8
Phlegm 2.1(1.4-3.3)
Cough 1.1(0.8- 1.7)
No Asthma / No Wheeze
Bronchitis 0.7 (0.4 - 1.0)
Phlegm 0.8(0.6- 1.3)
Cough 0.9(0.7- 1.2)
O
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
United States (cont'd)
McConnell et al. (2002)
12 Southern California communities
1994-1997
4-year mean cone. PM10 Mg/m3
High community: 43.3(12.0)
Low community: 21.6(3.8)
Dockeryetal. (1996)
24 communities in the U. S. and Canada.
PM10, PM25, sulfate fraction, H+, ozone, SO2,
and other measures of acid were monitored.
PM was measured using a Harvard impactor.
PM10 ranged form 15.4 to 32.7 with a mean of
23.8. PM25 ranged form 5.8 to 20.7 Mg/m3
with a mean of 14.5.
Raizenne et al. (1996)
24 communities in the U.S. and Canada
Pollutants measured for at least one year prior
to lung function tests: PM10, PM2l, particle
strong acidity, O3, NO2, and SO2. PM was
measured with a Harvard impactor. For
pollutant ranges, see Dockery et al. (1996).
In 3,535 children assessed, the association of
playing team sports with subsequent
development of asthma during 4 yrs of follow-
up. Comparing high pollutant communities to
low pollutant communities. Relative risks of
asthma adjusted for ethnic origin were
evaluated for every pollutant with a
multivariate proportional hazards model. See
also Peters et al. (1999b,c).
Respiratory health effects among 13,369 white
children aged 8 to 12 yrs analyzed in relation
to PM indices. Two-stage logistic regression
model used to adjust for gender, history of
allergies, parental asthma, parental education,
smoking in home.
Cross-sectional study of lung function. City
specific adjusted means for FEV and FVC
calculated by regressing the natural logarithm
of the measure on sex, In height, and In age.
These adjusted means were then regressed on
the annual pollutant means for each city.
Across all communities there was a
1.8-fold increased risk (95% CI
1.2-2.8) for asthma in children who
had played three or more team
sports in the previous year. In high
ozone (10:00 h to 18:00 h mean
concentration) communities, there
was a 3.3-fold increase risk of
asthma in children playing three or
more sports, an increase not seen in
low ozone communities.
Although bronchitis endpoint was
significantly related to fine PM
sulfates, no endpoints were related
to PM10 levels.
PM measures (e.g., particle strong
acidity) associated with FEV and
FVC decrement.
The effect of team sports was similar in
communities with high and low PM with
a small increase in asthma among children
playing team sports.
O
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
fe
H
6
o
o
H
O
O
H
W
O
Europe
Ackermann-Liebrich et al. (1997)
Eight Swiss regions
Pollutants: SO2, NO2, TSP, O3, and PM10.
PM was measured with a Harvard impactor.
PM10 ranged from 10 to 53 /
of37.
3 with a mean
Braun-Fahrlander et al. (1997)
10 Swiss communities
Pollutants: PM10, NO2, SO2, and O3.
PM was measured with a Harvard impactor.
PM10 ranged from 10 to 33 ,ug/m3.
Zempetal. (1999)
8 study sites in Switzerland.
Pollutants: TSP, PM10, SO2, NO2, and O3.
PM was measured with a Harvard impactor.
PM10 ranged from 10 to 33 ,ug/rn3 with a mean
of21.
Long-term effects of air pollution studied in
cross-sectional population-based sample of
adults aged 18 to 60 yrs. Random sample of
2,500 adults in each region drawn from
registries of local inhabitants. Natural
logarithms of FVC and FEV[ regressed against
natural logarithms of height, weight, age,
gender, atopic status, and pollutant variables.
Impacts of long-term air pollution exposure on
respiratory symptoms and illnesses were
evaluated in cross-sectional study of Swiss
school children, (aged 6 to 15 years).
Symptoms analyzed using a logistic regression
model including covariates of family history
of respiratory and allergic diseases, number of
siblings, parental education, indoor fuels,
passive smoking, and others.
Logistic regression analysis of associations
between prevalences of respiratory symptoms
in random sample of adults and air pollution.
Regressions adjusted for age, BMI, gender,
parental asthma, education, and foreign
citizenship.
Significant and consistent effects on
FVC and FEV were found for PM10,
NO, and SO,.
Respiratory endpoints of chronic
cough, bronchitis, wheeze and
conjunctivitis symptoms were all
related to the various pollutants.
The colinearity of the pollutants
including NO2, SO2, and O3,
prevented any causal separation.
Chronic cough and chronic phlegm
and breathlessness were related to
TSP,PM10andN02.
Estimated regression coefficient for PM10
versus FVC = -0.035 (95% CI -0.041,
-0.028). Corresponding value for FEV[
-0.016 (95% CI -0.023 to -0.01). Thus,
10 Mg/ni3 PM10 increase estimated to lead
to estimated 3.4 percent decrease in FVC
and 1.6 percent decrease in FEV[.
PM10
Chronic cough OR 11.4 (2.8, 45.5)
Bronchitis OR 23.2 (2.8, 45.5)
Wheeze OR 1.41 (0.55, 3.58)
Chronic cough, chronic phlegm and
breathlessness were related to PM10, and
TSP.
O
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
ij
oo
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Heinrich et al. (1999)
Bitterfeld, Zerbstand Hettstedt areas of former
East Germany,
During Sept. 1992 to July 1993 TSP ranged
from 44 to 65 /^ig/m3',
PM10 measured October 1993 - March 1994
ranged from 33 to 40; and BS ranged from
26 to 42 ,ug/m3. PM was measured with a
Harvard impactor.
Heinrich et al. (2000)
Three areas of former E. Germany
Pollution measures: SO2, TSP, and some
limited PM10 data. TSP decreased from 65,
48, and 44 ,ug/m3 to 43, 39, and 36 Mg/m3 in
the three areas. PM was measured with a
Harvard impactor.
Kramer etal. (1999)
Six East and West Germany communities
(Leipzig, Halle, Maddeburg, Altmark,
Duisburg, Borken)
Between 1991 and 1995 TSP levels in six
communities ranged from 46 to 102 ,ug/m3.
Each East Germany community had decrease
in TSP between 1991 and 1995. TSP was
measured using a low volume sampler.
Parents of 2470 school children ( 5-14 yr)
completed respiratory health questionnaire.
Children excluded from analysis if had lived
< 2 years in their current home, yielding an
analysis group of 2,335 children. Outcomes
studied: physician diagnosis for asthma,
bronchitis, symptom, bronchial reactivity, skin
prick test, specific IgE. Multiple logistic
regression analyses examined regional effects.
Cross-sectional study of children (5-14 yr).
Survey conducted twice, in 1992-1993 and
1995-1996; 2335 children surveyed in first
round, and 2536 in second round. Only 971
children appeared in both surveys. The
frequency of bronchitis, otitus media, frequent
colds, febrile infections studied. Because
changes measured over time in same areas,
covariate adjustments not necessary.
The study assessed relationship between TSP
and airway disease and allergies by parental
questionnaires in yearly surveys of children
(5-8 yr) between February and May. The
questions included pneumonia, bronchitis ever
diagnosed by physician, number of colds,
frequent cough, allergic symptoms.
In all, 19,090 children participated. Average
response was 87%. Analyses were conducted
on 14,144 children for whom information on
all covariates were available. Variables
included gender; parent education, heating
fuel, ETS. Logistic regression used to allow
for time trends and SO2 and TSP effects.
Regression coefficients were converted to odds
ratios.
Controlling for medical, socio-
demographic, and indoor factors,
children in more polluted area had
circa 50% increase for bronchitic
symptoms and physician-diagnosed
allergies compared to control area
and circa twice the respiratory
symptoms (wheeze, shortness of
breath and cough). Pulmonary
function tests suggested slightly
increased airway reactivity to cold
for children in polluted area.
PM and SO2 levels both decreased
in the same areas; so results are
confounded.
TSP and SO2 simultaneously
included in the model. Bronchitis
ever diagnosed showed a significant
association. A decrease in raw
percentage was seen between the
start of the study and the end for
bronchitis. Bronchitis seemed to be
associated only with TSP in spite of
huge differences in mean SO2
levels.
No single pollutant could be separated out
as being responsible for poor respiratory
health.
The prevalence of all respiratory
symptoms decreased significantly in all
three areas over time.
Bronchitis ever diagnosed
TSP per 50 /-ig/m3
OR 1.63 CI (1.37- 1.93)
Halle (East) %
TSP Mg/m3 Bronchitis
1991 102 60.5
1992 73 54.7
1993 62 49.6
1994 52 50.4
1995 46 51.9
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
Europe (cont'd)
oo
td
ij
VO
fe
H
6
o
o
H
O
O
H
W
O
O
Baldietal. (1999)
24 areas of seven French towns 1974-1976
Pollutants: TSP, BS, and SO2, NO4
3-year average TSP-mean annual values
ranging 45-243 ^g/m3. TSP was measured by
the gravimetric method.
Zeghnounetal. (1999)
La Havre, France during 1993 and 1996.
Daily mean BS levels measured in three
stations ranged 12-14 ^g/m3.
Leonardi et al. (2000)
17 cities of Central Europe
Yearly average concentration (Nov. 1995 -
Oct. 1996) across the 17 study areas varied
from 41 to 96 /j,g/m3 for PM10, from 29 to 67
Mg/m3 for PM25, and from 12 to 38 ,ug/rn3 for
PM ,„.,,.
Reanalysis of Pollution Atmospheric of
Affection Respiratory Chroniques (PAARC)
survey data to search for relationships between
mean annual air pollutant levels and
prevalence of asthma in 1291 adult (25-59 yrs)
and 195 children (5-9 yrs) asthmatics. Random
effects logistic regression model used and
included age, smoking, and education level in
the final model.
Respiratory drug sales for mucolytic and
anticough medications (most prescribed by a
physician) were evaluated versus BS, SO2, and
NO2 levels. An autoregressive Poisson
regression model permitting overdispersion
control was used in the analysis.
Cross-sectional study collected blood and
serum samples from 10-61 school children
aged 9 to 11 in each community 11 April to
May 1996. Blood and serum samples
examined for parameters in relation to PM.
Final analysis group of 366 examined for
peripheral lymphocyte type and total
immunoglobulin classes. Association between
PM and each log transformed biomarker
studied by linear regression in two-stage
model with adjustment for confounding factors
(age, gender, number of smokers in house,
laboratory, and recent respiratory illness). This
survey was conducted within the frame work of
the Central European study of Air Quality and
Respiratory Health (CEASAR) study.
Only an association between SO2
and asthma in adults observed. No
other pollutant was associated. Nor
was relationship with children seen.
Meteorological variables and O3
not evaluated.
Respiratory drug sales associated
with BS, NO2, and SO2 levels. Both
an early response (0 to 3 day lag)
and a longer one (lags of 6 and
9 days) were associated.
Number of lymphocytes (B, CD4+,
CD8d, and NK) increased with
10 increasing concentration of PM
adjusted for confounders. The
adjusted regression slopes are
largest and statistically significant
for PM2 5 as compared to PM10, but
small and non statistically signif.
for PM10_25. Positive relationship
found between concentration of IgG
in serum and PM2 5 but not for PM10
or PM10_25. Two other models
produced similar outcomes: a
multi-level linear regression model
and an ordinal logistic regression
model.
For a 50 /j.g/w? increase in
TSP
Adult asthma prevalence
OR 1.01 CI 0.92-1.11
SO2
Adult asthma prevalence
OR 1.26 CI 1.04-1.53
Adjusted
Regression slope
PM,.
CD4+
80% 95% CI (34; 143)
p< 0.001
Total IgG
24%
95% CI (2; 52)
p 0.034
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
oo
o
fe
H
6
o
o
H
O
O
H
W
O
O
Europe (cont'd)
Tumovska and Kostiranev (1999)
Dimitrovgrad, Bulgaria, May 1996
Total suspended particulate matter (TSPM)
mean levels were 520 ± I 61 Aig/m3 in 1986
and 187 ± 9 Mg/m3 in 1996. SO2, H2S, and
NO, also measured.
Jedrychowski et al. (1999)
In Krakow, Poland in 1995 and 1997
Spacial distributions for BS and SO2 derived
from network of 17 air monitoring stations.
BS 52.6 Mg/m ±53.98 in high area and
33.23 ±35.99 in low area.
Jedrychowski and Flak (1998)
In Kracow Poland, in 1991-1995
Daily 24 h concentration of SPM (black
smoke) measured at 17 air monitoring
stations.
High areas had 52.6 /-ig/m3 mean compared to
low areas at 33.2 ^g/m3.
Respiratory function of 97 schoolchildren
(mean age 10.4 ± 0.6 yr) measured in May
1996 as a sample of 12% of all four-graders in
Dimitrovgrad. The obtained results were
compared with reference values for Bulgarian
children aged 7 to 14 yr, calculated in the same
laboratory in 1986 and published
(Gherghinova et al., 1989; Kostianev et al.,
1994). Variation analysis technique were used
to treat the data.
Effects on lung function growth studied in
preadolescent children. Lung function growth
rate measured by gain in FVC and FEV[ and
occurrence of slow lung function growth
(SLFG) over the 2 yr period defined as lowest
quintile of the distribution of a given test in
gender group. 1129 children age 9 participated
in first year and 1001 in follow-up 2 years
later. ATS standard questionnaire and PFT
methods used. Initially univariate descriptive
statistics of pulmonary function indices and
SLFG were established, followed by
multivariate linear regression analyses
including gender, ETS, parental education,
home heating system and mold. SO2 also
analyzed.
Respiratory health survey of 1,129 school
children (aged 9 yr). Respiratory outcomes
included chronic cough, chronic phlegm,
wheezing, difficulty breathing and asthma.
Multi-variable logistic regression used to
calculate prevalence OR for symptoms
adjusted for potential confounding.
Vital capacity and FEV[ were
significantly lower (mean value. =
88.54% and 82.5% respectfully)
comparing values between 1986 and
1996. TSPM pollution had
decreased by 2.74 times to levels
still higher than Bulgarian and
WHO standards.
Statistically significant negative
association between air pollution
level and lung function growth
(FVC and FEVO over the follow up
in both gender groups. SLFG was
significantly higher in the more
polluted areas only among boys.
In girls there was consistency in the
direction of the effect, but not stat.
significant. Could not separate BS
and SO2 effects on lung function
growth. Excluding asthma subjects
subsample (size 917) provided
similar results.
The comparison of adjusted effect
estimates revealed chronic phlegm
as unique symptom related neither
to allergy nor to indoor variable but
was associated significantly with
outdoor air pollution category
(APL). No potential confounding
variable had major effect.
Boys
SLFG (FVC)
OR = 2.15 (CI 1.25 - 3.69)
SLFG (FEVO
OR=1.90(CI1.12- 3.25)
Girls
FVC OR =1.50 (CI 0.84-2.68)
FEV1 OR= 1.39 (CI 0.78 - 2.44)
It was not possible to assess separately the
contribution of the different sources of air
pollutants to the occurrence of respiratory
symptoms. ETS and household heating
(coal vs. gas vs. central heating) appeared
to be of minimal importance.
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
oo
fe
H
6
o
o
H
O
O
H
W
O
Latin America
Calderon-Garciduenas et al. (2000)
Southwest Metropolitan Mexico City
(SWMMC) winter of 1997 and summer of
1998.
Australia
Lewis etal. (1998)
Summary measures of PM10 and SO2
estimated for each of 10 areas in steel cities of
New South Wales. PM10 was measured using
a high volume sampler with size-selective
inlets.
Asia
Wong etal. (1999)
Hong Kong, 1989 to 1991
Sulfate concentrations in respirable particles
fell by 38% after implementing legislation
reducing fuel sulfur levels.
Study of 59 SWMMC children to evaluate
relationship between exposure to ambient
pollutants (O3 and PM10) and chest x-ray
abnormalities. Fishers exact test used to
determine significance in a 2x2 task between
hyperinflation and exposure to SWMMC
pollutant atmosphere and to control, low-
pollutant city atmosphere.
Cross-sectional survey of children's health and
home environment between Oct 1993 and Dec
1993 evaluated frequency of respiratory
symptoms (night cough, chest colds, wheeze,
and diagnosed asthma). Covariates included
parental education and smoking, unflued gas
heating, indoor cats, age, sex, and maternal
allergy. Logistic regression analysis used
allowing for clustering by GEE methods.
3405 nonsmoking, women (mean age 36.5 yr;
SD ± 3.0) in a polluted district and a less
polluted district were studied for six
respiratory symptoms via self-completed
questionnaires. Binary latent variable
modeling used.
Bilateral symmetric mild lung
hyperinflation was significantly
associated with exposure to the
SWMMC air pollution mixture
(p>0.0004). This raises concern for
development of chronic disease
outcome in developing lungs.
SO2 was not related to differences in
symptom rates, but adult indoor
smoking was.
Night cough OR 1.34 (1.18, 1.53)
Chest colds OR 1.43 (1.12, 1.82)
Wheeze OR 1.13 (0.93, 1.38)
Comparison was by district; no PM
measurements reported. Results
suggest control regulation may have
had some (but not statistically
significant) impact.
O
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
oo
to
fe
H
6
o
o
H
O
O
H
W
O
O
Asia (cont'd)
Wang etal. (1999)
Kaohsiung and Panting, Taiwan
October 1995 to June 1996
TSP measured at 11 stations, PM10 at 16
stations. PM10 annual mean ranged from 19.4
to 112.81 ,ug/m3 (median =91.00 Mg/m3)
TSP ranged from 112.81 to 237.82 Mg/m3
(median = 181.00). CO,NO2, SO2,
hydrocarbons and O3 also measured.
Guo etal. (1999)
Taiwan, October 1955 and May 1996
PM10 measured by beta-gauge.
Also monitoring for SO2, NO2, O3, CO.
PM10 ranged from 40 to 110 Mg/ni3 with a
mean of 69.
Wang etal. (1999)
Chongquing, China
April to July 1995
Dichot samplers used to measure PM25.
Mean PM2 5 level high in both urban
(143 ^ig/m3) and suburban (139 /L/g/m3) area.
SO2 also measured
Relationship between asthma and air pollution
examined in cross-sectional study among
165,173 high school students (11-16 yr).
Evaluated wheeze, cough and asthma
diagnosed by doctor. Video determined if
student displayed signs of asthma. Only
155,283 students met all requirements for
study analyses and, of these, 117,080 were
covered by air monitoring stations. Multiple
logistic regression analysis used to determine
independent effects of risk factors for asthma
after adjusting for age, gender, ETS, parents
education, area resident, and home incense use.
Study of asthma prevalence and air pollutants.
Survey for respiratory disease and symptoms in
middle-school students age < 13 to > 15 yr.
Total of 1,018,031 (89.3%) students and their
parents responded satisfactorily to the
questionnaire. Schools located with 2 km of
55 monitoring sites. Logistic regression
analysis conducted, controlling for age, hx
eczema, parents education.
Study examined relationship between PFT and
air pollution. Pulmonary function testing
performed on 1,075 adults (35 - 60 yr) who
had never smoked and did not use coal stoves
for cooking. Generalized additive model used
to estimate difference, between two areas for
FEVl5 FVC, and FEVj/FVCro with adjustment
for confounding factors (gender; age, height,
education, passive smoking, and occupational
exposures).
Asthma significantly related to high
levels of TSP, NO2, CO, O3 and
airborne dust. However PM10 and
SO2 not associated with asthma.
The lifetime prevalence of asthma
was 18.5% and the 1-year
prevalence was 12.5%.
Because of close correlation among
air pollutants, not possible to
separate effects of individual ones.
Factor analysis used to group into
two classes (traffic-related and
stationary fossil fuel-related). No
association found between lifetime
asthma prevalence and nontraffic
related air pollutants (SO2, PM10).
Mean SO2 concentration in the
urban and suburban area highly
statistically significant different
(213 and 103 /-ig/m3 respectfully).
PM2 5 difference was small, while
levels high in both areas. Estimated
effects on FEV1 statistically
different between the two areas.
Adjusted OR
PM10
1.00(0.96-1.05)
TSP
1.29(1.24-1.34)
Difference between urban and suburban
area excluding occupational exposures:
FEV,
B- 119.79
SE28.17
t - 4.25
p<0.01
FVC
B - 57.89
SE 30.80
t- 1.88
p<0.05
-------
to
o
o
to
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
Effect estimates as reported by study
authors. Negative coefficients for lung
function and ORs greater than 1 for other
endpoints suggest effects of PM
Reference citation, location, duration, type of
study, sample size, pollutants measured,
summary of values
Health outcomes measured, analysis design,
covariates included, analysis problems
Results and Comments
Effects of co-pollutants
oo
td
oo
fe
H
6
o
o
H
O
O
H
W
O
Asia (cont'd)
Zhang etal. (1999)
4 areas of 3 Chinese Cities (1985 - 1988)
TSP levels ranged from an annual arithmetic
mean 137 /-ig/m3 to 1250 Mg/m3 using
gravimetric methods.
Qian et al. (2000)
4 China cities
The 4 year average TSP means were 191, 296,
406, and 1067 Mg/m3. SO2 and NO2
measurements were also available.
TSP was measured gavimetrically.
A pilot study of 4 districts of 3 Chinese cities
in for the years 1985-1988, TSP levels and
respiratory health outcomes studied. 4,108
adults (< 49 yrs) examined by questionnaires
for couth, phlegm, wheeze, asthma, and
bronchitis. Categorical logistic—regression
model used to calculate odds ratio. SO2 and
NO2 were also examined. Other potential
confounding factors (age, education level,
indoor ventilation, and occupation) examined
in the multiple logistic regression model.
Pilot cross-sectional survey of 2789 elementary
school children in four Chinese communities
chosen for their PM gradient. Frequency of
respiratory symptoms (cough, phlegm, wheeze,
and diagnosed asthma, bronchitis, or
pneumonia) assessed by questionnaire.
Covariates included parental occupation,
education and smoking. The analysis used
logistic regression, controlling for age, sex,
parental smoking, use of coal in home, and
home ventilation.
Results suggested that the OR's for
cough, phlegm, persistent cough
and phlegm and wheeze increased
as outdoor TSP concentrations did. .
Wheeze produced largest OR for both
mothers and fathers in all locations.
Results not directly related to
pollution levels, but symptom rates
were highest in highest pollution
area for cough, phlegm,
hospitalization for respiratory
disease, bronchitis, and pneumonia.
No gradient correlating with
pollution levels found for the three
lower exposure communities.
o
-------
i 9. INTEGRATIVE SYNTHESIS
2
3
4 9.1 INTRODUCTION
5 This chapter focuses on integration of key information on exposure-dose-response risk
6 assessment components drawn from the preceding detailed chapters, to provide a coherent
7 framework for assessment of human health risks posed by ambient particulate matter (PM) in the
8 United States. As such, the chapter updates the integrated assessment of available scientific
9 information regarding ambient PM sources, exposures, and health risks as they pertain to the
10 United States that was provided in the 1996 Particulate Matter Air Quality Criteria Document
11 (1996 PM AQCD; U.S. Environmental Protection Agency, 1996a).
12 This chapter mainly uses the 10 Questions from the National Research Council (NRC)
13 Particulate Matter (PM) Research Agenda (NRC, 1998, 2001) as an organizing principle to
14 summarize and integrate key points derived from the material presented in detail in Chapters 1 to
15 8 of this document. After providing certain background information, the chapter is then basically
16 organized to follow the Risk Assessment Framework (as shown in Figure 9-1), and it addresses
17 the NRC questions noted earlier in Chapter 1 within the context of discussing general topic areas
18 that follow the flow of that framework from sources/emissions to effects. Some additional topics
19 in addition to the 10 NRC questions are also addressed.
20 Unlike the other criteria pollutants (O3, CO, NO2, SO2, and Pb), PM is not a specific
21 chemical entity but is a mixture of particles of different sizes, compositions, and properties.
22 Therefore, it is useful to present some background on the size, chemistry and physics of PM
23 before entering the Risk Assessment Framework. Thus, this chapter first provides background
24 information on key features of atmospheric particles, highlighting important distinctions between
25 fine- and coarse-mode particles with regard to size, chemical composition, sources, atmospheric
26 behavior, and potential human exposure relationships—distinctions that collectively continue to
27 suggest that fine- and coarse-mode particles should be treated as two distinct subclasses of air
28 pollutants. Recent trends in U.S. concentrations of different ambient PM size and composition
29 fractions (e.g., PM10, PM2 5, and PM10_2 5) and ranges of variability seen in U.S. regions and urban
30 airsheds are also summarized to place the ensuing human exposure and health effects discussions
April 2002 9-1 DRAFT-DO NOT QUOTE OR CITE
-------
Sources of Airborne
Particulate Matter
Or Gaseous
Precursor Emissions
Indicator in
Ambient (Outdoor) Air
(e.g. Mass
Concentration)
Mechanism determining emissions,
chemical transformation (including
formation of secondary particles from
gaseous precursors), and
transport in air
Human time-activity
patterns, Indoor (or
microenvironmental)
sources and sinks of
particuiate matter
Deposition,
clearance, retention
and disposition of
particuiate matter
presented to an
individual
Mechanisms of
damage and repair
Figure 9-1. A general framework for integrating particulate-matter research. Note that
this figure is not intended to represent a framework for research management.
Such a framework would include multiple pathways for the flow of
information.
Source: National Research Council (2001), as modified from NRC (1983, 1994), Lioy (1990), and Sexton et al.
(1992).
1 in perspective. After discussing human exposure aspects, the chapter next summarizes key points
2 regarding respiratory tract dosimetry, followed by a discussion of the extensive PM health
3 database that has expanded greatly during recent years. The latter includes numerous new
4 epidemiologic studies of populations throughout the world published since the 1996 PM AQCD
5 that provide further evidence that serious health effects (mortality, exacerbation of chronic
6 disease, increased hospital admissions, etc.) are associated with exposures to ambient levels of
7 PM found in contemporary U.S. urban air sheds. Evaluations of other possible explanations for
8 the reported PM epidemiology results (e.g., other co-pollutants, choice of models, etc.) also are
9 discussed, ultimately leading to the conclusion that the reported associations of PM exposure and
10 effects are valid.
11 New toxicologic evidence (derived from controlled exposure studies of humans and
12 laboratory animals) is also discussed, which elucidates likely mechanisms of action and other
13 information that greatly enhances the plausibility of the epidemiologic findings in comparison to
14 1996. Quantitative evidence is then discussed that (a) further substantiates associations of such
15 serious health effects with U.S. ambient PM10 levels, (b) also more strongly establishes fine
April 2002
9-2
DRAFT-DO NOT QUOTE OR CITE
-------
1 particles (as indexed by various indicators, e.g., PM25) as likely being important contributors to
2 the observed human health effects, and (c) now provides additional information on associations
3 between coarse-fraction (PM10_2 5) particles and adverse health impacts. The overall coherence of
4 the newer epidemiologic database also is discussed, which strengthens the 1996 PM AQCD
5 evaluation suggesting a likely causal role of ambient PM in contributing to the reported effects.
6 The nature of the observed effects and the biological mechanisms that might underlie such
7 effects then are discussed. The discussion of potential mechanisms of injury examines ways in
8 which PM could induce health effects. The increased, but still limited, availability of new
9 experimental evidence necessary to evaluate or directly substantiate the viability of hypothesized
10 mechanisms is noted. Information concerning possible contributions of particular classes of
11 specific ambient PM constituents also is summarized.
12 The chapter also provides information on the identification of susceptible population
13 groups at special risk for ambient PM effects and factors placing them at increased risk, which
14 need to be considered in generating risk estimates for the possible occurrence of PM-related
15 health events in the United States.
16
17
18 9.2 BACKGROUND
19 9.2.1 Basic Concepts
20 Atmospheric particles originate from a variety of sources and possess a range of
21 morphological, chemical, physical, and thermodynamic properties. Sources include combustion,
22 photochemical oxidation of precursors, and soil dust. Atmospheric particles contain inorganic
23 ions, metallic compounds, elemental carbon, organic compounds, and crustal compounds. Some
24 atmospheric particles are hygroscopic and contain particle-bound water. The organic fraction is
25 especially complex, containing hundreds of organic compounds. Individual particles may be
26 composed by any number of the above and other components.
27
28 9.2.2 Particle Size Distributions
29 As discussed in Chapter 2, the distribution of particles with respect to size is an important
30 physical parameter governing their behavior. Atmospheric particles vary in density and often are
April 2002 9-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 not spherical. Therefore, their diameters are often described by an "equivalent" diameter (i.e.,
2 that of a unit density sphere that would have the same physical behavior). The aerodynamic
3 diameter (Da) depends on the density of the particle and is defined as the diameter of a spherical
4 particle with a density of 1 g/cm3 but with a settling velocity equal to that of the particle in
5 question. The atmospheric deposition rates of particles, and therefore, their residence times in
6 the atmosphere, are a strong function of their aerodynamic diameters. The aerodynamic diameter
7 also influences deposition patterns of particles within the lung. The effects of atmospheric
8 particles on visibility, radiative balance, and climate, will also be influenced by the size
9 distribution of the particles. Atmospheric particles cover several orders of magnitude in particle
10 size. Therefore, size distributions often are expressed in terms of the logarithm of the particle
11 diameter on the X-axis and the measured differential concentration on the Y-axis. If the
12 differential concentration is plotted on a linear scale, the number of particles (per cm3 of air), or
13 the surface area, the volume, or the mass of particles (per m3 of air) having diameters in the size
14 range from log D to log(D + AD), will be proportional to the area under that part of the size
15 distribution curve.
16 Averaged atmospheric size distributions are shown in Figure 9-2. Figure 9-2a shows the
17 number distributions of particles, on a logarithmic scale, as a function of particle diameter for
18 several aerosols. The particle volume distributions for two of these are shown in Figure 9-2b.
19 These distributions show that most of the particles are quite small, below 0.1 //m; whereas most
20 of the particle volume (and therefore most of the mass) is found in particles larger than 0.1 //m.
21
22 9.2.3 Definitions of Particle Size Fractions
23 Aerosol scientists use four different approaches or conventions in the classification of
24 particles by size: (1) modes, based on the observed size distributions and formation mechanisms;
25 (2) cut point, usually based on the 50% cut point of the specific sampling device; (3) dosimetry
26 or occupational health sizes, based on the entrance into various compartments of the respiratory
27 system; and (4) legally specified, regulatory sizes for air quality standards.
28
29 Modal. The modal classification, first proposed by Whitby (1978), is shown in Figure 9-3.
30 In polluted atmospheres, the nuclei mode can be seen clearly in the volume distribution only in
31 traffic or near traffic or other sources of nuclei mode particles. The observed modal structure is
April 2002 9-4 DRAFT-DO NOT QUOTE OR CITE
-------
1,000,000 -
X
10,000 -
03
"55
E
CD
b
_0)
o
'•c
CD
Q_
•••,,-N
100 -
1 -
0.01 -
.2 0.0001 -
5
z
T3
0.000001 -
\
•
-------
O)
I
7
6
5
4 -
2
1
Mechanically
Generated
DGV = 0.018
o =1.6
1 ' I ""I
0.002 0.01
Nuclei Mode
0.1 1
Particle Diameter, Dp(|jm)
Accumulation Mode
Fine-Mode Particles
1 I ' '"I ' r
10
Coarse Mode
Coarse-Mode Particles
100
Figure 9-3. Volume size distribution, measured in traffic, showing fine-mode and
coarse-mode particles and the nuclei and accumulation modes within the
fine-particle mode. DGV (geometric mean diameter by volume, equivalent to
volume median diameter) and og (geometric standard deviation) are shown for
each mode. Also shown are transformation and growth mechanisms (e.g.,
nucleation, condensation, and coagulation).
Source: Adapted from Wilson and Suh (1997).
1 0.1 //m. Toxicologists and epidemiologists use the term "ultrafme" and aerosol physicists and
2 material scientists use the term "nanoparticles" to refer to particles in the nuclei-mode size range.
3 Accumulation Mode: That portion of the fine particle mode with diameters above about 0.1 //m.
4 The major processes that influence the formation and growth of particles in the three modes
5 are also shown in Figure 9-3. New particles may be formed by nucleation from gas phase
6 material. Particles may grow by condensation as gas phase material condenses on existing
7 particles. Particles also may grow by coagulation as two particles combine to form one. Gas
April 2002
9-6
DRAFT-DO NOT QUOTE OR CITE
-------
1 phase material condenses preferentially on smaller particles, and the rate constant for coagulation
2 of two particles decreases as the particle size increases. Therefore, nuclei mode particles grow
3 into the accumulation mode, but accumulation mode particles do not normally grow into the
4 coarse mode.
5 Over the years, the terms fine and coarse, as applied to particle sizes, have lost the precise
6 meaning given in Whitby's (1978) definition. In any given article, therefore, the meaning of fine
7 and coarse, unless defined, must be inferred from the author's usage. In particular, PM2 5 and
8 fine-mode particles are not equivalent. In this document, the term "mode" is used with fine and
9 coarse when it is desired to specify the distribution of fine-mode particles or coarse-mode
10 particles as shown in Figure 9-3.
11
12 Size-Selective and Occupational Health Size Fractions
13 Size-selective sampling refers to the collection of particles below or within a specified
14 aerodynamic size range, usually defined by the upper 50% cut point size, and has arisen in an
15 effort to measure particle size fractions with some special significance (e.g., health, visibility,
16 source apportionment, etc.). An example of a PM10 and a PM25 size cut are shown in Figure 9-4.
17 The subscripts, 10 and the 2.5, signify the 50% cut size, i.e., the size at which 50% of the
18 particles are collected and 50% of the particles are rejected. As can be seen, the cut is not
19 perfectly sharp. Some particles larger than the 50% cut point are collected; neither are all
20 particles smaller than the 50% cut point collected.
21 The occupational health community has defined size fractions for use in the protection of
22 human health. This convention classifies particles into inhalable, thoracic, and respirable
23 particles according to their upper size cuts (also shown in Figure 9-4). However, these size
24 fractions may also be characterized in terms of their entrance into various compartments of the
25 respiratory system. Thus, inhalable particles enter the respiratory tract, including the head
26 airways. Thoracic particles travel past the larynx and reach the lung airways and the
27 gas-exchange regions of the lung. Respirable particles are a subset of thoracic particles that are
28 more likely to reach the gas-exchange region of the lung.
29
30 Regulatory Size Cuts. In 1987, the NAAQS for PM were revised to use PM10, rather than
31 total suspended particulate matter (TSP), as the indicator for the NAAQS for PM (Federal
April 2002 9-7 DRAFT-DO NOT QUOTE OR CITE
-------
100
0
APM10
• IPM
• TPM
O RPM
VPM25
4 10 20 50
Aerodynamic Diameter (|jm)
100
Figure 9-4. Specified particle penetration (size-cut curves) through an ideal (no-particle-
loss) inlet for five different size-selective sampling criteria. Regulatory size
cuts are defined in the Code of Federal Regulations; PM25 (2001a), PM10
(2001b). PM2 5 is also defined in the Federal Register (1997). Size-cut curves
for inhalable particulate matter (IPM), thoracic particulate matter (TPM) and
respirable particulate matter (RPM) size cuts are computed from definitions
given by American Conference of Governmental and Industrial Hygienists
(1994).
1 Register, 1987). The use of PM10 as an indicator is an example of size-selective sampling based
2 on a regulatory size cut (Federal Register, 1987). The selection of PM10 as an indicator was
3 based on health considerations and was intended to focus regulatory concern on those particles
4 small enough to enter the thoracic region of the human respiratory tract. The PM2 5 standard set
5 in 1997 is also an example of size-selective sampling based on a regulatory size cut (Federal
6 Register, 1997). The PM25 standard was based primarily on epidemiological studies using
7 concentrations measured with PM2 5 samplers as an exposure index. However, the PM2 5 sampler
April 2002
9-8
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
was not designed to collect respirable particles. It was designed to collect fine-mode particles.
EPA is currently considering the possibility of a thoracic coarse particle standard with PM10_2 5 as
an indicator. Examples of regulatory size cuts are shown in Figure 9-5. Note also that, in the
range of particle aerodynamic diameter (Da) between 1.0 and 2.5 //m, there is overlap between
fine- and coarse-mode particles. The degree of overlap depends on prevailing conditions of
humidity and the amount of soil dust in the atmosphere.
E
:o
50 -
40 -
30 -
20 -
10 -
0.1
Coarse-Mode Particles
Fine-Mode Particles
0.2
i • • i • i i i ^ r
0.5 1.0 2 5 10 20
Aerodynamic Particle Diameter (|jm)
Total Suspended Particles (TSP)
PM
10
PM
2.5
PM
10.2.5
100
Figure 9-5. An idealized distribution of ambient particulate matter showing fine-mode
particles and coarse-mode particles and the fractions collected by size-selective
samplers. (WRAC is the Wide Range Aerosol Classifier which collects the
entire coarse mode [Lundgren and Burton, 1995].)
Source: Adapted from Wilson and Suh (1997).
April 2002
9-9
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.3 CHARACTERIZATION OF EMISSION SOURCES
2 What are the size distribution, chemical composition, and mass-emission rates of
3 paniculate matter emitted from the collection of primary-particle sources in the United States,
4 and what are the emissions of reactive gases that lead to secondary particle formation through
5 atmospheric chemical reactions?
6
1 The linkages between airborne PM and its sources are not as well defined as they are for
8 many other pollutants. In large part this is because PM is not a well defined chemical entity but
9 represents a complex mixture of primary and secondary components. PM is called "primary" if it
10 is in the same chemical form in which it was emitted into the atmosphere. PM is called
11 "secondary" if it is formed by chemical reactions in the atmosphere. Primary coarse particles are
12 usually formed by mechanical processes, such as the abrasion of surfaces or by the suspension of
13 soil or biological material. This includes material emitted in particulate form, such as wind-
14 blown dust, sea salt, road dust, and combustion-generated particles such as fly ash and soot.
15 PM10_25 is mainly primary in origin. Primary fine particles are emitted from sources either
16 directly as particles or as vapors that rapidly condense to form ultrafme or nuclei-mode particles.
17 Secondary PM is formed by chemical reactions of free, adsorbed, or dissolved gases. Most
18 secondary fine PM is formed from condensable vapors generated by chemical reactions of
19 gas-phase precursors. Secondary formation processes can result in either the formation of new
20 particles or the addition of condensable vapor to preexisting particles. Most of the sulfate and
21 nitrate and a portion of the organic compounds in atmospheric particles are formed by chemical
22 reactions in the atmosphere. Because precursor gases undergo mixing during transport from their
23 sources, it is difficult to identify individual sources of secondary constituents of PM.
24 Table 9-1 summarizes anthropogenic and natural sources for the major primary and
25 secondary aerosol constituents of fine and coarse particles. Anthropogenic sources can be further
26 divided into stationary and mobile sources. Stationary sources include fuel combustion for
27 electrical utilities, residential space heating and industrial processes; construction and
28 demolition; metals, minerals, and petrochemicals; wood products processing; mills and elevators
29 used in agriculture; erosion from tilled lands; waste disposal and recycling; and fugitive dust
30 from paved and unpaved roads. Mobile, or transportation-related, sources include direct
31 emissions of primary PM and secondary PM precursors from highway and off-highway vehicles
April 2002 9-10 DRAFT-DO NOT QUOTE OR CITE
-------
>
TABLE 9-1. CONSTITUENTS OF ATMOSPHERIC PARTICLES AND THEIR MAJOR SOURCES1
*— Sources
§ Primary (PM <2 . 5 ^m) Primary (PM >2 . 5 ^m) Secondary PM Precursors (PM <2 . 5 /j,m)
Aerosol
species Natural Anthropogenic Natural
SO4= Sea spray Fossil fuel combustion Sea spray
Sulfate
Anthropogenic Natural
— Oxidation of reduced sulfur
gases emitted by the oceans and
wetlands and SO2 and H2S
emitted by volcanism and forest
fires
Anthropogenic
Oxidation of SO2 emitted
from fossil fuel combustion
NO3-
Nitrate
Minerals
Oxidation of NO,, produced by
soils, forest fires, and lighting
Oxidation of NO,, emitted
from fossil fuel combustion
and in motor vehicle
exhaust
Erosion and Fugitive dust paved
re-entrainment and unpaved roads,
agriculture, and
forestry
Erosion and re-entrainment
Fugitive dust, paved
and unpaved road
dust, agriculture, and
forestry
^
o
3>
'•Tj
H
6
o
2
0
H
O
c
o
~1
w
o
hrl
7s
H
W
NH4+
Ammonium
Organic
carbon (OC)
Elemental
carbon
(EC)
Metals
Bioaerosols
—
Wild fires
Wild fires
Volcanic
activity
Viruses and
bacteria
'Dash (-) indicates either very
—
Prescribed burning,
wood burning, motor
vehicle exhaust, and
cooking
Motor vehicle exhaust
wood burning, and
cooking
— —
— Tire and asphalt wear
and paved road dust
, — Tire and asphalt wear
and paved road dust
Emissions of NH3 from wild Emissions of NH3 from
animals, and undisturbed soil animal husbandry, sewage,
and fertilized land
Oxidation of hydrocarbons Oxidation of hydrocarbons
emitted by vegetation (terpenes, emitted by motor vehicles,
waxes) and wild fires prescribed burning, and
wood burning
— —
Fossil fuel combustion, Erosion, re-entrainment, — — —
smelting, and brake
wear
minor source or no known
and organic debris
Plant and insect fragments, —
pollen, fungal spores, and
bacterial agglomerates
source of component.
-------
1 and nonroad sources. In addition to fossil fuel combustion, biomass in the form of wood is
2 burned for fuel. Vegetation is burned to clear new land for agriculture and for building
3 construction, to dispose of agricultural and domestic waste, to control the growth of animal or
4 plant pests, and to manage forest resources (prescribed burning). Also shown are sources for
5 precursor gases whose oxidation forms secondary particulate matter.
6 In general, the sources of fine PM are very different from those for coarse PM. Some of the
7 mass in the fine size fraction has been formed during combustion from material that volatilized
8 in combustion chambers and then recondensed before emission into the atmosphere. By and
9 large, however, most ambient PM2 5 is secondary, having been formed in the atmosphere from
10 photochemical reactions involving precursor gases. Transport and transformations of precursors
11 can occur over distances of hundreds of kilometers. The coarse PM constituents have shorter
12 lifetimes in the atmosphere, so their effects tend to be more localized. Only major sources for
13 each constituent within each broad category shown at the top of Table 9-1 are listed. Not all
14 sources are equal in magnitude. Chemical characterizations of primary particulate emissions for
15 a wide variety of natural and anthropogenic sources (as shown in Table 9-1) were given in
16 Chapter 5 of the 1996 PM AQCD. Summary tables of the composition of source emissions
17 presented in the 1996 PM AQCD and updates to that information are provided in Appendix 3D
18 of Chapter 3 in this document. The profiles of source composition are based largely on results of
19 various studies that collected signatures for use in source apportionment studies.
20 Natural sources of primary PM include windblown dust from undisturbed land, sea spray,
21 and plant and insect debris. The oxidation of a fraction of terpenes emitted by vegetation and
22 reduced sulfur species from anaerobic environments leads to secondary PM formation.
23 Ammonium (NH4+) ions, which play a major role in regulating the pH of particles, are derived
24 from emissions of ammonia (NH3) gas. Source categories for NH3 have been divided into
25 emissions from undisturbed soils (natural) and emissions that are related to human activities
26 (e.g., fertilized lands, domestic and farm animal waste). There is ongoing debate about
27 characterizing emissions from wild fires (i.e., unwanted fire) as either natural or anthropogenic.
28 Wildfires have been listed in Table 9-1 as natural in origin, but land management practices and
29 other human actions affect the occurrence and scope of wildfires. For example, fire suppression
30 practices allow the buildup of fire fuels and increase the susceptibility of forests to more severe
31 and infrequent fires from whatever cause, including lightning strikes. Similarly, prescribed
April 2002 9-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 burning is listed as anthropogenic, but can viewed as a substitute for wildfires that would
2 otherwise eventually occur on the same land.
3 The precursors to secondary PM have natural and anthropogenic sources, just as primary
4 PM has natural and anthropogenic sources. Whereas the major atmospheric chemical
5 transformations leading to the formation of particulate nitrate and sulfate have been relatively
6 well studied, those involving the formation of secondary aerosol organic carbon are still under
7 active investigation. A large number of organic precursors are involved, many of the kinetic
8 details still need to be determined, and many of the actual products of the oxidation of
9 hydrocarbons have yet to be identified.
10 However, over the past decade, a significant amount of research has been carried out to
11 improve the understanding of the atmospheric chemistry of secondary organic PM (SOPM)
12 formation. Although additional sources of SOPM might still be identified, there appears to be a
13 general consensus that biogenic compounds (monoterpenes, sesquiterpenes) and aromatic
14 compounds (toluene, ethylbenzene) are the most significant SOPM precursors. A large number
15 of compounds have been detected in biogenic and aromatic SOPM, although the chemical
16 composition of these two categories has not been fully established, especially for aromatic
17 SOPM. Transformations that occur during the aging of particles are still not adequately
18 understood. There are still large gaps in current understanding of a number of key processes
19 relating to the partitioning of semivolatile compounds between the gas phase and ambient
20 particles containing organic compounds, liquid water, inorganic salts, and acids. In addition,
21 there is a general lack of reliable analytical methods for measuring multifunctional oxygenated
22 compounds in the gas and aerosol phases.
23 Emissions estimates for primary PM2 5 components shown in Table 9-1 are provided in
24 Table 9-2 and emissions of precursors of secondary PM2 5 are shown in Table 9-3. The values
25 shown are annual averages for the entire United States. As can be seen from a comparison of the
26 entries in the two tables, the emissions of precursor gases of secondary PM are much larger than
27 those for primary PM. It should be noted here that the emissions estimates given above are
28 subject to a considerable degree of uncertainty, which varies from species to species. In addition,
29 there can be a great deal of temporal variability in the emissions. See NARSTO (2002) for
30 further details regarding the calculation of emissions inventories.
31
April 2002 9-13 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-2. EMISSIONS OF PRIMARY PM7< BY VARIOUS SOURCES IN 1999
Source
Emissions
(109 kg/y) Maj or PM Components
Notes
On-road vehicle 0.21
exhaust
Non-road vehicle 0.37
exhaust
Organic compounds,
elemental carbon
Organic compounds,
elemental carbon
Fossil fuel
combustion
0.36 Crustal elements, trace
metals
Industrial 0.35
processes
Biomass burning 1.2
Waste disposal 0.48
Fugitive dust 3.3
Windblown dust NA1
Other 0.02
Total 6.2
Metals, crustal material,
organic compounds
Organic compounds,
elemental carbon
Organic compounds,
trace metals
Crustal elements
Crustal elements
Organic compounds,
elemental carbon
Exhaust emissions from diesel (72%) and
gasoline vehicles (28%).
Exhaust emissions from off-road diesel (57%)
and gasoline vehicles (20%); ships and boats
(10%); aircraft (7%); railroads (6%).
Fuel burning in stationary sources such as
power plants (33%); industries (39%);
businesses and institutions (25%); residences
(3%).
Metals processing (29%); mineral products
(27%); chemical mfg. (11%); other industries
(33%).
Managed burning (47%); residential wood
burning (28%); agricultural burning (7%);
wildfires (18%).
Open burning (91%); incineration (9%).
Dust raised by vehicles on paved (19%) and
unpaved roads (40%); construction (15%),
dust from raising crops (24%) and livestock
(2%).
Dust raised by wind on bare land.
Structural fires
'NA = not available.
Source: Adapted from U. S. Environmental Protection Agency (2001).
1 Although most emphasis in this section has been placed on sources within the United
2 States, it also should be remembered that sources outside the United States contribute to ambient
3 PM levels that can, at times, exceed the ambient NAAQS. Dense hazes, composed mainly of
4 dust, occur frequently during the summer in southern Florida. This dust has been emitted in the
5 Sahara Desert and then transported across the Atlantic Ocean. Large-scale dust storms in the
6 deserts of central Asia recently have been found to contribute to PM levels in the Northwest on
7 an episodic basis. Not only dust but microbial pathogens and various pollutants are transported
April 2002
9-14
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-3. EMISSIONS OF PRECURSORS TO SECONDARY PM2 5 FORMATION
BY VARIOUS SOURCES IN 1999
Precursor
Emissions
(109kg/y)
Secondary PM
Component
Notes
S02
NO''2
Biogenic
VOCs1
NH,
17
Anthropogenic 16
VOCs
44
45
Sulfate
26 Nitrate
Various mainly
unidentified
compounds of 'OC'
Various mainly
unidentified
compounds of 'OC'
Ammonium
Exhaust from on-road (2%) and non-road (5%) engines
and vehicles; fossil fuel combustion by electrical utilities,
industries, other sources (85%); various industrial
processes (7%); and other minor sources (1%).
Exhaust from on-road (34%) and non-road (22%)
engines and vehicles; fossil fuel combustion by electrical
utilities, industries, other sources (39%); lightning (4%);
soils (4%); and other minor sources (5%).
Evaporative and exhaust emissions from on-road (29%)
and non-road (18%) vehicles; evaporation of solvents
and surface coatings (27%); biomass burning (9%);
storage and transport of petroleum and volatile
compounds (7%); chemical and petroleum industrial
processes (5%); other sources i
Approximately 98% emitted by vegetation. Isoprene
(35%); monoterpenes (25%); all other reactive and
non-reactive compounds (40%).
Exhaust from on-road and non-road engines and vehicles
(5%); chemical manufacturing (3%); waste disposal,
recycling, and other minor sources (5%); livestock
(82%); and fertilizer application (18%).
'Includes estimates of natural sources from Guenther et al. (2000).
2Emissions expressed in terms of NO2.
Source: Adapted from U. S. Environmental Protection Agency (2001).
1
2
3
4
5
6
during these events. Uncontrolled biomass burning in central America and Mexico may have
contributed to elevated PM levels that exceeded the daily NAAQS level for PM in Texas; and
wildfires throughout the United States, Canada, Mexico, and Central America all contribute to
PM background concentrations in the United States.
April 2002
9-15
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.4 AMBIENT CONCENTRATIONS
2 What are the basic characteristics of ambient monitoring data used to draw inferences
3 about the relations between health outcomes and air pollution?
4
5 9.4.1 Measurement of Particulate Matter
6 It is possible to measure a variety of PM indicators with high precision. However, the
7 absolute accuracy of a PM monitoring techniques cannot be established because no standard
8 reference calibration material or procedure has been developed for suspended, atmospheric PM.
9 Therefore, accuracy is defined as the degree of agreement between a field PM sampler and a
10 collocated PM reference method audit sampler. Intercomparison studies, therefore, are very
11 important for establishing the reliability of PM measurements.
12 One important measurement problem arises from the presence of semivolatile components
13 (i.e., species that exist in the atmosphere in dynamic equilibrium between the condensed phase
14 and gas phase) in atmospheric PM. Important examples include ammonium nitrate, semivolatile
15 organic compounds, and particle-bound water. Most filter-weighing techniques for PM,
16 including the U.S. Federal Reference Methods (FRM), require equilibration of collected material
17 at fixed, near-room temperature (25 °C) and moderate relative humidity (40%) to reduce particle-
18 bound water. This also causes the loss of an unknown, but possibly significant fraction, of
19 ammonium nitrate and semivolatile organic compounds. Some modest amount of particle-bound
20 water may be present at the 40 % relative humidity at which filter samples are equilibrated.
21 However, to avoid measurement of large amounts of particle-bound water that would be present
22 at higher relative humidities, continuous measurement techniques must reduce particle-bound
23 water in situ. One technique is to stabilize PM at a specified temperature high enough to remove
24 all, or almost all, particle-bound water. This results in loss of much of the semivolatile PM.
25 Examples include the tapered element oscillating microbalance (TEOM) operated at 50 °C and
26 beta gauge monitors with heated inlets. Another technique is the use of a diffusion denuder to
27 remove water vapor without heating. Examples include the Brigham Young absorptive sampler
28 and Harvard pressure drop monitor. The three approaches give different mass concentrations,
29 especially in air sheds with high nitrate, wood smoke, or secondary organic aerosols. Current
30 PM standards are based on health effects studies mainly using filter techniques. However, the
April 2002 9-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 need to provide new real time information to the public and the economic pressure to replace
2 filter samplers with continuous monitors will require a better understanding of the physics and
3 chemistry of the semivolatile components of PM and studies of the potential health effects of
4 these components.
5
6 9.4.2 Mass Concentrations
7 Data for ambient PM2 5 and PM10 concentrations are obtained routinely by networks
8 operated by various state and local agencies. Data are also collected as part of research efforts by
9 governmental, academic and industrial groups. Data from state and local agencies are stored in
10 the AIRS (Aerometric Information Retrieval System) data base, maintained by the U.S.
11 Environmental Protection Agency. Concentrations of PM10_2 5 based on FRM PM10 and PM25
12 monitors are estimated by taking the difference between these two measurements. The spatial
13 coverage and frequency of sampling depends on the resources of the agency carrying out the
14 monitoring. Thus, the amount of data collected in a given urban area varies across the United
15 States.
16 The median PM2 5 concentration was 13 //g/m3 in the United States on a county basis, for
17 1999 and 2000. The corresponding median PM10_25 concentration was about 10 //g/m3 for the
18 same period. However, there was a good deal of variability in the annual means in different
19 environments in the United States. The mean PM2 5 concentration was below 7 //g/m3 in 5% and
20 below 18 //g/m3 in 95% of counties that met minimum AIRS data completeness criteria for
21 calculation of an annual mean concentration (at least 11 days data for each calendar quarter).
22 The mean PM10_2 5 concentration was below 4 //g/m3 in 5% and below 21 //g/m3 in 95% of
23 counties meeting the criteria given above. Mean PM2 5 and PM10_2 5 concentrations reported by
24 the IMPROVE network were considerably lower than the lowest 5th percentile values reported by
25 state and local agencies.
26 An adequate characterization of the PM concentrations found in urban areas cannot be
27 obtained by considering only annual average concentrations for the whole urban area. There can
28 be considerable spatial and temporal variability in the concentration fields. Typically, annual
29 mean concentrations are within 5 //g/m3 of each other in urban areas (MSAs). The spread in
30 values can be much greater if CMSAs are considered. Even within some MSAs, concentrations
31 measured at separate sites on individual days can differ by over 100 //g/m3.
April 2002 9-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 Pairs of sites within MSAs are correlated with each other to varying degrees, depending on
2 the urban area. There are some very general regional patterns evident in the data base in which
3 sites tend to be more highly correlated with each other in the eastern United States and less well
4 correlated with each other in the western United States. Figure 9-6 shows an example for
5 Philadelphia, PA. The exceptions are frequent enough to prevent extrapolation from one city to
6 another without first examining the data. Although sites may be highly correlated with each
7 other within an MSA, this does not mean that the concentration fields are uniform, as illustrated
8 by Figure 9-7 for three urban areas. Concentrations for the three site pairs chosen are all well
9 correlated with each other (r>0.9), but the concentrations display different degrees of uniformity.
10
11 9.4.3 Physical and Chemical Properties of Ambient PM
12 Physical and chemical properties of fine-mode and coarse-mode particles that are produced
13 by sources listed in Table 9-1 are summarized in Table 9-4. It can readily be seen that fine- and
14 coarse-mode particles show striking differences in the nature of their sources, their composition,
15 and hence, their chemical properties, and in their removal processes. Differences in sources and
16 removal processes for fine- and coarse-mode particles account for many differences in their
17 behavior in the atmosphere. The much shorter atmospheric lifetimes of coarse particles
18 compared to fine particles implies that fine particles can travel much further in the atmosphere
19 than coarse particles. The more sporadic nature of the sources of coarse particles, in addition,
20 implies that coarse PM should be more highly spatially variable than fine PM. Elemental
21 compositions, including trace elements by X-ray flourescence analysis, for PM25 and PM10_2 5 in
22 two cities with different fine/coarse relationships are given in Table 9-5. The major chemical
23 components of PM25 from several sites in the eastern, interior, and western parts of the United
24 States are shown in Figure 9-8.
25
26
27
April 2002 9-18 DRAFT-DO NOT QUOTE OR CITE
-------
Phildelphia, PA MSA
AIRS Site ID
Site A
SiteB
SiteC
SiteD
SiteE
34-007-0003
34-007-1007
42-045-0002
42-101-0004
42-101-0136
b.
100 km
3400700031 3400710071 4204500021 4210100041 4210101361
Mean
Obs
SD
50-
40-
14.941
108
8.524
15.427
103
8.749
15.992
112
8.265
14.823
284
8.537
14.718
278
8.295
1234
c. Site A
B
D
A
B
C
D
E
0.964 0.868
1 (3.3,0.082) (6.3,0.155)
95 98
0.849
1 (6.9,0.158)
94
1
0.88
(4.8, 0.129)
81
0.894
(3.7, 0.135)
77
0.868
(5.0,0.149)
85
1
0.868
(5.4, 0.147)
80
0.857
(6.4, 0.148)
79
0.818
(6.6, 0.154)
83
0.918
(4.9,0.13)
246
1
Figure 9-6. Philadelphia, PA-NJ MSA. (a) Locations of sampling sites by AIRS ID#;
(b) Quarterly distribution of 24-h average PM2 5 concentrations; (c) Intersite
correlation statistics, for each data pair, the correlation coefficient, (P90,
coefficient of divergence) and number of measurements are given.
April 2002
9-19
DRAFT-DO NOT QUOTE OR CITE
-------
Columbia SC1999 & 2000
1
CD
O
O
0
"o
L_
CD
|
z
CO
8
CD
ecu mi
0
"5
L_
CD
.Q
E
z
120 -
100 -
80 -
60 -
40 -
20 -
Q
45-079-0007 vs 45-079-001
r = 0.97
COD = 0.06
Psf> = 2.7figlm>
h
(V nf N^ CK' 9} ^ ^ *> N *> I/
N? N$> *$> <$>
Chicago IL2000
30 -,
25 -
20 -
15 -
10 -
5 -
n -.
17-031-2001 vs 17-031-4201
r = 0.94
COD = 0.1 4
•
•
1 .
II. II -
Detroit Ml 2000
8 12-
c
£ 10-
3 R
o 8 -
0
•5 6-
te 4 -
1 2 -
z n
|
26-099-0009 vs 26-163-0033
r = 0.93
COD = 0.22
1
1
P90= 12.7/(g/m3
_
1
1 1
nil. ,i . i
Concentration Difference (//g/m3)
Figure 9-7. Occurrence of differences between pairs of sites in three MSAs. The absolute
differences in daily average PM2 5 concentrations between sites are shown on
the x-axis and the number of occurrences on the y-axis. The MSA, years of
observations, AIRS site I.D. numbers for the site pairs, Pearson correlation
coefficients (r), coefficients of divergence (COD), and 90th percentile (P90)
difference in concentration between concurrent measurements are also shown.
Source: Pinto et at. (2002).
April 2002
9-20
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-4. COMPARISON OF AMBIENT PARTICLES,
FINE MODE (Nuclei Mode Plus Accumulation Mode) AND COARSE MODE
Fine
Coarse
Nuclei
Accumulation
Formation
Processes:
Formed by:
Composition:
Solubility:
Atmospheric
half-life:
Removal
Processes:
Travel
distance:
Combustion, high-temperature
processes, and atmospheric reactions
Nucleation
Condensation
Coagulation
Sulfates
Elemental Carbon
Metal compounds
Organic compounds
with very low
saturation vapor
pressure at ambient
temperature
Probably less soluble
than accumulation
mode
Minutes to hours
Grows into
accumulation mode
-------
TABLE 9-5. CONCENTRATIONS OF PM2 5, PM10 2 5 AND SELECTED ELEMENTS
IN THE PM?, AND PM,n,, SIZE RANGE
10-2.5
Phoenix, AZ (n = 164)
Species
Mass
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Co
Ni
Cu
Zn
As
Se
Br
Pb
Concentration
PM25
11,200
125
330
11
487
19
110
129
11
0.7
0.6
5.7
177
-0.4
0.6
5.2
17
1.9
0.4
3.8
6.6
(ng/m3)
PM10.2.5
27,600
1879
535
37
131
208
561
1,407
130
2.0
2.6
29
1,211
1.2
1.8
10.3
25
0.6
-0.02
0.8
4.6
Philadelphia, PA (n =
Species
Mass
Al
Si
P
S
Cl
K
Ca
Ti
V
Cr
Mn
Fe
Co
Ni
Cu
Zn
As
Se
Br
Pb
Concentration
PM25
29,800
109
191
15
3,190
23
68
63
8.7
9.7
1.4
3.2
134
0.8
8.5
7.7
56
0.4
1.3
14
28
20)
(ng/m3)
PM10.2.5
8,400
325
933
28
38
47
100
421
30
3.2
1.0
6.3
352
-0.2
2.0
14
52
0
-0.1
3.0
13
Source: Zweidinger et al. (1998); Pinto et al. (1995).
April 2002
9-22
DRAFT-DO NOT QUOTE OR CITE
-------
Figure 9-8. Major chemical components of PM2 5 as determined in the pilot study for
EPA's national speciation network.
1
2
3
4
5
6
7
8
9
10
11
12
9.5 AIR QUALITY MODEL DEVELOPMENT AND TESTING
What are the linkages between emissions sources and the biologically important
components of paniculate matter?
Atmospheric models that address this question fit into two general categories. Either they
are process oriented and attempt to predict variables of interest based on the solution of equations
describing basic physical and chemical processes or they are statistically oriented and rely on the
statistical analysis of atmospheric data to infer information about the nature and relative
importance of different sources. Although there are many sub-categories within each of these
two broad categories, the two main types of models that are under active development and
application are chemistry-transport models (CTMs) and receptor models.
April 2002
9-23
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
9
10
11
12
13
14
15
The main components of a CTM are summarized in Figure 9-9. Models such as the
CMAQ (Community Model for Air Quality) system and MAQSIP (Multiscale Air Quality
Simulation Platform) incorporate the processes shown in Figure 9-9 as numerical algorithms to
predict time dependent concentration fields of a wide variety of gaseous and paniculate phase
pollutants. Also shown in Figure 9-9 is the meteorological model used to provide the inputs for
calculating the transport of species in the CTM. The meteorological models such as the MM5
model, which supples these inputs to the CTMs mentioned above, also provide daily weather
forecasts. The domains of these models extend typically over several thousand kilometers by
several thousand kilometers. Because these models are computationally intensive, it is often
impractical to run them over larger domains without sacrificing some features. For these
reasons, both the meteorological model and the CTM must have boundary conditions that allow
the effects of processes occurring outside the model domain to be felt. The entire system of
meteorological model emissions processors and output processors constituents the framework of
EPA'sModels-3.
Meteorological
Model
Emissions
Model
Anthropogenic
(point, area sources)
&
V Biogenic Emissions
Chemistry Transport Model
Visualization of Output
Process Analyses
Figure 9-9. Main components of a comprehensive atmospheric chemistry modeling
system, such as Models 3.
April 2002
9-24
DRAFT-DO NOT QUOTE OR CITE
-------
1 The performance of models such as these must be evaluated by comparison with field data
2 as part of a cycle of model improvements and subsequent evaluations. Discrepancies between
3 model predictions and observations can be used to point out gaps in current understanding of
4 atmospheric chemistry. Very often, however, the algorithms in the model are 'tuned' to improve
5 agreement between the model predictions and a particular set of observations. Model evaluation
6 does not merely involve a straightforward comparison between model predictions and the
7 concentration field of the pollutant of interest. Even this task is not straightforward in the case of
8 PM, because PM is composed of a number of different substances with different chemical and
9 physical properties. A comparison of model predicted PM25 mass with measured PM25 mass
10 may not be very meaningful because there can be compensating errors in the model calculations
11 and there are significant artifacts affecting the collection and retention of a number of PM
12 components such as semi-volatile organic compounds and ammonium nitrate. Because of the
13 number and complexity of the parameterizations used in CTMs, there may be compensating
14 errors and tests of these parameterizations must be made for individual physical and chemical
15 processes.
16 Another issue relates to the averaging time that is used for both the observations and the
17 model outputs. Model predictions can be made with time steps shorter than an hour, however, as
18 noted in Chapters 2 and 3, there is a considerable degree of uncertainty associated with individual
19 hourly observations by continuous monitors. Emissions inventories, as shown in Tables 9-2 and
20 9-3, represent annual averages; and it is impractical, except in a few cases, to increase that
21 resolution down to even a few days. At least for modeling ozone, it has been found that
22 agreement between model and observations is improved if seasonal averages, rather than
23 episodic averages are considered.
24 Models such as the CTMs discussed above have been under development for a number of
25 years. Discussions of these models have not been included in the earlier chapters because these
26 models are not yet being used to provide information about human exposures that could be
27 incorporated into this document. CTMs are being used to develop emissions control strategies
28 and to aid in implementation of existing air quality standards. The reader is referred to NARSTO
29 documents (NARSTO, 2002) for further details.
30 There are two main approaches to receptor modeling. Receptor models such as the
31 chemical mass balance (CMB) model relate source category contributions to ambient
April 2002 9-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentrations based on analyses of the composition of ambient PM and source emissions
2 samples. This technique has been developed for apportioning source categories of primary PM
3 and was not formulated to include the processes of secondary PM formation. In the second
4 approach, various forms of factor analysis are used. They rely on the analysis of time series of
5 compositional data from ambient samples to derive both the composition of sources and the
6 source contributions. Standard approaches such as factor analysis or Principal Component
7 Analysis (PCA) can apportion only the variance and not the mass in an aerosol composition data
8 set. Positive matrix factorization (PMF) is a recently developed multivariate technique that
9 overcomes many of the limitations of standard techniques, such as principal components analysis
10 (PCA), by allowing for the treatment of missing data and data near or below detection limits.
11 This is accomplished by weighting elements inversely according to their uncertainties. Standard
12 methods such as PCA weight elements equally regardless of their uncertainty. Solutions also are
13 constrained to yield nonnegative factors. Both the CMB and the PMF approaches find a solution
14 based on least squares fitting and minimize an object function. Both methods provide error
15 estimates for the solutions based on estimates of the errors in the input parameters. It should be
16 remembered that the error estimates often contain subjective judgments. For a complete
17 apportionment of mass, all of the major sources affecting a monitoring site must be sampled for
18 analysis by CMB, whereas there is no such restriction in the use of PMF.
19 Among other approaches, the UNMIX model takes a geometric approach that exploits the
20 covariance of the ambient data to determine the number of sources, the composition and
21 contributions of the sources, and the uncertainties (Henry, 1997). A simple example may help
22 illustrate the approach taken by UNMIX. For example, in a two-element scatter plot of ambient
23 Al and Si, a straight line and a high correlation for Al versus Si can indicate a single source for
24 both species (soil), while the slope of the line gives information on the composition of the soil
25 source. In the same data set, iron may not plot on a straight line against Si, indicating other
26 sources of Fe in addition to soil. More importantly, the Fe-Si scatter plot may reveal a lower
27 edge. The points defining this edge represent ambient samples collected on days when the only
28 significant source of Fe was soil. Success of the UNMIX model hinges on the ability to find
29 these "edges" in the ambient data from which the number of sources and the source compositions
30 are extracted. UNMIX uses principal component analysis to find edges in m-dimensional space,
31 where m is the number of ambient species. UNMIX does not make explicit use of errors or
April 2002 9-26 DRAFT-DO NOT QUOTE OR CITE
-------
1 uncertainties in the ambient concentrations, unlike the methods outlined above. This is not to
2 imply that the UNMIX approach regards data uncertainty as unimportant, but rather that the
3 UNMIX model results implicitly incorporate error in the ambient data. The underlying
4 philosophy here is that the uncertainties are often unquantifiable, and hence it is best to make no
5 a priori assumptions about what they are.
6 For most practical purposes, the relative contributions of sources to ambient PM samples
7 are determined by receptor models. Receptor models have most successfully been applied to the
8 determination of sources of primary PM. The process based models are more flexible and could
9 be used for determining sources of secondary PM. However, they are computationally much
10 more intensive, and they rely on a large number of inputs, with varying degrees of uncertainty.
11 Arguably, emissions inventories represent the major source of uncertainty in the application of
12 CTMs (see e.g., Calvert et al., 1993). However, significant uncertainty also exists in
13 photochemical transformations, in part because of a lack of data for many key reactions. Further
14 uncertainty is added in the methods that are used to reduce the literally thousands of reactions
15 involving hundreds of species occurring in the atmosphere to more tractable numbers through the
16 use of idealized chemical mechanisms. Issues concerning the gas phase mechanisms are relevant
17 because the free radicals that are involved in the formation of photochemical oxidants are also
18 involved in the formation of secondary PM.
19
20
21 9.6 EXPOSURE TO PARTICULATE MATTER AND COPOLLUTANTS
22 What are the quantitative relationships between concentrations of particulate matter and
23 gaseous copollutants measured at stationary outdoor air-monitoring sites and the contributions
24 of these concentrations to actual personal exposures, especially for subpopulations and
25 individuals?
26
27 It will be useful to separate these relationships into two components: (a) central site to
28 outdoor concentrations; and (b) outdoor concentrations to personal exposures.
29
30
31
April 2002 9-27 DRAFT-DO NOT QUOTE OR CITE
-------
l 9.6.1 Central Site to Outdoor
2 The first component to be examined is the relationship between ambient PM concentrations
3 measured by a central monitor, located at a site presumably representative of the community (or
4 the average of several such sites), and the ambient PM concentration just outside an indoor
5 microenvironment such as a home.
6
7 9.6.1.1 Exposure for Acute Epidemiology
8 In acute time-series studies, daily deaths (or other health effects) are regressed against the
9 daily ambient PM concentrations as measured at a single site (or the average of several sites) in a
10 city. Spatial variations in daily exposure can lead to errors in the estimated relative risk. Under
11 the assumption of a linear relationship between exposure and effect, analysis of exposure error
12 suggests that a key indicator of the effect on epidemiologic results of spatial variations in
13 exposure will be the strength of the daily site-to-site correlations of ambient PM concentrations.
14 Chapter 3 presents a substantial body of new monitoring data from AIRS. A range of
15 correlations of PM25 concentrations were found between monitoring sites in the cities chosen for
16 analysis. PM10 and TSP sites were frequently chosen to monitor specific local point or area
17 sources. However, PM25 sites are chosen primarily to be representative of community
18 exposures. Still it would be wise to check the representativeness of a site before choosing a site
19 or group of sites to provide a representative community concentration for exposure or
20 epidemiologic studies. As shown in Figure 9-10, site-to-site correlations tend to be higher for a
21 site pair where both of which are dominated by regional PM than for a site pair where one of
22 which is more strongly influenced by local sources.
23
24 9.6.1.2 Exposure for Chronic Epidemiology
25 In chronic studies, total or annual deaths in large cohorts in different cities are regressed
26 against long-term or annual average concentrations in the different cities. Few analyses of
27 exposure error have been performed for this case. However, the key consideration for chronic
28 studies might be differences in the annual (or seasonal) averages in different parts of a city.
29 Prior to NRC-1, there was little information on the variations of long term PM concentration
30 averages across cities. Some information on the spatial variations in long-term (seasonal)
31 averages are reported in Chapter 3 of this document, based on data from AIRS.
April 2002 9-28 DRAFT-DO NOT QUOTE OR CITE
-------
CD
'O
it
0
O
O
c
o
•j= 0.80 •
1C
•0
b
o
Chicago, IL
X XX x
XXXv X Y X
X ** / #
X
40 60
Distance (km)
Washington, DC-MD-VA-WV
1.00-
^,
•g 0 90 -
CD
0
Sfc
CD
0
O
o
13
CD
O
O
0.70-
Xv v K
X «vX$X X v
\ xxxX
x Xf X X X
x * x x xx
% x x
X X
x x x xx
v X
-X
X
X
x x
X
X
X
0 20 40 60
100 120
Distance (km)
Pittsburgh, PA
Los Angeles - Long Beach, CA
1.0-
0.9-
.^
CD °°~
'o
£
0
0
O 0.7-
c
.0
15
CD 06-
O
O
05-
0.4-
« J x.x xy xXj( x „ ^x
X * y Y y X
A A "AW
xx xx x
X xxXX X
X A
A
" X
x x
X
X j,
x x x
X
X
x x
1 1 1 1
1.0-
0.9-
• — '
'§ 0.7-
0
o
O 0.6-
c
,0
IS 0.5-
0
b
0 Q.4-
0.3-
0.2-
X y X
XXX v
X~ v
" X
XX
x
x x x
x x
x X x x x x
x x x
X
X
X
I I 1 I
Distance (km)
Distance (km)
Figure 9-10. Correlograms showing the variation in site-to-site correlation coefficient for
PM2 5 as a function of distance between sites for several cities.
Source: Fitz-Simons et al. (2000).
April 2002
9-29
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.6.2 Home Outdoor Concentrations Versus Ambient Concentrations
2 Indoors and the Ambient Contribution to Total Personal Exposure
3 What is the relationship between the concentration of ambient PM outside a home and the
4 concentration of ambient PM that has infiltrated into the home?
5
6 9.6.2.1 Mass Balance Model
7 It will be useful to review some concepts derived from the equilibrium mass balance
8 model, discussed in detail in Chapter 5. The ratio of the ambient PM concentration outdoors, C,
9 to the concentration of ambient PM that has infiltrated indoors, C(AI), is given by the infiltration
10 factor where P is the particle penetration efficiency, a is the air exchange rate, and k is the
11 deposition rate.
12
13 C(AI)/C = Pa/(a+k) = Fmp (the infiltration factor) (9-1)
14
15 As will be discussed later, P and k are functions of the particle size, so FINP will also depend on
16 particle size. The mass balance equation may be modified to include particle removal by air
17 handling systems and to account for nonequilibrium behavior.
18 While indoors, a person will be exposed to a concentration of ambient pollution given by
19 C • FINP. However, while outdoors a person will be exposed to the full ambient concentration.
20 The infiltration factor and the fraction of time outdoors may be used with the ambient
21 concentration to estimate the ratio of the ambient PM exposure (while indoors and outdoors) to
22 the ambient PM concentration, where y = the fraction of time spent outdoors,
23
24 A/C =y + (l-y)FINP =y + (l-y)Pa/(a+k) = a (the attenuation factor). (9-2)
25
26 Since y and a may vary from day to day and person to person and P and k will vary with particle
27 size, a will also be a variable.
28 It is necessary to understand the infiltration factor, used to estimate the concentration of
29 ambient PM concentration indoors [C(AI) = C • FINP\, and the attenuation factor, used to estimate
30 the ambient exposure, i.e., personal exposure to particles of ambient original, [A = C • a],
April 2002 9-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 because they may be estimated from exposure measurements and used to estimate A, the ambient
2 component of total personal exposure.
3
4 9.6.2.2 Separation of Total Personal Exposure into its Ambient and Nonambient
5 Components
6 A person's total exposure to PM or other pollutants includes a nonambient component,
7 usually divided into a component due to indoor-generated pollutants that are evenly distributed
8 through out the house and a component, sometimes called the personal cloud, due to activities of
9 the person that generate pollutants which influence that person more than other persons in the
10 same house. Thus, total personal exposure, T, equals the sum of ambient exposure, A, and
11 nonambient exposure, N:
12
13 T = A+N (9-3)
14
15 As NRC Topic 1 makes clear, a key variable of interest is A, the ambient exposure, i.e., the
16 contributions of particulate matter and gaseous copollutants measured at stationary outdoor
17 air-monitoring sites to actual personal exposures, not T, the total personal exposures due to
18 ambient and indoor-generated pollutants. However, it is not possible to measure A or N directly.
19 Only T and C can be measured directly. It is necessary to understand the infiltration factor, used
20 to estimate the concentration of ambient PM concentration indoors, \C(AI) = C • FINF], and the
21 attenuation factor, used to estimate the ambient exposure, [A = C • a], because these factors may
22 be estimated from exposure measurements and used to estimate A, the ambient component of
23 total personal exposure.
24 In recent years, the need to separate personal exposure into ambient and nonambient
25 components has been recognized (Wilson and Suh, 1997), techniques for separating total
26 personal exposure into its ambient and nonambient components have been recommended
27 (Wilson et al., 2000), several papers have reported average values of a and N, and one paper has
28 reported individual values of A.
29
30 Average Values
31 As shown in Figure 9-11, regression of individual measurements of personal exposure on
32 the corresponding measurements of ambient concentrations yields two components of total
April 2002 9-31 DRAFT-DO NOT QUOTE OR CITE
-------
o
Q.
LJJ
"cc
c
o
250
,^ 200-
•» 150
cu
*, 100-
0)
Q_
1
50-
0
r= 0.373
R2= 0.142
T, = 40.5+0.711C,
#= 147
• T= 90 + 01 C or T = N + aC = N + A
0
50 100 150
Ambient PM Concentration, Ct
200
250
Figure 9-11. Regression analysis of daytime total personal exposures to PM10 versus
ambient PM10 concentrations using data from the PTEAM study. The
slope of the regression line is interpreted by exposure analysts as the
average a, where aC = A.
Source: Wilson et al. (2000)
1 exposure, one dependent on concentration, one not (T= 60 + 6jC; Zeger et al., 2000). Exposure
2 analysts associate the component independent of concentration, 60, with cohort average
3 nonambient exposure and the component dependent on concentration, 6l3 with alpha, a, the ratio
4 of ambient exposure to ambient concentration (T = N + aC = N + A; Dockery and Spengler,
5 1981; Ott et al., 2000; Wilson et al., 2000). Most exposure studies report the correlation between
6 ambient concentrations and personal exposure, and many of these also report the slope of the
7 relationship. Since the slope may be interpreted as the average alpha there are a number of
8 studies from which estimates of the average alpha may be estimated. However, the slope may
9 not accurately reflect the average alpha unless the data has been examined for outliers. Several
April 2002
9-32
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
9
10
11
12
13
14
15
studies have interpreted the slope and reported the average FINP or a for cohorts (Ott et al., 2000;
Wilson et al., 2000; Patterson and Eatough, 2000; Landis et al., 2001).
Individual Values
The high correlations found between ambient sulfate and personal sulfate (which has few
indoor sources) suggest that a better relationship may be found between ambient concentrations
and ambient exposures than between ambient concentrations and total personal exposures to PM
(Figure 9-12). The PTEAM study provided sufficient information to permit estimation of
individual values of ambient PM10 exposure, A. These individual values of A were found to be
highly correlated with the corresponding ambient PM10 concentration, C (Figure 9-13) (Wilson
et al., 2000). It is also important to determine whether or not the nonambient exposure, N, is a
function of C, since if TV is not correlated with C, N cannot be a confounder in a regression of
health effects on ambient concentration (Figure 9-14) (Zeger et al., 2000).
1.00 -
0.75 -
£ 0.50 -
0)
0 0.25 -
1
o
0 0.00 -
0
•5; -0.25 -
O
f i
U -0.50 -
-0.75 -
_-i nn -
Ebeltetal.,2000
Pearson's "r"
pt
PM26 y
1
Sulfate
Percentile
90th Percentile
75th Percentile
Median
25th Percentile
10th Percentile
Sarnatetal., 2000
Spearman's "r"
PM25
r
i
T
—
1
gl
T
Sulfate
PM2 5 Sulfate
PM2 5 Sulfate
Figure 9-12. Comparison of correlation coefficients for longitudinal analyses of personal
exposure for individual subjects versus ambient concentrations of PM25 and
sulfate.
Source: Ebelt et al. (2000), Sarnat et al. (2000).
April 2002
9-33
DRAFT-DO NOT QUOTE OR CITE
-------
100
E 75-
§.<
uj 5
50-
25-
50 100 150 200
Ambient PM Concentration, Ct (|jg/m3)
250
Figure 9-13. Regression analysis of daytime exposures to the ambient component of
personal exposure to PM10 (ambient exposure) versus ambient PM10
concentrations.
Source: Wilson et al. (2000).
CO
-t-» ^
CD 5
s_ _
0 -100-
UJ i
1 « 50
CD <:
°" § o-
^n
c
r = 0.051 *
R!= 0.0026 *
N = 39.S+0.086C •
# = 147
. . * . *
* * *
4 4 * +* *
4 74^f/ tr **: *
^vAxi^V *::^^ * 4 *
* * *
i i i
) 50 100 150 200 25
Ambient PM Concentration, C, (|jg/m3)
Figure 9-14. Regression analysis of daytime exposures to the nonambient component of
personal exposure to PM10 (nonambient exposure) versus ambient PM10
concentrations. The two variables are unrelated.
Source: Mage et al. (1999)
April 2002
9-34
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.6.3 Variability in the Relationship Between Concentrations and
2 Personal Exposures
3 The values of the infiltration factor and alpha may vary from person-to-person as shown by
4 the distribution of the infiltration factor and alpha in the PTEAM study (Figure 9-15) (Wilson
5 et al., 2000). The average value of alpha may vary from season-to-season and from city-to-city.
6 The variation in average alpha across cities, as estimated by city-to-city air-conditioning use, can
7 explain some of the variation in the quantitative effects of particles on health across cities
8 (Figure 9-16) (Janssen et al., 2002). For a given PM component, the air exchange rate, a, is a
9 major factor in determining the relationship between outdoor and personal exposure. This has
10 been shown in a study in which personal exposure data were classified into three groups based on
11 home ventilation status. High values of alpha and high correlations were found for the well-
12 ventilated homes, lower values for moderately well-ventilated homes, and much lower values for
13 poorly ventilated homes.
14
15 9.6.4 Exposure Relations for Co-Pollutants
16 The key issue is whether the gaseous co-pollutants (CO, NO2, SO2, and O3) contribute to
17 the health effects attributed to PM or whether they merely serve as surrogates for PM. If the
18 gaseous co-pollutants were responsible for some or all of the health effects attributed to PM in a
19 single pollutant, community time-series epidemiologic analysis, they would be contributors, and
20 the health effects due to PM would be overestimated. However, if the gaseous co-pollutants
21 were surrogates for PM, i.e., significantly correlated with PM but not contributing to the health
22 effects attributed to PM in the analysis, in a multiple regression, the surrogate would share some
23 of the health effect with the causal agent, especially if the surrogate were measured more
24 accurately than the causal agent. Thus, use of a surrogate in a multiple regression would result in
25 an underestimation of the health effects due to PM.
26 In community, time-series epidemiology, in which daily, community-average health effects
27 are regressed against daily ambient concentrations, there are several requirements that must be
28 met in order for a gaseous co-pollutant to be a contributor to the health effects attributed to PM.
29 (1) The gaseous co-pollutant must be capable of causing the effect at the level of the community
30 exposure, (2) the daily ambient concentrations of the gaseous co-pollutant must be related to (i.e.,
31 correlated with) the daily ambient concentrations of the PM indicator, and (3) the daily ambient
April 2002 9-3 5 DRAFT-DO NOT QUOTE OR CITE
-------
Fraction of Ambient PM10 Found Indoors (F,NF = C(AI)/C)
0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90
Fraction of Ambient PM10 Found in Personal Exposure (a =A/C)
0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90
Figure 9-15. Distribution of individual, daily values of the infiltration factor, F^ =
C(AI)/C and the attenuation factor, a = A/C, estimated using data from the
PTEAM study. The distribution of the attenuation factor is shifted to higher
values compared to the infiltration factor because people are exposed to the
full ambient concentration when outdoors.
Source: Wilson et al. (2000).
April 2002
9-36
DRAFT-DO NOT QUOTE OR CITE
-------
0.0025
0.0020 -
• 0.0015 -
o
o
Q 0.0010 -
O
0.0005 -
0.0000
Winter peaking cities
Non-winter peaking cities
O
10 20 30 40 50 60
Central Air Conditioning (%)
70 80
Figure 9-16. Percentage of homes with air conditioning versus the regression coefficient
for the relationship of cardiovascular-related hospital emissions to ambient
PM10 concentrations. The higher the percent air conditioning, the lower the
amount of personal exposure to ambient PM per unit of ambient PM
concentration, i.e., lower a, and there a lower regression coefficient (increase
in risk per until PM10 exposure).
Source: Janssen et al. (2002).
1 concentrations of the gaseous co-pollutant must be related to (correlated with) the personal
2 exposures to that gaseous co-pollutant. Requirements 1 and 2 are also requirements for being a
3 "confounder" in epidemiologic and biostatistics terminology. Whether or not requirement 3 is
4 also a requirement for confounding will depend on the exact definition used for confounding.
5 A fourth requirement, that may apply to confounding, is that the gaseous co-pollutant not be in
6 the formation pathway of the PM. Since SO2 and NO2 are in the formation pathway for the
7 sulfate and nitrate components of PM and O3 is a key chemical reactant in the formation of the
8 sulfate, nitrate, and organic components of PM, this fourth requirement has implications for
9 possible confounding of PM by gaseous co-pollutants that have not yet been adequately analyzed.
10 The exposure analyst is concerned with requirements 2 and 3. How well are the daily
11 ambient concentrations of the gaseous co-pollutants correlated with the daily ambient
April 2002
9-37
DRAFT-DO NOT QUOTE OR CITE
-------
1 concentrations of PM (or specific PM components or indicators) and are the daily ambient
2 concentrations of the gaseous co-pollutants correlated with the daily personal exposures to the
3 ambient? In order to answer these questions quantitatively, information would be needed on the
4 spatial variability of PM indicators and the gaseous co-pollutants and on the variability of the
5 factors which control the infiltration factors (penetration factor and deposition or removal rates).
6 Exposure relationships for gaseous co-pollutants were not reviewed in the exposure chapter
7 (Chapter 5) of this document. Although there have been many exposure studies of the gaseous
8 co-pollutants, there has been little analysis of the experimental data in terms relevant to
9 epidemiology. Exposure studies for CO, NO2, and O3 have been reviewed in the respective Air
10 Quality Criteria Documents (U.S. Environmental Protection Agency, 1993, 1996b, 2000a) and
11 exposure studies of the gaseous co-pollutants and PM components have been reviewed by Monn
12 (2001). Qualitative information on exposure relationships which may be inferred from the
13 studies reviewed in these publications are given in Table 9-6.
14
15
TABLE 9-6. QUALITATIVE ESTIMATES OF EXPOSURE VARIABLES
Highest
High
Medium
Low
Lowest
Spatial Homogeneity1
scv
PM25
NO2, O3, PM10.2 5, SO2
CO, EC4
UF5
Infiltration Factor2
CO
PM2 5, SO4=, EC4
NO2
PM10.2.5
UF5, 03, S02
Stability of the
Infiltration Factor3
CO
PM2 5, SO4=, EC4
N02, PM10.2 5, UF5
03, S02
1. As indicated by the inverse size of the site-to-site correlation coefficient.
2. As indicated by the value of the infiltration factor, inferred in the case of gaseous co-pollutants from
indoor/outdoor ratios for homes without known indoor sources.
3. As indicated by the inverse sensitivity of the deposition or removal rate to the surface to volume ratio and
the chemical composition of the surface.
4. Elemental carbon.
5. Ultrafine particles.
April 2002 9-3 8 DRAFT-DO NOT QUOTE OR CITE
-------
1 Based on the estimates in Table 9-6, it might be expected that the correlation between daily
2 ambient concentrations of PM25 and sulfate and personal exposure to PM25 and sulfate would be
3 high and statistically significant but that this relationship would not be as significant for the
4 gaseous co-pollutants. Two recent studies (Sarnat et al., 2000, 2001) provide new information
5 relevant to the possible contribution of gaseous co-pollutants to the health effects attributed to
6 PM. Personal exposure measurements were made of NO2, O3, and sulfate (winter and summer)
7 and of SO2 and EC (winter only). Ambient measurements were made of these species (same
8 seasons) and of CO (both seasons). Personal exposures to ambient PM2 5 were estimated by
9 using the daily, individual ratios of personal exposure to sulfate to ambient concentrations of
10 sulfate as an estimate of the attenuation factor for PM25. Correlations among ambient
11 concentrations, among personal exposures, and between ambient concentrations and personal
12 exposures were examined.
13 Daily personal exposures to NO2 and O3 were not significantly correlated with daily
14 ambient concentrations of those gaseous co-pollutants in either summer or winter. This suggests
15 that NO2 and O3 cannot be contributors to the health effects attributed to PM in an epidemiologic
16 analysis using daily ambient concentrations. In the winter, daily personal exposures to SO2 were
17 negatively correlated with daily ambient concentrations of SO2. Personal exposures to CO were
18 not reported. During summer, O3 and NO2 were positively and significantly associated with
19 PM25; the association with CO was positive but not significant. During winter, CO and NO2
20 were positively and significantly associated with PM2 5 while O3 was negatively and significantly
21 associated with PM25; the association with SO2 was negative but not significant. Similar
22 association of gaseous co-pollutants were found with personal exposure to PM2 5 except that the
23 winter association with SO2 became significant. Also, the significant associations were more
24 significant with personal exposure to ambient PM2 5. This indicates that daily ambient
25 concentrations of CO, NO2, O3 and SO2 can be surrogates for daily ambient concentrations of
26 PM2 5 but that exposure and epidemiologic analyses including O3 and SO2 need to examine
27 relationships on a seasonal basis. These studies also indicate that daily ambient concentrations of
28 PM2 5, CO, NO2, O3 and SO2 serve as surrogates for daily personal exposures to PM2 5 and are
29 even better surrogates for daily personal exposures to ambient PM2 5. Thus, in a multiple
30 regression using PM and a gaseous copollutant, both variables would be surrogates for personal
31 exposure to ambient PM.
April 2002 9-39 DRAFT-DO NOT QUOTE OR CITE
-------
1 Sarnat et al. (2001) point out that "it is inappropriate to treat one variable as a confounder
2 of another when both variables are actually surrogates of the same thing." While the exposure
3 results from these studies are based on a small number of non-randomly chosen subjects and
4 therefore cannot be extrapolated with assurance to other situations, they do indicate the value of
5 exposure analysis in identifying which of several collinear variables could possibly be causal.
6 The work also suggests that neither NO2, O3, nor SO2 are likely to be the causal factor in the
7 reported associations of ambient PM with health effects. No information was found on the
8 correlation of ambient CO with personal exposure to CO in homes with no indoor CO sources.
9 However, the low spatial homogeneity of ambient CO concentrations suggests that the
10 relationship would be weak. Therefore, it seems likely, but not certain, that exposure
11 relationships would also indicate that CO is unlikely to be a contributor to the health effects
12 attributed to PM. It is important to understand that this does not indicate that these ambient
13 pollutants do not cause health effects of the type associated with PM in epidemiologic analyses.
14 It only indicates that community, time-series epidemiology using ambient concentrations cannot
15 provide information on the possible health effects of pollutants whose ambient concentrations are
16 not significantly correlated with personal exposure to that ambient pollutant.
17 Sarnat et al. (2001) also suggest that some of the gaseous co-pollutants may be acting as
18 surrogates for specific PM25 source categories or components. "For subjects with COPD,
19 ambient CO and NO2 were not significantly associated with total personal PM2 5, but were
20 significantly associated with personal exposure to PM2 5 of ambient origin and also to personal
21 elemental carbon (EC). These significant associations may be due to the fact that motor vehicles
22 are a major source of CO, NO2, EC, and, to a lesser degree, to PM2 5 of ambient origin.
23 Conversely, ambient CO and NO2 were not significantly associated with personal sulfate, a
24 pollutant not associated with motor vehicle emissions. O3, in contrast, was predominantly
25 associated with personal sulfate (positively in summer and negatively in winter) . . ." Thus, CO,
26 NO2, EC, and PM2 5 may be surrogates for personal exposure to pollutants from motor vehicles
27 and O3 may be a surrogate for regional sulfate. It should be noted that since PM2 5, CO, NO2, EC,
28 and PM associated with motor vehicles are all significantly correlated with each other, a
29 community, time-series epidemiologic analysis, in one community for one time period, cannot
30 tell whether a variable is actually responsible for relationship between concentration and health
31 effects observed in the analysis, or whether the variable is a surrogate for the causal variable.
April 2002 9-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 In order to more clearly differentiate between contributor and surrogate, it will be necessary to
2 integrate information from toxicology and exposure analysis, as well as from epidemiologic
3 studies in different time periods and different communities.
4
5 9.6.5 Summary
6 For most cities, site-to-site correlations are high for PM2 5. However, the spatial
7 distribution of PM should be investigated before beginning long term monitoring for exposure or
8 epidemiologic studies. The relationship between the concentrations of an ambient pollutant
9 outdoors and the contribution of that ambient pollutant to personal exposure is given by the mass
10 balance model (Equation 9-2) and depends on the outdoor concentration, the time spent outdoors
11 and indoors, the air exchange rate, and penetration factor, and the indoor deposition or removal
12 rate. For a given PM component, the major cause of variability in the relationship is the air
13 exchange rate. For gaseous co-pollutants, if the correlation between the ambient concentrations
14 and the personal exposures to the ambient concentrations are not statistically significant, that
15 gaseous co-pollutant cannot contribute to the health effect attributed to PM in a community,
16 time-series epidemiologic analysis. However, if ambient concentration of the gaseous
17 co-pollutant is significantly correlated with the ambient concentration of PM, it may be as good
18 or better indicator of personal exposure to the lexicologically active component of PM as the
19 ambient PM concentration; and, thus, it may be a surrogate, i.e., it will falsely show an effect due
20 to its correlation with the personal exposure to the active component and, in a multiple
21 regression, it will appear to reduce the effect associated with PM. Therefore, correlations among
22 ambient concentrations and ambient concentration-personal exposure relationships for PM and
23 co-pollutants are useful in interpreting the results of epidemiologic studies.
24
25
26 9.7 EXPOSURE TO BIOLOGICALLY IMPORTANT
27 CHARACTERISTICS OF PARTICULATE MATTER
28 What are the exposures to biologically important constituents and specific characteristics
29 of paniculate matter that cause responses in potentially susceptible subpopulations and the
30 general population?
31
April 2002 9-41 DRAFT-DO NOT QUOTE OR CITE
-------
1 In their discussion of Topic 2, the NRC notes that in order to make such investigations
2 practicable, it will be necessary to characterize susceptible subpopulations more fully, identify
3 lexicologically important chemical constituents or particle-size fractions, develop and field-test
4 exposure-measurement techniques for relevant properties of PM, and design comprehensive
5 studies to determine population exposures.
6
7 9.7.1 Exposure Relationships for Susceptible Subpopulations
8 Children, the elderly, and people with pre-existing diseases such as diabetes, respiratory
9 disease, and cardiovascular disease appear to constitute susceptible subpopulations. A number of
10 studies of small cohorts drawn from these and other subpopulations have been conducted
11 recently by EPA and other organizations. Correlations between ambient concentrations and total
12 personal exposure have been presented for a few of these. However, most of the studies have not
13 yet been published, most of the studies have not reported the ambient exposure, and the studies
14 have not been analyzed to determine if there are indeed exposure differences between susceptible
15 groups and the general population.
16 An analysis of cohort exposure studies available in 1998 (Wallace, 2000) concluded that
17 the personal cloud component of nonambient exposure was less for subjects with COPD than for
18 the general population, healthy elderly subjects or children, presumably because of the higher
19 activity level of younger or healthier subjects. However, the relationship between ambient
20 concentrations and personal exposure for COPD patients was not better than that for other
21 cohorts. Wallace (2000) noted that the desirable correlation is that "between personal exposure
22 to particles originating outdoors and outdoor concentrations." However, at that time there was
23 no information on the ambient component of personal exposure. Unfortunately, there is still no
24 published information that would suggest differences in exposure relationships for healthy versus
25 susceptible populations.
26
27 9.7.2 Toxicologically Important Components of PM
28 Inherent in the NRC research agenda (NRC, 1998) was the consideration that one, or
29 perhaps a few, characteristics of PM would be associated with toxicity, and exposure monitoring
30 could concentrate on these components. However, it has not yet been possible to identify any
April 2002 9-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM characteristic as not being of toxicologic importance. Table 9-7 lists characteristics of PM
2 that have been found to be associated with toxicity either through epidemiologic or toxicologic
3 studies.
4
5
TABLE 9-7. PARTICULATE MATTER CHARACTERISTICS POTENTIALLY
RELEVANT TO HEALTH
Particle number
Particle surface area
Mass-ultrafme PM [PM0 J
Mass-fine PM [PM2 5 or PMl 0]
Mass-thoracic coarse PM [PM10_2 5 or PM1(M]
Sulfate
Strong acidity (H+)
Nitrate
Elemental carbon
Organic carbon (many different compounds)
Transition metals
Specific toxic metals
Bioaerosols
1 9.7.3 Exposure-Measurement Techniques
2 Measurement techniques, suitable for stationary monitors with 24-hour collection periods,
3 exist for the characteristics of PM listed in Table 9-7. For many of these measurements,
4 continuous or 1-hour-average stationary monitors also exist or are in development. However,
5 personal monitoring is usually limited to either PM2 5 or PM10. A few studies have included
6 passive monitors for NO2, O3, SO2, and CO. A roll-around monitor, which can be rolled around
7 to follow a person and thus simulate a personal exposure measurement with a more complete
April 2002 9-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 suite of measurements, has been used in recent studies. However, with the exception of personal
2 light-scattering monitors, which do not have well-established relationships with PM mass, there
3 are still no adequate personal monitors for continuous measurement of mass of other PM
4 characteristics.
5
6 9.7.4 Comprehensive Studies to Determine Population Exposure
7 Chapter 5 reports only four exposure studies that have even attempted to provide
8 statistically representative studies of population exposure of the general population or susceptible
9 subgroups. However, only in the case of the PTEAM study has the exposure data been used to
10 estimate ambient and nonambient exposure separately (for PM10). Even though statistically
11 representative studies are limited, available data from small cohorts allow some inferences
12 regarding differences in concentration-exposure relationships among different characteristics of
13 PM. PM may be classified by particle size, by chemical composition, or by sources.
14 Concentration - exposure relationships may be different for different classes of particles.
15
16 Central Site to Outdoors
17 The 1996 PM AQCD reported information from a few cities (mostly eastern US) that
18 suggested that site-to-site correlation coefficients, r, were high for sulfate and PM25 in some
19 cities; were lower but still relatively high for PM10 and TSP; but were low for PM10_2 5 (Figure
20 9-17). However, there was little information on site-to-site correlations of chemical components
21 of PM (except sulfate and strong acidity) or of orthogonal source-category factors. New site-to
22 site correlation studies, using PM data from the AIRS data base, are presented in Chapter 3.
23 Some examples of the differences in site-to-site correlations for PM2 5 and PM10_25, derived from
24 the data in Chapter 3, are shown in Figure 9-18. It should be noted that the PM2 5 data is from
25 1999 and 2000 and satisfies certain criteria for number of days of data per season. The PM10_2 5
26 data is from 2000 only and is less complete. In addition, some information on the site-to-site
27 correlations of PM25 components and source contributions are now available (Figures 9-19 and
28 9-20). In order to reduce spatial variability, some cohort studies have used the concentration at
29 the nearest monitoring site or the distance to major traffic sources for exposure information.
30
31
April 2002 9-44 DRAFT-DO NOT QUOTE OR CITE
-------
1.0
0.9
0.8
0.7
o 0.6
£
-------
-0.2
Site Pairs
- Mass j
- Sulfate |
-OC |
-EC j
-Crustal |
- Lead I
Figure 9-19. Site-to-site correlation coefficients for PM25 mass and some chemical
components of PM25 in 1994 in Philadelphia, PA.
Source: Pinto et al. (1995).
o 1
10 11
Site Pair
- Mass |
- Secondary I
- Motor Vehicles j
- Crustal |
- Residual Oil I
Figure 9-20. Site-to-site correlation coefficients for PM2 5 mass and several source category
factors in 1986 in the South Coast Basin (Los Angeles area).
Source: Wongphatarakul et al. (1998)
April 2002
9-46
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
9
10
11
12
13
14
15
Outdoors to Indoors
Information on the infiltration rate, FINP , as a function of particle size may be obtained as
follows. Indoor and outdoor measurements of PM concentrations as a function of particle size
are made during the night when it is assumed that there are no indoor activities occurring that
might generate indoor PM. Under this assumption the indoor concentration measurement is
C(AI) and C(AI)/C = FINP (Long et al., 2000). As can be seen in Figure 9-21, FINP is low for
ultrafme and coarse particles but high for accumulation mode particles. FINP also depends on the
air exchange rate, a, FINP increases when a increases. The variation of P and k as a function of
particle size can also be determined by this technique (Figure 9-22) (Long et al., 2000). There is
little information on ambient concentration - exposure relationships for specific chemical
components, except sulfate, or for specific source categories, other than what would be inferred
from the size distributions. Infiltration ratios are low for components like strong acidity (FT) that
are neutralized by indoor-generated ammonia or like ammonium nitrate (NH4NO3) that evaporate
indoors.
1.1
o
ro
u.
c
_o
'•C
2
+-
t
1.0 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0.0
0.1
Summer Fall
LO CNI CO
(Dp T—
0 O o
O
co
CD
I--.
CD
Particle Diameter (|jm)
Source: Long, Sun and Koutrakis (2000)
Figure 9-21. Values of geometric mean infiltration factor, FINF = A/C, as a function of
particle diameter for hourly nighttime data (assuming no indoor sources) for
summer and fall seasons. Distribution of air exchange rates, a, for each
season are shown in the insert.
Source: Long et al. (2000).
April 2002
9-47
DRAFT-DO NOT QUOTE OR CITE
-------
5-
CD
'o
£
ts
C
d>
Q.
1.2
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
S §
q CD ^
q ,1 ^
LO CN CO
•si- LO CD
m -t ui
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
i
O
O
Q
Size Interval (pm)
Figure 9-22. Values of penetration efficiency and deposition rate as a function of particle
diameter estimated from model of average nighttime indoor-outdoor
concentration data.
Source: Long et al. (2000).
1
2
3
4
5
9
10
11
12
13
14
9.7.5 Air Pollutants Generated Indoors
The NRC discussion of Research Topic 2 is clear that the primary purpose of the
investigations recommended should be "to examine the outdoor contributions to measurements
of total personal exposure." However, they also recommended determining the exposure to "air
pollutants generated indoors." Total personal exposure includes both ambient and nonambient
sources. Important sources of indoor PM are smoking, cooking, and cleaning. Because of the
variation of Finfwith particle size, ambient-infiltrated PM tends to be primarily in the
accumulation mode. As shown in Table 9-8, however, indoor PM is generated primarily in the
ultrafme mode (smoking, other combustion sources, most cooking) or the coarse mode (cleaning,
sauteing). Another, possibly important indoor source, is the reaction of ambient-infiltrated ozone
with indoor emissions of terpenes from air fresheners or cleaning agents, e.g., cleaning with Pine
Sol. These particles are also generated largely in the ultrafme mode. Ambient and indoor
generated PM also differ somewhat in their chemical composition as shown in Table 9-9.
April 2002
9-48
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-8. VOLUME MEAN DIAMETER (VMD) OF INDOOR
PARTICLE SOURCESab
Particle Source
Cooking
Baking (Electric)
Baking (Gas)
Toasting
Broiling
Stir-Frying
Frying
Barbecuing
Sauteing, fine
Sauteing, coarse
Cleaning
Dusting
Vacuuming
Cleaning with Pine Sol
General Activities
Walking Vigorously (w/Carpet)
Sampling w/Carpet
Sampling w/o Carpet
Burning Candles
N
8
24
23
4
3
20
2
13
13
11
10
5
15
52
26
7
Indoor Activity - Mean VMD
(//m)
0.189
0.107
0.138
0.114
0.135
0.173
0.159
0.184
3.48
5.38
3.86
0.097
3.96
4.25
4.28
0.311
Notes:
Includes only individual particle events that were unique for a given time period and could be detected above
background particle levels.
Fine particle sizes calculated for PV0 02.0 5 using SMPS data; coarse particle sizes calculated for PV0 7_10 using
APS data.
Source: Long et al. (2000).
April 2002 9-49 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-9. CONCENTRATION DIFFERENCES BETWEEN CONSTITUENTS OF
NONAMBIENT (INDOOR-GENERATED) AND AMBIENT PM
Higher Concentration in Nonambient PM
Higher Concentration in Ambient PM
Mold Spores
Endotoxin
Animal Dander
Biological Fragments
(from insects, etc)
Environmental Tobacco Smoke
Resuspended Soil and House Dust
Ultrafine Particles and Coarse-Mode Particles
Pollen
Transition Metals (non-soil Fe, Mn)
Other Metals (Se, As, Ni, Cu)
Oxygenated and Nitrated
Polyaromatic Compounds
Other Oxygenated Organic Compounds
Sulfates and Nitrates
Accumulation-Mode Particles
1 9.8 DOSIMETRY: DEPOSITION AND FATE OF PARTICLES IN THE
2 RESPIRATORY TRACT
3 What are the deposition patterns and fate of particles in the respiratory tract of individuals
4 belonging to presumed susceptible subpopulations?
5 Knowledge of the dose of particles delivered to a target site or sites in the respiratory tract
6 is important for understanding possible health effects associated with human exposure to ambient
7 PM and for extrapolating and interpreting data obtained from studies of laboratory animals. The
8 dosimetry of particles of different sizes are subject to large differences in regional respiratory
9 tract deposition, translocation, and clearance mechanisms and pathways and, consequently,
10 retention times. The following sections summarize the current understanding of the physical
11 characteristics of particles and the biological determinants that affect particle dosimetry
12 mechanisms and pathways, as discussed in Chapter 6.
13
14 9.8.1 Particle Deposition in the Respiratory Tract
15 For dosimetry purposes, the respiratory tract can be divided into three regions:
16 (1) extrathoracic (ET), (2) tracheobronchial (TB), and (3) alveolar (A). The ET region consists
17 of head airways (i.e., nasal and oral passages) through the larynx and represents the areas through
18 which inhaled air first passes. In humans, inhalation can occur through the nose or mouth (or
April 2002
9-50
DRAFT-DO NOT QUOTE OR CITE
-------
1 both, known as oronasal breathing). However, most laboratory animals commonly used in
2 respiratory toxicological studies are obligate nose breathers.
3 From the ET region, inspired air enters the TB region at the trachea. From the level of the
4 trachea, the conducting airways then undergo branching for a number of generations. The
5 terminal bronchiole is the most peripheral of the distal conducting airways and these lead,
6 in humans, to the respiratory bronchioles, alveolar ducts, alveolar sacs, and alveoli (all of which
7 comprise the A region). All of the conducting airways, except the trachea and portions of the
8 mainstem bronchi, are surrounded by parenchymal tissue. This is composed primarily of the
9 alveolated structures of the A region and associated blood and lymphatic vessels. It should be
10 noted that the respiratory tract regions are comprised of various cell types and that there are
11 distinct differences in the cells of airway surfaces in the ET, TB, and A regions.
12 Particles deposit in the respiratory tract by five mechanisms: (1) inertial impaction,
13 (2) sedimentation, (3) diffusion, (4) electrostatic precipitation, and (5) interception. Sudden
14 changes in airstream direction and velocity cause inhaled particles to impact onto airway
15 surfaces. The ET and upper TB airways are dominant sites of inertial impaction, a key
16 mechanism for particles with aerodynamic diameter (Da) >1 //m. Particles with Da > 0.5 //m
17 mostly are affected by sedimentation out of the airstream. Both sedimentation and inertial
18 impaction influence deposition of particles in the same size range and occur in the ET and TB
19 regions, with inertial impaction dominating in the upper airways and gravitational settling
20 (sedimentation) increasingly more dominant in lower conducting airways. Particles with actual
21 physical diameters < 1 //m are increasingly subjected to diffusive deposition due to random
22 bombardment by air molecules, resulting in contact with airway surfaces. Particles between
23 0.3 and 0.5 //m in size are small enough to be little influenced by impaction or sedimentation and
24 large enough to be minimally influenced by diffusion, and so, they undergo the least respiratory
25 tract deposition. The interception potential of any particle depends on its physical size; fibers are
26 of chief concern for interception, their aerodynamic size being determined mainly by their
27 diameter. Electrostatic precipitation is deposition related to particle charge; effects of charge on
28 deposition are inversely proportional to particle size and airflow rate. This type of deposition is
29 likely small compared to effects of other deposition mechanisms and is generally a minor
30 contributor to overall particle deposition, but one recent study found it to be a significant TB
31 region deposition mechanism for ultrafine, and some fine, particles.
April 2002 9-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 The ET region acts as an efficient filter that reduces penetration of inhaled particles to the
2 TB and A regions of the lower respiratory tract. Total respiratory tract deposition increases with
3 particle size for particles >1.0 //m Da, is at a minimum for particles 0.3 to 0.5 //m, and increases
4 as particle size decreases below that range. The ET deposition is higher with nose breathing than
5 for mouth breathing, with increased ventilation rates associated with increasing levels of physical
6 activity or exercise leading to more oronasal breathing and increased delivery of inhaled particles
7 to TB and A regions in the lung.
8 Hygroscopicity, the propensity of a material for taking up and retaining moisture, is a
9 property of some ambient particle species and affects respiratory tract deposition. Such particles
10 can increase in size in humid air in the respiratory tract and, when inhaled, deposit according to
11 their hydrated size rather than their initial size. Compared to nonhygroscopic particles of the
12 same initial size, deposition of hygroscopic aerosols in different regions varies, depending on
13 initial size: hygroscopicity generally increases total deposition for particles with initial sizes
14 larger than «0.5 //m, but decreases deposition for particles between «0.01 and 0.5 and again
15 increases deposition for particles <0.01 //m.
16 Enhanced particle retention occurs on carinal ridges in the trachea and throughout the
17 segmental bronchi; and deposition "hot spots" occur at airway bifurcations or branching points.
18 Peak deposition sites shift from distal to proximal sites as a function of particle size, with greater
19 surface dose in conducting airways than in the A region for all particle sizes. Whereas both fine
20 (<2.5 //m) and thoracic coarse (2.5 to 10 //m) particles deposit to about the same extent on a
21 percent particle mass basis in the trachea and upper bronchi, a distinctly higher percent of fine
22 particles deposit in the A region. However, surface number dose (particles/cm2/day) is much
23 higher for fine than for coarse particles, indicating much higher numbers of fine particles
24 depositing, with the fine fraction contributing upwards of 10,000 times greater particle number
25 per alveolar macrophage.
26 Ventilation rate, gender, age, and respiratory disease status are all factors that affect total
27 and regional respiratory tract particle deposition. In general, because of somewhat faster
28 breathing rates and likely smaller airway size, women have somewhat greater deposition of
29 inhaled particles than men in upper TB airways, but somewhat lower A region deposition than
30 for men. Children appear to show four effects: (1) greater total respiratory tract deposition than
31 adults (possibly as much as 50% greater for those <14 years old than for adults >14 years),
April 2002 9-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 (2) distinctly enhanced ET region deposition (decreasing with age from 1 year), (3) enhanced TB
2 deposition for particles < 5 //m, and (4) enhanced A region deposition (also decreasing with age).
3 Overall, given that children have smaller lungs and higher minute volumes relative to lung size,
4 they likely receive greater doses of particles per lung surface area than adults for comparable
5 ambient PM exposures. This and the propensity for young children to generally exhibit higher
6 activity levels and associated higher breathing rates than adults likely contribute to enhanced
7 susceptibility to ambient particle effects resulting from particle dosimetry factors. In contrast,
8 limited available data on respiratory tract deposition across adult age groups (18 to 80 years) with
9 normal lung function do not indicate age-dependent effects (e.g., enhanced deposition in healthy
10 elderly adults). Altered PM deposition patterns due to respiratory disease status may put certain
11 groups of adults (including some elderly) and children at greater risk for PM effects.
12 Both information noted in the 1996 PM AQCD and newly published findings discussed in
13 this document indicate that respiratory disease status is an especially important determinant of
14 respiratory tract particle deposition. Importantly, the pathophysiologic characteristics of chronic
15 obstructive pulmonary disease (COPD) contribute to more heterogenous deposition patterns and
16 differences in regional deposition. One study indicates that people with COPD tend to breath
17 faster and deeper than those with normal lungs (i.e., about 50% higher resting ventilation) and
18 had about 50% greater deposition than age-matched healthy adults under typical breathing
19 conditions, with average deposition rates 2.5 times higher under elevated ventilation rates.
20 Enhanced deposition appears to be associated more with the chronic bronchitic than the
21 emphysematous component of COPD. In this and other new studies, fine-particle deposition
22 increased markedly with increased degree of airway obstruction (ranging up to 100% greater with
23 severe COPD). With increasing airway obstruction and uneven airflow because of irregular
24 obstruction patterns, particles tend to penetrate more into remaining better ventilated lung areas,
25 leading to enhanced focal deposition at airway bifurcations and alveoli in those A region areas.
26 In contrast, TB deposition increases with increasingly more severe bronchoconstrictive states, as
27 occur with asthmatic conditions.
28 Differences between species in particle deposition patterns were summarized in the 1996
29 PM AQCD and more recently by Schlesinger et al. (1997), as discussed in Chapter 6 of this
30 document. These differences should be considered when relating biological responses obtained
31 in laboratory animal studies to effects in humans. Various species used in inhalation toxicology
April 2002 9-53 DRAFT-DO NOT QUOTE OR CITE
-------
1 studies serving as the basis for dose-response assessment may not receive identical doses in a
2 comparable respiratory tract region (i.e., ET, TB, A) when exposed to the same aerosol at the
3 same inhaled concentration. This is illustrated by mathematical modeling studies that evaluate
4 interspecies differences in respiratory tract deposition. For example, Hofmann et al. (1996)
5 found total deposition efficiencies for all particles (0.01, 1, and 10 //m) at upper and lower
6 airway bifurcations to be comparable for rats and humans, but when higher penetration
7 probabilities from preceding airways in the human lung were considered, bronchial deposition
8 fractions were mostly higher for humans. For all particle sizes, deposition at rat bronchial
9 bifurcations was less enhanced on the carinas than in human airways. Numerical simulations of
10 three-dimensional particle deposition patterns within selected (species-specific) bronchial
11 bifurcations indicated that interspecies differences in morphologic asymmetry is a major
12 determinant of local deposition patterns. The dependence of deposition on particle size is similar
13 in rats and humans, with deposition minima in the 0.1- to l-//m size range for both total
14 deposition and deposition in the TB and A regions, but total respiratory tract and TB deposition
15 was consistently higher in the human lung. Alveolar regional deposition in humans was lower
16 than in rat for 0.001- to 10-//m particles (deposition of such particles being highest in the upper
17 bronchial airways), whereas it was higher for 0.1- and l-//m particles in more peripheral airways
18 (i.e., bronchiolar airways in rat, respiratory bronchioles in humans). In a histology study, Nikula
19 et al. (2000) examined particle retention in rats (exposed to diesel soot) and humans (exposed to
20 coal dust). In both, the volume density of deposition increased with increasing dose. In rats,
21 diesel exhaust particles were found mainly in lumens of the alveolar duct and alveoli, whereas in
22 humans, retained dust was mainly in interstitial tissue. Thus, in the two species, different lung
23 cells appear to contact retained particles and may result in different biological responses with
24 chronic exposure.
25 The probability of any biological effect of PM in humans or animals depends on particle
26 dosimetry, and subsequent particle retention, as well as underlying dose-response relationships.
27 Interspecies dosimetric extrapolation must, therefore, consider differences in deposition,
28 clearance, translocation, and dose-response. Even similar deposition patterns may not result in
29 similar effects in different species, because dose also is affected by clearance mechanisms and
30 species sensitivity. Total number of particles deposited in the lung may not be the most relevant
31 dose metric by which to compare species; rather, the number of deposited particles per unit
April 2002 9-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 surface area may determine response. Even if deposition is similar in rats and humans, there
2 would be a higher deposition density in the rat because of the smaller surface area of the rat lung.
3 Thus, species-specific differences in deposition density are important when attempting to
4 extrapolate health effects observed in laboratory animals to humans.
5
6 9.8.2 Particle Clearance and Translocation
7 Particles depositing on airway surfaces may be cleared from the respiratory tract completely
8 or translocated to other sites within this system by regionally specific clearance mechanisms, as
9 follow: ETregion—mucocialiary transport, sneezing, nose wiping and blowing, and dissolution
10 and absorption into blood; TB region—mucociliary transport, endocytosis by macrophages and
11 epithelial cells, coughing, and dissolution and absorption into blood and lymph; A region—
12 macrophages, epithelial cells, interstitial, and dissolution and absorption into blood and lymph.
13 Regionally specific clearance defense mechanisms operate to clear deposited particles of
14 varying particle characteristics (size, solubility, etc.) from the ET, TB, and A regions and are
15 variously affected by different disease states. For example, particles are cleared from the ET
16 region by mucociliary transport to the nasopharynx area, dissolution and absorption into the
17 blood, or sneezing, wiping or blowing of the nose; but such clearance is slowed by chronic
18 sinusitis, bronchiectasis, rhinitis, and cystic fibrosis. Also, in the TB region, poorly soluble
19 particles are cleared mainly by upward mucociliary transport or by phagocytosis by airway
20 macrophages that move upward on the mucociliary blanket, followed by swallowing. Soluble
21 particles in the TB region are absorbed mostly into the blood and some by mucociliary transport.
22 Although TB clearance is generally fast and much material is cleared in <24 h, the slow
23 component of TB clearance (likely associated with bronchioles 24 h and clearance
25 half-times of about 50 days. Bronchial mucous transport is slowed by bronchial carcinoma,
26 chronic bronchitis, asthma, and various acute respiratory infections; these are disease conditions
27 that logically would be expected to increase retention of deposited particle material and, thereby,
28 increase the probability of toxic effects from inhaled ambient PM components reaching the TB
29 region. Also, spontaneous coughing, an important TB region clearance mechanism, does not
30 appear to fully compensate for impaired mucociliary clearance in small airways and may become
31 depressed with worsening airway disease, as seen in COPD.
April 2002 9-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 Clearance of particles from the A region by alveolar macrophages and their mucociliary
2 transport is usually rapid (<24 h). However, penetration of uningested particles into the
3 interstitium increases with increasing particle load and results in increased translocation to lymph
4 nodes. Soluble particles not absorbed quickly into the blood stream and translocated to
5 extrapulmonary organs (e.g., the heart) within minutes may also enter the lymphatic system, with
6 lymphatic translocation probably being increased as other clearance mechanisms (e.g., removal
7 by macrophages) are taxed or overwhelmed under "particle overload" conditions. Insoluble
8 particles <2 //m clear to the lymphatic system at a rate independent of size; particles of this size,
9 more so than those >5.0 //m, are deposited significantly in the A region. Translocation into the
10 lymphatic system is quite slow, and elimination from lymph nodes even slower (half-times
11 estimated in decades). Focal accumulations of reservoirs of potentially toxic materials and their
12 slow release for years after initial ambient PM exposure may account partially for the observation
13 in epidemiologic studies that higher relative risks are associated with long-term ambient PM
14 exposure than can be accounted for by additive effects of acute PM exposures. Alveolar region
15 clearance rates are decreased in human COPD sufferers and slowed by acute respiratory
16 infections, and the viability and functioning of alveolar macrophages are reduced in human
17 asthmatics and in animals with viral lung infections. These observations suggest that persons
18 with asthma or acute lung infections are likely at increased risk for ambient PM exposure effects.
19 Differences in regional and total clearance rates between some species reflect differences in
20 mechanical clearance processes. The importance of interspecies clearance differences is that
21 retention of deposited particles can differ between species and may result in differences in
22 response to similar PM exposures. Hsieh and Yu (1998) summarize existing data on pulmonary
23 clearance of inhaled, poorly soluble particles in the rat, mouse, guinea pig, dog, monkey, and
24 human. Two clearance phases, "fast" and "slow," in the A region are associated with mechanical
25 clearance along two pathways, the former with the mucociliary system and the latter with lymph
26 nodes. Rats and mice are fast clearers, compared to other species. Increasing initial lung burden
27 results in an increasing mass fraction of particles cleared by the slower phase. As lung burden
28 increases beyond 1 mg particles/g lung, the fraction cleared by the slow phase increases to almost
29 100% for all species. The rate for the fast phase is similar in all species, not changing with
30 increasing lung burden, whereas the slow phase rate decreases with increasing lung burden.
31 At elevated burdens, the "overload" effect on clearance rate is greater in rats than in humans.
April 2002 9-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.8.3 Deposition and Clearance Patterns of Particles Administered by
2 Inhalation Versus Intratracheal Instillation
3 Inhalation is the most directly relevant exposure route for evaluating PM toxicity, but many
4 studies deliver particles by intratracheal instillation. Because particle disposition is a determinant
5 of dose, it is important to compare deposition and clearance of particles delivered by instillation
6 versus inhalation. It is difficult to compare particle deposition and clearance among different
7 inhalation and instillation studies because of differences in experimental methods and in
8 quantification of particle deposition and clearance. Key points from a recent detailed evaluation
9 (Driscoll et al., 2000) of the role of instillation in respiratory tract dosimetry and toxicology
10 studies are informative. In brief, inhalation may result in deposition within the ET region, the
11 extent of which depends on the size of the particles used, but intratracheal instillation bypasses
12 this portion of the respiratory tract and delivers particles directly to the TB tree. Although some
13 studies indicate that short (0 to 2 days) and long (100 to 300 days postexposure) phases of
14 clearance of insoluble particles delivered either by inhalation or intratracheal instillation are
15 similar, others indicate that the percent retention of particles delivered by instillation is greater
16 than for inhalation, at least up to 30 days postexposure. Another salient finding is that inhalation
17 generally results in a fairly homogeneous distribution of particles throughout the lungs, but
18 instillation is typified by heterogeneous distribution (especially in the A region) and high levels
19 of focal particles. Most instilled material penetrates beyond the major tracheobronchial airways,
20 but the lung periphery is often virtually devoid of particles. This difference is reflected in
21 particle burdens within macrophages, those from animals inhaling particles being burdened more
22 homogeneously and those from animals with instilled particles showing some populations of
23 cells with no particles and others with heavy burdens, and is likely to impact clearance pathways,
24 dose to cells and tissues, and systemic absorption. Exposure method, thus, clearly influences
25 dose distribution that argues for caution in interpreting results from instillation studies.
26
27 9.8.4 Inhaled Particles as Potential Carriers of Toxic Agents
28 It has been proposed that particles also may act as carriers to transport toxic gases into the
29 deep lung. Water-soluble gases, which would be removed by deposition to wet surfaces in the
30 upper respiratory system during inhalation, could dissolve in particle-bound water and be carried
31 with the particles into the deep lung. Equilibrium calculations indicate that particles do not
April 2002 9-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 increase vapor deposition in human airways. However, these calculations do show that soluble
2 gases are carried to higher generation airways (deeper into the lung) in the presence of particles
3 than in the absence of particles. In addition, species such as SO2 and formaldehyde react in
4 water, reducing the concentration of the dissolved gas-phase species and providing a kinetic
5 resistence to evaporation of the dissolved gas. Thus, the concentration of the dissolved species
6 may be greater than that predicted by the equilibrium calculations. Also, certain other toxic
7 species (e.g., nitric oxide [NO], nitrogen dioxide [NO2], benzene, polycyclic aromatic
8 hydrocarbons [PAH], nitro-PAH, a variety of allergens) may be absorbed onto solid particles and
9 carried into the lungs. Thus, ambient particles may play important roles not only in inducing
10 direct health impacts of their constituent components but also in facilitating delivery of toxic
11 gaseous pollutants or bioagents into the lung and may, thereby, serve as key mediators of health
12 effects caused by the overall air pollutant mix.
13
14 9.8.5 Summary of Particle Dosimetry
15 Although the current understanding of basic mechanisms of particle dosimetry, clearance,
16 and retention has not changed since the 1996 PM AQCD, additional information has become
17 available on the role of certain biological determinants of these processes, such as gender and
18 age; and there has been an expansion of previous knowledge about the relationship between
19 regional deposition and translocation in regard to specific particle size ranges of significance to
20 ambient particulate exposure scenarios. There also has been significant improvement in the
21 mathematical and computational fluid dynamic modeling of particle dosimetry in the respiratory
22 tract of humans. Although the models have become more sophisticated and versatile, validation
23 of the models is still needed.
24 One of the areas that has improved since the 1996 PM ACQD is consideration of specific
25 and relevant ambient size particle ranges in deposition studies. One such size mode is the nuclei
26 mode or ultrafine particles (< 0.1 //m). While further information on respiratory deposition for
27 this size mode is still needed, there has been an improvement in the understanding of total
28 deposition as a function of particle size and breathing pattern and of certain aspects of regional
29 deposition of ultrafine particles. This new information indicates that the ET region, especially
30 the nasal passages, is a very efficient "filter" for these particles, reducing the amount which
31 would be available for deposition in the TB and A regions of the respiratory tract. Within the
April 2002 9-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 thoracic region, the deposition distribution of ultrafme particles is highly skewed towards the
2 proximal airway regions and resembles that of coarse particles. In other words, deposition
3 patterns of ultrafme particles are very much like those of coarse particles. Another example
4 involves studies which attempt to evaluate the contribution of fine- and coarse-mode particles to
5 deposition in various parts of the respiratory tract, although there have been only a few of these.
6 It always has been clear that certain host factors affect deposition, and there has been
7 improvement since the 1996 PM AQCD in the understanding of some of these factors,
8 specifically gender, age, and health status. Recent information suggests that there are significant
9 gender differences in the homogeneity of deposition as well as the deposition rate, and this could
10 affect susceptibility. In regard to age, recent evaluations employed both mathematical models as
11 well as experimental studies, and most involved comparison of deposition in children compared
12 to adults. These studies generally indicate that children would receive greater doses of particles
13 per lung surface area than would adults. Unfortunately, deposition studies in another potentially
14 susceptible population, namely the elderly, are still lacking although there have been a number of
15 studies examining effects of chronic pulmonary disease on deposition. These studies confirmed
16 that significant increases in deposition could occur in obstructed lungs.
17 Once deposited on airway surfaces, particles are subjected to translocation and clearance.
18 While the general pathways of clearance have been known for years, recent information has
19 improved the understanding of translocation of particles within size ranges which may be of
20 specific concern for ambient exposures. One such size mode, as noted above, is the ultrafme;
21 and recent studies indicate that ultrafme particles can be rapidly cleared from the lungs into the
22 systemic circulation and reach extrapulmonary organs. This provides a mechanism whereby
23 inhaled particles may affect cardiovascular function, as noted in various epidemiological studies.
24 As with experimental studies, the major improvements in mathematical modeling of
25 dosimetry involve evaluation of realistic size modes for ambient conditions, as well as
26 improvements in the precision of these models for more realistic depictions of respiratory tract
27 airflow patterns and detailed airway structures that may result in deposition "hot spots". These
28 improvements include more detailed evaluations of enhanced deposition at airway bifurcations,
29 use of parameters that allow determination of age differences in dosimetry, and improvement in
30 the modeling of clearance mechanisms.
April 2002 9-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 Thus, in general, while our understanding of specific aspects of particle dosimetry has
2 improved since the 1996 PM AQCD, there are still areas in need of further evaluation. These
3 include dosimetry in susceptible humans, better models for extrapolation between humans and
4 animals used in inhalation studies, and better understanding of differences in the manner in
5 which particles of different and relevant ambient size modes are handled following deposition.
6 This latter research need is important for determining the potential of various particle types to
7 exert effects systemically, rather than just locally within the respiratory tract.
10 9.9 ASSESSMENT OF PARTICULATE-MATTER PROPERTIES LINKED
11 TO HEALTH EFFECTS
12 What is the role ofphysicochemical characteristics ofparticulate matter in eliciting
13 adverse health effects?
14
15 9.9.1 Introduction
16 Ambient PM comprises a complex mix of constituents derived from many sources, both
17 natural and anthropogenic. Hence, the physicochemical composition of PM generally reflects the
18 major contributing sources locally and regionally. Within this framework of source or origin,
19 PM composition also varies significantly by the size-mode within which it is classified (ultrafme,
20 accumulation, or coarse). It should be clear that any given particle can differ appreciably from
21 another individual particle of similar size, but that the region of origin with all of its contributing
22 sources determines the general composition of the generic PM in that classification mode. By its
23 nature then, exposure to airborne ambient PM constitutes an exposure to what is very clearly a
24 mixture of different particles of differing composition and to other gaseous co-pollutants that
25 coexist in that air-shed.
26 The epidemiology information reviewed in the 1996 PM AQCD and updated in this
27 document convincingly shows that a positive correlation exists between the levels of ambient PM
28 pollution and mortality/morbidity. However, this correlation is based mainly on a mass metric,
29 which is somewhat counter-intuitive considering the complexities in composition of PM and
30 given the perceptively low concentrations of most PM constituents, even when fractionated by
31 PM size. What has evolved since the 1996 PM AQCD is the advance in our understanding that
April 2002 9-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 the linkages between PM exposure and health impacts is most strongly related to accumulation
2 mode particles, with combustion-derived PM typically being the most active of the source-based
3 contributors. It is also appreciated that discovery of a "magic bullet" regarding PM
4 physicochemical attributes is not likely to occur, and perhaps the sources from which the PM
5 derive may be the best linkage one can achieve.
6 Approaches to elucidating "causation" and "biological plausibility" have attempted to
7 integrate the wealth of epidemiological data with the growing body of toxicology to reveal
8 coherence among the findings to encourage the pursuit of sound hypotheses. Thus, while it is
9 often difficult to separate the physicochemical attributes of PM that may be of health significance
10 from the mechanisms by which individual factor(s) may function in the response, a number of
11 hypotheses have evolved espousing various PM characteristics as potentially significant
12 contributors to the observed health effects (reviewed by Dreher, 2000). Each of the attribute-
13 based hypotheses has a sufficient data base to merit consideration and further investigation.
14 As the science progresses, it is important that any hypothesis be critically evaluated in the context
15 of the problem, and that the hypothesis provide reasonable responses to at least the following
16 generic, yet pertinent questions (Chapman et al., 1997).
17 • Are there environmental sources that would lead to exposure to PM with the putative
18 constituent(s) or characteristic?
19 • Is there evidence of personal exposure involving PM with that attribute and effect?
20 • Does the putative attribute possess or contribute to a toxic potential?
21 • Is there evidence of an exposure-response relationship, especially at the low
22 concentrations found in the ambient environment?
23 • How well does the hypothesis generalize from one PM sample, exposure, or locale to
24 another?
25 To date, toxicologic studies on PM have provided important, albeit still limited, evidence
26 for specific PM attributes being primarily or essentially responsible for the cardiopulmonary
27 effects linked to ambient PM. In most cases, however, exposure concentrations in laboratory
28 studies have been inordinately high compared to the exposures at which epidemiologic studies
29 have found effects. Reasons for this dosimetric discrepancy range from the limited numbers of
30 animals or human subjects that can be practically studied, the uncertainty and narrow range of
31 responsiveness of the study groups and especially the typically limited use of young, elderly,
April 2002 9-61 DRAFT-DO NOT QUOTE OR CITE
-------
1 unhealthy, or otherwise at-high-risk animals or humans, especially in light of poorly understood
2 risk factors. Thus, most of the toxicology data-base resides in the "Hazard-Identification"
3 compartment of the Risk Assessment paradigm. However, sufficient coherence in the
4 epidemiological and toxicological data has provided a level of "plausibility" to the observational
5 studies and have opened new avenues for investigation to link PM properties and constituents to
6 specific sources and to health outcomes. The primary PM properties thought to be related to
7 health effects are discussed below.
8
9 9.9.2 Specific Properties of Ambient PM Linked to Health Effects
10 9.9.2.1 Physical Properties
11 Acid Aerosols: There is relatively little new information on the effects of acid aerosols,
12 and the basic conclusions of the the 1996 PM AQCD remain unchanged. It previously was
13 concluded that acid aerosols cause little or no change in pulmonary function in healthy subjects,
14 but asthmatics may experience small decrements in pulmonary function. Long-term exposures of
15 animals to acid aerosols, on the other hand, have been shown to alter airway morphology with
16 epithelial cell desquamation and an increase in secretory cells, but these changes have been
17 considered relatively minor. The conclusions about the acute health effects, however, are
18 supported by a recent study by Linn and colleagues (1997), in which healthy children (and
19 children with allergy or asthma) were exposed to sulfuric acid aerosol (100 //g/m3) for 4 hours.
20 While there were no significant effects on symptoms or pulmonary function when the entire
21 group was analyzed, the allergy group did have significant acid-related increases in symptoms,
22 although the acid concentrations were distinctly higher than typical ambient concentrations.
23 These findings were consistent with those reported for adolescent asthmatics exposed to acid
24 aerosols in earlier studies reported in the 1996 PM AQCD.
25 Although pulmonary effects of acid aerosols have been the subject of extensive research,
26 the cardiovascular effects of acid aerosols have received little attention. One example, which
27 raises the issue is a study of acetic acid fumes where reflex mediated increases in blood pressure
28 were found in normal and spontaneously hypertensive rats (Zhang et al., 1997). Similarly, acidic
29 residual oil fly ash (ROFA) PM (which also contains a considerable amount of metal sulfates)
30 was found to alter ecocardiogram (ECG) patterns in the same strain of rats at high air
31 concentrations (Kodavanti et al., 2000). Thus, acidic components should not be entirely
April 2002 9-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 dismissed as possible mediators of ambient PM health effects, since so little is known about
2 potential cardiovascular impacts or impacts in compromised subjects.
3
4 Ultrafme Particles (Size. Surface Area. Number): The physical attributes of PM - size,
5 surface area and number - are intimately interrelated. These properties influence lung deposition,
6 penetrance and persistence in lung tissues, and systemic transport, and, in several studies,
7 apparently the inherent toxicity of the particle itself. While a few epidemiological studies
8 (Wichmann et al., 2000) show correlations between health outcomes and ultrafme (<100 nm)
9 ambient PM, the bulk of the information regarding its toxic potential, and the role of surface
10 area, has derived from studies of surrogate insoluble particles, such as mineral oxides (e.g., TiO2)
11 and carbon black (Oberdorster et al., 1994; Osier and Oberdorster, 1997; Li et al., 1997, 1999).
12 These studies have shown that on an equivalent mass exposure-dose metric, ultrafme PM can
13 induce more acute lung injury than fine PM. Similarly, surrogate PM with high surface areas
14 induced more toxicity than those of like composition, but having smaller surface areas (Lison
15 et al., 1997). On the other hand, studies have shown that composition also matters; for example
16 MgO ultrafmes produce less injury than ZnO (Kuschner et al., 1997), as did sparked carbon
17 versus similarly generated metal oxides (Elder et al., 2000).
18 As with acid aerosols, studies of ultrafme particles have focused largely on effects in the
19 lung, but inhaled ultrafme particles may also have the potential to be distributed systemically and
20 have effects that are independent of lung effects. Recent epidemiological studies evaluating
21 blood viscosity as a biologic correlate of ultrafme exposures, have reported slight increases that
22 raise the prospect of potential cardiovascular implications (Wichmann et al., 2000).
23
24 Fine and Thoracic Coarse Particles: In contrast to ultrafme particles, the respective roles of
25 fine (<2.5 //m) and thoracic coarse (2.5-10 //m) particles in defining health outcomes have
26 garnered considerable research attention because they are the most frequently measured size-
27 fractions of ambient PM and for which most health effects data exist. The fine fraction
28 comprises most of the combustion-related constituents discussed below under chemicals and
29 most readily penetrates deeply into the respiratory tract - at least in terms of a mass metric dose.
30 Naturally, the fine fraction had greater surface area than the thoracic coarse fraction, but much
31 less surface area and particle number than the ultrafme fraction. To the extent that inhaled PM
April 2002 9-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 may carry chemicals or reactive species on their surfaces, these smaller size fractions may have
2 an additional dimension to their toxicity (in terms of surface chemical bioavailablilty) that is not
3 found with coarse PM. For example, acute exposure to sulfate-coated carbon black was found to
4 impair alveolar macrophage phagocytosis and intrapulmonary bactericidal activity in mice (Jakab
5 et al., 1996; Clarke et al., 2000). On the other hand, coarse PM usually is of mineral (earthen) or
6 biologic (discussed below) origin and, thus, has a less complex bioavailable chemical matrix than
7 the finer PM mode. The relative toxicity of most earthen-derived PM has been observed to be
8 less than that of the finer combustion-derived or surrogate ultrafme particles. However, because
9 ambient coarse PM would tend to impact on the airways of humans, it is thought this fraction
10 may be adverse to those with airways sensitivities or disease (e.g., asthma).
11
12 9.9.2.2 Chemical Properties
13 Inorganic Constituents: The inorganic constituents of ambient PM comprise a number of
14 compounds and elements that derive from either natural or combustion sources. The earthen or
15 natural constituents of PM are typically silicates that contain surface and matrix bound metals
16 such as calcium, magnesium, aluminum, and iron. As noted above, most of these silicates do not
17 appear to contribute much toxicity to ambient PM, as considered in this document. Sulfate and
18 nitrate anions derived from combustion or photochemical processes usually complex with other
19 constituents in PM - often more water-soluble ammonium ions or organic acids, as well as
20 elemental cations, such as metals. The intrinsic, independent toxicities of sulfates (as per above)
21 and nitrates appear to be rather low, but they may influence the toxicity or bioavailability of other
22 PM components. Of the cations, metals represent a potential class of causal constituents for
23 PM-associated health effects that have received considerable attention (discussed in more detail
24 below). Sulfate, nitrate, ammonium, and metals make up a substantial part of the mass of
25 ambient PM, often with a silicate or carbonaceous (see below) core, layering, or matrix. The
26 majority of PM-associated metals in fine PM are derived from stationary or mobile combustion
27 sources whereas particle sulfate, nitrate and ammonium originate from secondary atmospheric
28 transformation reactions of involving SO2, NOX and biomass ammonia emissions. Organic PM
29 has both primary and secondary sources.
30
April 2002 9-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 Metals: The 1996 PM AQCD relied on data from occupational exposures to initially
2 evaluate the potential toxicity of metals in PM air pollution. Since that time, in vivo and in vitro
3 studies using ROFA or soluble transition metals have contributed substantial new information on
4 the health effects of PM-associated soluble metals. The metals of most interest, notably the
5 transition metals of iron, vanadium, copper, nickel, chromium, cadmium, arsenic, are ubiquitous
6 constituents of PM-derived from anthropogenic fossil fuel emissions. Exposure seems to be
7 widespread with studies in autopsy specimens (1980's) showing dramatic increases in the content
8 of the first row transition metals in lung tissues of Mexico City residents since the 1950's
9 consistent with industrialization and pollution (Fortoul et al., 1996). Similar studies in North
10 America show metals in the lung tissues of urban dwellers. Although there remain uncertainties
11 about the differential effects of one transition metal versus another, water-soluble or bioavailable
12 metals leached from ROFA or bulk ambient PM cause a variety of biological effects. Many
13 studies show that the action of instilled ROFA and constituent metals are pro-inflammatory
14 (cells, mediators, and molecular signaling processes - in vivo and in vitro), and recently, they
15 have been shown to induce cardiac arrhythmias in animal models (both healthy and diseased).
16 In studies in which various ambient and emission source PM were instilled into rats, the soluble
17 metal content appeared to be the primary determinant of lung injury (Costa and Dreher, 1999).
18 However, these and the related findings on metal toxicity generally have derived from relatively
19 high dose instillation or inhalation exposures, lending them to criticism as to their relevancy for
20 ambient PM that is low in metal content.
21 Nevertheless, a series of studies associated with the closing of a metal smelter in Utah
22 Valley, where ambient PM extracts (containing metals and other soluble constituents) were
23 instilled into the lungs of humans (Ohio and Devlin, 2001) and animals (Dye et al., 2001), as
24 well as tested in vitro (Frampton et al., 1999), showed remarkable coherence with
25 epidemiological studies of hospitalization and mortality (Pope, 1989; Pope et al., 1999b) in the
26 same area and at the same times of the PM samples used in the laboratory studies. The response
27 patterns in each study paralleled the metal content. Furthermore, recent application of novel
28 statistical approaches to the study of source-associated constituents (often metals are the
29 elemental markers) have shown promise in linking sources with their associated emission
30 profiles (including metals) to health outcomes in both humans (Laden et al., 2000) and animals
April 2002 9-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 (Clarke et al., 2000). Thus, while metals appear to be one component involved in PM associated
2 health effects, the full story is incomplete.
3
4 Organic Constituents: Published research on the acute effects of PM-associated organic
5 carbon constituents is conspicuous by its relative absence, except for diesel exhaust particles
6 (DEP). Like metals, organics are common constituents of combustion-generated PM and are
7 found in ambient PM samples over a wide geographical range. Organic carbon constituents
8 comprise a substantial portion of the mass of ambient PM (10 to 60% of the total dry mass
9 [Turpin, 1999]). Although the organic fraction of PM is a poorly characterized heterogeneous
10 mixture of a widely varying number of different compounds, strategies have been proposed for
11 examining the health effects of potentially important organic constituents (Turpin, 1999).
12 In contrast, the mutagenic effects of ambient PM and evidence of DNA-adducts have had more
13 extensive study and have been linked to specific organic fractions (Binkova et al., 1999; Chorqzy
14 et al., 1994; Izzotti et al., 1996). The extent to which organic constituents of ambient PM
15 contribute to adverse health effects identified by current epidemiology studies is not known.
16 Nevertheless, organic constituents remain of concern regarding PM health effects due in large
17 part to the contribution of DEP to the fine PM fraction and the health effects associated with
18 exposure to these particles.
19
20 Diesel Exhaust Particles (DEP): There is growing toxicological evidence that DEP
21 exacerbates the allergic response to inhaled antigens. The organic fraction of diesel exhaust has
22 been linked to eosinophil degranulation and induction of cytokine production suggesting that the
23 organic constituents of DEP are responsible for the immune effects. It is known that the
24 adjuvant-like activity of DEP is not unique, and that certain metals have analogous adjuvant
25 effects (Lambert et al., 2000). It is important to compare the immune effects of other source-
26 specific emissions, as well as concentrated ambient PM, to DEP to determine the extent to which
27 exposure to diesel exhaust may contribute to the incidence and severity of allergic rhinitis and
28 asthma. Other types of noncancer and carcinogenic (especially lung cancer) effects are of
29 concern with regard to DEP exposures, as discussed in a separate EPA Health Assessment
30 Document for Diesel Exhaust (U.S. Environmental Protection Agency, 2002).
31
April 2002 9-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 Biogenic Constituents: Recent studies support the conclusion of the 1996 PM AQCD that
2 bioaerosols, at the concentrations present in the ambient environment, are unlikely to account for
3 the health effects of ambient PM. Dose-response inhalation studies in healthy volunteers
4 exposed to 0.55 and 50 //g endotoxin showed the threshold for pulmonary and systemic effects
5 for endotoxin to be between 0.5 and 5.0 //g (Michel et al., 1997). Urban ambient air PM contains
6 variable amounts of endotoxin, but the levels typically are several orders of magnitude less. The
7 in vitro toxicological studies that have shown endotoxin associated with ambient PM to be pro-
8 inflammatory, inducing cytokine expression in human and rat alveolar macrophages, appear to
9 relate to the endotoxin dose to cell ratio (Becker et al., 1996; Dong et al., 1996). However,
10 endotoxin content does appear to vary by size-mode. Monn and Becker (1999) demonstrated
11 cytokine induction by human monocytes, characteristic of endotoxin activity, in the coarse size
12 fraction of outdoor PM, but not in the fine fraction. Interestingly, while studies in animals
13 models also require more endotoxin than typically found in ambient PM to induce inflammation,
14 recent studies suggest endotoxin may have a priming effect on PM-induced inflammatory
15 processes (Imrich et al., 1999). Thus, the role of biogenic material like endotoxin may have a
16 subtle role that is poorly understood.
17
18 9.9.2.3 Summary
19 Toxicological studies have provided considerable supportive evidence that certain
20 physicochemical particle attributes can provide elements of "causality" to observed health effects
21 of ambient PM. A primary causative attribute may not exist but rather many attributes may
22 contribute to a complex mechanism driven by the nature of a given PM and its contributing
23 sources. The multiple interactions that may occur in eliciting a response in a host may make the
24 identification of any single causal component difficult and may account for the fact that mass as
25 the most basic metric shows the relationships to health outcomes that it does.
26
27 9.9.3 Chemical Components and Source Categories Associated with Health
28 Effects in Epidemiologic Studies
29 Epidemiologic studies using either individual chemical species or classes or using source
30 category factors (SCF) derived from factor analysis have identified a variety of species whose
April 2002 9-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 ambient concentrations are statistically associated with either total mortality or more specific
2 mortality groupings.
3
4 9.9.3.1 Individual Chemical Species
5 Table 9-10 lists the various gaseous co-pollutants, size fractions, chemical element or ions,
6 and organic fractions that have been found to be associated with mortality in regressions using
7 one pollutant species at a time.
TABLE 9-10. CHEMICAL SPECIES ASSOCIATED WITH MORTALITY IN
EPIDEMIOLOGIC STUDIES
Co-Pollutants
CO
NO2
SO2
03
PM Size Fractions
TSP
PM10
PM25
PM10.2.5
PM0,
number
Ions/Elements
SO4=
NO3
Ni
Pb
Carbon/Organic Fractions
TC (Total Carbon)
EC (elemental Carbon)
BC (Black Carbon)
COH (Coefficient of Haze)
OC Organic Carbon)
CX (Cyclohexene-extractable Carbon)
1 9.9.3.2 Source Category Factors
2 There are also three studies in which factor analysis has been used to identify several
3 specific source category factors. In two cases (Laden et al., 2000 and Tsai et al., 2000), the
4 source category factors (SCF) were then used in a multiple regression, the nonsignificant factors
5 were eliminated, and the multiple regression was rerun with only the significant factors. In the
6 third case (Mar et al., 2000), relative risk values are reported for regression with SCF one at a
7 time but the paper states that "Regression analysis with all of the factors included in a
8 multi-source model produced similar results." The similar results in single and multiple
9 regressions and the low correlation between SCF indicates that there is low potential for
10 confounding among the various SCFs.
April 2002 9-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 Source categories that have been found to be significantly associated (p < 0.05) with total,
2 cardiovascular, or cardiovascular plus respiratory mortality in one or more cities are shown in
3 Table 9-11. A source category associated with motor vehicles was found in all four studies. The
4 epidemiological studies do not provide sufficient information to determine whether the causal
5 factor is one or both of the gaseous co-pollutants (CO and NO2); soot particles from cars
6 (indexed by BS, COH, or EC); organic PM from vehicles, transition metals emitted by vehicle
7 (Mn, Fe, Zn); or other particles generated or resuspended by vehicular traffic.
TABLE 9-11. SOURCE CATEGORIES ASSOCIATED WITH MORTALITY IN
EPIDEMIOLOGIC STUDIES
Source Category Tracers
Tsai et al. (2000)
Motor vehicles CO
Fuel Oil Combustion Ni, V
Sulfate S
Industrial Zn, Cd
Laden et al. (2000)
Motor Vehicles Pb
Coal Burning (sulfate) Se, (S)
Mar et al. (2000)
Motor Vehicles CO, NO2; EC, OC; Mn, Fe, Zn, Pb
Vegetative Burning OC, non-soil K
Sulfate S
Ozkaynak et al. (1996)
Motor vehicles CO, COH, NO2
1 The three studies that investigated multiple source categories also found a sulfate factor.
2 The factor reported by Laden et al. (2000) as "coal burning" contains high loadings of both
3 selenium and sulfur and could also have been called "regional sulfate". Mar et al. (2000) refer to
April 2002 9-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 the factor with high sulfate specifically as "regional sulfate". They were able to make this
2 connection because they also had a factor with a high loading of SO2 which they called a "local
3 SO2" factor. The regression with the chemical species S (assumed due to sulfate) was not
4 significant, but the regression with the regional sulfate factor was significant. This may be
5 because the factor analysis will tend to remove other more localized sulfate sources such as
6 CaSO4 and Na2SO4, leaving only acid sulfates ([NH4]2SO4, NH4HSO4, and H2SO4) for a regional
7 sulfate factor. (In Phoenix, there was a modest loading of S in the soil factor.) Therefore, all
8 three sulfate factors should be considered as regional sulfate.
9 The studies of specific chemical components and source categories are especially important
10 because they indicate the association of health effects with the three major components of PM
11 mass: sulfate, nitrate, and organic PM. Examination of PM25 and nitrate effects, alone and in
12 multiple regressions, indicates that PM2 5 and nitrate were not confounded by NO2, CO or O3 in
13 Santa Clara, CA (Fairley, 1999). Examination of the lag structure from the Phoenix study reveals
14 that neither the regional sulfate factor nor the vegetative burning factor was confounded by NO2,
15 CO, SO2, or O3. The epidemiologic results suggest the need for toxicologic studies of the sulfate,
16 nitrate, and organic components of PM, including studies with compromised or susceptible
17 subjects.
18 All of the studies that investigated multiple source categories found a soil or crustal source
19 that was negatively associated with mortality. This suggests that the components of natural soil
20 may have minimal toxicity unless contaminated by anthropogenic sources, such transition metals
21 or polyaromatic hydrocarbons. In any event, the epidemiologic associations suggest additional
22 PM components that should be investigated in toxicologic studies.
23
24
25 9.10 SUSCEPTIBLE SUBPOPULATIONS
26 What subpopulations are at increased risk of adverse health outcomes from particulate
27 matter?
28
29
April 2002 9-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.10.1 Introduction
2 The 1996 PM AQCD identified several population groups potentially being at increased
3 risk for experiencing health impacts of ambient PM exposure. Elderly individuals (>65 years)
4 were most clearly identified, along with those having preexisting cardiovascular or respiratory
5 disease conditions. Smokers and ex-smokers likely comprise a large percentage of individuals
6 with cardiovascular and respiratory disease, e.g., chronic obstructive pulmonary disease (COPD).
7 Individuals with asthma, especially children, also were identified as a potential susceptible
8 population group. The studies appearing since the 1996 PM AQCD provide additional evidence
9 to substantiate the above named groups as likely being at increased risk for ambient PM-related
10 morbidity or mortality effects. There is even evidence, though quite limited at this time, of
11 prenatal effects on cardiac development and potential mortality impacts on infants in the first two
12 years of life.
13 While the identification of susceptible population groups is a critical element of the risk
14 paradigm, characterizing risk factors that underlie susceptibility and that may be common to
15 multiple groups would better substantiate risk estimates and provide better predictability to PM
16 responsiveness. Information relating to these factors, as gleaned from recent epidemiology and
17 toxicology studies, suggests contributing host attributes that may be useful in gaining perspective
18 on their relative public health impact.
19
20 9.10.2 Preexisting Disease as a Risk Factor for Particulate Matter
21 Health Effects
22 The information reviewed in the 1996 PM AQCD is now augmented by numerous new
23 studies which substantiate the finding that preexisting disease conditions represents an important
24 risk factor for ambient PM health effects. Cardiovascular and respiratory diseases continue to
25 appear to be of greatest concern in relation to increasing risk for PM mortality and morbidity.
26 Indeed, the fact that these disease 'entities' often involve both organ systems, albeit to varying
27 degrees, might argue for their compilation under a broader classification of 'cardiopulmonary'
28 disease. Nevertheless, as they are diagnosed and reported separately, Table 9-12 shows the 1996
29 numbers of U.S. cases reported for COPD, asthma, heart disease, and hypertension.
30
31
April 2002 9-71 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-12. INCIDENCE OF SELECTED CARDIORESPIRATORY DISORDERS BY AGE AND
BY GEOGRAPHIC REGION, 1996
to
o
o
to
oo
1
to
o
!>
H
6
O
0
H
O
O
O
o
H
W
(reported as incidence per thousand population and as number of cases in thousands)
Chronic Condition/Disease
COPD*
Incidence/1,000 persons
No. cases x 1,000
Asthma
Incidence/1,000 persons
No. cases x 1,000
Heart Disease
Incidence/1,000 persons
No. cases x 1,000
HD-ischemic
Incidence/ 1,000 persons
No. cases x 1,000
HD -rhythmic
Incidence/ 1,000 persons
No. cases x 1,000
Hypertension
Incidence/1,000 persons
No. cases x 1,000
All Ages
60.4
15,971
55.2
14,596
78.2
20,653
29
7,672
33
8,716
107.1
28,314
Under 45
50.6
9,081
58.9
10,570
33.1
5,934
2.5
453
24.3
4,358
30.1
5,391
Age
45-64
72.3
3,843
48.6
2,581
116.4
6,184
51.6
2,743
40.7
2,164
214.1
11,376
Over 65
95.9
3,047
45.5
1,445
268.7
8,535
140.9
4,476
69.1
2,195
363.5
11,547
Over 75
99.9
1,334
48.0
641
310.7
4,151
154.6
2,065
73.1
977
373.8
4,994
Regional
NE MW S W
57.8 67.6 59.4 56.6
61.8 56.6 51.8 52.9
88.5 78.0 77.0 70.4
28.9 30.0 30.7 25.0
40.2 34.0 28.1 32.9
109.3 108.2 113.5 93.7
'Total chronic bronchitis and emphysema.
Source: Adams etal. (1999).
-------
1 9.10.2.3 Ambient PM Exacerbation of Respiratory Disease Conditions
2 Many time-series studies have shown that pre-existent chronic lung diseases as a group (but
3 especially chronic obstructive pulmonary disease - COPD) constitutes a risk factor for mortality
4 with PM exposure. Studies with humans that might reveal more specific data have been limited
5 both ethically, as well as by the absence of good biomarkers of response (such as ECG's serve
6 cardiac disease). Measures of blood-gas saturation and lung function appear not to be
7 sufficiently revealing or sensitive to mild physiologic changes in those with moderate disease
8 conditions who might be amenable to lab study. In the field, assessing the degree of underlying
9 disease and how that relates to responsiveness of these biomarkers is unclear. However, subjects
10 with COPD and asthma have been studied with inert aerosols for the purpose of assessing
11 distribution of PM within the lung, and it is now quite clear that airways disease leads to very
12 heterogeneous distribution of PM deposited within the lung. Studies have shown up to 10-fold
13 higher than normal deposition at airway bifurcations, thus creating "hot-spots" that may well
14 have biologic implications, especially if the individual already has diminished function or other
15 debility due to the underlying disease, even CVD. Thus the dosimetry of PM within the lung
16 must be considered an important element of the susceptibility paradigm with most any
17 cardiopulmonary disease condition.
18 There are several reports of associations between short-term fluctuations in ambient PM
19 and day to day frequency of respiratory illness. In most cases, notably in children and young
20 people, exacerbation of preexisting respiratory illness and related symptoms has been assessed
21 rather than de novo acute respiratory infections, with asthma apparently an additional risk factor.
22 The use of inhalers has also been shown to increase in many young asthmatics in response to air
23 pollution, with PM often noted as the primary correlate, and as a result school absenteeism
24 increases, again especially in asthmatic children. Interestingly, acute respiratory infections in the
25 elderly with cardiopulmonary disease appears to result in complications of underlying cardiac
26 disorders when PM exposure is involved (Zanobetti et al., 2000), and likewise is linked to
27 subsequent hospitalization. Animal studies with surrogate PM, however, show varied impact on
28 the induction of infection, but in general can alter lung phagocyte functions, which might worsen
29 the condition. Thus, while there appears to a strong likelihood that infections may be worsened
30 by exposure to PM, general statements regarding interaction of PM with response to infectious
April 2002 9-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 agents are difficult given the unique attributes of various infectious agents and the immune status
2 of the host.
3 The underlying biology of lung diseases might also lead to heightened sensitivity to PM
4 (apart from the dose issue noted above), but this attribute of disease remains hypothetical in the
5 context of PM. The functional linkages with the cardiac system for maintenance of adequate gas
6 exchange and fluid balance notwithstanding, the role of inflammation in the diseased respiratory
7 tract (airways and alveoli) could play a key role. Studies in animals genetically or exogenously
8 altered to induce inflammation are sometimes intrinsically more responsive to surrogate or
9 concentrated ambient PM. While a PM-induced response may on the one hand be cumulative
10 with the underlying injury or condition, the responses may, on the other hand, be magnified by
11 any number of mechanisms that are poorly understood. There is sufficient basic biological data
12 to hypothesize that the exudated fluids in the airspaces may either interact differently with
13 deposited PM (e.g., to generate oxidants - Costa and Dreher, 1999; Ohio et al., 2001) to augment
14 injury, or predispose the lung (e.g., sensitize receptors - Undem and Carr, 2002) to enhance the
15 response to a stereotypic PM stimulus through otherwise normal pathways. Less appreciated is
16 the loss of reserve - functional or biochemical - where the susceptible individual is incapable of
17 sufficient compensation (e.g., antioxidant responses - Kodavanti et al., 2000). Any of these or
18 related mechanisms may contribute to "susceptibility" and may indeed be a common factor that
19 can be attributable to other susceptible groups. Understanding these will ultimately aid in
20 addressing true risk of susceptible groups to PM.
21 Again, even a small percentage reduction in PM health impacts on respiratory-related
22 diseases could calculate out to a large number of avoided cases. In 1997, there were 3,475,000
23 U.S. hospital discharges for respiratory diseases: 38% for pneumonia, 14% for asthma, 13% for
24 chronic bronchitis, 8% for acute bronchitis, and the remainder not specified (Lawrence and Hall,
25 1999). Of the 195,943 deaths recorded as caused by respiratory diseases, 44% resulted from
26 acute infections, 10% from emphysema and bronchitis, 2.8% from asthma, and 42% from
27 unspecified COPD (Hoyert et al., 1999).
28
29 9.10.2.4 Ambient PM Exacerbation of Cardiovascular Disease Conditions
30 Exacerbation of cardiovascular (CVD) has been associated epidemiologically, not only
31 with ambient PM, but also with other combustion-related ambient pollutants such as CO. Thus,
April 2002 9-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 while leaving little doubt that ambient PM exposures importantly affect CVD mortality and
2 morbidity, the quantitation of the proportion of risk for such exacerbation specifically attributable
3 to ambient PM exposure is difficult. Recent studies (e.g., concentrated ambient particle studies
4 [CAPS]) have demonstrated cardiovascular effects in response to ambient particle exposures, and
5 studies utilizing animals and other approaches also have produced results suggesting plausible
6 mechanisms leading to cardiovascular effects. However, much remains to be resolved with
7 regard to delineation of dose-response relationships for the induction and extrapolation of such
8 effects to estimate appropriate and effective human equivalent PM (or specific constituent/s)
9 exposures.
10 The recent appreciation for underlying cardiovascular dysfunction as a risk factor for PM
11 health effects derives from a growing and diverse body of literature. While many time-series
12 studies have revealed stronger associations between PM exposures and mortality when a
13 subpopulation was segregated for pre-existent cardiac disease, no direct and plausible evidence
14 had been available. However, recent panel studies of human subjects with CVD (Peters et al.,
15 2000) have shown correlations between air pollution levels, notably PM, and intervention
16 discharge frequency of implanted cardiac defribrillators. Analogously, Pope and colleagues
17 (2001) have noted altered autonomic control of cardiac electrocardiograms (in terms of Heart
18 Rate Variability) over a wide age- range of ostensibly healthy subjects when they were
19 introduced into a room with active smokers. Evidence of vascular narrowing with exposure to
20 concentrated ambient PM (CAPS) has likewise been reported suggesting parallel cardiovascular
21 responses (Brook et al., 2002). Collectively, these and previous studies that have shown ambient
22 PM-induced alterations in cardiac physiology (Pope et al, 1999a,b; Liao et al., 1999; Peters et al.,
23 1999a; Gold et al., 2000) in human subjects, complemented with animal studies (Godleski et al.,
24 1996; Watkinson et al., 1998, 2001; Kodavanti et al., 2000), reinforce the notion of significant
25 cardiac responses to PM. Moreover, indications of changes in plasma viscosity (Peters et al.,
26 1997a) and other factors involved in clotting function (Ohio et al., 2000) provide a plausible
27 cascade of events that could culminate in a sudden cardiac events in some individuals.
28 The HEI report on an epidemiologic study in Montreal, Canada by Goldberg et al. (2000),
29 provides interesting new information regarding types of medical conditions potentially
30 predisposing susceptible individuals to increased risk for PM-associated mortality. It is
31 specifically suggestive that other diseases involving cardiovascular complications could also
April 2002 9-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 contribute to PM risk. First, the immediate causes of death, as listed on death certificates, were
2 evaluated in relation to various ambient PM indices (TSP, PM10 estimated PM2 5, COH, sulfates,
3 and extinction coefficients) lagged for 0 to 4 days. Significant associations were seen between
4 each of the PM measures and total nonaccidental deaths, respiratory diseases, and diabetes, with
5 an approximate 2% increase in excess nonaccidental mortality being observed per 9.5 //g/m3
6 interquartile increase in 3-day mean estimated PM2 5 exposure. When underlying clinical
7 conditions identified in the decedents' medical records were then evaluated in relation to ambient
8 PM measures, all three measures (COH, sulfate, and estimated PM2 5) were associated with acute
9 lower respiratory disease, congestive heart failure, and any cardiovascular disease. Predicted
10 PM25 and COH also were reported to be associated with cancer, chronic coronary artery disease,
11 and any coronary artery disease, whereas sulfate was associated with acute and chronic upper
12 respiratory disease. None of the three PM measures were related to airways disease, acute
13 coronary artery disease, or hypertension. These results both tend to confirm previous findings
14 identifying those with preexisting cardiopulmonary diseases as being at increased risk for
15 ambient PM effects and implicate another possible risk factor, diabetes (which involves
16 cardiovascular complications as it progresses), as a potential susceptibility condition putting
17 individuals at increased risk for ambient PM effects. Zanobetti and Schwartz (2001) have
18 likewise found, perhaps more directly, that those with diabetes are at increased risk, presumably
19 related to the cardiac and vascular complications associated with this disease.
20 To the extent that the observed associations between ambient PM and heart disease
21 exacerbation are causal and specific, the impact on public health could be dramatic. In 1997,
22 there were about 4,188,000 U.S. hospital discharges with heart disease as the first-listed
23 diagnosis (Lawrence and Hall, 1999). Among these, about 2,090,000 (50%) were for ischemic
24 heart disease, 756,000 (18%) for myocardial infarction or heart attack (a subcategory of ischemic
25 heart disease), 957,000 (23%) for congestive heart failure, and 635,000 (15%) for cardiac
26 dysrhythmias. Also, there were 726,974 deaths from heart disease (Hoyert et al., 1999). Thus,
27 even a small percentage reduction in PM-associated admissions or deaths from heart disease
28 would predict a large number of avoided cases.
29
30
April 2002 9-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.10.3 Age-Related At-Risk Population Groups: The Elderly and Children
2 The very young and the very old apparently constitute another group especially affected by
3 PM air pollution. As noted above, a major factor in increased susceptibility to air pollution is the
4 presence of a preexisting illness, as discussed by Zanobetti and Schwartz (2000).
5 The impact of PM pollution is well-documented in time-series studies with mortality risk in
6 studies where age is a factor in the analysis, risk increases above the age of 45 and continues to
7 increase significantly throughout the remainder of life. Cardiopulmonary diseases more common
8 to the elderly play into the risk within older age groups, but panel studies of morbidity focusing
9 on generally healthy people in retirement homes or elderly volunteers exposed to concentrated
10 ambient PM in chambers show subtle alterations of autonomic control of cardiac function (i.e.,
11 slight depression of heart rate variability) and blood factors concordant with a putative response
12 to ambient PM levels. Though small, these changes are considered clinically significant based on
13 studies of risk in cardiac patients and general population studies of cardiac disease progression.
14 Moreover, these changes are in contrast to the lack of similar physiologic changes in healthy
15 young people. Over the long term, innate differences in metabolism or other mechanisms may
16 impact the likelihood of chronic outcomes, e.g., COPD or lung cancer. To what extent
17 progression occurs with repeated PM exposures and how much disease or other risk factors add
18 to or complicate the magnitude of response remains uncertain.
19 Although infection as a risk factor for PM has already been discussed, it is important to
20 emphasize that there are clear age differences in both the incidence and type of infections across
21 age groups. Young children have the highest rates of respiratory illnesses related to infection
22 (notably respiratory synctial virus), while adults are affected by other infectious agents such as
23 influenza that may also lend susceptibility to PM. Data to fully address the importance of these
24 differences is incomplete. The distribution of infectious lung diseases in the U.S. in 1996,
25 summarized in the Table 9-13, provides a good overview of the diversity of this category of
26 preexisting lung disease.
27 In addition to their higher incidences of preexisting respiratory conditions, several other
28 factors may render children and infants more susceptible to PM exposures, including more time
29 spent outdoors, greater activity levels and ventilation, higher doses per body weight and lung
30 surface area, and the potential for irreversible effects on the developing lung. For example, PM
31 doses on a per kilogram body weight basis are much higher for children than for adults as is
April 2002 9-77 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-13. NUMBER OF ACUTE RESPIRATORY CONDITIONS PER
100 PERSONS PER YEAR, BY AGE: UNITED STATES, 1996
45 Years and Over
Type of Acute Condition
Respiratory Conditions
Common Cold
Other Acute Upper Respiratory
Infections
Influenza
Acute Bronchitis
Pneumonia
Other Respiratory Conditions
All
Ages
78.9
23.6
11.3
36.0
4.6
1.8
1.7
Under 5
Years
129.4
48.6
13.1
53.7
*7.2
*3.9
*2.9
5-17
Years
101.5
33.8
15.0
44.3
4.3
*1.7
*2.4
18-24
Years
86.0
23.8
16.1
40.5
*3.9
*1.4
*0.4
25-44
Years
76.9
18.7
11.6
38.1
5.1
*1.3
*2.0
Total
53.3
16.1
7.0
23.3
3.8
*2.0
*1.1
45-64
Years
55.9
16.4
7.5
26.1
3.5
*0.9
*1.5
65 Years
and Over
49.0
15.7
6.1
18.6
*4.4
*3.8
*0.5
Source: Adams etal. (1999).
1 displayed graphically in Figure 9-23. The amount of air inhaled per kilogram body weight
2 decreases dramatically with increasing age, due in part to ventilation differences (in cubic meters
3 per kilogram a day) of a 10-year-old being roughly twice that of a 30-year-old person, even
4 without the consideration of activity level. Child-adult dosage disparities are even greater when
5 viewed on a per lung surface-area basis.
6 As to potential lung developmental impacts of PM, there exist both experimental and
7 epidemiologic data, which although limited, suggest that the early post-neonatal period of lung
8 development is a time of high susceptibility for lung damage by environmental toxicants.
9 In experimental animals, for example, elevated neonatal susceptibility to lung-targeted toxicants
10 has been reported at doses "well below the no-effects level for adults" (Plopper and Fanucchi,
11 2000); and acute injury to the lung during early postnatal development may impair normal repair
12 processes, such as down-regulation of cellular proliferation (Smiley-Jewel et al., 2000, Fanucchi
13 et al., 2000). These results in animals appear concordant with recent findings for young children
14 growing in the Los Angeles area where both oxidants and high PM prevail (Gauderman et al.,
15 2000).
16 These and other types of health effects in children are emerging as potentially more
17 important than appreciated in the 1996 PM AQCD. Unfortunately, relatively little is known
April 2002
9-78
DRAFT-DO NOT QUOTE OR CITE
-------
0.6
0.5
E 0.4
(D
-I— »
as
03
0.3
-^ 0.2
T
0.1
0
I
10
i
20
30
40
i
50
60
i
70
80
Age (y)
Figure 9-23. Inhalation rates on a per body-weight basis for males (•) and females
by age (Layton, 1993).
1 about the relationship of PM to these and other serious health endpoints (low birth weight,
2 preterm birth, neonatal and infant mortality, emergency hospital admissions and mortality in
3 older children). The recent report by Ritz et al. (2002) linking CO exposures of mothers in
4 Los Angeles with fetal cardiac defects raises concerns for PM, which was inconclusively linked
5 in the study. Similarly, little is yet known about the involvement of PM exposure in the
6 progression from less serious childhood conditions, such as asthma and respiratory symptoms, to
7 more serious disease endpoints later in life. Thus, the loss of productive life-years that add to the
8 costs to society may be more than just those indexed by PM-related mortality and/or hospital
9 admissions/visits.
10 In summary, host variability may come to be the most important factor in determining the
11 response profile of any population exposed to PM. Studies to date suggest that certain
12 subpopulations are indeed more acutely responsive to PM, perhaps due to differences in lung
April 2002
9-79
DRAFT-DO NOT QUOTE OR CITE
-------
1 deposition (either in terms of dose and/or intrapulmonary distribution) or other biologic aspects
2 of the cardiopulmonary system or disease thereof. The role of innate attributes of risk grounded
3 in one's genetic code is largely unknown but potentially of great importance. Animal models
4 have been used to show clear differences in response to PM and other pollutants, and the critical
5 involvement of varied genes in the induction of asthma, emphysema, and many other ailments is
6 widely accepted, but poorly understood.
7
8
9 9.11 MECHANISMS OF INJURY
10 What are the mechanisms by which acute exposure to PM causes adverse health effects?
11
12 Numerous epidemiologic studies have shown statistically significant associations between
13 ambient PM levels and a variety of human health endpoints, including mortality, hospital
14 admissions, emergency department visits, respiratory illness, and symptoms measured in
15 community surveys. These associations have been observed with both short and long-term PM
16 exposure. There was little information available in the 1996 PM AQCD to provide biologically
17 plausible mechanisms to support the epidemiologic observations. However, in the intervening
18 years significant progress has been made in identifying pathophysiological effects in humans and
19 animals exposed to various PM that can provide insight into the mechanisms by which PM may
20 exert its effects. Potential mechanisms include neural mechanisms affecting the autonomic
21 nervous system (ANS) via direct pulmonary reflexes or through pulmonary inflammatory
22 processes, direct effects of PM or its components on ion channel function in myocardial cells,
23 ischemic responses of the myocardium, or systemic responses including inflammation that can
24 trigger endothelial cell dysfunction, and thrombosis via alterations in the coagulation cascade.
25 The interactions between these pathways which may lead to sudden cardiac death is shown in the
26 Figure 9-24. However, it must be noted that PM is a complex mixture of many different
27 components and it is possible that different components may stimulate different mechanistic
28 pathways. Thus exposure to PM may result in one or more pathways being activated, depending
29 on the chemical and physical makeup of the PM.
30
April 2002 9-80 DRAFT-DO NOT QUOTE OR CITE
-------
, ,
Pulmonary Reflexes 9 Pulmonary Inflammation
i
Autonomic Nervous
System ~^*C Heart ) •«— Systemic Inflammation
/~^ \
* Endothe ial Cell
Conduction/Repolarization Dysfunction
1 1 1
Heart Rate Cardiac Rhythm Plaque
1 1
Brad yea rdia Thro
Tachycardia *
Ventricular Fibrillation ^V /
Sudden Cardiac
Death
i
^ Platelet
Activation
1
Rupture
1
.
Clotting
F acto rs
Viscosity
Figure 9-24. Schematic representation of potential pathophysiological pathways and
mechanisms by which ambient PM may increase risk of cardiovascular
morbidity and/or mortality.
1 There is now ample evidence that inhaled particles can affect the heart through the ANS.
2 Direct input from the lungs to the ANS via pulmonary afferent fibers can affect both heart rate
3 (HR) and heart rate variability (HRV). The heart is under the constant influence of both
4 sympathetic and parasympathetic innervation from the ANS; and monitoring changes in HR and
5 HRV can provide insight into the balance between those two ANS subdivisions. During recent
6 decades a large clinical database has developed describing a significant relationship between
7 autonomic dysfunction and sudden cardiac death. One measure of this dysfunction, low HRV,
8 has been implicated as a predictor of increased cardiovascular morbidity and mortality. Several
9 independent epidemiologic panel studies of elderly volunteers (some having cardiovascular or
10 pulmonary disease) have reported associations between PM concentrations and various measures
11 of HR and HRV. Although there are some differences among the studies, in general they report
12 an association between PM levels and a reduction in the standard deviation of normal to normal
13 beat intervals (SDNN), a time-domain variable of which the reduction was associated in the
April 2002
9-81
DRAFT-DO NOT QUOTE OR CITE
-------
1 Framingham Heart Study with a higher risk of death. Some studies also reported an association
2 between PM and decreased HRV in the high frequency (HF) range, which is a reflection of
3 parasympathetic modulation of the heart. Other studies have reported a positive association
4 between PM and HR; elevated HR has been associated with hypertension, coronary heart disease,
5 and death. Thus taken as a whole, evidence from panel studies indicates that PM can directly
6 affect the ANS in such as way as to alter heart rate and heart rate variability. However, it should
7 be noted that lowered HRV has primarily been used as a predictor of subsequent increased
8 mortality and morbidity. It is not clear whether a single reversible acute change in HRV places a
9 person more at risk for an immediate adverse cardiac event. Whether changes in HRV associated
10 with exposure to PM represent an independent risk or is just a marker of exposure is not yet
11 known.
12 PM as also been shown to induce changes in conductance and repolarization of the heart as
13 well. Repolarization duration and morphology may reflect subtle changes in myocardial
14 substrate and vulnerability governed by changes in ion channel function. There is considerable
15 evidence linking changes in T wave morphology, QT and T wave variability, T wave Alternans,
16 and changes in ST segment height, to the risk of sudden death. In some studies, rodent models of
17 susceptibility (monocrotaline injected, spontaneously hypertensive) exposed to ROFA showed
18 exacerbated ST segment depression, a factor reflecting T wave morphology during repolarization
19 and which as been useful in diagnosing patients with ischemic heart disease. Healthy dogs
20 exposed to CAPS also showed changes in ST segment elevation; this was exacerbated in dogs
21 with coronary artery occlusion.
22 While PM-induced changes in HRV and HR, as well as changes associated with
23 repolarization and conductance, have the potential to progress to malignant arrhythmias, there is
24 now evidence from both human and animal studies that PM exposure may be linked with severe
25 events directly associated with sudden cardiac death. A recent epidemiology study of patients
26 with implanted cardiac defibrillators reported associations between PM and increased
27 defibrillator discharges. Presumably, some of these patients would have suffered a fatal event
28 had they not had an implanted defibrillator. A second study reported that the risk for myocardial
29 infarction (MI) onset increased in association with PM levels in the 2 hours preceding the MI.
30 PM exposure has also been linked with malignant arrhythmia in some toxicology studies.
31 Healthy rodents exposed to ROFA demonstrated an increase in serious arrhythmic events,
April 2002 9-82 DRAFT-DO NOT QUOTE OR CITE
-------
1 including bradycardia. Rats treated with monocrotaline had significantly exacerbated
2 arrhythmias, and some animals even died within 24 hours following exposure. Older rats,
3 exposed to both ROFA and PM collected from Ottowa, also experienced increased arrhythmias.
4 Dogs exposed to CAPS experienced a slight bradycardia following exposure. Some of these
5 studies involved instillation of a specific PM component (ROFA) at high concentrations, making
6 it uncertain that these observations would hold true using ambient PM at more realistic
7 concentrations. Nevertheless, at least one study used ambient particles collected from Ottowa,
8 and other studies exposed animals by inhalation to CAPS. Taken as a whole, these studies
9 provide convincing evidence that exposure of animals to high levels of PM can affect
10 conductance and repolarization, potentially leading to fatal arrhythmias. However, it remains to
11 be seen if these mechanisms, that can potentially explain acute mortality associated with PM
12 exposure, operate at the lower concentrations of ambient PM to which most people are exposed.
13 Particulate matter could potentially affect the ANS by direct interaction with nerve ending
14 in the lung, or indirectly through the production of inflammatory mediators. Numerous studies
15 have documented that exposure of rodents to ROFA results in substantial lung inflammation and
16 injury. However, due to the levels of ROFA used in many of these studies and the fact that
17 ROFA only makes up a small portion of most airsheds, studies with ambient air particles may be
18 more relevant. There are several studies in which humans, dogs, or rodents have been exposed to
19 CAPS and mild pulmonary inflammation observed. Other studies have shown similar effects
20 when ambient PM collected on filters was used. However, the level of inflammation was quite
21 low in most of these studies, certainly lower than reported in humans or animals exposed to
22 ozone, and it is not yet clear whether lung inflammation plays a role in PM-induced changes in
23 the ANS.
24 In addition to affecting the ANS via the lung, it is also possible that PM or its components
25 could directly attack the myocardium. There is substantial evidence that chronic exposure to
26 fibers encountered in the workplace (e.g., asbestos) result in deposition of fibers in organs other
27 than the lung. Some recent studies have suggested that ultrafme PM may exit the lung and
28 deposit in other organs, including the liver and heart. So far these studies have used sources of
29 particles not naturally found in the air (e.g., silver colloid, latex) so it is not yet clear to what
30 extent PM actually leaves the lung or, if it does, how it interacts directly with the heart.
31 However, there is some evidence of direct changes in the myocardium following PM exposure.
April 2002 9-83 DRAFT-DO NOT QUOTE OR CITE
-------
1 For example, rats exposed to ROFA, which is made up mostly of soluble transition metals, have
2 increased pro-inflammatory cytokine expression in the left ventricle. In another study, dogs
3 living in highly-polluted Mexico City had histopathology changes in heart tissue compared with
4 dogs living in areas with low air pollution. Substantial deposits of particulate matter could be
5 seen throughout the myocardium in the Mexico City dogs. Though preliminary, these
6 observations point to a need for additional work to better define PM-induced changes in
7 myocardial tissue.
8 Acute coronary events frequently occur as a result of thrombus formation in the site of a
9 ruptured atherosclerotic plaque. Increased levels of clotting and coagulation factors, platelet
10 aggregability, and blood viscosity, together with reduced fibrinolytic activity and endothelial cell
11 dysfunction can promote a pro-coagulant state which could potentially contribute to thrombus
12 formation. C reactive protein, a marker of systemic inflammation which correlates with some
13 cardiac events, is positively associated with PM in several panel studies. Some of these studies
14 also report associations between PM and enhanced blood viscosity or increased fibrinogen, a
15 known risk factor for ischemic heart disease. Controlled human and animal exposure studies
16 have also reported that exposure to CAPS (in humans) or ROFA (in animals) results in increased
17 levels of blood fibrinogen. These studies suggest that PM may alter the coagulation pathways in
18 such a way as to trigger cardiovascular events in susceptible individuals.
19 Panel studies have also reported associations between PM and changes in white blood cells,
20 although these findings are not easy to interpret since some studies report positive associations
21 while others report negative associations. Animal studies are similarly unclear, with some
22 studies (rodents exposed to CAPS) reporting increased numbers of blood platelets and white
23 blood cells and others (rodents exposed to ROFA) reporting decreased numbers of white blood
24 cells. In one study, rabbits instilled with colloidal carbon had an increase in neutrophils released
25 from the bone marrow. The same research group found an association between PM and elevated
26 band neutrophil counts (a marker for bone marrow precursor release) in humans exposed to high
27 levels of carbon from biomass burning during the 1997 Southeast Asian smoke-haze episodes.
28 Endothelial cell dysfunction may contribute to myocardial ischemia in some susceptible
29 populations. The vascular endothelium secretes multiple factors that control vascular tone,
30 modulate platelet activity, and influence thrombogenesis. A recent study has reported endothelial
31 cell dysfunction in humans exposed to CAPS, as measured by dilation of the brachial artery.
April 2002 9-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 This vasoconstriction could be caused by an increase in circulating endothelin-1, which has been
2 described in rats exposed to PM.
3 Taken as a whole, these studies are difficult to interpret but clearly indicate that PM can
4 affect the circulatory system. However, a complete understanding of the pathways by which very
5 small concentrations of inhaled ambient PM can produce vascular changes that can contribute to
6 increased mortality/morbidity remains to be more fully elucidated.
7
8
9 9.12 HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER
10 OBSERVED IN POPULATION STUDIES
11 How are exposures to ambient PM quantitatively related to increased risks of health effects
12 (mortality/morbidity) among general human populations and susceptible subgroups ?
13
14 9.12.1 Introduction
15 This section assesses available scientific evidence regarding the physiologic and health
16 effects of exposure to ambient PM as observed in epidemiologic (human population) studies.
17 The main objectives of this evaluation are (1) to summarize and evaluate strengths and
18 limitations of available epidemiologic findings; (2) to summarize quantitative relationships
19 between ambient PM exposures and increased human health risks; (3) to assess the biomedical
20 coherence of findings across studied endpoints; and (4) to note the increased biologic plausibility
21 of the available epidemiologic evidence in light of (a) linkages between specific PM components
22 and health effects and (b) various dosimetric, mechanistic, and pathophysiologic considerations
23 discussed earlier in this chapter.
24 Numerous epidemiologic studies have shown statistically significant associations of
25 ambient PM levels with a variety of human health endpoints, including mortality, hospital
26 admissions, emergency department visits, other medical visits, respiratory illness and symptoms
27 measured in community surveys, and physiologic changes in pulmonary function. Associations
28 have been consistently observed between both short- and long-term PM exposure and these
29 endpoints. The general internal consistency of the epidemiologic database and available findings
30 demonstrate well that notable human health effects are associated with exposures to ambient PM
31 at concentrations currently found in many geographic locations across the United States.
April 2002 9-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 However, many difficulties still exist with regard to delineating the magnitudes and variabilities
2 of risk estimates for ambient PM, the ability to attribute observed health effects to specific PM
3 constituents, the time intervals over which PM health effects are manifested, the extent to which
4 findings in one location can be generalized to other locations, and the nature and magnitude of
5 the overall public health risk imposed by ambient PM exposure.
6 The etiology of most air pollution-related health outcomes is highly multifactorial, and the
7 impact of ambient air pollution exposure on these outcomes is often small in comparison to that
8 of other etiologic factors (e.g., smoking). Also, ambient PM exposure usually is accompanied by
9 exposure to many other pollutants, and PM itself is composed of numerous physical/chemical
10 components. Assessment of the health effects attributable to PM and its constituents within an
11 already-subtle total air pollution effect is difficult even with well-designed studies. Indeed,
12 statistical partitioning of separate pollutant effects may not characterize fully the etiology of
13 effects that actually depend on simultaneous exposure to multiple air pollutants. In this regard,
14 several viewpoints existed at the time of the 1996 PM AQCD regarding how best to interpret the
15 epidemiology data: one saw the PM exposure indicators as surrogate measures of complex
16 ambient air pollution mixtures and the reported PM-related effects as representative of those of
17 the overall mixture; another held that reported PM-related effects are attributable to PM
18 components (per se) of the air pollution mixture and reflect independent PM effects; and a third
19 viewpoint holds that PM can be viewed both as a surrogate indicator, as well as a specific cause
20 of health effects.
21 Several other key issues and problems also must be considered when attempting to interpret
22 the data reviewed in this document. For example, although the epidemiology data provide strong
23 support for the associations mentioned above, questions remain regarding potential underlying
24 mechanisms. Although much progress has been made toward identification of anatomic sites at
25 which particles trigger specific health effects and elucidation of biological mechanisms that
26 underlie induction of such effects, this area of scientific inquiry is still at an early stage. Still,
27 compared to the lack of much solid evidence available in the 1996 PM AQCD, there now is a
28 stronger basis for assessing biologic plausibility of the epidemiologic observations given notable
29 improvement in conceptual formulation of reasonable mechanistic hypotheses and evidence
30 bearing on such hypotheses. New evidence related to several hypotheses was discussed earlier
31 with regard to possible mechanisms by which ambient PM may exert human health effects,
April 2002 9-86 DRAFT-DO NOT QUOTE OR CITE
-------
1 which tends to support the likelihood of a causal relationship between low ambient
2 concentrations of PM and observed increased mortality or morbidity risks. At the same time,
3 much still remains to be done to identify more confidently specific causal agents among typical
4 ambient PM constituents.
5
6 9.12.2 Community-Health Epidemic logic Evidence for Ambient Particulate
7 Matter Effects
8 In recent years, epidemiologic studies showing associations of ambient air pollution
9 exposure with mortality, exacerbation of preexisting illness, and pathophysiologic changes have
10 increased concern about the extent to which exposure to ambient air pollution exacerbates or
11 causes harmful health outcomes at pollutant concentrations now experienced in the United
12 States. The PM epidemiology studies assessed in the 1996 PM AQCD implicated ambient PM
13 as a likely key contributor to mortality and morbidity effects observed epidemiologically to be
14 associated with ambient air pollution exposures. New studies appearing since the 1996 PM
15 AQCD are important in extending results of earlier studies to many more cities and in confirming
16 earlier findings.
17 In epidemiologic studies of ambient air pollution, small positive estimates of air pollutant
18 health effects have been observed quite consistently, frequently being statistically significant at
19 p < 0.05. If ambient air pollution promotes or produces harmful health effects, relatively small
20 effect estimates from current PM concentrations in the United States and many other countries
21 would generally be expected on biological and epidemiologic grounds. Also, magnitudes and
22 significance levels of observed air pollution-related effects estimates would be expected to vary
23 somewhat from place to place, if the observed epidemiologic associations denote actual effects,
24 because (a) not only would the complex mixture of PM vary from place to place, but also
25 (b) affected populations may differ in characteristics that could affect susceptibility to air
26 pollution health effects. Such characteristics include sociodemographic factors, underlying
27 health status, indoor-outdoor activities, diet, medical care access, exposure to risk factors other
28 than ambient air pollution (such as extreme weather conditions), and variations in factors (e.g.,
29 air-conditioning) affecting human exposures to ambient-generated PM.
30 Although it has been argued by some that the observed effects estimates for ambient air
31 pollution are not sufficiently constant across epidemiologic studies and that epidemiologic
April 2002 9-87 DRAFT-DO NOT QUOTE OR CITE
-------
1 studies are trustworthy only if they show relatively large effects estimates (e.g., large relative
2 risks), these arguments have only limited weight in relation to ambient air pollution studies.
3 Also, in any large population exposed to ambient air pollution, even a small relative risk for a
4 widely prevalent health disorder could result in a substantial public health burden attributable to
5 air pollution exposure, as was noted earlier (see Section 9.10).
6 As noted above, small health effects estimates generally have been observed for ambient air
7 pollutants, as would be expected on biological and epidemiologic grounds. In contrast to effects
8 estimates derived for the 1952 London smog episode with relative risk (RR) exceeding 4.0 (i.e.,
9 400% increase over baseline) for extremely high (>2 mg/m3) ambient PM concentrations, effects
10 estimates in most current epidemiology studies at distinctly lower PM concentrations (often
11 < 100 //g/m3) are relatively small. The statistical estimates (1) are more often subject to small
12 (but proportionately large) differences in estimated effects of PM and other pollutants; (2) may
13 be sensitive to a variety of methodological choices; and (3) sometimes may not be statistically
14 significant, reflecting low statistical power of the study design to detect a small but real effect.
15 The ambient atmosphere contains numerous air pollutants, and it is important to continue to
16 recognize that health effects associated statistically with any single pollutant may actually be
17 mediated by multiple components of the complex ambient mix. Specific attribution of effects to
18 any single pollutant may therefore be overly simplistic. Particulate matter is one of many air
19 pollutants derived from combustion sources, including mobile sources. These pollutants include
20 PM, carbon monoxide, sulfur oxides, nitrogen oxides, and ozone, all of which have been
21 considered in various epidemiologic studies to date. Many volatile organic compounds (VOCs)
22 or semivolatile compounds (SVOCs) also emitted by combustion sources or formed in the
23 atmosphere have not yet been systematically considered in relation to noncancer health outcomes
24 usually associated with exposure to criteria air pollutants. In many newly available
25 epidemiologic studies, harmful health outcomes are often associated with multiple combustion-
26 related or mobile-source-related air pollutants, and some investigators have raised the possibility
27 that PM may be a key surrogate or marker for a larger subset of the overall ambient air pollution
28 mix. This possibility takes on added potential significance to the extent that ambient aerosols
29 indeed may not only exert health effects directly attributable to their constituent components, per
30 se, but also serve as carriers for more efficient delivery of water soluble toxic gases (e.g., O3,
31 NO2, SO2) deeper into lung tissue, as noted earlier in Section 9.8.4. This suggests that airborne
April 2002 9-88 DRAFT-DO NOT QUOTE OR CITE
-------
1 particle effects may be enhanced by the presence of other toxic agents or mistakenly attributed to
2 them if their respective concentrations are highly correlated temporally. Thus, although
3 associations of PM with harmful effects continue to be observed consistently across most new
4 studies, the newer findings do not fully resolve issues concerning relative contributions to the
5 observed epidemiologic associations of (a) PM acting alone, (b) PM acting in combination with
6 gaseous co-pollutants, (c) the gaseous pollutants per se, and (d) the overall ambient pollutant
7 mix.
8 It seems likely that, for pollutants whose concentrations are not highly correlated, effects
9 estimates in multipollutant models would be more biologically and epidemiologically sound than
10 those in single-pollutant models, although it is conceivable that single-pollutant models also
11 might be credible if independent biological plausibility evidence supported designation of PM or
12 some other single pollutant as likely being the key toxicant in the ambient pollutant mix
13 evaluated. Because neither of these possibilities have been definitively demonstrated and there is
14 not yet full scientific consensus as to optimal interpretation of modeling outcomes for time
15 series-air pollution studies, the choice of appropriate effects estimates to employ in risk
16 assessments for ambient PM effects remains a difficult issue. Issues related to confounding by
17 co-pollutants, along with issues related to time scales of exposure and response and
18 concentration-response function, still apply to new epidemiologic studies relating concentrations
19 of PM or correlated ambient air pollutants to hospital admissions, exacerbation of respiratory
20 symptoms, and asthma in children, to reduced pulmonary function in children and adults, and to
21 changes in heart rate, and heart rate variability in adults. However, with considerable new
22 experimental evidence now in hand, it is possible to hypothesize various ways in which ambient
23 exposure to PM acting alone or in combination with others could plausibly be involved in the
24 complex chain of biological events leading to harmful health effects in the human population.
25 This newer experimental evidence, coupled with new exposure analyses results, add considerable
26 support for interpreting the epidemiologic findings discussed below as likely being indicative of
27 causal relationships between exposures to ambient PM and consequent associated increased
28 morbidity and mortality risks.
29
30
31
April 2002 9-89 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.12.2.1 Short-Term Particulate Matter Exposure Effects on Mortality
2 This section focuses primarily on discussion of short-term PM exposure effects on
3 mortality, but also highlights some morbidity effects in relation to the mortality findings.
4 Morbidity effects are discussed more fully after discussion of long-term mortality effects in the
5 section following this one.
6
7 Summary of Previous Findings on Short-Term Particulate Matter Exposure-Mortality Effects
8 Time series mortality studies reviewed in the 1996 PM AQCD provided strong evidence
9 that ambient PM air pollution is associated with increased daily mortality. The 1996 PM AQCD
10 summarized about 35 PM-mortality time series studies published between 1988 and 1996. The
11 available information from those studies was consistent with the hypothesis that PM is a causal
12 agent in the mortality impacts of air pollution. The PM10 relative risk estimates derived from the
13 PM10 studies reviewed in the 1996 PM AQCD suggested that an increase of 50 //g/m3 in the 24-h
14 average of PM10 is associated with an increased risk of premature total mortality (total deaths
15 minus accidents and injuries) mainly on of the order of relative risk (RR) = 1.025 to 1.05 (i.e.,
16 2.5 to 5.0% excess risk) in the general population, with statistically significant increases being
17 reported more broadly across the range of 1.5 to 8.5% per 50 //g/m3 PM10. Higher relative risks
18 were indicated for the elderly and for those with preexisting respiratory conditions. Also, based
19 on the then recently published Schwartz et al. (1996) analysis of Harvard Six City data, the 1996
20 PM AQCD found the relative risk for excess total mortality in relation to 24-h fine-particle
21 concentrations to be in the range of RR = 1.026 to 1.055 per 25 Mg/m3 PM25 (i.e., 2.6 to 5.5%
22 excess risk per 25 //g/m3 PM2 5). Relative risk estimates for morbidity and mortality effects
23 associated with standard increments in ambient PM10 concentrations and for fine-particle
24 indicators (e.g., PM25, sulfates, etc.) were presented in Chapters 12 and 13 of the 1996 PM
25 AQCD (see Appendix 9 A), and those effect estimates are updated below in light of the extensive
26 newly available evidence discussed in Chapter 8 of this document.
27 Although numerous studies reported PM-mortality associations, several important issues
28 needed to be addressed in interpreting those relative risks. The 1996 PM AQCD extensively
29 discussed the following critical issues: (1) seasonal confounding and effect modification,
30 (2) confounding by weather, (3) confounding by co-pollutants, (4) measurement error,
April 2002 9-90 DRAFT-DO NOT QUOTE OR CITE
-------
1 (5) functional form and threshold, (6) harvesting and life shortening; and (7) the roles of specific
2 PM components.
3 Season-specific analyses are often not feasible because of small magnitudes of expected
4 effect size or small sample sizes (low power) available for some studies. Some studies had
5 earlier suggested possible season-specific variations in PM coefficients, but it was not clear if
6 these were caused by peak variations in PM effects from season to season, varying extent of PM
7 correlations with other co-pollutants, or weather factors during different seasons. The likelihood
8 of PM effects being accounted for mainly by weather factors was addressed by various methods
9 that controlled for weather variables in most studies (including some involving sophisticated
10 synoptic weather pattern evaluations), and that possibility was found to be very unlikely.
11 Many early PM studies considered at least one co-pollutant in the mortality regression, and
12 an increasing number have examined multiple pollutants. Usually, when PM indices were
13 significant in single-pollutant models, addition of a co-pollutant diminished the PM effect size
14 somewhat, but did not eliminate PM associations. In multiple-pollutant models performed by
15 season, the PM coefficients became less stable, again possibly because of varying correlations of
16 PM with co-pollutants among seasonal or smaller sample sizes. However, in many studies, PM
17 indices showed the highest significance in both single- and multiple-pollutant models. Thus,
18 PM-mortality associations did not appear to be seriously distorted by co-pollutants.
19 Interpretation of the relative significance of each pollutant in mortality regression in
20 relation to its relative causal strength was difficult, however, because of lack of quantitative
21 information on pertinent exposure measurement errors among the air pollutants. Measurement
22 errors can influence the size and significance of air pollution coefficients in time series
23 regression analyses, an issue also important in assessing confounding among multiple pollutants,
24 because the varying extent of such errors among pollutants may influence corresponding relative
25 significance. The 1996 PM AQCD discussed several types of exposure measurement and
26 characterization errors, including site-to-site variability and site-to-person variability. These
27 errors are thought to bias the estimated PM coefficients downward in most cases, but there was
28 insufficient quantitative information available at the time to allow estimation of such bias.
29 The 1996 PM AQCD also reviewed evidence for threshold and various other functional
30 forms of short-term PM mortality associations. Some studies indicated that associations were
31 seen monotonically to even below the PM standards. It was considered difficult, however, to
April 2002 9-91 DRAFT-DO NOT QUOTE OR CITE
-------
1 statistically identify a threshold from available data because of low data density at lower ambient
2 PM concentrations, potential influence of measurement error, and adjustments for other
3 covariates. Thus, use of relative risk (rate ratio) derived from log-linear Poisson models was
4 deemed adequate.
5 The extent of prematurity of death, i.e., mortality displacement (or harvesting) in observed
6 PM-mortality associations has important public health policy implications. At the time of the
7 1996 PM AQCD review, only a few studies had investigated this issue. Although one of the
8 studies suggested that the extent of such prematurity might be only a few days, this may not be
9 generalized because this estimate was obtained for identifiable PM episodes. Insufficient
10 evidence then existed to suggest the extent of prematurity for nonepisodic periods, from which
11 most of the recent PM relative risks were derived.
12 Only a few PM-mortality studies had analyzed fine particles and chemically specific
13 components of PM. The Harvard Six Cities Study (Schwartz et al., 1996) analyzed size-
14 fractionated PM (PM25, PM10/15, and PM10/15.2 5) and PM chemical components (sulfates and H+).
15 The results suggested that PM2 5 was associated most significantly with mortality among the PM
16 components. Although H+ was not significantly associated with mortality in this and earlier
17 analyses, the smaller sample size for H+ than for other PM components made direct comparison
18 difficult. Also, certain respiratory morbidity studies showed associations between hospital
19 admissions and visits with components of PM in the fine-particle range. Thus, the 1996 PM
20 AQCD concluded that there was adequate evidence to suggest that fine particles play especially
21 important roles in observed PM mortality effects.
22 Overall, then, the outcome of assessment of the above key issues in the 1996 PM AQCD
23 can be thusly summarized: (1) observed PM effects are not likely seriously biased by inadequate
24 statistical modeling (e.g., control for seasonality); (2) observed PM effects are not likely
25 significantly confounded by weather; (3) observed PM effects may be confounded or modified to
26 some extent by co-pollutants, and such extent may vary from season to season; (4) determining
27 the extent of confounding and effect modification by co-pollutants requires knowledge of relative
28 exposure measurement/characterization error among pollutants (there was not sufficient
29 information on this); (5) no clear evidence for any threshold for PM-mortality associations was
30 reported (statistically identifying a threshold from existing data also was considered difficult, if
31 not impossible); (6) some limited evidence for harvesting, a few days of life-shortening, was
April 2002 9-92 DRAFT-DO NOT QUOTE OR CITE
-------
1 reported for episodic periods (no study was conducted to investigate harvesting in nonepisodic
2 U.S. data); and (7) only a relatively limited number of studies suggested a causal role of fine
3 particles in PM-mortality associations, but in light of historical data, biological plausibility, and
4 results from morbidity studies, a greater role for fine particles than coarse particles was suggested
5 as being likely.
6
7 Updated Epidemiologic Findings for Short-Term Ambient Particulate Matter
8 Exposure Effects on Mortality
9 With regard to updating the assessment of PM effects in light of new epidemiologic
10 information published since the 1996 PM AQCD, the most salient key points on relationships
11 between short-term PM exposure and mortality (drawn from Chapter 8 discussions in this
12 document) can be summarized as follows.
13 Since the 1996 PM AQCD, there have been more than 80 new time-series PM-mortality
14 analyses, several of which investigated multiple cities using consistent data analytical
15 approaches. With only few exceptions, the estimated mortality relative risks in these studies are
16 generally positive, many are statistically significant, and they generally comport well with
17 previously reported PM-mortality effects estimates delineated in the 1996 PM AQCD. There are
18 also now numerous additional studies demonstrating associations between short-term (24-h) PM
19 exposures and various morbidity endpoints.
20 Several new studies conducted time series analyses in multiple cities. The major advantage
21 of these studies over meta-analyses for multiple "independent" studies is the consistency in data
22 handling and model specifications, thus eliminating variation in results attributable to study
23 design. Also, many of the cities included in these studies were ones for which no earlier time
24 series analyses had been conducted. Therefore, unlike regular meta-analysis, they likely do not
25 suffer from omission of negative studies caused by publication bias. Furthermore, any spatial or
26 geographic variability of air pollution effects can be systematically evaluated in such multi-city
27 analyses.
28
29 PM10 Effect Size Estimates. In the NMMAPS (Samet et al., 2000a,b) analysis of the
30 90 largest U.S. cities, the combined nationwide relative risk estimate was about a 2.3% increase
31 in total mortality per 50-//g/m3 increase in PM10. The NMMAPS effect size estimates did vary
April 2002 9-93 DRAFT-DO NOT QUOTE OR CITE
-------
1 somewhat by U.S. region (see Figures 8-3 and 8-5), with the largest estimate being for the
2 Northeast (4.5% for a 1-day lag, the lag typically showing maximum effect size for most U.S.
3 regions). Various other U.S. multi-city analyses, as well as single-city analyses, obtained PM10
4 effect sizes mainly in the range of 2.5 to 5.0% per 50-//g/m3 increase in PM10. There is some
5 evidence that, if the effects over multiple days are considered, the effect size may be larger.
6 What heterogeneity existed for the estimated PM10 risks across NMMAPS cities could not be
7 explained with the city-specific explanatory variables (e.g., as the mean levels of pollution and
8 weather), mortality rate, sociodemographic variables (e.g., median household income),
9 urbanization, or variables related to measurement error.
10 Original results reported for the multi-city APHEA study showed generally consistent
11 associations between mortality and both SO2 and PM indices in western European cities, but not
12 for central and eastern European cities. More recent studies from APHEA n analyses, however,
13 found analogous increased risks to be associated with PM exposures in central and eastern
14 Europe as in western European cities. The pooled estimate of PM10-mortality relative risks for
15 European cities comport well with estimates derived from U.S. data.
16 Certain other individual-city studies using similar methodology in analyses for each city
17 (but not generating combined overall pooled effect estimates) also report variations in PM effect
18 size estimates between cities and in their robustness to inclusion of gaseous copollutants in
19 multi-pollutant models. Thus, one cannot entirely rule out that real differences may exist in
20 excess risk levels associated with varying size distributions, number, or mass of the chemical
21 constituents of ambient PM; the combined influences of varying co-pollutants present in the
22 ambient air pollution mix from location to location or season to season; or to variations in the
23 relationship between exposure and ambient PM concentration.
24 Nevertheless, there still appears to be reasonably good consistency among the results
25 derived from those several new multi-city studies providing pooled analyses of data combined
26 across multiple cities (thought to yield the most precise effect size estimates). Such analyses
27 indicate the percent excess total (nonaccidental) deaths estimated per 50 //g/m3 increase in 24-h
28 PM10 to be 2.3% in the 90 largest U.S. cities (4.5% in the Northeast region); 3.4% in 10 U.S.
29 cities; 3.5% in the eight largest Canadian cities; and about 2.0% in European cities (using PM10
30 = TSP*0.55). These combined estimates are reasonably consistent with the range of PM10
31 estimates previously reported in the 1996 PM AQCD (i.e., 1.5 to 8.5% per 50 //g/m3 PM10).
April 2002 9-94 DRAFT-DO NOT QUOTE OR CITE
-------
1 These and other excess risk estimates from many other individual-city studies comport well with
2 a number of new studies confirming increased cause-specific cardiovascular- and respiratory-
3 related mortality, and those noted below as showing ambient PM associations with increased
4 cardiovascular and respiratory hospital admissions and medical visits.
5
6 Fine and Coarse Particle Effect Size Estimates. Table 9-14 summarizes effects
7 estimates (RR values) for increased mortality and/or morbidity associated with variable
8 increments in short-term (24-h) exposures to ambient fine particles indexed by various fine PM
9 indicators (PM25, sulfates, H+, etc.) in U.S. and Canadian cities. Table 9-15 shows analogous
10 effect size estimates for inhalable thoracic fraction coarse particles (i.e., PM10_25). In both tables,
11 studies that were highlighted in comparable tables in the 1996 PM AQCD are indicated by
12 italics. For purposes of comparison across studies, results of single-pollutant models are
13 presented in these tables; co-pollutant model results are presented and discussed in more detail in
14 Chapter 8.
15 The effect size estimates derived for PM25 as an ambient fine particle indicator (especially
16 those based on directly measured versus estimated PM2 5 levels) generally appear to fall in the
17 range of 2.0 to 8.5% increase in total (nonaccidental) deaths per 25-//g/m3 increment in 24-h
18 PM25 for U.S. and Canadian cities. Cause-specific effects estimates appear to fall mainly in the
19 range of 3.0 to 7.0% per 25 //g/m3 24-h PM25 for cardiovascular or combined cardiorespiratory
20 mortality and 2.0 to 7.0% per 25 //g/m3 24-h PM2 5 for respiratory mortality in U.S. cities.
21 In the 1996 PM AQCD, there was only one study, the Harvard Six Cities study, in which
22 the relative importance of fine and coarse particles was examined. That study suggested that fine
23 particles, but not coarse particles, were associated with daily mortality. Now, more than
24 10 studies have analyzed both PM2 5 and PM10_25 for their associations with mortality (see
25 Figure 9-25). Although some of these studies (e.g., the Santa Clara County, CA, analysis and the
26 eight largest Canadian cities analysis) suggest that PM2 5 is more important than PM10_2 5 in
27 predicting mortality fluctuations, several others (e.g., the Mexico City and Santiago, Chile
28 studies) seem to suggest that PM10_2 5 may be as important as PM2 5 in certain locations (some
29 shown to date being drier, more arid areas). Seasonal dependence of PM components'
30 associations observed in some of the locations (e.g., higher coarse [PM10_25] fraction estimates for
31 summer than winter in Santiago, Chile) hint at possible contributions of biogenic materials (e.g.,
April 2002 9-95 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14. EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Indicator
RR(±CI)**per25-,wg/m3
PM Increase or 15-^g/m3
SOJ Increase or 75-nmol/m3 H+ increase
Reported PM
Levels Mean
(Min, Max)***
Acute Total Mortality
Six City:A
Portage, WI
Topeka, KS
Boston, MA
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
Overall Six-City Results
Six U.S. Cities8
Santa Clara County, CAC
Buffalo, NYD
Philadelphia, PAE
Detroit, MIF
Phoenix, AZG
Phoenix, AZH
Los Angeles, CA1
San Bernadino and
Riverside Counties, CAJ
Coachella Valley, CAK
Boston, MAL
Three New Jersey Cities :M
Newark, NJ
Camden, NJ
Elizabeth, NJ
Eight Canadian CitiesN
PM25
PM25
PM25
PM2.5
PM25
PM25
PM25
PM25
PM25
so;
PM25
PM25
PM25
PM25
PM25
Est. PM2 5
PM25
PM25
PM25
PM25
PM25
PM25
1.030(0.993, 1.071)
1.020(0.951, 1.092)
1.056(1.038, 1.0711)
1.028(1.010, 1.043)
1.035 (1.005, 1.066)
1.025(0.998, 1.053)
1.015 (1.011, 1.019)
Overall 1.010 (1.028, 1.053)
Mobile 1.087(1.042, 1.134)
Coal 1.028 (1.006, 1.050)
Crustal 0.944 (0.863, 1.032)
1.13 (p< 0.01)
1.034 (1.009, 1.062)
1.042 (p< 0.055)
1.031(0.994, 1.069)
1.030 (1.000, 1.076)
(>25 Mg/m3) 2.868 (1.126, 7.250)
(<25 ^g/m3) 0.779 (0.610, 0.995)
1. 06 (NS, from figure)
1.003 (0.992, 1.015)
1.118(1.013, 1.233)
1.053 (1.018, 1.090)
1.043 (1.028, 1.059)
1.057(1.001, 1.115)
1.018 (0.946, 1.095)
1.030(1.011, 1.050)
11.2 (±7.8)
12. 2 (±7.4)
15.7 (±9.2)
18.7 (±10.5)
20.8 (±9.6)
29.6 (±2 1.9)
Median 14.7
Means 11.3-30.5
13 (2, 105)
61.7(0.78,390.5)
nmol/m3
17.28 (-0.6, 72.6)
18 (6, 86)
13.0 (0, 42)
NR
22 (4, 86)
32.5(9.3, 190.1)
16.8 (5, 48)
15.6 (±9.2)
42.1 (±22.0)
39.9 (±18.0)
37.1 (±19.8)
13. 3 (max 86)
April 2002
9-96
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Toronto, Canada0
Montreal, Canada1"
Indicator
Est. PM2.5
PM25
RR(±CI)**per25-,wg/m3
PM Increase or 15-^g/m3
SO; Increase or 75-nmol/m3 H+ increase
1.048 (1.033, 1.064)
1.058 (1.034, 1.083)
Reported PM
Levels Mean
(Min, Max)***
18.0 (8, 90)
17.4 (2.2, 72.0)
Cause-Specific Mortality
Cardiorespiratorv:
Three New Jersey Cities :M
Newark, NJ
Camden, NJ
Elizabeth, NJ
Total Cardiovascular:
Santa Clara County, CAC
Buffalo, NY)D
Philadelphia, PAF
(seven-county area)
Detroit, MIG
Phoenix, AZH
Los Angeles, CA1
San Bernadino and
Riverside Counties, CA1
Coachella Valley, CAK
Cerebrovascular:
Los Angeles, CA1
Total Respiratory:
Santa Clara County, CAC
Buffalo, NYD
Philadelphia, PAF
(seven-county area)
Detroit, MIG
San Bernadino and
Riverside Counties, CA1
PM25
PM25
PM25
PM25
so;
PM25
PM25
PM25
PM25
Est. PM2 5
PM25
PM25
PM25
so;
PM25
PM25
Est. PM2 5
1.051(1.031, 1.072)
1.062(1.006, 1.121)
1.023 (0.950, 1.101)
1.07 (p> 0.05)
1.040 (0.995, 1.088)
1.028 (p< 0.055)
1.032 (0.977, 1.089)
1.187(1.057, 1.332)
1.027 (1.003, 1.048)
1.007 (0.997, 1.017)
1.086 (0.937, 1.258)
1.036 (0.994, 1.080)
1.13(p>0.05)
1.108(1.007, 1.219)
1.014 (p> 0.055)
1.023 (0.897, 1.166)
1.021 (0.997, 1.045)
42.1 (±22.0)
39.9 (±18.0)
37.1 (±19.8)
13 (2, 105)
61.7(0.78,390.5)
nmol/m3
17.28 (-0.6, 72.6)
18 (6, 86)
13.0 (0, 42)
22 (4, 86)
32.5(9.3, 190.1)
16.8 (5, 48)
22 (4, 86)
13 (2, 105)
61.7(0.78,390.5)
nmol/m3
17.28 (-0.6, 72.6)
18 (6, 86)
32.5(9.3, 190.1)
April 2002
9-97
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES*
Study Location
COPD:
Los Angeles, CA1
Increased Hospitalization
Ontario, Canada®
Ontario, Canada1^
NYC/Buffalo, NYS
Toronto, Canada3
Total Respiratory:
King County, WAT
Toronto, Canada"
Buffalo, NYD
Montreal, Canadav
Montreal, Canadaw
St. John, Canadax
Pneumonia:
Detroit, MIF
Respiratory infections:
Toronto, Canada"
COPD:
Atlanta, GAZ
Detroit, MIF
King County WA^
Los Angeles, CABB
Toronto, CanadaY
Indicator
PM25
so=4
so:
03
so=4
H+ (Nmol/m3)
SO^
PM25
PMj
PM25
SOJ
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
RR(±CI)**per25-,wg/m3
PM Increase or 15-^g/m3
SOJ Increase or 75-nmol/m3 H+ increase
1.027 (0.966, 1.091)
1.03(1.02, 1.04)
1.03(1.02, 1.04)
1.03(1.02, 1.05)
1.05 (1.01, 1.10)
1.16(1.03, 1.30)'
1.12(1.00, 1.24)
1.15 (1.02, 1.28)
1.058(1.011, 1.110)
1.085 (1.034, 1.138)
1.082(1.042, 1.128)
1.239 (1.048, 1.428)
1.137(0.998, 1.266)
1.057(1.006, 1.110)
1.125 (1.037, 1.220)
1.108(1.072, 1.145)
1.124(0.921, 1.372)
1.055 (0.953, 1.168)
1.064(1.009, 1.121)
1.051(1.009, 1.094) (65+ y.o.)
1.04(0.99, 1.09) ((0-19 y.o.)
1.06(1.02, 1.09) (20-64 y.o.)
1.048(0.998, 1.100)
Reported PM
Levels Mean
(Min, Max)***
22 (4, 86)
R = 3. 1-8.2
R = 2. 0-7.7
NR
28.8 (NR/391)
7. 6 (NR, 48.7)
18.6 (NR, 66.0)
NR
16.8(1,66)
61.7(0.78,390.5)
nmol/m3
Summer 93
12.2 (max 31)
18.6 (SD 9.3)
Summer 93
8.5 (max 53.2)
18 (6, 86)
18.0 (max 90)
19.4 (±9.35)
18 (6, 86)
18.1 (3,96)
Median 22 (4, 86)
18.0 (max 90)
April 2002
9-98
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Asthma:
Atlanta, GAZ
Seattle, WACC
Seattle, WADD
Toronto, CanadaY
Total Cardiovascular:
Atlanta, GAZ
Buffalo, NYD
Los Angeles, CAEE
St. John, Canadax
Toronto, Canada"
Ischemic Heart Disease:
Detroit, MIF
Toronto, CanadaY
Dvsrhvthmias:
Atlanta, GAZ
Detroit, MIF
Toronto, CanadaY
Heart Failure:
Detroit, MIF
Toronto, CanadaY
Cerebrovascular:
Los Angeles, CAEE
Toronto, CanadaY
Peripheral circulation diseases:
Toronto, CanadaY
Indicator
PM25
PM25
Est. PM2 5
PM25
PM25
SOI
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
PM25
RR(±CI)**per25-,wg/m3
PM Increase or 15-^g/m3
SOJ Increase or 75-nmol/m3 H+ increase
1.023 (0.852, 1.227)
1.087(1.033, 1.143)
1.445 (1.217, 1.714)
1.064(1.025, 1.106)
1.061 (0.969, 1.162)
1.015 (0.987, 1.043)
(65+) 1.043 (1.025, 1.061)
(<65) 1.035 (1.018, 1.053)
1.151 (0.998, 1.328)
1.072(0.994, 1.156)
1.043 (0.986, 1.104)
1.080(1.054, 1.108)
1.061 (0.874, 1.289)
1.032(0.934, 1.140)
1.061(1.019, 1.104)
1.091 (1.023, 1.162)
1.066(1.025, 1.108)
1.015 (0.992, 1.038)
"NEG" reported
"NEG" reported
Reported PM
Levels Mean
(Min, Max)***
19.4 (±9.35)
16.7 (6, 32)
4.8 (1.2, 32.4)
18.0 (max 90)
19.4 (±9.35)
61.7(0.78,390.5)
nmol/m3
Median 22 (4, 86)
Summer 93
8.5 (max 53.2)
16.8(1,66)
18 (6, 86)
18.0 (max 90)
19.4 (±9.35)
18 (6, 86)
18.0 (max 90)
18 (6, 86)
18.0 (max 90)
Median 22 (4, 86)
18.0 (max 90)
18.0 (max 90)
April 2002
9-99
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Stroke:
Detroit, MIF
Increased Respiratory Symptoms
Southern California1'1'
Six Cities00
(Cough)
Six Cities00
(Lower Resp. Symp.)
Uniontown, PAm
(Evening Cough)
Six Cities ReanalysesKK
(Lower Resp. Symptoms)
(Cough)
Connecticut summer camp11
State College, PAJJ
(Wheeze)
State College, PAJJ
(Cold)
State College, PAJJ
(Cough)
Decreased Lung Function
Uniontown, PAm
Uniontown, PA1^
(Reanalysis)
State College, PA835
(Reanalysis)
Connecticut summer camp11
Indicator
PM25
SO^
PM25
so^
H+
PM25
so:
H+
PM25
PM25
so;
PM21
PM21
PM21
PM25
PM25
PM25
so;
RR(±CI)**per25-Mg/m3
PM Increase or 15-^g/m3
SO; Increase or 75-nmol/m3 H+ increase
1.018 (0.947, 1.095)
Odd Ratio (95% CI) per 25-^g/m3
PM Increase or 15-^g/m3
SO; Increase or 75-nmol/m3 H+ increase
1.48(1.14, 1.91)
1.24(1.00, 1.54)
1.86(0.86, 4.03)
1.19(0.66,2.15)
1.58(1.18, 2.10)
6.82(2.09, 17.35)
1.16(0.10, 13.73)
1.45 (1.07, 1.97)
1.61 (1.20,2.16)
1.28(0.98, 1.67)
1.71(1.30,2.25)
1.59(0.94,2.71)
1.61 (1.21,2.17)
1.48(1.17, 1.88)
PEFR change (L/min) per 25-^g/m3
PM Increase or 15-^g/m3
SO; Increase or 75-nmol/m3 H+ increase
PEFR -1.38 (-2.77, 0.02)
pmPEFR -1.52, (-2.80, -0.24)
pm PEFR -0.93 (-1.88, 0.01)
PEFR -5.4 (-12.3, 1.52)
Reported PM
Levels Mean
(Min, Max)***
18 (6, 86)
R = 2-37
18.0 (max 86.0)
2.5 (max 15.1)
18.1 (max 37 1.1)
nmol/m3
18.0 (max 86.0)
2.5 (max 15.1)
18.1 (max 37 1.1)
nmol/m3
24.5 (max 88.1)
18.0 (max 86.0)
2.5 (max 15.1)
18.1 (max 37 1.1)
nmol/m3
7.0 (1.1,26.7)
23. 5 (max 85. 8)
23. 5 (max 85. 8)
23. 5 (max 85. 8)
24.5 (max 88.1)
24.5 (max 88.1)
23. 5 (max 85. 8)
7.0(1.1,26.7)
April 2002
9-100
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Southwest, VALL
State College, PAJJ
Philadelphia, PAMM
Indicator
PM25
PM21
PM25
RR(±CI)**per25-Mg/m3
PM Increase or 15-^g/m3
SOJ Increase or 75-nmol/m3 H+ increase
amPEFR -1.825 (-3.45, -0.21)
pmPEFR -0.63 (-1.73, 0.44)
amPEFR -3.28 (-6.64, 0.07)
pmPEFR -0.91 (-4.04, 2.21)
Reported PM
Levels Mean
(Min, Max)***
21.62 (3.48, 59.65)
23. 5 (max 85. 8)
22.2 (IQR 16.2)
* Studies highlighted in the 1996 CD are in italics; new studies in plain text. For purposes of comparison across
studies, results of single-pollutant models are presented in these tables; co-pollutant model results are presented and
discussed in more detail in Chapter 8.
**Relative Risk (95% Confidence Interval), except for Fairley (1999) and Lipfert et al. (2000), where insufficient
data were available to calculate confidence intervals so p-value is given in parentheses.
***Min, Max 24-h PM indicator level shown in parentheses unless otherwise noted as (±S.D.), NR = not reported,
or R = range of values from min-max, no mean value reported.
References:
ASchwartz et al. (1996)
BLaden et al. (2000)
GFairley (1999)
DGwynn et al. (2000)
ELipfert et al. (2000a)
FLippmann et al. (2000)
GMaretal. (2000)
HSmith et al. (2000)
'Moolgavkar (2000a)
JOstro (1995)
KOstro et al. (2000)
LSchwartz (2000)
MTsai et al. (2000)
NBurnett et al. (2000)
°Burnett et al. (1998)
FGoldberg et al. (2000)
QBurnett et al. (1994)
RBurnett et al. (1995)
sThurston et al. (1992, 1994)
TLumley and Heagerty (1999)
"Burnett etal. (1997)
vDelfino et al. (1997)
wDelfino et al. (1998)
xStieb et al. (2000)
YBurnett et al. (1999)
zTolbert et al. (2000)
^Moolgavkar et al. (2000)
BBMoolgavkar (2000b)
CGSheppard et al. (1999)
DDNorris et al. (1999)
EEMoolgavkar (2000c)
FFOstroetal. (1993)
GGSchwartz et al. (1994)
HHNeasetal. (1995)
"Thurston et al. (1997)
"Neasetal. (1996)
^Schwartz and Neas (2000)
LLNaeher et al. (1999)
MMNeasetal. (1999)
April 2002
9-101
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-15. EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
CONCENTRATIONS OF COARSE-FRACTION PARTICLES (PM10.2.S)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Indicator
RR (±CI)** per 25-,wg/m3
Increase
Reported PM
Levels Mean
(Min, Max)***
Acute Mortality
Six Cities:A
Portage, WI
Topeka, KS
Boston, MA
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
Overall Six-City Results
Coachella Valley, CAB
Detroit, MIC
Philadelphia, PAD
Phoenix, AZE
Phoenix, AZF
Santa Clara County, CAG
Eight Canadian Cities11
PM10_2,
PM10_25
PM10_25
PM10_2,
PM10_2,
PM10_25
PM10_25
PM10.2.5
PM10.2.5
PM10.2.5
PM10.,5
PM10.2.5
PM10.2.5
PM1(M,
1.013 (0.970, 1.058)
0.968(0.920, 1.015)
1.005 (0.985, 1.030)
1.005 (0.983, 1.028)
1.025 (0.985, 1.066)
1.061 (1.013, 1.111)
1.004(0.999, 1.010)
1.013 (0.994, 1.032)
1.040 (0.988, 1.094)
1.052 (p> 0.055)
1.030 (0.995, 1.066)
(>25 Mg/m3) 1.185 (1.069, 1.314)
(<25 Mg/m3) 1.020 (1.005, 1.035)
1.03 (p>0.05))
1.018 (0.992, 1.044)
6.6 (±6.8)
14.5 (±12.2)
8. 8 (±7.0)
11.9 (±8.5)
11.2 (±7.4)
16.1 (±13.0)
Median 9.0
17.9 (0, 149)
13 (4, 50)
6.80 (-20.0, 28.3)
33.5 (5, 187)
NR
11(0,45)
12.9 (max 99)
Cause-Specific Mortality
Total Cardiovascular:
Coachella Valley, CAB
Detroit, MIC
Philadelphia, PAD
(seven-county area)
Phoenix, AZE
Santa Clara County, CAG
Total Respiratory:
Coachella Valley, CAB
Detroit, MID
Philadelphia, PAD
(seven-county area)
Santa Clara County, CAG
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.,5
PM,^<
1.026 (1.006, 1.045)
1.078(1.000, 1.162)
1.034 (p> 0.055)
1.064(1.014, 1.117)
1.03 (p> 0.05)
1.026 (1.006, 1.045)
1.074 (0.910, 1.269)
1.030 (p> 0.055)
1.16(p>0.05)
17.9 (0, 149)
13 (4, 50)
6.80 (-20.0, 28.3)
33.5 (5, 187)
11(0,45)
17.9 (0, 149)
13 (4, 50)
6.80 (-20.0, 28.3)
11 (0,45)
April 2002
9-102
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-15 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF COARSE-FRACTION PARTICLES (PM10.2.S)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Indicator
RR(±CI)**per25-,ag/m3
Increase
Reported PM
Levels Mean
(Min, Max)***
Increased Hospitalization
Total Respiratory:
Toronto, Canada1
Pneumonia:
Detroit, MIC
Respiratory infections:
Toronto, Canada1
COPD:
Atlanta, GAK
Detroit, MIC
Los Angeles'3
Toronto, Canada1
Total Cardiovascular:
Atlanta, GAK
Toronto, Canada1
Ischemic Heart Disease:
Detroit, MIC
Toronto, Canada1
Dysrhythmias:
Detroit, MIC
Atlanta, GAK
Toronto, Canada1
Heart Failure:
Detroit, MIC
Toronto, Canada1
Stroke:
Detroit, MIC
Cerebrovascular:
Toronto, Canada1
PM10.,5
PM10.2.5
PM10.,5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
1.125 (1.052, 1.20)
1.119(1.006, 1.244)
1.093(1.046, 1.142)
0.770 (0.493, 1.202)
1.093 (0.958, 1.247)
1.17(1.09, 1.26) (0-19 y.o.)
1.09(1.03, 1.15) (20-64 y.o.)
1.05(0.99, 1.11) (65+ y.o.)
1.128(1.049, 1.213)
1.176(0.954, 1.450)
1.205(1.082, 1.341)
1.105(1.027, 1.189
1.037 (1.013, 1.062))
1.002(0.877, 1.144)
1.532(1.021,2.30)
1.051(0.998, 1.108)
1.052(0.967, 1.144)
1.079(1.023, 1.138)
1.049(0.953, 1.155)
"NEG" reported
11.6(1,56)
13 (4, 50)
12.2 (max 68)
9.39 (±4.52)
13 (4, 50)
NR
12.2 (max 68)
9.39 (±4.52)
11.6(1,56)
13 (4, 50)
12.2 (max 68)
13 (4, 50)
9.39 (±4.52)
12.2 (max 68)
13 (4, 50)
12.2 (max 68)
13 (4, 50)
12.2 (max 68)
April 2002
9-103
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-15 (cont'd). EFFECT ESTIMATES PER VARIABLE INCREMENTS IN
24-HOUR CONCENTRATIONS OF COARSE-FRACTION PARTICLES (PM10.2.S)
FROM U.S. AND CANADIAN STUDIES*
Study Location
Indicator
RR(±CI)**per25-,ag/m3
Increase
Reported PM
Levels Mean
(Min, Max)***
Peripheral Circulation Diseases:
Toronto, Canada1 PM10_2.5
Asthma:
1.056(1.003, 1.112)
12.2 (max 68)
Seattle, WAL
Toronto, Canada1
Increased Respiratory Symptoms
Six U.S. CitiesM
(Lower Respiratory
Symptoms)
Six U.S. CitiesM
(Cough)
Southwest VirginiaN
(Runny or Stuffy Nose)
Decreased Lung Function
Southwest Virginia0
Uniontown, PAM
(Reanalysis)
State College, PAM
(Reanalysis)
Philadelphia, PAP
PM10.2.5
PM10.2.5
PM10.2.5
PM10.,5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.2.5
PM10.25
1.111(1.028, 1.201)
1.111 (1.058, 1.166)
Odds Ratio (95% CI) per 25-,wg/m3
PM Increase
1.51(0.66,3.43)
1.77(1.24,2.55)
2.62(1.16,5.87)
PEFR change (L/min) per 25-^g/m3
PM Increase
am PEFR 5. 3 (2.6, 8.0)
pm PEFR +1.73 (5.67, -2.2)
pm PEFR -0.28 (2.86, -3.45)
am PEFR -4.31 (-11.44, 2.75)
16.2 (6, 29)
12.2 (max 68)
NR
NR
NR
27.07 (4.89, 69.07)
NR
NR
9.5(IQR5.1)
* Studies highlighted in the 1996 CD are in italics; new studies in plain text. For purposes of comparison across
studies, results of single-pollutant models are presented in these tables; co-pollutant model results are presented
and discussed in more detail in Chapter 8.
** Relative Risk (95% Confidence Interval), except for Fairley (1999) and Lipfert et al. (2000), where insufficient
data were available to calculate confidence intervals so p-value is given in parentheses.
*** Min, Max 24-h PM indicator level shown in parentheses unless otherwise noted as (±S.D.), NR = not reported,
or R = range of values from min-max, no mean value reported.
References:
ASchwartz et al. (1996)
BOstro et al. (2000)
cLippmann et al. (2000)
DLipfert et al. (2000a)
EMar et al. (2000)
FSmith et al. (2000)
GFairley (1999)
HBurnett et al. (2000)
'Burnett etal. (1997)
'Burnett etal. (1999)
KTolbert et al. (2000)
LSheppard et al. (1999)
MSchwartz and Neas (2000)
NNaeheretal. (1999)
°Zhang et al. (2000)
FNeasetal. (1999)
QMoolgavkar (2000b)
April 2002
9-104
DRAFT-DO NOT QUOTE OR CITE
-------
>
to
o
o
to
H
6
o
o
H
O
Percent excess death (total unless otherwise noted) per
25 ug/m3 increase in PM2.5 (•) or PM10.2.5 (o).
Harvard 6 Cities (recomputed)
Steubenville, OH
8 Canadian Cities
Chock et al (2000)
Pittsburgh, PA
Klemm and Mason (2000)
Atlanta, GA
Lipfert et al (2000a)
Philadelphia, PA ~
Lippmann et al (2000)
Detroit, Ml
Santa Clara Co.
Ostro et al. (2000)
Coachella Valley, CA
Castillejos et al (2000)
Mexico City, Mexico
Cifuentes et al. (2000)
5-4-3-2-10 1 234 56 7 8 9 10 11 12 13 14 1
i i i i i i i i i i i i i i i i i i i i
g / 0
w — ,
• } aye .* * 5
'-+•
n total "
v mortality
_
Laq 5 day MA 1 • Q
V Q } All year
Laa 2 dav MA } ^ — } Winter
o
HH
H
W
Figure 9-25. Percent excess risks estimated per 25-jUg/m3 increase in PM2 5 or PM10_2 5 from new studies evaluating both
PM2 5 and PM10_2 5 data for multiple years. All lags = 1 day, unless indicated otherwise.
-------
1 molds, endotoxins, etc.) to the observed coarse particle effects in at least some locations.
2 Overall, for U.S. and Canadian cities, effect size estimates for the coarse fraction (PM10_25) of
3 those inhalable thoracic particles capable of depositing in TB and A regions of the respiratory
4 tract generally appear to fall in the range of 0.5 to 6.0% excess total (nonaccidental) deaths per
5 25 //g/m3 of 24-h PM10_25. Respective increases for cause-specific mortality are 3.0 to 8.0% for
6 cardiovascular and 3.0 to 16.0% for respiratory causes per 25-//g/m3 increase in 24-h PM10_25.
7
8 Chemical Components of Particulate Matter. Several new studies examined the role of
9 specific chemical components of PM in relation to mortality risks. Studies of U.S. and Canadian
10 cities showed mortality associations with one or more of several specific fine particle
11 components of PM, including H+, sulfate, nitrate, as well as COH; but their relative importance
12 varied from city to city, likely depending, in part, on their concentrations (e.g., no clear
13 associations in those cities where H+ and sulfate levels were very low [i.e., circa nondetection
14 limits]). Figure 9-26 depicts relatively consistent estimates of total mortality excess risk
15 resulting from a 5-//g/m3 increase in sulfate, possibly reflecting impacts of sulfate per se or
16 perhaps sulfate serving as a surrogate for fine particles in general. Sulfate effect size estimates
17 generally fall in the range of 1 to 4% excess total mortality per 5-//g/m3 increase for U.S. and
18 Canadian cities.
19 A significant factor in some western cities is the occasional occurrence of high levels of
20 windblown crustal particles that constitute the major part of the coarse PM fraction and a
21 substantial fraction of intermodal fine particles (PM^.j). The small-size tail of the windblown
22 crustal particles extends into the PM2 54 size range (intermodal), at times contributing
23 significantly to PM2 5. Claiborn et al. (2000) report that in Spokane, WA, PM2 5 constitutes about
24 30% of PM10 on dust event days, but 48% on days preceding the dust event. The intermodal
25 fraction represents about 51% of PM2 5 during windblown dust events, about 28% on preceding
26 days. However, PMX in Spokane often shows little change during dust events, when coarse
27 particles (presumably crustal particles) are transported into the region. The lack of increased
28 mortality during periods of time with high wind speeds and presumably high crustal material
29 concentrations was shown by Schwartz et al. (1999) for Spokane, and by Pope et al. (1999b) for
30 three cities in the Wasatch front region of Utah. Other recent studies suggest that coarse
31 particles, as well as fine particles, may be associated with excess mortality in certain U.S.
April 2002 9-106 DRAFT-DO NOT QUOTE OR CITE
-------
Percent excess death (total mortality, unless otherwise noted)
per 5 |jg/m3 increase in sulfate
Schwartz et al. (1996)
Six Cities
Burnett etal. (1998)
Toronto, Canada
Burnett et al. (2000)
Q 1 -jrnea
-------
TABLE 9-16. SUMMARY OF SOURCE-ORIENTED EVALUATIONS OF
PARTICULATE MATTER COMPONENTS IN RECENT STUDIES
Author, City
Source Categories and Species with High Factor
Loadings Used to Suggest the Source Categories
Source Categories Associated with
Mortality. Comments.
Laden et. al.,(2000)
Harvard Six Cities
1979-1988
Soil and crustal material: Si
Motor vehicle emissions: Pb
Coal combustion (Regional Sulfate): Se, S
Fuel oil combustion: V
Salt: Cl
Note: the trace elements are from PM2 5 samples
The strongest increase in daily mortality
was associated with the mobile source
factor. The coal combustion factor was
positively associated with mortality in all
metropolitan areas, with the exception of
Topeka. The crustal factor from the fine
particles was not associated with
mortality.
Coal and mobile sources account for the
majority of fine particles in each city.
Mar et al. (2000).
Phoenix, AZ
1995-1997
PM25 (from DFPSS) trace elements:
Motor vehicle emissions and resuspended road dust:
Mn, Fe, Zn, Pb, OC, EC, CO, and NO2
Soil: Al, Si, and Fe
Vegetative burning: OC and Ks (soil-corrected
potassium)
Local SO, sources: SO2
Regional sulfate: S
PM10_25 (from dichot) trace elements:
Soil: Al, Si, K, Ca, Mn, Fe, Sr, and Rb
A source of coarse fraction metals: Zn, Pb, and Cu
A marine influence: Cl
PM, s factors results: Soil factor and local
SO2 factor were negatively associated with
total mortality. Regional sulfate was
positively associated with total mortality
on the same day, but negatively associated
on the lag 3 day. Motor vehicle factor,
vegetative burning factor, and regional
sulfate factor were significantly positively
associated with cardiovascular mortality.
Factors from dichot PM10_2 5 trace elements
were not analyzed for their associations
with mortality because of the small sample
size (every-third-day samples from June
1996).
Ozkaynak et al.
(1996).
Toronto, Canada.
Motor vehicle emissions: CO, COH, and NO2
Motor vehicle factor was a significant
predictor for total, cancer, cardiovascular,
respiratory, and pneumonia deaths.
Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ.
1981-1983.
Motor vehicle emissions: Pb and CO
Geological (Soil): Mn and Fe
Oil burning: V and Ni
Industrial: Zn, Cu, and Cd (separately)
Sulfate/secondary aerosol: Sulfate
Note: The trace elements are from PM15 samples.
Oil burning, industry, secondary aerosol,
and motor vehicle factors were associated
with mortality.
1 not positively associated with total mortality, with Mar et al. (2000) reporting a negative
2 association between the crustal component of PM2 5 and cardiovascular mortality.
3 However, these source-category-oriented evaluation results are derived from relatively
4 limited underlying analytic bases for resolving source categories and the identification of souce
April 2002
9-108
DRAFT-DO NOT QUOTE OR CITE
-------
1 categories must be viewed with caution at this time. For example, whereas Laden et al. (2000)
2 had 6211 days of every-other-day data from the Harvard Six City Study of eastern/midwest U.S.
3 cities, they had only elements in PM2 5 analyzed by X-ray fluorescence (XRF) spectroscopy (no
4 organic PM or gases). They used factors in the regression analysis and used Pb as a tracer to
5 identify a motor vehicle source category, Se to identify a coal combustion source category, and Si
6 as a tracer for soil. Since the "coal combustion" factor had a high loading of S as well as Se, it
7 could equally as well have been identified as the regional sulfate source category. The "motor
8 vehicle" and "coal combustion" sources were statistically significant for total mortality as well as
9 mortality resulting from ischemic heart disease and respiratory diseases (COPD plus pneumonia).
10 The crustal component had a negative association with total mortality.
11 The Mar et al. (2000) study had 3 years of pollutant data for Phoenix, AZ. In addition to
12 elements determined by XRF, they had pollutant gases (CO, NO2, SO2, and O3) and total,
13 organic, and elemental carbon. They were able to identify five factors and attributed them to five
14 source categories. Motor vehicles (plus resuspended road dust), vegetative burning, and regional
15 sulfate all had statistically significant associations with cardiovascular mortality, but soil
16 (indexed by Si and Al, as crustal markers) had a statistically significant negative association.
17 Also of importance, Mar et al. (2000) found significant associations between cardiovascular
18 mortality and PM25 and marginally significant (p < 0.10) associations between total mortality and
19 PM10.2.5.
20 Tsai et al. (2000) had only 156 days of data and used measurements of CO, sulfate, and
21 some elements; and they did not have Si, Ca, Al, or Mg as soil tracers nor Se as a tracer of coal
22 combustion, although much of the sulfate probably came from coal combustion. They had three
23 fractions of extractable organic matter, but these did not appear to be useful in determining
24 factors. Statistically significant (p > 0.05) factors for both total daily deaths and combined
25 cardiovascular and respiratory daily deaths in at least one or another of the three New Jersey
26 cities studied (Newark, Camden, and Elizabeth) were attributed to motor vehicles, oil burning,
27 and sulfate. Also, an industrial source containing Zn and Cd was statistically significant for total
28 deaths in Newark; and an industrial source containing Cd was marginally statistically significant
29 for cardiorespiratory disease in Elizabeth.
April 2002 9-109 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ozkaynak et al. (1996) had only TSP, coefficient of haze (COH), and gases; however, they
2 reported that a factor with COH, CO, and NO2 (considered to be representative of motor vehicle
3 emissions) was associated with mortality in Toronto, Canada.
4 None of these studies had measurements of nitrate or semivolatile organic compounds nor
5 did they use the newest, and most effective, techniques for source apportionment. For example,
6 using positive matrix factorization, Ramadan et al. (2000) were able to determine eight factors
7 using the same data set as Mar et al. (2000). In spite of these deficiencies, all four studies were
8 able to associate one or more types of mortality with motor vehicles, several with coal
9 combustion, and three with sulfate.
10 Factor analyses also were described briefly in a report by Lippmann et al. (2000). In that
11 study, neither sulfate nor acid aerosols were related significantly to morbidity or mortality, but
12 the concentrations were extremely low (with about 70% of the acid measurements below
13 detection limit).
14 It is difficult to compare these source-categories-related assessments. They are based on
15 different regions of the country over different periods of time when the sources of particles,
16 marker elements such as Pb, and other urban air pollutants were changing greatly. Also, each of
17 these studies constructed factors based on city-specific data. Thus, the factors in each study are
18 based on the idiosyncrasies of the specific data set for each city in the study, so the factors may
19 indeed represent different sources in different locations. Nevertheless, although somewhat
20 limited at this time, the new factor analysis results appear to implicate ambient PM derived from
21 fossil fuel (oil, coal) combustion and vegetative burning, as well as secondarily formed sulfates,
22 as important contributors to observed mortality effects, but not crustal particles.
23 In summary, the new evidence suggests that exposure to particles from several different
24 source categories, and of different composition and size, may have independent associations with
25 health outcomes. The excess risks from different types of combustion sources (coal, oil,
26 gasoline, wood, and vegetation) may vary from place to place and from time to time, so that
27 substantial intra-regional and inter-regional heterogeneity would be expected. Likewise,
28 although earlier evaluations in the 1996 PM AQCD seemed to indicate coarse particles and
29 intermodal particles of crustal composition as not likely being associated with adverse health
30 effects, there are now some reasonably credible studies suggesting that coarse particles (although
31 not necessarily those of crustal composition) may be associated with excess mortality in at least
April 2002 9-110 DRAFT-DO NOT QUOTE OR CITE
-------
1 some locations. These notably include areas where past deposition of fine PM metals from
2 smelter (Phoenix) or steel mills (Steubenville) onto surrounding soils may result in enhanced
3 toxicity of later resuspended coarse (PM10_25) particles.
4
5 Updated Epidemiologic Findings for Long-Term Particulate Matter Exposure
6 Effects on Mortality
7 The 1996 PM AQCD indicated that past epidemiologic studies of chronic PM exposures
8 collectively indicate increases in mortality to be associated with long-term exposure to airborne
9 particles of ambient origins (see appendix Table 9A-3). The PM effect size estimates for total
10 mortality from these studies also indicated that a substantial portion of these deaths reflected
11 cumulative PM impacts above and beyond those exerted by acute exposure events. Table 9-17
12 shows long-term exposure effects estimates (RR values) per variable increments in ambient PM
13 indicators in U.S. and Canadian cities, including results from newer analyses since the 1996 PM
14 AQCD.
15 One of the most important advances since the 1996 PM AQCD is the substantial
16 verification and extension of the findings of the Six City prospective cohort study (Dockery
17 et al., 1993) and the cohort study relating American Cancer Society (ACS) health data to
18 fine-particle data from 50 cities and sulfate data from 151 cities (Pope et al., 1995). The
19 reanalyses, sponsored by the Health Effects Institute (HEI), included a data audit, replication of
20 the original investigators' findings, and additional analyses to explore the sensitivity of the
21 original findings to other model specifications. The investigators of the HEI Reanalysis Project
22 (Krewski et al., 2000) first performed a data audit, using random samples to verify the accuracy
23 of the data sets used in the original Six City analyses, including death certificate data, air
24 pollution data, and socioeconomic data. In general, the air pollution data were reproducible and
25 correlated highly with the original aerometric data in Pope et al. (1995).
26 The reanalyses substantially verified the findings of the original investigators, with PM25 or
27 sulfate relative risk (RR) estimates for total mortality and for cardiopulmonary mortality differing
28 at most by ±0.02 (±2% excess risk) from the least polluted to the most polluted cities in the
29 study. A larger difference was noted for the PM2 5 lung cancer relative risk in the Six Cities
30 study, 1.37 originally and 1.43 in the reanalysis, neither estimate being statistically significant.
31 The sensitivity analyses for the Six Cities study found generally similar results with other
April 2002 9-111 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-17. EFFECT ESTIMATES PER INCREMENTSA IN LONG-TERM MEAN
LEVELS OF FINE AND INHALABLE PARTICLE INDICATORS FROM U.S. AND
CANADIAN STUDIES
Type of Health
Effect and Location
Increased Total Mortality
Six City11
ACS Study0
(151 U.S. SMSA)
Six City ReanalysisD
ACS Study ReanalysisD
ACS Study Extended
Analyses'3
Southern CaliforniaE
Indicator
in Adults
PM15/10(20^g/m3)
PM25(10^g/m3)
SO=4 (15 /ug/m3)
PM25(10^g/m3)
SO, (15 /ug/m3)
PM15/10 (20 Mg/m3)
PM25 (10 Mg/m3)
PM15/10 (20 ^g/m3)
(SSI)
PM25 (10 Mg/rn3)
PM25 (10 Mg/rn3)
PM10 (50 Mg/m3)
PM10 (cutoff =
30 days/year
>100 Mg/m3)
PM10 (50,wg/m3)
PM10 (cutoff =
30 days/year
>100 Mg/m3)
Increased Bronchitis in Children
Six at/
Six City0
24 City"
24 City"
24 City"
24 City"
Southern California1
12 Southern California
communities1
(all children)
12 Southern California
communitiesK
(children with asthma)
PM15/10(50^g/m3)
TSP (100 ^g/m3)
H+ (WOnmol/m3)
S0=4 (15 ^g/m3)
PM21 (25 ^g/m3)
PM10(50^g/m3)
S0=4(15^ig/m3)
PM10 (25 Mg/m3)
Acid vapor (1.7 ppb)
PM10 (19 Mg/m3)
PM25 (15 Mg/m3)
Acid vapor (1.8 ppb)
Change in Health Indicator per
Increment in PMa
Relative Risk (95% CI)
1.18(1.06-1.32)
1.13 (1.04-1.23)
1.46(1.16-2.16)
1.07 (1.04-1.10)
1.10(1.06-1.16)
1.19(1.06-1.34)
1.13 (1.04-1.23)
1.02 (0.99-1.04)
1.07(1.04-1.10)
1.04(1.01-1.08)
1.242 (0.955-1.616) (males)
1.082 (1.008-1. 162) (males)
0.879 (0.713-1.085) (females)
0.958 (0.899-1.021) (females)
Odds Ratio (95% CI)
3.26(1.13, 10.28)
2.80(1.17, 7.03)
2.65 (1.22, 5.74)
3.02(1.28, 7.03)
1.97 (0.85, 4.51)
3.29(0.81, 13.62)
1.39 (0.99, 1.92)
0.94(0.74, 1.19)
1.16(0.79, 1.68)
1.4(1.1, 1.8)
1.4(0.9,2.3)
1.1 (0.7, 1.6)
Range of City
PM Levels *
Means (^g/m3)
18-47
11-30
5-13
9-34
4-24
18.2-46.5
11.0-29.6
58.7(34-101)
9.0-33.4
21.1 (SD=4.6)
51 (±17)
51 (±17)
20-59
39-114
6.2-41.0
18.1-67.3
9.1-17.3
22.0-28.6
—
28.0-84.9
0.9-3.2 ppb
13.0-70.7
6.7-31.5
1.0-5.0 ppb
April 2002
9-112
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-17 (cont'd). EFFECT ESTIMATES PER INCREMENTSA IN LONG-TERM
MEAN LEVELS OF FINE AND INHALABLE PARTICLE INDICATORS FROM U.S.
AND CANADIAN STUDIES
Type of Health
Effect and Location
Increased Cough in Children
12 Southern California
communities'
(all children)
12 Southern California
communitiesK
(children with asthma)
Indicator
PM10 (25 Mg/m3)
Acid vapor (1.7 ppb)
PM10 (19 Mg/rn3)
PM25 (15 Mg/rn3)
Acid vapor (1.8 ppb)
Change in Health Indicator per
Increment in PMa
Odds Ratio (95% CI)
1.06 (0.93, 1.21)
1.13 (0.92, 1.38)
1.1 (0.0.8, 1.7)
1.3 (0.7, 2.4)
1.4(0.9,2.1)
Range of City
PM Levels *
Means Cwg/m3)
28.0-84.9
0.9-3.2 ppb
13.0-70.7
6.7-31.5
1.0-5.0 ppb
Increased Obstruction in Adults
Southern California1"
Decreased Lung Function in
Six at/
Six City0
24 City**
24 Cit/4
24 Cit/4
24 City**
12 Southern California
communitiesN
(all children)
12 Southern California
communitiesN
(all children)
12 Southern California
communities0
(4th grade cohort)
12 Southern California
communities0
(4th grade cohort)
PM10 (cutoff of
42 days/year
>100 Mg/m3)
Children
PM15/10(50^g/m3)
TSP (100 ^ig/m3)
H+ (52 nmoles/m3)
PM21(15^g/m3)
SO=(7 ^g/m3)
PM10(17^g/m3)
PM10 (25 Mg/m3)
Acid vapor (1.7 ppb)
PM10 (25 A^g/m3)
Acid vapor (1.7 ppb)
PM10(51.5Mg/m3)
PM25 (25.9 Mg/m3)
PM10_25 (25.6 Mg/m3)
Acid vapor (4.3 ppb)
PM10(51.5Mg/m3)
PM25(25.9,wg/m3)
PM10.2.5(25.6Mg/m3)
Acid vapor (4.3 ppb)
1.09(0.92, 1.30)
NS Changes
NS Changes
-3.45% (-4.87, -2. 01) FVC
-3.21% (-4.98, -1.41) FVC
-3.06% (-4.50, -1.60) FVC
-2.42% (-4.30, -.0.51) FVC
-24.9 (-47.2, -2.6) FVC
-24.9 (-65.08, 15.28) FVC
-32.0 (-58.9, -5.1) MMEF
-7.9 (-60.43, 44.63) MMEF
-0.58 (-1.14, -0.02) FVC growth
-0.47 (-0.94, 0.01) FVC growth
-0.57 (-1.20, 0.06) FVC growth
-0.57 (-1.06, -0.07) FVC growth
-1.32 (-2.43, -0.20) MMEF growth
-1.03 (-1.95, -0.09) MMEF growth
-1.37 (-2.57, -0.15) MMEF growth
-1.03 (-2.09, 0.05) MMEF growth
NR
20-59
39-114
6.2-41.0
18.1-67.3
9.1-17.3
22.0-28.6
28.0-84.9
0.9-3. 2 ppb
28.0-84.9
0.9-3. 2 ppb
NR
NR
April 2002
9-113
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-17 (cont'd). EFFECT ESTIMATES PER INCREMENTSA IN LONG-TERM
MEAN LEVELS OF FINE AND INHALABLE PARTICLE INDICATORS FROM U.S.
AND CANADIAN STUDIES
Type of Health
Effect and Location
Decreased Lung Function in
Southern Californiap
(% predicted FEVb
females)
Southern California1"
(% predicted FEVb males)
Southern California1"
(% predicted FEVb males
whose parents had asthma,
bronchitis, emphysema)
Indicator
Adults
PM10 (cutoff of
54.2 days/year
>100 Mg/m3)
PM10 (cutoff of
54.2 days/year
>100 Mg/m3)
PM10 (cutoff of
54.2 days/year
>100 Mg/m3)
Range of City
Change in Health Indicator per PM Levels *
Increment in PMa Means (^g/m3)
+0.9 % (-0.8, 2.5) FEVj 52.7 (21.3, 80.6)
+0.3 % (-2.2, 2.8) FEVj 54.1 (20.0, 80.6)
-7.2 % (-11.5, -2.7) FEVj 54.1 (20.0, 80.6)
Southern California1"
(% predicted FEVj,
females)
Southern California1"
(% predicted FEVb males)
Not reported
-1.5% (-2.9, -0.1)FEV!
7.4(2.7, 10.1)
7.3 (2.0, 10.1)
*Range of mean PM levels given unless, as indicated, studies reported overall study mean (min, max), or mean
(±SD); NR=not reported.
AResults calculated using PM increment between the high and low levels in cities, or other PM increments given
in parentheses; NS Changes = No significant changes.
References:
BDockery et al. (1993)
GPope et al. (1995)
DKrewski et al. (2000)
EAbbey et al. (1999)
FDockery et al. (1989)
GWareetal. (1986)
HDockery et al. (1996)
'Abbey et al. (1995a,b,c)
JPetersetal. (1999b)
KMcConnell et al. (1999)
LBerglund et al. (1999)
MRaizenne et al. (1996)
NPeters et al. (1999a)
°Gauderman et al. (2000)
FAbbeyetal. (1998)
QPope et al. (2002)
1 individual covariates included. The time-dependent covariate model for total mortality (taking
2 into account higher postexposures in early years of the study and changes over time to the last
3 years of the study) had a substantially lower RR than the model without time-dependent
4 covariates. Educational level made a large difference, with individuals having less than a high
5 school education at much greater risk for mortality than those with any postsecondary education.
April 2002
9-114
DRAFT-DO NOT QUOTE OR CITE
-------
1 Among the ecological covariates, sulfates adjusted for artifact had little effect on the risk
2 estimates for total mortality compared to that without adjustment, but, in the ACS study, the filter
3 adjustment actually increased the relative risk for all causes and cardiopulmonary mortality,
4 while substantially reducing the estimated sulfate effect on lung cancer. Inclusion of SO2 as an
5 additional ecological covariate greatly reduced the estimated PM2 5 and sulfate effects in the ACS
6 study, whereas a spatial model including SO2 effects caused only a modest reduction of the
7 estimated PM2 5 and sulfate effects. However, the SO2 effects were reduced greatly when sulfates
8 were included in the model. Sulfur dioxide and sulfates often are highly correlated, because of
9 the formation of secondary sulfates.
10 Many model selection issues in the prospective cohort studies are analogous to those in the
11 time series analyses. One issue of particular concern is whether the exposure indices used in the
12 analyses adequately characterize the exposure of the participants in the study during the months
13 or years preceding death. This question is particularly conspicuous in regard to the Pope et al.
14 (1995) study, in which PM2 5 and sulfate data were collected in the 1979 to 1982 period from the
15 EPA AIRS database and the Inhalable Particle Network, largely preceding the collection of the
16 ACS cohort data by only a few years, and so possibly not adequately reflecting exposure to
17 presumably much higher PM concentrations occurring long before the cohort was recruited, nor
18 exposure to presumably lower concentrations during the study. This issue was raised in the 1996
19 PM AQCD. However, the Six Cities Study did have air pollution data and repeated survey data
20 over time, with PM2 5 and sulfate data measured every other day and sometimes daily, and so the
21 new investigators were able to use the information about time-dependent cumulative PM
22 concentrations during the course of the study. Changes in smoking status and body mass index
23 over the 10 to!2 years of the study had little effect on risk estimates, but taking into account the
24 decrease in particle concentrations from the earlier years to the later years reduced the effect size
25 estimate substantially, although it remained statistically significant. Nevertheless, overall, the
26 reanalyses of the ACS and Harvard Six-Cities studies (Krewski et al., 2000) "replicated the
27 original results, and tested those results against alternative risk models and analytic approaches
28 without substantively altering the original findings of an association between indicators of
29 particulate matter air pollution and mortality."
30 The shape of the relationship of concentration to mortality also was explored. Preliminary
31 findings suggest some possible nonlinearity, but further study is needed. Among the most
April 2002 9-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 important new findings of the study are spatial relationships between mortality and air pollution,
2 discussed later below.
3 Recently reported extension of the ACS analyses (Pope et al., 2002) to include additional
4 years of data provides further substantiation of originally reported findings for total, respiratory,
5 and cardiovascular mortality. Also of great importance, these new analyses provide much
6 stronger evidence substantiating links between long-term ambient fine PM exposures and lung
7 cancer. This is consistent with findings of increased lung cancer risk being associated with
8 exposure with diesel exhaust particles, an important constituent of PM2 5 in many U.S. urban
9 areas.
10 With regard to the role of various PM constituents in the PM-mortality association, past
11 cross-sectional studies generally have found that the fine particle component, as indicated either
12 by PM25 or sulfates, was the PM constituent most consistently associated with chronic PM
13 exposure-mortality. Although the relative measurement errors of the various PM constituents
14 must be further evaluated as a possible source of bias in these estimate comparisons, the Harvard
15 Six-Cities study and the latest reported AHSMOG prospective semi-individual study results
16 (Abbey, et al., 1999; McDonnell et al., 2000) are both indicative of the fine mass components of
17 PM likely being associated more strongly with the mortality effects of PM than coarse PM
18 components. The ACS study, its reanalyses, and its recent extension all further substantiate
19 ambient fine particle effects, including increased risk not only of cardiopulmonary-related
20 mortality but lung cancer mortality as well.
21 Several other new studies report epidemiologic evidence indicating that: (a) PM exposure
22 early in pregnancy (during the first month) may be associated with slowed intrauterine growth
23 leading to low birth weight events (Dejmek et al., 1999); and (b) early postnatal PM exposures
24 may lead to increased infant mortality (Woodruff et al., 1997; Boback and Leon, 1999; Loomis
25 et al., 1999; Lipfert et al., 2000b).
26
27 9.12.2.2 Relationships of Ambient Particulate Matter Concentrations to
28 Morbidity Outcomes
29 New epidemiology studies add greatly to the overall database relating morbidity outcomes
30 to ambient PM levels. These include much additional evidence for cardiovascular and
31 respiratory diseases being related to ambient PM. The newer epidemiology studies expand the
April 2002 9-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 evidence on cardiovascular (CVD) disease and are discussed first below, followed by discussion
2 of respiratory disease effects with particular emphasis on newly enhanced evidence for
3 PM-asthma relationships.
4
5 Cardiovascular Effects of Ambient Particulate Matter Exposures
6 About 75% of all U.S. deaths occur in persons at least 65 years old, and, of these, nearly
7 40% are for cardiac causes (nearly 45%, if deaths from cerebrovascular causes are also included).
8 Thus, if ambient PM exposure indeed produces increased total mortality in the elderly, it would
9 seem possible that cardiovascular (CVD) deaths may be involved.
10
11 Cardiovascular Hospital Admissions. Just two studies were available for review in the 1996
12 PM AQCD that provided data on acute cardiovascular morbidity outcomes (Schwartz and
13 Morris, 1995; Burnett et al., 1995). Both studies were of ecologic time series design using
14 standard statistical methods. Analyzing 4 years of data on the >65-year-old Medicare population
15 in Detroit, MI, Schwartz and Morris (1995) reported significant associations between ischemic
16 heart disease admissions and PM10, controlling for environmental covariates. Based on an
17 analysis of admissions data from 168 hospitals throughout Ontario, Canada, Burnett and
18 colleagues (1995) reported significant associations between particle sulfate concentrations, as
19 well as other air pollutants, and daily cardiovascular admissions. The relative risk because of
20 sulfate particles was slightly larger for respiratory than for cardiovascular hospital admissions.
21 The 1996 PM AQCD concluded on the basis of these studies that, "There is a suggestion of a
22 relationship to heart disease, but the results are based on only two studies and the estimated
23 effects are smaller than those for other endpoints." The PM AQCD went on to state that acute
24 impacts on CVD admissions had been demonstrated for elderly populations (i.e., >65), but that
25 insufficient data existed to assess relative impacts on younger populations.
26 Although the literature still remains relatively sparse, an important new body of data now
27 exists that both extends the available quantitative information on relationships between ambient
28 PM pollution and hospital CVD admissions, and that, more intriguingly, illuminates some of the
29 physiological changes that may occur on the mechanistic pathway leading from PM exposure to
30 adverse cardiac outcomes. Figure 9-27 depicts excess risk estimates derived from 10 studies of
31 acute PM10 exposure effects on CVD admissions in U.S. cities. Although new studies depicted
April 2002 9-117 DRAFT-DO NOT QUOTE OR CITE
-------
Sametetal. (2000a,b) -
14US Cities
Schwartz (1999) -
8 US Counties
Moolgavkar (2000c) -
Maricopa, AZ
Moolgavkar (2000c) -
LA.CA
Moolgavkar (2000c) -
Cook County
Linn et al. (2000) -
LA.CA
Schwartz (1997) -
Tucson,AZ
Tolbert et al. (2000) -
Atlanta
Morris and Naumova (1998) -
Chicago
Lippmannet al.(2000) -
Total CVD
Period 1 (AIRS Data)
I <
CHF
i » i
Period 2 (Supersite Data)
I » 1
-15 -10 -5 0 5 10
Reconstructed Excess Risk Percentage
50 ug/m3 Increase in
Figure 9-27. Acute cardiovascular hospitalizations and PM exposure excess risk estimates
derived from selected U.S. PM10 studies. CVD = cardiovascular disease and
CHF = congestive heart failure.
1 in Figure 9-27 have reported generally consistent associations between daily hospitalizations for
2 cardiovascular disease and measures of PM, the data not only implicate PM, but also CO and
3 NO2 as well, possibly because of covarying of PM and these other gaseous pollutants derived
4 from common emission sources (e.g., motor vehicles). Taken as a whole, this body of evidence
5 suggests that PM is likely an important risk factor for cardiovascular hospitalizations in the
6 United States.
April 2002
9-118
DRAFT-DO NOT QUOTE OR CITE
-------
1 For example, in the recently published NMMAPS 14-city analysis of daily CVD hospital
2 admissions in persons 65 and older in relation to PM10 (Samet et al., 2000a,b). The mean risk
3 estimate (for average 0-1 day lag) was a 8.5% increase in CVD admissions per 50 //g/m3 PM10
4 (95% CI: 1.0 to 33.0%). No relationship was observed between city-specific risk estimates and
5 measures of socioeconomic status, including percent living in poverty, percent non-white, and
6 percent with college educations. In another study, remarkably consistent PM10 associations with
7 cardiovascular admissions were observed across eight U.S. metropolitan areas, with a 25 //g/m3
8 increase in PM10 associated with between 1.8 and 4.2 percent increases in admissions (Schwartz,
9 1999). Also, in a study of Los Angeles data from 1992-1995, PM10, CO, and NO2 were all
10 significantly associated with increased cardiovascular admission in single-pollutant models
11 among persons 30 and older (Linn et al., 2000). Moolgavkar (2000c) analyzed PM10, CO, NO2,
12 O3, and SO2 in relation to daily total cardiovascular (CVD) and total cerebrovascular admissions
13 for persons 65 and older from three urban counties (Cook, IL; Los Angeles, CA; Maricopa, AZ),
14 and found that, in univariate regressions, PM10 (and PM25 in LA) was associated with CVD
15 admissions in Cook and LA counties but not in Maricopa county. On the other hand, in
16 two-pollutant models in Cook and LA counties, the PM risk estimates diminished and/or were
17 rendered nonsignificant.
18 The recent NMMAPS study of PM10 concentrations and hospital admissions by persons
19 65 and older in 14 U.S. cities provides particularly important findings of positive and significant
20 associations, even when concentrations are below 50 //g/m3 (Samet et al., 2000a,b). As noted in
21 Table 9-18, this study indicates PM10 effects similar to other cities, but with narrower confidence
22 bands, because of its greater power derived by combining multiple cities in the same analysis.
23 This allows significant associations to be identified, despite the fact that many of the cities
24 considered have relatively small populations and that each of the 14 cities had mean PM10 below
25 50 //g/m3.
26
27 Physiologic Measures of Cardiac Function. Several very recent studies by independent groups
28 of investigators have also reported longitudinal associations between ambient PM concentrations
29 and physiologic measures of cardiovascular function. These studies measure outcomes and most
30 covariates at the individual level, making it possible to draw conclusions regarding individual
31 risks, as well as to explore mechanistic hypotheses. For example, several studies recently have
April 2002 9-119 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-18. PERCENT INCREASE IN HOSPITAL ADMISSIONS PER 10-^g/m3
INCREASE IN 24-HOUR PMin IN 14 U.S. CITIES
CVD
COPD
Pneumonia
Increase
(95% CI)
Increase
(95% CI)
Increase
(95% CI)
Constrained Lag Models (Fixed Effect Estimates)
One-day meana
Previous-day mean
Two-day meanb
PM10<50Mg.m3
(2 -day mean)b
Quadratic distributed lag
Unconstrained Distributed
Fixed effects estimate
Random effects estimate
1.07
0.68
1.17
1.47
1.18
Lag
1.19
1.07
(0.93, 1.22)
(0.54,0.81)
(1.01, 1.33)
(1.18, 1.76)
(0.96, 1.39)
(0.97, 1.41)
(0.67, 1.46)
1.44
1.46
1.98
2.63
2.49
2.45
2.88
(1.00, 1.89)
(1.03, 1.88)
(1.49, 2.47)
(1.71,3.55)
(1.78, 3.20)
(1.75,3.17)
(0.19,5.64)
1.57
1.31
1.98
2.84
1.68
1.90
2.07
(1.27, 1.87)
(1.03, 1.58)
(1.65,2.31)
(2.21, 3.48)
(1.25,2.11)
(1.46,2.34)
(0.94, 3.22)
aLag.
bMean of lag 0 and lag 1.
Source: Samet et al. (2000a,b).
1 reported temporal associations between PM exposures and various electrocardiogram (ECG)
2 measures of heart beat or rhythm in panels of elderly subj ects. Reduced HR variability is a
3 predictor of increased cardiovascular morbidity and mortality risks. Three independent studies
4 reported decreases in HR variability associated with PM in elderly cohorts, although r-MSSD
5 (one measure of high-frequency HR variability) showed elevations with PM in one study.
6 Differences in methods used and results obtained across the studies argue for caution in
7 drawing any strong conclusions yet regarding PM effects from them, especially in light of the
8 complex intercorrelations that exist among measures of cardiac physiology, meteorology, and air
9 pollution (Dockery et al., 1999). Still, the new heart rhythm results, in general, comport well
10 with other findings of cardiovascular mortality and morbidity endpoints being associated with
11 ambient PM. Chapter 5 discusses available exposure studies of elderly subjects with CVD, such
12 as the Sarnat et al. (2000) Baltimore study. Less active groups tend to have lower exposure to
13 nonambient PM because of reduced personal activity. However, Williams et al. (2000a,b,c)
April 2002
9-120
DRAFT-DO NOT QUOTE OR CITE
-------
1 report a very high pooled correlation coefficient between PM2 5 personal exposure and outdoor
2 concentrations. These exposure studies tend to enhance the plausibility of panel study findings
3 of impacts on HR variability being caused by exposure to ambient-generated PM.
4
5 Changes in Blood Characteristics. Additional epidemiologic findings (Peters et al., 1997a)
6 also provide new evidence for ambient PM exposure effects on blood characteristics (e.g.,
7 increased c-reactive protein in blood) thought to be associated with increased risk of serious
8 cardiac outcomes (e.g., heart attacks).
9
10 Key Conclusions Regarding PM-CVD Morbidity
11 Overall, the newly available studies of PM-CVD relationships appear to support the
12 following conclusions regarding several key issues:
13
14 Temporal Patterns of Response. The evidence from recent time series studies of CVD
15 admissions suggests rather strongly that PM effects are likely maximal at lag 0, with some
16 carryover to lag 1.
17
18 Physical and Chemical Attributes Related to Particulate Matter Health Effects. The
19 characterization of ambient PM attributes associated with acute CVD is incomplete. Insufficient
20 data exist from the time series CVD hospital admissions literature or from the emerging
21 individual-level studies to provide clear guidance as to which PM attributes, defined either on the
22 basis of size or composition, determine potency. The epidemiologic studies published to date
23 have been constrained by the limited availability of multiple PM metrics. Where multiple PM
24 metrics exist, they often are of differential quality because of differences in numbers of
25 monitoring sites and in monitoring frequency. Until more extensive and consistent data become
26 available for epidemiologic research, the question of PM size and composition, as they relate to
27 acute CVD impacts, will remain open.
28
29 Susceptible Subpopulations. Because they lack data on individual subject characteristics,
30 ecologic time series studies provide only limited information on susceptibility factors based on
31 stratified analyses. The relative impact of PM on cardiovascular (and respiratory) admissions
April 2002 9-121 DRAFT-DO NOT QUOTE OR CITE
-------
1 reported in ecologic time series studies is generally somewhat higher than those reported for total
2 admissions. This provides some limited support for the hypothesis that acute effects of PM
3 operate via cardiopulmonary pathways or that persons with preexisting cardiopulmonary disease
4 have greater susceptibility to PM, or both. Although there is some data from the ecologic time
5 series studies showing larger relative impacts of PM on cardiovascular admissions in adults 65
6 and over as compared with younger populations, the differences are neither striking nor
7 consistent. Some individual-level studies of cardiophysiologic function suggest that elderly
8 persons with preexisting cardiopulmonary disease are susceptible to subtle changes in heart rate
9 variability (HRV) in association with PM exposures. However, because younger and healthier
10 populations have not yet been assessed, it is not possible to say at present whether the elderly
11 have clearly increased susceptibility compared to other groups, as indexed by cardiac
12 pathophysiological indices such as HRV.
13
14 Role of Other Environmental Factors. The ecologic time series morbidity studies published since
15 1996 generally have controlled adequately for weather influences. Thus, it is unlikely that
16 residual confounding by weather accounts for the PM associations observed. With one possible
17 exception (Pope et al., 1999a), the roles of meteorological factors have not been analyzed
18 extensively as yet in the individual-level studies of cardiac physiologic function. Thus, the
19 possibility of confounding in such studies as yet cannot be discounted totally or readily.
20 Co-pollutants have been analyzed rather extensively in many of the recent time series studies of
21 hospital admissions and PM. In some studies, PM clearly carries an independent association
22 after controlling for gaseous co-pollutants. In others, the "PM effects" are reduced markedly
23 once co-pollutants are added to the model. Among the gaseous criteria pollutants, CO has
24 emerged as the most consistently associated with cardiovascular (CVD) hospitalizations. The
25 CO effects are generally robust in the multi-pollutant model, sometimes as much so as PM
26 effects. However, the typically low levels of ambient CO concentrations in most such studies
27 and minimal expected impacts on carboxyhemoglobin levels and consequent associated hypoxic
28 effects thought to underlie CO CVD effects complicate interpretation of the CO findings and
29 argue for the possibility that CO may be serving as a general surrogate for combustion products
30 (e.g., PM) in the ambient pollution mix. See the most recent EPA CO Criteria Document (U.S.
31 Environmental Protection Agency, 2000a).
April 2002 9-122 DRAFT-DO NOT QUOTE OR CITE
-------
1 Respiratory Effects of Ambient Particulate Matter Exposures
2 The number of studies examining hospitalization and emergency department visits for
3 respiratory-related causes and other respiratory morbidity endpoints has increased markedly since
4 the 1996 PM AQCD. In addition to evaluating statistical relationships for PM10, quite a few new
5 studies also evaluated other PM metrics. Those providing estimates of increased risk in U.S. and
6 Canadian cities for respiratory-related morbidity measures (hospitalizations, respiratory
7 symptoms, etc.) in relation to 24-h increments in ambient fine particles (PM2 5) or coarse fraction
8 (PM10_25) of inhalable thoracic particles are included in Tables 9-12 and 9-13, respectively.
9
10 Respiratory-Related Hospital Admission/Visits. PM hospital admissions/ visit studies that
11 evaluated excess risks in relation to PM10 measures are still quite informative. Maximum excess
12 risk estimates for PM10 associations with respiratory-related hospital admissions and visits in
13 U.S. cities are shown in Figure 9-28. Nearly all the studies showed positive, statistically
14 significant relationships between ambient PM10 and increased risk for respiratory-related doctors'
15 visits and hospital admissions. Overall, the results substantiate well ambient PM10 impacts on
16 respiratory-related hospital admissions/visits. The excess risk estimates fall most consistently in
17 the range of 5 to 25.0% per 50 //g/m3 PM10 increment, with those for asthma hospital admissions
18 and doctor's visits being higher than for COPD and pneumonia hospitalization. Other, more
19 limited, new evidence (not depicted in Figure 9-10) shows excess risk estimates for overall
20 respiratory-related or COPD hospital admissions falling in the range of 5 to 15.0% per 24-h
21 25 //g/m3 increment in PM25 or PM10_25. Larger estimates are found for asthma admissions or
22 physician visits, ranging up to ca. 40 to 50% for children <18 yr old in one study.
23 Of particular note in Figure 9-28 are the large effect size estimates now being reported for
24 asthma hospitalizations and visits. Very importantly, these hospital admission/visit studies and
25 other new studies on respiratory symptoms and lung function decrements in asthmatics are
26 emerging as possibly indicative of ambient PM likely being a notable contributor to exacerbation
27 of asthma. Additional evidence for PM-asthma effects is also emerging from panel studies of
28 lung function and respiratory symptoms, as discussed below.
29 New panel studies of lung function and respiratory symptoms in asthmatic subjects have
30 been conducted by more than 10 research teams in various locations world-wide. As a group, the
31 studies examine health outcome effects that are similar, such as pulmonary peak flow rate
April 2002 9-123 DRAFT-DO NOT QUOTE OR CITE
-------
Tolbert et al. (2000) Atlanta -
Morris et al. (2000) Seattle -
Morris et al. (2000) Spokane -
Morris etal. (1999) Seattle -
Choudhury et al. (1997) Anchorage -
Nauenberg and Basu (1999) LA.CA -
Sheppard et al. (1999) Seattle -
Zanobetti et al. (2000a) Chicago -
Sametetal. (2000a) 14 US Cities -
Moolgavkar (2000b) Phoenix -
Moolgavkar (2000b) LA.CA -
Moolgavkar (2000b) Chicago -
Moolgavkar et al. (2000) King C -
Moolgavkar et al. (1997) Minn-SP -
Moolgavkar et al. (1997) Birm. -
Chen et al. (2000) Reno.NV -
Zanobetti et al. (2000a) Chicago -
Sametetal. (2000a) 14 US Cities -
,
Asthma Visits
.
Asthma Hospital Admissions
I-H
«_! COPD Hospital Admissions
— '
(-»-!
«H
1 * 1
w Pneumonia Hospital Admissions
-25
25 50 75 100
Excess Risk, %
125
150
Figure 9-28. Maximum excess risk in selected studies of U.S. cities relating PM10 estimate
of exposure (50 jUg/m3) to respiratory-related hospital admissions and visits.
1 (PEFR); and the studies typically characterize the clinical-symptomatic aspects in a sample of
2 mild to moderate asthmatics (mainly children aged 5 to 16 yrs) observed in their natural setting.
3 Their asthma typically is being treated to keep them symptom free (with "normal" pulmonary
4 function rates, and activity levels) and to prevent recurrent exacerbations of asthma. Severity of
5 their asthma is characterized by symptom, pulmonary function, and medication use and would be
6 classified to include mild intermittent to mild persistent asthma suffers (National Institutes of
7 Health, 1997). As a group, they may thusly differ from asthmatics examined in studies of
8 hospitalization or doctor visits for acute asthmatic episodes, who may have more severe asthma.
9 Most studies reported ambient PM10 results, but PM2 5 was examined in two studies. Other
10 ambient PM measures (BS and SO4) also were used. For these studies, mean PM10 levels range
April 2002
9-124
DRAFT-DO NOT QUOTE OR CITE
-------
1 from a low of 13 //g/m3 in Finland to a high of 167 //g/m3 in Mexico City. The Mexico City
2 level is over three times more than each of the other levels and is unique compared to the others.
3 Related 95% CI for these means or ranges show 1-day maximums above 100 //g/m3 in four
4 studies, with two of these above 150 //g/m3. Hence, these studies mainly evaluated different PM
5 metrics indexing PM concentrations in the range found in U.S. cities (see Chapter 3). All the
6 studies controlled for temperature, and several controlled for relative humidity.
7 Many panel studies are analyzed using a design that takes advantage of the repeated
8 measures on the same subject. Study subject number (N) varied from 12 to 164, with most
9 having N >50; and all gathered adequate subject-day data to provide sufficient power for their
10 analyses. Linear models often are used for lung function and logistic models for dichotomous
11 outcomes. Meteorological variables are used as covariates; and medication use is also sometimes
12 evaluated as a dependent variable or treated as an important potential confounder. However,
13 perhaps the most critical choice in the model is selection of the lag for the pollution variable.
14 Presenting lag periods with only the strongest associations introduces potential bias, because the
15 biological basis for lag structure may be related to effect. No biological bases for pertinent lag
16 periods are known, but some hypotheses can be proposed. Acute asthmatic reactions can occur
17 4 to 6 h after exposure and, thus, 0-day lag may be more appropriate than 1-day lags for that
18 acute reaction. Lag 1 may be more relevant for morning measurement of asthma outcome from
19 PM exposure the day before, and longer term lags (i.e., 2 to 5 days) may represent the outcome of
20 a more prolonged inflammatory mechanism; but too little information is now available to
21 predetermine appropriate lag(s).
22 Chapter 8 noted that people with asthma tend to have greater TB deposition than do healthy
23 people, but this data was not derived from the younger age group studied in most asthma panel
24 studies. The Peters et al. (1997b) study is unique for two reasons: (1) they studied the size
25 distribution of the particles in the range 0.01 to 2.5 //m and (2) examined the number of particles.
26 They reported that asthma-related health effects of 5-day means of the number of ultrafine
27 particles were larger than those of the mass of the fine particles. In contrast, Pekkanen et al.
28 (1997) also examined a range of PM sizes, but PM10 was more consistently associated with PEF.
29 Delfino et al. (1998) is unique in that they report larger effects for 1- and 8-h maximum PM10
30 than for the 24-h mean.
April 2002 9-125 DRAFT-DO NOT QUOTE OR CITE
-------
1 The results for the asthma panels of the peak flow analysis consistently show small
2 decrements for both PM10 and PM2 5. The effects using 2- to 5-day lags averaged about the same
3 as did the 0 to 1 day lags. Stronger relationships often were found with ozone. The analyses
4 were not able to clearly separate co-pollutant effects. The effects on respiratory symptoms in
5 asthmatics also tended to be positive. Most studies showed increases in cough, phlegm,
6 difficulty breathing, and bronchodilator use. The only endpoint more strongly related to longer
7 lag times was bronchodilator use, which was observed in three studies. The peak flow
8 decrements and respiratory symptoms are indicators for asthma episodes.
9 For PM10, nearly all of the point estimates showed decreases, but most were not statistically
10 significant, as shown in Figure 9-29 as an example of PEF outcomes. Lag 1 may be more
11 relevant for morning measurement of asthma outcome from the previous day. The figure
12 presents studies that provided this data. The results were consistent for both AM and PM peak
13 flow analyses. Similar results were found for the PM25 studies, although there were fewer
14 studies. Several studies included PM25 and PM10 independently in their analyses of peak flow.
15 Of these, Gold et al. (1999), Naeher et al. (1999), Tiittanen et al. (1999), Pekkanen et al. (1997),
16 and Romieu et al. (1996) all found similar results for PM25 and PM10. The study of Peters et al.
17 (1997b) found slightly larger effects for PM25. The study of Schwartz and Neas (2000) found
18 larger effects for PM2 5 than for PM10_2 5. Naeher et al. (1999) found that FT was related
19 significantly to a decrease in morning PEF. Thus, there is no evidence here for a stronger effect
20 of PM2 5 when compared to PM10. Also, of studies that provided analyses that attempted to
21 separate out effects of PM10 and PM2 5 from other pollutants, Gold et al. (1999) studied possible
22 interactive effects of PM25 and ozone on PEF; they found independent effects of the two
23 pollutants, but the joint effect was slightly less than the sum of the independent effects.
24 The effects on respiratory symptoms in asthmatics also tended to be positive, although
25 much less consistent than the lung function effects. Most studies showed increases in cough,
26 phlegm, difficulty breathing, and bronchodilator use (although generally not statistically
27 significant), as shown in Figure 9-30 for cough as an example. Three studies included both PM10
28 and PM2 5 in their analyses. The studies of Peters et al. (1997c) and Tiittanen et al. (1999) found
29 comparable effects for the two measures. Only the Romieu et al. (1996) found slightly larger
30 effects for PM2 5. These studies also give no good evidence for a stronger effect of PM2 5 when
31 compared to PM10.
April 2002 9-126 DRAFT-DO NOT QUOTE OR CITE
-------
Romieu etal. (1996)
(Mexico)
Pekkannen etal. (1997)
(Finland)
Gielenetal. (1997)
(Netherlands)
Romieu et al. (1997)
(Mexico)
-10 -5 0 5
Change in Pulmonary Function, L/min
Figure 9-29. Selected acute pulmonary function change studies of asthmatic children.
Effect of 50 Aig/m3 PM10 on morning peak flow lagged 1 day.
Vedaletal. (1998) -
(Canada)
Romieu etal. (1997)
(Mexico)
Gielenetal. (1997)
(Netherlands)
Peters et al. (1997c)
(Czech Republic)
-\ 1 1 1 1-
12345
Odds Ratios for Cough
Figure 9-30. Odds ratios for cough for a 50-//g/m3 increase in PM10 for selected asthmatic
children studies, with lag 0 with 95% CI.
April 2002
9-127
DRAFT-DO NOT QUOTE OR CITE
-------
1 The results of PM10 peak flow analyses for nonasthmatic populations were inconsistent.
2 Fewer studies reported results in the same manner as the asthmatic studies. Many of the point
3 estimates showed increases rather than decreases. PM2 5 studies found similar results. The
4 effects on respiratory symptoms in nonasthmatics were similar to those in asthmatics: most
5 studies showed that PM10 increases cough, phlegm, and difficulty breathing, but these increases
6 were generally not statistically significant. Schwartz and Neas (2000) found that PM10_2 5 was
7 significantly related to cough. Tiittanen et al. (1999) found that 1-day lag of PM10_25 was related
8 to morning PEF, but not evening PEF. Neas et al. (1999) found no association of PM10_25 with
9 PEF in non-asthmatic subjects.
10
11 Long-Term Particulate Matter Exposure Effects on Lung Function and Respiratory
12 Symptoms
13 In the 1996 PM AQCD, the available respiratory disease studies were limited in terms of
14 conclusions that could be drawn. At that time, three studies based on a similar type of
15 questionnaire administered at three different times as part of the Harvard Six-City and 24-City
16 Studies provided data on the relationship of chronic respiratory disease to PM. All three studies
17 suggest a chronic PM exposure effect on respiratory disease. The analysis of chronic cough,
18 chest illness, and bronchitis tended to be significantly positive for the earlier surveys described
19 by Ware et al. (1986) and Dockery et al. (1989). Using a design similar to the earlier one,
20 Dockery et al. (1996) expanded the analyses to include 24 communities in the United States and
21 Canada. Bronchitis was found to be higher (odds ratio = 1.66) in the community with highest
22 exposure of strongly acidic particles when compared with the least polluted community. Fine
23 PM sulfate was also associated with higher reporting of bronchitis (OR = 1.65, 95% CI1.12,
24 2.42).
25 The studies by Ware et al. (1986), Dockery et al. (1989), and Neas et al. (1994) all had
26 good monitoring data and well-conducted standardized pulmonary function testing over many
27 years, but showed no effect on children of PM pollution indexed by TSP, PM15, PM25, or
28 sulfates. In contrast, the later 24-city analyses reported by Raizenne et al. (1996) found
29 significant associations of effects on FEVj or FVC in U.S. and Canadian children with both
30 acidic particles and other PM indicators. Overall, the available studies provided limited evidence
April 2002 9-128 DRAFT-DO NOT QUOTE OR CITE
-------
1 suggestive of pulmonary lung function decrements being associated with chronic exposure to PM
2 indexed by various measures (TSP, PM10, sulfates, etc.).
3 A number of studies have been published since 1996 which evaluate the effects of
4 long-term PM exposure on lung function and respiratory symptoms, as presented in Chapter 8.
5 The methodology in the long-term studies varies much more than the methodology in the short-
6 term studies. Some studies reported highly significant results (related to PM), whereas others
7 reported no significant results. Of particular note are several studies reporting associations
8 between long-term PM exposures (indexed by various measures) or changes in such exposures
9 over time and chronic bronchitis rates, consistent with the findings on bronchitis from the
10 Dockery et al. (1996) study noted above.
11 Unfortunately, the cross-sectional studies often are potentially confounded, in part, by
12 unexplained differences in geographic regions; and it is difficult to separate out results consistent
13 with a PM gradient from any other pollutants or factors having the same gradient. The studies
14 that looked for a time trend also are confounded by other conditions that changed over time. The
15 most credible cross-sectional study remains that described by Dockery et al. (1996) and Raizenne
16 et al. (1996). Whereas most studies include two to six communities, this study included 24
17 communities and is considered to provide the most credible estimates of long-term PM exposure
18 effects on lung function and respiratory symptoms.
19
20 9.12.2.3 Methodological Issues
21 Chapter 8 discussed several still important methodological issues related to assessment of
22 the overall PM epidemiologic database. These include, especially, issues related to model
23 specifications and consequent adequacy of control for potentially confounding of PM effects by
24 co-pollutants, evaluations of possible source relationships to pollutant effects that may be useful
25 in sorting out better effects attributable to PM versus other co-pollutants or both, and other issues
26 such as lag structure. Key points are discussed concisely below.
27
28 Time Series Studies: Confounding by Co-Pollutants in Individual Cities
29 The co-pollutant issue was discussed at length in the 1996 document and still remains an
30 important issue. It must be recognized that there are large differences in concentrations of
31 measured gaseous co-pollutants (and presumably unmeasured pollutants as well) in different
April 2002 9-129 DRAFT-DO NOT QUOTE OR CITE
-------
1 parts of the United States, as well as the rest of the world; and the concentrations are often
2 correlated with concentrations of PM and its components because of commonality in source
3 emissions, wind speed and direction, atmospheric processes, and other human activities and
4 meteorological conditions. Large sources in the United States include motor vehicle emissions
5 (gasoline combustion, diesel fuel combustion, evaporation, particles generated by tire wear, etc.),
6 coal combustion, fuel oil combustion, industrial processes, residential wood burning, solid waste
7 combustion, and so on. Thus, one might reasonably expect some large correlations among PM
8 and co-pollutants, but possibly with substantial differences in relation by season in different
9 cities or regions. Statistical theory suggests that PM and co-pollutant effect size estimates will be
10 highly unstable and often insignificant in multi-pollutant models when collinearity exists. Many
11 recent studies demonstrate this effect, for both hospital admissions (Moolgavkar, 2000b) and
12 mortality (Moolgavkar, 2000a; Chock et al., 2000). Because the problem seems largely insoluble
13 in studies in single cities, the new multi-city studies (Samet et al., 2000a,b; Schwartz, 1999;
14 Schwartz and Zanobetti, 2000) have provided important new insights. See discussions of
15 NMMAPS analysis in Chapter 8 and below for discussion of issues related to control for
16 co-pollutant effects. Overall, although such issues may warrant further evaluation, it now
17 appears unlikely that such confounding accounts for the vast array of effects attributed to ambient
18 PM based on the rapidly expanding PM epidemiology database.
19 Numerous new studies have reported associations not only between PM, but also gaseous
20 pollutants (O3, SO2, NO2, and CO), and mortality. In many of these studies, simultaneous
21 inclusion of one or more gaseous pollutants in regression models did not markedly affect PM
22 effect size estimates, as was generally the case in the NMMAPS analyses for 90 cities (see
23 Figure 9-31). On the other hand, some studies reporting positive and statistically significant
24 effects for gaseous copollutants (e.g., O3, NO2, SO2, CO) found varying degrees of robustness of
25 their effects estimates or those of PM in multi-pollutant models as discussed in Chapter 8
26 (Section 8.4). Thus, it is likely that there are independent health effects of PM and gaseous
27 pollutants, there is not yet sufficient evidence by which to confidently separate out fully the
28 relative contributions of PM versus those of other gaseous pollutants or by which to quantitate
29 modifications of PM effects by other co-pollutants, including possible synergistic interactions
30 that may vary seasonally or from location to location. Overall, it appears, however, that ambient
31 PM and O3 can be most clearly separated out as likely having independent effects, their
April 2002 9-130 DRAFT-DO NOT QUOTE OR CITE
-------
PM10
PM10 + 03
PM10 + 03 + N02
PM10 + 03+SO2
PM10 + 03 + CO
_ _ _ _
— . —
_.._.._
1 .UU
1 .UU
1.00
1.00
1.00
0.0 0.2 0.4 0.6
% Change in Mortality per 10 |jg/m3 Increase in PM10
1.0
Figure 9-31. Marginal posterior distributions for effect of PM10 on total mortality at lag 1,
with and without control for other pollutants, for the 90 cities. The numbers
in the upper right legend are the posterior probabilities that the overall
effects are greater than 0.
Source: Samet et al. (2000a,b).
1 concentrations often not being highly correlated. More difficulty is encountered, at times, in
2 sorting out whether NO2, CO, or SO2 are exerting independent effects in cities where they tend to
3 be highly correlated with ambient PM concentrations, possibly because of derivation of
4 important PM constituents from the same source (e.g., NO2, CO, PM from mobile sources) or a
5 gaseous pollutant (e.g., SO2) serving as a precursor for a significant PM component (e.g.,
6 sulfate). However, other information discussed in Section 8.4 on conceptual frameworks for
7 evaluating possible confounding makes it clear that diagnostic evaluations of inflation or
8 deflation of PM effect size estimates by addition of gaseous co-pollutants into multiple pollutant
April 2002
9-131
DRAFT-DO NOT QUOTE OR CITE
-------
1 models, at best, may indicate potential confounding of PM effects in a given analysis. Other
2 independently-derived exposure analyses, i.e., Sarnat et al. (2000, 2001), however, strongly
3 suggest a very low probability of observed PM effects being due to confounding with gaseous
4 criteria pollutants (CO, NO2, SO2, O3).
5
6 Time Series Studies: Model Selection for Lags, Moving Averages, and Distributed Lags
1 A number of different approaches have been used to evaluate the temporal dependence of
8 mortality or morbidity on time-lagged PM concentrations, including unweighted moving
9 averages of PM concentrations over one or more days, general weighted moving averages, and
10 polynomial distributed moving averages. Unless there are nearly complete daily data, each
11 different lag will be using a different set of mortality data corresponding to spaced PM
12 measurement; for example, for lag 0 with every-sixth-day PM measurements, the mortality data
13 are on the same day as the PM data, for lag 1 the mortality data are on the next day after the PM
14 data, and so on. Although this effect is likely to be small, it should nonetheless be kept in mind.
15 The issue of dealing with lag structure, which may not necessarily be the same for all cities
16 or for all regions, can be illustrated by NMMAPS findings. As shown in Table 9-19, the rank
17 ordering of effects by lag days differs somewhat among NMMAPS regions. The combined data
18 set suggests that lag 1 provides the best fit, but with some regional differences. This raises the
19 question as to whether a single lag model should be assumed to characterize a diverse set of
20 regional findings. Because the particle constituents, co-pollutants, susceptible subpopulations,
21 and meteorological covariates are likely to differ substantially from one region to another, the
22 timing of the largest mortality effects also may be presumed to differ in at least some cases. This
23 undoubtedly contributes to the variance of the estimated effects.
24 The distributed lag models used in the NMMAPS II morbidity studies are a noteworthy
25 methodological advance. The fitted distributed lag models showed significant heterogeneity
26 across cities for COPD and pneumonia, however (see Table 15 therein), again raising the
27 question of how heterogeneous effects can best be combined so as not to obscure potentially real
28 city-specific or region-specific differences.
29 Only three cities with nearly complete daily PM10 data were used to evaluate more general
30 multi-day lag models (Chicago, Minneapolis/St. Paul, Pittsburgh), and these show somewhat
31 different patterns of effect, with lag 0 < lag 1 and lag 1 »lag 2 for Chicago, lag 0 = lag 1 > lag 2
April 2002 9-132 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-19. PERCENT INCREASE IN MORTALITY PER 10 //g/m3 PM10
IN SEVEN U.S. REGIONS (from Figure 23 in NMMAPS II)
Region Rank Order of Effects by Lags
Northwest lag 0 < lag 1 = lag 2
Southwest lag 0 < lag 1< lag 2
Southern California lag 0 < lag 1, lag 1 > lag 2, lag 0 < lag 2
Upper Midwest lag 0 > lag 1, lag 0 > lag 2, lag 1 < lag 2
Industrial Midwest lag 0 < lag 1, lag 1 > lag 2
Northeast lag 0 < lag 1, lag 1 » lag 2
Southeast lag 0 « lag 1, lag 1 > lag 2
Combined lag 0 < lag 1, lag 1 > lag 2
1 for Minneapolis, and lag 0 < lag 1 = lag 2 for Pittsburgh. The 7-day distributed lag model is
2 significant for Pittsburgh, but less so in the other cities. The remaining data are limited
3 intrinsically in what they can reveal about temporal structure.
4
5 Time Series Studies: Model Selection for Concentration-Response Functions
6 Given the number of analyses that needed to be performed, it is not surprising that most of
7 the NMMAPS studies focused on linear concentration-response models. More recent studies
8 (Daniels et al., 2000) for the 20 largest U.S. cities have found posterior mean effects of 2 to 2.7%
9 excess risk of total daily mortality per 50 //g/m3 24-h PM10 at lags 0, 1, 0+1 days; 2.4 to 3.5%
10 excess risk of cardiovascular and respiratory mortality; and 1.2 to 1.7% for other causes of
11 mortality. The posterior 95% credible regions are all significantly greater than 0. However, the
12 threshold models gave distinctly different estimates of 95% credible regions for the threshold for
13 total mortality (15 //g/m3 at lag 1, range 10 to 20), cardiovascular and respiratory mortality
14 (15 //g/m3 at lag 0+1, range 0 to 20), and other causes of mortality (65 //g/m3 at lag 0+1, range
15 50 to 75 //g/m3).
16 Another problem is that the shape of the relationship between mortality and PM10 may
17 depend, to some extent, on the associations of PM10 with gaseous co-pollutants. The association
April 2002 9-133 DRAFT-DO NOT QUOTE OR CITE
-------
1 is not necessarily linear, and is indeed likely to have both seasonal and secular components that
2 depend on the city location. Thus, further elaborations of these models may be desirable.
3
4 Effects of Exposure Error in Daily Time Series Epidemiology
5 There has been considerable controversy over how to deal with the nonambient component
6 of personal exposure. Recent biostatistical analyses of exposure error have indicated that the
7 nonambient component will not bias the statistically calculated risk in community time-series
8 epidemiology, provided that the nonambient component of personal exposure is independent of
9 the ambient concentration. Consideration of the random nature of nonambient sources and recent
10 studies, in which estimates of a, ambient-generated PM divided by ambient PM concentrations,
11 have been used to estimate separately the ambient-generated and nonambient components of
12 personal exposure, support the assumption that the nonambient exposure is independent of the
13 ambient concentration. Therefore, it is reasonable to conclude that community time series
14 epidemiology describes statistical associations between health effects and exposure to ambient-
15 generated PM, but does not provide any information on possible health effects resulting from
16 exposure to nonambient PM (e.g., indoor-generated PM).
17 From the point of view of exposure error, it is also significant to note that, although
18 ambient concentrations of a number of gaseous pollutants (O3, NO2, SO2) often are found to be
19 highly correlated with various PM parameters, personal exposures to these gases are not
20 correlated highly with personal exposure to PM indicators. The correlations of the ambient
21 concentrations of these gases also are not correlated highly with the personal exposure to these
22 gases. Therefore, when significant statistical associations are found between these gases and
23 health effects, it could be that these gases may, at times, be serving as surrogates for PM rather
24 than being causal themselves. Pertinent information on CO has not been reported.
25 The attenuation factor, a, is a useful variable. For relatively constant a, the risk because of
26 a personal exposure to 10 //g/m3 of ambient PM is equal I/a times the risk from a concentration
27 of 10 //g/m3 of ambient PM, where a varies from a low of 0.1 to 0.2 to a maximum of 1.0. (The
28 health risk for an interquartile change in ambient concentration of PM is the same as that for an
29 interquartile change in exposure to ambient PM). Differences in a among cities, reflecting
30 differences in air-exchange rates (e.g., because of variation in seasonal temperatures and in extent
31 of use of air conditioners) and differences in indoor/outdoor time ratios, may, in part, account for
April 2002 9-134 DRAFT-DO NOT QUOTE OR CITE
-------
1 any differences in risk estimates based on statical associations between ambient concentrations
2 and health effects for different cities or regions. If a were 0.3 in city A, but 0.6 in city B, and the
3 risks for an increase in personal exposure of 10 //g/m3 were identical, then a regression of health
4 effects on ambient concentrations would yield a health risk for city B that would be twice that
5 obtained for city A.
6 A number of exposure analysts have discussed the PM exposure paradox (i.e., that
7 epidemiology yields statistically significant associations between ambient concentrations and
8 health effects even though there is a near zero correlation between ambient concentrations and
9 personal exposure in many studies). Several explanations have been advanced to resolve this
10 paradox. First, personal exposure contains both an ambient-generated and a nonambient
11 component. Community time series epidemiology yields information only on the ambient-
12 generated component of exposure. Therefore, the appropriate correlation to investigate is the
13 correlation between ambient concentration and personal exposure to ambient-generated PM, not
14 between ambient concentrations and total personal exposure (i.e., the sum of ambient-generated
15 and nonambient PM). Second, biostatistical analysis of exposure error indicates that if the risk
16 function is linear in the PM indicator, the average of the sum of the individual risks (risk function
17 times individual exposure) may be replaced by the risk function times the community average
18 exposure. Thus, the appropriate correlation (of ambient concentrations and ambient-generated
19 exposure) is not the pooled correlation of different days and different people but the correlation
20 between the daily ambient concentrations and the community average daily personal exposure to
21 ambient-generated PM. Because the nonambient component is not a function of the ambient
22 concentration, its average will tend to be similar each day. Therefore, the correlation coefficient
23 will depend on a but not on the nonambient exposure. These types of correlation yield high
24 correlation coefficients.
25 A few studies have conducted simulation analyses of effects of measurement errors on the
26 estimated PM mortality effects. These studies suggest that ambient PM excess risk effects are
27 more likely underestimated than overestimated, and that spurious PM effects (i.e., qualitative
28 bias such as change in the sign of the coefficient) because of transferring of effects from other
29 covariates require extreme conditions and are therefore very unlikely. The error because the
30 difference between the average personal exposure and the ambient concentration is likely the
April 2002 9-135 DRAFT-DO NOT QUOTE OR CITE
-------
1 major source of bias in the estimated relative risk. One study also suggested that apparent linear
2 exposure-response curves are unlikely to be artifacts of measurement error.
3 In conclusion, for time-series epidemiology, ambient concentration is a useful surrogate for
4 personal exposure to ambient-generated PM, although the risk per unit ambient PM
5 concentration is biased low by the factor a compared to the risk per unit exposure to ambient-
6 generated PM. Epidemiologic studies of statistical associations between long-term effects and
7 long term ambient concentrations compare health outcome rates across cities with different
8 ambient concentrations. Ordinarily, PM exposure measurement errors are not expected to
9 influence the interpretation of findings from either the community time-series or long-term
10 epidemiologic studies that have used ambient concentration data if they include sufficient
11 adjustments for seasonality and key personal and geographic confounders. When individual level
12 health outcomes are measured in small cohorts, to reduce exposure misclassification errors, it is
13 essential that better real-time exposure monitoring techniques be used and that further speciation
14 of indoor-generated, ambient, and personal PM mass be accomplished. This should enable
15 measurement (or estimation) of both ambient and nonambient components of personal exposure
16 and evaluation of the extent to which personal exposure to ambient-generated PM, personal
17 exposure to nonambient PM, or total personal exposure (to ambient-generated plus nonambient
18 PM) contribute to observed health effects.
19
20 9.12.3 Coherence of Reported Epidemic logic Findings
21 Interrelationships Between Health Endpoints. Considerable coherence exists across
22 newly available epidemiologic study findings. For example, it was earlier noted that effects
23 estimates for total (nonaccidental) mortality generally fall in the range of 2.5 to 5.0% excess
24 deaths per 50 //g/m3 24-h PM10 increment. These estimates comport well with those found for
25 cause-specific cardiovascular- and respiratory-related mortality. Furthermore, larger effect sizes
26 for cardiovascular (in the range of 3 to 6% per 50 //g/m3 24-h PM10 increment) and respiratory (in
27 the range of 5 to 25% per 50 //g/m3 24-h PM10) hospital admissions and visits are found, as
28 would be expected versus those for PM10-related mortality. Also, several independent panel
29 studies, evaluating temporal associations between PM exposures and measures of heart beat
30 rhythm in elderly subjects, provide generally consistent indications of decreased heart rate (HR)
31 variability being associated with ambient PM exposure (decreased HR variability being an
April 2002 9-136 DRAFT-DO NOT QUOTE OR CITE
-------
1 indicator of increased risk for serious cardiovascular outcomes, e.g., heart attacks). Other studies
2 point toward changes in blood characteristics (e.g., increased C-reactive protein levels) related to
3 increased risk of ischemic heart disease as also being associated with ambient PM exposures.
4
5 Spatial Interrelationships. Both the NMMAPS and Cohort Reanalyses studies had a
6 sufficiently large number of cities to allow considerable resolution of regional PM effects within
7 the "lower 48" states, but this approach was taken much farther in the Cohort Reanalysis studies
8 than in NMMAPS. There were 88 cities with PM10 effect size estimates in NMMAPS; 50 cities
9 with PM25 and 151 cities with sulfates in Pope et al. (1995) and in the reanalyses using the
10 original data; and, in the additional analyses by the cohort study reanalysis team, 63 cities with
11 PM25 data and 144 cities with sulfate data. The relatively large number of data points allowed
12 estimation of surfaces for elevated long-term concentrations of PM2 5, sulfates, and SO2 with
13 resolution on a scale of a few tens to hundreds of kilometers. Information drawn from the maps
14 presented in Figures 16-21 in Krewski et al. (2000) is summarized below.
15 The patterns are similar, but not identical. In particular, the modeled PM25 surface
16 (Krewski, Figure 18) has peak levels in the industrial midwest, including the Chicago and
17 Cleveland areas, the upper Ohio River Valley, and around Birmingham, AL. Lower, but
18 elevated, PM25 is found almost everywhere else east of the Mississippi, as well as in southern
19 California. This is rather similar to the modeled sulfate surface (Krewski, Figure 16), with the
20 absence of a peak in Birmingham and an emerging sulfate peak in Atlanta. The only region with
21 elevated SO2 concentrations is the Cleveland-Pittsburgh area. A preliminary evaluation is that
22 secondary sulfates in particles derived from local SO2 is more likely to be important in the
23 industrial midwest, south from the Chicago-Gary region and along the upper Ohio River region.
24 This intriguing pattern may be related to the combustion of high-sulfur fuels in the subject areas.
25 The overlay of mortality and air pollution is also of interest. The spatial overlay of long-
26 term PM25 and mortality (Krewksi, Figure 21) is highest for the upper Ohio River region, but
27 also includes a significant association over most of the industrial midwest from Illinois to the
28 eastern noncoastal parts of North Carolina, Virginia, Pennsylvania, and New York. This is
29 reflected, in diminished form, by the sulfates map (Krewski, Figure 19) where the peak sulfate-
30 mortality associations occur somewhat east of the peak PM2 5-mortality associations. The SO2
31 map (Krewski, Figure 20) shows peak associations similar to, but slightly east of, the peak
April 2002 9-137 DRAFT-DO NOT QUOTE OR CITE
-------
1 sulfate associations. This suggests that, although SO2 may be an important precursor of sulfates
2 in this region, there may be other considerations (e.g., metals) in the association between PM2 5
3 and long-term mortality, embracing a wide area of the midwest and northeast (especially
4 noncoastal areas).
5 It should be noticed that, although a variety of spatial modeling approaches were discussed
6 in the NMMAPS methodology report (NMMAPS Part I, pp. 66-71), the primary spatial analyses
7 in the 90-city study (NMMAPS, Part II) were based on a simpler seven-region breakdown of the
8 contiguous 48 states. The 20-city results reported for the spatial model in NMMAPS I show a
9 much smaller posterior probability of a PM10 excess risk of short-term mortality, with a spatial
10 posterior probability versus a nonspatial probability of a PM10 effect of 0.89 versus 0.98 at lag 0,
11 of 0.92 versus 0.99 at lag 1, and of 0.85 versus 0.97 at lag 2. The evidence that PM10 is
12 associated with an excess short-term mortality risk is still moderately strong with a spatial model,
13 but much less strong than with a nonspatial model. In view of the sensitivity of the strength of
14 evidence to the spatial model, the model assumptions warrant additional study. Even so, there is
15 a considerable degree of coherence between the long-term and short-term mortality findings of
16 the studies, with stronger evidence of a modest but significant short-term PM10 effect and a larger
17 long-term fine particle (PM25 or sulfate) effect in the industrial midwest. The short-term effects
18 are larger but less certain in southern California and the northeast, whereas the long-term effects
19 seem less certain there.
20
21
22 9.13 EVALUATION OF STATISTICAL AND MEASUREMENT
23 ERROR ISSUES
24 9.13.1 Errors Related to Concentration, Exposure, and Dose
25 What Is the Effect of Measurement Error and Misclassification on Estimates of the
26 Association Between Air Pollution and Health?
27 In PM epidemiology, statistical models are developed that relate heath effects to some
28 measurement of ambient PM. However, if PM is toxic, the most direct relationship should be
29 between health effects and PM dose. Therefore, in PM epidemiology, ambient PM
30 concentrations must be considered a surrogate for PM dose. In going from ambient PM
31 concentrations to PM dose, there are many possibilities for introducing error or variability. This
April 2002 9-138 DRAFT-DO NOT QUOTE OR CITE
-------
1 section will discuss such possibilities and, to the extent information is available, the influence of
2 such errors on the variability in epidemiologic results.
3 Figure 9-32 shows an expanded version of the Risk Assessment Framework giving in more
4 detail the various processes involved in going from PM sources to PM dose. In Figure 9-32,
5 variables that can be measured directly are enclosed in hexagons; variables that cannot be
6 measured directly but can be estimated are enclosed in diamonds; and processes that influence
7 the relationship between ambient PM concentrations and PM dose are enclosed in ovals. There
8 are many opportunities for error in going from ambient concentrations to dose.
9
10 9.13.1.1 Opportunities for Error in the Use of Ambient PM Concentration as a
11 Surrogate for PM Dose in Epidemiologic Studies
12 Measurement ofPM Concentrations
13 As discussed in Chapter 2, Section 2.2.2.6, since there is no standard reference material that
14 can represent suspended PM, there cannot be any real determination of the accuracy with which
15 the concentration of suspended PM is measured. The precision of the measurement can be
16 determined by comparison of results from several collocated monitors. The mass of PM,
17 collected on a filter and equilibrated for 24 hours at 25 C and 40% relative humidity according to
18 the Federal Reference Method, can be measured with high precision. The precision of a
19 measurement of PM10_2 5 is normally less than that of PM2 5 but can be nearly as high if special
20 care is taken. The measurement of ultrafme PM (PM01) presents special problems and little is
21 known about the accuracy or precision of such measurements.
22 As discussed in Chapter 2, Section 2.2.2.1, a major problem in the measurement of PM10
23 and especially PM2 5, is the variable loss of semivolatile components of PM. The most important
24 are particle-bound water (PBW), ammonium nitrate, and semivolatile organic compounds.
25 During equilibration, much of the PBW is lost and the remainder is stable at the low, constant
26 relative humidity of equilibration. However, variable fractions of the other semivolatile
27 components are also lost during sampling or equilibration (Figure 9-33). For continuous
28 monitors, the collection surface must be changed at least every hour or the PM must be dried
29 in situ. Otherwise, changes in relative humidity will cause changes in the amount of PBW which
30 will cause large changes in perceived mass. Techniques which use heating to remove PBW may
31 also remove portions of the ammonium nitrate and semivolatile organic compounds which in
April 2002 9-139 DRAFT-DO NOT QUOTE OR CITE
-------
Source to Exposure
Cone, in Other
Ambient ues
Transport
Transformation
and removal
Sources of
Ambient PM
Gaseous Pre-
coursors of
Secondary PM
Ambient
Community
Cone.
Transport
Transformation
and removal
ue Outdoor at
Home, i.e.
Backyard ue
Cone.
ndoor Cone.
of Ambient-
Infiltrated
Modifications by
Indoor Chemistry
Sources of
Indoor-
Generated PM
Indoor Cone.
of Indoor-
Generated
PM
Secondary Indoor
Time-Activity Patterns, i.e
Time in Various Ambient
and Nonambient ues
Personal Cloud
or Personal
Activity PM
Total
Personal
Respiratory
Exposure
Exposure to Dose
Figure 9-32. An expanded version of the Risk Assessment Framework: (a) PM sources to
PM exposure, (b) PM exposure to PM dose.
April 2002
9-140
DRAFT-DO NOT QUOTE OR CITE
-------
Should be
retained
Particle-bound water
should be removed
(NH4)XS04
X= 0 to 2
Aerodynamic Diameter, |jm
::!: Semivolatile components subject to evaporation during or after sampling
Figure 9-33. Schematic showing major nonvolatile and semivolatile components of PM25.
Semivolatile components are subject to partial to complete loss during
equilibration or heating. The optimal technique would be to remove all
particle-bound water but no ammonium nitrate or semivolatile organic PM.
1 some cases could be on the order of 50% of the total suspended PM mass. Thus, current
2 measurements of PM mass have uncertainties and variability relative to the mass of PM
3 suspended in the atmosphere. Since most of the semivolatile components are in the
4 accumulation mode, this is a more serious problem for PM2 5 measurements than for PM10_2 5
5 measurements. As discussed in Chapter 2, Section 2.3, several new techniques are being tested
6 which may allow removal of PBW without loss of the semivolatile components of PM.
7 However, no such measurements of PM mass have been used in epidemiologic studies. Most
8 currently available epidemiologic studies used PM indicators which measure only the relatively
9 nonvolatile components of PM. Likewise, except for the new studies using in-situ concentrated
10 ambient air particles, most toxicologic studies of ambient air particles have used filter-collected
11 material which contains only the relatively nonvolatile components of PM. Therefore, little
12 information is available on the possible health effects of the semivolatile components of PM.
April 2002
9-141
DRAFT-DO NOT QUOTE OR CITE
-------
1 Errors Due to Inadequate Resolution of PMMeasurements by Size, Composition, and Source
2 Another source of error, with implications for epidemiology, is lumping together PM
3 components that behave differently with respect to processes that influence the relationship
4 between concentration and exposure, dose, and toxicity. This includes use of PM10
5 measurements instead of separate measurements of PM2 5 and PM10_2 5, measurement of PM mass
6 rather than individual chemical components, and measurement of mass instead of contributions
7 from specific source categories. Examples of the results obtained form improved resolution are
8 shown in Tables 9-20 through 9-22. As shown in Table 9-20, only two studies found a
9 statistically significant relationship (t > 1.96) for both PM25 and PM10_25. In most cases, only one
10 or the other size fraction was significant. However, in each case the most significant fraction
11 showed a higher % excess risk per//g/m3 and a greater t-statistic than was found for PM10. When
12 chemical components are treated separately (Table 9-21) the statistical significance may be
13 reduced compared to PM2 5, but the % excess risk per //g/m3 is higher than for PM2 5. Similarly,
14 when PM25 is split into orthogonal factors, representative of different source categories (Table
15 9-22), the % excess risk increases even though the t values are slightly smaller. The inclusion of
16 some fraction of coarse mode particles in PM2 5 may be a source of error in locations with high
17 and variable concentrations of thoracic coarse PM. A related error may occur from the use of 24-
18 hour average concentrations if the health effect is more closely related to peak dose than to
19 integrated dose.
20
21 Frequency of PM Measurements
22 Most epidemiologic studies have relied on monitoring data from existing networks, some
23 of which provide only every-6th-day monitoring data. This represents a loss of information
24 compared to having every day monitoring data. In addition, a level of uncertainty is introduced
25 into the estimation of the lag structure (the variation of PM effect as a function the number of
26 days between exposure and the observation of the health effect) since each lag day is based on a
27 different day of health effects. If everyday measurements of PM are available, each lag day is
28 based on health effects measured on the same day. The use of every-sixth day measurements
29 may also lead to errors in estimating annual and seasonal averages and distributions.
30
31
April 2002 9-142 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-20. PERCENT EXCESS RISK (t-statistic) PER 10 //g/m3
INCREASE IN PM FOR THE RELATIONSHIP OF VARIOUS INDICATORS
OF PM WITH VARIOUS TYPES OF MORTALITY (CV = cardiovascular)
IN SEVERAL DIFFERENT LOCATIONS. IN ONLY ONE CASE WERE
BOTH PM2 5 AND PM10 2 5 SIGNIFICANT. IN MOST CASES, THE MORE
SIGNIFICANT OF THE PM2 5 OR PM10 2 5 SIZE FRACTION HAD
A LARGER % EXCESS RISK AND T-STATISTIC THAN PM10.
Location
Mortality
PM10
PM25
PMin.,,
Phoenix1
CV
1.9(2.5)
7.1(2.9)
2.3 (2.5)
Mexico2
Total
1.8 (4.2)
1.5(1.9)
4.1 (5.0)
Mexico2 Santa Clara, Co3 Boston4
CV
2.0 (2.4)
1.6(1.1)
4.8 (4.4)
Total
1.6*
3.1**
I 5***
Total
1.3 (4.9)
2.2 (6.3)
0.2 (0.6)
Steubenville4
Total
0.9 (2.2)
1.0(1.8)
2.4 (2.4)
6-Cities4
Total
0.8(5.8)
1.5 (7.4)
0.4(1.5)
1. Mar et al., 2000. 2. Castillejos etal., 2000. 3. Fairley, 1999. 4. Swartzetal., 1996.
Significant at: *, p = 0.05; **, p = 0.01; ***, not significant.
TABLE 9-21. EXAMPLES OF HOW % EXCESS RISK PER 10 //g/m3 INCREASE IN
PM INDICATOR INCREASES FOR SPECIFIC CHEMICAL COMPONENTS OF PM.
IN THIS CASE, THE T-STATISTICS TEND TO BE LOWER.
Location Santa Clara, Co1 6-Cities2
PM25 3.1 1.5
Sulfate 17.4 2.2
Nitrate 8.8 —
1. Fairley, 1999. 2. Swartzetal., 1996.
TABLE 9-22. PERCENT EXCESS RISK (t-statistic) PER INTERQUARTILE
INCREASE IN PM INDICATOR FOR THE RELATIONSHIP OF VARIOUS
INDICATORS OF PM WITH CARDIOVASCULAR MORTALITY FOR PHOENIX
(Mar et al., 2000). FACTORS, ESTIMATED USING A FACTOR ANALYSIS SOURCE
APPORTIONMENT MODEL, ARE VEHICLE EXHAUST AND RESUSPENDED
ROAD DUST (vehicle), VEGETATIVE BURNING (wood), AND REGIONAL
SULFATE (R. SO4=). SOURCE CATEGORIES WERE SIGNIFICANT ON
DIFFERENT LAG DAYS.
PM25
Vehicle
Wood
R. SO4
Lag Day
1
1
O
0
%ER (t)
6.0 (2.9)
5.8 (2.6)
5.0 (2.7)
5.7 (2.0)
April 2002 9-143 DRAFT-DO NOT QUOTE OR CITE
-------
1 Spatial Variation
2 Most epidemiologic analyses assume that the PM concentration is uniform across the
3 spatial area in which health effects are measured or that the temporal variations in various parts
4 of the spatial area are highly correlated. As discussed in Chapter 3, Sections 3.2.5 and 3A, a lack
5 of uniformity may lead to error due to low site-to-site correlations between daily concentrations
6 or to spatial differences in long-term average concentrations. The site-to-site correlation is most
7 important for acute epidemiologic studies that relate daily concentrations to daily health effects.
8 Data from the PM2 5 monitoring network in 1999 and 2000 indicates relative high site-to-site
9 correlations in many cities. However, site-to-site correlations may not be as high for chemical
10 components or source category contributions. The small amount of data available suggest lower
11 site-to-site correlations for PM10_2 5. In some cities, where PM air pollution is heavily influenced
12 by local point sources, site-to-site correlations of PM25 may be low. Such cities may not provide
13 the best data for time-series epidemiology. Spatial differences in average concentration may be
14 more important for studies of the effects of long term exposure to PM on longevity or rates of
15 disease. Spatial inhomogeneity, as found in cities with local sources of primary PM25, may be
16 more important for health effects that are nonlinear with PM dose in the range of PM dose
17 experienced.
18
19 The Difference Between Ambient PM Concentration and Ambient PM Exposure
20 As discussed in Chapter 5, Section 5.3, there are two sources of variability in the
21 relationship between ambient concentrations and exposure to ambient PM (also called ambient
22 PM exposure). The indoor environment is protective, in that the concentration of ambient PM
23 indoors is generally less than the concentration of ambient PM outdoors. The relationship
24 depends on the particle size and on the air exchange rate. Thus, there will be differences between
25 ultra fine, fine, and thoracic coarse PM and between air-conditioned and un-air-conditioned
26 homes. The second source of variability is the fraction of time spent outdoors. These two
27 sources of variability are combined into the attenuation factor, a, the ratio of ambient exposure to
28 ambient concentration. The product of a and the ambient PM concentration, C, yields the
29 ambient PM exposure, A (i.e., A = a C) where a is different for the different particle-size
30 fractions. Since a may vary from person-to-person and time-to-time, due to variations in the air
31 exchange rate, the relationship between A and C may vary across the population and across
April 2002 9-144 DRAFT-DO NOT QUOTE OR CITE
-------
1 seasons. In spite of this variability, the correlation between A and C was high in the one study in
2 which individual daily values of A were estimated. Variations in a, as indexed by air
3 conditioning use, may explain some of the heterogeneity in excess rates of health effects
4 observed in epidemiologic studies of PM10 in different cities.
5
6 Seasonal Variations
1 Few epidemiological studies have had enough data for season-by-season analyses. Such
8 differences might be expected due to seasonal variations in the relative concentrations of
9 pollutants and PM components, the average a, correlations between ambient concentrations and
10 exposure, and correlations among potential surrogates and confounders. Most recent studies do
11 attempt to adjust for seasonal influences in their statistical models.
12
13 The Difference Between Ambient PM Exposure and Total PM Exposure
14 Total exposure to PM, as measured by a monitor worn by a person, is composed of an
15 ambient exposure component and a nonambient exposure component. The former includes
16 exposure to ambient pollution while outdoors and exposure to a fraction of the ambient pollution
17 while indoors. The latter is composed of primary and secondary indoor-generated PM and
18 personal cloud PM. The nonambient exposure is found to be variable from day-to-day for a
19 given subject and to be variable from subject-to-subject on a given day. However, the average
20 daily nonambient exposure may be relatively constant not only within a given city, but from city
21 to city within developed countries.
22
23 Ambient Concentration—Personal Exposure Relationships for Gaseous Co-Pollutants
24 Ambient concentrations of gaseous co-pollutants, such as CO, NO2, SO2 and O3, are
25 sometimes used in epidemiologic analyses. Only a few studies have examined the correlations of
26 ambient copollutant concentrations with (1) ambient PM concentrations, (2) personal exposures
27 to co-pollutants, and (3) either ambient or total personal PM exposure. These studies find that
28 the ambient concentrations of NO2, SO2, and O3 are not well correlated with personal exposure to
29 these gaseous co-pollutants. Rather, the concentrations of these gaseous co-pollutants and CO
30 are well correlated with the ambient concentrations of PM25, the ambient exposures to PM25, and
31 the total exposures to PM2 5. Therefore, these studies conclude that ambient concentrations of
April 2002 9-145 DRAFT-DO NOT QUOTE OR CITE
-------
1 NO2, SO2, O3, and likely CO, are not confounders but rather surrogates for ambient PM exposure
2 or more likely of ambient exposure to source categories with which the gases are correlated, i.e.,
3 NO2 and CO with motor vehicle associated PM and SO2 and O3 with regional sulfate.
4
5 The Difference Between Exposure and Dose
6 As discussed in Chapter 6 there are several causes of variability between exposure and
7 dose. The relationship between exposure and dose is highly dependent on particle size. Not only
8 total deposition, but also the location of deposition, varies with particle size, as shown in
9 Figure 9-34. The deposition fraction and location also depends on the size of the lung and the
10 breathing rate and is different for nose breathing and mouth breathing. Thus, deposition is higher
11 during exercise than normal activity. Also, children, with smaller lungs and higher breathing
12 rates than adults, will have higher deposition fractions than adults. Deposition fraction and
13 deposition location may be different in people with compromised lungs. Very importantly,
14 deposition per unit surface area may be higher in the healthy sections of their lungs. It is not
15 currently known which of the various deposition parameters are most important, i.e., deposition
16 could be estimated as mass per body mass, particle surface area per lung surface area, or number
17 of particles per number of alveolar cells. The importance of exercise in influencing dose was
18 demonstrated in a recent study of asthma. Exposure to O3 was related to increased prevalence of
19 asthma but only among children who participated in outdoor activities which involved exercise.
20
21 9.13.2 Possible Errors Related to Health and Epidemiology
22 Resolution of Health Effects
23 As of 2001, the majority of PM epidemiology data was based on the relationship between
24 PM10 mass and total mortality. However, it is possible that different kinds of particles may cause
25 different kinds of health effects and with different times between exposure and death. Thus,
26 lumping all nonaccidental deaths together may obscure useful information. Similar arguments
27 apply to morbidity. Some studies that consider classes of mortality tend to find higher excess
28 risks for cardiovascular and respiratory mortality than for total mortality.
29
30
April 2002 9-146 DRAFT-DO NOT QUOTE OR CITE
-------
.001 .002
.005 .01
.05 .1 .2 .512
Particle Diameter (|jm)
10 20
50 100
Figure 9-34. The percent deposition of inhaled particles in the tracheobronchial (TB) and
alveolar (A) regions of the lung as a function of particle size. The graph is
based on calculations using the ICRP model for a young adult with an
inhalation volume of 500 ml and a breathing frequency of 15 breaths a
second for spherical particles with a density of 1 g/ce.
1 Variation in Time Between Exposure and Appearance of Health Effects
2 Variations in toxicity of various types of PM or variations in the health status of members
3 of the exposed population may lead to variations in the time lag between exposure and
4 appearance of a response. Therefore, studies need to account for responses that may lag exposure
5 by several days. A dose of PM may also cause a health effect on more than one lag day. If so,
6 and if day to day concentrations are correlated, as is the usual case for PM, the use of only one
7 lag day will overestimate the risk on the lag day selected, due to autocorrelation, but will
8 underestimate the total risk. To obtain the total risk it is necessary to integrate the risk over
9 several days by a multiple regression model that accounts for health effects persisting for several
April 2002
9-147
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
days. This technique, with either a constrained or unconstrained lag structure, or using a running
average of daily concentrations, accounts for autocorrelation in the day-to-day concentrations and
leads to higher estimated excess risks.
9.13.3 Apportioning Health Effects to PM (by size, chemical component,
or source category) and Gaseous Co-Pollutants
One of the important technical problems in air pollution epidemiology is properly
apportioning health effects related to air pollution to the proper PM size fraction, chemical
component, or source component or to one or more gaseous co-pollutants. A major problem in
epidemiology is that a study may attribute an effect to a measured variable used as a regressor
when another measured or unmeasured variable is really the causal agent. The incorrect
attribution of effect (or part of the effect) to a variable used as a regressor, to another variable is
known as confounding. The potential for confounding exists anytime the concentration of a
causal agent is significantly correlated with the measured concentration of the regressor. The
proper apportionment of effect in air pollution epidemiology is difficult because PM and the
gaseous co-pollutants, NO2, CO, SO2, and O3 are often significantly correlated with each other.
The concepts of confounding; over-, under-, and mis-filtering, and effects modification are
discussed in Sections 8.1 and 8.4. Consider the relationship diagramed in Figure 9-35 for two air
pollutants, A and B. Lines with double arrows indicate a statistically significant association
between the two variables. There are many possible variations in the relationships among these
variables.
A
True Ambient
Concentration
1
B
True Ambient
Exposure
2
8
A
Measured
Ambient
Concentration
B
Measured
Ambient
Concentration
3
7
A
Exposure
B
Exposure
4
6
A
Outcome
5
B
Outcome
Figure 9-35. Diagram showing relationships (correlations) between A and B and between
various concentration, exposure, and outcome measures.
April 2002
9-148
DRAFT-DO NOT QUOTE OR CITE
-------
1 1. All relationships, paths 1-8, are significant.
2 la. In a multiple regression, A and B will share the health effects due to A and B. The
3 split will depend on the differential error in the paths between concentrations and outcomes.
4 Depending on the error structure and the relative strengths of the relationships, a portion of the
5 health effect due to the pollutant with the higher error will be transferred to the pollutant with
6 the lower error. It will not be possible to accurately apportion the health effects between A and B
7 (confounding).
8 Ib. If A is used as a single regressor, some of the effect of A will be transferred to B and
9 the effect of A will be overestimated (A is confounded by B, under-fitting).
10
11 2. Pollutant A does not cause the outcome of interest at the exposure level encountered.
12 Pathways 4 and 5 and outcome A disappear.
13 2a. Using B as a single regressor, the correct value is obtained for the association of B with
14 the outcomes due to B.
15 2b. Using A as a single regressor, a false positive value is obtained for the effect of A due
16 to the correlation of A with B (A is a surrogate for B, mis-fitting).
17 2c. If A and B are used in a multiple regression, some of the effect of B will be transferred
18 to A and the true effect of B will be underestimated (over-fitting).
19
20 3. Pollutants A and B are independent and cause independent health effects. Pathways 1
21 and 5 disappear. Since the concentration-outcome relationships are independently, an analysis
22 with either single or multiple regressors will give the correct association for each pollutant.
23 Situation 3 is the desirable situation.
24
25 Unfortunately, it is not possible, on the basis of one epidemiologic study in one community
26 during one time period, to tell whether A or B is responsible for the health effects; or if both are
27 responsible, to correctly apportion the effects to each. To correctly apportion the health effects
28 between A and B it is necessary to seek other sources of information.
29
30 Toxicity. If we know that the potential confounder, A in situation 2, does not cause
31 outcome A at the levels of exposure, we know that a single regression with B as the regressor
April 2002 9-149 DRAFT-DO NOT QUOTE OR CITE
-------
1 will yield the correct value for the association of B with outcome B (Situation 2a). However, a
2 multiple regression using A and B would underestimate the association of outcome B with B
3 because some of the health effects due to B would be incorrectly transferred to A (Situation 2c,
4 over-fitting). Thus, if we know that A does not cause the outcome at the exposure level of A,
5 then the health effect may confidently be assigned to B (realizing that there could always be an
6 unknown variable C, correlated with B, that is really the causal agent).
7
8 Exposure. Useful information can also be obtained from concentration-exposure
9 relationships. Suppose that A and B are significantly correlated but that the ambient
10 concentration of A is not significantly correlated with either the ambient, nonambient, or total
11 personal exposure to A. This can occur, as discussed in Chapter 5 and Section 9.6.4, if the
12 spatial distribution of A is inhomogeneous of if A is very rapidly removed once it is penetrates
13 indoors. In this case pathway 4 disappears. Even is A is capable of causing the outcome at the
14 levels of exposure, a regression using A will not show any association because the exposure
15 causing the outcomes are not correlated with the concentration used as the regressor.
16
17 Lag structure. If the effect due to A peaks on the day of exposure (lag day 0) but the
18 effect of B peaks on the day after exposure (lag day 1), we can conclude that both A and B cause
19 independent effects. This conclusion holds only if the concentration of A is not correlated with
20 the concentration of B on the prior day.
21
22 Multiple regression. In a number of studies the effects attributed to A (PA, the slope of the
23 regressions of A on outcomes) and B (PB) in a multiple regression are compared to those in single
24 regressions to obtain information on possible confounding. An increase in the variance of PAM
25 and PBM, from the multiple regression, over those PAS and PBS, obtained from single regression; a
26 reduction in the values of either PAM and PBM compared to PAS and PBS, or a decrease in the value
27 of P for the combined action of A and B relative to sum of PAS and PBS is evidence for potential
28 confounding. However, this information is not sufficient to determine whether A is a confounder
29 of B or whether A is a surrogate for B. If, PBM equals PBS and PAM = 0, it may be assumed that B
30 is causal and A is not. In order for this to happen, the correlation between A and B would have
31 to be non-significant. Thus, if A and B are uncorrelated, PA is zero or non-significant, and PB is
April 2002 9-150 DRAFT-DO NOT QUOTE OR CITE
-------
1 positive and significant, a multiple regression, yielding similar values of PA and PB, will confirm
2 that A is noncausal and B is causal. If the correlation is between 0 and 1, the multiple regression
3 may provide some idea of how much of the effect of A could be due to potential confounding by
4 B, but it cannot determine whether the confounding is real or only potential. If a community is
5 studied where the correlation of A and B is low or nonsignificant, a comparison of single and
6 multiple regression can demonstrate that significant confounding is not occurring if that is the
7 case.
8
9 Orthogonal regressors. Various types of source apportion models have been developed to
10 assist implementation of PM standards by identifying the sources of PM in a given airshed. The
11 process involves application of statistical techniques such a factor analysis to daily concentration
12 values of PM components to generate orthogonal factors, containing various loadings of PM
13 components and in some cases, also containing gaseous co-pollutants. In some case, these
14 factors can be identified with specific source categories. This source category factor (SCF) can
15 be used to determine the daily contributions of these source factors to the PM concentration.
16 Since the SCF are orthogonal (i.e., independent and uncorrelated) we have situation 3 and either
17 single regressions with each source factor or a multiple regression with all SCF should give
18 correct values of the relationship of each SCF to the health outcomes with which it is related.
19 There will be no potential for confounding since the SCF are uncorrelated.
20 The concept of SCF can help explain some of the results from multiple regression of PM
21 and a gaseous co-pollutant. Consider the situation shown in Figure 9-36. The vehicular
22 traffic-related (VTR) SCF and the regional sulfate (RSO4) SCF are uncorrelated and we will
23 assume they are causal. The VTR SCF contains contributions from CO, NO2, and PM2 5. The
24 RSO4 SCF contain contributions from regional sulfate in the form of (NH4)2SO4, NH4H2SO4, and
25 H2SO4, but not local sulfate in the form of CaSO4. PM2 5 is correlated with both SCFs but CO
26 and NO2 are only correlated with the VTR SCF.
27 Now considered a multiple regression using PM10 and CO (or NO2) as variables. Since
28 PM2 5 contains both RSO4 and VTR components, it is possible that the ambient concentrations of
29 CO (or NO2) will be more highly correlated than PM2 5 with the ambient concentrations of the
30 VTR SCF. Thus, in a multiple regression, the effects of the VTR SCF would be transferred
31 largely to CO (or NO2) and only the effects of RSO4 would be transferred to the PM2 5. Thus, CO
April 2002 9-151 DRAFT-DO NOT QUOTE OR CITE
-------
True Ambient
Concentration
N<~>2 nr 4
\ CO *
TT
\ \ Motor
\ \ Vehicle
\ Related PM
"/ (MVR)
\ /
X Regional
/ Sulfate (RS)
* PM *
Measured
Ambient
Concentration
* * \
?
/
Exposure
/\ b 4
\
\ * *
Outcomes
Outcome NO2 or
i
CO
Outcome MVR
Outcome RS
i
Figure 9-36. Diagram showing concentrations—exposure—outcome relationships
(correlations for CO or NO2, PM2 5, and source category factors for vehicular
traffic related PM and regional sulfate.
1
2
3
4
5
6
1
8
9
10
11
12
13
(or NO2) might show a higher relative risk than PM2 5 (or possibly than PM2 5 in a single
regression) and the relative risk associated with PM2 5 would be reduced in the multiple
regression. In this case, CO (or NO2) would be confounded by the VTR SCF.
While a regression with SCFs will give excess risk values, unconfounded by other SCFs,
it will not be possible to tell from epidemiology alone whether the CO, the NO2, or the PM
component of the VTR pollution is truly causal. However, in view of the anticipated low
correlation between CO and NO2 with their respective personal exposures, and the unlikelihood
that CO or NO2 cause acute mortality as the very low values of personal exposures for most of
the population, the identification of the PM component of VTR pollution as the most likely
causal agent is reasonable.
April 2002
9-152
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.14 IMPLICATIONS OF HEALTH EFFECTS OF LONG-TERM
2 EXPOSURES TO PARTICULATE MATTER
3 What are the implications of observed effects of long-term exposure to paniculate matter
4 and other pollutants for life expectancy?
5
6 9.14.1 Methodological Issues
7 Closed-cohort studies of ambient air pollutants are methodologically similar to typical
8 epidemiological studies of occupational cohorts and, in some respects, to experimental trials.
9 Subjects are enrolled, characterized as to their exposures and other relevant health factors, and
10 followed over time as they experience adverse health outcomes. Methodological issues
11 regarding the loss of subjects to follow-up, the movement of subjects between exposure groups
12 or levels, and the characterization of exposure are well-understood and are adequately handled by
13 standard epidemiologic methods.
14 The assignment of exposure in both environmental and occupational studies is generally
15 based on area rather than personal sampling and any consequential exposure misclassification
16 will generally bias effect estimates towards the null. With appropriate individual-level
17 assessment and analysis of other risk factors, the assignment of a common exposure to a group
18 does not give raise to an ecological fallacy (Kunzli and Tager, 1997). The current PM AQCD
19 has avoided a reliance on purely ecological analyses of county-level data that lack
20 individual-level data on non-environmental determinants of mortality.
21 A key difference between epidemiologic closed-cohorts studies and experimental trials is
22 the lack of randomization of subjects to exposure. In observational studies, randomization is
23 replaced by a careful consideration and analytic correction for differences in other salient health
24 factors other than the exposure of interest. A potential confounder must be (a) an independent
25 determinant of the outcome of interest among unexposed subjects, (b) non-causally associated
26 with the exposure of interest, and (c) not a part of the causal pathway linking the exposure and
27 outcome. Once these potential confounders have been controlled, differences in survival, that is,
28 in the relative rates of mortality, are attributed to differences in subjects' exposure histories.
29
30
April 2002 9-153 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.14.2 Overall Survival and Life Expectancy
2 Our current knowledge of the adverse health effects of long-term exposures to ambient
3 particulate matter is based on a small number of epidemiological studies that compare differences
4 in the survival of well-characterized closed-cohorts of free-living human subjects with air
5 pollution levels in their cities of residence (AQCD Section 8.2.3). Compared with the more
6 intricate methodological aspects of epidemiological studies of short-term particulate matter
7 exposures using time-series methods to examine non-enumerated open-cohorts, the design,
8 conduct and analysis of closed-cohorts is straightforward. However, such survival studies are
9 much less common due to the difficulty and expense of enrolling and maintaining follow-up of
10 an enumerated cohort.
11 At the time of the 1996 PM AQCD, three closed-cohort (survival) studies of particulate
12 matter had been published in the peer-review literature. Two of these survival studies were
13 national in scope, the Harvard Six-Cities Adult Cohort Study (Dockery et al., 1993) and the
14 American Cancer Society Cohort Study (Pope et al., 1995), and one focused solely on California,
15 the Adventist Health Study of Smog or AHSMOG (Abbey et al., 1991). The American Cancer
16 Society Cohort Study was a secondary analysis of a very extensive cohort of 552,138 subjects in
17 151 cities whose exposures were characterized by routinely collected air quality data and who
18 were followed for seven years. The Harvard Six-Cities Adult Cohort Study enrolled 8,111
19 subjects in six cities, characterized their exposures with investigator-conducted measurements of
20 size-fractionated particulate matter, and followed these subjects for 14 to 16 years. The
21 Adventist Health Study on Smog enrolled 6,340 non-smoking subjects, grouped into three major
22 urban areas and the remainder of California, whose exposures were characterized by routinely
23 collected air quality data, and who were followed for an average of 10 years.
24 The two national studies found strong associations between higher particulate matter levels
25 and decreased survival. For non-external causes of mortality, a 25 //g/m3 increment in PM25 was
26 associated with increases in the rate of mortality: 36 percent in the Harvard Six-Cities Adult
27 Study and 18 percent in the American Cancer Society Cohort Study. The California study did
28 not initially find any statistically significant overall mortality effects.
29 After the 1996 PM AQCD was completed, concerns were expressed regarding the adequacy
30 of the conduct and analysis of these survival studies of particulate matter (Gamble, 1998). Many
April 2002 9-154 DRAFT-DO NOT QUOTE OR CITE
-------
1 concerns related to standard methodological issues regarding the assessment of exposure,
2 geographic mobility, and adequacy of control for potential confounders.
3
4 9.14.3 Verification and Sensitivity Analyses
5 Since the 1996 PM AQCD, the two national studies have been critically reanalyzed by
6 independent researchers under the auspices of the Health Effects Institute (Krewski et al., 2000).
7 In addition to the replication and validation of the original findings of the Harvard Six-Cities
8 Adult Study, Krewski et al. considered the sensitivity of the original findings to alternative risk
9 models and analytic approaches. Generally this sensitivity analysis found that the original results
10 were robust to changes in model specification and the inclusion of other community-level
11 covariates. Both the original and the reanalyses found a 13% increase in risk of mortality per
12 10 //g/m3 increment in PM25. The HEI reanalysis project both confirmed and extended the
13 results of the American Cancer Society Cohort Study. Both the reanalysis and the extension
14 found a 7% increase in risk of mortality per 10 //g/m3 in PM2 5.
15 Since the conclusion of the reanalysis project, these three survival cohorts have been
16 extended by the original investigators to additional years of follow-up and alternative exposure
17 measures. Using airport visibility records to estimate exposures to PM25, the Adventist Health
18 Study on Smog reported an 8.5 percent increase in the rate of non-external mortality associated
19 with a 10 //g/m3 increment in PM25 (McDonnell et al., 2000).
20 Thus, the relative risk estimates for these three survival cohorts have converged in the
21 range of 7 to 13 percent increase in the non-external mortality rate associated with a 10 //g/m3
22 increment in a long-term average of PM2 5. Methodological criticisms of these studies have been
23 largely resolved in favor of the validity of their original findings of a strong association between
24 long-term exposures to particulate matter and decreased survival (Bates, 2000).
25
26 9.14.4 Impact on Life-Expectancy
27 The increased rate of non-external mortality found in these three survival cohorts is greater
28 than the mere accumulation of the adverse effects of short-term exposures for a few days.
29 Conceptually, particulate matter may be associated with both the long-term development of
30 underlying health problems ("Frailty") and with the short-term variations in timing of mortality
April 2002 9-155 DRAFT-DO NOT QUOTE OR CITE
-------
1 among a susceptible population with some underlying health condition (Kunzli et al. 2001).
2 Epidemiologic studies of the mortality effects of short-term exposure to particulate matter using
3 unenumerated open-cohorts ("time-series studies") can only capture particulate matter's
4 association with short-term variations in mortality and, therefore, must systematically
5 underestimate the proportion of total mortality attributable to particulate matter. A recent
6 time-series study that examined the contribution of daily particulate matter levels over an
7 extended lag period (42 days) could only partially bridge the gap between the effects of
8 short-term and long-term exposures to particulate matter (Zanobetti et al., 2002).
9 Recent investigations of the public health implications of effect estimates for long-term PM
10 exposures also were reviewed in Chapter 8. Life table calculations by Brunekreef (1997) found
11 that relatively small differences in long-term exposure to airborne PM of ambient origin can have
12 substantial effects on life expectancy. For example, a calculation for the 1969 to 1971 life table
13 for U.S. white males indicated that a chronic exposure increase of 10 //g/m3 PM was associated
14 with a reduction of 1.31 years for the entire population's life expectancy at age 25. The new
15 evidence noted above of infant mortality associations with PM exposure suggests that life
16 shortening in the entire population from long-term PM exposure could well be significantly
17 larger than estimated by Brunekreef (1997).
18
19 9.14.5 Specific Causes of Death
20 The increase in non-external mortality cannot be explained by increases in chronic
21 respiratory diseases since chronic non-malignant lower respiratory disease accounts for only
22 5.6 percent and lung cancer for only another 6.9 percent of all deaths over age 24 years due to
23 non-external causes. Cardiovascular diseases, which account for 43 percent of non-external
24 mortality, must play the leading role in the decreased survival associated with exposure to
25 ambient PM. It is nevertheless useful to highlight the newer results of the extension of the ACS
26 study analyses (that include more years of participant follow-up and address previous criticisms
27 of the earlier ACS analyses), which provide the strongest evidence to date that long-term ambient
28 PM exposures are associated with increased risk of lung cancer. That increased risk appears to
29 be in about the same range as that seen for a non-smoker residing with a smoker and, therefore,
30 passively exposed chronically to tobacco smoke, with any consequent life-shortening impacts
31 due to lung cancer.
April 2002 9-156 DRAFT-DO NOT QUOTE OR CITE
-------
REFERENCES
Abbey, D. E.; Ostro, B. E.; Petersen, F.; Burchette, R. J. (1995) Chronic respiratory symptoms associated with
estimated long-term ambient concentrations of fine particulates less than 2.5 microns in aerodynamic
diameter (PM2 5) and other air pollutants. J. Exposure Anal. Environ. Epidemiol. 5: 137-159.
Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
158: 289-298.
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
Care Med. 159: 373-382.
Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
Survey, 1996. Vital Health Stat. 10(200).
Bigger, J. T., Jr.; Fleiss, J. L.; Steinman, R. C.; Rolnitzky, L. M.; Kleiger, R. E.; Rottman, J. N. (1992) Frequency
domain measures of heart period variability and mortality after myocardial infarction. Circulation
85: 164-171.
Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
Broday, D. M.; Georgopoulos, P. G. (2000) Growth and deposition of hygroscopic paniculate matter in the human
lungs. Aerosol Sci. Technol.: submitted.
Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M.;
Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of
respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
142: 15-22.
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105: 614-620.
Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
Ozkaynak, H. (1998) The association between ambient carbon monoxide levels and daily mortality in
Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12(suppl. 4): 15-39.
Castillejos, M.; Borja-Aburto, V. H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
mortality. In: Inhalation Toxicology: proceedings of the third colloquium on paniculate air pollution and
human health; June, 1999; Durham, NC. Inhalation Toxicology 12(suppl. 1): 61-72.
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52: 113-117.
Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
50: 1287-1298.
Claiborn, C. S.; Finn, D.; Larson, T. V.; Koenig, J. Q. (2000) Windblown dust contributes to high PM2 5
concentrations. J. Air Waste Manage. Assoc. 50: 1440-1445.
April 2002 9-157 DRAFT-DO NOT QUOTE OR CITE
-------
Clark, W. E.; Whitby, K. T. (1975) Measurements of aerosols produced by the photochemical oxidation of SO2 in
air. J. Colloid Interface Sci. 51: 477-490.
Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
Code of Federal Regulations. (1991a) Appendix J to Part 50-reference method for the determination of paniculate
matter as PM10 in the atmosphere. C. F. R. 40: §50.
Code of Federal Regulations. (1991b) Ambient air monitoring reference and equivalent methods. C. F. R. 40: §53.
Costa, D. L.; Dreher, K. L. (1999) What do we need to know about airborne particles to make effective risk
management decisions? A toxicology perspective. Hum. Ecol. Risk Assess. 5: 481-491.
Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
152: 397-406.
Dejmek, J.; Selevan, S. G.; Benes, L; Solansky, L; Sram, R. J. (1999) Fetal growth and maternal exposure to
paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998) Symptoms inpediatric asthmatics and air pollution:
differences in effects by symptom severity, anti-inflammatory medication use and particulate averaging time.
Environ. Health Perspect. 106: 751-761.
Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
Dockery, D. W.; Pope, C. A., Ill; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
(1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
329: 1753-1759.
Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
symptoms. Environ. Health Perspect. 104: 500-505.
Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
saturation and pulse rate associated with particulate air pollution and barometric pressure. Cambridge, MA:
Health Effects Institute; research report no. 83.
Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
limitations. Toxicol. Sci. 55: 24-35.
Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000) Endotoxin priming affects the lung
response to ultrafine particles and ozone in young and old rats. In: Inhalation Toxicology: proceedings of the
third colloquium on particulate air pollution and human health; June, 1999; Durham, NC. Inhalation
Toxicology 12(suppl. 1): 85-98.
Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
Perspect. 107: 637-641.
Fanucchi, M. V.; Plopper, C. G. (1997) Pulmonary developmental responses to toxicants. In: Roth, R. A., ed.
Toxicology of the Respiratory System, v. 8 of Comprehensive Toxicology. New York, NY: Pergamon,
p. 203-220.
Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B. K.; Van Winkle, L. S.; Evans, M. J.; Plopper, C. G. (2000)
Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
rhesus monkeys. Am. J. Respir. Crit. Care Med. 161: A615.
Federal Register. (1987) Revisions to the national ambient air quality standards for particulate matter. F. R. (July 1)
52: 24,634-24,669.
Fitz-Simons, T. S.; Mathias, M.; Rizzo, M. (2000) Analyses of 1999 PM data for the PM NAAQS review.
U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards; September 29.
Gardner, S. Y.; Lehmann, J. R.; Costa, D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in
rats. Toxicol. Sci. 56: 175-180.
Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
April 2002 9-158 DRAFT-DO NOT QUOTE OR CITE
-------
Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. I; Brunekreef, B. (1997) Acute effects of
summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
155:2105-2108.
Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
53: 817-822.
Godleski, J. J.; Sioutas, C.; Katler, M.; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
Godleski, J. J.; et al. (1998) Increased cardiac vulnerability during exposure to inhaled environmental particles
[abstract]. Presented at: Health Effects Institute annual meeting; May; Boston, MA. Boston, MA: Health
Effects Institute.
Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms of morbidity and mortality from
exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
Gold, A.; Litonjua, J.; Schwartz, M.; Verrier, R.; Milstein, A.; Larson, E.; Lovett, B. (1998) Cardiovascular
vulnerability to paniculate pollution. Presented at: 1998 International conference [American Thoracic
Society]; April; Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A261.
Gold, D. R.; Damokosh, A. I.; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
Castillejos, M. (1999) Particulate and ozone pollutant effects on the respiratory function of children in
southwest Mexico City. Epidemiology 10: 8-16.
Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm [15
February, 2001].
Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
(1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
Inhalation Toxicol. 11: 981-998.
Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
Hayano, J.; Sakakibara, Y.; Yamada, M.; Ohte, N.; Fujinami, T.; Yokoyama, K.; Watanabe, Y.; Takata, K. (1990)
Decreased magnitude of heart rate spectral components in coronary artery disease. Its relation to
angiographic severity. Circulation 81: 1217-1224.
Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
Hoyert, D. L.; Kochanek, K. D.; Murphy, S. L. (1999) Deaths: final data for 1997. Atlanta, GA: U.S. Centers for
Disease Control and Prevention, National Center for Health Statistics. (National vital statistics reports: v. 47,
no. 19).
Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
Inhalation Toxicol. 10: 121-130.
Kamens, R.; Jang, M.; Chien, C.-J.; Leach, K. (1999) Aerosol formation from the reaction of a-pinene and ozone
using a gas-phase kinetics-aerosol partitioning model. Environ. Sci. Technol. 33: 1430-1438.
Keywood, M. D.; Ayers, G. P.; Gras, J. L.; Gillett, R. W.; Cohen, D. D. (1999) Relationships between size
segregated mass concentration data and ultrafine particle number concentrations in urban areas.
Atmos. Environ. 33: 2907-2913.
Kim, Y. J.; Boatman, J. F.; Gunter, R. L.; Wellman, D. L.; Wilkison, S. W. (1993) Vertical distribution of
atmospheric aerosol size distribution over south-central New Mexico. Atmos. Environ. Part A
27: 1351-1362.
Kittelson, D. B. (1998) Engines and nanoparticles: a review. J. Aerosol Sci. 29: 575-588.
Kleiger, R. E.; Miller, J. P.; Bigger J. T. Jr.; Moss, A. J.; Multicenter Post-infarction Research Group. (1987)
Decreased heart rate variability and its association with increased mortality after acute myocardial infarction.
Am. J. Cardiol. 59: 256-262.
April 2002 9-159 DRAFT-DO NOT QUOTE OR CITE
-------
Kleinman, M. T.; Leaf, D. A.; Kelly, E.; Caiozzo, V.; Osann, K.; O'Niell, T. (1998) Urban angina in the mountains:
effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
Health 53: 388-397.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
and daily mortality statistical modeling-interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
50: 1215-1222.
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
W. H.. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of
paniculate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
Project. Cambridge, MA: Health Effects Institute.
Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
Perspect. 105: 1234-1237.
Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
Lawrence, L.; Hall, M. J. (1999) 1997 summary: national hospital discharge survey. Hyattsville, MD: U.S.
Department of Health & Human Resources, Centers for Disease Control and Prevention, National Center for
Health Statistics. (Advance data from vital and health statistics: no. 308).
Layton, D. W. (1993) Metabolically consistent breathing rates for use in dose assessments. Health Phys. 64: 23-36.
Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
paniculate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000b) Infant mortality and air pollution: a comprehensive analysis of U.S.
data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambndge, MA: Health Effects Institute; research report
no. 95.
Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
mortality in Mexico City. Epidemiology 10: 118-123.
Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
J. R. Stat. Soc. B 61(part 2): 459-477.
Lundgren, D. A.; Burton, R. M. (1995) Effect of particle size distribution on the cut point between fine and coarse
ambient mass fractions. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air
pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 131-148.
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
Martin, G. J.; Magid, N. M.; Myers, G.; Barnett, P. S.; Schaad, J. W.; Weiss, J. S.; Lesch, M.; Singer, D. H. (1987)
Heart rate variability and sudden death secondary to coronary artery disease durtng ambulatory
electrocardiographic monitoring. Am. J. Cardiol. 60: 86-89.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
California children with asthma. Environ. Health Perspect. 107: 757-760.
Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
April 2002 9-160 DRAFT-DO NOT QUOTE OR CITE
-------
Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
exposed to fine (PM2 5) and coarse particles (PM10.2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
155: 245-252.
Moolgavkar, S. H. (2000a) Air pollution and daily mortality in three U.S. counties. Environ. Health Perspect.
108: 777-784.
Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health.
Inhalation Toxicol. 12(suppl. 4): 75-90.
Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and
admissions for chronic respiratory disease in King County, Washington. In: Inhalation Toxicology:
proceedings of the third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC.
Inhalation Toxicology 12(suppl. 1): 157-171.
Morawska, L.; Thomas, S.; Bofinger, N.; Wainwright, D.; Neale, D. (1998) Comprehensive characterization of
aerosols in a subtropical urban atmosphere: particle size distribution and correlation with gaseous pollutants.
Atmos. Environ. 32: 2467-2478.
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
Healthy women's PEF variations with ambient summer concentrations of PM10, PN2 5, SO42_, H+, and O3. Am.
J. Respir. Crit. Care Med. 160: 117-125.
Nakamura, T.; Hayashida, Y. (1992) Autonomic cardiovascular responses to smoke exposure in conscious rats. Am.
J. Physiol. 262(5 pt. 2): R73 8-745.
National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
publication no. 97-4051.
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
and exposure to air pollution. Public Health Rep. 114: 135-148.
Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
139: 1088-1099.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
141: 111-122.
Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and
other determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
Philadelphia. Environ. Health Perspect. 107: 629-631.
Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
paniculate material in rat and human lungs. Presented at: PM2000: paniculate matter and health—the
scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
Pittsburgh, PA: Air & Waste Management Association.
Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
Perspect. 107: 489-493.
Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
stagnant air. Thorax 55: 466-470.
Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Branson, K.; Jakab, G. J.; Kleeberger, S. R. (2000) Interstrain variation in
murine susceptibility to inhaled acid-coated particles. Am. J. Physiol. L469-L476.
April 2002 9-161 DRAFT-DO NOT QUOTE OR CITE
-------
Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols. Am.
J. Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Krupnick, A.; Harrington, W. (1993) Air pollution and respiratory
morbidity among adults in Southern California. Am. J. Epidemiol. 137: 691-700.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, California: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10: 412-419.
Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
Environmental Health; March 45.
Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
Perry, K. D.; Cahill, T. A.; Eldred, R. A.; Dutcher, D. D.; Gill, T. E. (1997) Long-range transport of North African
dust to the eastern United States. J. Geophys. Res. [Atmos] 102: 11,225-11,238.
Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
episode: a link to mortality? Lancet 349: 1582-1587.
Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b) Respiratory effects are associated with the
number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c) Short-term effects of paniculate air pollution on
respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities with
differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
Med. 159: 760-767.
Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
159: 768-775.
Plopper, C. G.; Fanucchi, M. V. (2000) Do urban environmental pollutants exacerbate childhood lung diseases?
Environ. Health Perspect. 108: A252-A253.
Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
Respir. Crit. Care Med. 151: 669-674.
Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
Front. Environ. Health Perspect.: 107: 567-573.
Pope, C. A.; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
G. M.; Gold, D. R.; Dockery, D. W. (1999b) Heart rate variability associated with paniculate air pollution.
Am. Heart J. 138: 890-899.
Raizenne, M.; Neas, L. M.; Damokosh, A. L; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
Health Perspect. 104: 506-514.
Ramadan, Z.; Song, X.-H.; Hopke, P. K. (2000) Identification of sources of Phoenix aerosol by positive matrix
factorization. J. Air Waste Manage. Assoc. 50: 1308-1320.
Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
154: 300-307.
Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
Mexico City. Arch. Environ. Health 52: 368-376.
Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (1999) Acute inflammatory
responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.
April 2002 9-162 DRAFT-DO NOT QUOTE OR CITE
-------
Samet, J. M; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and
air pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute;
research report no. 94.
Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
(2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
Schlesinger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
emissions. In: Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
health. Washington, DC: National Academy Press; pp. 239-298.
Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
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. (2000) Harvesting and long term exposure effects in the relation between air pollution and mortality.
Am. J. Epidemiol. 151: 440-448.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
Michigan. Am. J. Epidemiol. 142: 23-35.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
with application to air pollution and daily death. Epidemiology 11: 666-672.
Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
Ferris, E.G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
Am. J. Respir. Crit. Care Med. 150: 1234-1242.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles?
J. Air Waste Manage. Assoc. 46: 927-939.
Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Singer, D. H.; Martin, G. J.; Magid, N.; Weiss, J. S.; Schaad, J. W.; Kehoe, R.; Zheutiin, T.; Fintel, D. J.; Hsieh,
A. M.; Lesch, M. (1988) Low heart rate variability and sudden cardiac death. J. Electrocardiol.
21(suppl.): S46-S55.
Smiley-Jewell, S. M.; Liu, F. J.; Weir, A. J.; Plopper, C. G. (2000) Acute injury to differentiating Clara cells in
neonatal rabbits results in age-related failure of bronchiolar epithelial repair. Toxicol. Pathol. 28: 267-276.
Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
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. Circulation 93: 1043-1065.
Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; VanEeden, S. F. (1997) Phagocytosis of small carbon particles
(PM10) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
Am. J. Respir. Crit. Care Med. 155: 1441-1447.
Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers.
J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
April 2002 9-163 DRAFT-DO NOT QUOTE OR CITE
-------
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
65: 271-290.
Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
Tolbert, P. E.; Klein, M.; Metzger, K. B.; Peel, J.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
Frumkin, H. (2000) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure
Anal. Environ. Epidemiol. 10: 446-460.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
Turpin, B. J. (1999) Options for characterizing organic paniculate matter. Environ. Sci. Technol. 33: 76A-79A.
U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/00laF-cF. 3v.
U.S. Environmental Protection Agency. (2000a) Air quality criteria for carbon monoxide. Research Triangle Park,
NC: National Center for Environmental Assessment; report no. EPA/600/P-99/001F. Available:
www.epa.gov/ncea/co/ [2000, October 6].
U.S. Environmental Protection Agency. (2000b) Health assessment document for diesel exhaust [external review
draft]. Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria
and Assessment Office; report nos. EPA/600/8-90/057E.
Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
Respir. Dis. 133:834-842.
Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U.P.; Highfill, J. W.; Costa, D. L. (2000)
Thermoregulatory effects following exposure to paniculate matter in healthy and
cardiopulmonary-compromisedrats. J. Therm. Biol. 25: 131-137.
Weschler, C. J.; Shields, H. C. (1999) Indoor ozone/terpene reactions as a source of indoor particles. Atmos.
Environ. 33:2301-2312.
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
Whitby, K. T.; Sverdrup, G. M. (1980) California aerosols: their physical and chemical characteristics. In: Hidy,
G. M.; Mueller, P. K.; Grosjean, D.; Appel, B. R.; Wesolowski, J. J., eds. The character and origins of smog
aerosols: a digest of results from the California Aerosol Characterization Experiment (ACHEX). New York,
NY: John Wiley & Sons, Inc.; pp. 477-517. (Advances in environmental science and technology: v. 9).
Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
Cambridge, MA: Health Effects Institute; Research Report no. 98.
Williams, R.; Suggs, J.; Zweidinger, R.; Evans, G.; Creason, J.; Kwok, R.; Rodes, C.; Lawless, P.; Sheldon, L.
(2000a) The 1998 Baltimore Paniculate Matter Epidemiology-Exposure Study: part 1. Comparison of
ambient, residential outdoor, indoor and apartment paniculate matter monitoring. J. Exposure Anal. Environ.
Epidemiol. 10: 518-532.
Williams, R.; Suggs, J.; Creason, J.; Rodes, C.; Lawless, P.; Kwok, R.; Zweidinger, R.; Sheldon, L. (2000b) The
1998 Baltimore Paniculate Matter Epidemiology-Exposure Study: part 2. Personal exposure assessment
associated with an elderly study population. J. Exposure Anal. Environ. Epidemiol. 10: 533-543.
Williams, R. W.; Creason, J.; Zweidinger, R.; Watts, R. R.; Shy, C. (2000c) Indoor, outdoor, and personal exposure
monitoring of paniculate air pollution: the Baltimore elderly epidemiology-exposure pilot study.
Atmos. Environ. 34: 4193-4204.
Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal
infant mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
April 2002 9-164 DRAFT-DO NOT QUOTE OR CITE
-------
Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
J. Occup. Environ. Med. 42: 469-474.
Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
Zhang, H.; Triche, E.; Leaderer, B. (2000) Model for the analysis of binary time series of respiratory symptoms.
Am. J. Epidemiol. 151: 1206-1215.
April 2002 9-165 DRAFT-DO NOT QUOTE OR CITE
-------
APPENDIX 9A
Key Quantitative Estimates of Relative Risk for Particulate Matter-Related
Health Effects Based on Epidemiologic Studies of U.S. and Canadian Cities
Assessed in the 1996 Particulate Matter Air Quality Criteria Document
April 2002 9A-1 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-1. EFFECT ESTIMATES PER 50-^g/m3 INCREASE
IN 24-HOUR PM,n CONCENTRATIONS FROM U.S. AND CANADIAN STUDIES
Study Location
RR (±CI)
Only PM
in Model
RR (±CI) Reported
Other Pollutants PM10 Levels
in Model Mean (Min/Max)*
Increased Total Acute Mortality
Six Cities"
Portage, WI
Boston, MA
Topeka, KS
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
St. Louis, MOC
Kingston, TNC
Chicago, ILh
Chicago, ILg
Utah Valley, UTb
Birmingham, ALd
Los Angeles, CAf
Increased Hospital Admissions (for Elderly
Respiratory Disease
Toronto, Canada1
Tacoma, WAJ
New Haven, CTJ
Cleveland, OHk
Spokane, WA1
COPD
Minneapolis, MNn
Birmingham, ALm
Spokane, WA1
Detroit, MI°
1.04 (0.98, 1.09)
1.06(1.04, 1.09)
0.98 (0.90, 1.05)
1.03 (1.00, 1.05)
1.05 (1.00, 1.09)
1.05 (1.00, 1.08)
1.08(1.01, 1.12)
1.09 (0.94, 1.25)
1.04 (1.00, 1.08)
1.03 (1.02, 1.04)
1.08(1.05, 1.11)
1.05(1.01, 1.10)
1.03 (1.00, 1.055)
> 65 years)
1.23 (1.02, 1.43)*
1.10(1.03, 1.17)
1.06(1.00, 1.13)
1.06(1.00, 1.11)
1.08(1.04, 1.14)
1.25(1.10, 1.44)
1.13(1.04, 1.22)
1.17(1.08, 1.27)
1.10(1.02, 1.17)
—
— 18 (±11.7)
— 24 (±12.8)
— 27 (±16.1)
— 31 (±16.2)
— 32 (±14.5)
— 46 (±32.3)
1.06(0.98,1.15) 28(1/97)
1.09 (0.94, 1.26 30 (4/67)
— 37 (4/365)
1.02 (1.01, 1.04) 38 (NR/128)
1.19(0.96,1.47) 47(11/297)
— 48 (21, 80)
1.02 (0.99, 1.036) 58( 15/177)
1.12(0.88,1.36)* 30-39*
1.11(1.02,1.20) 37(14,67)
1.07(1.01,1.14) 41(19,67)
— 43 (19, 72)
— 46 (16, 83)
— 36 (18, 58)
— 45 (19, 77)
— 46 (16, 83)
— 48 (22, 82)
April 2002
9A-2
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-1 (cont'd). EFFECT ESTIMATES PER 50-^g/m3 INCREASE
IN 24-HOUR PMnCONCENTRATIONS FROM U.S. AND CANADIAN STUDIES
Study Location
Pneumonia
Minneapolis, MNn
Birmingham, ALm
Spokane, WA1
Detroit, MI°
Ischemic HP
Detroit, MP
RR (±CI)
OnlyPM
in Model
1.08(1.01, 1.15)
1.09(1.03, 1.15)
1.06(0.98, 1.13)
—
1.02(1.01, 1.03)
RR (±CI) Reported
Other Pollutants PM10 Levels
in Model Mean (Min/Max)1
— 36 (18,58)
— 45 (19, 77)
— 46 (16, 83)
1.06(1.02,1.10) 48(22,82)
1.02 (1.00, 1.03) 48 (22, 82)
Increased Respiratory Symptoms
Lower Respiratory
Six Cities5
Utah Valley, UTr
Utah Valley, UTS
Cough
Denver, COX
Six Cities5
Utah Valley, UTS
Decrease in Lung Function
Utah Valley, UTr
Utah Valley, UTS
Utah Valley, UTW
2.03 (1.36, 3.04)
1.28(1.06, 1.56)T
1.01 (0.81, 1.27)71
1.27 (1.08, 1.49)
1.09(0.57,2.10)
1.51(1.12,2.05)
1.29(1.12, 1.48)
55 (24, 86)**
30 (10, 50)**
29(7,51)***
Similar RR 30(13,53)
— 46(11/195)
— 76(7/251)
— 22 (0.5/73)
Similar RR 30(13,53)
— 76(7/251)
— 46(11/195)
— 76(7/251)
— 55(1,181)
References:
"Schwartz etal.(1996a).
bPopeetal. (1992, 1994)/O3.
'Dockery et al. (1992)/O3.
•"Schwartz (1993).
8Ito and Thurston (1996)/O3.
fKinney et al. (1995)/O3, CO.
hStyeretal. (1995).
'Thurston et al. (1994)/O3.
jSchwartz (1995)/SO2.
"•Schwartz et al. (1996b).
'Schwartz (1996).
"Schwartz (1994a).
"Schwartz (1994b).
"Schwartz (1994c).
"Schwartz and Morris (1995)/O3, CO, SO2.
"Schwartz etal. (1994).
Tope etal. (1991).
Tope and Dockery (1992).
'Schwartz (1994d).
Tope and Kanner (1993).
"Ostro etal. (1991).
fMin/Max 24-h PM10 in parentheses unless noted
otherwise as standard deviation (±SD), 10 and
90 percentile (10, 90). NR = not reported.
"Children.
"Asthmatic children and adults.
'Means of several cities.
"PEFR decrease in mlVs.
"FEV! decrease.
*RR refers to total population, not just >65 years.
April 2002
9A-3
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-2. EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM2 5, SO^, H+)
FROM U.S. AND CANADIAN STUDIES
Acute Mortality
Six City3
Portage, WI
Topeka, KS
Boston, MA
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
Increased Hospitalization
Ontario, Canadab
Ontario, Canada0
NYC/Buffalo, NYd
Torontod I
Increased Respiratory Symptoms
Southern Californiaf
Six Cities8
(Cough)
Six Cities8
(Lower Resp. Symp.)
Decreased Lung Function
Uniontown, PAe
Indicator
PM25
PM25
PM25
PM25
PM25
PM,,
so;
so;
03
so;
f (Nmol/m3)
so;
PM,,
so;
PM25
PM2 5 Sulfur
H+
PM25
PM2 5 Sulfur
H+
PM25
RR (±CI) per 25 i^g/m3
PM Increase
1.030(0.993, 1.071)
1.020(0.951, 1.092)
1.056(1.038, 1.0711)
1.028(1.010, 1.043)
1.035 (1.005, 1.066)
1.025 (0.998, 1.053)
1.03 (1.02, 1.04)
1.03 (1.02, 1.04)
1.03 (1.02, 1.05)
1.05(1.01, 1.10)
1.16(1.03, 1.30)*
1.12(1.00, 1.24)
1.15 (1.02, 1.78)
1.48(1.14, 1.91)
.19(1.01, 1.42)"
.23 (0.95, 1.59)"
.06 (0.87, 1.29)"
.44(1.15-1.82)**
.82 (1.28-2.59)**
.05 (0.25-1.30)**
PEFR 23.1 (-0.3, 36.9) (per 25 ,wg/m3)
Reported PM
Levels Mean
(Mm/Max)1
11.2 (±7.8)
12.2 (±7.4)
15.7 (±9.2)
18.7 (±10.5)
20.8 (±9.6)
29.6 (±21.9)
R = 3. 1-8.2
R = 2.0-7.7
NR
28.8 (NR/391)
7.6 (NR, 48.7)
18.6 (NR, 66.0)
R = 2-37
18.0 (7.2, 37)"*
2.5(3.1,61)*"
18.1 (0.8,5.9)*"
18.0 (7.2, 37)"*
2.5 (0.8, 5.9)"*
18.1(3.1,61)"*
25/88 (NR/88)
References:
"Schwartz etal. (1996a).
bBurnettetal. (1994).
cBurnettetal. (1995) O3.
dThurston et al. (1992, 1994).
dNeas et al. (1995).
fOstro etal. (1993).
BSchwartzetal. (1994).
fMin/Max 24-h PM indicator level shown in parentheses unless
otherwise noted as (±SD), 10 and 90 percentile (10,90) or
R = range of values from min-max, no mean value reported.
'Change per 100 nmoles/m3
"Change per 20 ^g/m3 for PM2 5; per 5 ^g/m3 for PM2 5 sulfur;
per 25 nmoles/m3 for H+.
***50th percentile value (10,90 percentile).
April 2002
9A-4
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-3. EFFECT ESTIMATES PER INCREMENTS3 IN
ANNUAL MEAN LEVELS OF FINE PARTICLE INDICATORS FROM
U.S. AND CANADIAN STUDIES
Type of Health
Effect and Location
Increased Total Chronic
Six Cityb
ACS Study0
(151 U.S. SMSA)
Indicator
Mortality in Adults
PM15/10
PM25
so;
PM25
so:
Increased Bronchitis in Children
Six Cityd
Six City6
24 Cityf
24 Cityf
24 Cityf
24 Cityf
Southern California8
PM15/10
TSP
H+
so;
PM21
PM10
so:
Change in Health Indicator per
Increment in PMa
Relative Risk (95% CI)
1.42(1.16-2.01)
1.31 (1.11-1.68)
1.46(1.16-2.16)
1.17(1.09-1.26)
1.10(1.06-1.16)
Odds Ratio (95% CI)
3.26(1.13, 10.28)
2.80(1.17,7.03)
2.65 (1.22, 5.74)
3.02(1.28,7.03)
1.97(0.85,4.51)
3.29(0.81, 13.62)
1.39 (0.99, 1.92)
Range of City
PM Levels
Means (,wg/m3)
18-47
11-30
5-13
9-34
4-24
20-59
39-114
6.2-41.0
18.1-67.3
9.1-17.3
22.0-28.6
—
Decreased Lung Function in Children
Six City4h
Six City6
24 Citylj
24 City1
24 City1
24 City1
PM15/10
TSP
H+ (52 nmoles/m3)
PM2 ! (15 Mg/m3)
SOI (7 Mg/m3)
PM10(17Mg/m3)
NS Changes
NS Changes
-3. 45% (-4.87, -2.01)FVC
-3.21% (-4.98, -1.41) FVC
-3.06% (-4.50, -1.60) FVC
-2.42% (-4.30, -.0.51) FVC
20-59
39-114
—
—
—
—
"Estimates calculated annual-average PM increments assume: a 100-^g/m3 increase for TSP; a 50-^g/m3
increase for PM10 and PM15; a 25-^g/m3 increase for PM2 5; and a 15-^g/m3 increase for SO:, except where
noted otherwise; a 100-nmole/m3 increase forH+.
bDockeryetal. (1993).
Tope etal. (1995).
dDockery et al. (1989).
6Wareetal. (1986).
TJockery et al. (1996).
BAbbey et al. (1995).
hNS Changes = No significant changes.
'Raizenne et al. (1996).
JPollutant data same as for Dockery et al. (1996).
April 2002
9A-5
DRAFT-DO NOT QUOTE OR CITE
-------
REFERENCES
Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
Adventists. Environ. Health Perspect. 94: 43-50.
Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
7: 19-34.
Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
2.5 microns in aerodynamic diameter (PM2.5) from airport visibility data in California. J. Exposure Anal.
Environ. Epidemiol. 5: 161-180.
Abbey, D. E.; Ostro, B. E.; Petersen, F.; Burchette, R. J. (1995c) Chronic respiratory symptoms associated with
estimated long-term ambient concentrations of fine particulates less than 2.5 microns in aerodynamic
diameter (PM2 5) and other air pollutants. J. Exposure Anal. Environ. Epidemiol. 5: 137-159.
Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
158: 289-298.
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
Care Med. 159: 373-382.
Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
Survey, 1996. Vital Health Stat. 10(200).
American Conference of Governmental Industrial Hygienists. (1994) Appendix D: particle size-selective sampling
criteria for airborne paniculate matter. In: 1994-1995 Threshold limit values for chemical substances and
physical agents and biological exposure indices. Cincinnati, OH: American Conference of Governmental
Industrial Hygienists; pp. 43-46.
Bates, D. V. (2000) Lines that connect: assessing the causality inference in the case of paniculate pollution.
Environ. Health Perspect. 108: 91-92.
Becker, S.; Soukup, J. M.; Gilmour, M. I.; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
Pharmacol. 141:637-648.
Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
Binkova, B.; Vesely, D.; Vesela, D.; Jelinek, R.; Sram, R. J. (1999) Genetoxicity and embryotoxicity of urban air
paniculate matter collected during winter and summer period in two different districts of the Czech Republic.
Mutat. Res. 440: 45-58.
Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
Brook, R. D.; Brook, J. R.; Urch, B.; Vincent, R.; Rajagopalan, S.; Silverman, F. (2002) Inhalation of fine
paniculate air pollution and ozone causes acute arterial vasoconstriction in health adults. Circulation 105:
1534-1536.
Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M.;
Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of
respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
142: 15-22.
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105: 614-620.
April 2002 9A-6 DRAFT-DO NOT QUOTE OR CITE
-------
Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
Ozkaynak, H. (1998) The association between ambient carbon monoxide levels and daily mortality in
Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12(suppl. 4): 15-39.
Calvert, J. G.; Heywood, J. B.; Sawyer, R. F.; Seinfeld, J. H. (1993) Achieving acceptable air quality: some
reflections on controlling vehicle emissions. Science (Washington, DC) 261: 37-45.
Castillejos, M.; Borja-Aburto, V. H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
mortality. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the third colloquium on paniculate air
pollution and human health (first special issue); June, 1999; Durham, NC. Inhalation Toxicol.
12(suppl. 1): 61-72.
Chapman, R. S.; Watkinson, W. P.; Dreher, K. L.; Costa, D. L. (1997) Ambient paniculate matter and respiratory
and cardiovascular illness in adults: particle-borne transition metals and the heart-lung axis. Environ.
Toxicol. Pharmacol. 4: 331-338.
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
Chor^zy, M.; Szeliga, J.; Strozyk, M.; Cimander, B. (1994) Ambient air pollutants in Upper Silesia: partial chemical
composition and biological activity. In: Symposium of risk assessment of urban air: emissions, exposure, risk
identification, and risk quantitation; May-June 1992; Stockholm, Sweden. Environ. Health Perspect.
102(suppl4):61-66.
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52: 113-117.
Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
50: 1287-1298.
Claiborn, C. S.; Finn, D.; Larson, T. V.; Koenig, J. Q. (2000) Windblown dust contributes to high PM25
concentrations. J. Air Waste Manage. Assoc. 50: 1440-1445.
Clarke, R. W.; Antonini, J. M.; Hemenway, D. R.; Frank, R.; Kleeberger, S. R.; Jakab, G. J. (2000) Inhaled
particle-bound sulfate: effects on pulmonary inflammatory responses and alveolar macrophage function.
Inhalation Toxicol. 12: 169-186.
Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
Code of Federal Regulations. (200 la) Appendix J to Part 50~reference method for the determination of paniculate
matter as PM10 in the atmosphere. C. F. R. 40: §50.
Code of Federal Regulations. (200Ib) Appendix L to Part 50~reference method for the determination of fine
paniculate matter as PM2.5 in the atmosphere. C. F. R. 40: §50.
Costa, D. L.; Dreher, K. L. (1999) What do we need to know about airborne particles to make effective risk
management decisions? A toxicology perspective. Hum. Ecol. Risk Assess. 5: 481-491.
Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
152: 397-406.
Dejmek, J.; Selevan, S. G.; Benes, I.; Solansky, I.; Sram, R. J. (1999) Fetal growth and maternal exposure to
paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998) Symptoms inpediatric asthmatics and air pollution:
differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging time.
Environ. Health Perspect. 106: 751-761.
April 2002 9A-7 DRAFT-DO NOT QUOTE OR CITE
-------
Dockery, D. W.; Spengler, J. D. (1981) Personal exposure to respirable particulates and sulfates. J. AirPollut.
Control Assoc. 31: 153-159.
Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
Dockery, D. W.; Pope, C. A., Ill; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
(1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
329: 1753-1759.
Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
symptoms. Environ. Health Perspect. 104: 500-505.
Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
Health Effects Institute; research report no. 83.
Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor "alpha" expression from
alveolar macrophages treated with urban air particles. Exp. Lung Res. 22: 577-592.
Dreher, K. L. (2000) Paniculate matter physicochemistry and toxicology: in search of causality—a critical
perspective. Inhalation Toxicol. 12(suppl. 3): 45-57.
Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
limitations. Toxicol. Sci. 55: 24-35.
Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. J.; Costa, D. L.
(2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
studies in Utah Valley residents. Environ. Health Perspect. 109(suppl. 3): 395-403.
Ebelt, S. T.; Petkau, A. J.; Vedal, S.; Fisher, T. V.; Brauer, M. (2000) Exposure of chronic obstructive pulmonary
disease patients to paniculate matter: relationships between personal and ambient air concentrations. J. Air
Waste Manage. Assoc. 50: 1081-1094.
Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000) Endotoxin priming affects the lung
response to ultrafine particles and ozone in young and old rats. In: Inhalation Toxicology: proceedings of the
third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC. Inhalation
Toxicology 12(suppl. 1): 85-98.
Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
Perspect. 107: 637-641.
Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B. K.; Van Winkle, L. S.; Evans, M. J.; Plopper, C. G. (2000)
Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
rhesus monkeys. Am. J. Respir. Crit. Care Med. 161: A615.
Federal Register. (1987) Revisions to the national ambient air quality standards for paniculate matter. F. R. (July 1)
52: 24,634-24,669.
Federal Register. (1997) National ambient air quality standards for paniculate matter; final rule. F. R. (July 18)
62: 38,652-38,752.
Fitz-Simons, T. S.; Mathias, M.; Rizzo, M. (2000) Analyses of 1999 PM data for the PM NAAQS review.
U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards; September 29.
Fortoul, T. I.; Osorio, L. S.; Tovar, A. T.; Salazar, D.; Castilla, M. E.; Olaiz-Fernandez, G. (1996) Metals in lung
tissue from autopsy cases in Mexico City residents: comparison of cases from the 1950s and the 1980s.
Environ. Health Perspect. 104: 630-632.
Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol. 277:
L960-L967.
Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98-MP26.03.
Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
Am. J. Respir. Crit. Care Med. 164: 704-708.
April 2002 9A-8 DRAFT-DO NOT QUOTE OR CITE
-------
Ohio, A. J.; Kim, C.; Devlin, R. B. (2000) Concentrated ambient air particles induce mild pulmonary inflammation
in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
Ohio, A. J.; Gilbey, J. G.; Roggli, V. L.; Richards, J. H.; McGee, J. K.; Carson, J. L.; Devlin, R. B.; Cascio, W. E.
(2001) Diffuse alveolar damage after exposure to an oil fly ash. Am. J. Respir. Crit. Care Med. 164:
1514-1518.
Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
155:2105-2108.
Godleski, J. J.; Sioutas, C.; Katler, M.; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
Gold, D. R.; Damokosh, A. L; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
Castillejos, M. (1999) Paniculate and ozone pollutant effects on the respiratory function of children in
southwest Mexico City. Epidemiology 10: 8-16.
Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm
[15 February, 2001].
Guenther, A.; Geron, C.; Pierce, T.; Lamb, B.; Harley, P.; Fall, R. (2000) Natural emissions of non-methane volatile
organic compounds, carbon monoxide, and oxides of nitrogen from North America. Atmos. Environ.
34: 2205-2230.
Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
Henry, R. C. (1997) History and fundamentals of multivariate air quality receptor models. Chemom. Intell. Lab.
Syst. 37: 37-42.
Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
Hoyert, D. L.; Kochanek, K. D.; Murphy, S. L. (1999) Deaths: final data for 1997. Atlanta, GA: U.S. Centers for
Disease Control and Prevention, National Center for Health Statistics. (National vital statistics reports: v. 47,
no. 19).
Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
Inhalation Toxicol. 10: 121-130.
Imrich, A.; Ning, Y. Y.; Koziel, H.; Coull, B.; Kobzik, L. (1999) Lipopolysaccharide priming amplifies lung
macrophage tumor necrosis factor production in response to air particles. Toxicol. Appl. Pharmacol.
159: 117-124.
Izzotti, A.; Camoirano, A.; DAgostini, F.; Sciacca, S.; De Naro Papa, F.; Cesarone, C. F.; De Flora, S. (1996)
Biomarker alterations produced in rat lung by intratracheal instillations of air paniculate extracts and
chemoprevention with oral N-acetyIcysteine. Cancer Res. 56: 1533-1538.
Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
Janssen, N. A. H.; Schwartz, J.; Zanobetti, A.; Suh, H. H. (2002) Air conditioning and source-specific particles as
modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ. Health Perspect.
110:43-49.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
and daily mortality statistical modeling-interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
50: 1215-1222.
Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Watkinson, W. P.; Campen, M. J.; Winsett, D. W.;
Richards, J. R.; Crissman, K. M.; Hatch, G. E.; Costa, D. L. (2000) The spontaneously hypertensive rat as a
model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress
from inhaled emission paniculate matter. Toxicol. Appl. Pharmacol. 164: 250-263.
April 2002 9A-9 DRAFT-DO NOT QUOTE OR CITE
-------
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M; White,
W. H.. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of
paniculate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
Project. Cambridge, MA: Health Effects Institute.
Ktinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared
to ecologic studies. Environ. HealthPerspect. 105: 1078-1083.
Ktinzli, N.; Medina, S.; Kaiser, R.; Quenel, P.; Horak, F., Jr.; Studnicka, M. (2001) Assessment of deaths
attributable to air pollution: should we use risk estimates based on time series or on cohort studies? Am. J.
Epidemiol. 153: 1050-1055.
Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
Perspect. 105: 1234-1237.
Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
Lambert, A. L.; Dong, W.; Selgrade, M. K.; Gilmour, M. I. (2000) Enhanced allergic sensitization by residual oil fly
ash particles is mediated by soluble metal constituents. Toxicol. Appl. Pharmacol. 165: 84-93.
Landis, M. S.; Norris, G. A.; Williams, R. W.; Weinstein, J. P. (2001) Personal exposures to PM2 5 mass and trace
elements in Baltimore, MD, USA. Atmos. Environ. 35: 6511-6524.
Lawrence, L.; Hall, M. J. (1999) 1997 summary: national hospital discharge survey. Hyattsville, MD: U.S.
Department of Health & Human Resources, Centers for Disease Control and Prevention, National Center for
Health Statistics. (Advance data from vital and health statistics: no. 308).
Layton, D. W. (1993) Metabolically consistent breathing rates for use in dose assessments. Health Phys. 64: 23-36.
Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
paniculate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
Li, X. Y.; Brown, D.; Smith, S.; MacNee, W.; Donaldson, K. (1999) Short-term inflammatory responses following
intratracheal instillation of fine and ultrafine carbon black in rats. Inhalation Toxicol. 11: 709-731.
Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lioy, P. J. (1990) Assessing total human exposure to contaminants: a multidisciplinary approach. Environ. Sci.
Technol. 24: 938-945.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000b) Infant mortality and air pollution: a comprehensive analysis of U.S.
data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report
no. 95.
Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
Long, C. M.; Suh, H. H.; Koutrakis, P. (2000) Characterization of indoor particle sources using continuous mass
and size monitors. J. Air Waste Manage. Assoc. 50: 1236-1250.
Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
mortality in Mexico City. Epidemiology 10: 118-123.
Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
J. R. Stat. Soc. B 61(part 2): 459-477.
Lundgren, D. A.; Burton, R. M. (1995) Effect of particle size distribution on the cut point between fine and coarse
ambient mass fractions. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air
pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 131-148.
April 2002 9A-10 DRAFT-DO NOT QUOTE OR CITE
-------
Mage, D.; Wilson, W.; Hasselblad, V.; Grant, L. (1999) Assessment of human exposure to ambient paniculate
matter. J. Air Waste Manage. Assoc. 49: 174-185.
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
California children with asthma. Environ. Health Perspect. 107: 757-760.
McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of
mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers.
J. Exposure Anal. Environ. Epidemiol. 10: 427-436.
Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
Monn, C. (2001) Exposure assessment of air pollutants: a review on spatial heterogeneity and
indoor/outdoor/personal exposure to suspended paniculate matter, nitrogen dioxide and ozone. Atmos.
Environ. 35: 1-32.
Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
155: 245-252.
Moolgavkar, S. H. (2000a) Air pollution and daily mortality in three U.S. counties. Environ. Health Perspect.
108: 777-784.
Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health.
Inhalation Toxicol. 12(suppl. 4): 75-90.
Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and
admissions for chronic respiratory disease in King County, Washington. In: Inhalation Toxicology:
proceedings of the third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC.
Inhalation Toxicology 12(suppl. 1): 157-171.
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
Healthy women's PEF variations with ambient summer concentrations of PM10, PN2 5, SO42_, H+, and O3.
Am. J. Respir. Crit. Care Med. 160: 117-125.
NARSTO. (2002) Fine particle assessment (2nd review draft). Available: http://www.cgenv.com/Narsto/ (13
February 2002).
National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
publication no. 97-4051.
National Research Council. (1983) Acid deposition: atmospheric processes in eastern North America, a review of
current scientific understanding. Washington, DC: National Academy Press.
National Research Council. (1994) Science and judgment in risk assessment. Washington, DC: National Academy
Press.
National Research Council. (1998) Research priorities for airborne paniculate matter. I. Immediate priorities and a
long-range research portfolio. Washington, DC: National Academy Press.
National Research Council. (2001) Research priorities for airborne paniculate matter. III. Early research progress.
Washington, DC: National Academy Press. Available: http://www.nap.edu/catalog/10065.html [30 April,
2002].
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
and exposure to air pollution. Public Health Rep. 114: 135-148.
Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
139: 1088-1099.
April 2002 9A-11 DRAFT-DO NOT QUOTE OR CITE
-------
Neas, L. M; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
141: 111-122.
Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and
other determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
Philadelphia. Environ. Health Perspect. 107: 629-631.
Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
paniculate material in rat and human lungs. Presented at: PM2000: paniculate matter and health—the
scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
Pittsburgh, PA: Air & Waste Management Association.
Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
Perspect. 107: 489-493.
Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
stagnant air. Thorax 55: 466-470.
Oberdorster, G.; Ferin, J.; Lehnert, B. E. (1994) Correlation between particle size, in vivo particle persistence, and
lung injury. Environ. Health Perspect. 102(suppl. 5): 173-179.
Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
Fundam. Appl. Toxicol. 40: 220-227.
Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
70: 98-104.
Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
Am. J. Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Krupnick, A.; Harrington, W. (1993) Air pollution and respiratory
morbidity among adults in Southern California. Am. J. Epidemiol. 137: 691-700.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, California: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10: 412-419.
Ott, W.; Wallace, L.; Mage, D. (2000) Predicting paniculate (PM10) personal exposure distributions using a random
component superposition statistical model. J. Air Waste Manage. Assoc. 50: 1390-1406.
Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
Environmental Health; March 45.
Patterson, E.; Eatough, D. J. (2000) Indoor/outdoor relationships for ambient PM2 5 and associated pollutants:
epidemiological implications inLindon, Utah. J. Air Waste Manage. Assoc. 50: 103-110.
Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
episode: a link to mortality? Lancet 349: 1582-1587.
Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b) Respiratory effects are associated with the
number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c) Short-term effects of paniculate air pollution on
respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities with
differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
159: 768-775.
Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
Med. 159: 760-767.
Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
Monahan, K.; Dockery, D. W. (2000) Air pollution and incidence of cardiac arrhythmia. Epidemiology 11:
11-17.
April 2002 9A-12 DRAFT-DO NOT QUOTE OR CITE
-------
Pinto, J. P.; Pleasant, M; Kellogg, R.; Torok, Sz.; Stiles, D.; Willis, R. D.; Koutrakis, P.; Allen, G. A. (1995)
Chemical characterization of the ambient aerosol in Philadelphia. Presented at: Paniculate matter: health and
regulatory issues: an A&WMA international specialty conference; April; Pittsburgh, PA. Pittsburgh, PA: Air
& Waste Management Association; submitted abstract.
Pinto, J. P.; Lefohn, A. S.; Shadwick, D. S. (2002) Aspects of the spatial variability of PM25 concentrations within
urban areas of the United States. Environ. Sci. Technol.: submitted.
Plopper, C. G.; Fanucchi, M. V. (2000) Do urban environmental pollutants exacerbate childhood lung diseases?
Environ. Health Perspect. 108: A252-A253.
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
Respir. Crit. Care Med. 151: 669-674.
Pope, C. A.; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
G. M.; Gold, D. R.; Dockery, D. W. (1999a) Heart rate variability associated with paniculate air pollution.
Am. Heart J. 138: 890-899.
Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999b) Paniculate air pollution and daily mortality on Utah's Wasatch
Front. Environ. Health Perspect.: 107: 567-573.
Pope, C. A., Ill; Eatough, D. J.; Gold, D. R.; Pang, Y.; Nielsen, K. R.; Nath, P.; Verrier, R. L.; Kanner, R. E. (2001)
Acute exposure to environmental tobacco smoke and heart rate variability. Environ. Health Perspect.
109:711-716.
Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
Assoc. 287: 1132-1141.
Raizenne, M.; Neas, L. M.; Damokosh, A. L; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
Health Perspect. 104: 506-514.
Ramadan, Z.; Song, X.-H.; Hopke, P. K. (2000) Identification of sources of Phoenix aerosol by positive matrix
factorization. J. Air Waste Manage. Assoc. 50: 1308-1320.
Ritz, B.; Yu, F.; Fruin, S.; Chapa, G.; Shaw, G. M.; Harris, J. A. (2002) Ambient air pollution and risk of birth
defects in southern California. Am. J. Epidemiol. 155: 17-25.
Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
154: 300-307.
Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
Mexico City. Arch. Environ. Health 52: 368-376.
Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and
air pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute;
research report no. 94.
Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, I.; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
(2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
Samet, J. M.; Hatch, G. E.; Horstman, D.; Steck-Scott, S.; Arab, L.; Bromberg, P. A.; Levine, M.; McDonnell,
W. F.; Devlin, R. B. (2001) Effect of antioxidant supplementation on ozone-induced lung injury in human
subjects. Am. J. Respir. Crit. Care Med. 164: 819-825.
Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
Sarnat, J. A.; Schwartz, J.; Catalano, P. J.; Suh, H. H. (2001) Gaseous pollutants in paniculate matter epidemiology:
confounders or surrogates? Environ. Health Perspect. 109: 1053-1061.
Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
8: 371-377.
April 2002 9A-13 DRAFT-DO NOT QUOTE OR CITE
-------
Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
10: 17-22.
Schwartz, J. (2000) Harvesting and long term exposure effects in the relation between air pollution and mortality.
Am. J. Epidemiol. 151: 440-448.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
Michigan. Am. J. Epidemiol. 142: 23-35.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
with application to air pollution and daily death. Epidemiology 11: 666-672.
Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
Ferris, E.G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
Am. J. Respir. Crit. Care Med. 150: 1234-1242.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles?
J. Air Waste Manage. Assoc. 46: 927-939.
Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
Sexton, K.; Selevan, S. G.; Wagener, D. K.; Lybarger, J. A. (1992) Estimating human exposures to environmental
pollutants: availability and utility of existing databases. Arch. Environ. Health 47: 398-407.
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Smiley-Jewell, S. M.; Liu, F. J.; Weir, A. J.; Plopper, C. G. (2000) Acute injury to differentiating Clara cells in
neonatal rabbits results in age-related failure of bronchiolar epithelial repair. Toxicol. Pathol. 28: 267-276.
Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers.
J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
65: 271-290.
Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine particulate air pollution,
resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
Tolbert, P. E.; Klein, M.; Metzger, K. B.; Peel, J.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
Frumkin, H. (2000) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure
Anal. Environ. Epidemiol. 10: 446-460.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne particulate matter. Inhalation Toxicol. 12(suppl.): 121-135.
Turpin, B. J. (1999) Options for characterizing organic particulate matter. Environ. Sci. Technol. 33: 76A-79A.
U.S. Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park,
NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525, and
PB95-124517.
U.S. Environmental Protection Agency. (1996a) Air quality criteria for particulate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/00laF-cF. 3v.
U.S. Environmental Protection Agency. (1996b) Air quality criteria for ozone and related photochemical oxidants.
Research Triangle Park, NC: Office of Research and Development; report nos. EPA/600/AP-93/004aF-cF.
3v. Available from: NTIS, Springfield, VA; PB96-185582, PB96-185590, andPB96-185608. Available
online at: www.epa.gov/ncea/ozone.htm.
April 2002 9A-14 DRAFT-DO NOT QUOTE OR CITE
-------
U.S. Environmental Protection Agency. (2000a) Air quality criteria for carbon monoxide. Research Triangle Park,
NC: National Center for Environmental Assessment; report no. EPA/600/P-99/001F. Available:
www.epa.gov/ncea/co/ [2000, October 6].
U.S. Environmental Protection Agency. (2000b) Health assessment document for diesel exhaust [SAB review
draft]. Washington, DC: Office of Research and Development, National Center for Environmental
Assessment-Washington Office; report no. EPA/600/8-90/057E. Available:
http://cfpub.epa.gov/ncea/cfm/nceahome.cfm [29 March, 2002].
U.S. Environmental Protection Agency. (2001) National air quality and emissions trends report, 1999. Research
Triangle Park, NC: Office of Air Quality Planning and Standards; report no. EPA/454/R-01-004. Available:
http://www.epa.gov/oar/aqtrnd99/ [11 March, 2002].
Undem, B. I; Carr, M. J. (2002) The role of nerves in asthma. Curr. Allergy Asthma Rep. 2: 159-165.
Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
Wallace, L. (2000) Correlations of personal exposure to particles with outdoor air measurements: a review of recent
studies. Aerosol Sci. Technol. 32: 15-25.
Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
Respir. Dis. 133:834-842.
Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Costa, D. L. (2001) Cardiovascular and systemic responses to
inhaled pollutants in rodents: effects of ozone and paniculate matter. Environ. Health Perspect.
109(suppl. 4): 539-546.
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
Whitby, K. T.; Sverdrup, G. M. (1980) California aerosols: their physical and chemical characteristics. In: Hidy,
G. M.; Mueller, P. K.; Grosjean, D.; Appel, B. R.; Wesolowski, J. J., eds. The character and origins of smog
aerosols: a digest of results from the California Aerosol Characterization Experiment (ACHEX). New York,
NY: John Wiley & Sons, Inc.; pp. 477-517. (Advances in environmental science and technology: v. 9).
Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
Cambridge, MA: Health Effects Institute; Research Report no. 98.
Williams, R.; Suggs, J.; Zweidinger, R.; Evans, G.; Creason, J.; Kwok, R.; Rodes, C.; Lawless, P.; Sheldon, L.
(2000a) The 1998 Baltimore Paniculate Matter Epidemiology-Exposure Study: part 1. Comparison of
ambient, residential outdoor, indoor and apartment paniculate matter monitoring. J. Exposure Anal. Environ.
Epidemiol. 10: 518-532.
Williams, R.; Suggs, J.; Creason, J.; Rodes, C.; Lawless, P.; Kwok, R.; Zweidinger, R.; Sheldon, L. (2000b) The
1998 Baltimore Paniculate Matter Epidemiology-Exposure Study: part 2. Personal exposure assessment
associated with an elderly study population. J. Exposure Anal. Environ. Epidemiol. 10: 533-543.
Williams, R. W.; Creason, J.; Zweidinger, R.; Watts, R. R.; Shy, C. (2000c) Indoor, outdoor, and personal exposure
monitoring of paniculate air pollution: the Baltimore elderly epidemiology-exposure pilot study.
Atmos. Environ. 34: 4193-4204.
Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
Wilson, W. E.; Mage, D. T.; Grant, L. D. (2000) Estimating separately personal exposure to ambient and
nonambient paniculate matter for epidemiology and risk assessment: why and how. J. Air Waste Manage.
Assoc. 50: 1167-1183.
Wongphatarakul, V.; Friedlander, S. K.; Pinto, J. P. (1998) A comparative study of PM25 ambient aerosol chemical
databases. Environ. Sci. Technol. 32: 3926-3934.
Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal
infant mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
J. Occup. Environ. Med. 42: 469-474.
Zanobetti, A.; Schwartz, J. (2001) Are diabetics more susceptible to the health effects of airborne particles? Am. J.
Respir. Crit. Care Med. 164: 831-833.
April 2002 9A-15 DRAFT-DO NOT QUOTE OR CITE
-------
Zanobetti, A.; Schwartz, J.; Gold, D. (2000) Are there sensitive subgroups for the effects of airborne particles?
Environ. Health Perspect. 108: 841-845.
Zanobetti, A.; Schwartz, J.; Samoli, E.; Gryparis, A.; Touloumi, G.; Atkinson, R.; Le Tertre, A.; Bobros, J.;
Celko, M.; Goren, A.; Forsberg, B.; Michelozzi, P.; Rabczenko, D.; Ruiz, E. A.; Katsouyanni, K. (2002)
The temporal pattern of mortality responses to air pollution: a multicity assessment of mortality displacement.
Epidemiology 13: 87-93.
Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
Perspect. 108: 419-426.
Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
Zhang, H.; Triche, E.; Leaderer, B. (2000) Model forthe analysis of binary time series of respiratory symptoms.
Am. J. Epidemiol. 151: 1206-1215.
Zweidinger, R.; Purdue, L.; Fitzgerald, K.; Carmichael, L.; Kellog, R. (1998) NERL PM research monitoring
platforms: Baltimore, Fresno and Phoenix, data report. Research Triangle Park, NC: U.S. Environmental
Protection Agency, National Exposure Research Laboratory; report no. EPA/600/R-98/138. Available from:
NTIS, Springfield, VA; PB99-105660.
April 2002 9A-16 DRAFT-DO NOT QUOTE OR CITE
------- |