United States
Environmental Protection
Agency
Fourth External Review Draft
of Air Quality Criteria for
Particulate Matter (June, 2003)
Volume II
-------
EPA/600/P-99/002aD
June 2003
Fourth External Review Draft
Air Quality Criteria for Particulate 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
13 list widespread air pollutants that reasonably may be expected to endanger public health or
14 welfare; (2) to issue air quality criteria for them that assess the latest available scientific
15 information on nature and effects of ambient exposure to them; (3) to set "primary" NAAQS to
16 protect human health with adequate margin of safety and to set "secondary" NAAQS to protect
17 against welfare effects (e.g., effects on vegetation, ecosystems, visibility, climate, manmade
18 materials, etc.); and (5) to periodically (every 5 years) review and revise, as appropriate, the
19 criteria and NAAQS for 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 jim) 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 |ig/m3, 24-h; 50 |ig/m3, annual average) were
28 retained in modified form and new standards (65 |ig/m3, 24-h; 15 |ig/m3, annual average) for
29 particles < 2.5 jim (PM25) were promulgated in July 1997.
30 This Fourth External Review Draft of revised Air Quality Criteria for Particulate Matter
31 assesses new scientific information that has become available mainly between early 1996
June 2003 II-iii DRAFT-DO NOT QUOTE OR CITE
-------
1 through April 2002. The present draft is being released for public comment and review by the
2 Clean Air Scientific Advisory Committee (CASAC) to obtain comments on the organization and
3 structure of the document, the issues addressed, the approaches employed in assessing and
4 interpreting the newly available information on PM exposures and effects, and the key findings
5 and conclusions arrived at as a consequence of this assessment. Public comments and CASAC
6 review recommendations will be taken into account in making any appropriate further revisions
7 to this document for incorporation into a final draft. Evaluations contained in the present
8 document will be drawn on to provide inputs to associated PM Staff Paper analyses prepared by
9 EPA's Office of Air Quality Planning and Standards (OAQPS) to pose alternatives for
10 consideration by the EPA Administrator with regard to proposal and, ultimately, promulgation of
11 decisions on potential retention or revision of the current PM NAAQS.
12 Preparation of this document was coordinated by staff of EPA's National Center for
13 Environmental Assessment in Research Triangle Park (NCEA-RTP). NCEA-RTP scientific
14 staff, together with experts from other EPA/ORD laboratories and academia, contributed to
15 writing of document chapters; and earlier drafts of this document were reviewed by experts from
16 federal and state government agencies, academia, industry, and non-governmental organizations
17 (NGOs) for use by EPA in support of decision making on potential public health and
18 environmental risks of ambient PM. The document describes the nature, sources, distribution,
19 measurement, and concentrations of PM in outdoor (ambient) and indoor environments. It also
20 evaluates the latest data on human exposures to ambient PM and consequent health effects in
21 exposed human populations (to support decision making regarding primary, health-related PM
22 NAAQS). The document also evaluates ambient PM environmental effects on vegetation and
23 ecosystems, visibility, and man-made materials, as well as atmospheric PM effects on climate
24 change processes associated with alterations in atmospheric transmission of solar radiation or its
25 reflectance from the Earth's surface or atmosphere (to support decision making on secondary
26 PM NAAQS).
27 The NCEA of EPA acknowledges the contributions provided by authors, contributors, and
28 reviewers and the diligence of its staff and contractors in the preparation of this document.
29
June 2003 Il-iv 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
APPENDIX 2A: Techniques for Measurement of Semivolatile Organic
Compounds 2A-1
APPENDIX 2B: Analytical Techniques 2B-1
3. CONCENTRATIONS, SOURCES, AND EMISSIONS 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
APPENDIX 3E: Variability Observed in PM2 5 and PM10_2 5 Concentrations
at IMPROVE Sites 3E-1
4. ENVIRONMENTAL EFFECTS OF AIRBORNE PARTICULATE
MATTER 4-1
APPENDIX 4A: Colloquial and Latin Names 4A-1
5. HUMAN EXPOSURE TO PARTICULATE MATTER AND ITS
CONSTITUENTS 5-1
June 2003 II-v DRAFT-DO NOT QUOTE OR CITE
-------
Air Quality Criteria for Particulate Matter
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 ASSOCIATED
WITH AMBIENT PARTICULATE MATTER 8-1
APPENDIX 8A: Short-Term PM Exposure-Mortality Studies:
Summary Tables 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
June 2003 Il-vi DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
Page
List of Tables I-xvi
List of Figures I-xxiii
Authors, Contributors, and Reviewers I-xxix
U.S. Environmental Protection Agency Project Team for Development of Air
Quality Criteria for Particulate Matter I-xxxvii
Abbreviations and Acronyms I-xl
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 6-3
6.2 PARTICLE DEPOSITION 6-5
6.2.1 Mechanisms of Deposition 6-5
6.2.2 Deposition Patterns in the Human Respiratory Tract 6-7
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 6-15
6.2.2.4 Local Distribution of Deposition 6-19
6.2.2.5 Deposition of Specific Size Modes of Ambient
Aerosol 6-23
6.2.3 Biological Factors Modulating Deposition 6-24
6.2.3.1 Gender 6-25
6.2.3.2 Age 6-27
6.2.3.3 Respiratory Tract Disease 6-31
6.2.3.4 Anatomical Variability 6-34
6.2.4 Interspecies Patterns of Deposition 6-36
6.3 PARTICLE CLEARANCE AND TRANSLOCATION 6-43
6.3.1 Mechanisms and Pathways of Clearance 6-43
6.3.1.1 Extrathoracic Region 6-43
6.3.1.2 Tracheobronchial Region 6-45
6.3.1.3 Alveolar Region 6-46
6.3.2 Clearance Kinetics 6-47
6.3.2.1 Extrathoracic Region 6-48
6.3.2.2 Tracheobronchial Region 6-48
6.3.2.3 Alveolar Region 6-50
6.3.3 Interspecies Patterns of Clearance 6-54
6.3.4 Factors Modulating Clearance 6-56
6.3.4.1 Age 6-56
6.3.4.2 Gender 6-56
6.3.4.3 Physical Activity 6-56
6.3.4.4 Respiratory Tract Disease 6-56
6.4 PARTICLE OVERLOAD 6-58
June 2003 II-vii 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 6-60
6.6 MODELING THE DISPOSITION OF PARTICLES IN THE
RESPIRATORY TRACT 6-66
6.6.1 Modeling Deposition, Clearance, and Retention 6-66
6.6.2 Models To Estimate Retained Dose 6-73
6.6.3 Fluid Dynamics Models for Deposition Calculations 6-76
6.6.4 Modeling Results Obtained with Models Available to the
Public 6-82
6.6.4.1 International Commission on Radiological
Protection 6-82
6.6.4.2 Multiple Path Particle Dosimetry Model 6-88
6.6.4.3 Comparisons of Deposition in Humans and Rats 6-96
6.7 SUMMARY AND CONCLUSIONS 6-102
6.7.1 Particle Dosimetry 6-102
6.7.2 Host Factors 6-102
6.7.3 Laboratory Animal Studies 6-104
6.7.4 Mathematical Models 6-104
REFERENCES 6-107
7. TOXICOLOGY OF PARTICIPATE MATTER IN
HUMANS AND LABORATORY ANIMALS 7-1
7.1 INTRODUCTION 7-1
7.2 RESPIRATORY EFFECTS OF PARTICULATE MATTER IN HEALTHY
HUMANS AND LABORATORY ANIMALS: IN VIVO EXPOSURES .... 7-3
7.2.1 Ambient Combustion-Related and Surrogate Particulate Matter 7-5
7.2.1.1 Ambient Particulate Matter 7-16
7.2.1.2 Diesel Particulate Matter 7-21
7.2.1.3 Complex Combustion-Related Particles 7-27
7.2.2 Acid Aerosols 7-31
7.2.3 Metal Particles, Fumes, and Smoke 7-33
7.2.4 Ambient Bioaerosols 7-37
7.3 CARDIOVASCULAR AND SYSTEMIC EFFECTS OF PARTICULATE
MATTER IN HUMANS AND LABORATORY ANIMALS: IN VIVO
EXPOSURES 7-39
7.4 PARTICULATE MATTER TOXICITY AND PATHOPHYSIOLOGY:
IN VITRO EXPOSURES 7-53
7.4.1 Introduction 7-53
7.4.2 Experimental Exposure Data 7-53
7.4.2.1 Ambient Particles 7-62
June 2003 II-viii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
7.4.2.2 Comparison of Ambient and Combustion-Related
Surrogate Particles 7-66
7.4.2.3 Mutagenicity 7-69
7.4.3 Potential Cellular and Molecular Mechanisms 7-69
7.4.3.1 Reactive Oxygen Species 7-69
7.4.3.2 Intracellular Signaling Mechanisms 7-76
7.4.3.3 Other Potential Cellular and Molecular Mechanisms ... 7-80
7.4.4 Specific Particle Size and Surface Area Effects 7-82
7.5 SUSCEPTIBILITY TO THE EFFECTS OF PARTICIPATE MATTER
EXPOSURE 7-87
7.5.1 Pulmonary Effects of Parti culate Matter in Compromised Hosts ... 7-87
7.5.2 Genetic Susceptibility to Inhaled Particles and their Constituents .. 7-93
7.5.3 Parti culate Matter Effects on Allergic Hosts 7-96
7.5.4 Resistance to Infectious Disease 7-102
7.6 RESPONSES TO PARTICIPATE MATTER AND GASEOUS
POLLUTANT MIXTURES 7-103
7.7 SUMMARY OF KEY FINDINGS AND CONCLUSIONS 7-114
7.7.1 Links Between Specific Particulate Matter Components and
Health Effects 7-115
7.7.1.2 Acid Aerosols 7-115
7.7.1.3 Metals 7-116
7.7.1.4 Diesel Exhaust Particles 7-117
7.7.1.5 Organic Compounds 7-117
7.7.1.6 Ultrafme Particles 7-117
7.7.1.7 Concentrated Ambient Particle Studies 7-118
7.7.1.8 Bioaerosols 7-119
7.7.2 Mechanisms of Action 7-119
7.7.2.1 Direct Pulmonary Effects 7-120
7.7.2.2 Systemic Effects Secondary to Lung Injury 7-123
7.7.2.3 Direct Effects on the Heart 7-125
7.7.3 Susceptibility 7-127
7.7.4 PM Interactions with Gaseous Co-Pollutants 7-127
REFERENCES 7-129
8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS ASSOCIATED WITH
AMBIENT PARTICULATE MATTER 8-1
8.1 INTRODUCTION 8-1
8.1.1 Approaches for Identifying and Assessing Studies 8-2
8.1.2 Types of Epidemiologic Studies Reviewed 8-5
8.1.3 Confounding and Effect Modification 8-8
8.1.4 GAM Convergence Issue 8-16
8.1.5 Ambient PM Increments Used to Report Risk Estimates 8-17
June 2003 Il-ix DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
.2 MORTALITY EFFECTS ASSOCIATED WITH AIRBORNE
PARTICIPATE MATTER EXPOSURE 8-18
8.2.1 Introduction 8-18
8.2.2 Mortality Effects of Short-Term Particulate Matter Exposure 8-19
8.2.2.1 Summary of 1996 Particulate Matter Criteria
Document Findings and Key Issues 8-19
8.2.2.2 Newly Available Information on Short-Term
Mortality Effects 8-23
8.2.2.3 New Multi-City Studies 8-30
8.2.2.4 The Role of Particulate Matter Components 8-50
8.2.2.5 New Assessments of Cause-Specific Mortality 8-70
8.2.2.6 Salient Points Derived from Assessment of Studies
of Short-Term Particulate Matter Exposure Effects
on Mortality 8-75
8.2.3 Mortality Effects of Long-Term Exposure to Ambient
Particulate Matter 8-78
8.2.3.1 Studies Published Prior to the 1996 Particulate
Matter Criteria Document 8-78
8.2.3.2 New Prospective Cohort Analyses of Mortality
Related to Chronic Particulate Matter Exposures 8-82
8.2.3.3 Studies by Particulate Matter Size-Fraction and
Composition 8-106
8.2.3.4 Population-Based Mortality Studies in Children 8-112
8.2.3.5 Salient Points Derived from Analyses of Chronic
Particulate Matter Exposure Mortality Effects 8-115
.3 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE .... 8-118
8.3.1 Cardiovascular Effects Associated with Acute Ambient
Particulate Matter Exposure 8-118
8.3.1.1 Introduction 8-118
8.3.1.2 Summary of Key Findings on Cardiovascular
Morbidity from the 1996 Particulate Matter Air
Quality Criteria Document 8-119
8.3.1.3 New Particulate Matter-Cardiovascular Morbidity
Studies 8-120
8.3.1.4 Issues in the Interpretation of Acute Cardiovascular
Effects Studies 8-146
8.3.2 Effects of Short-Term Particulate Matter Exposure on the
Incidence of Respiratory-Related Hospital Admissions and
Medical Visits 8-147
8.3.2.1 Introduction 8-147
June 2003 II-x DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
8.3.2.2 Summary of Key Respiratory Hospital Admissions
Findings from the 1996 Particulate Matter Air Quality
Criteria Document 8-148
8.3.2.3 New Respiratory-Related Hospital Admissions
Studies 8-149
8.3.2.4 Key New Respiratory Medical Visits Studies 8-163
8.3.2.5 Identification of Potential Susceptible
Subpopulations 8-165
8.3.2.6 Summary of Salient Findings on Acute Particulate
Matter Exposure and Respiratory-Related Hospital
Admissions and Medical Visits 8-167
8.3.3 Effects of Particulate Matter Exposure on Lung Function
and Respiratory Symptoms 8-168
8.3.3.1 Effects of Short-Term Particulate Matter Exposure
on Lung Function and Respiratory Symptoms 8-169
8.3.3.2 Long-Term Particulate Matter Exposure Effects
on Lung Function and Respiratory Symptoms 8-185
.4 DISCUSSION OF EPIDEMIOLOGIC STUDIES OF HEALTH
EFFECTS OF AMBIENT PARTICULATE MATTER 8-189
8.4.1 Introduction 8-189
8.4.2 GAM Issue and Reanalyses Studies 8-192
8.4.2.1 Impact of Using the More Stringent GAM Model
on PM Effect Estimates for Mortality 8-193
8.4.2.2 Impact of Using the More Stringent GAM Model
on PM Effect Estimates for Respiratory Hospital
Admissions 8-197
8.4.2.3 HEI Commentaries 8-201
8.4.3 Assessment of Confounding by Co-Pollutants 8-204
8.4.3.1 Introduction 8-204
8.4.3.2 Conceptual Issues in Assessing Confounding 8-206
8.4.3.3 Statistical Issues in the Use of Multi-Pollutant
Models 8-207
8.4.3.4 Epidemiologic Studies of Ambient Air Pollution
Interventions 8-213
8.4.4 Role of Particulate Matter Components 8-219
8.4.4.1 Fine- and Coarse-Particle Effects on Mortality 8-219
8.4.4.2 PM10, PM25 (Fine), and PM10.25 (Coarse) Particulate
Matter Effects on Morbidity 8-229
8.4.5 The Question of Lags 8-234
8.4.6 Concentration-Response Relationships for Ambient PM 8-238
8.4.7 Heterogeneity of Particulate Matter Effects Estimates 8-241
June 2003 Il-xi DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
8.4.7.1 Evaluation of Heterogeneity of Particulate Matter
Mortality Effect Estimates 8-243
8.4.7.2 Comparison of Spatial Relationships in the
NMMAPS and Cohort Reanalyses Studies 8-247
8.4.8 New Assessments of Measurement Error Consequences 8-248
8.4.8.1 Theoretical Framework for Assessment of
Measurement Error 8-248
8.4.8.2 Spatial Measurement Error Issues That May Affect
the Interpretation of Multi-Pollutant Models with
Gaseous Co-Pollutants 8-255
8.4.8.3 Measurement Error and the Assessment of
Confounding by Co-Pollutants in Multi-Pollutant
Models 8-263
8.4.8.4 Air Pollution Exposure Proxies in Long-Term
Mortality Studies 8-264
8.4.9 Implications of Airborne Particle Mortality Effects 8-268
8.4.9.1 Short-Term Exposure and Mortality Displacement .... 8-268
8.4.9.2 Life-Shortening Estimates Based on Semi-Individual
Cohort Study Results 8-273
8.4.9.3 Potential Effects of Infant Mortality on
Life-Shortening Estimates 8-274
8.5 SUMMARY OF KEY FINDINGS AND CONCLUSIONS DERIVED
FROM PARTICULATE MATTER EPIDEMIOLOGY STUDIES 8-274
REFERENCES 8-282
APPENDIX 8A: Short-Term PM Exposure-Mortality Studies: Summary Table 8A-1
APPENDIX 8B: Particulate Matter-morbidity Studies: Summary Tables 8B-1
Appendix 8B.1: PM-Cardiovascular Admissions Studies 8B-2
Appendix 8B.2. PM-Respiratory Hospitalization Studies 8B-19
Appendix 8B.3: PM-Respiratory Visits Studies 8B-42
Appendix 8B.4: Pulmonary Function Studies 8B-53
Appendix 8B.5: Short-Term PM Exposure Effects On Symptoms in Asthmatic
Individuals 8B-59
Appendix 8B.6: Short-Term PM Exposure Effects on Pulmonary Function
in Nonasthmatics 8B-65
Appendix 8B.7: Short-Term PM Exposure Effects on Symptoms in
Nonasthmatics 8B-73
Appendix 8B.8: Long-Term PM Exposure Effects on Respiratory Health
Indicators, Symptoms, and Lung Function 8B-78
9. INTEGRATIVE SYNTHESIS 9-1
June 2003 II-xii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
9.1 INTRODUCTION 9-1
9.1.1 Legislative Requirements and PastNAAQS Reviews 9-1
9.1.2 Organization of the Chapter 9-4
9.2 BACKGROUND 9-6
9.2.1 Basic Concepts 9-6
9.2.2 Particle Size Distributions 9-6
9.2.3 Definitions of Particle Size Fractions 9-7
9.3 CHARACTERIZATION OF PM SOURCES 9-15
9.4 AMBIENT CONCENTRATIONS 9-19
9.4.1 Measurement of Particulate Matter 9-19
9.4.2 Mass Concentrations 9-21
9.4.3 Physical and Chemical Properties of Ambient PM 9-21
9.5 EXPOSURE TO PARTICULATE MATTER AND CO-POLLUTANTS .... 9-25
9.5.1 Central Site to Outdoor Relationships 9-25
9.5.1.1 Exposure for Acute Epidemiology 9-25
9.5.1.2 Exposure for Chronic Epidemiology 9-26
9.5.2 Home Outdoor Concentrations Versus Concentrations of
Ambient PM Infiltrated Indoors 9-28
9.5.2.1 Mass Balance Model 9-28
9.5.2.2 Separation of Total Personal Exposure into its
Ambient and Nonambient Components 9-29
9.5.3 Variability in the Relationship Between Outdoor Concentrations
and Personal Exposures 9-32
9.5.4 Exposure Relations for Co-Pollutants 9-33
9.5.5 Exposure Relationships for Susceptible Subpopulations 9-38
9.5.6 Air Pollutants Generated Indoors 9-38
9.6 DOSIMETRY: DEPOSITION AND FATE OF PARTICLES IN THE
RESPIRATORY TRACT 9-39
9.6.1 Particle Deposition in the Respiratory Tract 9-39
9.6.2 Particle Clearance and Translocation 9-45
9.6.3 Dosimetric Considerations in Comparing Dosages for Inhalation,
Instillation, and Exposure of Cultured Cells 9-47
9.6.4 Inhaled Particles as Potential Carriers of Toxic Agents 9-49
9.7 TOXICOLOGIC ASSESSMENT OF PARTICULATE-MATTER
PROPERTIES LINKED TO HEALTH EFFECTS 9-50
9.7.1 Chemical Components and Source Categories Associated with
Health Effects in Epidemiologic Studies 9-52
9.7.1.1 Toxicologically Important Components of PM 9-52
9.7.1.2 Source Category Factors 9-53
9.7.2 Specific Properties of Ambient PM Linked to Health Effects 9-55
9.7.2.1 Physical Properties 9-55
9.7.2.2 Chemical Properties 9-57
June 2003 II-xiii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
9.7.2.3 Summary 9-61
9.7.3 Mechanisms of Action Underlying PM Cardiovascular Effects .... 9-61
9.8 HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER
OBSERVED IN HUMAN POPULATION STUDIES 9-66
9.8.1 Introduction 9-66
9.8.1.2 GAM Convergence Issue 9-70
9.8.1.3 Ambient PM Increments Used to Report Ri sk
Estimates 9-71
9.8.2 Short-Term Particulate Matter Exposure Effects on Mortality 9-72
9.8.2.1 Methodological Issues for Short-Term Exposure
Studies 9-111
9.8.3 Health Effects of Long-Term Exposures to Particulate Matter .... 9-118
9.8.4 Coherence of Reported Epidemiologic Findings 9-127
9.9 SUSCEPTIBLE SUBPOPULATIONS AND IMPLICATIONS OF
EFFECTS OF AMBIENT PM EXPOSURE ON HUMAN HEALTH 9-129
9.9.1 Introduction 9-129
9.9.2 Preexisting Disease as a Risk Factor for Particulate Matter
Health Effects 9-130
9.9.2.1 Ambient PM Exacerbation of Cardiovascular Disease
Conditions 9-130
9.9.2.2 Ambient PM Exacerbation of Respiratory Disease
Conditions 9-133
9.9.3 Age-Related At-Risk Population Groups: The Elderly and
Children 9-135
9.9.4 Impact on Life-Expectancy 9-138
9.10 INTEGRATIVE SYNTHESIS OF KEY FINDINGS FOR
ENVIRONMENTAL EFFECTS OF AMBIENT AIRBORNE PM 9-139
9.10.1 Introduction 9-139
9.10.2 Effects of Ambient Airborne PM on Vegetation and Natural
Ecosystems 9-139
9.10.2.1 Ecological Attributes 9-141
9.10.2.2 Ecosystem Exposures - Particle Deposition 9-143
9.10.2.3 Direct and Indirect Effects on Ecosystems 9-143
9.10.3 Visibility Effects of Airborne Particles 9-149
9.10.4 Materials Damage Related to Airborne Particulate Matter 9-151
9.10.5 Atmospheric Particle Effects on Global Warming Processes and
Transmission of Solar Ultraviolet Radiation 9-152
REFERENCES 9-154
June 2003 II-xiv DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
Page
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
June 2003 II-xv DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
Number Page
6-1 Effects of Age on Particle Deposition in Respiratory Tract 6-28
6-2 Overview of Respiratory Tract Particle Clearance and Translocation
Mechanisms 6-44
6-3 Respiratory Parameters Used in LUDEP Model 6-84
6-4 Levels of Physical Exertion for Adult, Corresponding Representative
Activities, and Breathing Parameters 6-90
6-5 Parameters Used in Age Dependent Calculations of the CIIT/RIVM
Deposition Model 6-92
6-6 Respiratory Parameters for Humans and Rats 6-97
6-7 Surface Areas of Tracheobronchial and Alveolar Regions for Humans
and Rats 6-100
7-la Respiratory Effects of Inhaled Ambient Particulate Matter in Controlled
Exposure Studies of Human Subjects and Laboratory Animals 7-6
7- Ib Respiratory Effects of Instilled Ambient Parti culate Matter in Laboratory
Animals and Human Subjects 7-7
7-2a Respiratory Effects of Instilled Complex Combustion-related Parti culate
Matter in Laboratory Animals 7-8
7-2b Respiratory Effects of Inhaled Complex Combustion-related Parti culate
Matter in Compromised Laboratory Animal Models 7-12
7-3 Respiratory Effects of Surrogate Parti culate Matter in Laboratory Animals .... 7-13
7-4 Respiratory Effects of Acid Aerosols in Humans and Laboratory Animals 7-32
7-5a Respiratory Effects of Instilled Metal Particles, Fumes, and Smoke in Human
Subjects and Laboratory Animals 7-34
7-5b Respiratory Effects of Inhaled Metal Particles, Fumes, and Smoke in Humans
and Laboratory Animals 7-35
7-6 Controlled Exposure Studies of Respiratory Effects of Inhaled Ambient
Bioaerosols 7-38
June 2003 II-xvi DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
7-7a Cardiovascular and Systemic Effects of Instilled Ambient and
Combustion-Related Particulate Matter 7-40
7-7b Cardiovascular and Systemic Effects of Inhaled Ambient and
Combustion-Related Particulate Matter 7-42
7-8 Physicochemical Properties of Particulate Matter 7-54
7-9 In Vitro Effects of Particulate Matter and Particulate Matter Constituents 7-55
7-10 Mutagenic/Carcinogenic Effects of Particulate Matter 7-70
7-11 Numbers and Surface Areas of Monodisperse Particles of Unit Density of
Different Sizes at a Mass Concentration of 10 |ig/m3 7-83
7-12 Respiratory and Cardiovascular Effects of PM and Gaseous Pollutant
Mixtures 7-105
8-1 Recent U.S. and Canadian Time-Series Studies of PM-Related Daily
Mortality 8-24
8-2 Synopsis of Short-Term Mortality Studies that Examined Relative
Importance of PM25 and PM10_25 8-52
8-3 Newly Available Studies of Mortality Relationships to PM Chemical
Components 8-62
8-4 Summary of Source-Oriented Evaluations of PM Components in Recent
Studies 8-67
8-5 Comparison of Six Cities and American Cancer Society Study Findings
from Original Investigators and Health Effects Institute Reanalysis 8-84
8-6 Relative Risk of All-Cause Mortality for Selected indices of Exposure to
Fine Particulate Matter (per 18.6 |ig/m3) Based on Multivariate Poisson
Regression Analysis, by Age Group, for Harvard Six City Study Data 8-88
8-7 Summary of Results from the Extended ACS Study 8-90
8-8 Relative Risk of Mortality from All Nonexternal Causes, by Sex and Air
Pollutant, for an Alternative Covariate Model in the ASHMOG Study 8-97
June 2003 II-xvii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
8-9 Relative Risk of Mortality from Cardiopulmonary Causes, by Sex and Air
Pollutant, for an Alternative Covariate Model in the ASHMOG Study 8-98
8-10 Relative Risk of Mortality from Lung Cancer by Air Pollutant and by
Gender for an Alternative Covariate Model 8-99
8-11 Comparison of Excess Relative Risks of Long-Term Mortality in the
Harvard Six Cities, ACS, AHSMOG, and VA Studies 8-104
8-12 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 8-107
8-13 Comparison of Reported SO4= and PM2 5 Relative Risks for Various
Mortality Causes in the American Cancer Society Study 8-108
8-14 Comparison of Total Mortality Relative Risk Estimates and T-Statistics for
Particulate Matter Components in Three Prospective Cohort Studies 8-109
8-15 Comparison of Cardiopulmonary Mortality Relative Risk Estimates and
T-Statistics for Particulate Matter Components in Three Prospective
Cohort Studies 8-110
8-16 Summary of Studies of PM10, PM10.2 5, or PM2 5 Effects on Total C VD
Hospital Admissions and Emergency Visits 8-121
8-17 Summary of United States PM10 Respiratory-Related Hospital Admission
Studies 8-150
8-18 Percent Increase in Hospital Admissions per 10-|ig/m3 Increase in PM10 in
14 U.S. Cities (original and reanalyzed results) 8-152
8-19 Summary of United States PM25 Respiratory-Related Hospital Admission
Studies 8-156
8-20 Summary of United States PM10_2 5 Respiratory-Related Hospital Admission
Studies ' 8-157
8-21 Intercomparison of Detroit Pneumonia Hospital Admission Relative Risks
(± 95% CI below) of PM Indices (per 5th-to-95th percentile pollutant
Increment) for Various Model Specifications 8-158
June 2003 II-xviii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
8-22 Summary of United States PM10, PM25, and PM10_2 5 Asthma Medical Visit
Studies 8-163
8-23 Summary of Quantitative PFT Changes in Asthmatics per 50 |ig/m3 PM10
Increment 8-171
8-24 Summary of PFT Changes in Asthmatics per 25 |ig/m3 PM25 Increment 8-172
8-25 Summary of Asthma PM10 Cough Studies 8-175
8-26 Summary of Asthma PM10 Phlegm Studies 8-176
8-27 Summary of Asthma PM10 Lower Respiratory Illness Studies 8-176
8-28 Summary of Asthma PM10 Bronchodilator Use Studies 8-177
8-29 Summary of Asthma PM25 Respiratory Symptom Studies 8-179
8-30 Summary of Non-Asthma PM10 PFT Studies 8-181
8-31 Summary of Non-Asthma PM10 Respiratory Symptom Studies 8-182
8-32 Summary of Non-Asthma PM25 Respiratory Outcome Studies 8-183
8-33 Summary of Non-Asthma Coarse Fraction Studies of Respiratory
Endpoints 8-184
8-34 PM10 Excess Risk Estimates from Reanalysis Studies for Total
Non-Accidental Mortality per 50 |ig/m3 Increase in PM10 8-194
8-35 Comparison of Maximum Single Day Lag Effect Estimates for PM2 5,
PM2 5_10, and PM10 for Seattle Asthma Hospital Admissions Based on
Original Gam Analyses Using Default Convergence Criteria Versus
Reanalyses Using GAM with More Stringent Convergence Criteria
and GLM 8-199
8-36 Comparison of Los Angeles COPD Hospital Admissions Maximum
Single Day Lag Effect Estimates for PM2 5 and PM10 from the Original
GAM Analyses Using Default Convergence Criteria Versus for Reanalyses
Using More Stringent Convergence Criteria and for Models Smoothed
with More Degrees of Freedom 8-201
June 2003 II-xix DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
8-37 Summary of Past Ecologic and Case-Control Epidemiologic Studies
of Outdoor Air and Lung Cancer 8-225
8-38 Maximum, Mean, 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
8-39 Summary of Within-City Heterogeneity by Region 8-260
8-40 Summary of ACS Pollution Indices: Units, Primary Sources, Number of
Cities and Subjects Available for Analysis, and the Mean Levels (standard
deviations) 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-20
8B-3 Acute Particulate Matter Exposure and Respiratory Medical Visits 8B-43
8B-4 Short-Term Particulate Matter Exposure Effects on Pulmonary Function
Tests in Studies of Asthmatics 8B-54
8B-5 Short-term Particulate Matter Exposure Effects on Symptoms in Studies
of Asthmatics 8B-60
8B-6 Short-Term Particulate Matter Exposure Effects on Pulmonary Function
Tests in Studies of Nonasthmatics 8B-66
8B-7 Short-Term Particulate Matter Exposure Effects on Symptoms in Studies
of Nonasthmatics 8B-74
8B-8 Long-Term Particulate Matter Exposure Respiratory Health Indicators:
Respiratory Symptom, Lung Function 8B-79
9-1 Constituents of Atmospheric Particles and Their Major Sources 9-16
June 2003 II-xx DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
9-2 Comparison of Ambient Particles, Fine (ultrafine plus accumulation mode)
and Coarse 9-22
9-3 Concentrations of PM25, PM10_25 and Selected Elements in the PM25 and
PM10.2 5 Size Range .' 9-23
9-4 Qualitative Estimates of Exposure Variables 9-36
9-5 Concentration Differences Between Constituents of Nonambient
(indoor-generated) and Ambient PM 9-39
9-6 Particulate Matter Associated with Mortality in Epidemiologic Studies 9-52
9-7 Source Categories Associated with Mortality in Epidemiologic Studies 9-54
9-8 Estimated Total, Cardiovascular and Respiratory Mortality Effect Sizes per
Increments in 24-h Concentrations of PM10, PM25 and PM10_25 from U.S.
and Canadian Studies 9-77
9-9 Summary of Source-oriented Evaluations of PM Components in Recent
Studies 9-89
9-10 Cardiovascular and Respiratory-Related Morbidity Effect Size Estimates per
Increment in 24-h Concentrations of PM10, PM25, and PM10_25 in U.S.
and Canadian Studies 9-91
9-11 Effect Estimates per Increments in Long-Term Mean Levels of Fine and
Coarse Fraction Particle Indicators from U.S. and Canadian Studies 9-119
9-12 Incidence of Selected Cardiorespiratory Disorders by Age and by Geographic
Region, 1996 (reported as incidence per thousand population and as number
of cases in thousands) 9-131
9-13 Number of Acute Respiratory Conditions per 100 Persons per Year, by
Age: United States, 1996 9-136
9-14 Essential Ecological Attributes and Reporting Categories 9-142
9-15 Effects of Reactive Nitrogen 9-148
9A-1 Effect Estimates per 50-|ig/m3 Increase in 24-Hour PM10 Concentrations
From U.S. and Canadian Studies 9A-2
June 2003 II-xxi DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number
9A-2 Effect Estimates per Variable Increments in 24-Hour Concentrations of Fine
Particle Indicators (PM25, SO^, FT) 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
June 2003 II-xxii DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
Number Page
6-1 Diagrammatic representation of respiratory tract regions in humans 6-4
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 (oral inhalation as percentage deposition of the
amount inhaled) in humans as a function of particle size 6-16
6-6 Alveolar deposition (as percentage deposition of the amount inhaled) in
humans as a function of particle size 6-16
6-7 Lung deposition fractions in the tracheobronchial (TB) and alveolar (A)
regions estimated by the bolus technique 6-18
6-8 Estimated lung deposition fractions in ten volumetric regions for particle sizes
ranging from ultrafine particle diameter (dp) of 0.04 to 0.01 jim (Panel A) to
fine (dp =1.0 |im; Panel B) and coarse (dp = 3 and 5 jim; Panels C and D) 6-22
6-9 Regional deposition fraction measured in laboratory animals as a function of
particle size for (a) upper respiratory tract, (b) tracheobronchial region, and
(c) pulmonary region 6-38
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 6-39
6-11 Major clearance pathways for particles deposited in the extrathoracic region
and tracheobronchial tree 6-44
6-12 Diagram of known and suspected clearance pathways for poorly soluble
particles depositing in the alveolar region 6-45
6-13 Percent deposition for total results of LUDEP model for an adult male worker
(default) showing total percent deposition in the respiratory tract (TOT) and
in the ET, TB, and A regions 6-85
June 2003 II-xxiii DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number Page
6-14 Percent deposition for total results of LUDEP model for a young adult (default)
showing total percent deposition in the respiratory tract (TOT) and in the ET,
TB, and A regions 6-86
6-15 Comparison of percent deposition in the TB and A regions for a worker (WK;
light exercise) and a young adult (YA; resting) 6-87
6-16 Dependency of aerosol deposition in human adults on physical exertion
expressed as minute ventilation for different particle sizes 6-91
6-17 Age dependency of human aerosol deposition for different particle sizes 6-93
6-18 Age dependency of human standardized aerosol deposition for different
particle sizes 6-95
6-19 Comparison of percent deposition for rats (nasal breathing) and humans (nasal
and mouth breathing) and the ratio of human to rat for nasal and mouth
breathing humans for the ET (a), TB (b), and A (c) regions of the respiratory
tract 6-97
6-20 Normalized deposition patterns for rats (nasal breathing) and humans (nasal and
mouth breathing) and the ratio of human to rat for nasal and mouth breathing
humans for the thoracic region (in terms of jig PM per g of lung) 6-99
6-21 Normalized deposition patterns arising from 8 hr exposure to 100 |ig/m3, based
on EPA default values of surface area, for rats (nasal breathing) and humans
(nasal and mouth breathing) and the ratio of human to rat (a) for the TB region
(in units of jig PM per m2 TB area) and (b) for the A region (in terms of jig
PM per m2 of A) 6-100
6-22 Normalized deposition patterns arising from 8 hr exposure to 100 |ig/m3, based
on surface area values from Winter-Sorkina and Cassee (2002), for rats (nasal
breathing) and humans (nasal and mouth breathing) and the ratio of human to
rat (a) for the TB region (in units of jig PM per m2 TB area) and (b) for the A
region (in terms of jig PM per m2 of A) 6-101
8-la Strong within-city association between PM and mortality, but no second-stage
association 8-11
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 8-11
June 2003 II-xxiv DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number Page
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 8-13
8-3 Estimated excess risks for PM mortality (1 day lag) for the 88 largest U.S.
cities as shown in the revised NMMAPS analysis 8-32
8-4 Map of the United States showing the 88 cities (the 20 cities are circled) and
the seven U.S. regions considered in the NMMAPS geographic analyses 8-33
8-5 Percent excess mortality risk (lagged 0, 1, or 2 days) estimated in the
NMMAPS 90-City Study to be associated with 10-|ig/m3 increases in PM10
concentrations in cities aggregated within U.S. regions shown in Figure 8-4 . . . 8-34
8-6 Marginal posterior distributions for effect of PM10 on total mortality at lag 1
with and without control for other pollutants, for the 90 cities 8-36
8-7 Percent excess risks estimated per 25 |ig/m3 increase in PM2 5 or PM10_2 5
from new studies evaluating both PM2 5 and PM10_2 5, based on single pollutant
(PM only) model 8-55
8-8 Excess risks estimated per 5 |ig/m3 increase in sulfate, based on the studies
in which both PM2 5 and PM10_2 5 data were available 8-65
8-9 Natural logarithm of relative risk for total and cause-specific mortality per
10 |ig/m3 PM25 (approximately the excess relative risk as a fraction), with
smoothed concentration-response functions 8-91
8-10 Relative risk of total and cause-specific mortality at 10 |ig/m3 PM25 (mean
of 1979-1983) of alternative statistical models 8-92
8-11 Relative risk of total and cause-specific mortality for particle metrics and
gaseous pollutants over different averaging periods (years 1979-2000 in
parentheses) 8-93
8-12 Acute cardiovascular hospitalizations and particulate matter exposure
excess risk estimates derived from selected U.S. PM10 studies based on
single-pollutant models 8-137
8-13 Percent change in hospital admission rates and 95% CIs for an IQR increase
in pollutants from single-pollutant models for asthma 8-160
June 2003 II-xxv DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number Page
8-14 Maximum excess risk of respiratory-related hospital admissions and visits
per 50 |ig/m3 PM10 increment in selected studies of U.S. cities based on
single-pollutant models 8-168
8-15 Selected acute pulmonary function change studies of asthmatic children 8-173
8-16 Odds ratios with 95% confidence interval for cough per 50-|ig/m3 increase
in PM10 for selected asthmatic children studies at lag 0 8-178
8-17 PM10 excess risk estimates for total non-accidental mortality for numerous
locations (and for cardiovascular mortality[*] for Coachella Valley, CA
and Phoenix, AZ), using: (1) GAM with default convergence criteria
(white circle); (2) GAM with stringent convergence criteria (black circle);
and, (3) GLM/natural splines (x) that approximate the original GAM model
from the GAM reanalysis studies 8-195
8-18 Comparison of GAM results for original (default) convergence case versus
those from reanalyses with a more stringent convergence criterion (10e-15)
for constrained lag respiratory model cases 8-198
8-19 Marginal posterior distribution for effects of PM10 on all cause mortality at
lag 0, 1, and 2 for the 90 cities 8-235
8-20 Particulate matter < 10 jim in aerodynamic diameter (PM10)-total mortality
concentration-response curves for total (TOTAL) mortality, cardiovascular
and respiratory (CVDRESP) mortality, and other causes (OTHERS) mortality,
20 largest US cities, 1987-1994 8-239
8-21 An EPA-derived plot showing relationship of PM10 total mortality effects
estimates and 95% confidence intervals for all cities in the Dominici et al.
(2000a; 2003) NMMAPS 90-cities analyses in relation to study size (i.e., the
natural logarithm or numbers of deaths times days of PM observations) 8-244
8-22 The EPA-derived plots showing Dominici et al. (2003) relationships of
PM10-mortality (total, nonaccidental) effects estimates and 95% confidence
intervals to study size (defined as in Figure 8-10) for cities broken out by
regions as per the NMMAPS regional analyses of Samet et al. (2000a,b) 8-245
8-23 Concentration of PM10 and NO2 versus distance 8-258
9-1 A general framework for integrating particulate-matter research 9-6
June 2003 II-xxvi DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number Page
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 9-8
9-3 Volume size distribution, measured in traffic, showing fine and coarse particles
and the nuclei and accumulation modes within the fine particles 9-9
9-4 Submicron number size distribution observed in a boreal forest in Finland
showing the tri-modal structure of fine particles 9-10
9-5 Specified particle penetration (size-cut curves) through an ideal (no-particle-
loss) inlet for five different size-selective sampling criteria 9-13
9-6 An idealized distribution of ambient particulate matter showing the
accumulation mode and the coarse mode and the size fractions collected by
size-selective samplers 9-14
9-7 Schematic showing major nonvolatile and semivolatile components of PM25 . . . 9-20
9-8 Major chemical components of PM25 as determined in the U.S. Environmental
Protection Agency's national speciation network from October 2001 to
September 2002 9-24
9-9 Occurrence of differences between pairs of sites in three MS As 9-27
9-10 Regression analysis of daytime total personal exposures to PM10 versus
ambient PM10 concentrations using data from the PTEAM study 9-30
9-11 Regression analysis of daytime exposures to the ambient component of
personal exposure to PM10 (ambient exposure) versus ambient PM10
concentrations 9-31
9-12 Percentage of homes with air conditioning versus the regression coefficient
for the relationship of cardiovascular-related hospital admissions to ambient
PM10 concentrations 9-33
9-13 Values of geometric mean infiltration factor, FINP = A/C, as a function of
particle diameter for hourly nighttime data (assuming no indoor sources)
for summer and fall seasons 9-34
June 2003 II-xxvii DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number Page
9-14 Percent deposition for total results of LUDEP model for an adult male worker
(default) showing total percent deposition in the respiratory tract (TOT) and
in the ET, TB, and A regions 9-42
9-15 Schematic representation of potential pathophysiological pathways and
mechanisms by which ambient PM may increase risk of cardiovascular
morbidity and/or mortality 9-62
9-16 Percent excess risks estimated per 25-|ig/m3 increase in PM2 5 or PM10_25 from
new studies evaluating both PM2 5 and PM10_2 5 data for multiple years 9-86
9-17 Relative risks estimated per 5-|ig/m3 increase in sulfate from U.S. and
Canadian studies in which both PM2 5 and PM10_2 5 data were available 9-88
9-18 Acute cardiovascular hospitalizations and PM exposure excess risk
estimates derived from selected U.S. PM10 studies 9-102
9-19 Maximum excess risk in selected studies of U.S. cities relating PM10
estimate of exposure (50 |ig/m3) to respiratory-related hospital admissions
and visits 9-106
9-20 Selected acute pulmonary function change studies of asthmatic children 9-109
9-21 Odds ratios for cough for a 50-|ig/m3 increase in PM10 for selected asthmatic
children studies, with lag 0 with 95% CI 9-110
9-22 Marginal posterior distributions for effect of PM10 on total mortality at lag
1, with and without control for other pollutants, for the NMMAPS 90 cities ... 9-113
9-23 Inhalation rates on a per body-weight basis for males (•) and females (*)
by age (Layton, 1993) 9-137
9-24 Illustration of the nitrogen cascade showing the movement of the
human-produced reactive nitrogen (Nr) as it cycles through the various
environmental reservoirs in the atmosphere, terrestrial ecosystems, and
acquatic ecosystems 9-146
June 2003 II-xxviii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
CHAPTER 6. DOSIMETRYOFPARTICULATEMATTER
Principal Authors
Dr. Ramesh Sarangapani—ICF Consulting, 3200 NC-54, Suite 101, P.O. Box 14348, Research
Triangle Park, NC 27709
Dr. Richard Schlesinger—New York University School of Medicine, Department of
Environmental Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. James McGrath—Chapel Hill, NC 27517
Dr. William Wilson—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributing Authors
Dr. John Stanek—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. William Bennett—University of North Carolina at Chapel Hill, Campus Box 7310,
Chapel Hill, NC 37599
Dr. Dan Costa—National Health and Environmental Effects Research Laboratory (B305-02),
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. Mark Frampton—University of Rochester, 601 Elmwood Avenue, Box 692, Rochester, NY
14642
Dr. Andrew Ohio—National Health and Environmental Effects Research Laboratory (MD-58D),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. John Godleski—Weston, MA 02493
June 2003 II-xxix DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont1 d)
Dr. Judith Graham—American Chemistry Council, 1300 Wilson Boulevard, Arlington, VA
22207
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-58C),
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. 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. Jim Samet—National Health and Environmental Effects Research Laboratory (MD-58D),
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
CHAPTER 7. TOXICOLOGY OF PARTICULATE MATTER IN HUMANS AND
LABORATORY 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
June 2003 II-xxx DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont1 d)
Principal Authors
(cont'd)
Dr. Terry Gordon—New York University Medical Center, Department of Environmental
Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. James McGrath—Chapel Hill, NC 27517
Dr. Christine Nadziejko—Department of Environmental Medicine, New York University School
of Medicine, Tuxedo, NY
Mr. James Raub—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. Susanne Becker—National Health and Environmental Effects Research Laboratory
(MD-58D), 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. Dan Costa—National Health and Environmental Effects Research Laboratory (B305-02),
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
Dr. Mark Frampton—University of Rochester, 601 Elmwood Avenue, Box 692, Rochester, NY
14642
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-92),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
June 2003 II-xxxi DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
Dr. John Godleski—Weston, MA 02493
Dr. Tony Huang—National Health and Environmental Effects Research Laboratory (MD-58D),
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-58C),
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. 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
Bill Russo—National Health and Environmental Effects Research Laboratory (B305-02),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
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. Jim Samet—National Health and Environmental Effects Research Laboratory (MD-58D),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
John J. Vandenberg—National Center for Environmental Assessment (8601 D),
U.S. Environmental Protection Agency, 1200 Pennsylvania Avenue, NW, Washington, DC
20460
Bellina Veronesi—National Health and Environmental Effects Research Laboratory (MD-74B),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
June 2003 II-xxxii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont1 d)
Contributors and Reviewers
(cont'd)
Dr. William Watkinson—National Health and Environmental Effects Research Laboratory
(MD-82), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
CHAPTER 8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM
AMBIENT PARTICULA TE MA TTER
Principal Authors
Dr. Lester D. Grant—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Vic Hasselblad—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 (B243-02),
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
Contributing Authors
Dr. Robert Chapman—Retired, formerly at the National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
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
June 2003 II-xxxiii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont.)
Dr. Raymond Carroll—Texas A & M University, Department of Statistics, College Station, TX
77843-3143
Dr. Steven Colome—Integrated Environmental Services, 5319 University Drive, #430, Irvine,
CA 92612
Dr. Ralph Delfino—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. Fred Lipfert—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
Dr. James Robins—Harvard School of Public Health, Department of Epidemiology,
Boston, MA 02115
Dr. Mary Ross—Office of Air Quality Planning and Standards (C539-01), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Dr. Lianne Sheppard—University of Washington, Box 357232, Seattle, WA 98195-7232
Dr. Leonard Stefanski—North Carolina State University, Department of Statistics, Box 8203,
Raleigh, NC 27695
June 2003 II-xxxiv DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont1 d)
Contributors and Reviewers
(cont'd)
Dr. William Wilson—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
CHAPTER 9. INTEGRATIVE SYNTHESIS: PARTICVLATE MATTER
ATMOSPHERIC SCIENCE, AIR QUALITY, HUMAN EXPOSURE,
DOSIMETRY, AND HEALTH RISKS
Principal Authors
Dr. William E. Wilson—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lester D. Grant—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Dennis J. Kotchmar—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Joseph P. Pinto—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. Robert Chapman—Retired, formerly at the National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dan Costa—National Health and Environmental Effects Research Laboratory (B305-02),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Bob Devlin—National Health and Environmental Effects Research Laboratory (MD-58),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Bill Russo—National Health and Environmental Effects Research Laboratory (B305-02),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
June 2003 II-xxxv DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont1 d)
Contributors and Reviewers
(cont'd)
John J. Vandenberg—National Center for Environmental Assessment (8601 D),
U.S. Environmental Protection Agency, 1200 Pennsylvania Avenue, NW, Washington, DC
20460
June 2003 II-xxxvi DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
Executive Direction
Dr. Lester D. Grant—Director, National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Scientific Staff
Dr. Robert W. Elias—PM Team Leader, Health Scientist, National Center for Environmental
Assessment (B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Dr. William E. Wilson—Air Quality Coordinator, Physical Scientist, National Center for
Environmental Assessment (B243-01), U.S. Environmental Protection Agency, Research
Triangle Park, NC 27711
Ms. Beverly Comfort—Health Scientist, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. J.H.B. Garner—Ecological Scientist, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Dennis J. Kotchmar—Medical Officer, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Joseph P. Pinto—Physical Scientist, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Technical Support Staff
Ms. Nancy Broom—Information Technology Manager, National Center for Environmental
Assessment (B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Mr. Douglas B. Fennell—Technical Information Specialist, National Center for Environmental
Assessment (B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Ms. Emily R. Lee—Management Analyst, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
June 2003 II-xxxvii DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Technical Support Staff
(cont'd)
Ms. Diane H. Ray—Program Specialist, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Donna Wicker—Administrative Officer, National Center for Environmental Assessment
(B243-01), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. Richard Wilson—Clerk, National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Document Production Staff
Dr. Carol A. Seagle—Technical Editor, Computer Sciences Corporation, 2803 Slater Road, Suite
220, Morrisville, NC 27560
Mr. Charles E. Gaul—Graphic Artist, Computer Sciences Corporation, 2803 Slater Road,
Suite 220, Morrisville, NC 27560
Mr. John M. Havel—Graphic Artist, Computer Sciences Corporation, 2803 Slater Road,
Suite 220, Morrisville, NC 27560
Ms. Jessica Long—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., 8200 Greensboro Drive, Suite 1450,
McLean, VA 22102
Ms. Laura A. Rhodes—Graphic Artist, Computer Sciences Corporation, 2803 Slater Road,
Suite 220, Morrisville, NC 27560
Mr. Keith A. Tarpley—Graphic Artist, Computer Sciences Corporation, 2803 Slater Road,
Suite 220, Morrisville, NC 27560
June 2003 II-xxxviii DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Technical Reference Staff
Mr. John A. Bennett—Technical Information Specialist, SANAD Support Technologies, Inc.,
11820 Parklawn Drive, Suite 400, Rockville, MD 20852
Ms. Lisa Greenbaum—Records Management Technician, Reference Retrieval and Database
Entry Clerk, InfoPro, Inc., 8200 Greensboro Drive, Suite 1450, McLean, VA 22102
Ms. Sandra L. Hughey—Technical Information Specialist, SANAD Support Technologies, Inc.,
11820 Parklawn Drive, Suite 400, Rockville, MD 20852
June 2003 II-xxxix DRAFT-DO NOT QUOTE OR CITE
-------
Abbreviations and Acronyms
4-POBN
A
ACE
ADS
AED
AHSMOG
AIC
AM
BAD
BAL
BALF
BAUS
BHR
BIC
BMI
BW
CAPs
CAT
CB
CESAR
CF
CFA
CFD
CHF
CUT
CL
a -(4-py ri dy 1 -1 -oxi de)-N-tert-buty Initrone
alveolar
acetone
anatomic dead space
aerodynamic equivalent diameter
Adventist Health Study on Smog
Akaike Information Criterion
alveolar macrophage
brachial artery diameter
bronchoalveolar lavage
brochoalveolar lavage fluid
brachial artery ultrasonography
bronchial hyperreactivity
Bayes Information Criterion
body mass index
bronchial wash
concentrated ambient particles
computer-aided tomography
chronic bronchitis
Central European Air Quality and Respiratory Health
cystic fibrosis
coal fly ash
computational fluid dynamics
congestive heart failure
Chemical Industry Institute of Technology
chemiluminescence
June 2003
II-xl
DRAFT-DO NOT QUOTE OR CITE
-------
CMD
CMP
CoH
COHb
CP
CPZ
CR
CRC
CrD
CSIRO
CVD
CVM
CX
DBF
DCFH
DCM
DE
DE
DEF
DEP
DHR
DMTU
DOFA
DPM
DRG
DTPA
DYS
ECG
count mean diameter
copper smelter dust
coefficient of haze
carboxyhemoglobin
coarse particle
capsazepine
concentration-response
contributing respiratory causes
cerebrovascular disease
Commonwealth Scientific and Industrial Research Organisation
cardiovascular disease
cardiovascular mortality
cyclohexane
diastolic blood pressure
di chl orofluore scin
dichloromethane
diesel exhaust
deposition efficiencies
Deferoxamine
diesel exhaust particles
dihy drorhodamine-123
dimethylthiourea
domestic oil fly ash
diesel particulate matter
dorsal root ganglia
techetium-diethylenetriamine-pentaaceticacid
dysrhythmias
el ectrocardi ogram
June 2003
DRAFT-DO NOT QUOTE OR CITE
-------
ED
EGA
EOF
EPEC
EPM
ER
ERK
ESR
ET
EU
FEF
FEVj
FMD
FP
FPD
FVC
G6PDH
GLM
GMCSF
GMPD
GP
GSF
GSH
HDM
HF
HR
LcBcc
ICAM-1
emergency department
evolved gas analysis
epidermal growth factor
Ecological Processes and Effects Committee
emission particulate matter
excess risk
extracellular receptor kinase
electron spin resonance
extrathoracic
endotoxin units
forced expiratory flow
forced expiratory volume in 1
flow-mediated dilation
fine particle
flame photometric detector
forced vital capacity
glucose-6-phosphate dehydrogenase
Generalized Linear Model
granulocyte macrophage colony stimulating factor
geometric mean particle diameter
general practice
Gessellschaft fur Strahlenforschung
glutathione
house dust mite
high frequency
heart rate
inhibitory kappa B alpha
intercellular adhesion molecule-1
June 2003
DRAFT-DO NOT QUOTE OR CITE
-------
ICD9
ICRP
IgE
IgG
IHD
IL
ip
IP
IQR
IUGR
INK
KS
LCL
LDH
LF
LFA-1
LN
LPS
LRD
LRI
MAPK
MAS
MC
MCM
MCT
MEK
MIP
MMAD
International Classification of Disease
International Commission on Radiological Protection
immunoglobin E
immunoglobin G
ischemic heart disease
interleukin
intraperitoneal
inhalable particle
interquartile range
intrauterine growth retardation
c-jun N-terminal kinase
soil-corrected potassium
lower 95th% confidence limit
lactate dehydrogenase
low frequency
leukocyte function-associated antigen-1
lymph nodes
lipopolysaccharide
lower respiratory disease
lower respiratory illness
mitogen-activated protein kinase
Mobil Aerosol Spectrometer
mass concentration
mass concentrations monitor
monocrotaline
mitogen-activated protein kinase
macrophage inflammatory protein
mean median aerodynamic diameter (see a )
June 2003
DRAFT-DO NOT QUOTE OR CITE
-------
MMD
MMPs
MPL
MPO
MPPD
MSH
NAC
NAL
NC
NCRP
NF
NF-KB
NHBE
NMD
NMD
NMMAPS
NMRI
Nn
NOPL
OAA
OLS
OTT
OVA
PB
PDGF
PDL
PEF
PFA
mass median diameter
matrix metalloproteinases
multipath lung
myeloperoxidase
multiple path particle dosimetry
Mount St. Helens
N-acetylcysteine (antioxidant)
nasal lavage fluid
number concentration
National Council on Radiation Protection and Measurement
nuclear factor
nuclear factor kappa B
normal human bronchial epithelial
nitroglycerine-mediated dilation
number mean diameter
National Morbidity, Mortality, and Air Pollution Study
Naval Medical Research Institute
numerical density of neutrophils
nasa-oro-pharyngo-laryngeal
Ottowa ambient air
ordinary least squares
Ottawa dust
ovalbumin
polymyxin-B
platelet-derived growth factor
polynomial distributed lag
peak expiratory flow
pulverized fuel ash
June 2003
DRAFT-DO NOT QUOTE OR CITE
-------
PFT
PHS-2
PMN
P°
poly I:C
PTFE
PVCs
QHIP
r-MSSD
RAIV
RAPS
RCAL
RIVM
ROFA
ROS
RR
RSP
RTE
SAD
SCA
SDANN
SDMM
SH
SHEDS
SIMEX
SIXE
SL
pulmonary function tests
prostaglandin H synthase-2
polymorphonuclear leukocytes
equilibrium vapor pressure
polyionosinic-polycytidilic acid
polytetrafluoroethylene
premature ventricular complexes
Quebec Health Insurance Plan
root mean squared differences between adjacent normal-to-normal
heartbeat intervals
rat-adapted influenza virus
Regional Air Pollution Study
Regression Calibration
Directorate-General for Environmental Protection
residual oil fly ash
reactive oxygen species
relative risk
respirable particulate matter
rat tracheal epithelial
small airway disease
sudden cardiac arrest
standard deviation of the average of normal-to-normal heartbeat
intervals
standard deviation of normal-to-normal heartbeat intervals
spontaneously hypertensive
Stochastic Human Exposure and Dose Simulation
Simulation Extrapolation
synchrotron induced X-ray emission
stochastic lung
June 2003
II-xlv
DRAFT-DO NOT QUOTE OR CITE
-------
SOD
SPM
Sp02
Stk
SWMMC
T(CO)
TB
TDF
TIMP
UAP
UCL
ufCB
UFP
URT
UVD
VA
VBE
VCAM-1
VMTD
VMTG
WEE
WIS
WKY
superoxide dismutase
synthetic polymer monomers
oxygen saturation
Stokes number
Southwest Metropolitan Mexico City
core temperature
tracheabronchial
total deposition fraction
tissue inhibitor of metalloproteinase
urban air particles
upper 95th% confidence limit
ultrafme carbon black
ultrafme fluorospheres
upper respiratory tract
Utah Valley dust
Veterans' Administration
Japanese B encephalitis
vascular cell adhesion molecule-1
vehicle miles of travel per mi2 per year by diesel
vehicle miles of travel per mi2 per year by gasoline
western equine encephalitis
Wistar
Wi star-Kyoto
June 2003
DRAFT-DO NOT QUOTE OR CITE
-------
i 6. DOSIMETRY OF PARTICULATE MATTER
2
3
4 6.1 INTRODUCTION
5 The proximal cause of any biological response to particulate matter (PM) is the dose
6 delivered to the target site rather than the external exposure. Characterization of the exposure-
7 dose-response continuum for PM requires an understanding of the mechanistic determinants of
8 inhaled particle dose. Furthermore, dosimetric information is critical to extrapolating to humans
9 health effects demonstrated by toxicological studies using experimental animals and for
10 comparing results from controlled clinical studies involving healthy human subjects and those
11 with preexisting respiratory disease.
12 Dose to target tissue depends on the initial deposition and subsequent retention of particles
13 within the respiratory tract. Once particles have deposited onto the surfaces of the respiratory
14 tract, they are subsequently subjected to either absorptive or nonabsorptive particulate removal
15 processes. This may result in their removal or translocation from airway surfaces, as well as
16 their removal from the respiratory tract itself. Clearance of deposited particles depends upon the
17 initial site of deposition and upon the physicochemical properties of the particles, both of which
18 affect specific translocation pathways. Retained particle burdens are determined by the dynamic
19 relationship between deposition and clearance rates.
20 This chapter is concerned with particle dosimetry, the study of the deposition,
21 translocation, clearance, and retention of particles within the respiratory tract and
22 extrapulmonary tissues. It summarizes basic concepts as presented in Chapter 10 of the 1996
23 EPA document, Air Quality Criteria for Particulate Matter or "PM AQCD" (U. S. Environmental
24 Protection Agency, 1996); and it updates the state of the science based upon new literature
25 appearing since publication of the 1996 PM AQCD. Although our understanding of the basic
26 mechanisms governing deposition and clearance of inhaled particles has not changed, there has
27 been significant additional information on the role of certain biological determinants of the
28 deposition/clearance processes, such as gender and age. Additionally, the understanding of
29 regional dosimetry within the respiratory tract and the particle size range over which this has
30 been evaluated has been expanded.
June 2003 6-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 The dose from inhaled particles deposited and retained in the respiratory tract is governed
2 by a number of factors. These include exposure concentration and exposure duration, respiratory
3 tract anatomy and ventilatory parameters, and physicochemical properties of the particles
4 themselves (e.g., particle size, hygroscopicity, and solubility in airway fluids and cellular
5 components). The basic characteristics of particles as they relate to deposition and retention, as
6 well as anatomical and physiological factors influencing particle deposition and retention, were
7 discussed in depth in the 1996 PM AQCD. Thus, in this chapter, only an overview of basic
8 information related to one critical factor in deposition, namely particle size, is provided (Section
9 6.1.1), so as to allow the reader to understand the different terms used in the remainder of this
10 chapter and in subsequent ones dealing with health effects. This is followed, in Section 6.1.2, by
11 a basic overview of respiratory tract structure as it relates to the deposition evaluation. The
12 ensuing major sections of this chapter provide updated information on particle deposition,
13 clearance, and retention in the respiratory tract of humans, as well as laboratory animals, which
14 are useful in the evaluation of PM health effects. Issues related to the phenomenon of particle
15 overload as it may apply to human exposure and the use of instillation of particle suspensions as
16 an exposure technique to evaluate PM health effects also are discussed. The final sections of the
17 chapter deal with mathematical models of particle disposition in the respiratory tract.
18 It must be emphasized that any dissection into discrete topics of factors that control dose
19 from inhaled particles tends to mask the dynamic and interdependent nature of the intact
20 respiratory system. For example, although deposition is discussed separately from clearance
21 mechanisms, retention (i.e., the actual amount of particles found in component regions of the
22 respiratory tract at any point in time) is, as noted previously, determined by the relative rates of
23 both deposition and clearance. Thus, assessment of overall dosimetry requires integration of
24 these various components of the overall process. In summarizing the literature on particle
25 dosimetry, when applicable, changes from control are described if they were statistically
26 significant at a probability (p) value less than 0.05 (i.e., p < 0.05). When trends are described, an
27 attempt will be made to provide the actual p values given in the published reports.
28
29
30
June 2003 6-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.1.1 Size Characterization of Inhaled Particles
2 Information about particle size distribution is important in the evaluation of effective
3 inhaled dose. Particle attributes, as well as some general definitions important in understanding
4 particle fate within the respiratory tract, are described in Chapter 2.
5 It is important to note that most aerosols present in natural and work environments are
6 polydisperse. This means that the constituent particles within an aerosol have a range of sizes
7 and are more appropriately described in terms of size distribution parameters. The log-normal
8 distribution (i.e., the situation in which the logarithms of particle diameter are distributed
9 normally) can be used for describing size distributions of most aerosols. The geometric mean is
10 the median of the distribution, and the metric of variability around this central tendency is the
11 geometric standard deviation (og). The og, a dimensionless term, is the ratio of the 84th (or 16th)
12 % particle size to the 50th % size. Thus, the only two parameters needed to describe a log normal
13 distribution of particle sizes for a specific aerosol are the median diameter and the geometric
14 standard deviation. However, the actual size distribution may be obtained in various ways. For
15 example, when a distribution is described by counting particles, the median is called the count
16 median diameter (CMD). On the other hand, the median of a distribution based on particle mass
17 in an aerosol is the mass median diameter (MMD). When using aerodynamic diameters, a term
18 that is encountered frequently is mass median aerodynamic diameter (MMAD), which refers to
19 the median of the distribution of mass with respect to aerodynamic equivalent diameter.
20 Although CMD might be more useful, most of the present discussion will focus on MMAD
21 because it is the most commonly used measure of aerosol distribution. However, alternative
22 distributions should be used for particles with actual physical sizes below about 0.5 jim because,
23 for these, aerodynamic properties become less important. One such metric is
24 thermodynamic-equivalent size, which is the diameter of a spherical particle that has the same
25 diffusion coefficient in air as the particle of interest.
26
27 6.1.2 Structure of the Respiratory Tract
28 A detailed discussion of respiratory tract structure was provided in the 1996 PM AQCD
29 (U.S. Environmental Protection Agency, 1996), and only a brief synopsis is presented here.
30 For dosimetry purposes, the respiratory tract can be divided into three regions (Figure 6-1):
31 (1) extrathoracic (ET), (2) tracheobronchial (TB), and (3) alveolar (A). The ET region consists
June 2003 6-3 DRAFT-DO NOT QUOTE OR CITE
-------
to
o
o
H
6
o
o
H
O
o
HH
H
W
T
Extrathoracic
Region
Pharynx
I
Posterior
Nasal Pasage
Nasal Part
Oral Part
Tracheo bronchial
Region
Alveolar
Region
Trachea
Main Bronchi
Bronchi
Bronchioles
Ciliated Bronchial
Epithelium
(Secretory and
Basal Cells)
Bronchiolar
Alveolar Interstitial
- Bronchioles
'Terminal Bronchioles
Respiratory Bronchioles
"Alveolar Duct +
Alveoli
Alveolar Endothelium,
Epithelium and Interstitium
(Endothelial Cells, Type II
Epithelial Cells and Clara Cells
Figure 6-1. Diagrammatic representation of respiratory tract regions in humans.
Source: U.S. Environmental Protection Agency (1996).
-------
1 of head airways (i.e., nasal and oral passages) through the larynx and represents the areas
2 through which inhaled air first passes. In humans, inhalation can occur through the nose or
3 mouth (or both, known as oronasal breathing). However, most laboratory animals commonly
4 used in respiratory toxicological studies are obligate nose breathers.
5 From the ET region, inspired air enters the TB region at the trachea. From the level of the
6 trachea, the conducting airways then undergo dichotomous branching for a number of
7 generations. The terminal bronchiole is the most peripheral of the distal conducting airways and,
8 in humans, leads to the gas-exchange region, which consists of respiratory bronchioles, alveolar
9 ducts, alveolar sacs, and alveoli (all of which comprise the A region). All of the conducting
10 airways, except the trachea and portions of the mainstem bronchi, are surrounded by
11 parenchymal tissue composed primarily of the alveolated structures of the A region and
12 associated blood and lymphatic vessels. It should be noted that the respiratory tract regions are
13 comprised of various cell types and that there are distinct differences in the distribution of cells
14 lining the airway surfaces in the ET, TB, and A regions. Although a discussion of cellular
15 structure of the respiratory tract is beyond the scope of this section, details may be found in a
16 number of sources (e.g., Crystal et al., 1997).
17
18
19 6.2 PARTICLE DEPOSITION
20 This section discusses the deposition of particles in the respiratory tract. It begins with an
21 overview of the basic physical mechanisms that govern deposition. This is followed by an
22 update on both total respiratory tract and regional deposition patterns in humans. Some critical
23 biological factors that may modulate deposition are then presented. The section ends with a
24 discussion of issues related to interspecies patterns of particle deposition.
25
26 6.2.1 Mechanisms of Deposition
27 Particles may deposit within the respiratory tract by five mechanisms: (1) inertial
28 impaction, (2) sedimentation, (3) diffusion, (4) electrostatic precipitation, and (5) interception.
29 Sudden changes in airstream direction and velocity may cause some particles to fail to
30 follow the streamlines of airflow. As a consequence, the particles contact, or impact, airway
31 surfaces. The ET and upper TB airways are characterized by high air velocities and sharp
June 2003 6-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 directional changes and, thus, dominate as sites of inertial impact!on. Impact!on is a significant
2 deposition mechanism for particles larger than 2 jim aerodynamic equivalent diameter (AED).
3 All aerosol particles are continuously influenced by gravity, but particles with an AED >
4 1 |im are affected to the greatest extent. A particle will acquire a terminal settling velocity when
5 a balance is achieved between the acceleration of gravity acting on the particle and the viscous
6 resistance of the air, and it is this settling out of the airstream that takes it into contact with
7 airway surfaces. Both sedimentation and inertial impaction can influence the deposition of
8 particles within the same size range. These deposition processes act together in the ET and TB
9 regions: inertial impaction dominates in the upper airways, and gravitational settling becomes
10 increasingly dominant in the smaller conducting airways.
11 Particles having actual physical diameters < 1 |im are subjected increasingly to diffusive
12 deposition because of random bombardment by air molecules, resulting in contact with airway
13 surfaces. The root mean square displacement that a particle experiences in a unit of time along a
14 given cartesian coordinate is a measure of its diffusivity. The density of a particle is unimportant
15 in determining particle diffusivity. Thus, instead of having an aerodynamic equivalent size,
16 diffusive particles of different shapes can be related to the diffusivity of a thermodynamic
17 equivalent size based on spherical particles.
18 The particle size range around 0.2 to 1.0 |im frequently is described as consisting of
19 particles that are small enough to be minimally influenced by impaction or sedimentation and
20 large enough to be minimally influenced by diffusion. Such particles are the most persistent in
21 inhaled air and undergo the lowest degree of deposition in the respiratory tract.
22 Interception is deposition by physical contact with airway surfaces. The interception
23 potential of any particle depends on its physical size. Fibers are of chief concern in relation to
24 the interception process. Their aerodynamic size is determined predominantly by their diameter,
25 but their length is the factor that influences probability of interception deposition.
26 Electrostatic precipitation is deposition related to particle charge. The minimum charge an
27 aerosol particle can have is zero. This condition rarely is achieved because of the random
28 charging of aerosol particles by air ions. Aerosol particles acquire charges by collisions with air
29 ions because of their random thermal motion. Many laboratory-generated aerosols are highly
30 charged and there are methods such as passage of the particle-containing airstream through a
31 Kr-85 charge neutralizer that eliminates the charge. In addition, these aerosols will generally
June 2003 6-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 lose their initial charge as they attract oppositely charged ions, and an equilibrium state of these
2 competing processes eventually is achieved. This Boltzmann equilibrium represents the charge
3 distribution of an aerosol in charge equilibrium with bipolar ions. The minimum amount of
4 charge is very small: there is a statistical probability that some particles within the aerosol will
5 have no charge and that others will have one or more positive and negative charges.
6 The electrical charge on some particles will result in an enhanced deposition over what
7 would be expected from size alone. This results from image charges induced on the surface of
8 the airway by these particles or to space-charge effects whereby repulsion of particles containing
9 like charges results in increased migration toward the airway wall. The effect of charge on
10 deposition is inversely proportional to particle size and airflow rate. This type of deposition is
11 often small compared to the effects of turbulence and other deposition mechanisms, and it
12 generally has been considered to be a minor contributor to overall particle deposition. However,
13 a study by Cohen et al. (1998), employing hollow airway casts of the human tracheobronchial
14 tree to assess deposition of ultrafine (0.02 jim) and fine (0.125 jim) particles, found the
15 deposition of singly charged particles to be 5 to 6 times that of particles having no charge and
16 2 to 3 times that of particles at Boltzmann equilibrium. This suggests that electrostatic
17 precipitation may, in certain situations such as workplace exposures or indoor tobacco smoke,
18 be a significant deposition mechanism for ultrafine, and some fine, particles within the TB
19 region. However, the influence of charge in the deposition of urban aerosols should be minimal.
20
21 6.2.2 Deposition Patterns in the Human Respiratory Tract
22 Knowledge of sites where particles of different sizes deposit in the respiratory tract and the
23 amount of deposition therein is necessary for understanding and interpreting the health effects
24 associated with exposure to particles. Particles deposited in the various respiratory tract regions
25 are subjected to large differences in clearance mechanisms and pathways and, consequently,
26 retention times. This section summarizes concepts of particle deposition in humans and
27 laboratory animals as reported in the 1996 PM AQCD (U.S. Environmental Protection Agency,
28 1996) and provides additional information based on studies published since that earlier
29 document.
30 Ambient air often contains particles too massive to be inhaled. The descriptor
31 "inhalability" is used to denote the overall spectrum of particle sizes that are potentially capable
June 2003 6-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 of entering the respiratory tract. Inhalability is defined as the ratio of the number concentration
2 of particles of a certain aerodynamic diameter that are inspired through the nose or mouth to the
3 number concentration of the same diameter particle present in ambient air (International
4 Commission on Radiological Protection, 1994). In general, for humans, unit density particles >
5 100 jim diameter have a low probability of entering the mouth or nose in still air, but there is no
6 sharp cutoff to zero probability. Additionally, there is no lower limit to inhalability, so long as
7 the particle exceeds a critical size where the aggregation of atomic or molecular units is stable
8 enough to endow it with "particulate" properties in contrast to those of free ions or gas
9 molecules.
10
11 6.2.2.1 Total Respiratory Tract Deposition
12 Total human respiratory tract deposition, as a function of particle size, is depicted in
13 Figure 6-2. These data were obtained by various investigators using different sizes of spherical
14 test particles in healthy male adults under different ventilation conditions; the large standard
15 deviations reflect inter-individual variability in airpath dimensions and airway branching and
16 breathing-pattern related variability of deposition efficiencies. Deposition in the ET region with
17 nose breathing is generally higher than that with mouth breathing because of the superior
18 filtration capabilities of the nasal passages which results in somewhat higher total deposition
19 with nasal breathing for particles > 1 |im. For particles with aerodynamic diameters greater than
20 1 |im, deposition is governed by impaction and sedimentation, and it increases with increasing
21 AED. When AED is > 10 |im, almost all inhaled particles are deposited. As the particle size
22 decreases from =0.5 jim, diffusional deposition becomes dominant and total deposition depends
23 more on the actual physical diameter of the particle. Decreasing particle diameter leads to an
24 increase in total deposition. Total deposition shows a minimum for particle diameters in the
25 range of 0.2 to 1.0 |im where, as noted above, neither sedimentation, impaction, or diffusion
26 deposition are very effective. Deposition never reaches zero because of mixing between
27 particle-rich tidal air and nearly particle-free residual lung air. The particles in the tidal air
28 remaining in the deep lung are gradually deposited.
29 Besides particle size, breathing pattern (tidal volume, breathing frequency, route of
30 breathing) is the most important factor affecting lung deposition. Kim (2000) reported total lung
31 deposition values in healthy adults for a wide range of breathing patterns, tidal volumes (375 to
June 2003 6-8 DRAFT-DO NOT QUOTE OR CITE
-------
100
90 -
80 -
70 -
Co"
e so -
c
o
:-E so H
w
o
§" 40 H
Q
30 -
20 -
10 -
0
Human (oral inhalation)
Human (nasal inhalation)
' I "I
0.01
0.1 1.0
Particle Diameter (pm)
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 um.
Source: Modified from Schlesinger (1989).
1 1500 mL), flow rates (150 to 1000 mL/s), and respiratory times (2 to 12 s). Total lung
2 deposition increased with increasing tidal volume at a given flow rate and with increasing flow
3 rate at a given respiratory time. Various deposition values were correlated with a single
4 composite parameter consisting of particle size, flow rate, and tidal volume.
5 The ultrafine mode (i.e., particles having diameters < 0.1 jam) is specifically being
6 evaluated for determination of its potential toxicity. There is, however, little information on
7 total respiratory tract deposition of such particles. Frampton et al. (2000) exposed healthy adult
June 2003
6-9
DRAFT-DO NOT QUOTE OR CITE
-------
1 human males and females, via mouthpiece, to 0.0267 jim diameter carbon particles (at 10 |ig/m3)
2 for 2 h at rest. The inspired and expired particle number concentration and size distributions
3 were evaluated. Total respiratory tract deposition fraction was determined for six particle size
4 fractions ranging from 0.0075 to 0.1334 jim. They found an overall total lung deposition
5 fraction of 0.66 (by particle number) or 0.58 (by particle mass), indicating that exhaled mean
6 particle diameter was slightly larger than inhaled diameter. There was no gender difference.
7 The deposition fraction decreased with increasing particle size within the ultrafine range, from
8 0.76 at the smallest size to 0.47 at the largest.
9 Jaques and Kim (2000) measured total deposition fraction (TDF) of ultrafine particles
10 (number median diameter [NMD] = 0.04-0.1 jim and og = 1.3) in 22 healthy adults (men and
11 women in equal number) under a variety of breathing conditions. The study was designed to
12 obtain a rigorous data set for ultrafine particles that could be applied to health risk assessment.
13 TDF was measured for six different breathing patterns: tidal volume (Vt) of 500 mL at
14 respiratory flow rates (Q) of 150 and 250 mL/s; V, = 750 mL at Q of 250 and 375 mL/s; V, = 1 L
15 at Q of 250 and 500 mL/s. Aerosols were monitored continuously by a modified condensation
16 nuclei counter during mouthpiece inhalation with the prescribed breathing patterns. For a given
17 breathing pattern, TDF increased as particle size decreased, regardless of the breathing pattern
18 used. For example, at V, = 500 mL and Q = 250 mL/s, TDF was 0.26, 0.30, 0.35, and 0.44 for
19 NMD = 0.10, 0.08, 0.06, and 0.04 |im, respectively (see Figure 6-3). For a given particle size,
20 TDF increased with an increase in V, and a decrease in Q, indicating an importance of breathing
21 pattern in assessing respiratory dose. The study also found that TDF was greater for women than
22 men at NMD = 0.04 |im within all breathing patterns used, but the difference was smaller or
23 negligible for larger-sized ultrafine particles. The results clearly demonstrate that the TDF of
24 ultrafine particles increases with a decrease of particle size and with breathing patterns of longer
25 respiratory time, a pattern that is consistent with deposition by diffusion mechanism. The results
26 also indicate that there is a differential lung deposition of ultrafine particles as small as 0.04 jam
27 for men versus women. These data are the only systematic human experimental data for
28 ultrafine particles reported since the 1996 PM AQCD.
29 A property of some ambient particulate species that affects deposition is hygroscopicity,
30 the propensity of a material for taking up and retaining moisture under certain conditions of
31 humidity and temperature. Ambient fine particles (sulfate, nitrate, and possibly organics) tend to
June 2003 6-10 DRAFT-DO NOT QUOTE OR CITE
-------
c
Q
1
2
LL
C
o
Q.
o>
Q
"i
4-*
|2
1.U ™
u.o
04 ™
n A -
T*
1
T
1
x'
>
^-
1
ujiu/iiui/jiijiuujj/nn/m
fftiniiminTmmtniifnmttrnmmiii 1
~L
it
^ ™ij
< Z±'i
, _j
' &~; i
:£•!
, j- — i 1
iy
—
5
^
x"^
1
^
*fl
^
i
^>'
$:!!
•5~iJ
"^^ * •
?!:
-r
• !S
i =
•i
cz;
EZ
cz
m
ZJ
13
m
ffi
Vt Q T f
1000ml, 250 ml/s. 8s, 7.5 bpm
1000ml, 500 mils. 8s 15 bpm
750ml. 250 ml/s, 6s. 10 bpm
750 ml, 375 ml/s, 4s, 15 bpm
500ml, 150 ml/s, 6.7s 9 bpm
500ml, 250 ml/s, 4s, 15 bpm
_
I
T T
£
1
^
B— .
^
t;
i
^
:H-
•£:
^
•3
=
; =
: 1
i S
T T .
I
i
f
i
!
i
1
^
i
t
i
i
TSc"
rii?
||
^
' ^ ?
: i
ii
0,02
0,04
006
008
0.10
0,12
Particle Diameter, pm
Figure 6-3. Total deposition fraction as a function of particle size in 22 healthy men and
women under six different breathing patterns. For each breathing pattern, the
total deposition fraction is different (p < 0.05) for two successive particle sizes.
Vt is tidal volume (mL); Q is respiratory flow rate (mL/s); T is respiratory
time (s); and f is breathing frequency in breaths/min (bpm).
Source: Jacques and Kim (2000).
1 be hygroscopic (see Chapter 2). Such particles can increase in size in the humid air within the
2 respiratory tract and, when inhaled, will deposit according to their hydrated size rather than their
3 initial size. The implications of hygroscopic growth on deposition have been reviewed
4 extensively by Morrow (1986) and Hiller (1991); whereas the difficulties of studying lung
5 deposition of hygroscopic aerosols have been reviewed recently by Kim (2000). In general,
6 compared to nonhygroscopic particles of the same initial size, the deposition of hygroscopic
1 aerosols in different regions of the lung may be higher or lower, depending on the initial size.
June 2003
6-11
DRAFT-DO NOT QUOTE OR CITE
-------
1 Thus, for particles with initial sizes larger than =0.5 |im, the influence of hygroscopicity would
2 be to increase total deposition with a shift from peripheral to central or extrathoracic regions;
3 whereas for smaller ones total deposition would tend to be decreased. See Chapter 2 for a
4 detailed description of particle hygroscopicity.
5
6 6.2.2.2 Deposition in the Extrathoracic Region
7 The fraction of inhaled particles depositing in the ET region is quite variable and
8 dependent on particle size, flow rate, breathing frequency, whether breathing is through the nose
9 or the mouth (Figure 6-4), and the cross-sectional area of the flow path. Mouth breathing
10 bypasses much of the filtration capabilities of the nasal airways and leads to increased deposition
11 in the lungs (TB and A regions). The ET region is clearly the site of first contact with particles
12 in the inhaled air and essentially acts as a "prefilter" for the lungs.
13 Since release of the 1996 PM AQCD, a number of studies have explored ET deposition
14 with in vivo studies, as well as in both physical and mathematical model systems. In one study,
15 the relative distribution of particle deposition between the oral and nasal passages was assessed
16 during "inhalation" by use of a physical model (silicone rubber) of the human upper respiratory
17 system, extending from the nostrils and mouth through the main bronchi (Lennon et al., 1998).
18 Monodisperse particles ranging in size from 0.3 to 2.5 jim were evaluated at flow rates ranging
19 from 15 to 50 L/min. Regional deposition in the oral passages, lower oropharynx-trachea, nasal
20 passages, and nasopharynx-trachea, as well as total deposition in the model, were assessed.
21 Deposition within the nasal passages was found to agree with available data obtained from a
22 human inhalation study (Heyder and Rudolf, 1977), being proportional to particle size, density,
23 and inspiratory flow rate. It also was found that for oral inhalation, the relative distribution of
24 particle deposition between the oral cavity and the oropharynx-trachea was similar; whereas for
25 nasal inhalation, the nasal passages contained most of the particles deposited in the model, with
26 only about 10% deposited in the nasopharynx-trachea region. Furthermore, the deposition
27 efficiency of the nasopharynx-trachea region was greater than that of the oropharynx-trachea
28 region. For simulated oronasal breathing, deposition in the ET region depended primarily on
29 particle size rather than flow rate. For all flows and for all breathing modes, total deposition in
30 the ET region increased as particle diameter increased. Such information on deposition patterns
31 in the ET region is useful in refining empirical deposition models.
June 2003 6-12 DRAFT-DO NOT QUOTE OR CITE
-------
100
90 -
80 -
70 -
St 60 -
c
o
S 50 H
CO
O
§• 40 -
Q
30 -
20 -
10 -
0
0.01
A Human (oral inhalation)
A Human (nasal inhalation)
0.1 1.0
Particle Diameter (|jm)
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 jim and geometric (or diffusion equivalent) for those < 0.5 um.
Source: Modified from Schlesinger (1989).
1 Deposition within the nasal passages was further evaluated by Kesavanathan and Swift
2 (1998), who examined the deposition of 1- to 10-um particles in the nasal passages of normal
3 adults under an inhalation regime in which the particles were drawn through the nose and out
4 through the mouth at flow rates ranging from 15 to 35 L/min. At any particle size, deposition
5 increased with increasing flow rate; whereas deposition increased with increasing particle size at
6 any flow rate. In addition, as was shown experimentally by Lennon et al. (1998) under oronasal
7 breathing conditions, deposition of 0.3- to 2.5-um particles within the nasal passages was
8 significantly greater than within the oral passages, and nasal inhalation resulted in greater total
June 2003
6-13
DRAFT-DO NOT QUOTE OR CITE
-------
1 deposition in the model than did oral inhalation. These results are consistent with other studies
2 discussed in the 1996 PM AQCD and with the known dominance of impaction deposition within
3 the ET region.
4 Rasmussen et al. (2000) measured deposition in the nasal cavity of normal adult humans of
5 0.7 |im particles consisting of sodium chloride and radioactively-labeled technetium-
6 diethylenetriamine-pentaacetic acid (DTPA). Each subject inhaled one liter for each inspiration
7 at flow rates ranging from 10-30 L/min. They found that the deposition fraction in the nasal
8 passages increased as flow rate increased and that an estimate of maximum linear air velocity
9 was the best single predictor of nasal deposition fraction.
10 For ultrafine particles (dp < 0.1 |im), deposition in the ET region is controlled by diffusion,
11 which depends only on the particle's geometric diameter. Prior to 1996, ET deposition for this
12 particle size range had not been studied extensively in humans, and this remains the case. In the
13 1996 PM AQCD, the only data available for ET deposition of ultrafine particles were from
14 hollow airway cast studies. More recently, deposition in the ET region was examined using
15 mathematical modeling. Three- dimensional numerical simulations of flow and particle
16 diffusion in the human upper respiratory tract, which included the nasal region, oral region,
17 larynx, and first two generations of bronchi, were performed by Yu et al. (1998). Deposition of
18 particles of 0.001 and 0.01 |im in these different regions was calculated under inspiratory and
19 expiratory flow conditions. Deposition efficiencies in the total model were lower on expiration
20 than inspiration although values for the former were quite high. About 75% of the ultrafine
21 particles were deposited. Nasal deposition accounted for up to 54% of total deposition in the
22 model system for 0.001-|im particles. With oral breathing, deposition efficiency was estimated
23 at 48% (of amount entering; Yu et al., 1998).
24 Swift and Strong (1996) examined the deposition of ultrafine particles, ranging in size
25 from 0.053 to 0.062 jim, in the nasal passages of normal adults during constant inspiratory flows
26 of 6 to 22 L/min. The results are consistent with results noted in studies above, namely that the
27 nasal passages are highly efficient collectors for ultrafine particles. In this case, fractional
28 deposition ranged from 94 to 99% (of amount inhaled). Only a weak dependence of deposition
29 on flow rate was found, which contrasts with results noted above (i.e., Lennon et al., 1998) for
30 particles > 0.3 jim, but is consistent with diffusion being the main deposition mechanism. This
31 report has important implications for assessing the toxicity of PM because the filtration
June 2003 6-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 efficiency of the nasal passages will lessen the probability of ultrafme particle deposition in the
2 lungs.
3 Cheng et al. (1997) examined oral airway deposition in a replicate cast of the human nasal
4 cavity, oral cavity, and laryngeal-tracheal sections. Particle sizes ranged from
5 0.005 to 0.150 |im, and constant inspiratory and expiratory flow rates of 7.5 to 30 L/min were
6 used. They noted that the deposition fractions within the oral cavity were essentially the same as
7 that in the laryngeal-tracheal sections for all particle sizes and flow rates. They ascribed this to
8 the balance between flow turbulence and residence time in these two regions. Svartengren et al.
9 (1995) examined the effect of changes in external resistance on oropharyngeal deposition of
10 3.6-|im particles in asthmatics. Under controlled mouthpiece breathing conditions (flow rate
11 0.5 L/s), the median deposition as a percentage of inhaled particles in the mouth and throat was
12 20% (mean = 33%; range 12 to 84%). Although the mean deposition fell to 22% with added
13 resistance, the median value remained at 20% (range 13 to 47%). Fiberoptic examination of the
14 larynx revealed that there was a trend for increased mouth and throat deposition associated with
15 laryngeal narrowing. On the basis of mathematical model calculations, Katz et al. (1999) found
16 that turbulence plays a key role in enhancing particle deposition in the larynx and trachea.
17 The results of all of the above studies support the previously known ability of the ET
18 region, especially the nasal passages, to act as an efficient filter for nanoparticles (< 0.1 jim) as
19 well as for larger ones (> 5 jim), potentially reducing the amount of particles within a wide size
20 range that are available for deposition in the TB and A regions.
21
22 6.2.2.3 Deposition in the Tracheobronchial and Alveolar Regions
23 Particles that do not deposit in the ET region of the respiratory tract enter the lungs;
24 however, their regional deposition within the lungs cannot be precisely measured. Much of the
25 available deposition data for the TB and A regions have been obtained from experiments with
26 radioactively labeled, poorly soluble particles (Figures 6-5 and 6-6, respectively). These have
27 been described previously (U.S. Environmental Protection Agency, 1996). Although there are
28 no new regional data obtained by means of the radioactive aerosol method since the publication
29 of that document, a novel serial bolus delivery method has been introduced. Using this bolus
30 technique, regional deposition has been measured for fine and coarse aerosols (Kim et al., 1996;
31 Kim and Hu, 1998) and for ultrafme aerosols (Kim and Jacques, 2000). The serial bolus method
June 2003 6-15 DRAFT-DO NOT QUOTE OR CITE
-------
c
g
"-t—t
O)
o
Q.
0)
Q
c
o
Q.
0)
Q
60-
50-
40-
30-
20-
10-
1.0
70-
60-
50-
40-
30-
20-
10-
A
Particle Diameter (|jm)
A A
A
A
0.1 1.0
Particle Diameter (|jm)
Figure 6-5.
Tracheobronchial deposition (oral
inhalation 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 urn and
geometric (or diffusion equivalent) for
those < 0.5 jim.
Source: Modified from Schlesinger (1989).
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 jim and geometric (or
diffusion equivalent) for those < 0.5 urn.
Source: Modified from Schlesinger (1989).
10
June 2003
6-16
DRAFT-DO NOT QUOTE OR CITE
-------
1 uses nonradioactive aerosols and can estimate regional deposition in a virtually unlimited
2 number of lung compartments. Because of experimental limitations of the technique, the
3 investigators estimated regional lung deposition in ten serial, 50-mL increments from the mouth
4 to the end of a typical 500-mL tidal volume. Deposition estimates in the TB and A regions were
5 obtained for both men and women for particles ranging from 0.04 to 5.0 jim in diameter.
6 It should be noted that particle deposition in the TB and A regions was based on volumetric
7 compartments of 50- to 150-mL and > 150 mL, respectively. Deposition in the ET region was
8 based on the 0- to 50-mL compartment. Lung deposition fractions are shown in Figure 6-7.
9 In men, 24-32% of total particle deposition (0.04-, 0.06-, 0.08-, and 0.10-|im particles) was
10 deposited in the TB region and 67-76% was deposited in the A region. In women, deposition of
11 these particles was consistently greater in the TB region (21-48%), but was comparable or
12 slightly smaller in the A region as compared to men. As a result, total lung deposition of
13 ultrafine particles was slightly greater in women than men (-5-14%). For 1-, 3-, and 5-|im
14 particles, 16-37% of total particle deposition, in men, was deposited in the TB region and
15 57-83% was deposited in the A region. Deposition of these size particles was consistently
16 greater in the TB region in women (27-68%), but was comparable or slightly smaller in the
17 A region as compared to men. As a result, total lung deposition was slightly greater in women
18 than men (-16-22%). Thus, deposition of ultrafine and coarse particles in the TB region was
19 greater for women than men.
20 Fine particles that penetrate to the gas exchange airways are deposited on airway
21 bifurcations at higher concentrations. The deposition diminishes rapidly with airway generation,
22 consistent with the concentration of streamlines near the bifurcations and the penetration depth
23 of convective tidal flow.
24 Brody and Roe (1983) studied the deposition pattern of 5 aerosolized dusts (chrysotile and
25 crocidolite asbestos, fiber glass, cc-quartz, and ash from Mt. St. Helens) in the lungs of rats.
26 Mice were exposed to chrysotile asbestos. Quantitative electron microscopy was carried out on
27 tissues fixed by vascular perfusion. Immediately following a brief exposure, a significantly
28 greater number of particles had deposited on alveolar duct bifurcations when compared with the
29 number of particles on duct surfaces adjacent to the bifurcations. Few particles were counted at
30 midpoints between bifurcations, and particles were rarely observed within alveoli. The data
31 show that regardless of mineral nature, shape, or concentration, inhaled particles small enough to
June 2003 6-17 DRAFT-DO NOT QUOTE OR CITE
-------
o
"G
ro
w
o
Q.
0)
Q
Male
Female
Vt = 500ml_
Q = 250 mL/s
0.04
0.06
0.08
0.10
Particle Diameter (pm)
Figure 6-7. Lung deposition fractions in the tracheobronchial (TB) and alveolar (A)
regions estimated 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 um in diameter,
respectively, for men. In comparison to men, TB deposition in women was
27-68% greater, whereas A deposition was comparable. For ultrafine
particles of 0.04 to 0.1 urn diameter, TB and A deposition in men ranged from
5.7 to 15.6% and 18.2 to 33.1%, respectively. In comparison to men, TB
deposition was 21-48% greater, whereas A deposition was comparable. 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).
June 2003
6-18
DRAFT-DO NOT QUOTE OR CITE
-------
1 pass through the conducting airways are deposited primarily at alveolar duct bifurcations. The
2 authors proposed that the alveolar deposition patterns are the result of airflow characteristics that
3 cause enhanced deposition of particles at alveolar duct bifurcations intersecting the flow and is
4 similar to deposition patterns that occur at bifurcations of conducting airways.
5 Brody et al. (1981) studied the initial deposition and subsequent translocation of chrysotile
6 asbestos in the lungs of rats exposed for 1 h. Using scanning and transmission electron
7 microscopy of tissue fixed by vascular perfusion, they determined that the majority of fibers that
8 pass through the conducting airways deposit at the bifurcations of alveolar ducts. The farther an
9 alveolar duct bifurcation was from its terminal bronchiole, the less asbestos was observed.
10 Warheit and Hartsky (1990) compared inhaled-particle-deposition patterns in alveolar
11 regions of four rodent species with differing airway branching patterns and poorly developed
12 respiratory bronchioles. Proximal alveolar regions of hamsters and guinea pigs contain
13 rudimentary respiratory bronchioles; whereas in rats and mice, terminal bronchioles lead directly
14 into alveolar ducts. Groups of animals from one strain each of rats, mice, hamsters, and guinea
15 pigs were exposed to aerosols of carbonyl iron (CI) particles (100 mg/m3) for 1 h. Total lung
16 deposition of iron particles was highest in mice and hamsters. Particle deposition patterns in the
17 proximal regions of the distal lung were similar for all species although greater numbers of CI
18 particles per bifurcation were deposited in rats and mice compared to hamsters, and greater
19 numbers were deposited in hamsters compared to guinea pigs. The data suggest that the
20 presence of undeveloped respiratory bronchioles in the lungs of hamsters and guinea pigs has
21 little influence on distal lung particle deposition patterns. It is not known whether inhaled
22 particles are deposited at similar sites in the lungs of species with well-developed respiratory
23 bronchioles such as cats, nonhuman primates, and humans.
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 in the 1996 PM AQCD. Basically, using deposition data from living
31 subjects as well as from mathematical and physical models, enhanced deposition has been shown
June 2003 6-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 to occur in the nasal passages and trachea and at branching points in the TB and A regions (see
2 Chapter 10 of U.S. Environmental Protection Agency, 1996). Churg and Vedal (1996) examined
3 retention of particles on carinal ridges and tubular sections of airways from lungs obtained at
4 necropsy. Results indicated significant enhancement of particle retention on carinal ridges
5 through the segmental bronchi; the ratios were similar in all airway generations examined.
6 Kim and Fisher (1999) studied local deposition efficiencies and deposition patterns of
7 aerosol particles (2.9 to 6.7 jim) in sequential double-bifurcation tube-models with two different
8 branching geometries: one with in-plane (A) and another with out-of-plane (B) bifurcation. The
9 deposition efficiencies (DE) in each bifurcation increased with increasing Stokes number (Stk).
10 (The Stokes number is used to characterize the ability of a particle to follow a streamline in
11 curvilinear motion. It is the ratio of the stopping distance of a particle to a characteristic
12 dimension of the obstacle. As the Stokes number increases, particles tend to become less able to
13 follow a streamline around an obstacle and more likely to impact the obstacle [Hinds, 1999]).
14 With symmetric flow conditions, DE was somewhat smaller in the second than the first
15 bifurcation in both models. DE was greater in the second bifurcation in model B than in model
16 A. With asymmetric flows, DE was greater in the low-flow side compared to the high-flow side;
17 this was consistent in both models. Deposition pattern analysis showed highly localized
18 deposition on and in the immediate vicinity of each bifurcation ridge, regardless of branching
19 and flow patterns.
20 Comer et al. (2000) used a three-dimensional computer simulation technique to investigate
21 local deposition patterns in sequentially bifurcating airway models that were previously used in
22 experiments by Kim and Fisher (1999). The simulation was for 3-, 5-, and 7-|im particles and
23 assumed steady, laminar, constant air flow with symmetry about the first bifurcation. The
24 overall trend of the particle deposition efficiency, i.e., an exponential increase with Stokes
25 number, was similar for all bifurcations; and deposition efficiencies in the bifurcation regions
26 agreed very well with experimental data. Local deposition patterns consistently showed that the
27 majority of the deposition occurred within the carinal region.
28 Deposition "hot spots" at airway bifurcations have undergone additional analyses using
29 mathematical modeling techniques. Using calculated deposition sites, a strong correlation has
30 been demonstrated between secondary flow patterns and deposition sites and density both for
31 large (10 jim) particles and for ultrafine particles (0.01 jim; Heistracher and Hofmann, 1997;
June 2003 6-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hofmann et al., 1996). This supports experimental work, noted in U.S. Environmental
2 Protection Agency (1996), indicating that, like larger particles, ultrafme particles also show
3 enhanced deposition at airway branch points — even in the upper tracheobronchial tree.
4 The pattern of particle distribution on a more regional scale was evaluated by Kim et al.
5 (1996) and Kim and Hu (1998). Deposition patterns were measured in situ in nonsmoking
6 healthy young adult males using an aerosol bolus technique that delivered 1-, 3-, or 5-|im
7 particles into specific volumetric depths within the lungs. The distribution of particle deposition
8 was uneven; and it was noted that sites of peak deposition shifted from distal to proximal regions
9 of the lungs with increasing particle size (Figure 6-8). Furthermore, the surface dose was found
10 to be greater in the conducting airways than in the alveolar region for all of the particle sizes
11 evaluated. Within the conducting airways, the largest airway regions (i.e., 50 to 100 mL volume
12 distal to the larynx) received the greatest surface doses.
13 Bennett et al. (1998) studied the effect of variable anatomic dead space (ADS) on aerosol
14 bolus delivery in healthy subjects inhaling radiolabeled, 99mTc-iron oxide particles (3.5 jim).
15 The subjects inhaled 40 mL aerosol boluses to a volumetric front depth of 70 mL into the lung at
16 a lung volume of 70% total lung capacity end-inhalation and estimated the fraction of the inhaled
17 boluses deposited in intrathoracic airways (IDF). ADS was also measured from 70% total lung
18 capacity. The IDF deposition fraction varied from 0.04 - 0.43 and increased with decreasing
19 ADS. The deposited dose in the IDF was lower in subjects with large ADS (> 250 mL).
20 A lower dose to the IDF was also noted in women due to a smaller IDF and smaller airspace
21 dimensions. They observed significantly greater deposition in the left (L) versus right lungs (R);
22 mean L/R (ratio of deposition in L lung to R lung, normalized to ratio of L-to-R lung volume)
23 was 1.58 ± 0.42. Retention of deposited particles at 2 h was independent of ADS or IDF. There
24 was significant retention of particles in the whole lung at 24 h post deposition and slow
25 clearance of these particles continued through 48 h post deposition. There was significant
26 retention of insoluble particles in large bronchial airways at 24 h post deposition (i.e., 24 h
27 central-to-peripheral ratio = 1.40 and 1.82 in the R and L lung, respectively).
28 Kim and Jaques (2000) used the respiratory bolus technique to estimate the deposition
29 distribution of ultrafme particles (0.04, 0.06, 0.08, and 0.1 jim) in young adults. Under normal
30 breathing conditions (tidal volume 500 mL, respiratory flow rate 250 mL/s), bolus aerosols were
31 delivered sequentially to a lung depth ranging from 50 to 500 mL in 50-mL increments. The
June 2003 6-21 DRAFT-DO NOT QUOTE OR CITE
-------
0.14
C
.0
"o
CO
c
o
'(7)
o
Q.
0
Q
"CD
o
o
0.2
0.1 •
0.0
Volumetric Lung Region (mL)
Figure 6-8. Estimated lung deposition fractions in ten volumetric regions for particle sizes
ranging from ultrafine particle diameter (dp) of 0.04 to 0.01 urn (Panel A) to
fine (dp = 1.0 um; Panel B) and coarse (dp = 3 and 5 urn; 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 Jaques (2000).
June 2003
6-22
DRAFT-DO NOT QUOTE OR CITE
-------
1 results indicate that regional deposition of ultrafme particles (0.4-1.0 jim) varies widely along
2 the depth of the lung. Regional deposition of larger particles (1.0-5.0 jim) is far less variable
3 (Figure 6-8). The deposition patterns for ultrafme particles, especially for very small ultrafme
4 particles, were similar to those for coarse particles. Peak deposition occurred in the lung regions
5 situated between 150 and 200 mL from the mouth, and sites of peak deposition shifted
6 proximally with decreasing particle size. Deposition dose per unit average surface area was
7 greatest in the proximal lung regions and decreased rapidly with increased lung depth. Peak
8 surface dose was 5 to 7 times greater than average lung dose. These results indicate that local
9 enhancement of dose occurs in healthy lungs, which could be an important factor in eliciting
10 pathophysiological effects.
11
12 6.2.2.5 Deposition of Specific Size Modes of Ambient Aerosol
13 Several recent modeling studies provide estimates of the deposition profiles of "real world"
14 particle size fractions. One such study using a lung-anatomical model (Venkataraman and Kao,
15 1999) examined the contribution of two specific size modes of the PM10 ambient aerosol, namely
16 the fine mode (defined as particles with diameters up to 2.5 jim) and the thoracic fraction of the
17 coarse mode (defined as particles with diameters 2.5 to 10 jim) to total lung and regional lung
18 doses (i.e., a daily dose expressed as jig/day, and a surface dose expressed a |ig/cm2/day)
19 resulting from a 24-h exposure to a particle concentration of 150 |ig/m3. The study also
20 evaluated deposition in terms of two metrics, namely mass dose and number dose. Deposition
21 was calculated using a mathematical model for a healthy human lung under both simulated
22 moderate exertion (1 L at 15 breaths/min) and vigorous exertion (1.5 L at 15 breaths/min) and
23 for a compromised lung (0.5 L at 30 breaths/min). Regional deposition values were obtained for
24 the ET, TB, and A regions. Because the exposure scenario used is quite unrealistic, only general
25 trends should be inferred from this study rather than actual deposition values. These estimates
26 would also be highly uncertain for the compromised lung.
27 The daily modeled mass dose peaked in the A airways for all breathing patterns; whereas
28 that for the coarse fractions was comparable in the TB and A regions. The mass per unit surface
29 area of various airways from the fine and coarse fractions was larger in the trachea and first few
30 generations of bronchi. It was suggested that these large surface doses may be related to
31 aggravation of upper respiratory tract illness.
June 2003 6-23 DRAFT-DO NOT QUOTE OR CITE
-------
1 The modeled daily number dose was different for fine and coarse fractions in all lung
2 airways: the dose from the fine fraction was higher by about 100 times in the ET and about 10s
3 times in internal lung airways. The surface number dose (particles/cm2/day) was 103 to 10s times
4 higher for fine than for coarse particles in all lung airways, indicating the larger number of fine
5 particles depositing. Particle number doses did not follow trends in mass doses and are much
6 higher for fine than coarse particles and are higher for different breathing patterns. It also was
7 concluded that the fine fraction contributes 10,000 times greater particle number per alveolar
8 macrophage than the coarse fraction particles. As noted, these results must be viewed with
9 caution because they were obtained using a pure mathematical model that must be validated in
10 terms of realistic physiologic conditions.
11 Another evaluation of deposition that included consideration of size mode of the ambient
12 aerosol was that of Broday and Georgopoulos (2001). In this case, a mathematical model was
13 used to account for particle hygroscopic growth, transport, and deposition in tracking the
14 changes in the size distribution of inhaled aerosols. It was concluded that different rates of
15 particle growth in the inspired air resulted in a change in the aerosol size distribution such that
16 increased mass and number fractions of inspired ultrafine particles (< 0.1 jim) were found in the
17 size range between 0.1 to 1 |im and, therefore, deposited to a lesser extent due to a decrease in
18 diffusion deposition. On the other hand, particles that were originally in the 0.1- to l-|im size
19 range when inhaled will undergo enhanced deposition because of their increase in size resulting
20 from hygroscopic growth. Hence, the initial size distribution of the inhaled poly disperse aerosol
21 affects the evolution of size distribution once inhaled and, thus, its deposition profile in the
22 respiratory tract. Hygroscopicity of respirable particles must be considered for accurate
23 predictions of deposition. Because different size fractions likely have different chemical
24 composition, such changes in deposition patterns will affect biological responses.
25
26 6.2.3 Biological Factors Modulating Deposition
27 Experimental deposition data in humans have been commonly derived using healthy adult
28 Caucasian males. Various factors can act to alter deposition patterns from those obtained in this
29 group. Evaluation of these factors is important to help understand potentially susceptible
30 subpopulations because differences in biological response following pollutant exposure may be
31 caused by dosimetry differences as well as by differences in innate sensitivity. The effects of
June 2003 6-24 DRAFT-DO NOT QUOTE OR CITE
-------
1 different biological factors on deposition were discussed in the 1996 PM AQCD (U.S.
2 Environmental Protection Agency, 1996) and are summarized below together with additional
3 information obtained from more recent studies.
4
5 6.2.3.1 Gender
6 Males and females have different body size, conductive airway size, and ventilatory
7 parameter distributions; therefore, it is expected that there would be gender differences in
8 deposition. In some of the controlled studies, however, men and women are breathing at the
9 same tidal volume and frequency. If the women are generally smaller than the men, the
10 increased minute ventilation compared to their normal ventilation would cause different changes
11 in deposition patterns. In these cases, it would be better for the investigators to have used size-
12 adjusted tidal volumes. This may help to explain some of the differing results discussed below.
13 Using particles in the 2.5- to 7.5-|im size range, Pritchard et al. (1986) indicated that, for
14 comparable particle sizes and inspiratory flow rates, females had higher ET and TB deposition
15 and smaller A deposition than did males. The ratio of A deposition to total thoracic deposition
16 in females also was found to be smaller. These differences were attributed to gender differences
17 in airway size.
18 In another study (Bennett et al., 1996), the total respiratory tract deposition of 2-|im
19 particles was examined in adult males and females aged 18 to 80 years who breathed with a
20 normal resting pattern. Deposition was assessed in terms of a deposition fraction, the difference
21 between the amount of particles inhaled and exhaled during oral breathing. Although there was
22 a tendency for a greater deposition fraction in females compared to males, and because males
23 had greater minute ventilation, the deposition rate (i.e., deposition per unit time) was greater in
24 males than in females.
25 Kim and Hu (1998) assessed regional deposition patterns in healthy adult males and
26 females using particles with median aerodynamic sizes of 1, 3, and 5 |im and a bolus delivery
27 technique that involved controlled breathing. The total fractional deposition in the lungs was
28 similar for both genders with the smallest particle size, but was greater in women for the 3- and
29 5-|im particles regardless of the inhalation flow rate used; this difference ranged from 9 to 31%,
30 with higher values associated with higher flow rates. The pattern of deposition was similar for
31 both genders although females showed enhanced deposition peaks for all three particle sizes.
June 2003 6-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 The volumetric depth location of these peaks was found to shift from peripheral (i.e., increased
2 volumetric depth) to proximal (i.e., shallow volumetric depth) regions of the lung with
3 increasing particle size, but the shift was greater in females than in males. Thus, deposition
4 appeared to be more localized in the lungs of females compared to those of males. These
5 differences were attributed to the smaller size of the upper airways, particularly of the laryngeal
6 structure, in females. Local deposition of l-|im particles was somewhat flow dependent but, for
7 larger (5-jim) particles, was largely independent of flow (flows did not include those that would
8 be typical of exercise).
9 In a related study, Kim et al. (2000) evaluated differences in deposition between males and
10 females under varying breathing patterns (simulating breathing conditions of sleep, resting, and
11 mild exercise). Using particles at the same size noted above and a number of breathing
12 conditions, total fractional lung deposition was comparable between men and women for l-|im
13 particles, but was slightly greater in women than men for 3- and 5-|im particles with all
14 breathing patterns. The gender difference was about 15% at rest and variable during exercise
15 depending on particle size. However, total lung deposition rate (i.e., deposition per unit time)
16 was found to be 3 to 4 times greater during moderate exercise than during rest for all particle
17 sizes. Thus, it was concluded that exercise may increase the health risk from particles because
18 of increased large airway deposition and that women may be more susceptible to this exercise-
19 induced change.
20 Jaques and Kim (2000) and Kim and Jaques (2000) expanded the evaluation of deposition
21 in males and females to particles < 1 jim. They measured total fractional lung deposition in
22 healthy adults using sizes in the ultrafine mode (0.04 to 0.1 jim) in addition to those having
23 diameters of 1 and 5 jim. Total fractional lung deposition was greater in females than in males
24 for 0.04- and 0.06-|im particles. The difference was negligible for 0.08- and 0.1-|im particles.
25 Therefore, the gender effect was particle-size dependent, showing a greater fractional deposition
26 in females for very small ultrafine and large coarse particles, but not for particles ranging from
27 0.08 to 1 |im. A local deposition fraction was determined in each volumetric compartment of the
28 lung to which particles are injected based on the inhalation procedure (Kim and Jaques, 2000).
29 The fractional deposition was found to increase with increasing lung depth from the mouth,
30 reach a peak value, and then decrease with further increase in lung volumetric depth. The height
31 of the peak and its depth varied with particle size and breathing pattern. Peak fractional
June 2003 6-26 DRAFT-DO NOT QUOTE OR CITE
-------
1 deposition for the 5-|im particles was more proximal than that for the l-|im particles; whereas
2 that for the ultrafine particles occurred between these two peaks. For the ultrafine particles, the
3 peak fractional deposition became more proximal as particle size decreased. Although this
4 pattern of deposition distribution was similar for both men and women, the region of peak
5 fractional deposition was shifted closer to the mouth and peak height was slightly greater for
6 women than for men for all exposure conditions.
7
8 6.2.3.2 Age
9 Airway structure and respiratory conditions vary with age, and these variations may alter
10 the deposition pattern of inhaled particles (Table 6-1). The limited experimental studies reported
11 in the 1996 PM AQCD (U. S. Environmental Protection Agency, 1996) indicated results ranging
12 from no clear dependence of total deposition on age to slightly higher deposition in children than
13 adults. However, children have a different resting ventilation than do adults. The experimental
14 studies must adjust for the higher minute ventilation per unit body weight in children when
15 comparing deposition results to those obtained in adults.
16
17 Modeled Deposition Patterns
18 Potential regional deposition differences between children and adults have been assessed to
19 a greater extent using mathematical models. These indicated that, if the entire respiratory tract
20 and a complete breathing cycle at normal rate are considered, then ET deposition in children
21 would be generally higher than that in adults. However, TB and A regional deposition in
22 children may be either higher or lower than that in adults, depending on particle size (Xu and
23 Yu, 1986). There is enhanced TB deposition in children for particles < 5 jim (Xu and Yu, 1986;
24 Hofmann et al., 1989a). Becquemin et al. (1991) compared nasal filtering efficiency in children
25 and adults; two groups of children (12 children, aged 5.5-11.5 y; 8 children, aged 12-15 y) were
26 studied along with 10 adults. The deposition of polystyrene beads (1, 2.05, 2.8 |im MMAD) was
27 measured for both nose and mouth breathing. Ventilation was controlled to scale breathing
28 patterns appropriate for each age either at rest or during moderate exercise. Anterior nasal
29 resistance and standard lung function were measured for each subject. For the same inhalation
30 flow rate, children had much higher nasal resistances than adults. Individually, nasal deposition
31 increased with particle size, ventilation flow rate and nasal resistance, from rest to exercise.
June 2003 6-27 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-1. EFFECTS OF AGE ON PARTICLE DEPOSITION IN RESPIRATORY TRACT
to
o
o
OJ
ON
to
oo
o
£
H
6
0
2;
o
H
0
0
H
W
0
O
1 — I
H
W
Type study
Inhalation
Inhalation
Inhalation
Inhalation
Airway models
Nasal casts
Model
Model
Model
Model
C.F = Cvstic fibrns
Particles
(MMAD) Summary
1 um Measured deposition of particles in children, adolescents, and adults. No differences in
deposition among three groups. Breath-to-breath fractional deposition in children increased
with increasing tidal volume. Rate of deposition normalized to lung surface area tended to be
35% greater in children compared to adolescents and adults.
4.5 um Particles inhaled via mouthpiece by children and adults with mild CF, but normal airway
anatomy. Extrathoracic deposition of particles 50% greater in children and tended to be
higher for younger ages. No significant difference in lung or total respiratory tract deposition.
1 um Examined deposition of particles in subjects aged 18-80 yrs. Fractional deposition not found to
be age-related but more depended on airway resistance and breathing patterns.
1, 2.05, 2.8 um For same flow rate, children had higher nasal resistance then adults. Nasal deposition
increased with particle size, ventilation flow rate, and nasal resistance. Average nasal
deposition percentages lower in children than in adults; differences increased with exercise.
Average nasal deposition percentages best correlated with airflow rate.
1, 5, 10, 15 um Airway models of trachea and first few generations of bronchial airways of children and adult;
total deposition in child model greater than in adult.
0.0046-0.2 um Nasal casts of children's airways; deposition efficiency for particles decreased with
increasing age.
0. 1-10 um Total fractional lung deposition comparable between children and adults for all sizes.
TB-deposition fraction greater in children; A deposition fraction reduced in children.
1.95 um Mass based deposition of ROFA decreased with age from 7 mo to adulthood; mass
deposition per unit surface area greater in children.
0.25-5 um A fractional deposition highest in children for all particle sizes; TB fractional deposition
monotonically decreasing function of age for all sizes; total fractional lung deposition higher
in children than adults.
ET deposition in children higher; TB and A may be lower or higher depending on particle size;
enhanced deposition for particles < 5 um in children.
;is
Author
Bennett and
Zeman(1998)
Bennett et al.
(1997a)
Bennett et al.
(1996)
Becquemin et al.
(1991)
Oldham et al.
(1997)
Cheng et al.
(1995)
Phalen and
Oldham (2001)
Musante and
Martonen (2000a)
Musante and
Martonen (1999)
Xu and Yu (1986)
-------
1 The average nasal deposition percentages were lower in children than in adults at rest;
2 these differences were even greater during exercise. The average nasal deposition percentages in
3 children and in adults for these particle sizes were better correlated with inspiratory airflow rate
4 than with resistances or pressure drops at rest and during moderate exercise. The authors
5 conclude that while the airways of children are narrower, they are also shorter and the inhalation
6 flow rate is reduced. This would mean that the thoracic airways of children are protected to a
7 lesser degree than those of adults.
8 An age dependent theoretical model to predict regional particle deposition in children's
9 lungs that incorporates breathing parameters and morphology of the growing lung was developed
10 by Musante and Martonen (1999). The model was used to compare deposition of monodisperse
11 aerosols, ranging from 0.25 to 5 jim, in the lungs of children (ages 7, 22, 48, and 98 mo) at rest
12 to that in adults (ages 30 years) at rest. Compared to adults, fractional deposition was highest in
13 the 48- and 98-mo subjects for all particle sizes; TB fractional deposition was found to be a
14 monotonically decreasing function of age for all sizes; and total fractional lung deposition (i.e.,
15 TB+A) was generally higher in children than adults, with children of all ages showing similar
16 total deposition fractions.
17 The model was later used by Musante and Martonen (2000a) to evaluate the deposition of a
18 residual oil fly ash (ROFA) having an MMAD of 1.95 jim, a geometric standard deviation of
19 2.19, and a CMD of 0.53 (assuming a particle density of 0.34 g/cm2). Deposition was evaluated
20 under resting breathing conditions. The mass-based deposition fraction of the particles was
21 found to decrease with age from 7 mo to adulthood, and the mass deposition per unit surface
22 area in the lungs of children could be significantly greater than that in the adult.
23 Phalen and Oldham (2001) calculated the respiratory deposition of particles with sizes
24 ranging from 0.1 to 10 jim in diameter for 20 year-old adults and 2 year-old children. Total
25 fractional lung deposition was comparable between adults and children for all particle sizes
26 tested; however, TB deposition fraction was much greater in children than in adults (from 13 to
27 81%, depending on particle size). Particle deposition fraction in the A region was significantly
28 reduced in children.
29 Cheng et al. (1995) examined deposition of ultrafine particles in replica casts of the nasal
30 airways of children aged 1.5 to 4 years. Particle sizes ranged from 0.0046 to 0.2 jim, and both
June 2003 6-29 DRAFT-DO NOT QUOTE OR CITE
-------
1 inspiratory and expiratory flow rates were used (3 to 16 L/min). Deposition efficiency was
2 found to decrease with increasing age for a given particle size and flow rate.
3 Oldham et al. (1997) examined the deposition of monodisperse particles having diameters
4 of 1, 5, 10, and 15 jim in hollow airway models that were designed to represent the trachea and
5 the first few bronchial airway generations of an adult, a 7-year-old child, and a 4-year-old child.
6 They noted that, in most cases, the total deposition efficiency was greater in the child-size
7 models than in the adult model.
8
9 Inhaled Deposition Patterns
10 Bennett et al. (1997a) analyzed the regional deposition of poorly soluble 4.5-|im particles
11 inhaled via mouthpiece. The subjects were children and adults with mild cystic fibrosis (CF),
12 but who likely had normal upper airway anatomy such that intra- and extrathoracic deposition
13 would be similar to that in healthy people. The mean age of the children was 13.8 years and
14 29.1 years for the adults. Extrathoracic deposition of the 4.5-|im particles, as a percentage of
15 total respiratory tract deposition, was higher by about 50% in children compared to adults
16 (30.7%, 20.1%, and 16.0%, respectively). There was an age dependence of ET deposition for
17 the 4.5-|im particles in the children in that the percentage ET deposition tended to be higher at a
18 younger age (p = 0.08); the younger group (< 14 years; p = 0.05) had almost twice the
19 percentage ET deposition of the older group (> 14 years). Additional analyses showed an
20 inverse correlation of extrathoracic deposition with body height. There was no significant
21 difference in lung or total respiratory tract deposition between the children and adults. Because
22 ET deposition was age dependent and total deposition was not, this suggests that, in children, the
23 ET region does a more effective job of filtering out particles that would otherwise reach the TB
24 region. However, because the lungs of children are smaller than are those of adults, children
25 may still have deposition per unit surface area comparable to adults. These results are consistent
26 with the predicted increase in head deposition of particle greater than 2 jim with decreasing age
27 reported by Xu and Yu (1986).
28 Bennett and Zeman (1998) measured the deposition of monodisperse 2-|im (MMAD)
29 particles in children (aged 7 to 14 years) and adolescents (aged 14 to 18 years) for comparison to
30 that in adults (19 to 35 years). Each subject inhaled the particles by following their previously
31 determined individual spontaneous resting breathing pattern. Deposition was assessed by
June 2003 6-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 measuring the amount of particles inhaled and exhaled. There was no age-related difference in
2 deposition within the children group. There was also no significant difference in deposition
3 between the children and adolescents between the children and adults or between the adolescents
4 and adults. However, the investigators noted that, because the children had smaller lungs and
5 higher minute volumes relative to lung size, they likely would receive greater doses of particles
6 per lung surface area compared to adults. Furthermore, breath-to-breath fractional deposition in
7 children did vary with tidal volume, increasing with increasing volume. The rate of deposition
8 normalized to lung surface area tended (p = 0.07) to be greater (35%) in children when compared
9 to the combined group of adolescents and adults. These additional studies still do not provide
10 unequivocal evidence for significant differences in deposition between adults and children, even
11 when considering differences in lung surface area. However, it should be noted that differences
12 in levels of activity between adults and children are likely to play a fairly large role in age-
13 related differences in deposition patterns of ambient particles. Children generally have higher
14 activity levels during the day and higher associated minute ventilation per lung size, which can
15 contribute to a greater size-specific dose of particles. Activity levels in relationship to exposure
16 are discussed more fully in Chapter 5.
17 Another subpopulation of potential concern related to susceptibility to inhaled particles is
18 the elderly. In the study of Bennett et al. (1996) in which the total respiratory tract deposition of
19 2-|im particles was examined in people aged 18 to 80 years, the deposition fraction in the lungs
20 of people with normal lung function was found to be independent of age, depending solely on
21 breathing pattern and airway resistance.
22
23 6.2.3.3 Respiratory Tract Disease
24 The presence of respiratory tract disease can affect airway structure and ventilatory
25 parameters, thus altering deposition compared to that occurring in healthy individuals. The
26 effect of airway diseases on deposition has been studied extensively, as described in the 1996
27 PM AQCD (U.S. Environmental Protection Agency, 1996). Studies described therein showed
28 that people with chronic obstructive pulmonary disease (COPD) had very heterogeneous
29 deposition patterns and differences in regional deposition compared to normals. People with
30 asthma and obstructive pulmonary disease tended to have greater TB deposition than did healthy
31 people. Furthermore, there tended to be an inverse relationship between bronchoconstriction and
June 2003 6-31 DRAFT-DO NOT QUOTE OR CITE
-------
1 the extent of deposition in the A region; whereas total respiratory tract deposition generally
2 increased with increasing degrees of airway obstruction. The described studies were performed
3 during controlled breathing; i.e., all subjects breathed with the same tidal volume and respiratory
4 rate. However, although resting tidal volume is similar or elevated in people with COPD
5 compared to healthy individuals, the former tend to breathe at a faster rate, resulting in higher
6 than normal tidal peak flow and resting minute ventilation. Thus, some of the reported
7 differences in the deposition of particles could have been caused by increased fractional
8 deposition with each breath. Although the extent to which lung deposition may change with
9 respect to particle size, breathing pattern, and disease status in people with COPD is still unclear,
10 some recent studies have attempted to provide additional insight into this issue.
11 Bennett et al. (1997b) measured the fractional deposition of insoluble 2-|im particles in
12 people with severe to moderate COPD (mix of emphysema and chronic bronchitis, mean age
13 62 years) and compared this to healthy older adults (mean age 67 years) under conditions where
14 the subjects breathed using their individual resting breathing pattern as well as a controlled
15 breathing pattern. People with COPD tended to have an elevated tidal volume and a faster
16 breathing rate than people with healthy lungs, resulting in about 50% higher resting minute
17 ventilation. Total respiratory tract deposition was assessed in terms of deposition fraction
18 (determined from measures of the amount of aerosol inhaled and exhaled) and deposition rate
19 (the amount of particulate deposited per unit time). Under typical breathing conditions, people
20 with COPD had about 50% greater deposition fraction than did age-matched healthy adults.
21 Because of the elevation in minute ventilation, people with COPD had average deposition rates
22 about 2.5 times that of healthy adults. Similar to previously reviewed studies (U.S.
23 Environmental Protection Agency, 1996), these investigators observed an increase in deposition
24 with an increase in airway resistance, suggesting that, at rest, COPD resulted in increased
25 deposition of fine particles in proportion to the severity of airway disease. The investigators also
26 reported a decrease in deposition with increasing mean effective airspace diameter; this
27 suggested that the enhanced deposition was associated more with the chronic bronchitic
28 component of COPD than with the emphysematous component. Greater deposition was noted
29 with natural breathing compared to the fixed pattern.
30 Kim and Kang (1997) measured lung deposition of l-|im particles inhaled via the mouth
31 by healthy adults (mean age 27 years) and by those with various degrees of airway obstruction,
June 2003 6-32 DRAFT-DO NOT QUOTE OR CITE
-------
1 namely smokers (mean age 27 years), smokers with small airway disease (SAD; mean age
2 37 years), asthmatics (mean age 48 years), and patients with COPD (mean age 61 years)
3 breathing under the same controlled pattern. Deposition fraction was obtained by measuring the
4 number of particles inhaled and exhaled, breath by breath. There was a marked increase in
5 deposition in people with COPD. Deposition was 16%, 49%, 59%, and 103% greater in
6 smokers, smokers with SAD, asthmatics and people with COPD, respectively, than in healthy
7 adults. Deposition in COPD patients was significantly greater than that associated with either
8 SAD or asthma; there was no significant difference in deposition between people with SAD and
9 asthma. Deposition fraction was found to be correlated with percent predicted forced expiratory
10 volume (FEVj) and forced expiratory flow (FEF25.750/0). Airway resistance was not correlated
11 strongly with total lung deposition. Kohlhaufl et al. (1999) showed increased deposition of fine
12 particles (0.9 jim) in women with bronchial hyperresponsiveness.
13 Brown et al. (2001) examined the relationship between regional lung deposition for coarse
14 particles (5 jim) and ventilation patterns in healthy adults and in patients with CF. They found
15 that deposition in the TB region was positively associated with regional ventilation in healthy
16 subjects, but negatively associated in CF patients. The relationships were reversed for
17 deposition in the A region. These data suggest that significant coarse particle deposition may
18 occur in the TB region of poorly ventilated lungs, as occurs in CF; whereas TB deposition
19 follows ventilation in healthy subjects.
20 Segal et al. (2000a) developed a mathematical model for airflow and particle motion in the
21 lung that was used to evaluate how lung cancer affects deposition patterns in the lungs of
22 children. It was noted that the presence of airway tumors could affect deposition by increasing
23 probability of inertial deposition and diffusion. The former would occur on upstream surfaces of
24 tumors and the latter on downstream surfaces. It was concluded that particle deposition is
25 affected by the presence of airway disease, that effects may be systematic and could be
26 predicted, and that, therefore, they could be incorporated into dosimetry models. Segal et al.
27 (2002) used a computer model to calculate the deposition fractions of PM within the lungs of
28 COPD patients. The original model was for a healthy lung with a total volume of 4800 mL. The
29 chronic bronchitis component of COPD was modeled by reducing airway diameters based on
30 airway resistance measurements in vivo. The emphysema component was modeled by
31 increasing alveolar volumes by 10 - 30%. The calculated results were compared with
June 2003 6-33 DRAFT-DO NOT QUOTE OR CITE
-------
1 experimental data obtained from COPD patients for controlled breathing trials (tidal volume of
2 500 mL, respiratory time of 1 s) with a particle size of 1 |im. The model successfully depicts
3 PM deposition patterns and their dependence on the severity of disease and indicate that airway
4 obstructions are the main cause for increased deposition in the COPD lung.
5 Thus, the database related to particle deposition and lung disease suggests that total lung
6 deposition generally is increased with obstructed airways, regardless of deposition distribution
7 between the TB and A regions. Airflow distribution is very uneven in diseased lungs because of
8 the irregular pattern of obstruction, and there can be closure of small airways. In this situation, a
9 part of the lung is inaccessible, and particles can penetrate deeper into other, better ventilated
10 regions. Thus, deposition can be enhanced locally in regions of active ventilation, particularly in
11 the A region.
12
13 6.2.3.4 Anatomical Variability
14 As indicated above, variations in anatomical parameters between genders and between
15 healthy people and those with obstructive lung disease can affect deposition patterns. However,
16 previous analyses generally have overlooked the effect on deposition of normal interindividual
17 variability in airway structure in healthy individuals. This is an important consideration in
18 dosimetry modeling, which often is based on a single idealized structure. Studies that have
19 become available since the 1996 PM AQCD have attempted to assess the influence of such
20 variation in respiratory tract structure on deposition patterns.
21 The ET region is the first to contact inhaled particles and, therefore, deposition within this
22 region would reduce the amount of particles available for deposition in the lungs. Variations in
23 relative deposition within the ET region will, therefore, propagate through the rest of the
24 respiratory tract, creating differences in calculated doses from individual to individual.
25 A number of studies have examined the influence of variations in airway geometry on deposition
26 in the ET region.
27 Cheng et al. (1996) examined nasal airway deposition in healthy adults using particles
28 ranging in size from 0.004 to 0.15 jim and at two constant inspiratory flow rates, 167 and
29 33 mL/s. Deposition was evaluated in relation to measures of nasal geometry as determined by
30 magnetic resonance imaging and acoustic rhinometry. They noted that interindividual variability
31 in deposition was correlated with the wide variation of nasal dimensions, in that greater surface
June 2003 6-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 area, smaller cross-sectional area, and increasing complexity of airway shape were all associated
2 with enhanced deposition.
3 Using a regression analysis of data on nasal airway deposition derived from Cheng et al.
4 (1996), Guilmette et al. (1997) noted that the deposition efficiency within this region was highly
5 correlated with both nasal airway surface area and volume; this indicated that airway size and
6 shape factors were important in explaining intra-individual variability noted in experimental
7 studies of human nasal airway aerosol deposition. Thus, much of the variability in measured
8 deposition among people resulted from differences in the size and shape of specific airway
9 regions.
10 Bennett et al. (1998) studied the role of anatomic dead space (ADS) in particle deposition
11 and retention in bronchial airways, using an aerosol bolus technique. They found that the
12 fractional deposition was dependant on the subject's ADS and that a significant number of
13 particles was retained beyond 24 h. This finding of prolonged retention of insoluble particles in
14 the airways is consistent with the findings of Scheuch et al. (1995) and Stahlhofen et al. (1986a)
15 and with the predictions of asymmetric stochastic human lung models (Asgharian et al., 2001).
16 Bennett et al. (1999) also found a lung volume-dependent asymmetric distribution of particles
17 between the left and right lung; the leftright ratio was increased at increased percentage of total
18 lung capacity (e.g., at 70% TLC, L:R was 1.60).
19 From the analysis of detailed deposition patterns measured by a serial-bolus mouth-
20 delivery method, Kim and Hu (1998) and Kim and Jaques (2000) found a marked enhancement
21 in deposition in the very shallow region (lung penetration depth < 150 mL) of the lungs in
22 females. The enhanced local deposition for both ultrafine and coarse particles was attributed to a
23 smaller size of the upper airways, particularly of the laryngeal structure.
24 Kesavanathan and Swift (1998) also evaluated the influence of geometry in affecting
25 deposition in the nasal passages of normal adults from two ethnic groups. Mathematical
26 modeling of the results indicated that the shape of the nostril affected particle deposition in the
27 nasal passages, but that there still remained large inter-subject variations in deposition when this
28 was accounted for, and which was likely caused by geometric variability in the mid and posterior
29 regions of the nasal passages.
30 Hofmann et al. (2000) examined the role of heterogeneity of airway structure in the rat
31 acinar region in affecting deposition patterns within this area of the lungs. By the use of
June 2003 6-3 5 DRAFT-DO NOT QUOTE OR CITE
-------
1 different morphometric models, they showed a substantial variability in predicted particle
2 deposition and concluded that the heterogeneity of acinar airway structure is primarily
3 responsible for the heterogeneity of acinar particle deposition.
4
5 6.2.4 Interspecies Patterns of Deposition
6 The primary purpose of this document is to assess the health effects of particles in humans.
7 As such, human dosimetry studies have been stressed. Such studies avoid uncertainties
8 associated with extrapolation of dosimetry from laboratory animals to humans. Nevertheless,
9 animal models have been and are currently being used in evaluations of health effects from
10 particulate matter because there are ethical limits to the types of studies that can be performed on
11 human subjects. Because of this, there is a considerable need to understand dosimetry in animals
12 and to understand dosimetric differences between animals and humans. In this regard, there are
13 a number of newly published studies that were designed to assess particle dosimetry in
14 commonly used animals and to relate this to dosimetry in humans.
15 The various species used in inhalation toxicological studies that serve as the basis for
16 dose-response assessment may not receive identical doses in a comparable respiratory tract
17 region (i.e., ET, TB, or A) when exposed to the same aerosol at the same inhaled concentration.
18 Such interspecies differences are important because any toxic effect is often related to the
19 quantitative pattern of deposition within the respiratory tract as well as to the exposure
20 concentration; this pattern determines not only the initial respiratory tract tissue dose, but also
21 the specific pathways by which deposited material is cleared and redistributed (Schlesinger,
22 1985). Differences in patterns of deposition between humans and animals were summarized
23 previously in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996) and by others
24 (Schlesinger et al., 1997). Such differences in initial deposition must be considered when
25 relating biological responses obtained in laboratory animal studies to effects in humans.
26 It is difficult to systematically compare interspecies deposition patterns obtained from
27 various reported studies because of variations in experimental protocols, measurement
28 techniques, definitions of specific respiratory tract regions, and so on. For example, tests with
29 humans are generally conducted under protocols that standardize the breathing pattern; whereas
30 those using laboratory animals involve a wider variation in respiratory exposure conditions (e.g.,
31 spontaneous breathing versus ventilated breathing or varying degrees of sedation). Much of the
June 2003 6-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 variability in the reported data for individual species may be due to the lack of normalization for
2 specific respiratory parameters during exposure. In addition, the various studies have used
3 different exposure techniques, such as nasal mask, oral mask, oral tube, or tracheal intubation.
4 Regional deposition is affected by the exposure route and delivery technique employed.
5 Figure 6-9 shows the regional deposition data versus particle diameter in commonly used
6 laboratory animals obtained by various investigators as compiled by Schlesinger (1988; 1989).
7 The results are described in detail in the 1996 PM AQCD (U.S. Environmental Protection
8 Agency, 1996). In general, there is much variability in the data; however, it is possible to make
9 some generalizations concerning comparative deposition patterns. The relationship between
10 total respiratory tract deposition and particle size is approximately the same in humans and most
11 of these animals: deposition increases on both sides of a minimum that occurs for particles of
12 0.2 to 1 jim. Interspecies differences in regional deposition occur due to anatomical and
13 physiological factors. In most laboratory animal species, deposition in the ET region is near 100
14 percent for a particle diameter (dp) greater than 5 jim (Raabe et al., 1988), indicating greater
15 efficiency than that seen in humans. In the TB region, there is a relatively constant, but lower,
16 deposition fraction for dp greater than 1 jim in all species compared to humans. Finally, in the
17 A region, deposition fraction peaks at a lower particle size (dp about 1 |im) in laboratory animals
18 than in humans.
19 One of the issues that must be considered in interspecies comparisons of hazards from
20 inhaled particles is inhalability of the aerosol in the atmosphere of concern. Inhalability is the
21 fraction of suspended PM in ambient air that actually enters the nose or mouth with the volume
22 of air inhaled and is a function of particle aerodynamic size, inspiratory flow rate, wind speed,
23 and wind direction. Although inhalability may not be an issue for humans per se as far as
24 exposure to ambient particles is concerned, it can be an important issue when attempting to
25 extrapolate to humans the results of studies using animal species commonly employed in
26 inhalation toxicological studies (Miller et al., 1995). For example, differences between rat and
27 human become very pronounced for particles > 5 jim, and some differences are also evident for
28 particles as small as 1 |im (Figure 6-10). Menache et al. (1995) have developed equations that
29 can be used to determine the inhalability adjustments needed as a function of particle size to
30 compare laboratory animal and human studies.
June 2003 6-37 DRAFT-DO NOT QUOTE OR CITE
-------
100
80 -I
c 60 H
Upper Respiratory Tract
O Rat
D Hamster
A Mouse
^ Guinea Pig
V Dog
40 -
20 -
0.01
0.1
Tracheobronchial Region
uu -
-
^
b 40 -
c
g
03
O
Q_
o> 20 -
Q
n -
Al 0
D
A
Rat
Hamster
Mouse
O Guinea Pig
V
V
Dog
<
^#
c
1
.
A«^fe
\
;
ui
0.01
60 -
0.1 1.0
Pulmonary Region
ORat
D Hamster
A Mouse
<> Guinea Pig
V Dog
0.01
0.1 1.0
Particle Diameter (|jm)
10
Figure 6-9. Regional deposition fraction measured in laboratory animals as a function of
particle size for (a) upper respiratory tract, (b) tracheobronchial region, and
(c) pulmonary region. Particle diameters are aerodynamic (MMAD) for those
> 0.5 jim and geometric (or diffusion equivalent) for those < 0.5 urn.
Source: Schlesinger(1988).
June 2003
6-38
DRAFT-DO NOT QUOTE OR CITE
-------
B
100
Total Respiratory Tract
Human
- Oral Breathing
0
100
o
o.
o>
Q
Human
Nasal Breathing
100 |- Rat
0
0.01 0.1 1.0
Particle Diameter (|jm)
10
Extrathoracic Region
10°r Human
. Oral Breathing
0
100
Human
- Nasal Breathing
100 r
Rat
).01 0.1 1.0
Particle Diameter (|jm)
10
Tracheobronchial Region
100r
Oral Breathing
Human
o
Q.
CD
Q
100
0.01 0.1 1.0
Particle Diameter (|jm)
10
Alveolar Region
100r Human
. Oral Breathing
0
100
O
CL
CD
Q
0
Human
- Nasal Breathing
100 r
Rat
0.01 0.1 1.0
Particle Diameter (|jm)
10
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 um and
geometric (or diffusion equivalent) for those < 0.5 urn.
Source: Modified from Schlesinger (1989).
June 2003
6-39
DRAFT-DO NOT QUOTE OR CITE
-------
1 A number of studies have addressed various aspects of interspecies differences in
2 respiratory tract deposition using mathematical modeling approaches. Hofmann et al. (1996)
3 compared deposition between rat and human lungs using three-dimensional asymmetric
4 bifurcation models and mathematical procedures for obtaining air flow and particle trajectories.
5 Deposition in segmental bronchi and terminal bronchioles was evaluated under both inspiration
6 and expiration at particle sizes of 0.01, 1.0, and 10 jim, which covers the range of deposition
7 mechanisms from diffusion to impaction. Total deposition efficiencies of all particles in the
8 upper and lower airway bifurcations were comparable in magnitude for both rat and human.
9 However, the investigators noted that penetration probabilities from preceding airways must be
10 considered. When considering the higher penetration probability in the human lung, the
11 resulting bronchial deposition fractions were generally higher in human than in rat. For all
12 particle sizes, deposition at rat bronchial bifurcations was less enhanced on the carinas compared
13 to that found in human airways.
14 Hofmann et al. (1996) attempted to account for interspecies differences in branching
15 patterns in deposition analyses. Numerical simulations of three-dimensional particle deposition
16 patterns within selected (species-specific) bronchial bifurcations indicated that morphologic
17 asymmetry was a major determinant of the heterogeneity of local deposition patterns. They
18 noted that many interspecies deposition calculations used morphometry that was described by
19 deterministic lung models (i.e., the number of airways in each airway generation is constant, and
20 all airways in a given generation have identical lengths and diameters). Such models cannot
21 account for variability and branching asymmetry of airways in the lungs. Thus, their study
22 employed computations that used stochastic morphometric models of human and rat lungs
23 (Koblinger and Hofmann, 1985, 1988; Hofmann et al., 1989b) and evaluated regional and local
24 particle deposition. Stochastic models of lung structure describe, in mathematical terms, the
25 inherent asymmetry and variability of the airway system, including diameter, length, and angle.
26 They are based on statistical analyses of actual morphometric analyses of lungs. The model also
27 incorporated breathing patterns for humans and rats. In a later analysis (Hofmann and
28 Bergmann, 1998), the dependence of deposition on particle size was found to be qualitatively
29 similar in both rats and humans: deposition minima were found for total deposition as well as
30 deposition within the TB and A regions in the size range of 0.1 to 1 |im. In addition, a
31 deposition maximum occurred at about 0.02 to 0.03 jim and between 3 and 5 jim in both species.
June 2003 6-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 The deposition decrease in the A region at the smallest and largest sizes resulted from the
2 filtering efficiency of upstream airways. Although deposition patterns were qualitatively similar
3 in rat and human, deposition in the human lung appeared to be consistently higher than in the rat
4 in all regions of the lung (TB and A) over the entire size range. Both species showed a similar
5 pattern of dependence of deposition on flow rate.
6 The above model also assessed local deposition. In both human and rat, deposition of
7 0.001-|im particles was highest in the upper bronchial airways; whereas 0.1- and l-|im particles
8 showed higher deposition in more peripheral airways, namely the bronchiolar airways in rat and
9 the respiratory bronchioles in humans. Deposition was variable within any branching generation
10 because of differences in airway dimensions, and regional and total deposition also exhibited
11 intrasubject variations. Airway geometric differences between rats and humans were reflected in
12 deposition. Because of the greater branching asymmetry in rats prior to about generation 12,
13 each generation showed deposition maxima at two particle sizes, reflecting deposition in major
14 and minor daughters. These geometric differences became reduced with depth into the lung;
15 beyond generation 12, these two maxima were no longer seen.
16 Another comparison of deposition in lungs of humans and rats was performed by Musante
17 and Martonen (2000b). An interspecies mathematical dosimetry model was used to determine
18 the deposition of ROFA in the lungs under sedentary and light activity breathing patterns. This
19 latter condition was mimicked in the rat by increasing the CO2 level in the exposure system. The
20 MMAD of the particle size distribution was 1.95 |im with a geometric standard deviation of
21 2.19. They noted that physiologically comparable respiratory intensity levels did not necessarily
22 correspond to comparable dose distribution in the lungs. Because of this, the investigators
23 speculate that the resting rat may not be a good model for the resting human. The ratio of
24 aerosol mass deposited in the TB region to that in the A region for the human at rest was 0.961,
25 indicating fairly uniform deposition throughout the lungs. On the other hand, in the resting rat,
26 the ratio was 2.24, indicating greater deposition in the TB region than in the A region. However,
27 by mimicking light activity in the rat, the ratio was reduced to 0.97, similar to the human. These
28 data underscore the need for dose-response studies and for models that are capable of adjusting
29 for the dosimetric differences between species.
30 The relative distribution of particles deposited within the bronchial and alveolar regions of
31 the airways may differ in the lungs of animals and humans for the same total amount of
June 2003 6-41 DRAFT-DO NOT QUOTE OR CITE
-------
1 deposited matter because of structural differences. The effect of such structural differences
2 between rat and human airways on particle deposition patterns was examined by Hofmann et al.
3 (1999; 2000) in an attempt to find the most appropriate morphometric parameter to characterize
4 local particle deposition for extrapolation modeling purposes. Particle deposition patterns were
5 evaluated as functions of three morphometric parameters, namely (1) airway generation,
6 (2) airway diameter, and (3) cumulative path length. It was noted that airway diameter was a
7 more appropriate morphometric parameter for comparison of particle deposition patterns in
8 human and rat lungs than was airway generation.
9 The manner in which particle dose is expressed, that is, the specific dose metric, may affect
10 relative differences in deposition between humans and other animal species. For example,
11 although deposition when expressed on a mass per unit alveolar surface area basis may not be
12 different between rats and humans, dose metrics based on particle number per various
13 anatomical parameters (e.g., per alveolus or alveolar macrophage) can differ between rats and
14 humans, especially for particles around 0.1 to 0.3 jim (Miller et al., 1995). Furthermore, in
15 humans with lung disease (such as asthma or COPD), differences between rat and human can be
16 even more pronounced.
17 The probability of any biological effect occurring in humans or animals depends on
18 deposition and retention of particles, as well as the underlying tissue sensitivity. Interspecies
19 dosimetric extrapolation must consider these differences in evaluating dose-response
20 relationships. Thus, even similar deposition patterns may not result in similar effects in different
21 species because dose also is affected by clearance mechanisms. In addition, the total number of
22 particles deposited in the lung may not be the most relevant dose metric for interspecies
23 comparisons. For example, it may be the number of deposited particles per unit surface area or
24 dose to a specific cell (e.g., alveolar macrophage) that determines response for specific regions.
25 More specifically, even if fractional deposition is similar in the rat and human, there would be
26 differences in deposition density because of the higher metabolic rate in the rat. Thus, species-
27 specific differences in deposition density should be considered when health effects observed in
28 laboratory animals are being evaluated for potential effects occurring in humans.
29
30
31
June 2003 6-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.3 PARTICLE CLEARANCE AND TRANSLOCATION
2 This section discusses the clearance and translocation of particles that have deposited in the
3 respiratory tract. First, a basic overview of biological mechanisms and pathways of clearance in
4 the various region of the respiratory tract is presented. This is followed by an update on regional
5 kinetics of particle clearance. Interspecies patterns of clearance are then addressed, followed by
6 new information on biological factors that may modulate clearance.
7
8 6.3.1 Mechanisms and Pathways of Clearance
9 Particles that deposit on airway surfaces may be cleared from the respiratory tract
10 completely or may be translocated to other sites within this system by various regionally distinct
11 processes. These clearance mechanisms, which are outlined in Table 6-2, can be categorized as
12 either absorptive (i.e., dissolution) or nonabsorptive (i.e., transport of intact particles) and may
13 occur simultaneously or with temporal variations. It should be mentioned that particle solubility
14 in terms of clearance refers to solubility within the respiratory tract fluids and cells. Thus, a
15 poorly soluble particle is considered to be one whose rate of clearance by dissolution is
16 insignificant compared to its rate of clearance as an intact particle. All deposited particles,
17 therefore, are subject to clearance by the same basic mechanisms, with their ultimate fate a
18 function of deposition site, physicochemical properties (including solubility and any toxicity),
19 and sometimes deposited mass or number concentration. Clearance routes from the various
20 regions of the respiratory tract have been discussed previously in detail (U.S. Environmental
21 Protection Agency, 1996; Schlesinger et al., 1997). They are schematically shown in
22 Figure 6-11 (for extrathoracic and tracheobronchial regions) and in Figure 6-12 (for poorly
23 soluble particle clearance from the alveolar region) and are reviewed only briefly below.
24
25 6.3.1.1 Extrathoracic Region
26 The clearance of poorly soluble particles deposited in the posterior portions of the nasal
27 passages occurs via mucociliary transport, with the general flow of mucus being towards the
28 nasopharynx. Mucus flow in the most anterior portion of the nasal passages is forward, clearing
29 deposited particles to the vestibular region where removal occurs by sneezing, wiping, or
30 blowing. Soluble material deposited on the nasal epithelium is accessible to underlying cells via
31 diffusion through the mucus. Dissolved substances may be translocated subsequently into the
June 2003 6-43 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-2. 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
Dissolution and absorption into blood/lymph
Source: Schlesinger(1995).
( Nasal Passages
Blood
J
Dissolution X""
•« ( 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).
June 2003
6-44
DRAFT-DO NOT QUOTE OR CITE
-------
Deposited Particle
P
i
Phagocytosis by "N
Alveolar Macrophages )
\ ^
Movement within .
Alveolar Lumen
1
Bronchiolar/ Bronchial *
Lumen -^
1
Mucociliary Blanket
1
Gl Tract
fc-
i
assage
veolar E
J
r
Throug
Epitheliu
' 1
Interstitium
mphatic Chann
Lymph Nodes
zndocyt
Type 1 A
Epithelis
h
m ^~
^—
osis t
Iveol
3lCe
els "^
-
3V
ar
r^ t
Passage through
Pulmonary Capillary
Endothelium
— -f Phagocytosis by"\
" 1 Interstitial 1
\^ Macrophages y
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 bloodstream. The nasal passages have a rich vasculature, and uptake into the blood from this
2 region may occur rapidly.
3 Clearance of poorly soluble particles deposited in the oral passages is by coughing and
4 expectoration or by swallowing into the gastrointestinal tract. Soluble particles are likely to be
5 rapidly absorbed after deposition, but it depends on the rate of dissolution of the particle and the
6 molecular size of the solute.
7
8 6.3.1.2 Tracheobronchial Region
9 Poorly soluble particles deposited within the TB region are cleared by mucociliary
10 transport towards the oropharynx, followed by swallowing. Poorly soluble particles also may
11 traverse the epithelium by endocytotic processes, entering the peribronchial region, be engulfed
12 via phagocytosis by airway macrophages (which can then move cephalad on the mucociliary
13 blanket), or enter the airway lumen from the bronchial or bronchiolar mucosa. Soluble particles
June 2003
6-45
DRAFT-DO NOT QUOTE OR CITE
-------
1 may be absorbed through the epithelium into the blood. It has been shown that blood flow
2 affects translocation from the TB region in that decreased bronchial blood flow is associated
3 with increased airway retention of soluble particles (Wagner and Foster, 2001). There is,
4 however, evidence that even soluble particles may be cleared by mucociliary transport (Bennett
5 and Ilowite, 1989; Matsui et al., 1998; Wagner and Foster, 2001).
6
7 6.3.1.3 Alveolar Region
8 Clearance from the A region occurs via a number of mechanisms and pathways. Particle
9 removal by macrophages comprises the main nonabsorptive clearance process in this region.
10 These cells, which reside on the epithelium, phagocytize and transport deposited material that
11 they contact by random motion or via directed migration under the influence of chemotactic
12 factors.
13 Although alveolar macrophages normally comprise up to about 3-19% of the total alveolar
14 cells in healthy, nonsmoking humans and other mammals (Crapo et al., 1982) the actual cell
15 count may be altered by particle loading. The magnitude of any increase in cell number is
16 related to the number of deposited particles rather than to total deposition by weight. Thus,
17 equivalent masses of an identically deposited substance would not produce the same response if
18 particle sizes differed, and the deposition of smaller particles would tend to result in a greater
19 elevation in macrophage number than would deposition of larger particles.
20 Particle-laden macrophages may be cleared from the A region along a number of pathways.
21 As noted in Figure 6-11, this includes cephalad transport via the mucociliary system after the
22 cells reach the distal terminus of the mucus blanket; movement within the interstitium to a
23 lymphatic channel; or perhaps traversing of the alveolar-capillary endothelium directly entering
24 the bloodstream. Particles within the lymphatic system may be translocated to tracheobronchial
25 lymph nodes, which can become reservoirs of retained material. Particles subsequently reaching
26 the postnodal lymphatic circulation will enter the blood. Once in the systemic circulation, these
27 particles or transmigrated macrophages can travel to extrapulmonary organs. Deposited particles
28 that are not ingested by alveolar macrophages may enter the interstitium where they are subject
29 to phagocytosis by resident interstitial macrophages, and may travel to perivenous,
30 peribronchiolar or subpleural sites where they become trapped, increasing particle burden. The
31 migration and grouping of particles and macrophages within the lungs can lead to the
June 2003 6-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 redistribution of initially diffuse deposits into focal aggregates. Some particles or components
2 can bind to epithelial cell membranes, macromolecules, or to other cell components, delaying
3 clearance from the lungs.
4 Churg and Brauer (1997) examined lung autopsy tissue from 10 people who had never
5 smoked from Vancouver, Canada. They noted that the geometric mean particle diameter
6 (GMPD) in lung parenchymal tissue was 0.38 jim (og = 2.4). Ultrafine particles accounted for
7 less than 5% of the total retained particulate matter. Metal particles had a GMPD of 0.17 jim
8 and silicates 0.49 jim. Ninety-six percent of retained PM was less than 2.5 jim. A subsequent
9 study considered retention of ambient particles in the lungs. Brauer et al. (2001) showed that
10 small particles could undergo significant steady-state retention within the lungs. Using lungs
11 obtained at autopsy from long-term, nonsmoking residents of an area having high levels of
12 ambient PM (Mexico City, Mexico) and those from an area with relatively low PM levels
13 (Vancouver, Canada), the investigators measured the particle concentration per gram of lung
14 within the parenchyma. They found that living in the high PM region resulted in significantly
15 greater retention of both fine and ultrafine particles within the lungs: levels in the lungs from
16 Mexico City contained over 7.4 times the concentration of these particles as did the lungs from
17 residents of Vancouver. These results indicate a clear relationship between ambient exposure
18 concentration and retention in the A region.
19 Clearance by the absorptive mechanism involves dissolution in the alveolar surface fluid
20 followed by transport through the epithelium and into the interstitium, and then diffusion into the
21 lymph or blood. Solubility is influenced by the particle's surface to volume ratio and other
22 properties, such as hydrophilicity and lipophilicity (Mercer, 1967; Morrow, 1973; Patton, 1996).
23
24 6.3.2 Clearance Kinetics
25 The kinetics of clearance have been reviewed in U.S. Environmental Protection Agency
26 (1996) and in a number of monographs (e.g., Schlesinger et al., 1997) and are discussed only
27 briefly here. The actual time frame over which clearance occurs affects the cumulative dose
28 delivered to the respiratory tract, as well as the dose delivered to extrapulmonary organs.
29
30
June 2003 6-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.3.2.1 Extrathoracic Region
2 Mucus flow rates in the posterior nasal passages are highly nonuniform, but the median
3 rate in a healthy adult human is about 5 mm/min, resulting in a mean anterior to posterior
4 transport time of about 10 to 20 min for poorly soluble particles (Rutland and Cole, 1981;
5 Stanley et al., 1985). Particles deposited in the anterior portion of the nasal passages are cleared
6 more slowly by mucus transport and are usually more effectively removed by sneezing, wiping,
7 or nose blowing (Fry and Black, 1973; Morrow, 1977).
8
9 6.3.2.2 Tracheobronchial Region
10 Mucus transport in the tracheobronchial tree occurs at different rates in different local
11 regions; the velocity of movement is fastest in the trachea, and it becomes progressively slower
12 in more distal airways. In healthy nonsmoking humans, using noninvasive procedures and no
13 anesthesia, average tracheal mucus transport rates have been measured at 4.3 to 5.7 mm/min
14 (Yeates et al., 1975, 1981; Foster et al., 1980; Leikauf et al., 1981, 1984); whereas that in the
15 main bronchi has been measured at =2.4 mm/min (Foster et al., 1980). Estimates for human
16 medium bronchi range between 0.2 to 1.3 mm/min; whereas those in the most distal ciliated
17 airways range down to 0.001 mm/min (Morrow et al., 1967; Cuddihy and Yeh, 1988; Yeates and
18 Aspin, 1978).
19 The total duration of bronchial clearance or some other time parameter often is used as an
20 index of mucociliary kinetics. Although clearance from the TB region is generally rapid, there is
21 experimental evidence, discussed in U.S. Environmental Protection Agency (1996), that a
22 fraction of material deposited in the TB region is retained much longer than the 24 h commonly
23 used as the outer range of clearance time for particles within this region (Stahlhofen et al.,
24 1986a,b; Scheuch and Stahlhofen, 1988; Smaldone et al., 1988). A study by Asgharian et al.
25 (2001) showed that it is not necessary to invoke a slow- and fast-phase for TB clearance to have
26 particles retained longer than 24 h. Based upon asymmetric stochastic human-lung modeling-
27 data, inter-subject variability in path length and the number of generations to the alveoli, which
28 may result in some material reaching the alveoli even with shallow breathing, can explain the
29 experimental observations while still fitting a single compartment clearance model. Other
30 studies described below, however, do support the concept that TB regional clearance consists of
31 both a fast and a slow component.
June 2003 6-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 Falk et al. (1997) studied clearance in healthy adults using monodisperse
2 polytetraflouethylene (PTFE; Teflon) particles (6.2 jim) inhaled at two flow rates. Each subject
3 inhaled twice at two flow rates (0.45 and 0.045 L/s). Theoretical calculations indicated that the
4 particles inhaled at 0.45 L/s should deposit mainly in large bronchi and in the alveolar region;
5 whereas the particles inhaled at 0.045 L/s should deposit mainly in small ciliated airways.
6 Twenty-four hours after inhalation about half of the particles inhaled with both modes of
7 inhalation had cleared. For the inhalation rate of 0.45 L/s, 15% cleared with a half time of
8 3.4 days and 85% with a half time of 190 days. For the inhalation rate of 0.045 L/s, 20% cleared
9 with a half time of 2.0 days and 80% with a half time of 50 days. The results indicate that a
10 considerable fraction of particles deposited in small ciliated airways had not cleared within 24 h,
11 and that these particles cleared differently from particles deposited in the alveolar region. The
12 authors observed that the experimental data agreed well with the theoretical predictions. Camner
13 et al. (1997) also noted that clearance from the TB region was incomplete by 24 h postexposure
14 and suggested that this may be caused by incomplete clearance from bronchioles. Healthy adults
15 inhaled teflon particles (6, 8, and 10 jim) under low flow rates to maximize deposition in the
16 small ciliated airways. The investigators noted a decrease in 24-h retention with increasing
17 particle size, indicating a shift toward either a smaller retained fraction, deposition more
18 proximally in the respiratory tract, or both. They calculated that a large fraction, perhaps as high
19 as 75% of particles depositing in generations 12 through 16, was still retained at 24 h
20 postexposure.
21 In a study to examine retention kinetics in the tracheobronchial tree (Falk et al., 1999),
22 nonsmoking healthy adults inhaled radioactively tagged 6. l-|im particles at both a normal flow
23 rate and a slow flow rate designed to deposit particles preferentially in small ciliated airways.
24 Lung retention was measured from 24 h to 6 mo after exposure. Following normal flow rate
25 inhalation, 14% of the particles retained at 24 h cleared with a half time of 3.7 days and 86%
26 with a half time of 217 days. Following slow flow rate inhalation, 35% of the particles retained
27 at 24 h cleared with a half time of 3.6 days and 65% with a half time of 170 days. Estimates
28 using a number of mathematical models indicated higher deposition in the bronchiolar region
29 (generations 9 through 15) with the slow rate inhalation compared to the normal rate. The
30 experimental data and predictions of the deposition modeling indicated that 40% of the particles
31 deposited in the conducting airways during the slow inhalation were retained after 24 h. The
June 2003 6-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles that cleared with the shorter half time were mainly deposited in the bronchiolar region,
2 but only about 25% of the particles deposited in this region cleared in this phase. This study
3 provided additional confirmation for a phase of slow clearance from the bronchial tree.
4 The underlying sites and mechanisms of long-term TB retention in the smaller airways are
5 not known. Some proposals were presented in the earlier 1996 PM AQCD (U.S. Environmental
6 Protection Agency, 1996). This slow clearing tracheobronchial compartment likely is associated
7 with bronchioles < 1 mm in diameter (Lay et al., 1995; Kreyling et al., 1999; Falk et al., 1999).
8 Based on a study in which an adrenergic agonist was used to stimulate mucus flow so as to
9 examine the role of mucociliary transport in the bronchioles, it was found that clearance from the
10 smaller airways was not influenced by the drug, suggesting to the investigators that mucociliary
11 transport was not as an effective clearance mechanism from this region as it is in larger airways
12 (Svartengren et al., 1998, 1999). Although slower or less effective mucus transport may result in
13 longer retention times in small airways, other factors may account for long-term TB retention.
14 One possibility is that particles are displaced into the gel phase because of surface tension forces
15 of the liquid lining of the small airways (Gehr et al., 1990, 1991). The issue of particle retention
16 in the tracheobronchial tree certainly is not resolved.
17 Long-term TB retention patterns are not uniform. There is an enhancement at bifurcation
18 regions (Radford and Martell, 1977; Henshaw and Fews, 1984; Cohen et al., 1988), the likely
19 result of both greater deposition and less effective mucus clearance within these areas. Thus,
20 doses calculated based on uniform surface retention density may be misleading, especially if the
21 material is lexicologically slow acting.
22
23 6.3.2.3 Alveolar Region
24 Particles deposited in the A region generally are retained longer than are those deposited in
25 airways cleared by mucociliary transport. There are limited data on alveolar clearance rates in
26 humans. Within any species, reported clearance rates vary widely because, in part, of different
27 properties of the particles used in the various studies. Furthermore, some chronic experimental
28 studies have employed high concentrations of poorly soluble particles that may have interfered
29 with normal clearance mechanisms, resulting in clearance rates different from those that would
30 typically occur at lower exposure levels. Prolonged exposure to high particle concentrations is
June 2003 6-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 associated with what is termed particle "overload." This is discussed in greater detail in
2 Section 6.4.
3 There are numerous pathways of A-region clearance, and the utilization of these may
4 depend on the nature of the particles being cleared. Little is known concerning relative rates
5 along specific pathways. Thus, generalizations about clearance kinetics are difficult to make.
6 Nevertheless, A-region clearance is usually described as a multiphasic process, with each phase
7 representing removal by a different mechanism or pathway and often characterized by increased
8 retention half times following toxicant exposure.
9 The initial uptake of deposited particles by alveolar macrophages is very rapid and
10 generally occurs within 24 h of deposition (Lehnert and Morrow, 1985; Naumann and
11 Schlesinger, 1986; Lay et al., 1998). The time for clearance of particle-laden alveolar
12 macrophages via the mucociliary system depends on the site of uptake relative to the distal
13 terminus of the mucus blanket at the bronchiolar level. Furthermore, clearance pathways and
14 subsequent kinetics may depend to some extent on particle size. For example, some smaller
15 ultrafine particles (< 0.02 jim) may be less effectively phagocytosed than larger ones
16 (Oberdorster, 1993).
17 Uningested particles may penetrate into the interstitium within a few hours following
18 deposition. This transepithelial passage seems to increase as particle loading increases,
19 especially to that level above which macrophage numbers increase (Ferin, 1977; Ferin et al.,
20 1992; Adamson and Bowden, 1981). It also may be particle size dependent because insoluble
21 ultrafine particles (< 0.1 jim diameter) of low intrinsic toxicity show increased access to the
22 interstitum and greater lymphatic uptake than do larger particles of the same material
23 (Oberdorster et al., 1992; Ferin et al., 1992). However, ultrafine particles of different materials
24 may not enter the interstitium to the same extent. Similarly, a depression of phagocytic activity,
25 a reduction in macrophage ability to migrate to sites of deposition (Madl et al., 1998), or the
26 deposition of large numbers of ultrafine particles may increase the number of free particles in the
27 alveoli, perhaps enhancing removal by other routes. In any case, free particles may reach the
28 lymph nodes perhaps within a few days after deposition (Lehnert et al., 1988; Harmsen et al.,
29 1985) although this route is not definitive and may be species dependent.
30 Kreyling et al. (2002) studied the translocation of insoluble ultrafine 192Ir radiolabeled
31 particles (15 and 80 nm count median diameter) inhaled by healthy, young adult, male rats
June 2003 6-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 ventilated for 1 h via an endotracheal tube. At time points ranging from 6 h to 7 d, rats were
2 sacrificed, and a complete balance of 192Ir activity retained in the body and cleared by excretion
3 was determined. Thoracic deposition fractions of inhaled 15 and 80 nm particles were 0.49 and
4 0.28, respectively. One week after inhalation, particles were predominantly cleared from the
5 lungs into the gastrointestinal tract and eliminated in feces. Minute particle translocation of <1%
6 of the deposited particles into secondary organs such as liver, spleen, heart, and brain was
7 measured after systemic uptake from the lungs. The translocated fraction of the 80-nm particles
8 was about an order of magnitude less than that of 15-nm particles. In further investigations, the
9 biokinetics of ultrafme particles and soluble 192Ir was studied after administration by either
10 gavage or intratracheal instillation or intravenous injection. These studies confirmed the low
11 solubility of the 192Ir particles and proved that (1) particles were neither dissolved nor absorbed
12 from the gut, (2) systemically circulating particles were rapidly and quantitatively accumulated
13 and retained in the liver and spleen, and (3) soluble 192Ir instilled in the lungs was rapidly
14 excreted via urine with little retention in the lungs and other organs. This study indicates that
15 only a rather small fraction of ultrafme 192Ir particles are translocated from peripheral lungs to
16 systemic circulation and extrapulmonary organs.
17 The extent of lymphatic uptake of particles may depend on the effectiveness of other
18 clearance pathways in that lymphatic translocation likely increases when the phagocytic activity
19 of alveolar macrophages decreases. This may be a factor in lung overload. However, it seems
20 that the deposited mass or number of particles must exceed some threshold below which
21 increases in loading do not affect translocation rate to the lymph nodes (Ferin and Feldstein,
22 1978; LaBelle and Brieger, 1961). In addition, the rate of translocation to the lymphatic system
23 may be somewhat particle-size dependent. Although no human data are available, translocation
24 of latex particles to the lymph nodes of rats was greater for 0.5- to 2-|im particles than for 5- and
25 9-|im particles (Takahashi et al., 1992), and particles within the 3- to 15-|im size range were
26 found to be translocated at faster rates than were larger sizes (Snipes and Clem, 1981). On the
27 other hand, translocation to the lymph nodes was similar for both 0.4-|im barium sulfate or
28 0.02-|im gold colloid particles (Takahashi et al., 1987). It seems that particles < 2 jim clear to
29 the lymphatic system at a rate independent of size; and it is particles of this size, rather than
30 those > 5 jim, that would have significant deposition within the A region following inhalation.
31 In any case, the normal rate of translocation to the lymphatic system is quite slow; and
June 2003 6-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 elimination from the lymph nodes is even slower, with half times estimated in tens of years
2 (Roy, 1989).
3 Soluble particles depositing in the A region may be cleared rapidly via absorption through
4 the epithelial surface into the blood. Actual rates depend on the size of the particle (i.e., solute
5 size), with smaller molecular weight solutes clearing faster than larger ones. Absorption may be
6 considered as a two-stage process: in the first stage deposited particles are dissociated into
7 material that can be absorbed into the circulation (i.e., dissolution); the second stage is uptake of
8 this material. Each of these stages may be time dependent. The rate of dissolution depends on a
9 number of factors, including particle surface area and chemical structure. A portion of the
10 dissolved material may be absorbed more slowly because of binding to respiratory tract
11 components. Accordingly, there is a very wide range for absorption rates, depending on the
12 physicochemical properties of the material deposited.
13 As indicated in both the toxicology and epidemiology chapters of this document
14 (Chapters 7 and 8), there is concern about how ambient PM affects the cardiovascular system.
15 Thus, an important dosimetric issue involves the pathways by which inhaled and deposited
16 particles in the lungs could affect extrapulmonary systems. Pathways by which PM, constituents
17 of PM, or cytokines released by the respiratory tract in response to PM could affect systems
18 distal to the respiratory tract occur have been recently described. Nemmar et al. (2001) instilled
19 hamsters with radioactively-labeled colloidal albumin particles (diameter < 0.080 jim) as a
20 model for ambient ultrafme particles and measured the label appearing in systemic blood and
21 various extrapulmonary organs up to 1 h postexposure. They found label in blood within
22 5 minutes after instillation. In their subsequent studies in which healthy volunteers were
23 challenged with inhalation of ""Technitum-labeled ultrafme (< 100 nm) carbon particles
24 (Nemmar et al., 2002), the radioactivity was detected in blood as early as 1 min, reaching a
25 maximum between 10 and 20 min after inhalation of the aerosol. While label was also noted in
26 the other extrapulmonary organs examined (namely liver, heart, spleen, kidneys, and brain), the
27 liver had the highest levels and these increased with increasing time postexposure while the
28 second highest levels were noted in the heart or kidney, depending upon the instilled
29 concentration. This suggests that ultrafme particles can rapidly diffuse from the lungs into the
30 systemic circulation, thus providing a pathway by which ambient PM may rapidly affect
31 extrapulmonary organs.
June 2003 6-53 DRAFT-DO NOT QUOTE OR CITE
-------
1 In another study, Takenaka et al. (2001) exposed rats by inhalation to 0.015 jim particles of
2 elemental silver and found elevated levels of silver (Ag) in various extrapulmonary organs up to
3 7 days postexposure. They found that the amount of Ag in the lungs decreased rapidly with
4 time, and, by day 7, only about 4% of the initial lung burden remained. At day 0, Ag was
5 already found in the blood. By 1 day postexposure, Ag had been distributed to the liver, kidney,
6 heart, and brain. The Ag concentration was highest in the kidney, followed by the liver, and then
7 the heart. This study also indicates that inhaled ultrafme particles were rapidly cleared from the
8 lungs. A similar clearance pattern was found after intratracheal instillation of AgNO3 solution.
9 Therefore, the investigators postulated that the rapid clearance of elemental silver particles was
10 due to a fast dissolution of ultrafme silver particles into the lung fluid and subsequent diffusion
11 into the blood stream although a possibility of direct translocation of solid particles into the
12 blood stream was not excluded. The investigators also instilled an aqueous suspension of
13 elemental silver particles (100+ jim) into some animals; in this case, there was more retention in
14 the lungs, which was ascribed to phagocytic accumulation of agglomerated particles in alveolar
15 macrophages and slow dissolution of particles in cells. Thus, this study also suggested that
16 particle size and the tendency of particles to aggregate can affect the translocation pathway from
17 the lungs. Earlier studies (Huchon et al., 1987; Peterson et al., 1989; Morrison et al., 1998)
18 investigated lung clearance of labeled macromolecule solutes with widely varying molecular
19 weight and labeled albumin as well as albumin ultrafme aggregates. Clearance rates found from
20 these earlier studies were much slower than recent studies described above, suggesting that the
21 possibility of a fast clearing pathway of solid ultrafme particles may need further study.
22
23 6.3.3 Interspecies Patterns of Clearance
24 The inability to study the retention of certain materials in humans for direct risk assessment
25 requires use of laboratory animals. Because dosimetry depends on clearance rates and routes,
26 adequate toxicological assessment necessitates that clearance kinetics in such animals be related
27 to those in humans. The basic mechanisms and overall patterns of clearance from the respiratory
28 tract are similar in humans and most other mammals. However, regional clearance rates can
29 show substantial variation between species, even for similar particles deposited under
30 comparable exposure conditions, as extensively reviewed elsewhere (U.S. Environmental
31 Protection Agency, 1996; Schlesinger et al., 1997; Snipes et al., 1989).
June 2003 6-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 In general, there are species-dependent rate constants for various clearance pathways.
2 Differences in regional and total clearance rates between some species are a reflection of
3 differences in mechanical clearance processes. For example, the relative proportion of particles
4 cleared from the A region in the short- and longer-term phases differs between laboratory
5 rodents and larger mammals, with a greater percentage cleared in the faster phase in rodents.
6 A recent study (Oberdorster et al., 1997) showed interstrain differences in mice and rats in the
7 handling of particles by alveolar macrophages. Macrophages of B6C3F1 mice could not
8 phagocytize 10-jim particles, but those of C57 black/61 mice did. In addition, the
9 nonphagocytized 10-|im particles were efficiently eliminated from the alveolar region; whereas
10 previous work in rats found that these large particles were retained persistently after uptake by
11 macrophages (Snipes and Clem, 1981; Oberdorster et al., 1992). The ultimate implication of
12 interspecies differences in clearance that need to be considered in assessing particle dosimetry is
13 that the retention of deposited particles can differ between species and may result in differences
14 in response to similar PM exposure atmospheres.
15 Hsieh and Yu (1998) summarized the existing data on pulmonary clearance of inhaled,
16 poorly soluble particles in the rat, mouse, guinea pig, dog, monkey, and human. Clearance at
17 different initial lung burdens, ranging from 0.001 to 10 mg particles/g lung, was analyzed using
18 a two-phase exponential decay function. Two clearance phases in the alveolar region, namely
19 fast and slow, were associated with mechanical clearance along two pathways, the former with
20 the mucociliary system and the latter with the lymph nodes. Rats and mice were fast clearers in
21 comparison to the other species. Increasing the initial lung burden resulted in an increasing mass
22 fraction of particles cleared by the slower phase. As lung burden increased beyond 1 mg
23 particles/g lung, the fraction cleared by the slow phase increased to almost 100% for all species.
24 However, the rate for the fast phase was similar in all species and did not change with increasing
25 lung burden of particles; whereas the rate for the slow phase decreased with increasing lung
26 burden. At elevated burdens, the effect on clearance rate was greater in rats than in humans, an
27 observation consistent with previous findings (Snipes, 1989).
28
29
30
June 2003 6-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.3.4 Factors Modulating Clearance
2 A number of factors have previously been assessed in terms of modulation of normal
3 clearance patterns, including age, gender, workload, disease, and irritant inhalation. Such factors
4 have been discussed in detail previously (U.S. Environmental Protection Agency, 1996).
5
6 6.3.4.1 Age
7 Studies described in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996)
8 indicated that there appeared to be no clear evidence for any age-related differences in clearance
9 from the lung or total respiratory tract, either from child to adult, or young adult to elderly.
10 Studies of mucociliary function have shown either no changes or some slowing in mucous
11 clearance function with age after maturity, but at a rate that would be unlikely to significantly
12 affect overall clearance kinetics.
13
14 6.3.4.2 Gender
15 Previously reviewed studies (U.S. Environmental Protection Agency, 1996) indicated no
16 gender-related differences in nasal mucociliary clearance rates in children (Passali and Bianchini
17 Ciampoli, 1985) nor in tracheal transport rates in adults (Yeates et al., 1975).
18
19 6.3.4.3 Physical Activity
20 The effect of increased physical activity on mucociliary clearance is unresolved:
21 previously discussed studies (U.S. Environmental Protection Agency, 1996) indicate either no
22 effect or an increased clearance rate with exercise. There are no data concerning changes in
23 A region clearance with increased activity levels. Breathing with an increased tidal volume was
24 noted to increase the rate of particle clearance from the A region, and this was suggested to
25 result from distension-related evacuation of surfactant into proximal airways resulting in a
26 facilitated movement of particle-laden macrophages or uningested particles because of the
27 accelerated motion of the alveolar fluid film (John et al., 1994).
28
29 6.3.4.4 Respiratory Tract Disease
30 Various respiratory tract diseases are associated with clearance alterations. Evaluation of
31 clearance in individuals with lung disease requires careful interpretation of results because
June 2003 6-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 differences in deposition of particles used to assess clearance function may occur between
2 normal individuals and those with disease; this would directly affect the measured clearance
3 rates, especially in the tracheobronchial tree. Studies reported in the 1996 PM AQCD (U.S.
4 Environmental Protection Agency, 1996) noted findings of (a) slower nasal mucociliary
5 clearance in humans with chronic sinusitis, bronchiectasis, rhinitis, or cystic fibrosis and
6 (b) slowed bronchial mucus transport associated with bronchial carcinoma, chronic bronchitis,
7 asthma, and various acute respiratory infections. However, a recent study by Svartengren et al.
8 (1996a) concluded, based on deposition and clearance patterns, that particles cleared equally
9 effectively from the small ciliated airways of healthy humans and those with mild to moderate
10 asthma; but, this similarity was ascribed to effective therapy for the asthmatics.
11 In another study, Svartengren et al. (1996b) examined clearance from the TB region in
12 adults with chronic bronchitis who inhaled 6-|im Teflon particles. Based on calculations,
13 particle deposition was assumed to be in small ciliated airways at low flow and in larger airways
14 at higher flow. The results were compared to those obtained in healthy subjects from other
15 studies. At low flow, a larger fraction of particles was retained over 72 h in people with chronic
16 bronchitis compared to healthy subjects, indicating that clearance resulting from spontaneous
17 cough could not fully compensate for impaired mucociliary transport in small airways. For
18 larger airways, patients with chronic bronchitis cleared a larger fraction of the deposited particles
19 over 72 h than did healthy subjects, but this was reportedly because of differences in deposition
20 resulting from airway obstruction.
21 An important mechanism of clearance from the tracheobronchial region, under some
22 circumstances, is cough. Although cough can be a reaction to an inhaled stimulus, in most
23 individuals with respiratory infections and disease, spontaneous coughing also serves to clear the
24 upper bronchial airways by dislodging mucus from the airway surface. Recent studies confirm
25 that this mechanism likely plays a significant role in clearance for people with mucus
26 hypersecretion, at least for the upper bronchial tree, and for a wide range of deposited particle
27 sizes (0.5 to 5 jim; Toms et al., 1997; Groth et al., 1997). There appears to be a general trend
28 towards an association between the extent (i.e., number) of spontaneous coughs and the rate of
29 particle clearance; faster clearance is associated with a greater number of coughs (Groth et al.,
30 1997). Thus, recent evidence continues to support cough as an adjunct to mucociliary movement
31 in the removal of particles from the lungs of individuals with COPD. However, some recent
June 2003 6-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 evidence suggests that, like mucociliary function, cough-induced clearance may become
2 depressed with worsening airway disease. Noone et al. (1999) found that the efficacy of
3 clearance via cough in patients with primary ciliary dyskinesia (who rely on coughing for
4 clearance because of immotile cilia) correlated with lung function (FEVj), in that decreased
5 cough clearance was associated with decreased percentage of predicted FEVj.
6 Earlier studies (U.S. Environmental Protection Agency, 1996) indicated that rates of
7 A region particle clearance were reduced in humans with chronic obstructive lung disease and in
8 laboratory animals with viral infections; whereas the viability and functional activity of
9 macrophages were impaired in human asthmatics and in animals with viral-induced lung
10 infections. However, any modification of functional properties of macrophages appears to be
11 injury-specific in that they reflect the nature and anatomic pattern of disease.
12 One factor that may affect clearance of particles is the integrity of the epithelial surface
13 lining of the lungs. Damage or injury to the epithelium may result from disease or from the
14 inhalation of chemical irritants or cigarette smoke. Earlier studies performed with particle
15 instillation showed that alveolar epithelial damage in mice at the time of deposition resulted in
16 increased translocation of inert carbon to pulmonary interstitial macrophages (Adamson and
17 Hedgecock, 1995). A similar response was observed in a more recent assessment (Adamson and
18 Prieditis, 1998), whereby silica (< 0.3 jim) was instilled into a lung having alveolar epithelial
19 damage (as evidenced by increased permeability) and particles were noted to reach the
20 interstitium and lymph nodes.
21
22
23 6.4 PARTICLE OVERLOAD
24 Experimental studies using some laboratory rodents have employed high exposure
25 concentrations of relatively nontoxic, poorly soluble particles. These particle loads interfered
26 with normal clearance mechanisms and produced clearance rates different from those that would
27 occur at lower exposure levels. Prolonged exposure to high particle concentrations is associated
28 with a phenomenon that has been termed particle "overload," defined as the overwhelming of
29 macrophage-mediated clearance by the deposition of particles at a rate that exceeds the capacity
30 of that clearance pathway. It has been suggested that, in the rat, overload is more dependent
31 upon the volume rather than the mass of particles (Iran et al., 2000) and that volumetric
June 2003 6-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 overloading will begin when particle retention approaches 1 mg particles/g lung tissue (Morrow,
2 1988). The importance of surface area to inflammation and the tumorogenic response is detailed
3 in an analysis performed by Driscoll (1995). He observed a positive tumor response associated
4 with pulmonary inflammation and epithelial cell proliferation in the rat. Moreover, there was a
5 significant relationship between lung particle dose, expressed as particle surface area/lung, and
6 the lung tumor response. There was a positive correlation between the surface area
7 characteristics of various chemically distinct particulate materials and tumorogenic activity.
8 Overload is a nonspecific effect noted in experimental studies using many different kinds of
9 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
11 of 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
25 particle clearance mechanisms in people with compromised lung status may occur whereby
26 clearance is overwhelmed and results in morbidity or mortality from particle exposure.
27
28
29
June 2003 6-59 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 particulate 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
7 effort, cost, and amount of test material than does inhalation; and can deliver a known, exact
8 dose of a toxicant to the lungs. It is also an extremely useful technique for mechanistic studies.
9 Because particle disposition is a determinant of dose, it is important to compare deposition and
10 clearance of particles delivered by these two routes in order to evaluate the relevance of studies
11 using instillation. However, in most instillation studies, the effect of this route of administration
12 on particle deposition and clearance per se was not examined. Although these parameters were
13 evaluated in some studies, it has been very difficult to compare particle deposition/clearance
14 between different inhalation and instillation studies because of differences in experimental
15 procedures and in the manner by which particle deposition/clearance was quantitated. Thus,
16 while instillation studies are valuable in providing mechanistic insights, inhalation studies are
17 more appropriate for risk assessment. A recent paper provides a detailed evaluation of the role
18 of instillation in respiratory tract dosimetry and toxicology studies (Driscoll et al., 2000).
19 A short summary derived from this paper is provided below in this section.
20 The pattern of initial regional deposition is strongly influenced by the exposure technique
21 used. Furthermore, the patterns within specific respiratory tract regions also are influenced in
22 this regard. Depending on particle size, inhalation results in varying degrees of deposition
23 within the ET airways, a region that is completely bypassed by instillation. Thus, differences in
24 amount of particles deposited in the lower airways will occur between the two procedures,
25 especially for those particles in the coarse mode. This is important if inhaled particles in
26 ambient air affect the upper respiratory tract and such responses are then involved in the
27 evaluation of health outcomes.
28 Exposure technique also influences the intrapulmonary distribution of particles, which
29 potentially would affect routes and rates of ultimate clearance from the lungs and dose delivered
30 to specific sites within the respiratory tract or to extrapulmonary organs. Intratracheal
31 instillation tends to disperse particles fairly evenly within the TB region but can result in
32 heterogeneous distribution in the A region; whereas inhalation tends to produce a more
June 2003 6-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 homogeneous distribution throughout the major conducting airways as well as the A region for
2 the same particles. Thus, inhalation results in a randomized distribution of particles within the
3 lungs; whereas intratracheal instillation produces an heterogeneous distribution, in that the
4 periphery of the lung receives little particle load and most of the instilled particles are found in
5 regions that have a short path length from the major airways. Furthermore, inhalation results in
6 greater deposition in apical areas of the lungs and less in basal areas; whereas intratracheal
7 instillation results in less apical than basal deposition. Thus, toxicological effects from instilled
8 materials may not represent those which would occur following inhalation, due to differences in
9 sites of initial deposition following exposure. In addition, instillation studies generally deliver
10 high doses to the lungs, much higher than those which would occur with realistic inhalation
11 exposure. This would also clearly affect the initial dose delivered to target tissue and its
12 relevance to ambient exposure.
13 Comparison of the kinetics of clearance of particles administered by instillation or
14 inhalation have shown similarities, as well as differences, in rates for different clearance phases
15 depending on the exposure technique used (Oberdorster et al., 1997). However, some of the
16 differences in kinetics may be explained by differences in the initial sites of deposition. One of
17 the major pathways of clearance involves particle uptake and removal via pulmonary
18 macrophages. Domes and Valberg (1992) noted that inhalation resulted in a lower percentage of
19 particles recovered in lavaged cells and a more even distribution of particles among
20 macrophages. More individual cells received measurable amounts of particles via inhalation
21 than via intratracheal instillation; whereas with the latter, many cells received little or no
22 particles and others received very high burdens. Furthermore, with intratracheal instillation,
23 macrophages at the lung periphery contained few, if any, particles; whereas cells in the regions
24 of highest deposition were overloaded, reflecting the heterogeneity of particle distribution when
25 particles are administered via instillation. Additionally, both the relative number of particles
26 phagocytized by macrophages as well as the percentage of these cells involved in phagocytosis
27 is affected by the burden of administered particles, which is clearly different in instillation and
28 inhalation (Suarez et al., 2001). Thus, when guinea pigs were administered latex microspheres
29 (1.52-3.97 jim MMAD) by inhalation or instillation, the percentage of cells involved in
30 phagocytosis, as well as the amount of particles per cell, were both significantly higher with the
June 2003 6-61 DRAFT-DO NOT QUOTE OR CITE
-------
1 latter route. The route of exposure, therefore, influences particle distribution in the macrophage
2 population and could, by assumption, influence clearance pathways and clearance kinetics.
3 In summary, inhalation may result in deposition within the ET region, and the extent of
4 deposition depends on the size of the particles used. Of course, intratracheal instillation
5 bypasses this portion of the respiratory tract and delivers particles directly to the
6 tracheobronchial tree. Although some studies indicate that short (0 to 2 days) and long (100 to
7 300 days postexposure) phases of clearance of insoluble particles delivered either by inhalation
8 or intratracheal instillation are similar, other studies indicate that the percentage retention of
9 particles delivered by instillation is greater than that for inhalation at least up to 30 days
10 postexposure. Thus, there is some inconsistency in this regard.
11 Perhaps the most consistent conclusion regarding differences between inhalation and
12 intratracheal instillation is related to the intrapulmonary distribution of particles. Inhalation
13 generally results in a fairly homogeneous distribution of particles throughout the lungs. On the
14 other hand, instillation results in a heterogeneous distribution, especially within the alveolar
15 region, and focally high concentrations of particles. The bulk of instilled material penetrates
16 beyond the major tracheobronchial airways, but the lung periphery is often virtually devoid of
17 particles. This difference is reflected in particle burdens within macrophages, with those from
18 animals inhaling particles having more homogeneous burdens and those from animals with
19 instilled particles showing groups of cells with no particles and others with heavy burdens. This
20 difference impacts on clearance pathways, dose to cells and tissues, and systemic absorption.
21 Exposure method, thus, clearly influences dose distribution.
22
23 Dosimetric Considerations in Comparing Dosages for Inhalation, Instillation, and
24 Exposure of Cultured Cells
25 There are three common experimental approaches for studying the biological effects of
26 particulate material: inhalation, instillation, and in vitro. Inhalation studies are the more
27 realistic physiologically, and thus the most applicable to risk assessment. However, because
28 they are expensive, time consuming and require specialized equipment and personnel, they must
29 be supplemented by other techniques. In vitro studies using live cells are cost-effective, provide
30 for precise dose delivery, and permit investigators who do not have access to inhalation
31 techniques to perform mechanistic and comparative toxicity studies of particulate material.
June 2003 6-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 Commonly, the initial information on likely mechanisms of action of particles is obtained
2 through in vitro techniques.
3 Instillation studies, in which particles suspended in a carrier such as physiological saline
4 are applied to the airways, have certain advantages over in vitro studies. The exposed cells have
5 normal attachments to basement membranes and adjacent cells, circulatory support, surrounding
6 cells and normal endocrine, exocrine and neuronal relationships. Thus, instillation experiments
7 can bridge between in vitro and inhalation studies as well as produce useful mechanistic and
8 comparative toxicity information (Benson et al., 1986; Dorries and Valberg, 1992; Henderson et
9 al., 1995; Kodavanti et al., 2002; Leong et al., 1998; Oberdorster et al., 1997; Osier and
10 Oberdorster, 1997; Pritchard et al., 1985; Sabaitis et al., 1999; Suarez et al., 2001; Warheit et al.,
11 1991). Although the tracheobronchial region is most heavily dosed, alveolar regions can also be
12 exposed via instillation techniques (Kodavanti et al., 2002; Leong et al., 1998; Oberdorster et al.,
13 1997; Pritchard et al., 1985; Suarez et al., 2001; Warheit et al., 1991). As for in vitro studies,
14 dose selection is important because it is easy to overwhelm normal defense mechanisms.
15 Selection of the doses of particles used in instillation studies is far from an exact process.
16 If the goal is to expose tracheobronchial tree cell populations to particle concentrations (on a
17 number of particles per unit surface area basis) that are similar to those occurring in human
18 environmental exposures, or a known multiple of such exposures, dosimetric calculations must
19 be performed. Such calculations require selecting characteristics associated with the particles,
20 the exposed subject and the environmental exposure scenario. Hence each study can present a
21 unique dosimetric analysis. In most cases, it will be useful to know the relationship between the
22 surface doses in instillation studies and realistic local surface doses that could occur in vivo in
23 human subpopulations receiving the maximum potential dose. Although these subpopulations
24 have not been completely defined (NRC, 2001), some characteristics of individuals do serve to
25 enhance the local surface deposition doses to respiratory tract cells. These characteristics
26 include: exercise and mouth breathing (ICRP, 1994; NCRP, 1997); non-uniform inhaled air
27 distribution such as occurs in chronic obstructive pulmonary disease and chronic bronchitis
28 (Smaldone et al., 1993; Subramaniam et al., 2003; Sweeney et al., 1995; Segal et al., 2002;
29 Brown et al., 2002; Kim and Kang, 1997); impaired particle clearance as occurs in some disease
30 states (Pavia, 1987; Pavia et al., 1980; Smaldone, 1993) and location near pollutant sources
31 (Adgate et al., 2002; Zhu et al., 2002). In addition, even normal subjects exposed by inhalation
June 2003 6-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 are expected to have numerous sites of high local particle deposition (specifically at bifurcations)
2 within the tracheobronchial tree (Balashazy et al., 1999; Oldham et al., 2000; Kaye et al., 2000).
3 It is difficult to provide precise estimates of dose. However, by considering the several
4 factors discussed above that enhance local surface doses, order of magnitude estimates can be
5 made. As an example, consider the scenario of a physically active nose breather with chronic
6 lung disease that lives near a PM source. The increase in minute ventilation during exercise, due
7 to an increase in breaths per minute and in tidal volume, results in an increase in the number of
8 particles inhaled per unit time. Even light exertion can double the minute ventilation, and heavy
9 exertion can produce a six-fold increase (Phalen et al., 1985). Exercise also causes a shift from
10 nasal to oral breathing which bypasses the filtering efficiency of the nose (ICRP, 1994; NCRP,
11 1997). The switch from nasal to oral breathing will lead to increased exposure of the TB and
12 alveolar regions in a particle size dependent fashion. As particle aerodynamic diameter
13 increases from 1 to 10 jim, nasal region deposition at rest increases from 17% to 71% (NCRP,
14 1997) allowing more particles in this size range to reach the TB and alveolar regions. Thus, it is
15 reasonable to assume that oral breathing can lead to a doubling of TB and alveolar deposition of
16 thoracic coarse particles (PM10_25) in many individuals (see Figure 13, % deposition as a function
17 of particle size for the ICRP default worker). In disease states that produce uneven distribution
18 of inhaled air, available measurements and models indicate that an enhancement factor of 2 to 5
19 is realistic for surface doses (Bennett et al., 1997; Brown et al., 2002; Kim and Kang, 1997;
20 Miller et al., 1995; Segal et al., 2002).
21 The most important factor that produces high surface deposition doses of inhaled particles
22 in the TB region is the disturbed airflow produced by airway bifurcations. An enhanced
23 deposition of particles (for all sizes that have been examined) is seen at bifurcations in the TB
24 tree (Balashazy et al., 1999; Bell and Friedlander, 1973; Kaye et al., 2000; Oldham et al., 2000;
25 Schlesinger et al., 1982). The dose enhancement factor is dependent on both inhaled particle
26 diameter and the size of the deposition region under consideration. Using the computational
27 fluid dynamic modeling in a physiologically realistic (human TB tree) 3-dimensional group of
28 bifurcations, Balashazy et al. (1999) provided numerical enhancement (over average airway
29 surface deposition doses) factors. For the smallest region considered, which would comprise
30 less than a few hundred epithelial cells, the enhancement factors ranged from 52-fold for 0.01
31 |im diameter particles up to 113-fold for 10 jim diameter particles. An enhancement factor of
June 2003 6-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 81-fold was calculated for 1 jim diameter particles. Thus, for the purposes of simulating the
2 exposure of the heavily dosed TB bifurcation cells to PM10/PM2 5, an enhancement factor of 80-
3 fold is reasonable. Taken together, the combined dose enhancing effects of increased ventilation
4 (2-fold), oral breathing (2-fold), lung disease (2-fold) and bifurcation effects (80-fold), one could
5 expect populations of epithelial cells to experience enhanced deposition (over average surface
6 deposition) of about 640-fold. Considering that clearance impairment may also be a factor in
7 subpopulations with some disease states, the buildup of particles at such TB bifurcations further
8 increases the dose in relation to healthy individuals.
9 As a final consideration in this susceptibility scenario, the proximity of exposure to sources
10 of PM may be important. Although data are sparse in this regard, Zhu et al. (2002) have
11 measured time-averaged concentrations of black carbon and particle number at various distances
12 downwind from freeways in Los Angeles. In comparison to upwind concentrations,
13 concentrations at 30m downwind were about 4-fold higher for black carbon, and about 3-fold
14 higher for particle number. A factor of 3 for increased dose over the average might be expected
15 for this subpopulation. By taking all of the above factors into account, it is not unreasonable to
16 expect local PM doses to groups of cells in potentially susceptible subpopulations to be 3-4,000
17 times greater than the average TB surface exposures for the general population. Other scenarios
18 could be evaluated that lead to greater, or lesser, local dose estimates.
19 An estimate of the average surface deposition dose in the TB tree of a individual (with
20 COPD) exposed to PM2 5 for 24 hours at the current 24 hour NAAQS (65 |ig/m3) can be
21 calculated using the NCRP (1997) report and values for the surface area of the TB region in
22 adults. Assuming 5% of the inhaled particles deposit on a TB surface of 2,470 cm2 (Mercer et
23 al., 1994), and that no clearance occurs, the average surface deposition would be about 0.02
24 jig/cm2 of epithelium. Applying an enhancement factor of 3,000 to represent the most heavily
25 exposed epithelial cells yields a surface deposition of 57 jig/cm2. Assuming a rat has a TB
26 surface area of 27.2 cm2 (Mercer et al., 1994) and that the instillation of a PM suspension
27 exposes 10% of this area (Pritchard et al., 1985), an instillation of 150 jig could be reasonable. It
28 should be noted that even greater delivered doses to respiratory tract cells would be expected in
29 less well developed regions of the world with significantly higher levels of particulate air
30 pollutants.
June 2003 6-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 In conclusion, well-conducted instillation studies are valuable for examining the relative
2 toxicity of particulate materials and for providing mechanistic information that is useful for
3 interpreting in vitro and inhalation studies. However, because mechanisms of injury may vary
4 with the delivered dose, it would be useful if instillation studies designed to provide information
5 relevant to human risk assessment were accompanied by dosimetric calculations.
6
7
8 6.6 MODELING THE DISPOSITION OF PARTICLES IN THE
9 RESPIRATORY TRACT
10 6.6.1 Modeling Deposition, Clearance, and Retention
11 Over the years, mathematical models for predicting deposition, clearance and, ultimately,
12 retention of particles in the respiratory tract have been developed. Such models help interpret
13 experimental data and can be used to make dosimetry predictions for cases where data are not
14 available. In fact, model predictions described below are estimates based on the best available
15 models at the time of publication and, except where noted, have not been verified by
16 experimental data.
17 A review of various mathematical deposition models was given by Morrow and Yu (1993)
18 and in U.S. Environmental Protection Agency (1996). There are three major elements involved
19 in mathematical modeling. First, a structural model of the airways must be specified in
20 mathematical terms. Second, deposition efficiency in each airway must be derived for each of
21 the various deposition mechanisms. Finally, a computational procedure must be developed to
22 account for the transport and deposition of the particles in the airways. As noted earlier, most
23 models are deterministic in that particle deposition probabilities are calculated using anatomical
24 and airflow information on an airway generation by airway generation basis. Other models are
25 stochastic, whereby modeling is performed using individual particle trajectories and finite
26 element simulations of airflow.
27 Recent reports involve modeling the deposition of ultrafine particles and deposition at
28 airway bifurcations. Zhang and Martonen (1997) used a mathematical model to simulate
29 diffusion deposition of ultrafine particles in the human upper tracheobronchial tree and
30 compared the results to those in a hollow cast obtained by Cohen et al. (1990). The model
31 results were in good agreement with experimental data. Zhang and Martonen (1997) studied the
June 2003 6-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 inertial deposition of particles in symmetric three-dimensional models of airway bifurcations,
2 mathematically examining effects of geometry and flow. They developed equations for use in
3 predicting deposition based on Stokes numbers, Reynolds numbers (a dimensionless number that
4 describes the tendency for a flowing fluid to change from laminar flow to turbulent flow), and
5 bifurcation angles for specific inflows.
6 Models for deposition, clearance, and dosimetry of the respiratory tract of humans have
7 been available for the past four decades. For example, the International Commission on
8 Radiological Protection (ICRP) has recommended three different mathematical models during
9 this time period (International Commission on Radiological Protection, 1960, 1979, 1994).
10 These models make it possible to calculate the mass deposition and retention in different parts of
11 the respiratory tract and provide, if needed, mathematical descriptions of the translocation of
12 portions of the deposited material to other organs and tissues beyond the respiratory tract.
13 A somewhat simplified variation of the 1994 ICRP dosimetry model was used by Snipes et al.
14 (1997) to predict average particle deposition in the ET, T and A regions and retention patterns in
15 the A region under a repeated exposure situation for two characterized environmental aerosols
16 obtained from Philadelphia, PA and Phoenix, AZ. Both of these aerosols contained both fine
17 and coarse particles. They found similar retention for the fine particles in both aerosols, but
18 significantly different retention for the coarse-mode particles. Because the latter type dominated
19 the aerosol in the Phoenix sample, this type of evaluation can be used to improve the
20 understanding of the relationship between exposures to ambient PM and retention patterns that
21 affect health endpoints in residents of areas where the particle distributions and particle
22 chemistry may differ.
23 A morphological model based on laboratory data from planar gamma camera and single-
24 photon emission tomography images has been developed (Martonen et al., 2000). This model
25 defines the parenchymal wall in mathematical terms, divides the lung into distinct left and right
26 components, derives a set of branching angles from experimental measurements, and confines
27 the branching network within the left and right components (so there is no overlapping of
28 airways). The authors conclude that this more physiologically realistic model can be used to
29 calculate PM deposition patterns for risk assessment.
30 Musante and Martonen (2000c) developed an age-dependent theoretical model to predict
31 dosimetry in the lungs of children. The model includes the dimensions of individual airways and
June 2003 6-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 the geometry of branching airway networks within developing lungs and breathing parameters
2 as a function of age. The model suggests that particle size, age, and activity level markedly
3 affect deposition patterns of inhaled particles. Simulations thus far predict a lung deposition
4 fraction of 38% in an adult and 73% (nearly twice as high) in a 7-mo-old for 2-|im particles
5 inhaled during heavy breathing. The authors conclude that this model will be useful for
6 estimating dose delivered to sensitive subpopulations such as children.
7 Martonen et al. (2001a) developed a three-dimensional (3D) physiologically realistic
8 computer model of the human upper-respiratory tract (URT). The URT morphological model
9 consists of the extrathoracic region (nasal, oral, pharyngeal, and laryngeal passages) and upper
10 airways (trachea and main bronchi) of the lung. The computer representation evolved from a
11 silicone rubber impression of a medical school teaching model of the human head and throat.
12 The final unified 3D computer model may have significant applications in inhalation toxicology
13 for evaluating lung injuries from the inhalation of particulate matter.
14 Segal et al. (2000a) developed a computer model for airflow and particle motion in the
15 lungs of children to study how airway disease, specifically cancer, affects inhaled PM
16 deposition. The model considers how tumor characteristics (size and location) and ventilatory
17 parameters (breathing rates and tidal volumes) influence particle trajectories and deposition
18 patterns. The findings indicate that PM may be deposited on the upstream surfaces of tumors
19 because of enhanced efficiency of inertial impaction. Additionally, submicron particles and
20 larger particles, respectively, may be deposited on the downstream surfaces of tumors because of
21 enhanced efficiency of diffusion and sedimentation. The mechanisms of diffusion and
22 sedimentation are functions of the particle residence times in airways. Eddies downstream of
23 tumors would trap particles and allow more time for deposition to occur by diffusion and
24 sedimentation. The authors conclude that particle deposition is complicated by the presence of
25 airway disease and that the effects are systematic and predictable.
26 Segal et al. (2000b) have used a traditional mathematical model based on Weibel's lung
27 morphology and calculated total lung deposition fraction of 1- to 5-|im diameter particles in
28 healthy adults. Airway dimensions were scaled by individual lung volume. Deposition
29 predictions were made with both plug flow and parabolic flow profiles in the airways. The
30 individualized airway dimensions improved the accuracy of the predicted values when compared
31 with experimental data. There were significant differences, however, between the model
June 2003 6-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 predictions and experimental data depending on the flow profiles used, indicating that use of
2 more realistic parameters is essential to improving the accuracy of model predictions.
3 Broday and Georgopoulos (2001) presented a model that solves a variant of the general
4 dynamic equation for size evolution of respirable particles within human tracheobronchial
5 airways. The model considers polydisperse aerosols with respect to size but heterosperse with
6 respect to thermodynamic state and chemical composition. The aerosols have an initial bimodal
7 log-normal size distribution that evolves with time in response to condensation-evaporation and
8 deposition processes. Simulations reveal that submicron size particles grow rapidly and cause
9 increased number and mass fractions of the particle population to be found in the intermediate
10 size range. Because deposition by diffusion decreases with increasing size, hygroscopic fine
11 particles may persist longer in the inspired air than nonhygroscopic particles of comparable
12 initial size distribution. In contrast, the enhanced deposition probability of hygroscopic particles
13 initially from the intermediate size range increases their fraction deposited in the airways. The
14 model demonstrates that the combined effect of growth and deposition tends to decrease the
15 nonuniformity of the persistent aerosol, forming an aerosol which is characterized by size
16 distribution of smaller variance. These factors also alter the deposition profile along airways.
17 Lazaridis et al. (2001) developed a deposition model for humans that was designed to
18 better describe the dynamics of respirable particles within the airways. The model took into
19 account alterations in aerosol particle size and mass distribution that may result from processes
20 such as nucleation, condensation, coagulation, and gas-phase chemical reactions. The airway
21 geometry used was the regular dichotomous model of Weibel, and it incorporated the influences
22 of airway boundary layers on particle dynamics although simplified velocity profiles were used
23 so as to maintain a fairly uncomplicated description of respiratory physiology. Thus, this model
24 was considered to be an improvement over previous models which did not consider either the
25 effects of boundary layers on both the airborne and deposited particles or the effects of gas-phase
26 transport processes because it can account for the polydispersity, multimodality, and
27 heterogeneous composition of common ambient aerosols. The authors indicate that the model
28 predictions were both qualitatively and quantitatively consistent with experimental data for
29 particle deposition within the TB and A regions.
30 Another respiratory tract dosimetry model was developed concurrently with the ICRP
31 model by the National Council on Radiation Protection and Measurements (NCRP, 1997).
June 2003 6-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 As with the ICRP model, the NCRP model addresses inhalability of particles, revised subregions
2 of the respiratory tract, dissolution-absorption as an important aspect of the model, body size,
3 and age. The NCRP model defines the respiratory tract in terms of a
4 naso-oro-pharyngo-laryngeal (NOPL) region, which is equivalent to the ICRP (1994) model's
5 ET region, a tracheobronchial (TB) region, a pulmonary (P) region (equivalent to the ICRP
6 model A region), and lung-associated lymph nodes (LN). Deposition and clearance are
7 calculated separately for each of these regions. As with the 1994 ICRP model, inhalability of
8 aerosol particles is considered, and deposition in the various regions of the respiratory tract is
9 modeled using methods that relate to mechanisms of inertial impaction, sedimentation, and
10 diffusion.
11 Fractional deposition in the NOPL region was developed from empirical relationships
12 between particle diameter and air flow rate. Deposition in the TB and P regions were projected
13 from model calculations based on geometric or aerodynamic particle diameter and physical
14 deposition mechanisms such as impaction, sedimentation, diffusion, and interception.
15 Deposition in the TB and P regions used the lung model of Yeh and Schum (1980) with a
16 method of calculation similar to that of Findeisen (1935) and Landahl (1950). This method was
17 modified to accomodate an adjustment of lung volume and substitution of realistic deposition
18 equations. These calculations were based on air flow information and idealized morphometry
19 and used a typical pathway model. Comparison of regional deposition fraction predictions
20 between the NCRP and ICRP models was provided in U.S. Environmental Protection Agency
21 (1996). The definition of inhalability was that of the American Conference of Governmental
22 Industrial Hygenists (1985). Breathing frequency, tidal volume, and functional residual capacity
23 were the ventilatory factors used to model deposition. These were related to body weight and to
24 three levels of physical activity (low activity, light exertion, and heavy exertion).
25 Clearance from all regions of the respiratory tract was considered to result from
26 competitive mechanical and absorptive mechanisms. Mechanical clearance in the NOPL and TB
27 regions was considered to result from mucociliary transport. This was represented in the model
28 as a series of escalators moving towards the glottis and where each airway had an effective
29 clearance velocity. Clearance from the P region was represented by fractional daily clearance
30 rates to the TB region, the pulmonary LN region, and the blood. A fundamental assumption in
31 the model was that the rates for absorption into blood were the same in all regions of the
June 2003 6-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 respiratory tract. The rates of dissolution-absorption of particles and their constituents were
2 derived from clearance data primarily from laboratory animals. The effect of body growth on
3 particle deposition also was considered in the model, but particle clearance rates were assumed
4 to be independent of age. Some consideration for compromised individuals was incorporated
5 into the model by altering normal rates for the NOPL and TB regions.
6 Mathematical deposition models for a number of nonhuman species have been developed;
7 these were discussed in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996).
8 Despite difficulties, modeling studies in laboratory animals remain a useful step in extrapolating
9 exposure-dose-response relationships from laboratory animals to humans.
10 Respiratory tract clearance begins immediately upon deposition of inhaled particles. Given
11 sufficient time, the deposited particles may be removed completely by these clearance processes.
12 However, single inhalation exposures may be the exception rather than the rule. It generally is
13 accepted that repeated or chronic exposures are common for environmental aerosols. As a result
14 of such exposures, accumulation of particles may occur. Chronic exposures produce respiratory
15 tract burdens of inhaled particles that continue to increase with time until the rate of deposition is
16 balanced by the rate of clearance. This is defined as the "equilibrium respiratory tract burden."
17 It is important to evaluate these accumulation patterns, especially when assessing ambient
18 chronic exposures, because they dictate what the equilibrium respiratory tract burdens of inhaled
19 particles will be for a specified exposure atmosphere. Equivalent concentrations can be defined
20 as "species-dependent concentrations of airborne particles which, when chronically inhaled,
21 produce equal lung deposits of inhaled particles per gram of lung during a specified exposure
22 period" (Schlesinger et al., 1997). Available data and approaches with which to evaluate
23 exposure atmospheres that produce similar respiratory tract burdens in laboratory animals and
24 humans were discussed in detail in the 1996 PM AQCD.
25 Several laboratory animal models have been developed to help interpret results from
26 specific studies that involved chronic inhalation exposures to nonradioactive particles (Wolff
27 et al., 1987; Strom et al., 1988; Stober et al., 1994). These models were adapted to data from
28 studies involving high level chronic inhalation exposures in which massive lung burdens of low
29 toxicity, poorly soluble particles were accumulated. Koch and Stober (2001) further adapted
30 clearance models for more relevant particle deposition in the pulmonary region. They published
31 a pulmonary retention model that accounts for dissolution and macrophage-mediated removal of
June 2003 6-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 deposited polydisperse aerosol particles. The model provides a mathematical solution for the
2 size distribution of particles in the surfactant layer of the alveolar surface and in the cell plasma
3 of alveolar macrophages and accounts for the different kinetics and biological effects in the two
4 compartments. It does not, however, account for particle penetration to the lung interstitium and
5 particle clearance by the lymph system.
6 Estimating regional particle deposition patterns is important for establishing the
7 comparability of animal models, for understanding interspecies differences in the expression of
8 chemical toxicities, and, ultimately, for the human risk assessment process. Different species
9 exposed to the same particle atmosphere may not receive identical initial doses in comparable
10 respiratory tract regions, and the selection of a certain species for toxicological evaluation of
11 inhaled particles may, thus, influence the estimated human lung or systemic dose, as well as its
12 relationship to potential adverse health effects. Asgharian et al. (1995) described a strategy for
13 summarizing published data on regional deposition of particles of different diameters and
14 calculating a deposited fraction for a specific particle size distribution. The authors constructed
15 nomograms for three species, namely the human, monkey, and rat, to allow estimation of
16 alveolar deposition fractions. They then developed a regression model to permit the calculation
17 of more exact deposition fractions. While this paper describes the procedure for one region of
18 the lungs, the authors maintain that the technique can be applied to other regions of the
19 respiratory tract or to the total system for which deposition data are available. The model is
20 somewhat constrained at present due to the limitations of the underlying deposition database.
21 Tran et al. (1999) used a mathematical model of clearance and retention in the A region of
22 rats lungs to determine the extent to which a sequence of clearance mechanisms and pathways
23 could explain experimental data obtained from inhalation studies using relatively insoluble
24 particles. These pathways were phagocytosis by macrophages with subsequent clearance,
25 transfer of particles into the interstitium and to lymph nodes, and overloading of defense
26 mechanisms. The model contained a description of the complete defense system in this region
27 using both clearance and transfer processes as represented by sets of equations. The authors
28 suggested that the model could be used to examine the consistency of various hypotheses
29 concerning the fate of inhaled particles and could be used for species other than the rat with
30 appropriate scaling.
June 2003 6-72 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
3 fixed airway geometry; the stochastic lung (SL) model with a randomly selected branching
4 structure; and a hybrid of the MPL and SL models. They calculated total and regional deposition
5 for a range of particle sizes during quiet and heavy breathing. Although the total bronchial and
6 acinar deposition fractions were similar for the three models, the SL and the hybrid models
7 predicted a substantial variation in particle deposition among different acini. Acinar deposition
8 variances in the MPL model were consistently smaller than in the SL and the hybrid lung
9 models. The authors conclude that the similarity of acinar deposition variations in the latter two
10 models and their independence of the breathing pattern suggest that the heterogeneity of the
11 acinar airway structure is primarily responsible for the heterogeneity of acinar particle
12 deposition.
13 The combination of MPL and SL models developed for the human lung takes into
14 consideration both intra- and inter-human variability in airway structure. The models also have
15 been developed to approximately the same level of complexity for laboratory animals and,
16 therefore, can be readily used for interspecies extrapolation (Asgharian et al., 1999). A variation
17 of these models will soon be developed for inclusion of the airway geometry of children. By
18 incorporating particle clearance in the TB region (Asgharian et al., 2001) and in the alveolar
19 region (Koch and Stober, 2001), this suite of models should prove to be very useful in better
20 predicting PM dosimetry in humans.
21
22 6.6.2 Models To Estimate Retained Dose
23 Models have been used routinely to express retained dose in terms of temporal patterns for
24 A region retention of acutely inhaled materials. Available information for a variety of
25 mammalian species, including humans, can be used to predict deposition patterns in the
26 respiratory tract for inhalable aerosols with reasonable degrees of accuracy. Additionally,
27 alveolar clearance data for non-human mammalian species commonly used in inhalation studies
28 are available from numerous experiments that involved inhaled radioactive particles.
29 An important factor in using models to predict retention patterns in laboratory animals or
30 humans is the dissolution-absorption rate of the inhaled material. Factors that affect the
31 dissolution of materials or the leaching of their constituents in physiological fluids and the
June 2003 6-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 subsequent absorption of these constituents are not fully understood. Solubility is known to be
2 influenced by the surface-to-volume ratio and other surface properties of particles (Mercer,
3 1967; Morrow, 1973). The rates at which dissolution and absorption processes occur are
4 influenced by factors that include the chemical composition of the material. Temperature history
5 of materials is also an important consideration for some metal oxides. For example, in
6 controlled laboratory environments, the solubility of oxides usually decreases when the oxides
7 are produced at high temperatures, which generally results in compact particles having small
8 surface-to-volume ratios. It is sometimes possible to accurately predict dissolution-absorption
9 characteristics of materials based on physical/chemical considerations, but predictions for in
10 vivo dissolution-absorption rates for most materials, especially if they contain multivalent
11 cations or anions, should be confirmed experimentally.
12 Phagocytic cells, primarily macrophages, clearly play a role in dissolution-absorption of
13 particles retained in the respiratory tract (Kreyling, 1992). Some particles dissolve within the
14 phagosomes because of the acidic milieu in those organelles (Lundborg et al., 1984, 1985), but
15 the dissolved material may remain associated with the phagosomes or other organelles in the
16 macrophage rather than diffuse out of the macrophage to be absorbed and transported elsewhere
17 (Cuddihy, 1984). This same phenomenon has been reported for organic materials. For example,
18 covalent binding of benzo[a]pyrene or metabolites to cellular macromolecules resulted in an
19 increased alveolar retention time for that compound after inhalation exposures of rats (Medinsky
20 and Kampcik, 1985). Understanding these phenomena and recognizing species similarities and
21 differences are important for evaluating alveolar retention and clearance processes and for
22 interpreting the results of inhalation studies.
23 Dissolution-absorption of materials in the respiratory tract is clearly dependent on the
24 chemical and physical attributes of the material. Although it is possible to predict rates of
25 dissolution-absorption, it is prudent to determine this important clearance parameter
26 experimentally. It is important to understand the effect of this clearance process for the lungs,
27 tracheobronchial lymph nodes, and other body organs that might receive particles or their
28 constituents that enter the circulatory system from the lung.
29 Additional research must be done to provide the information needed to evaluate properly
30 retention of particles in conducting airways. However, a number of earlier studies, discussed in
31 the 1996 document and in Section 6.2.2.2 herein, noted that some particles were retained for
June 2003 6-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 relatively long times in the tracheobronchial regions, effectively contradicting the general
2 conclusion that almost all inhaled particles that deposit in the TB region clear within hours or
3 days. These studies have demonstrated that variable portions of the particles that deposit in, or
4 are cleared through, the TB region are retained with half times on the order of weeks or months.
5 Long-term retention and clearance patterns for particles that deposit in the ET and TB regions
6 must continue to be thoroughly evaluated because of the implications of this information for
7 respiratory tract dosimetry and risk assessment.
8 Model projections are possible for the A region using the cumulative information in the
9 scientific literature relevant to deposition, retention, and clearance of inhaled particles.
10 Clearance parameters for six laboratory animal species were summarized in U.S. Environmental
11 Protection Agency (1996). Nikula et al. (1997) evaluated results in rats and monkeys exposed to
12 high levels of either diesel soot or coal dust. Although the total amount of retained material was
13 similar in both species, the rats retained a greater portion in the lumens of the alveolar ducts and
14 alveoli than did monkeys; whereas the monkeys retained a greater portion of the material in the
15 interstitium. The investigators concluded that intrapulmonary retention patterns in one species
16 may not be predictive of those in another species at high levels of exposure, but this may not be
17 the case at lower levels of exposure.
18 The influence of exposure concentration on the pattern of particle retention in rats (exposed
19 to diesel soot) and humans (exposed to coal dust) was examined by Nikula et al. (2000) using
20 histological lung sections obtained from both species. The exposure concentrations for diesel
21 soot were 0.35, 3.5, or 7.0 |ig/m3; and exposure duration was 7 h/day, 5 days/week for 24 mo.
22 The human lung sections were obtained from nonsmoking nonminers, nonsmoking coal miners
23 exposed to levels < 2 mg dust/m3 for 3 to 20 years, or nonsmoking miners exposed to 2 to
24 10 mg/m3 for 33 to 50 years. In both species, the amount of retained material (using
25 morphometric techniques based on the volume density of deposition) increased with increasing
26 dose (which is related to exposure duration and concentration). In rats, the diesel exhaust
27 particles were found to be primarily in the lumens of the alveolar duct and alveoli; whereas in
28 humans, retained dust was found primarily in the interstitial tissue within the respiratory acini.
29 Dosimetric models may be used to adjust for differences in the exposure-dose relationship
30 in different species, thus allowing for comparison of lung responses at different doses. In a
31 series of papers Kuempel (Kuempel 2000, 200la; Kuempel 200Ib) presents a biologically based
June 2003 6-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 human dosimetric lung model to describe the fate of respirable particles in the lungs of humans.
2 The model uses data from coal miners and assumptions about the overloading of alveolar
3 clearance from studies in rats. The form of the model that provides the best fit to the lung dust
4 burden data in the coal miners includes a first-order interstitialization process and either a no
5 dose-dependent decline in alveolar clearance or a much lower decline than expected from the
6 rodent studies. These findings were consistent with particle retention patterns observed
7 previously in the lungs of primates.
8
9 6.6.3 Fluid Dynamics Models for Deposition Calculations
10 The available models developed to simulate particulate matter deposition in the lung are
11 based on simplifying assumptions about the morphometry of the lung and the fluid dynamics of
12 inspired air through a branching airway system. As new models are developed, they will better
13 predict particle deposition patterns in a more realistic airway geometry under realistic flow
14 conditions that can result in local inhomogeneities of particle deposition and the formation of hot
15 spots. One example is the model of ventilation distribution in the human lung developed by
16 Chang and Yu (1999). This model was designed as an improvement over those that assumed
17 uniform ventilation in the lungs because it better simulated the effect of airway dynamics on the
18 distribution of ventilation under different conditions which may occur in the various lobes of the
19 lungs and under various inspiratory flow rates. The authors indicated that the results of the
20 model compared favorably with experimental data and that the model will be incorporated into a
21 particle deposition model which will allow for the evaluation of the nonuniformity of deposition
22 within the lungs resulting from the physiological situation of nonuniform distribution of
23 ventilation. Computational fluid dynamics (CFD) modeling adds another step to better model
24 development by providing increased ability to predict local airflow and particle deposition
25 patterns and provide a better representation of extrathoracic deposition in the human respiratory
26 tract. The CFD models developed to date, however, also are limited in scope because they are
27 unable to simulate flow in the more complex gas exchange regions. Due to a lack of more
28 realistic simulations for the lower airways, they impose another "idealized" boundary condition
29 at the distal end of the human respiratory tract.
30 Airflow patterns within the lung are determined by the interplay of structural and
31 ventilatory conditions. These flow patterns govern the deposition kinetics of entrained particles
June 2003 6-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 in the inspired air. A number of CFD software programs are available to simulate airflow
2 patterns in the lung by numerically solving the Navier-Stokes equations (White, 1974). The
3 CFD modeling requires a computer reconstruction of the appropriate lung region and the
4 application of boundary conditions. The flow field resulting from the CFD modeling is
5 represented by velocity vectors in the grid points of a two- or three-dimensional mesh.
6 Numerical models of particle deposition patterns are computed by simulating the trajectories of
7 particles introduced into these flow streams after solving for the particles' equation of motion.
8 Such CFD models have been developed for different regions of the respiratory tract, including
9 the nasal cavity (Yu et al., 1998; Sarangapani and Wexler, 2000); larynx (Martonen et al. 1993;
10 Katz et al., 1997; Katz, 2001); major airway bifurcations (Gradon and Orlicki, 1990; Balashazy
11 and Hofmann, 1993a,b, 1995, 2001; Heistracher and Hofmann, 1995; Lee et al., 1996; Zhang
12 et al., 1997, 2000, 2001, 2002; Comer et al., 2000, 2001a,b); and alveoli (Tsuda et al., 1994a,b;
13 Darquenne, 2001).
14 Kimbell (2001) has recently reviewed the literature on CFD models of the upper
15 respiratory tract (URT). Most of these models have focused on characterizing the airflow
16 patterns in the URT and have not included simulation of particulate dosimetry. Keyhani et al.
17 (1995) were the first to use computer-aided tomography (CAT) scans of the human nasal cavity
18 to construct an anatomically accurate three-dimensional airflow model of the human nose.
19 Subramaniam et al. (1998) used MRI scan data to extend these CFD studies to include the
20 nasopharynx. However, neither of these studies investigated particle deposition in the upper
21 respiratory tract.
22 Yu et al. (1998) have developed a three-dimensional CFD model of the entire human upper
23 respiratory tract, including the nasal airway, oral airway, laryngeal airway, and the first two
24 generations of the tracheobronchial airway. They have used this CFD model to investigate the
25 effect of breathing pattern, i.e., nasal breathing, oral breathing, and simultaneous nasal and oral
26 breathing, on airflow and ultrafine particle deposition. They concluded that the ultrafine particle
27 deposition simulated using the CFD model was in reasonable agreement with the corresponding
28 experimental measurements. In a study led by Sarangapani and Wexler (2000), an upper
29 respiratory tract CFD model that included the nasal cavity, nasopharynx, pharynx, and larynx
30 was developed to study the deposition efficiency of hygroscopic and non-hygroscopic particles
31 in this region. They used the CFD model to simulate the temperature and water vapor conditions
June 2003 6-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 in the upper airways and predicted high relative humidity conditions in this region. They also
2 simulated particle trajectories for 0.5 |im, 1 |im, and 5 jim particles under physiologically
3 realistic flow rates. The predictions of the CFD model indicated that high relative humidity
4 conditions contribute to rapid growth of hygroscopic particles and would dramatically alter the
5 deposition characteristics of ambient hygroscopic aerosols.
6 Stapleton et al. (2000) investigated deposition of a poly disperse aerosol (MMD = 4.8 jim
7 and GSD = 1.65) in a replica of a human mouth and throat using both experimental results and
8 3-D CFD simulation. They found that CFD results were comparable with experimental results
9 for a laminar flow case, but were more than 200% greater for a turbulent flow case. The results
10 suggest that accurate predictions of particle deposition in a complex airway geometry requires a
11 careful evaluation of geometric and fluid dynamic factors in developing CFD models.
12 Due to the complex structural features and physiological conditions of the human laryngeal
13 region, only a limited number of modeling studies have been conducted to evaluate laryngeal
14 fluid dynamics and particle deposition. A high degree of inter-subject variability, a compliant
15 wall that presents challenges in setting appropriate boundary conditions, and a complex turbulent
16 flow field are some of the difficulties encountered in developing CFD models of the laryngeal
17 airways. Martonen et al. (1993) investigated laryngeal airflow using a two-dimensional CFD
18 model and concluded that laryngeal morphology exerts a pronounced influence on regional flow,
19 as well as fluid motion in the trachea and the main bronchi. In this study, the glottal aperture
20 (defined by the geometry of the vocal folds) was allowed to change in a prescribed manner with
21 the volume of inspiratory flow (Martonen and Lowe, 1983), and three flow rates corresponding
22 to different human activity were examined.
23 In a subsequent CFD analysis, a three-dimensional model of the larynx based on
24 measurements of human replica laryngeal casts (Martonen and Lowe, 1983; Katz and Martonen,
25 1996; Katz et al., 1997) simulated the flow field in the larynx and trachea under steady
26 inspiratory flow conditions at three flow rates. They observed that the complex geometry
27 produces jets, recirculation zones, and circumferential flow that may directly influence particle
28 deposition at select sites within the larynx and tracheobronchial airways. The primary
29 characteristics of the simulated flow field were a central jet penetrating into the trachea created
30 by the ventricular and vocal folds, a recirculating zone downstream of the vocal folds, and a
31 circumferential secondary flow. Recently, a computational model for fluid dynamics and
June 2003 6-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 particle motion for inspiratory flow through the human larynx and trachea has been described
2 (Katz, 2001). This model calculates the trajectory of single particles introduced at the entrance
3 to the larynx using a stochastic model for turbulent fluctuations incorporated into the particles'
4 equation of motion and time-averaged flow fields in the larynx and trachea. The effects of flow
5 rate and initial particle location on overall deposition were presented in the form of probability
6 density histograms of final particle deposition sites. At present, however, there are no
7 experimental data to validate results of such modeling.
8 A number of CFD models have been developed to study fluid flow and particle deposition
9 patterns in airway bifurcations. The bifurcation geometries that have been modeled include
10 two-dimensional (Li and Ahmadi, 1995); idealized three-dimensional using circular airways
11 (Kinsara et al., 1993) or square channels (Asgharian and Anjilvel, 1994); symmetric bifurcations
12 (Balashazy and Hofmann, 1993a,b); or physiologically realistic asymmetric single (Balashazy
13 and Hofmann, 1995; Heistracher and Hofmann, 1995) and multiple bifurcation models (Lee
14 et al., 1996; Heistracher and Hofmann, 1997; Comer et al., 2000, 2001a,b; Zhang et al., 2000,
15 2001, 2002) with anatomical irregularities such as cartilaginous rings (Martonen et al., 1994a)
16 and carinal ridge (Martonen et al., 1994b; Comer et al., 2001a) shapes incorporated. The CFD
17 flow simulations in the bifurcating geometry models show distinct asymmetry in the axial
18 (primary) and radial (secondary) velocity profile in the daughter and parent airway during
19 inspiration and expiration, respectively. In a systematic investigation of flow patterns in airway
20 bifurcations, numerical simulations were performed to study primary flow (Martonen et al.,
21 2001b), secondary currents (Martonen et al., 2001c), and localized flow conditions (Martonen
22 et al., 2001d) for different initial flow rates. The effects of inlet conditions, Reynolds numbers,
23 ratio of airway diameters, and branching angles with respect to intensity of primary flow, vortex
24 patterns of the secondary currents, and reverse flow in the parent-daughter transition region were
25 investigated. These simulated flow patterns match experimentally observed flow profiles in
26 airway bifurcations (Schroter and Sudlow, 1969).
27 Gradon and Orlicki (1990) computed the local deposition flux of submicron size particles
28 in a three-dimensional bifurcation model for both inhalation and exhalation, and they found
29 enhanced deposition in the carinal ridge region during inspiration and in the central zone of the
30 parent airway during expiration. Numerical models of particle deposition in symmetric three-
31 dimensional bifurcations were developed by Balashazy and Hofmann (1993a,b), and these were
June 2003 6-79 DRAFT-DO NOT QUOTE OR CITE
-------
1 subsequently extended to incorporate effects of asymmetry in airway branching (Balashazy and
2 Hofmann, 1995) and physiologically realistic shapes of the bifurcation transition zone and the
3 carinal ridge (Heistracher and Hofmann, 1995; Balashazy and Hofmann, 2001). In these
4 numerical models, three-dimensional airflow patterns were computed by finite difference or
5 finite volume methods, and the trajectories of particles entrained in the airstream were simulated
6 using Monte Carlo techniques considering the simultaneous effects of gravitational settling,
7 inertial impaction, Brownian motion, and interception. The spatial deposition pattern of inhaled
8 particles was examined for a range of particle sizes (0.01-10 jim) and flow rates (16-32 L/min)
9 by determining the intersection of particle trajectories with the surrounding surfaces. The
10 overall deposition rates derived using the CFD models correspond reasonably with experimental
11 data (Kim and Iglesias, 1989). These simulations predict deposition hot spots at the inner side of
12 the daughter airway downstream of the carinal ridge during inspiration, corresponding to the
13 secondary fluid motion of the inhaled air stream. During exhalation, the CFD models predict
14 enhanced deposition at the top and bottom parts of the parent airway, consistent with secondary
15 motion in the exhaled air stream. These studies indicate that secondary flow patterns within the
16 bifurcating geometry play a dominant role in determining highly non-uniform local particle
17 deposition patterns.
18 Zhang et al. (1997) numerically simulated particle deposition in three-dimensional
19 bifurcating airways (having varying bifurcation angles) due to inertial impaction during
20 inspiration for a wide range of Reynolds numbers (100-1000). Inlet velocity profile, flow
21 Reynolds number, and bifurcation angle had a substantial effect on particle deposition
22 efficiency. Based on the simulated results, equations were derived for particle deposition
23 efficiency as a function of nondimensional parameters, such as Stokes number, Reynolds
24 number, and bifurcation angle, and were shown to compare favorably with available
25 experimental results. More recently, Comer et al. (2000) have estimated the deposition
26 efficiency of 3-, 5-, and 7-|im particles in a three-dimensional double bifurcating airway model
27 for both in-plane and out-of-plane configurations for a wide range of Reynolds numbers (500-
28 2000). They demonstrated deposition in the first bifurcation to be higher than in the second
29 bifurcation, with deposition mostly concentrated near the carinal region. The non-uniform flow
30 generated by the first bifurcation had a dramatic effect on the deposition pattern in the second
31 bifurcation. Based on these results, they concluded that use of single bifurcation models is
June 2003 6-80 DRAFT-DO NOT QUOTE OR CITE
-------
1 inadequate to capture the complex fluid-particle interactions that occur in multigeneration airway
2 systems.
3 Comer et al. (2001a) further investigated detailed characteristics of the axial and secondary
4 flow in a double bifurcation airway model using 3-D CFD simulation. Effects of carina shape
5 (sharp versus rounded) and bifurcation plane (planar versus non-planar) were examined. Particle
6 trajectories and deposition patterns were subsequently investigated in the same airway model
7 (Comer et al, 2001b). There was a highly localized deposition at and near the carina both in the
8 first and second bifurcation, and deposition efficiency was much lower in the second bifurcation
9 than in the first bifurcation as demonstrated in the earlier study (Comer et al, 2000). They found
10 that deposition patterns were not much different between the sharp versus rounded carina shape
11 at Stokes numbers of 0.04 and 0.12. However, deposition patterns were altered significantly for
12 these particles when the bifurcation plane was rotated, suggesting that a careful consideration of
13 realistic airway morphology is important for accurate prediction of particle deposition by CFD
14 modeling.
15 Zhang et al. (2000, 2001) extended the studies of Comer et al. described above and
16 investigated effects of angled inlet tube as well as asymmetric flow distribution between
17 daughter branches. The flow asymmetry caused uneven deposition between downstream
18 daughter branches. Also noted was that the absolute deposition amount was higher, but
19 deposition efficiency per se was lower in the high flow branch than in the low flow branch. The
20 intriguing relationship between flow asymmetry and deposition was in fact consistent with
21 experimental data of Kim and Fisher (1999), indicating that the CFD model could correctly
22 simulate complicated airflow and particle dynamics that may occur in the respiratory airways.
23 Most CFD models use constant inspiratory or expiratory flows for simplicity and practical
24 reasons. However, the respiratory airflow is cyclic, and such flow characteristics cannot be fully
25 described by constant flows. Recent studies of Zhang et al. (2002) investigated particle
26 deposition in a triple bifurcation airway model under cyclic flow conditions mimicking resting
27 and light activity breathing. Deposition dose was obtained for every mm square area. They
28 found that deposition patterns were similar to those obtained with constant flows. However,
29 deposition efficiencies were greater with the cyclic flows than constant flows, and the difference
30 could be as high as 50% for 0.02 < mean Stk < 0.12 during normal breathing. The CFD results
31 are qualitatively comparable to experimental data (Kim and Garcia, 1991) that showed about
June 2003 6-81 DRAFT-DO NOT QUOTE OR CITE
-------
1 25% increase in deposition with cyclic flows. With further improvement of airway morphology
2 and computational scheme, CFD modeling could be a valuable tool for exploring the
3 microdosimetry in the airway structure.
4 Current CFD models of the acinar region are limited due to the complex and dynamic
5 nature of the gas exchange region. Flow simulation in a linearly increasing volume of a
6 spherical truncated two-dimensional alveolus model show distinct velocity maxima in the
7 alveolar ducts close to the entrance and exit points of the alveolus and a radial velocity profile in
8 the interior space of the alveolus (Tsuda et al., 1996). This is in contrast to simulations based on
9 a rigid alveolus (Tsuda, 1994a,b) and suggests that a realistic simulation of the flow pattern in
10 the acinar region should involve application of time-dependent methods with moving boundary
11 conditions. Nonuniform deposition patterns, with higher deposition near the alveolar entrance
12 ring, have been predicted using numerical models (Tsuda, 1994a,b, 1996).
13 Recent studies of Darquenne (2001) examined aerosol transport and deposition in
14 6-generation alveolated ducts using 2-D computer simulation. Particle trajectories and
15 deposition patterns were obtained for one complete breathing cycle (2 s inspiration and
16 2s expiration). There were large non-uniformities in deposition between generations, between
17 ducts of a given generation, and within each alveolated duct, suggesting that local deposition
18 dose can be much greater than the mean acinar dose.
19
20 6.6.4 Modeling Results Obtained with Models Available to the Public
21 Two relatively user-friendly computer models for calculating percent deposition in various
22 compartments of the respiratory tract as a function of particle size are publicly available.
23 Several model runs have been done to demonstrate the outputs of the models. Published results
24 from one model are also presented. Both model calculations are for particles of density of
25 1 g/cm3 so aerodynamic and Stokes diameter are the same.
26
27 6.6.4.1 International Commission on Radiological Protection (ICRP)
28 The LUDEP (Lung Dose Evaluation Program; National Radiologic Protection Board,
29 1994) model was developed concurrently with the ICRP (International Commission on
30 Radiological Protection, 1994) respiratory tract model mainly to help the ICRP Task Group
31 examine the model in detail by testing the predictions of deposition, clearance and retention of
June 2003 6-82 DRAFT-DO NOT QUOTE OR CITE
-------
1 inhaled radionuclides against experimental data, and by determining the model's implications for
2 doses to the respiratory tract (ICRP, 1994; NRPB, 1994). This model was designed to represent
3 the deposition of inhaled particles in the respiratory tract, the subsequent biokinetic behavior of
4 inhaled radionuclides, and the doses delivered to the respiratory tract. Although created for
5 calculating the internal dose of radionuclides, the model is useful for determining the deposition
6 of nonradioactive materials, but not for describing clearance of nonradioactive particles.
7 In particular, the model has wide applicability for calculating the regional deposition of particles
8 in the respiratory tract based on particle size, body size (age), breathing rate, activity patterns,
9 and exposure environment. The overall dosimetric model for the respiratory tract consists of
10 several critical elements important for dose calculations including detailed descriptions of
11 morphometry, respiratory physiology, and deposition. The morphometric element of the model
12 describes the structure of the respiratory tract and its dimensions. A description of respiratory
13 physiology provides the rates and volumes of inhaled and exhaled air which determines the
14 amount of material that can be deposited in the respiratory tract. Deposition characterizes the
15 initial distribution of the inhaled material within the different regions of the respiratory tract
16 specific to the age and gender of the subject and the physiological parameters. The ICRP model
17 covers the particle size range from 0.001 to 100 jim.
18 Two simulations were run to demonstrate some aspects of deposition as predicted by the
19 ICRP model. Respiratory parameters for a worker with a moderately high activity level and a
20 young adult with a lower activity level are given in Table 6-3. Each simulation was run for nasal
21 breathing and mouth breathing. The ICRP model calculates deposition in five compartments:
22 - ET1 - the extrathoracic region comprising the anterior nose;
23 - ET2 - the extrathoracic region comprising the posterior nasal passages, larynx, pharynx
and mouth;
24 - BB - the bronchial region;
25 - bb - the bronchiolar region consisting of bronchioles and terminal bronchioles; and
26 - Al - the alveolar-interstitial region consisting of the respiratory bronchioles, the alveolar
ducts with their alveoli and the interstitial connective tissue.
27 In the presentation of the model results, ET1 and ET2 are combined to give an ET (extrathoracic
28 region), BP and bb are combined to give a TB (tracheobronchial) region, and Al gives the A
June 2003 6-83 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-3. RESPIRATORY PARAMETERS USED IN LUDEP MODEL
Activity Related Physiological
Parameters
Ventilation Rate Frequency Tidal Volume
Activity Percent (m3/hr) (breaths/min) (mL)
Adult Male ICRP Defaults for Environmental Outdoor Exposure
Sleep
Sitting
Light Exercise
Heavy Exercise
0
50
38
12
0.45
0.54
1.5
3
12
12
20
26
625
750
1250
1923
Mean 1.2
Young Adult
Sitting 100 .45 15 500
1 (alveolar) region. Results are shown in Figures 6-13 to 6-15. Figure 6-13 shows the total and
2 regional deposition as a function of particle size for the worker: nasal breathing (13a), mouth
3 breathing (13b), and a comparison of nasal and mouth breathing for the TB and A regions (13c).
4 Figure 6-14 gives similar results for the young adult. For both simulations, the deposition is a
5 minimum between 0.1 and 1 |im diameter (the accumulation mode size range) and increases for
6 larger (coarse mode) and smaller (ultrafme particle) size ranges. For ultrafme particles, TB
7 deposition peaks between 0.001 and 0.01 jim and A deposition peaks between 0.01 and 0.1 jim.
8 The comparisons of nasal and mouth breathing in Figures 6-13c and 6-14c show almost no
9 difference in deposition between 0.01 and 1 |im. Below 0.1 more particles are removed by
10 diffusion in the extrathoracic (ET) region while above 1.0 more particles are removed by
11 impaction in the ET region. Therefore, mouth breathing leads to greater deposition of coarse
12 mode particles (Da > 1 jim) and of the smaller ultrafme particles (Dp < 0.01 jim). The
13 A deposition approaches zero as particle size increases to 10 jim. However, TB deposition
14 continues for larger particle sizes.
15 The TB and A deposition patterns of the worker under moderate activity and the young
16 adult under low activity are compared in Figure 6-15a and b. Increased activity lowers the
June 2003 6-84 DRAFT-DO NOT QUOTE OR CITE
-------
0.001
0.01 0.1 1 2.5 5 10 25
Particle Diameter, |jm
100
100
C
_o
+jjj
'w
o
Q.
0
O
10
0.001
0.01 0.1 1 2.5 5 10 25 100
Particle Diameter,
0.001 0.01 0.1 1 2.5 5 10 25
Diameter, pm
100
Figure 6-13. Percent deposition for total results of LUDEP model for an adult male
worker (default) showing total percent deposition in the respiratory tract
(TOT) and in the ET, TB, and A regions. Respiratory parameters given in
Table 6-3. (a) nasal breathing (NB), (b) mouth breathing (MB),
(c) comparison of nasal and mouth breathing for TB and A regions.
June 2003
6-85
DRAFT-DO NOT QUOTE OR CITE
-------
0.001
0.01 0.1 1 2.5 5 10 25 100
Particle Diameter, \im
100
0.001
0.01 0.1
Particle Diameter,
100
60
50
40
...TB(MB)
^ A (MB)
AAATB(NB)
•«• A (NB)
0.01
0.1 1 2.5 5 10 25 100
Diameter, |jm
Figure 6-14. Percent deposition for total results of LUDEP model for a young adult
(default) showing total percent deposition in the respiratory tract (TOT) and
in the ET, TB, and A regions. Respiratory parameters given in Table 6-3.
(a) nasal breathing (NB), (b) mouth breathing (MB), (c) comparison of nasal
and mouth breathing for TB and A regions.
June 2003
6-86
DRAFT-DO NOT QUOTE OR CITE
-------
•••TB(YA)
AAAA(YA)
.... TB(WK)
1 2.5 5 10 25
Diameter,
TB(YA)
AAAA(YA)
.... TB(WK)
0.01
0.1 1 2.5 5 10 25
Diameter, [Jin
100
100
Figure 6-15. Comparison of percent deposition in the TB and A regions for a worker
(WK; light exercise) and a young adult (YA; resting), (a) nasal breathing
and (b) mouth breathing.
1 A deposition of coarse particles for nasal breathing and lowers both A and TB deposition of
2 coarse particles for mouth breathing. It also shifts the maximum deposition for coarse particles
3 to smaller sizes. Increased activity increases A deposition of ultrafme particles and shifts the
4 maximum deposition to larger sizes. Increased activity also increases the A deposition of
5 accumulation-mode particles.
June 2003
6-87
DRAFT-DO NOT QUOTE OR CITE
-------
1 6.6.4.2 Multiple Path Particle Dosimetry Model (MPPD)
2 Some results from this model, developed by CUT (the Chemical Industry Institute of
3 Technology, USA) and RIVM (Directorate-General for Environmental Protection, The
4 Netherlands), will be taken from a RIVM report (Winter-Sorkina and Cassee, 2002). The MPPD
5 model allows calculation of PM deposition fractions and exposure doses for humans and rats,
6 and includes age-specific human lung models. The MPPD model covers the particle size range
7 from 0.01 to 10 jim. The model may be used to improve understanding of the exposure-dose-
8 response relationships observed in environmental epidemiological studies and for extrapolation
9 of studies in experimental animals to humans. In addition, factors resulting in increased
10 susceptibility can be studied. The report describes the results of monodisperse aerosol
11 deposition calculations with the MPPD model and its sensitivity to various parameters.
12 Deposition of inhaled PM depends primarily on exposure concentrations, physical characteristics
13 of the particles, lung morphometry, and breathing parameters, and cannot easily be measured.
14 Therefore, computer models such as the MPPD model have proven to be important tools to
15 analyze PM dosimetry. Because these models use an explicit set of equations which describe
16 real-life processes, either empirically or based on first principles, they are especially suited to
17 analyze effects of scenarios such as particulate exposure control strategies. The age of the
18 subject, the functional capacity of the lungs, and breathing parameters as well as the individual
19 lung morphometry are factors that significantly affect the particle deposition and can explain the
20 susceptibility of subpopulations. Results depicting deposition as a function of minute ventilation
21 (a surrogate for exertion or exercise level) and as a function of age by particle size for various
22 respiratory tract regions will be shown.
23
24 6.6.4.2.1 Deposition as a function of physical exertion
25 Earlier studies indicate that PM deposition depends on the level of physical exertion.
26 Information on this dependency as well as on activity patterns is necessary for an estimate of the
27 actual exposure of a whole population. Winter-Sorkina and Cassee (2002) used the MPPD
28 model with Yeh-Schum 5-lobe limited multiple-pass particle deposition to calculate aerosol
29 deposition in the human adult at different levels of physical exertion. The model uses data made
30 available by Yeh and Schum (1980) that characterizes individual airways at the level of the
June 2003 6-88 DRAFT-DO NOT QUOTE OR CITE
-------
1 segmental bronchi, but describes the airways within each lobe in a single-path manner.
2 A separate symmetric tree represents each of the five lobes.
3
4 Levels of physical exertion for adults, corresponding representative activities and
5 corresponding minute ventilation (CARB, 1987) used in the calculation are presented in
6 Table 6-4. The breathing frequency and tidal volume for different physical exertion levels
7 (Table 6-4) are calculated from minute ventilation keeping the ratio of breathing frequency and
8 tidal volume nearly constant. For normal augmenters, the switch to oronasal breathing
9 (combined nose and mouth breathing) is considered to occur at a minute ventilation of
10 35.3 L/min. Partitions of airflow between the nose and mouth as given by Niinimaa et al. (1981)
11 are used for the oronasal breathing. The partitioning flow is assumed to be the same for inhaled
12 and exhaled air. For minute ventilation lower than this value, breathing is only through the nose,
13 therefore, the calculations present a discontinuity at this point. Calculations are performed for
14 monodisperse aerosol particles with 10 different aerodynamic diameters ranging from 0.01 um to
15 10 um and with a particle density of 1 g/cm3. The deposited mass rates were calculated for an
16 aerosol concentration of 140 ug/m3.
17 Results on aerosol deposition as a function of physical exertion for different particle sizes
18 are shown in Figure 6-16. The head deposition fractions for 1.3 um, 2.5 um and 5 um particles
19 increase from rest to light exercise. They decrease with a factor of respectively 2.3, 1.8, and
20 1.5 and further stay about constant when breathing is changed from nasal to oronasal at modest
21 and heavy exercise with minute ventilation of 40 L/min and higher. The head deposition fraction
22 of ultrafine particles decreases slightly from rest to light exercise. Tracheobronchial deposition
23 fractions for ultrafine particles of 0.01 um, 0.02 um, and 0.04 um decrease from rest to light
24 exercise, decrease slightly further to heavy exercise for 0.01 um particles and stay constant for
25 0.04 um particles.
26 Tracheobronchial deposition fraction for coarse particles decreases slightly from rest to
27 light exercise and rises when breathing is changed from nasal to oronasal. It increases from
28 modest to heavy exercise especially for 5 um particles. Tracheobronchial deposition fraction of
29 ultrafine particles is larger than deposition fraction of coarse particles at rest, light and modest
30 exercise, however, at heavy exercise the deposition fraction of 5 um particles is larger than that
31 of ultrafine particles. Pulmonary or alveolar deposition fraction of ultrafine particles increases
32 from rest to light exercise, deposition fraction of coarse 2.5 um and 5 um particles decreases
33 from rest to light exercise, rises when breathing is changed from nasal to oronasal and decreases
34 slightly from modest to heavy exercise. Thoracic deposition fraction shows a light increase for
35 0.01 um and 0.02 um particles and a decrease for 2.5 um and 5 um particles from rest to light
36
June 2003 6-89 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-4. LEVELS OF PHYSICAL EXERTION FOR ADULT, CORRESPONDING
REPRESENTATIVE ACTIVITIES, AND BREATHING PARAMETERS
Minute
ventilation,
L/min
5
7.5
13
19
25
30
35
40
59 (55-63)
72
85
100 (> 100)
Source: CARB
Breathing
frequency,
min"1
10
12
16
19
22
24
26
28
34
37
40
44
(1987).
Tidal
Volume,
mL
500
625
813
1,000
1,136
1,250
1,346
1,429
1,735
1,946
2,125
2,273
Exertion Level
Rest
Rest
Light
Light
Light
Modest
Modest
Modest
Heavy
Very heavy
Very heavy
Extremely heavy
Representative activity
Sleep
Awake
Walk (4 km/h); washing clothes
Walk (5 km/h); bowling; scrubbing floors
Dance; push a 15 kg wheelbarrow; building
activities; piling firewood; walk (7 km/h)
Quiet cycling; pushing a 75 kg wheelbarrow;
using a sledgehammer
Climb 3 stairs; play tennis; digging soil
Cycle (23 km/h); walk in snow; digging a
trench; jogging
Skiing cross-country; mountaineering;
climbing stairs with weight
Squash and handball; chopping wood
Running (18 km/h); cycle racing
Marathon; triathlon; cross-country ski race
1 exercise. Deposited thoracic mass rate increases with increasing physical exertion, faster for
2 heavy exercise. At light exercise with a minute ventilation of 25 L/min the deposited thoracic
3 mass rate is 13 times larger than at rest awake (7.5 L/min) for 0.01 um particles and 4 times
4 larger for 5 um particles. At modest exercise with minute ventilation of 40 L/min the deposited
5 thoracic mass rate is 36 times larger than at rest awake (7.5 L/min) for 0.01 um particles and
6 44 times larger for 5 um particles.
7
8 6.1.4.2.2 Deposition as a function of age
9 An important issue in risk assessment is the age dependency of PM deposition, especially
10 for children. The CIIT/RIVM particle deposition model includes age-specific lung models.
11 Winter-Sorkina and Cassee (2002) used CIIT/RIVM particle deposition mode to calculate age
12 dependent deposition for the ages and respiratory parameters given in Table 6-5.
June 2003
6-90
DRAFT-DO NOT QUOTE OR CITE
-------
Total Deposition
1.0
0.8
0.6
0.4
8.2
0.0
20
40 60 80
® Ventilation, t/min
100
C Tracheobronchial (TB) Deposition
0.3
c 0.2
0.1
0.0
20 40 SO 80 100
Minute ¥@ntiiatiai\ 1/iBin
Thoracic (TB+P) Deposition
0.6
0.5
0.4
0.3
0.2
0.1
0.0
""±£S*
O
20 40 60 80 100
Ventilation, I/iron
1.0
0.8
Head Deposition
0.$
8 °-4
8-
o
0.2
0.0
O
B
0 20 40 60 80
Minute Ventilation, 1/mtn
d Pulmonary (P( Deposition
100
0.5
0.4
0.3
0.2
0.1
0,0
20 40 60 80
fiiinute VentHatioet, l/min
Thoracic Mass Rate
100
100
Figure 6-16. Dependency of aerosol deposition in human adults on physical exertion
expressed as minute ventilation for different particle sizes. Aerosol
concentration used for mass calculation is 140 ug/m3.
Source: Winter-Sorkina and Cassee (2002).
June 2003
6-91
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-5. PARAMETERS USED IN AGE DEPENDENT CALCULATIONS OF
THE CIIT/RIVM DEPOSITION MODEL
Age
3 month
21 month
23 month
27 month
3 years
8+ years
9+ years
14 years
18 years
21 years
FRC, mL
27.36
64.46
78.45
100.67
95.43
437.34
513.12
881.47
1,935.34
1,854.54
URT volume,
mL
2.45
6.52
6.94
7.92
9.47
21.03
22.44
30.63
37.38
42.27
Breathing
frequency, min"1
39
28
27
26
24
17
17
16
15
14
Tidal
volume, mL
30.44
81.22
86.79
100.1
121.3
278.2
295.8
388.1
446.7
477.2
Minute ventilation,
L/min
1.19
2.27
2.34
2.60
2.91
4.73
5.03
6.21
6.70
6.68
Source: Winter-Sorkina and Cassee (2002).
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Results of age-dependent deposition using the parameters given in Table 6-5 are
shown in Figure 6-17. The head impaction are based on the data from Becquemin et al.
(1991). For coarse particles the adult (here 18 and 21 years old) head deposition fractions are
larger than the head deposition fractions in children. The thoracic deposition fraction (which
is a sum of tracheobronchial and pulmonary deposition fractions) of ultrafine particles does
not change with age. For coarse particles (5 um and 10 um) tracheobronchial and thoracic
deposition fractions are significantly larger for children (ages of 0-15 years old) than for
adults, mainly due to the increase in head deposition from children to adults. The difference
in tracheobronchial and thoracic deposition fractions between children and adults increases
with particle size.
Pulmonary or alveolar deposition fractions of 5 um particles for 8-14 years old
children are higher than for adults. Deposited aerosol mass rate in the thoracic region
increases with age for ultrafine particles. For coarse particles the deposited aerosol mass rate
in the thoracic region increases with age up to the age of 14 years. The increase of deposited
mass rate is due to the growing tidal volume (Table 6-5). For coarse particles the deposited
aerosol mass rate in the thoracic region of 8-14 years old children for 5 um particles and of
2-14 years old children for 10 um particles is higher than in adults (18 and 21 years old).
June 2003
6-92
DRAFT-DO NOT QUOTE OR CITE
-------
Head Deposition
XX
C Tracheobronchial {TB) Deposition
0.6 -
* 0.2 Pk
a
0.1
0.0
D
-H-
*
XX ft X X
O A A
0 5 10 15 20
Ag®r Yesrs
6 Thoracic (TEH-P| Deposition
20
0 S 10 IS
Ag*> Years
d Pulmonary (P) Deposition
0.5
o
w
^ 0.3
5
|o,2
a
0,1
00
"x Xx
>| * * *
t«$c o
^ ++ + *
j-^ D
-,£3 , 43. . n
0 5 10 15
Age, Years
20
X
O
A
,n
20
f Thoracic Deposited Mass Rate
0.6
c
i
1,0.5
£X
&
S 0.4
a.
m 0.3
"O
- 0.2
w
o
_
JK
^r JK
jy ffl
aP o
0° 0
^ A
>w\
Hr
a 'w
5°-1i a
0 Ql i t t
0 5 10 15
«
i^
O
A
Q
L
20
Figure 6-17. Age dependency of human aerosol deposition for different particle sizes.
Total (a), head (b), tracheobronchial (c), pulmonary (d) and thoracic
(e) deposition fractions, deposited thoracic mass rate (f). Aerosol
concentration used for mass calculation is 140 ug/m3.
Source: Winter-Sorkina and Cassee (2002).
June 2003
6-93
DRAFT-DO NOT QUOTE OR CITE
-------
1 Here it should be emphasized that the age dependency of deposited mass is determined by the
2 age dependencies of head deposition and minute ventilation (tidal volume multiplied by
3 breathing frequency), and that the age dependency of head deposition is based on a limited
4 number of measurements.
5
6 It is also useful to examine particle deposition normalized to some parameter such as lung
7 mass, surface area, or number of alveoli. Aerosol deposition normalized to surface area and
8 alveoli is shown in Figure 6-18.
9
10 The CIIT/RIVM model calculates lung surface area per airway generation, the first
11 16 generations belong to the tracheobronchial region. The tracheobronchial lung area,
12 tracheobronchial deposition fractions per unit of surface area and deposited tracheobronchial
13 mass rates per unit surface area are shown in the top of the Figure 6-18. The
14 tracheobronchial surface area grows monotonously from about 197 cm2 at birth to about
15 1,554 cm2 at the age of 21 years. Tracheobronchial deposition fractions per unit surface area
16 are decreasing with age for all particle sizes due to increasing tracheobronchial lung area with
17 age. Tracheobronchial deposition fractions per unit surface area are up to 10 times (for
18 ultrafine particles) and 68 times (for coarse particles) higher for 3 month old babies compared
19 to adults; up to 4 times (ultrafine) and 27 times (coarse) higher for the age of 2 years
20 compared to adults. For ultrafine particles the deposited aerosol mass rates in
21 tracheobronchial region per unit surface area for 3 month old babies are up to 1.8 times larger
22 than for adults and seem to decrease monotonously with progressing age. However, small
23 deviation in tracheobronchial surface area at the age of 2.3 and 3 years, due to the differences
24 in lung geometry, leads to almost same tracheobronchial deposited mass rates per unit surface
25 area as for adults. For coarse particles of 2.5 um, 5 um and 10 um the deposited aerosol mass
26 rates in tracheobronchial region per unit surface area for 3 month old babies are respectively
27 2.5, 8, and 12 times larger than for adults, for 2.3 years old children respectively 1.2, 2.2, and
28 6.2 times larger than for adults, and for 8 years old children respectively 1.7, 3.5, and
29 11 times larger than for adults. The age dependency of tracheobronchial deposited mass per
30 unit surface area is determined by the age dependencies of head deposition, minute
31 ventilation, and tracheobronchial surface area.
32 The total number of alveoli, pulmonary deposition fractions per alveolus and deposited
33 pulmonary mass rates per alveolus as a function of age are shown in the bottom part of the
34 Figure 6-18. There are approximately 50 • 106 alveoli at birth and about 85% of alveoli are
3 5 added after birth, the adult number of about 3 00 • 106 is attained by 20 years (Mauderly,
June 2003 6-94 DRAFT-DO NOT QUOTE OR CITE
-------
8 Tracheobronchial fTB) Lung Area
b TB Deposition I TB Lung Area
3
I
8.15
„
E
©
E 0.10
«t
S
0.05
0 0
o
o
o
-
o
<*> °
"$
>
r 30
e
fas
a
»- 20
£1
£ 15
c
1 10
ffl
n
J
^ -|- 0.01pm
U ^ 0.02pm
X 0.04pm
Rr^i **\
- U \f 2.5pm
<$ A 5pm
q-j n 1*m
i i. i. ^ a
0 5 10 15 20 0 5 10 15 20
C TB Mass Rats 1 TB Lung Area d Number of Alveoli
«t
1
~a
a
«f 4
I
" 3
5
a
1 2
E
1 1
1
§•
-JU
A *
Lj
J- i
* jP "N-
fe_
^O" ^.^ ^. «1_
tg5 -f-
vjt » ?R 5j£, *
%\,?C? Xs . _A_
T1* sni *^
t s i
«,»io-
10*10*
2.5x10*
| 2,0)rt08
5
1.5x10*
1.0x1 a8
*.*«?
" 4fi& A4 A A A
-
.
^
-
*
2 0 S 10 15 20 0 S 10 15 20
Age, Years Age, Years
e Pulmonary Deposition per Alveolus c f Pulmonary Mass Rate per Alveolus
i~t
,~s
0
<
-
J*
K
i
-
t^ i M M
I* n. v »
rin B n S
* 1.2x10-*
1 1.0x10"*
3
*3 *n
|
JL
O
1
< XX
Xx i * ^
5 ^S A
m o o
^ +H + i *
™ «SK ^
i-^t. °
AjjjTt r°T.
jnij i-Q ^ D ,n
0 5 10 15 20 0 i 10 15 20
Ago, Years *§», Years
Figure 6-18. Age dependency of human standardized aerosol deposition for different
particle sizes. Tracheobronchial (TB) lung area (a), TB deposition fraction
per unit of TB lung area (b), deposited TB mass rate per unit of TB lung area
(c), total number of alveoli (d), pulmonary (P) deposition fraction per
alveolus (e), deposited P mass rate per alveolus (f). Aerosol concentration
used for mass calculation is 140 ug/m3.
Source: Winter-Sorkina and Cassee (2002).
June 2003
6-95
DRAFT-DO NOT QUOTE OR CITE
-------
1 1979). Alveolar multiplication is extremely rapid in the first few years of life and then slows
2 down. Pulmonary deposition fractions per alveolus are up to 5 times (for ultrafine particles)
3 and 6 times (for coarse particles) higher for 3 month old babies compared to adults.
4 For children of the age of about 2 years and older the pulmonary deposition fraction per
5 alveolus does not change significantly. Deposited pulmonary mass rates per alveolus are
6 lower for the age of 2-3 years compared to adults for ultrafine and 2.5 um particles and 1.8 to
7 2.2 times higher for 8-14 years old children compared to adults for 5 um particles. The age
8 dependency of pulmonary deposition per alveolus is determined by the age dependencies of
9 head deposition, minute ventilation, and alveolar multiplication.
10 Alveolar surface area obtained from 8 normal adult human lungs by Gehr et al. (1978)
11 is 143 ± 12 in2. The airway surface area for generations above 16 belonging to the
12 pulmonary 2 region calculated from the model is 9.35 m2. Therefore, the adult
13 tracheobronchial deposition fraction and mass rate per unit surface area are 2,078 to 377
14 times (for 0.01 um to 0.04 um particles) and 223 to 6,238 times (for 2.5 um to 10 um
15 particles) larger than adult pulmonary deposition fraction and mass rate per unit surface area.
16 Progressive morphological changes in the senescent adult lung result primarily in a loss of
17 alveolar surface area and altered elastic properties. Alveolar septal membranes weaken and
18 stretch, causing an enlargement of alveoli and a reduced surface area. Changes occurring in
19 the alveolar septal wall result in a nearly linear decrease of surface area between the ages of
20 20 and 80 years, such that by 80 years the surface area is reduced by approximately 30%
21 (Mauderly, 1979). There is little age-related change of breathing patterns of adults at rest
22 although there is a slight trend toward a larger minute ventilation with age. The minute
23 ventilation during exercise increases with age (Mauderly, 1979). Thus, the pulmonary
24 deposition fraction, mass rate and deposited mass per unit surface area increase nearly linear
25 between the ages of 20 and 80 years by approximately 30%.
26
27 6.6.4.3 Comparisons of Deposition in Humans and Rats
28 Dosimetric issues are important in the use of animal to human extrapolation in risk
29
30
31
32
33
34
35
assessment. The MPPD model was used to compare deposition in humans and rats. The MPPD
model uses the multiple-path aerosol deposition model for a rat (Anjilvel and Asgharian, 1995)
which incorporates asymmetry in the lung branching structure and calculates deposition at the
individual airway level. Deposition calculations were performed with the 5 lobe lung model for
humans for light exercise. Respiratory parameters used in the model runs are shown in
Table 6-6. The percent deposition for human mouth breathing, human nasal breathing, and rat
nasal breathing (rats are obligate nose breathers) are shown in Figure 6-19a, b, and c for ET, TB,
June 2003 6-96 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-6. RESPIRATORY PARAMETERS FOR HUMANS AND RATS
Rat
Human
Breaths
min"1
102
20
Tidal Volume
mL
2.1
1,250
FRM
mL
4
3,300
URT
mL
.42
50
Lung mass
g
4.34
1,100
90-
-^^^^^
y
X
vs.
•••J**-**^
-»- ET(NB)
-•-ET(MB)
-A- ET(RAT)
P
# i
tmtf&sltn
jf
/
/
^ j^M^B^ ^
/ X
A
t^*
0.01 0.1 1 2.5 5 1
Particle Diameter, prn
b-1
Particle Diameter, pm
Diameter, jjrn
Diameter, prn
Figure 6-19. Comparison of percent deposition for rats (nasal breathing) and humans
(nasal and mouth breathing) and the ratio of human to rat for nasal and
mouth breathing humans for the ET (a), TB (b), and A (c) regions of the
respiratory tract.
June 2003
6-97
DRAFT-DO NOT QUOTE OR CITE
-------
1 and A deposition. Figure 6-19 also shows the ratios of percent deposition for human to rat for
2 mouth breathing and nasal breathing humans.
3 ET deposition is shown in Figure 6-19a. Deposition of coarse mode particles in the ET
4 region increases significantly with particle size because of impaction. However, increased
5 inertia poses a limitation to the ability of particles to enter the ET region. This reduction in the
6 inhaled fraction of the aerosol is relevant for particle sizes larger than 3-4 jim for rats and sizes
7 larger than about 8 jim for humans and is more significant for rats than for humans. The
8 inhalability adjustment (Menache et al., 1995) used in the MPPD model does not change
9 deposition results for humans significantly, the tracheobronchial deposition fraction reduces
10 3.5% and thoracic deposition fraction 2.5% for 10 jim particles. For rats accounting for
11 inhalability reduces the nasal deposition fraction about 1.5 times for 5 jim particles and more
12 than 2 times for 10 jim particles. As a result tracheobronchial and pulmonary deposition
13 fractions are reduced about 25% for 5 jim particles. ET percent deposition is greater for humans
14 than rats, above about 0.15 jim for nose breathing and 0.3 jim for mouth breathing, except that
15 for mouth breathing, human percent deposition drops below that of rats at about 3 jim. This
16 leads to a peak in the human/rat ratio at 1 jim. The fraction TB percent deposition
17 (Figure 6-19b) is much lower for rats than humans in the accumulation mode size range.
18 However, between 1.5 and 5 jim the percent deposition for the rat is greater than that for the
19 nasal breathing human. Above about 2.5 jim, the percent deposition for the mouth breathing
20 human increases rapidly relative to that of the rat. For A deposition (Figure 6-19c), rats and
21 humans have almost the same percent deposition in the accumulation mode region. However,
22 the percent deposition for the nasal breathing human and the rat fall off for particles above about
23 2.5 jim, the rat more rapidly than the human. These differences are borne out in the human/rat
24 ratios which become very high for particles above 2.5 jim.
25 The percent deposition values for human and rat, shown in Figure 6-19, can be used with
26 respiratory parameters and respiratory tract surface areas or lung mass to normalize deposition to
27 lung mass, TB surface area, or A surface area provided those parameters are known.
28 Figure 6-20a compares deposition of PM by size in humans and rats normalized to lung mass for
29 thoracic (TH = TB + A) deposition. The deposition, in terms of jig of PM deposited per gram of
30 lung is greater for humans than rats for particles below about 2 jim for mouth breathing humans
31 and for particles below about 5 jim for nasal breathing humans. As can be seen in 6-20b and c,
June 2003 6-98 DRAFT-DO NOT QUOTE OR CITE
-------
Particle Diameter, pm
Diameter, (jrn
Figure 6-20. Normalized deposition patterns for rats (nasal breathing) and humans (nasal
and mouth breathing and the ratio of human to rat for nasal and mouth
breathing humans for the thoracic region (in terms of jig PM per g of lung).
Quantity of PM deposited based on 8 hour exposure to 100 ug/m3.
1 the ratio of human to rat deposition, especially for mouth breathing, increases to very high values
2 for particles above about 2.5 jim.
3 From the above comparison of rats and humans, it would appear that for inhalation studies
4 with accumulation mode aerosols, as might be done using concentrated air particles, equivalent
5 TH deposition in rats could be obtained with 0.5 to 0.75 of concentrations for humans.
6 However, for coarse particles the deposition ratios are very sensitive to particle size. Thus, for
7 coarse particles resuspended from bulk material particle size distribution measurements would
8 be needed and very high concentration ratio might be needed for equivalent deposition on a per
9 gram of lung b asi s.
10 There is some variation in the reported values for the surface areas of the various portions
11 of the human and rat respiratory tract as listed in Table 6-7. The results using the U.S. EPA
12 default surface areas are shown in Figure 6-21. For TB deposition in terms of jig of PM per cm2
13 of bronchial surface, shown in Figure 6-2 la, the human percent deposition is greater than that of
14 the rat except for particles between about 1.5 and 5 jim. In this size range the rat deposition is
15 greater than that of the nasal breathing human. Again, the ratio increases rapidly, especially for
16 mouth breathing and for larger particles. The A deposition (Figure 6-2Ib) for nasal breathing
17 humans and rats is similar — between 0.05 and 3 jim with rat deposition dropping for particles
18 above 3 jim. However, the ratios increase above 3 jim and rapidly above 5 jim.
19
June 2003
6-99
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-7. SURFACE AREAS OF TRACHEOBRONCHIAL AND ALVEOLAR
REGIONS FOR HUMANS AND RATS
Surface Areas
Human Rat
TB A TB A
EPA Default3 .269 54 .00225 0.34
CIIT/RIVM Model b .1554C 150.3d .00124 e .55C
Human/Rat Ratios
TB A
119.6 158.8
125.3 273.3
aU.S. EPA (1996) based on U.S. EPA 1994).
b As reported in Winter-Sorkina and Cassee (2002).
cMauderly (1979).
dGehr et al. (1978). (143 m2 alveolar + 7.3 m2 respiratory bronchioles).
e Calculated from human/rat ratio in Winter-Sorkina and Casse (2002).
a-2
a-3
b-2
ffi
&L
o: 20
c
ffl
E
•r 10
1 2.5 5 10
0.1
Particle Diameter,
b-3
c 30
ra
E
Diameter, (jrn
Diameter, |jm
Figure 6-21. Normalized deposition patterns arising from 8 hr exposure to 100 ug/m3,
based on EPA default values of surface area, for rats (nasal breathing) and
humans (nasal and mouth breathing) and the ratio of human to rat (a) for
the TB region (in units of ug PM per m2 TB area) and (b) for the A region (in
terms of ug PM per m2 of A).
June 2003
6-100
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
The results using the CIIT/RIVM surface area values are shown in Figure 6-22. As would
be expected for the changes in surface area, the TB deposition amounts are larger and the
A deposition amounts are smaller (Figure 6-22a-l and 22b-l) and the A deposition for mouth
breathing is not as much greater for humans than rats for coarse particles. However, the ratios
(Figure 22a-2,3 and 22b-2,3) are not greatly different.
a-2
a-3
ffi
E
u
30
__________________
,
,
;
/
i __/_
ZZIZIII^ZZIZIZ
TTT^r^I^^
Particle Diameter, Mm
Particie Diameter, |j
Diameter, [jm
Figure 6-22. Normalized deposition patterns arising from 8 hr exposure to 100 ug/m3,
based on surface area values from Winter-Sorkina and Cassee (2002), for
rats (nasal breathing) and humans (nasal and mouth breathing) and the ratio
of human to rat (a) for the TB region (in units of jig PM per m2 TB area) and
(b) for the A region (in terms of ug PM per m2 of A).
1 The human/rat comparisons, whether normalized by lung mass or by either sets of surface
2 areas, indicate that for fine particles normalized human and rat deposition are comparable.
3 However, for coarse particles much higher exposures may be required for rats to obtain
4 equivalent normalized doses.
June 2003
6-101
DRAFT-DO NOT QUOTE OR CITE
-------
1 6.7 SUMMARY AND CONCLUSIONS
2 6.7.1 Particle Dosimetry
3 Understanding the mechanisms of action and ultimate biological effects of inhaled
4 particulate matter requires knowledge of the dosimetry of such material. This is because the
5 proximal cause of the biological response is the dose of particles delivered to and retained at the
6 target site, rather than the exposure concentration. Deposition, clearance, translocation, and
7 retention comprise the essential elements of dosimetry.
8 Dosimetry of inhaled particles is essential for extrapolating effects found in controlled
9 exposure studies of laboratory animals to those observed in human exposure studies, and for
10 relating effects in healthy individuals to those in potentially susceptible persons.
11 Understanding of total deposition as a function of particle size and breathing pattern and of
12 certain aspects of regional deposition of particles has improved since publication of the 1996 PM
13 AQCD. The ET region, especially the nasal passages, is a moderately efficient filter for ultrafme
14 and coarse particles. Accordingly, particles removed in the ET region are not available for
15 deposition in the TB and A regions of the respiratory tract. Within the thoracic region, the
16 deposition distribution of ultrafme particles is highly skewed towards the proximal airway
17 regions and resembles that of coarse particles. Thus, the deposition patterns for ultrafme
18 particles are similar to those of coarse-mode particles with significant fractional deposition in all
19 three regions. Particles in the accumulation mode size range (0.1 to 1.0 jim) have low fractional
20 deposition in all three regions.
21
22 6.7.2 Host Factors
23 Certain host factors have a marked effect on particle dosimetry and can affect the
24 biological response to inhaled particulate matter.
25
26 Gender
27 There are significant gender differences in the homogeneity of deposition as well as the
28 deposition rate of particles. These differences derive from differences between males and
29 females in body size, conductive airway size, and ventilatory parameters. Females have a
30 greater deposition of coarse mode particles in the ET and TB regions, and lower deposition in
31 the A region. This gender effect appears to be particle size dependent showing a greater
June 2003 6-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 fractional deposition in females for very small ultrafine and large coarse particles. Total
2 fractional lung deposition for 0.04 and 0.06 jim particles also appears to be greater in females
3 than males but only negligibly so for particles in the size range 0.8 - 1.0 |im. As the particle size
4 increases (3 to 5 jam), total fractional deposition increases in females. While deposition appears
5 to be more localized in females than males, deposition rate appears to be greater in males.
6
7 Exercise
8 Exercise may also increase the potential health risks of inhaled particles because exercise
9 increases the rate of oxygen consumption and changes ventilatory parameters affecting airflow
10 rate and breathing patterns. The switch from nose breathing to mouth breathing, which occurs as
11 exercise intensity increases, leads to an increase in fractional deposition of coarse particles in the
12 TB and A regions. The higher breathing rate and larger tidal volume lead to a greater amount of
13 deposition. Total lung deposition rate may be 3 to 4 times greater during exercise. The more
14 rapid breathing of children also leads to a greater amount of deposition.
15
16 Age
17 Airway structure and physiological function vary with age and health status of the
18 respiratory tract. Such variations may alter the deposition patterns for inhaled particles.
19 Significant age differences have been predicted by mathematical models and observed in
20 experimental studies. These studies generally indicate that ET and TB deposition is greater in
21 children, and children receive greater doses of particles per lung surface area than adults.
22 Unfortunately, deposition studies in another susceptible population, the elderly, are still lacking.
23
24 Lung Disease
25 A number of studies have examined particle deposition in chronic lung disease. These
26 studies indicate that total lung deposition is generally increased with obstructed airways.
27 Airflow distribution is very uneven in diseased lungs, and deposition can be enhanced locally in
28 areas of active ventilation.
29
30
31
June 2003 6-103 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.7.3 Laboratory Animal Studies
2 It is difficult to systematically compare deposition patterns in laboratory animals used in
3 dosimetric studies. Deposition patterns are similar between laboratory animals and humans but
4 there are absolute differences in deposition fractions. In most laboratory animal species,
5 deposition in the ET region is near 100% for particles greater than 5 |im, indicating greater
6 efficiency than that seen in humans. In the TB region, there is a relatively constant, but lower
7 deposition fraction for particles greater than 1 jim compared to humans. Finally, in the A
8 region, deposition fraction peaks at a lower particle size (~1 |im) in laboratory animals than in
9 humans.
10 Clearance processes are similar in animals and humans but the clearance rate for particles
11 is typically faster in laboratory animals.
12 There is a need for better laboratory models of susceptible human populations. Once
13 particles are deposited on the surface of the airways, they are subsequently cleared from the
14 respiratory tract completely or translocated to other sites within the system by distinct regional
15 processes. Ultrafine particles can be rapidly cleared from the lungs into the systemic circulation
16 where they can be transported to extrapulmonary regions. Such transport could provide a
17 mechanism whereby particles could affect cardiovascular function as reported in the
18 epidemiology studies (Chapter 8).
19
20 6.7.4 Mathematical Models
21 There has been significant improvement in the mathematical and computational fluid
22 dynamic modeling of particle dosimetry in the respiratory tract of humans. Although the
23 models have become more sophisticated and adaptable, validation of the models by
24 experimental data is still required.
25
26 Key Points
27 • Dosimetry establishes the relationship between PM exposure and the dose of PM
delivered to and retained at the target site. Deposition, clearance, translocation,
and retention comprise the essential elements of dosimetry.
June 2003 6-104 DRAFT-DO NOT QUOTE OR CITE
-------
• Dosimetric information is critical to extrapolating effects found in controlled exposure
studies of laboratory animals to those observed in human exposure studies and for relating
effects in normal healthy persons to those in potentially susceptible persons.
• Dosimetry separates the respiratory tract into three regions, extrathoracic (ET),
tracheobronchial (TB), and alveolar (A), based on anatomical features and particle
deposition and clearance phenomena that occur within each region.
• Particles in the accumulation mode size range (0.1 to 1.0 jim Dp) have the lowest
deposition fraction in all three regions.
• Coarse and ultrafine particles have higher fractional deposition. For coarse particles,
fractional deposition peaks between 5 and 10 jim Dp for the TB region and 2.5 and 5 jim
Dp for the A region.
• For ultrafine particles, fractional deposition peaks between 0.0025 and 0.005 |im Dp for
the TB region and between 0.01 and 0.05 for the A region.
• A significant fraction of ultrafine and coarse particles, but not particles in the
accumulation-mode size range, are deposited in the ET region.
• Once particles are deposited on the surface of the airways, they are subsequently cleared
from the respiratory tract completely or translocated to other sites within the system by
distinct regional processes. Ultrafine particles can be rapidly cleared from the lungs into
the systemic circulation where they can be transported to extrapulmonary regions. Such
transport could provide a mechanism whereby particles could affect cardiovascular
function as reported in the epidemiologic studies
• Fractional deposition, as a function of particle size, depends on lung size, tidal volume,
and breathing rate. Exercising subjects receive higher doses of particles per cm2 of lung
surface than subjects at rest.
• Airway structure and physiological function vary with age. Such variations may alter the
deposition patterns for inhaled particles. Airflow distribution is very uneven in diseased
lungs, and deposition can be enhanced locally in areas of active ventilation. Total lung
deposition is generally increased by obstructed airways so that particle deposition is
enhanced in people with chronic lung disease. Unfortunately, deposition studies in
another susceptible population, the elderly, are still lacking.
June 2003 6-105 DRAFT-DO NOT QUOTE OR CITE
-------
1 • Computational models allow calculation of fractional deposition and dose per cm2 of lung
surface as a function of particle size and respiratory parameters for humans and some
animals (including the laboratory rat). Such calculations can be used to predict the
exposures needed to produce comparable doses for animal to human extrapolation.
2 • Computational models have been improved in recent years but experimental validation of
model predictions is still required.
3
4
June 2003 6-106 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 Adgate, J. L.; Ramachandran, G.; Pratt, G. C.; Waller, L. A.; Sexton, K. (2002) Spatial and temporal variability in
9 outdoor, indoor, and personal PM2.5 exposure. Atmos. Environ. 36: 3255-3265.
10 American Conference of Governmental Industrial Hygienists (ACGIH). (1985) Particle size-selective sampling in
11 the workplace: report of the ACGIH technical committee on air sampling procedures. Cincinnati, OH:
12 American Conference of Governmental Industrial Hygienists.
13 Anjilvel, S.; Asgharian, B. (1995) A multiple-path model of particle deposition in the rat lung. Fundam. Appl.
14 Toxicol. 28: 41-50.
15 Asgharian, B.; Anjilvel, S. (1994) Inertia! and gravitational deposition of particles in a square cross section
16 bifurcating airway. Aerosol Sci. Technol. 20: 177-193.
17 Asgharian, B.; Wood, R.; Schlesinger, R. B. (1995) Empirical modeling of particle deposition in the alveolar region
18 of the lungs: abasis for interspecies extrapolation. Fundam. Appl. Toxicol. 27: 232-238.
19 Asgharian, B.; Miller, F. J.; Subramaniam, R. P. (1999) Dosimetry software to predict particle deposition in
20 humans and rats. CUT Activities 19(3). Available:
21 http://www.ciit.org/ACT99/ACTIVITIESMAR99/mar99.html [19 December, 2001].
22 Asgharian, B.; Hofmann, W.; Miller, F. J. (2001) Mucociliary clearance of insoluble particles from the
23 tracheobronchial airways of the human lung. Aerosol Sci. 32: 817-832.
24 Balashazy, I.; Hofmann, W. (1993a) Particle deposition in airway bifurcations-I: inspiratory flow. J. Aerosol Sci.
25 24:745-772.
26 Balashazy, I.; Hofmann, W. (1993b) Particle deposition in airway bifurcations-II: expiratory flow. J. Aerosol Sci.
27 24:773-786.
28 Balashazy, I.; Hofmann, W. (1995) Deposition of aerosols in asymmetric airway bifurcations. J. Aerosol Sci.
29 26:273-292.
30 Balashazy, I.; Hofmann, W. (2001) Fluid dynamics and related particle deposition patterns in human airway
31 bifurcations. In: Martonen, T. B., ed., Medical Applications of Computer Modeling: The Respiratory
32 System. WIT Press, Southampton, Boston, p. 84-108.
3 3 Balashazy, I.; Hofmann, W.; Heistracher, T. (1999) Computation of local enhancement factors for the
34 quantification of particle deposition patterns in airway bifurcations. J. Aerosol Sci. 30: 185-203.
35 Becquemin, M. H.; Swift, D.L.; Bouchikhi, A.; Roy, M.; Teillac, A. (1991) Particle deposition and resistance in the
36 noses of adults and children. Eur. Resp. J. 4: 694-702.
37 Bell, K. A.; Friedlander, S. K. (1973) Aerosol deposition in models of a human lung bifurcation. Staub Reinhalt.
38 Luft 33: 178-182.
39 Bennett, W. D.; Ilowite, J. S. (1989) Dual pathway clearance of 99mTc-DTPA from the bronchial mucosa. Am. Rev.
40 Respir. Dis. 139: 1132-1138.
41 Bennett, W. D.; Zeman, K. L. (1998) Deposition of fine particles in children spontaneously breathing at rest.
42 Inhalation Toxicol. 10: 831-842.
43 Bennett, W. D.; Zeman, K. L.; Kim, C. (1996) Variability of fine particle deposition in healthy adults: effect of age
44 and gender. Am. J. Respir. Crit. Care Med. 153: 1641-1647.
45 Bennett, W. D.; Zeman, K. L.; Kang, C. W.; Schechter, M. S. (1997a) Extrathoracic deposition of inhaled, coarse
46 particles (4.5 um) in children vs. adults. In: Cherry, N.; Ogden, T., eds. Inhaled particles VIII: proceedings
47 of an international symposium on inhaled particles organised by the British Occupational Hygiene Society;
48 August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 497-502.
49 Bennett, W. D.; Zeman, K. L.; Kim, C.; Mascarella, J. (1997b) Enhanced deposition of fine particles in COPD
50 patients spontaneously breathing at rest. Inhalation Toxicol. 9: 1-14.
51 Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1998) Bronchial
52 airway deposition and retention of particles in inhaled boluses: effect of anatomic dead space. J. Appl.
53 Physiol. 85: 685-694.
54 Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1999) Regional
5 5 deposition and retention of particles in shallow, inhaled boluses: effect of lung volume. J. Appl. Physiol.
56 86: 168-173.
June 2003 6-107 DRAFT-DO NOT QUOTE OR CITE
-------
1 Benson, J. M; Henderson, R. F.; McClellan, R. O.; Hanson, R. L.; Rebar, A. H. (1986) Comparative acute toxicity
2 of four nickel compounds to F344 rat lung. Fundam. Appl. Toxicol. 7: 340-347.
3 Brauer, M.; Avila-Casado, C.; Fortoul, T. L; Vedal, S.; Stevens, B.; Churg, A. (2001) Air pollution and retained
4 particles in the lung. Environ. Health Perspect. 109: 1039-1043.
5 Broday, D. M.; Georgopoulos, P. G. (2001) Growth and deposition of hygroscopic paniculate matter in the human
6 lungs. Aerosol Sci. Technol. 34: 144-159.
7 Brody, A. R.; Roe, M. W. (1983) Deposition pattern of inorganic particles at the alveolar level in the lungs of rats
8 and mice. Am. Rev. Respir. Dis. 128: 724-729.
9 Brody, A. R.; Hill, L. H.; Adkins, B., Jr.; O'Connor, R. W. (1981) Chrysotile asbestos inhalation in rats: deposition
10 pattern and reaction of alveolar epithelium and pulmonary macrophages. Am. Rev. Respir. Dis. 123:
11 670-679.
12 Brown, J. S.; Kirby, Z. L.; Bennett, W. D. (2001) Regional deposition of coarse particles and ventilation
13 distribution in healthy subjects and patients with cystic fibrosis. J. Aerosol Med. 14: 443-454.
14 Brown, J. S.; Zeman, K. L.; Bennett, W. D. (2002) Ultrafine particle deposition and clearance in the healthy and
15 obstructed lung. Am. J. Respir. Crit. Care Med. 166: 1240-1247.
16 California Air Resources Board. (1987) Ambient air quality standard for ozone: health and welfare effects.
17 Sacramento, CA: California Air Resources Board.
18 Camner, P.; Anderson, M.; Philipson, K.; Bailey, A.; Hashish, A.; Jarvis, N.; Bailey, M.; Svartengren, M. (1997)
19 Human bronchiolar deposition and retention of 6-, 8-, and 10-^ particles. Exp. Lung Res. 23: 517-535.
20 Chang, Y. H.; Yu, C. P. (1999) A model of ventilation distribution in the human lung. Aerosol Sci. Technol.
21 30:309-319.
22 Cheng, Y.-S.; Smith, S. M.; Yeh, H.-C; Kim, D.-B.; Cheng, K.-H.; Swift, D. L. (1995) Deposition of ultrafine
23 aerosols and thoron progeny in replicas of nasal airways of young children. Aerosol. Sci. Technol.
24 23:541-552.
25 Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Guilmette, R. A.; Simpson, S. Q.; Yang, Y.-H.; Swift, D. L. (1996)
26 In vivo measurements of nasal airway dimensions and ultrafine aerosol deposition in the human nasal and
27 oral airways. J. Aerosol Sci. 27: 785-801.
28 Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Swift, D. L. (1997) An experimental method for measuring aerosol
29 deposition efficiency in the human oral airway. Am. Ind. Hyg. Assoc. J. 58: 207-213.
30 Churg, A.; Brauer, M. (1997) Human lung parenchyma retains PM2 5. Am. J. Respir. Crit. Care Med.
31 155:2109-2111.
32 Churg, A.; Vedal, S. (1996) Carinal and tubular airway particle concentrations in the large airways of non-smokers
33 in the general population: evidence for high particle concentration at airway carinas. Occup. Environ. Med.
34 53: 553-558.
35 Cohen, B. S.; Harley, N. H.; Schlesinger, R. B.; Lippmann, M. (1988) Nonuniform particle deposition on
36 tracheobronchial airways: implications for lung dosimetry. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.;
37 Fisher, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung
38 dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32 (suppl. 1): 1045-1053.
39 Cohen, B. S.; Sussman, R. G.; Lippmann, M. (1990) Ultrafine particle deposition in a human tracheobronchial cast.
40 Aerosol Sci. Technol. 12: 1082-1091.
41 Cohen, B. S.; Xiong, J. Q.; Fang, C.-P.; Li, W. (1998) Deposition of charged particles on lung airways. Health
42 Phys. 74: 554-560.
43 Comer, J. K.; Kleinstreuer, C.; Hyun, S.; Kim, C. S. (2000) Aerosol transport and deposition in sequentially
44 bifurcating airways. J. Biomech. Eng. 122: 152-158.
45 Comer, J. K.; Kleinstreuer, C.; Zhang, Z. (2001a) Flow structures and particle deposition patterns in double
46 bifurcation airway models. Part 1. Air flow fields. J. Fluid Mech. 435: 24-54.
47 Comer, J. K.; Kleinstreuer, C.; Kim, C. S. (200Ib) Flow structures and particle deposition patterns in double
48 bifurcation airway models. Part 2. Aerosol transport and deposition. J. Fluid Mech. 435: 55-80.
49 Crapo, J. D.; Barry, B. E.; Gehr, P.; Bachofen, M.; Weibel, E. R. (1982) Cell characteristics of the normal human
50 lung. Am. Rev. Respir. Dis. 125: 740-745.
51 Crystal, R. G., West, J. B.; Barnes, P. J.; Weibel, E. R., eds. (1997) The lung: scientific foundations. Volume 1.
52 Section III: major components. 2nd ed. Philadelphia, PA: Lippincott-Raven; chapters 30-67, pp. 445-991.
53 Cuddihy, R. G. (1984) Mathematical models for predicting clearance of inhaled radioactive materials. In: Smith,
54 H.; Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a meeting jointly
5 5 organized by the Commission of the European Communities and the National Radiological Protection
56 Board; March; Oxford, United Kingdom. Luxembourg: Commission of the European Communities; pp.
57 167-179; report no. EUR 9384 EN.
June 2003 6-108 DRAFT-DO NOT QUOTE OR CITE
-------
1 Cuddihy, R. G.; Yeh, H. C. (1988) Respiratory tract clearance of particles and substances dissociated from
2 particles. In: Mohr, U.; Dungworth, D.; Kimmerle, G.; Lewkowski, I; McClellan, R.; Stober, W., eds.
3 Inhalation toxicology: the design and interpretation of inhalation studies and their use in risk assessment.
4 New York, NY: Springer-Verlag; pp. 169-193.
5 Darquenne, C. (2001) A realistic two-dimensional model of aerosol transport and deposition in the alveolar zone of
6 the human lung. J. Aerosol Sci. 32: 1161-1174.
7 Dorries, A. M; Valberg, P. A. (1992) Heterogeneity of phagocytosis for inhaled versus instilled material. Am. Rev.
8 Respir.Dis. 146:831-837.
9 Driscoll, K. E. (1995) The toxicology of crystalline silica studied in vitro. Appl. Occup. Environ. Hyg.
10 10: 1118-1125.
11 Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
12 Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
13 limitations. Toxicol. Sci. 55: 24-35.
14 Falk, R.; Philipson, K.; Svartengren, M.; Jarvis, N.; Bailey, M.; Camner, P. (1997) Clearance of particles from
15 small ciliated airways. Exp. Lung Res. 23: 495-515.
16 Falk, R.; Philipson, K.; Svartengren, M.; Bergmann, R.; Hofmann, W.; Jarvis, N.; Bailey, M.; Camner, P. (1999)
17 Assessment of long-term bronchiolar clearance of particles from measurements of lung retention and
18 theoretical estimates of regional deposition. Exp. Lung Res. 25: 495-516.
19 Ferin, J. (1977) Effect of particle content of lung on clearance pathways. In: Sanders, C. L.; Schneider, R. P.;
20 Dagle, G. E.; Ragan, H. A., eds. Pulmonary macrophages and epithelial cells: proceedings of the sixteenth
21 annual Hanford biology symposium; September 1976; Richland, WA. Oak Ridge, TN: Energy Research and
22 Development Administration; pp. 414-423. Available from: NTIS, Springfield, VA; CONF-760927.
23 (ERDA symposium series 43).
24 Ferin, J.; Feldstein, M. L. (1978) Pulmonary clearance and hilar lymph node content in rats after particle exposure.
25 Environ. Res. 16: 342-352.
26 Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
27 Respir. Cell Mol. Biol. 6: 535-542.
28 Findeisen, W. (1935) Uber das Absetzen kleiner, in der Luft suspendierter Teilchen in der menschlichen Lunge bei
29 der Atmung [The deposition of small airborne particles in the human lung during respiration]. Pfluegers
30 Arch. Gesamte Physiol. Menschen Tiere 236: 367-379.
31 Foster, W. M.; Langenback, E.; Bergofsky, E. H. (1980) Measurement of trachea! and bronchial mucus velocities in
32 man: relation to lung clearance. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 48: 965-971.
33 Frampton, M. W.; Chalupa, D.; Morrow, P. E.; Gibb, F. R.; Oberdorster, G.; Speers, D. M.; Zareba, W.; Utell, M.
34 J. (2000) Deposition and effects of inhaled ultrafine carbon particles in healthy subjects at rest. Presented at:
3 5 PM2000: paniculate matter and health—the scientific basis for regulatory decision-making, specialty
36 conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
37 Fry, F. A.; Black, A. (1973) Regional deposition and clearance of particles in the human nose. J. Aerosol Sci.
38 4: 113-124.
39 Gehr, P.; Bachofen, M.; Weibel, E. R. (1978) The normal human lung: ultrastructure and morphometric estimation
40 of diffusion capacity. Respir. Physiol. 32: 121-140.
41 Gehr, P.; Schurch, S.; Berthaiume, Y.; Im Hof, V.; Geiser, M. (1990) Particle retention in airways by surfactant.
42 J. Aerosol Med. 3: 27-43.
43 Gehr, P.; Im Hof, V.; Geiser, M.; Schurch, S. (1991) The fate of particles deposited in the intrapulmonary
44 conducting airways. J. Aerosol Med. 4: 349-362.
45 Gradoii, L.; Orlicki, D. (1990) Deposition and inhaled aerosol particles in a generation of the tracheobronchial tree.
46 J. Aerosol Sci. 21:3-19.
47 Green, F. H. Y. (2000) Pulmonary responses to inhaled poorly soluble paniculate in the human. In: Gardner, D. E.,
48 ed. ILSI Risk Science Institute Workshop: The Relevance of the Rat Lung Response to Particle Overload for
49 Human Risk Assessment; March, 1998. Inhalation Toxicol. 12: 59-95.
50 Groth, M. L.; Macri, K.; Foster, W. M. (1997) Cough and mucociliary transport of airway paniculate in chronic
51 obstructive lung disease. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an
52 international symposium on inhaled particles organised by the British Occupational Hygiene Society;
53 August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 515-521.
54 Guilmette, R. A.; Cheng, Y. S.; Griffith, W. C. (1997) Characterising the variability in adult human nasal airway
55 dimensions. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international
56 symposium on inhaled particles organised by the British Occupational Hygiene Society; August 1996;
57 Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 491-496.
June 2003 6-109 DRAFT-DO NOT QUOTE OR CITE
-------
1 Harmsen, A. G.; Muggenburg, B. A.; Snipes, M. B.; Bice, D. E. (1985) The role of macrophages in particle
2 translocation from lungs to lymph nodes. Science (Washington, DC) 230: 1277-1280.
3 Heistracher, T.; Hofmann, W. (1995) Physiologically realistic models of bronchial airway bifurcation. J. Aerosol
4 Sci. 26: 497-509.
5 Heistracher, T.; Hofmann, W. (1997) Flow and deposition patterns in successive airway bifurcations. In: Cherry,
6 N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international symposium on inhaled particles
7 organised by the British Occupational Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg.
8 41(suppl.): 537-542.
9 Henderson, R. F.; Driscoll, K. E.; Harkema, J. R.; Lindenschmidt, R. C.; Chang, I.-Y.; Maples, K. R.; Barr, E. B.
10 (1995) A comparison of the inflammatory response of the lung to inhaled versus instilled particles inF344
11 rats. Fundam. Appl. Toxicol. 24: 183-197.
12 Henshaw, D. L.; Fews, A. P. (1984) The microdistribution of alpha emitting particles in the human lung.
13 In: Smith, H.; Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a
14 meeting jointly organized by the Commission of the European Communities and the National Radiological
15 Protection Board; March; Oxford, United Kingdom. Luxembourg: Commission of the European
16 Communities; pp. 199-218; report no. EUR 9384 EN.
17 Heyder, J.; Rudolf, G. (1977) Deposition of aerosol particles in the human nose. In: Walton, W. H.; McGovern, B.,
18 eds. Inhaled particles IV: proceedings of an international symposium, part 1; September 1975; Edinburgh,
19 United Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 107-126.
20 Killer, F. C. (1991) Health implications of hygroscopic particle growth in the human respiratory tract. J. Aerosol
21 Med. 4:1-23.
22 Hinds, W. C. (1999) Aerosol technology: properties, behavior, and measurement of airborne particles. 2nd ed.
23 New York, NY: John Wiley & Sons.
24 Hofmann, W.; Bergmann, R. (1998) Predictions of particle deposition patterns in human and rat airways. Inhalation
25 Toxicol. 10: 557-583.
26 Hofmann, W.; Martonen, T. B.; Graham, R. C. (1989a) Predicted deposition of nonhygroscopic aerosols in the
27 human lung as a function of subject age. J. Aerosol Med. 2: 49-68.
28 Hofmann, W.; Koblinger, L.; Martonen, T. B. (1989b) Structural differences between human and rat lungs:
29 implications for Monte Carlo modeling of aerosol deposition. In: Mahaffey, J. A., ed. 26th Hanford life
30 sciences symposium, modeling for scaling to man: biology, dosimetry, and response. Health Phys.
31 57(suppl. 1): 41-47.
32 Hofmann, W.; Balashazy, L; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns
33 in bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
34 Hofmann, W.; Bergmann, R.; Koblinger, L. (1999) Characterization of local particle deposition patterns in human
3 5 and rat lungs by different morphometric parameters. J. Aerosol Sci. 30: 651-667.
36 Hofmann, W.; Asgharian, B.; Bergmann, R.; Anjilvel, S.; Miller, F. J. (2000) The effect of heterogeneity of lung
37 structure on particle deposition in the rat lung. Toxicol. Sci. 53: 430-437.
38 Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
39 Inhalation Toxicol. 10: 121-130.
40 Huchon, G. J.; Montgomery, A. B.; Lipavsky, A.; Hoeffel, J. M.; Murray, J. F. (1987) Respiratory clearance of
41 aerosolized radioactive solutes of varying molecular weight. J. Nucl. Med. 28: 894-902.
42 International Commission on Radiological Protection. (1960) Report of Committee II on permissible dose for
43 internal radiation (1959). Health Phys. 3.
44 International Commission on Radiological Protection. (1979) Limits for intakes of radionuclides by workers.
45 Oxford, United Kingdom: Pergamon Press; ICRP publication 30, part 1.
46 International Commission on Radiological Protection. (1994) Human respiratory tract model for radiological
47 protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
48 United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP: v. 24, pp. 1-482.).
49 Jaques, P. A.; Kim, C. S. (2000) Measurement of total lung deposition of inhaled ultrafine particles in healthy men
50 and women. Inhalation Toxicol. 12:715-731.
51 John, J.; Wollmer, P.; Dahlback, M.; Luts, A.; Jonson, B. (1994) Tidal volume and alveolar clearance of insoluble
52 particles. J. Appl. Physiol. 76: 584-588.
53 Katz, I. M. (2001) Computer modeling of fluid dynamics and particle motion in the larynx and trachea. In:
54 Martonen, T. B., ed. Medical Applications of Computer Modeling: The Respiratory System. Boston, MA:
55 WIT Press, pp. 47-63.
56 Katz, I. M.; Martonen, T. B. (1996) Three-dimensional fluid particle trajectories in the human larynx and trachea.
57 J. Aerosol Med. 9:513-20.
June 2003 6-110 DRAFT-DO NOT QUOTE OR CITE
-------
1 Katz, I. M, Martonen, T. B., and Flaa, W. (1997) Three-dimensional computational study of inspiratory aerosol
2 flow through the larynx: the effect of glottal aperture modulation. J. Aerosol Sci. 28: 1073-83.
3 Katz, I. M.; Davis, B. M.; Martonen, T. B. (1999) Anumerical study of particle motion within the human larynx
4 and trachea. J. Aerosol Sci. 30: 173-183.
5 Kaye, S. R.; Phillips, C. G.; Winlove, C. P. (2000) Measurement of non-uniform aerosol deposition patterns in the
6 conducting airways of the porcine lung. J. Aerosol Sci. 31: 849-866.
7 Kesavanathan, J.; Swift, D. L. (1998) Human nasal passage particle deposition: the effect of particle size, flow rate,
8 and anatomical factors. Aerosol Sci. Technol. 28: 457-463.
9 Keyhani, K.; Scherer, P. W.; Mozell, M. M. (1995) Numerical simulation of airflow in the human nasal cavity.
10 J. Biomech. Eng. 117: 429-41.
11 Kim, C. S. (2000) Methods of calculating lung delivery and deposition of aerosol particles. Respir. Care
12 45:695-711.
13 Kim, C. S.; Fisher, D. M. (1999) Deposition characteristics of aerosol particles in sequentially bifurcating airway
14 models. Aerosol Sci. Technol. 31:198-220.
15 Kim, C. S.; Garcia, L. (1991) Particle deposition in cyclic bifurcating tube flow. Aerosol Sci. Technol. 14: 302-315.
16 Kim, C. S.; Hu, S. C. (1998) Regional deposition of inhaled particles in human lungs: comparison between men and
17 women. J. Appl. Physiol. 84: 1834-1844.
18 Kim, C. S.; Iglesias, A. J. (1989) Deposition of inhaled particles in bifurcating airway models: I. inspiratory
19 deposition. J. Aerosol Med. 2: 1-14.
20 Kim, C. S.; Jaques, P. A. (2000) Respiratory dose of inhaled ultrafine particles in healthy adults. Phil. Trans. Roy.
21 Soc. London A 358: 2693-2705.
22 Kim, C. S.; Kang, T. C. (1997) Comparative measurement of lung deposition of inhaled fine particles in normal
23 subjects and patients with obstructive airway disease. Am. J. Respir. Crit. Care Med. 155: 899-905.
24 Kim, C. S.; Hu, S. C.; DeWitt, P.; Gerrity, T. R. (1996) Assessment of regional deposition of inhaled particles in
25 human lungs by serial bolus delivery method. J. Appl. Physiol. 81: 2203-2213.
26 Kim, C. S.; Hu, S. C.; Jaques, P.; Ding, J.; DeWitt, P. (2000) Measurement of regional deposition dose of inhaled
27 particles in the size range of 0.04-5 n in human lungs. Presented at: PM2000: paniculate matter and
28 health~the scientific basis for regulatory decision-making, specialty conference & exhibition; January;
29 Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
30 Kimbell, J. S. (2001) Computational fluid dynamics of the extrathoracic airways. In: Martonen, T. B., ed. Medical
31 Applications of Computer Modeling: The Respiratory System. Boston, MA: WIT Press; pp. 11-45.
32 Kinsara, A. A.; Tompson, R. V.; Loyalka, S. K. (1993) Computational flow and aerosol concentration profiles in
33 lung bifurcations. Health Phys. 64: 13-22.
34 Koblinger, L.; Hofmann, W. (1985) Analysis of human lung morphometric data for stochastic aerosol deposition
35 calculations. Phys. Med. Biol. 30: 541-556.
36 Koblinger, L.; Hofmann, W. (1988) Stochastic morphological model of the rat lung. Anat. Rec. 221: 533-539.
37 Koch, W.; Stober, W. (2001) A simple pulmonary retention model accounting for dissolution and
3 8 macrophage-mediated removal of deposited polydisperse particles. Inhalation Toxicol. 13: 129-148.
39 Kodavanti, U. P.; Schladweiler, M. C. J.; Ledbetter, A. D.; Hauser, R.; Christiani, D. C.; Samet, J. M.; McGee, J.;
40 Richards, J. H.; Costa, D. L. (2002) Pulmonary and systemic effects of zinc-contining emission particles in
41 three rat strains: multiple exposure scenarios. Toxicol. Sci. 70: 73-85.
42 Kohlhaufl, M.; Brand, P.; Scheuch, G.; Meyer, T. S.; Schulz, H.; Haussinger, K.; Heyder, J. (1999) Increased fine
43 particle deposition in women with asymptomatic nonspecific airway hyperresponsiveness. Am. J. Respir.
44 Crit. Care Med. 159: 902-906.
45 Kreyling, W. G. (1992) Intracellular particle dissolution in alveolar macrophages. Environ. Health Perspect.
46 97:121-126.
47 Kreyling, W. G.; Blanchard, J. D.; Godleski, J. J.; Haeussermann, S.; Heyder, J.; Hutzler, P.; Schulz, H.; Sweeney,
48 T. D.; Takenaka, S.; Ziesenis, A. (1999) Anatomic localization of 24-and 96-h particle retention in canine
49 airways. J. Appl. Physiol. 87: 269-284.
50 Kreyling, W. G.; Semmler, M.; Erbe, F.; Mayer, P.; Takenaka, S.; Schulz, H.; Oberdorster, G.; Ziesenis, A. (2002)
51 Translocation of ultrafine insoluble iridium particles from lung epithelium to extrapulmonary organs is size
52 dependent but very low. J. Toxicol. Environ. Health Part A 65: 1513-1530.
53 Kuempel, E. D. (2000) Comparison of human and rodent lung dosimetry models for particle clearance and
54 retention. Drug Chem. Toxicol. 23: 203-222.
55 Kuempel, E. D.; O'Flaherty, E. J.; Stayner, L. T.; Smith, R. J.; Green, F. H. Y.; Vallyathan, V. (2001a) A
5 6 biomathematical model of particle clearance and retention in the lungs of coal miners. I. Model
57 development. Regul. Toxicol. Pharmacol. 34: 69-87.
June 2003 6-111 DRAFT-DO NOT QUOTE OR CITE
-------
1 Kuempel, E. D.; Tran, C.-L.; Smith, R. J.; Bailer, A. J. (200Ib) A biomathematical model of particle clearance and
2 retention in the lungs of coal miners. II. Evaluation of variability and uncertainty. Regul. Toxicol.
3 Pharmacol. 34: 88-101.
4 LaBelle, C. W.; Brieger, H. (1961) Patterns and mechanisms in the elimination of dust from the lung. In: Davies,
5 C. N., ed. Inhaled particles and vapours: proceedings of an international symposium; March-April 1960;
6 Oxford, United Kingdom. New York, NY: Pergamon Press; pp. 356-368.
7 Landahl, H. D. (1950) On the removal of air-borne droplets by the human respiratory tract: I. the lung. Bull. Math.
8 Biophys. 12: 43-56.
9 Lay, J. C.; Berry, C. R.; Kim, C. S.; Bennett, W. D. (1995) Retention of insoluble particles after local intrabronchial
10 deposition in dogs. J. Appl. Physiol. 79: 1921-1929.
11 Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular
12 distribution of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol.
13 18:687-695.
14 Lazaridis, M.; Broday, D. M.; Hov, 0.; Georgopoulos, P. G. (2001) Integrated exposure and dose modeling and
15 analysis system. 3. Deposition of inhaled particles in the human respiratory tract. Environ. Sci. Technol.
16 35: 3727-3734.
17 Lee, J. W.; Goo, J. H.; Chung, M. K. (1996) Characteristics of inertia! deposition in a double bifurcation. J. Aerosol
18 Sci. 27: 119-138.
19 Lehnert, B. E.; Morrow, P. E. (1985) Association of 59iron oxide with alveolar macrophages during alveolar
20 clearance. Exp. Lung Res. 9: 1-16.
21 Lehnert, B. E.; Valdez, Y. E.; Bomalaski, S. H. (1988) Analyses of particles in the lung free cell, tracheobronchial
22 lymph nodal, and pleural space compartments following their deposition in the lung as related to lung
23 clearance mechanisms. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fisher, D. R., eds. Inhaled particles
24 VI: proceedings of an international symposium and workshop on lung dosimetry; September 1985;
25 Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 125-140.
26 Leikauf, G.; Yeates, D. B.; Wales, K. A.; Spektor, D.; Albert, R. E.; Lippmann, M. (1981) Effects of sulfuric acid
27 aerosol on respiratory mechanics and mucociliary particle clearance in healthy nonsmoking adults. Am. Ind.
28 Hyg. Assoc. J. 42: 273-282.
29 Leikauf, G. D.; Spektor, D. M.; Albert, R. E.; Lippmann, M. (1984) Dose-dependent effects of submicrometer
30 sulfuric acid aerosol on particle clearance from ciliated human lung airways. Am. Ind. Hyg. Assoc. J.
31 45:285-292.
32 Lennon, S.; Zhang, Z.; Lessmann, R.; Webster, S. (1998) Experiments on particle deposition in the human upper
33 respiratory system. Aerosol Sci. Technol. 28: 464-474.
34 Leong, B. K. J.; Coombs, J. K.; Sabaitis, C. P.; Rop, D. A.; Aaron, C. S. (1998) Quantitative morphometric analysis
35 of pulmonary deposition of aerosol particles inhaled via intratracheal nebulization, intratracheal instillation
36 or nose-only inhalation in rats. J. Appl. Toxicol. 18: 149-160.
37 Li, A.; Ahmadi, G. (1995) Computer simulation of particle deposition in the upper tracheobronchial tree. Aerosol
38 Sci. Tech. 23:201-223.
39 Lundborg, M.; Lind, B.; Camner, P. (1984) Ability of rabbit alveolar macrophages to dissolve metals. Exp. Lung
40 Res. 7:11-22.
41 Lundborg, M.; Eklund, A.; Lind, B.; Camner, P. (1985) Dissolution of metals by human and rabbit alveolar
42 macrophages. Br. J. Ind. Med. 42: 642-645.
43 Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
44 macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
45 Pharmacol. 153: 28-38.
46 Martonen, T. B.; Lowe, J. (1983) Assessment of aerosol deposition patterns in human respiratory tract casts.
47 In: Marple, V. A.; Liu, B. Y. H., eds. Aerosols in the Mining and Industrial Work Environments Vol. 1
48 Fundamentals and Status, p. 151-164. Ann Arbor, MI: Ann Arbor Science Publishers.
49 Martonen, T. B.; Zhang, Z.; Lessmann, R. (1993) Fluid dynamics of the human larynx and upper tracheobronchial
50 airways. Aerosol Sci. Technol. 19: 133-156.
51 Martonen, T. B.; Yang, Y.; Xue, Z. Q. (1994a) Influences of cartilagious rings on tracheobronchial fluid dynamics.
52 Inhalation Toxicol. 6: 185-203.
53 Martonen, T. B.; Yang, Y.; Xue, Z. Q. (1994b) Effects of carinal ridge shapes on lung airstreams. Aerosol Sci.
54 Technol. 21: 119-136.
55 Martonen, T. B.; Schroeter, J. D.; Hwang, D.; Fleming, J. S.; Conway, J. H. (2000) Human lung morphology
56 models for particle deposition studies. In: Grant, L. D., ed. PM2000: paniculate matter and health.
57 Inhalation Toxicol. 12(suppl. 4): 109-121.
June 2003 6-112 DRAFT-DO NOT QUOTE OR CITE
-------
1 Martonen, T. B.; Zhang, Z; Yu, G.; Musante, C. J. (2001a) Three-dimensional computer modeling of the human
2 upper respiratory tract. Cell Biochem. Biophys. 35: 255-261.
3 Martonen, T. B.; Guan, X.; Schreck, R. M. (200Ib) Fluid dynamics in airway bifurcations: I. Primary flows.
4 Inhalation Toxicol. 13:261-79.
5 Martonen, T. B.; Guan, X.; Schreck, R. M. (2001c) Fluid dynamics in airway bifurcations: II. Secondary currents.
6 Inhalation Toxicol. 13:281-9.
7 Martonen, T. B.; Guan, X.; Schreck, R. M. (2001d) Fluid dynamics in airway bifurcations: III. Localized flow
8 conditions. Inhalation Toxicol. 13: 291-305.
9 Matsui, H.; Randell, S. H.; Peretti, S. W.; Davis, C. W.; Boucher, R. C. (1998) Coordinated clearance of periciliary
10 liquid and mucus from airway surfaces. J. Clin. Invest. 102: 1125-1131.
11 Mauderly, J. L. (1979) Effect of age on pulmonary structure and function of immature and adult animals and man.
12 Fed. Proc. 38: 173-177.
13 Medinsky, M. A.;Kampcik, S. J. (1985) Pulmonary retention of [14C]benzo/o/pyrene in rats as influenced by the
14 amount instilled. Toxicology 35: 327-336.
15 Menache, M. G.; Miller, F. J.; Raabe, O. G. (1995) Particle inhalability curves for humans and small laboratory
16 animals. Ann. Occup. Hyg. 39: 317-328.
17 Mercer, T. T. (1967) On the role of particle size in the dissolution of lung burdens. Health Phys. 13: 1211-1221.
18 Mercer, R. R.; Russell, M. L.; Roggli, V. L.; Crapo, J. D. (1994) Cell number and distribution in human and rat
19 airways. Am. J. Respir. Cell Mol. Biol. 10: 613-624.
20 Miller, F. J.; Anjilvel, S.; Menache, M. G.; Asgharian, B.; Gerrity, T. R.. (1995) Dosimetric issues relating to
21 paniculate toxicity. Inhalation Toxicol. 7: 615-632.
22 Morrison, D.; Skwarski, D.; Millar, A. M.; Adams, W.; MacNee, W. (1998) A comparison of three methods of
23 measuring 99mTc-DTPA lung clearance and their repeatability. Eur. Respir. J. 11: 1141-1146.
24 Morrow, P. E. (1973) Alveolar clearance of aerosols. Arch. Intern. Med. 131: 101-108.
25 Morrow, P. E. (1977) Clearance kinetics of inhaled particles. In: Brain, J. D.; Proctor, D. F.; Reid, L. M., eds.
26 Respiratory defense mechanisms (in two parts), part II. New York, NY: Marcel Dekker, Inc.; pp. 491-543.
27 (Lenfant, C., ed. Lung biology in health and disease: v. 5).
28 Morrow, P. E. (1986) Factors determining hygroscopic aerosol deposition in airways. Physiol. Rev. 66: 330-376.
29 Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
30 10:369-384.
31 Morrow, P. E. (1994) Mechanisms and significance of "particle overload." In: Mohr, U.; Dungworth, D. L.;
32 Mauderly, J. L.; Oberdorster, G., eds. Toxic and carcinogenic effects of solid particles in the respiratory
33 tract: [proceedings of the 4th international inhalation symposium]; March 1993; Hannover, Germany.
34 Washington, DC: International Life Sciences Institute Press; pp. 17-25.
35 Morrow, P. E.; Yu, C. P. (1993) Models of aerosol behavior in airways and alveoli. In: Moren, F.; Dolovich, M. B.;
36 Newhouse, M. T.; Newman, S. P., eds. Aerosols in medicine: principles, diagnosis and therapy. 2nd rev. ed.
37 Amsterdam, The Netherlands: Elsevier; pp. 157-193.
38 Morrow, P. E.; Gibb, F. R.; Gazioglu, K. M. (1967) A study of paniculate clearance from the human lungs.
39 Am. Rev. Respir. Dis. 96: 1209-1221.
40 Musante, C. J.; Martonen, T. B. (1999) Predicted deposition patterns of ambient paniculate air pollutants in
41 children's lungs under resting conditions. In: Proceedings of the third colloquium on paniculate air pollution
42 and human health; June; Durham, NC. Irvine, CA: University of California, Air Pollution Health Effects
43 Laboratory, p. 7-15 - 7-20.
44 Musante, C. J.; Martonen, T. B. (2000a) Computer simulations of particle deposition in the developing human lung.
45 J. Air Waste Manage. Assoc. 50: 1426-1432.
46 Musante, C. J.; Martonen, T. B. (2000b) An extrapolation model to aid in lexicological studies of paniculate air
47 pollutants. Presented at: PM2000: paniculate matter and health—the scientific basis for regulatory
48 decision-making, specialty conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste
49 Management Association.
50 Musante, C. J.; Martonen, T. B. (2000c) Paniculate matter deposition in the lungs of children and adults:
51 predictions of an age-dependent computer model. Presented at: PM2000: paniculate matter and health — the
52 scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
53 Pittsburgh, PA: Air & Waste Management Association.
54 National Council on Radiation Protection and Measurements (NCRP). (1997) Deposition, retention and dosimetry
55 of inhaled radioactive substances. Bethesda, MD: National Council on Radiation Protection and
56 Measurements; report no. 125.
June 2003 6-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 National Radiologic Protection Board (NRPB). (1994) National Radiologic Protection Board Report R287,
2 implementation of the ICRP publication 66 respiratory tract model: LUDEP 2.0 (LUng Dose Evaluation
3 Program). Richland, WA: ACJ & Associates, Inc. Software information available: www.acj-associates.com
4 (13 June 2003).
5 National Research Council. (2001) Research priorities for airborne paniculate matter. III. Early research progress.
6 Washington, DC: National Academy Press. Available: http://www.nap.edu/books/0309073375/html/ (4 June
7 2003).
8 Naumann, B. D.; Schlesinger, R. B. (1986) Assessment of early alveolar particle clearance and macrophage
9 function following an acute inhalation of sulfuric acid mist. Exp. Lung Res. 11: 13-33.
10 Nemmar, A.; Vanbilloen, H.; Hoylaerts, M. F.; Hoet, P. H. M.; Verbruggen, A.; Nemery, B. (2001) Passage of
11 intratracheally instilled ultrafine particles from the lung into the systemic circulation in hamster. Am. J.
12 Respir. Crit. Care Med. 164: 1665-1668.
13 Nemmar, A.; Hoet, H. M.; Vanquickenborne, B.; Dinsdale, D.; Thomeer, M.; Hoylaerts, M. F.; Vanbilloen, H.;
14 Mortelmans, L.; Nemery, B. (2002) Passage of inhaled particles into the blood circulation in humans.
15 Circulation 105: 411-414.
16 Nikula, K. J.; Avila, K. J.; Griffith, W. C.; Mauderly, J. L. (1997) Lung tissue responses and sites of particle
17 retention differ between rats and cynomolgus monkeys exposed chronically to diesel exhaust and coal dust.
18 Fundam. Appl. Toxicol. 37: 37-53.
19 Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
20 paniculate material in rat and human lungs. Presented at: PM2000: paniculate matter and health - the
21 scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
22 Pittsburgh, PA: Air & Waste Management Association.
23 Niinimaa, V.; Cole, P.; Mintz, S.; Shephard, R. J. (1981) Oronasal distribution of respiratory airflow. Respir.
24 Physiol. 43:69-75.
25 Noone, P. G.; Bennett, W. D.; Regnis, J. A.; Zeman, K. L.; Carson, J. L.; King, M.; Boucher, R. C.; Knowles, M.
26 R. (1999) Effect of aerosolized uridine-5'-triphosphate on airway clearance with cough in patients with
27 primary ciliary dyskinesia. Am. J. Respir. Crit. Care Med.: 160: 144-149.
28 Oberdorster, G. (1993) Lung dosimetry: pulmonary clearance of inhaled particles. Aerosol Sci. Technol.
29 18:279-289.
30 Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
31 lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
32 Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer
33 particles for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
34 Oldham, M. J.; Mannix, R. C.; Phalen, R. F. (1997) Deposition of monodisperse particles in hollow models
35 representing adult and child-size tracheobronchial airways. Health Phys. 72: 827-834.
36 Oldham, M. J.; Phalen, R. F.; Heistracher, T. (2000) Computational fluid dynamic predictions and experimental
37 results for particle deposition in an airway model. Aerosol Sci. Technol. 32: 61-71.
3 8 Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
39 Fundam. Appl. Toxicol. 40: 220-227.
40 Passali, D.; Bianchini Ciampoli, M. (1985) Normal values of mucociliary transport time in young subjects. Int. J.
41 Pediatr. Otorhinolaryngol. 9: 151-156.
42 Patton, J. S. (1996) Mechanisms of macromolecule absorption by the lungs. Adv. Drug Delivery Rev. 19: 3-36.
43 Pavia, D. (1987) Acute respiratory infections and mucociliary clearance. Eur. J. Respir. Dis. 71: 219-226.
44 Pavia, D.; Bateman, J. R. M.; Clarke, S. W. (1980) Deposition and clearance of inhaled particles. Bull. Eur.
45 Physiopath. Respir. 16: 335-366.
46 Peterson, B. T.; Dickerson, K. D.; James, H. L.; Miller, E. J.; McLarty, J. W.; Holiday, D. B. (1989) Comparison of
47 three tracers for detecting lung epithelial injury in anesthetized sheep. J. Appl. Physiol. 66: 2374-2383.
48 Phalen, R. F.; Oldham, M. J.; Beaucage, C. B.; Crocker, T. T.; Mortensen, J. D. (1985) Postnatal enlargement of
49 human tracheobronchial airways and implications for particle deposition. Anat. Rec. 212: 368-380.
50 Phalen, R. F.; Oldham, M. J. (2001) Methods for modeling particle deposition as a function of age. Respir. Physiol.
51 128:119-130.
52 Pritchard, J. N.; Holmes, A.; Evans, J. C.; Evans, N.; Evans, R. J.; Morgan, A. (1985) The distribution of dust in the
53 rat lung following administration by inhalation and by single intratracheal instillation. Environ. Res.
54 36:268-297.
55 Pritchard, J. N.; Jefferies, S. J.; Black, A. (1986) Sex differences in the regional deposition of inhaled particles in
56 the 2.5-7.5 urn size range. J. Aerosol Sci. 17: 385-389.
June 2003 6-114 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, I; 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, I; 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 Sabaitis, C. P.; Leong, B. K. J.; Rop, D. A.; Aaron, C. S. (1999) Validation of intratracheal instillation as an
16 alternative for aerosol inhalation toxicity testing. J. Appl. Toxicol. 19: 133-140.
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 in boli 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.; Gurman, J. L.; Lippmann, M. (1982) Particle deposition within bronchial airways: comparisons
32 using constant and cyclic inspiratory flows. Ann. Occup. Hyg. 26: 47-64.
33 Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
34 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
35 Schroter, R. C.; Sudlow, M. F. (1969) Flow patterns in models of the human bronchial airways. Respir. Physiol.
36 7:341-55.
37 Segal, R. A.; Guan, X.; Shearer, M; Martonen, T. B. (2000a) Mathematical model of airflow in the lungs of
38 children I: effects of tumor sizes and locations. J. Theor. Med. 2:199-213.
39 Segal, R. A.; Martonen, T. B.; Kim, C. S. (2000b) Comparison of computer simulations to total lung deposition to
40 human subject data in healthy test subjects. J. Air Waste Manage. Assoc. 50: 1262-1268.
41 Segal, R. A.; Martonen, T. B.; Kim, C. S.; Shearer, M. (2002) Computer simulations of particle deposition in the
42 lungs of chronic obstructive pulmonary disease patients. Inhalation Toxicol. 14: 705-720.
43 Smaldone, G. C.; Perry, R. J.; Bennett, W. D.; Messina, M. S.; Zwang, J.; Ilowite, J. (1988) Interpretation of
44 "24 hour lung retention" in studies of mucociliary clearance. J. Aerosol Med. 1: 11-20.
45 Smaldone, G. C.; Foster, W. M.; Oriordan, T. G.; Messina, M. S.; Perry, R. J.; Langenback, E. G. (1993) Regional
46 impairment of mucociliary clearance in chronic obstructive pulmonary disease. Chest 103: 1390-1396.
47 Snipes, M. B. (1989) Long-term retention and clearance of particles inhaled by mammalian species. CRC Crit. Rev.
48 Toxicol. 20: 175-211.
49 Snipes, M. B.; Clem, M. F. (1981) Retention of microspheres in the rat lung after intratracheal instillation.
50 Environ. Res. 24: 33-41.
51 Snipes, M. B.; McClellan, R. O.; Mauderly, J. L.; Wolff, R. K. (1989) Retention patterns for inhaled particles in the
52 lung: comparisons between laboratory animals and humans for chronic exposures. Health Phys.
53 57(suppl. 1): 69-78.
54 Snipes, M. B.; James, A. C.; Jarabek, A. M. (1997) The 1994 ICRP66 human respiratory tract dosimetry model as a
55 tool for predicting lung burdens from exposures to environmental aerosols. Appl. Occup. Environ. Hyg.
56 12:547-554.
June 2003 6-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G.; Philipson, K. (1986a) Clearance from the human airways of
2 particles of different sizes deposited from inhaled aerosol boli. In: Aerosols: formation and reactivity,
3 proceedings of the second international aerosol conference; September; Berlin, Federal Republic of
4 Germany. Oxford, United Kingdom: Pergamon Press; pp. 192-196.
5 Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G. (1986b) Measurement of lung clearance with pulses of
6 radioactively-labelled aerosols. J. Aerosol Sci. 17: 333-336.
7 Stanley, P. J.; Wilson, R.; Greenstone, M. A.; Mackay, I. S.; Cole, P. J. (1985) Abnormal nasal mucociliary
8 clearance in patients with rhinitis and its relationship to concomitant chest disease. Br. J. Dis. Chest
9 79:77-82.
10 Stapleton, K. W.; Guentsch, E.; Hoskinson, M. K.; Finlay, W. H. (2000) On the suitability of K-E turbulence
11 modeling for aerosol deposition in the mouth and throat: a comparison with experiment. J. Aerosol Sci.
12 31:739-749.
13 Stober, W.; Morrow, P. E.; Koch, W.; Morawietz, G. (1994) Alveolar clearance and retention of inhaled insoluble
14 particles in rats simulated by a model inferring macrophage particle load distributions. J. Aerosol Sci.
15 25:975-1002.
16 Strom, K. A.; Chan, T. L.; Johnson, J. T. (1988) Pulmonary retention of inhaled submicron particles in rats: diesel
17 exhaust exposures and lung retention model. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.; Fischer, D. R.,
18 eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung dosimetry;
19 September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 645-657.
20 Suarez, S.; Kazantseva, M.; Bhat, M.; Costa, D.; Hickey, A. J. (2001) The influence of suspension nebulization or
21 instillation on particle uptake by guinez pig alveolar macrophages. Inhalation Toxicol. 13: 773-788.
22 Subramaniam, R. P.; Richardson, R. B.; Morgan, K. T.; Guilmette, R. A.; Kimbell, J. S. (1998) Computational fluid
23 dynamics simulations of inspiratory airflow in the human nose and nasopharynx. Inhalation Toxicol.
24 10:91-120.
25 Subramaniam, R. P.; Asgharian, B.; Miller, F. J.; Anjilvel, S.; Freijer, J. I. (2003) Analysis of lobar differences in
26 particle deposition in the human lung. Inhalation Toxicol. 15: 1-21.
27 Svartengren, K.; Lindestad, P.; Svartengren, M.; Philipson, K.; Bylin, G.; Camner, P. (1995) Added external
28 resistance reduces oropharyngeal deposition and increases lung deposition of aerosol particles in asthmatics.
29 Am. J. Respir. Crit. Care Med. 152: 32-37.
30 Svartengren, K.; Philipson, K.; Svartengren, M.; Anderson, M.; Camner, P. (1996a) Tracheobronchial deposition
31 and clearance in small airways in asthmatic subjects. Eur. Respir. J. 9: 1123-1129.
32 Svartengren, K.; Ericsson, C. H.; Svartengren, M.; Mossberg, B.; Philipson, K.; Camner, P. (1996b) Deposition and
33 clearance in large and small airways in chronic bronchitis. Exp. Lung Res. 22: 555-576.
34 Svartengren, K.; Philipson, K.; Svartengren, M.; Camner, P. (1998) Effect of adrenergic stimulation on clearance
35 from small ciliated airways in healthy subjects. Exp. Lung Res. 24: 149-158.
36 Svartengren, M.; Svartengren, K.; Aghaie, F.; Philipson, K.; Camner, P. (1999) Lung deposition and extremely
37 slow inhalations of particles. Limited effect of induced airway obstruction. Exp. Lung Res. 25: 353-366.
38 Sweeney, T. D.; Skornik, W. A.; Brain, J. D.; Hatch, V.; Godleski, J. J. (1995) Chronic bronchitis alters the pattern
39 of aerosol deposition in the lung. Am. J. Respir. Crit. Care Med. 151: 482-488.
40 Swift, D. L.; Strong, J. C. (1996) Nasal deposition of ultrafine 218Po aerosols in human subjects. J. Aerosol Sci.
41 27: 1125-1132.
42 Takahashi, S.; Asaho, S.; Kubota, Y.; Sato, H.; Matsuoka, O. (1987) Distribution of 198Au and 133Ba in thoracic and
43 cervical lymph nodes of the rat following the intratracheal instillation of 198Au-colloid and 133BaSO4.
44 J. Radiat. Res. 28:227-231.
45 Takahashi, S.; Kubota, Y.; Hatsuno, H. (1992) Effect of size on the movement of latex particles in the respiratory
46 tract following local administration. Inhalation Toxicol. 4: 113-123.
47 Takenaka, S.; Karg, E.; Roth, C.; Schulz, H.; Ziesenis, A.; Heinzmann, U.; Schramel, P.; Heyder, J. (2001)
48 Pulmonary and systemic distribution of inhaled ultrafine silver particles in rats. Environ. Health Perspect.
49 109(suppl. 4): 547-551.
50 Toms, N.; Hasani, A.; Pavia, D.; Clarke, S. W.; Agnew, J. E. (1997) Effect of mucus hypersecretion on initial time
51 course of inert particle clearance from the lung. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII:
52 proceedings of an international symposium on inhaled particles organised by the British Occupational
53 Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 509-514.
54 Tran, C. L.; Jones, A. D.; Cullen, R. T.; Donaldson, K. (1999) Exploration of the mechanisms of retention and
5 5 clearance of low-toxicity particles in the rat lung using a mathematical model. Inhalation Toxicol.
56 11: 1077-1108.
June 2003 6-116 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.
3 12: 1113-1126.
4 Tsuda, A.; Butler, J. P.; Fredberg, J. J. (1994a) Effects of alveolated duct structure on aerosol kinetics. I.
5 Diffusional deposition in the absence of gravity. J. Appl. Physiol. 76: 2497-2509.
6 Tsuda, A.; Butler, J. P.; Fredberg, J. J. (1994b) Effects of alveolated duct structure on aerosol kinetics. II.
7 Gravitational sedimentation and inertia! impaction. J. Appl. Physiol. 76: 2510-2516.
8 Tsuda, A.; Henry, F. S.; Otani, Y.; Haber, S.; Butler, J. P. (1996) Aerosol transport and deposition in the
9 rhythmically expanding pulmonary acinus. J. Aerosol. Med. 9: 389-408.
10 U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
11 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
12 Venkataraman, C.; Kao, A. S. (1999) Comparison of particle lung doses from the fine and coarse fractions of urban
13 PM-10 aerosols. Inhalation Toxicol. 11: 151-169.
14 Wagner, E. M; Foster, W. M. (2001) The role of the bronchial vasculature in soluble particle clearance. Environ.
15 Health Perspect. 109(suppl. 4): 563-565.
16 Warheit, D. B.; Hartsky, M. A. (1990) Species comparisons of proximal alveolar deposition patterns of inhaled
17 particulates. Exp. Lung Res. 16: 83-99.
18 Warheit, D. B.; Carakostas, M. C.; Hartsky, M. A.; Hansen, J. F. (1991) Development of a short-term inhalation
19 bioassay to assess pulmonary toxicity of inhaled particles: comparisons of pulmonary responses to carbonyl
20 iron and silica. Toxicol. Appl. Pharmacol. 107: 350-368.
21 White, F. M. (1974) Viscous fluid flow. New York, NY: McGraw-Hill, pp. 65-72.
22 Winter-Sorkina, R. de; Cassee, F. R. (2002) From concentration to dose: factors influencing airborne paniculate
23 matter deposition in humans and rats. Bilthoven, The Netherlands: National Institute of Public Health and
24 the Environment (RIVM); report no. 650010031/2002. Available:
25 http://www.rivm.nl/bibliotheek/rapporten/650010031.html (13 June 2003).
26 Wolff, R. K.; Henderson, R. F.; Snipes, M. B.; Griffith, W. C.; Mauderly, J. L.; Cuddihy, R. G.; McClellan, R. O.
27 (1987) Alterations in particle accumulation and clearance in lungs of rats chronically exposed to diesel
28 exhaust. Fundam. Appl. Toxicol. 9: 154-166.
29 Xu, G. B.; Yu, C. P. (1986) Effects of age on deposition of inhaled aerosols in the human lung. Aerosol Sci.
30 Technol. 5: 349-357.
31 Yeates, D. B.; Aspin, M. (1978) A mathematical description of the airways of the human lungs. Respir. Physiol.
32 32:91-104.
33 Yeates, D. B.; Aspin, N.; Levison, H.; Jones, M. T.; Bryan, A. C. (1975) Mucociliary tracheal transport rates in
34 man. J. Appl. Physiol. 19: 487-495.
35 Yeates, D. B.; Pitt, B. R.; Spektor, D. M.; Karron, G. A.; Albert, R. E. (1981) Coordination of mucociliary transport
36 in human trachea and intrapulmonary airways. J. Appl. Physiol. 51: 1057-1064.
37 Yeh, H.-C.; Schum, G. M. (1980) Models of human lung airways and their application to inhaled particle
38 deposition. Bull. Math. Biol. 42: 461-480.
39 Yu, G.; Zhang, Z.; Lessmann, R. (1998) Fluid flow and particle diffusion in the human upper respiratory system.
40 Aerosol Sci. Technol. 28: 146-158.
41 Zhang, Z.; Martonen, T. (1997) Deposition of ultrafine aerosols in human tracheobronchial airways. Inhalation
42 Toxicol. 9:99-110.
43 Zhang, L.; Asgharian, B.; Anjilvel, S. (1997) Inertia! deposition of particles in the human upper airway
44 bifurcations. Aerosol Sci. Tech. 26: 97-110.
45 Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2000) Effects of asymmetric branch flow rates on aerosol deposition in
46 bifurcating airways. J. Med. Eng. Technol. 24: 192-202.
47 Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2001) Effects of curved inlet tube on air flow and particle deposition in
48 bifurcating lung models. J. Biomechanics 34: 659-669.
49 Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2002) Cyclic micron-size particle inhalation and deposition in a triple
50 bifurcation lung airway model. J. Aerosol Sci. 33: 257-281.
51 Zhu, Y.; Hinds, W. C.; Kim, S.; Shen, S.; Sioutas, C. (2002) Study of ultrafine particles near a major highway with
52 heavy-duty diesel traffic. Atmos. Environ. 36: 4323-4335.
53
June 2003 6-117 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 airborne particulate matter (PM) during the past five years or so
7 has focused strongly on addressing several interrelated questions, such as: (1) what
8 characteristics (size, chemical composition, etc.) of ambient PM cause or contribute to health
9 effects; (2) what evidence is available for elucidating potential mechanisms underlying PM
10 health effects; (3) what susceptible subgroups are at increased risk for ambient PM health effects
11 and what types of factors contribute to their increased susceptibility; and (4) what evidence exist
12 that illustrates examples of interactive effects of particles and gaseous co-pollutants?
13 A variety of research approaches have been and continue to be used to address these
14 questions, including studies of human volunteers exposed to PM under controlled conditions;
15 in vivo studies of laboratory animals including nonhuman primates, dogs, and rodent species;
16 and in vitro studies of tissue, cellular, genetic, and biochemical systems. Similarly, a wide
17 variety of exposure conditions and exposure concentrations/doses have been employed,
18 including whole body and nose-only inhalation exposures to laboratory-generated PM or
19 concentrated ambient PM, intratracheal instillation, and in vitro exposure to test materials in
20 solution or suspension. These research approaches have been targeted mainly to test hypotheses
21 to provide improved understanding of the role of PM in producing health effects identified by
22 epidemiologic studies. Thus, most of the toxicological studies have been designed to address the
23 question of biologic plausibility of epidemiologically-demonstrated effects, rather than providing
24 dose-response quantification for experimentally-induced toxic effects. Much care should
25 therefore be taken when attempting to extrapolate effects seen in these studies to humans under
26 "real world" exposure conditions.
27 Particulate matter is a broad term that encompasses myriad physical and chemical species,
28 some of which have been investigated in the controlled laboratory animal or human studies.
29 However, a full discussion of all types of particles that have been studied is beyond the scope of
30 this chapter (see Chapter 2). Thus, specific criteria were used to select topics for presentation.
31 High priority was placed on studies that (1) elucidate health effects of ambient PM or its major
June 2003 7-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 common constituents and/or (2) may contribute to enhanced understanding of PM epidemiologic
2 study results. Diesel particulate matter (DPM) generally fits the above criteria; however,
3 because it is described in other documents in great detail (U. S. Environmental Protection
4 Agency, 1999; Health Effects Institute, 1995), only limited aspects (e.g., chronic animal studies,
5 controlled human studies, and immune effects) are covered in this chapter. Particles with high
6 inherent toxicity, such as silica, that are of concern mostly because of occupational exposure, are
7 excluded from this chapter and are discussed in detail in other documents and reports (e.g., U.S.
8 Environmental Protection Agency, 1996b; Gift and Faust, 1997).
9 Because of the sparsity of toxicological data on ambient PM at the time of the previous PM
10 Air Quality Criteria Document or "PM AQCD" (U.S. Environmental Protection Agency, 1996a),
11 the discussion of toxicologic effects of PM was organized there into specific chemical
12 components of ambient PM or model "surrogate" particles (e.g., acid aerosols, metals, ultrafine
13 particles, bioaerosols, "other particle matter"). Many of the newer toxicological studies evaluate
14 potential toxic effects of combustion-related particles. The main reason for this extensive
15 interest in combustion particles is that these particles, along with the secondary aerosols that they
16 form, are typically among the most dominant components represented in the fine fraction of
17 ambient air PM.
18 This chapter is organized as follows. The respiratory effects of specific components of
19 ambient PM or surrogate particles delivered by controlled in vivo exposures of both humans and
20 laboratory animals are discussed first (Section 7.2), followed by discussion of cardiovascular and
21 systemic effects of in vivo PM exposure (Section 7.3). In vitro exposure studies are discussed
22 next (Section 7.4) and are valuable in providing information on potential hazardous constituents
23 and mechanisms of PM injury. Studies of PM effects in laboratory animal models that mimic
24 human disease are then discussed (Section 7.5) as providing information useful for
25 characterizing factors affecting susceptibility to PM effects. Section 7.6 assesses controlled-
26 exposure studies evaluating health effects of mixtures of ambient PM or PM surrogates with
27 gaseous pollutants. This organization provides the underlying data for interpretive evaluation in
28 the final section (Section 7.7), but it may not fully convey the extensive and intricate linkages
29 among the pulmonary, cardiac, and nervous systems, all of which may be involved individually
30 and/or in concert in mediating PM exposure effects.
31
June 2003 7-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.2 RESPIRATORY EFFECTS OF PARTICULATE MATTER IN
2 HEALTHY HUMANS AND LABORATORY ANIMALS: IN VIVO
3 EXPOSURES
4 This section assesses the respiratory effects of (a) controlled human exposure to various
5 types of PM and (b) controlled laboratory animal PM exposures. Related in vitro studies using
6 animal or human respiratory cells are discussed in Section 7.4.
7 The biological responses occurring in the respiratory tract following controlled PM
8 inhalation include changes in pulmonary inflammation and systemic effects resulting from direct
9 effects on lung tissue. The observed responses may be dependent on the physicochemical
10 characteristics of the PM, the exposure, and the health status of the host. Many of the responses
11 are usually seen only at the higher concentrations typical of occupational and laboratory animal
12 exposures and not necessarily at (typically much lower) ambient particle concentrations.
13 Moreover, there are substantial differences in the inhalability and deposition profiles of PM in
14 humans and rodents (see Chapter 6 for details). Observed responses and dose-response
15 relationships also are very dependent on the specific biological response being measured.
16 Most of the laboratory animal studies summarized here used high particulate mass
17 concentrations administered by inhalation or by intratracheal instillation. The doses used are
18 generally quite high when compared to ambient exposure levels, even when laboratory animal -
19 to-human dosimetric differences are considered. Such high doses may be necessary, however,
20 in laboratory animal studies that explore potentially toxic effects using numbers of subjects
21 (animals) that are orders of magnitude fewer than numbers of human subjects included in most
22 epidemiology analyses. Further research on particle dose extrapolation is thusly needed to
23 determine species differences and to delineate the importance of exercise and other factors
24 influencing particle deposition in humans that, together, can account for large (possibly 50-fold
25 or more) differences in dose. Another important consideration is that healthy animals are most
26 typically used in controlled-exposure toxicology studies, whereas epidemiologic findings often
27 reflect ambient pollutant effects on susceptible or compromised humans (e.g., children or those
28 with one or another chronic disease). A key question, then, is the extent to which high-dose PM
29 exposures in healthy animals or even in acutely damaged animals exert toxic effects via similar
30 mechanisms operating in humans in response to exposures to doses of ambient PM.
31 As noted earlier, data available in the 1996 PM AQCD were from studies that evaluated
32 respiratory effects of specific components of ambient PM or surrogate particles, e.g., pure
June 2003 7-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 sulfuric acid droplets. More recently, pulmonary effects of controlled exposures to ambient PM
2 have been investigated by the use of particles collected from emission source bag filters or
3 ambient samplers (e.g., impactors; diffusion denuders) and by the use of aerosol concentrators
4 (e.g., Sioutas et al., 1995a,b, 2000; Gordon et al., 1998; Chang et al., 2000, Kim et al., 2000a,b).
5 Particles from ambient air samplers are collected on filters or other media, stored, and
6 resuspended in an aqueous medium for use in inhalation, intratracheal installation, or in vitro
7 studies. Both ambient PM and concentrated ambient particles (CAPs) have been used to
8 evaluate effects in normal and compromised laboratory animals and humans. Some ambient PM
9 has been standardized as a reference material and compared to existing dust and soot standards
10 [e.g., National Institutes of Standards and Technology (NIST)].
11 Particle concentrators provide a technique for exposing animals or humans by inhalation to
12 concentrated ambient particles (CAPs) at levels higher than typical ambient PM concentrations.
13 The development of particle concentrators has permitted the study of ambient real-world
14 particles under controlled conditions. This strength is somewhat weakened by the inability of
15 CAPs studies to precisely control the mass concentration and day-to-day variability in ambient
16 particle composition. Nonetheless, these studies are invaluable in the attempt to understand the
17 biological mechanisms responsible for the cardiopulmonary response to inhaled PM. Because
18 the composition of concentrated ambient PM varies in both time and location, a thorough
19 physical-chemical characterization is necessary to compare results among studies or even among
20 exposures within studies or to link particle composition to effect.
21 The in vivo studies discussed here and in vitro studies discussed later have almost
22 exclusively used PM10 or PM2 5 as particle size cutoffs for studying the adverse effects of
23 ambient PM. Studying particles in such size ranges is justified based in part on interests in
24 evaluating the bases for existing PM10 and PM25 standards. In addition, collection of these size
25 fractions has been made easier by widespread availability of ambient sampling equipment for
26 PM10 and PM25. Unfortunately, the study of other important size fractions, such as the coarse
27 fraction (PM10_2 5) and PMX 0 has been largely ignored, and only limited toxicology data are
28 available to specifically address these potentially important particle sizes. Similarly, although
29 organic compounds often comprise 20 to 60% of the dry fine particle mass of ambient PM
30 (Chapter 3), little research has addressed mechanisms by which this organic fraction contributes
31 to adverse effects associated with ambient PM exposures. The potential contribution of organics
June 2003 7-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 in mutagenesis and carcinogenesis has been studied, but these extensive findings are only briefly
2 discussed in this chapter (Section 7.4.3.2), which mainly focuses on studies aimed at evaluating
3 the biological plausibility of epidemiologic evidence for increased cardiopulmonary morbidity
4 and mortality being associated with exposure to ambient PM.
5
6 7.2.1 Ambient Combustion-Related and Surrogate Participate Matter
7 Some new in vivo toxicology studies utilizing inhalation exposure as a technique for
8 evaluating the respiratory effects of ambient particles in humans and laboratory animals have
9 been conducted with CAPs and with DPM. However, the vast majority of the new in vivo
10 exposure studies have utilized intratracheal instillation techniques. The pros and cons of this
11 technique in comparison to inhalation are covered in Chapter 6 (Section 6.5), and these issues
12 have also been reviewed elsewhere (Driscoll et al., 2000; Oberdorster et al., 1997; Osier and
13 Oberdorster, 1997). In most of the studies, PM samples were collected on filters, resuspended in
14 a vehicle (usually saline), and a small volume of the suspension was instilled intratracheally into
15 the animals. The physiochemical characteristics of the collected PM may be altered by
16 deposition and storage on a filter and resuspension in an aqueous medium. In addition, the doses
17 used in these instillation studies are generally high compared to ambient concentrations, even
18 when laboratory animal-to-human dosimetric differences are considered. Therefore, in terms of
19 direct extrapolation to humans in ambient exposure scenarios, greater importance should be
20 placed on inhalation studies. However, delivery of PM by instillation has the advantages that
21 much less material is needed and that the dose is accurate even though the particle deposition
22 and distribution patterns differ somewhat from that of inhalation. Instillation studies have
23 proven valuable in comparing the effects of different types of PM and for investigating some of
24 the mechanisms by which particles may cause inflammation and lung injury. Tables 7-la,b,
25 7-2a,b, and 7-3 summarize studies in which various biological endpoints were measured
26 following exposures to CAPs, ambient PM extracts, complex combustion-related PM, or
27 laboratory-derived surrogate PM, respectively.
28 There were only limited data available from human studies or laboratory animal studies on
29 ultrafine particles and even less on coarse particles at the time of the release of the previous
30 criteria document (U.S. Environmental Protection Agency, 1996a). In vitro studies had shown
31 that ultrafine particles have the capacity to cause injury to cells of the respiratory tract. High
June 2003 7-5 DRAFT-DO NOT QUOTE OR CITE
-------
•— H
3
O
^j
i
ON
o
>
>Tl
H
6
0
2|
0
H
O
0
H
W
O
HH
H
W
TABLE 7-la. RESPIRATORY EFFECTS OF INHALED AMBIENT PARTICULATE MATTER IN CONTROLLED
EXPOSURE STUDIES OF HUMAN SUBJECTS AND LABORATORY ANIMALS
Species, Gender, Strain,
Age, etc.
Humans, healthy
nonsmokers;
18to40yrold
Humans, healthy;
n=4, 19-41 yr old
Rats, male S-D
200-225 g, control
and SCytreated
Mongrel dogs, some
with balloon occluded
LAD coronary artery
n= 14
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, male
F344
Particle"
CAPs
(Chapel Hill)
CAPs
(LA)
Concentrated
ambient
particles
(CAPs)
(Boston)
CAPs
(Boston)
CAPs
(NY)
CAPs
(RTF)
CAPs
(NY)
PEE = Peak expiratory flow
TV = tidal volume
Exposure Exposure
Technique Concentration Particle Size Duration
Inhalation 23.1 to 311.1 ug/m3 0.65 um 2 h; analysis
og = 2.35 at!8h
Inhalation 148-246 ug/m3 PM25 2h
Inhalation; 73.5 to 733 ug/m3 0.18 and 5 h/day for
Harvard/EPA for Days 1-3; 0.27 um 3 days
fine particle 29 °C, og = 2.9
concentrator; 47 and 59% RH
animals
restrained
in chamber
Inhalation via 69-828 ug/m3 0.23 to 6 h/day x
tracheostomy 0.34 um 3 days
og = 0.2 to
2.9
Inhalation 132 to 919 ug/m3 0.2tol.2um 1 x 3 h or
og = 0.2 to 3 x 6 h
3.9
Inhalation 650 ug/m3 6 h/day x
2-3 days
Inhalation 100-350 ug/m3 0.4 um 3 h
(mean 225 ug/m3) og = 2.5
SaO2 = arterial oxygen saturation
Effect of Particles/Comments
Increased BAL neutrophils in both bronchial and
alveolar fractions. Particles were concentrated 6- to
10-fold at the inlet of the chamber.
No significant changes in lung function, symptoms,
SaO2, or Holler ECGs were observed. The maximum
steady slale fine particle concenlralion in Ihe brealhing
zone was lypically seven limes Ihe ambienl
concenlralion.
PEE and TV increased in CAPS exposed animals.
Increased prolein and percenl neulrophils and
lymphocytes in lavage fluid after CAPS exposure.
Responses were grealer in SO2-bronchilis animals. No
changes in LDH. No dealhs occurred. Exposures
were lo 30-40 limes grealer PM concenlralions lhan
found in ambienl air.
Decreased respiratory rale and increased lavage fluid
neulrophils in normal dogs. Sludy utilized Harvard
ambienl particle concenlralor. Ambienl particles
concenlraled by approximately 30-fold.
No inflammatory responses, no cell damage responses,
no PET changes. The PM mean concenlralion factor
(gravimelric) was 19.5 ± 18.6.
No significanl changes in heallhy rals; increased
BALE prolein and neulrophil influx in bronchilic rals;
responses were variable belween exposure regimens.
Basal levels of superoxide ('O2~) reduced by 90% 72 h
poslexposure; zymosan-slimulaled O2~ formation
increased by > 150% after 24 h; basal level H2O2
production by PAM depressed 90% 3 h following
exposure and remained 60% below levels al leasl 24 h;
zymosan-slimulaled H2O2 unaffected.
Reference
Ohio el al.
(2000a)
Gong el al.
(2000)
Clarke el al.
(1999)
Saldiva el al.
(2002)
Godleski
el al. (2000)
Gordon el al.
(2000)
Kodavanli
etal.
(2000a)
Zelikoff
elal.
(2003)
LDH = lactic dehydrogenase
-------
TABLE 7-lb. RESPIRATORY EFFECTS OF INSTILLED AMBIENT PARTICIPATE MATTER IN
LABORATORY ANIMALS AND HUMAN SUBJECTS
to
o
o
-^
^
o
hj
fe
H
6
O
0
0
0
H
W
O
O
H
W
Species, Gender,
Strain, Age, etc.
Rats, male S-D
60 days
Rats, S-D
60 days
Rats, Wis
(HAN strain)
Rats, S-D
Humans, healthy
nonsmokers;
21 M, 3 F;
26.4 ± 2.2 yr old
Exposure
Particle" Technique
Provo, UT, Intratracheal
TSP filters instillation
(10 years old)
Provo, UT, Intratracheal
TSP filters instillation
(10 years old),
soluble and
insoluble extracts
Ambient PM Intratracheal
Edinburgh, CB, CB instillation
Ultrafme (UCB)
DEP Intratracheal
instillation
Provo, UT, Intrabronchial
PM10 filters instillation
(10 years old)
Concentration Particle Size
0.25,1.0,2.5, N/A
S.OmgofPM
extract in 0.3 mL
saline
100-1000 ug N/A
of PM extract in
0. 5 mL saline
50-125 ug in PM10
.2 mL CB = (200-500 nm)
UCB = 20 nm
500 ug in 0.5 ml N/A
saline
SOOugofPM N/A
extract in lOmL
saline
Exposure
Duration Effect of Particles/Comments
24 h Inflammation (PMN) and pulmonary injury
produced by particles collected during steel
mill operation was greater than for during
period of mill closure.
24 h Inflammation (PMN) and lavage fluid protein
was greater with the soluble fraction containing
more metal (Zn, Fe, Cu).
Sacrificed at Increased PMN, protein, and LDH following
6 h PM10; greater response with ultrafme CB but
not CB; decreased GSH level in BAL; free
radical activity (deplete supercoil DNA);
leukocytes from treated animals produced
greater NO and TNF.
3 times/wk, Decreased concentration of lavage ascorbate.
2 wk Urate and glutathione concentrations
unchanged; elevated MIP-2 and TNF; total cell
count increased; lavage protein and LDH
increased; increased total lavage iron
concentration.
24 h BAL Inflammation (PMN) and pulmonary injury
produced by particles collected during steel
mill operation was greater than for during
period of mill closure.
Reference
Dye et al.
(2001)
Ohio et al.
(1999a)
Li et al.
(1996, 1997)
Ohio et al.
(2000b)
Ohio and
Devlin
(2001)
"PEF = Peak expiratory flow
TV = tidal volume
LDH = lactic dehydrogenase
SaO2 = arterial oxygen saturation
-------
to
O
o
TABLE 7-2a. RESPIRATORY EFFECTS OF INSTILLED COMPLEX COMBUSTION-RELATED
PARTICIPATE MATTER IN LABORATORY ANIMALS
oo
fe
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Species, Gender,
Strain, Age, etc.
Hamsters, Syrian
golden, male,
90-125 g
Mice, female,
NMRI, 28-32 g
Rats, male, S-D,
60 days old
Rats, male, S-D,
60 days old
Rats, S-D, 5-day-
old
Rats, male, S-D
rats
60 days old
Particle'
Kuwaiti oil fire
particles;
urban particles
from St. Louis, MO
CFA
CMP
we
Emission source
PM (ROFA,
DOFA, CFA)
Ambient airshed
PM
ROFA
ROFA
ROFA
#6 ROFA,
volcanic ash
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
Instillation
Intratracheal
Instillation
Concentration
0.15, 0.75, and
3.75 mg/100 g
CMP: 20 jig arsenic/kg,
or CMP: 100 mg
particles/kg,
WC alone (100 mg/kg),
CFA alone (100 mg/kg
[i.e., 20 ug arsenic/kg]),
CMP mixed with WC
(CMP, 13.6 mg/kg [(i.e.,
20 ug arsenic/kg]),
WC (86.4 mg/kg) and
Ca3(AsO4)2 mixed with
WC (20 jig arsenic/kg),
WC (100 mg/kg)
Total mass: 2.5 mg/rat
Total transition metal:
46 ug/rat
8.33 mg/mL
0.3 mL/rat
500 jig/rat
0.3, 1.7
8.3 mg/mL
8.3 mg/mL
Particle Size
Oil fire
particles:
< 3.5 um,
10daysof24-h
samples
N/A
Emission PM:
1.78-4. 17 urn
Ambient PM:
3.27-4.09 urn
1.95 urn
1.95 um
1.95 urn
og = 2.19
1.4 um
Exposure
Duration
Sacrificed 1
and 7 days
postinstillation
1, 5, 30 days
post-treatment,
lavage for total
protein content,
inflammatory
cell number
and type, and
TNF-a
production
particle
retention
Analysis at
24 and 96 h
following
instillation
Analysis at
24 and 96 h
24h
24 h
Effect of Particles/Comments
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.
Mild inflammation for WC; Ca3(AsO4)2 caused
significant inflammation; 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, and
eosinophils following exposure to emission and ambient
particles; induction of injury by emission and ambient
PM samples is determined primarily by constituent
metals and their bioavailability.
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 reactive oxygen
species.
Plasma fibrinogen elevated after ROFA instillation
but not volcanic ash
Reference
Brain et al.
(1998)
Broeckaert
etal. (1997)
Costa and
Dreher
(1997)
Dreher et al.
(1997)
Dye et al.
(1997)
Gardner
et al. (2000)
-------
TABLE 7-2a (cont'd). RESPIRATORY EFFECTS OF INSTILLED COMPLEX COMBUSTION-RELATED
PARTICIPATE MATTER IN LABORATORY ANIMALS
to
o
o
^J
^b
O
p>
~
H
1
O
o
2|
0
H
0
Cj
0
H
W
O
HH
H
W
Species, Gender,
Strain, Age, etc.
Rats, male, S-D,
5-day-old
Rats, male,
S-D, 60 days old
Mice, female,
Balb/cJ
7-15 weeks
Rats, male, S-D
Mice, normal and
Hp, 105 days old
Rats, male,
S-D, 60 days old
Rats, male,
S-D and F-344
(60 days old)
Particle'
lo-S
#6 ROFA,
volcanic ash
saline
Two ROFA
samples
Rl had 2 x saline-
leachable sulfate,
Ni, and V and
40 x Fe as R2; R2
had 3 1 x higher Zn
#6 ROFA, lo-S
ROFA
ROFA
ROFA
ROFA
Exposure
Technique
Intratracheal
Instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Concentration Particle Size
0.3, 1.7, 1.95 um
8.3 mg/kg BW og=1.95
in saline 1.4um
8.3 mg/kg BW
1 mL/kg BW
2.5mgin0.3mL Rl: 1.88 um
R2: 2.03 um
60 ug in 50 uL < 2.5
(dose 3 mg/kg)
500 ug/animal 3.6 um
50 ug 1.95 um
1.0 mg in 0.5 mL saline 1.95 um
8.3 mg/kg 1.95 um
og = 2.14
Exposure
Duration
24 h
Analysis at
4 days
Analysis at 1,
3, 8, 15 days
after instillation
Analyzed
4 and 96 h
postexposure
Analysis at
24 h
Analysis at
24 h
Sacrificed at
24 h
Effect of Particles/Comments
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 Rls 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 PMNs, protein.
Increase in neutrophils in both S-D and F-344 rats;
a time-dependent increase in eosinophils occurred in
S-D rats but not in F-344 rats.
Reference
Gardner
et al. (2000)
Gavett et al.
(1997)
Gavett et al.
(1999)
Ohio et al.
(1998a)
Ohio et al.
(2000c)
Kadiiska
etal. (1997)
Kodavanti
etal. (1996)
-------
to
O
o
TABLE 7-2a (cont'd). RESPIRATORY EFFECTS OF INSTILLED COMPLEX COMBUSTION-RELATED
PARTICIPATE MATTER IN LABORATORY ANIMALS
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Species, Gender,
Strain, Age, etc. Particle"
Rats, male, S-D, ROFA
WIS, and F-344
(60 days old)
Exposure
Technique Concentration
mtratracheal 8.3 mg/kg
instillation
Exposure
Particle Size Duration
1.95 urn Sacrificed at 6,
og = 2.14 24, 48, and
72 h; 1, 3,
and 12 weeks
Effect of Particles/Comments
Inflammatory cell infiltration, as well as alveolar,
airway, and interstitial thickening in all three rat strains;
a sporadic incidence of focal alveolar 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.
Reference
Kodavanti
etal.
(1997a)
Rats, male, S-D,
60 days old
Rats, male, S-D,
60 days old
Rats, male, S-D
60-day-old treated
with MCT
(60 mg/kg)
Rats, male, WKY
andSH, 11-13
weeks old
ROFA Intratracheal 8.33 mg/kg 1.95 urn
instillation og = 2. 14
Fe2(SO4)3, ROFA-equivalent dose
VSO4, of metals
NiSO4
10 compositionally Intratracheal 0.833, 3.33, 8.3 mg/kg 1.99-2.59 urn
different ROFA instillation
particles from a
Boston power plant
ROFA Intratracheal 0, 0.83, 3.3 mg/kg 1.95 urn
instillation og = 2. 14
ROFA Intratracheal 3.33 mg/mL/kg 1.95 urn
instillation
VSO4, l.Sumol/kg og = 2.14
NiSO4, or saline
Analysis at 3,
24, and 96 h,
postinstillation
Sacrificed at
24 h
24-96 h
1 and 4 days;
postinstillation
analysis at 6 or
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 LDH, 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.
IT rats showed inflammatory responses (IL-6,
MIP-2 genes upregulated). 58% of rats exposed
to ROFA died within 96 h.
Increased BALF protein and LDH alveolitis with
macrophage accumulation in alveoli; increased
neutrophils in BAL. Increased pulmonary protein
leakage and inflammation in SH rats. Effects of metal
constituents of ROFA were strain specific; vanadium
caused pulmonary injury only in WKY rats; nickel was
toxic in both SH and WKY rats.
Kodavanti
etal.
(1997b)
Kodavanti
etal.
(1998a)
Kodavanti
etal. (1999)
Kodavanti
etal. (2001)
-------
•— ^
3
to
o
o
OJ
.
O
H
6
0
0
H
0
0
w
o
HH
H
W
TABLE 7-2a (cont'd). RESPIRATORY EFFECTS OF INSTILLED COMPLEX COMBUSTION-RELATED
PARTICULATE MATTER IN LABORATORY ANIMALS
Species, Gender,
Strain, Age, etc. Particle1*
Rats, Brown ROFA
Norway
Rats, male, S-D, #6 ROFA from
60-day-old Florida
Rats, male, S-D, NC ROFA;
60-day-old Domestic oil fly
ash
Rats, male, S-D; #6 ROFA (Florida)
60 days old NiSO4
VS04
Rats, male, S-D, ROFA
60-day-old
Rats, male, S-D, LoS,
60-day-old #6 ROFA
Rats, male, S-D; Ottawa dust,
60-day-old; WKY ROFA, and
and SH; volcanic ash
cold-stressed SH,
ozone-exposed
SH, and MCT-
treated SH
"CFA = Coal fly ash
CMP = Copper smelter dust
WC = Tungsten carbide
MCT = Monocrotaline
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
DOFA = Fly ash from a domestic oil-burning furnace
ROFA = Residual oil fly ash
Exposure
Concentration Particle Size Duration
200 ug N/A N/A
100 ug
1000 ug in 0.5 mL saline 1.95±0.18um 15 min to 24 h
1000 ug in 0.5 mL saline 15 min to 24 h
3.3 mg/mL/kg; ROFA 1.9 um 3 or 24 h
equivalent dose of og = 2. 14
metals
400-1000 ug/mL N/A 12 h post-IT
500 ug in 0.5 mL saline 3.6 um 1,4, or 24 h
Dose: IT 0,0.25, 1.0, 1.95 um 96 h post-IT
and 2.5 mg/rat
Fe2(SO4) = Iron sulfate
VSO4 = Vanadium sulfate
NiSO4 = Nickel sulfate
LoS = low sulfur
OVA = Ovalbumin
Effect of Particles/Comments
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.
Production of acetaldehyde increased at 2 h
postinstillation.
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.
Inflammatory and stress responses were upregulated;
the numbers of genes upregulated were correlated with
metal type and ROFA
ROFA increased PGE2 via cycloxygenase expression.
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
Lambert
etal. (1999)
Madden
etal. (1999)
Madden
etal. (1999)
Nadadur
et al. (2000);
Nadadur and
Kodavanti
(2002)
Samet et al.
(2000)
Silbajoris
et al. (2000)
Watkinson
etal.
(2000a,b)
-------
•— H
3
to
o
o
OJ
^
to
o
s>
^rj
H
6
o
21
0
H
r-i
*• — (
o
H
W
O
O
H
W
TABLE 7-2b. RESPIRATORY EFFECTS OF INHALED COMPLEX COMBUSTION-RELATED
PARTICULATE MATTER IN COMPROMISED LABORATORY ANIMAL MODELS
Species, Gender, Exposure
Strain, Age, etc. Particle" Technique
Rats, male WISTAR Coal oil fly Inhalation
Bor: WISW strain ash (chamber)
Mice, BALB/C, Aerosolized Nose-only
2-day-old, sensitized ROFA inhalation
to ovalbumin (OVA) leachate
Rats, S-D, 250 g ROFA Inhalation
MCT
Rats, male, S-D ROFA Nose-only
60-day-old treated inhalation
with MCT
(60 mg/kg)
Rats, male, WKY ROFA Nose-only
and SH, 11-13 weeks Inhalation
old
"CFA = Coal fly ash
CMP = Copper smelter dust
WC = Tungsten carbide
MCT = Monocrotaline
DOFA = Fly ash from a domestic oil-burning furnace
ROFA = Residual oil fly ash
Exposure
Concentration Particle Size Duration
0, 11, 32, and 1.9-2.6 urn 6 h/day,
103 mg/m3 og=1.6-1.8 5days/week,
4 weeks
50 mg/mL N/A 30 min
580 ± 2.06 urn 6 h/day for 3 days
HOug/m3 og=1.57
15 mg/m3 1.95 um 6 h/day for 3 days
og = 2. 14 analysis at 0 or
18h
15 mg/m3 1.95 um 6 h/day x 3 day,
og = 2. 14 analysis at 0 or
18h
Fe2(SO4) = Iron sulfate
VSO4 = Vanadium sulfate
NiSO4 = Nickel sulfate
LoS = low sulfur
OVA = Ovalbumin
Effect of Particles/Comments
At the highest concentration, type II cell proliferation and mild
fibrosis occurred and 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.
Increased airway response to methylcholine and to OVA in
ROFA exposed mice; increased airway inflammation also.
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.
No mortality occurred by inhalation. ROFA exacerbated lung
lesions (edema, inflammation, alveolar thickening) and gene
expression in MCT rats. Rats showed inflammatory responses
(IL-6, MIP-2 genes upregulated).
More pulmonary injury in SH rats. Increased RBCs in BALF
of SH rats. ROFA increased airway reactivity to Acctylcholine
in both SH and WKY 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.
Reference
Dormans
et al. (1999)
Hamada
et al. (1999)
Killingswort
h et al.
(1997)
Kodavanti
et al. (1999)
Kodavanti
etal.
(2000b)
-------
TABLE 7-3. RESPIRATORY EFFECTS OF SURROGATE PARTICIPATE MATTER IN LABORATORY ANIMALS
to
O
o
Species, Gender, Strain,
Age, etc.
Inhalation
Hamsters, Syrian golden
900 male, 900 female,
4-wks-old
Mice, C57B1/6J
Rats, male, F-344
200-230 g
Particle3
Toner
(carbon)
TiO2
Silica
PTFE
Ti02
PTFE
Fumes
Exposure
Technique
Nose-only
inhalation
Inhalation
Whole
body
inhalation
Concentration Particle Size
1.5, 6.0, or 4.0 nm
24 mg/m3
40 mg/m3 1.1 nm
3 mg/m3 1.4 nm
PTFE: PTFE: 18nm
1.25,2.5, or TiO2-F: 200 nm
5 x 105 particles/cc TiO2-D: 10 nm
TiO2-F: 10 mg/m3
NiO: 5 mg/m3
Ni3S2: 0.5 mg/m3
1,2.5, or 5 x 105 18 nm
particles/cm3
Exposure
Duration
3, 9, 15 mo
6h/day
5days/week
30 min or
6 h/day,
5days/week, 6 mo
15 min, analysis
4 h postexposure
Effect of Particles
Retention increased with increased
exposure. Clearance half-times
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
Reference
Creutzenberg
etal. (1998)
Finkelstein
etal. (1997)
Johnston et al.
(1996)
Mice, male, C57BL/6J, PTFE
8 weeks and 8-mo-old Fumes
Whole
body
inhalation
1, 2.5, or 5 x 105
particles/cm3
18 nm
expressed around all airways and
interstitial regions; PMN expressed
IL-6, MT, and TNF-a; AM and
epithelial cells were actively
involved.
30-min exposure, Increased PMN, lymphocytes, and
analysis 6 h protein levels in old mice over young
following mice; increased TNF-a mRNA in
exposure old mice over young mice;
no difference in LDH and
p-Glucuronidase.
Johnston et al.
(1998)
-------
TABLE 7-3 (cont'd). RESPIRATORY EFFECTS OF SURROGATE PARTICIPATE MATTER IN LABORATORY ANIMALS
to
O
o
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Species, Gender, Strain,
Age, etc.
Inhalation (cont'd)
Rats, male, S-D,
MCT-treated
Mice, male,
Swiss- Webster, 6-8 weeks
old (A/I AKR/J,
B6C3F1/J, BALB/cJ,
C3H/HeJ-C3, CSHeOuJ,
CSTBL/6J-B6, SJL/J,
SWR/J, 129/J) strains
raised in a pathogen free
laboratory
Exposure
Particle3 Technique Concentration
Fluorescent Inhalation 3. 85 ±0.81
micro sphere mg/m3
s
Carbon Nose only 10 mg/m3
black inhalation 285 |ig/m3
Regal 660
Carbon-
associated
S04=
Particle Size
1.38 ±0.10 |mi
ag= 1.8 ±0.28
0.29 |im
±2.7 |mi
Exposure
Duration Effect of Particles
3 h/day x 3 days Monocrotaline-treated animals
contained fewer microspheres in
their macrophages, probably because
of impaired chemo taxis.
4 h Differences in inflammatory
responses (PMN) across strains.
Appears to be genetic component to
the susceptibility.
Reference
Madl et al.
(1998)
Ohtsuka et al.
2000a,b
Instillation
Rats, male, S-D
(200g)
Diesel, Intratrache 1 mg in 0.4 mL.
SiO2, al
carbon instillation
black
DEP Collected
as TSP-
disaggregated in
solution by
sonication
(20 nm); SiO2
(7nm);
carbon black
Necropsy at 2, 7, Amorphous SiO2 increased
21, 42, and permeability, and neutrophilic
84 days inflammation. Carbon black
postinstillation and DEP translocated to interstitum
and lymph nodes by 12 weeks.
Murphy et al.
(1998)
aPTFE = polytetrafluoroethylene
TiO2 = titanium oxide
SiO, = silicon dioxide
-------
1 levels of ultrafme particles, as metal or polymer "fume," are associated with toxic respiratory
2 responses in humans and other mammals. Such exposures are associated with cough, dyspnea,
3 pulmonary edema, and acute inflammation. At concentrations less than 50 |ig/m3, freshly
4 generated insoluble ultrafme PTFE fume particles can be severely toxic to the lung. However,
5 it is not clear as to what roles in the observed effects may have been played by fume gases which
6 adhered to the particles. Newer data from controlled exposures have demonstrated that particle
7 composition, in addition to particle size, may be responsible for the adverse health effects
8 associated with ambient PM exposures.
9 Toxicological studies of other types of PM species were also discussed in the previous
10 criteria document (U.S. Environmental Protection Agency, 1996a). These studies included
11 exposures to fly ash, volcanic ash, coal dust, carbon black, and miscellaneous other particles,
12 either alone or in mixture. Some of the particles discussed were considered to be models of
13 "respirable low toxicity particles" and were used in instillation studies to delineate nonspecific
14 particle effects from effects of known toxicants. A number of studies on "other PM" examined
15 effects of up to 50,000 |ig/m3 of respirable particles with inherently low toxicity. Although there
16 was no mortality, some mild pulmonary function changes after exposure to 5,000 to 10,000
17 Hg/m3 of inert particles were observed in rats and guinea pigs. Lung morphology studies
18 revealed focal inflammatory responses, some epithelial hyperplasia, and fibrotic responses after
19 exposure to > 5,000 |ig/m3. Changes in macrophage clearance after exposure to > 10,000 |ig/m3
20 were equivocal (no host defense effects). In studies of mixtures of particles and other pollutants,
21 effects varied depending on the toxicity of the associated pollutant. In humans, co-exposure to
22 carbon particles appeared to increase responses to formaldehyde but not to acid aerosol. None of
23 the "other" particles mentioned above are present in ambient air in more than trace quantities.
24 Thus, it was concluded that the relevance of any of these studies to standard setting for ambient
25 PM may be extremely limited (see also Chapter 6, Section 4, Particle Overload m this draft
26 document).
27 Newer studies, on the other hand, appear to provide evidence of likely greater relevance to
28 understanding ambient PM exposure effects and underlying mechanisms.
29
30
31
June 2003 7-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.2.1.1 Ambient Particulate Matter
2 New studies that examined the acute effects of intratracheal instillation of ambient PM
3 obtained from specific ambient locations have shown clearly that PM can cause lung
4 inflammation and injury.
5 Costa and Dreher (1997) showed that instillation of relatively high concentrations of PM
6 samples from three emission sources (two oil and one coal fly ash) and four ambient airsheds
7 (St. Louis, MO; Washington, DC; Dusseldorf, Germany; and Ottawa, Canada) resulted in
8 increases in lung polymorphonuclear leucocytes (PMNs) and eosinophils in rats 24 h after
9 instillation. Biomarkers of permeability (total protein and albumin) and cellular injury, lactic
10 dehydrogenase (LDH), also were increased. Animals were dosed with (1) an equal dose by mass
11 (nominal 2.5 mg/rat) of each PM mixture or (2) normalization of each PM mass to a metal
12 content of 46 mg/dose and 35.5 jig of total metals (Cu, Fe, V, Zn) for the ambient PM and
13 ROFA comparison. This study demonstrated that the lung dose of bioavailable transition metal,
14 not instilled PM mass, was the primary determinant of the acute inflammatory response.
15 Kennedy et al. (1998) reported a similar dose-dependent inflammation (i.e., increase in
16 protein and PMN in lavage fluid, proliferation of bronchiolar epithelium, and intraalveolar
17 hemorrhage) in rats instilled with water-extracted particles (TSP) collected in Provo, UT. The
18 particulate mixture was composed of 1.0 mg/g Zn, 0.04 mg/g Ni, 2.2 mg/g Fe, 0.01 mg/g Vn,
19 1.4 mg/g Cu, 1.7 mg/g Pb, and 78 mg/g SO4= in 500 mL saline solution. This study also
20 indicated that the metal constituent, in this case PM-associated Cu, was a plausible cause of the
21 outcome based on IL-8 secretion and enhanced activation of the transcription factor NF-kB in
22 cultured epithelium.
23 Further toxicological studies of ambient PM collected around Provo, UT (Utah Valley) in
24 the late 1980s are particularly interesting (Ohio and Devlin, 2001; Dye et al., 2001; Wu et al.,
25 2001; Soukup et al., 2000; Frampton et al., 1999). Epidemiologic studies by Pope (1989, 1991)
26 had shown that exposures to PM10 during closure of an open-hearth steel mill over a 13-mo
27 period beginning in 1987 were associated with reductions in several health endpoints, e.g.,
28 hospital admissions for respiratory diseases, as discussed in the 1996 PM AQCD (U.S.
29 Environmental Protection Agency, 1996a). Ambient PM was collected near the steel mill during
30 the winter of 1986 (before closure), 1987 (during closure), and again in 1988 (after plant
31 reopening). The fibrous glass hi-vol filters were stored, folded PM-side inward, in plastic
June 2003 7-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 sleeves at room temperature and humidity (Dye et al., 2001). A description of the in vivo
2 toxicological studies follows; the in vitro studies are discussed in Section 7.5.2.1.
3 Ohio and Devlin (2001) investigated the biologic effect of PM from the Utah Valley to
4 determine if the biological responses mirrored the epidemiologic findings, with greater injury
5 occurring after exposure to an equal mass of particles from those years in which the mill was in
6 operation. Aqueous extracts of the filters collected prior to closure of the steel mill, during the
7 closure and after its reopening, were instilled through a bronchoscope into the lungs of
8 nonsmoking human volunteers. Twenty-four hours later, the same subsegment was lavaged.
9 Exposure to aqueous extracts of PM collected before closure and after reopening of the steel mill
10 provoked a greater inflammatory response than PM extract acquired during the plant shutdown.
11 These results indicate that the pulmonary effects observed after experimental exposure of
12 humans to the Utah Valley PM can be correlated with health outcomes observed in
13 epidemiologic studies of the same material under normal exposure conditions.
14 Dye et al. (2001) similarly examined the relationship between Utah Valley ambient PM
15 and respiratory health effects but in laboratory animals. Sprague-Dawley rats were
16 intratracheally instilled with equivalent masses of aqueous extracts from filters originally
17 collected during the winter before, during, and after closure of the steel mill. Twenty-four hours
18 after instillation, rats exposed to extracts of particles collected when the plant was open
19 developed significant pulmonary injury and neutrophilic inflammation. Additionally, 50% of
20 rats exposed to these extracts had increased airway responsiveness to acetylcholine, compared to
21 17 and 25% of rats exposed to saline or the extracts of particles collected when the plant was
22 closed. By 96 hr, these effects were largely resolved except for increases in lung lavage fluid
23 neutrophils and lymphocytes in rats exposed to PM extracts from prior to the plant closing.
24 Analogous effects were observed with lung histologic assessment. Extract analysis
25 demonstrated that nearly 70% of the mass in all three extracts appeared to be sodium-based salts
26 derived from the glass filter matrix. Extracts of particles collected when the plant was open
27 contained more sulfate, cationic salts (i.e., calcium, potassium, magnesium), and certain metals
28 (i.e., copper, zinc, iron, lead, strontium, arsenic, manganese, nickel). Although total metal
29 content was ~ 1% of the extracts by mass, the greater quantity detected in the extracts of particles
30 collected when the plant was open suggests that metals may be important determinants of the
31 observed pulmonary toxicity. The authors concluded that the pulmonary effects induced in rats
June 2003 7-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 by exposure to aqueous extracts of local ambient PM filters were in good accord with the
2 epidemiologic reports of adverse respiratory health effects in Utah Valley residents.
3 Molinelli et al. (2002) exposed human airway epithelial cell line (BEAS-2B) cultures for
4 24 h to an aqueous extract of PM collected in the Utah Valley. A portion of the extract was
5 treated with Chelex, an agent that removes transition metals from solution. Cells incubated with
6 the untreated extract showed a significant concentration-dependent increase in the inflammatory
7 mediator interleukin-8 (IL-8) when compared to the control cells. However, cells incubated with
8 Chelex-treated extract produced no change (relative to control) in IL-8. They exposed rats
9 in vivo for 24 h to the same treatments as the cells and found significant increases in lactate
10 dehydrogenase (LDH) and total protein in the rats exposed to the untreated extract and to the
11 Chel ex-treated extract with metals added back to achieve original concentrations. There was an
12 attenuation of the observed LDH and total protein increases in the rats instilled with the
13 Chelex-treated extract. The authors concluded that removal of metal cations attenuates cellular
14 responses to the aqueous extract and suggest a role for transition metal involvement in
15 PM-associated increases in morbidity and mortality.
16 In parallel work on potential importance of metals in mediating ambient PM effects,
17 Kodavanti et al. (2002) examined the role of zinc in PM-induced health effects in several
18 different animal models. Male Sprague-Dawley rats were instilled IT with an oil combustion
19 emission PM (EPM) in saline (0.0, 0.8, 3.3, or 8.3 mg/kg); and, in order to examine the potential
20 role of EPM teachable zinc, additional rats were instilled with either saline, whole EPM
21 suspension, the saline teachable fraction of EPM, the particulate fraction of EPM (8.3 mg/kg,
22 soluble Zn = 14.5 ug/mg EPM), or ZnSO4 (0.0, 33.0, or 66.0 ug/kg Zn). Three rat strains of
23 differing PM susceptibility (male SD, normotensive Wistar-Kyoto (WKY), and spontaneously
24 hypertensive (SH) rats (90 days old)) were exposed nose-only to either filtered air or EPM (2, 5,
25 or 10 mg/m3 for 6 h/day x 4 days/week x 1 week; or 10 mg/m3 for 6 h/day x 1 day/week for 1, 4,
26 or 16 weeks) and assessed at 2 days postexposure. Intratracheal exposures to whole EPM
27 suspensions were associated with a dose-dependent increase in protein/albumin permeability and
28 neutrophilic inflammation. Pulmonary protein/albumin leakage and neutrophilic inflammation
29 caused by the teachable fraction of EPM and ZnSO4 were comparable to the effects of the whole
30 suspension. However, protein/albumin leakage was not associated with the parti culate fraction,
31 although significant neutrophilic inflammation did occur following instillation. With EPM nose-
June 2003 7-18 DRAFT-DO NOT QUOTE OR CITE
-------
1 only inhalation, acute exposures (10 mg/m3 only) for 4 days resulted in small increases in
2 bronchoalveolar lavage fluid (BALF) protein and n-acetyl glucosaminidase activities
3 (approximately 50% above control). Unlike IT exposures, no neutrophilic influx was detectable
4 in BALF from any of the inhalation groups. The only major effect of acute and long-term EPM
5 inhalation was a dose- and time-dependent increase in alveolar macrophages (AM) regardless of
6 the rat strain. Histological evidence also showed dose- and time-dependent accumulations of
7 particle-loaded AM. Particles were also evident in interstitial spaces, and in the lung-associated
8 lymph nodes following the inhalation exposures (SH > WKY= SD). There were strain-related
9 differences in peripheral white blood cell counts and plasma fibrinogen with no major EPM
10 inhalation effect. The authors attributed the critical differences in pulmonary responsiveness to
11 EPM between IT and inhalation exposures to the dose of bioavailable zinc. EPM IT exposures,
12 but not acute and long-term inhalation of up to 10 mg/m3, caused neutrophilic inflammation.
13 Also of interest are some other new instillation study results. For example, Li et al. (1996,
14 1997) reported that instillation of ambient PM10 collected in Edinburgh, Scotland, also caused
15 pulmonary injury and inflammation in rats. In addition, Brain et al. (1998) examined the effects
16 of instillation of particles that resulted from the Kuwaiti oil fires in 1991 compared to effects of
17 urban PM collected in St. Louis (NIST SRM 1648, collected in a bag house in the early 1980s)
18 and showed that, on an equal mass basis, the acute toxicity of the Kuwaiti oil fire particles was
19 similar to that of urban particles collected in the United States.
20 The fact that instillation of ambient PM collected from different geographical areas and
21 from a variety of emission sources consistently caused pulmonary inflammation and injury tends
22 to corroborate epidemiologic studies that report increased PM-associated respiratory effects in
23 populations living in many different geographical areas and climates. On the other hand, there is
24 a potential that more "realistic" doses of metals may activate cells and signaling pathways in a
25 manner that are not observed at doses that are magnitudes greater than present in ambient air,
26 such that these mechanisms may be overwhelmed. Thus, high-dose instillation studies may
27 produce different effects on the lung than inhalation exposures at more relevant concentrations.
28 With regard to inhalation studies more directly mimicking ambient exposures, Ohio et al.
29 (2000a) exposed 38 healthy volunteers exercising intermittently at moderate levels of exertion
30 for 2 h to either filtered air or particles concentrated from the air in Chapel Hill, NC (23 to
31 311 |ig/m2). Analysis of cells and fluid obtained 18 h after exposure showed a mild increase in
June 2003 7-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 neutrophils in the bronchial and alveolar fractions of bronchoalveolar lavage (BAL) in subjects
2 exposed to the highest quartile concentration of concentrated PM (mean of 206.7 |ig/m3).
3 Lavage protein did not increase, and there were no other indicators of pulmonary injury.
4 No respiratory symptoms or decrements in pulmonary function were found after exposure to
5 CAPs. The 38 human volunteers reported on by Ohio et al. (2000a) were also examined for
6 changes in host defense and immune parameters in BAL and blood (Harder et al., 2001). There
7 were no changes in the number of lymphocytes or macrophages, subcategories of lymphocytes
8 (according to surface marker analysis by flow cytometry), cytokines IL-6 and IL-8, or
9 macrophage phagocytosis. Similarly, there was no effect of concentrated ambient PM exposure
10 on lymphocyte subsets in blood. Thus, a mild inflammatory response to concentrated ambient
11 PM was not accompanied by an effect on immune defenses as determined by lymphocyte or
12 macrophage effects. The increase in neutrophils may represent an adaptive response of the lung
13 to particles, although the presence of activated neutrophils may release biochemical mediators
14 which produce lung injury. Whether this mild inflammatory increase in neutrophils constitutes a
15 biologically significant injury to the lung is an ongoing controversial issue.
16 Other human inhalation studies with CAPs are limited by the small numbers of subjects
17 studied. Petrovic et al. (1999) exposed four healthy volunteers (aged 18 to 40) under resting
18 conditions to filtered air and 3 concentrations of concentrated ambient PM (23 to 124 |ig/m3) for
19 2 hours using a face mask. The exposure was followed by 30 minutes of exercise. No cellular
20 signs of inflammation were observed in induced sputum samples collected at 2 or 24 hours after
21 exposure. There was a trend toward an increase in nasal lavage neutrophils although no
22 statistical significance was presented. The only statistically significant change in pulmonary
23 function was a 6.4% decrease in thoracic gas volume after exposure to 124 |ig/m3 PM versus a
24 5.6% increase after air. A similar, small pilot study has been reported (Gong et al., 2000) in
25 which no changes in pulmonary function or symptoms were observed in four subjects aged 19 to
26 41 after a 2 hour exposure to air or mean concentrations of 148 to 246 |ig/m3 concentrated
27 ambient PM in Los Angeles, CA. Both of these laboratories are currently expanding on these
28 preliminary findings, but no additional data are available at this time.
29 Saldiva et al. (2002) studied the effects on the lungs of CAPs from Boston. The study was
30 designed (1) to determine whether short-term exposures to CAPs cause pulmonary inflammation
31 in normal rats and rats with chronic bronchitis (CB); (2) to identify the site within the lung
June 2003 7-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 parenchyma where CAPs-induced inflammation occurs; and (3) to characterize the component(s)
2 of CAPs significantly associated with development of the inflammatory reaction. Four groups of
3 animals were studied: (1) air treated, filtered air exposed (air-sham); (2) sulfur dioxide treated
4 (CB), filtered air exposed (CB-sham); (3) air treated, CAPs exposed (air-CAPs); and (4) sulfur
5 dioxide treated, CAPs exposed (CB-CAPs). Chronic bronchitis and normal rats were exposed by
6 inhalation either to filtered air or CAPs during 3 consecutive days (5 hours/day). CAPs (as a
7 binary exposure term) and CAPs mass (in regression correlations) induced a significant increase
8 in bronchoalveolar lavage (BAL) neutrophils and in normal and CB animals. Numerical density
9 of neutrophils (Nn) in the lung tissue significantly increased with CAPs in normal animals only.
10 Greater Nn was observed in central, compared with peripheral, regions of the lung. A significant
11 dose-dependent association was found between CAPs components and BAL neutrophils or
12 lymphocytes, but only vanadium and bromine concentrations had significant associations with
13 both BAL neutrophils and Nn in CAPs-exposed groups analyzed together. The authors
14 concluded that (a) short-term exposures to CAPs from Boston induce a significant inflammatory
15 reaction in rat lungs and (b) the reaction is influenced by particle composition.
16
17 7.2.1.2 Diesel Particulate Matter
18 Other controlled human exposures of ambient PM that may be relevant to this discussion
19 were the DPM studies previously examined in detail in separate assessment documents (U.S.
20 Environmental Protection Agency, 2000; Health Effects Institute, 1995). Briefly, the data from
21 work shift studies suggest that the principle noncancer human hazard from exposure to diesel
22 exhaust (DE) includes increased acute sensory and respiratory symptoms (e.g., cough, phlegm,
23 chest tightness, wheezing) that are more sensitive indicators of possible health risks from
24 exposure to DE than pulmonary function decrements. Immunological changes also have been
25 demonstrated under short-term exposure scenarios to either DE or diesel particulate (DPM), and
26 the evidence indicates that these immunological effects are caused by both the non-extractable
27 carbon core and the adsorbed organic fraction of the diesel particle. While noncancer effects
28 from long-term exposure to a high concentration of DPM in several laboratory animal species
29 include pulmonary histopathology and chronic inflammation, noncancer effects in humans from
30 long-term chronic exposure to DPM are not evident. The mode of action of DPM is not
31 completely understood but the effects on the upper respiratory tract, observed in acute studies,
June 2003 7-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 suggest a non-inflammatory irritant response while the effects on the lung, observed in chronic
2 studies, indicate an underlying inflammatory response. Available data suggest that the
3 carbonaceous core of the diesel particle, or metabolites of metal components of the particle, are
4 possible causative agents for the noncancer lung effects which are mediated, at least in part, by a
5 progressive impairment of alveolar macrophage function. The noncancer lung effects occur in
6 response to DPM in several species and occur in rats at doses lower than those inducing particle
7 overload.
8 Diesel particulate matter, therefore, can be relevant to the urban environment, particularly
9 in urban micro-environments with heavy diesel engine traffic. The findings of controlled-
10 exposure studies of DPM are discussed both here and in Section 7.5.3 (Particulate Matter Effects
11 on Allergic Hosts).
12 Pulmonary function and inflammatory markers (as assayed in induced sputum samples or
13 BAL) have been studied in human subjects exposed to either resuspended or freshly generated
14 and diluted DPM. In a controlled human study, Sandstrom and colleagues (Rudell et al., 1994)
15 exposed eight healthy subjects in an exposure chamber to diluted exhaust from a diesel engine
16 for 1 h with intermittent exercise. Dilution of the DE was controlled to provide a median NO2
17 level of approximately 1.6 ppm. Median particle number was 4.3 x 106 /cm3, and median levels
18 of NO and CO were 3.7 and 27 ppm, respectively (particle size and mass concentration were not
19 provided). There were no effects on spirometry or on airway closing volume. Five of eight
20 subjects experienced unpleasant smell, eye irritation, and nasal irritation during exposure. BAL
21 was performed 18 hours after exposure and was compared with a control BAL performed 3
22 weeks prior to exposure. There was no control air exposure. Small yet statistically significant
23 reductions were seen in BAL mast cells, AM phagocytic function, and lymphocyte CD4 to
24 CD8+ cell ratios. A small increase in neutrophils was also observed. These findings suggest
25 that DE may induce mild airway inflammation in the absence of spirometric changes. Although
26 this early study provided important information on the effect of DE exposure in humans, only
27 one exposure level was used, the number of subjects was low, and a limited range of endpoints
28 was reported. Several follow-up studies have been done by the same and other investigators.
29 Rudell et al. (1996) later exposed 12 healthy volunteers to DE for 1 h in an exposure
30 chamber. Light work on a bicycle ergometer was performed during exposure. Random, double-
31 blinded exposures included air, DE, or DE with particle numbers reduced 46% by a particle trap.
June 2003 7-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 The engine used was a new Volvo model 1990, a six-cylinder direct-injection turbocharged
2 diesel with an intercooler, which was run at a steady speed of 900 rpm during the exposures.
3 It is difficult to compare this study with others, because neither exhaust dilution ratios nor
4 particle concentrations were reported. Concentrations of 27-30 ppm CO and of 2.6-2.7 ppm NO,
5 however, suggested DPM concentrations may have equaled several mg/m3. The most prominent
6 symptoms during exposure were irritation of the eyes and nose, accompanied by an unpleasant
7 smell. Both airway resistance and specific airway resistance increased significantly during the
8 exposures. Despite the 46% reduction in particle numbers by the trap, effects on symptoms and
9 lung function were not significantly reduced.
10 A follow-up study on the usefulness of a particle trap confirmed the lack of effect of the
11 filter on DE-induced symptoms (Rudell et al., 1999). In this study, 10 healthy volunteers also
12 underwent BAL 24 hours after exposure. Exposure to DE produced inflammatory changes in
13 BAL, as evidenced by increases in neutrophils and decreases in macrophage phagocytic function
14 in vitro. A 50% reduction in the particle number concentration by the particle trap did not alter
15 these BAL cellular changes. Salvi et al. (1999) exposed healthy human subjects to diluted DE
16 (DPM = 300 |ig/m3 ) for 1 h with intermittent exercise. As reported in the studies by Rudell and
17 Sandstrom (Rudell et al., 1990, 1996, 1999; Blomberg et al., 1998; Salvi et al., 1999) significant
18 increases in neutrophils and B lymphocytes, as well as histamine and fibronectin in airway
19 lavage fluid, were not accompanied by decrements in pulmonary function. Bronchial biopsies
20 obtained 6 h after DE exposure showed a significant increase in neutrophils, mast cells, and
21 CD4+ and CD8+ T lymphocytes, along with upregulation of the endothelial adhesion molecules
22 ICAM-1 and vascular cell adhesion molecule-1 (VCAM-1) and increases in the number of
23 leukocyte function-associated antigen-1 (LFA-1+) in the bronchial tissue. Importantly, extra-
24 pulmonary effects were observed in these subjects. Significant increases in neutrophils and
25 platelets were found in peripheral blood following exposure to DE.
26 Several DE toxicity studies cited in the EPA Health Effects of Diesel Exhaust Health
27 Assessment (2000) compared the effects of whole, unfiltered exhaust to those produced by the
28 gaseous components of the exhaust. Heinrich et al. (1982) compared the toxic effects in
29 hamsters and rats exposed to whole and filtered DE. Exposures were to 3.9 mg/m3 for 7 to
30 8 hrs/day and 5 days/week. Rats exposed for 24 mo to either whole or filtered exhaust exhibited
31 no significant changes in respiratory frequency, respiratory minute volume, compliance or
June 2003 7-23 DRAFT-DO NOT QUOTE OR CITE
-------
1 resistance, as measured by whole body plethysmography, and heart rate. In the hamsters,
2 histological changes (adenomatous proliferations) were seen in the lungs of animals exposed to
3 either whole or filtered exhaust; however, in all groups exposed to the whole exhaust, the
4 number of hamsters exhibiting such lesions was significantly higher than for the corresponding
5 groups exposed to filtered exhaust or clean air. Severity of the lesions was not reported.
6 In a second study Heinrich et al. (1986) and Stober (1986) compared the toxic effects of
7 whole and filtered DE on hamsters, rats, and mice. The test animals (96 per test group) were
8 exposed to 4.24 mg/m3 DPM for 19 hrs/day, 5 days/week for 120 (hamsters and mice) or 140
9 (rats) weeks. Body weights of hamsters were unaffected by either exposure; whereas those of
10 rats and mice were reduced by the whole exhaust but not by the filtered exhaust. Exposure-
11 related higher mortality rates occurred in mice after 2 years of exposure to whole exhaust. After
12 1 year of exposure to whole exhaust, hamsters exhibited increased lung weights, a significant
13 increase in airway resistance, and a nonsignificant reduction in lung compliance. For the same
14 time period, rats exhibited increased lung weights, a significant decrease in dynamic lung
15 compliance, and a significant increase in airway resistance. Test animals exposed to filtered
16 exhaust did not exhibit these effects. In hamsters, filtered exhaust caused no significant
17 histopathological effects in the lung; but whole exhaust caused thickened alveolar septa,
18 bronchiolo-alveolar hyperplasia, and emphysematous lesions. In mice, whole exhaust, but not
19 filtered exhaust, caused multifocal bronchi olo-alveolar hyperplasia, multifocal alveolar
20 lipoproteinosis, and multifocal interstitial fibrosis. In rats, there were no significant
21 morphological changes in the lungs following exposure to filtered exhaust. In rats exposed to
22 whole exhaust, there were severe inflammatory changes in the lungs, thickened alveolar septa,
23 foci of macrophages, crystals of cholesterol, and hyperplastic and metaplastic lesions.
24 Biochemical studies of lung lavage fluids of hamsters and mice indicated that exposure to
25 filtered exhaust caused fewer changes than did exposure to whole exhaust. The latter produced
26 significant increases in lactate dehydrogenase, alkaline phosphatase, glucose-6-phosphate
27 dehydrogenase (G6PDH), total protein, protease (pH 5.1), and collagen. The filtered exhaust
28 had a slight but nonsignificant effect on G6PDH, total protein, and collagen. Similarly,
29 cytological studies showed that while the filtered exhaust had no effect on differential cell
30 counts, the whole exhaust resulted in an increase in leukocytes (161 ± 43.3/|iL versus 55.7 ±
31 12.8/|iL controls), a decrease in AMs (30.0 ± 12.5 versus 51.3 ± 12.5/|iL in the controls), and an
June 2003 7-24 DRAFT-DO NOT QUOTE OR CITE
-------
1 increase in granulocytes (125 ± 39.7 versus 1.23 ± 1.14/|iL in the controls). The differences
2 were significant for each cell type. There was also a small increase in lymphocytes (5.81 ± 4.72
3 versus 3.01 ± 1.23 jiL in the controls).
4 Iwai et al. (1986) exposed rats (24 per group) to whole or filtered DE 8 h/day, 7 days/week
5 for 24 mo. The whole exhaust was diluted to a concentration of 4.9 or 1.6 mg/m3 DPM. Body
6 weights in the whole exhaust group began to decrease after 6 mo and in both exposed groups
7 after 18 mo. Lung-to-body weight ratios of the rats exposed to the whole exhaust showed a
8 significant increase (p < 0.01) after 12 mo in comparison with control values. Spleen-to-body
9 weight ratios of both exposed groups were higher than control values after 24 mo. After 6 mo of
10 exposure to whole exhaust, DPM accumulated in AMs, and Type II cell hyperplasia was
11 observed. After 2 years of exposure, the alveolar walls had become fibrotic with mast cell
12 infiltration and epithelial hyperplasia. In rats exposed to filtered exhaust, after 2 years there
13 were only minimal histologic changes in the lungs, with slight hyperplasia and stratification of
14 bronchiolar epithelium and infiltration of atypical lymphocytic cells in the spleen.
15 Brightwell et al. (1986) evaluated the toxic effects of whole and filtered DE on rats and
16 hamsters. Three exhaust dilutions were tested, producing concentrations of 0.7, 2.2, and
17 6.6 mg/m3 DPM. The test animals (144 rats and 312 hamsters per exposure group) were exposed
18 for five 16-h periods per week for 2 years. The four exhaust types were gasoline, gasoline
19 catalyst, diesel, and filtered diesel. The results presented were limited to statistically significant
20 differences between exhaust-exposed and control animals. The inference from the discussion
21 section of the paper was that there was a minimum of toxicity in the animals exposed to filtered
22 DE: "It is clear from the results presented that statistically significant differences between
23 exhaust-exposed and control animals are almost exclusively limited to animals exposed to either
24 gasoline or unfiltered diesel exhaust."
25 Heinrich et al. (1995) exposed female NMRI and C57BL/6N mice to a DE dilution that
26 resulted in a DPM concentration of 4.5 mg/m3 and to the same dilution after filtering to remove
27 the particles. This study is focused on the carcinogenic effects of DPM exposure, and
28 inadequate information was presented to compare noncancer effects in filtered versus unfiltered
29 exhaust.
30 A comparison of the toxic responses in laboratory animals exposed to whole exhaust or
31 filtered exhaust containing no particles demonstrates across studies that, when the exhaust is
June 2003 7-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 sufficiently diluted to limit the concentrations of gaseous irritants (NO2 and SO2 ), irritant vapors
2 (aldehydes), CO, or other systemic toxicants, the diesel particles are the prime etiologic agents of
3 noncancer health effects, although additivity or synergism with the gases cannot be ruled out.
4 These toxic responses are both functional and pathological and represent a cascading sequelae of
5 lung pathology based on concentration and species. The diesel particles plus gas exposures
6 produced biochemical and cytological changes in the lung that are much more prominent than
7 those evoked by the gas phase alone. Such marked differences between whole and filtered DE
8 are also evident from general toxicological indices, such as decreases in body weight and
9 increases in lung weights, pulmonary function measurements, and pulmonary histopathology
10 (e.g., proliferative changes in Type II cells and respiratory bronchiolar epithelium fibrosis).
11 Hamsters, under equivalent exposure regimens, have lower levels of retained DPM in their lungs
12 than rats and mice and, consequently, less pulmonary function impairment and pulmonary
13 pathology. These differences may result from lower DPM inspiration and deposition during
14 exposure, greater DPM clearance, or lung tissue less susceptible to the cytotoxicity of deposited
15 DPM.
16 In a follow-up investigation of potential mechanisms underlying the DE-induced airway
17 leukocyte infiltration, Salvi et al. (2000) exposed healthy human volunteers to diluted DE on two
18 separate occasions for 1 h each, in an exposure chamber. Fiber-optic bronchoscopy was
19 performed 6 h after each exposure to obtain endobronchial biopsies and bronchial wash (BW)
20 cells. These workers observed that diesel exhaust (DE) exposure enhanced gene transcription of
21 interleukin-8 (IL-8) in the bronchial tissue and BW cells and increased growth-regulated
22 oncogene-cc protein expression and IL-8 in the bronchial epithelium; there was also a trend
23 toward an increase in interleukin-5 (IL-5) mRNA gene transcripts in the bronchial tissue.
24 Nightingale et al. (2000) have reported inflammatory changes in healthy volunteers
25 exposed to 200 |ig/m3 resuspended DPM under resting conditions in a double-blinded study.
26 Small but statistically significant increases in neutrophils and myeloperoxidase (an index of
27 neutrophil activation) were observed in sputum samples induced 4 hours after exposure to DPM
28 in comparison to air. Exhaled carbon monoxide was measured as an index of oxidative stress
29 and was found to increase maximally at 1 hour after exposure. These biochemical and cellular
30 changes occurred in the absence of any decrements in pulmonary function, thus confirming that
31 markers of inflammation are more sensitive than pulmonary function measurements.
June 2003 7-26 DRAFT-DO NOT QUOTE OR CITE
-------
1 Because of the considerable concern about inhalation of ambient particles by sensitive
2 subpopulations, Sandstrom's laboratory also studied the effect of a 1 hour exposure to 300 |ig/m3
3 DPM on 14 atopic asthmatics with stable disease and on inhaled corticosteroid treatment
4 (Nordenhall et al., 2001). At 6 hours after exposure, there was a significant increase in IL-6 in
5 induced sputum. At 24 hours after exposure, there was a significant increase in the nonspecific
6 airway responsiveness to inhaled methacholine. Although the exposure level was high relative
7 to ambient PM levels, these findings may be important in terms of supporting epidemiologic
8 evidence for increased asthma morbidity associated with episodic exposure to ambient PM.
9 The role of antioxidant defenses in protecting against acute diesel exhaust exposure has
10 also been studied. Blomberg et al. (1998) investigated changes in the antioxidant defense
11 network within the respiratory tract lining fluids of human subjects following diesel exhaust
12 exposure. Fifteen healthy, nonsmoking, asymptomatic subjects were exposed to filtered air or
13 diesel exhaust (DPM 300 mg/m3) for 1 h on two separate occasions at least 3 weeks apart. Nasal
14 lavage fluid and blood samples were collected prior to, immediately after, and 5.5 h post-
15 exposure. Bronchoscopy was performed 6 h after the end of diesel exhaust exposure. Nasal
16 lavage ascorbic acid concentration increased tenfold during diesel exhaust exposure, but returned
17 to basal levels 5.5 h post-exposure. Diesel exhaust had no significant effects on nasal lavage uric
18 acid or GSH concentrations and did not affect plasma, bronchial wash, or bronchoalveolar
19 lavage antioxidant concentrations or malondialdehyde or protein carbonyl concentrations. The
20 authors concluded that the acute increase in ascorbic acid in the nasal cavity induced by diesel
21 exhaust may help prevent further oxidant stress in the upper respiratory tract of healthy
22 individuals.
23
24 7.2.1.3 Complex Combustion-Related Particles
25 Because emission sources contribute to the overall ambient air particulate burden (Spengler
26 and Thurston, 1983), many of the studies investigating the response of laboratory animals to
27 particle exposures have used complex combustion-related particles (see Table 7-2).
28 For example, the residual oil fly ash (ROFA) samples used in toxicological studies have been
29 collected from a variety of sources, e.g., boilers, bag houses used to control emissions from
30 power plants, and from particles emitted downstream of such collection devices. ROFA has a
31 high content of water soluble sulfate and metals, accounting for 82 to 92% of water-soluble
June 2003 7-27 DRAFT-DO NOT QUOTE OR CITE
-------
1 mass, while the water-soluble mass fraction in ambient air varies from low teens to more than
2 60% (Costa and Dreher, 1997; Prahalad et al., 1999). More than 90% of the metals in ROFA are
3 transition metals; whereas these metals are only a small subtraction of the total ambient PM
4 mass. Transition metals generate reactive oxygen species and are relevant to understanding the
5 mechanisms of toxicity and the components contributing to the toxic responses. Thus, the dose
6 of bioavailable metal that is delivered to the lung when ROFA is instilled into a laboratory
7 animal can be orders of magnitude greater than an ambient PM dose, even under a worst-case
8 scenario.
9 Intratracheal instillation of various doses of ROFA suspension has been shown to produce
10 severe inflammation, an indicator of pulmonary injury that includes recruitment of neutrophils,
11 eosinophils, and monocytes into the airway. The biological effects of ROFA in rats have been
12 shown to depend on aqueous teachable chemical constituents of the particles (Dreher et al.,
13 1997; Kodavanti et al., 1997b). A leachate prepared from ROFA, containing predominantly Fe,
14 Ni, V, Ca, Mg, and sulfate, produced similar lung injury to that induced by the complete ROFA
15 suspension (Dreher et al., 1997). Depletion of Fe, Ni, and V from the ROFA leachate eliminated
16 its pulmonary toxicity. Correspondingly, minimal lung injury was observed in animals exposed
17 to saline-washed ROFA particles. A surrogate transition metal sulfate solution containing Fe, V,
18 and Ni largely reproduced the lung injury induced by ROFA. Interestingly, ferric sulfate and
19 vanadium sulfate antagonized the pulmonary toxicity of nickel sulfate. Interactions between
20 different metals and the acidity of PM were found to influence the severity and kinetics of lung
21 injury induced by ROFA and its soluble transition metals.
22 To further investigate the response to ROFA with differing metal and sulfate composition,
23 male Sprague-Dawley rats (60 days old) were intratracheally instilled with ROFA (2.5 mg/rat) or
24 metal sulfates (iron -0.54 jimole/rat, vanadium -1.7 jimole/rat, and nickel -1.0 jimole/rat,
25 individually or in combination) (Kodavanti et al., 1997b). Transition metal sulfate mixtures
26 caused less injury than ROFA or Ni alone, suggesting metal interactions. This study also
27 showed that V-induced effects were less severe than that of Ni and were transient. Ferric sulfate
28 was least pathogenic. Cytokine gene expression was induced prior to the pathology changes in
29 the lung, and the kinetics of gene expression suggested persistent injury by nickel sulfate.
30 Another study by the same investigators was performed using 10 different ROFA samples
31 collected at various sites within a power plant burning residual oil (Kodavanti et al., 1998a).
June 2003 7-28 DRAFT-DO NOT QUOTE OR CITE
-------
1 Animals received intratracheal instillations of either saline (control), or a saline suspension of
2 whole ROFA (< 3.0 |im MMAD for all ground PM) at three doses (0.833, 3.33, or 8.33 mg/kg).
3 This study showed that ROFA-induced PMN influx was associated with its water-leachable V
4 content; but protein leakage was associated with water-leachable Ni content. ROFA-induced
5 in vitro activation of alveolar macrophages (AMs) was highest with ROFA containing teachable
6 V but not with Ni plus V, suggesting that the potency and the mechanism of pulmonary injury
7 may differ between emissions containing bioavailable V and Ni.
8 Other studies have shown that soluble metal components play an important role in the
9 toxicity of emission source particles. Gavett et al. (1997) investigated the effects of two ROFA
10 samples of equivalent diameters, but having different metal and sulfate content, on pulmonary
11 responses in Sprague-Dawley rats. ROFA sample 1 (Rl) (the same emission particles used by
12 Dreher et al. [1997]) had approximately twice as much saline-leachable sulfate, nickel, and
13 vanadium, and 40 times as much iron as ROFA sample 2 (R2); whereas R2 had a 31-fold higher
14 zinc content. Rats were instilled with suspensions of 2.5 mg R2 in 0.3 mL saline, the
15 supernatant of R2 (R2s), the supernatant of 2.5 mg Rl (Rls), or saline only. By 4 days after
16 instillation, 4 of 24 rats treated with R2s or R2 had died. None treated with Rls or saline died.
17 Pathological indices, such as alveolitis, early fibrotic changes, and perivascular edema, were
18 greater in both R2 groups. In surviving rats, baseline pulmonary function parameters and airway
19 hyperreactivity to acetylcholine were significantly worse in the R2 and R2s groups than in the
20 Rls groups. Other than BAL neutrophils, which were significantly higher in the R2 and R2s
21 groups, no other inflammatory cells (macrophages, eosinophils, or lymphocytes) or biochemical
22 parameters of lung injury were significantly different between the R2 and R2s groups and the
23 Rl s group. Although (a) soluble forms of zinc had been found in guinea pigs to produce a
24 greater pulmonary response than other sulfated metals (Amdur et al., 1978) and (b) the level of
25 zinc was 30-fold greater in R2 than Rl, the precise mechanisms by which zinc may induce such
26 responses are unknown. Still, these results show that the composition of soluble metals and
27 sulfate is critical in the development of airway hyperractivity and lung injury produced by
28 ROFA, albeit at very high instilled doses.
29 Dye et al. (1997) pretreated rats with an intraperitoneal injection of 500 mg/kg
30 dimethylthiourea (DMTU) or saline, followed 30 min later by intratracheal instillation of either
31 acidic saline (Ph = 3.3) or an acidified suspension of ROFA (500 |ig/rat). Dimethylthiourea
June 2003 7-29 DRAFT-DO NOT QUOTE OR CITE
-------
1 reduces the activity of the reactive oxygen species. The systemic administration of DMTU
2 impeded development of the cellular inflammatory response to ROFA but did not ameliorate
3 biochemical alterations in BAL fluid. In a subsequent study, it was determined that oxidant
4 generation, possibly induced by soluble vanadium compounds in ROFA, is responsible for the
5 subsequent rat tracheal epithelial cells gene expression, inflammatory cytokine production
6 (MIP-2 and IL-6), and cytotoxicity (Dye et al., 1999).
7 In addition to transition metals, other components in fly ash also may cause lung injury.
8 The effects of arsenic compounds in coal fly ash or copper smelter dust on the lung integrity and
9 on the ex vivo release of TNFcc by alveolar phagocytes were investigated by Broeckaert et al.
10 (1997). Female Naval Medical Research Institute (NMRI) mice were instilled with different
11 particles normalized for the arsenic content (20 |ig/kg body weight [i.e., 600 ng arsenic/mouse])
12 and the particle load (100 mg/kg body weight [i.e., 3 mg/mouse]). Mice received tungsten
13 carbide (WC) alone, coal fly ash (CFA) alone, copper smelter dust (CMP) mixed with WC, and
14 Ca3(AsO4)2 mixed with WC (see Table 7-2 for concentration details). Copper smelter dust
15 caused a severe but transient inflammatory reaction; whereas a persisting alveolitis (30 days
16 postexposure) was observed after treatment with coal fly ash. In addition, TNFcc production in
17 response to lipopolysaccharide (LPS) by alveolar phagocytes were significantly inhibited at day
18 1 but was still observed at 30 days after administration of CMP and CFA. Although arsenic was
19 cleared from the lung tissue 6 days after Ca3(AsO4)2 administration, a significant fraction
20 persisted (10 to 15% of the arsenic administered) in the lung of CMP- and CFA-treated mice at
21 Day 30. It is possible that suppression of TNF-cc production is dependent upon the slow
22 elimination of the particles and their metal content from the lung.
23 In summary, intratracheally instilled high doses of ROFA produced acute lung injury and
24 inflammation. Water soluble metals in ROFA appear to play a key role in the acute effects of
25 instilled ROFA through the production of reactive oxygen species. Although studies done with
26 ROFA clearly show that combustion-generated particles with a high metal content can cause
27 substantial lung injury, there are still insufficient data to extrapolate the high dose effects to the
28 low levels of particle-associated transition metals in ambient PM.
29
30
June 2003 7-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.2.2 Acid Aerosols
2 There have been extensive studies of the effects of controlled exposures to aqueous acid
3 aerosols on various aspects of lung function in humans and laboratory animals. Many of these
4 studies were reviewed in the 1996 PM AQCD (U.S. Environmental Protection Agency 1996a)
5 and in the Acid Aerosol Issue Paper (U.S. Environmental Protection Agency, 1989); some of the
6 more recent studies are summarized in this document (Table 7-4). Methodology and
7 measurement methods for controlled human exposure studies have been reviewed elsewhere
8 (Folinsbeeetal., 1997).
9 The studies summarized in the 1996 PM AQCD illustrate that aqueous acidic aerosols have
10 minimal effects on symptoms and mechanical lung function in young healthy adult volunteers at
11 concentrations as high as 1000 |ig/m3. However, at concentrations as low as 100 |ig/m3, acid
12 aerosols can alter mucociliary clearance. Brief exposures (< 1 h) to low concentrations
13 (=100 |ig/m3) may accelerate clearance while longer (multihour) exposures to higher
14 concentrations (> 100 |ig/m3) can depress clearance. Asthmatic subjects appear to be more
15 sensitive to the effects of acidic aerosols on mechanical lung function. Responses have been
16 reported in adolescent asthmatics at concentrations as low as 68 |ig/m3, and modest
17 bronchoconstriction has been seen in adult asthmatics exposed to concentrations >400 |ig/m3, but
18 the available data are not consistent.
19 Acid aerosol exposure in humans (1000 |ig/m3 H2SO4) did not result in airway
20 inflammation (Frampton et al., 1992), and there was no evidence of altered macrophage host
21 defenses. Zelikoff et al. (1997) compared the responses of rabbits and humans exposed to
22 similar concentrations of H2SO4 aerosol. For both rabbits and humans, there was no evidence of
23 PMN infiltration into the lung and no change in BAL fluid protein level, although there was an
24 increase in LDH in rabbits but not in humans. Macrophages showed less antimicrobial activity
25 in rabbits; insufficient data were available for humans. Macrophage phagocytic activity was
26 slightly reduced in rabbits but not in humans. Superoxide production by macrophages was
27 somewhat depressed in both species. No respiratory effects of long-term exposure to acid
28 aerosol were found in dogs (Heyder et al., 1999). Thus, recent studies do not provide any
29 additional evidence clearly demonstrating that relevant concentrations of aqueous acid aerosols
30 contribute to the acute cardiopulmonary effects of ambient PM.
31
June 2003 7-31 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-4. RESPIRATORY EFFECTS OF ACID AEROSOLS IN HUMANS AND LABORATORY ANIMALS
to
o
o
OJ
^j
OJ
to
o
|7d
l-rj
H
6
0
!2
O
H
0
Species, Gender,
Strain Age, etc.
Dogs, beagle,
healthy;
n=16
Humans,
asthmatic; 13 M,
11F
Rats, female,
Fischer 344;
Guinea Pigs,
female, Hartley
Humans, healthy
nonsmokers;
10 M, 2 F; 21-37
years old
Rabbits, New
Zealand white
Humans, healthy
nonsmokers;
Exposure
Particle Technique
Neutral Inhalation
sulfite
aerosol
Acidic Inhalation
sulfate
aerosol
H2SO4 Inhalation
aerosol by
NH+4/SCT24 face mask
aerosol
H2SO4 Inhalation
aerosol
H2SO4 Inhalation
aerosol
H2SO4 Inhalation
Concentrati
on Particle Size
1.5 mg/m3 1.0 um
MMAD
ag = 2.2
5.7 mg/m3 1.1 um
MMAD
ag = 2.0
500 ug/m3 9 um
MMAD
7 um
MMAD
94 mg/m3 0.80 ±1.89
43 mg/m3 ag
0.93 ±2.11
ag
1,000 ug/m3 0.8-0.9 um
MMAD
1,000 ug/m3
Exposure
Duration Effects of Particles
16.5 h/day Long-term exposure to particle-associated
for 13 mo sulfur and hydrogen ions caused only subtle
respiratory responses and no change in lung
pathology.
6 h/day for
13 mo
1 h Exposure to simulated natural acid fog did not
induce bronchoconstriction nor change
bronchial responsiveness in asthmatics.
4h Acid aerosol increased surfactant film
compressibility in guinea pigs.
3 h No inflammatory responses; LDH activity in
BAL was elevated. Effect on bacterial killing
by macrophages was inconclusive; latex
particle phagocytosis was reduced 28%.
2 h No inflammatory response; antibody mediated
cytotoxicity of AM increased by H2SO4; no
alterations in antimicrobial defense.
Reference
Heyder et al.
(1999)
Leduc et al.
(1995)
Lee et al.
(1999)
Frampton
et al. (1992)
Zelikoff
et al. (1997)
C 12m, 20-39 years
0
H
W
O
old
H2SO4 = Sulfuric
acid
^ BAL = Bronchoalveolar lavage
O
HH
H
W
LDH = Lactate dehydrogenase
MMAD = Mass median aerodynamic diameter
MMD = Mass median diameter
ag = Geometric standard deviation
-------
1 7.2.3 Metal Particles, Fumes, and Smoke
2 Data from occupational and laboratory animal studies reviewed in the previous criteria
3 document (U. S. Environmental Protection Agency, 1996a) indicated that acute exposures to
4 very high levels (hundreds of |ig/m3 or more) or chronic exposures to lower levels (as low as
5 15 |ig/m3) of metallic particles could have an effect on the respiratory tract. Therefore, it was
6 concluded on the basis of data available at that time that the metals at typical concentrations
7 present in the ambient atmosphere (1 to 14 |ig/m3) were not likely to have a significant acute
8 effect in healthy individuals. The metals include arsenic, cadmium, copper, nickel, vanadium,
9 iron, and zinc. Other metals found at concentrations less than 0.5 |ig/m3 were not reviewed in
10 the previous criteria document. However, more recently published data from high-dose
11 laboratory animal studies tend to indicate that particle-associated metals are among likely
12 potential candidates for inducing adverse effects attributed to ambient PM.
13 Controlled human exposure studies have been performed with particles other than acid
14 aerosols (details are in Table 7-5a,b). Controlled inhalation exposure studies to high
15 concentrations of two different fume particles, MgO and ZnO, demonstrate the differences in
16 response based on particle metal composition (Kuschner et al., 1997; Kuschner et al., 1995).
17 Up to 6400 mg/m3/min cumulative dose of MgO had no effect on lung function (spirometry,
18 DLCO), symptoms of metal fume fever, or changes in inflammatory mediators or cells recovered
19 by BAL. However, lower concentrations of ZnO fume (166 to 1110 mg/m3/min) induced a
20 neutrophilic inflammatory response in the airways 20 h postexposure. Lavage fluid PMNs,
21 TNF-cc, and IL-8 were increased by ZnO exposure. Although the concentrations used in these
22 exposure studies exceed ambient levels by more than 1000-fold, the absence of a response to an
23 almost 10-fold higher concentration of MgO compared with ZnO indicates that differential metal
24 composition, in addition to particle size (ultrafine/fine), is likely an important determinant of
25 observed health responses to inhaled ambient PM.
26 Several metals (e.g., zinc, chromium, cobalt, copper, and vanadium) have been shown to
27 stimulate cytokine release in cultured human pulmonary cells. Boiler makers, exposed
28 occupationally to -400 to 500 |ig/m3 of fuel oil ash, containing high levels of soluble metals,
29 showed acute nasal inflammatory responses characterized by increased myeloperoxidase (MPO)
30 and IL-8 levels; these changes were associated with increased vanadium levels in the upper
31
June 2003 7-33 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 7-5a. RESPIRATORY EFFECTS OF INSTILLED METAL PARTICLES,
FUMES, AND SMOKE IN HUMAN SUBJECTS AND LABORATORY ANIMALS
Species, Gender,
Strain, Age, etc.
Humans, healthy
nonsmokers;
12 M, 4 F;
18-3 5 years old
Humans, healthy
nonsmokers; 27 M,
7F;
20-36 years old
Mice, Swiss
Rats, Fischer 344.
(250 g)
Mice, NMRI;
Mouse peritoneal
macrophage
Rats, M, F344,
175-225 g
Rats, M. F344,
175-225 g
Rats
Exposure
Particle Technique
Colloidal Bronchial
iron oxide instillation
Fe2O3 Intrapulmonary
instillation
EHC-93 Intratracheal
soluble instillation
metal salts
Fe2O3 Intratracheal
instillation
MnO2 Intratracheal
instillation;
in vitro
TiO2 Intratracheal
inhalation and
Intratracheal
instillation
TiO2 Intratracheal
inhalation and
Intratracheal
instillation
NaVO3 Intratracheal
VOSO4 instillation
V205
Concentration
5 mg in 10 mL
3 x 108 microspheres in
10 mL saline.
1 mg in 0.1 ml
7.7 x 107 microspheres
in 5 mL saline
0.037,0.12,0.75,
2.5 mg/animal
Inhalation at 125 mg/m3
for 2 h; Instillation at
500 ug for fine,
750 ug for ultrafine
Inhalation at 125 mg/m3
for 2 h; Instillation at
500 ug for fine, 750 ug
for ultrafine
21 or210 ugV/kg
(NaVO3, VOSO4
soluble)
42 or 420 ug V/kg
(V2OS) less soluble
Particle Size Exposure Duration
2.6 |im 1, 2, and 4 days after
instillation
2.6 urn N/A
0.8 ±0.4 urn 3 days
2.6 urn N/A
surface area of Sacrificed at 5 days
0.16,0.5; 17,
62 mVg
Fine: 250 nm Inhalation exposure,
Ultrafine: 2 h;
21 nm sacrificed at 0, 1, 3,
and 7 days
postexposure for
both techniques
Fine: 250 nm Inhalation exposure,
Ultrafine: 2 h; sacrificed at 0,
21 nm 1,3, and 7 days
postexposure for
both techniques
N/A 1 h or 10 days
following instillation
Effect of Particles
L-ferritin increased after iron oxide particle
exposure; transferrin was decreased. Both
lactoferrin and transferrin receptors were increased.
Transient inflammation induced initially
(neutrophils, protein, LDH, IL-8) was resolved by 4
days postinstillation.
Solution containing all metal salts (Al, Cu, Fe, Pb,
Mg, Ni, Zn) or ZnCl alone increased BAL
inflammatory cells and protein.
Transient inflammation at 1 day postinstillation.
LDH, protein and cellular recruitment increased
with increasing surface area; freshly ground particles
had enhanced cytotoxicity.
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.
Reference
Ohio et al.
(1998b)
Lay et al.
(1998)
Adamson
et al. (2000)
Lay et al.
(1998)
Lison et al.
(1997)
Osier and
Oberdorster
(1997)
Osier et al.
(1997)
Pierce et al.
(1996)
CdO = Cadmium oxide
Fe2O3 = Iron oxide
MgO = Magnesium oxide
MnO2 = Manganese oxide
NaVO3 =
TiO2 = Titanium oxide
VOSO4 = Vanadyl sulfate
V,O, = Vanadium oxid
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
-------
•— ^
3
to
o
o
OJ
OJ
o
H
6
0
0
H
0
0
H
W
O
O
H
W
TABLE 7-5b. B
Species, Gender, Strain,
Age, etc. Particle
Rats, SD; 60 days old VSO4
NiSO4
Rats, WISTAR Furth; CdO
7-week-old, Fume
Mice, C57BL6 and
DBA3NCR
Humans, boilermakers ROFA
(1 8 M), 26-61 years old,
and utility worker
controls (11 M),
30-55 years old
Humans, vanadium plant V2O5
workers; 40 M; 19-60
years old
Humans, healthy MgO
nonsmokers;
4 M, 2 F;
2 1-43 years old
Humans, healthy Fe2O3
nonsmokers; 8 M, 8 F;
18-34 years old
CdO = Cadmium oxide
Fe2O3 = Iron oxide
MgO = Magnesium oxide
MnO2 = Manganese oxide
NaV03 =
TiO2 = Titanium oxide
VOSO4 = Vanadyl sulfate
V2O5 = Vanadium oxid
RESPIRATORY EFFECTS OF INHALED METAL PARTICLES, FUMES, AND SMOKE
IN HUMANS AND LABORATORY ANIMALS
Exposure
Technique Concentration Particle Size
Inhalation 0.3 - 2.4 mg/m3 N/A
Nose-only 1.04 mg/m3 CMD = 0.008
inhalation Rats dose = 18.72 ug umog=l.l
Mouse dose = 4.59 ug
Inhalation of 0.4-0.47 mg/m3 10 urn
fuel-oil ash 0.1-0.13 mg/m3
Inhalation < 0.05-1.53 mg/m3 N/A
Inhalation 5.8-230 mg/m3 99% < 1.8 urn
29% < 0.1 urn
Inhalation 12. 7 mg/m3 1.5 um
og = 2.1
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
Exposure
Duration Effect of Particles
6h/day x 4 days 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.
1 x 3 h Mice created more metallothionein than rats, which may
be protective of tumor formation.
6 weeks Exposure to fuel-oil ash resulted in acute upper airway
inflammation, possibly mediated by increased IL-8 and
PMNs.
Variable 12/40 workers had bronchial hyperreactivity that
persisted in some for up to 23 mo.
15-45 min 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.
30 min No significant difference in 98mTc-DTPA clearance
half-times, DLCO, or spirometry
Reference
Campen
etal. (2001)
McKenna
et al. (1998)
Woodin
et al. (1998)
Irsigler et al.
(1999)
Kuschner
et al. (1997)
Lay et al.
(2001)
-------
1 airway (Woodin et al., 1998). Also, Irsigler et al. (1999) reported that V2O5 can induce asthma
2 and bronchial hyperreactivity in exposed workers.
3 Autopsy data suggest that chronic exposure to urban air pollution leads to an increased
4 retention of metals in human tissues. A comparison of autopsy cases in Mexico City from the
5 1950s with the 1980s indicated substantially higher (5- to 20-fold) levels of Cd, Co, Cu, Ni, and
6 Pb in lung tissue from the 1980s (Fortoul et al., 1996). Similar studies have examined metal
7 content in human blood and lung tissue (Tsuchiyama et al., 1997; Osman et al., 1998), with
8 similar results.
9 Iron is the most abundant of the elements capable of catalyzing oxidant generation and is
10 also present in ambient urban particles. Lay et al. (1998) and Ohio et al. (1998b) tested the
11 hypothesis that the human respiratory tract will attempt to diminish the added, iron-generated
12 oxidative stress. They examined cellular and biochemical responses of human subjects instilled,
13 via the intrapulmonary route, with a combination of iron oxyhydroxides that introduced an
14 oxidative stress to the lungs. Saline alone and iron-containing particles suspended in saline were
15 instilled into separate lung segments of human subjects. Subjects underwent bronchoalveolar
16 lavage at 1 to 91 days after instillation of 2.6-|im diameter iron oxide (approximately 5 mg or
17 2.1 x 108 particles) agglomerates. Lay and colleagues found iron-oxide-induced inflammatory
18 responses in both the alveolar fraction and the bronchial fraction of the lavage fluid at 1 day
19 postinstillation. Lung lavage 24 h after instillation revealed decreased transferrin concentrations
20 and increased ferritin and lactoferrin concentrations, consistent with a host-generated response to
21 decrease the availability of catalytically reactive iron (Ohio et al., 1998b). Normal iron
22 homeostasis returned within 4 days of the iron particle instillation. The same iron oxide
23 preparation, which contained a small amount of soluble iron, produced similar pulmonary
24 inflammation in rats. In contrast, instillation of rats with two iron oxide preparations that
25 contained no soluble iron failed to produce injury or inflammation, thus suggesting that soluble
26 iron was responsible for the observed intrapulmonary changes.
27 In a subsequent inhalation study, Lay et al. (2001) studied the effect of iron oxide particles
28 on lung epithelial cell permeability. Healthy, nonsmoking human subjects inhaled 12.7 mg/m3
29 low- and high-solubility iron oxide particles (MMAD =1.5 jim and og = 2.1) for 30 minutes.
30 Neither pulmonary function nor alveolar epithelial permeability, as assessed by pulmonary
31 clearance of technetium-labeled DPTA, was changed at 0.5 or 24 hours after exposure to either
June 2003 7-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 type of iron oxide particle. Because the exposure concentration was so high, the data suggest
2 that iron may play little role in the adverse effects of ambient, urban PM. Ohio et al. (2001)
3 have reported a case study, however, in which acute exposure to oil fly ash from a domestic oil-
4 burning stove produced diffuse alveolar damage, difficulty in breathing, and symptoms of
5 angina. While steroid treatment led to rapid improvement in symptoms and objective
6 measurements, this report suggests that the high metal content of oil fly ash can alter the
7 epithelial cell barrier in the alveolar region.
8
9 7.2.4 Ambient Bioaerosols
10 Ambient bioaerosols include fungal spores, pollen, bacteria, viruses, endotoxins, and plant
11 and animal debris. Such biological aerosols can produce various health effects including
12 irritation, infection, hypersensitivity, and toxic response. Bioaerosols present in the ambient
13 environment have the potential to cause disease in humans under certain conditions. However, it
14 was concluded in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) that
15 bioaerosols, at the concentrations present in the ambient environment, would not likely
16 contribute to the observed effects of PM on human mortality and morbidity reported in PM
17 epidemiologic studies. Moreover, bioaerosols generally represent a rather small fraction of the
18 measured urban ambient PM mass and are typically present even at lower concentrations during
19 the winter months when notable ambient PM effects have been demonstrated. Bioaerosols tend
20 to be in the coarse fraction of PM, but some bioaerosols, including nonagglomerated bacteria
21 and fragmented pollens, are found in the fine fraction.
22 More recent inhalation studies on ambient bioaerosols are summarized in Table 7-6. The
23 majority of these studies have focused on endotoxin, because little research on other bioaerosol
24 components has been conducted. In vitro studies on particle-associated endotoxin are discussed
25 in Section 7.5.2.2. Endotoxin, a cell wall component of gram negative bacteria, is ubiquitous in
26 the environment. Although there is strong evidence that inhaled endotoxin plays a major role in
27 the toxic effects of bioaerosols encountered in the work place (Vogelzang et al., 1998; Castellan
28 et al., 1984, 1987), it is not clear whether ambient concentrations of endotoxin are sufficient to
29 produce toxic pulmonary or systemic effects in healthy or compromised individuals.
30 Michel et al. (1997) examined the dose-response relationship to inhaled lipopolysaccharide
31 (LPS: the purified derivative of endotoxin) in normal healthy volunteers exposed to 0, 0.5, 5, and
June 2003 7-37 DRAFT-DO NOT QUOTE OR CITE
-------
c
3
oo
TABLE 7-6. CONTROLLED EXPOSURE STUDIES OF RESPIRATORY EFFECTS OF
INHALED AMBIENT BIO AEROSOLS
to
2 Species, Gender, Exposure Particle
Strain, Age, etc. Particle Technique Concentration Size
Rats, Fischer 344, LPS Inhalation 70 EU 0.72 um
8 weeks to (endotoxin) ag=1.6
22 months old,
N = 3 /group
Humans, healthy; LPS Inhalation 0.5 ug 1-4 um
5 M, 4 F, 24 to (endotoxin) 5.0 ug MMAD
50 years of age 50 ug
^i
OJ
Exposure
Duration Effect of Particles
12 min Significant increase in PMNs in
bronchoalveolar lavage (BAL) in LPS
exposed animals. LPS significantly affected
the reactive oxygen species activity in BAL.
Effects were age-dependent.
30 min Significant decrease in PMN luminol-
enhanced chemiluminescence with 0.5 ug
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 ug
LPS.
Reference
Elder et al.
(2000a,b)
Michel
etal. (1997)
Humans, healthy;
32 M, 32 F, 16 to
50 years of age
Humans, pig
farmers,
82 symptomatic
and
89 asymptomatic
n=171
Indoor pool
water spray
Dust
Endotoxin
Inhalation
N/A
0.1-7.5 um
N/A
Inhalation 2.63 mg/m3
ag=1.3
105 ng/m3
ag=1.5
N/A
5h/day
average
lifetime
exposure
Recurring outbreaks of pool-associated Rose et al.
granulomatous pneumonitis (n = 33); case (1998)
patients had higher cumulative work hours.
Analysis indicated increased levels of
endotoxin in pool air and water.
Large decline in FEVj (73 mL/year) and FVC Vogelzang
(55 mL/year) associated with long-term et al. (1998)
average exposure to endotoxin.
Humans, potato Endotoxin
plant workers, low
exposures (37 M),
high exposures
(20 M)
Inhalation 21.2 EU/m3 low N/A
ag= 1.6
55.7 EU/m3 high
ag = 2.1
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
8h.
Zock et al.
(1998)
-------
1 50 jig of LPS. Inhalation of 5 or 50 jig of LPS resulted in increased PMNs in blood and sputum
2 samples. At the higher concentration, a slight (3%) but not significant decrease in FEVj was
3 observed. Cormier et al. (1998) reported an approximate 10% decline in FEVj and an increase
4 in methacholine airway responsiveness after a 5-h exposure inside a swine containment building.
5 This exposure induced significant neutrophilic inflammation in both the nose and the lung.
6 Although these exposures are massive compared to endotoxin levels in ambient PM in U.S.
7 cities, these studies serve to illustrate the effects of endotoxin and associated bioaerosol material
8 in healthy, nonsensitized individuals.
9 Some health effects have been observed after occupational exposure to complex aerosols
10 containing endotoxin at concentrations relevant to ambient levels. Zock et al. (1998) reported a
11 decline in FEVj (~ 3%) across a shift in a potato processing plant with up to 56 endotoxin units
12 (EU)/m3 in the air. Rose et al. (1998) reported a high incidence (65%) of BAL lymphocytes in
13 lifeguards working at a swimming pool where endotoxin levels in the air were on the order of
14 28 EU/m3. Although these latter two studies may point towards pulmonary changes at low
15 concentrations of airborne endotoxin, it is not possible to rule out the contribution of other
16 agents in these complex organic aerosols. The contribution of endotoxin to the toxicity of
17 ambient PM has been studied in vitro, and these studies provide some evidence that endotoxin
18 contaminates in ambient PM may play a role in the observed in vitro effects (discussed in
19 Section 7.5).
20
21
22 7.3 CARDIOVASCULAR AND SYSTEMIC EFFECTS OF
23 PARTICULATE MATTER IN HUMANS AND LABORATORY
24 ANIMALS: IN VIVO EXPOSURES
25 A growing number of epidemiology studies have demonstrated that increases in cardiac-
26 related deaths are associated with exposure to PM (U.S. Environmental Protection Agency,
27 1996a) and that PM-related cardiac deaths appear to be as great or greater than those attributed
28 to respiratory causes (see Chapter 8). The toxicological consequences of inhaled particles on the
29 cardiovascular system had not been extensively investigated prior to 1996. Since then (see
30 Table 7-7a,b), Costa and colleagues (e.g., Costa and Dreher, 1997) have demonstrated that
31 intratracheal instillation of high levels of ambient particles can increase or accelerate death in an
32 animal model of cardiorespiratory disease that followed monocrotaline administration in rats.
June 2003 7-39 DRAFT-DO NOT QUOTE OR CITE
-------
•— H
3
to
o
o
OJ
-U
O
C
H
1
O
o
H
O
0
H
W
O
o
1 — I
w
TABLE 7-7 a. CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INSTILLED AMBIENT
AND COMBUSTION-RELATED PARTICULATE MATTER
Species, Gender,
Strain Age, or
Body Weight
Rats, male, S-D,
60 days old,
MCT-treated and
healthy, n = 64
Rats, male, S-D,
60 days old,
MCT-treated,
and healthy
Rats, male, S-D;
60 days old
Rats, male SH and
WKY; 12- 13 weeks
old
Rabbits, female,
New Zealand White,
1.8 to 2.4 kg
Rats, male, S-D,
MCT-treated
Exposure
Particle3 Technique
ROFA Instillation
Emission Instillation
source PM
Ambient
airshed PM
ROFA
ROFA Instillation
ROFA from a Intratracheal
precipitator of instillation
an oil-burning
power plant
Colloidal Instillation
carbon
ROFA Instillation
Mass
Concentration Particle Size
0.0,0.25,1.0, 1.95 \im
and 2.5 mg/rat
Total mass: Emission PM:
2. 5 mg/rat 1. 78-4.17 |im
Total Ambient PM:
transition 3.27-4.09 \im
metal: 46
Hg/rat
0.3, 1.7, or 1.95 urn
8. 3 mg/kg ag = 2.19
1.5 urn
1 and 5 mg/kg ag = 1.5
2 mL of 1% < 1 nm
colloidal
carbon
(20 mg)
0.25, 1.0, or 1.95 nm
2.5 mg in MMAD
0.3 mL saline ag = 2.19
Exposure
Duration
Analysis at
96 h
Analysis at
24 and 96 h
following
instillation
Analysis at
24 h
Analysis at 1,
2, and 4 days
Examined for
24 to 192 h
after
instillation
Monitored for
96 h after
instillation of
ROFA
particles
Cardiovascular Effects
Dose-related hypothermia and bradycardia in
healthy rats, potentiated by compromised
models.
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.
Exposure increased plasma fibrinogen and
decreased peripheral lymphocytes in both SH
and WKY rats.
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.
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
Campen
etal.
(2000)
Costa and
Dreher
(1997)
Gardner
etal.
(2000)
Kodavanti
etal.
(2002)
Terashima
etal.
(1997)
Watkinson
etal.
(1998)
-------
to
o
TABLE 7-7 a (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INSTILLED AMBIENT
AND COMBUSTION-RELATED PARTICULATE MATTER
H
6
O
o
H
C>
O
H
W
O
O
HH
H
W
Species, Gender,
Strain Age, or
Body Weight
(1) Rats, S-D
healthy and
cold-stressed,
ozone-treated,
and MCT-treated
(2) Rats, SH,
15-mo-old
(3) Rats, S-D
MCT-treated
Particle3
ROFA
OTT
ROFA
MSH
Fe2(S04)3
VS04
NiSO2
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Mass
Concentration
0.0,0.25,1.0,
or 2. 5 mg/rat
2.5 mg
0.5 mg
2.5 mg
105 ng
245 ng
262.5 ng
Exposure
Particle Size Duration Cardiovascular Effects
1 . 95 urn Monitored for ( 1 ) Healthy rats exposed IT to ROFA
ag = 2.19 96 h after demonstrated dose-related hypothermia,
instillation bradycardia, and increased 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 IT.
(2) Older rats exposed IT to OTT showed a
pronounced biphasic hypothermia and a
severe drop in HR accompanied by increased
arrhythmias; exposure to ROFA caused less
pronounced, but similar effects. No cardiac
effects were seen with exposure to MSH.
(3) Ni and V showed the greatest toxicity;
Fe-exposed rats did not differ from controls.
Reference
Watkinson
etal.
(2000a,b);
Watkinson
etal (2001)
"ROFA = Residual oil fly ash
OTT = Ottawa dust
Fe2(SO4)3 = Iron sulfate
MSH = Mt. St. Helen's volcanic ash
VSO4 = Vanadium sulfate
NiSO, = Nickel sulfate
-------
3
to
2 Species, Gender,
Strain Age, or
Body Weight
Dogs, beagles,
10.5-year- old,
healthy, n = 4
Rats, male, F-344;
200-250 g
Dogs, female
mongrel,
14 to 17 kg
TABLE 7-7b. CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INHALED AMBIENT
AND COMBUSTION-RELATED PARTICULATE MATTER
Exposure Mass
Particle8 Technique Concentration
ROFA Oral 3 mg/m3
inhalation
OTT Nose-only 40 mg/m3
Inhalation
CAPs Inhalation via 3-360 ug/m3
tracheostomy
Particle Size
2.22 um
MMAD
og = 2.71
4 to 5 um
MMAD
0.2 to 0.3 um
Exposure
Duration
3 h/day for
3 days
4h
6 h/day for
3 days
Cardiovascular Effects
No consistent changes in ST segment, the form or
amplitude of the T wave, or arrhythmias; slight
bradycardia during exposure.
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.
Peripheral blood parameters were related to specific
particle constituents. Factor analysis from paired
and crossover experiments showed that hematologic
Reference
Muggenburg
et al. (2000a)
Bouthillier
etal. (1998)
Clarke etal.
(2000a)
changes were not associated with increases in total
CAP mass concentration.
I
[\
to
o
?d
m
H
1
O
o
0
H
O
0
H
W
O
^
o
HH
H
W
Humans, healthy CAPs
nonsmokers,
1 8 to 40 years old
Dogs, mongrel, CAPs
some with balloon
occluded LAD
coronary artery,
n= 14
Rats CAPs
Rats, male, F-344, CAPs
MCT-treated
Hamsters, 6-8 mo
old; Bio TO-2
Rats, S-D, FOFA
MCT-treated,
250 g
Inhalation 23.1 to 0.65 um
3 11.1 ug/m3 og = 2.35
Inhalation via 69-828 ug/m3 0.23 to 0.34 um
tracheostomy og = 0.2 to 2.9
Nose-only 11 0-3 50 ug/m3 N/A
inhalation
Inhalation 132-919 ug/m3 0.2-1.2 um
og = 0.2-3.9
Inhalation 580 ± 2.06 um
110 ug/m3 MMAD
og=1.57
2 h, analysis
at 18 h
6 h/day for
3 days
3h
3h,
evaluated at
3 and 24 h
6 h/day for
3 days
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.
Ghio et al.
(2000a)
Godleski
et al. (2000)
Gordon etal.
(1998)
Gordon etal.
(2000)
Killingswort
h et al.
(1997)
-------
to
o
o
OJ
TABLE
Species, Gender,
Strain Age, or
Body Weight Particle8
Rats, male WKY ROFA
and SH, 12 to
13- week-old
7-7b (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INHALED
AMBIENT AND COMBUSTION-RELATED PARTICULATE MATTER
Exposure Mass
Technique Concentration
Nose-only 1 5 mg/m3
inhalation
Exposure
Particle Size Duration
N/A 6 h/day for
3 days
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.
Reference
Kodavanti
etal.
(2000b)
Rats, Wistar
Rats, S-D, SH rats,
WKY rats, healthy
and MCT-treated
Ottawa
ambient
(EHC-93)
(ECH-93L)
Diesel soot
(DPM)
Carbon black
(CB)
Inhalation 48 mg/m
(nose only) 49 mg/m
5 mg/m3
5 mg/m3
ROFA
Inhalation 15 mg/m3
36,56,80,100, 4h
and 300 um
1.95 um
MMAD
6 h/day for
3 days
EHC-93 elevated blood pressure and ET-1 and Vincent et al.
ET-3 levels. (2001)
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.
Pulmonary hypertensive (MCT-treated S-D) and Watkinson
systemically hypertensive (SH) rats exposed to ROFA et al.
by inhalation demonstrated similar effects, but of (2000a,b)
diminished amplitude. There were no lethalities by
the inhalation route.
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Rats, male, SH
and WKY; 12 to
13 weeks old
Rats, male, S-D,
healthy and MI
ROFA from Inhalation 1 5 mg/m3
a precipitator
of an oil
burning
power plant
Boston Inhalation 3 mg/m3
ROFA
Carbon black
1.5 um
og=1.5
1.81 um
0.95 um
6 h/d, 3 d/wk
for 1, 2, or
4wk
Ih
One week exposure increased plasma fibrogen in SH
rats only; longer exposure caused pulmonary injury
but no changes in figrogen.
ROFA increased arrhythmia frequency in animals
with preexisting premature ventricular complexes and
decreased heart rate variability. Other exposed groups
not affected.
Kodavanti
et al. (2002)
Wellenius
et al. (2002)
aROFA = Residual oil fly ash
OTT = Ottawa dust
MSH = Mt. St. Helen's volcanic ash
Fe2(SO4)3 = Iron sulfate
VSO4 = Vanadium sulfate
NiSO2= Nickel sulfate
MI - Myocardial infraction
-------
1 These deaths did not occur with all types of ambient particles tested. Some dusts, such as
2 volcanic ash from Mount Saint Helens, were relatively inert; whereas other ambient dusts,
3 including those from urban sites, were toxic. These early observations suggested that particle
4 composition plays an important role in the adverse health effects associated with episodic
5 exposure to ambient PM, despite the "general particle" effect attributed to the epidemiologic
6 associations of ambient PM exposure and increased mortality in many regions of the United
7 States (i.e., regions with varying particle composition). Work that examines the role of inherent
8 susceptibility to the adverse effects of PM in compromised animal models of human
9 pathophysiology provides a potentially important link to epidemiologic observations and is
10 discussed below.
11 To date, studies examining the systemic and cardiovascular effects of particles have used a
12 number of compromised animal models, largely rodent models. Two studies in normal or
13 compromised dogs (Godleski et al., 2000; Muggenburg et al., 2000a) also have been published
14 as well as the preliminary results from studies in which human subjects were exposed to
15 concentrated ambient PM (see Section 7.4.1). Muggenburg et al. (2000b) described several
16 potential animal models of cardiac disease (monocrotaline-induced pulmonary hypertension,
17 dilated cardiomyopathy, viral and mycoplasmal myocarditis, and ischemic heart disease),
18 including a discussion of the advantages and disadvantages in the use of animal models in the
19 study of cardiac disease and air pollution. Pulmonary hypertension in humans may result from
20 airway and vascular effects from COPD, asthma, and cystic fibrosis. The monocrotaline (MCT)-
21 induced vascular disease model exhibits common features of chronic obstructive pulmonary
22 disease in humans. The mechanism of injury includes selective pulmonary endothelial damage
23 and progressive pulmonary arterial muscularization. Pulmonary hypertension develops as the
24 blood flow is impeded. Right ventricular hypertrophy follows the pulmonary hypertension. To
25 produce pulmonary hypertension, animals are injected subcutaneously with 50-60 mg/kg
26 monocrotaline. Within two weeks following treatment, experimental animals, primarily rats,
27 develop pulmonary hypertension (Kodavanti et al., 1998a). The majority of animal studies
28 examining the systemic effects of PM have used metal-laden ROFA as a source particle,
29 a growing number of studies have used collected and stored ambient PM or real-time generated
30 concentrated ambient particles. The following discussion of the systemic effects of PM first
June 2003 7-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 describes the studies using ROFA and then compares these findings with the ambient PM
2 studies.
3 Killingsworth et al. (1997) used fuel oil fly ash to examine the adverse effects of a model
4 urban particle using the MCT model of cardiorespiratory disease. They observed 42% mortality
5 in MCT rats exposed to -580 |ig/m3 fly ash for 6 h/day for 3 consecutive days. Deaths did not
6 occur in MCT rats exposed to filtered air or in saline-treated rats exposed to fly ash.
7 The increase in MCT/fly ash group deaths was accompanied by increased neutrophils in lavage
8 fluid and increased immunostaining of MIP-2 in the heart and lungs of the MCT/fly ash animals.
9 Cardiac immunohistochemical analysis indicated increased MIP-2 in cardiac macrophages. The
10 fly ash-induced deaths did not result from a change in pulmonary arterial pressure and the cause
11 of death was not identified.
12 In a similar experimental model, Watkinson et al. (1998) examined the effects of
13 intratracheally instilled ROFA (0.0, 0.25, 1.0, 2.5 mg in 0.3 mL saline) on ECG measurements in
14 control and MCT rats. They observed a dose-related increase in the incidence and duration of
15 arrhythmic events in control animals exposed to ROFA particles, and these effects were clearly
16 exacerbated in the MCT animals. Similar to the results of Killingsworth et al. (1997), healthy
17 animals treated with ROFA suffered no deaths, but there were 1, 3, and 2 deaths in the low-,
18 medium-, and high-dose MCT groups, respectively. Thus, ROFA PM was linked to the
19 conductive and hypoxemic arrhythmias associated with cardiac-related deaths in the MCT
20 animals.
21 To examine the biological relevance of intratracheal instillation of ROFA particles,
22 Kodavanti et al. (1999) exposed MCT rats to ROFA by either instillation (0.83 or 3.33 mg/kg) or
23 nose-only inhalation (15 mg/m3, 6 h/day for 3 consecutive days). Similar to Watkinson et al.
24 (1998), intratracheal instillation of ROFA in MCT rats resulted in =50% mortality. Notably, no
25 mortality occurred in MCT rats exposed to ROFA by the inhalation route despite the high
26 exposure concentration (15 mg/m3). In addition, no mortality occurred in healthy rats exposed to
27 ROFA or in MCT rats exposed to clean air. Despite the fact that mortality was not associated
28 with ROFA inhalation exposure of MCT rats, exacerbation of lung lesions and pulmonary
29 inflammatory cytokine gene expression, as well as ECG abnormalities, clearly were evident.
30 Watkinson and colleagues further examined the effect of instilled ROFA in rodents
31 previously exposed to ozone or housed in the cold (Watkinson et al., 2000a,b; Watkinson et al.,
June 2003 7-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 2001; Campen et al., 2000). The effect of ozone-induced pulmonary inflammation (preexposure
2 to 1 ppm ozone for 6 h) or housing in the cold (10 °C) on the response to instilled ROFA in rats
3 was similar to that produced with MCT. Bradycardia, arrhythmias, and hypothermic changes
4 were consistently observed in the ozone exposed and hypothermic animals treated with ROFA,
5 although, unlike in the MCT animals, no deaths occurred. Thus, in rodents with
6 cardiopulmonary disease/stress, instillation of 0.25 mg or more of ROFA can produce systemic
7 changes that may be used to study potential mechanisms of toxicity that are consistent with the
8 epidemiology and panel studies showing cardiopulmonary effects in humans.
9 While studies of instilled residual oil fly ash demonstrated immediate and delayed
10 responses, consisting of bradycardia, hypothermia, and arrhythmogenesis in conscious,
11 unrestrained rats (Watkinson et al., 1998; Campen et al., 2000), further study of instilled ROFA-
12 associated transition metals showed that vanadium induced the immediate responses, while
13 nickel was responsible for the delayed effects (Campen et al., 2002). Moreover, Ni, when
14 administered concomitantly, potentiated the immediate effects caused by V.
15 In another study, Campen et al. (2001) examined the responses to these metals in conscious
16 rats by whole-body inhalation exposure. The authors attempted to ensure valid dosimetric
17 comparisons with the instillation studies, by using concentrations of V and Ni ranging from
18 0.3-2.4 mg/m3. The concentrations used in this study incorporated estimates of total inhalation
19 dose derived using different ventilatory parameters. Heart rate (FIR), core temperature (T[CO]),
20 and electrocardiographic (ECG) data were measured continuously throughout the exposure.
21 Animals were exposed to aerosolized Ni, V, or Ni + V for 6 h per day for 4 days, after which
22 serum and bronchoalveolar lavage samples were taken. While Ni caused delayed bradycardia,
23 hypothermia, and arrhythmogenesis at concentrations > 1.2 mg/m3, V failed to induce any
24 significant change in HR or T (CO), even at the highest concentration. When combined, Ni and
25 V produced observable delayed bradycardia and hypothermia at 0.5 mg/m3 and potentiated these
26 responses at 1.3 mg/m3, to a greater degree than were produced by the highest concentration of
27 Ni (2.1 mg/m3) alone. Although these studies were performed at metal concentrations that were
28 orders of magnitude greater than ambient concentrations, the results indicate a possible
29 synergistic relationship between inhaled Ni and V.
30 Watkinson et al. (2000a,b) also sought to examine the relative toxicity of different particles
31 on the cardiovascular system of spontaneously hypertensive rats. They instilled 2.5 mg of
June 2003 7-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 representative particles from ambient (Ottawa) or natural (Mount Saint Helens volcanic ash)
2 sources and compared the response to 0.5 mg ROFA. Instilled particles were either mass
3 equivalent dose or adjusted to produce equivalent metal dose. They observed adverse changes in
4 ECG, heart rate, and arrhythmia incidence that were much greater in the Ottawa- and ROFA-
5 treated rats than in the Mount Saint Helens-treated rats. The cardiovascular changes observed
6 with the Ottawa particles were actually greater than with the ROFA particles. These
7 experiments by Watkinson and colleagues clearly demonstrate: (a) that instillation of ambient
8 air particles, albeit at a very high concentration, can produce cardiovascular effects; and (b) that
9 exposures of equal mass dose to particle mixes of differing composition did not produce the
10 same cardiovascular effects, suggesting that PM composition rather than just mass was
11 responsible for the observed effects.
12 To more closely mimic environmental exposures, Kodavanti et al. (2000b) exposed
13 spontaneously hypertensive (SH) and normotensive (WKY) rats to 15 mg/m3 ROFA for 6 h/day
14 for 3 days. The exposure concentration, while 100 times or more higher than usual current U.S.
15 ambient air PM concentrations, was selected to produce a frank but non-lethal injury and to
16 allow comparison to the intratracheal approaches. Exposure to ROFA produced alterations in
17 the ECG waveform of spontaneously hypertensive (SH) but not normotensive rats. Although the
18 ST segment area of the ECG was depressed in the SH rats exposed to air, further depressions in
19 the ST segment were observed at the end of the 6-h exposure to ROFA on Days 1 and 2. The
20 enhanced ST segment depression was not observed on the third day of exposure, suggesting that
21 adaptation to the response had occurred. Thus, exposure to a very high concentration of ROFA
22 exacerbated a defect in the electroconductivity pattern of the heart in an animal model of
23 hypertension. This ROFA-induced alteration in the ECG waveform was not accompanied by an
24 enhancement in the monocytic cell infiltration and cardiomyopathy that also develop in SH rats.
25 Further work is necessary to determine the relevance of this ROFA study to PM at
26 concentrations relevant to ambient exposures.
27 Godleski and colleagues (2000a) have performed a series of experiments examining the
28 cardiopulmonary effects of inhaled concentrated ambient PM on normal mongrel dogs and on
29 dogs with coronary artery occlusion. Dogs were exposed by inhalation via a tracheostomy tube
30 to concentrated ambient PM for 6 h/day for 3 consecutive days. The investigators found little
31 biologically-relevant evidence of pulmonary inflammation or injury in normal dogs exposed to
June 2003 7-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM (daily range of mean concentrations was -100 to 1,000 |ig/m3). The only statistically
2 significant effect was a doubling of the percentage of neutrophils in lung lavage. Despite the
3 absence of major pulmonary effects, a significant increase in heart rate variability (an index of
4 cardiac autonomic activity), a decrease in heart rate, and an increase in T alternans (an index of
5 vulnerability to ventricular fibrillation) were seen. Exposure assessment of particle composition
6 produced no specific components of the particles that were correlated with the day-to-day
7 variability in response. The significance of these effects is not yet clear, because the effects did
8 not occur on all exposure days. For example, the change in heart rate variability was observed
9 on only 10 of the 23 exposure days. Although the heart rate variability change and the increase
10 in T alternans suggest a possible proarrhythmic response to inhaled concentrated ambient PM,
11 the clinical significance of this effect is currently unknown.
12 The most important finding of Godleski et al. (2000) was the observation of a potential
13 increase in ischemic stress of the cardiac tissue from repeated exposure to concentrated ambient
14 PM. During coronary occlusion in four dogs exposed to PM, they observed a significantly more
15 rapid development of ST elevation of the ECG waveform. Also, the peak ST-segment elevation
16 was greater after PM exposure. Together, these changes suggest that concentrated ambient PM
17 can augment the ischemia associated with coronary artery occlusion in this dog model. More
18 work in more dogs as well as other species is necessary to determine the significance of these
19 findings to the human response to ambient PM.
20 Muggenburg and colleagues (2000a) reported that inhalation exposure to high
21 concentrations of ROFA produces no consistent changes in amplitude of the ST-segment, form
22 of the T wave, or arrhythmias in dogs. In their studies, four beagle dogs were exposed to
23 3 mg/m3 ROFA particles for 3 h/day for 3 consecutive days. They noted a slight but variable
24 decrease in heart rate, but the changes were not statistically or biologically significant. The
25 transition metal content of the ROFA used by Muggenburg was -15% by mass, a value on the
26 order of a magnitude higher than that found in ambient urban PM samples. Although the study
27 did not specifically address the effect of metals, it suggests that inhalation of high concentrations
28 of metals may have little effect on the cardiovascular system of a healthy individual.
29 In a series of studies, (Gordon et al., 2000) examined the response of the rodent
30 cardiovascular system to concentrated ambient PM derived from New York City air. Particles of
31 0.2 to 2.5 jim diameter were concentrated up to 10 times their levels in ambient air (=150 to
June 2003 7-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 900 |ig/m3) to maximize possible differences in effects between normal and cardiopulmonary-
2 compromised laboratory animals. ECG changes were not detected in normal Fischer 344 rats or
3 hamsters exposed by inhalation to concentrated ambient PM for 1 to 3 days. Similarly, no
4 deaths or ECG changes were seem in MCT rats or cardiomyopathic hamsters exposed to PM.
5 In contrast, to the nonsignificant decrease in heart rate observed in dogs exposed to concentrated
6 ambient PM (Godleski et al., 2000), heart rate was increased significantly in both normal and
7 MCT rats exposed to PM. The increase was approximately 5% and statistically significant, but
8 was not observed on all exposure days. Thus, extrapolation of the heart rate changes in these
9 animal studies to human health effects is difficult, although the increase in heart rate in rats is
10 similar to that observed in some human population studies.
11 Gordon and colleagues (1998) have reported other cardiovascular effects in animals
12 exposed to inhaled CAP. Increases in peripheral blood platelets and neutrophils were observed
13 in control and MCT rats at 3 h, but not 24 h, after exposure to 150 to 400 |ig/m3 concentrated
14 ambient PM (CAP). This neutrophil effect did not appear to be dose-related and did not occur
15 on all exposure days, suggesting that day-to-day changes in particle composition may play an
16 important role in the systemic effects of inhaled particles. The number of studies reported was
17 small; and, it is therefore not possible to statistically determine if the day-to-day variability was
18 truly due to differences in particle composition or even to determine the size of this effect.
19 Terashima et al. (1997) also examined the effect of particles on circulating neutrophils. They
20 instilled rabbits with 20 mg colloidal carbon, a relatively inert particle (< 1 jim), and observed a
21 stimulation of the release of 5'-bromo-2'deoxyuridine (BrdU)-labeled PMNs from the bone
22 marrow at 2 to 3 days after instillation. Because the instilled supernatant from rabbit AMs
23 treated in vitro with colloidal carbon also stimulated the release of PMNs from the bone marrow,
24 the authors hypothesized that cytokines released from activated macrophages could be
25 responsible for this systemic effect. The same research group (Tan et al., 2000) looked for
26 increased white blood cell counts as a marker for bone marrow PMN precursor release in
27 humans exposed to very high levels of carbon from biomass burning during the 1997 Southeast
28 Asian smoke-haze episodes. They found a significant association between PM10 (1-day lag) and
29 elevated band neutrophil counts expressed as a percentage of total PMNs. The biological
30 relevance of this latter study more usual urban PM exposure-induced systemic effects is unclear;
31 however, because of the high dose of carbon particles.
June 2003 7-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 The results of epidemiology studies suggest that homeostatic changes in the vascular
2 system can occur after episodic exposure to ambient PM. Studies by Vincent et al. (2001)
3 indicate that urban particles inhaled by laboratory rats can affect blood levels of endothelin and
4 cause a vasopressor response without causing acute lung injury. Moreover, the potency to
5 influence hemodynamic changes can be modified by removing the polar organic compounds and
6 soluble elements from the particles. Frampton (2001) exposed healthy, nonsmoking subjects (18
7 to 55 years old) to 10 |ig/m3 ultrafine carbon while resting. Subjects were exposed to the
8 ultrafine carbon through a mouthpiece for 2 h with a ten minute break between each hour
9 exposure. The exposure concentration (10 |ig/m3) corresponded to 2 x 106 particles/cm3.
10 Subjects were assessed for respiratory symptoms, spirometry, blood pressure, pulse-oximetry,
11 blood markers, and exhaled NO before, immediately following, and 3.5 and 21 h post-exposure.
12 Blood markers focused on parameters related to acute response, blood coagulation, circulating
13 leukocyte activation, including complete blood leukocyte counts and differentials, IL-6,
14 fibrinogen, and clotting factor VII. Heart rate variability and repolarization phenomena were
15 evaluated by continuous 24-h Hotter monitoring. Preliminary findings indicated no particle-
16 related symptoms. In a study described previously (Section 7.2.3), Ohio et al. (2000a) also
17 showed that inhalation of concentrated PM in healthy nonsmokers causes increased levels of
18 blood fibrinogen. They exposed 38 volunteers exercising intermittently at moderate levels of
19 exertion for 2 h to either filtered air or particles concentrated from the air in Chapel Hill, NC
20 (23 to 311 |ig/m3). Blood obtained 18 h after exposure contained significantly more fibrinogen
21 than blood obtained before exposure. The observed effects in blood may be associated with the
22 mild pulmonary inflammation also found 18 h after exposure to CAP (see Section 7.2.3).
23 Zelikoff et al. (2003) reported that CAPs had relatively little effect on the pulmonary or
24 systemic immune defense mechanisms in Fisher rats exposed to 0 or 90 to 600 |ig/m3 for 3 h
25 prior to IT instillation of Streptococcus pneumoniae (2 - 4 x 107 organisms delivered dose). The
26 number of lavageable cells, PAM and PMN, increased in both experimental groups but were
27 twice as high in the CAPs exposed groups and were elevated faster and remained elevated
28 longer. Lymphocyte values and WPC were significantly increased 24 and 72 h postinfection in
29 both groups. CAPs exposure retarded the decline of TNFcc and IL-6 levels three days
30 postinfection compared to bacteria only exposed rats, however, the differences were not
June 2003 7-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 significant. CAPs exposure significantly increase the bacterial burdens at 24 h postinfection.
2 Thereafter, CAPs-exposed animals exhibited significantly lower bacterial burdens.
3 In another set of experiments, Zelokoff et al. (2003) evaluated the effects of CAPs
4 exposure in rats following a single 5 h exposure to IT instilled Streptococcus pneumoniae. CAPs
5 exposure significantly reduced the percentages of lavageable PMN 24 h following CAPs
6 exposure and remained well below the match counterparts for up to 3 days. Lavageable PAM
7 was significantly increased in the CAPs exposed animals. CAPs exposure reduced the levels of
8 TNFcc, IL-1, and IL-6. The bacterial burden reduced in both exposed groups over time, however,
9 CAPs exposed animals had a significantly greater burden after 24 h than did control rats. Levels
10 of lymphocytes and monocytes were unaffected by CAPs exposure.
11 Gardner et al. (2000) examined whether the instillation of particles would alter blood
12 coagulability factors in laboratory animals. Sprague-Dawley rats were instilled with 0.3, 1.7, or
13 8.3 mg/kg of ROFA or 8.3 mg/kg Mount Saint Helens volcanic ash. Because fibrinogen is a
14 known risk factor for ischemic heart disease and stroke, the authors suggested that this alteration
15 in the coagulation pathway could take part in the triggering of cardiovascular events in
16 susceptible individuals. Elevations in plasma fibrinogen, however, were observed in healthy rats
17 only at the highest treatment dose (8.3 mg/kg); and no other changes in clotting function were
18 noted. Because the lower treatment doses are known to cause pulmonary injury and
19 inflammation, albeit to a lesser extent, the absence of plasma fibrinogen changes at the lower
20 doses suggests that only high levels of pulmonary injury are able to produce an effect in healthy
21 test animals.
22 To establish the temporal relationship between pulmonary injury, increased plasma
23 fibrinogen, and changes in peripheral lymphocytes, Kodavanti et al. (2002) exposed
24 spontaneously hypertensive (SH) and Wistar-Kyoto (WKY) rats to ROFA using both
25 intratracheal and inhalation exposure (acute and long-term) scenarios. Increases in plasma
26 fibrinogen and decreases in circulating white blood cells were found during the acute phase
27 responses to ROFA exposure and were temporally associated with acute, but not long-term, lung
28 injury. A bolus intratracheal instillation of ROFA increased plasma fibrinogen in both SH and
29 WKY rats; whereas the increase was evident only in SH rats after acute ROFA inhalation. The
30 increased fibrinogen in SH rats was associated with greater pulmonary injury and inflammation
31 than was found in the WKY rats.
June 2003 7-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 Nemmar et al. (2002) investigated the effect of ultrafme (60 nm) polystyrene particles on
2 thrombus formation in a hamster model after IT administration of unmodified, carboxylate-
3 polystyrene, or amine-polystyrene particles. Unmodified and carboxylate-polystyrene particles
4 (5 mg/kg) did not modify significantly the intensity of thrombosis formed. In contrast the
5 administration of 5 mg/kg amine-polystyrene particles significantly enhanced thrombosis
6 formation. The authors concluded that the presence of ultrafme particles in the circulation may
7 affect hemostasis and that this phenomenon is dependent on the surface properties of the
8 particles.
9 Suwa et al. (2002) studied the effect of PM10 on the progression of atherosclerosis in
10 rabbits. They exposed Watanabe heritable hyperlipidemic rabbits to PM10 (n = 10) or vehicle
11 (n = 6) for four weeks, and both measured bone marrow stimulation and used quantitative
12 histologic methods to determine the morphologic features of the atherosclerotic lesions.
13 Exposure to PM10 caused an increase in circulating polymorphonuclear leukocytes (PMN) band
14 cell counts and an increase in the size of the bone marrow mitotic pool of PMNs. Exposure to
15 PM10 also caused progression of atherosclerotic lesions toward a more advanced phenotype. The
16 volume fraction (vol/vol) of the coronary atherosclerotic lesions was increased by PM10
17 exposure. The vol/vol of atherosclerotic lesions correlated with the number of alveolar
18 macrophages that phagocytosed PM10. Exposure to PM10 also caused an increase in plaque cell
19 turnover and extracellular lipid pools in coronary and aortic lesions, as well as in the total
20 amount of lipids in aortic lesions.
21 In summary, controlled laboratory animal studies, to date, have provided evidence
22 indicating that high concentrations of inhaled or instilled particles can have systemic, especially
23 cardiovascular, effects. In the case of MCT rats, these effects can be lethal. Controlled human
24 exposure studies also have shown that ambient levels of inhaled PM can produce some
25 biochemical and cellular changes in the blood. Although some of these biochemical changes
26 have been used as clinical "markers" for cardiovascular diseases, the causal relationship between
27 these changes and the potential life-threatening diseases remains to be established.
28 Understanding the pathways by which very small concentrations of inhaled ambient PM can
29 produce systemic, life-threatening changes also is far from clear. Among the hypotheses that
30 have been proposed to account for the nonpulmonary effects of PM are activation of neural
31 reflexes, cytokine effects on heart tissue (Killingsworth et al., 1997), alterations in coagulability
June 2003 7-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 (Seaton et al., 1995; Sjogren, 1997), perturbations in both conductive and hypoxemic
2 arrythmogenic mechanisms (Watkinson et al., 1998; Campen et al., 2000), and altered endothelin
3 levels (Vincent et al., 2001). A great deal of research using controlled exposures of laboratory
4 animals and human subjects to PM will be necessary to test further such mechanistic hypotheses
5 generated to date, as well as those that are likely to be proposed in the future.
6
7
8 7.4 PARTICULATE MATTER TOXICITY AND PATHOPHYSIOLOGY:
9 IN VITRO EXPOSURES
10 7.4.1 Introduction
11 Toxicological studies play an integral role in determining the biological plausibility for the
12 health effects associated with ambient PM exposure. At the time of 1996 PM AQCD (U.S.
13 Environmental Protection Agency, 1996a) little was known about potential mechanisms that
14 could explain the morbidity and mortality observed in populations exposed to PM. One of the
15 difficulties in trying to sort out possible mechanisms is the nature of particles themselves.
16 Ambient PM has diverse physicochemical properties (Table 7-8) ranging from the physical
17 characteristics of the particle to the chemical components in or on the surface of the particle.
18 Any one of these properties could change at any time in the ambient exposure atmosphere,
19 making it hard to replicate the actual properties in a controlled experiment. As a result,
20 controlled exposure studies as yet have not been able to unequivocally determine the particle
21 properties and the specific mechanisms by which ambient PM may affect biological systems.
22 Despite these underlying difficulties, a number of toxicological studies have become available
23 since 1996 to help explain how ambient particles may exert toxic effects on the cardiovascular
24 and respiratory systems. The following section discusses the more recently published studies
25 that provide an approach toward identifying potential mechanisms by which PM mediates health
26 effects. The remaining sections discuss potential mechanisms in relation to PM characteristics
27 based on these available data.
28
29 7.4.2 Experimental Exposure Data
30 In vitro exposure is a useful technique to provide information on potential hazardous PM
31 constituents and mechanisms of PM injury, especially when only limited quantities of the test
June 2003 7-53 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-8. PHYSICOCHEMICAL PROPERTIES OF PARTICIPATE MATTER
Physical Characteristics Chemical Components
• particle mass (size, shape, density) • elemental and organic carbon
• particle number • semivolatile organics
• surface area • metals (Fe, Cd, Co, Cu, Mn, Ni, Pb, Ti, V, Zn)
• surface chemistry • biologicals (e.g., pollen, microbes)
• surface charge • sulfates
• acidity • nitrates
• pesticides
1 material are available. In addition, in vitro exposure allows the examination of the response to
2 particles in only one or two cell types. Respiratory epithelial cells that line the airway lumen are
3 the initial targets of airborne pollutants. These cells have been featured in numerous studies
4 involving airborne pollutants and show inflammatory responses similar to that of human primary
5 epithelial cultures. Limitations of in vitro studies include difficulty in extrapolating dose-
6 response relationships and from in vitro to in vivo biological response and mechanistic
7 extrapolations. Besides alterations in physiochemcial characteristics of PM because of the
8 collection and resuspension processes, these exposure conditions do not simulate the air-cell
9 interface that actually exists within the lungs, and, thus, the exact dosage delivered to target cells
10 in vivo is not known. Furthermore, unless an in vitro exposure system that is capable of
11 delivering particles uniformly to monolayers of airway epithelial cells cultured in an air-liquid
12 interface system is used (Chen et al., 1993), conventional incubation systems alter the
13 microenvironment surrounding the cells and may alter the mechanisms of cellular injury induced
14 by these agents.
15 Doses delivered in vitro, like intratracheal administration, are very high on a cellular basis,
16 making it very difficult to extrapolate to in vivo exposure conditions. It would be useful if
17 in vitro studies included, in addition to the high doses, doses comparable to environmental doses
18 predicted to occur under in vivo conditions at the cellular level. Even with these limitations,
19 in vitro studies do provide an approach to identify potential cellular and molecular mechanisms
20 by which PM mediates health effects. These mechanisms can then be evaluated in vivo. In vitro
21 studies published since the 1996 PM AQCD was completed are summarized in Table 7-9.
22
June 2003 7-54 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-9. IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE MATTER CONSTITUENTS
to
O
o
Species, Cell Particle or
Type, etc." Constituent'
Human bronchial DPM
epithelial cells,
asthmatic (ASTH)
nonasthmatic
(NONA)
Human bronchial DPM
epithelial cells
(smokers)
Human and Four Urban air
rat AM particles:
ROFA
DPM
Volcanic ash
Silica
Cell Count Concentration
10-100 ug/mL
10-100 ug/mL
2.5 x 105 cells/mL Urban and DPM:
12,27, 111,333, or
1000 ug/mL
SiO2 and TiO2:
4, 12, 35, or
167 ug/mL
Fe2O3: 1:1,3:1;
10:1 particles/cell
ratio
Particle Size
0.4 um
0.4 um
Urban particles:
0.3-0.4 urn
DPM: 0.3 um
ROFA: 0.5 um
Volcanic ash:
1.8 um
Silica: 05-10 um
TiO2: <5\im
Latex: 3.8 um
Exposure Duration
2, 4, 6, 24 h
24 h
2 h for cytotoxicity,
16-18 h for cytokine
assay;
chemiluminescence
at 30 minutes
Effect of Particles
DPM caused no gross cellular damage.
Ciliary beat frequency was attentuated
at all doses. DPM caused IL-8 release
at lower dose in ASTH than NONA.
Higher concentrations of DPM
suppressed IL-8, GM-CSF, and
RANTES in ASTH cells.
DPM attenuated ciliary beating.
Release of IL-8, protein, GM-CSF, and
SIC AM- 1 increased after DPM
exposure.
UAP-induced cytokine production
(TNF, IL-6) in AM of both species that
is not related to respiratory burst or
transition metals but may be related to
LPS (blocked by polymyxin B but not
DEF) ROFA induced strong
chemiluminescence but had weak
effects on TNF production.
Reference
Bayram et al.
(1998a)
Bayram et al.
(1998b)
Becker et al.
(1996)
Human AM and
blood monocytes
Rat AM
NHBE cells
Urban air particles;
St. Louis SRM 1648;
Washington, DC, SRM
1649; Ottawa, Canada,
EHC-93
PM10
Mexico City 1993;
volcanic ash (MSHA)
ROFA
33 or 100 ug/mL
10 ug/cm2
0.2 to 0.7 um
: 10 um
3, 6, or 18-20 h
24 h
0, 5, 50, or
200 ug/mL (actual
dose delivered 1.6 —
60 ug/cm2)
: 10 um
Analysis at 2 and 24 h
postexposure
Phagocytosis was inhibited by UAP at Becker and
18 h. UAP caused decreased Soukup (1998)
expression of P2-integrins involved in
antigen presentation and phagocytosis.
PM10 stimulated alveolar macrophages Bonner et al.
to induce up-regulation of PDGF •* (1998)
receptor on myofiboroblasts.
Endotoxin and metal components of
PM10 stimulate release of IL-p. This
is a possible mechanism for
PM10-induced airway remodeling.
Increase in expression of the cytokines Carter et al.
IL-6, IL-8, and TNF-a; inhibition by (1997)
DMTU or deferoxamine.
-------
to
O
o
TABLE 7-9 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Species, Cell
Type, etc."
Human
erythrocytes;
RAW 264.7 cells
Supercoiled
DNA
Rat AM
Primary cultures
ofRTE
Particle or
Constituent'
PM10.2.5; PM2.5
from Rome, Italy
PM10 from
Edinburgh,
Scotland
UAP
DPM
ROFA
Cell Count Concentration
1 x 106 50 ± 45 ug/m3
cells/mL 31±24ug/m3
19 ± 20 ug/m3
996.2 ± 181.8
ug/filter in 100 uL
1 x 106 50 to 200 ug/mL
cells/mL
3x10" 5, 10, or20ug/cm2
cells/cm2
Particle Size
PM10
PM2.5
PM10.,5
PM10
DPM: 1.1 -1.3 urn
UAP: St Louis,
between 1974 and
1976 in a baghouse,
sieved through
200-mesh (125 urn)
1.95 umMMAD
Exposure
Duration
Ih
24 h
8h
2 h exposure;
supernatant
collected 18 h
postexposure
Analysis at 6 and
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.
PM10 caused damage to DNA; mediated by hydro xyl
radicals (inhibited by mannitol) and iron (inhibited by
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.
Dose dependent increase in TNF-a, IL-6, CINC, MIP-2
gene expression by urban particles but not with DPM;
cytokine production were not related to ROS; cytokine
production can be inhibited by polymyxin B; LPS was
detected on UAP but not DPM; endotoxin is responsible
for the cytokine gene expression induced by UAP in AM.
Particle induced epithelial-cell detachment and lytic cell
injury; alterations in the permeability of the cultured RTE
Reference
Diociaiuti
etal. (2001)
Donaldson
etal. (1997)
Dong et al.
(1996)
Dye et al.
(1997)
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.
Primary cultures
of RTE
Peripheral blood
monocytes
BEAS-2B
ROFA; metal
solutions
Organic extract
of TSP, Italy
Provo PM10
extract
5, 10, or 20 |ig/cm2
1 x 10" 5.3,10.6,21.2,
cells/mL 42.5, 85, 340 ug
residue/m3
(acetone)
125, 250,
500 ug/mL
1.95 urn 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
aparticulate concentration of 0.17mg/mL (340 ug) 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
etal. (1999)
-------
to
O
o
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Rat AM
NHBE
BEAS-2B
BEAS-2B
respiratory
epithelial cells
BEAS-2B
0X174 RF1
DNA
TABLE 7-9 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
Species, Cell
Type, etc."
Particle or
Constituent'
Cell Count
Concentration
Particle Size
Exposure
Duration
Effect of Particles
Reference
ROFA, iron sulfate,
0.5-1.0
nickel sulfate, vanadyl 106 cells/mL
sulfate
Latex particles with
metal complexed on
the surface
ROFA
ROFA
Provo
TSP soluble and
insoluble extract
PM10 from Edinburgh,
Scotland
0.01-1.Omg/mL 3.6umMMAD Up to
400 min
Increase chemiluminescence, inhibited by DEF and hydroxyl Ohio et al.
radical scavengers; solutions of metal sulfates and metal- (1997a)
complexed latex particles similarly elevated chemiluminescence
in a dose- and time-dependent manner.
5, 50, 200 ug/mL 3.6 urn
100 ug/mL 3.6 urn
500 ug/mL TSP
3.7or7.5ng/mL PM10
2 and 24 h mRNA for ferritin did not change; ferritin protein increase; Ohio et al.
mRNA for transferrin receptor decreased, mRNA for lactoferrin (1998c)
increased; transferrin decreased whereas lactoferrin increased;
deferoxamine alone increased lactoferrin mRNA.
5 min - 1 h
24 h
8h
Lactoferrin binding with PM metal occurred within 5 min. V
and Fe (m), but not Ni, increased the concentration of lactoferrin
receptor.
Ohio et al.
(1999b)
Water soluble fraction caused greater release of IL-than Ohio et al.
insoluble fraction. The effect was blocked by deferoxamine and (1999a)
presumably because of metals (Fe, Cu, Zn, Pb).
Significant free radical activity on degrading supercoiled DNA; Gilmour et al.
mainly because of hydro xyl radicals (inhibited by mannitol); Fe (1996)
involvement (DEF-B conferred protection); more Fe3+ was
released compared to Fe2+, especially at pH 4.6 than at 7.2.
Hamster AM
Hamster AM
ROFA or CAPs
CAPs, ROFA, and
their water-soluble
and particulate
fractions
0.5 x 106 ROFA: 0, 25, 50,
cells/mL 100, or
200 ug/mL
CAPS: 1:15,
1:10, 1:20
(described as 4,
10, 20 ug/mL)
0.5 x 106 ROFA: 25, 50,
cells/mL 100, 200 ug/mL
CAPS: 38-180
Hg/mL
CAPs:
0.1-2.5 urn
(from Harvard
concentrator)
TiO2: 1 urn
CAPs = 0.125 urn
ROFA= 1.0 urn
30 min
incubation,
analysis
immediately
following
30 min
Dose-dependent increase in AM oxidant stress with both ROFA
and CAP. Increase in particle uptake; Mac-type SR mediate a
substantial proportion of AM binding; particle-associated
components (e.g., transition metals) are likely to mediate
intracellular oxidant stress and proinflammatory activation.
ROFA and CAPs (water soluble components) 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.
Goldsmith
etal. (1997)
Goldsmith
etal. (1998)
-------
to
O
o
TABLE 7-9 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
oo
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Species, Cell
Type, etc."
AMs from
female CD rats
Human PMN
Particle or Constituent'
Vanadyl chloride sodium
metavanadate
Aqueous and organic
extracts of TSP in
Dusseldorf and Duisburg,
Germany
Cell Count Concentration
2-2.5x10' 10-1000 uM
cells/mL metavanadate
1 x 106 0.42-0.78 mg
cells/mL dust/mL
Particle Size
N/A
Collected by high volume
sampler, 90% < 5 um,
50% < lum, maximum at
0.3-0.45 urn
Extracted using water and then
dichloromethane to yield
aqueous and organic extracts
Exposure
Duration
30min
Up to
35 min
Effect of Particles
Metavanadate caused increased production of
ROS. The LOEL was 50 uM.
PM extract alone significantly stimulated the
production and release of ROS in resting but not
in zymosan-stimulated PMN. The effects of the
PM extracts were inhibited by SOD, catalase and
sodium azide (NaN3);
Zymosan-induced LCL is inhibited by both types
of extracts, but aqueous extracts have a stronger
inhibitory effect.
Reference
Grabowski
etal. (1999)
Hitzfeld et al.
(1997)
Human AM UAP
(#1648, 1649)
Volcanic ash
ROFA
Primary GPTE ROFA
cells DOFA
STL
WDC
OT
MSH
BEAS-2B TSP collected in Provo
1 x 106 0, 25, 100, or
cells/mL 200 ug/mL
2-5 x 105 6.25,12.5,25,
cells/cm2 and 50 ug/cm2
2 x 105 TSP filter
cells/mL samples
(36.5 mg/mL)
agitated in
deionized H2O2
for 96 h,
centrifuged at
1 200 g for
30 min,
lyophylized and
resuspended in
deionized H2O2
or saline
Volume median diameter: 24 h ROFA highly toxic; urban PM toxic at
ROFA 1.1 um 200ug/mL; ROFA produced significant apoptosis
#1648: 1.4 |im as low as 25 ug/mL; UAP produced apoptosis at
#1649: 1.1 um 100 ug/mL; UAP and ROFA also affect AM
volcanic ash 2.3 um phenotype: increased immune stimulatory,
whereas decreased
immune suppressor phenotype.
N/A 4, 8, and ROFA was the most toxic particle, enhancing
24 h mucin secretion and causing toxicity, assessed by
LDH release.
N/A (TSP samples, comprised Sacrificed at Provo particles caused cytokine-induced
50 to 60% PM10) 24 h neutrophil-chemoattractant-dependent
inflammation of rat lungs; Provo particles
stimulated IL-6 and IL-8 production, increased
IL-8 mRNA and ICAM-1 in BEAS-2B cells, and
stimulated IL-8 secretion in primary cultures of
BEAS-2B cells; cytokine secretion was preceded
by activation of NF-KB and was reduced by
SOD, DBF, or NAC; quantities of Cu2+ found in
Provo particles replicated the effects
Holian et al.
(1998)
Jiang et al.
(2000)
Kennedy
etal. (1998)
-------
to
O
o
TABLE 7-9 (cont'd). IN VITRO EFFECTS OF PARTICIPATE MATTER AND PARTICIPATE
MATTER CONSTITUENTS
Species, Cell Type,
etc."
Human lung
mucoepidermoid
carcinoma cell line,
NCI-H292
BEAS-2B
Male (Wistar) rat
lung macrophages
Human blood
monocytes and
neutrophils (PMN)
BEAS-2B
BEAS-2B
Particle or
Constituent' Cell Count Concentration
ROFA 1 x 106 10, 30, 100 ug/mL
cells/mL
ROFA 5 x 106 0,0.5, or 2.0 mg in
cells/mL 10 mL
Urban dust SRM 2 x 105 0-100 ug/mL
1649, TiO2, cells/mL
quartz
Ambient air 2 x 105 cells/ 100 jig
particles, carbon 0.2 mL 50, 100, 150, 200 ug
black, oil fly ash,
coal fly ash
ROFA 0, 6, 12, 25, or
50 ug/mL
ROFA 2, 20, or 60 ug/cm2
Exposure
Particle Size Duration Effect of Particles
N/A 1 and 24 h Epithelial cells secreted increased mucin and lysozyme;
effect time- and concentration-dependent; caused by V-
rich fraction (18. 8%).
1.95 urn Ih ROFA induced production of acetaldehyde in dose-
dependant fashion.
0.3 -0.6 urn 18 h Cytotoxicity ranking was quartz > SRM 1649 > TiO2,
based on cellular ATP decrease and LDH, acid
phosphatase, and p-glucuronidase release.
N/A 40 min. 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.
1.96 |im 1 to 24 h 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.
1.96 |im 2or24-h Epithelial cells exposed to ROFA for 24 h secreted
exposure substantially increased amounts of the PHS products
Reference
Longphre et al.
(2000)
Madden et al.
(1999)
Nadeau et al.
(1996)
Prahalad et al.
(1999)
Quay et al.
(1998)
Samet et al.
(1996)
l-rj BEAS-2B ROFA
H
6
0
o
^ BEAS-2B ROFA
,-, Synthetic ROFA
^H (soluble Ni, Fe,
Q and V)
H
W
O
H
W
2, 20, or 60 ug/cm2
ROFA: 0-200 ug/mL
Synthetic ROFA
(100 ug/mL):
Ni, 64 uM
Fe, 63 uM
V, 370 mM
1.96 um 2or24-h Epithelial cells exposed to ROFA for 24 h secreted
exposure substantially increased amounts of the PHS products
prostaglandins E2 and F2o; ROFA-induced increase in
prostaglandin synthesis was correlated with a marked
increase in PHS activity.
ROFA: 1.96 um 5 min to 24 h Tyrosine phosphatase activity, which was known to
Synthetic ROFA: be inhibited by vanadium ions, was markedly diminished
N/A (soluble) after ROFA treatment; ROFA exposure induces
vanadium ion-mediated inhibition of tyrosine
phosphatase activity, leading to accumulation of protein
phosphotyrosines in cells.
Samet et al.
(1996)
Samet
et al. (1997)
-------
•— H
3
to
o
o
OJ
^j
ON
O
O
!>
CTJ
H
6
0
2;
o
H
O
0
H
O
Species, Cell type,
etc."
Human airway
epithelium-derived
cell lines BEAS-2B
A549
0X174 RFI DNA
Human AMs
Human AMs
Rat (Wistar) AM
RAM cells
(a rat AM cell line)
A549
TABLE 7-9 (cont'd). IN VITRO EFFECTS OF PARTICIPATE
MATTER CONSTITUENTS
Particle or Exposure
Constituent11 Cell Count Concentration Particle Size Duration
Particle 500 uM of As, F, N/A (soluble) 20 min and
components As, Cr (III), Cu, V, Zn 6 and 24 h
Cr, Cu, Fe, Ni, V,
andZn
Urban particles: 20,000 cells/cm2 Img/mLforFe SRM 1648: Up to 25 h
SRM 1648, mobilization assay 50% < 10 urn
St. Louis SRM 1649:
SRM 1649, 30% < 10 urn
Washington, DC
Provo PM10 2 x 105 cells/mL 500 ug PM10 24 h
extract
Chapel Hill PM 2 x 107 cells/mL 100 ug/mL PM2 5 24 h
extract; both H20 PM10_25
soluble(s) and
insoluble(is)
TiO2 1 x 106 cells/mL 20, 50, or N/A 4 h
80 ug/mL
ROFA, a-quartz, 2.5 x 105 1 mg/mL N/A 60 min
TiO2 cells/mL
MATTER AND PARTICULATE
Effect of Particles
Noncytotoxic concentrations of As, V, and Zn induced
a rapid phosphorylation of MAPK in cells; activity
assays confirmed marked activation of ERK, JNK, and
P38 in 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 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 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.
AM phagocytosis of (FITC)-labeled Saccharomyces
cerevisiae inhibited 30% by particles collected before
steel mill closure.
Increased cytokine production (IL-6, TNFa, MCP-1);
isPM10 > sPM10 > isPM25; sPM25 was inactive; endotoxin
was partially responsible.
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.
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.
Reference
Samet et al.
(1998)
Smith and
Aust (1997)
Soukup et al.
(2000)
Soukup and
Becker (2001)
Stringer and
Kobzik
(1996)
Stringer and
Kobzik (1998)
O
HH
H
W
-------
•— H
3
to
O
o
OJ
ON
^^
O
H
M
^^
o
H
o
H
W
O
O
H
W
TABLE 7-9 (cont'd). IN VITRO EFFECTS OF PARTICIPATE
MATTER CONSTITUENTS
Species, Cell Type, Particle or Exposure
etc." Constituent' Cell Count Concentration Particle Size Duration
A549 TiO2, Fe2O3, 3 x 105 cells/mL TiO2 [40 ug/mL], N/A 24 h
CAP, and the Fe2O3 [100
fibrogenic particle ug/mL], a-quartz
a-quartz [200 ug/mL], or
CAP [40 ng/mL]
RLE-6TN cells PM25, Burlington, 1 x 106 cells/mL 1, 2.5, 5, or PM25: 39 nm 24 and 48 h
(type II like cell VT; 10 ug/mL Fine TiO2: 159 nm exposure
line) Fine/ultrafine UF TiO2: 37 nm
TiO2
Rat, Long Evans CFA 1 x 10" cells/100 2.6 urn 3 h
epithelial cells PFA uL 17.7 urn
a-quartz. 2.5 um
BEAS-2B ROFA 100 ug/mL N/A 2-6 h
Birmingham, AL.
188mg/gofVO
NHBE Utah Valley PM10 50, 100, PM10 24 h
BEAS-2B extract 200 ug/mL
MATTER AND PARTICULATE
Effect of Particles
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.
Increases in c-Jun kinase activity, levels of
phosphorylated c-Jun immunoreactive protein, and
transcriptional activation of activator protein-
1 -dependent gene expression; elevation in number
of cells incorporating 5'-bromodeoxyuridine.
CFA produced highest level of hydro xyl radicals;
iron content is more important than quartz content.
ROFA caused increased intracellular Ca++, IL-6, IL-and
TNF-a through activation of capsicin- and pH-sensitive
receptors.
Dose-dependent increase in expression of IL-8 produced
by particles collected when the steel mill was in
operation.
Reference
Stringer et al.
(1996)
Timblin et al.
(1998)
Van Maanen
et al. (1999)
Veronesi et al.
(1999a)
Wu et al.
(2001)
"Cell types: RTE = Rat tracheal epithelial cells; GPTE = Guinea pig tracheal epithelial cells; NHBE = Normal human bronchial epithelial; A549 = Human lung epithelial cell line.
bDEF = Deferoxamine
ROFA = Residual oil fly ash
UAP = Urban air particulates
TSP = Total suspended particles
CAP = Concentrated air particles
DOFA = Domestic oil fly ash
VO = Vanadate oxide
CFA = Coal fly ash
PFA = Pulverized fuel ash
TiO2 = Titanium oxide
-------
1 7.4.2.1 Ambient Particles
2 Several studies have exposed airway epithelial cells, alveolar macrophages, or blood
3 monocytes and erythrocytes to aqueous extracts of ambient PM to investigate cellular processes
4 such as oxidant generation and cytokine production that may contribute to the
5 pathophysiological response seen in vivo. Among the ambient PM being examined were
6 samples collected from Boston, MA (Goldsmith et al., 1998); North Provo, UT (Ohio et al.,
7 1999a,b); St. Louis, MO (SRM 1648, Dong et al., 1996; Becker and Soukup, 1998); Washington,
8 DC (SRM 1649, Becker and Soukup, 1998); Ottawa, Canada (EHC-93, Becker and Soukup,
9 1998); Dusseldorf and Duisburg, Germany (Hitzfeld et al., 1997), Mexico City (Bonner et al.,
10 1998), Terni, Italy (Fabiani et al., 1997); and Rome, Italy (Diociaiuti et al., 2001). In any
11 in vitro study, however, potential exists for contamination of ambient PM by biologic material
12 during collection on filters. Endotoxin contamination, in particular, can occur at any time in the
13 manufacture of the filter media or during handling of the filter samples before, during, and after
14 the particle collection process. This potential inadvertent contamination of filter samples can
15 make extrapolation of the study results difficult, although careful handling, characterization, and
16 controls can eliminate these concerns.
17 Because soluble metals of ambient surrogates like ROFA have been associated with
18 biological effect and toxicity, several studies have investigated whether the soluble components
19 of ambient PM may have the same biological activities. Extracts of ambient PM samples
20 collected from North Provo, UT, (during 1981 and 1982) were used to test whether the soluble
21 components or ionizable metals, which accounted for approximately 0.1% of the mass, are
22 responsible for the biological activity of the extracted PM components. The oxidant generation
23 (thiobarbituric acid reactive products), release of IL-8 from BEAS-2B cells, and PMN influx in
24 rats exposed to these samples correlated with sulfate content and the ionizable concentrations of
25 metals in these PM extracts (Ohio et al., 1999a,b). In addition, these extracts stimulated IL-6
26 and IL-8 production as well as increased IL-8 mRNA and enhanced expression of intercellular
27 adhesion molecule-1 (ICAM-1) in BEAS-2B cells (Kennedy et al., 1998). Cytokine secretion
28 was preceded by activation of nuclear factor kappa B (NF-KB) and was reduced by treatment
29 with superoxide dismutase (SOD), Deferoxamine (DEF), or N-acetylcysteine. The addition of
30 similar quantities of Cu+2 as found in the Provo extract replicated the biological effects observed
31 with particles alone. When normal constituents of airway lining fluid (mucin or ceruloplasmin)
June 2003 7-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 were added to BEAS cells, particulate-induced secretion of IL-8 was modified. Mucin reduced
2 IL-8 secretion; whereas ceruloplasmin significantly increased IL-8 secretion and activation of
3 NF-KB. The authors suggest that copper ions may cause some of the biologic effects of inhaled
4 PM in the Provo region and may provide an explanation for the sensitivity of asthmatics to
5 Provo PM seen in epidemiologic studies.
6 Frampton et al. (1999) examined the effects of the same ambient PM samples collected
7 from Utah Valley in the late 1980s (see Section 7.2.1). Aqueous extracts of the filters were
8 analyzed for metal and oxidant production and added to cultures of human respiratory epithelial
9 cells (BEAS-2B) for 2 or 24 h. Particles collected in 1987, when the steel mill was closed had
10 the lowest concentrations of soluble iron, copper, and zinc and showed the least oxidant
11 generation. Ambient PM collected before and after plant closing induced expression of IL-6 and
12 IL-8 in a dose-response relationship (125, 250, and 500 |ig/mL). Ambient PM collected after
13 reopening of the steel mill also caused cytotoxicity, as demonstrated by microscopy and LDH
14 release at the highest concentration used (500 jig/mL).
15 Soukup et al. (2000) used similar ambient PM extracts as Frampton et al. (1999) to
16 examine effects on human alveolar macrophages. The phagocytic activity and oxidative
17 response of AMs was measured after segmental instillation of aqueous extracts from the Utah
18 Valley or after overnight in vitro cell culture. Ambient PM collected before closure of the steel
19 mill inhibited AM phagocytosis of (FITC)-labeled Saccharomyces cerevisiae by 30%; no
20 significant effect on phagocytosis was seen with the other two extracts. Furthermore, although
21 extracts of ambient PM collected before and after plant closure inhibited oxidant activity of AMs
22 when incubated overnight in cell culture, only the former particles caused an immediate
23 oxidative response in AMs. Host defense effects were attributed to apoptosis which was most
24 evident in particles collected before plant closure. Interpretation of loss of these effects by
25 chelation removal of the metals was complicated by the observed differences in apoptosis
26 despite similar metal contents of ambient PM collected during the steel mill operation.
27 Wu et al. (2001) investigated the intracellular signaling mechanisms for the pulmonary
28 responses to Utah Valley PM extracts. Human primary airway epithelial cells were exposed to
29 aqueous extracts of PM collected from the year before, during, and after the steel mill closure in
30 Utah Valley. Transfection with kinase-deficient extracellular signal-regulated kinase (ERK)
31 constructs partially blocked the PM-induced interleukin (IL)-8 promoter reporter activity. The
June 2003 7-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 mitogen-activated protein kinase/ERK kinase (MEK) activity inhibitor PD-98059 significantly
2 abolished IL-8 released in response to the PM, as did the epidermal growth factor (EGF)
3 receptor kinase inhibitor AG-1478. Western blotting showed that the PM-induced
4 phosphorylation of EGF receptor tyrosine, MEK1/2, and ERK1/2 could be ablated with AG-
5 1478 or PD-98059. The results indicate that the potency of Utah Valley PM collected during
6 plant closure was lower than that collected while the steel mill was in operation and imply that
7 Utah Valley PM can induce IL-8 expression partially through the activation of the EGF receptor
8 signaling.
9 There are regional as well as daily variations in the composition of ambient PM and, hence,
10 its biological activities. For example, concentrated ambient PM (CAP, from Boston urban air)
11 has substantial day-to-day variability in its composition and oxidant effects (Goldsmith et al.,
12 1998). Similar to Utah PM, the water-soluble component of Boston CAPs significantly
13 increased AM oxidant production and inflammatory cytokine (MIP2 and TNFcc) production over
14 negative control values. These effects can be blocked by metal chelators or antioxidants. The
15 regional difference in biological activity of ambient PM has been shown by Becker and Soukup
16 (1998). The oxidant generation, phagocytosis, as well as the expressions of receptors important
17 for phagocytosis in human alveolar macrophage and blood monocyte were reduced significantly
18 by PM exposure.
19 Becker and Soukup (1998) and others (Dong et al., 1996, Becker et al., 1996) have
20 suggested that the biological activity of the ambient PM may result from the presence of
21 endotoxin on the particles rather than metal-associated oxidant generation. Using the same
22 urban particles (SRM 1648), cytokine production (TNF-cc, IL-1,11-6, CINC, and MIP-2) was
23 increased in macrophages following treatment with 50 to 200 |ig/mL of urban PM (Dong et al.,
24 1996). The urban particle-induced TNF-cc secretion was abrogated completely by treatment with
25 polymyxin B, an antibiotic that blocks LPS-associated activities, but not with antioxidants.
26 The involvement of endotoxin, at least partially, in PM induced biological effects was
27 supported more recently by Bonner et al. (1998) and Soukup and Becker (2001). Urban PM10
28 collected from north, south, and central regions of Mexico City was used with SD rat AM to
29 examine PM effects on platelet-derived growth factor (PDGF) receptors on lung myofibroblasts
30 (Bonner et al., 1998). Mexico City PM10 (but not volcanic ash) stimulated secretion of
31 upregulatory factors for the PDGF a, receptor, possibly via IL-1 p. In the presence of an
June 2003 7-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 endotoxin-neutralizing protein, the Mexico City PM10 effect on PDGF was blocked partially,
2 suggesting that LPS was responsible partially for the effect of the PM10 on macrophages.
3 In addition, both LPS and vanadium (both present in the PM10) acted directly on lung
4 myofibroblasts. However, the V levels in Mexico City PM10 were probably not high enough to
5 exert an independent effect. The authors concluded that PM10 exposure could lead to airway
6 remodeling by enhancing myofibroblast replication and chemotaxis.
7 Soukup and Becker (2001) collected fresh PM25 and PM10_25 from the ambient air of
8 Chapel Hill, NC, and compared the activity of these two particle size fractions. Both water
9 soluble and insoluble components were assessed for cytokine production, inhibition of
10 phagocytosis, and induction of apoptosis. The most potent fraction was the insoluble PM10_2 5
11 thus suggesting the importance of the coarse fraction in the investigation of ambient PM's health
12 effects. Endotoxin was responsible for much of the cytokine production, while inhibition of
13 phagocytosis was induced by other moieties in the coarse material. None of the activities were
14 inhibited by the metal chelator deferoxamine.
15 The effects of water soluble as well as organic components (extracted in dichloromethane)
16 of ambient PM were investigated by exposing human PMN to PM extracts (Hitzfeld et al.,
17 1997). PM was collected with high-volume samplers in two German cities, Dusseldorf and
18 Duisburg; these sites have high traffic and high industrial emissions, respectively. Organic, but
19 not aqueous, extracts of PM alone significantly stimulated production and release of ROS in
20 resting human PMN. The effects of the PM extracts were inhibited by SOD, catalase, and
21 sodium azide (NaN3). Similarly, the organic fraction (extractable by acetone) of ambient PM
22 from Terni, Italy, was shown to produce cytotoxicity, superoxide release in response to PMA
23 and zymosan in peripheral monocytes (Fabiani et al., 1997).
24 Diociaiuti et al. (2001) compared the in vitro toxicity of coarse (PM10_25) and fine (PM25)
25 particulate matter, collected in an urban area of Rome. The in vitro toxicity assays used included
26 human red blood cell hemolysis, cell viability, and nitric oxide (NO) release in the RAW 264.7
27 macrophage cell line. There was a dose-dependent hemolysis in human erythrocytes when they
28 were incubated with fine and coarse particles. The hemolytic potential was greater for the fine
29 particles than for the coarse particles in equal mass concentration. However, when data were
30 expressed in terms of PM surface area per volume of suspension, the hemolytic activity of the
31 fine fraction was equal to the coarse fraction. This result suggested that the oxidative stress
June 2003 7-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 induced by PM on the cell membranes could be due mainly to the interaction between the
2 particle surfaces and the cell membranes. Although RAW 264.7 cells challenged with fine and
3 coarse particles showed decreased viability and an increased release of NO, a key inflammatory
4 mediator, both effects were not dose-dependent in the tested concentration range. The fine
5 particles were the most effective in inducing these effects when the data were expressed as mass
6 concentration or as surface area per unit volume. The authors concluded that these differences in
7 biological activity were due to the differing physicochemical nature of the particles.
8
9 7.4.2.2 Comparison of Ambient and Combustion-Related Surrogate Particles
10 In vitro toxicology studies utilizing alveolar macrophages as target cells (Imrich et al.,
11 2000; Long et al., 2001; Ning et al., 2000; Mukae et al., 2000, 2001; Van Eeden et al., 2001)
12 have found that urban air particles are much more potent for inducing cellular responses than
13 individual combustion particles such as diesel and ROFA. Similar to the results described above
14 in Section 7.5.2.1, these studies also show that when cytokine responses are measured,
15 LPS/endotoxin is found to be responsible for most of the activity. Metals, on the other hand, do
16 not seem to affect cytokine production, as confirmed by studies showing that ROFA does not
17 induce macrophage cytokine production. These results are important because LPS is an
18 important component associated with both coarse and fine particles (Menetrez et al., 2001).
19 In fact, in one study (Long et al., 2001), cytokine responses in the alveolar macrophages were
20 correlated with LPS content and more LPS was found associated with indoor PM2 5 than outdoor
21 PM25.
22 Imrich et al. (2000) found that when mice alveolar macrophages were stimulated with
23 CAPs (PM2 5), the resulting TNF responses could be inhibited by the use of an endotoxin
24 neutralizing agent [e.g., polymyxin-B (PB)]. Because the MIP-2 response (IL-8) was only partly
25 inhibited by PB; however, the authors concluded that endotoxin primed cells to respond to other
26 particle components. In a related study (Ning et al., 2000), the use of PB showed that particle-
27 absorbed endotoxin in CAPs suspensions caused activation of normal (control) AMs, while other
28 (nonendotoxin) components were predominantly responsible for the enhanced cytokine release
29 observed by primed AMs incubated with CAPs. The non-LPS component was not identified in
30 this study, however, the AM biological response did not correlate with any of a panel of
June 2003 7-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 elements quantified within the insoluble CAPs samples (e.g., Al, Cd, Cr, Cu, Fe, Mg, Mn, Ni, S,
2 Ti, V).
3 Van Eeden et al. (2001) compared ROFA, the atmospheric dust sample EHC-93, and
4 different size latex particles for cytokine induction on human alveolar macrophages. The
5 EHC-93 particles produced greater than 8-fold induction of various cytokines, including IL-1,
6 TNF, GMCSF; the other particles induced these cytokines approximately 2-fold. Using the same
7 EHC-93 particles, Mukae et al. (2000, 2001) found that inhalation exposure stimulated bone
8 marrow band cell-granulocyte precursor production. They also found that the magnitude of the
9 response was correlated with the amount of phagocytosis of the particles by alveolar
10 macrophages. These results may indicate that macrophages produce factors which stimulate
11 bone marrow, including IL-6 and GMCSF. In fact, alveolar macrophages exposed in vitro to
12 these particles released cytokines; and when the supernatant of PM-stimulated macrophages was
13 instilled into rabbits, the bone marrow was stimulated.
14 In a series of studies using the same ROFA samples, several in vitro experiments have
15 investigated the biochemical and molecular mechanisms involved in ROFA induced cellular
16 injury. Prostaglandin metabolism in cultured human airway epithelial cells (BEAS-2B and
17 NHBE) exposed to ROFA was investigated by Samet et al. (1996). Epithelial cells exposed to
18 ROFA for 24 h secreted substantially increased amounts of prostaglandins E2 and F2 a. The
19 ROFA-induced increase in prostaglandin synthesis was correlated with a marked increase in
20 activity of the prostaglandin H synthase-2 (PHS-2) as well as mRNA coded for this enzyme.
21 In contrast, expression of the PHS1 form of the enzyme was not affected by ROFA treatment of
22 airway epithelial cells. These investigators further demonstrated that noncytotoxic levels of
23 ROFA induced a significant dose- and time-dependent increase in protein tyrosine phosphate, an
24 important index of signal transduction activation leading to a broad spectrum of cellular
25 responses. ROFA-induced increases in protein phosphotyrosines were associated with its
26 soluble fraction and were mimicked by V-containing solutions but not iron or nickel solutions
27 (Samet etal., 1997).
28 ROFA also stimulates respiratory cells to secrete inflammatory cytokines such as IL-6,
29 IL-8, and TNF. Normal human bronchial epithelial (NHBE) cells exposed to ROFA produced
30 significant amounts of IL-8, IL-6, and TNF, as well as mRNAs coding for these cytokines
31 (Carter et al., 1997). Increases in cytokine production were dose-dependent. The cytokine
June 2003 7-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 production was inhibited by the addition of metal chelator, DEF, or the free radical scavenger
2 dimethylthiourea (DMTU). Similar to the data of Samet et al. (1997), V but not Fe or Ni
3 compounds were responsible for these effects. Cytotoxicity, decreased cellular glutathione
4 levels in primary cultures of rat tracheal epithelial (RTE) cells exposed to suspensions of ROFA
5 indicated that respiratory cells exposed to ROFA were under oxidative stress. Treatment with
6 buthionine sulfoxamine (an inhibitor of y-glutamyl cysteine synthetase) augmented ROFA-
7 induced cytotoxicity; whereas treatment with DMTU inhibited ROFA-induced cytoxicity further
8 suggested that ROFA-induced cell injury may be mediated by hydroxyl-radical-like reactive
9 oxygen species (ROS) (Dye et al., 1997). Using BEAS-2B cells, a time- and dose-dependent
10 increase in IL-6 mRNA induced by ROFA was shown to be preceded by the activation of
11 nuclear proteins, for example, nuclear factor-KB (NF-KB) (Quay et al., 1998). Taken together,
12 ROFA exposure increases oxidative stress, perturbs protein tyrosine phosphate homeostasis,
13 activates NF-KB, and up-regulates inflammatory cytokine and prostaglandin synthesis and
14 secretion to produce lung injury.
15 Stringer and Kobzik (1998) observed that "primed" lung epithelial cells exhibited
16 enhanced cytokine responses to PM. Compared to normal cells, exposure of tumor necrosis
17 factor (TNF)-cc-primed A549 cells to ROFA or a -quartz caused increased IL-8 production in a
18 concentration-dependent manner for particle concentrations ranging from 0-200 |ig/mL.
19 Addition of the antioxidant N-acetylcysteine (NAC) (1.0 mM) decreased ROFA and cc -quartz-
20 mediated IL-8 production by approximately 50% in both normal and TNF-cc-primed A549 cells.
21 Exposure of A549 cells to ROFA caused an increase in oxidant levels that could be inhibited by
22 NAC. These data suggest that (1) lung epithelial cells primed by inflammatory mediators show
23 increased cytokine production after exposure to PM and (2) oxidant stress is an important
24 mechanism for this response.
25 In summary, exposure of lung epithelial cells to ambient PM or ROFA leads to increased
26 production of cytokines and the effects may be mediated, at least in part, through production of
27 ROS. Day-to-day variations in the components of PM, such as soluble transition metals (which
28 may be critical to eliciting the response) are suggested. The involvement of organic components
29 in ambient PM was also suggested by some studies.
30
31
June 2003 7-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.4.2.3 Mutagenicity
2 The majority of recent PM research has focused on the acute cardiopulmonary effects
3 which have been documented to occur following episodic exposure to ambient PM. However,
4 epidemiologic investigations have recently linked chronic exposure to ambient PM not only to
5 increases in long term cardiopulmonary mortality but also to lung cancer effects (Pope, 2002).
6 Also, a limited number of recent studies have examined the mutagenic potential of ambient PM
7 and, in general, they have shown some degree of evidence that appears to support the biologic
8 plausibility of the long-term lung cancer effects.
9 These in vitro studies, discussed in Table 7-10, have focused on the ability of the organic
10 fraction of ambient PM to induce mutagenic effects in mammalian cell lines and bacteria. The
11 organic fractions produced increase mutations (revertants) in the Ames assay (Bunger, 2000) as
12 well as sister chromatid exchanges in mammalian cells (Hornberg, 1996, 1998). Seemayer and
13 colleagues (1998) also observed increases in SV40 transformation of hamster kidney cells
14 treated with extracts of ambient PM collected with a high volume sampler. Investigators have
15 also compared the mutagenic potential of the combustion products of high and low sulfur
16 content diesel fuel with plant derived fuels. In the Ames assay, the number of revertants was
17 significantly elevated in bacteria treated with high versus low sulfur diesel fuel. Moreover, the
18 high sulfur fuel caused more mutations than the plant-derived fuels.
19 Although each of the above studies demonstrates the mutagenic potential of ambient PM
20 and fuel combustion products, these in vitro studies are generally lacking in details regarding the
21 dose of PM extract delivered to the cells in vitro. In general, equal volumes of air or amounts of
22 time were sampled, but little to no characterization of the amount of PM mass or size were
23 determined. Thus, the relevance of these mutagenicity studies is still quite limited in terms of
24 substantiating the biologic plausibility of, or elucidating potential mechanisms underlying, the
25 reported associations between long-term exposure to PM and increases in lung cancer.
26
27 7.4.3 Potential Cellular and Molecular Mechanisms
28 7.4.3.1 Reactive Oxygen Species
29 Ambient particulate matter contains transition metals, such as iron (most abundant),
30 copper, nickel, zinc, vanadium, and cobalt. These metals are capable of catalyzing the
31 one-electron reductions of molecular oxygen necessary to generate reactive oxygen
June 2003 7-69 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-10. MUTAGENIC/CARCINOGENIC EFFECTS OF PARTICIPATE MATTER
to
O
o
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Particle
Ambient PM
Ambient PM10 and
PM25 collected in
industrial and rural
regions
Ambient particles and
particles from diesel
exhaust, rubber and
metal industries, and
biologic sources
(poultry/swine farming)
Ambient PM
Diesel exhaust particles
Species, Gender,
Strain Age, or Body Exposure
Weight Technique
Cultured tracheal in vitro
epithelial cells from
Hamster, Syrian
golden, young
Human in vitro
bronchioepithelial
cell line (BEAS-2B)
Liver tumor cell line in vitro
(HEPAlclc?)
Kidney cells from in vitro
hamster,
Syrian golden,
8-10 weeks old
Ames assay with in vitro
and without
activation
Mass
Concentration Particle
(|ig/mL) or Characteristics
(jig/m ) Size (Jim); ^ig
Not given Dichloromethane
extraction of high
volume samples.
Not given Dichloromethane
in |_ig/mL extraction of coarse
(PM10) and fine
(PM25) fractions.
6 to 226 |_ig/mL Aqueous and organic
extraction of
particles collected
with high volume
samplers.
Not given Dichloromethane
extraction of high
volume samples.
Not given Dichloromethane
extraction of
particles collected
Exposure Duration
Dilutions of extracted
organic phase of
particles incubated with
cells for 48 hours.
Dilutions of extracted
organic phase of size-
segregated particles
incubated with cells for
72 hours.
Not given.
Dilutions of extracted
organic phase of
particles incubated with
cells for 18 hours
followed by infection
with simian virus
SV-40.
48 hours incubation
with TA98 and TA100
strains.
Adverse Effects of Particles
on Mammalian Cells or
Bacteria
Dose-related increases in sister
chromatid exchanges were
observed.
Significant increases in sister
chromatid exchanges were
greater in PM2 5 from all
sampling sites. Extraction
phase of coarse particles
produced fewer sister
chromatid exchanges than did
the fine particles.
Inhibition of gap-junctional
intercellular communication
was significant only in cells
treated with aqueous extract of
diesel, compost, or rubber
particles.
Significantly greater SV-40-
induced transformation of
hamster kidney cells
pre-treated with organic
extractions of urban particles.
Revertants were 2 to 10-fold
higher with high sulfur diesel
fuel particles.
Reference
Hornberg
(1996)
Hornberg
(1998)
Alink
(1998)
Seemayer
and
Hornberg
(1998)
Hunger
(2000)
Ambient PM
Cultured hepatoma
cells
in vitro
Not given
from diesel engine
run with diesel fuels
with low or high
sulfur and 2 plant oil
fuels.
Acetone/dichloromet
hane extraction of
high volume
samples.
Dilutions of extracted
organic phase of
particles incubated with
cells for 6 or 48 hours.
Extracts of ambient PM both
upwind and downwind of
highway have genotoxic
effects although PAH content
was greater in downwind
samples.
Hamers
(2000)
-------
1 species (ROS). These reactions can be demonstrated by the iron-catalyzed Haber-Weiss
2 reactions that follow.
Reductant11 + Fe(III) -> Reductantn+1 + Fe(II) (1)
3
4 Fe(II) + O2 -> Fe(III) + O2 (2)
5
6 HO2 + O2 + H+ ^ O2 + H2O2 (3)
7
8 Fe(II) + H2O2 -> Fe(III)+ *OH + HCT (Fentan Reaction) (4)
9
10 Iron will continue to participate in the redox cycle in the above reactions as long as there is
11 sufficient O2 or H2O2 and reductants.
12 Soluble metals from inhaled PM dissolved into the fluid lining of the airway lumen can
13 react directly with biological molecules (acting as a reductant in the above reactions) to produce
14 ROS. For example, ascorbic acid in the human lung epithelial lining fluid can react with Fe(III)
15 from inhaled PM to cause single strand breaks in supercoiled plasmid DNA, c|)X174 RFI (Smith
16 and Aust, 1997). The DNA damage caused by a PM10 suspension can be inhibited by mannitol,
17 an hydroxyl radical scavenger, further confirming the involvement of free radicals in these
18 reactions (Gilmour et al., 1996; Donaldson et al., 1997; Li et al., 1997). Because the clear
19 supernatant of the centrifuged PM10 suspension contained all of the suspension activity, the free
20 radical activity is derived either from a fraction that is not centrifugable (10 min at 13,000 rpm
21 on a bench centrifuge) or the radical generating system is released into solution (Gilmour et al.,
22 1996; Donaldson et al., 1997; Li et al., 1997).
23 In addition to measuring the interactions of ROS and biomolecules directly, the role of
24 ROS in PM-induced lung injury also can be assessed by measuring the electron spin resonance
25 (ESR) spectrum of radical adducts or fluorescent intensity of dichlorofluorescin (DCFH), an
26 intracellular dye that fluoresces on oxidation by ROS. Alternatively, ROS can be inhibited using
27 free radical scavengers, such as dimethylthiourea (DMTU); antioxidants, such as glutathione or
28 N-acetylcysteine (NAC); or antioxidant enzymes, such as superoxide dismutase (SOD). The
29 diminished response to PM after treatment with these antioxidants may indicate the involvement
June 2003 7-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 of ROS; however, some antioxidants (e.g., thiol-containing) can interact with metal ions
2 directly.
3 As described earlier, Kadiiska et al. (1997) used the ESR spectra of 4-POBN [cc-(4-pyridyl
4 l-oxide)-N-tert-butylnitrone] adducts to measure ROS in rats instilled with ROFA and
5 demonstrated the association between ROS production within the lung and soluble metals in
6 ROFA. Using DMTU to inhibit ROS production, Dye et al. (1997) had shown that systemic
7 administration of DMTU impeded development of the cellular inflammatory response to ROFA,
8 but did not ameliorate biochemical alterations in BAL fluid. Goldsmith et al. (1998), as
9 described earlier, showed that ROFA and CAPs caused increases in ROS production in AMs.
10 The water-soluble component of both CAPs and ROFA significantly increased AM oxidant
11 production over negative control values. In addition, increased PM-induced cytokine production
12 was inhibited by NAC. Li et al. (1996, 1997) instilled rats with PM10 particles (collected on
13 filters from an Edinburgh, Scotland, monitoring station). Six hours after intratracheal instillation
14 of PM10, they observed a decrease in glutathione (GSH) levels in the BAL fluid. Although this
15 study does not describe the composition of the PM10, the authors suggest that changes in GSH,
16 an important lung antioxidant, support the contention that the free radical activity of PM10 is
17 responsible for its biological activity in vivo.
18 In addition to ROS generated directly by PM, resident or newly recruited AMs or PMNs
19 also are capable of producing these reactive species on stimulation. The ROS produced during
20 the oxidative burst can be measured using a chemiluminescence (CL) assay. With this assay,
21 AM CL signals in vitro have been shown to be greatest with ROFA containing primarily soluble
22 V and were less with ROFA containing Ni plus V (Kodavanti et al., 1998a). As described
23 earlier, exposures to Dusseldorf and Duisburg PM increased the resting ROS production in
24 PMNs, which could be inhibited by SOD, catalase, and sodium azide (Hitzfeld et al., 1997).
25 Stringer and Kobzik (1998) showed that addition of NAC (1.0 mM) decreased ROFA-mediated
26 IL-8 production by approximately 50% in normal and TNF-cc-primed A549 cells. In addition,
27 exposures of A549 cells to ROFA caused a substantial (and NAC inhibitable) increase in oxidant
28 levels as measured by DCFH oxidation. In human AMs, Becker et al. (1996) found a CL
29 response for ROFA, but not urban air particles (Ottawa and Dusseldorf) or volcanic ash.
30 Metal compounds of PM are the most probable species capable of catalyzing ROS
31 generation on exposure to PM. To determine elemental content and solubility in relation to their
32 ability to generate ROS, PMN or monocytes were exposed to a wide range of ambient air
June 2003 7-72 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles from divergent sources (one natural dust, two types of oil fly ash, two types of coal fly
2 ash, five different ambient air samples, and one carbon black sample), and CL production was
3 measured over a 20-min period postexposure (Prahalad et al., 1999). Percent of sample mass
4 accounted for by XRF detectable elements was 1.2% (carbon black); 22 to 29% (natural dust and
5 ambient air particles); 13 to 22% (oil fly ash particles); and 28 to 49% (coal fly ash particles).
6 The major proportion of elements in most of these particles were aluminosilicates and insoluble
7 iron, except oil derived fly ash particles in which soluble vanadium and nickel were in highest
8 concentration, consistent with particle acidity as measured in the supernatants. All particles
9 induced CL response in cells, except carbon black. The CL response of PMNs in general
10 increased with all washed particles, with oil fly ash and one urban air particle showing statistical
11 differences between deionized water washed and unwashed particles. These CL activities were
12 significantly correlated with the insoluble Si, Fe, Mn, Ti, and Co content of the particles.
13 No relationship was found between CL and soluble transition metals such as V, Cr, Ni, and Cu.
14 Pretreatment of the particles with a metal ion chelator, deferoxamine, did not affect CL
15 activities. Particle sulfate content and acidity of the particle suspension did not correlate with
16 CL activity.
17 Soluble metals can be mobilized into the epithelial cells or AMs to produce ROS
18 intracellularly. Size-fractionated coal fly ash particles (2.5, 2.5 to 10, and < 10 |im) of
19 bituminous b (Utah coal), c (Illinois coal), and lignite (Dakota coal) were used to compare the
20 amount of iron mobilization in A549 cells and by citrate (1 mM) in cell-free suspensions (Smith
21 et al., 1998). Iron was mobilized by citrate from all three size fractions of all three coal types.
22 More iron, in Fe(III) form, was mobilized by citrate from the < 2.5-|im fraction than from the
23 > 2.5-|im fractions. In addition, the amount of iron mobilized was dependent on the type of coal
24 used to generate the fly ash (Utah coal > Illinois coal = Dakota coal) but was not related to the
25 total amount of iron present in the particles. Ferritin (an iron storage protein) levels in A549
26 cells increased by as much as 12-fold in cells treated with coal fly ash (Utah coal > Illinois
27 coal > Dakota coal). More ferritin was induced in cells treated with the < 2.5-|im fraction than
28 with the > 2.5-jim fractions. Mossbauer spectroscopy of a fly ash sample showed that the
29 bioavailable iron was assocated with the glassy aluminosilicate fraction of the particles (Ball
30 et al., 2000). As with the bioavailability of iron, there was an inverse correlation between the
31 production of IL-8 and fly ash particle size, with the Utah coal fly ash being the most potent.
June 2003 7-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 Using ROFA and colloidal iron oxide, Ohio et al. (1997b; 1998a,b,c; 1999c; 2000c) have
2 shown that exposures to these particles disrupted iron homeostasis and induced the production of
3 ROS in vivo and in vitro. Treatment of animals or cells with metal-chelating agents such as
4 DEF with an associated decrease in response has been used to infer the involvement of metal in
5 PM-induced lung injury. Metal chelation by DEF (1 mM) caused significant inhibition of
6 particulate-induced AM oxidant production, as measured using DCFH (Goldsmith et al., 1998).
7 DEF treatment also reduced NF-icB activation and cytokine secretion in a human bronchial
8 epithelial cell line (BEAS-2B cells) exposed to Provo PM (Kennedy et al., 1998). However,
9 treatment of ROFA suspension with DEF was not effective in blocking teachable metal induced
10 acute lung injury (Dreher et al., 1997). Dreher et al. (1997) indicated that DEF could chelate
11 Fe(III) and V(II), but not Ni(II), suggesting that metal interactions played a significant role in
12 ROFA-induced lung injury.
13 Other than Fe, several V compounds have been shown to increase mRNA levels for
14 selected cytokines in BAL cells and induce pulmonary inflammation (Pierce et al., 1996).
15 NaVO3 and VOSO4, highly soluble forms of V, tended to induce pulmonary inflammation and
16 inflammatory cytokine mRNA expression more rapidly and more intensely than the less soluble
17 form, V2O5, in rats. Neutrophil influx was greatest following exposure to VOSO4 and lowest
18 following exposure to V2O5. However, metal components of fly ash have not been shown to
19 consistently increase ROS production from bovine AM treated with combustion particles
20 (Schliiter et al., 1995). For example, As(III), Ni(II), and Ce(III), which are major components of
21 fly ash, had been shown to inhibit the secretion of superoxide anions (O2") and hydrogen
22 peroxide. In the same study, O2" were lowered by Mn(II) and Fe(II); whereas V(IV) increased
23 O2" and H2O2. In contrast, Fe(III) increased O2" production, demonstrating that the oxidation state
24 of metal may influence its oxidant generating properties. Other components of fly ash, such as
25 Cd(II), Cr(III), and V(V), had no effects on ROS.
26 It is likely that a combination of several metals rather than a single metal in PM is
27 responsible for the PM-induced cellular response. For example, V and Ni+V but not Fe or Ni
28 alone (in saline with the final pH at 3.0) resulted in increased epithelial permeability, decreased
29 cellular glutathione, cell detachment, and lytic cell injury in rat tracheal epithelial cells exposed
30 to soluble salts of these metals at equivalent concentrations found in ROFA (Dye et al., 1999).
31 Treatment of V-exposed cells with buthionine sulfoximine further increased cytotoxicity.
32 Conversely, treatment with radical scavenger dimethyl thiourea inhibited the effects in a
June 2003 7-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 dose-dependent manner. These results suggest that soluble metal or combinations of several
2 metals in ROFA may be responsible for these effects.
3 Similar to combustion particles such as ROFA, the biological response to exposure to
4 ambient PM also may be influenced by the metal content of the particles. Human subjects were
5 instilled with 500 jig (in 20 mL sterile saline) of Utah Valley dust (UVD1, 2, 3, collected during
6 3 successive years) on the left segmental bronchus and on the right side with sterile saline as
7 control. A second bronchoscopy was performed 24 hours post-instillation and phagocytic cells
8 were obtained from the segmental bronchi on both sides. Alveolar macrophage from subjects
9 instilled with UVD, obtained by bronchoaveolar lavage 24 h post-instillation, were incubated
10 with fluoresceinated yeast (Saccharomyces cerevisiae) to assess their phagocytic ability.
11 Although the same proportion of AMs were exposed to UVD phagocytized yeast, AMs exposed
12 to UVD1, which were collected while a local steel mill was open, took up significantly less
13 particles than AMs exposed to other extracts (UVD2 when the steel mill was closed and UVD3
14 when the plant reopened). AMs exposed to UVD1 also exhibited a small decrease in oxidant
15 activity (using dihydrorhodamine-123, DFIR). AMs from healthy volunteers were incubated
16 in vitro with the various UVD extracts to assess whether similar effects on human AMs function
17 could be observed to those seen following in vivo exposure. The percentage of AMs that
18 engulfed yeast particles was significantly decreased by exposure to UVD1 at 100 jig/mL, but not
19 at 25 |ig/mL. However, the amount of particles engulfed was the same following exposure to all
20 three UVD extracts. AMs also demonstrated increased oxidant stress (using
21 chemiluminescence) after in vitro exposure to UVD1, and this effect was not abolished with
22 pretreatment of the extract with the metal chelator deferoxamine. As with the AMs exposed to
23 UVD in vivo, AM exposed to UVD in vitro had a decreased oxidant activity (DHR assay).
24 UVD1 contains 61 times and 2 times the amount of Zn compared to UVD 2 and UVD3,
25 respectively; whereas UVD3 contained 5 times more Fe than UVD1. Ni and V were present
26 only in trace amounts. Using similarly extracted samples, Frampton et al. (1999) exposed
27 BEAS-2B cells for 2 and 24 h. Similar results were observed for oxidant generation in these
28 cells (i.e., UVD 2, which contains the lowest concentrations of soluble iron, copper, and zinc,
29 produced the least response). Only UVD 3 produced cytotoxicity at a dose of 500 |ig/mL. UVD
30 1 and 3, but not 2, induced expression of IL-6 and 8 in a dose-dependent fashion. Taken
31 together, the above results showed that the biological response to ambient particle extracts is
32 heavily dependent on the source and, hence, the chemical composition of PM.
June 2003 7-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.4.3.2 Intracellular Signaling Mechanisms
2 In has been shown that the intracellular redox state of the cell modulates the activity of
3 several transcription factors, including NF-KB, a critical step in the induction of a variety of
4 proinflammatory cytokine and adhesion-molecule genes. NF-KB is a heterodimeric protein
5 complex that in most cells resides in an inactive state in the cell cytoplasm by binding to
6 inhibitory kappa B alpha (LcEcc). On appropriate stimulation by cytokines or ROS, LcBa is
7 phosphorylated and subsequently degraded by proteolysis. The dissociation of LcBa from NF-
8 KB allows the latter to translocate into the nucleus and bind to appropriate sites in the DNA to
9 initiate transcription of various genes. Two studies in vitro have shown the involvement of
10 NF-KB in particulate-induced cytokine and intercellular adhesion molecule-1 (ICAM-1)
11 production in human airway epithelial cells (BEAS-2B) (Quay et al., 1998; Kennedy et al.,
12 1998). Cytokine secretion was preceded by activation of NF-KB and was reduced by treatment
13 with antioxidants or metal chelators. These results suggest that metal-induced oxidative stress
14 may play a significant role in the initiation phase of the inflammatory cascade following PM
15 exposure.
16 A second well-characterized human transcription factor, AP-1, also responds to the
17 intracellular ROS concentration. AP-1 exists in two forms, either in a homodimer of c-jun
18 protein or a heterodimer consisting of c-jun and c-fos. Small amounts of AP-1 already exist in
19 the cytoplasm in an inactive form, mainly as phosphorylated c-jun homodimer. Many different
20 oxidative stress-inducing stimuli, such as UV light and IL-1, can activate AP-1. Exposure of rat
21 lung epithelial cells to ambient PM in vitro resulted in increases in c-jun kinase activity, levels of
22 phosphorylated c-jun immunoreactive protein, and transcriptional activation of AP-1-dependent
23 gene expression (Timblin et al., 1998). This study demonstrated that interaction of ambient
24 particles with lung epithelial cells initiates a cell signaling cascade related to aberrant cell
25 proliferation.
26 Early response gene transactivation has been linked to the development of apoptosis,
27 a potential mechanism to account for PM-induced changes in cellular response. Apoptosis of
28 human AMs exposed to ROFA (25 |ig/mL) or urban PM was observed by Holian et al. (1998).
29 In addition, both ROFA and urban PM upregulated the expression of the RFD1+ AM phenotype;
30 whereas only ROFA decreased the RFDl+7+ phenotype. It has been suggested that an increase in
31 the AM phenotype ratio of RFDl+/RFDl+7+ may be related to disease progression in patients
32 with inflammatory diseases. These data showed that ROFA and urban PM can induce apoptosis
June 2003 7-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 of human AMs and increase the ratio of AM phenotypes toward a higher immune active state
2 and may contribute to or exacerbate lung inflammation.
3 Inhaled fine and coarse particles are trapped by impaction in the epithelial lining of the
4 nasal and tracheal airways. Somatosensory neurons located in the dorsal root ganglia (DRG)
5 innervate the upper thoracic region of the airways and extend their terminals over and between
6 the epithelial lining of the lumen. Given this anatomical proximity, the sensory fibers and the
7 tracheal epithelial cells that they innervate encounter inhaled pollutants, such as PM, early
8 during inhalation. The differential responses of these cell types to PM derived from various
9 sources (i.e., industrial, residential, volcanic) were examined with biophysical and
10 immunological endpoints (Veronesi et al., 2002a). Although the majority of PM tested
11 stimulated IL-6 release in both BEAS-2B epithelial cells and DRG neurons in a receptor-
12 mediated fashion, the degree of these responses was markedly higher in sensory neurons.
13 Epithelial cells are damaged or denuded in many common health disorders (e.g., asthma, viral
14 infections), allowing PM particles to directly encounter the sensory terminals and their acid-
15 sensitive receptors.
16 Another intracellular signaling pathway that could lead to diverse cellular responses such
17 as cell growth, differentiation, proliferation, apoptosis, and stress responses to environmental
18 stimuli, is the phosphorylation-dependent, mitogen-activated protein kinase (MAPK).
19 Significant dose- and time-dependent increases in protein tyrosine phosphate levels have been
20 seen in BEAS cells exposed to 100 |ig/mL ROFA for periods ranging from 5 min to 24 h (Samet
21 et al., 1997). In a subsequent study, the effects of As, Cr, Cu, Fe, Ni, V, and Zn on the MAPK,
22 extracellular receptor kinase (ERK), c-jun N-terminal kinase (INK), and P38 in BEAS cells were
23 investigated (Samet et al., 1998). Arsenic, V, and Zn induced a rapid phosphorylation of MAPK
24 in BEAS cells. Activity assays confirmed marked activation of ERK, INK, and P38 in BEAS
25 cells exposed to As, V, and Zn; Cr and Cu exposure resulted in a relatively small activation of
26 MAPK; whereas Fe and Ni did not activate MAPK. Similarly, the transcription factors c-Jun
27 and ATF-2, substrates of INK and P38, respectively, were markedly phosphorylated in BEAS
28 cells treated with As, Cr, Cu, V, and Zn. The same acute exposure to As, V, or Zn that activated
29 MAPK was sufficient to induce a subsequent increase in IL-8 protein expression in BEAS cells.
30 All exposures were non-cytotoxic based on measurement of lactate dehydrogenase release and
31 microscopic examination of trypan blue or propidium iodide exclusion (Samet et al., 1996).
32 These data suggest that MAPK may mediate metal-induced expression of inflammatory proteins
June 2003 7-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 in human bronchial epithelial cells. The ability of ROFA to induce activation of MAPKs in vivo
2 was demonstrated by Silbajoris et al. (2000; see Table 7-3). In addition, Gercken et al. (1996)
3 showed that the ROS production induced by PM was markedly decreased by the inhibition of
4 protein kinase C as well as phospholipase A2. Comparisons of in vitro and in vivo exposures of
5 ROFA to airway epithelial cells requires consideration of in vivo dosimetry and ambient
6 concentrations. Therefore, such extrapolations must be made with caution.
7 The major cellular response downstream of ROS and the cell signaling pathways described
8 above is the production of inflammatory cytokines or other reactive mediators. In an effort to
9 determine the contribution of cyclooxygenase to the pulmonary responses to ROFA exposure
10 in vivo, Samet et al. (2000) intratracheally instilled Sprague-Dawley rats with ROFA (200 or
11 500 jig in 0.5 mL saline). These animals were pretreated ip with 1 mg/kg NS398, a specific
12 prostaglandin H synthase 2 (COX2) inhibitor, 30 min prior to intratracheal exposure. At 12 h
13 after intratracheal instillations, ip injections (1 mL of NS398 in 20% ethanol in saline) were
14 repeated. ROFA treatment induced a marked increase in the level of PGE2 recovered in the BAL
15 fluid, which was effectively decreased by pretreating the animals with the COX2 inhibitor.
16 Immunohistochemical analyses of rat airway showed concomitant expression of COX2 in the
17 proximal airway epithelium of rats treated with soluble fraction of ROFA. This study further
18 showed that, although COX2 products participated in ROFA induced lung inflammation, the
19 COX metabolites are not involved in IL-6 expression nor the influx of PMN influx into the
20 airway. However, the rationale for the use of intraperitoneal challenge was not elaborated.
21 The production of cytokines and mediators also has been shown to depend on the type of
22 PM used in the experiments. A549 cells (a human airway epithelial cell line) were exposed
23 in vitro to several particulate materials: carbon black (CB, Elftex-12, Cabot Corp.), diesel soot
24 from two sources (ND from NIST, LD produced from General Motors LH 6.2 V8 engine at light
25 duty cycle), ROFA (from the heat exchange section of the Boston Edison), OAA (Ottawa
26 ambient air PM, EHC-93), SiO2, andM3S2 at 0.01, 0.03, 0.1, 0.3, 1.0, 3.0, 100, 300, 1,000
27 jig/cm2 for 18 h (Seagrave and Nikula, 2000). Endpoints included loss of adherence to tissue
28 culture substratum as evaluated by crystal violet staining, cell death measured by lactate
29 dehydrogenase release, release of interleukin-8 (IL-8) measured by enzyme-linked
30 immunosorbent assay, mitotic fraction and apoptosis, and release of alkaline phosphatase
31 measured by enzymatic activity using paranitrophenol phosphate. Results indicated that (1) SiO2
32 and Ni3S2 caused dose dependent acute toxicity and apototic changes; (2) ROFA and ND were
June 2003 7-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 acutely toxic only at the highest concentrations; (3) SiO2 (30, 100, 300 jig/cm2) and Ni3S2 (10,
2 30, 100, 300 jig/cm2) increased IL-8 (three and eight times over the control, respectively) but
3 suppressed IL-8 release at the highest concentration; (4) OAA and ROFA also induced IL-8 but
4 to a lesser degree; and (5) both diesel soots suppressed IL-8 production. The authors speculated
5 that the suppression of IL-8 release may contribute to increased respiratory disease as a result of
6 decreased response to infectious agents. Silicon dioxide and Ni3S2 increased the release of
7 alkaline phosphatase, a marker of toxic responses, only slightly. The less acutely toxic
8 compounds caused significant release of alkaline phosphatase. The order of potency in alkaline
9 phosphatase production is OAA > LD = ND > ROFA » SiO2 = Ni3S2. These results
10 demonstrated that the type of particle used has a strong influence on the biological response.
11 Dye et al. (1999) carried out reverse transcriptase-polymerase chain reactions on RNA
12 from rat tracheal epithelial cells to evaluate changes in steady-state gene expression of IL-6,
13 MIP-2, and iNOS in cells exposed for 6 h to ROFA (5 |ig/cm2) and Ni, V, or Ni and V(water-
14 soluble equivalent metal solution [pH 3.0]). Expression of MIP-2 and IL-6 genes was
15 significantly upregulated as early as 6 h post-ROFA-exposure in rat tracheal epithelial cells;
16 whereas gene expression of iNOS was maximally increased 24 h postexposure. Vanadium but
17 not Ni appeared to be mediating the effects of ROFA on gene expression. Treatment with
18 dimethylthiourea (4 and 40 mm) inhibited both ROFA and V induced gene expression in a dose-
19 dependent manner.
20 It appears that many biological responses are produced by PM whether it is composed of a
21 single component or a complex mixture. The newly developed gene array monitors the
22 expressions of many mediator genes that regulate complex and coordinated cellular events
23 involved in tissue injury and repair. Using an array consisting of 27 rat genes representing
24 inflammatory and anti-inflammatory cytokines, growth factors, adhesion molecules, stress
25 proteins, metalloproteinases, vascular tone regulatory molecules, transcription factors, surfactant
26 proteins and antioxidant enzymes, Nadadur et al. (2000) measured pulmonary effects in rats 3
27 and 24 h following intratracheal instillation of ROFA (3.3 mg/kg), NiSO4 (1.3 |imol/kg), and
28 VSO4 (2.2 jimol/kg). Their data revealed a two- to three-fold increase in the expression of IL-6
29 and TIMP-1 at 24 h post-Ni exposure. The expression of cellular fibronectin (cFn-EIIIA) and
30 iNOS increased 24 h following ROFA exposure. Cellular fibronectin, interferon, iNOS, ICAM-
31 1 was increased 24 h following Ni exposure and IL-6 was increased 24 h postexposure in V
32 exposed animals. There was a modest increase in the expression of SP-S and p-actin genes.
June 2003 7-79 DRAFT-DO NOT QUOTE OR CITE
-------
1 There was a 2-fold increase in the expression of IL-6 24 h following exposure to ROFA, Ni, and
2 V using the Northern blot analysis. A densitometric scan of an autoradiograph of blots stripped
3 and reprobed with SP-A cDNA insert indicated a minimal increase in the expression of SP-A,
4 both 3 and 24 h postexposure in all test groups. The findings in this study suggest that gene
5 array may provide a tool for screening the expression profile of tissue specific markers following
6 exposure to PM. However, care should be taken in reviewing such findings because of the
7 variations in dose, instillation versus inhalation, and the time-course for gene expression.
8 To investigate the interaction between respiratory cells and PM, Kobzik (1995) showed
9 that scavenger receptors are responsible for AM binding of unopsonized PM and that different
10 mechanisms mediate binding of carbonaceous dusts such as DPM. In addition, surfactant
11 components can increase AM phagocytosis of environmental particles in vitro, but only slightly
12 relative to the already avid AM uptake of unopsonized particles (Stringer and Kobzik, 1996).
13 Respiratory tract epithelial cells are also capable of binding with PM to secrete cytokine IL-8.
14 Using a respiratory epithelial cell line (A549), Stringer et al. (1996) found that binding of
15 particles to epithelial cells was calcium-dependent for TiO2 and Fe2O3, while cc-quartz binding
16 was not calcium dependent. In addition, as observed in AMs, PM binding by A549 cells also
17 was mediated by scavenger receptors, albeit those distinct from the heparin-insensitive
18 acetylated-LDL receptor. Furthermore, cc-quartz, but not TiO2 or CAPs, caused a dose-
19 dependent production of IL-8 (range 1 to 6 ng/mL), demonstrating a particle-specific spectrum
20 of epithelial cell cytokine (IL-8) response.
21
22 7.4.3.3 Other Potential Cellular and Molecular Mechanisms
23 A potential mechanism involving in the alteration of surface tension may be related to
24 changes in the expression of matrix metalloproteinases (MMPs), such as pulmonary matrilysin
25 and gelatinase A and B, and tissue inhibitor of metalloproteinase (TIMP) (Su et al., 2000a,b).
26 Sprague-Dawley rats exposed to ROFA by intratracheal injection (2.5 mg/rat) had increased
27 mRNA levels of matrilysin, gelatinase A, and TIMP-1. Gelatinase B, not expressed in control
28 animals, was increased significantly from 6 to 24 h following ROFA exposure. Alveolar
29 macrophages, epithelial cells, and inflammatory cells were major cellular sources for the
30 pulmonary MMP expression. The expression of Gelatinase B in rats exposed to the same dose
31 of ambient PM (< 1.7 jim and 1.7 to 3.7 jim) collected from Washington, DC, was significantly
32 increased as compared to saline control; whereas the expression of TIMP-2 was suppressed.
June 2003 7-80 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ambient PM between 3.7 and 20 jim also increased the Gelatinase B expression. Increases in
2 MMPs, which degrade most of the extracellular matrix, suggest that ROFA and ambient PM can
3 similarly increase the total pool of proteolytic activity to the lung and contribute in the
4 pathogenesis of PM-induced lung injury.
5 The role of sensory nerve receptors in the initiation of PM inflammation has been
6 described in a series of recent studies. Neuropeptide and acid-sensitive sensory irritant (i.e.,
7 capsaicin, VR1) receptors were first identified on human bronchial epithelial cells (i.e., BEAS-
8 2B). To address whether PM could initiate airway inflammation through these acid sensitive
9 sensory receptors, BEAS-2B cells were exposed to ROFA and responded with an immediate
10 increase in [Ca+2]; followed by a concentration-dependent release of inflammatory cytokine (i.e.,
11 IL-6, IL-8, TNFcc) and their transcripts (Veronesi et al., 1999b). To test the relevance of
12 neuropeptide or capsaicin VR1 receptors to these changes, BEAS-2B cells were pretreated with
13 neuropeptide receptor antagonists or capsazepine (CPZ), the antagonist for the capsaicin (i.e.,
14 VR1) receptor. The neuropeptide receptor antagonists reduced ROFA-stimulated cytokine
15 release by 25%-50%. However, pretreatment of cells with CPZ inhibited the immediate
16 increases in [Ca+2];, diminished transcript (i.e., IL-6, IL-8, TNFcc) levels and reduced IL-6
17 cytokine release to control levels (Veronesi et al., 1999a). The above studies suggested that
18 ROFA inflammation was mediated by acid sensitive VR1 receptors located on the sensory nerve
19 fibers that innervate the airway and on epithelial target cells.
20 Colloidal particles carry an inherently negative surface charge (i.e., zeta potential) that
21 attracts protons from their vaporous milieu. These protons form a neutralizing, positive ionic
22 cloud around the individual particle (Hunter, 1981). Since VR1 irritant receptors respond to
23 acidity (i.e., protonic charge), experiments were designed to determine if the surface charge
24 carried by ROFA and other PM particles could biologically activate cells and stimulate
25 inflammatory cytokine release. The mobility of ROFA particles was measured in an electrically
26 charged field (i.e., micro-electrophoresis) microscopically and their zeta potential calculated.
27 Next, synthetic polymer microspheres (SPM) (i.e., polymethacrylic acid nitrophenyl aery late
28 microspheres) were prepared with attached carboxyl groups to yield SPM particles with a
29 geometric diameter of 2 ± 0.1 and 6 ± 0.3 jim and with zeta potentials similar to ROFA
30 (-29 + 0.9 mV) particles. These SPM acted as ROFA surrogates with respect to their size and
31 surface charge, but lacked all other contaminants thought to be responsible for its toxicity (e.g.,
32 transition metals, sulfates, volatile organics and biologicals). Similar concentrations of SPM and
June 2003 7-81 DRAFT-DO NOT QUOTE OR CITE
-------
1 ROFA particles were used to test BEAS-2B cells and mouse dorsal root ganglia (DRG) sensory
2 neurons, both targets of inhaled PM. Equivalent degrees of biological activation (i.e., increase in
3 intracellular calcium, [Ca+2];, IL-6 release) occurred in both cell types in response to either
4 ROFA or SPM, and both responses could be reduced by antagonists to VR1 receptors or acid-
5 sensitive pathways. Neutrally charged SPM (i.e., zeta potential of 0 mV), however, failed to
6 stimulate increases in [Ca+2]; or IL-6 release (Oortgiesen et al., 2000). To expand on these data, a
7 larger set of PM was obtained from urban (St. Louis, Ottawa), residential (wood stove), volcanic
8 (Mt. St. Helen), and industrial (oil fly ash, coal fly ash) sources. Each PM sample was described
9 physicochemically (i.e., size and number of visible particles, acidity, zeta potential) and used to
10 test BEAS-2B epithelial cells. The resulting biological effect (i.e., increases in [Ca+2];, IL-6
11 release) was related to their physicochemical descriptions. When examined by linear regression
12 analysis, the only measured physicochemical property that correlated with increases in [Ca+2];
13 and IL-6 release was the zeta potential of the visible particles (r2 > 0.97) (Veronesi et al.,
14 2002b).
15 Together, the above studies have demonstrated a neurogenic basis for PM inflammation by
16 which the proton cloud associated with negatively-charged colloidal PM particles can activate
17 acid-sensitive VR1 receptors found on human airway epithelial cells and sensory terminals. This
18 activation results in an immediate influx of calcium and the release of inflammatory
19 neuropeptides and cytokines which proceed to initiate and sustain inflammatory events in the
20 airways through the pathophysiology of neurogenic inflammation (Veronesi and Oortgiesen,
21 2001).
22
23 7.4.4 Specific Particle Size and Surface Area Effects
24 Most particles used in laboratory animal toxicology studies are greater than 0.1 |im in size.
25 However, the enormous number and huge surface area of ultrafine particles highlight the
26 importance of considering the size of the particle in assessing response. Ultrafine particles with
27 a diameter of 20 nm, when inhaled at the same mass concentration, have a number concentration
28 that is approximately 6 orders of magnitude higher than for a 2.5-|im diameter particle; particle
29 surface area is also greatly increased (Table 7-11).
30 Many studies summarized in 1996 PM AQCD (U.S. Environmental Protection Agency,
31 1996a), as well as in this document, suggest that the surface of particles or substances that are
32 released from the surface (e.g., transition metals, organics) interact with the biological system,
June 2003 7-82 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 jig/m3
Particle Diameter
(|im)
0.02
0.1
0.5
1.0
2.5
Particle Number
(per cm3 air)
2,400,000
19,100
153
19
1.2
Particle Surface Area
(|im2 per cm3 air)
3,016
600
120
60
24
Source: Oberdorster (1996a).
1 and that surface-associated free radicals or free radical-generating systems may be responsible
2 for toxicity. Thus, if ultrafine particles were to cause toxicity by a transition metal-mediated
3 mechanism, for example, then the relatively large surface area for a given mass of ultrafine
4 particles would mean high concentrations of transition metals being available to cause oxidative
5 stress to cells.
6 Two groups have examined toxicity differences between fine and ultrafine particles, with
7 the general finding that ultrafine particles show a significantly greater response at similar mass
8 doses (Oberdorster et al., 1992; Li et al., 1996, 1997, 1999). However, only a few studies have
9 investigated the ability of ultrafine particles to generate a greater oxidative stress when compared
10 to fine particles of the same material. Studies by Gilmour et al. (1996) have shown that, at equal
11 mass, ultrafine TiO2 caused more plasmid DNA strand breaks than fine TiO2. This effect could
12 be inhibited with mannitol. Osier and Oberdorster (1997) compared the response of rats (F344)
13 exposed by intratracheal inhalation to "fine" (-250 nm) and "ultrafine" (-21 nm) TiO2 particles
14 with rats exposed to similar doses by intratracheal instillation. Animals receiving particles
15 through inhalation showed a smaller pulmonary response, measured by BAL parameters, in both
16 severity and persistence, when compared with those animals receiving particles through
17 instillation. Ultrafine TiO2 particles consistently had a significantly greater response than did the
18 fine TiO2 particles. These results demonstrate a difference in pulmonary response to an inhaled
19 versus an instilled dose, which may result from differences in dose rate, particle distribution,
20 particle surface activity, or altered clearance between the two methods.
June 2003 7-83 DRAFT-DO NOT QUOTE OR CITE
-------
1 Consistent with these in vivo studies, Finkelstein et al. (1997) has shown that exposing
2 primary cultures of rat Type II cells to 10 |ig/mL ultrafine TiO2 (20 nm) causes increased TNF
3 and IL-1 release throughout the entire 48-h incubation period. In contrast, fine TiO2 (200 nm)
4 had no effect. In addition, ultrafine polystyrene carboxylate-modified microspheres (UFP,
5 fluorospheres, molecular probes 44 ± 5 nm) have been shown to induce a significant
6 enhancement of both substance P and histamine release after administration of capsaicin (10"4
7 M), to stimulate C-fiber, and carbachol (10"4 M), a cholinergic agonist in rabbit intratracheally
8 instilled with UFP (Nemmar et al., 1999). A significant increase in histamine release also was
9 recorded in the UFP-instilled group following the administration of both Substance P (10"6 M)
10 plus thiorpan (10'5 M) and compound 48/80 (C48/80, 10'3 M) to stimulate mast cells.
11 Bronchoalveolar lavage analysis showed an influx of PMN, an increase in total protein
12 concentration, and an increase in lung wet weight/dry weight ratio. Electron microscopy showed
13 that both epithelial and endothelial injuries were observed. The pretreatment of rabbits in vivo
14 with a mixture of either SR 140333 and SR 48368, a tachykinin NKl and NK2 receptor
15 antagonist, or a mixture of terfenadine and cimetidine, a histamine Hx and H2 receptor
16 antagonist, prevented UFP-induced PMN influx and increased protein and lung WW/DW ratio.
17 It is believed that ultrafine particles cause greater cellular injury because of the relatively
18 large surface area for a given mass. In addition, the fate of ultrafines after deposition is also
19 different in that they interact more rapidly with epithelial target cells rather than to be
20 phagocytized by alveolar macrophages. However, in a study that compared the response to
21 carbon black particles of two different sizes, Li et al. (1999) demonstrated that in the instillation
22 model, a localized dose of particle over a certain level causes the particle mass to dominate the
23 response, rather than the surface area. Ultrafine carbon black (ufCB, Printex 90), 14 nm in
24 diameter, and fine carbon black (CB, Huber 990), 260 nm in diameter, were instilled
25 intratracheally in rats, and BAL profile at 6 h was assessed. At mass of 125 jig or below, ufCB
26 generated a greater response (increase LDH, epithelial permeability, decrease in GSH, TNF, and
27 NO production) than fine CB at various times postexposure. However, higher doses of CB
28 caused more PMN influx than the ufCB. In contrast to the effect of CB, which showed dose-
29 related increasing inflammatory response, ufCB at the highest dose caused less of a neutrophil
30 influx than at the lower dose, confirming earlier work by Oberdorster et al. (1992). Moreover,
31 when the PMN influx was expressed as a function of surface area, CB produced greater response
32 than ufCB at all doses used in this study. Although particle insterstitialization with a consequent
June 2003 7-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 change in the chemotatic gradient for PMN was offered as an explanation, these results need
2 further scrutiny. Moreover, these findings imply that mass is relatively less important than
3 surface area and that the latter metric may be more useful for assessing PM toxicity. However, it
4 is unclear if this finding is restricted to the particular endpoints addressed and/or carbon black,
5 the PM compound studied.
6 Oberdorster et al. (2000) recently completed a series of studies in rats and mice using
7 ultrafine particles of various chemical composition. In rats sensitized with endotoxin (70 EU)
8 and exposed to ozone (1 ppm) plus ultrafine carbon particles (-100 |ig/m3), they found a nine-
9 fold greater release of reactive oxygen species in old rats (20 mo) than in similarly treated young
10 rats (10 wk). Exposure to ultrafine PM alone in sensitized old rats also caused an inflammatory
11 response.
12 Although the exact mechanism of ultrafine-induced lung injury remains unclear, it is likely
13 that ultrafine particles, because of their small size, are not effectively phagocytized by alveolar
14 macrophages and can easily penetrate the airway epithelium, gaining access to the interstitium.
15 This is particularly significant for ultrafine droplets of acids that do not persist as particles once
16 deposited. However, organic ultrafine particles may persist longer depending on organic
17 components. Using electron microscopy, Churg et al. (1998) examined particle uptake in rat
18 tracheal explants. Explants were submerged in a 5 mg/mL suspension of either fine (0.12 jim) or
19 ultrafine (0.021 jim) TiO2 particles in Dulbecco's minimal Eagle's medium, without serum and
20 examined after 3 or 7 days. They found both size particles in the epithelium at both time points;
21 but, in the subepithelial tissues, only at day 7. The volume proportion (the volume of TiO2 over
22 the entire volume of epithelium or subepithelium area) of both fine and ultrafine particles in the
23 epithelium increased from 3 to 7 days. It was greater for ultrafine at 3 days but was greater for
24 fine at 7 days. The volume proportion of particles in the subepithelium at day 7 was equal for
25 both particles, but the ratio of epithelial to subepithelial volume proportion was 2:1 for fine and
26 1:1 for ultrafine. Ultrafine particles persisted in the tissue as relatively large aggregates; whereas
27 the size of fine particle aggregates became smaller over time. Ultrafine particles appeared to
28 enter the epithelium faster and, once in the epithelium, a greater proportion of them were
29 translocated to the subepithelial space compared to fine particles. However, the authors assumed
30 that the volume proportion is representative of particle number and the number of particles
31 reaching the interstitial space is directly proportional to the number applied (i.e., there is no
32 preferential transport from lumen to interstitium by size). These data are in contrast to the
June 2003 7-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 results of instillation or inhalation of fine and ultrafine TIO2 particles reported earlier (Ferin
2 et al., 1990, 1992). However, the explant and intratracheal instillation test systems differ in
3 many aspects, making direct comparisons difficult. Limitations of the explant test system
4 include traumatizing the explanted tissue, introducing potential artifacts through the use of liquid
5 suspension for exposure, the absence of inflammatory cells, and possible overloading of the
6 explants with dust.
7 Only two studies examined the influence of specific surface area on biological activity
8 (Lison et al., 1997; Oettinger et al., 1999). The biological responses to various MnO2 dusts with
9 different specific surface area (0.16, 0.5, 17, and 62 m2/g) were compared in vitro and in vivo
10 (Lison et al., 1997). In both systems, the results show that the amplitude of the response is
11 dependent on the total surface area that is in contact with the biological system, indicating that
12 surface chemistry phenomena are involved in the biological reactivity. Freshly ground particles
13 with a specific surface area of 5 m2/g also were examined in vitro. These particles exhibited an
14 enhanced cytotoxic activity that was almost equivalent to that of particles with a specific surface
15 area of 62 m2/g, indicating that undefined reactive sites produced at the particle surface by
16 mechanical cleavage also may contribute to the toxicity of insoluble particles. In another study,
17 two types of carbon black particles, Printex 90 (P90, Degussa, Germany, formed by controlled
18 combustion, consists of defined granules with specific surface area of 300 m2/g and particle size
19 of 14 nm) and FR 101 (Degussa, Germany, with specific surface area of 20 m2/g and particle size
20 of < 95 nm, has a coarse structure, and the ability to adsorb poly cyclic and other carbons) were
21 used in the study (Oettinger et al., 1999). Exposure of AMs to 100 |ig/106 cells of FR 101 and
22 P90 resulted in a 1.4- and 2.1-fold increase in ROS release. These exposures also caused a
23 fourfold up-regulation of NF-icB gene expression. These studies indicated that PM of single
24 component with larger surface area produce greater biological response than similar particles
25 with smaller surface area. By exposing bovine AMs to metal oxide coated silica particles,
26 Schluter et al. (1995) showed that most of the metal coatings (As, Ce, Fe, Mn, Ni, Pb, and V)
27 had no effect on ROS production by these cells. However, coating with CuO markedly lowered
28 the O2" and H2O2, whereas V(IV) increases both reactive oxygen intermediates (ROI). This study
29 demonstrated that, in addition to specific surface area, chemical composition of the particle
30 surface also influences its cellular response.
31 Thus, ultrafine particles apparently have the potential to significantly contribute to the
32 adverse effects of PM. These studies, however, have overlooked the portion of ambient ultrafine
June 2003 7-86 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles that are not solid in form. Droplets (e.g., sulfuric acid droplets) and organic based
2 ultrafine particles do exist in the ambient environment, but their role in the adverse effects of
3 ultrafine particles has been ignored. Moreover, the ability of these droplet ultrafine particles to
4 spread, disperse, or dissolve after contact with liquid surface layers must be considered.
5 Accordingly, all of the hypotheses and studies should be critically analyzed with regard to the
6 concentrations/doses used, models used, and the specific PM tested.
7
8
9 7.5 SUSCEPTIBILITY TO THE EFFECTS OF PARTICULATE
10 MATTER EXPOSURE
11 Susceptibility of an individual to adverse health effects of PM can vary depending on a
12 variety of host factors such as age, physiological activity profile, genetic predisposition, or
13 preexistent disease. The potential for preexistent disease to alter pathophysiological responses to
14 toxicant exposure is widely acknowledged but poorly understood. Epidemiologic studies have
15 demonstrated that the effects of PM exposure tend to be more evident in populations with pre-
16 existing disease; and it is logical that important mechanistic differences may exist among these
17 populations. However, because of inherent variability (necessitating large numbers of subjects)
18 and ethical concerns associated with using diseased subjects in clinical research studies, a solid
19 database on human susceptibilities is lacking. For more control over both host and
20 environmental variables, animal models often are used. Many laboratory studies have
21 demonstrated alterations in a variety of endpoints in experimental animals following exposure to
22 laboratory-generated particles. These findings (e.g., increased pulmonary inflammation,
23 increased airway resistance, and decrements in pulmonary host defenses) may be of limited
24 value because of uncertainties in extrapolating between the laboratory-generated particles and
25 actual ambient air particle mixes. Thus, care must be taken in extrapolation from animal models
26 of human disease to humans. Rodent models of human disease, their use in toxicology, and the
27 criteria for judging their appropriateness as well as their limitations must be considered
28 (Kodavanti et al., 1998b; Kodavanti and Costa, 1999; Costa, 2000).
29
30 7.5.1 Pulmonary Effects of Particulate Matter in Compromised Hosts
31 Epidemiologic studies suggest there may be subsegments of the population that are
32 especially susceptible to effects from inhaled particles (see Chapter 8). The elderly with chronic
June 2003 7-87 DRAFT-DO NOT QUOTE OR CITE
-------
1 cardiopulmonary disease, those with pneumonia and possibly other lung infections, and those
2 with asthma (at any age) appear to be at higher risk than healthy people of similar age.
3 Unfortunately, most toxicology studies have used healthy adult animals. However, an increasing
4 number of newer studies have started to examine effects of ambient particles in compromised
5 host models. For example, Costa and Dreher (1997) used a rat model of cardiopulmonary
6 disease to explore the question of susceptibility and the possible mechanisms by which PM
7 effects are potentiated. Rats with advanced monocrotaline (MCT)-induced pulmonary
8 vasculitis/hypertension were given intratracheal instillations of ROFA (0, 0.25, 1.0, and
9 2.5mg/rat). A brief description of the model appears above, in Section 7.3. The MCT animals
10 had a marked neutrophilic inflammation. In the context of this inflammation, ROFA induced a
11 four- to fivefold increase in BAL PMNs. There was increased mortality at 96 h that was ROFA-
12 dose dependent. The results of this study indicate that particles, albeit at a high concentration,
13 enhanced mortality in MCT animals but not in healthy animals.
14 As discussed previously, Kodavanti et al. (1999) also studied PM effects in the MCT rat
15 model of pulmonary disease. Rats treated with 60 mg/kg MCT were exposed to 0, 0.83 or
16 3.3 mg/kg ROFA by intratracheal instillation and to 15 mg/m3 ROFA by inhalation. Both
17 methods of exposure caused inflammatory lung responses; and ROFA exacerbated the lung
18 lesions, as shown by increased lung edema, inflammatory cells, and alveolar thickening.
19 The manner in which MCT can alter the response of rats to inhaled particles was examined
20 by Madl and colleagues (1998). Rats were exposed to fluorescent colored microspheres (1 |im)
21 2 weeks after treatment with MCT. In vivo phagocytosis of the microspheres was altered in the
22 MCT rats in comparison with control animals. Fewer microspheres were phagocytized in vivo
23 by alveolar macrophages, and there was a concomitant increase in free microspheres overlaying
24 the epithelium at airway bifurcations. The decrease in in vivo phagocytosis was not
25 accompanied by a similar decrease in vitro. Macrophage chemotaxis, however, was impaired
26 significantly in MCT rats compared with control rats. Thus, MCT appeared to impair particle
27 clearance from the lungs via inhibition of macrophage chemotaxis.
28 Chronic bronchitis is the most prevalent of the COPD-related illnesses. In humans, chronic
29 bronchitis is characterized by pathologic airway inflammation and epithelial damage, mucus cell
30 hyperplasia and hypersecretion, airway obstruction and in advance cases, airway fibrosis. The
31 most widely used animal models of bronchitis (rat and dog) are those produced by subchronic
32 exposure to high concentrations of SO2 (150 to 600 ppm) for 4 to 6 weeks. Exposure to SO2
June 2003 7-88 DRAFT-DO NOT QUOTE OR CITE
-------
1 produces changes in the airways similar to those of chronic bronchitis in humans. There is an
2 anatomical difference between the rat and the human in the absence of submucosal glands in the
3 rat. However, like humans, rats exhibit increased airway responsiveness to inhaled
4 bronchoconstricting agonists. Sulfur dioxide-induced lesions include increased numbers of
5 epithelial mucus-producing cell, loss of cilia, airway inflammation, increased pro-inflammatory
6 cytokine expression, and thickening of the airway epithelium. When the cause of the chronic
7 bronchitis is removed, the pathology slowly reverses. The time course and the extent of reversal
8 differs between the human and rodent. Consequently, care should be exercised when applying
9 this model (Kodavanti et al., 1998b).
10 Respiratory infections are common in all individuals. The infections are generally cleared
11 quickly, depending on the virulence of the organism, however, in individuals with immunologic
12 impairment or lung diseases such a COPD, the residence time in the lung is extended. A variety
13 of viral and bacterial agents have been used to develop infection models in animals. Viral
14 infection models primarily use mice and rats. The models focus on the proliferation and
15 clearance of the microorganisms and the associated pulmonary effect. The models range from
16 highly virulent and lethal (influenza A/Hong Kong/8/68, H3N2) to nonlethal (rat-adapted
17 influenza virus model [RAIV]). The lethal model terminates in extensive pneumonia and lung
18 consolidation. Less virulent models (A/Port Chalmers/1/73 and H3N2) exhibit airway epithelial
19 damage and immune responses. The non-lethal model exhibits airway reactivity that subsides,
20 with recovery being complete in about 2 weeks (Kodavanti et al., 1998b). Bacterial infection
21 models mimic the chronic bacterial infections experienced by humans with other underlying
22 disease conditions. The models develop signs similar to those in humans but to a milder degree.
23 To mimic the more chronic infections, the bacteria are encased in agar beads to prevent rapid
24 clearance. Generally, the models involve pre-exposure to the irritant followed by the bacterial
25 challenge. More recently, bacterial infection models have involved pre-exposure by the bacteria
26 followed by exposure to the irritant (Kodavanti et al., 1998b).
27 Elder et al. (2000a,b) exposed 8 week to 22 month old Fischer 344 rats and 14- to
28 17-month-old Tsk mice to 100 |ig/m3 of ultrafine carbon (UF) and/or 1.0 ppm O3 for six hours
29 following a 12 minute exposure to a low dose (70 EU) of endotoxin (lipopolysaccharide, LPS).
30 The ultrafine carbon had a small effect on lung inflammation and inflammatory cell activation.
31 The effects were enhanced in the compromised lung and in older animals. The greatest effect
32 was in the compromised lung exposed to both ultrafine carbon and ozone.
June 2003 7-89 DRAFT-DO NOT QUOTE OR CITE
-------
1 The sulfur dioxide (SO2)-induced model of chronic bronchitis has also been used to
2 examine the potential interaction of PM with preexisting lung injury. Clarke and colleagues
3 pretreated Sprague-Dawley rats for 6 weeks with air or 170 ppm SO2 for 5 h/day and
4 5 days/week (Clarke et al., 1999; Saldiva et al., 2002). Exposure to concentrated ambient air
5 particles (CAPs) for 5 h/day for 3 days to concentrations ranging from 73.5 to 733 |ig/m3
6 produced significant changes in both cellular and biochemical markers in lavage fluid.
7 In comparison to control animal values, protein was increased approximately threefold in
8 SO2-pretreated animals exposed to concentrated ambient PM. Lavage fluid neutrophils and
9 lymphocytes were increased significantly in both groups of rats exposed to concentrated ambient
10 PM, with greater increases in both cell types in the SO2-pretreated rats. Thus, exposure to
11 concentrated ambient PM produced adverse changes in the respiratory system, but no deaths, in
12 both normal rats and in a rat model of chronic bronchitis.
13 Clarke et al. (2000b) next examined the effect of concentrated ambient PM from Boston,
14 MA, in normal rats of different ages. Unlike the earlier study that used Sprague-Dawley rats,
15 4- and 20-mo-old Fischer 344 rats were examined after exposure to concentrated ambient PM for
16 5 h/day for 3 consecutive days. They found that exposure to the daily mean concentrations of
17 80, 170, and 50 |ig/m3 PM, respectively, produced statistically significant increases in total
18 neutrophil counts (over 10-fold) in lavage fluid of the young, but not the old, rats. Thus,
19 repeated exposure to relatively low concentrations of ambient PM produced an inflammatory
20 response, although the actual percent neutrophils in the concentrated ambient PM-exposed
21 young adult rats was low (approximately 3%). On the other hand, Gordon et al. (2000) found no
22 evidence of neutrophil influx in the lungs of normal and monocrotaline-treated Fischer 344 rats
23 exposed in nine separate experiments to concentrated ambient PM from New York, NY at
24 concentrations as high as 400 |ig/m3 for a 6-h exposure or 192 |ig/m3 for three daily 6-h
25 exposures. Similarly, normal and cardiomyopathic hamsters showed no evidence of pulmonary
26 inflammation or injury after a single exposure to the same levels of concentrated ambient PM.
27 Gordon and colleagues did report a statistically significant doubling in protein concentration in
28 lavage fluid in monocrotaline-treated rats exposed for 6 h to 400 |ig/m3 concentrated ambient
29 PM.
30 Kodavanti and colleagues (1998b) also have examined the effect of concentrated ambient
31 PM in normal rats and rats with sulfur dioxide-induced chronic bronchitis. Among the four
32 separate exposures to PM, there was a significant increase in lavage fluid protein in bronchitic
June 2003 7-90 DRAFT-DO NOT QUOTE OR CITE
-------
1 rats from only one exposure protocol in which the rats were exposed to 444 and 843 |ig/m3 PM
2 on 2 consecutive days (6 h/day). Neutrophil counts were increased in bronchitic rats exposed to
3 concentrated ambient PM in three of the four exposure protocols, but was decreased in the fourth
4 protocol. No other changes in normal or bronchitic rats were observed, even in the exposure
5 protocols with higher PM concentrations. Thus, rodent studies have demonstrated that
6 inflammatory changes can be produced in normal and compromised animals exposed to
7 concentrated ambient PM. These findings are important because only a limited number of
8 studies have used real-time inhalation exposures to actual ambient urban PM.
9 Pulmonary function measurements are often less invasive than other means to assess the
10 effects of inhaled air pollutants on the mammalian lung. After publication of the 1996 PM
11 AQCD, a number of investigators examined the response of rodents and dogs to inhaled ambient
12 particles. In general, these investigators have demonstrated that ambient PM has minimal effects
13 on pulmonary function. Gordon et al. (2000) exposed normal and monocrotaline-treated rats to
14 filtered air or 181 |ig/m3 concentrated ambient PM for 3 h. For both normal and monocrotaline-
15 treated rats, no differences in lung volumes or diffusion capacities for carbon monoxide were
16 observed between the air or PM exposed animals at 3 or 24 h after exposure. Similarly, in
17 cardiomyopathic hamsters, concentrated ambient PM had no effect on these same pulmonary
18 function measurements.
19 Other pulmonary function endpoints have been studied in animals exposed to concentrated
20 ambient PM. Clarke et al. (1999) observed that tidal volume was increased slightly in both
21 control rats and rats with sulfur dioxide-induced chronic bronchitis exposed to 206 to 733 |ig/m3
22 PM on 3 consecutive days. No changes in peak expiratory flow, respiratory frequency, or
23 minute volume were observed after exposure to concentrated ambient PM. In the series of dog
24 studies by Godleski et al. (2000) (also see Section 7.3), no signficant changes in pulmonary
25 function were observed in normal mongrel dogs exposed to concentrated ambient PM, although
26 a 20% decrease in respiratory frequency was observed in dogs that underwent coronary artery
27 occlusion and were exposed to PM. Thus, studies using normal and compromised animal
28 models exposed to concentrated ambient PM have found minimal biological effects of ambient
29 PM on pulmonary function.
30 Johnston et al. (1998) exposed 8-week-old mice (young) and 18-mo-old mice (old) to
31 polytetrafluoroethylene (PTFE) fumes (0, 10, 25, and 50 |ig/m3) for 30 min. Lung lavage
32 endpoints (PMN, protein, LDH, and p-glucuronidase) as well as lung tissue mRNA levels for
June 2003 7-91 DRAFT-DO NOT QUOTE OR CITE
-------
1 various cytokines, metallothionein and for Mn superoxide dismutase were measured 6 h
2 following exposure. Protein, lymphocyte, PMN, and TNF-cc mRNA levels were increased in
3 older mice when compared to younger mice. These findings suggest that the inflammatory
4 response to PTFE fumes is altered with age, being greater in the older animals. Although
5 ultrafme PTFE fumes are not a valid surrogate for ambient ultrafine particles (Oberdorster et al.,
6 1992), this study provides evidence supporting the hypothesis that particle-induced pulmonary
7 inflammation differs between young and old mice. Other studies on age-related PM effects are
8 described in Section 7.6 (Responses to PM and Gaseous Pollutant Mixtures).
9 Kodavanti et al. (2000b; 2001) used genetically predisposed spontaneously hypertensive
10 (SH) rats as a model of cardiovascular disease to study PM-related susceptibility. The SH rats
11 were found to be more susceptible to acute pulmonary injury from intratracheal ROFA exposure
12 than normotensive control Wistar Kyoto (WKY) rats (Kodavanti et al., 2001). The primary
13 metal constituents of ROFA, V and Ni, caused differential species-specific effects. Vanadium,
14 which was less toxic than Ni in both strains, caused inflammatory responses only in WKY rats;
15 whereas Ni was injurious to both WKY and SH rats (SH > WKY). This differential
16 responsiveness of V and Ni was correlated with their specificity for airway and parenchymal
17 injury, discussed in another study (Kodavanti et al., 1998b). When exposed to the same ROFA
18 by inhalation (15 mg/m3, 6 h/d, 3 days), SH rats were more sensitive than WKY rats in regards to
19 vascular leakage (Kodavanti et al., 2000b). The SH rats exhibited a hemorrhagic response to
20 ROFA. Oxidative stress was much higher in ROFA exposed SH rats than matching WKY rats.
21 Also, SH rats, unlike WKY rats, showed a compromised ability to increase BALF glutathione in
22 response to ROFA, suggesting a potential link to increased susceptibility. However, lactate
23 dehydrogenase and n-acetylglucosaminidase activities were higher in WKY rats. Lactate
24 dehydrogenase was slightly higher in SH rats instilled with ROFA (Kodavanti et al., 2001).
25 Cardiovascular effects were characterized by ST-segment area depression of the ECG in ROFA-
26 exposed SH but not WKY rats. When the same rats were exposed to ROFA by inhalation to
27 15 mg/m3, 6 h/d, 3 d/wk for 1,2, or 4 wk compared to intratracheal exposure to 0, 1.0, 5.0 mg/kg
28 in saline (Kodavanti et al., 2002), differences in effects were dependent on the length of
29 exposure. After acute exposure, increased plasma fibrinogen was associated with lung injury;
30 longer-term, episodic ROFA exposure resulted in progressive protein leakage and inflammation
31 that was significantly worse in SH rats when compared to WKY rats. These studies demonstrate
32 the potential utility of cardiovascular disease models for the study of PM health effects and show
June 2003 7-92 DRAFT-DO NOT QUOTE OR CITE
-------
1 that genetic predisposition to oxidative stress and cardiovascular disease may play a role in
2 increased sensitivity to PM-related cardiopulmonary injury.
3 On the basis of in vitro studies, Sun et al. (2001) predicted that the antioxidant and lipid
4 levels in the lung lining fluid may determine susceptibility to inhaled PM. In a subsequent study
5 from the same laboratory, Norwood et al. (2001) conducted inhalation studies on guinea pigs to
6 test this hypothesis. On the basis of dietary supplementation or depletion of ascorbic acid (C)
7 and glutathione (GSH) the guinea pigs were divided into four groups: (+C + GSH),
8 (+C - GSH), (-C + GSH), and (-C - GSH). All groups were exposed (nose-only) to clean air or
9 19-25 mg/m3 ROFA (< 2.5 |im) for 2 h. Nasal lavage and BAL fluid and cells were examined at
10 Oh and 24 h postexposure. Exposure to ROFA increased lung injury in the (-C-GSH) group
11 only (as shown by increased BAL fluid protein, LDH, and PMNs and decreased BAL
12 macrophages) and resulted in lower antioxidant concentrations in BAL fluid than were found
13 with single deficiencies.
14 In summary, although more of these studies are just beginning to emerge and are only now
15 being replicated or followed more thoroughly to investigate underlying mechanisms, they do
16 provide evidence suggestive of enhanced susceptibility to inhaled PM in "compromised" hosts.
17
18 7.5.2 Genetic Susceptibility to Inhaled Particles and their Constituents
19 A key issue in understanding adverse health effects of inhaled ambient PM is identification
20 of which classes of individuals are susceptible to PM. Although factors such as age and health
21 status have been studied in both epidemiology and toxicology studies, some investigators have
22 begun to examine the importance of genetic susceptibility in the response to inhaled particles
23 because of evidence that genetic factors play a role in the response to inhaled pollutant gases.
24 To accomplish this goal, investigators typically have studied the interstrain response to particles
25 in rodents. The response to ROFA instillation in different strains of rats has been investigated by
26 Kodavanti et al. (1996, 1997a). In the first study, male Sprague-Dawley (SD) and Fischer-344
27 (F-344) rats were instilled intratracheally with saline or ROFA particles (8.3 mg/kg). ROFA
28 instillation produced an increase in lavage fluid neutrophils in both SD and F-344 rats; whereas a
29 time-dependent increase in eosinophils occurred only in SD rats. In the subsequent study
30 (Kodavanti et al., 1997a), SD, Wistar (WIS), and F-344 rats (60 days old) were exposed to saline
31 or ROFA (8.3 mg/kg) by intratracheal instillation and examined for up to 12 weeks. Histology
32 indicated focal areas of lung damage showing inflammatory cell infiltration as well as alveolar,
June 2003 7-93 DRAFT-DO NOT QUOTE OR CITE
-------
1 airway, and interstitial thickening in all three rat strains during the week following exposure.
2 Trichrome staining for fibrotic changes indicated a sporadic incidence of focal alveolar fibrosis
3 at 1, 3, and 12 weeks in SD rats; whereas WIS and F-344 rats showed only a modest increase in
4 trichrome staining in the septal areas. One of the isoforms of fibronectin mRNA was
5 upregulated in ROFA-exposed SD and WIS rats, but not in F-344 rats. Thus, in rats there
6 appears to be a genetic based difference in susceptibility to lung injury induced by instilled
7 ROFA.
8 Differences in the degree of pulmonary inflammation have been described in rodent strains
9 exposed to airborne pollutants. To understand the underlying causes, signs of airway
10 inflammation (i.e., airway hyper-responsiveness, inflammatory cell influx) were established in
11 responsive (BABL/c) and non-responsive (C57BL/6) mouse strains exposed to ROFA (Veronesi
12 et al., 2000). Neurons taken from the ganglia (i.e., dorsal root ganglia) that innervate the nasal
13 and upper airways were cultured from each mouse strain and exposed to 25 or 50 |ig/mL ROFA
14 for 4 h. The difference in inflammatory response noted in these mouse strains in vivo was
15 retained in culture, with C57BL/6 neurons showing significantly lower signs of biological
16 activation (i.e., increased intracellular calcium levels) and cytokine (i.e., IL-6, IL-8) release
17 relative to BALB/c mice. RT-PCR and immunocytochemistry indicated that the BALB/c mouse
18 strain had a significantly higher number of neuropeptide and acid-sensitive (i.e., NK1, VR1)
19 sensory receptors on their sensory ganglia relative to the C57BL/6 mice. Such data indicate that
20 genetically-determined differences in sensory inflammatory receptors can influence the degree
21 of PM-induced airway inflammation.
22 Kleeberger and colleagues have examined the role that genetic susceptibility plays in the
23 effect of inhaled acid-coated particles on macrophage function. Nine inbred strains of mice were
24 exposed nose-only to carbon particles coated with acid (10 mg/m3 carbon with 285 |ig/m3
25 sulfate) for 4 h (Ohtsuka et al., 2000a). Significant inter-strain differences in Fc-receptor-
26 mediated macrophage phagocytosis were seen with C57BL/6J mice being the most sensitive.
27 Although neutrophil counts were increased more in C3H/HeOuJ and C3H/HeJ strains of mice
28 than in the other strains, the overall magnitude of change was small and not correlated with the
29 changes in macrophage phagocytosis. In follow-up studies using the same type particle, Ohtsuka
30 et al. (2000a,b) performed a genome-wide scan with an intercross cohort derived from C57BL/6J
31 and C3H/HeJ mice. Analyses of phenotypes of segregant and nonsegregant populations derived
32 from these two strains indicate that two unlinked genes control susceptibility. They identified a
June 2003 7-94 DRAFT-DO NOT QUOTE OR CITE
-------
1 3-centiMorgan segment on mouse chromosome 17 that contains an acid-coated particle
2 susceptibility locus. Interestingly, this quantitative trait locus (a) overlaps with those described
3 for ozone-induced inflammation (Kleeberger et al., 1997) and acute lung injury (Prows et al.,
4 1997) and (b) contains several promising candidate genes that may be responsible for the
5 observed genetic susceptibility for macrophage dysfunction in mice exposed to acid-coated
6 particles.
7 Leikauf and colleagues (Leikauf et al., 2000; Wesselkamper et al., 2000; McDowell et al.,
8 2000; Prows and Leikauf, 2001; Leikauf et al., 2001) have identified a genetic susceptibility in
9 mice that is associated with mortality following exposures to high concentrations (from 15 to
10 150 i-ig/m3) of a N1SO4 aerosol (0.22 |im MMAD) for up to 96 h. These studies also have
11 preliminarily identified the chromosomal locations of a few genes that may be responsible for
12 this genetic susceptibility. This finding is particularly significant in light of the toxicology
13 studies demonstrating that bioavailable, first-row transition metals participate in acute lung
14 injury following exposure to emission and ambient air particles. Similar genes may be involved
15 in human responses to particle-associated metals; but additional studies are needed to determine
16 whether the identified metal susceptibility genes are involved in human responses to ambient
17 levels of particulate-associated metals.
18 One study has examined the interstrain susceptibility to ambient particles. C57BL/6J and
19 C3H/HeJ mice were exposed to 250 |ig/m3 concentrated ambient PM2 5 for 6 h and examined at
20 0 and 24 h after exposure for changes in lavage fluid parameters and cytokine mRNA expression
21 in lung tissue (Shukla et al., 2000). No interstrain differences in response were observed.
22 Surprisingly, although no indices of pulmonary inflammation or injury were increased over
23 control values in the lavage fluid, increases in cytokine mRNA expression were observed in both
24 murine strains exposed to PM2 5. Although the increase in cytokine mRNA expression was
25 generally small (approximately twofold), the effects on IL-6, TNF-cc, TGF-P2, and y-interferon
26 were consistent.
27 Thus, a handful of studies have begun to demonstrate that genetic susceptibility can play a
28 role in the response to inhaled particles. However, the doses of PM administered in these
29 studies, whether by inhalation or instillation, were extremely high when compared to ambient
30 PM levels. Similar strain differences in response to inhaled metal particles have been observed
31 by other investigators (McKenna et al., 1998; Wesselkamper et al., 2000), although the
32 concentration of metals used in these studies were also more relevant to occupational rather than
June 2003 7-95 DRAFT-DO NOT QUOTE OR CITE
-------
1 environmental exposure levels. The extent to which genetic susceptibility plays as significant a
2 role in the adverse effects of ambient PM as does age or health status remains to be determined.
3
4 7.5.3 Participate Matter Effects on Allergic Hosts
5 Relatively little is known about the effects of inhaled particles on humoral (antibody) or
6 cell-mediated immunity. Alterations in the response to a specific antigenic challenge have been
7 observed in animal models at high concentrations of acid sulfate aerosols (above 1,000 |ig/m3)
8 (Pinto et al., 1979; Kitabatake et al., 1979; Fujimaki et al., 1992). Several studies have reported
9 an enhanced response to nonspecific bronchoprovocation agents, such as acetylcholine and
10 histamine, after exposure to inhaled particles. This nonspecific airway hyperresponsiveness,
11 a central feature of asthma, occurs in animals and human subjects exposed to sulfuric acid under
12 controlled conditions (Gearhart and Schlesinger, 1986; Utell et al., 1983). Although, its
13 relevance to specific allergic responses in the airways of atopic individuals is unclear, it
14 demonstrates that the airways of asthmatics may become sensitized to either specific or
15 nonspecific triggers that could result in increases in asthma severity and asthma-related hospital
16 admissions (Peters et al., 1997; Jacobs et al., 1997; Lipsett et al., 1997). Combustion particles
17 also may serve as carrier particles for allergens (Knox et al., 1997).
18 A number of in vivo and in vitro studies have demonstrated that diesel particles (DPM) can
19 alter the immune response to challenge with specific antigens and suggest that DPM may act as
20 an adjuvant. These studies have shown that treatment with DPM enhances the secretion of
21 antigen-specific IgE in mice (Takano et al., 1997) and in the nasal cavity of human subjects
22 (Diaz-Sanchez et al., 1996, 1997; Ohtoshi et al., 1998; Nel et al., 2001). Because IgE levels play
23 a major role in allergic asthma (Wheatley and Platts-Mills, 1996), upregulation of its production
24 could lead to an increased response to inhaled antigen in particle-exposed individuals.
25 Van Zijverden et al. (2000) and Van Zijverdan and Granum (2000) used mouse models to
26 assess the potency of particles (diesel, carbon black, silica) to adjuvate an immune response to a
27 protein antigen. All particles exert an adjuvant effect on the immune response to co-
28 administered antigen, apparently stimulated by the particle core rather than the attached chemical
29 factors. Different particles, however, stimulate distinct types of immune responses. In one
30 model (Van Zijverden et al., 2001), BALB/c mice were intranasally treated with a mixture of
31 antigen (model antigen TNP-Ovalbumin, TNP-OVA) and particles on three consecutive days.
32 On day 10 after sensitization, mice were challenged with the antigen TNP-OVA alone, and five
June 2003 7-96 DRAFT-DO NOT QUOTE OR CITE
-------
1 days later the immune response was assessed. Diesel particulate matter, as well as carbon black
2 particles (CB), were capable of adjuvating the immune response to TNP-OVA as evidenced by
3 an increase of TNP-specific antibody (IgGl and IgE) secreting B cells antibodies in the lung-
4 draining lymph nodes. Increased antigen-specific IgGl, IgG2a, and IgE isotypes were measured
5 in the serum, indicating that the response resulted in systemic sensitization. Importantly, an
6 increase of eosinophils in the bronchio-alveolar lavage was observed with CB. Companion
7 studies with the intranasal exposure model showed that the adjuvant effect of CB particles was
8 even more pronounced when the particles were given during both the sensitization and challenge
9 phases; whereas administration during the challenge phase caused only marginal changes in the
10 immune response. These data show that PM can increase both the sensitization and challenge
11 responses to a protein antigen, and the immune stimulating activity of particles appears to be a
12 time-dependent process, suggesting that an inflammatory microenvironment (such as may be
13 created by the particles) is crucial for enhancing sensitization by particles.
14 Only a small number of studies have examined mechanisms underlying the enhancement
15 of allergic asthma by ambient urban particles. Ohtoshi et al. (1998) reported that a coarse size-
16 fraction of resuspended ambient PM, collected in Tokyo, induced the production of granulocyte
17 macrophage colony stimulating factor (GMCSF), an upregulator of dendritic cell maturation and
18 lymphocyte function, in human airway epithelial cells in vitro. In addition to increased GMCSF,
19 epithelial cell supernatants contained increased IL-8 levels when incubated with DPM, a
20 principal component of ambient particles collected in Tokyo. Although the sizes of the two
21 types of particles used in this study were not comparable, the results suggest that ambient PM, or
22 at least the DPM component of ambient PM, may be able to upregulate the immune response to
23 inhaled antigen through GMCSF production. Similarly, Takano et al. (1998) has reported airway
24 inflammation, airway hyperresponsiveness, and increased GMGSF and IL-5 in mice exposed to
25 diesel exhaust.
26 In a study by Walters et al. (2001), PM10 was found to induce airway hyperresponsiveness,
27 suggesting that PM exposure may be an important factor in increases in asthma prevalence.
28 Naive mice were exposed to a single dose (0.5 mg/ mouse) of ambient PM, coal fly ash, or diesel
29 PM. Exposure to PM10 induced increases in airway responsiveness and BAL cellularity; whereas
30 diesel PM induced significant increases in BAL cellularity, but not airway responsiveness.
31 On the other hand, coal fly ash exposure did not elicit significant changes in either of these
32 parameters. Ambient PM-induced airway hyperresponsiveness was sustained over 7 days. The
June 2003 7-97 DRAFT-DO NOT QUOTE OR CITE
-------
1 increase in airway responsiveness was preceded by increases in BAL eosinophils; whereas a
2 decline in airway responsiveness was associated with increases in macrophages. Thus, ambient
3 PM can induce asthma-like parameters in naive mice.
4 Several other studies have examined in greater detail the contribution of the particle
5 component and the organic fraction of DPM to allergic asthma. Tsien et al. (1997) treated
6 transformed IgE-producing human B lymphocytes in vitro with the organic extract of DPM. The
7 organic phase extraction had no effect on cytokine production but did increase IgE production.
8 In these in vitro experiments, DPM appeared to be acting on cells already committed to IgE
9 production, thus suggesting a mechanism by which the organic fraction of combustion particles
10 can directly affect B cells and influence human allergic asthma.
11 Cultured epithelial cells from atopic asthmatics show a greater response to DPM exposure
12 when compared with cells from nonatopic nonasthmatics. IL-8, GM-CSF, and soluble ICAM-1
13 increased in response to DPM at a concentration of 10 |ig/mL DPM (Bayram et al., 1998a,b).
14 This study suggests that particles could modulate airway disease through their actions on airway
15 epithelial cells. This study also suggests that bronchial epithelial cells from asthmatics are
16 different from those of nonasthmatics in regard to their mediator release in response to DPM.
17 Sagai and colleagues (1996) repeatedly instilled mice with DPM for up to 16 weeks and
18 found increased numbers of eosinophils, goblet cell hyperplasia, and nonspecific airway
19 hyperresponsiveness, changes which are central features of chronic asthma (National Institutes
20 of Health, 1997). Takano et al. (1997) extended this line of research and examined the effect of
21 repeated instillation of DPM on the antibody response to antigen OVA in mice. They observed
22 that antigen-specific IgE and IgG levels were significantly greater in mice repeatedly instilled
23 with both DPM and OVA. Because this upregulation in antigen-specific immunoglobulin
24 production was not accompanied by an increase in inflammatory cells or cytokines in lavage
25 fluid, it would suggest that, in vivo, DPM may act directly on immune system cells, as described
26 in the work by Tsien et al. (1997). Animal studies have confirmed that the adjuvant activity of
27 DPM also applies to the sensitization of Brown-Norway rats to timothy grass pollen
28 (Steerenberg et al., 1999).
29 Diaz-Sanchez and colleagues (1996) have continued to study the mechanism of DPM-
30 induced upregulation of allergic response in the nasal cavity of human subjects. In one study,
31 a 200 jiL aerosol bolus containing 0.15 mg of DPM was delivered into each nostril of subjects
32 with or without seasonal allergies. In addition to increases in IgE in nasal lavage fluid (NAL),
June 2003 7-98 DRAFT-DO NOT QUOTE OR CITE
-------
1 they found an enhanced production of IL-4, IL-6, and IL-13, cytokines known to be B cell
2 proliferation factors. The levels of several other cytokines also were increased, suggesting a
3 general inflammatory response to a nasal challenge with DPM. In a following study, these
4 investigators delivered ragweed antigen, alone or in combination with DPM, on two occasions,
5 to human subjects with both allergic rhinitis and positive skin tests to ragweed (Diaz-Sanchez
6 et al., 1997). They found that the combined challenge with ragweed antigen and DPM produced
7 significantly greater antigen-specific IgE and IgG4 in NAL. A peak response was seen at 96 h
8 postexposure. The combined treatment also induced expression of IL-4, IL-5, IL-10, and IL-13,
9 with a concomitant decrease in expression of Thl-type cytokines. Although the treatments were
10 not randomized (antigen alone was given first to each subject), the investigators reported that
11 pilot work showed no interactive effect of repeated antigen challenge on cellular and
12 biochemical markers in NAL. Diesel particulate matter also resulted in the nasal influx of
13 eosinophils, granulocytes, monocytes, and lymphocytes, as well as the production of various
14 inflammatory mediators. The combined DPM plus ragweed exposure did not increase the
15 rhinitis symptoms beyond those of ragweed alone. Thus, diesel exhaust (particles and gases) can
16 produce an enhanced response to antigenic material in the nasal cavity.
17 Extrapolation of these findings of enhanced allergic response in the nose to the human lung
18 would suggest that ambient combustion particles containing DPM may have significant effects
19 on allergic asthma. A study by Nordenhall et al. (2001) has addressed the effects of diesel PM
20 on airway hyperresponsiveness, lung function and airway inflammation in a group of atopic
21 asthmatics with stable disease. All were hyperresponsive to methacholine. Each subject was
22 exposed to DPM (300 |ig/m3) and air for 1 h on two separate occasions. Lung function was
23 measured before and immediately after the exposures. Sputum induction was performed 6 h,
24 and methacholine inhalation test 24 h, after each exposure. Exposure to DE was associated with
25 a significant increase in the degree of hyperresponsiveness, as compared to after air, a significant
26 increase in airway resistance and in sputum levels of interleukin (IL)-6 (p=0.048). No changes
27 were detected in sputum levels of methyl-histamine, eosinophil cationic protein,
28 myeloperoxidase, and IL-8.
29 These studies provide biological plausibility support for the exacerbation of allergic asthma
30 likely being associated with episodic exposure to PM. Although DPM may make up only a
31 fraction of the mass of urban PM, because of their small size, DPM may represent a significant
32 fraction of the ultrafine particle mode in urban air, especially in cities and countries that rely
June 2003 7-99 DRAFT-DO NOT QUOTE OR CITE
-------
1 heavily on diesel-powered vehicles; and number concentrations of ultrafme DPM may be
2 increasing due to manufacture and use of modern diesel engines, in contrast to decreases in DPM
3 mass concentrations over the past decades.
4 In an examination of the effect of concentrated ambient PM on airway responsiveness in
5 mice, Goldsmith et al. (1999) exposed control and ovalbumin-sensitized mice to an average
6 concentration of 787 |ig/m3 PM for 6 h/day for 3 days. Although ovalbumin sensitization itself
7 produced an increase in the nonspecific airway responsiveness to inhaled methylcholine,
8 concentrated ambient PM did not change the response to methylcholine in ovalbumin-sensitized
9 or control mice. For comparison, these investigators examined the effect of inhalation of an
10 aerosol of the active soluble fraction of ROFA on control and ovalbumin-sensitized mice and
11 found that ROFA could produce nonspecific airway hyperresponsiveness to methylcholine in
12 both control and ovalbumin-sensitized mice. Similar increases in airway responsiveness have
13 been observed after exposure to ROFA in normal and ovalbumin-sensitized rodents (Gavett
14 et al., 1997, 1999; Hamada et al., 1999, 2000).
15 Gavett et al. (1999) have investigated the effects of ROFA (intratracheal instillation) in
16 ovalbumin (OVA) sensitized and challenged mice. Instillation of 3 mg/kg (approximately 60
17 jig) ROFA induced inflammatory and physiological responses in the OVA mice that were related
18 to increases in Th2 cytokines (IL-4, IL-5). Compared to OVA sensitization alone, ROFA
19 induced greater than additive increases in eosinophil numbers and in airway responsiveness to
20 methylcholine.
21 Hamada et al. (1999, 2000) have examined the effect of a ROFA leachate aerosol in a
22 neonatal mouse model of allergic asthma. In the first study, neonatal mice sensitized by
23 intraperitoneal (ip) injection with OVA developed airway hyperresponsiveness, eosinophilia, and
24 elevated serum anti-ovalbumin IgE after a challenge with inhaled OVA. Exposure to the ROFA
25 leachate aerosol had no marked effect on the airway responsiveness to inhaled methacholine in
26 nonsensitized mice, but did enhance the airway hyperresponsiveness to methylcholine produced
27 in OVA-sensitized mice. No other interactive effects of ROFA exposure with OVA were
28 observed. In a subsequent study, Hamada et al. clearly demonstrated that, whereas inhaled OVA
29 alone was not sufficient to sensitize mice to a subsequent inhaled OVA challenge, pretreatment
30 with a ROFA leachate aerosol prior to the initial exposure to aerosolized OVA resulted in an
31 allergic response to the inhaled OVA challenge. Thus, exposure to a ROFA leachate aerosol can
June 2003 7-100 DRAFT-DO NOT QUOTE OR CITE
-------
1 alter the immune response to inhaled OVA both at the sensitization stage at an early age and at
2 the challenge stage.
3 Lambert et al. (1999) and Gilmour et al. (2001) also examined the effect of ROFA on a
4 rodent model of pulmonary allergy. Rats were instilled intratracheally with 200 or 1,000 jig
5 ROFA 3 days prior to sensitization with house dust mite (FIDM) antigen. FIDM sensitization
6 after 1,000 jig ROFA produced increased eosinophils, LDH, BAL protein, and IL-10 relative to
7 HDM alone. Although ROFA treatment did not affect antibody levels, it did enhance pulmonary
8 eosinophil numbers. The immediate bronchoconstrictive and associated antigen-specific IgE
9 response to a subsequent antigen challenge was increased in the ROFA-treated group in
10 comparison with the control group. Together, these studies suggest the components of ROFA
11 can augment the immune response to antigen.
12 Evidence that metals are responsible for the ROFA-enhancement of an allergic
13 sensitization was demonstrated by Lambert et al. (2000). In this follow-up study, Brown
14 Norway rats were instilled with 1 mg ROFA or the three main metal components of ROFA (iron,
15 vanadium, or nickel) prior to sensitization with instilled house dust mite. The three individual
16 metals were found to augment different aspects of the immune response to house dust mite.
17 Nickel and vanadium produced an enhanced immune response to the antigen as seen by higher
18 house dust mite-specific IgE serum levels after an antigen challenge at 14 days after
19 sensitization. Nickel and vanadium also produced an increase in the lymphocyte proliferative
20 response to antigen in vitro. In addition, the antigen-induced bronchoconstrictive response was
21 greater only in nickel-treated rats. Thus, instillation of metals at concentrations equivalent to
22 those present in the ROFA leachate mimicked the response to ROFA, suggesting that the metal
23 components of ROFA are responsible for the increased allergic sensitization observed in ROFA-
24 treated animals.
25 Although these studies demonstrate that inhalation or instillation of ROFA augments the
26 immune response in allergic hosts, the applicability of these findings to ambient PM is an
27 important consideration. Goldsmith et al. (1999) have compared the effect of inhalation of
28 concentrated ambient PM for 6 h/day for 3 days versus the effect of a single exposure to a ROFA
29 leachate aerosol on the airway responsiveness to methylcholine in OVA-sensitized mice.
30 Exposure to ROFA leachate aerosols significantly enhanced the airway hyperresponsiveness in
31 OVA-sensitized mice; whereas exposure to concentrated ambient PM (average concentration of
32 787 |ig/m3) had no effect on airway responsiveness in six separate experiments. Thus, the effect
June 2003 7-101 DRAFT-DO NOT QUOTE OR CITE
-------
1 of the ROFA leachate aerosols on the induction of airway hyperresponsiveness in allergic mice
2 was significantly different than that of a high concentration of concentrated ambient PM.
3 Although airway responsiveness was examined at only one post-exposure time point, these
4 findings do suggest that a great deal of caution should be used in interpreting the results of
5 studies using ROFA particles or leachates in the attempt to investigate the biologic plausibility
6 of the adverse health effects of PM.
7
8 7.5.4 Resistance to Infectious Disease
9 The development of an infectious disease requires both the presence of the appropriate
10 pathogen, as well as host susceptibility to the pathogen. There are numerous specific and
11 nonspecific host defenses against microbes, and the ability of inhaled particles to modify
12 resistance to bacterial infection could result from a decreased ability to clear or kill microbes.
13 Rodent infectivity models frequently have been used to examine the effect of inhaled particles
14 on host defense and infectivity. Mice or rats are challenged with a bacterial or viral load either
15 before or after exposure to the particles (or gas) of interest; mortality rate, survival time, or
16 bacterial clearance are then examined. A number of studies that have used the infectivity model
17 to assess the effect of inhaled PM were discussed previously (U.S. Environmental Protection
18 Agency, 1982, 1989, 1996a). In general, acute exposure to sulfuric acid aerosols at
19 concentrations up to 5,000 |ig/m3 were not very effective in enhancing mortality in a bacterially
20 mediated murine model. In rabbits, however, sulfuric acid aerosols altered anti-microbial
21 defenses after exposure for 2 h/day for 4 days to 750 |ig/m3 (Zelikoff et al., 1994). Acute or
22 short-term repeated exposures to high concentrations of relatively inert particles have produced
23 conflicting results. Carbon black (10,000 |ig/m3) was found to have no effect on susceptibility to
24 bacterial infection (Jakab, 1993); whereas TiO2 (20,000 |ig/m3) decreased the clearance of
25 microbes and the bacterial response of lymphocytes isolated from mediastinal lymph nodes
26 (Gilmour et al., 1989a,b). In addition, exposure to DPM (2 mg/m3, 7h/d, 5d/wk for 3 and 6 mo)
27 has been shown to enhance the susceptibility of mice to the lethal effects of some, but not all,
28 microbial agents (Hahon et al., 1985). Thus, the pulmonary response to microbial agents has
29 been shown to be altered at relatively high particle concentrations in animal models. Moreover,
30 these effects appear to be highly dependent on the microbial challenge and the test animal
31 studied. Pritchard et al. (1996) observed in rats exposed to particles with a high concentration of
June 2003 7-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 metals (e.g., ROFA), that the increased mortality rate after streptococcus infection was
2 associated with the amount of metal in the PM.
3 There are few recent studies that have examined mechanisms potentially responsible for
4 the effect of PM on infectivity. In one study, Cohen and colleagues (1997) examined the effect
5 of inhaled vanadium (V) on immunocompetence. Healthy rats were repeatedly exposed to
6 2 mg/m3 V, as ammonium metavanadate, and then instilled with polyinosinic-polycytidilic acid
7 (poly I:C), a double-stranded polyribonucleotide that acts as a potent immunomodulator.
8 Induction of increases in lavage fluid protein and neutrophils was greater in animals preexposed
9 to V. Similarly, IL-6 and interferon-gamma were increased in V-exposed animals. Alveolar
10 macrophage function, as determined by zymosan-stimulated superoxide anion production and by
11 phagocytosis of latex particles, was depressed to a greater degree after poly I:C instillation in V-
12 exposed rats as compared to filtered air-exposed rats. These findings provide evidence that
13 inhaled V, a trace metal found in combustion particles and shown to be toxic in vivo in studies
14 using instilled or inhaled ROFA (Dreher et al., 1997; Kodavanti et al., 1997b, 1999), has the
15 potential to inhibit the pulmonary response to microbial agents. However, it must be
16 remembered that these effects were found at very high exposure concentrations of V, and as with
17 many studies, care must be taken in extrapolating the results to the ambient exposure of healthy
18 individuals or those with preexisting cardiopulmonary disease to trace concentrations (~3 orders
19 of magnitude lower concentration) of metals in ambient PM.
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
25 following discussion examines effects of mixtures of ambient PM, or PM surrogates, with
26 gaseous pollutants. Ambient PM co-exists in indoor and outdoor air with a number of co-
27 pollutant gases, including ozone, sulfur dioxide, oxides of nitrogen, and carbon monoxide and
28 innumerable other non-PM components that are not routinely measured. Toxicological
29 interactions between PM and gaseous co-pollutants may be antagonistic, additive, or synergistic
30 (Mauderly, 1993). The presence and nature of any interaction appears to depend on the chemical
31 composition, size, concentration and ratios of pollutants in the mixture, exposure duration, and
32 the endpoint being examined. It may be difficult to predict a priori from the presence of certain
June 2003 7-103 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
5 synergy between particles and gases, especially if effects occur at concentrations at which no
6 effects occur when individual constituents are inhaled. Various physical and chemical
7 mechanisms may underlie synergism. For example, physical adsorption or absorption of some
8 material on a particle could result in transport to more sensitive sites, or sites where this material
9 would not normally be deposited in toxic amounts. This physical process may explain the
10 interaction found in studies of mixtures of carbon black and formaldehyde or of carbon black
11 and acrolein (Jakab, 1992, 1993).
12 Chemical interactions between PM and gases can occur on particle surfaces, thus forming
13 secondary products whose surface layers may be more active lexicologically than the primary
14 materials and that can then be carried to a sensitive site. The hypothesis of such chemical
15 interactions has been examined in the gas and particle exposure studies by Amdur and
16 colleagues (Amdur and Chen, 1989; Chen et al., 1992) and Jakab and colleagues (Jakab and
17 Hemenway, 1993; Jakab et al., 1996). These investigators have suggested that synergism occurs
18 as secondary chemical species are produced, especially under conditions of increased
19 temperature and relative humidity.
20 Another potential mechanism of gas-particle interaction may involve a pollutant-induced
21 change in the local microenvironment of the lung, enhancing the effects of the co-pollutant.
22 For example, Last et al. (1984) suggested that the observed synergism between ozone (O3) and
23 acid sulfates in rats was due to a decrease in the local microenvironmental pH of the lung
24 following deposition of acid, enhancing the effects of O3 by producing a change in the reactivity
25 or residence time of reactants, such as radicals, involved in O3-induced tissue injury. Likewise,
26 Pinkerton et al. (1989) showed increased retention of the mass and number of asbestos fibers in
27 rats exposed to O3, suggesting an increase in lung fiber burden due to exposure to this gaseous
28 pollutant.
29 Vincent et al. (1997) exposed rats to 0.8 ppm O3 in combination with 5 or 50 mg/m3 of
30 resuspended urban particles for 4 h. Although PM alone caused no change in cell proliferation
31 (3H-thymidine labeling), co-exposure to either concentration of resuspended PM with O3 greatly
32 potentiated the proliferative effects of exposure to O3 alone. These interactive changes occurred
June 2003 7-104 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-12. RESPIRATORY AND CARDIOVASCULAR EFFECTS OF PM AND GASEOUS POLLUTANT MIXTURES
to
O
o
Species, Gender,
Strain Age, or Body
Weight
Rats, Fischer NNia,
male, 22 to 24 mo old
Gases and PM
Carbon,
ammonium
bisulfate,
andO3
Exposure Technique
Inhalation
Mass Concentration
50 ug/m3 carbon +
70 ug/m3 ammonium
bisulfate + 0.2 ppm
O3 or 100 ug/m3
carbon +140 ug/m3
ammonium bisulfate
+ 0.2 ppm O3
Exposure
Particle Size Duration
0.4 urn MMAD 4 h/day,
ag = 2.0 3 days/week
for 4 weeks
Cardiopulmonary Effects of Inhaled
PM and Gases
No changes in protein concentration in
lavage fluid or in prolyl 4-hydroxylase
activity in blood. Slight, but
statistically significant decreases in
plasma fibronectin in animals exposed
to the combined atmospheres compared
to animals exposed to O3 alone.
Reference
Bolarin et al.
(1997)
fe
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Rats
Humans; healthy
15 M, 10 F,
34.9+-10 years of age
Humans; healthy
children
Humans; 59 healthy
children in Mexico
City; 19 controls in
Gulf port town
Humans; 1 5 healthy
children in Mexico
City; 1 1 children in
Veracruz; 4-15 years
of age
O3 and Ottawa
urban dust
CAPs
Ambient gases
and particles
Ambient gases
and particles
Ambient gases
and particles
Inhalation 40,000 ug/m3 and 4.5 um
0.8 ppm O3 MMAD
Inhalation 150 ug/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
Single 4-h Co-exposure to particles potentiated
exposure O3-induced septal cellurity. Enhanced
followed by septal thickening associated with
20 h clean elevated production of macrophage
air 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.
Bouthillier
etal. (1998)
Brook et al.
(2002)
Calderon-
Garciduefias
et al. (2000a)
Calderon-
Garciduefias
et al. (2000b)
Calderon-
Garciduefias
etal. (200 la)
-------
to
O
o
TABLE 7-12 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF PM AND
GASEOUS POLLUTANT MIXTURES
fe
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Species, Gender,
Strain Age, or Body
Weight
Humans; 83 healthy
children in Mexico
City; 24 children in
Isla Mujeres;
6-12 years of age
Dogs, 109 healthy
male and female
mongrels form
Mexico City; 43 dogs
from less-polluted
cities
Mice, Swiss, female,
5 weeks old
Rats, S-D, male,
250-300 g
Particle Exposure
Gases and PM Exposure Technique Mass Concentration Size Duration
Ambient gases Natural 24 h exposure
and particles in SWMMC
compared to low
pollution Caribbean
Ambient gases Natural 24 h exposure
and particles in SWMMC and
NWMMC compared
to low pollution cities
SO2 and carbon Inhalation, 10,000 ug/m3 carbon O.SumMMAD Single 4-h
flow-past, with or without 5 to ag = 2.7 exposure
nose-only 20 ppm SO2 at 1 0%
or 85% RH
H2SO4 and O3 Inhalation, 500 ug/m3 H2SO4 Fine (0.3 urn 4 h/day for
nose-only aerosol (two different MMD, ag = 2 days
particle sizes), 1 .7) and
with or without ultrafine
0.6 ppm O3 (0.06 urn,
ag = 1.4)
Cardiopulmonary Effects of Inhaled
PM and Gases
Nasal biopsies revealed p53
accumulation by immunochemistry;
increased upper and lower respiratory
symptoms.
LM and EM of lungs exhibited patchy
chronic mononuclear cell infiltrates and
AMs loaded with particles; bronchiolar
and smooth muscle hyperplasia;
peribronchiolar fibrosis; BAL
demonstrated proliferating AMs.
LM and EM of heart exhibited
increased myocardial abnormalities and
including apototic myocytes,
endothelial and immune effector cells,
degramulated mast cells, and clusters
of adipocytes.
Macrophage phagocytosis was
depressed only in animals exposed to
the combination of SO2 and carbon at
85% humidity. This inhibition
in macrophage function lasted at least
7 days after exposure.
The volume percentage of injured
alveolar septae was increased only in
the combined ultrafine acid/O3 animals.
BrdU labeling in the periacinar region
was increased in a synergistic manner
in the combined fine acid/O3 animals.
Reference
Calderon-
Garciduefias
etal. (200 Ib)
Calderon-
Garciduefias
etal. (2001 c)
Calderon-
Garciduefias
etal. (200 Id)
Jakab et al.
(1996)
Clark et al.
(2000)
Kimmel et al.
(1997)
-------
fe
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
TABLE 7-12 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF PM AND
GASEOUS POLLUTANT MIXTURES
O Species,
00 Gender, Strain
Age, or Body Exposure
Weight Gases and PM Technique
Rats, S-D O3 and H2SO4-coated Inhalation,
300 g carbon nose-only
Rats O3 + elemental carbon Inhalation
+ ammonium bisulfate
Mass Concentration
0.2 ppm O3
+ 50 ug/m3 C
+ 100 ug/m3 H2SO4
0.4 ppm O3
+250 ug/m3 C
+500 ug/m3 H2SO4
0.2 ppm O3 +
carbon 50 um/m3
ammonium
Bisulfate 70 ug/m3
Particle
Size
0.26 um
ag = 2.2
0.46 um
0.3 um
Exposure
Duration
4 h/day for 1 day
or 5 days
4hr/d
3d/wk
4wk
Cardiopulmonary Effects of Inhaled
PM and Gases
No airway inflammation at low dose.
Greater inflammatory response at high
dose; greater response at 5 days than
1 day. Contrasts with O3 alone where
inflammation was greatest at 0.40 ppm
on Day 1.
Increased macrophage phagocytosis and
increased respiratory burst; decreased
lung collagen.
Reference
Kleinman
etal. (1999)
Kleinman
etal.
(2000)
Mice, BALB/c,
3 days old
CAPs
(Boston)
O,
Inhalation
63-1569 ug/m3
0.3 ppm
PM25
5h
A small increase in pulmonary
resistance and airway responsiveness
was found in both normal mice and
mice with ovalbumin-induced asthma
immediately after exposure to CAPs, but
not O3; no evidence of synergy; activity
attributed to the AISi PM component.
Kobzik
etal. (2001)
Rats
Humans,
children,
healthy and
asthmatic
Pigeons
(Columba livia)
H2SO4 and O3
H2S04,
SO2, and O3
Ambient gases and
particles
Inhalation,
whole body
Inhalation
Natural 24-h
exposure in urban
and rural areas
around Madrid,
Spain
20 to 150 ug/m3 0.4 to
H2SO4and0.12or 0.8 urn
0.2 ppm O3
60 to 140 ug/m3 0.6 urn
H2SO4, 0.lppm H2SO4
SO2, and 0.1 ppm
03
Intermittent
(12 h/day) or
continuous
exposure for up
to 90 days
Single 4-h
exposure with
intermittent
exercise
Continuous
ambient
exposure
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.
Increased number of AMs and
decreased number of lamellar bodies in
type II epithelial cells in urban pigeons.
Last and
Pinkerton
(1997)
Linn et al.
(1997)
Lorz and
Lopez
(1997)
-------
*-H
to
2 Species,
Gender, Strain
Age, or Body
Weight
Rats, F344/N
male
Rats, F344,
9-weeks-old,
male and
female
TABLE 7-12
Gases and PM
O3 + nitric acid
NO2 + carbon particles
+ ammonium
bisulfate
Ambient gases and
particles
(cont'd). RESPIRATORY AND CARDIOVASCULAR
GASEOUS POLLUTANT MIXTURES
Exposure
Technique
Inhalation
Natural 23 h/day
exposure to filtered
and unfiltered
Mexico City air.
Mass Concentration
0.018ppmO3
3.3 ppb CH20
0.068 mg/m3 TSP
0.032 mg/m3 PM10
0.016 mg/m3 PM2,
Particle Exposure
Size Duration
4h/d
3d/wk
4wk
23 h/day
for 7 weeks
EFFECTS OF PM AND
Cardiopulmonary Effects of Inhaled
PM and Gases
Decreases in macrophage Fc-receptor
mediated-phagocytosis, increased
epithelial permeability and proliferation,
altered breathing pattern.
Histopathology examination revealed no
nasal lesions in exposed or control rats;
tracheal and lung tissue from both
groups showed similar levels of minor
abnormalities.
Reference
Mautz et al.
(2001)
Moss et al.
(2001)
1
o
oo
O
h>
£
H
6
o
0
H
0
H
W
O
o
i — 1
H
W
Rats, F344/N O3 + nitric acid NO2 +
male carbon particles +
ammonium bisulfate
Dogs Ambient gases and
particles
Rats O3 and resuspended
urban PM
Inhalation
Natural 24-h
exposure in four
urban areas of
Mexico City and
one rural area
Inhalation, 0.8ppmO3and
whole-body 5,000 or
50,000 ug/m3 PM
4h/d
3d/wk
40 wk
Continuous
ambient
exposure
Single 4-h
exposure
Increased lung putrescine content.
No significant differences in AMs or
total cell counts in lavage from dogs
studied among the five regions. A
significant increase in lavage fluid
neutrophils and lymphocytes in the
southwest region, where the highest O3
levels were recorded, compared to the
two industrial regions with the highest
PM levels.
PM alone caused no change in cell
proliferation in bronchioles or
parenchyma. Co-exposure with O3
greatly potentiated the proliferative
changes induced by O3 alone. These
changes were greatest in the epithelium
of the terminal bronchioles and alveolar
ducts.
Sindhu
etal.
(1998)
Vanda et al.
(1998)
Vincent
etal.
(1997)
-------
1 in epithelial cells of the terminal bronchioles and the alveolar ducts. These findings using
2 resuspended dusts, although at high concentrations, are consistent with studies demonstrating
3 interaction between sulfuric acid (H2SO4) aerosols and O3. Kimmel and colleagues (1997)
4 examined the effect of acute co-exposure to O3 (0.6 ppm) and fine (MMD = 0.3 jim) or ultrafme
5 (MMD = 0.06 |im) H2SO4 aerosols (0.5 mg/m3) on rat lung morphology. They determined
6 morphometrically that alveolar septal volume was increased in animals co-exposed to O3 and
7 ultrafme, but not fine, H2SO4. Interestingly, cell labeling, an index of proliferative cell changes,
8 was increased only in animals co-exposed to fine H2SO4 and O3, as compared to animals exposed
9 to O3 alone. Importantly, Last and Pinkerton (1997) extended their previous work and found that
10 subchronic exposure to acid aerosols (20 to 150 |ig/m3 H2SO4) had no interactive effect on the
11 biochemical and morphometric changes produced by either intermittent or continuous O3
12 exposure (0.12 to 0.2 ppm). Thus, the interactive effects of O3 and acid aerosol co-exposure in
13 the lung disappeared during the long-term exposure.
14 Kleinman et al. (1999) examined the effects of O3 (0.2 and 0.4 ppm) plus fine
15 (MMAD = 0.26 |im), H2SO4-coated, carbon particles (100, 250, and 500 |ig/m3) for 1 or 5 days.
16 They found the inflammatory response with the O3-particle mixture was greater after 5 days
17 (4 h/day) than after Day 1. This contrasted with O3 exposure alone (0.4 ppm), which caused
18 marked inflammation on acute exposure, but no inflammation after 5 consecutive days of
19 exposure.
20 Kleinman et al. (2000) examined the effects of a mixture of elemental carbon particles
21 (50 |ig/m3), O3 (0.2 ppm), and ammonium bisulfate (70 |ig/m3) on rat lung collagen content and
22 macrophage activity. Decreases in lung collagen, and increases in macrophage respiratory burst
23 and phagocytosis were observed relative to other pollutant combinations. Mautz et al. (2001)
24 used a similar mixture (i.e., elemental carbon particles, O3, ammonium bisulfate, but with NO2
25 also) and exposure regimen as Kleinman et al. (2000). There were decreases in pulmonary
26 macrophage Fc-receptor binding and phagocytosis and increases in acid phosphatase staining.
27 Bronchoalveolar epithelial permeability cell proliferation were increased. Altered breathing-
28 patterns were also observed, with some adaptations occurring.
29 Studies have examined interactions between carbon particles and gaseous co-pollutants.
30 Jakab et al. (1996) and Clarke et al. (2000c) challenged mice with a single 4-h exposure to a high
31 concentration of carbon particles (10 mg/m3) in the presence of 10 ppm SO2 (-140 jig cpSO42") at
32 low and high relative humidities. Macrophage phagocytosis was depressed significantly only in
June 2003 7-109 DRAFT-DO NOT QUOTE OR CITE
-------
1 mice exposed to the combined pollutants under high relative humidity (85%) conditions. There
2 was no evidence of an inflammatory response based on total cell counts and differential cell
3 counts from BAL; however, macrophage phagocytosis remained depressed for 7 to 14 days.
4 Intrapulmonary bactericidal activity also was suppressed and remained suppressed for 7 days.
5 This study suggests that fine carbon particles can serve as an effective carrier for acidic sulfates
6 where chemical conversion of adsorbed SO2 to acid sulfate species occurred. Interestingly, the
7 depression in macrophage function was present as late as 7 days postexposure. Bolarin et al.
8 (1997) exposed rats to only 50 or 100 |ig/m3 carbon particles in combination with ammonium
9 bisulfate and O3. Despite 4 weeks of exposure, they observed no changes in protein
10 concentration in lavage fluid or blood prolyl 4-hydroxylase, an enzyme involved in collagen
11 metabolism. Slight decreases in plasma fibronectin were present in animals exposed to the
12 combined pollutants versus O3 alone. Thus as, previously noted, the potential for adverse effects
13 in the lungs of animals challenged with a combined exposure to particles and gaseous pollutants
14 is dependent on numerous factors, including the gaseous co-pollutant, concentration, and time.
15 In a complex series of exposures, Oberdorster and colleagues examined the interaction of
16 ultrafine carbon particles (100 |ig/m3) and O3 (1 ppm) in young and old Fischer 344 rats that
17 were pretreated with aerosolized endotoxin (Elder et al., 2000a,b). In old rats, exposure to
18 carbon and O3 produced an interaction that resulted in a greater influx in neutrophils than that
19 produced by either agent alone. This interaction was not seen in young rats. Oxidant release
20 from lavage fluid cells was also assessed and the combination of endotoxin, carbon particles, and
21 O3 produced an increase in oxidant release in old rats. This combination produced the opposite
22 response in the cells recovered from the lungs of the young rats, indicating that the lungs of the
23 aged animals underwent greater oxidative stress in response to this complex pollutant mix of
24 particles, O3, and a biogenic agent.
25 Wagner et al. (2001) examined the synergistic effect of co-exposure to O3 and endotoxin
26 on the transition and respiratory epithelium of rats that also was mediated, in part, by
27 neutrophils. Fisher 344 rats (10 to 12 week old) exposed to 0.5 ppm O3, 8 h per day, for 3 days,
28 developed mucous cell metaplasia in the nasal transitional epithelium, an area normally devoid
29 of mucous cells; whereas, intratracheal instillation of endotoxin (20 jig) caused mucous cell
30 metaplasia rapidly in the respiratory epithelium of the conducting airways. A synergistic
31 increase of intraepithelial mucosubstances and morphological evidence of mucous cell
June 2003 7-110 DRAFT-DO NOT QUOTE OR CITE
-------
1 metaplasia were found in rat maxilloturbinates upon exposure to both ozone and endotoxin,
2 compared to each pollutant alone.
3 The effects of O3 modifying the biological potency of PM (diesel PM and carbon black)
4 was examined by Madden et al. (2000). Reaction of NIST Standard Reference Material # 2975
5 diesel PM with 0.1 ppm O3 for 48 hr increased the potency (compared to unexposed or
6 air-exposed diesel PM) to induce neutrophil influx, total protein, and LDH in lung lavage fluid in
7 response to intratracheal instillation. Exposure of the diesel PM to high, non-ambient O3
8 concentration (1.0 ppm) attenuated the increased potency, suggesting destruction of the bioactive
9 reaction products. Unlike the diesel particles, carbon black particles exposed to 0.1 ppm O3 did
10 not exhibit an increase in biological potency, which suggested that the reaction of organic
11 components of the diesel PM with O3 were responsible for the increased potency. Reaction of
12 particle components with O3 was ascertained by chemical determination of specific classes of
13 organic compounds.
14 The interaction of PM and O3 was further examined in a murine model of ovalbumin
15 (OVA)-induced asthma. Kobzik et al. (2001) investigated whether coexposure to inhaled,
16 concentrated PM from Boston, MA and to O3 could exacerbate asthma-like symptoms. On days
17 7 and 14 of life, half of the BALB/c mice used in this study were sensitized by ip injection of
18 OVA and then exposed to OVA aerosol on three successive days to create the asthma phenotype.
19 The other half received the ip OVA, but were exposed to a phosphate-buffered saline aerosol
20 (controls). The mice were further subdivided (n > 6I/group) and exposed for 5 h to CAPs,
21 ranging from 63 to 1,569 |ig/m3, 0.3 ppm O3, CAPs + O3, or to filtered air. Pulmonary resistance
22 and airway responsiveness to an aerosolized MCh challenge were measured after exposures.
23 A small, statistically significant increase in pulmonary resistance and airway responsiveness,
24 respectively, was found in both normal and asthmatic mice immediately after exposure to CAPs
25 alone and to CAPs + O3, but not to O3 alone or to filtered air. By 24 h after exposure, the
26 responses returned to baseline levels. There were no significant increases in airway
27 inflammation after any of the pollutant exposures. In this well-designed study of a small-animal
28 model of asthma, O3 and CAPs did not appear to be synergistic. In further analysis of the data
29 using specific elemental groupings of the CAPs, the acutely increased pulmonary resistance was
30 found to be associated withe the AISi fraction of PM. Thus, some components of concentrated
31 PM2 5 may affect airway caliber in sensitized animals, but the results are difficult to extrapolate
32 to people with asthma.
June 2003 7-111 DRAFT-DO NOT QUOTE OR CITE
-------
1 Linn and colleagues (1997) examined the effect of a single exposure to 60 to 140 |ig/m3
2 H2SO4, 0.1 ppm SO2, and 0.1 ppm O3 in healthy and asthmatic children. The children performed
3 intermittent exercise during the 4-h exposure to increase the inhaled dose of the pollutants.
4 An overall effect on the combined group of healthy and asthmatic children was not observed.
5 A positive association between acid concentration and symptoms was seen, however, in the
6 subgroup of asthmatic children. The combined pollutant exposure had no effect on spirometry in
7 asthmatic children, and no changes in symptoms or spirometry were observed in healthy
8 children. Thus, the effect of combined exposure to PM and gaseous co-pollutants appeared to
9 have less effect on asthmatic children exposed under controlled laboratory conditions in
10 comparison with field studies of children attending summer camp (Thurston et al., 1997).
11 However, prior exposure to H2SO4 aerosol may enhance the subsequent response to O3 exposure
12 (Linn et al., 1994; Frampton et al., 1995); and the timing and sequence of the exposures may be
13 important.
14 Six unique animal studies have examined the adverse cardiopulmonary effects of complex
15 mixtures in urban and rural environments of Italy (Gulisano et al., 1997), Spain (Lorz and Lopez,
16 1997), and Mexico (Vanda et al., 1998; Calderon-Garciduefias et al., 2001c,d; Moss et al., 2001).
17 Five of these studies, identified in Table 7-11, have taken advantage of the differences in
18 pollutant mixtures of urban and rural environments to report primarily morphological changes in
19 the nasopharynx (Calderon-Garciduefias et al., 2001c), the lower respiratory tract (Gulisano
20 et al., 1997; Lorz and Lopez, 1997; Calderon-Garciduefias et al., 2001c) and in the heart
21 (Calderon-Garciduefias et al., 2001d) of lambs, pigeons, and dogs, respectively, after natural,
22 continuous exposures to ambient pollution. Each study has provided evidence that animals
23 living in urban air pollutants have greater pulmonary and cardiac changes than would occur in a
24 rural and presumably cleaner, environment. The study by Moss et al. (2001) examined the nasal
25 and lung tissue of rats exposed (23 h/day) to Mexico City air for up to 7 weeks and compared
26 them to controls similarly exposed to filtered air. No inflammatory or epithelial lesions were
27 found using quantitative morphological techniques; however, the concentrations of pollutants
28 were low (see Table 7-11). Extrapolation of these results to humans is restricted, however, by
29 uncontrolled exposure conditions, small sample sizes, and other unknown exposure and
30 nutritional factors in the studies in mammals and birds, and the negative studies in rodents. They
31 also bring up the issue of which species of "sentinel" animals is more useful for predicting
32 pollutant effects in humans. Thus, in these field studies, it is difficult to assign a specific role to
June 2003 7-112 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM (or to any other component of the mixture) in the significant cardiopulmonary effects
2 reported.
3 Similar morphological changes (Calderon-Garciduefias et al., 2000a; 2001a,b) and chest
4 X-ray evidence of mild lung hyperinflation (Calderon-Garciduefias et al., 2000b) have been
5 reported in children residing in urban and rural areas of Mexico City. The ambient air in urban
6 areas, particularly in Southwest Metropolitan Mexico City (SWMMC), is a complex mixture of
7 particles and gases, including high concentrations of O3 and aldehydes that previously have been
8 shown to cause airway inflammation and epithelial lesions in humans (e.g., Calderon-
9 Garciduefias et al., 1992, 1994, 1996) and laboratory animals (Morgan et al., 1986; Heck et al.,
10 1990; Harkema et al., 1994, 1997a,b). The described effects demonstrate a persistent, ongoing
11 upper and lower airway inflammatory process and chest X-ray abnormalities in children residing
12 predominantly in SWMMC. Again, extrapolation of these results to urban populations of the
13 United States is difficult because of the complexity of urban air in Mexico City, and the altitude,
14 the uncontrolled exposure conditions, and other unknown exposure and nutritional factors.
15 However, these results may represent an upper bound on what might be the effects of PM in the
16 United States.
17 Only one controlled study has examined the effect of a combined inhalation exposure to
18 CAPs and O3 in human subjects. In a randomized, double-blind crossover study, Brook et al.
19 (2002) exposed 25 healthy male and female subjects, 34.9 ± 10 (SD) years of age, to filtered
20 ambient air containing 1.6 |ig/m3 PM25 and 9 ppb O3 (control) or to unfiltered air containing
21 150 |ig/m3 CAPs and 120 ppb O3 while at rest for 2 h. Blood pressure was measured and high-
22 resolution brachial artery ultrasonography was performed prior to and 10 min after exposure.
23 The brachial artery ultrasonography (B AUS) technique was used to measure brachial artery
24 diameter (BAD), endothelium-dependent flow-mediated dilation (FMD), and endothelial-
25 independent nitroglycerine-mediated dilation (NMD). Although no changes in blood pressure or
26 endothelial-dependent or independent dilatation were observed, a small (2.6%) but statistically
27 significant (p = 0.007) decrease in BAD was observed in CAPs plus O3 exposures (-0.09 mm)
28 when compared to filtered air exposures (+0.01 mm). Pre-exposure BAD showed no significant
29 day-to-day variation (0.03 mm), and no significant exposure differences were found for other
30 gaseous pollutants (CO, NOX, SO2) in the ambient air. This finding suggests that combined
31 exposure to a mixture of CAPs and O3 produces vasoconstriction, potentially via autonomic
32 reflexes or as a result of an increase in circulating endothelin, as has been described in rats
June 2003 7-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposed to urban PM (Vincent et al., 2001). It is not known, however, whether this effect is
2 caused by CAPS or O3 alone, or if vasoactive responses would be found at PM25 and O3
3 concentrations typically found in most urban locations in North America.
4 The effects of gaseous pollutants on PM-mediated responses also have been examined by
5 in 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 O3 may increase the penetration of some types of PM into
8 epithelial cells. Additionally, Madden et al. (2000) demonstrated a greater potency for ozonized
9 diesel PM to induce prostaglandin E2 production from human epithelial cell cultures, suggesting
10 that O3 can modify the biological activity of PM derived from diesel exhaust.
11
12
13 7.7 SUMMARY OF KEY FINDINGS AND CONCLUSIONS
14 Toxicological studies can play an integral role in addressing several key issues regarding
15 ambient PM health effects:
16 (1) What characteristics (size, chemical composition, etc.) of ambient PM cause or contribute
to health effects?
17 (2) What evidence is available for elucidating potential mechanisms underlying PM health
effects?
18 (3) What susceptible subgroups are at increased risk for ambient PM health effects and what
types of factors contribute to their increased susceptibility?
19 (4) What evidence exists that illustrates examples of interactive effects of particles and
gaseous copollutants?
20 This summary focuses on highlighting salient findings that reflect the notable progress that
21 toxicological studies have made towards addressing these questions. All these questions have
22 especially important implications bearing on the issue of biological plausibility of
23 epidemiologically-observed ambient PM effects.
24
25
26
June 2003 7-114 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.7.1 Links Between Specific Particulate Matter Components and
2 Health Effects
3 Key to the validity of the biological plausibility is the need to understand the linkage
4 between the components of airborne PM responsible for the adverse effects and the individuals
5 at risk. The plausibility of epidemiologically-demonstrated associations between ambient PM
6 and increases in morbidity and mortality has been questioned because adverse cardiopulmonary
7 effects have been observed among human populations at very low ambient PM concentrations.
8 To date, toxicology studies on PM have provided only limited evidence for specific PM
9 components potentially being responsible for observed cardiopulmonary effects of ambient PM.
10 Studies have shown that some components of particles are more toxic than others. For example,
11 high concentrations of ROFA and associated soluble metals have produced clinically significant
12 effects (including death) in compromised animals. The relevance of these findings to
13 understanding the adverse effects of PM components is tempered, however, by the large
14 difference between metal concentrations delivered to the test animals and metal concentrations
15 present in the ambient urban environment. Such comparisons must be applied to the
16 interpretation of all studies that examine the individual components of ambient urban PM. Key
17 findings regrading potential contributions of individual physical/chemical factors of particles to
18 cardiopulmonary effects are summarized below.
19
20 7.7.1.2 Acid Aerosols
21 There is relatively little new information on the effects of acid aerosols, and the
22 conclusions of the 1996 PM AQCD are unchanged. It was previously concluded that acid
23 aerosols cause little or no change in pulmonary function in healthy subjects, but asthmatics may
24 develop small changes in pulmonary function. This conclusion is supported by the recent study
25 of Linn and colleagues (1997) in which children (26 children with allergy or asthma and
26 15 healthy children) were exposed to sulfuric acid aerosol (100 |ig/m3) for 4 h. There were no
27 significant effects on symptoms or pulmonary function when data from the entire group were
28 analyzed, but the allergy group had a significant increase in symptoms after the acid aerosol
29 exposure. Accordingly, acid aerosol health effects may represent a possible causal physical
30 property for PM-related health effects. However, it is unlikely that particle acidity alone could
31 account for the pulmonary function effects (Dreher, 2000).
June 2003 7-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 Although pulmonary effects of acid aerosols have been the subject of extensive research in
2 past decades, the cardiovascular effects of acid aerosols have received little attention. Zhang
3 et al. (1997) reported that inhalation of acetic acid fumes caused reflex-mediated increases in
4 blood pressure in normal and spontaneously hypertensive rats. Thus, acid components should
5 not be ruled out as possible mediators of PM health effects. In particular, the cardiovascular
6 effects of acid aerosols at realistic concentrations need further investigation.
7
8 7.7.1.3 Metals
9 The 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) mainly relied on
10 data related to occupational exposures to evaluate the potential toxicity of metals in particulate
11 air pollution. Since that time, newly published in vivo and in vitro studies using ROFA or
12 soluble transition metals have contributed substantial further information on the health effects of
13 particle-associated soluble metals. Although there are some uncertainties about differential
14 effects of one transition metal versus another, water soluble metals leached from ROFA or
15 ambient filter extracts have been shown consistently (albeit at high concentrations) to cause cell
16 injury and inflammatory changes in vitro and in vivo.
17 Even though it is clear that combustion particles that have a high content of soluble metals
18 can cause lung injury and even death in compromised animals and correlate well with
19 epidemiological findings in some cases (e.g., Utah Valley Studies), it has not been established
20 that the small quantities of metals associated with ambient PM are sufficient to cause health
21 effects. Moreover, it cannot be assumed that metals are the primary toxic component of ambient
22 PM, nor that there is a single primary toxic component. Rather there may be many such
23 components. In studies in which various ambient and emission source particulates were instilled
24 into rats, the soluble metal content did appear to be the primary determinant of lung injury
25 (Costa and Dreher, 1997). However, one published study (Kodavanti et al., 2000a) has
26 compared the effects of inhaled ROFA (at 1 mg/m3) to concentrated ambient PM (four
27 experiments, at mean concentrations of 475 to 900 |ig/m3) in normal and SO2-induced bronchitic
28 rats. A statistically significant increase in at least one lung injury marker was seen in bronchitic
29 rats with only one out of four of the concentrated ambient exposures; whereas inhaled ROFA
30 had no effect, even though the content of soluble iron, vanadium, and nickel was much higher in
31 the ROFA sample than in the concentrated ambient PM.
June 2003 7-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 Nevertheless, particularly interesting new findings point toward ambient PM exacerbation
2 of allergic airway hyperresponsiveness and/or antigen-induced immune responses. Both metal
3 and diesel particles have been implicated with an expanding array of new studies showing DPM
4 in particular as being effective in exacerbating allergic asthmatic responses.
5
6 7.7.1.4 Diesel Exhaust Particles
7 As described in Section 7.5.3, there is growing toxicological evidence that diesel PM
8 exacerbates the allergic response to inhaled antigens. The organic fraction of diesel exhaust has
9 been linked to eosinophil degranulation and induction of cytokine production, suggesting that the
10 organic constituents of diesel PM are the responsible part for the immune effects. It is not
11 known whether the adjuvant-like activity of diesel PM is unique or whether other combustion
12 particles have similar effects. It is important to compare the immune effects of other source-
13 specific emissions, as well as concentrated ambient PM, to diesel PM to determine the extent to
14 which exposure to diesel exhaust may contribute to the incidence and severity of allergic rhinitis
15 and asthma.
16
17 7.7.1.5 Organic Compounds
18 Published research on the acute effects of particle-associated organic carbon constituents is
19 conspicuous by its relative absence, except for diesel exhaust particles. Like metals, organics are
20 common constituents of combustion-generated particles and have been found in ambient PM
21 samples over a wide geographical range. Organic carbon constituents comprise a substantial
22 portion of the mass of ambient PM (10 to 60% of the total dry mass [Turpin, 1999]). The
23 organic fraction of ambient PM has been evaluated for its mutagenic effects. Although the
24 organic fraction of ambient PM is a poorly characterized heterogeneous mixture of an unknown
25 number of different compounds, organic compounds remain a potential causal property for PM
26 health effects due to the contribution of diesel exhaust particles to the fine PM fraction (Dreher,
27 2000). Strategies have been proposed for examining the health effects of this potentially
28 important constituent (Turpin, 1999).
29
30 7.7.1.6 Ultrafine Particles
31 When this subject was reviewed in the 1996 PM AQCD (U. S. Environmental Protection
32 Agency, 1996a), it was not known whether the pulmonary toxicity of freshly generated ultrafme
June 2003 7-117 DRAFT-DO NOT QUOTE OR CITE
-------
1 polytetraflouethylene (PTFE; teflon) particles was due to particle size or a result of adsorbed
2 fumes. Subsequent studies with other ultrafine particles have demonstrated a significantly
3 greater inflammatory response than that seen with fine particles of the same chemical
4 composition at similar mass doses (Oberdorster et al., 1992; Li et al., 1996, 1997, 1999). In
5 other more limited studies, ultrafmes also have generated greater oxidative stress in experimental
6 animals. Inhalation exposure of normal rats to ultrafine carbon particles generated by electric
7 arc discharge (100 |ig/m3 for 6 h) caused minimal lung inflammation per unit mass (Elder et al.,
8 2000a,b), compared to ultrafine PTFE or metal particles. On the other hand, instillation of
9 125 jig of ultrafine carbon black (20 nm) caused substantially more inflammation per unit mass
10 than did the same dose of fine particles of carbon black (200 to 250 nm), suggesting that
11 ultrafine particles may cause more inflammation per unit mass than larger particles (Li et al.,
12 1997). However, the chemical constituents of the two sizes of carbon black used in this study
13 were not analyzed, and it cannot be assumed that the chemical composition was the same for the
14 two sizes since composition may vary with particle size. Further, when the particle surface area
15 is used a dosimetric, the inflammatory response to both fine and ultrafine particles may be
16 basically the same (Oberdorster, 1996b, 2000; Li et al., 1996). Thus, there is still insufficient
17 toxicological evidence to conclude that ambient concentrations of ultrafine particles contribute to
18 the health effects of particulate air pollution. With acid aerosols, studies of low concentrations
19 of ultrafine sulfuric acid and metal oxide particles have demonstrated effects in the lung.
20 However, it is possible that inhaled ultrafine particles may have systemic effects that are
21 independent of effects on the lung.
22
23 7.7.1.7 Concentrated Ambient Particle Studies
24 Concentrated ambient particle (CAPS) studies should be among the most relevant in
25 helping to understand the characteristics of PM producing toxicity, susceptibility of individuals
26 to PM, and the underlying mechanisms. Studies have used collected urban PM for intratracheal
27 administration to healthy and compromised animals. Despite the difficulties in extrapolating
28 from the bolus delivery used in such studies, they have provided strong evidence that the
29 chemical composition of ambient particles can have a major influence on toxicity. More recent
30 work with inhaled concentrated ambient PM has observed cardiopulmonary changes in rodents
31 and dogs at high concentrations of fine PM. No comparative studies to examine the effects of
32 ultrafine and coarse ambient PM have been done, although a new ambient particle concentrator
June 2003 7-118 DRAFT-DO NOT QUOTE OR CITE
-------
1 developed by Sioutas and colleagues should permit the direct toxicological comparison of
2 various ambient particle sizes. Importantly, it has become evident that, although the
3 concentrated ambient PM studies can provide important dose-response information, identify
4 susceptibility factors in animal models, and permit examination of mechanisms related to PM
5 toxicity, they are not particularly well suited for the identification of toxic components in urban
6 PM. Because only a limited number of exposures using concentrated ambient PM can be
7 reasonably conducted by a given laboratory in a particular urban environment, there may be
8 insufficient information to conduct a factor analysis on an exposure/response matrix. This may
9 also hinder principal component analysis techniques that are useful in identifying particle
10 components responsible for adverse outcomes. New particle concentrator systems now coming
11 on-line at the U.S. EPA and elsewhere that permit selective concentration of ultrafine, fine, and
12 thoracic coarse PM hold promise for enhanced understanding of PM characteristics producing
13 toxicity.
14
15 7.7.1.8 Bioaerosols
16 Recent studies support the conclusion of the 1996 PM AQCD (U. S. Environmental
17 Protection Agency, 1996a), which stated that bioaerosols, at concentrations present in the
18 ambient environment, would not account for the reported health effects of ambient PM.
19 However, it is possible that bioaerosols could contribute to the health effects of PM.
20 Dose-response studies in healthy volunteers exposed to 0.55 and 50 jig endotoxin, by the
21 inhalation route, showed a threshold for pulmonary and systemic effects for endotoxin between
22 0.5 and 5.0 jig (Michel et al., 1997). Monn and Becker (1999) examined effects of size
23 fractionated outdoor PM on human monocytes and found cytokine induction characteristic of
24 endotoxin activity in the coarse-size fraction but not in the fine fraction. Available information
25 suggests that ambient concentrations of endotoxin are very low and do not exceed 0.5 ng/m3.
26 However, there are numerous bioaerosols present in the ambient air including pollens and
27 allergens. Their contribution to the potential health effects of PM are largely unknown.
28
29 7.7.2 Mechanisms of Action
30 The mechanisms that underlie biological responses to ambient PM are not yet clear.
31 Findings since 1996 have provided evidence supporting many hypotheses for PM effects; and
32 this body of evidence has grown substantially. Various toxicologic studies using PM having
June 2003 7-119 DRAFT-DO NOT QUOTE OR CITE
-------
1 diverse physicochemical characteristics have shown that these characteristics have a great impact
2 on the specific response that is observed. Thus, there are multiple biological mechanisms that
3 may be responsible for observed morbidity/mortality due to exposure to ambient PM, and these
4 mechanisms may be highly dependent on the type of particle in the exposure atmosphere.
5 It should be noted that many animal controlled-exposure studies used particle concentrations
6 much higher than those typically occurring in ambient air, whereas clinical concentrator studies
7 have shown responses at levels similar to and higher than those occurring in ambient air (e.g.,
8 Ohio et al., 2000a). It is not known if the mechanisms elicited are the same across exposure
9 levels. Clearly, controlled-exposure studies have not as yet been able to delineate fully particle
10 characteristics and the toxicological mechanisms by which ambient PM may affect biological
11 systems. Nevertheless, as discussed in preceding sections of this chapter, much progress has
12 been made since the 1996 PM AQCD in evaluating pathophysiological mechanisms involved in
13 PM-associated cardiovascular and respiratory health effects. Key findings derived from the
14 newly emerging toxicological evidence for these potential pathophysiological mechanisms are
15 summarized below.
16
17 7.7.2.1 Direct Pulmonary Effects
18 When the 1996 PM AQCD was written, the lung was thought to be the primary organ
19 affected by particulate air pollution. Although the lung still is a primary organ affected by PM
20 inhalation, there is growing toxicological and epidemiologic evidence that the cardiovascular
21 system is also affected and may be a co-primary organ system related to certain health endpoints
22 such as mortality. Nonetheless, understanding how particulate air pollution causes or
23 exacerbates respiratory disease remains an important goal. There is some toxicological evidence
24 for the following three hypothesized mechanisms for PM inducing direct pulmonary effects.
25
26 Particulate Air Pollution Causes Lung Injury and Inflammation
27 Particularly compelling evidence pointing towards ambient PM causing lung injury and
28 inflammation derives from the study of ambient PM materials on filter extracts collected from
29 community air monitors before, during the temporary closing of a steel mill in Utah Valley, and
30 after its reopening. Ohio and Devlin (2001) found that intratracheal instillation of filter extract
31 materials in human volunteers provoked greater lung inflammatory responses for materials
32 obtained before and after the temporary closing versus that collected during the plant closing.
June 2003 7-120 DRAFT-DO NOT QUOTE OR CITE
-------
1 The instilled dose of 500 jig of extract material was calculated by Ohio and Devlin to result in
2 focal lung deposition in the lingula roughly equivalent to 5 times more than would be deposited
3 if an active person experienced 24-h inhalation exposure to 100 |ig/m3 PM10 (during wintertime
4 temperature inversions in Utah Valley 24-h PM10 levels can exceed 100 |ig/m3). Moreover, 100
5 jig of filter extract collected during the winter before the temporary plant closure similarly
6 instilled into the lungs of human volunteers also increased levels of neutrophils, protein, and
7 inflammatory cytokines. Ohio and Devlin (2001) indicated that these results and calculations
8 suggest that biologic effects found in their study could be experienced during a typical winter
9 inversion in the Utah Valley.
10 Further, the instillation in rats (Dye et al., 2001) of extract materials from before and after
11 the plant closing resulted in a 50% increase in air way hyperresponsiveness to acetylcholine
12 compared to 17 or 25% increases with saline or extract materials for the period when the plant
13 was closed, respectively. Analysis of the extract materials revealed notably greater quantities of
14 metals for when the plant was opened suggesting that such metals (e.g., Cu, Zn, Fe, Pb, As, Mn,
15 Ni) may be important contributors to the pulmonary toxicity observed in the controlled exposure
16 studies, as well as to health effects shown epidemiologically to vary with PM exposures of Utah
17 Valley residents before, during, and after the steel mill closing.
18 Still other toxicological studies point towards lung injury and inflammation being
19 associated with exposure of lung tissue to complex combustion-related PM materials, with
20 metals again being likely contributors. For example, in the last few years, numerous studies
21 have shown that instilled and inhaled ROFA, a product of fossil fuel combustion, can cause
22 substantial lung injury and inflammation. The toxic effects of ROFA are largely caused by its
23 high content of soluble metals, and some of the pulmonary effects of ROFA can be reproduced
24 by equivalent exposures to soluble metal salts. In contrast, controlled exposures of animals to
25 sulfuric acid aerosols, acid-coated carbon, and sulfate salts cause little lung injury or
26 inflammation, even at high concentrations. Inhalation of concentrated ambient PM (which
27 contains only small amounts of metals) by laboratory animals at concentrations in the range of
28 100 to 1000 |ig/m3 have been shown in some (but not all) studies to cause mild pulmonary injury
29 and inflammation. Rats with SO2-induced bronchitis and monocrotaline-treated rats have been
30 reported to have a greater inflammatory response to concentrated ambient PM than normal rats.
31 These studies suggest that exacerbation of respiratory disease by ambient PM may be caused in
32 part by lung injury and inflammation.
June 2003 7-121 DRAFT-DO NOT QUOTE OR CITE
-------
1 Particulate Air Pollution Causes Increased Susceptibility to Respiratory Infections
2 Antonini et al. (2002) investigated the effect of preexposure to ROFA on lung defenses and
3 injury after pulmonary challenge with Listeria monocytogenes, a bacterial pathogen. Male
4 Sprague-Dawley rats were dosed IT at day 0 with saline (control) or ROFA (0.2 or 1 mg/100 g
5 body weight). Three days later, both groups of rats were instilled IT with a low (5 x 103) or high
6 (5 x O5) dose of L. monocytogenes. Chemiluminescence (CL) and nitric oxide (NO) production,
7 two indices of alveolar macrophage (AM) function, were measured on cells recovered from the
8 right lungs by bronchoalveolar lavage. The left lungs and spleens were homogenized, cultured,
9 and colony-forming units were counted after overnight incubation. Exposure to ROFA and the
10 high dose of L. monocytogenes led to marked lung injury and inflammation as well as to an
11 increase in mortality, compared with rats treated with saline and the high dose of
12 L. monocytogenes. Preexposure to ROFA significantly enhanced injury and delayed the
13 pulmonary clearance of L. monocytogenes at both bacterial doses when compared to the saline-
14 treated control rats. ROFA had no effect on AM CL but caused a significant suppression of AM
15 NO production. The authors concluded that acute exposure to ROFA slowed pulmonary
16 clearance of L. monocytogenes and altered AM function. They postulated that these changes
17 could lead to increased susceptibility to lung infection in exposed populations.
18 Ohtsuka et al. (2000a,b) have also shown that a single 4 h exposure of mice to acid-coated
19 carbon particles at a mass concentration of 10,000 |ig/m3 carbon black causes decreased
20 phagocytic activity of alveolar macrophages, even in the absence of lung injury.
21
22 Particulate Air Pollution Increases Airway Reactivity and Exacerbates Asthma
23 The strongest evidence supporting this hypothesis is from studies on diesel particulate
24 matter (DPM). Diesel particulate matter has been shown to increase production of antigen-
25 specific IgE in mice and humans (summarized in Section 7.2.1.2). In vitro studies have
26 suggested that the organic fraction of DPM is involved in the increased IgE production. ROFA
27 leachate also has been shown to enhance antigen-specific airway reactivity in mice (Goldsmith
28 et al., 1999), indicating that soluble metals can also enhance an allergic response. However, in
29 this same study, exposure of mice to concentrated ambient PM did not affect antigen-specific
30 airway reactivity. It is premature to conclude from the Goldsmith experiment that concentrated
31 ambient PM does not exacerbate allergic airways disease because the chemical composition of
June 2003 7-122 DRAFT-DO NOT QUOTE OR CITE
-------
1 the PM (as indicated by studies with DPM and ROFA) may be more important than the mass
2 concentration.
3
4 7.7.2.2 Systemic Effects Secondary to Lung Injury
5 When the 1996 PM AQCD was written, it was thought that cardiovascular-related
6 morbidity and mortality most likely would be secondary to impairment of oxygenation or some
7 other consequence of lung injury and inflammation. Newly available toxicologic studies provide
8 some additional evidence regarding such possibilities.
9
10 Lung Injury from Inhaled Particulate Matter Causes Impairment of Oxygenation and
11 Increased Work of Breathing That Adversely Affects the Heart
12 Instillation of ROFA (0, 0.25, 1.0, 2.5 mg) has been shown to cause a 50% mortality rate in
13 monocrotaline-treated rats (Watkinson et al., 2000a,b). Although blood oxygen levels were not
14 measured in this study, there were ECG abnormalities consistent with severe hypoxemia in about
15 half of the rats that subsequently died. Given the severe inflammatory effects of instilled ROFA
16 and the fact that monocrotaline-treated rats have increased lung permeability as well as
17 pulmonary hypertension, it is plausible that instilled ROFA can cause severe hypoxemia leading
18 to death in this rat model. Results from studies in which animals (normal and compromised)
19 were exposed to concentrated ambient PM (at concentrations many times higher than would be
20 encountered in the United States) indicate that ambient PM is unlikely to cause severe
21 disturbances in oxygenation or pulmonary function. However, even a modest decrease in
22 oxygenation can have serious consequences in individuals with ischemic heart disease.
23 Kleinman et al. (1998) has shown that a reduction in arterial blood saturation from 98 to 94% by
24 either mild hypoxia or by exposure to 100 ppm CO significantly reduced the time to onset of
25 angina in exercising volunteers. Thus, information is needed on the effects of PM on arterial
26 blood gases and pulmonary function to fully address the above hypothesis.
27
28 Lung Inflammation and Cytokine Production Cause Adverse Systemic Hemodynamic Effects
29 It has been suggested that systemic effects of particulate air pollution may result from
30 activation of cytokine production in the lung (Li et al., 1997). In support of this idea,
31 monocrotaline-treated rats exposed to inhaled ROFA (15,000 |ig/m3, 6 h/day for 3 days) showed
32 increased pulmonary cytokine gene expression, bradycardia, hypothermia, and increased
June 2003 7-123 DRAFT-DO NOT QUOTE OR CITE
-------
1 arrhythmias (Watkinson et al., 2000a,b). However, spontaneously hypertensive rats had a
2 similar cardiovascular response to inhaled ROFA (except that they also developed ST segment
3 depression) with no increase in pulmonary cytokine gene expression. Studies in dogs exposed to
4 concentrated ambient PM (322 |ig/m3, MMAD = 0.23-.034 jim) showed minimal pulmonary
5 inflammation and no positive staining for IL-8, IL-1, or TNF in airway biopsies. However, there
6 was a significant decrease in the time of onset of ischemic ECG changes following coronary
7 artery occlusion in PM-exposed dogs compared to controls (Godleski et al., 2000). Thus, the
8 link between changes in the production of cytokines in the lung and cardiovascular function is
9 not clear-cut, and basic information on the effects of mild pulmonary injury on cardiovascular
10 function is needed to understand the mechanisms by which inhaled PM affects the heart. In this
11 regard, Wellenius et al. (2002) have developed and tested a model for investigating the effects of
12 inhaled PM on arrhythmias and heart rate variability (HRV) in rats with acute myocardial
13 infarction. Left-ventricular MI was induced in Sprague-Dawley rats by thermocoagulation of the
14 left coronary artery. Diazepam-sedated rats were exposed (1 h) to residual oil fly ash (ROFA),
15 carbon black, or room air at 12-18 h after surgery. Each exposure was immediately preceded
16 and followed by a 1-h exposure to room air (baseline and recovery periods, respectively). Lead-
17 II electrocardiograms were recorded. In the MI group, 41% of rats exhibited one or more
18 premature ventricular complexes (PVCs) during the baseline period. Exposure to ROFA, but not
19 to carbon black or room air, increased arrhythmia frequency in animals with preexisting PVCs.
20 Furthermore, MI rats exposed to ROFA, but not to carbon black or room air, decreased HRV.
21 There was no difference in arrhythmia frequency or HRV among sham-operated animals. The
22 authors concluded that this model may be useful for elucidating the physiologic mechanisms of
23 particle-induced cardiovascular arrhythmias and contribute to defining the specific constituents
24 of ambient particles responsible for arrhythmias.
25
26 Lung Inflammation from Inhaled Particulate Matter Causes Increased Blood Coagulability
27 That Increases the Risk of Heart Attacks and Strokes
28 There is abundant evidence linking risk of heart attacks and strokes to small prothrombotic
29 changes in the blood coagulation system. However, the published toxicological evidence that
30 moderate lung inflammation causes increased blood coagulability is inconsistent. Ohio et al.
31 (2000a) have shown that inhalation of concentrated ambient PM in healthy nonsmokers causes
32 increased levels of blood fibrinogen. Gardner et al. (2000) have shown that a high dose
June 2003 7-124 DRAFT-DO NOT QUOTE OR CITE
-------
1 (8,300 |ig/kg) of instilled ROFA in rats causes increased levels of fibrinogen, but no effect was
2 seen at lower doses. Exposure of dogs to concentrated ambient PM had no effect on fibrinogen
3 levels (Godleski et al., 2000). The coagulation system is as multifaceted and complex as the
4 immune system, and there are many other sensitive and clinically significant parameters that
5 should be examined in addition to fibrinogen. Thus, it is premature to draw any conclusions
6 about the relationship between PM and blood coagulation.
7
8 Interaction of Particulate Matter with the Lung Affects Hematopoiesis
9 Terashima et al. (1997) found that instillation of fine carbon particles (20,000 jig/rabbit)
10 stimulated release of PMNs from bone marrow. In further support of this hypothesis, Gordon
11 and colleagues reported that the percentage of PMNs in the peripheral blood increased in rats
12 exposed to ambient PM in some but not all exposures. On the other hand, Godleski et al. (2000)
13 found no changes in peripheral blood counts of dogs exposed to concentrated ambient PM.
14 Thus, consistent evidence that PM ambient concentrations can affect hematopoiesis remains to
15 be demonstrated.
16
17 7.7.2.3 Direct Effects on the Heart
18 Changes in heart rate, heart rate variability, and conductance associated with ambient PM
19 exposure have been reported in animal studies (Godleski et al., 2000; Gordon et al., 2000;
20 Watkinson et al., 2000a,b; Campen et al., 2000), in several human panel studies (described in
21 Chapter 8), and in a reanalysis of data from the MONICA study (Peters et al., 1997). Some of
22 these studies included endpoints related to respiratory effects but few significant adverse
23 respiratory changes were detected. This raises the possibility that ambient PM may have effects
24 on the heart that are independent of adverse changes in the lung. There is certainly precedent for
25 this idea. For example, tobacco smoke (which is a mixture of combustion-generated gases and
26 PM) causes cardiovascular disease by mechanisms that are independent of its effect on the lung.
27 Two types of hypothesized direct effects of PM on the heart are noted below.
28
29 Inhaled Particulate Matter Affects the Heart by Uptake of Particles into the Circulation
30 or Release of a Soluble Substances into the Circulation
31 Drugs can be rapidly and efficiently delivered to the systemic circulation by inhalation.
32 This implies that the pulmonary vasculature absorbs inhaled materials, including charged
June 2003 7-125 DRAFT-DO NOT QUOTE OR CITE
-------
1 substances such as small proteins and peptides. Nemmar et al. (2001) studied the movement of
2 radioactively labeled ultrafine particles out of the lungs of hamsters receiving a single IT
3 instillation of albumin nanocolloid particles (<80 nm) labeled with 99mTc and killed after 5, 15,
4 30, and 60 min. Blood radioactivity, at 5, 15, 30, and 60 min, respectively, expressed as
5 percentage of total body radioactivity per gram blood, was 2.88 ± 0.80%, 1.30 ± 0.17%, 1.52 ±
6 0.46%, and 0.21 ± 0.06%. Liver radioactivity, at 5, 15, 30, and 60 min, respectively, expressed
7 as percentage of total radioactivity per organ, was 0.10 ± 0.07%, 0.23 ± 0.06%, 1.24 ± 0.27%,
8 and 0.06 ± 0.02%. Lower values were observed in the heart, spleen, kidneys, and brain. Dose
9 dependence was assessed at 30 min following instillation of 10 jig and 1 jig 99mTc-albumin per
10 animal (n = 3 at each dose), and values of the same relative magnitudes as after instillation of
11 100 jig were obtained. The authors concluded that a significant fraction of ultrafine 99mTc-
12 albumin diffuses rapidly from the lungs into the systemic circulation.
13 Nemmar et al. (2002) investigated the extent inhaled particles entered into the systemic
14 circulation, in 5 healthy volunteers, after inhaling "Technegas," an aerosol consisting mainly of
15 ultrafine 99mTc -labeled carbon particles (< 100 nm). Radioactivity detected in blood at 1 minute,
16 reached a maximum between 10 and 20 minutes, and remained at this level up to 60 minutes.
17 Thin layer chromatography of blood showed that in addition to a species corresponding to
18 oxidized 99mTc (i.e., pertechnetate) there was also a species corresponding to particle-bound
19 99mjc Qamma camera images showed substantial radioactivity over the liver and other areas of
20 the body. These workers conclude that inhaled 99mTc-labeled ultrafine carbon particles pass
21 rapidly into the systemic circulation.
22
23 Inhaled Particulate Matter Affects Autonomic Control of the Heart and
24 Cardiovascular System
25 There is growing evidence for this idea as described above. This raises the question of
26 how inhaled particles could affect the autonomic nervous system. Activation of neural receptors
27 in the lung is a logical area to investigate. Studies in conscious rats have shown that inhalation
28 of wood smoke causes marked changes in sympathetic and parasympathetic input to the
29 cardiovascular system that are mediated by neural reflexes (Nakamura and Hayashida, 1992).
30 Although research on airway neural receptors and neural-mediated reflexes is a well established
31 discipline, the cardiovascular effects of stimulating airway receptors continue to receive less
32 attention than the pulmonary effects. Previous studies of airway reflex-mediated cardiac effects
June 2003 7-126 DRAFT-DO NOT QUOTE OR CITE
-------
1 usually employed very high doses of chemical irritants, and the results may not be applicable to
2 air pollutants. There is a need for basic physiological studies to examine effects on
3 cardiovascular system when airway and alveolar neural receptors are stimulated in a manner
4 relevant to air pollutants.
5
6 7.7.3 Susceptibility
7 Progress has been made in understanding the role of individual susceptibility to ambient
8 PM effects. Studies have consistently shown that older animals or animals with certain types of
9 compromised health, either genetic or induced, are more susceptible to instilled or inhaled
10 particles, although the increased animal-to-animal variability in these models has created greater
11 uncertainty for the interpretation of the findings (Clarke et al., 1999, 2000; Kodavanti et al.,
12 1998, 2000b, 2001; Gordon et al., 2000; Ohtsuka et al., 2000c; Wesselkamper et al., 2000;
13 Leikauf et al., 2000; Saldiva et al., 2002). Moreover, because PM seems to affect broad
14 categories of disease states, ranging from cardiac arrhythmias to pulmonary infection, it can be
15 difficult to know what disease models to use in evaluating the biological plausibility of adverse
16 health effects of PM.
17 Nevertheless, particularly interesting new findings point toward ambient PM exacerbation
18 of allergic airway hyperresponsiveness and/or antigen-induced immune responses. Both metals
19 and diesel particles have been implicated, with an expanding array of new studies showing DPM
20 in particular as being effective in exacerbating allergic asthma responses (Takano et al., 1997;
21 Nel et al., 2001; Van Zijverden et al., 2000, 2001; Walters et al., 2001; Nordenhall et al., 2001;
22 Hamada et al., 1999, 2000; Lambert et al., 1999; Gilmour et al., 2001).
23
24 7.7.4 PM Interactions with Gaseous Co-Pollutants
25 Several new studies have examined possible cardiopulmonary effects of complex air
26 pollution mixtures in Mexico, Spain, and Italy. These studies, taking advantage of differences in
27 pollutant mixtures and concentrations in relatively "clean" rural areas versus urban environments
28 found morphological changes in the nasopharynx (Calderon-Garciduefias et al., 2001c), the
29 lower respiratory tract (Gulisano et al., 1997; Lorz and Lopez, 1997; Calderon-Garciduefias
30 et al., 2001c) and in the heart (Calderon-Garciduefias et al., 2001c) of lambs, pigeons, and dogs,
31 respectively, experiencing long-term continuous natural exposures to elevated ambient air
32 pollution. Each study provided evidence suggesting that animals living in urban environments
June 2003 7-127 DRAFT-DO NOT QUOTE OR CITE
-------
1 with higher air pollution levels have greater pulmonary and cardiac changes than those living in
2 cleaner rural areas. It is difficult, however, to (a) assign relative specific roles to PM or other
3 components of the urban air mixtures in producing the observed effects or (b) extrapolate the
4 findings to U.S. urban situations having typically much lower air pollutant concentrations (e.g.
5 especially the case for notably higher PM and O3 levels observed in Mexico City than in U.S.
6 cities).
7 Two well-conducted new controlled human exposure studies do provide somewhat more
8 readily interpretable results. In one, a randomized double-blind crossover study by Brook et al.
9 (2002) observed increased brachial artery constriction in adult males and females, mean age
10 = 34.9 yr ± 10 SD, exposed for 2 hr to filtered ambient air containing 150 |ig/m3 CAPS and
11 120 ppb O3 while at rest. Another study, by Linn et al. (1997) found a positive association
12 between acid concentration and respiratory symptoms (but not spirometry) among asthmatic
13 children following a single 4-hr exposure to 60 to 140 |ig/m3 H2SO4, 0.1 ppm SO2, and 0.1 ppm
14 O3 while undergoing intermittent exercise. No changes were seen among healthy children.
15
16
June 2003 7-128 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 Alink, G. M; Sjogren, M; Bos, R. P.; Doekes, G.; Kromhout, H.; Scheepers, P. T. J. (1998) Effect of airborne
5 particles from selected indoor and outdoor environments on gap-junctional intercellular communication.
6 Toxicol. Lett. 96/97: 209-213.
7 Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects.
8 In: Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC. Environ.
9 Health Perspect. 79: 147-150.
10 Amdur, M. O.; Dubriel, M.; Creasia, D. A. (1978) Respiratory response of guinea pigs to low levels of sulfuric acid.
11 Environ. Res. 15: 418-423.
12 Ball, B. R.; Smith, K. R.; Veranth, J. M.; Aust, A. E. (2000) Bioavailability of iron from coal fly ash: mechanisms of
13 mobilization and of biological effects. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
14 Toxicol. 12(suppl. 4): 209-225.
15 Bayram, H.; Devalia, J. L.; Khair, O. A.; Abdelaziz, M. M.; Sapsford, R. J.; Sagai, M.; Davies, R. J. (1998a)
16 Comparison of ciliary activity and inflammatory mediator release from bronchial epithelial cells of nonatopic
17 nonasthmatic subjects and atopic asthmatic patients and the effect of diesel exhaust particles in vitro.
18 J. Allergy Clin. Immunol. 102: 771-782.
19 Bayram, H.; Devalia, J. L.; Sapsford, R. J.; Ohtoshi, T.; Miyabara, Y.; Sagai, M.; Davies, R. J. (1998b) The effect of
20 diesel exhaust particles on cell function and release of inflammatory mediators from human bronchial
21 epithelial cells in vitro. Am. J. Respir. Cell Mol. Biol. 18: 441-448.
22 Becker, S.; Soukup, J. M. (1998) Decreased GDI IB expression, phagocytosis, and oxidative burst in urban
23 paniculate pollution-exposed human monocytes and alveolar macrophages. J. Toxicol. Environ. Health Part A
24 55:455-477.
25 Becker, S.; Soukup, J. M.; Gilmour, M. L; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
26 by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
27 Pharmacol. 141:637-648.
28 Blomberg, A.; Sainsbury, C.; Rudell, B.; Frew, A. J.; Holgate, S. T.; Sandstrom, T.; Kelly, F. J. (1998) Nasal cavity
29 lining fluid ascorbic acid concentration increases in healthy human volunteers following short term exposure
30 to diesel exhaust. Free Radical Res. 28: 59-67.
31 Bolarin, D. M.; Bhalla, D. K.; Kleinman, M. T. (1997) Effects of repeated exposures of geriatric rats to ozone and
32 particle-containing atmospheres: an analysis of bronchoalveolar lavage and plasma proteins. Inhalation
33 Toxicol. 9: 423-434.
34 Bonner, J. C.; Rice, A. B.; Lindroos, P. M.; O'Brien, P. O.; Dreher, K. L.; Rosas, L; Alfaro-Moreno, E.;
35 Osornio-Vargas, A. R. (1998) Induction of the lung myofibroblast PDGF receptor system by urban ambient
36 particles from Mexico City. Am. J. Respir. Cell Mol. Biol. 19: 672-680.
37 Bouthillier, L.; Vincent, R.; Goegan, P.; Adamson, I. Y. R.; Bjarnason, S.; Stewart, M.; Guenette, J.; Potvin, M.;
3 8 Kumarathasan, P. (1998) Acute effects of inhaled urban particles and ozone: lung morphology, macrophage
39 activity, and plasma endothelin-1. Am. J. Pathol. 153: 1873-1884.
40 Brain, J. D.; Long, N. C.; Wolfthal, S. F.; Dumyahn, T.; Dockery, D. W. (1998) Pulmonary toxicity in hamsters of
41 smoke particles from Kuwaiti oil fires. Environ. Health Perspect. 106: 141-146.
42 Brightwell, J.; Fouillet, X.; Cassano-Zoppi, A.-L.; Gatz, R.; Duchosal, F. (1986) Neoplastic and functional changes
43 in rodents after chronic inhalation of engine exhaust emissions. In: Ishinishi, N.; Koizumi, A.; McClellan,
44 R. O.; Stober, W., eds. Carcinogenic and mutagenic effects of diesel engine exhaust: proceedings of the
45 international satellite symposium on lexicological effects of emissions from diesel engines; July; Tsukuba
46 Science City, Japan. Amsterdam, Holland: Elsevier Science Publishers B. V.; pp. 471-485. (Developments in
47 toxicology and environmental science: v. 13).
48 Broeckaert, F.; Buchet, J. P.; Huaux, F.; Lardot, C.; Lison, D.; Yager, J. W. (1997) Reduction of the ex vivo
49 production of tumor necrosis factor alpha by alveolar phagocytes after administration of coal fly ash and
50 copper smelter dust. J. Toxicol. Environ. Health 51: 189-202.
51 Brook, R. D.; Brook, J. R.; Urch, B.; Vincent, R.; Rajagopalan, S.; Silverman, F. (2002) Inhalation of fine paniculate
52 air pollution and ozone causes acute arterial vasoconstriction in health adults. Circulation 105: 1534-1536.
53 Btinger, J.; Krahl, J.; Baum, K.; Schroder, O.; Mtiller, M.; Westphal, G.; Ruhnau, P.; Schulz, T. G.; Hairier, E.
54 (2000) Cytotoxic and mutagenic effects, particle size and concentration analysis of diesel engine emissions
55 using biodiesel and petrol diesel as fuel. Arch. Toxicol. 74: 490-498.
June 2003 7-129 DRAFT-DO NOT QUOTE OR CITE
-------
1 Calderon-Garciduenas, L.; Osorno-Velaquez, A.; Bravo-Alvarez, H.; Delgado-Chavez, R.; Barrios-Marquez, R.
2 (1992) Histopathologic changes of the nasal mucosa in southwest metropolitan Mexico City inhabitants.
3 Am. J. Pathol. 140: 225-232.
4 Calderon-Garciduenas, L.; Rodriguez-Alcaraz, A.; Garcia, R.; Sanchez, G.; Barragan, G.; Camacho, R.; Ramirez, L.
5 (1994) Human nasal mucosal changes after exposure to urban pollution. Environ. Health Perspect.
6 102: 1074-1080.
7 Calderon-Garciduenas, L.; Osnaya-Brizuela, N.; Ramirez-Martinez, L.; Villarreal-Calderon, A. (1996) DNA strand
8 breaks in human nasal respiratory epithelium are induced upon exposure to urban pollution. Environ. Health
9 Perspect. 104: 160-168.
10 Calderon-Garciduenas, L.; Devlin, R. B.; Miller, F. J. (2000a) Respiratory tract pathology and cytokine imbalance in
11 clinically healthy children chronically and sequentially exposed to air pollutants. Med. Hypotheses
12 55: 373-378.
13 Calderon-Garciduenas, L.; Mora-Tascareno, A.; Chung, C. J.; Valencia, G.; Fordham, L. A.; Garcia, R.; Osnaya, N.;
14 Romero, L.; Acuna, H.; Villarreal-Calderon, A.; Devlin, R. B.; Koren, H. S. (2000b) Exposure to air pollution
15 is associated with lung hyperinflation in healthy children and adolescents in southwest Mexico City: a pilot
16 study. Inhalation Toxicol. 12: 537-561.
17 Calderon-Garciduenas, L.; Valencia-Salazar, G.; Rodriguez-Alcaraz, A.; Gambling, T. M.; Garcia, R.; Osnaya, N.;
18 Villarreal-Calderon, A.; Devlin, R. B.; Carson, J. L. (2001a) Ultrastructural nasal pathology in children
19 chronically and sequentially exposed to air pollutants. Am. J. Respir. Cell Mol. Biol. 24: 132-138.
20 Calderon-Garciduenas, L.; Rodriguez-Alcaraz, A.; Valencia-Salazar, G.; Mora-Tascareno, A.; Garcia, R.; Osnaya,
21 N.; Villarreal-Calderon, A.; Devlin, R. B.; Van Dyke, T. L. (200Ib) Nasal biopsies of children exposed to air
22 pollutants. Toxicol. Pathol. 29: 558-564.
23 Calderon-Garciduenas, L.; Mora-Tiscareno, A.; Fordham, L. A.; Chung, C. J.; Garcia, R.; Osnaya, N.; Hernandez,
24 J.; Acuna, H.; Gambling, T. M.; Villarreal-Calderon, A.; Carson, J.; Koren, H. S.; Devlin, R. B. (200Ic)
25 Canines as sentinel species for assessing chronic exposures to air pollutants: part 1. respiratory pathology.
26 Toxicol. Sci. 61:342-355.
27 Calderon-Garciduenas, L.; Gambling, T. M.; Acuna, H.; Garcia, R.; Osnaya, N.; Monroy, S.; Villarreal-Calderon,
28 A.; Carson, J.; Koren, H. S.; Devlin, R. B. (2001d) Canines as sentinel species for assessing chronic
29 exposures to air pollutants: part 2. cardiac pathology. Toxicol. Sci. 61: 356-367.
30 California Environmental Protection Agency. (1998) Report to the Air Resources Board on the proposed
31 identification of diesel exhaust as a toxic air contaminant. Part A: exposure assessment. Sacramento, CA:
32 Air Resources Board. Available: http://www.arb.ca.gov/regact/diesltac/diesltac.htm [16 May, 2003].
33 Campen, M. J.; Costa, D. L.; Watkinson, W. P. (2000) Cardiac and thermoregulatory toxicity of residual oil fly ash
34 in cardiopulmonary-compromised rats. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the third
3 5 colloquium on paniculate air pollution and human health (second special issue); June, 1999; Durham, NC.
36 Inhalation Toxicol. 12(suppl. 2): 7-22.
37 Campen, M. J.; Nolan, J. P.; Schladweiler, M. C. J.; Kodavanti, U. P.; Evansky, P. A.; Costa, D. L.; Watkinson,
38 W. P. (2001) Cardiovascular and thermoregulatory effects of inhaled PM-associated transition metals:
39 a potential interaction between nickel and vanadium sulfate. Toxicol. Sci. 64: 243-252.
40 Campen, M. J.; Nolan, J. P.; Schladweiler, M. C. J.; Kodavanti, U. P.; Costa, D. L.; Watkinson, W. P. (2002)
41 Cardiac and thermoregulatory effects of instilled paniculate matter-associated transition metals in healthy and
42 cardiopulmonary-compromised rats. J. Toxicol. Environ. Health Part A 65:1615-1631.
43 Carter, J. D.; Ohio, A. J.; Samet, J. M.; Devlin, R. B. (1997) Cytokine production by human airway epithelial cells
44 after exposure to an air pollution particle is metal-dependent. Toxicol. Appl. Pharmacol. 146: 180-188.
45 Castellan, R. M.; Olenchock, S. A.; Hankinson, J. L.; Millner, P. D.; Cocke, J. B.; Bragg, C. K.; Perkins, H. H., Jr.;
46 Jacobs, R. R. (1984) Acute bronchoconstriction induced by cotton dust: dose-related responses to endotoxin
47 and other dust factors. Ann. Intern. Med. 102: 157-163.
48 Castellan, R. M.; Olenchock, S. A.; Kinsley, K. B.; Hankinson, J. L. (1987) Inhaled endotoxin and decreased
49 spirometric values: an exposure-response relation for cotton dust. N. Engl. J. Med. 317: 605-610.
50 Chang, M. C.; Sioutas, C.; Kim, S.; Gong, H., Jr.; Linn, W. S. (2000) Reduction of nitrate losses from filter and
51 impactor samplers by means of concentration enrichment. Atmos. Environ. 34: 85-98.
52 Chen, L. C.; Fine, J. M.; Qu, Q.-S.; Amdur, M. O.; Gordon, T. (1992) Effects of fine and ultrafine sulfuric acid
53 aerosols in guinea pigs: alterations in alveolar macrophage function and intracellular pH. Toxicol. Appl.
54 Pharmacol. 113: 109-117.
55 Chen, L. C.; Fang, C. P.; Qu, Q. S.; Fine, J. M.; Schlesinger, R. B. (1993) A novel system for the in vitro exposure
56 of pulmonary cells to acid sulfate aerosols. Fundam. Appl. Toxicol. 20: 170-176.
57 Churg, A.; Brauer, M.; Keeling, B. (1996) Ozone enhances the uptake of mineral particles by tracheobronchial
58 epithelial cells in organ culture. Am. J. Respir. Crit. Care Med. 153: 1230-1233.
June 2003 7-130 DRAFT-DO NOT QUOTE OR CITE
-------
1 Churg, A.; Stevens, B.; Wright, J. L. (1998) Comparison of the uptake of fine and ultrafine TiO2 in a tracheal explant
2 system. Am. J. Physiol. 274: L81-L86.
3 Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.; Sioutas, C.; Paulauskis, J.; Coull, B.;
4 Ferguson, S.; Godleski, J. J. (1999) Urban air paniculate inhalation alters pulmonary function and induces
5 pulmonary inflammation in a rodent model of chronic bronchitis. Inhalation Toxicol. 11: 637-656.
6 Clarke, R. W.; Coull, B.; Reinisch, U.; Catalano, P.; Killingsworth, C. R.; Koutrakis, P.; Kavouras, L; Murthy,
7 G. G. K.; Lawrence, J.; Lovett, E.; Wolfson, J. M; Verrier, R. L.; Godleski, J. J. (2000a) Inhaled concentrated
8 ambient particles are associated with hematologic and bronchoalveolar lavage changes in canines. Environ.
9 Health Perspect. 108: 1179-1187.
10 Clarke, R. W.; Catalano, P.; Coull, B.; Koutrakis, P.; Krishna Murthy, G. G.; Rice, T.; Godleski, J. J. (2000b)
11 Age-related responses in rats to concentrated urban air particles (CAPs). In: Phalen, R. F., ed. Inhalation
12 Toxicology: proceedings of the third colloquium on paniculate air pollution and human health (first special
13 issue); June, 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 73-84.
14 Clarke, R. W.; Antonini, J. M.; Hemenway, D. R.; Frank, R.; Kleeberger, S. R.; Jakab, G. J. (2000c) Inhaled
15 particle-bound sulfate: effects on pulmonary inflammatory responses and alveolar macrophage function.
16 Inhalation Toxicol. 12: 169-186.
17 Cohen, M. D.; Becker, S.; Devlin, R.; Schlesinger, R. B.; Zelikoff, J. T. (1997) Effects of vanadium upon
18 polyl:C-induced responses in rat lung and alveolar macrophages. J. Toxicol. Environ. Health 51: 591-608.
19 Cormier, Y.; Laviolette, M.; Bedard, G.; Dosman, J.; Israel-Assayag, E. (1998) Effect of route of breathing on
20 response to exposure in a swine confinement building. Am. J. Respir. Crit. Care Med. 157: 1512-1521.
21 Costa, D. L.; Dreher, K. L. (1997) Bioavailable transition metals in paniculate matter mediate cardiopulmonary
22 injury in healthy and compromised animal models. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of
23 the sixth international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
24 September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1053-1060.
25 Creutzenberg, O.; Bellmann, B.; Muhle, H.; Dasenbrock, C.; Morrow, P.; Mermelstein, R. (1998) Lung clearance
26 and retention of toner, TiO2, and crystalline silica, utilizing a tracer technique during chronic inhalation
27 exposure in Syrian Golden hamsters. Inhalation Toxicol. 10: 731-751.
28 Diaz-Sanchez, D.; Tsien, A.; Casillas, A.; Dotson, A. R.; Saxon, A. (1996) Enhanced nasal cytokine production in
29 human beings after in vivo challenge with diesel exhaust particles. J. Allergy Clin. Immunol. 98: 114-123.
30 Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997) Combined diesel exhaust paniculate and ragweed
31 allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
32 production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
33 Diociaiuti, M.; Balduzzi, M.; De Berardis, B.; Cattani, G.; Stacchini, G.; Ziemacki, G.; Marconi, A.; Paoletti, L.
34 (2001) The two PM25 (fine) andPM25-10 (coarse) fractions: evidence of different biological activity. Environ.
35 Res. A 86: 254-262.
36 Donaldson, K.; Brown, D. M.; Mitchell, C.; Dineva, M.; Beswick, P. H.; Gilmour, P.; MacNee, W. (1997) Free
37 radical activity of PM10: iron-mediated generation of hydroxyl radicals. In: Driscoll, K. E.; Oberdorster, G.,
3 8 eds. Proceedings of the sixth international meeting on the toxicology of natural and man-made fibrous and
39 non-fibrous particles; September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5):
40 1285-1289.
41 Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor a expression from alveolar
42 macrophages treated with urban air particles. Exp. Lung Res. 22: 577-592.
43 Dormans, J. A. M. A.; Steerenberg, P. A.; Arts, J. H. E.; Van Bree, L.; De Klerk, A.; Verlaan, A. P. J.; Bruijntjes,
44 J. P.; Beekhof, P.; Van Soolingen, D.; VanLoveren, H. (1999) Pathological and immunological effects of
45 respirable coal fly ash in male wistar rats. Inhalation Toxicol. 11:51-69.
46 Dreher, K. L. (2000) Paniculate matter physicochemistry and toxicology: in search of causality — a critical
47 perspective. Inhalation Toxicol. 12(suppl. 3): 45-57.
48 Dreher, K. L.; Jaskot, R. H.; Lehmann, J. R.; Richards, J. H.; McGee, J. K.; Ohio, A. J.; Costa, D. L. (1997)
49 Soluable transition metals mediate residual oil fly ash induced acute lung injury. J. Toxicol. Environ. Health
50 50:285-305.
51 Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
52 Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
53 limitations. Toxicol. Sci. 55: 24-35.
54 Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1997) Epithelial injury induced by exposure to residual oil
55 fly-ash particles: role of reactive oxygen species? Am. J. Respir. Cell Mol. Biol. 17: 625-633.
56 Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1999) Role of soluble metals in oil fly ash-induced airway
57 epithelial injury and cytokine gene expression. Am. J. Physiol. 277: L498-L510.
June 2003 7-131 DRAFT-DO NOT QUOTE OR CITE
-------
1 Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. I; Costa, D. L.
2 (2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
3 studies in Utah Valley residents. Environ. Health Perspect. Suppl. 109(3): 395-403.
4 Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000a) Endotoxin priming affects the lung
5 response to ultrafine particles and ozone in young and old rats. 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): 85-98.
8 Elder, A. C. P.; Gelein, R.; Finkelstein, J. N.; Cox, C.; Oberdorster, G. (2000b) Pulmonary inflammatory response to
9 inhaled ultrafine particles is modified by age, ozone exposure, and bacterial toxin. In: Grant, L. D., ed.
10 PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 227-246.
11 Fabiani, R.; Pampanella, L.; Minelli, A.; Mezzasoma, L; Vecchiarelli, A.; Morozzi, G. (1997) Effect of airborne
12 particulate extracts on monocyte oxidative metabolism. J. Environ. Pathol. Toxicol. Oncol. 16: 195-199.
13 Ferin, J.; Oberdorster, G.; Penney, D. P.; Soderholm, S. C.; Gelein, R.; Piper, H. C. (1990) Increased pulmonary
14 toxicity of ultrafine particles? I. Particle clearance, translocation, morphology. J. Aerosol Sci. 21: 381-384.
15 Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
16 Respir. Cell Mol. Biol. 6: 535-542.
17 Finkelstein, J. N.; Johnston, C.; Barrett, T.; Oberdorster, G. (1997) Particulate-cell interactions and pulmonary
18 cytokine expression. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
19 on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
20 NY. Environ. Health Perspect. Suppl. 105(5): 1179-1182.
21 Folinsbee, L. J.; Kim, C. S.; Kehrl, H. R.; Prah, J. D.; Devlin, R. B. (1997) Methods in human inhalation toxicology.
22 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press LLC; pp. 607-670.
23 Fortoul, T. L; Osorio, L. S.; Tovar, A. T.; Salazar, D.; Castilla, M. E.; Olaiz-Fernandez, G. (1996) Metals in lung
24 tissue from autopsy cases in Mexico City residents: comparison of cases from the 1950s and the 1980s.
25 Environ. Health Perspect. 104: 630-632.
26 Frampton, M. W. (2001) Systemic and cardiovascular effects of airway injury and inflammation: ultrafine particle
27 exposure in humans. Environ. Health Perspect. 109(suppl. 4): 529-532.
28 Frampton, M. W.; Voter, K. Z.; Morrow, P. E.; Roberts, N. J., Jr.; Culp, D. J.; Cox, C.; Utell, M. J. (1992) Sulfuric
29 acid aerosol exposure in humans assessed by bronchoalveolar lavage. Am. Rev. Respir. Dis. 146: 626-632.
30 Frampton, M. W.; Morrow, P. E.; Cox, C.; Levy, P. C.; Condemi, J. J.; Speers, D.; Gibb, F. R.; Utell, M. J. (1995)
31 Sulfuric acid aerosol followed by ozone exposure in healthy and asthmatic subjects. Environ. Res. 69: 1-14.
32 Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
3 3 extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol.
34 277: L960-L967.
35 Fujimaki, H.; Katayama, N.; Wakamori, K. (1992) Enhanced histamine release from lung mast cells of guinea pigs
36 exposed to sulfuric acid aerosols. Environ. Res. 58: 117-123.
37 Gardner, S. Y.; Lehmann, J. R.; Costa, D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in rats.
38 Toxicol. Sci. 56: 175-180.
39 Gavett, S. H.; Madison, S. L.; Dreher, K. L.; Winsett, D. W.; McGee, J. K.; Costa, D. L. (1997) Metal and sulfate
40 composition of residual oil fly ash determines airway hyperreactivity and lung injury in rats. Environ. Res.
41 72: 162-172.
42 Gavett, S. H.; Madison, S. L.; Stevens, M. A.; Costa, D. L. (1999) Residual oil fly ash amplifies allergic cytokines,
43 airway responsiveness, and inflammation in mice. Am. J. Respir. Crit. Care Med. 160: 1897-1904.
44 Gearhart, J. M.; Schlesinger, R. B. (1986) Sulfuric acid-induced airway hyperresponsiveness. Fundam. Appl.
45 Toxicol. 7: 681-689.
46 Gercken, G.; Berg, I.; Dorger, M.; Schluter, T. (1996) Mechanisms of particle-induced activation of alveolar
47 macrophages. Toxicol. Lett. 88: 121-129.
48 Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
49 Am. J. Respir. Crit. Care Med. 164: 704-708.
50 Ohio, A. J.; Meng, Z. H.; Hatch, G. E.; Costa, D. L. (1997a) Luminol-enhanced chemiluminescence after in vitro
51 exposures of rat alveolar macrophages to oil fly ash is metal dependent. Inhalation Toxicol. 9: 255-271.
52 Ohio, A. J.; Piantadosi, C. A.; Crumbliss, A. L. (1997b) Hypothesis: iron chelation plays a vital role in neutrophilic
53 inflammation. BioMetals 10: 135-142.
54 Ohio, A. J.; Richards, J. H.; Dittrich, K. L.; Samet, J. M. (1998a) Metal storage and transport proteins increase after
5 5 exposure of the rat lung to an air pollution particle. Toxicol. Pathol. 26: 3 88-3 94.
56 Ohio, A. J.; Carter, J. D.; Richards, J. H.; Brighton, L. E.; Lay, J. C.; Devlin, R. B. (1998b) Disruption of normal
57 iron homeostasis after bronchial instillation of an iron-containing particle. Am. J. Physiol. 274: L396-L403.
June 2003 7-132 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ohio, A. I; Carter, J. D.; Samet, J. M; Reed, W.; Quay, J.; Dailey, L. A.; Richards, J. H.; Devlin, R. B. (1998c)
2 Metal-dependent expression of ferritin and lactoferrin by respiratory epithelial cells. Am. J. Physiol.
3 274: L728-L736.
4 Ohio, A. J.; Stonehuerner, J.; Dailey, L. A.; Carter, J. D. (1999a) Metals associated with both the water-soluble and
5 insoluble fractions of an ambient air pollution particle catalyze an oxidative stress. Inhalation Toxicol.
6 11:37-49.
7 Ohio, A. J.; Carter, J. D.; Dailey, L. A.; Devlin, R. B.; Samet, J. M. (1999b) Respiratory epithelial cells demonstrate
8 lactoferrin receptors that increase after metal exposure. Am. J. Physiol. 276: L933-L940.
9 Ohio, A. J.; Stoneheurner, J.; McGee, J. K.; Kinsey, J. S. (1999c) Sulfate content correlates with iron concentrations
10 in ambient air pollution particles. Inhalation Toxicol. 11:293-307.
11 Ohio, A. J.; Kim, C.; Devlin, R. B. (2000a) Concentrated ambient air particles induce mild pulmonary inflammation
12 in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
13 Ohio, A. J.; Richards, J. H.; Carter, J. D.; Madden, M. C. (2000b) Accumulation of iron in the rate lung after trachea!
14 instillation of diesel particles. Toxicol. Pathol. 28: 619-627.
15 Ohio, A. J.; Carter, J. D.; Richards, J. H.; Crissman, K. M.; Bobb, H. H.; Yang, F. (2000c) Diminished injury in
16 hypotransferrinemic mice after exposure to a metal-rich particle. Am. J. Physiol. 278: L1051-L1061.
17 Ohio, A. J.; Gilbey, J. G.; Roggli, V. L.; Richards, J. H.; McGee, J. K.; Carson, J. L.; Devlin, R. B.; Cascio, W. E.
18 (2001) Diffuse alveolar damage after exposure to an oil fly ash. Am. J. Respir. Crit. Care Med.
19 164: 1514-1518.
20 Gift, J. S.; Faust, R. A. (1997) Noncancer inhalation toxicology of crystalline silica: exposure-response assessment.
21 J. Exposure Anal. Environ. Epidemiol. 7: 345-358.
22 Gilmour, M. L; Taylor, F. G. R.; Baskerville, A.; Wathes, C. M. (1989a) The effect of titanium dioxide inhalation on
23 the pulmonary clearance of Pasteurella haemolytica in the mouse. Environ. Res. 50: 157-172.
24 Gilmour, M. L; Taylor, F. G. R.; Wathes, C. M. (1989b) Pulmonary clearance of Pasteurella haemolytica and
25 immune responses in mice following exposure to titanium dioxide. Environ. Res. 50: 184-194.
26 Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
27 effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
28 53:817-822.
29 Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms of morbidity and mortality from
30 exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
31 Goldsmith, C.-A.; Frevert, C.; Imrich, A.; Sioutas, C.; Kobzik, L. (1997) Alveolar macrophage interaction with air
32 pollution particulates. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
33 on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
34 NY. Environ. Health Perspect. Suppl. 105(5): 1191-1195.
35 Goldsmith, C.-A. W.; Imrich, A.; Danaee, H.; Ning, Y. Y; Kobzik, L. (1998) Analysis of air pollution
36 particulate-mediated oxidant stress in alveolar macrophages. J. Toxicol. Environ. Health Part A 54: 529-545.
37 Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
3 8 (1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
39 Inhalation Toxicol. 11: 981-998.
40 Gong, H., Jr.; Sioutas, C.; Linn, W. S.; Clark, K. W.; Terrell, S. L.; Terrell, L. L.; Anderson, K. R.; Kim, S.; Chang,
41 M.-C. (2000) Controlled human exposures to concentrated ambient fine particles in metropolitan
42 Los Angeles: methodology and preliminary health-effect findings. In: Phalen, R. F., ed. Inhalation
43 Toxicology: proceedings of the third colloquium on paniculate air pollution and human health (first special
44 issue); June, 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 107-119.
45 Gordon, T.; Nadziejko, C.; Schlesinger, R.; Chen, L. C. (1998) Pulmonary and cardiovascular effects of acute
46 exposure to concentrated ambient paniculate matter in rats. Toxicol. Lett. 96-97: 285-288.
47 Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
48 hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
49 Grabowski, G. M.; Paulauskis, J. D.; Godleski, J. J. (1999) Mediating phosphorylation events in the
50 vanadium-induced respiratory burst of alveolar macrophages. Toxicol. Appl. Pharmacol. 156: 170-178.
51 Gulisano, M.; Marceddu, S.; Barbara, A.; Pacini, A.; Buiatti, E.; Martini, A.; Pacini, P. (1997) Damage to the
52 nasopharyngeal mucosa induced by current levels of urban air pollution: a field study in lambs. Eur. Respir. J.
53 10:567-572.
54 Hahon, N.; Booth, J. A.; Green, F.; Lewis, T. R. (1985) Influenza virus infection in mice after exposure to coal dust
55 and diesel engine emissions. Environ. Res. 37: 44-60.
56 Hamada, K.; Goldsmith, C.-A.; Kobzik, L. (1999) Increased airway hyperresponsiveness and inflammation in a
57 juvenile mouse model of asthma exposed to air-pollutant aerosol. J. Toxicol. Environ. Health 58: 129-143.
June 2003 7-133 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hamada, K.; Goldsmith, C.-A.; Goldman, A.; Kobzik, L. (2000) Resistance of very young mice to inhaled allergen
2 sensitization Is overcome by coexposure to an air-pollutant aerosol. Am. J. Respir. Crit. Care Med.
3 161: 1285-1293.
4 Hamers, T.; Van Schaardenburg, M. D.; Felzel, E. C.; Murk, A. J.; Koeman, J. H. (2000) The application of reporter
5 gene assays for the determination of the toxic potency of diffuse air pollution. Sci. Total Environ.
6 262: 159-174.
7 Harder, S. D.; Soukup, J. M.; Ohio, A. J.; Devlin, R. B.; Becker, S. (2001) Inhalation of PM25 does not modulate
8 host defense or immune parameters in blood or lung of normal human subjects. Environ. Health Perspect.
9 Suppl. 109(4): 599-604.
10 Harkema, J. R.; Morgan, K. T.; Gross, E. A.; Catalano, P. J.; Griffith, W. C. (1994) Consequences of prolonged
11 inhalation of ozone on F344/N rats: collaborative studies. Part VII: effects on the nasal mucociliary apparatus.
12 Cambridge, MA: Health Effects Institute; research report no. 65.
13 Harkema, J. R.; Barr, E. B.; Hotchkiss, J. A. (1997a) Responses of rat nasal epithelium to short- and long-term
14 exposures of ozone: image analysis of epithelial injury, adaptation and repair. Microsc. Res. Tech.
15 36:276-286.
16 Harkema, J. R.; Hotchkiss, J. A.; Griffith, W. C. (1997b) Mucous cell metaplasia in rat nasal epithelium after a
17 20-month exposure to ozone: a morphometric study of epithelial differentiation. Am. J. Respir. Cell Mol.
18 Biol. 16: 521-530.
19 Health Effects Institute. (1995) Diesel exhaust: a critical analysis of emissions, exposure, and health effects:
20 a special report of the Institute's Diesel Working Group. Cambridge, MA: Health Effects Institute.
21 Heck, H. d'A.; Casanova, M.; Starr, T. B. (1990) Formaldehyde toxicity — new understanding. Crit. Rev. Toxicol.
22 20:397-426.
23 Heinrich, U.; Peters, L.; Funcke, W.; Pott, F.; Mohr, U.; Stober, W. (1982) Investigation of toxic and carcinogenic
24 effects of diesel exhaust in long-term inhalation exposure of rodents. In: Lewtas, J., ed. Toxicological effects
25 of emissions from diesel engines: proceedings of the Environmental Protection Agency diesel emissions
26 symposium; October 1981; Raleigh, NC. New York, NY: Elsevier Biomedical; pp. 225-242. (Developments
27 in toxicology and environmental science: v. 10).
28 Heinrich, U.; Muhle, H.; Takenaka, S.; Ernst, H.; Fuhst, R.; Mohr, U.; Pott, F.; Stober, W. (1986) Chronic effects on
29 the respiratory tract of hamsters, mice, and rats after long-term inhalation of high concentrations of filtered
30 and unfiltered diesel engine emissions. J. Appl. Toxicol. 6: 383-395.
31 Heinrich, U.; Fuhst, R.; Rittinghausen, S.; Creutzenberg, O.; Bellmann, B.; Koch, W.; Levsen, K. (1995) Chronic
32 inhalation exposure of Wistar rats and two different strains of mice to diesel engine exhaust, carbon black,
33 and titantium dioxide. Inhalation Toxicol. 7: 533-556.
34 Heyder, J.; Beck-Speier, L; Busch, B.; Dirscherl, P.; Heilmann, P.; Perron, G. A.; Josten, M.; Karg, E.; Kreyling,
35 W. G.; Lenz, A.-G.; Maier, K. L.; Miaskowski, U.; Platz, S.; Reitmeir, P.; Schulz, H.; Takenaka, S.;
36 Ziesenis, A. (1999) Health effects of sulfur-related environmental air pollution. I. Executive summary.
37 Inhalation Toxicol. 11: 343-359.
38 Hitzfeld, B.; Friedrichs, K. H.; Ring, J.; Behrendt, H. (1997) Airborne paniculate matter modulates the production of
39 reactive oxygen species in human polymorphonuclear granulocytes. Toxicology 120: 185-195.
40 Holian, A.; Hamilton, R. F., Jr.; Morandi, M. T.; Brown, S. D.; Li, L. (1998) Urban particle-induced apoptosis and
41 phenotype shifts in human alveolar macrophages. Environ. Health Perspect. 106: 127-132.
42 Hornberg, C.; Maciuleviciute, L.; Seemayer, N. H. (1996) Sister chromatid exchanges in rodent trachea! epithelium
43 exposed in vitro to environmental pollutants. Toxicol. Lett. 88: 45-53.
44 Hornberg, C.; Maciuleviciute, L.; Seemayer, N. H.; Kainka, E. (1998) Induction of sister chromatid exchanges
45 (SCE) in human trachea! epithelial cells by the fractions PM-10 and PM-2.5 of airborne particulates. Toxicol.
46 Lett. 96/97: 215-220.
47 Hunter, R. J. (1981) Zeta potential in colloid science: principles and applications. Orlando, FL: Academic Press,
48 pp. 1-124. (Ottewill, R. H.; Rowell, R. L., eds. Colloid Science).
49 Imrich, A.; Ning, Y. Y.; Kobzik, L. (2000) Insoluble components of concentrated air particles mediate alveolar
50 macrophage responses in vitro. Toxicol. Appl. Pharmacol. 167: 140-150.
51 Irsigler, G. B.; Visser, P. J.; Spangenberg, P. A. L. (1999) Asthma and chemical bronchitis in vanadium plant
52 workers. Am. J. Ind. Med. 35: 366-374.
53 Iwai, K.; Udagawa, T.; Yamagishi, M.; Yamada, H. (1986) Long-term inhalation studies of diesel exhaust on F344
54 SPF rats. Incidence of lung cancer and lymphoma. In: Ishinishi, N.; Koizumi, A.; McClellan, R. O.;
55 Stober, W., eds. Carcinogenic and mutagenic effects of diesel engine exhaust: proceedings of the international
56 satellite symposium on lexicological effects of emissions from diesel engines; July; Tsukuba Science City,
57 Japan. Amsterdam, Holland: Elsevier Science Publishers B. V.; pp. 349-360. (Developments in toxicology
58 and environmental science: v. 13).
June 2003 7-134 DRAFT-DO NOT QUOTE OR CITE
-------
1 Jacobs, I; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
2 1983-1992. Environ. HealthPerspect. 105: 980-985.
3 Jakab, G. J. (1992) Relationship between carbon black particulate-bound formaldehyde, pulmonary antibacterial
4 defenses, and alveolar macrophage phagocytosis. Inhalation Toxicol. 4: 325-342.
5 Jakab, G. J. (1993) The toxicologic interactions resulting from inhalation of carbon black and acrolein on pulmonary
6 antibacterial and antiviral defenses. Toxicol. Appl. Pharmacol. 121: 167-175.
7 Jakab, G. J.; Hemenway, D. R. (1993) Inhalation coexposure to carbon black and acrolein suppresses alveolar
8 macrophage phagocytosis and TNF-a release and modulates peritoneal macrophage phagocytosis. Inhalation
9 Toxicol. 5: 275-289.
10 Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
11 acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
12 Jiang, N.; Dreher, K. L.; Dye, J. A.; Richards, J. H.; Martin, L. D.; Adler, K. B. (2000) Residual oil fly ash induces
13 cytotoxicity and mucin secretion by guinea pig trachea! epithelial cells via an oxidant-mediated mechanism.
14 Toxicol. Appl. Pharmacol. 163: 221-230.
15 Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Baggs, R.; Oberdorster, G. (1996) Characterization of the early
16 pulmonary inflammatory response associated with PTFE fume exposure. Toxicol. Appl. Pharmacol.
17 140: 154-163.
18 Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Oberdorster, G. (1998) Pulmonary inflammatory responses and
19 cytokine and antioxidant mRNA levels in the lungs of young and old C57BL/6 mice after exposure to Teflon
20 fumes. Inhalation Toxicol. 10: 931-953.
21 Kadiiska, M. B.; Mason, R. P.; Dreher, K. L.; Costa, D. L.; Ohio, A. J. (1997) In vivo evidence of free radical
22 formation in the rat lung after exposure to an emission source air pollution particle. Chem. Res. Toxicol.
23 10: 1104-1108.
24 Kennedy, T.; Ohio, A. J.; Reed, W.; Samet, J.; Zagorski, J.; Quay, J.; Carter, J.; Dailey, L.; Hoidal, J. R.; Devlin,
25 R. B. (1998) Copper-dependent inflammation and nuclear factor-icB activation by paniculate air pollution.
26 Am. J. Respir. Cell Mol. Biol. 19: 366-378.
27 Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
28 (1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly ash
29 inhalation. Inhalation Toxicol. 9: 541-565.
30 Kim, S.; Chang, M.-C.; Kim, D.; Sioutas, C. (2000a) A new generation of portable coarse, fine, and ultrafine particle
31 concentrators for use in inhalation toxicology. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the
32 third colloquium on paniculate air pollution and human health (first special issue); June, 1999; Durham, NC.
33 Inhalation Toxicol. 12(suppl. 1): 121-137.
34 Kim, S.; Sioutas, C.; Chang, M.-C.; Gong, H., Jr. (2000b) Factors affecting the stability of the performance of
35 ambient fine-particle concentrators. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
36 Toxicol. 12(suppl. 4): 281-298.
37 Kimmel, T. A.; Chen, L. C.; Bosland, M. C.; Nadziejko, C. (1997) Influence of acid aerosol droplet size on
3 8 structural changes in the rat lung caused by acute exposure to sulfuric acid and ozone. Toxicol. Appl.
39 Pharmacol. 144: 348-355.
40 Kitabatake, M.; Imai, M.; Kasama, K.; Kobayashi, L; Tomita, Y.; Yoshida, K. (1979) Effects of air pollutants on the
41 experimental induction of asthma attacks in guinea pigs: sulfuric acid mist and mixture of the mist and sulfur
42 dioxide. Mie Med. J. 29: 29-36.
43 Kleeberger, S. R.; Levitt, R. C.; Zhang, L.-Y.; Longphre, M.; Harkema, J.; Jedlicka, A.; Eleff, S. M.; DiSilvestre, D.;
44 Holroyd, K. J. (1997) Linkage analysis of susceptibility to ozone-induced lung inflammation in inbred mice.
45 Nat. Genet. 17: 475-478.
46 Kleinman, M. T.; Leaf, D. A.; Kelly, E.; Caiozzo, V.; Osann, K; O'Niell, T. (1998) Urban angina in the mountains:
47 effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
48 Health 53: 388-397.
49 Kleinman, M. T.; Mautz, W. J.; Bjarnason, S. (1999) Adaptive and non-adaptive responses in rats exposed to ozone,
50 alone and in mixtures, with acidic aerosols. Inhalation Toxicol. 11: 249-264.
51 Kleinman, M. T.; Bufalino, C.; Rasmussen, R.; Hyde, D.; Bhalla, D. K.; Mautz, W. J. (2000) Toxicity of chemical
52 components of ambient fine paniculate matter (PM2 5) inhaled by aged rats. J. Appl. Toxicol. 20: 357-364.
53 Knox, R. B.; Suphioglu, C.; Taylor, P.; Desai, R.; Watson, H. C.; Peng, J. L.; Bursill, L. A. (1997) Major grass
54 pollen allergen Lol p 1 binds to diesel exhaust particles: implications for asthma and air pollution. Clin. Exp.
55 Allergy 27: 246-251.
56 Kobzik, L. (1995) Lung macrophage uptake of unopsonized environmental particulates: role of scavenger-type
57 receptors. J. Immunol. 155: 367-376.
June 2003 7-135 DRAFT-DO NOT QUOTE OR CITE
-------
1 Kobzik, L.; Goldsmith, C.-A. W.; Ning, Y. Y.; Qin, G.; Morgan, B.; Imrich, A.; Lawrence I; Murthy, G. G. K.;
2 Catalano, P. J. (2001) Effects of combined ozone and air pollution particle exposure in mice. Boston, MA:
3 Health Effects Institute; research report no. 106. Available: http://www.healtheffects.org/Pubs/Kobzik.pdf
4 [27 January 2003].
5 Kodavanti, U. P.; Costa, D. L. (1999) Animal models to study for pollutant effects. In: Holgate, S. T.; Samet, J. M.;
6 Koren, H. S.; Maynard, R. L., eds. Air pollution and health. New York, NY: Academic Press; pp. 165-197.
7 Kodavanti, U. P.; Jaskot, R. H.; Bonner, J.; Badgett, A.; Dreher, K. L. (1996) Eosinophilic lung inflammation in
8 particulate-induced lung injury: technical consideration in isolating RNA for gene expression studies.
9 Exp. Lung Res. 22: 541-554.
10 Kodavanti, U. P.; Jaskot, R. H.; Su, W. Y.; Costa, D. L.; Ohio, A. J.; Dreher, K. L. (1997a) Genetic variability in
11 combustion particle-induced chronic lung injury. Am. J. Physiol. 272: L521-L532.
12 Kodavanti, U. P.; Jaskot, R. H.; Costa, D. L.; Dreher, K. L. (1997b) Pulmonary proinflammatory gene induction
13 following acute exposure to residual oil fly ash: roles of particle-associated metals. Inhalation Toxicol.
14 9:679-701.
15 Kodavanti, U. P.; Hauser, R.; Christian!, D. C.; Meng, Z. H.; McGee, J.; Ledbetter, A.; Richards, J.; Costa, D. L.
16 (1998a) Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals.
17 Toxicol. Sci. 43:204-212.
18 Kodavanti, U. P.; Costa, D. L.; Bromberg, P. A. (1998b) Rodent models of cardiopulmonary disease: their potential
19 applicability in studies of air pollutant susceptibility. Environ. Health Perspect. Suppl. 106(1): 111-130.
20 Kodavanti, U. P.; Jackson, M. C.; Ledbetter, A. D.; Richards, J. R.; Gardner, S. Y.; Watkinson, W. P.; Campen,
21 M. J.; Costa, D. L. (1999) Lung injury from intratracheal and inhalation exposures to residual oil fly ash in a
22 rat model of monocrotaline-induced pulmonary hypertension. J. Toxicol. Environ. Health Part A 57: 101-121.
23 Kodavanti, U. P.; Mebane, R.; Ledbetter, A.; Krantz, T.; McGee, J.; Jackson, M. C.; Walsh, L.; Hilliard, H.; Chen,
24 B. Y.; Richards, J.; Costa, D. L. (2000a) Variable pulmonary responses from exposure to concentrated
25 ambient air particles in a rat model of bronchitis. Toxicol. Sci. 54: 441-451.
26 Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Watkinson, W. P.; Campen, M. J.; Winsett, D. W.;
27 Richards, J. R.; Crissman, K. M.; Hatch, G. E.; Costa, D. L. (2000b) The spontaneously hypertensive rat as a
28 model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress
29 from inhaled emission particulate matter. Toxicol. Appl. Pharmacol. 164: 250-263.
30 Kodavanti, U. P.; Schladweiler, M. C. J.; Richards, J. R.; Costa, D. L. (2001) Acute lung injury from intratracheal
31 exposure to fugitive residual oil fly ash and its constituent metals in normo- and spontaneously hypertensive
32 rats. Inhalation Toxicol. 13: 37-54.
33 Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Hauser, R.; Christiani, D. C.; McGee, J.; Richards, J. R.;
34 Costa, D. L. (2002) Temporal association between pulmonary and systemic effects of particulate matter in
3 5 healthy and cardiovascular compromised rats. J. Toxicol. Environ. Health Part A 65: 1545-1569.
36 Kuschner, W. G.; D'Alessandro, A.; Wintermeyer, S. F.; Wong, H.; Boushey, H. A.; Blanc, P. D. (1995) Pulmonary
37 responses to purified zinc oxide fume. J. Invest. Med. 43: 371-378.
38 Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
3 9 experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
40 Perspect. 105: 1234-1237.
41 Lambert, A. L.; Dong, W.; Winsett, D. W.; Selgrade, M. K.; Gilmour, M. I. (1999) Residual oil fly ash exposure
42 enhances allergic sensitization to house dust mite. Toxicol. Appl. Pharmacol. 158: 269-277'.
43 Lambert, A. L.; Dong, W.; Selgrade, M. K.; Gilmour, M. I. (2000) Enhanced allergic sensitization by residual oil fly
44 ash particles is mediated by soluble metal constituents. Toxicol. Appl. Pharmacol. 165: 84-93.
45 Last, J. A.; Pinkerton, K. E. (1997) Chronic exposure of rats to ozone and sulfuric acid aerosol: biochemical and
46 structural responses. Toxicology 116: 133-146.
47 Last, J. A.; Hyde, D. M.; Chang, D. P. Y. (1984) A mechanism of synergistic lung damage by ozone and a respirable
48 aerosol. Exp. Lung Res. 7: 223-235.
49 Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular distribution
50 of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol. 18: 687-695.
51 Lay, J. C.; Zeman, K. L.; Ohio, A. J.; Bennett, W. D. (2001) Effects of inhaled iron oxide particles on alveolar
52 epithelial permeability in normal subjects. Inhalation Toxicol. 13: 1065-1078.
53 Leduc, D.; Fally, S.; De Vuyst, P.; Wollast, R.; Yernault, J.-C. (1995) Acute exposure to realistic acid fog: effects on
54 respiratory function and airway responsiveness in asthmatics. Environ. Res. 71: 89-98.
55 Lee, M. M.; Green, F. H. Y.; Roth, S. H.; Karkhanis, A.; Bjarnason, S. G.; Schtirch, S. (1999) Sulfuric acid aerosol
56 induces changes in alveolar surface tension in the guinea pig but not in the rat. Exp. Lung Res. 25: 229-244.
June 2003 7-136 DRAFT-DO NOT QUOTE OR CITE
-------
1 Leikauf, G. D.; McDowell, S. A.; Gammon, K.; Wesselkamper, S. C.; Bachurski, C. J.; Puga, A.; Wiest, J. S.;
2 Leikauf, J. E.; Prows, D. R. (2000) Functional genomics of particle-induced lung injury. Inhalation Toxicol.
3 12:59-73.
4 Leikauf, G.; McDowell, S. A.; Wesselkamper, S. C.; Miller, C. R.; Hardie, W. D.; Gammon, K.; Biswas, P. P.;
5 Korfhagen, T. R.; Bachurski, C. J.; Wiest, J. S.; Willeke, K.; Bingham, E.; Leikauf, J. E.; Aronow, B. J.;
6 Prows, D. R. (2001) Pathogenomic mechanisms for paniculate matter induction of acute lung injury and
7 inflammation in mice. Boston, MA: Health Effects Institute; research report no. 105. Available:
8 http://www.healtheffects.org/pubs-research.htm [15 May, 2003].
9 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1996) Free radical activity and pro-inflammatory effects of
10 particulate air pollution (PM10) in vivo and in vitro. Thorax 51: 1216-1222.
11 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
12 particulate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international
13 meeting on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996;
14 Lake Placid, NY. Environ. HealthPerspect. Suppl. 105(5): 1279-1283.
15 Li, X. Y.; Brown, D.; Smith, S.; MacNee, W.; Donaldson, K. (1999) Short-term inflammatory responses following
16 intratracheal instillation of fine and ultrafine carbon black in rats. Inhalation Toxicol. 11: 709-731.
17 Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D. (1994) Effects of prolonged,
18 repeated exposure to ozone, sulfuric acid, and their combination in healthy and asthmatic volunteers. Am. J.
19 Respir. Crit. Care Med. 150: 431-440.
20 Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
21 mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
22 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
23 California. Environ. Health Perspect. 105: 216-222.
24 Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
25 insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
26 Long, C. M.; Suh, H. H.; Kobzik, L.; Catalano, P. J.; Ning, Y. Y.; Koutrakis, P. (2001) A pilot investigation of the
27 relative toxicity of indoor and outdoor fine particles: in vitro effects of endotoxin and other particulate
28 properties. Environ. Health Perspect. 109: 1019-1026.
29 Longphre, M.; Li, D.; Matovinovic, E.; Gallup, M.; Samet, J. M.; Basbaum, C. B. (2000) Lung mucin production is
30 stimulated by the air pollutant residual oil fly ash. Toxicol. Appl. Pharmacol. 162: 86-92.
31 Lorz, C.; Lopez, J. (1997) Incidence of air pollution in the pulmonary surfactant system of the pigeon (Columbia
32 livid). Anat. Rec. 249: 206-212.
33 Madden, M. C.; Thomas, M. J.; Ohio, A. J. (1999) Acetaldehyde (CH3CHO) production in rodent lung after
34 exposure to metal-rich particles. Free Radical Biol. Med. 26: 1569-1577.
35 Madden, M. C.; Richards, J. H.; Dailey, L. A.; Hatch, G. E.; Ohio, A. J. (2000) Effect of ozone on diesel exhaust
36 particle toxicity in rat lung. Toxicol. Appl. Pharmacol. 168: 140-148.
37 Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
3 8 macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
39 Pharmacol. 153: 28-38.
40 Mauderly, J. L. (1993) Toxicological approaches to complex mixtures. Environ. Health Perspect.
41 Suppl. 101(4): 155-165.
42 Mautz, W. J.; Kleinman, M. T.; Bhalla, D. K.; Phalen, R. F. (2001) Respiratory tract responses to repeated inhalation
43 of an oxidant and acid gas-particle air pollutant mixture. Toxicol. Sci. 61: 331-341.
44 McDowell, S. A.; Gammon, K.; Bachurski, C. J.; Weist, J. S.; Leikauf, J. E.; Prows, D. R. Leikauf, G. D. (2000)
45 Differential gene expression in the initiation and progression of nickel-induced acute lung injury. Am. J.
46 Respir. Cell Mol. Biol. 23: 466-474.
47 McKenna, I. M.; Gordon, T.; Chen, L. C.; Anver, M. R.; Waalkes, M. P. (1998) Expression of metallothionein
48 protein in the lungs of Wistar rats and C57 and DBA mice exposed to cadmium oxide fumes. Toxicol. Appl.
49 Pharmacol. 153: 169-178.
50 Menetrez, M. Y.; Foarde, K. K.; Ensor, D. S. (2001) An analytical method for the measurement of nonviable
51 bioaerosols. J. Air Waste Manage. Assoc. 51: 1436-1442.
52 Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
53 relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
54 Molinelli, A. R.; Madden, M. C.; McGee, J. K.; Stonebuerner, J. G.; Ohio, A. J. (2002) Effect of metal removal on
55 the toxicity of airborne particulate matter from the Utah Valley. Inhalation Toxicol. 14: 1069-1086.
56 Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
57 exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
58 155:245-252.
June 2003 7-137 DRAFT-DO NOT QUOTE OR CITE
-------
1 Morgan, K. T.; Patterson, D. L.; Gross, E. A. (1986) Responses of the nasal mucociliary apparatus of F-344 rats to
2 formaldehyde gas. Toxicol. Appl. Pharmacol. 82: 1-13.
3 Moss, O. R.; Gross, E. A.; James, R. A.; Janszen, D. B.; Ross, P. W.; Roberts, K. C.; Howard, A. M.; Harkema,
4 J. R.; Calderon-Garciduenas, L.; Morgan, K. T. (2001) Respiratory tract toxicity in rats exposed to Mexico
5 City air. Cambridge, MA: Health Effects Institute; research report no. 100. Available:
6 http://www.healtheffects.org/pubs-research.htm [15 May, 2003].
7 Muggenburg, B. A.; Barr, E. B.; Cheng, Y. S.; Seagrave, JC.; Tilley, L. P.; Mauderly, J. L. (2000a) Effect of inhaled
8 residual oil fly ash on the electrocardiogram of dogs. In: Grant, L. D., ed. PM2000: paniculate matter and
9 health. Inhalation Toxicol. 12(suppl. 4): 189-208.
10 Muggenburg, B. A.; Tilley, L.; Green, F. H. Y. (2000b) Animal models of cardiac disease: potential usefulness for
11 studying health effects of inhaled particles. Inhalation Toxicol. 12: 901-925.
12 Mukae, H.; Hogg, J. C.; English, D.; Vincent, R.; VanEeden, S. F. (2000) Phagocytosis of paniculate air pollutants
13 by human alveolar macrophages stimulates the bone marrow. Am. J. Physiol. 279: L924-L931.
14 Mukae, H.; Vincent, R.; Quinlan, K.; English, D.; Hards, J.; Hogg, J. C.; Van Eeden, S. F. (2001) The effect of
15 repeated exposure to paniculate air pollution (PM10) on the bone marrow. Am. J. Respir. Crit. Care Med.
16 163:201-209.
17 Murphy, S. A.; BeruBe, K. A.; Pooley, F. D.; Richards, R. J. (1998) The response of lung epithelium to well
18 characterised fine particles. Life Sci. 62: 1789-1799.
19 Nadadur, S. S.; Kodavanti, U. P. (2002) Altered gene expression profiles of rat lung in response to an emission
20 paniculate and its metal constituents. J. Toxicol. Environ. Health Part A 65:1333-1350.
21 Nadadur, S. S.; Schladweiler, M. C. J.; Kodavanti, U. P. (2000) A pulmonary rat gene array for screening altered
22 expression profiles in air pollutant-induced lung injury. Inhalation Toxicol. 12: 1239-1254.
23 Nadeau, D.; Vincent, R.; Kumarathasan, P.; Brook, J.; Dufresne, A. (1996) Cytotoxicity of ambient air particles to
24 rat lung macrophages: comparison of cellular and functional assays. Toxicol. in Vitro 10: 161-172.
25 Nakamura, T.; Hayashida, Y. (1992) Autonomic cardiovascular responses to smoke exposure in conscious rats.
26 Am. J. Physiol. 262: R73 8-745.
27 National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
28 Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
29 publication no. 97-4051.
30 Nel, A. E.; Diaz-Sanchez, D.; Li, N. (2001) The role of paniculate pollutants in pulmonary inflammation and
31 asthma: evidence for the involvement of organic chemicals and oxidative stress. Curr. Opin. Pulm. Med.
32 7:20-26.
33 Nemmar, A.; Delaunois, A.; Nemery, B.; Dessy-Doize, C.; Beckers, J.-F.; Sulon, J.; Gustin, P. (1999) Inflammatory
34 effect of intratracheal instillation of ultrafme particles in the rabbit: role of C-fiber and mast cells.
35 Toxicol. Appl. Pharmacol. 160: 250-261.
36 Nemmar, A.; Vanbilloen, H.; Hoylaerts, M. F.; Hoet, P. H. M.; Verbruggen, A.; Nemery, B. (2001) Passage of
37 intratracheally instilled ultrafine particles from the lung into the systemic circulation in hamster. Am. J.
38 Respir. Crit. Care Med. 164: 1665-1668.
39 Nemmar, A.; Hoet, P. H. M.; Vanquickenborne, B.; Dinsdale, D.; Thomeer, M.; Hoylaerts, M. F.; Vanbilloen, H.;
40 Mortelmans, L.; Nemery, B. (2002) Passage of inhaled particles into the blood circulation in humans.
41 Circulation 105: 411-414.
42 Nightingale, J. A.; Maggs, R.; Cullinan, P.; Donnelly, L. E.; Rogers, D. F.; Kinnersley, R.; Chung, K. F.; Barnes,
43 P. J.; Ashmore, M.; Newman-Taylor, A. (2000) Airway inflammation after controlled exposure to diesel
44 exhaust particulates. Am. J. Respir. Crit. Care Med. 162: 161-166.
45 Ning, Y.; Imrich, A.; Goldsmith, C. A.; Qin, G.; Kobzik, L. (2000) Alveolar macrophage cytokine production in
46 response to air particles in vitro: role of endotoxin. J. Toxicol. Environ. Health A 59: 165-180.
47 Nordenhall, C.; Pourazar, J.; Ledin, M. C.; Levin, J. O.; Sandstrom, T.; Adelroth, E. (2001) Diesel exhaust enhances
48 airway responsiveness in asthmatic subjects. Eur. Respir. J. 17: 909-915.
49 Norwood, J., Jr.; Ledbetter, A. D.; Doerfler, D. L.; Hatch, G. E. (2001) Residual oil fly ash inhalation in guinea pigs:
50 influence of absorbate and glutathione depletion. Toxicol. Sci. 61: 144-153.
51 Oberdorster, G. (1996a) Effects of ultrafine particles in the lung and potential relevance to environmental particles.
52 In: Marijnissen, J. C. M.; Gradon, L., eds. Proceedings of Aerosol inhalation, lung transport, deposition and
53 the relation to the environment: recent research frontiers, an international workshop; September, 1995;
54 Warsaw, Poland. London, United Kingdom: Kluwer Academic; pp. 165-174.
5 5 Oberdorster, G. (1996b) Significance of particle parameters in the evaluation of exposure-dose-response
56 relationships of inhaled particles. Inhalation Toxicol. 8(suppl.): 73-89.
57 Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
58 lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
June 2003 7-138 DRAFT-DO NOT QUOTE OR CITE
-------
1 Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer particles
2 for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
3 Oberdorster, G.; Finkelstein, J. N.; Johnston, C.; Gelein, R.; Cox, C.; Baggs, R.; Elder, A. C. P. (2000) Acute
4 pulmonary effects of ultrafine particles in rats and mice. Cambridge, MA: Health Effects Institute; research
5 report no. 96.
6 Oettinger, R.; Drumm, K.; Knorst, M; Krinyak, P.; Smolarski, R.; Kienast, K. (1999) Production of reactive oxygen
7 intermediates by human macrophages exposed to soot particles and asbestos fibers and increase in NF-kappa
8 B p50/p!05 mRNA. Lung 177: 343-354.
9 Ohtoshi, T.; Takizawa, H.; Okazaki, H.; Kawasaki, S.; Takeuchi, N.; Ohta, K.; Ito, K. (1998) Diesel exhaust
10 particles stimulate human airway epithelial cells to produce cytokines relevant to airway inflammation
11 in vitro. J. Allergy Clin. Immunol. 101: 778-785.
12 Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Branson, K.; Jakab, G. J.; Kleeberger, S. R. (2000a) Interstrain variation in
13 murine susceptibility to inhaled acid-coated particles. Am. J. Physiol. L469-L476.
14 Ohtsuka, Y.; Branson, K. J.; Jedlicka, A. E.; Mitzner, W.; Clarke, R. W.; Zhang, L.-Y.; Eleff, S. M.; Kleeberger,
15 S. R. (2000b) Genetic linkage analysis of susceptibility to particle exposure in mice. Am. J. Respir. Cell Mol.
16 Biol. 22: 574-581.
17 Oortgiesen, M.; Veronesi, B.; Eichenbaum, G.; Kiser, P. F.; Simon, S. A. (2000) Residual oil fly ash and charged
18 polymers activate epithelial cells and nociceptive sensory neurons. Am. J. Physiol. 278: L683-L695.
19 Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
20 Fundam. Appl. Toxicol. 40: 220-227.
21 Osier, M.; Baggs, R. B.; Oberdorster, G. (1997) Intratracheal instillation versus intratracheal inhalation: influence of
22 cytokines on inflammatory response. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
23 international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles; September
24 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1265-1271.
25 Osman, K.; Zejda, J. E.; Schtitz, A.; Mielzynska, D.; Blinder, C. G.; Vahter, M. (1998) Exposure to lead and other
26 metals in children from Katowice district, Poland. Int. Arch. Occup. Environ. Health71: 180-186.
27 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
28 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
29 Petrovic, S.; Urch, B.; Brook, J.; Datema, J.; Purdham, J.; Liu, L.; Lukic, Z.; Zimmerman, B.; Tofler, G.;
30 Downar, E.; Corey, P.; Tarlo, S.; Broder, I.; Dales, R.; Silverman, F. (1999) Cardiorespiratory effects of
31 concentrated ambient PM2.5: a pilot study using controlled human exposures. In: Phalen, R. F., ed. Inhalation
32 Toxicology: proceedings of the third colloquium on paniculate air pollution and human health (first special
33 issue); June, 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 173-188.
34 Pierce, L. M.; Alessandrini, F.; Godleski, J. J.; Paulauskis, J. D. (1996) Vanadium-induced chemokine mRNA
35 expression and pulmonary inflammation. Toxicol. Appl. Pharmacol. 138: 1-11.
36 Pinkerton, K. E.; Brody, A. R.; Miller, F. J.; Crapo, J. D. (1989) Exposure to low levels of ozone results in enhanced
37 pulmonary retention of inhaled asbestos fibers. Am. Rev. Respir. Dis. 140: 1075-1081.
38 Pinto, M.; Birnbaum, S. C.; Kadar, T.; Goldberg, G. M. (1979) Lung injury in mice induced by factors acting
39 synergistically with inhaled paniculate antigen. Clin. Immunol. Immunopathol. 13: 361-368.
40 Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
41 Am. J. Public Health 79: 623-628.
42 Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and
43 Cache Valleys. Arch. Environ. Health 46: 90-97.
44 Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
45 cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
46 Assoc. 287: 1132-1141.
47 Prahalad, A. K.; Soukup, J. M.; Inmon, J.; Willis, R.; Ohio, A. J.; Becker, S.; Gallagher, J. E. (1999) Ambient air
48 particles: effects on cellular oxidant radical generation in relation to paniculate elemental chemistry. Toxicol.
49 Appl. Pharmacol. 158: 81-91.
50 Pritchard, R. J.; Ohio, A. J.; Lehmann, J. R.; Winsett, D. W.; Tepper, J. S.; Park, P.; Gilmour, M. I.; Dreher, K. L.;
51 Costa, D. L. (1996) Oxidant generation and lung injury after paniculate air pollutant exposure increase with
52 the concentrations of associated metals. Inhalation Toxicol. 8: 457-477.
53 Prows, D. R.; Shertzer, H. G.; Daly, M. J.; Sidman, C. L.; Leikauf, G. D. (1997) Genetic analysis of ozone-induced
54 acute lung injury in sensitive and resistant strains of mice. Nat. Genet. 17: 471-474.
55 Prows, D. R.; Leikauf, G. D. (2001) Quantitative trait analysis of nickel-induced acute lung injury in mice. Am. J.
56 Respir. Cell Mol. Biol. 24: 740-746.
57 Quay, J. L.; Reed, W.; Samet, J.; Devlin, R. B. (1998) Air pollution particles induce IL-6 gene expression in human
58 airway epithelial cells via NF-KB activation. Am. J. Respir. Cell Mol. Biol. 19: 98-106.
June 2003 7-139 DRAFT-DO NOT QUOTE OR CITE
-------
1 Rose, C. S.; Martyny, J. W.; Newman, L. S.; Milton, D. K.; King, T. E., Jr.; Beebe, J. L.; McCammon, J. B.;
2 Hoffman, R. E.; Kreiss, K. (1998) "Lifeguard lung": endemic granulomatous pneumonitis in an indoor
3 swimming pool. Am. J. Public Health 88: 1795-1800.
4 Rudell, B.; Sandstrom, T.; Stjernberg, N.; Kolmodin-Hedman, B. (1990) Controlled diesel exhaust exposure in an
5 exposure chamber: pulmonary effects investigated with bronchoalveolar lavage. J. Aerosol Sci.
6 21(suppl. 1):S411-S414.
7 Rudell, B.; Sandstrom, T.; Hammarstrom, U.; Ledin, M.-L.; Horstedt, P.; Stjernberg, N. (1994) Evaluation of an
8 exposure setup for studying effects of diesel exhaust in humans. Int. Arch. Occup. Environ. Health 66: 77-83.
9 Rudell, B.; Ledin, M.-C.; Hammarstrom, U.; Stjernberg, N.; Lundback, B.; Sandstrom, T. (1996) Effects on
10 symptoms and lung function in humans experimentally exposed to diesel exhaust. Occup. Environ. Med. 53:
11 658-662.
12 Rudell, B.; Blomberg, A.; Helleday, R.; Ledin, M.-C.; Lundback, B.; Stjernberg, N.; Horstedt, P.; Sandstrom, T.
13 (1999) Bronchoalveolar inflammation after exposure to diesel exhaust: comparison between unfiltered and
14 particle trap filtered exhaust. Occup. Environ. Med. 56: 527-534.
15 Sagai, M.; Furuyama, A.; Ichinose, T. (1996) Biological effects of diesel exhaust particles (DEP). III. Pathogenesis
16 of asthma like symptoms in mice. Free Radical Biol. Med. 21: 199-209.
17 Saldiva, P. H. N.; Clarke, R. W; Coull, B. A.; Stearns, R. C.; Lawrence, J.; Krishna-Murthy, G. G.; Diaz, E.;
18 Koutrakis, P.; Suh, H.; Tsuda, A.; Godleski, J. J. (2002) Lung inflammation induced by concentrated ambient
19 air particles is related to particle composition. Am. J. Respir. Crit. Care Med. 165: 1610-1617.
20 Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (1999) Acute inflammatory
21 responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
22 volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.
23 Salvi, S. S.; Nordenhall, C.; Blomberg, A.; Rudell, B.; Pourazar, J.; Kelly, F. J.; Wilson, S.; Sandstrom, T.; Holgate,
24 S. T.; Frew, A. J. (2000) Acute exposure to diesel exhaust increases IL-8 and GRO-aproduction in healthy
25 human airways. Am. J. Respir. Crit. Care Med. 161: 550-557.
26 Samet, J. M.; Reed, W.; Ohio, A. J.; Devlin, R. B.; Carter, J. D.; Dailey, L. A.; Bromberg, P. A.; Madden, M. C.
27 (1996) Induction of prostaglandin H synthase 2 in human airway epithelial cells exposed to residual oil fly
28 ash. Toxicol. Appl. Pharmacol. 141: 159-168.
29 Samet, J. M.; Stonehuerner, J.; Reed, W.; Devlin, R. B.; Dailey, L. A.; Kennedy, T. P.; Bromberg, P. A.; Ohio, A. J.
30 (1997) Disruption of protein tyrosine phosphate homeostasis in bronchial epithelial cells exposed to oil fly
31 ash. Am. J. Physiol. 272: L426-L432.
32 Samet, J. M.; Graves, L. M.; Quay, J.; Dailey, L. A.; Devlin, R. B.; Ohio, A. J.; Wu, W.; Bromberg, P. A.; Reed, W.
33 (1998) Activation of MAPKs in human bronchial epithelial cells exposed to metals. Am. J. Physiol.
34 275:L551-L558.
35 Samet, J. M.; Ohio, A. J.; Costa, D. L.; Madden, M. C. (2000) Increased expression of cyclooxygenase 2 mediates
36 oil fly ash-induced lung injury. Exp. Lung Res. 26: 57-69.
37 SchTuter, T.; Berg, I.; Dorger, M.; Gercken, G. (1995) Effect of heavy metal ions on the release of reactive oxygen
38 intermediates by bovine alveolar macrophages. Toxicology 98: 47-55.
39 Schwartz, J. (1994) What are people dying of on high air pollution days? Environ. Res. 64: 26-35.
40 Seagrave, J. C.; Nikula, K. J. (2000) Multiple modes of responses to air pollution paniculate materials in
41 A549 alveolar type II cells. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
42 12(suppl. 4): 247-260.
43 Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Paniculate air pollution and acute health effects. Lancet
44 (8943): 176-178.
45 Seemayer, N. H.; Hornberg, C. (1998) Malignant transformation of Syrian hamster kidney cells in vitro by
46 interaction of airborne particulates and simian virus (SV-) 40. Toxicol. Lett. 96/97: 231-238.
47 Shukla, A.; Timblin, C.; BeruBe, K.; Gordon, T.; McKinney, W.; Driscoll, K.; Vacek, P.; Mossman, B. T. (2000)
48 Inhaled paniculate matter causes expression of nuclear factor (NF)-KB-related genes and oxidant-dependent
49 NF-KB activation in vitro. Am. J. Respir. Cell Mol. Biol. 23: 182-187.
50 Silbajoris, R.; Ohio, A. J.; Samet, J. M.; Jaskot, R.; Dreher, K. L.; Brighton, L. E. (2000) In vivo and in vitro
51 correlation of pulmonary map kinase activation following metallic exposure. Inhalation Toxicol. 12: 453-468.
52 Sindhu, R. K.; Mautz, W. J.; Kikkawa, Y. (1998) Chronic exposure to ozone and nitric acid vapor results in
53 increased levels of rat pulmonary putrescine. Arch. Toxicol. 72: 445-449.
54 Sioutas, C.; Koutrakis, P.; Burton, R. M. (1995a) A technique to expose animals to concentrated fine ambient
55 aerosols. Environ. Health Perspect. 103: 172-177.
56 Sioutas, D.; Koutrakis, P.; Ferguson, S. T.; Burton, R. M. (1995b) Development and evaluation of a prototype
57 ambient particle concentrator for inhalation exposure studies. Inhalation Toxicol. 7: 633-644.
June 2003 7-140 DRAFT-DO NOT QUOTE OR CITE
-------
1 Sioutas, C.; Kim, S.; Chang, M; Terrell, L. L.; Gong, H., Jr. (2000) Field evaluation of a modified DataRAM MIE
2 scattering monitor for real-time PM2 5 mass concentration measurements. Atmos. Environ. 34: 4829-4838.
3 Sjogren, B. (1997) Occupational exposure to dust: inflammation and ischaemic heart disease. Occup. Environ. Med.
4 54; 466-469.
5 Smith, K. R.; Aust, A. E. (1997) Mobilization of iron from urban particulates leads to generation of reactive oxygen
6 species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chem. Res. Toxicol.
7 10:828-834.
8 Smith, K. R.; Veranth, J. M.; Lighty, J. S.; Aust, A. E. (1998) Mobilization of iron from coal fly ash was dependent
9 upon the particle size and the source of coal. Chem. Res. Toxicol. 11: 1494-1500.
10 Soukup, J. M.; Becker, S. (2001) Human alveolar macrophage responses to air pollution particulates are associated
11 with insoluble components of coarse material, including paniculate endotoxin. Toxicol. Appl. Pharmacol.
12 171:20-26.
13 Soukup, J. M.; Ohio, A. J.; Becker, S. (2000) Soluble components of Utah Valley paniculate pollution alter alveolar
14 macrophage function in vivo and in vitro. Inhalation Toxicol. 12:401-414.
15 Spengler, J. D.; Thurston, G. D. (1983) Mass and elemental composition of fine and coarse particles in six U.S.
16 cities. J. AirPollut. Control Assoc. 33: 1162-1171.
17 Steerenberg, P. A.; Dormans, J. A. M. A.; Van Doom, C. C. M.; Middendorp, S.; Vos, J. G.; VanLoveren, H.
18 (1999) A pollen model in the rat for testing adjuvant activity of air pollution components. Inhalation Toxicol.
19 11: 1109-1122.
20 Stober, W. (1986) Experimental induction of tumors in hamsters, mice and rats after long-term inhalation of filtered
21 and unfiltered diesel engine exhaust. In: Ishinishi, N.; Koizumi, A.; McClellan, R. O.; Stober, W., eds.
22 Carcinogenic and mutagenic effects of diesel engine exhaust: proceedings of the international satellite
23 symposium on lexicological effects of emissions from diesel engines; July; Tsukuba Science City, Japan.
24 Amsterdam, The Netherlands: Elsevier Science Publishers B. V.; pp. 421-439. (Developments in toxicology
25 and environmental science: v. 13).
26 Stringer, B.; Kobzik, L. (1996) Alveolar macrophage uptake of the environmental paniculate titanium dioxide:
27 role of surfactant components. Am. J. Respir. Cell Mol. Biol. 14: 155-160.
28 Stringer, B.; Kobzik, L. (1998) Environmental particulate-mediated cytokine production in lung epithelial cells
29 (A549): role of preexisting inflammation and oxidant stress. J. Toxicol. Environ. Health Part A 55: 31-44.
30 Stringer, B.; Imrich, A.; Kobzik, L. (1996) Lung epithelial cell (A549) interaction with unopsonized environmental
31 particulates: quantitation of particle-specific binding and IL-8 production. Exp. Lung Res. 22: 495-508.
32 Su, W.-Y.; Jaskot, R. H.; Richards, J.; Abramson, S. R.; Woessner, J. F., Jr.; Yu, W.-H.; Dreher, K. L. (2000a)
3 3 Induction of pulmonary matrilysin expression by combustion and ambient air particles. Am. J. Physiol.
34 279: L152-L160.
35 Su, W.-Y.; Jaskot, R. H.; Dreher, K. L. (2000b) Paniculate matter induction of pulmonary gelatinase A, gelatinase
36 B, tissue inhibitor of metalloproteinase expression. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings
37 of the third colloquium on paniculate air pollution and human health (second special issue); June, 1999;
3 8 Durham, NC. Inhalation Toxicol. 12(suppl. 2): 105-119.
39 Sun, G.; Crissman, K.; Norwood, J.; Richards, J.; Slade, R.; Hatch, G. E. (2001) Oxidative interactions of synthetic
40 lung epithelial lining fluid with metal-containing paniculate matter. Am. J. Physiol. 281: L807-L815.
41 Suwa, T.; Hogg, J. C.; Quinlan, K. B.; Ohgami, A.; Vincent, R.; VanEeden, S. F. (2002) Paniculate air pollution
42 induces progression of atherosclerosis. J. Am. Coll. Cardiol. 39: 935-942.
43 Takano, H.; Yoshikawa, T.; Ichinose, T.; Miyabara, Y.; Imaoka, K.; Sagai, M. (1997) Diesel exhaust particles
44 enhance antigen-induced airway inflammation and local cytokine expression in mice. Am. J. Respir. Crit.
45 Care Med. 156: 36-42.
46 Takano, H.; Ichinose, T.; Miyabara, Y.; Shibuya, T.; Lim, H.-B.; Yoshikawa, T. Sagai, M. (1998) Inhalation of
47 diesel exhaust enhances allergen-related eosinophil recruitment and airway hyperresponsiveness in mice.
48 Toxicol. Appl. Pharmacol. 150: 328-337.
49 Tan, W. C.; Qiu, D.; Liam, B. L.; Ng, T. P.; Lee, S. H.; VanEeden, S. F.; D'Yachkova, Y.; Hogg, J. C. (2000)
50 The human bone marrow response to acute air pollution caused by forest fires. Am. J. Respir. Crit. Care Med.
51 161:1213-1217.
52 Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; VanEeden, S. F. (1997) Phagocytosis of small carbon particles
53 (PM10) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
54 Am. J. Respir. Crit. Care Med. 155: 1441-1447.
55 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
56 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
57
June 2003 7-141 DRAFT-DO NOT QUOTE OR CITE
-------
1 Timblin, C.; BeruBe, K.; Chrug, A.; Driscoll, K.; Gordon, T.; Hemenway, D.; Walsh, E.; Cummins, A. B.;
2 Vacek, P.; Mossman, B. (1998) Ambient paniculate matter causes activation of the c-jun
3 kinase/stress-activated protein kinase cascade and DNA synthesis in lung epithelial cells. Cancer Res.
4 58: 4543.4547.
5 Tsien, A.; Diaz-Sanchez, D.; Ma, I; Saxon, A. (1997) The organic component of diesel exhaust particles and
6 phenanthrene, a major polyaromatic hydrocarbon constituent, enhances IgE production by IgE-secreting
7 EBV-transformed human B cells in vitro. Toxicol. Appl. Pharmacol. 142: 256-263.
8 Tsuchiyama, F.; Hisanaga, N.; Shibata, E.; Aoki, T.; Takagi, H.; Ando, T.; Takeuchi, Y. (1997) Pulmonary metal
9 distribution in urban dwellers. Int. Arch. Occup. Environ. Health 70: 77-84.
10 Turpin, B. J. (1999) Options for characterizing organic paniculate matter. Environ. Sci. Technol. 33: 76A-79A.
11 U.S. Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
12 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
13 Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
14 U.S. Environmental Protection Agency. (1989) An acid aerosols issue paper: health effects and aerometrics.
15 Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
16 Assessment Office; EPA report no. EPA-600/8-88-005F. Available from: NTIS, Springfield, VA;
17 PB91-125864.
18 U.S. Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park,
19 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
20 U.S. Environmental Protection Agency. (1996b) Ambient levels and noncancer health effects of inhaled crystalline
21 and amorphous silica: health issue assessment [draft]. Research Triangle Park, NC: National Center for
22 Environmental Assessment; EPA report no. EPA/600/R-95/115.
23 U.S. Environmental Protection Agency. (1999) Discussion paper for CASAC-diesel health assessment issues.
24 Washington, DC: Office of Research and Development, National Center for Environmental Assessment;
25 paper no. NCEA-W-0634. Available: http://cfpub.epa.gov/ncea/ [16 May, 2003].
26 U.S. Environmental Protection Agency. (2000) Health assessment document for diesel exhaust [SAB review draft].
27 Washington, DC: Office of Research and Development, National Center for Environmental
28 Assessment-Washington Office; report no. EPA/600/8-90/057E. Available:
29 http://cfpub.epa.gov/ncea/cfm/nceahome.cfm [29 March, 2002].
30 Utell, M. J.; Morrow, P. E.; Hyde, R. W. (1983) Latent development of airway hyperreactivity in human subjects
31 after sulfuric acid aerosol exposure. J. Aerosol Sci. 14: 202-205.
32 Van Eeden, S. F.; Tan, W. C.; Suwa, T.; Mukae, H.; Terashima, T.; Fujii, T.; Qui, D.; Vincent, R.; Hogg, J. C.
33 (2001) Cytokines involved in the systemic inflammatory response induced by exposure to paniculate matter
34 air pollutants (PM10). Am. J. Respir. Crit. Care Med. 164: 826-830.
35 VanMaanen, J. M.; Borm, P. J.; Knaapen, A.; VanHerwijnen, M.; Schilderman, P. A.; Smith, K. R.; Aust, A. E.;
36 Tomatis, M.; Fubini, B. (1999) in vitro effects of coal fly ashes: hydroxyl radical generation, iron release,
37 and DNA damage and toxicity in rat lung epithelial cells. Inhalation Toxicol. 11: 1123-1141.
38 Van Zijverden, M.; Granum, B. (2000) Adjuvant activity of paniculate pollutants in different mouse models.
39 Toxicology 152: 69-77.
40 Van Zijverden, M.; Van der Fiji, A.; Bol, M.; Van Pinxteren, F. A.; De Haar, C.; Penninks, A. H.; Van Loveren, H.;
41 Pieters, R. (2000) Diesel exhaust, carbon black, and silica particles display distinct Thl/Th2 modulating
42 activity. Toxicol. Appl. Pharmacol. 168: 131-139.
43 Van Zijverden, M.; De Haar, C.; Van Beelen, A.; Van Loveren, H.; Penninks, A.; Pieters, R. (2001)
44 Co-administration of antigen and particles optimally stimulates the immune response in an intranasal
45 administration model in mice. Toxicol. Appl. Pharmacol. 177: 174-178.
46 Vanda, B.; de Buen, N.; Jasso, R.; Valero, G.; Vargas, M. H.; Olmos, R.; Arreola, J. L.; Santillan, P.; Alonso, P.
47 (1998) Inflammatory cells and ferruginous bodies inbronchoalveolar lavage in urban dogs. Acta Cytol.
48 42: 939-944.
49 Veronesi, B.; Oortgiesen, M. (2001) Neurogenic inflammation and paniculate matter (PM) air pollutants.
50 Neurotoxicology 22: 795-810.
51 Veronesi, B.; Oortgiesen, M.; Carter, J. D.; Devlin, R. B. (1999a) Paniculate matter initiates inflammatory cytokine
52 release by activation of capsaicin and acid receptors in a human bronchial epithelial cell line. Toxicol. Appl.
53 Pharmacol. 154: 106-115.
54 Veronesi, B.; Carter, J. D.; Devlin, R. B.; Simon, S. A.; Oortgiesen, M. (1999b) Neuropeptides and capsaicin
5 5 stimulate the release of inflammatory cytokines in a human bronchial epithelial cell line. Neuropeptides
56 33:447-456.
June 2003 7-142 DRAFT-DO NOT QUOTE OR CITE
-------
1 Veronesi, B.; Oortgiesen, M.; Roy, I; Carter, J. D.; Simon, S. A.; Gavett, S. H. (2000) Vanilloid (capsaicin)
2 receptors influence inflammatory sensitivity in response to paniculate matter. Toxicol. Appl. Pharmacol.
3 169:66-76.
4 Veronesi, B.; De Haar, C.; Roy, J.; Oortgiesen, M. (2002a) Paniculate matter inflammation and receptor sensitivity
5 are target cell specific. Inhalation Toxicol. 14: 159-183.
6 Veronesi, B.; De Haar, C.; Lee, L.; Oortgiesen, M. (2002b) The surface charge of paniculate matter related to
7 cytokine release in human bronchial epithelial cells (BEAS-2B). Toxicol. Appl. Pharmacol. 178: 144-154.
8 Vincent, R.; Bjarnason, S. G.; Adamson, I. Y. R.; Hedgecock, C.; Kumarathasan, P.; Guenette, J.; Potvin, M.;
9 Goegan, P.; Bouthillier, L. (1997) Acute pulmonary toxicity of urban paniculate matter and ozone. Am. J.
10 Pathol. 151: 1563-1570.
11 Vincent, R.; Kumarathasan, P.; Goegan, P.; Bjarnason, S. G.; Guenette, J.; Berube, D.; Adamson, I. Y.; Desjardins,
12 S.; Burnett, R. T.; Miller, F. J.; Battistini, B. (2001) Inhalation toxicology of urban ambient paniculate matter:
13 acute cardiovascular effect in rats. Boston, MA: Health Effects Institute; research report no. 104. Available:
14 http://www.healtheffects.org/Pubs/Vincent.pdf [29 January 02].
15 Vogelzang, P. F. J.; VanDer Gulden, J. W. J.; Folgering, H.; Kolk, J. J.; Heederik, D.; Preller, L.; Tielen, M. J. M.;
16 Van Schayck, C. P. (1998) Endotoxin exposure as a major determinant of lung function decline in pig
17 farmers. Am. J. Respir. Crit. Care Med. 157: 15-18.
18 Wagner, J. G.; Hotchkiss, J. A.; Harkema, J. R. (2001) Effects of ozone and endotoxin coexposure on rat airway
19 epithelium: potentiation of toxicant-induced alterations. Environ. Health Perspect. 109(suppl. 4): 591-598.
20 Walters, D. M.; Breysse, P. N.; Wills-Karp, M. (2001) Ambient urban Baltimore particulate-induced airway
21 hyperresponsiveness and inflammation in mice. Am. J. Respir. Crit. Care Med. 164: 1438-1443.
22 Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil fly
23 ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
24 Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Kodavanti, U. P.; Dreher, K. L.; Su, W.-Y.; Highfill, J. W.; Costa,
25 D. L. (2000a) Cardiovascular effects following exposure to paniculate matter in healthy and
26 cardiopulmonary-compromised rats. In: Heinrich, U.; Mohr, U., eds. Relationships between acute and chronic
27 effects of air pollution. Washington, DC: ISLI Press, pp. 447-463.
28 Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U.P.; Highfill, J. W.; Costa, D. L. (2000b)
29 Thermoregulatory effects following exposure to paniculate matter in healthy and
30 cardiopulmonary-compromised rats. J. Therm. Biol. 25: 131-137.
31 Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Costa, D. L. (2001) Cardiovascular and systemic responses to
32 inhaled pollutants in rodents: effects of ozone and paniculate matter. Environ. Health Perspect.
33 109(suppl. 4): 539-546.
34 Wellenius, G. A.; Saldiva, P. H. N.; Batalha, J. R. F.; Murthy, G. G. K.; Coull, B. A.; Verrier, R. L.; Godleski, J. J.
3 5 (2002) Electrocardiographic changes during exposure to residual oil fly ash (ROFA) particles in a rat model
36 of myocardial infarction. Toxicol. Sci. 66: 327-335.
37 Wesselkamper, S. C.; Prows, D. R.; Biswas, P.; Willeke, K.; Bingham, E.; Leikauf, G. D. (2000) Genetic
38 susceptibility to irritant-induced acute lung injury in mice. Am. J. Physiol. 279: L575-L582.
39 Wheatley, L. M.; Platts-Mills, T. A. E. (1996) Perennial allergens and the asthma epidemic. Sci. Med. 3: 6-13.
40 Woodin, M. A.; Hauser, R.; Liu, Y.; Smith, T. J.; Siegel, P. D.; Lewis, D. M.; Tollerud, D. J.; Christian!, D. C.
41 (1998) Molecular markers of acute upper airway inflammation in workers exposed to fuel-oil ash. Am. J.
42 Respir. Crit. Care Med. 158: 182-187.
43 Wu, W.; Samet, J. M.; Ohio, A. J.; Devlin, R. B. (2001) Activation of the EOF receptor signaling pathway in human
44 airway epithelial cells exposed to Utah Valley PM. Am. J. Physiol. 281: L483-L489.
45 Zelikoff, J. T.; Sisco, M.; Cohen, M. D.; Frampton, M. W.; Utell, M. J; Schlesinger, R. B. (1994) Interspecies
46 comparison of immunotoxicity of inhaled sulfuric acid. II. New Zealand white rabbits. In: 1994 international
47 conference sponsored by the American Lung Association and the American Thoracic Society; May; Boston,
48 MA. Am. J. Respir. Crit. Care Med. 149: A621.
49 Zelikoff, J. T.; Frampton, M. W.; Cohen, M. D.; Morrow, P. E.; Sisco, M.; Tsai, Y.; Utell, M. J.; Schlesinger, R. B.
50 (1997) Effects of inhaled sulfuric acid aerosols on pulmonary immunocompetence: a comparative study in
51 humans and animals. Inhalation Toxicol. 9: 731-752.
52 Zelikoff, J. T.; Chen, L. C.; Cohen, M. D.; Fang, K.; Gordon, T.; Li, Y.; Nadziejko, C.; Schlesinger, R. B. (2003)
53 Effects of inhaled ambient paniculate matter on pulmonary antimicrobial immune defense. Inhalation
54 Toxicol. 15: 131-150.
5 5 Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
56 normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
57 Zock, J.-P.; Hollander, A.; Heederik, D.; Douwes, J. (1998) Acute lung function changes and low endotoxin
58 exposures in the potato processing industry. Am. J. Ind. Med. 33: 384-391.
June 2003 7-143 DRAFT-DO NOT QUOTE OR CITE
-------
i 8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS
2 ASSOCIATED WITH AMBIENT
3 PARTICULATE MATTER
4
5
6 8.1 INTRODUCTION
7 Epidemiologic studies linking community ambient PM concentrations to health effects
8 played an important role in the 1996 PM Air Quality Criteria Document (PM AQCD; U.S.
9 Environmental Protection Agency, 1996a). Many of those studies reported that measurable
10 excesses in pulmonary function decrements, respiratory symptoms, hospital and emergency
11 department admissions, and mortality in human populations are associated with ambient levels
12 of PM25, PM10_25, PM10, and/or other indicators of PM exposure. Numerous more recent
13 epidemiologic studies discussed in this chapter have also evaluated ambient PM relationships to
14 morbidity and mortality and, thereby, provide an expanded basis for assessment of health effects
15 associated with exposures to airborne PM at concentrations currently encountered in the United
16 States.
17 The epidemiology studies assessed here are best considered in combination with
18 information on ambient PM concentrations presented in Chapter 3, studies of human PM
19 exposure (Chapter 5), and PM dosimetry and toxicology (Chapters 6 and 7). The epidemiology
20 studies contribute important information on associations between health effects and exposures of
21 human populations to "real-world" ambient PM and also help to identify susceptible subgroups
22 and associated risk factors. Chapter 9 provides an interpretive synthesis of information drawn
23 from this and other chapters.
24 This chapter opens with discussion of approaches used for selecting studies, followed by a
25 brief overview of key general features of the several types of epidemiologic studies assessed and
26 discussion of important general methodological issues that need to be considered in their critical
27 assessment. Then, Section 8.2 assesses epidemiologic studies of PM effects on mortality; and
28 Section 8.3 evaluates studies of morbidity as a health endpoint. Section 8.4 provides an
29 interpretive assessment of the overall PM epidemiologic data base reviewed in Sections 8.2 and
30 8.3 in relation to various key issues. The overall key findings and conclusions for this chapter
31 are then summarized in Section 8.5.
June 2003 8-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.1.1 Approaches for Identifying and Assessing Studies
2 Numerous PM epidemiologic papers have been published since completion of the 1996 PM
3 AQCD, and U.S. EPA (NCEA-RTP) has used a systematic approach to identifying pertinent
4 epidemiologic studies for consideration in this chapter. In general, an ongoing continuous
5 Medline search has been employed in conjunction with other strategies to identify PM literature
6 pertinent to developing criteria for PM NAAQS. The literature search method is similar to those
7 used by others (e.g., Basu and Samet, 1999). A publication base was first established by using
8 Medline and other data bases and a set of key words (particles, air pollution, mortality,
9 morbidity, cause of death, PM, etc.) in a search strategy which was later reexamined and
10 modified to enhance identification of pertinent published papers. Since literature searches
11 encounter not a static but a changing, growing stream of information, searches are not run just
12 for the most recent calendar quarter but are backdated in an attempt to capture references added
13 to that time period since the previous search was conducted. Papers were also added to the
14 publication base by EPA staff (a) through review of advance tables of contents of thirty journals
15 in which relevant papers are published and (b) by requesting scientists known to be active in the
16 field to identify papers recently accepted for publication.
17 While the above search regime builds a certain degree of redundancy into the system,
18 which ensures good coverage of the relevant literature and lessens the possibility of important
19 papers being missed, additional approaches have augmented traditional search methods. First, at
20 the beginning of the process, a Federal Register Notice was issued, requesting information and
21 published papers from the public at large. Next, non-EPA chapter authors are expert in this
22 field; and, while EPA provides them with the outcomes of searches, the authors are also charged
23 with identifying the literature on their own. Finally, a keystone in the literature identification
24 process is that, at several review stages in the process, both the public and CASAC offer
25 comments which also often identify potentially relevant publications.
26 The publication of new PM studies has been and is proceeding at a prodigious rate; and the
27 acquisition and evaluation of pertinent literature in this PM AQCD development process is an
28 ongoing process which continues to identify new information for consideration. Efforts have
29 been made to assess here pertinent new studies accepted for publication through April, 2002,
30 as well as some published since then (if such recent new papers provide particularly important
31 information helpful in addressing key scientific issues).
June 2003 8-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 Those epidemiologic studies that relate measures of ambient air PM to human health
2 outcomes are assessed in this chapter, whereas studies of (typically much higher) occupational
3 exposures are not considered here. Criteria used for selecting literature for the present
4 assessment include mainly whether a given study includes information on: (1) ambient PM
5 indices (e.g., PM10, PM25, PM10_25, etc.) as a key element; (2) analyses of health effects of
6 specific PM chemical or physical constituents (e.g., metals, sulfates, nitrates or ultrafine
7 particles, etc.); (3) evaluation of health endpoints and populations not previously extensively
8 researched; (4) multiple pollutant analyses; and/or (5) for long-term effects, mortality
9 displacement information.
10 To produce a thorough appraisal of the evidence, the authors first concisely highlight key
11 points derived from the 1996 PM AQCD assessment of the available information. Then, key
12 new information is presented in succinct text summary tables for important new studies that have
13 become available since the prior PM AQCD. More detailed information on methodological
14 features and results for these and other numerous newly available studies is summarized in
15 tabular form in Appendices 8A and 8B. These appendix tables are generally organized to
16 include: (1) information about study location and ambient PM levels; (2) description of study
17 methods employed; (3) results and comments; and (4) quantitative outcomes for PM measures.
18 In the main body of the chapter, greater emphasis is placed on integrating and interpreting
19 findings from the array of evidence provided by the more important newer studies than on
20 detailed evaluation of each of the numerous newly available studies.
21 Particular emphasis is focused in the text on those studies and analyses thought to provide
22 information most directly applicable for U.S. standard setting purposes. Specifically, North
23 American studies conducted in the U.S. or Canada are generally accorded more text discussion
24 than those from other geographic regions; and analyses using gravimetric (mass) measurements
25 are generally accorded more text attention than those using non-gravimetric ambient PM
26 measures, e.g., black smoke (BS) or coefficient of haze (CoH). In addition, emphasis is placed
27 on text discussion of (a) new multi-city studies that employ standardized methodological
28 analyses for evaluating PM effects across several or numerous cities and often provide overall
29 effects estimates based on combined analyses of information pooled across multiple cities and/or
30 (b) other studies providing quantitative PM effect-size estimates for populations of interest.
31
June 2003 8-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 While efforts have been made to acquire and evaluate all pertinent newly available
2 published studies presenting acceptable statistical analysis of health outcomes in relation to
3 quantitative gravimetric measures of exposure to PM25, PM10_25, PM10, etc., this does not
4 necessarily ensure that all possible studies have been found and summarized in appendix tables
5 or assessed in the main text. Nevertheless, the large database considered, containing such
6 numerous studies, tends to insulate the integration of the body of evidence from the potential
7 impacts of omitting one or another study that may not necessarily be key in and of itself. The
8 interpretation and integration presented are done with the goal of producing an objective
9 appraisal of the evidence, including weighing of alternative views on controversial issues.
10 In assessing the relative scientific quality of epidemiologic studies reviewed here and to
11 assist in the interpretations of their findings, the following types of questions were considered, as
12 was done in the 1996 PM AQCD:
13 (1) Was the quality of the aerometric data used sufficient to allow for meaningful
characterization of geographic or temporal differences in study population pollutant
exposures in the range(s) of pollutant concentrations evaluated?
14 (2) Were the study populations well defined and adequately selected so as to allow for
meaningful comparisons between study groups or meaningful temporal analyses of health
effects results?
15 (3) Were the health endpoint measurements meaningful and reliable, including clear
definition of diagnostic criteria utilized and consistency in obtaining dependent variable
measurements?
16 (4) Were the statistical analyses used appropriate and properly performed and interpreted,
including accurate data handling and transfer during analyses?
17 (5) Were likely important confounding or covarying factors adequately controlled for or
taken into account in the study design and statistical analyses?
18 (6) Were the reported findings internally consistent, biologically plausible, and coherent in
terms of consistency with other known facts?
19 These guidelines provide benchmarks for judging the relative quality of various studies and
20 for selecting the best for use in criteria development. Detailed critical analysis of all
21 epidemiologic studies on PM health effects, especially in relation to all of the above questions, is
June 2003 8-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 beyond the scope of this document. Of most importance for present purposes are those studies
2 which provide useful qualitative or quantitative information on exposure-effect or
3 exposure-response relationships for health effects associated with ambient air levels of PM
4 currently likely to be encountered in the United States.
5
6 8.1.2 Types of Epidemiologic Studies Reviewed
7 Definitions of various types of epidemiologic studies assessed here were provided in the
8 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) and are briefly summarized
9 here. Briefly, the epidemiologic studies are divided into mortality studies and morbidity studies.
10 Mortality studies evaluating PM effects on total (non-accidental) mortality and cause-specific
11 mortality provide the most unambiguous evidence related to a clearly adverse endpoint. The
12 morbidity studies further evaluate PM effects on a wide range of health endpoints, such as
13 cardiovascular and respiratory-related hospital admissions, medical visits, reports of respiratory
14 symptoms, self-medication in asthmatics, changes in pulmonary function tests (PFT), low
15 birthweight infants, etc.
16 The epidemiologic strategies most commonly used in PM health studies are of four types:
17 (1) ecologic studies; (2) time-series semi-ecologic studies; (3) longitudinal panel and
18 prospective cohort studies; and (4) case-control and crossover studies. In addition, time-series
19 analyses or other analytic approaches have been used in intervention studies. All of these are
20 observational studies rather than experimental studies. In general, the exposure of the participant
21 is not directly observed; and the concentration of airborne particles and other air pollutants at
22 one or more stationary air monitors is used as a proxy for individual exposure to ambient air
23 pollution.
24 In ecologic studies, the responses are at a community level (for example, annual mortality
25 rates), as are the exposure indices (for example, annual average PM concentrations) and
26 covariates (for example, the percentage of the population greater than 65 years of age).
27 No individual data are used in the analysis; therefore, the relationship between health effect and
28 exposure calculated across different communities may not reflect individual-level associations
29 between health outcome and exposure. The use of proxy measures for individual exposure and
30 covariates or effect modifiers may also bias the results, and within-city or within-unit
31 confounding may be overlooked.
June 2003 8-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 Time-series studies are more informative because they allow the study of associations
2 between changes in a health outcome and changes in exposure indicators preceding or
3 simultaneous with the outcome. The temporal relationship supports a conclusion of a causal
4 relation, even when both the outcome (for example, the number of non-accidental deaths in a
5 city during a day) and the exposure (for example, daily air pollution concentration) are
6 community indices.
7 Prospective cohort (or panel) studies use data from individuals, including health status
8 (where available), individual exposure (not usually available), and individual covariates or risk
9 factors, observed over time. The participants in a prospective cohort study are ideally recruited
10 (using a simple or stratified random sample) so as to represent a target population for which
11 individual or community exposure of the participants is known before and during the interval up
12 to the time the health endpoint occurs. The use of individual-level data is believed to give
13 prospective cohort studies greater inferential strength than other epidemiologic strategies. The
14 use of community-level or estimated exposure data, if necessary, may weaken this advantage, as
15 it does in time-series studies.
16 Case-control studies are retrospective studies in that exposure is determined after the
17 health endpoint occurs (as is common in occupational health studies). As Rothman and
18 Greenland (1998) describe it, "Case-control studies are best understood by defining a source
19 population, which represents a hypothetical study population in which a cohort study might have
20 been conducted ... In a case-control study, the cases are identified and their exposure status is
21 determined just as in a cohort study . . . [and] a control group of study subjects is sampled from
22 the entire source population that gives rise to the cases ... the cardinal requirement of control
23 selection is that the controls must be sampled independently of their exposure status."
24 The case-crossover design is suited to the study of a transient effect of an intermittent
25 exposure on the subsequent risk of an acute-onset health effect hypothesized to occur a short
26 time after exposure. In the original development of the method, effect estimates were based on
27 within-subject comparisons of exposures associated with incident disease events with exposures
28 at times before the occurrence of disease, using matched case-control methods or methods for
29 stratified follow-up studies with spare data within each stratum. The principle of the analysis is
30 that the exposures of cases just before the event are compared with the distribution of exposure
31 estimated from some separate time period. This distribution is assumed to be representative of
June 2003 8-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 the distribution of exposures for those individuals while they were at risk of developing the
2 outcome of interest.
3 When measurements of exposure or potential effect modifiers are available on an
4 individual level, it is possible to incorporate this information into a case-crossover study (unlike
5 a time-series analysis). A disadvantage of the case-crossover design, however, is the potential
6 for bias due to time trends in the exposure time-series. Because case-crossover comparisons are
7 made between different points in time, the case-crossover analysis implicitly depends on an
8 assumption that the exposure distribution is stable over time (stationary). If the exposure time-
9 series is non-stationary and case exposures are compared with referent exposures systematically
10 selected from a different period in time, a bias may be introduced into estimates of the measure
11 of association for the exposure and disease. These biases are particularly important when
12 examining the small associations that appear to exist between PM and health outcomes.
13 Intervention studies (often involving features of time-series or other above types of
14 analyses) provide a particularly powerful additional approach for evaluating possible causal
15 relationships between ambient air pollution variables (e.g., PM) and health effects in human
16 populations. In such studies, the effects of active interventions that result in reductions of one or
17 another or several air pollutants (constituting essentially a "natural experiment") are evaluated in
18 relation to changes in mortality or morbidity outcomes among population groups affected by the
19 reduction in air pollution exposure. To date, only a few epidemiological studies have evaluated
20 the consequences of interventions which allow for comparison of PM-health outcome
21 relationships before and after certain relatively discrete events resulting in notable changes in
22 ambient PM concentrations. Given that etiology of health outcomes related to PM or other air
23 pollutants are typically also affected by other risk factors, it is important in intervention studies
24 not only to measure air pollution exposure and health status before and after air pollution
25 reductions but also to identify and evaluate potential effects of other risk factors before and after
26 air pollution reductions.
27 The proposition that intervention studies can provide strong support for causal inferences
28 was emphasized by Hill (1965). In his classic monograph (The Environment and Disease:
29 Association or Causation?), Hill (1965) addressed the topic of preventive action and its
30 consequences under Aspect 8, stating:
31
June 2003 8-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 "Experiment: Occasionally it is possible to appeal to experimental, or semi-experimental,
2 evidence. For example, because of an observed association some preventive action is taken.
3 Does it in fact prevent? The dust in the workshop is reduced, lubricating oils are changed,
4 persons stop smoking cigarettes. Is the frequency of the associated events affected? Here the
5 strongest support for the causation hypothesis may be revealed."
6
7 8.1.3 Confounding and Effect Modification
8 A pervasive problem in the analysis of epidemiologic data, no matter what design or
9 strategy, is the unique attribution of the health outcome to the nominal causal agent (i.e.,
10 airborne particles in this document). The health outcomes attributed to particles are not specific
11 (for example, mortality in a broad range of [International Classification of Disease ] ICD-9
12 categories); and, as such, they may also be attributable to high or low temperatures, influenza
13 and other diseases, and/or exposure to other air pollutants. Many of the other factors can be
14 measured directly or by proxies. Some of these co-variables may be confounders and others
15 effect modifiers. The distinctions are important.
16 Confounding is "... 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
population (the population at risk from which the cases are derived).
22 (3) A confounding factor must not be affected by the exposure or the disease (i.e., it cannot
be an intermediate step in the causal path between the exposure and the disease).
23 Thus, the possible confounder should both be a risk indicator by itself and also be associated
24 with the exposure of interest in the study.
25 Causal events occur prior to some initial bodily response. A causal association may
26 usually be defined as an association in which alteration in the frequency or quality of one
27 category is followed by a change in the other. The concept of the chain mechanism is that many
28 variables may be related to a single effect through a direct-indirect mechanism. In fact, events
29 are not dependent on single causes. A given chain of causation may represent only a fraction of
June 2003 8-8 DRAFT-DO NOT QUOTE OR CITE
-------
1 a web (MacMahon and Pugh, 1970). A causal pathway refers to the network of relationships
2 among factors in one or more causal chains in which the members of the population are exposed
3 to causal agents that produce the observed health effect. The primary cause may be mediated by
4 secondary causes (possibly proximal to exposure) and may have either a direct effect on
5 exposure or an indirect effect through the secondary causes, or both, as illustrated below.
6 A non-causal pathway may involve factors that are not associated with the health effect or for
7 which there is no population exposure, so that the factors are not potential confounders.
8 The determination of whether a potential confounder is an actual confounder may be
9 elucidated from biological or physical knowledge about its exposure and health effects. Patterns
10 of association in epidemiology may be helpful in suggesting where to look for this knowledge,
11 but do not replace it. Gaseous criteria pollutants (CO, NO2, SO2, O3) are candidates for
12 confounders because all of these have at least some adverse health effects also associated with
13 particles (CO more often being associated with cardiovascular effects and the others with
14 respiratory effects, including symptoms and hospital admissions). In addition, the gaseous
15 criteria pollutants may be associated with particles for several reasons, including common
16 sources and correlated changes in response to wind and weather. Lastly, SO2 and NO2 may be
17 precursors to sulfate and nitrate components of ambient particle mixes, while NO2 contributes
18 also to the formation of organic aerosols during photochemical transformations.
19 The problem of disentangling the effects of other pollutants is especially difficult when
20 high correlation exists between one or more of them and ambient PM measurements.
21 A common source, such as combustion of gasoline in motor vehicles emitting CO, NO2, and
22 primary particles (and often resulting in high correlations), may play an important role in
23 confounding among these pollutants, as do weather and seasonal effects. Even though O3 is a
24 secondary pollutant also associated with emission of NO2, it is often more variably correlated
25 with ambient PM concentrations, depending on location, season, etc. Levels of SO2 in the
26 western U.S. are often quite low, so that secondary formation of particle sulfates plays a much
27 smaller role there, resulting in usually relatively little confounding of SO2 with PM mass
28 concentration in the West. On the other hand, in the industrial Midwest and northeastern states,
29 SO2 and sulfate levels during many of the epidemiology studies were relatively high and highly
30 correlated with fine particle mass concentrations, such that criterion 3 (no causal path leading
31 from confounder to exposure, or exposure to confounder to health effect) may not be strictly true
June 2003 8-9 DRAFT-DO NOT QUOTE OR CITE
-------
1 for SO2 versus sulfate or overall fine particle mass. If the correlation with PM and SO2 is not too
2 high, it may be possible to estimate some part of their independent effects which depend on the
3 assumption of independence under the particular model analyzed. If there is a causal pathway,
4 then it may be difficult to determine whether the observed relationship of exposure to health
5 effect is a direct effect of the exposure (to sulfate or fine PM in the example), an indirect effect
6 mediated by the potential confounder (i.e., exposure to SO2), or a mixture of these.
7 Consideration of additional (e.g., exposure, dosimetric, toxicologic) information beyond narrow
8 reliance on observed correlations among the PM measure(s), other pollutants, and health
9 outcome indicators is often useful in helping to elucidate the plausibility of PM or other
10 pollutants being causally related to statistically-associated health effects. As an example, of
11 much relevance is the extent to which the population in a community time-series study or the
12 participants in a prospective cohort study are exposed to measurable levels of the potential
13 confounder, particularly the ambient gaseous co-pollutants. If there is little or no exposure, then
14 the potential confounder does not satisfy the requirement that it is related to both exposure and
15 outcomes. This is discussed in Section 8.4 in connection with the role of exposure measurement
16 errors in air pollution epidemiology.
17 Some extraneous variables fall into the category of effect modifiers. "Effect-measure
18 modification differs from confounding in several ways. The main difference is that, whereas
19 confounding is a bias that the investigator hopes to prevent or remove from the effect estimate,
20 effect-measure modification is a property of the effect under study ... In epidemiologic analysis
21 one tries to eliminate confounding but one tries to detect and estimate effect-measure
22 modification." (Rothman and Greenland, 1998, p. 254). Examples of effect modifiers in some
23 of the studies evaluated in this chapter include environmental variables (such as temperature or
24 humidity in time-series studies), individual risk factors (such as education, cigarette smoking
25 status, age in a prospective cohort study), and community factors (such as percent of population
26 > 65 years old). It is often possible to stratify the relationship between health outcome and
27 exposure by one or more of these risk factor variables.
28 Effect modifiers may be encountered (a) within single-city time-series studies or (b) across
29 cities in a two-stage hierarchical model or meta-analysis. Figure 8-1 illustrates some
30 possibilities, using hypothetical examples with four cities in which a co-pollutant of the PM
31 index is to be evaluated as a possible effect modifier. In the examples in Figure 8-1, the
June 2003 8-10 DRAFT-DO NOT QUOTE OR CITE
-------
Q_
S-_
£
E
0)
"o5
tn
§
x
LU
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.
Q_
| ,
if
.1 0
IS
0) 1
tn
§
City 4
City 3
Co-Pollutant Concentration'
City 2
Cityl
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.
1 co-pollutant is assumed to have a relatively high positive correlation with the PM index. It is
2 also assumed that the excess relative risk for PM is calculated in a model in which PM is the
3 only air pollutant. For any given co-pollutant concentration within each city, there is likely to be
4 only a modest range of values of the PM index and the associated excess relative risk, as
June 2003
8-11
DRAFT-DO NOT QUOTE OR CITE
-------
1 suggested by the ellipses in Figure 8-1. The relationship between mortality and PM in
2 Figure 8-la is assumed to be the same and positive in all four cities; thus, with increasing
3 co-pollutant concentration within each city, the excess relative risk increases because the
4 co-pollutant is strongly correlated with the PM index. However, in the hypothetical 8-la, the
5 co-pollutant is not an effect modifier for PM, as can be shown by a regression of the estimated
6 mean PM effect on the mean co-pollutant concentration across the four cities.
7 In Figure 8-lb, the relationship between PM and mortality is assumed to differ across the
8 four cities, ranging from strongly negative in City 1 to strongly positive in City 4. Thus, with
9 increasing co-pollutant concentration within each city, the excess relative risk decreases in
10 City 1 and City 2 (but increases in City 3 and City 4) because the co-pollutant is strongly
11 correlated with the PM index. In Figure 8-lb, the co-pollutant is a hypothetical effect modifier
12 for PM, as can be shown by a regression of the estimated mean PM effect on the mean
13 co-pollutant concentration across the four cities, even though the simple mean of the excess
14 relative risks across the four cities is nearly zero. A relationship would be found if all within-
15 city effects were positive or if the across-city ecological regression were negative. Stratification
16 by levels of the putative effect modifier is also often useful.
17 Potential confounding (Figure 8-2a) is more difficult to identify and several statistical
18 methods are available, none of them being completely satisfactory. The usual methods include
19 the following:
20 Within a city:
21 (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;
22 (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;
23 Using data from several cities:
24 (C) Proceed as in Method A and pool the effect size estimates across cities for single-
and multi-pollutant models;
25 (D) Carry out a hierarchical regression of the PM effects versus the mean co-pollutant
concentration and determine if there is a relationship; and
June 2003 8-12 DRAFT-DO NOT QUOTE OR CITE
-------
Confounder
Figure 8-2a
, Outcomes
Exposure
Outcomes
Modifier
Figure 8-2b
Exposure
Outcomes
Primary Cause (s)
Secondary Cause(s) —• Exposure
Indirect Effect
Figure 8-2c
Outcomes
Secondary Cause 1
Primary Cause(s)
Secondary Cause 2
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.
2
3
(E) First carry out a regression of PM versus the co-pollutant concentration within each
city and the regression coefficient of mortality versus PM for each city. Then fit a
second-stage model regressing the mortality-PM coefficient versus the
PM-co-pollutant coefficient, concluding that the co-pollutant is a confounder if there
is an association at the second stage (See Figure 8-2c).
Each of the above methods (A through E) are subject to one or more disadvantages. The
multi-pollutant regression coefficients in method A, for example, may be unstable and have
June 2003
8-13
DRAFT-DO NOT QUOTE OR CITE
-------
1 greatly inflated standard errors, weakening their interpretation. In method B, the factors may be
2 sensitive to the choice of co-pollutants and the analysis method, and may be difficult to relate to
3 real-world entities. In method C, as with any meta-analysis, it is necessary to consider the
4 heterogeneity of the within-city effects before pooling them. Some large multi-city studies have
5 revealed unexpected heterogeneity, not fully explained at present. While method D is sometimes
6 interpreted as showing confounding if the regression coefficient is non-zero, this is an argument
7 for effect modification, not confounding. Method E is sensitive to the assumptions being made;
8 for instance, if PM is the primary cause in Figure 8-2c and the co-pollutant the secondary cause,
9 then the two-stage approach may be valid. However, if the model is mis-specified and there are
10 two or more secondary causes, some of which may not be identified, then the method may give
11 misleading results.
12 Given the wide array of considerations and possibilities discussed above, it is extremely
13 important to recognize that there is no single "correct" approach to modeling ambient PM-health
14 effects associations that will thereby provide the "right" answer with regard to precise
15 quantification of PM effect sizes for different health outcomes. Rather, it is clear that emphasis
16 needs to be placed here on (a) looking for convergence of evidence derived from various
17 acceptable analyses of PM effects on a particular type of health endpoint (e.g., total mortality,
18 respiratory hospital admissions, etc.); (b) according more weight to those well-conducted
19 analyses having greater power to detect effects and yielding narrower confidence intervals; and
20 (c) evaluating the coherence of findings across pertinent health endpoints and effect sizes for
21 different health outcomes. With regard to the latter, for example, the credibility of the overall
22 array of epidemiologically-demonstrated health effects being causally related to ambient PM
23 exposure is greatly enhanced to the extent that effect sizes for hospital admissions are larger than
24 those for PM-mortality effects and those for physician and emergency department visits are at
25 least as large as those for hospital admissions, and so on for respiratory symptoms, asthma
26 medication use, etc.
27 The issue of what PM effect sizes should be the main focus of presentation and discussion
28 in ensuing text - i.e., those derived from single-pollutant models including only PM or effect
29 sizes derived from multi-pollutant models that include one or more other copollutants along with
30 the PM indicator(s) - is an important one. Again, there is not necessarily any single "correct"
31 answer on this point. Implicit in arguments asserting that multi-pollutant model results must be
June 2003 8-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 reported and accorded equal or more weight than single-pollutant model PM results is
2 a functional construct that has generally been used in epidemiologic modeling of health effects
3 of air pollution, a functional construct that considers the various air pollutants mainly
4 independently of one another in terms of their health effects, which may not necessarily be the
5 case. This may be causing either over- or under-estimation of PM health effects, depending on
6 the modeling choices made by the investigator and the study situation. For example, ozone and
7 PM2 5 can share some similar oxidative formation and effect pathways in exerting adverse health
8 effects on the lung, yet are often modeled as independent pollutants or are placed in models
9 simultaneously, even though they may have high correlations over space and time and in their
10 health effects on the human body. Another complication is that other pollutants can be derived
11 from like sources and may serve less as a measure of direct effects than as a marker of pollution
12 from a specific source. As an example noted earlier, SO2 and PM2 5 are often predominantly
13 derived from the same sources in a locale (e.g., coal-fired power plants in the mid-western U.S.),
14 so that putting these two pollutants in a model simultaneously may cause a diminution of the
15 PM2 5 coefficient that may be misleading.
16 One approach that has been taken is to look at pollutant interactions (either multiplicative
17 or additive, depending on the model assumed), but until we understand (and appropriately
18 model) the biological mechanisms, such models are assumptions on the part of the researcher.
19 Present modeling practices represent the best methods now available and provide useful
20 assessments of PM health effects. However, ultimately, more biological-plausibility based
21 models are needed that more accurately model pollutant interactions and allow more
22 biologically-based interpretations of modeling results, rather than simply relying on a statistical
23 model specification or specific modeling criteria to determine the "winner" co-pollutant.
24 Until more is known about multiple pollutant interactions, it is important to avoid over-
25 interpreting model results regarding the relative sizes and significance of specific pollutant
26 effects, but instead to use biological plausibility in interpreting model results. For example, as
27 discussed later, Krewski et al (2000) found significant associations for both PM and SO2 in their
28 reanalysis for the Health Effects Institute of the ACS data set published by Pope et al. (1995).
29 Regarding these pollutant associations, they concluded that: "The absence of a plausible
30 toxicological mechanism by which sulfur dioxide could lead to increased mortality further
31 suggests that it might be acting as a marker for other mortality-associated pollutants." (Note:
June 2003 8-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 Annual mean SO2 averaged < 10 ppb across ca. 125 cities in ACS data set.) Rather than letting
2 statistical significance be the sole determinant of the "most important" pollutant, the authors
3 utilized biological plausibility to conclude which association was most likely driving the
4 pollution-health effects association in question. In the future, such biological
5 plausibility/mechanistic considerations need to be similarly considered in modeling and
6 weighing pollutant interactions in evaluating the health effects of PM. In the meantime, the
7 results from single-pollutant models of PM effects are emphasized here, as being those most
8 likely reflecting overall effects exerted by ambient PM either acting alone and/or in combination
9 with other ambient air pollutants.
10
11 8.1.4 GAM Convergence Issue
12 In the spring of 2002, the original investigators of a key newly available multi-city study
13 (the National Mortality and Morbidity Air Pollution Study; NMMAPS) cosponsored by the
14 Health Effects Institute (HEI) reported that use of the default convergence criteria setting used in
15 the GAM routine of certain widely-used statistical software (Splus) could result in biased
16 estimates of air pollution effects when at least two non-parametric smoothers are included in the
17 model (Health Effects Institute letter, May 2002). The NMMAPS investigators also reported
18 (Dominici et al., 2002), as determined through simulation, that such bias was larger when the
19 size of risk estimate was smaller and when the correlation between the PM and the covariates
20 (i.e., smooth terms for temporal trend and weather) was higher. While the NMMAPS
21 investigators reported that reanalysis of the 90 cities air pollution-mortality data (using stringent
22 convergence criteria) did not qualitatively change their original findings (i.e., the positive
23 association between PM10 and mortality; lack of confounding by gaseous pollutants; regional
24 heterogeneity of PM, etc.), the reduction in the PM10 risk estimate was apparently not negligible
25 (dropping, upon reanalysis, from 2.1% to 1.4% excess deaths per 50 |ig/m3 increase in PM10).
26 Issues surrounding potential bias in PM risk estimates from time-series studies using GAM
27 analyses and default convergence criteria were raised by EPA and discussed in July 2002 at the
28 CAS AC review of the Third External Review Draft of this PM AQCD. In keeping with a follow
29 up consultation with CASAC in August 2002, EPA encouraged investigators for a number of
30 important published studies to reanalyze their data by using GAM with more stringent
31 convergence criteria, as well as by using Generalized Linear Model (GLM) analyses with
June 2003 8-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 parametric smoothers that approximated the original GAM model. EPA, working closely with
2 HEI, also arranged for (a) the resulting reanalyses first to be discussed at an EPA-sponsored
3 open Workshop on GAM-Related Statistical Issues in PM Epidemiology held in November
4 2002; (b) then for any revamping of the preliminary analyses in light of the workshop
5 discussions; before (c) submittal by the investigators of short communications describing the
6 reanalyses approaches and results to EPA and HEI for peer-review by a special panel assembled
7 by HEI; and (d) the publication of the short communications on the reanalyses, along with
8 commentary by the HEI peer-review panel, in an HEI Special Report (2003a). Some of the
9 short-communications included in the HEI Special Report (2003a) included discussion of
10 reanalyses of data from more than one original publication because the same data were used to
11 examine different issues of PM-mortality associations (e.g., concentration/response function,
12 harvesting, etc.). In total, reanalyses were reported for more than 35 originally published
13 studies.
14
15 8.1.5 Ambient PM Increments Used to Report Risk Estimates
16 The effect of mortality from exposure to PM or other pollutants is usually expressed in this
17 document as a relative risk or risk rate (RR) relative to a baseline mortality or morbidity rate.
18 The crude mortality rates in 88 cities in 48 contiguous states in the NMMAPS study ranged from
19 about 8 deaths per day per million population in Denver, CO to about 40 per day per million in
20 St. Petersburg, FL. As reported in Samet et al. (2000a), there was little association between
21 PM10 effect size and crude mortality rate in the continental U.S. cities.
22 The PM increments used in this document to convert regression coefficients into
23 meaningful increments of excess risk are based on data from the U.S. fine particle monitoring
24 network for 1999 and 2001, the most recent years available. The difference between the annual
25 mean and the annual 95th percentile was used to characterize annual variation within each site;
26 and the average across all sites was used to select an appropriate increment for short-term
27 studies, about 50 |ig/m3 for PM10 and 25 |ig/m3 for PM2 5 and PM10_2 5, after rounding for ease of
28 calculation. The difference between the average of annual mean PM concentrations across all
29 sites and the average of the annual 95th percentiles across all sites was about 20 |ig/m3 for PM10
30 and 10 |ig/m3 for PM2 5 and PM10_2 5, values used here for PM increments in long-term studies.
June 2003 8-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 Thus, the pollutant concentration increments utilized here to report Relative Risks (RR's)
2 or Odds Ratio for various health effects are as follow for short-term (< 24 h) exposure studies:
3 50 |ig/m3 for PM10; 25 |ig/m3 for PM2 5 and PM10.2 5; 155 nmoles/m3 (15 |ig/m3) for SO4'2; and
4 75 nmoles/m3 (3.6 |ig/m3, if as H2SO4) for H+. The increments for short-term studies are the
5 same as were used in the 1996 PM AQCD, a choice now driven by more current data. In the
6 1996 PM AQCD, the same increments were used for the long- and short-term exposure studies.
7 However, for long-term exposure studies, 20 |ig/m3 is the increment used here for PM10 and
8 10 |ig/m3 for PM2 5 and PM10_2 5 for long-term exposure studies. These latter increments, derived
9 from new 1999-2001 data, are smaller than those used in the 1996 PM AQCD for long-term
10 studies.
11
12
13 8.2 MORTALITY EFFECTS ASSOCIATED WITH AIRBORNE
14 PARTICULATE MATTER EXPOSURE
15 8.2.1 Introduction
16 The relationship of PM and other air pollutants to excess mortality has been studied
17 extensively and represents an important issue addressed in previous PM criteria assessments
18 (U.S. Environmental Protection Agency, 1986, 1996a). Recent findings are evaluated here
19 mainly for the two most important epidemiology designs by which mortality is studied: time-
20 series mortality studies (Section 8.2.2) and prospective cohort studies (Section 8.2.3). The time-
21 series studies mostly assess acute responses to short-term PM exposure, although some recent
22 work suggests that time-series data sets can also be useful in evaluating responses to exposures
23 over a longer time scale. Time-series studies use community-level air pollution measurements to
24 index exposure and community-level response (i.e., the total number of deaths each day by age
25 and/or by cause of death). Prospective cohort studies usefully complement time-series studies;
26 they typically evaluate human health effects of long-term PM exposures indexed by community -
27 level measurements, using individual health records with survival lifetimes or hazard rates
28 adjusted for individual risk factors.
29
30
31
June 2003 8-18 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.2 Mortality Effects of Short-Term Particulate Matter Exposure
2 8.2.2.1 Summary of 1996 Particulate Matter Criteria Document Findings and Key Issues
3 The time-series mortality studies reviewed in the 1996 and other past PM AQCD's
4 provided much evidence that ambient PM air pollution is associated with increases in daily
5 mortality. The 1996 PM AQCD assessed about 35 PM-mortality time-series studies published
6 between 1988 and 1996. Of these studies, only five studies used GAM with default convergence
7 criteria. Recent reanalyses (Schwartz, 2003a; Klemm and Mason, 2003) using GAM with
8 stringent convergence criteria and other non-GAM approaches for one of these five studies, i.e.,
9 the Harvard Six cities time-series analysis (the only multi-city study among the five studies),
10 essentially confirmed the original findings. Thus, information provided in the 1996 PM AQCD
11 can be summarized without major concern with regard to the GAM convergence issue.
12 Information derived from those studies was generally consistent with the hypothesis that PM is a
13 causal agent in contributing to short-term air pollution exposure effects on mortality.
14 The PM10 relative risk estimates derived from short-term PM10 exposure studies reviewed
15 in the 1996 PM AQCD suggested that an increase of 50 |ig/m3 in the 24-h average of PM10 is
16 most clearly associated with an increased risk of premature total non-accidental mortality (total
17 deaths minus those from accident/injury) on the order of relative risk (RR) = 1.025 to 1.05 in the
18 general population or, in other words, 2.5 to 5.0% excess deaths per 50 |ig/m3 PM10 increase.
19 Higher relative risks were indicated for the elderly and for those with pre-existing
20 cardiopulmonary conditions. Also, based on the Schwartz et al. (1996a) analysis of Harvard Six
21 City data (as later confirmed in the reanalysis by Schwartz [2003a] and Klemm and Mason
22 [2003]), the 1996 PM AQCD found the RR (combined across the six cities) for excess total
23 mortality in relation to 24-h fine particle concentrations to be about 3% excess risk per 25 |ig/m3
24 PM2 5 increment.
25 While numerous studies reported PM-mortality associations, important issues needed to be
26 addressed in interpreting their findings. The 1996 PM AQCD evaluated in considerable detail
27 several critical issues, including: (1) seasonal confounding and effect modification; (2)
28 confounding by weather; (3) confounding by co-pollutants; (4) measurement error; (5) functional
29 form and threshold; (6) harvesting and life shortening; and (7) the role of PM components.
30 As important issues related to model specification became further clarified, more studies began
31 to address the most critical issues, some of which were at least partially resolved, whereas others
June 2003 8-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 required still further investigation. The next several paragraphs summarize the status of these
2 issues at the time of the 1996 PM AQCD publication.
3 One of the most important components in time-series model specification is adjustment for
4 seasonal cycles and other longer-term temporal trends. Residual over-dispersion and
5 autocorrelation result from inadequate control for these temporal trends, and not adequately
6 adjusting for them could result in biased RRs. Modern smoothing methods allow efficient fits of
7 temporal trends and reduce such statistical problems (it did introduce additional issues as
8 discussed in later sections). Most recent studies controlled for seasonal and other temporal
9 trends, and it was considered unlikely that inadequate control for such trends seriously biased
10 estimated PM coefficients. Effect modification by season was examined in several studies.
11 Season-specific analyses are often not feasible in small-sized studies (due to marginally
12 significant PM effect size), but some studies (e.g., Samet et al., 1996; Moolgavkar and Luebeck,
13 1996) suggested that estimated PM coefficients varied from season to season. It was not fully
14 resolved, however, whether these results represent real seasonal effect modifications or are due
15 to varying extent of correlation between PM and co-pollutants or weather variables by season.
16 While most available studies included control for weather variables, some reported
17 sensitivity of PM coefficients to weather model specification, leading some investigators to
18 speculate that inadequate weather model specifications may still have erroneously ascribed
19 residual weather effects to PM. Two PM studies (Samet et al., 1996; Pope and Kalkstein, 1996)
20 involved collaboration with a meteorologist and utilized more elaborate weather modeling, e.g.,
21 use of synoptic weather categories. Both of these studies used GAM, presumably with default
22 convergence criteria, and therefore need to be interpreted with caution. However, these studies
23 found that estimated PM effects were essentially unaffected by the synoptic weather variables
24 and also indicated that the synoptic weather model did not provide better model fits in predicting
25 mortality when compared to other weather model specifications used in previous PM-mortality
26 studies. Thus, these results suggested that the reported PM effects were not explained by more
27 sophisticated synoptic weather models.
28 Many earlier PM studies considered at least one co-pollutant in the mortality regression,
29 and some also examined several co-pollutants. In most cases, when PM indices were significant
30 in single pollutant models, addition of a co-pollutant diminished the PM effect size somewhat,
31 but did not eliminate the PM associations. When multiple pollutant models were performed by
June 2003 8-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 season, the PM coefficients became less stable, again, possibly due to PM's varying correlation
2 with co-pollutants among season and/or smaller sample sizes. However, in many studies, PM
3 indices showed the highest significance (versus gaseous co-pollutants) in single and multiple
4 pollutant models. Thus, it was concluded that PM-mortality associations were not seriously
5 distorted by co-pollutants, but interpretation of the relative significance of each pollutant in
6 mortality regression as relative causal strength was difficult because of limited quantitative
7 information on relative exposure measurement/characterization errors among air pollutants.
8 Measurement error can influence the size and significance of air pollution coefficients in
9 time-series regression analyses and is also important in assessing confounding among multiple
10 pollutants, as varying the extent of such error among the pollutants could also influence the
11 corresponding relative significance. The 1996 PM AQCD discussed several types of such
12 exposure measurement or characterization errors, including site-to-site variability and site-to-
13 person variability — errors thought to bias the estimated PM coefficients downward in most
14 cases. However, there was not sufficient quantitative information available to estimate such
15 bias.
16 The 1996 PM AQCD also reviewed evidence for threshold and various other functional
17 forms of short-term PM mortality associations. Several studies indicated that associations were
18 seen monotonically below the existing PM standards. It was considered difficult, however, to
19 statistically identify a threshold from available data because of low data density at lower ambient
20 PM concentrations, potential influence of measurement error, and adjustments for other
21 covariates. Thus, the use of relative risk (rate ratio) derived from the log-linear Poisson models
22 was considered adequate and appropriate.
23 The extent of prematurity of death (i.e., mortality displacement or "harvesting") in
24 observed PM-mortality associations has important public-health-policy implications. At the
25 time of the!996 PM AQCD review, only a few studies had investigated this issue. While one of
26 the studies suggested that the extent of such prematurity might be only a few days, this may not
27 be generalizable because this estimate was obtained for identifiable PM episodes. There was not
28 sufficient evidence to suggest the extent of prematurity for non-episodic periods from which
29 most of the recent PM relative risks were derived. The 1996 PM AQCD concluded:
30
31 In summary, most available epidemiologic evidence suggests that increased mortality results
32 from both short-term and long-term ambient PM exposure. Limitations of available evidence
June 2003 8-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 prevent quantification of years of life lost to such mortality in the population.
2 Life shortening, lag time, and latent period of PM-mediated mortality are almost certainly
3 distributed over long time periods, although these temporal distributions have not been
4 characterized, (p. 13-45)
5
6 Only a limited number of PM-mortality studies analyzed fine particles and chemically
7 specific components of PM. The Harvard Six Cities Study (Schwartz et al., 1996a) analyzed
8 size-fractionated PM (PM2 5, PM10/15, and PM10/15.2 5) and PM chemical components (sulfates and
9 H+). The results suggested that, among the components of PM, PM25 was most significantly
10 associated with mortality. Because the original study was conducted using GAM with default
11 convergence criteria, the data were recently reanalyzed by Schwartz (2003a), who reanalyzed
12 only PM25 and by Klemm and Mason (2003), who analyzed PM25, PM10/15, PM10/15.25, and
13 sulfate. Although the excess risk estimates were somewhat lower than those in the original
14 study, Klemm and Mason's reanalysis confirmed the original findings with regard to the relative
15 importance of fine versus coarse particles. While H+ was not significantly associated with
16 mortality in the original and an earlier analysis (Dockery et al., 1992), the smaller sample size
17 for H+ than for other PM components made a direct comparison difficult. The 1996 PM AQCD
18 also noted that mortality associations with BS or CoH reported in earlier studies in Europe and
19 the U.S. during the 1950s to 1970s most likely reflected contributions from fine particles, as
20 those PM indices had low 50% cut-points (< 4.5 jim). Furthermore, certain respiratory
21 morbidity studies showed associations between hospital admissions/visits with components of
22 PM in the fine particle range. Thus, the U.S. EPA 1996 PM AQCD concluded that there was
23 adequate evidence to suggest that fine particles play especially important roles in observed PM
24 mortality effects.
25 Overall, then, the status of key issues raised in the 1996 PM AQCD can be summarized as
26 follows: (1) the observed PM effects are unlikely to be seriously biased by inadequate statistical
27 modeling (e.g., control for seasonality); (2) the observed PM effects are unlikely to be seriously
28 confounded by weather (at least by synoptic weather models); (3) the observed PM effects may
29 be to some extent confounded or modified by co-pollutants, and such extent may vary from
30 season to season; (4) determining the extent of confounding and effect modification by co-
31 pollutants requires knowledge of relative exposure measurement characterization error among
32 pollutants (there was not sufficient information on this); (5) no clear evidence for any threshold
June 2003 8-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 for PM-mortality associations was reported (statistically identifying a threshold from existing
2 data was also considered difficult, if not impossible); (6) some limited evidence for harvesting,
3 a few days of life-shortening, was reported for episodic periods (no study was conducted to
4 investigate harvesting in non-episodic U.S. data); (7) only a relatively limited number of studies
5 suggested a causal role of fine particles in PM-mortality associations, but in the light of
6 historical data, biological plausibility, and the results from morbidity studies, a greater role for
7 fine particles than coarse particles was suggested in the 1996 PM AQCD as being likely. The
8 AQCD concluded:
9
10 The evidence for PM-related effects from epidemiologic studies is fairly strong, with most
11 studies showing increases in mortality, hospital admissions, respiratory symptoms, and
12 pulmonary function decrements associated with several PM indices. These epidemiologic
13 findings cannot be wholly attributed to inappropriate or incorrect statistical methods,
14 mis-specification of concentration-effect models, biases in study design or implementation,
15 measurement of errors in health endpoint, pollution exposure, weather, or other variables, nor
16 confounding of PM effects with effects of other factors. While the results of the
17 epidemiologic studies should be interpreted cautiously, they nonetheless provide ample
18 reason to be concerned that there are detectable human health effects attributable to PM at
19 levels below the current NAAQS. (p. 13-92)
20
21 8.2.2.2 Newly Available Information on Short-Term Mortality Effects
22 Since the 1996 PM AQCD, numerous new studies have examined short-term associations
23 between PM indices and mortality. Of these studies (over 80 studies), nearly 70% used GAM
24 (presumably with default convergence criteria). In the summer of 2002, U.S. EPA asked the
25 original investigators of some of these studies to reanalyze the data using GAM with more
26 stringent convergence criteria and GLM with parametric smoothers such as natural splines.
27 Because the extent of possible bias caused by the default criteria setting in the GAM models is
28 difficult to estimate for individual studies, the discussion here will focus only on those studies
29 that did not use GAM Poisson models and those studies that have reanalyzed data using more
30 stringent convergence criteria and/or alternative approaches. Newly available U.S. and Canadian
31 studies on relationships between short-term PM exposure and daily mortality that meet these
32 criteria are summarized in Table 8-1. More detailed summaries of all the short-term exposure
33 PM-mortality studies, including other geographic areas (e.g., Europe, Asia, etc) are described in
June 2003 8-23 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 8-1. RECENT U.S. AND CANADIAN TIME-SERIES STUDIES OF
PM-RELATED DAILY MORTALITY*
Reference
Type** Location(s)/period
Pollutants
Comments
Multi- City Mortality Studies in the U.S. and Canada
oo
to
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
PM10 studies using NMMAPS data
Samet et al. (2000a, b, c); A
Dominici et al. (2000a, b);
Samet (2000);
Dominici et al. (2003)
Daniels et al. (2000); A
Dominici et al. (2003)
Dominici et al. (2002)
Dominici et al. (2003)
Studies using every day PM10 data
Schwartz (2000a); A
Schwartz (2003b)
Schwartz (2000b);
Schwartz (2003b).
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
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.
Same ten U.S. cities as in
(Schwartz, 2000a)
PM10,03, CO,
NO2, SO2
PM10 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, Northeast regions, and
overall, but non-linear (usually concave) in the
other regions. Possible thresholds in Southeast.
Pooled PM10 (0 and 1 day lag average) mortality
estimates for the ten cities were presented.
Confounding and/or effect modification was
examined for season, co-pollutants, in- versus
out-of-hospital deaths.
Several pooled estimates across cities evaluated
for single day, moving average, and distributed
lags.
-------
to
O
o
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY*
Reference
Type** Location(s)/period
Pollutants
Comments
Multi- City Mortality Studies in the U.S. and Canada (cont'd)
oo
to
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Studies using every day PM10 data (cont 'd)
Bragaetal. (2001);
Schwartz (2003b)
Moolgavkar (2000a);
Moolgavkar (2003).
Laden etal.. (2000);
Schwartz (2003a)
Klemm et al., (2000);
Klemm and Mason
(2003)
Tsaietal. (1999, 2000)
A Same ten U.S. cities as in
(Schwartz, 2000a)
A Three large U.S. counties (cities):
Cook Co., IL; Los Angeles Co.,
CA; Maricopa Co., (Phoenix), AZ,
1987-1995 in the original analysis.
In the reanalysis, Maricopa Co. was
not analyzed.
A 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.
A Same six cities as (Laden etal.,
2000), 1979-1988.
B Camden, Elizabeth, and Newark,
NJ, 1981-1983.
PM10 only.
PM10 in all three; PM2 5 in
Los Angeles. O3, CO,
NO2, and SO2 in some
models. In the GAM
reanalysis, O3 was not
analyzed.
Chemically speciated
PM2 5 and factors aligned
with putative sources for
each city identified by
specific chemical
elements as tracers.
PM10, PM2.5, PM10.2.5,
sulfates
PM2 5, PM15, sulfates,
trace elements.
Pooled estimates across cities evaluated for
deaths due to pneumonia, COPD,
cardiovascular, and myocardial infarction using
distributed lags models.
Gaseous pollutants were at least as significantly
associated as PM indices. In particular, CO was
the best single index of air pollution association
with mortality in Los Angeles.
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.
Replicated Schwartz et al. (1996a) with
additional sensitivity analyses.
Significant effects of PM25, PM10, and sulfates
in Newark, Camden at most lags, but not
Elizabeth. Source-specific factors (oil burning,
automobiles) were also associated with
mortality.
-------
to
O
o
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY*
Reference
Type** Location(s)/period
Pollutants
Comments
Multi- City Mortality Studies in the U.S. and Canada (cont'd)
Studies using every day PM10 data (cont 'd)
Clyde et al. (2000)
Burnett et al. (2000);
Burnett and Goldberg
(2003)
B Phoenix, AZ, May, 1995- March,
1998. Seattle, WA, 1990-1995.
Eight Canadian cities: Montreal,
Ottawa, Toronto, Windsor, Calgary,
Edmonton, Winnipeg, Vancouver,
1986-1996.
PM2 5, PM10.2 5 in
Phoenix. PM10, PM25,
nephelometer, SO2 in
Seattle.
PM10, PM2.5, PM10.2.5,
sulfates, O3, CO, NO2,
SO,.
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.
The results of reanalysis indicate no clear
difference in association with mortality between
PM2 5 and PM10.2 5.
Single-City Mortality Studies in the U.S. and Canada
oo
to
a\
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Ostroetal. (1999a, 2000);
Ostro et al. (2003)
Fairley (1999);
Fairley (2003)
Schwartz etal. (1999)
Lippmann et al. (2000);
Ito (2003)
Chock et al. (2000)
Coachella Valley (Palm Springs),
CA, 1989-1998.
A Santa Clara County (San Jose), CA,
1989-1996.
B Spokane, WA, 1989-1995.
Detroit, MI, 1985-1990; 1992-1994
(separate analysis for two periods).
PM10 in earlier study,
PM10 (-65% of which was coarse particles) and
PM2 5 and PM10.2 5 in later PM10_2 5 (missing values predicted from PM10)
B
Pittsburgh, PA, 1989-1991.
study; O3, CO, NO2.
Reanalysis reported PM
risk estimates only.
PM10, PM2 5, PM10.2.5,
sulfates, nitrates, O3, CO,
NO2.
PM10 only.
PM10, PM2.5, PM10.2.5,
sulfates, acidity, TSP, O3,
CO, NO2, SO2
PM10,PM25,PM10.25,03,
CO, NO2, SO2
were associated with cardiovascular mortality.
PM2 5 was available for shorter period.
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
concentrations of crustal particles.
PM mass indices were more strongly associated
mortality than sulfate or acidity. The extent of
association with health outcomes was similar for
PM2 5 and PM10.25.
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
but not for other size fractions, likely because of
smaller sample size.
-------
to
O
o
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY*
Reference
Type** Location(s)/period
Pollutants
Comments
Single-City Mortality Studies in the U.S. and Canada (cont'd)
oo
to
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Klemm and Mason (2000)
Schwartz (2000c);
Schwartz (2003a)
Lipfert et al. (2000a)
Levy (1998)
Mar et al. (2000);
Mar et a. (2003)
Clyde et al. (2000)
Smith et al. (2000)
Gamble (1998)
B
A
B
Atlanta, GA, 1998-1999 (one year). PM10, PM2 5, PM10.2 5,
oxygenated hydrocarbons
(HC), elemental carbon
(EC),
organic carbon (OC),
Boston, MA, 1979-1986.
B Philadelphia, PA- Camden, NJ
seven- county area, 1995-1997.
B King County (Seattle), WA,
1990-1994.
Phoenix, AZ, near the EPA
platform monitor, 1995-1997.
B Phoenix, AZ, 1995-1997.
sulfates, acidity
PM25
PM10, PM2.5, PM10.2.5,
sulfates, acidity, metals,
O3, CO, NO2, SO2
PMj (nephelometer),
PM10, CO, SO2
PM10,PM25,PM10.25,
PM2 5 metals, EC, OC,
O3, CO, NO2, SO2, and
source-apportioned factor
scores.
PM2 5 and PM10.2 5
B Phoenix, AZ (within city and within PM2 5 and PM10_2 5
county), 1995-1997.
Dallas, TX, 1990-1994.
PM10, O3, CO, NO2, SO2
No significant effects likely due to short time-
series (ca. one year).
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 reduced PM coefficients.
PMj associated only with out- of- hospital
ischemic heart disease deaths; total mortality
with neither PM10 nor PM[
Only cardiovascular mortality was reanalyzed; it
was significantly
associated with PM10, PM2 5, PM10.2 5, EC, OC,
factors associated with motor vehicle,
vegetative-burning, and regional sulfate.
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, threshold for elderly mortality 20-
25 ug/m .
O3, CO, NO2 significantly associated with
mortality, PM10 and NO2 not associated
-------
to
O
o
TABLE 8-1 (cont'd). RECENT U.S. AND CANADIAN TIME-SERIES STUDIES
OF PM-RELATED DAILY MORTALITY*
Reference
Type** Location(s)/period
Pollutants
Comments
Single-City Mortality Studies in the U.S. and Canada (cont'd)
oo
to
oo
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Ostro (1995)
Murray and Nelson (2000)
B San Bernardino and Riverside
Counties, CA, 1980-1986.
B Philadelphia, PA, 1973-1990
PM2 5 estimated from
visual range, O3
TSP only
Neasetal. (1999)
Goldberg et al.
(2001a,b,c,d; 2003);
Goldberg and Burnett
(2003)
Ozkaynak et al. (1996)
B Philadelphia, PA
1973-1980
A
Montreal, PQ, Canada, 1984- 1995
TSP only
CoH and extinction were
available daily. PM2 5
and PM10 every sixth day
until 1992, daily through
1993.
B Toronto, ON, Canada 1970-1991 TSP, CoH, O3, CO, NO2,
SO,
Positive, significant PM2 s association only in
summer.
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 risk population.
Case- crossover study. Significant TSP
mortality associations reported.
Reanalysis indicated attenuation of PM risk
estimates, especially sensitive to weather model
specification. Congestive heart failure, as
classified based on medical records from
insurance plan, was associated with CoH, SO2,
and NO2.
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
most cause- specific deaths.
*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, 1996a). 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.
**Type: Type of studies: (A) Original study used GAM model including non-parametric smoothing terms with default or other lax convergence criteria, but
was reanalyzed using stringent convergence criteria and/or using parametric smoothers; (B) Original study used GLM with parametric smoothers or other
approaches, or used GAM but with only one non-parametric smoother.
-------
1 Appendix Table 8A-1. These include the studies that apparently used GAM with default
2 convergence criteria, and these studies are noted as such. Information on study location and
3 period, levels of PM, health outcomes, methods, results, and reported risk estimates and lags is
4 provided in Table 8A-1. In addition to these summary tables, discussion in the text below
5 highlights findings from several multi-city studies. Discussion of implications of new study
6 results for types of issues identified in foregoing text is mainly deferred to Section 8.4.
7 The summary of studies in Table 8-1 and 8A-1 (and in other tables) is not meant to imply
8 that all listed studies should be accorded equal weight in the overall interpretive assessment of
9 evidence regarding PM-associated health effects. In general, for those studies not clearly flawed
10 and having adequate control for confounding increasing scientific weight should be accorded to
11 in proportion to the precision of their estimate of a health effect. Small studies and studies with
12 an inadequate exposure gradient generally produce less precise estimates than large studies with
13 an adequate exposure gradient. Therefore, the range of exposures (e.g., as indicated by the IQR),
14 the size of the study as indexed by the total number of observations (e.g., days) and total number
15 of events (i.e., total deaths), and the inverse variance for the principal effect estimate are all
16 important indices useful in determining the likely precision of health effects estimates and in
17 according relative scientific weight to the findings of a given study. As can be seen in
18 Tables 8-1 and 8A-1, nearly all of the newly reported analyses with a few exceptions continue to
19 show statistically significant associations between short-term (24 h) PM exposures indexed by a
20 variety of ambient PM measurements and increases in daily mortality in numerous U.S. and
21 Canadian cities, as well as elsewhere around the world. Also, the effects estimates from the
22 newly reported studies are generally consistent with those derived from the 1996 PM AQCD
23 assessment, the newly reported PM risk estimates generally falling within the range of ca. 1 to
24 8% increase in excess deaths per 50 |ig/m3 PM10 and ca. 2 to 6% increase per 25 |ig/m3 PM2 5.
25 Several newly available PM epidemiologic studies that conducted time-series analyses in
26 multiple cities are of particular interest, as discussed below. Multi-city studies, such as the
27 NMMAPS study, avoid potential publication bias, because the cities were selected on the basis
28 of population size and the presence of PM monitoring data. In addition, because use of uniform
29 statistical analytical methods, findings cannot be attributed to different analytical approaches.
30
June 2003 8-29 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.2.3 New Multi-City Studies
2 The new multi-city studies are of particular interest here due to their evaluation of a wide
3 range of PM exposures and large numbers of observations holding promise of providing more
4 precise effects estimates than most smaller scale independent studies of single cities. Another
5 major advantage of the multi-city studies, over meta-analyses for multiple "independent" studies,
6 is the consistency in data handling and model specifications that eliminates variation due to
7 study design. Further, unlike regular meta-analysis, they clearly do not suffer from potential
8 omission of negative studies due to "publication bias." Furthermore, geographic patterns of air
9 pollution effects can be systematically evaluated in multiple-city analyses. Thus, the results
10 from multi-city studies can provide especially valuable evidence regarding the consistency
11 and/or heterogeneity, if any, of PM-health effects relationships across geographic locations.
12 Also, many of the cities included in these multi-city studies were ones for which no time-series
13 analyses had been previously reported. Most of these new multi-city studies used GAM Poisson
14 models, but the data sets have recently been reanalyzed using GAM models with more stringent
15 convergence criteria, as well as by GLM with parametric smoothers.
16
17 8.2.2.3.1 U.S. Multi-City Studies
18 U.S. PM10 90-Cities NMMAPS Analyses
19 The National Morbidity, Mortality, and Air Pollution Study (NMMAPS) focused on time-
20 series analyses of PM10 effects on mortality during 1987-1994 in the 90 largest U.S. cities
21 (Samet et al., 2000a,b), in the 20 largest U.S. cities in more detail (Dominici et al., 2000a), and
22 PM10 effects on emergency hospital admissions in 14 U.S. cities (Samet et al., 2000a,b). These
23 NMMAPS analyses are marked by extremely sophisticated statistical approaches addressing
24 issues of measurement error biases, co-pollutant evaluations, regional spatial correlation, and
25 synthesis of results from multiple cities by hierarchical Bayesian meta-regressions and
26 meta-analyses. These analyses provide extensive new information of much importance and
27 relevance to the setting of U.S. PM standards, because no other study has examined as many
28 U.S. cities in such a consistent manner. That is, NMMAPS used only one consistent PM index
29 (PM10) across all cities (noted PM10 samples were only collected every 6 days in most of the
30 90 cities); death records were collected in a uniform manner; and demographic variables were
31 uniformly addressed. The 90-cities analyses studies employ multi-stage models (see Table 8-1)
June 2003 8-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 in which heterogeneity in individual city's coefficients in the first stage Poisson models were
2 evaluated in the second stage models with city- or region-specific explanatory variables.
3 As noted earlier, the original investigators of the NMMAPS study reported in 2002 a
4 potential problem with using the GAM Poisson models with default convergence criteria
5 available in popular statistical software in estimating air pollution risks (Dominici et al., 2002).
6 The default convergence criteria were too lax to attain convergence in the setting of air pollution,
7 weather, and mortality/morbidity parameters where "small" PM regression coefficients were
8 estimated and at least two covariates were modeled with non-parametric smoothers. Their
9 simulation analysis also suggested that the extent of bias could be more serious when the
10 magnitude of risk coefficient was smaller and when PM's correlation with covariates was
11 stronger. The investigators since then reanalyzed the 90 cities data, using more stringent
12 convergence criteria as well as using fully parametric smoothers, and reported revised results.
13 The following description of the NMMAPS mortality study therefore focuses on the results of
14 the reanalysis of the 90 cities study.
15 In the original and reanalyzed 90 cities studies, the combined estimates of PM10
16 coefficients were positively associated with mortality at all the lags examined (0, 1, and 2 day
17 lags), although the 1-day lag PM10 resulted in the largest overall combined estimate. Figure 8-3
18 shows the reanalyzed results for the estimated percent excess total deaths per 10 |ig/m3 PM10 at
19 lag 1 day in the 88 (90 minus Honolulu and Anchorage) largest cities, as well as (weighted
20 average) combined estimates for U.S. geographic regions depicted in Figure 8-4. The majority
21 of the coefficients were positive for the various cities listed along the left axis of Figure 8-3. The
22 estimates for the individual cities were first made separately. The cities were then grouped into
23 the 7 regions seen in Figure 8-4 (based on characteristics of the ambient PM mix typical of each
24 region, as delineated in the 1996 PM AQCD). The bolded segments represent the posterior
25 means and 95% posterior intervals of the pooled regional effects without borrowing information
26 from other regions. The triangle and bolded segment at the bottom of Figure 8-3 display the
27 combined estimate of overall nationwide effects of PM10 for all the cities.
28 Note that there appears to be some regional-specific variation in the overall combined
29 estimates for all the cities in a given region. This can be discerned more readily in Figure 8-5,
30 which depicts overall region-specific excess risk estimates for 0, 1, and 2 day lags. For example,
31 the coefficients for the Northeast are generally higher than for other regions. The NMMAPS
June 2003 8-31 DRAFT-DO NOT QUOTE OR CITE
-------
% Increase in Mortality per 10 Unit Increase in
-4-202
I I I
^olt L^l'c Cit" o
K J J i
oi • ~~
Overall Ulympia
El Paso —
Overall ^^^^-.-,
overall
Minneapolis / St. Paul e
Wichita n
w
Overall _
Chicago^
Cleveland — e-
^t LAM!S ~~
it !ui jn jpuu j
Madison r
^
Overall
DJlLirnUi L-
I pjH,
JL-l jOj U-IL^
Nortolk n
vj
Anlnoton n
•_.
Overall
i ,X]_ jji u
°
Pahigh 'o
Little Rock 0
""" n
lr/m mva, jiifj-, ,-•.
Jackson n
Overall
Overall
I S I
Q Seattle
^^ ^mj 4.^
n Colorado Sonnas
u
-e — Spokane
•*-
Phnpni--'
u i tiuyni/v
"^ o ^u^tin Souit
/~'^v^^,,- /~-U^^.t.I
•fl-
9 — Los Angeles
^ nm, f- „! j
0 o Topeka
•-
~e — Detroit
e — Pittsburgh
Cincinnati indu
° o Lour-HIc m<^
r~ r.,4- •,!• /~, ,^ -
Q Evansvtile
_m_
e New York
n Pnrhoc1-Qr
,-, n i
0 Pra-idPnrP N°rt^
P| o,/rnr, .^.p
u -»v/rdCU^u
n Richmond
VJ
Q Kingston
. •
,-., Ll^-,. ,^.*-m^-.
^ D;^^-,i^.^,k-.^,
o o Dirmmgnam Sout
! 'J'JjIi v HiU
^ M^,,,. /^rlrt«^.^
^ j-5 t ,-^
or bt-ri.^Doro
n Shreveport
^
-•-
-»-
west
iwest
iern
ornla
>er
vest
strial
west
least
least
Figure 8-3. Estimated excess risks for PM mortality (1 day lag) for the 88 largest U.S.
cities as shown in the revised NMMAPS analysis.
Source: Dominici et al. (2002; 2003).
June 2003
8-32
DRAFT-DO NOT QUOTE OR CITE
-------
Northwest
Southern
California
Upper
Midwest
Industrial
Midwest
Northeast
Figure 8-4. Map of the United States showing the 88 cities (the 20 cities are circled) and
the seven U.S. regions considered in the NMMAPS geographic analyses.
1 investigators noted that the extent of the regional heterogeneity in the reanalysis result was
2 reduced slightly compared to the original finding (between-city standard deviation changed from
3 0.112 to 0.088 in the unit of percent excess deaths per 10 |ig/m3 PM10), but the pattern of
4 heterogeneity remained the same. The overall national combined estimate (i.e., at lag 1 day,
5 1.4% excess total deaths per 50 |ig/m3 increase in PM10 using GAM with stringent convergence
6 criteria) for the 90 cities is somewhat lower than the range of estimates for the cities reported in
7 the 1996PM AQCD.
8 In the original 90 cities study, the weighted second-stage regression included five types of
9 county- specific variables: (1) mean weather and pollution variables; (2) mortality rate (crude
10 mortality rate); (3) sociodemographic variables (% not graduating from high school and median
11 household income);(4) urbanization (public transportation); and (5) variables related to
12 measurement error (median of all pair-wise correlations between monitors). Some of these
13 variables were apparently correlated (e.g., mean PM10 and NO2, household income and
14 education) so that the sign of coefficients in the regression changed when correlated variables
15 were included in the model. Thus, while some of the county-specific variables were statistically
16 significant (e.g., mean NO2 levels), interpreting the role of these county-specific variables may
June 2003
8-33
DRAFT-DO NOT QUOTE OR CITE
-------
% Increase in Mortality per 10 Unit Increase in PM10
-1
-0.5
I
0.5
I
Northwest
NW.IagO
NW.Iag2
NW.Iagl
Southwest
S W.I agO
SW.Iag2
Southern
California
SC.IagO
SC.Iag2
SC.Iagl
Upper UMW.iagl
Midwest
UMW.IagO
UMW.IagZ
Industrial
Midwest
IM W.I agO
IWlW.lag2
IMW.Iagl
Northeast NE.Iagl
NE.IagO
NE.Iag2
Southeast
SE.IagO
SE.Iag2
SE.Iagl
Overall Overall.lag 1
Overall.lagO
Overall.Iag2
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-ug/m3 increases in PM10
concentrations in cities aggregated within U.S. regions shown in Figure 8-4.
Source: Dominici et al. (2002; 2003).
June 2003
8-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 require caution. Regarding the heterogeneity of PM10 coefficients, the investigators concluded
2 that they "did not identify any factor or factors that might explain these differences."
3 Another important finding from Samet and coworkers' analyses was the weak influence of
4 gaseous co-pollutants on the PM10 effect size estimates (see Figure 8-6). In the reanalysis of
5 90 cities data, PM10 coefficients slightly increased when O3 was added to regression models.
6 Additions of a third pollutant (i.e., PM10 + O3 + another gaseous pollutant) hardly changed the
7 posterior means of PM10 effect size estimates, but widened the distribution. However, the
8 posterior probabilities that the overall PM10 effects are greater than zero remained at or above
9 0.96. The gaseous pollutants themselves in single-, two-, and three-pollutant models were less
10 consistently associated with mortality than PM10. Ozone was not associated with mortality using
11 year-round data; but, in season-specific analyses, it was associated with mortality negatively in
12 winter and positively in summer. SO2, NO2, and CO were weakly associated with mortality, but
13 additions of PM10 and other gaseous pollutants did not always reduce their coefficients, possibly
14 suggesting their independent effects. As noted in Section 8.1, CO and NO2 from motor vehicles
15 are likely confounders of PM25 and, thus, of PM10 when it is not dominated by the coarse particle
16 fraction. The investigators stated that the PM10 effect on mortality "was essentially unchanged
17 with the inclusion of either O3 alone or O3 with additional pollutants."
18 The reanalyses of the 90 cities data by the original NMMAPS investigators also included a
19 sensitivity analysis of lag Iday PM10 GLM results to the alternative degrees of freedom for
20 adjustment of the confounding factors: season, temperature, and dewpoint. The degrees of
21 freedom for each of these three smoothing terms was either doubled or halved, resulting in nine
22 scenarios in addition to the degrees of freedom in the original GLM model. The PM10 effect
23 posterior means were generally higher when the degrees of freedom were halved for season, and
24 lower when they were doubled, ranging between 1.6% to 0.9% (the main GLM result was 1.1%)
25 excess total mortality per 50 |ig/m3 PM10 increase. These results underscore the fact that the
26 magnitude of sensitivity of the results due to model specification (in this case, degrees of
27 freedom alone) can be as great as the potential bias caused by the GAM convergence problem.
28 HEI (2003a) states that the revised NMMAPS 90 individual-city mortality results show
29 that, in general, the estimates of PM effect are shifted downward and the confidence intervals are
30 widened. In the revised analyses, a second stage meta-analysis was used to combine results on
31 effects of PM and other pollutants on health outcomes across cities. Tightening the convergence
June 2003 8-35 DRAFT-DO NOT QUOTE OR CITE
-------
0.0 0.2S OJ O.T5
% in per 10 pg/m9 in PM.
'10
Figure 8-6. 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: Dominici et al. (2003).
1
2
3
4
5
criteria in GAM obtained a substantially lower estimate of effect of PM10 combined over all
cities, and use of GLM with natural splines decreased the estimate further. The revised analyses
yielded a small, but statistically significant, effect of PM10 at lag 1 on total mortality, now esti-
mated to be 0.21% per 10 |ig/m3, with a posterior standard error of 0.06%. HEI (2003a) agrees
with the investigators' conclusions that the qualitative conclusions of NMMAPS II have not
changed although the evidence for an effect of PM10 at lag 0 and lag 2 is less convincing under
the new models. The NMMAPS II report found that the PM10 effect remained when copollutants
were introduced into the model (Samet et al., 2000a); and this conclusion has not changed.
June 2003
8-36
DRAFT-DO NOT QUOTE OR CITE
-------
1 The extent of reduction in PM10 excess risk estimate due to the change in the convergence
2 criteria (2.3% per 50 |ig/m3 PM10 using default versus 1.4% using stringent) using GAM models
3 in the 90 cities study appears to be greater than those reported in most of other reanalysis studies.
4 This may be in part due to the smaller risk estimate (2.3%) in the original study compared to
5 other studies (> 3%), as the smaller coefficient is likely more strongly affected as a relative
6 reduction. This may also be in part due to the more "aggressive" adjustment for possible
7 weather effects (discussed later) used in this study, which may have increased the concurvity
8 between PM and the covariates (which included four smoothing terms for weather adjustment).
9 Dominici et al. (2002) reported that the higher the concurvity, the larger the potential bias that a
10 GAM model with default convergence criteria could produce.
11 In summary, the 90-cities NMMAPS study provides extremely useful information
12 regarding the following: (1) the magnitude of combined PM10 risk estimate; (2) the lack of
13 sensitivity of PM10 risk estimates to gaseous co-pollutants; (3) indications of some regional
14 heterogeneity in PM10 risk estimates across the U.S.; (4) the shape of concentration-response
15 relationship (discussed in a later section); and (5) the range of sensitivity of PM10 risk estimates
16 to the extent of smoothing of covariates in their original weather model specification. One major
17 uncertainty that has not been examined in this study is the sensitivity of the PM10 risk estimates
18 to different weather model specifications (e.g., use of two temperature terms, rather than four).
19
20 U.S. 10-Cities Studies
21 In another set of multi-city analyses, Schwartz (2000a,b), Schwartz and Zanobetti (2000),
22 Zanobetti and Schwartz (2000), Braga et al. (2000), and Braga et al. (2001) analyzed 1987-1995
23 air pollution and mortality data from ten U.S. cities (New Haven, CT; Birmingham, AL;
24 Pittsburgh, PA; Detroit, MI; Canton, OH; Chicago, IL; Minneapolis-St. Paul, MN; Colorado
25 Springs, CO; Spokane, WA; and Seattle, WA.) or subsets (4 or 5 cities) thereof. The selection of
26 these cities was based on the availability of daily (or near daily) PM10 data. All of these original
27 studies utilized GAM Poisson models with default convergence criteria. Of these studies,
28 Schwartz (2003) reanalyzed the data from Schwartz (2000a), Schwartz (2000b), and Braga et al.
29 (2001) using GAM with stringent convergence criteria as well as alternative models such as
30 GLM with natural cubic splines or penalized splines, both of which are expected to give correct
31 standard errors. The main original results of the study were presented in the Schwartz (2000a)
June 2003 8-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 paper; and the other studies noted above focused on each of several specific issues, including
2 potential confounding, effect modification, distributed lag, and threshold. In this section, the
3 results for the three reanalysis studies noted above are discussed.
4 In the reanalysis (Schwartz, 2003b) of the main results (Schwartz, 2000a), daily total (non-
5 accidental) mortality in each of the 10 cities was fitted using a GAM Poisson model (with
6 stringent convergence criteria) or a GLM Poisson model with natural splines, adjusting for
7 temperature, dewpoint, barometric pressure, day-of-week, season, and time. The data were also
8 analyzed by season (November through April as heating season). The inverse-variance weighted
9 averages of the ten cities' estimates were used to combine results. PM10 (average of lag 0 and 1
10 days) was significantly associated with total deaths, and the effect size estimates were
11 comparable in summer and winter. Adjusting for other pollutants did not substantially change
12 the PM10 effect size estimates. The combined percent-excess-death estimate for total mortality
13 was 3.4% (95% CI = 2.6 - 4.1) per 50 |ig/m3 increase in the average of lag 0 and 1 days PM10
14 (essentially unchanged from the original study) using GAM with stringent convergence criteria.
15 The PM10 risk estimate using GLM with natural splines was 2.8% (95% CI = 2.0 - 3.6).
16 In the reanalysis (Schwartz, 2003b) of the study of multi-day effects of air pollution
17 (Schwartz, 2000b), constrained (quadratic model over 0 through 5 day lags) and unconstrained
18 (0 through 5 day lags) distributed lag models were fitted in each city. The overall estimate was
19 computed using the inverse-variance weighted average of individual city estimates. Among the
20 results obtained using GAM with stringent convergence criteria, the PM10 effect size estimate
21 was 6.3% (95% CI = 4.9 - 7.8) per 50 |ig/m3 increase for the quadratic distributed lag model,
22 and 5.8% (95% CI = 4.4 - 7.3) for the unconstrained distributed lag model. Corresponding
23 values using the penalized splines were somewhat smaller (~ 5.3%). These values are about
24 twice the effect-size estimate for single-day PM10 in the original report or the two-day mean
25 PM10 reported in the reanalysis above (this reanalysis did not report results for single-day or 2-
26 day mean PM10). These results suggest a possibility that PM effects may be underestimated
27 when only single-day PM indices are used.
28 Schwartz (2003b) also reanalyzed the data from Braga et al.'s (2001) study to examine the
29 lag structure of PM10 association with specific cause of mortality in the 10 cities. Unconstrained
30 distributed lags for 0 through 5 days as well as two-day mean were fitted in each city for COPD,
31 pneumonia, all cardiovascular, and myocardial infarction deaths using GAM with stringent
June 2003 8-38 DRAFT-DO NOT QUOTE OR CITE
-------
1 convergence criteria and penalized spline models. Combined estimates by lag were obtained
2 across the 10 cities. The distributed lag estimates were generally larger than the two-day mean
3 estimates for COPD and pneumonia mortality, but they were comparable for all cardiovascular
4 and myocardial infarction mortality. For example, in the results using GAM with stringent
5 convergence criteria, the PM10 effect size estimate was 11.0% (95% CI = 7.2 - 14.8) per
6 50 |ig/m3 increase for two-day mean model, and 16.8% (95% CI = 8.3 - 25.9) for the
7 unconstrained distributed lag model. Note that these values are substantially larger than those
8 reported for total non-accidental deaths.
9 The PM10 risk estimates from these 10 cities studies appear to be larger than those from the
10 90 cities study. Aside from the difference in the number of cities analyzed, the difference in
11 weather model specification and the extent of smoothing for temporal trends may have
12 contributed to the difference in the size of PM10 risk estimates. This issue is further discussed in
13 Section 8.2.2.3.5.
14
15 Reanalyses of Harvard Six Cities Study
16 Both the original Harvard Six Cities Study time-series analysis (Schwartz et al., 1996a) and
17 the replication analysis by Klemm et al. (2000), which essentially replicated Schwartz et al.'s
18 original findings, used GAM Poisson models with default convergence criteria. Schwartz
19 (2003a) and Klemm and Mason (2003) conducted reanalyses of the Harvard Six Cities data to
20 address the GAM statistical issues.
21 Schwartz (2003a) reported the risk estimates for PM25 only, but provided results using
22 several other spline smoothing methods (natural splines, B-splines, penalized splines, and thin
23 plate splines) in addition to GAM with stringent convergence criteria. The risk estimate
24 combined across the six cities per 25 |ig/m3 in PM25 (average of lag 0 and 1 day) using GAM
25 with stringent convergence criteria was 3.5% (95% CI = 2.5 - 4.5), as compared to the original
26 value of 3.7% (95% CI = 2.7 - 4.7). The corresponding value from a GLM model with natural
27 splines was 3.3% (95% CI = 2.2 - 4.3). The values using B-splines, penalized splines, and thin
28 plate splines were somewhat lower (3.0%, 2.9%, and 2.6%, respectively). However, when the
29 Harvard Six Cities were examined individually in the reanalysis of Schwartz using GLM and
30 penalized splines, Boston and St. Louis gave significant associations with PM25 and Steubenville
31 gave a significant association with coarse PM.
June 2003 8-39 DRAFT-DO NOT QUOTE OR CITE
-------
1 Klemm and Mason's reanalysis (2003) reported risk estimates for PM25, PM10_2 5, PM10
2 (PM15 or PM10), and SO4"2. They also conducted sensitivity analyses using GLM with natural
3 splines that approximated the degrees of freedom used in the LOESS smoothers in the GAM
4 models, as well as 12 knots per year and 4 knots per year for smoothing of temporal trends. The
5 PM2 5 and PM10_2 5 total non-accidental mortality risk estimates combined across the six cities per
6 25 |ig/m3 (average of lag 0 and 1 day) using GAM with stringent convergence criteria were 3.0%
7 (95% CI = 2.1 - 4.0) and 0.8% (95% CI = -0.5, 2.0), respectively. The corresponding PM10
8 mortality excess risk estimate per 50 |ig/m3 (average of lag 0 and 1 day) was 3.6% (95% CI =
9 2.1, 5.0). In their sensitivity analysis, increasing the degrees of freedom for temporal trends for
10 natural splines in GLM models from 4 knots/year to 12 knots/year markedly reduced PM risk
11 estimates. For example, the PM2 5 risk estimate per 25 |ig/m3 was reduced from 2% in the 4
12 knots/year model to 1% in the 12 knots/year model. The results showing the smaller PM risk
13 estimates for larger degrees of freedom for smoothing of temporal trends are consistent with
14 similar findings reported for the reanalysis of 90 cities study.
15 Although PM effect estimates from the Klemm and Mason (2003) reanalysis are somewhat
16 smaller than those from Schwartz (2003; e.g., 3.5% by Schwartz versus 3.0% by Klemm and
17 Mason for PM25 using strict convergence criteria), the results are essentially comparable. Both
18 studies also showed that the comparable GLM models produced smaller risk estimates than
19 GAM models.
20
21 U.S. 3-Cities Study
22 Moolgavkar (2000a) evaluated associations between short-term measures of major air
23 pollutants and daily deaths in three large U.S. metropolitan areas (Cook Co., IL, encompassing
24 Chicago; Los Angeles Co., CA; and Maricopa Co., AZ, encompassing Phoenix) during a 9-year
25 period (1987-1995). Moolgavkar (2003) reanalyzed the data for Cook Co. and Los Angeles Co.,
26 but not Maricopa Co. using GAM with stringent convergence criteria as well as GLM with
27 natural splines. Ozone was analyzed in the original analysis but not in the reanalysis (it was only
28 positive and significant in Cook county in the original analysis). This section describes the
29 results from the reanalysis. Total non-accidental deaths, deaths from cardiovascular disease
30 (CVD) and chronic obstructive lung disease (COPD) were analyzed in relation to 24-h readings
31 for PM, CO, NO2, and SO2 averaged over all monitors in a given county. Cerebrovascular
June 2003 8-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 mortality was analyzed in the original analysis but not in the reanalysis (its association with air
2 pollution was weak in the original analysis). The results of cause-specific mortality analyses are
3 described in a later section. Daily readings were available for each of the gaseous pollutants in
4 both Cook Co. and Los Angeles Co., as were PM10 values for Cook Co. However, PM10 and
5 PM2 5 values were only available every sixth day in Los Angeles Co. PM values were highest in
6 summer in Cook Co. and in the winter and fall in Los Angeles Co.; whereas the gases (except for
7 O3) were highest in winter in both counties. The PM indices were moderately correlated
8 (r = 0.30 to 0.73) with CO, NO2, and SO2 in Cook Co. and Los Angeles Co. Total
9 non-accidental, CVD, and COPD deaths were all highest during winter in both counties.
10 Adjusting for temperature and relative humidity effects in separate analyses for each
11 mortality endpoint for these two counties, varying patterns of results were found, as noted in
12 Table 8A-1. Moolgavkar (2003) also reported sensitivity of results to different degrees of
13 freedom (df) for smoothing of temporal trends (30 df and 100 df).
14 As for Cook Co. results, PM10 was significantly associated with total non-accidental
15 mortality at lag 0 (most significant) and 1 day in GAM models with both 30 df andlOO df for
16 smoothing of temporal trends, as well as in a GLM model with 100 df for smoothing of temporal
17 trends. The gaseous pollutants were also significantly associated with total non-accidental
18 mortality at various lags (wider lags than PM10), but most significant at lag 1 day. These
19 associations did not appear to be sensitive to the extent of smoothing for temporal trends, at least
20 at their most significant lags. In two pollutant models (results were not shown in tables but
21 described in text), the PM10 association remained "robust and statistically significant" at lag 0
22 day; whereas the coefficients for the gases became non-significant. However, at lag 1 day, the
23 PM10 association became non-significant and the gases remained significant. Thus, some extent
24 of "sharing" of the association is apparent, and whichever pollutant is more strongly associated
25 than the other at that lag tended to prevail in the two pollutant models in this data set.
26 For Los Angeles Co., CO was more significantly associated (positive and significant at lag
27 0 through 3 days) with mortality than PM10 (positive and significant at lag 2) or PM2 5 (positive
28 and significant at lag 1). In two pollutant models in which CO and PM indices were included
29 simultaneously at PM indices' "best" lags, CO remained significant; whereas PM coefficients
30 became non-significant (and negative for cases with 30 df for temporal smoothing). For Los
31 Angeles data, the PM coefficients appeared to be more sensitive to the choice of the degrees of
June 2003 8-41 DRAFT-DO NOT QUOTE OR CITE
-------
1 freedom than to the default versus stringent convergence criteria. GLM models tended to
2 produce smaller risk estimates than GAM models. Moolgavkar also reported that these
3 associations were robust to varying the extent of smoothing for weather covariates.
4 The results for these two cities do not reflect a common pattern. In Cook Co., all the
5 pollutants were associated with mortality, and their relative importance varied depending on the
6 lag day; whereas CO showed the strongest mortality associations in Los Angeles. Moolgavkar
7 concluded that, considering the substantial differences that can result from different analytic
8 strategies, no particular numeric estimates were too meaningful, although the patterns of
9 associations appeared to be robust.
10
11 8.2.2.3.2 Canadian Multicity Studies
12 Burnett et al. (2000) analyzed various PM indices (PM10, PM25, PM10_25, sulfate, CoH, and
13 47 elemental component concentrations for fine and coarse fractions) and gaseous air pollutants
14 (NO2, O3, SO2, and CO) for association with total mortality in the 8 largest Canadian cities:
15 Montreal, Ottawa-Hull, Toronto, Windsor, Winnipeg, Calgary, Edmonton, and Vancouver. This
16 study differs from Burnett et al. (1998a) in that it included fewer cities but more recent years of
17 data (1986-1996 versus 1980-1991) and detailed analyses of particle mass components by size
18 and elemental composition. Each city's mortality, pollution, and weather variables were
19 separately filtered for seasonal trends and day-of-week patterns. The residual series from all
20 cities were then combined and analyzed in a GAM Poisson model. In Burnett and Goldberg's
21 reanalysis (2003) of the eight cities data, they only examined the PM indices PM25, PM10_25, and
22 PM10 using GAM models with more stringent convergence criteria. The reanalysis used co-
23 adjustment regression (i.e., simultaneous regression), rather than the regression with pre-filtered
24 data that was the main approach of the original analysis. The reanalysis also considered several
25 sensitivity analyses including models with and without day-of-week adjustment and several
26 alternative approaches (fitting criteria and extent of smoothing) to adjust for temporal trends
27 using natural splines.
28 Adjusting for temporal trends, smoothing of same-day temperature, pressure, and day-of-
29 week effects, the pooled PM effect estimates across the eight Canadian cities were: 3.7% (95%
30 CI = 1.4-6.0) per 25 |ig/m3 increase in PM25; 2.1% (0.1-4.2) per 25 |ig/m3 increase PM10.25; and
31 3.6% (95% CI = 1.3-5.8) per 50 |ig/m3 increase PM10. These effect size estimates are fairly close
June 2003 8-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 to the estimates reported in the original study, despite the differences in the regression approach
2 (pre-filtering and GAM with default convergence criteria in the original study versus co-
3 adjustment and using GAM with stringent convergence criteria). The temporal adjustment of the
4 above model used LOESS smoothing with span of approximately 0.022 (= 90 days/4012 study
5 days). Sensitivity analysis included several choices of degrees of freedom for natural splines of
6 temporal trend, with two fitting criteria (i.e., Bartlett's test for white noise and AIC) and either
7 using the same degrees of freedom for all the eight cities or varying degrees of freedom for each
8 city. The PM risk estimates based on natural splines were generally smaller than those based on
9 LOESS smoothers. The PM risk estimates also varied inversely with the number of knots for
10 temporal trend. That is, the more details of the temporal trend were described by natural splines,
11 the smaller the PM risk estimates became. The reported PM2 5 risk estimates per 25 |ig/m3
12 increase were 3.0% (t=3.12), 2.8% (t=2.28), 2.2% (t=2.14), 2.1% (t=2.07), and 1.9% (t=1.72) for
13 knot/year, knot/6 months, knot/3 months, knot/2 months, and knot/1 month, respectively. The
14 corresponding values for 25 |ig/m3 increase in PM10.25 were 3.9% (t=3.42), 2.9% (t=2.52), 2.1%
15 (t=1.69), 1.8% (t=1.46), and 1.2% (t=0.91), suggesting greater sensitivity of PM10.25 risk
16 estimates to the extent of temporal smoothing. The authors suggested that this was likely due to
17 the stronger correlation between (and temporal trends in) mortality and mass concentrations for
18 PM10_2 5 (average correlation among cities of-0.45) than for PM2 5 (-0.36). Because the relative
19 significance and size of PM25 and PM10_25 risk estimates varied depending on the model and
20 extent of smoothing for temporal trend, it is difficult to determine the relative importance of the
21 two size-fractionated PM indices in this study.
22
23 8.2.2.3.3 European Multi-City APHEA Study Analyses
24 The Air Pollution and Health: A European Approach (APHEA) project is a multi-center
25 study of short-term effects of air pollution on mortality and hospital admissions within and
26 across a number of European cities having a wide range of geographic, climatic,
27 sociodemographic, and air quality patterns. The obvious strength of this approach is its ability to
28 evaluate potential confounders or effect modifiers in a consistent manner. It should be noted that
29 PM indices measured in those cities varied. In APHEA1, the PM indices measured were mostly
30 black smoke (BS), except for Paris, Lyon (PM13); Bratislava, Cologne, and Milan (TSP); and
31 Barcelnoa (BS and TSP). In APHEA2, 10 out of the 29 cities used actual PM10 measurements;
June 2003 8-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 and, in 11 additional cities, PM10 levels were estimated based on regression models relating
2 collocated PM10 measurements to BS or TSP. In the remaining 8 cities, only BS measurements
3 were available (14 cities had BS measurements). As discussed below, there have been several
4 papers published that present either a meta-analysis or pooled summary estimates of these multi-
5 city mortality results: (1) Katsouyanni et al. (1997) — SO2 and PM results from 12 cities; (2)
6 Touloumi et al. (1997) — ambient oxidants (O3 and NO2) results from six cities; (3) Zmirou
7 et al. (1998) — cause-specific mortality results from 10 cities (see Section 8.2.2.5); (4) Samoli
8 et al. (2001) — a reanalysis of APHEA1 using a different model specification (GAM) to control
9 for long-term trends and seasonality; and (5) Katsouyanni et al. (2001) — APHEA2, with
10 emphasis on the examination of confounding and effect modification. The original APHEA
11 protocol used sinusoidal terms for seasonal adjustment and polynomial terms for weather
12 variables in Poisson regression models. Therefore, publications 1 through 3 above are not
13 subject to the GAM default convergence issue. Publications 4 and 5 did use GAM Poisson
14 model with default convergence criteria, but the investigators have reanalyzed the data using
15 GAM with more stringent convergence criteria, as well as GLM with natural splines (Katsouyani
16 et al., 2003; Samoli et al., 2003). The discussions presented below on publications 4 and 5 are
17 focused on the results from the reanalyses.
18
19 APHEA1 Sulfur Dioxide and Particulate Matter Results for 12 Cities
20 The Katsouyanni et al. (1997) analyses evaluated data from the following cities: Athens,
21 Barcelona, Bratislava, Cracow, Cologne, Lodz, London, Lyons, Milan, Paris, Poznan, and
22 Wroclaw. In the western European cities, an increase of 50 |ig/m3 in SO2 or BS was associated
23 with a 3% (95% CI = 2.0, 4.0) increase in daily mortality; and the corresponding figure was 2%
24 (95% CI = 1.0, 3.0) for estimated PM10 (they used conversion: PM10 = TSP*0.55). In the 31
25 central/eastern European cities, the increase in mortality associated with a 50 ji g/m3 change was
26 0.8% (CI = 0.1, 2.4) for SO2 and 0.6% (CI = 0.1, 1.1) per 50 |ig/m3 change in BS. Estimates of
27 cumulative effects of prolonged (two to four days) exposure to air pollutants were comparable to
28 those for one day effects. The effects of both pollutants (BS, SO2) were stronger during the
29 summer and were mutually independent. Regarding the contrast between the western and
30 central/eastern Europe results, the authors speculated that this could be due to differences in
31 exposure representativeness; differences in pollution toxicity or mix; differences in proportion of
June 2003 8-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 sensitive sub-population; and differences in model fit for seasonal control. Bobak and Roberts
2 (1997) commented that the heterogeneity between central/eastern and western Europe could be
3 due to the difference in mean temperature. However, Katsouyanni and Touloumi (1998) noted
4 that, having examined the source of heterogeneity, other factors could apparently explain the
5 difference in estimates as well as or better than temperature.
6
7 APHEA1 Ambient Oxidants (Ozone and Nitrogen Dioxide) Results for Six Cities
8 Touloumi et al. (1997) reported on additional APHEA data analyses, which evaluated
9 (a) short-term effects of ambient oxidants on daily deaths from all causes (excluding accidents),
10 and (b) impacts on effect estimates for NO2 and O3 of including a PM measure (BS) in
11 multi-pollutant models. Six cities in central and western Europe provided data on daily deaths
12 and NO2 and/or O3 levels. Poisson autoregressive models allowing for overdispersion were
13 fitted. Significant positive associations were found between daily deaths and both NO2 and O3.
14 Increases of 50 |ig/m3 in NO2 (1-hour maximum) or O3 (1-hour maximum) were associated with
15 a 1.3% (95% CI = 0.9-1.8) and 2.9% (95% CI = 1.0-4.9) increase in the daily mortality,
16 respectively. There was a tendency for larger effects of NO2 in cities with higher levels of BS:
17 when BS was included in the model, the coefficient for NO2 was reduced by half (but remained
18 significant) whereas the pooled estimate for the O3 effect was only slightly reduced. The authors
19 speculated that the short-term effects of NO2 on mortality might be confounded by other vehicle-
20 derived pollutants (e.g., airborne ambient PM indexed by BS measurements). Thus, while this
21 study reports only relative risk levels for NO2 and O3 (but not for BS), it illustrates the
22 importance of confounding of NO2 and PM effects and the relative limited confounding of O3
23 and PM effects.
24
25 APHEA1: A Sensitivity Analysis for Controlling Long-Term Trends and Seasonality
26 The original study (Samoli et al., 2001) attempted to examine the sensitivity of APHEA1
27 results to how the temporal trends were modeled (i.e., sine/cosine in the APHEA1 versus LOESS
28 smoother using GAM with default convergence criteria). Samoli et al. (2003) reanalyzed the
29 data using GAM with more stringent convergence criteria, as well as GLM with natural splines.
30 Thus, the reanalysis allowed a comparison of results across a fixed functional model
31 (sine/cosine), a non-parametric smoother (GAM with LOESS), and a parametric smoother (GLM
June 2003 8-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 with natural splines). The combined estimate across cities for percent excess in total non-
2 accidental mortality per 50 |ig/m3 increase in BS using GAM with stringent convergence criteria
3 (2.3%; 95% CI = 1.9-2.7) was bigger than that using sine/cosine (1.3%; 95% CI = 0.9-1.7). The
4 GAM with stringent convergence criteria reduced the combined estimate by less than 10%
5 compared to that from GAM with default convergence criteria. The corresponding estimate
6 using GLM with natural splines (1.2%; 95% CI = 0.7-1.7) was comparable to that from the
7 sine/cosine model but smaller than that using GAM. The contrast between western and eastern
8 Europe in the original APHEA1 study (2.9% for west versus 0.6% for east) was less clear in the
9 results using GAM with stringent convergence criteria (2.7% versus 2.1%) or GLM with natural
10 splines (1.6% versus 1.0%). These results indicate that the apparent regional heterogeneity
11 found in the original APHEA1 study could be sensitive to model specification. Because the
12 number of cities used in the APHEA1 study is relatively small (eight western and five central-
13 eastern cities), the apparent regional heterogeneity found in the earlier publications could also be
14 due to chance. These reanalysis results also suggest that the results are somewhat sensitive to
15 the model specification of temporal trends.
16
17 APHEA2: Confounding and Effect Modification Using Extended Data
18 The APHEA2 original study (Katsouyanni et al. 2001) included more cities (29 cities) and
19 a more recent study period (variable years in 1990-1997, as compared to 1975-1992 in
20 APHEA1). Also, the APHEA2 original study used a GAM (with default convergence criteria)
21 Poisson model with LOESS smoothers to control for season and trends. Katsouyanni et al.
22 (2003) reanalyzed the data using GAM with more stringent convergence criteria, as well as two
23 parametric approaches: natural splines and penalized splines. Because the reanalysis GAM
24 results changed the PM10 risk estimates only slightly from the original estimates and the
25 investigators mention that the patterns of effect modification were preserved in their reanalyses
26 regardless of model specification, the qualitative description of the effect modification below
27 relies on the original study. The PM10 estimates for various models are from the reanalysis
28 results.
29 The analyses put emphasis on effect modification by city-specific factors. Thus, the city-
30 specific coefficients from the first stage of Poisson regressions were modeled in the second stage
31 regression using city-specific characteristics as explanatory variables. Inverse-variance
June 2003 8-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 weighted pooled estimates (fixed-effects model) were obtained as part of this model. When
2 substantial heterogeneity was observed, the pooled estimates were obtained using random-effects
3 models. These city-specific variables included (1) air pollution level and mix, such as average
4 air pollution levels and PM/NO2 ratio (as an indicator of traffic-generated PM); (2) climatic
5 variables, such as mean temperature and relative humidity; (3) health status of the population,
6 such as the age-adjusted mortality rates, the percentage of persons over 65 years of age, and
7 smoking prevalence; and (4) geographic area (three regions: central-eastern, southern, and
8 north-western). The study also addressed the issue of confounding by simultaneous inclusion of
9 gaseous co-pollutants in city-specific regressions and obtained the pooled PM estimates for each
10 co-pollutant included. Unlike APHEA1, in which the region (larger PM estimates in western
11 Europe than in central-eastern Europe) was highlighted as the important factor, APHEA2 found
12 several effect modifiers. NO2 (i.e., index of high pollution from traffic) was an important one.
13 The cities with higher NO2 levels showed larger PM effects as did the cities with a warmer
14 climate. The investigators noted that this might be due to the better estimation of population
15 exposures with outdoor community monitors (because of more open windows). Also, the cities
16 with low standardized mortality rate showed larger PM effects. The investigators speculated that
17 this may be because a smaller proportion of susceptible people (to air pollution) are available in
18 a population with a large age-standardized mortality rate. Interestingly, in the pooled PM risk
19 estimates from models with gaseous pollutants, it was also NO2 that affected (reduced) PM risk
20 estimates most. For example, in the fixed-effects models, approximately 50% reductions in both
21 PM10 and BS coefficients were observed when NO2 was included in the model. SO2 only
22 minimally reduced PM coefficients; whereas O3 actually increased PM coefficients. Thus, in
23 this analysis, NO2 was implicated both as a confounder and an effect modifier. The overall
24 random-effects model combined estimate for total mortality for 50 |ig/m3 increase in PM10 were
25 3.0% (95% CI = 2.0, 4.1), 2.1% (95% CI = 1.2, 3.0), and 2.8% (95% CI = 1.8, 3.8), for GAM
26 (stringent convergence criteria), natural splines, and penalized splines models, respectively. The
27 original estimate using GAM with default convergence criteria (3.1%) was thus reduced by 4%.
28 While the effect estimates varied somewhat depending on the choice of GAM with LOESS,
29 natural splines, or penalized splines, the investigators reported that the patterns of effect
30 modification (by NO2, etc.) were preserved.
31
June 2003 8-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.2.3.4 Comparison of Effects Estimates from Multi-City Studies
2 Based on different pooled analyses of data combined across multiple cities, the percent
3 excess (total, non-accidental) deaths estimated per 50 |ig/m3 increase in PM10 in the above multi-
4 city studies were (1) 1.4% using GAM (1.1% using GLM) at lag 1-day in the 90 largest U.S.
5 cities (the Northeast region results being about twice as high); (2) 3.4% using GAM (2.8% using
6 GLM) for average of 0 and 1 day lags in 10 U.S. cities; (3) 3.6% using GAM (2.7% using GLM)
7 for 1 day lag PM10 in the 8 largest Canadian cities; and (4) 3.0% using GAM (2.1% using GLM)
8 in APHEA2 for average of 0 and 1 day lags for 29 European cities during 1990-1997.
9 Note that the estimate for the NMMAPS 90 cities study is somewhat smaller than those for
10 the rest of the multi-city studies and the range reported in the previous PM AQCD (2.5 to 5%).
11 There may be several possible explanations for this, but model specification for weather is likely
12 one major factor. The 90 cities study used much more "aggressive" adjustment for possible
13 weather effects than most studies. The 90 cities analysis included four separate weather terms:
14 (1) smoothing splines (natural splines when GLM was used) of same-day temperature with
15 6 degrees of freedom; (2) smoothing splines of the average of lag 1 through 3 day temperature
16 with 6 degrees of freedom; (3) smoothing splines of same-day dewpoint with 3 degrees of
17 freedom; and, (4) smoothing splines of the average of lag 1 through 3 day dewpoint with
18 3 degrees of freedom. In contrast, most of the other studies used only one or two terms for
19 weather variables. For example, the Harvard Six Cites Study used a LOESS smoother (or
20 natural splines or other smoothers in reanalysis) of same-day temperature with a span of 0.5 and
21 aLOESS smoother of same-day dewpoint with a span of 0.5. Note that the 90 cities study not
22 only used more terms for weather effects, but it also used more degrees of freedom for
23 temperature than Schwartz et al.'s analysis (according to Klemm and Mason's reanalysis, the
24 span of 0.5 in LOESS corresponds to approximately 3.5 degrees of freedom). It should also be
25 noted here that the purpose of the inclusion of dewpoint in these models is often explained as "to
26 adjust for possible effects of humidity"; but, in fact, dewpoint and temperature are highly
27 correlated (r > 0.9) in most cities. Thus, although the inclusion of these terms may statistically
28 (i.e., by AIC, etc.) provide a better fit, the epidemiologic implications of the use of these terms is
29 not yet clear. While extreme temperature, hot or cold, is known to cause excess mortality, it is
30 not clear at this time whether these models are adequately modeling the weather effects in the
31 more moderate range (which is much of the data). Thus, the inclusion in the NMMAPS
June 2003 8-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 modeling of several weather terms with more degrees of freedom most likely provides
2 "conservative" PM risk estimates. That is, the NMMAPS excess risk estimates of 1.1% or 1.4%
3 per 50 |ig/m3 PM10 increase may well underestimate the PM10-total mortality effect-size
4 suggested by two other well conducted multicity studies to fall in the range of 2.7% to 3.6% per
5 50 |ig/m3 PM10 increment for U.S. and Canadian cities.
6 Another factor that may contribute to the difference in PM risk estimates is the extent of
7 smoothing to adjust for temporal trends. Several of the reanalysis studies (Dominici et al., 2002;
8 Burnett and Goldberg, 2003; Ito, 2003; Klemm and Mason, 2003; Molgavkar, 2003) consistently
9 reported, though to varying extents, that using more degrees of freedom for temporal trends
10 tended to reduce PM coefficients. That is, when more details in the short-term fluctuations of
11 mortality were ascribed to temporal trends, PM risk estimates were reduced. For example, in
12 Dominici et al.'s (2002) sensitivity analysis, the PM10 risk estimate was larger (1.6% per
13 50 |ig/m3 increase in PM10) for the GLM model with 3 degrees of freedom per year that the
14 estimate using 7 degrees of freedom (1.1%). Note that, in general, the presumed objective of
15 including temporal trends in the mortality regression is to adjust for potential confounding
16 (measured or unmeasured) by time-varying factors that change seasonally or in shorter time
17 spans (e.g., influenza epidemics). However, ascribing "too short" temporal fluctuations to these
18 "confounding temporal trends" may inadvertently take away PM effects. Because the "right"
19 extent of smoothing is not known, these sensitivity analyses are useful. In the reanalyses
20 mentioned above, the PM risk estimates could change by a factor of two when a range of degrees
21 of freedom was applied even for a model specification in which all the other terms were kept
22 unchanged.
23 Based on the results from the reanalysis studies, it has become apparent that different
24 smoothing approaches can also affect PM risk estimates. For example, the models with natural
25 splines (parametric smoothing) appear, in general but not always, to result in smaller PM risk
26 estimates than GAM models with LOESS or smoothing splines. GAM models may possibly
27 suffer from biased standard error of risk estimates, but they also seem to fit the data better (i.e.,
28 based on AIC) than GLM models with natural splines. Thus, it is not clear which smoothers
29 provide the most appropriate PM risk estimates. In any case, the choice of these smoothers does
30 not seem to affect PM risk estimates (~ 10 to 30%) as much as the range of weather model
June 2003 8-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 specifications or the range of the degrees of freedom for temporal trends adjustment do (as large
2 as a factor of two).
3 A less explored issue is the effect of multi-day effects of PM. The PM10 risk estimates
4 summarized above are either for a single-day lag (U.S. 90 cities study, Canadian 8 cities study,
5 and APHEA1), or an average of two days (U.S. 10 cities study and APHEA2). However, the
6 reanalysis of U.S. 10 cities study data suggests that the multi-day PM effect, accounting for 0
7 through 5 day lag, could be twice as large as the effect sizes estimated from single or two-day
8 average models and even bigger (~ 3 to 4 fold) when more specific cause of death categories
9 were examined. This issue warrants further investigation.
10 In summary, considering all the options in model specifications that can affect the PM risk
11 estimates, the reported combined PM10 total non-accidental mortality risk estimates from multi-
12 city studies are in good agreement, in the range of 1.0 to 3.5% per 50 |ig/m3 increase in single or
13 two-day average PM10. The U.S. 90 cities study provides estimates towards the lower end of this
14 range. Combinations of choices in model specifications (the number of weather terms and
15 degrees of freedom for smoothing of mortality temporal trends) alone may explain the extent of
16 the difference in PM10 risk estimates across studies. The range for these newly available
17 combined estimates from multi-cities studies overlap with the range of PM10 estimates (2.5 to
18 5%, obtained from single cities studies) previously reported in the 1996 PM AQCD, but extends
19 to somewhat lower values.
20
21 8.2.2.4 The Role of Particulate Matter Components
22 Delineation of the roles of specific ambient PM components in contributing to associations
23 between short-term PM exposures and mortality requires evaluation of several factors, e.g., size,
24 chemical composition, surface characteristics, and the presence of gaseous co-pollutants. While
25 possible combinations of these factors can in theory be limitless, the actual data tend to cover
26 definable ranges of aerosol characteristics and co-pollutant environments due to typical source
27 characteristics (e.g., fine particles tend to be combustion products in most cities). Newly
28 available studies conducted in the last few years have begun to provide more extensive
29 information on the roles of PM components; and their results are discussed below in relation to
30 three topics: (1) PM particle size (e.g., PM25 versus PM10_25); (2) chemical components; and
31 (3) source oriented evaluations.
June 2003 8-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 The ability to compare the relative roles of different PM size fractions and various PM
2 constituents is restricted by the limitations of the available studies. Comparisons nevertheless
3 can be attempted, using such information as the relative level of significance and/or the strength
4 of correlation between component estimate and health outcome. The relative significance across
5 cities/studies is influenced by the sample size and the level of the pollutants. The width of the
6 confidence band also needs to be taken into account, according more weight for studies with
7 narrower confidence bands. Caution in interpretation of such information, however, is warranted
8 because of potential measurement error and possible high correlations between indices being
9 compared. Additionally, limitations of single-city studies must be recognized.
10
11 8.2.2.4.1 Particulate Matter Particle Size Evaluations
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 only one acute mortality study (Schwartz et al., 1996a) had examined this issue. That
15 study (which used GAM with default convergence criteria in analyzing Harvard Six-City study
16 data) suggested that fine particles (PM2 5), distinctly more so than coarse fraction (PM10_2 5)
17 particles, were associated with daily mortality. Recent reanalyses using GAM with more
18 stringent convergence criteria have yielded only slightly smaller PM25 effect-size estimates
19 (Schwartz et al., 2003). It should also be noted that (a) the Klemm et al. (2000) reanalysis
20 reconstructed the data and replicated the original analyses (using GAM with default convergence
21 criteria) and (b) the Klemm and Mason (2003) reanalysis, using GAM with stringent
22 convergence criteria and GLM with parametric smoothers, also essentially reproduced the
23 original investigators' results.
24 Since the 1996 PM AQCD, several new studies have used size-fractionated PM data to
25 investigate the relative importance of fine (PM25) versus coarse (PM10_25) fraction particles.
26 Table 8-2 provides synopses of those studies with regard to the relative importance of the two
27 size fractions, as well as some characteristics of the data. The average levels of PM25 ranged
28 from about 13 to 30 |ig/m3 in the U.S. cities, but much higher average levels were measured in
29 Santiago, Chile (64.0 |ig/m3 ). As can be seen in Table 8-2, in the northeastern U.S. cities
30 (Philadelphia, PA and Detroit, MI), there was more PM25 mass than PM10_25 mass on the
31 average; whereas in the western U.S. (Phoenix, AZ; Coachella Valley, CA; Santa Clara County,
June 2003 8-51 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 (jig/m3); ratio
ofPM25toPM10;and
correlation between
PM2 5 and PM10_2.5
Results regarding relative importance of
PM25 versus PM10_2S and comments.
Fairley (1999 &
2003)*
Santa Clara
County, CA
Ostro et al.
(2000 & 2003)*
Coachella
Valley, CA
Clyde et al.
(2000) Phoenix,
AZ
Mar et al.
(2000 & 2003)*
Phoenix, AZ
1995-1997
Smith et al.
(2000)
Phoenix, AZ
Lippmann et al.
(2000);
Ito, (2003)*
Detroit, MI
1992-1994
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;
PM2 5/PM10 = 0.30;
r=0.65.
PM2 5 (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;
PM2 5/PM10 =0.58;
r=0.42.
Of the various pollutants (including PM10, PM25, PM10_25,
sulfates, nitrates, CoH, CO, NO2, and O3), the 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. Winter PM2 5 level is more
than twice that in summer. The daily number of O3 ppb-hours
above 60 ppb was also significantly associated with mortality.
Coarse particles dominate PM10 in this locale. PM2 5 was
available only for the last 2.5 years; and a predictive model could
not be developed, so that a direct comparison of PM2 5 and
PM10_2 5 results is difficult. Cardiovascular mortality was
significantly associated with PM10 (and predicted PM10_2 5),
whereas PM2 5 was mostly negatively (and not significant) at the
lags examined.
Using the Bayesian Model Averaging that incorporates model
selection uncertainty with 29 covariates (lags 0- to 3-day), the
effect of coarse particle (most consistent at lag 1 day) was
stronger than that for fine particles. The association was for
mortality defined for central Phoenix area where fine particles
(PM2 5) are expected to be uniform.
Cardiovascular mortality was significantly associated with both
PM2 5 (lags 1, 3, and 4) and PM10_25 (lag 0) with similar effect size
estimates. Of all the pollutants (SO2, NO2, and elemental carbon
were also associated), CO was most significantly associated with
cardiovascular mortality.
In linear PM effect model, the authors found a statistically
significant mortality association with PM10_2 5, but not with PM2 5.
In the models allowing for a threshold, they found evidence of a
threshold for PM2 5 (in the range of 20-25), but not for PM10.2 5.
A seasonal interaction in the PM10_2 5 effect was also reported: the
effect is highest in spring and summer when the anthropogenic
concentration of PM10_2 5 is lowest.
Both PM2 5 and PM10_2 5 were positively (but not significantly)
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.
Lipfert et al.
(2000a)
Philadelphia, PA
1992-1995.
PM25mean=17.3;
PM25/PM10 =0.72.
The authors conclude that no systematic differences were seen
according to particle size or chemistry. However, when PM2 5
and PM10_25 were compared, PM25 (at lag 1 or average of lag 0
and 1) was more significantly (with larger attributable risk
estimates) associated with cardiovascular mortality than PM10_2 5.
June 2003
8-52
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 (jig/m3); ratio of
PM2 5 to PM10; and
correlation between
PM2 5 and PM10.2 5
Results regarding relative importance of
PM2S versus PM10_2s and comments
Klemm and
Mason (2000)
Atlanta, GA
Klemm et al.
(2000); Klemm
and Mason
(2003)*
6 U.S. cities
Chock et al.
(2000)
Pittsburgh, PA
PM25mean= 19.9;
PM2 5/PM10 = 0.65
Mean PM2 s ranges from
11.3 to 29.6;
MeanPM10_25 ranges
from 6.6 to 16.1;
Mean PM2 5/PM10 ranges
from 50.1% to 66% in
the six cities.
Data distribution not
reported.
PM2 5/PM10 = 0.67
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.
This reanalysis of the Harvard Six-Cities time-series analysis by
Schwartz et al. (1996a) found significant associations between
total mortality and PM2 5 in 3 cities and in pooled effect, but no
significant association with PM10_2 5 in the reanalysis of the
replication study for any city. These results essentially confirmed
the findings of the original study by Schwartz et al. (1996a).
Seasonal dependence of correlation among pollutants, multi-
collinearity among pollutants, and instability of coefficients
were all emphasized in discussion and conclusion. These
considerations and the small size of the data set (stratified by age
group and season) limit confidence in finding of no consistently
significant associations for any size fractions.
Burnett et al.
(2000); Burnett
and Goldberg
(2003)*
8 Canadian
cities
Cifuentes et al.
(2000)
Santiago, Chile
1988-1996
PM25 mean=13.3;
PM25/PM10=0.51;
r=0.37.
PM25mean=64.0;
PM2 5/PM10 =0.58;
r=0.52.
Both PM2 5 and PM10_2 5 were significantly associated with total
non-accidental mortality. Results using varying extent of
smoothing of mortality temporal trends show that there is no
consistent pattern of either PM mass index being more important.
The authors note that PM10_2 5 was more sensitive to the type of
smother and amount of smoothing.
In GLM results for the whole years, only PM2 5 and NO2 were
consistently significantly associated with total non-accidental
mortality.
Note: * next to author name indicates that the study was originally analyzed using GAM models only with default
convergence criteria using at least two non-parametric smoothing terms.
1 CA) the average PM10_2 5 levels were higher than PM2 5 levels. It should be noted that the three
2 Phoenix studies in Table 8-2 use much the same data set; all used fine and coarse particle data
3 from EPA's 1995-1997 platform study. Seasonal differences in PM component levels should
4 also be noted. For example, in Santa Clara County and in Santiago, Chile, winter PM2 5 levels
5 averaged twice those during summer. The temporal correlation between PM2 5 and PM10_2 5
6 ranged between 0.30 and 0.65. Such differences in ambient PM mix features from season to
June 2003
8-53
DRAFT-DO NOT QUOTE OR CITE
-------
1 season or from location to location complicates assessment of the relative importance of PM25
2 and PM10_2 5.
3 To facilitate a quantitative overview of the effect size estimates and their corresponding
4 uncertainties from these studies, the percent excess risks are plotted in Figure 8-7. These
5 excluded the Clyde et al. study (for which the model specification did not obtain RRs for PM2 5
6 and PM10_2 5 separately) and the Smith et al. study (which did not present linear term RRs for
7 PM25 and PM10_25). Note that, in most of the original studies, the RRs were computed for
8 comparable distributional features (e.g., interquartile range, mean, 5th -10-95* percentile, etc.).
9 However, the increments derived and their absolute values varied across studies; therefore, the
10 RRs used in deriving the excess risk estimates delineated in Figure 8-7 were re-computed for
11 consistent increments of 25 |ig/m3 for both PM25 and PM10_25. Note also that re-computing the
12 RRs per 25 |ig/m3 in some cases changed the relative effect size between PM2 5 and PM10_2 5, but
13 it did not affect the relative significance. All of the studies found positive associations between
14 both the fine and coarse PM indices and increased mortality risk. However, most of the studies
15 did not have large enough sample sizes to separate out what often appear to be relatively small
16 differences in effect size estimates; but two of the studies do show distinctly larger mortality
17 associations with PM25 than for non-significant PM10_2 5 effects. For example, the Klemm et al.
18 (2000) and Klemm and Mason's (2003) re-computation of the Harvard Six Cities time-series
19 study reconfirmed the original Schwartz et al. (1996a) finding that PM2 5 was significantly
20 associated with excess mortality, but PM10_2 5 across all cities was not (although the Schwartz
21 [2003a] reanalyses reconfirmed the original findings of statistically significant PM10_2 5-mortality
22 relationship in Steubenville, OH). Similar findings of PM2 5 being significantly associated with
23 mortality were obtained in Santa Clara County (Fairley, 1999; Fairley 2003). Two studies
24 suggested that PM10_2 5 was more important than PM2 5: Coachella Valley, CA (Ostro et al., 2000
25 & 2003) and Phoenix, AZ (Clyde et al., 2000). There were five studies in which the importance
26 of PM2 5 and PM10_2 5 were considered to be similar or, at least, not distinguishable: Philadelphia,
27 PA (Lipfert et al., 2000a); Detroit, MI (Lippmann et al., 2000; reanalysis by Ito 2003); Phoenix,
28 AZ (Mar et al., 2000 and reanalysis in 2003); Eight Canadian cities (Burnett at al., 2000;
29 reanalysis by Burnett and Goldberg, 2003); and Santiago, Chile (Cifuentes et al., 2000).
30 In the reanalysis (Burnett and Goldberg, 2003) of the Canadian 8-city study (Burnett et al.,
31 2000), the relative importance of PM2 5 and PM10_2 5 was not clear, but both PM indices were
June 2003 8-54 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
oo
H
6
o
o
H
O
O
H
W
O
O
HH
H
W
Klemm and Mason (2003)*
Harvard 6 Cities -
(recomputed)
Burnett and Goldberg (2003)*
8 Canadian Cities
Chock et al. (2000)
Pittsburgh, PA ~
Klemm and Mason (2000)
Atlanta, GA ~
Lipfert et al. (2000a)
Philadelphia, PA ~
Ito (2003)*
Detroit, Ml
Mar et al. (2003)*
Phoenix, AZ
Fairley (2003)*
Santa Clara Co., CA
Cifuentes et al.
Santiago, Chile
-2
i
Percent excess death (total unless otherwise noted) per
25 |jg/m3 increase in PM2.5 (•) or PM-io-2.5 (O)
0 2 4 6 8 10 12 14 16 18
Lag 0+1 day
Lag 1 day
Lag 0 day
Lag 0 day
Lag 1 day Cardiovascular mortality
Lag 3day
Lag 1 day"
Lag 1 day
Lag 0 day
Lag 0 day
Lag 1+2 day
e-
-Q-
} age less than 75
} age 75 and over
-e-
-e-
-e-
Cardiovascular mortality
-e-
-e-
-e-
Figure 8-7. Percent excess risks estimated per 25 jig/m3 increase in PM2 5 or PM10_2 5 from new studies
evaluating both PM2 5 and PM10_2 5, based on single pollutant (PM only) models. The asterisk
next to reference indicates reanalysis of data using GLM with natural splines. Other studies
used GLM or OLS.
-------
1 significant in single pollutant models. In GAM models (stringent convergence criteria) with
2 LOESS smoothers, PM25 was more significant and showed larger risk estimates than PM10_2 5.
3 However, in sensitivity analysis in which varying degrees of freedom for mortality temporal
4 trends were applied in GLM models, the effect size and significance for these PM indices were
5 often comparable. The authors commented that PM10_2 5 coefficient was more sensitive to the
6 extent of temporal smoothing than PM2 5.
7 The Lippmann et al. (2000) results and a reanalysis (Ito, 2003) for Detroit are also
8 noteworthy in that additional PM indices were evaluated besides those depicted in Figure 8-6,
9 and the overall results obtained may be helpful in comparing fine- versus coarse-mode PM
10 effects. In analyses of 1985 to 1990 data, PM-mortality relative risks and their statistical
11 significance were generally in descending order: PM10, TSP-SO4"2, and TSP-PM10. For the
12 1992-1994 period, relative risks for equivalent distributional increment (e.g., IQR) were
13 comparable among PM10, PM25, and PM10_2 5 for both mortality and hospital admissions
14 categories; and SO4"2 was more strongly associated with most outcomes than FT. Consideration
15 of the overall pattern of results led the authors to state that the mass of the smaller size index
16 could explain a substantial portion of the variation in the larger size indices. In these data, on
17 average, PM2 5 accounted for 60% of PM10 (up to 80% on some days) and PM10 for 66% of TSP
18 mass. The temporal correlation between TSP and PM2 5 was r = 0.63, and that for PM25 and
19 PM10 was r = 0.90, suggesting that much of the apparent larger particle effects may well be
20 mainly driven by temporally covarying smaller PM2 5 particles. The stronger associations for
21 sulfates than FT, suggestive of non-acid fine particle effects, must be caveated by noting the very
22 low FT levels present (often at or near non-detection limit).
23 Three research groups, using different methods, have examined the same Phoenix, AZ data
24 set. While these groups used somewhat different approaches, there is some consistency among
25 their results in that PM10_2 5 appeared to emerge as the likely more important predictor of
26 mortality versus PM25. In the Clyde et al. (2000) analysis, PM-mortality associations were
27 found only for the geographic area where PM2 5 was considered uniformly distributed, but the
28 association was with PM10_2 5, not PM2 5. Based on the Bayes Information Criterion, the highly
29 ranked models consistently included 1-day lagged PM10_25. Smith et al. (2000) analyses found
30 that, based on a linear PM effect, PM10_2 5 was significantly associated with total mortality, but
31 PM2 5 was not. However, Smith et al.'s finding that PM2 5 may have a threshold effect further
June 2003 8-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 complicates a simple comparison of the two size-fractionated mass concentration indices. In the
2 Mar et al. (2000 & 2003) analyses, cardiovascular mortality (CVM) was significantly associated
3 with both PM2 5 and PM10_2 5. CVM was also significantly associated with a motor vehicle source
4 category with loading of PM25, EC, OC, CO, NO2, and some trace metals, as shown by the factor
5 analyses discussed later. The PM2 5 in Phoenix is mostly generated from motor vehicles,
6 whereas PM10_25 consists mainly of two types of particles: (a) crustal particles from natural
7 (wind blown dust) and anthropogenic (construction and road dust) processes, and (b) organic
8 particles from natural biogenic processes (endotoxin and molds) and anthropogenic (sewage
9 aeration) processes. The crustal particles, however, are also likely contaminated with metals
10 secondarily deposited over many years as the result of emissions from smelters operating until
11 recently in the Phoenix area.
12 In summary, the issue regarding the relative importance of PM2 5 and PM10_25 has not yet
13 been fully resolved. Caution in interpreting size-fraction PM studies is warranted due to the
14 problem of measurement error and the correlation between the two size fractions. Limitations of
15 single-city studies have been noted. While the limited sample size prevented clear statistical
16 distinction of the relative roles played by PM25 and PM10_25, recent studies show mixed results,
17 with some studies suggesting coarse particle effects. The relative importance may also vary
18 depending on the chemical constituents in each size fraction, which may vary from city to city.
19 Nevertheless, a number of studies published since the 1996 PM AQCD do appear to substantiate
20 associations between PM2 5 and increased total and/or CVD mortality. Consistent with the 1996
21 PM AQCD findings, effect-size estimates from the new studies generally fall within the range of
22 about 2 to 6% excess total mortality per 25 |ig/m3 PM2 5. The coarse particle (PM10_2 5) effect-
23 size estimates also tend to fall in the same range.
24
25 Crustal Particle Effects
26 Since the 1996 PM AQCD, several studies have yielded interesting new information
27 concerning possible roles of crustal wind-blown particles or crustal particles within the fine
28 particle fraction (i.e., PM25) in contributing to observed PM-mortality effects.
29 Schwartz et al. (1999), for example, investigated the association of coarse particle
30 concentrations with non-accidental deaths in Spokane, WA, where dust storms elevate coarse
31 PM concentrations. During the 1990-1997 period, 17 dust-storm days were identified. The
June 2003 8-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM10 levels during those storms averaged 263 |ig/m3, compared to 39 |ig/m3 for the entire period.
2 The coarse particle domination of PM10 data on those dust-storm days was confirmed by a
3 separate measurement of PM10 and PMX 0 during a dust storm in August, 1996: the PM10 level
4 was 187 |ig/m3, while PMX 0 was only 9.5 |ig/m3. The deaths on the day of a dust storm were
5 contrasted with deaths on control days (n = 95 days in the main analysis and 171 days in the
6 sensitivity analysis), which are defined as the same day of the year in other years when dust
7 storms did not occur. The relative risk for dust-storm exposure was estimated using Poisson
8 regressions, adjusting for temperature, dewpoint, and day of the week. Various sensitivity
9 analyses considering different seasonal adjustment, year effects, and lags were conducted. The
10 expected relative risk for these storm days with an increment of 221 |ig/m3 would be about 1.04,
11 based on PM10 relative risk from past studies, but the estimated RR for high PM10 days was
12 found to be only 1.00 (95% CI = 0.95-1.05) per 50 |ig/m3 PM10 change in this study. Schwartz
13 et al. concluded that there was no evidence to suggest that coarse (presumably crustal) particles
14 were associated with daily mortality.
15 Ostro et al. (2000 & 2003) analyzed the Coachella Valley, CA data for 1989-1998. This
16 desert valley, where coarse particles of geologic origin comprise circa 50-60% of annual-average
17 PM10 (> 90% during wind episodes throughout the year), includes the cities of Palm Springs and
18 Indio, CA. Cardiovascular deaths were analyzed using GAM (with stringent convergence
19 criteria) and GLM Poisson models adjusting for temperature, humidity, day-of-week, season,
20 and time. The actual PM25 and PM10_25 data were available for the last 2.5 years. Predictive
21 models for PM25 and PM10_25 concentrations were developed for earlier years, but the model for
22 PM25 was not considered successful and, therefore, was not used. Thus, a strict comparison of
23 risk estimates for PM2 5 and PM10_2 5 in this data set is difficult. Cardiovascular mortality was
24 positively associated with both PM10 and PM10_2 5 at multiple lags between 0 and 2 day lags;
25 whereas PM2 5 coefficient was positive only at lag 4 day. These results hint at crustal particle
26 effects possibly being important in this desert situation, but the ability to discern more clearly the
27 role of fine particles would likely be improved by analyses of more years of actual data for
28 PM25.
29 Laden et al. (2000) and Schwartz (2003b) analyzed Harvard Six-Cities Study data and Mar
30 et al. (2000) analyzed the Phoenix data to investigate the influence of crustal particles in PM2 5
31 samples on daily mortality. These studies are discussed in more detail in Section 8.2.2.4.3 on the
June 2003 8-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 source-oriented evaluation of PM; and only the basic results regarding crustal particles are
2 mentioned here. The elemental abundance data (from X-ray fluorescence spectroscopy analysis
3 of daily filters) were analyzed to estimate the concentration of crustal particles in PM2 5 using
4 factor analysis. Then the association of mortality with fine crustal mass was estimated using
5 Poisson regression (regressing mortality on factor scores for "crustal factor"), adjusting for time
6 trends and weather. No positive association was found between fine crustal mass factor and
7 mortality.
8 The above results, overall, mostly suggest that crustal particles (coarse or fine) per se are
9 not likely associated with daily mortality. However, as noted in the previous section, three
10 analyses of Phoenix, AZ data suggested that PM10_2 5 was associated with mortality. The results
11 from one of the three studies (Smith et al., 2000) suggest that coarse particle mortality
12 associations are stronger in spring and summer, when the anthropogenic portion of PM10_25 is
13 lowest as determined by factor analysis. However, during spring and summer, biogenic
14 processes (e.g., wind-blown endotoxins and molds) may contribute more to the PM10_25 fraction
15 in the Phoenix area, clouding any attribution of observed PM10_25 effects there to crustal
16 particles, per se.
17
18 Ultrafine Particle Effects
19 Wichmann et al. (2000) evaluated the attribution of PM effects to specific size fractions,
20 including both the number concentration (NC) and mass concentration (MC) of particles in a
21 given size range. To respond to the GAM convergence issues, Stolzel et al. (2003) reanalyzed
22 the data, using GAM with stringent convergence criteria and GLM with natural splines. The
23 study was carried out in the small German city of Erfurt (pop. 200,000) in the former German
24 Democratic Republic. Erfurt was heavily polluted by particles and SO2 in the 1980s, and excess
25 mortality was attributed to high levels of TSP by Spix et al. (1993). Concentrations of PM and
26 SO2 have markedly dropped since then. The present study provides a much more detailed look
27 at the health effects of ultrafme particles (diameter < 0.1 jim) than earlier studies and enables
28 examination of effects in relation to number counts for fine and ultrafme particles, as well as in
29 relation to their mass.
30 The Mobile Aerosol Spectrometer (MAS), developed by Gessellschaft fur
31 Strahlenforschung (GSF), produces number and mass concentrations in three size classes of
June 2003 8-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 ultrafmes (0.01 to 0.1 jim) and three size classes of larger fine particles (0.1 jim to 2.5 jim). The
2 mass concentration MC0 01_2 5 is well correlated with gravimetric PM2 5, and the number
3 concentration NC001.25 is well correlated with total particle counts from a condensation particle
4 counter (CPC). Mortality data were coded by cause of death, with some discrimination between
5 underlying causes and prevalent conditions of the deceased. In the reanalysis, daily mortality
6 data were fitted using a Poisson GAM (with stringent convergence criteria) and GLM, with
7 adjustments for weather variables, time trends, day of week, and particle indices. Weekly data
8 for all of Germany on influenza and similar diseases was also included in the model. In the
9 original analysis, two types of models were fitted; one used the best single-day lag for air
10 pollution and a second used the best polynomial distributed lag (PDL) model for air pollution.
11 Both linear (i.e., raw) and log-transformed pollution indices were examined. PDL models in the
12 original analysis generally had larger and more significant PM effects than single-day lag
13 models, but the reanalysis by Stolzel et al. (2003) focused on single-day lag results only.
14 Therefore, the numerical results in the following discussion will only include the single day lag
15 results from the reanalysis. It should be noted that, unlike most of the recent reanalyses that
16 have been conducted to address the GAM conversion issue, the reanalysis results from this study
17 were virtually unchanged from the original results.
18 Both mass and number concentrations at the size ranges examined were mostly positively
19 (and significantly or nearly significantly) associated with total non-accidental mortality. The
20 best single-day lags reported were mostly 0 or 1 day lag for mass concentrations and the 4 day
21 lag for number concentrations. For example, the estimated excess risk for MC001_2 5 at lag 1 day
22 was about 3.9% (CI = 0, 7.7) per 25 |ig/m3. The corresponding number for smaller fine particles,
23 MCooj.j 0, was 3.5% (CI = -0.4, 7.7). For number concentration, the estimated excess risk for
24 NC001.25 at lag 4 day was about 4.1% (CI = -0.9, 9.3) per IQR (13,269 particles/cm3). The
25 corresponding number for smaller fine particles, NC0014 0, was 4.6% (CI = -0.3, 9.7) per IQR
26 (12,690 particles/cm3). An examination of the all the results for MC001_25 and NC001.0 x shown
27 for lags 0 through 5 days indicates that the associations were mostly positive for these mass and
28 number concentrations, except for the "dip" around 2 or 3 day lags.
29 The estimated excess risks are reduced, sometimes drastically, when co-pollutants
30 (especially SO2 and NO2) are included in a two-pollutant model. This is not surprising, as the
31 number and mass concentrations of various ultrafine and fine particles in all size ranges are
June 2003 8-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 rather well correlated with gaseous co-pollutants, except for the intermodal size range MCL0.2.5.
2 The number correlations range from 0.44 to 0.62 with SO2, from 0.58 to 0.66 with NO2, and
3 from 0.53 to 0.70 with CO. The mass correlations range from 0.53 to 0.62 with SO2, from 0.48
4 to 0.60 with NO2, and from 0.56 to 0.62 with CO. The authors found that ultrafme particles, CO
5 and NO2 form a group of pollutants strongly identified with motor vehicle traffic. Immediate
6 and delayed effects seemed to be independent in two-pollutant models, with single-day lags of 0
7 to 1 days and 4 to 5 days giving 'best fits' to data. The delayed effect of ultrafme particles was
8 stronger than that for NO2 or CO. The large decreases in excess risk for number concentration,
9 particularly when NO2 is a co-pollutant with NC0 01.0 b clearly involves a more complex structure
10 than simple correlation. The large decrease in excess risk when SO2 is a co-pollutant with
11 MC001_25 is not readily explained and is discussed in some detail in Wichmann et al. (2000).
12 SO2 is a strong predictor of excess mortality in this study; and its estimated effect is little
13 changed when different particle indicators are included in a two-pollutant model. The authors
14 noted ". . .the [LOESS] smoothed dose response curve showed most of the association at the left
15 end, below 15 |ig/m3, a level at which effects were considered biologically implausible. . ."
16 Replacement of sulfur-rich surface coal has reduced mean SO2 levels in Erfurt from 456 |ig/m3
17 in 1988 to 16.8 |ig/m3 during 1995 to 1998 and to 6 |ig/m3 in 1998. The estimated
18 concentration-response functions for SO2 are very different for these time periods, comparing
19 Spix et al. (1993) versus Wichmann et al. (2000) results. Wichmann et al. concluded "These
20 inconsistent results for SO2 strongly suggested that SO2 was not the causal agent but an indicator
21 for something else." The authors offered no specific suggestions as to what the "something else"
22 might be, but they did finally conclude that their studies from Germany strongly supported PM
23 air pollution as being more relevant than SO2 to observed mortality outcomes.
24
25 8.2.2.4.2 Chemical Components
26 Eight new studies from the U.S. and Canada examined mortality associations with specific
27 chemical components of ambient PM. Table 8-3 shows the chemical components examined in
28 these studies, the mean concentrations for Coefficient of Haze (CoH), sulfate, and H+, as well as
29 indications of those components found to be associated with increased mortality.
30
31
June 2003 8-61 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-3. NEWLY AVAILABLE STUDIES OF MORTALITY
RELATIONSHIPS TO PM CHEMICAL COMPONENTS
Author, City
Mean
CoH
(1000ft)
Mean
SO/
(ug/m3)
Mean H+
(nmol/m3)
Other PM
components analyzed
Specific PM
components found to be
associated with mortality
(comments).
Burnett etal. (2000); 0.26 2.6
Burnett and Goldberg
(2003)* 8 largest
Canadian cities, 1986-
1996.
Fairley(1999& 0.5 1.8
2003)*; Santa Clara
County, CA.
Goldberg et al. (2000); 0.24 3.3
Goldberg and Burnett
(2003); Goldberg et al.
(2003)* Montreal,
Quebec, Canada.
1984-1993.
Lipfert et al, (2000a) 0.28 5.1
Philadelphia, PA.
1992-1995.
Lippmannetal. 5.2
(2000); Ito (2003)*
Detroit, MI.
1992-1994.
Klemm and Mason 5.2
(2000)
Atlanta, GA
1998-1999
8.0
PM10, PM25, PM1(K5,
and 47 trace elements
PM10, PM25, PM10.25, and
nitrate
Predicted PM2 5, and
extinction coefficient
(visual- range derived).
Nepherometry, NH4+,
TSP, PM10, PM25, and
PM10.2.5
PM10, PM25, CoH, sulfate, Zn, Ni,
and Fe were significantly
associated with total mortality in
the original analysis. The
reanalysis only analyzed mass
concentration indices.
CoH, sulfate, nitrate, PM10, and
PM2 5 were associated with
mortality. PM25 and nitrate most
significant.
CoH and extinction coefficient
were associated with the deaths
that were classified as having
congestive heart failure before
death based on medical records.
Associations were stronger in
warm season.
Essentially all PM components
were associated with mortality.
PM10, PM25, and PM10.25 PM10, PM25, and PM10.25 were
more significantly associated with
mortality outcomes than sulfate or
Nitrate, EC, OC,
oxygenated HC, PM10,
PM25, and PM10.25
"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
PM9V
Mar et al. (2000 &
2003)* Phoenix, AZ.
1995-1997.
Tsai et al. (2000). 12.7
Newark, Elizabeth,
and Camden, NJ.
1981-1983.
Hoeketal. 3.8
(2000 & 2003)* (median)
The Netherlands.
1986-1994.
EC, OC, TC, PM10,
PM25, and PM10.25
PM15, PM25,
cyclohexane-solubles
(CX), dichloromethane-
solubles (DCM), and
acetone-solubles (ACE).
PM10, BS, and nitrate
EC, OC, TC, PM10, PM25, and
PM10_25 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 was PM10.
*Note: The study was originally analyzed by GAM models only using default convergence criteria and at least two non-parametric
smoothing terms and was recently reanalyzed by GAM using stringent convergence criteria and/or other non-GAM analyses.
June 2003
8-62
DRAFT-DO NOT QUOTE OR CITE
-------
1 Coefficient of Haze, Elemental Carbon, and Organic Carbon
2 CoH is highly correlated with elemental carbon (EC) and is often considered as a good PM
3 index for motor vehicle sources, although other combustion processes such as space heating
4 likely also contribute to CoH levels. Several studies (Table 8-3) examined CoH; and, in most
5 cases, positive and significant associations with mortality outcomes were reported. In terms of
6 relative significance of CoH in comparison to other PM components, CoH was not the clearly
7 most significant PM component in most of these studies. The average level of CoH in these
8 studies ranged from 0.24 (Montreal, Quebec) to 0.5 (Santa Clara County, CA) 1000 linear feet.
9 The correlations between CoH and NO2 or CO in these studies (8 largest Canadian cities; Santa
10 Clara County, CA) were moderately high (r .0.7 to 0.8) and suggested a likely motor vehicle
11 contribution. Both EC and OC were significant predictors of cardiovascular mortality in the
12 Phoenix study; their effect sizes per IQR were comparable to those for PM10, PM2 5, and PM10_2 5.
13 Also, both EC and OC represented major mass fractions of PM25 (11% and 38%, respectively)
14 and were correlated highly with PM25 (r = 0.84 and 0.89, respectively). They were also highly
15 correlated with CO and NO2 (r = 0.8 to 0.9), indicating their associations with an "automobile"
16 factor. Thus, the CoH and EC/OC results from the Mar et al. (2000 and 2003) study suggest that
17 PM components from motor vehicle sources are likely associated with mortality. In a recent
18 study in Montreal, Quebec, by Goldberg et al. (2000 and 2003), CoH appeared to be correlated
19 with the congestive heart failure mortality (as classified based on medical records) more strongly
20 than other PM indices such as the visual-range derived extinction coefficient (considered to be a
21 good indicator of sulfate). However, the main focus of the study was the role of cardio-
22 respiratory risk factors for air pollution, and the investigators warned against comparing the
23 relative strength of associations among PM indices, pointing out complications such as likely
24 error involved in the visual range measurements. Additionally, the estimated PM2 5 values were
25 predicted from other PM indices, including CoH and extinction coefficient, making it difficult to
26 compare straightforwardly the relative importance of PM indices.
27
28 Sulfate and Hydrogen Ion
29 Sulfate and H+, markers of acidic components of PM, have been hypothesized to be
30 especially harmful components of PM (Lippmann and Thurston, 1996). The newly available
31 studies that examined sulfate are shown in Table 8-3; two of them also analyzed H+ data. The
June 2003 8-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 sulfate concentrations ranged from 1.8 |ig/m3 (Santa Clara County, CA) to 12.7 |ig/m3(three NJ
2 cities). Aside from the west versus east coast contrast, the higher levels observed in the three NJ
3 cities are likely due to their study period coverage of the early 1980's, when sulfate levels were
4 higher. Sulfate explained 25 to 30% of PM25 mass in eastern U.S. and Canadian cities, but it
5 was only 14% of PM2 5 mass in Santa Clara County, CA. The H+ levels measured in Detroit and
6 Philadelphia were low. The mean H+ concentration for Detroit, MI (the H+ was actually
7 measured in Windsor, a Canadian city a few miles from downtown Detroit), 8.8 nmol/m3, was
8 low as compared to the reported detection limit of 15.1 nmol/m3 (Brook et al., 1997) for the
9 measurement system used in the study. Note that the corresponding detection limit for sulfate
10 was 3.6 nmol/m3 (or 0.34 |ig/m3); and the mean sulfate level for Detroit was 54 nmol/m3 (or
11 5.2 jig/m3), so that the signal-to-noise ratio is expected to be higher for sulfate than for H+.
12 Thus, the ambient levels and possible relative measurement errors for these data should be
13 considered in interpreting the relative strength of mortality associations in these data.
14 Sulfate was a statistically significant predictor of mortality, at least in single pollutant
15 models, in: Santa Clara County, CA; Philadelphia, PA; Newark, NJ; and Camden, NJ, but not in
16 Elizabeth, NJ; Detroit, MI; or Montreal, CN. However, it should be noted that the relative
17 significance across the cities is influenced by the sample size (both the daily mean death counts
18 and number of days available), as well as the range of sulfate levels and should be interpreted
19 with caution. Figure 8-8 shows the excess risks (± 95% CI) estimated per 5 |ig/m3 increase in
20 24-h sulfate reported in these studies compared to the reanalysis results of the earlier Six Cities
21 Study result by Klemm and Mason (2003). The largest estimate was seen for Santa Clara
22 County, CA; but the wide confidence band (possibly due to the small variance of the sulfate,
23 because its levels were low) should be taken into account. In addition, the sulfate effect in the
24 Santa Clara County analysis was eliminated once PM2 5 was included in the model, perhaps
25 being indicative of sulfate mainly serving as a surrogate for fine particles in general there.
26 In any case, more weight should be accorded to estimates from other studies with narrower
27 confidence bands. In the other studies, the effect size estimates mostly ranged from about 1 to
28 4% per 5 |ig/m3 increase in 24-h sulfate.
29 The relative significance of sulfate and FT compared to other PM components is not
30 clear in the existing small number of publications. Because each study included different
31 combinations of co-pollutants that had different extents of correlation with sulfate and because
June 2003 8-64 DRAFT-DO NOT QUOTE OR CITE
-------
Percent excess death (total non-accidental mortality)
per 5 pig/m3 increase in sulfate
-2 0 2 4 6 8 10
Klemm and Mason (2003)* -
Harvard 6 cities (recomputed)
Santa Clara Co.
Atlanta, GA
Philadelphia, PA
Detroit, Ml
3 NJ cities
k
' % Newark
' Q
: Elizabeth
Figure 8-8. Excess risks estimated per 5 ug/m3 increase in sulfate, based on the studies in
which both PM2 5 and PM10_2 5 data were available.
1 multiple mortality outcomes were analyzed, it is difficult to assess the overall importance of
2 sulfate across the available studies. The fact that the Lippmann et al. (2000) study and the
3 reanalysis by Ito (2003) found that Detroit, MI data on H+ and sulfate were less significantly
4 associated with mortality than the size-fractionated PM mass indices may be due to acidic
5 aerosols levels being mostly below the detection limit in that data. In this case, it appears that
6 the Detroit PM components show mortality effects even without much acidic input.
7 In summary, assessment of new study results for individual chemical components of PM
8 suggest that an array of PM components (mainly fine particle constituents) are associated with
9 mortality outcomes, including CoH, EC, OC, sulfate, and nitrate. The variations seen with
10 regard to the relative significance of these PM components across studies may be in part due to
11 differences in their concentrations from locale to locale. This issue is further discussed below as
12 part of the assessment of new studies involving source-oriented evaluation of PM components.
June 2003
8-65
DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.2.4.3 Source-Oriented Evaluations
2 Several new studies have conducted source-oriented evaluation of PM components.
3 In these studies, daily concentrations of PM components (i.e., trace elements) and gaseous
4 co-pollutants were analyzed using factor analysis to estimate daily concentrations due to
5 underlying source types (e.g., motor vehicle emissions, soil, etc.), which are weighted linear
6 combinations of associated individual variables. The mortality outcomes were then regressed on
7 those factors (factor scores) to estimate the effect of source types rather than just individual
8 variables. These studies differ in terms of specific objectives/focus, the size fractions from
9 which trace elements were extracted, and the way factor analysis was used (e.g., rotation). The
10 main findings from these studies regarding the source-types identified (or suggested) and their
11 associations with mortality outcomes are summarized in Table 8-4.
12 The Laden et al. (2000) analysis of Harvard Six Cities data for 1979-1988 (reanalyzed by
13 Schwartz, 2003) aimed to identify distinct source-related fractions of PM25 and to examine each
14 fraction's association with mortality. Fifteen elements in the fine fraction samples were
15 routinely found above their detection limits and included in the data analysis. For each of the six
16 cities, up to 5 common factors were identified from among the 15 elements, using specific
17 rotation factor analysis. Using the Procrustes rotation (a type of oblique rotation), the projection
18 of the single tracer for each factor was maximized. This specification of the tracer element was
19 based on (a) knowledge from previous source apportionment research; (b) the condition that the
20 regression of total fine mass on that element must result in a positive coefficient; and (c) the
21 identifications of additional local source factors that positively contributed to total fine mass
22 regression. Three source factors were identified in all six cities: (1) a soil and crustal material
23 factor with Si as a tracer; (2) a motor vehicle exhaust factor with Pb as a tracer; and (3) a coal
24 combustion factor with Se as a tracer. City-specific analyses also identified a fuel combustion
25 factor (V), a salt factor (Cl), and selected metal factors (Ni, Zn, or Mn). In the original analysis
26 by Laden et al., a GAM Poisson regression model (with default convergence criteria), adjusting
27 for trend/season, day-of-week, and smooth function of temperature/dewpoint, was used to
28 estimate impacts of each source type (using absolute factor scores) simultaneously for each city.
29 In the reanalysis reported by Schwartz (2003a), GAM models with LOESS smoothers were
30 replaced with penalized splines. Summary estimates across cities were obtained by combining
31 the city-specific estimates, using inverse-variance weights. The identified factors and their
June 2003 8-66 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-4. SUMMARY OF SOURCE-ORIENTED EVALUATIONS OF PM
COMPONENTS IN RECENT STUDIES
Author, City
Source types identified (or suggested)
and associated variables
Source types associated with mortality
(Comments)
Laden et al.,
(2000);
Schwartz (2003)*
Harvard Six Cities.
1979-1988.
Mar et al.
(2000 & 2003)*
Phoenix, AZ.
1995-1997.
5*0/7 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
PM2S (fromDFPSS) trace elements:
Motor vehicle emissions and re-suspended
road dust: Mn, Fe, Zn, Pb, OC, EC, CO,
andNO2
5*0/7: Al, Si, and Fe
Vegetative burning: OC, and Ks
(soil-corrected potassium)
Local SO2 sources: SO2
Regional sulfate: S
Strongest increase in daily mortality was
associated with the mobile source factor.
Coal combustion factor was also positively
associated with mortality. Crustal factor
from fine particles not associated (negative
but not significant) with mortality. Coal
and mobile sources account for the majority
of fine particles in each city.
PM2 5 factors results: Motor vehicle factor
(1 day lag), vegetative burning factor (3 day
lag), and regional sulfate factor (0 day lag)
were significantly positively associated
with cardiovascular mortality.
Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ.
1981-1983.
Ozkaynak et al.
(1996).
Toronto, Canada.
PM10_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
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
Motor vehicle emissions: CO, CoH, and
NO,
Factors from dichot PM10_2 5 trace elements
not analyzed for their associations with
mortality because of the small sample size
(every 3rd-day samples from June 1996).
Oil burning, industry, secondary aerosol,
and motor vehicle factors were associated
with mortality.
Motor vehicle factor was a significant
predictor for total, cancer, cardiovascular,
respiratory, and pneumonia deaths.
*Note: The study was originally analyzed using GAM models only with default convergence criteria using at
least two non-parametric smoothing terms, but was later reanalyzed using more stringent convergence criteria
and/or other approaches.
1 tracers are listed in Table 8-4. The reanalysis using penalized splines changed somewhat the risk
2 estimates for source-apportioned mass concentrations in each city compared to those in the
3 original GAM results (increasing estimates in some cities and reducing them in others), but the
4 combined estimates across the six cities did not change substantially. The combined estimates
June 2003
8-67
DRAFT-DO NOT QUOTE OR CITE
-------
1 indicated that the largest increase in daily mortality was associated with the mobile source
2 associated fine mass concentrations, with an excess death risk increase of 9.3% (95% CI: 4.0,
3 14.9) per 25 |ig/m3 source-apportioned PM25 (average of 0 and 1 day lags). The corresponding
4 value for the PM25 mass apportioned for the coal combustion factor was 2.0% (95% CI: -0.3,
5 4.4). The crustal factor was not associated with mortality (-5.1%; 95% CI = -13.9, 4.6).
6 Mar et al. (2000) analyzed PM10, PM10_2 5, PM25 measured by two methods, and various
7 sub-components of PM25 for their associations with total (non-accidental) and cardiovascular
8 deaths in Phoenix, AZ during 1995-1997, using both individual PM components and factor
9 analysis-derived factor scores. In the original analysis, GAM Poisson models (with default
10 convergence criteria) were used and adjusted for season, temperature, and relative humidity.
11 In the reanalysis (Mar et al., 2003), GAM models with stringent convergence criteria and GLM
12 models with natural splines were used. Only cardiovascular mortality was analyzed in the
13 reanalysis; and the results for that category are summarized here. The evaluated air pollution
14 variables included O3, SO2, NO2, CO, TEOM PM10, TEOM PM25, TEOM PM10.25, DFPSS PM25,
15 S, Zn, Pb, soil, soil-corrected K (KS), nonsoil PM, OC, EC, and TC. Lags 0 to 4 days were
16 evaluated. A factor analysis conducted on the chemical components of DFPSS PM25 (Al, Si, S,
17 Ca, Fe, Zn, Mn, Pb, Br, KS, OC, and EC) identified factors for motor vehicle emissions/re-
18 suspended road dust; soil; vegetative burning; local SO2 sources; and regional sulfate (see Table
19 8-4). The results of mortality regression with these factors suggested that the motor vehicle
20 factor (lag 1 day), vegetative burning factor (3 day lag), and regional sulfate factor (0 day lag)
21 were each had significant positive associations with cardiovascular mortality. The PM25 mass
22 was not apportioned to these factors in this study; so information on the excess-deaths estimate
23 per source-apportioned PM2 5 concentrations were not available. The authors also analyzed
24 elements from dichot PM10_25 samples and identified soil, a source of coarse fraction metals
25 (industry), and marine influence factors. However, these factors were not analyzed for their
26 associations with mortality outcomes due to the short measurement period (starting in June 1996
27 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
June 2003 8-68 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 may not necessarily be due to anthropogenic components of the coarse particles,
4 the biogenically-generated coarse particles perhaps being key during the warmer months (as
5 noted earlier).
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. A factor analysis of trace elements and sulfate was first conducted and
12 identified several 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, an alternative approach was also used, in which the
15 inhalable particle mass (IPM, D50 < 15 jim) 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. Oil burning (V, Ni), various industrial sources (Zn, Cd), motor
19 vehicle (Pb, CO), and secondary aerosols, as well as the individual PM indices IPM, FPM (D50
20 < 3.5 |im), and sulfates, were all associated with total and/or cardiorespiratory mortality in
21 Newark and Camden, but not in Elizabeth. In Camden, the RRs for the source-oriented PM were
22 higher (1.10) than those for individual PM indices (1.02).
23 Ozkaynak et al. (1996) had earlier analyzed 21 years of mortality and air pollution data for
24 Toronto, Canada. In addition to the usual simultaneous inclusion of multiple pollutants in
25 mortality regressions, they also conducted a factor analysis of all the air pollution and weather
26 variables, including TSP, SO2, CoH, NO2, O3, CO, relative humidity and temperature. The factor
27 with the largest variance contribution (50%) had the highest factor loadings for CO, CoH, and
28 NO2 and was considered by them to be representative of motor vehicle emissions, since this
29 pollution grouping was also consistent with the emission inventory information for that city.
30 After filtering out seasonal cycles and adjusting for temperature and day-of-week effects, they
31 then regressed mortality on the factor scores (a linear combination of standardized scores for the
June 2003 8-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 covariates). The estimated effects of motor vehicle pollution on mortality ranged from 1 to 6%
2 for different specific health outcomes.
3 In summary, these source-oriented factor analyses studies suggest that a number of source
4 types are associated with mortality, including motor vehicle emissions, coal combustion, oil
5 burning, and vegetative burning. The crustal factor from fine particles was not associated with
6 mortality in the Harvard Six Cities data. In Phoenix, where coarse particles were reported to be
7 associated with mortality, the associations between the factors related to coarse particles (soil,
8 marine influence, and anthropogenic elements) and mortality could not be evaluated due to the
9 small sample size. Thus, although some unresolved issues remain (mainly due to the lack of
10 sufficient data), the limited results from the source-oriented evaluation approach (using factor
11 analysis) thus far seem to implicate fine particles of anthropogenic origin as being most
12 important (versus crustal particles of geologic origin) in contributing to increased mortality risks.
13
14 8.2.2.5 New Assessments of Cause-Specific Mortality
15 Consistent with similar findings described in the 1996 PM AQCD, most of the newly
16 available studies summarized in Tables 8-1 and 8A-1 that examined non-accidental total,
17 circulatory, and respiratory mortality categories (e.g., Samet et al., 2000a,b and the reanalysis by
18 Dominici et al., 2002 and 2003) found significant PM associations with both cardiovascular
19 and/or respiratory-cause mortality. Several studies (e.g., Fairley, 1999), his reanalysis, 2003;
20 Wordley et al., 1997; Prescott et al., 1998) reported estimated PM effects that were generally
21 higher for respiratory deaths than for circulatory or total deaths. Once again, the NMMAPS
22 results for U. S. cities are among those of particular note here due to the large study size and the
23 combined, pooled estimates derived for various U.S. regions.
24 The NMMAPS 90-cities analyses not only examined all-cause mortality (excluding
25 accidents), but also evaluated cardio-respiratory and other remaining causes of deaths. Results
26 were presented for all-cause, cardio-respiratory, and "other" mortality for lag 0, 1, and 2 days.
27 The investigators commented that, compared to the result for cardio-respiratory deaths showing
28 1.6% (CI = 0.8, 2.4) increase per 50 |ig/m3PM10 in a GLM model (versus 1.1% for total non-
29 accidental mortality using GLM), there was less evidence for non-cardio-respiratory deaths.
30 However, the estimates for "other" mortality, though less than half those for cardio-respiratory
31 mortality, were nevertheless positive, with a fairly high posterior probability (e.g., 0.92 at lag 1
June 2003 8-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 day) that the overall effects were greater than zero. It should be noted that the "other" (other
2 than cardio-respiratory) underlying cause of mortality may include deaths that had contributing
3 cardiovascular or respiratory causes. For example, Lippmann et al. (2000) noted that the "other"
4 (non-circulatory and non-respiratory) mortality showed seasonal cycles and apparent influenza
5 peaks, suggesting that this series may have also been influenced by respiratory contributing
6 causes. Thus, interpretation of the observed associations between PM and broad "specific"
7 categories of underlying causes of death may not be straightforward.
8 Another U.S. study, that of Moolgavkar (2000a), evaluated possible PM effects on cause-
9 specific mortality across a broad range of lag times (0-5 days) in Cook Co., IL; Los Angeles Co.,
10 CA; and Maricopa Co., AZ. Total non-accidental mortality, as well as deaths related to
11 cardiovascular disease (CVD), cerebrovascular disease (CRV), and chronic obstructive lung
12 disease (COPD) were analyzed in the original study. The data for Cook Co. and Maricopa Co.
13 were reanalyzed using GAM model with stringent convergence criteria and GLM model with
14 natural splines (Moolgavkar, 2003). Cerebrovascular disease mortality was not reanalyzed
15 because there was little evidence of association for PM with this category at any lag in any of the
16 three counties analyzed. Moolgavkar reported that varying patterns of results were obtained for
17 PM indices in evaluations of daily deaths related to CVD and COPD in the two counties. In the
18 Cook Co. (Chicago) area, the association of PM10 with CVD mortality was statistically
19 significant at a lag of 3 days based on a single-pollutant analysis and remained significantly
20 associated with CVD deaths with a 3-day lag in two pollutant models including one or another of
21 CO, NO2, SO2, or O3. In Los Angeles single-pollutant analyses, CVD mortality was significantly
22 associated with PM10 (2 day lag) and PM2 5 (0 and 1 day lag). Their percent excess risk estimates
23 were up to twice those for total non-accidental mortality. In a two-pollutant model with CO
24 (most strongly positively associated with mortality in Los Angeles Co. among the pollutants),
25 PM10 risk estimates were reduced. However, PM2 5 excess risk estimates in the two-pollutant
26 model with CO nearly doubled (2.5% per 25|ig/m3 increase in PM2 5 to 4.8% using GLM);
27 whereas that for CO became significantly negative. Obviously, CO and PM2 5 were correlated (r
28 ~ 0.58), and the estimated associations were likely confounded between these two pollutants in
29 this locale. With regard to COPD deaths, PM10 was significantly associated with COPD
30 mortality (lag 2 days) in Cook Co., but in Los Angeles Co., both PM10 and (especially) PM25
31 showed erratic associations with COPD mortality at varying lags, alternating positive and
June 2003 8-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 negative (significantly, at lag 3 day) coefficients. The combination of the every 6th-day PM data
2 in Los Angeles (versus daily PM10 in Cook Co.) and relatively small daily counts for COPD
3 (median = 6/day versus 57/day for CVD) makes the effective sample size of COPD mortality
4 analysis small and the results unstable.
5 Zmirou et al. (1998) presented cause-specific mortality analyses results for 10 of the
6 12 APHEA European cities (APHEA1). Using Poisson autoregressive models parametrically
7 adjusting for trend, season, influenza epidemics, and weather, each pollutant's relative risk was
8 estimated for each city and "meta-analyses" of city-specific estimates were conducted. The
9 pooled excess risk estimates for cardiovascular mortality were 1.0% (0.3, 1.7) per
10 25 |ig/m3increase in BS and 2.0% (0.5, 3.0) per 50 |ig/m3increase in SO2 in western European
11 cities. The pooled risk estimates for respiratory mortality in the same cities were 2.0% (0.8, 3.2)
12 and 2.5% (1.5, 3.4) for BS and SO2, respectively.
13 Seeking unique cause-specificity of effects associated with various pollutants has been
14 difficult because the "cause specific" categories examined are typically rather broad (usually
15 cardiovascular and respiratory) and overlap and because cardiovascular and respiratory
16 conditions tend to occur together. Examinations of more specific cardiovascular and respiratory
17 subcategories may be necessary to test hypotheses about any specific mechanisms, but smaller
18 sample sizes for more specific sub-categories may make a meaningful analysis difficult. The
19 Hoek et al. (2000 and 2001) study and its reanalysis by Hoek (2003) took advantage of a larger
20 sample size to examine cause-specific mortality. The large sample size, including the whole
21 population of the Netherlands (mean daily total deaths -330, or more than twice that of Los
22 Angeles County), allowed examination of specific cardiovascular causes of deaths. The
23 reanalysis using GAM with stringent convergence criteria as well as GLM with natural splines
24 either did not change or even increased the effect estimates. Deaths due to heart failure,
25 arrhythmia, and cerebrovascular causes were more strongly (~2 to 4 times larger excess risks)
26 associated with air pollution than the overall cardiovascular deaths. The investigators concluded
27 that specific cardiovascular causes (such as heart failure) were more strongly associated with air
28 pollution than total cardiovascular mortality, but noted that the largest contribution to the
29 association between air pollution and cardiovascular mortality was from ischemic heart disease
30 (about half of all CVD deaths). The analyses of specific respiratory causes, COPD, and
31 pneumonia yielded even larger risk estimates (e.g., ~ 6 to 10 times, respectively, larger than that
June 2003 8-72 DRAFT-DO NOT QUOTE OR CITE
-------
1 for overall cardiovascular deaths). Estimated PM10 excess risks per 50 |ig/m3 PM10 (average of
2 0 through 6 day lags) were 1.2% (0.2, 2.3), 0.9% (-0.8, 2.7), 2.7% (-4.2, 10.1), 2.4% (-2.3, 7.4),
3 6.1% (1, 11.4), and 10.3% (3.7, 17.2), respectively, for total non-accidental, cardiovascular,
4 arrhythmia, heart failure, COPD, and pneumonia, using GAM models with stringent
5 convergence criteria. Thus, the results from this study with a large effective sample size also
6 confirm past observations that PM risk estimates for specific causes of cardiovascular or
7 respiratory mortality can be larger than those estimated for total non-accidental mortality.
8 As mentioned earlier in the multi-cities results section, Schwartz (2003) reanalyzed data
9 from Braga et al. (2001) to examine the lag structure of PM10 associations with specific causes of
10 mortality in ten U.S. cities. The pattern of larger PM10 excess risk estimates for respiratory
11 categories than for cardiovascular categories found in this study was similar to that in the Hoek
12 et al. analyses noted above. For example, the combined risk estimates across 10 cities per
13 50 |ig/m3 increase in PM10 (2-day mean) were 4.1% (2.5, 5.6), 7.7% (4.1, 11.5), and 11.0% (7,
14 15.1) for cardiovascular, COPD, and pneumonia, respectively, using GAM with stringent
15 convergence criteria. These values were even larger for unconstrained distributed lag models.
16 The Goldberg et al. (2000) study, and its reanalyses (Goldberg et al., 2003; Goldberg and
17 Burnett, 2003) in Montreal, CN, investigated the role of co-morbidity prior to deaths in
18 PM-mortality associations for various subcategories, including cancer, acute lower respiratory
19 disease, chronic coronary artery disease, and congestive heart failure (CHF). They could
20 classify deaths into these subcategories using medical records from the universal Quebec Health
21 Insurance Plan (QHIP). This way of classifying deaths would presumably take into account
22 more detailed information on the disease condition prior to death than the "underlying cause" in
23 the death records. Thus, the PM-mortality associations could be compared by using
24 subcategories classified from death records versus those classified from QHIP medical records.
25 The Goldberg and Burnett (2003) reanalysis found that total non-accidental mortality (which
26 was significantly associated with PM indices in the original report using GAM with default
27 convergence criteria) was not associated with PM indices in GLM models. They reported that
28 the associations between PM and non-accidental mortality were rather sensitive to weather
29 model specification and did not find significant PM associations with most of the subcategories
30 as defined from either QHIP or underlying cause. However, they did find significant
31 associations between CoH, NO2, and SO2 and the CHF deaths as defined from QHIP, but not the
June 2003 8-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 CHF deaths as defined from underlying cause. The association was even stronger in warm
2 seasons. It should be noted, however, that while the period for this study was relatively long
3 (-10 years) and the counts for the total non-accidental deaths were not small (median = 36
4 deaths per day), the counts for various subcategories were quite small (e.g., CHF underlying
5 cause mortality mean = 0.75 per day).
6 A recent study (Gouveia and Fletcher, 2000), using data from Sao Paulo, Brazil, 1991-
7 1993, examined child mortality (age under 5 years). The Poisson auto-regressive model
8 included parametric terms (e.g., quadratic, two-piece linear temperature etc.) to adjust for
9 weather and temporal trends. Although Gouveia and Fletcher found significant associations
10 between air pollution and elderly mortality, they did not find statistically significant associations
11 between air pollution and child respiratory mortality (the PM10 coefficient was negative and not
12 significant). However, it should be noted that the average daily respiratory mortality counts for
13 this study were relatively small (~2.4/day). With the modest length of observations (3 years),
14 the statistical power of the data was likely less than desirable, and there may not have been
15 sufficient power to elucidate the range of short-term PM effects on child respiratory mortality.
16 Again, evaluation of the role of varying contributing conditions to PM-mortality associations are
17 often challenged by the sample size problem.
18 Overall, then, the above assessment of newly available studies provides interesting
19 additional new information with regard to cause-specific mortality related to ambient PM. That
20 is, a growing number of studies continue to report increased cardiovascular- and respiratory -
21 related mortality risks as being significantly associated with ambient PM measures at one or
22 another varying lag times. When specific subcategories of cardiovascular disease were
23 examined in a large population (The Netherlands study by Hoek et al.), some of the
24 subcategories such as heart failure were more strongly associated with PM and other pollutants
25 than total cardiovascular mortality. Largest effect estimates are most usually reported for 0-1
26 day lags (with some studies also now noting a second peak at 3-4 day lags). A few of the newer
27 studies also report associations of PM metrics with "other" (i.e., non-cardiorespiratory) causes,
28 as well. However, at least some of these "other" associations may also be due to seasonal cycles
29 that include relationships to peaks in influenza epidemics that may imply respiratory
30 complications as a contributing cause to the "other" deaths. Alternately, the "other" category
31 may include sufficient numbers of deaths due to diabetes or other diseases which may also
June 2003 8-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 involve cardiovascular complications as contributing causes. Varying degrees of robustness of
2 PM effects are seen in the newer studies, as typified by PM estimates in multiple pollutant
3 models containing gaseous co-pollutants. That is, some studies show little effect of gaseous
4 pollutant inclusion on estimated PM effect sizes, some show larger reductions in PM effects to
5 non-significant levels upon such inclusion, and a number also report significant associations of
6 cardiovascular and respiratory effects with one or more gaseous co-pollutants. Thus, the newer
7 studies both further substantiate PM effects on cardiovascular- and respiratory-related mortality,
8 while also pointing toward possible significant contributions of gaseous pollutants to such cause-
9 specific mortality. The magnitudes of the PM effect size estimates are consistent with the range
10 of estimates derived from the few earlier available studies assessed in the 1996 PM AQCD.
11
12 8.2.2.6 Salient Points Derived from Assessment of Studies of Short-Term Particulate
13 Matter Exposure Effects on Mortality
14 The most salient key points to be extracted from the above discussion of newly available
15 information on short-term PM exposures relationships to mortality can be summarized as follow:
16 PM10 effects estimates. Since the 1996 PM AQCD, there have been more than 80 new
17 time-series PM-mortality analyses published. Estimated mortality relative risks in these studies
18 are generally positive, statistically significant, and consistent with the previously reported PM-
19 mortality associations. However, due to the concerns regarding the GAM convergence issue,
20 quantitative evaluations were made here based only on the studies that either did not use GAM
21 Poisson model with default convergence criteria or on those studies that have reanalyzed the data
22 using more stringent convergence criteria and/or used fully parametric approaches. Of particular
23 importance are several studies which evaluated multiple cities using consistent data analytical
24 approaches. The NMMAPS analyses for the largest 90 U.S. cities (Samet et al., 2000a,b;
25 Dominici et al., 2002 and 2003), derived a combined nationwide excess risk estimate of about
26 1.4% (1.1% using GLM) increase in total (non-accidental) mortality per 50 |ig/m3increase in
27 PM10. Other well-conducted multi-city analyses, as well as various single city analyses, obtained
28 larger PM10-effect size estimates for total non-accidental mortality, generally falling in the range
29 of 2 to 3.5% per 50 |ig/m3increase in PM10. This is consistent with, but somewhat lower than,
30 the range of PM10 risk estimates given in the 1996 PM AQCD. However, somewhat more
31 geographic heterogeneity is evident among the newer multi-city study results than was the case
32 among the fewer studies assessed in the 1996 PM AQCD. In the NMMAPS analysis of the 90
June 2003 8-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 largest U.S. cities data, for example, the risk estimates varied by U.S. geographic region, with
2 the estimate for the Northeast being the largest (approximately twice the nation-wide estimates).
3 The observed heterogeneity in the estimated PM risks across cities/regions could not be
4 explained by city-specific explanatory variables, such as mean levels of pollution and weather,
5 mortality rate, sociodemographic variables (e.g., median household income), urbanization, or
6 variables related to measurement error. Notable apparent heterogeneity was also seen among
7 effects estimates for PM (and SO2) indices in the multi-city APHEA studies conducted in
8 European cities. In APHEA2, they found that several city-specific characteristics, such as NO2
9 levels and warm climate, were important effect modifiers. The issue of heterogeneity of effect
10 estimates is discussed further in Section 8.4.
11 Model specification Issue: The investigations of the GAM convergence issue also led to
12 examination of the sensitivity of the PM risk estimates to different model specifications. Several
13 reanalyses examined the sensitivity of results to varying the degrees of freedom for smoothing of
14 weather and temporal trends. PM risk estimates were often reduced when more degrees of
15 freedom were given to model temporal trends. While what constitutes an "adequate" extent of
16 smoothing (from an epidemiologic viewpoint) is currently not known, the overall assessment of
17 PM risk estimates should take into consideration the range of sensitivity of results to this aspect
18 of model specification.
19 Confounding and effect modification by other pollutants. Numerous new short-term PM
20 exposure studies not only continue to report significant associations between various PM indices
21 and mortality, but also between gaseous pollutants (O3, SO2, NO2, and CO) and mortality.
22 In most of these studies, simultaneous inclusions of gaseous pollutants in the regression models
23 did not meaningfully affect the PM-effect size estimates. This was the case for the NMMAPS
24 90 cities study with regard to the overall combined U.S. regional and nationwide risk estimates
25 derived for that study. The issue of confounding is discussed further in Section 8.4.
26 Fine and coarse particle effects. Newly available studies provide generally positive (and
27 often statistically significant) PM2 5 associations with mortality, with effect size estimates falling
28 in the range reported in the 1996 PM AQCD. New results from Germany appear to implicate
29 both ultrafine (nuclei-mode) and accumulation-mode fractions of urban ambient fine PM as
30 being important contributors to increased mortality risks. As to the relative importance of fine
31 and coarse particles, in the 1996 PM AQCD there was only one acute mortality study (Schwartz
June 2003 8-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 et al., 1996a) that examined this issue. The results of that study of six U.S. cities suggested that
2 fine particles (PM2 5), were associated with daily mortality, but not coarse particles (PM10_2 5),
3 except for in Steubenville, OH.. Now, eight studies have analyzed both PM25 and PM10_25 for
4 their associations with mortality. While the results from some of these new studies (e.g., the
5 Santa Clara County, CA analysis [Fairley, 1999]) did suggest that PM2 5 was more important
6 than PM10_2 5 in predicting mortality fluctuations, other studies (e.g., Phoenix, AZ analyses
7 [Clyde et al., 2000; Mar et al., 2000; Smith et al., 2000]) suggest that PM10_2 5 may also be
8 important in at least some locations. Seasonal dependence of size-related PM component effects
9 observed in some of the studies complicates interpretations.
10 Chemical components ofPM. Several new studies have examined the role of specific
11 chemical components of PM. The studies conducted in U.S., Canadian, and European cities
12 showed mortality associations with specific fine particle components of PM, including sulfate,
13 nitrate, and CoH; but their relative importance varied from city to city, likely depending on their
14 levels (e.g., no clear associations in those cities where H+ and sulfate levels were very low, i.e.,
15 circa non-detection limits). The results of several studies that investigated the role of crustal
16 particles, although somewhat mixed, overall do not appear to support associations between
17 crustal particles and mortality (see also the discussion of source-oriented evaluations presented
18 below).
19 Source-oriented evaluations. Several studies conducted source-oriented evaluations of PM
20 components using factor analysis. The results from these studies generally indicated 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 total non-accidental mortality in the Harvard Six Cities data,
24 and the soil (i.e., crustal) factor from fine particles in the Phoenix data was not associated with
25 cardiovascular mortality. Thus, the source-oriented evaluations seem to implicate fine particles
26 of anthropogenic origin as being most important in contributing to increased mortality, but
27 generally do not support increased mortality risks being related to short-term exposures to crustal
28 materials in U.S. ambient environments.
29 Cause-specific mortality. Findings for new results concerning cause-specific mortality
30 comport well with those for total (non-accidental) mortality, the former showing generally larger
31 effect size estimates for cardiovascular, respiratory, and/or combined cardiorespiratory excess
June 2003 8-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 risks than for total mortality risks. An analysis of specific cardiovascular causes in a large
2 population (The Netherlands) suggested that specific causes of deaths (such as heart failure)
3 were more strongly associated with PM (and other pollutants) than total cardiovascular
4 mortality.
5 Lags. In general, maximum effect sizes for total mortality appear to be obtained with 0-1
6 day lags, with some studies indicating a second peak for 3-4 days lags. There is also some
7 evidence that, if effects distributed over multiple lag days are considered, the effect size may be
8 larger than for any single maximum-effect-size lag day. Lags are discussed further in
9 Section 8.4.
10 Threshold. Few new short-term mortality studies explicitly address the issue of thresholds.
11 One study that analyzed Phoenix, AZ data (Smith et al., 2000) did report some limited evidence
12 suggestive of a possible threshold for PM25. However, several different analyses of larger PM10
13 data sets across multiple cities (Dominici, et al., 2002; Daniels et al., 2000; and reanalysis by
14 Dominici et al., 2003) generally provide little or no support to indicate a threshold for PM10
15 mortality effects. Threshold issues are discussed further in Section 8.4.
16
17 8.2.3 Mortality Effects of Long-Term Exposure to Ambient
18 Particulate Matter
19 8.2.3.1 Studies Published Prior to the 1996 Particulate Matter Criteria Document
20 8.2.3.1.1 Aggregate Population Cross-Sectional Chronic Exposure Studies
21 Mortality effects associated with chronic, long-term exposure to ambient PM have been
22 evaluated in cross-sectional studies and, more recently, in prospective cohort studies. A number
23 of older cross-sectional studies from the 1970s provided indications of increased mortality
24 associated with chronic (annual average) exposures to ambient PM, especially with respect to
25 fine mass or sulfate (SO4"2) concentrations. However, questions unresolved at that time
26 regarding the adequacy of statistical adjustments for other potentially important covariates (e.g.,
27 cigarette smoking, economic status, etc.) across cities tended to limit the degree of confidence
28 that was placed by the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) on such
29 purely "ecological" studies or on quantitative estimates of PM effects derived from them.
30 Evidence comparing the toxicities of specific PM components was relatively limited, although
31 the sulfate and acid components were discussed in detail in the 1986 PM AQCD (U.S.
32 Environmental Protection Agency, 1986).
June 2003 8-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.3.1.2 Semi-Individual (Prospective Cohort) Chronic Exposure Studies
2 Prospective cohort, semi-individual studies of mortality associated with chronic exposures
3 to air pollution of outdoor origins have yielded especially valuable insights into the adverse
4 health effects of long-term PM exposures. Such semi-individual cohort studies using subject-
5 specific information about relevant covariates (such as cigarette smoking, occupation, etc.)
6 typically are capable of providing more certain findings of long-term PM exposure effects than
7 are purely "ecological studies" (Kiinzli and Tager, 1997). The new, better designed cohort
8 studies, as discussed below, have largely confirmed the magnitude of PM effect estimates
9 derived from past cross-sectional studies.
10 The extensive Harvard Six-Cities Study (Dockery et al., 1993) and the American Cancer
11 Society (ACS) Study (Pope et al., 1995) agreed in their findings of statistically significant
12 positive associations between fine particles and excess mortality, although the ACS study did not
13 evaluate the possible contributions of other air pollutants. Neither study considered multi-
14 pollutant models, although the Six-City study did examine various PM and gaseous pollutant
15 indices (including total particles, PM25, SO4"2, H+, SO2, and ozone), and found that sulfate and
16 PM2 5 fine particles were most strongly associated with mortality. The excess RR estimates
17 originally reported for total mortality in the Six-Cities study (and 95 percent confidence
18 intervals, CI) per increments in PM indicator levels were: Excess RR = 18% (CI = 6.8%, 32%)
19 for 20 |ig/m3 PM10; excess RR = 13.0% (CI = 4.2%, 23%) for 10 |ig/m3 PM25; and excess RR =
20 13.4% (CI = 5.1%, 29%) for 5 |ig/m3 SO4'2. The estimates for total mortality derived from the
21 ACS study were excess RR = 6.6% (CI = 3.5%, 9.8%) for 10 |ig/m3 PM25 and excess RR 3.5%
22 (CI = 1.9%, 5.1%) for 5 |ig/m3 SO4'2. The ACS pollutant RR estimates were smaller than those
23 from the Six-Cities study, although their 95% confidence intervals overlap. In some cases in
24 these studies, the life-long cumulative exposure of the study cohorts included distinctly higher
25 past PM exposures, especially in cities with historically higher PM levels (e.g., Steubenville,
26 OH); but more current PM measurements were used to estimate the chronic PM exposures.
27 In the ACS study, the pollutant exposure estimates were based on concentrations at the start of
28 the study (during 1979-1983). In addition, the average age of the ACS cohort was 56, which
29 could overestimate the pollutant RR estimates and perhaps underestimate the life-shortening
30 associated with PM associated mortality. Still, although caution must be exercised regarding use
31 of the reported quantitative risk estimates, the Six-Cities and ACS semi-individual studies
June 2003 8-79 DRAFT-DO NOT QUOTE OR CITE
-------
1 provided consistent evidence of significant mortality associations with long-term exposure to
2 ambient PM.
3 In contrast to the Six-Cities and ACS studies, early results reported by Abbey et al. (1991)
4 and Abbey et al. (1995a) from another prospective cohort study, the Adventist Health Study on
5 Smog (AHSMOG), found no significant mortality effects of previous PM exposure in a
6 relatively young cohort of California nonsmokers. However, these analyses used TSP as the PM
7 exposure metric, rather than more health-relevant PM metrics such as PM10 or PM2 5, included
8 fewer subjects than the ACS study, and considered a shorter follow-up time than the Six-Cities
9 study (ten years versus 15 years for the Six-Cities study). Further, the AHSMOG study included
10 only nonsmokers (indicated by the Six-Cities Study as having lower pollutant RR's than
11 smokers), suggesting that a longer follow-up time than considered in the past (10 years) might be
12 required to have sufficient power to detect significant pollution effects than would be needed in
13 studies that include smokers (such as the Six-Cities and ACS studies). Thus, greater emphasis
14 was placed in the 1996 PM AQCD on the results of the Six-Cities and ACS studies.
15 Overall, the previously available chronic PM exposure studies collectively indicated that
16 increases in mortality are associated with long-term exposure to ambient airborne particles; and
17 effect size estimates for total mortality associated with chronic PM exposure indices appeared to
18 be much larger than those reported from daily mortality PM studies. This suggested that a major
19 fraction of the reported mortality relative risk estimates associated with chronic PM exposure
20 likely reflects cumulative PM effects above and beyond those exerted by the sum of acute
21 exposure events (i.e., assuming that the latter are fully additive over time). The 1996 PM AQCD
22 (Chapter 12) reached several conclusions concerning four key questions about the prospective
23 cohort studies, as noted below:
24
25 (1) Have potentially important confounding variables been omitted?
26 "While it is not likely that the prospective cohort studies have overlooked plausible
27 confounding factors that can account for the large effects attributed to air pollution, there may be
28 some further adjustments in the estimated magnitude of these effects as individual and
29 community risk factors are included in the analyses." These include individual variables such as
30 education, occupational exposure to dust and fumes, and physical activity, as well as ecological
June 2003 8-80 DRAFT-DO NOT QUOTE OR CITE
-------
1 (community) variables such as regional location, migration, and income distribution. Further
2 refinement of the effects of smoking status may also prove useful."
3
4 (2) Can the most important pollutant species be identified?
5 "The issue of confounding with co-pollutants has not been resolved for the prospective
6 cohort studies . . . Analytical strategies that could have allowed greater separation of air pollutant
7 effects have not yet been applied to the prospective cohort studies." The ability to separate the
8 effects of different pollutants, each measured as a long-term average on a community basis, was
9 clearly most limited in the Six Cities study. The ACS study offered a much larger number of
10 cities, but did not examine differences attributable to the spatial and temporal differences in the
11 mix of particles and gaseous pollutants across the cities. The AHSMOG study constructed time-
12 and location-dependent pollution metrics for most of its participants that might have allowed
13 such analyses, but no results were reported.
14
15 (3) Can the time scales for long-term exposure effects be evaluated?
16 "Careful review of the published studies indicated a lack of attention to this issue. Long-
17 term mortality studies have the potential to infer temporal relationships based on characterization
18 of changes in pollution levels over time. This potential was greater in the Six Cities and
19 AHSMOG studies because of the greater length of the historical air pollution data for the cohort
20 [and the availability of air pollution data throughout the study]. The chronic exposure studies,
21 taken together, suggest that there may be increases in mortality in disease categories that are
22 consistent with long-term exposure to airborne particles, and that at least some fraction of these
23 deaths are likely to occur between acute exposure episodes. If this interpretation is correct, then
24 at least some individuals may experience some years of reduction of life as a consequence of PM
25 exposure."
26
27 (4) Is it possible to identify pollutant thresholds that might be helpful in health assessments?
28 "Model specification searches for thresholds have not been reported for prospective cohort
29 studies. . . . Measurement error in pollution variables also complicates the search for potential
30 threshold effects. . . . The problems that complicate threshold detection in the population-based
31 studies have a somewhat different character for the long-term studies."
June 2003 8-81 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.3.2 New Prospective Cohort Analyses of Mortality Related to Chronic Particulate
2 Matter Exposures
3 Considerable further progress has been made towards addressing the above issues. As an
4 example, extensive reanalyses (Krewski et al., 2000) of the Six-Cities and ACS Studies
5 (sponsored by HEI), indicate that the published findings of the original investigators (Dockery
6 et al., 1993; Pope et al., 1995) are based on substantially valid data sets and statistical analyses.
7 The HEI reanalysis project demonstrated that small corrections in input data have very little
8 effect on the findings and that alternative model specifications further substantiate the robustness
9 of the originally reported findings. In addition, some of the above key questions have been
10 further investigated by Krewski et al. (2000) via sensitivity analyses (in effect, new analyses) for
11 the Six City and ACS studies data sets, including consideration of a much wider range of
12 confounding variables. Newly published analyses of ACS data for more extended time periods
13 (Pope et al., 2002) further substantiate original findings and also provide much clearer, stronger
14 evidence for ambient PM exposure relationships with increased lung cancer risk. Newer
15 published analyses of AHSMOG data (Abbey et al., 1999; Beeson et al., 1998) also extend the
16 ASHMOG findings and show some analytic outcomes different from earlier analyses reported
17 out from the study. Results from the Veterans' Administration- Washington University
18 (hereafter called "VA") prospective cohort study are also now available (Lipfert et al., 2000b).
19 Other additional, new studies suggestive of possible effects of sub-chronic PM exposures on
20 fetal and infant development/mortality (Woodruff et al., 1997; Bobak and Leon, 1998; Lipfert,
21 2000; Chen et al., 2002) are also discussed below.
22
23 8.2.3.2.1 Health Effects Institute Reanalyses of the Six-Cities and ACS Studies
24 The overall objective of the HEI "Particle Epidemiology Reanalysis Project" was to
25 conduct a rigorous and independent assessment of the findings of the Six Cities (Dockery et al.,
26 1993) and ACS (Pope et al., 1995) Studies of air pollution and mortality. The following
27 description of approach, key results, and conclusions is largely extracted from the Executive
28 Summary of the HEI final report (Krewski et al., 2000). The HEI-sponsored reanalysis effort
29 was approached in two steps:
30
June 2003 8-82 DRAFT-DO NOT QUOTE OR CITE
-------
1 • Part I: Replication and Validation. The Reanalysis Team sought to test (a) whether the
original studies could be replicated via a quality assurance audit of a sample of the original
data and (b) whether the original numeric results could be validated.
2 • Part II: Sensitivity Analyses. The Reanalysis Team tested the robustness of the original
analyses to alternate risk models and analytic approaches.
3 The Part I audit of the study population data for both the Six Cities and ACS Studies and of
4 the air quality data in the Six Cities Study revealed that data were of generally high quality with
5 few exceptions. In both studies, a few errors were found in the data coding for and exclusion of
6 certain subjects; but when those subjects were included in the analyses, they did not materially
7 change the results from those originally reported. Because the air quality data used in the ACS
8 Study could not be audited, a separate air quality database was constructed for the sensitivity
9 analyses in Part II.
10 The Reanalysis Team was able to replicate the original results for both studies using the
11 same data and statistical methods as used by the original investigators, as shown in Table 8-5.
12 The Reanalysis Team confirmed the original point estimates. For the Six Cities Study, they
13 reported the excess relative risk of mortality from all causes associated with an increase in fine
14 particles of 10 |ig/m3 to be 14%, close to the 13% reported by the original investigators. For the
15 ACS Study, they reported the relative risk of all-cause mortality associated with a 10 |ig/m3
16 increase in fine particles to be 7.0% in the reanalysis, close to the original 6.6% value.
17 The Part II sensitivity analysis applied an array of different models and variables to
18 determine whether the original results would remain robust to different analytic assumptions and
19 model specifications. The Reanalysis Team first applied the standard Cox model used by the
20 original investigators and included variables in the model for which data were available from
21 both original studies, but had not been used in the published analyses (e.g., physical activity,
22 lung function, marital status). The Reanalysis Team also designed models to include interactions
23 between variables. None of these alternative models produced results that materially altered the
24 original findings.
25 Next, for both the Six Cities and ACS Studies, the Reanalysis Team investigated the
26 possible effects of fine particles and sulfate on a range of potentially susceptible subgroups of
27 the population. These analyses did not find differences in PM-mortality associations among
28 subgroups based on various personal characteristics (e.g., including gender, smoking status,
June 2003 8-83 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-5. 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'
Findings
Six Cityb
Six Cityb
ACS Study0
Indicator
PM25
PM15/10
PM25
Mortality Risk per Increment in PMa
Total Mortality
Excess Relative Risk (95% CI)
13% (4.2%, 23%)
18% (6.8%, 32%)
6.6% (3. 5%, 9.8%)
Cardiopulmonary Mortality
Excess Relative Risk (95% CI)
18% (6.0%, 32%)
e
12% (6.7%, 17%)
HEI reanalysis Phase I:
Replication
Six City Reanalysisd
ACS Study Reanalysisd
PM25
PM15
PM25
PM15 (dichot)
PM15 (SSI)
14% (5.4%, 23%)
19% (6.1%, 34%)
7.0% (3.9%, 10%)
4.1% (0.9%, 7.4%)
1.6% (-0.8%, 4.1%)
19% (6.5%, 33%)
20% (2.9%, 41%)
12% (7.4%, 17%)
7.3% (3.0%, 12%)
5.7% (2.5%, 9.0%)
"Estimates calculated on the basis of differences between the most-polluted and least-polluted cities, scaled to
increments of 20 ug/m3 increase for PM10 and 10 ug/m3 increments for PM15 and PM2 5.
bDockeryetal. (1993).
cPopeetal. (1995).
dKrewski et al. (2000).
eResults presented only by smoking category subgroup.
1 exposure to occupational dusts and fumes, and marital status). However, estimated effects of
2 fine particles did vary with educational level: the association between an increase in fine
3 particles and mortality tended to be higher for individuals without a high school education than
4 for those with more education. The Reanalysis Team postulated that this finding could be
5 attributable to some unidentified socioeconomic effect modifier. The authors concluded "The
6 Reanalysis Team found little evidence that questionnaire variables had led to confounding in
7 either study, thereby strengthening the conclusion that the observed association between fine
8 particle air pollution and mortality was not the result of a critical covariate that had been
9 neglected by the Original Investigators." (Krewski et al., 2000, pp. 219-220).
10 In the ACS study, the Reanalysis Team tested whether the relationship between ambient
11 concentrations and mortality was linear. They found some indications of both linear and
June 2003
8-84
DRAFT-DO NOT QUOTE OR CITE
-------
1 nonlinear relationships, depending upon the analytic technique used, suggesting that the shapes
2 of the concentration-response relationships warrant additional research in the future.
3 One of the criticisms of both original studies has been that neither analyzed the effects of
4 change in pollutant levels over time. In the Six Cities Study, for which such data were available,
5 the Reanalysis Team tested whether effect estimates changed when certain key risk factors
6 (smoking, body mass index, and air pollution) were allowed to vary over time. In general, the
7 reanalysis results did not change when smoking and body mass index were allowed to vary over
8 time. The Reanalysis Team did find for the Six Cities Study, however, that when the general
9 decline in fine particle levels over the monitoring period was included as a time-dependent
10 variable, the association between fine particles and all-cause mortality was reduced (Excess
11 RR = 10.4%, 95% CI = 1.5%, 20%). This would be expected, because the most polluted cities
12 would likely have the greatest decline as pollution controls were applied. Despite this
13 adjustment, the PM25 effect estimate continued to be positive and statistically significant.
14 To test the validity of the original ACS air quality data, the Reanalysis Team constructed
15 and applied its own air quality dataset from available historical data. In particular, sulfate levels
16 with and without adjustment were found to differ by about 10% for the Six Cities Study. Both
17 the original ACS Study air quality data and the newly constructed dataset contained sulfate
18 levels inflated by 50% due to artifactual sulfate. For the Six Cities Study, the relative risks of
19 mortality were essentially unchanged with adjusted or unadjusted sulfate. For the ACS Study,
20 adjusting for artifactual sulfate resulted in slightly higher relative risks of mortality from all
21 causes and cardiopulmonary disease compared with unadjusted data, while the relative risk of
22 mortality from lung cancer was lower after the data had been adjusted. Thus, the Reanalysis
23 Team found essentially the same results as the original Harvard Six-Cities and ACS studies,
24 even after using independently developed pollution data sets and adjusting for sulfate artifact.
25 Because of the limited statistical power to conduct most model specification sensitivity
26 analyses for the Six Cities Study, the Reanalysis Team conducted the majority of its sensitivity
27 analyses using only the ACS Study dataset that considered 151 cities. When a range of city -
28 level (ecologic) variables (e.g., population change, measures of income, maximum temperature,
29 number of hospital beds, water hardness) were included in the analyses, the results generally did
30 not change. The only exception was that associations with fine particles and sulfate were
31 reduced when city-level measures of population change or SO2 were included in the model.
June 2003 8-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 A major product of the Reanalysis Project is the determination that both pollutant variables
2 and mortality appear to be spatially correlated in the ACS Study dataset. If not identified and
3 modeled correctly, spatial correlation could cause substantial errors in both the regression
4 coefficients and their standard errors. The Reanalysis Team identified several methods for
5 addressing this, each of which resulted in some reduction in the estimated regression
6 coefficients. The full implications and interpretations of spatial correlations in these analyses
7 have not been resolved and were noted to be an important subject for future research.
8 When the Reanalysis Team sought to take into account both the underlying variation from
9 city to city (random effects) and variation from the spatial correlation between cities, positive
10 associations were still found between mortality and sulfates or fine particles. Results of various
11 models, using alternative methods to address spatial autocorrelation and including different
12 ecologic covariates, found fine particle-mortality associations that ranged from 1.11 to 1.29 (the
13 RR reported by original investigators was 1.17) per 24.5 |ig/m3 increase in PM2 5. With the
14 exception of SO2, consideration of other pollutants in these models did not alter the associations
15 found with sulfates. The authors reported associations that were stronger for SO2 than for
16 sulfate, which may indicate that artifactual sulfate was "picking up" some of the SO2 association,
17 perhaps because the sulfate artifact is in part proportional to the prevailing SO2 concentration
18 (Coutant, 1977). It should be recognized that the Reanalysis Team did not use data adjusted for
19 artifactual sulfate for most alternative analyses. When they did use adjusted sulfate data, relative
20 risks of mortality from all causes and cardiopulmonary disease increased. This result suggests
21 that more analyses with adjusted sulfate might result in somewhat higher relative risks associated
22 with sulfate. The Reanalysis Team concluded: "it suggests that uncontrolled spatial
23 autocorrelation accounts for 24% to 64% of the observed relation. Nonetheless, all our models
24 continued to show an association between elevated risks of mortality and exposure to airborne
25 sulfate" (Krewski et al., 2000, p. 230).
26 In summary, the reanalyses generally confirmed the original investigators' findings of
27 associations between mortality and long-term exposure to PM, while recognizing that increased
28 mortality may be attributable to more than one ambient air pollution component. Regarding the
29 validity of the published Harvard Six-Cities and ACS Studies, the HEI Reanalysis Report
30 concluded that "Overall, the reanalyses assured the quality of the original data, replicated the
31 original results, and tested those results against alternative risk models and analytic approaches
June 2003 8-86 DRAFT-DO NOT QUOTE OR CITE
-------
1 without substantively altering the original findings of an association between indicators of
2 participate matter air pollution and mortality."
3 In a further analyses of the Harvard Six City study cohort using a Poisson regression
4 model, Villeneuve et al. (2002) evaluated the relationship between fixed-in-time and time-
5 dependent measures of PM25 and the risk of mortality among adult, Caucasian participants. The
6 RR of mortality using the Poisson method based upon city-specific exposures that remained
7 constant during the follow up was 1.31 (CI =1.12- 1.52), which is similar to results derived
8 from the Cox model used in the original analysis. However, the authors report that "The RR of
9 mortality due to PM2 5 exposure decreased when time-dependent measures of air pollution were
10 modeled (Table 8-6). Specifically, when the mean PM25 level within each city during each
11 period of follow-up was modeled, the RR was 1.16 (95% CI = 1.02 - 1.32). The authors noted
12 that "there were considerable variations in mortality rates across the calendar periods that were
13 modeled," and that "the magnitude of these variations in mortality rates may have dampened any
14 real PM2 5 effect on mortality." Villeneuve et al. (2002) concluded that the "attenuated risk of
15 mortality that was observed with a time-dependent index of PM2 5 is due to the combined
16 influence of city-specific variations in mortality rates and decreasing levels of air pollution that
17 occurred during follow-up."
18 Similar results were observed by Villeneuve et al. (2002) irrespective of the exposure
19 window considered. They used various time-dependent indices denoting exposures received in
20 the last two years of follow-up and (b) for exposures lagged 3-4 and > 5 years. Effect
21 modification was evaluated by fitting interaction terms that consisted of PM25 exposure and
22 individual risk factors (body mass index, education, smoking, age, gender, and occupational
23 exposure to dusts). The significance of this term was formally tested by constructing a
24 likelihood ratio test statistic. An interaction effect between PM2 5 exposure and age was
25 observed (p < 0.05), and they therefore presented stratified analysis by age group (< 60,
26 > 60 years). For each index of PM25, the RR of all-cause mortality was more pronounced among
27 subjects < 60 years old. There was no effect modification between PM25 and the other
28 individual risk factors. The RR for PM-associated mortality did not depend on when exposure
29 occurred in relation to death, possibly because dof little variation between the time-dependent
30 city-specific PM25 exposure indices (r > 0.9) and the fact that the rank ordering of the cities
31 changed little during follow-up.
June 2003 8-87 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-6. RELATIVE RISK3 OF ALL-CAUSE MORTALITY FOR
SELECTED INDICES OF EXPOSURE TO FINE PARTICULATE MATTER
(per 18.6 jig/m3) BASED ON MULTIVARIATE POISSON REGRESSION ANALYSIS,
BY AGE GROUP, FOR HARVARD SIX CITY STUDY DATAB
Age Group (years)
Model PM2 s Exposure City Specific Index Total <60 :> 60
1 Exposure to PM2 5 remained fixed over 1.31(1.12-1.52) 1.89(1.32-2.69) 1.21(1.02-1.43)
the entire follow up period.
2 Exposure to PM2 5 was defined according 1.19(1.04-1.36) 1.52(1.15-2.00) 1.11(0.95-1.29)
to 13 calendar periods (no smoothing).3
3 Exposure to PM25 was defined according 1.16 (1.02 - 1.32) 1.43 (1.10 - 1.85) 1.09 (0.93 - 1.26)
to 13 calendar periods (smoothed).15
4 Time dependent estimate of PM2 5 1.16(1.02-1.31) 1.42(1.09-1.82) 1.08(0.94-1.25)
received during the previous two years.
5 Time dependent estimate of PM2 5 1.14(1.02-1.27) 1.35(1.08-1.87) 1.08(0.95-1.22)
received 3-5 years before current year.
6 Time dependent estimate of PM2 5 1.14(1.05-1.23) 1.34(1.11-1.59) 1.09(0.99-1.20)
received > 5 years before current year.
a Relative risks were adjusted by age, gender, body mass, index, education, number of years smoked (at baseline),
occupational exposures and number of cigarettes smoked weekly.
b For each city, exposure to PM2 5 was estimated for 13 calendar periods using loglinear regression based on
annual mean PM2 5 levels. The calendar periods used were: 1970-1978, 1979, 1981,. . . 1989, and 1990+.
PM2.5 associations with all-cause mortality assessed for male Caucasian participants in Six Cities Study.
Source: Villeneuve et al. (2002).
1 8.2.3.2.2 The ACS Study Extension
2 Pope et al. (2002) extended the analyses (Pope et al., 1995) and reanalyses (Krewski et al.,
3 2000) of the ACS CPS-II cohort to include an additional eight years of follow-up data. The new
4 study has a number of advantages over the previous analyses, in that it (a) doubles the follow-up
5 time from eight to sixteen years and triples the number of deaths; (b) expands the ambient air
6 pollution data substantially, including two recent years of fine particle data and adding data on
7 gaseous co-pollutants; (c) improves statistical adjustments for occupational exposure;
8 (d) incorporates data on dietary covariates believed to be important factors in mortality,
9 including total fat consumption, and consumption of vegetables, citrus fruit, and high-fiber
10 grains; and (e) uses recent developments in non-parametric spatial smoothing and random effects
11 statistical models as input to the Cox proportional hazards model. Each participant was
June 2003 8-88 DRAFT-DO NOT QUOTE OR CITE
-------
1 identified with a specific metropolitan area, and mean pollutant concentrations were calculated
2 for all metropolitan areas with ambient air monitors in the one to two years prior to enrollment.
3 Ambient pollution during the follow-up period was extracted from the AIRS data base.
4 Averages of daily averages of the gaseous pollutants were used except for ozone, where the
5 average daily 1-hour maximum was calculated for the whole year and for the typical peak ozone
6 quarter (July, August, September). Mean sulfate concentrations for 1990 were calculated from
7 archived quartz filters, virtually eliminating the historical sulfate artifact leading to
8 overestimation of sulfate concentrations.
9 The Krewski et al. (2000), Burnett et al. (200la), and Pope et al. (2002) studies were
10 concerned that survival times of participants in nearby locations might not be independent of
11 each other, due to missing, unmeasured, or mis-measured risk factors or their surrogates that
12 may be spatially correlated with air pollution, thus violating an important assumption of the Cox
13 proportional hazards model. Thus, model fitting proceeded in two stages, the first of which was
14 an adjusted relative risk model with a standard Cox proportional hazards model including
15 individual-specific covariates and indicator variables for each metropolitan area, but not air
16 pollutants. In the second stage, the adjusted log(relative risks) were fitted to fine particle
17 concentrations or other air pollutants by a random effects linear regression model.
18 Models were estimated separately for each of four mortality (total, cardiopulmonary, lung
19 cancer, and causes other than cardiopulmonary or lung cancer deaths) endpoints for the entire
20 follow-up period and for fine particles in three time periods (1979-1983, 1999-2000, and the
21 average of the mean concentrations in these two periods). The results are shown in Table 8-7.
22 Figures 8-9, 8-10, and 8-11 show the results displayed in Figures 2, 3, and 5 of Pope et al.
23 (2002). Figure 8-9 shows that a smooth non-parametric model can be reasonably approximated
24 by a linear model for all-cause mortality, cardiopulmonary mortality, and other mortality; but the
25 log(relative risk) model for lung cancer appears to be non-linear, with a steep linear slope up to
26 an annual mean concentration of about 13 |ig/m3 and a flatter linear slope at fine particle
27 concentrations > 13 |ig/m3.
28 Figure 4 in Pope et al. (2002) shows results for the stratified first-stage models: ages
29 < 60 and > 69 yr are marginally significant for total mortality; ages > 70 are significant for
30 cardiopulmonary mortality; and ages 60-69 for lung cancer mortality. Men are at significantly
31 higher risk for total and lung cancer mortality than are women, but slightly less so for
June 2003 8-89 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-7. SUMMARY OF RESULTS FROM THE EXTENDED ACS STUDY*
Cause of death
All causes
Cardiopulmonary
Lung cancer
Other
PM2 s, average over
1979-1983
4.1% (0.8,7.5%)
5.9% (1.5, 10.5%)
8.2% (1.1, 15.8%)
0.8% (-3.0,4.8%)
PM2 s, average over
1999-2000
5.9% (2.0, 9.9%)
7.9% (2.3, 14.0%)
12.7% (4.1,21.9%)
0.9% (-3.4,5.5%)
PM2 s, average over all
seven years
6.2% (1.6, 11.0%)
9.3% (3.3, 15.8%)
13.5% (4.4,23.4%)
0.5% (-4.8,6.1%)
'Adjusted mortality excess risk ratios (95% confidence limits) per 10 ug/m3 PM25 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 cardiopulmonary mortality (although still significant). Log(RR) decreases significantly from
2 individuals with less than to those with more than a high school education, replicating findings
3 in Krewski et al. (2000), but with twice the time on study. Including smoking status showed
4 increased fine particle RR for cardiopulmonary and lung cancer mortality in never-smokers and
5 least effect in current smokers; however, for total mortality, significant or near-significant effects
6 occurred in both current and never-smokers, but not former smokers.
7 The second-stage random effects models on the right side of Figure 8-10 have much wider
8 confidence intervals than the first-stage models, but are still statistically significant for total,
9 cardiopulmonary, and lung cancer mortality. Spatial smoothing decreased the magnitude and
10 significance of the fine particle effect for total mortality. For cardiopulmonary mortality, spatial
11 smoothing increased the magnitude of the RR and its significance by reducing the width of the
12 confidence intervals in the "50%-span" and "lowest variance" smoothing methods. For lung
13 cancer mortality, spatial smoothing little changed the magnitude of the RR, but increased its
14 significance by reducing the width of confidence intervals in the "50%-span" and "lowest
15 variance" smoothing methods.
16 Figure 8-11 shows statistically significant relationships between fine particles and total,
17 cardiopulmonary, and lung cancer mortality no matter which averaging span was used for PM25
18 and slightly larger effect estimates for the average concentration of the 1979-1983 and 1999-
19 2000 intervals. PM15 for 1979-1983 is significantly associated with cardiopulmonary mortality
June 2003 8-90 DRAFT-DO NOT QUOTE OR CITE
-------
01
%
0
0.2
0.1
0.0
-0.1
-0.3
-0.4
All Cause
0.2 -
^ 0.1 -
en
| °-°-
1g -0.1 -
0)
o
-1 -0.3 -
-0.4 -
Cardiopulmonary
10
15
20
10
15
20
jjg/m3
Q)
*-
0.2
0.1
-0.1 -
I
-0.3
-0.4
Lung Cancer
0.2
0.1
>
j3 -0.1
0)
-0.4 -
All Other
10
15
20
10
15
20
Figure 8-9. Natural logarithm of relative risk for total and cause-specific mortality per
10 ug/m3 PM2 5 (approximately the excess relative risk as a fraction), with
smoothed concentration-response functions. Based on Pope et al. (2002) mean
curve (solid line) with pointwise 95% confidence intervals (dashed lines).
1 and marginally with total mortality; whereas 1987-1996 PM15 is not quite significantly
2 associated with cardiopulmonary mortality. Coarse particles (PM15.2 5) and TSP are not
3 significantly associated with any endpoint, but are positively associated with cardiopulmonary
4 mortality. Sulfate particles are very significantly associated with all endpoints, including
5 mortality from all other causes, but only marginally for lung cancer mortality using 1990 filters.
6 Figure 8-11 also shows highly positive significant relationships between SO2 and total,
1 cardiopulmonary, and other-causes mortality, but a weaker SO2 association with lung cancer
8 mortality. Only ozone using only the third quarter for 1982-1998 showed a marginally
9 significant relationship with cardiopulmonary mortality, but not the year-round average. The
10 other criteria pollutants, CO and NO2, are neither significantly nor positively related to any
11 mortality endpoint, unlike some findings for acute PM exposure-mortality studies.
June 2003
8-91
DRAFT-DO NOT QUOTE OR CITE
-------
a, Random Effects
1.15-
1.10-
1.05-
1.00-
0.95-
1.15-
1.10-
1.05-
1.00-
0.95-
1.15-
1.10-
1.05-
1.00-
0.95-
1.15-
tS Standard COX Panrfnm (with all covariates)
_ . KaMUum ninr- Oiii*i-il
|| „,„ Models ^ PffprtQ plus Spatial
& ^ °° - C Cl 1 CCIS On-inntKinn
|| | || _ |i (with an , Smoothing
If 1 ll I 1 !| .1 cova"ates) || || ||
j + ys s? < °° ° T -r =5 f>
]
f i
S i
I 1
> 1
f ^
\ i
4
>•
*
4
• )
; ^
f
All-Cause
h
ij
\ ]
> ]
r ^
[ ]
f ;
•
> ^
:H
>
Cardiopulmonary
^
^ 1
• .
. „
•4
> 4
» 4
* <
t
Lung Cancer
1.10-
1.05-
1.00-
0.95 .
* { ]
f f i-
All Other
i i :
*
<
\ 1 }
Figure 8-10. Relative risk of total and cause-specific mortality at 10 ug/m3 PM25 (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.
Source: Based on Pope et al. (2002).
June 2003
8-92
DRAFT-DO NOT QUOTE OR CITE
-------
Fine
Particles
00 O B
™ O "^
Inhalable, Coarse,
& Total Suspended
„ Particles
Sulfate
Particles
Gaseous
1.25-
1.20-
1.15-
1.10-
1.05-
1.00-
0.95-
0.90-
1.25-
1.20-
1.15-
1.10-
1.05-
1 on -
i.v v —
0.95-
0.90-
1.25-
1.20-
1.15-
1.10-
1.05-
1.00-
0.95-
0.90-
1.25-
1.20-
1.15-
1.10-
1.05-
1.00-
0.95-
0.90-
^^ O) r^j CD gs ^f
in S. E T tt *?
C1 5
4
'. i
1 4
I
4
>
n ,
Cp
— » ff
T
-, O
e~*l 1
-, O
-o> ^~.
3 ^< CO ^
'UN
Q ^ ,« /w^ D^J oo o O f^ * os
£• c
. o.
F
^ 4
4
r
L
^ o 0- « % *
e s. |2 a 4
, £
M
f 1
i
4
[ ]
P ?
'I
%, S. o" S '
« "S as £. e
O f~\ '~~f^ r*-' rt
p Vt % Jf1 C
UlOIIlb ^
Jo" 0°
Oi i
CJ W
_ ? CO S-
? g cp "S ^
11 w^ 1 w ^^
J 3T (M CO i-
T O u » *m
LI t"1""'*!"]
4
p
t»
•^
All-Cause
^
k
4
4
>
»•
4
-"-
4
I-
^
41
-*- ^
r <
^
r
p
4
'
» <
.
. 4
>
. .
p j
T X T
L T ....
^
•4]
^•••j'"4
4
p
^
p
^
Cardiopulmonary I
^
^ ^|
4
^
p
•^ 4
4
r
k^
4
-
4
••4
r
4
4
p
>• -
4
•i'f]
'""41
4
P .
T
k
T T....I
4r
^ Lung Cancer
. i
*
4
P
p
f 1LH
4
»
i4
!•
4
4
p
Ih
4,
4
p
-1- All Other 1
. •
"•<
r-
»• •"•
Figure 8-11. Relative risk of total and cause-specific mortality for particle metrics and
gaseous pollutants over different averaging periods (years 1979-2000 in
parentheses).
Source: Based on Pope et al. (2002).
June 2003
8-93
DRAFT-DO NOT QUOTE OR CITE
-------
1 This paper is noteworthy because it confirms that the general pattern of findings in the first
2 eight years of the study (Pope et al., 1995; Krewski et al., 2000) can be reasonably extrapolated
3 to the patterns that remain present with twice the length of time on study and three times the
4 number of deaths. As shown later in Table 8-11, the excess relative risk estimate (95% CI) per
5 10 |ig/m3 PM25 for total mortality in the original ACS study (Pope et al., 1995) was 6.6% (3.6,
6 9.9%); in the ACS reanalysis (Krewski et al., 2000) it was 7.0% (3.9, 10%); and, in the extended
7 ACS data set (Pope et al., 2002), it was 4.1% (0.8, 7.5%) using the 1979-1983 data and 6.2%
8 (1.6, 11%) using the average of the 1979-1983 and 1999-2000 data. The excess relative risk
9 estimate (95% CI) per 10 |ig/m3 PM25 for cardiopulmonary mortality in the original ACS study
10 (Pope et al., 1995) was 12% (6.7, 17%); in the ACS reanalysis (Krewski et al., 2000), it was 12%
11 (7.4, 17%); and, in the extended ACS data set (Pope et al., 2002), it was 5.9% (1.5, 10%) using
12 the 1979-1983 data and 9.3% (3.3, 16%) using the average of the 1979-1983 and 1999-2000
13 data. Thus, the additional data and statistical analyses reported in Pope et al. (2002) yield
14 somewhat smaller estimates than the original study (Pope et al., 1995), but are similar to
15 estimates from the (Krewski et al. (2000) reanalysis of the original ACS data set.
16 Based on the above patterns of results, the authors drew the following conclusions:
17 (1) The apparent association between long-term exposure to fine particle pollution and
mortality persists with longer follow-up as the participants in the cohort grow older and
more of them die.
18 (2) The estimated fine particle effect on cardiopulmonary mortality and cancer mortality
remained relatively stable even after adjustment for smoking status, although the
estimated effect was larger and more significant for never-smokers versus former or
current smokers. The estimates were relatively robust against inclusion of many
additional covariates: education, marital status, body mass index (BMI), alcohol
consumption, occupational exposure, and dietary factors. However, as the authors note,
the data on individual risk factors were collected only at the time of enrollment and have
not been updated, so that changes in these factors since 1982 could introduce risk-factor
exposure mis-classification and a consequent loss of precision in the estimates that might
limit the ability to characterize time dependency of effects. Moreover, it is noteworthy
that this study found education to be an effect modifier, with larger and more statistically
significant PM effect estimates for persons with less education. This may be due to the
June 2003 8-94 DRAFT-DO NOT QUOTE OR CITE
-------
fact that less-education is a marker for lower socio-economic status and, therefore,
poorer health status and greater pollution susceptibility. These results may also be an
indicator that the mobility of the less educated provides better estimates of effects in this
study (with no follow up of address changes) than for the more mobile well-educated.
In either case, because this cohort comprises a much higher percentage of well-educated
persons than the general public, the education effect modification seen suggests that the
overall PM effect estimates are likely underestimated by this study cohort versus that
which would be found for the general public.
19 (3) Additional assessments for potential spatial or regional differences not controlled in the
first-stage model were evaluated. If there are unmeasured or inadequately modeled risk
factors that are different across locations or spatially clustered, then PM risk estimates
may be biased. If the clustering is independent or random or independent across areas,
then adding a random-effects component to the Cox proportional hazards model can
address the problem. However, if location is associated with air pollution, then the
spatial correlation may be evaluated using non-parametric smoothing methods.
No significant spatial auto-correlation was found after controlling for fine particles.
Even after adjusting for spatial correlation, the estimated PM2 5 effects were significant
and persisted for cardiopulmonary mortality and lung cancer mortality and were
borderline significant for total mortality, but with much wider confidence intervals after
spatial smoothing.
20 (4) Fine particles (PM2 5) were associated with elevated total, cardiopulmonary, and lung
cancer mortality risks, but not other-cause mortality. PM10 for 1987-1996 and PM15 for
1979-1983 were just significantly associated with cardiopulmonary mortality, but
PM10_2 5 and TSP were not associated with total or any cause-specific mortality. All
endpoints but lung cancer mortality were very significantly associated with sulfates,
except for lung cancer with 1990 sulfate data. All endpoints except lung cancer
mortality were significantly associated with SO2 using 1980 data as were total and other
mortality using the 1982-1998 SO2 data; but cardiopulmonary and lung cancer mortality
had only a borderline significant association with the 1982-1998 SO2 data. None of the
other gaseous pollutants showed significant positive associations with any endpoint.
June 2003 8-95 DRAFT-DO NOT QUOTE OR CITE
-------
Thus, neither coarse thoracic particles nor TSP were significantly associated with
mortality; nor were CO and NO2 on a long-term exposure basis.
21 (5) The concentration-response curves estimated using non-parametric smoothers were all
monotonic and nearly linear (except for lung cancer). However, the shape of the curve
may become non-linear at much higher concentrations.
22 (6) The excess risk from PM2 5 exposure is much smaller than that estimated for cigarette
smoking for current smokers in the same cohort (Pope et al., 1995): RR = 2.07 for total
mortality, RR = 2.28 for cardiopulmonary mortality, and RR = 9.73 for lung cancer
mortality. In the more polluted areas of the United States, the relative risk for substantial
obesity (a known risk factor for cardiopulmonary mortality) is larger than that for PM2 5,
but 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), a third major U.S. prospective cohort
26 study of chronic PM exposure-mortality effects, started with enrollment in 1977 of
27 6,338 non-smoking non-Hispanic white Seventh Day Adventist residents of California, ages
28 27 to 95 years. All had resided for at least 10 years within 5 miles (8 km) of their then-current
29 residence locations, either within one of 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
32 and its initial results earlier reported elsewhere (Hodgkin et al., 1984; Abbey et al., 1991; Mills
33 etal., 1991).
34 In more recent AHSMOG analyses (Abbey et al., 1999), the mortality status of subjects
35 after ca. 15-years of follow-up (1977-1992) was determined by various tracing methods and
36 1,628 deaths (989 female, 639 male) were found in the cohort. This 50% percent increase during
37 the follow-up period (versus previous AHSMOG reports) enhances the power of the latest
38 analyses over past published ones. Of 1,575 deaths from all natural (non-external) causes,
39 1,029 were cardiopulmonary, 135 were non-malignant respiratory (ICD9 codes 460-529), and
40 30 were lung cancer (ICD9 code 162) deaths. Abbey et al. (1999) also created another death
41 category, contributing respiratory causes (CRC), which included any mention of nonmalignant
42 respiratory disease as an underlying or "contributing cause" on the death certificate. Numerous
June 2003 8-96 DRAFT-DO NOT QUOTE OR CITE
-------
1 analyses were done for the CRC category, due to the large numbers and relative specificity of
2 respiratory causes as a factor in the deaths. Education was used to index socio-economic status,
3 rather than income. Physical activity and occupational exposure to dust were also used as
4 covariates. Cox proportional hazard models adjusted for a variety of covariates or stratified by
5 sex were used. The "time" variable used in most of the models was survival time from date of
6 enrollment, except that age on study was used for lung cancer effects due to the expected lack of
7 short-term effects. Many covariate adjustments were evaluated, yielding results for all non-
8 external mortality as shown in Table 8-8.
9
10
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
Increment
30 days/yr
20 ug/m3
5 ug/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 UCL = Upper 95% confidence limit
Source: Abbey et al. (1999).
1 As for cause-specific mortality analyses of the AHSMOG data, positive and statistically
2 significant effects on deaths with underlying contributing respiratory causes were also found for
3 30 day/yr > 100 |ig/m3 PM10 (RR =1.14, 95% CI = 1.03-1.56) in models that included both sexes
4 and adjustment for age, pack-years of smoking, and BMI. Subsets of the cohort had elevated
5 risks: (a) former smokers had higher RR's than never-smokers (RR for PM10 exceedances for
6 never-smokers was marginally significant by itself); (b) subjects with low intake of anti-oxidant
7 vitamins A, C, E had significantly elevated risk of response to PM10, whereas those with
8 adequate intake did not (suggesting that dietary factors or, possibly, other socio-economic or life
June 2003 8-97 DRAFT-DO NOT QUOTE OR CITE
-------
1 style factors for which they are a surrogate may be important covariates); and (c) there also
2 appeared to be a gradient of PM10 risk with respect to time spent outdoors, with those who had
3 spent at least 16 h/wk outside being at greater risk from PM10 exceedances. The extent to which
4 time spent outdoors is a surrogate for other variables or is a modifying factor reflecting temporal
5 variation in exposure to ambient air pollution is not clear, e.g., if the males spent much more
6 time outdoors than the females, outdoor exposure time could be confounded with gender. When
7 the cardiopulmonary analyses are broken down by gender (Table 8-9), the RR's for female
8 deaths were generally smaller than that for males, but none of the risks for PM indices or
9 gaseous pollutants were statistically significant at p < 0.05.
10
11
TABLE 8-9. RELATIVE RISK OF MORTALITY FROM CARDIOPULMONARY
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
O3 mean
SO2 mean
Pollution
Increment
30 days/yr
20 ug/m3
5 ug/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 UCL = Upper 95% confidence limit
Source: Abbey et al. (1999).
1 The AHSMOG cancer analyses yielded very mixed results for lung cancer mortality
2 (Table 8-10). For example, RR's for lung cancer deaths were statistically significant for males
3 for PM10 and O3 metrics, but not for females. In contrast, such cancer deaths were significant for
4 mean NO2 only for females (but not for males), but lung cancer metrics for mean SO2 were
5 significant for both males and females. This pattern is not readily interpretable, but is reasonably
6 attributable to the very small numbers of cancer-related deaths (18 for females and 12 for males),
7 resulting in wide RR confidence intervals and very imprecise effects estimates.
June 2003 8-98 DRAFT-DO NOT QUOTE OR CITE
-------
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
Increment
30 days/yr
20 ug/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
smoker
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. UCL = Upper 95% confidence limit.
Source: Abbey et al. (1999).
1 The analyses reported by Abbey et al. (1999) attempted to separate PM10 effects from those
2 of other pollutants by use of two-pollutant models, but no quantitative findings from such
3 models were reported. Abbey et al. did mention that the PM10 coefficient for CRC remained
4 stable or increased when other pollutants were added to the model. Lung cancer mortality
5 models for males evaluated co-pollutant effects in detail and indicated that NO2 was
6 non-significant in all two-pollutant models but the other pollutant coefficients were stable. The
7 PM10 and O3 effects remained stable when SO2 was added, suggesting possible independent
8 effects, but PM10 and O3 effects were hard to separate because these pollutants were highly
9 correlated in this study. Again, however, the very small number of lung cancer observations and
10 likely great imprecision of reported effects estimates markedly limit the weight that should be
11 accorded to these results.
June 2003
8-99
DRAFT-DO NOT QUOTE OR CITE
-------
1 Other analyses, by Beeson et al. (1998), evaluated essentially the same data as in Abbey
2 et al. (1999), but focused on lung cancer incidence (1977-1992). There were only 20 female and
3 16 male lung cancer cases among the 6,338 subjects. Exposure metrics were constructed to be
4 specifically relevant to cancer, these being the annual average of monthly exposure indices from
5 January, 1973 through the following months but ending 3 years before date of diagnosis (i.e.,
6 representing a 3-year lag between exposure and diagnosis of lung cancer). The covariates in the
7 Cox proportional hazards model were pack-years of smoking and education, and the time
8 variable was attained age. Many additional covariates were evaluated for inclusion, but only
9 'current use of alcohol' met criteria for inclusion in the final model. Pollutants evaluated were
10 PM10, SO2, NO2, and O3. No interaction terms with the pollutants proved to be significant,
11 including outdoor exposure times. The RR estimates for male lung cancer cases were:
12 (a) positive and statistically significant for all PM10 indicators; (b) positive and mostly
13 significant for O3 indicators, except for mean O3, number of O3 exceedances > 60 ppb, and in
14 former smokers; (c) positive and significant for mean SO2, except when restricted to proximate
15 monitors; and (d) positive but not significant for mean NO2. When analyses are restricted to the
16 use of air quality data within 32 km of the residences of subjects, the RR over the IQR of
17 24 |ig/m3 in the full data set is 5.21 (or RR= 1.99 per 10 |ig/m3 PM10). The female RR's were all
18 much smaller than for males, their being significant for mean SO2 but not for any indicator of
19 PM10orO3.
20 The AHSMOG investigators also attempted to compare effects of fine versus coarse
21 particles (McDonnell et al, 2000). For AHSMOG participants living near an airport (n = 3,769),
22 daily PM2 5 concentrations were estimated from airport visibility using previously-described
23 methods (Abbey et al, 1995b). Given the smaller numbers of subjects in these subset analyses, it
24 is not necessarily surprising that no pollutants were found to be statistically significant in these
25 regressions, even based on analysis for the male subset near airports (n = 1266). It is important
26 to caveat that (a) the PM2 5 exposures were estimated from visibility measurements (increasing
27 exposure measurement error) and yielded a very uneven and clustered distribution of estimated
28 exposures and; (b) the PM10_25 values were calculated from the differencing of PM10 and PM25,
29 likely contributing to additional measurement error for the coarse particle (PM10_2 5) variable used
30 in the analyses.
31
June 2003 8-100 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.3.2.4 The EPRI- Washington University Veterans' Cohort Mortality Study
2 Lipfert et al. (2000b) reported preliminary results from large-scale mortality analyses for a
3 prospective cohort of up to 70,000 men assembled by the U.S. Veterans Administration (VA) in
4 the mid-1970s. While much smaller than the ACS cohort, this VA study group is similar in that
5 it was not originally formed to study air pollution, but was later linked to air pollution data
6 collected separately, much of it subsequent to the start of the study. The AHSMOG and Six City
7 studies were designed as prospective studies to evaluate long-term effects of air pollution and
8 had concurrent air pollution measurements. The ACS study was also a prospective study, using
9 air pollution data obtained at about the approximate time of enrollment but not subsequently
10 (Pope et al., 1995). The extended ACS data incorporated much more air pollution data,
11 including TSP data back to the 1960s and more recent fine particle data. The VA PM25 data set
12 was smaller than the TSP data set and similar to the ACS data.
13 The VA study cohort was male, middle-aged (51 ± 12 years) and included a larger
14 proportion of African-Americans (35%) than the U.S. population as a whole and a large
15 percentage of current or former smokers (81%). The cohort was selected at the time of
16 recruitment as being mildly to moderately hypertensive, with screening diastolic blood pressure
17 (DBF) in the range 90 to 114 mm Hg (mean 96, about 7 mm more than the U.S. population
18 average) and average systolic blood pressure (SBP) of 148 mm Hg. The subjects had all been
19 healthy enough to be in the U.S. armed forces at one time. A comparison of their pre-existing
20 health status at time of study recruitment versus the initial health status of the other cohorts
21 would be of interest. The study that led to the development of this clinical cohort (Veterans
22 Administration Cooperative Study Group on Antihypertensive Agents, 1970; 1967) was a
23 "landmark" VA cooperative study demonstrating that anti-hypertensive treatment markedly
24 decreased morbidity and mortality (Perry et al., 1982). The clinical cohort itself involved actual
25 clinical rather than research settings. Some differences between the VA cohort and other
26 prospective cohorts are noted below.
27 Pollutant levels of the county of residence at the time of entry into the study were used for
28 analyses versus levels at the VA hospital area. Contextual socioeconomic variables were also
29 assembled at the ZIP-code and county levels. The ZIP-code level variables were average
30 education, income, and racial mix. County-level variables included altitude, average annual
31 heating-degree days, percentage Hispanic, and socioeconomic indices. Census-tract variables
June 2003 8-101 DRAFT-DO NOT QUOTE OR CITE
-------
1 included poverty rate and racial mix. County-wide air pollution variables included TSP, PM10,
2 PM2 5, PM15, PM15_2 5, SO4, O3, CO, and NO2 levels at each of the 32 VA clinics where veterans
3 were enrolled. Besides considering average exposures over the entire period, three sequential
4 mortality follow-up periods (1976-81, 1982-88, 1989-96) were also evaluated in separate
5 statistical analyses that attempted to relate mortality in each of those periods to air pollution in
6 different preceding, concurrent, or subsequent periods (i.e., up to 1975, 1975-81, 1982-88, and
7 1989-86, for TSP in the first three periods, PM10 for the last, and NO2, 95th percentile O3, and
8 95th percentile CO for all four periods). Mortality in the above-noted periods was also evaluated
9 in relation to SO4 in each of the same four periods noted for NO2, O3, and CO, and to PM2 5,
10 PM15, and PM15.25 in 1979-81 and 1982-84.
11 The participants in the VA Cohort clearly formed an "at-risk" population, and the results
12 by Vasan et al. (2001) make more plausible the hypothesis stated in Lipfert et al. (2000b, p. 62)
13 that ". . . the relatively high fraction of mortality within this cohort may have depleted it of
14 susceptible individuals in the later periods of follow-up." The use of diastolic and systolic blood
15 pressure in the reported regression results may require further evaluation. The role of DBF and
16 SBP as predictors in regression models in the VA Cohort may be considered as closer to the
17 endpoint (mortality) than as a more distal behavioral, environmental, or contextual predictor of
18 mortality such as air pollution, temperature, smoking behavior, BMI, etc. Personal-level
19 variables tend to interact only with each other, as do county-level variables, with little
20 correlation across spatial scales.
21 The estimated mean risk of cigarette smoking in this cohort (RR = 1.43) is also smaller
22 than that of the Six City cohort (RR =1.59) and the ACS cohort (RR = 2.07 for current
23 smokers). Some possible differences include the higher proportion of former or current smokers
24 in this cohort (81%) versus 51% in the ACS study and 42 to 53% in the Six City study.
25 A possibly more important factor may be the difference in education levels, as only 12% of the
26 ACS participants had less than a high school education vs 28% of the Six City cohort. Education
27 level was not reported for the VA Cohort. Education differences may be associated with
28 smoking behavior, and the large number of interaction terms used in the VA study model may
29 also partially to account for differences in results obtained across the three ACS, Six-City, VA)
30 studies.
June 2003 8-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 The preliminary screening models used proportional hazards regression models (Miller
2 et al., 1994) to identify age, SBP, DBF, BMI (nonlinear), age and race interaction terms, and
3 present or former smoking as baseline predictors, with one or two pollution variables added.
4 In the final model using 233 terms (of which 162 were interactions of categorized SBP, DBF,
5 and BMI variables with age), the most significant non-pollution variables were SBP, DBF, BMI,
6 and their interactions with age, smoking status, average education, race, poverty, height, and a
7 clinic-specific effect. Lipfert et al. (2000b) noted that the risk of current cigarette smoking
8 (1-43) that they found was lower than reported in other studies. The most consistently positive
9 effects were found for O3 and NO2 exposures in the immediately preceding years. This study
10 used peak O3 rather than mean O3 as in some other cohort studies. This may account for the
11 higher O3 and NO2 effects here. While the PM analyses considering segmented (shorter) time
12 periods gave differing results (including significantly negative mortality coefficients for some
13 PM metrics), when methods consistent with the past studies were used (i.e., many- year average
14 PM concentrations), similar results were reported: the authors found that "(t)he single-mortality-
15 period responses without ecological variables are qualitatively similar to what has been reported
16 before (SO4 > PM2 5 > PM15)." With ecological variables included, the only significant PM
17 effect was that of TSP up to 1981 on 1976-81 mortality. It might be instructive to evaluate more
18 parsimonious regression models with fewer ecological covariates and interaction terms. It is
19 noteworthy that estimated PM effects appear to be smaller in the later years of the study rather
20 than in the earlier years. This may also be due to cohort depletion.
21 Overall, the authors concluded that "the implied mortality risks of long-term exposure to
22 air pollution were found to be sensitive to the details of the regression model, the time period of
23 exposure, the locations included, and the inclusion of ecological as well as personal variables."
24
25 8.2.3.2.5 Relationship ofAHSMOG, Six Cities, ACS and VA Study Findings
26 The results of the more recent AHSMOG mortality analyses (Abbey et al., 1999;
27 McDonnell et al., 2000) are compared here with findings from the earlier Six Cities study
28 (Dockery et al., 1993), the ACS study (Pope et al., 1995), the HEI reanalyses of the latter two
29 studies, the extension of the ACS study (Pope et al., 2002), and the VA study (Lipfert et al.,
30 2000b). Table 8-11 compares the estimated RR for total, cardiopulmonary, and cancer mortality
31 among the studies. The number of subjects in these studies varies greatly: 8,111 subjects in the
June 2003 8-103 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-11. 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
VA10
PM1
PM25
PM25
PM25
PM25
PM15.2.5
PM10/15
Dichot
PM10/15 SSI
PM25
1979-83
PM25
1999-000
PM2 5 Avg.
PM10/15
PM25
PM10.25
PM25
Total
Ex.
RR2
13%
14%
6.6%
7.0%
0.4%
4.1%
1.6%
4.1%
5.9%
6.2%
2.1%
8.5%
5.2%
- 10.0%
Mortality
95% CI
(4.2, 23%)
(5.4, 23%)
(3.5, 9.8%)
(3.9, 10%)
(-1.4,2.2%)
(0.9, 7.4%)
(-0.8,4.1%)
(0.8, 7.5%)
(2.0, 9.9%)
(1.6, 11%)
(-4.5, 9.2%)
(-2.3, 21%)
(-8.3, 21%)
(-15, -4.6%)
Cardiopulmonary
Mortality
Ex.
RR
18%
19%
12%
12%
0.4%
7.3%
5.7%
5.9%
7.9%
9.3%
0.6%
23%
20%
95% CI
(6.0, 32%)
(6.5, 33%)
(6.7, 17%)
(7.4, 17%)
(-2.2%, 3.1%)
(3.0, 12%)
(2.5, 9.0%)
(1.5, 10%
(2.3, 14%)
(3.3, 16%)
(-7.8, 10%)
(-3.0,55%)
(-13,64%)
Lung Cancer
Mortality
Ex.
RR
18%
21%
1.2%
0.8%
-1.2%
0.8%
-1.6%
8.2%
12.7%
13.5%
81%
39%
26%
95% CI
(-11,57%)
(-8.4,60%)
(-8,7, 12%)
(-8.7, 11%)
(-7.3%, 5.1%)
(-8.1, 11%)
(-9.1,6.4%)
(1.1, 16%)
(4.1,22%)
(4.4, 23%)
(14, 186%)
(-21, 150%)
(-38, 155%)
'Increments are 10 ug/m3 forPM25 and 20 ug/m3 forPM10/15.
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 (10 or 20) by the equation.
RR = exp(log(RR for range) x /range).
3From (Dockery et al., 1993; Krewski et al., 2000, Part II, Table 21a), original model.
4From (Krewski et al., 2000), Part I, Table 21c.
5From (Krewski et al., 2000), Part I, Table 25a.
6From (Krewski et al., 2000), Part I, Table 25c.
7From (Pope et al., 2002).
8From (Abbey et al., 1999), pooled estimate for males and females.
9From (McDonnell et al., 2000), using two-pollutant (fine and coarse particle) models; males only.
10Males only, exposure period 1979-81, mortality 1982-88 from Table 7 (Lipfert et al., 2000b).
June 2003
8-104
DRAFT-DO NOT QUOTE OR CITE
-------
1 Six-Cities Study; 295,223 subjects in the 50 fine particle (PM25) cities and 552,138 subjects in
2 the 151 sulfate cities of the ACS Study; 6,338 in the AHSMOG Study; and 70,000 in the VA
3 study. This may partially account for differences among their results.
4 The Six Cities study found significant associations of PM25 with total and cardiopulmonary
5 (but not lung cancer) mortality, but not with coarse particle indicators. In the Krewski et al.
6 (2000) reanalysis of the ACS study data, significant associations were found for both PM25 and
7 PM15 (excess relative risks of 6.6% for 10 |ig/m3 PM2 5 and 4% for 20 |ig/m3 increments in
8 annual PM10/15, respectively). The results most recently reported for the AHSMOG study (Abbey
9 et al., 1999; McDonnell et al., 2000) used PM10 as its PM mass index and found some significant
10 associations with total mortality and deaths with contributing respiratory causes, even after
11 controlling for potentially confounding factors (including other pollutants). However no pattern
12 of consistent, statistically significant associations between mortality and long-term PM exposure
13 was found. The VA study (Lipfert et al., 2000b), also did not find any association with PM25.
14 The lack of consistent findings in the AHSMOG study and negative results of the VA study, do
15 not negate the findings of the Six Cities and ACS studies: the ACS studies had a substantially
16 larger study population, and both the Six Cities and ACS studies were based on measured PM
17 data (in contrast with AHSMOG PM estimates based on TSP or visibility measurements) and
18 have been validated through exhaustive reanalyses. The results of these studies, including the
19 reanalyses results for the Six Cities and ACS studies and the results of the ACS study extension,
20 provide substantial evidence for positive associations between long-term ambient PM (especially
21 fine PM) exposure and mortality.
22 There is no clear consistency in relationships among PM effect sizes, gender, and smoking
23 status across these studies. The AHSMOG study cohort is a primarily nonsmoker group while
24 the VA study cohort had a large proportion of smokers and former smokers in an all-male
25 population. The ACS results, show similar and significant associations with total mortality for
26 both "never smokers" and "ever smokers", although the ACS cohort may include a substantial
27 number of long-term former smokers with much lower risk than current smokers. The Six Cities
28 study cohort shows the strongest evidence of a higher PM effect in current smokers than in non-
29 smokers, with female former smokers having a higher risk than male former smokers. This
30 study suggests that smoking status may be viewed as an effect modifier for ambient PM, just as
31 smoking may be a health effect modifier for ambient O3 (Cassino et al., 1999).
June 2003 8-105 DRAFT-DO NOT QUOTE OR CITE
-------
1 When the ACS study results are compared with the AHSMOG study results for SO4"2
2 (PM10_2 5 and PM10 were not considered in the ACS study, but were evaluated in ACS reanalyses
3 [Krewski et al., 2000; Pope et al, 2002]), the total mortality effect sizes per 15 |ig/m3 SO4"2 for
4 the males in the AHSMOG population fell between the Six-Cities and the ACS effect-size
5 estimates for males (RR = 1.28 for AHSMOG male participants; RR=1.61 for Six-Cities Study
6 male non-smokers; and RR = 1.10 for never smoker males in the ACS study), and the AHSMOG
7 study 95% confidence intervals encompass both of those other studies' sulfate RR's.
8
9 8.2.3.2.6 The S-Plus GAM Convergence Problem and Cohort Studies
10 The long-term pollution-mortality effect study results discussed above in this section were
11 unaffected by the GAM default convergence issue reported by Dominici et al. (2002) and
12 discussed earlier in this chapter, because they did not use such a model specification. Instead,
13 the cohort studies of long-term PM exposures used Cox Proportional Hazards models. For
14 example, in the recent Pope et al. study (2002), the baseline models were random effects Cox
15 Proportional Hazards models without the inclusion of nonparametric smooths. However, Pope
16 et al. (2002) did include a non-parametric spatial smooth in the model as part of a more extended
17 sensitivity analysis to evaluate more aggressive control of spatial differences in mortality. They
18 found that the estimated pollution-mortality effects were not sensitive to this additional spatial
19 control, so the final reported results did not include the smooth; and this study's results, like
20 those from the other cohort studies discussed above, were not affected by the S-Plus
21 convergence issue.
22
23 8.2.3.3 Studies by Particulate Matter Size-Fraction and Composition
24 8.2.3.3.1 Six Cities, ACS, and AHSMOG Study Results
25 Ambient PM consists of mixtures that may vary in composition over time and from place
26 to place. This should logically affect the relative toxicity of PM indexed by mass at different
27 times or locations. Some semi-individual chronic exposure studies have investigated relative
28 roles of various PM components in contributing to observed air pollution associations with
29 mortality. However, only a limited number of the chronic exposure studies have included direct
30 measurements of chemical-specific constituents of the PM mixes indexed by mass measurements
31 used in their analyses.
June 2003 8-106 DRAFT-DO NOT QUOTE OR CITE
-------
1 As shown in Table 8-12, the Harvard Six-Cities Study (Dockery et al., 1993) results
2 indicated that the PM25 and SO4"2 RR associations (as indicated by their respective 95% CFs and
3 t-statistics) were more consistent than those for the coarser mass components. Further, the
4 effects of sulfate and non-sulfate PM2 5 are quite similar. Acid aerosol (H+) exposure was also
5 considered by Dockery et al. (1993), but only less than one year of measurements collected near
6 the end of the follow-up period were available in most cities; consequently, the Six-Cities results
7 were much less conclusive for the acidic component of PM than for the other PM metrics
8 measured over many years during the study.
9
10
TABLE 8-12. COMPARISON OF ESTIMATED RELATIVE RISKS FOR
ALL-CAUSE MORTALITY IN SFX U.S. CITIES ASSOCIATED WITH
THE REPORTED INTER-CITY RANGE OF CONCENTRATIONS OF
VARIOUS PARTICULATE MATTER METRICS
PM Species
S04=
PM2 5 - SO4=
PM25
PM15.2.5
TSP-PM15
Concentration Range
Oig/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).
1 Table 8-13 presents comparative PM25 and SO4"2 results from the ACS study, indicating
2 that both had substantial, statistically significant effects on all-cause and cardiopulmonary
3 mortality. On the other hand, the RR for lung cancer was notably larger (and substantially more
4 significant) for SO4"2 than PM2 5 (not significant). The most recent AHSMOG analyses also
5 considered SO4"2 as a PM index for all health outcomes studied except lung cancer, but SO4"2 was
6 not as strongly associated as PM10 with mortality and was not statistically significant for any
7 mortality category.
8 Also, very extensive results were reported in Lipfert et al. (2000b) for various components:
9 TSP, PM10, PM2 5, PM15_2 5, PM15, SO4"2. There were no significant positive effects for any
June 2003 8-107 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-13. COMPARISON OF REPORTED SO4= AND PM25 RELATIVE RISKS
FOR VARIOUS MORTALITY CAUSES IN THE AMERICAN CANCER
SOCIETY (ACS) STUDY
Mortality Cause
All Cause
Cardiopulmonary
Lung Cancer
SO4=
(Range = 19.9 fig/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 fig/m3)
Relative
Risk
1.17
1.31
1.03
RR
95% CI
(1.09-1.26)
(1.17-1.46)
(0.80-1.33)
RR
t-Statistic
4.24
4.79
0.38
Source: Pope et al. (1995).
1 exposure period concurrent or preceding the mortality period for any PM component, but there
2 was for O3.
3 Results from the Harvard Six Cities, the ACS, and the AHSMOG studies are compared in
4 Table 8-14 (for total mortality) and Table 8-15 (for cause-specific mortality). Results for the VA
5 study are not shown in Tables 8-14 and 8-15 for two reasons. First, the VA cohort is all male
6 and largely consists of current or former smokers (81%) and is thusly not comparable to the total
7 or male non-smoker populations of the other studies. Secondly, the VA study analyzed a wide
8 variety of exposure periods and mortality periods, making it difficult to summarize or compare
9 the VA results.
10 Estimates for Six Cities parameters were calculated in two ways: (1) mortality RR for the
11 most versus least polluted city in Table 3 of Dockery et al. (1993), adjusted to standard
12 increments; and (2) ecological regression fits in Table 12-18 of U.S. Environmental Protection
13 Agency (1996a). The Six Cities study of eastern and mid-western U.S. cities suggests a strong
14 and highly significant relationship for fine particles and sulfates, a slightly weaker but still
15 highly significant relationship to PM10, and a marginal relationship to PM10_25. The ACS study
16 looked at a broader spatial representation of cities, and found a stronger statistically significant
17 relationship to PM2 5 than to sulfate (no other pollutants were examined). The AHSMOG study
18 at California sites (where sulfate levels are typically low) found significant effects in males for
19 PM10 100 |ig/m3 exceedances and a marginal effect of mean PM10, but no PM effects for females
20 or with sulfates. On balance, the overall results shown in Tables 8-14 and 8-15 suggest
June 2003 8-108 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-14. COMPARISON OF TOTAL MORTALITY RELATIVE RISK
ESTIMATES AND T-STATISTICS FOR PARTICULATE MATTER COMPONENTS
IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM10 (50 ug/m3)
PM2 5 (25 ug/m3)
S04 = (15 ug/m3)
Days/yr. with
PM10 > 100 ug/m3
(30 days)
PM10.2.5 (25 ug/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
1.50a; 1.53b
1.28a
1.24
1.36a; 1.38b
1.21a
1.17
1.25
1.50a; 1.57b
1.35
1.11
1.10
1.28
1.08
1.81a; 1.56b
1.43a
t Statistic
2.94a; 3.27b
0.81a
1.61
2.94a; 3.73b
0.81a
4.35
1.96
2.94a; 3.67b
0.81a
5.11
1.59
0.96
2.18
2.94a'c 1.81b
0.81a
"Method 1 compares Portage versus 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 statistically significant relationships between long-term exposures to PM10, PM2 5, and/or sulfates
2 and excess total and cause-specific cardiopulmonary mortality.
3 The semi-individual long-term PM exposure studies conducted to date collectively appear
4 to 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 PM25 population exposure
7 measurement (both analytical and spatial) relative to PM10_2 5 makes conclusions regarding their
June 2003
8-109
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-15. COMPARISON OF CARDIOPULMONARY MORTALITY RELATIVE
RISK ESTIMATES AND T-STATISTICS FOR PARTICULATE MATTER
COMPONENTS IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM10 (50 ug/m3)
PM2 5 (25 ug/m3)
S04= (15 ug/m3)
Days/yr. with
PM10 > 100 (30 days)
PM10.2.5 (25 ug/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 versus 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
6
7
relative contributions to observed PM10-related associations less certain than if the effect of their
relative errors of measurement could be addressed.
8.2.3.3.2 Lipfert and Morris (2002): An Ecological Study
Although reasons were identified for preferring to use prospective cohort studies to assess
the long-term exposure effects of particles and gases, additional useful information may still be
derived from ecological studies, particularly by repeated cross-sectional studies that may provide
another tool for examining changes in air-pollution-attributable mortality over time. Lipfert and
June 2003
8-110
DRAFT-DO NOT QUOTE OR CITE
-------
1 Morris (2002) carried out cross-sectional regressions for five time periods using published data
2 on mortality, air pollution, climate, and socio-demographic factors using county- level data.
3 Data were available for TSP and gaseous co-pollutants as far back as 1960 and for PM2 5, PM15,
4 and SO4= from the inhalable particular network (IPN). Attributable mortality at ages 45+ for
5 1979-1981 was reported to be associated with 1960-64 TSP, less strongly with 1970-1974 TSP,
6 but not with concurrent (1979-1981) TSP. Attributable mortality for ages 45+in 1979-1981 was
7 associated with PM2 5 and SO4"2 but not with PM15 for 1979-1984. However, SO4"2 for most
8 intervals from 1960-64 up to 1979-1981 was associated with mortality for most ages.
9 Concurrent SO2 (1979-1981) was associated with mortality, but much less for earlier years.
10 Pollution-attributable mortality in 1989-91 was no longer significantly associated with
11 TSP, but remained significantly associated with PM2 5 and SO4"2 for ages 45+ for most time
12 intervals: 1979-84 and 1999 for PM25; 1970-74, 1979-81, 1979-84 for fine); and 1982-88 for
13 SO4"2. Pollution--attributable mortality in 1995-1997 had little association with present or
14 previous PM2 5 and PM10, but a reasonably consistent and positive relationship to SO4"2. There
15 appeared to be a systematic decrease in the TSP, IPN, PM25, and PM10 effects from the 1960s to
16 the 1990s and in the AIRS and IPN SO4"2 effect over time, but an increase in the AIRS PM2 5
17 effect and in the NO2 and peak O3 effects.
18 One of the journal editors (Ayres, 2002) notes that this study uses some other ecological
19 variables that may improve the model. Two of the ecological variables, vehicle miles of travel
20 per square mile per year by gasoline (VMTG) and diesel (VMTD) vehicles, respectively, in a
21 county (also used in Janssen et al., 2002) are likely to have important associations with air
22 pollution. As noted earlier, some ambient pollutants associated with fuel combustion have
23 higher concentrations near main roads, such as PM10_2 5 (EC if from diesel exhaust) , NO2, and
24 CO; whereas other pollutants (such as O3) may have higher concentrations away from major
25 highways. Similarly, some models employed included the percentage of air conditioning in a
26 county, a factor that may well be correlated with greater secondary aerosol formation in warmer
27 temperatures and is likely associated with diminished exposure to air pollution, resulting in
28 smaller acute health effects per |ig/m3 of PM pollution (Janssen et al, 2002). Given these
29 potentially confounding terms in this study's model, it is not surprising that the authors find
30 somewhat lower percentage increases in mortality per |ig/m3 of PM than in the above-discussed
31 cohort studies.
June 2003 8-111 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.3.4 Population-Based Mortality Studies in Children
2 Some older cross-sectional mortality studies reviewed in the 1996 PM AQCD suggested
3 that the young may represent a susceptible sub-population for PM-related mortality.
4 For example, Lave and Seskin (1977) found mortality among those 0-14 years of age to be
5 significantly associated with TSP. More recently, Bobak and Leon (1992) studied neonatal (ages
6 < 1 mo) and post-neonatal mortality (ages 1-12 mo) in the Czech Republic and reported
7 significant and robust associations between post-neonatal mortality and PM10, even after
8 considering other pollutants. Post-neonatal respiratory mortality showed highly significant
9 associations for all pollutants considered, but only PM10 remained significant in simultaneous
10 regressions. The exposure duration was longer than a few days, but shorter than in the adult
11 prospective cohort studies. Thus, the limited available studies reviewed in the 1996 PM AQCD
12 were highly suggestive of an association between ambient PM concentrations and infant
13 mortality, especially among post-neonatal infants.
14 More recent studies since the 1996 PM AQCD have focused specifically on ambient PM
15 relationships to (a) intrauterine mortality and morbidity and (b) early post neonatal mortality.
16 In a study by Pereira et al. (1998) of intrauterine (pre-natal) mortality during one year
17 (1991-1992) in Brazil, PM10 was not found to be a significant predictor, but involvement of CO
18 was suggested by an association between increased carboxyhemoglobin (CoHb) in fetal blood
19 and ambient CO levels on the day of delivery measured in a separate study. Another study
20 (Dejmek et al., 1999) evaluated possible impacts of ambient PM10 and PM25 exposure
21 (monitored by EPA-developed VAPS methods) during pregnancy on intrauterine growth
22 retardation (IUGR) risk in the highly polluted Teplice District of Northern Bohemia in the Czech
23 Republic during three years (1993-1996). Mean levels of pollutants (PM, NO2, SO2) were
24 calculated for each month of gestation and three concentration intervals (low, medium, high)
25 were derived for each pollutant. Preliminary analyses found significant associations of IUGR
26 with SO2 and PM10 early in pregnancy but not with NO2. Odds ratios for IUGR for PM10 and
27 PM2 5 levels were determined by logistic regressions for each month during gestation, after
28 adjusting for potential confounding factors (e.g., smoking, alcohol consumption during
29 pregnancy, etc.). Definition of an IUGR birth was any one for which the birth weight fell below
30 the 10th percentile by gender and age for live births in the Czech Republic (1992-93). The ORs
31 for IUGR were significantly related to PM10 during the first month of gestation: that is, as
June 2003 8-112 DRAFT-DO NOT QUOTE OR CITE
-------
1 compared to low PM10, the medium level PM10 OR = 1.47 (CI 0.99-2.16), and the high level
2 PM10 OR = 1.85 (CI 1.29-2.66). PM2 5 levels were highly correlated with PM10 (r = 0.98) and
3 manifested similar patterns (OR = 1.16, CI 0.08-0.69 for medium PM2 5 level; OR = 1.68, CI
4 1.18-2.40 for high PM2 5 level). These results suggest effects of PM exposures (probably
5 including fine particles such as sulfates, acid aerosols, and PAHs in the Teplice ambient mix)
6 early in pregnancy (circa embryo implantation) on fetal growth and development.
7 More consistent results indicating likely early post-natal PM exposure effects on neonatal
8 infant mortality have emerged from other new studies. Woodruff et al. (1997), for example,
9 used cross-sectional methods to evaluate possible association of post-neonatal mortality with
10 ambient PM10 pollution. This study involved an analysis of a cohort of circa 4 million infants
11 born during 1989-1991 in 86 U.S. metropolitan statistical areas (MSAs). Data from the National
12 Center for Health Statistics-linked birth/infant death records were combined at the MSA level
13 with PM10 data from EPA's Aerometric database. Infants were categorized as having high,
14 medium, or low exposures based on tertiles of PM10 averaged over the first 2 postnatal months.
15 Relationships between this early neonatal PM10 exposure and total and cause-specific post-
16 neonatal mortality rates (from 1 mo to 1 y of age) were examined using logistic regression
17 analyses, adjusting for demographic and environmental factors. Overall post-neonatal mortality
18 rates per 1,000 live births were 3.1 among infants in areas with low PM10 exposures, 3.5 among
19 infants with medium PM10 exposures, and 3.7 among highly PM exposed infants. After
20 adjustment for covariates, the OR and 95% confidence intervals for total post-neonatal mortality
21 for the high versus the low exposure group was 1.10 (CI = 1.04-1.16). For normal birth weight
22 infants, high PM10 exposure was associated with mortality for respiratory causes (OR = 1.40,
23 CI = 1.05-1.85) and sudden infant death syndrome (OR= 1.26, CI = 1.14-1.39). Among low
24 birth weight babies, high PM10 exposure was positively (but not significantly) associated with
25 mortality from respiratory causes (OR = 1.18, CI=0.86-1.61). However, other pollutants (e.g.,
26 CO) were not considered as possible confounders. This study provides results consistent with
27 some earlier reports indicating that outdoor PM air pollution may be associated with increased
28 risk of post-neonatal mortality (e.g., Bobak and Leon, 1992), but lack of consideration of other
29 air pollutants as potential confounders in this new study reduces the certainty that PM is the
30 specific causal outdoor air pollutant in this case.
June 2003 8-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 Lipfert et al. (2000c) have reported replicating the basic findings of Woodruff et al. (1997)
2 using a similar modeling approach but annual average PM10 air quality data for one year (1990)
3 instead of PM10 averaged over the first two post natal months during 1989-1991. The
4 quantitative relationship between the individual risk of infant mortality did not differ among
5 infant categories (by age, by birthweight, or by cause), but PM10 risks for SIDs deaths were
6 higher for babies of smoking mothers. SO4"2 was a strong negative predictor of SIDs mortality
7 for all age and birth weight categories. The authors (a) noted difficulties in ascribing the
8 reported PM10 and SO4"2 associations to effects of the PM pollutants per se versus the results
9 possibly reflecting interrelationships between the air pollution indices, a strong well-established
10 East-West gradient in U.S. SIDS cases, and/or underlying sociodemographic factors (e.g., the
11 socioeconomic or education level of parents) and (b) hypothesized that a parallel gradient in use
12 of wood burning in fireplaces or woodstoves and consequent indoor wood smoke exposure
13 might explain the observed cross-sectional study results.
14 The basic findings from Woodruff et al. (1997) also appear to be bolstered by a more
15 recent follow-up study by Bobak and Leon (1999), who conducted a matched population-based
16 case-control study covering all births registered in the Czech Republic from 1989 to 1991 that
17 were linked to death records. They used conditional logistic regression to estimate the effects of
18 suspended particles and nitrogen oxides on risk of death in the neonatal and early post-neonatal
19 period, controlling for maternal socioeconomic status and birth weight, birth length, and
20 gestational age. The effects of all pollutants were strongest in the post-neonatal period and
21 specific for respiratory causes. Only PM showed a consistent association when all pollutants
22 were entered in one model. Thus, in this study, it appears that long-term exposure to PM is the
23 air pollutant metric most strongly associated with excess post-neonatal deaths.
24 Chay and Greenstone (2001a,b) also conducted a study of changes in annual air pollution
25 and infant mortality over time (rather than spatially) in the U.S. for the period 1981-1982. These
26 studies used sharp, differential air quality changes across sites attributable to geographic
27 variation in the effects of the 1981-1982 recession to estimate the relationship between PM air
28 pollution and infant mortality. During the narrow period of these two years, there was
29 substantial variation across counties in changes in particulate (TSP) pollution and these
30 differential pollution reductions appeared to be independent of changes in numerous
31 socioeconomic and health care factors that may be related to infant mortality. The authors found
June 2003 8-114 DRAFT-DO NOT QUOTE OR CITE
-------
1 that a 1 ug/m3 reduction in TSP resulted in about 4-8 fewer infant deaths per 100,000 live births
2 at the county level (a 0.35-0.45 elasticity), the estimates being remarkably stable across a variety
3 of specifications. The estimated effects in this study were driven almost entirely by fewer deaths
4 occurring within one month and one day of birth (i.e., neonatal), suggesting that fetal exposure to
5 pollution (via the mother) may have adverse health consequences. Findings of the population
6 reductions in infant birth weight in this study provide evidence consistent with the infant
7 mortality effects found, suggestive of a causal relationship between PM exposure and infant
8 mortality.
9 The study by Loomis et al. (1999) of infant mortality in Mexico City during 1993-1995
10 adds additional interesting information pointing towards likely fine particle effects on infant
11 mortality. That is, in Mexico City (where mean 24-h PM2 5 = 27.4 |ig/m3), infant mortality was
12 found to be associated with PM25, NO2, and O3 in single pollutant GAM Poisson models, but
13 much less consistently with NO2 and O3 than PM2 5 in multipollutant models. The estimated
14 excess risk for PM25-related infant mortality lagged 3-5 days was 18.2% (CI = 6.4-30.7) per
15 25 |ig/m3 PM2 5. The extent to which such a notable increased risk for infant mortality might be
16 extrapolated to U.S. situations is not clear, however, due to possible differences in prenatal
17 maternal or early postnatal infant nutritional status.
18
19 8.2.3.5 Salient Points Derived from Analyses of Chronic Particulate Matter Exposure
20 Mortality Effects
21 A review of the studies summarized in the previous PM AQCD (U.S. Environmental
22 Protection Agency, 1996a) indicates that past epidemiologic studies of chronic PM exposures
23 collectively indicate increases in mortality to be associated with long-term exposure to airborne
24 particles of ambient origins. The PM effect size estimates for total mortality from these studies
25 also indicate that a substantial portion of these deaths reflected cumulative PM effects above and
26 beyond those exerted by acute exposure events.
27 The recent HEI-sponsored reanalyses of the ACS and Harvard Six-Cities studies (Krewski
28 et al., 2000) "replicated the original results, and tested those results against alternative risk
29 models and analytic approaches without substantively altering the original findings of an
30 association between indicators of particulate matter air pollution and mortality." Several
31 questions, including the questions (1-4) posed at the outset of this Section (8.2.3) were
32 investigated by the Krewski et al. (2000) sensitivity analyses for the Six City and ACS studies
June 2003 8-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 data sets. Key results emerging from the HEI reanalyses and other new chronic PM mortality
2 studies are as follow:
3 (1) A much larger number of confounding variables and effects modifiers were considered
4 in the Reanalysis Study than in the original Six City and ACS studies. The only significant air
5 pollutant other than PM2 5 and SO4 in the ACS study was SO2, which greatly decreased the PM2 5
6 and sulfate effects when included as a co-pollutant (Krewski et al., 2000, Part II, Tables 34-38).
7 A similar reduction in particle effects occurred in any multi-pollutant model with SO2. The most
8 important new effects modifier was education. The AHSMOG study suggested that other
9 metrics for air pollution, and other personal covariates such as time spent outdoors and
10 consumption of anti-oxidant vitamins, might be useful. Both individual-level covariates and
11 ecological-level covariates shown in (Krewski et al., 2000, Part II, Table 33) were evaluated.
12 (2) Specific attribution of excess long-term mortality to any specific particle component or
13 gaseous pollutant was refined in the reanalysis of the ACS study. Both PM25 and sulfate were
14 significantly associated with excess total mortality and cardiopulmonary mortality and to about
15 the same extent whether the air pollution data were mean or median long-term concentrations or
16 whether based on original investigator or Reanalysis Team data. The association of mortality
17 with PM15 was much smaller, though still significant; and the associations with the coarse
18 fraction (PM15.25) or TSP were even smaller and not significant. The lung cancer effect was
19 significant only for sulfate with the original investigator data or for new investigators with
20 regional sulfate artifact adjustment for the 1980-1981 data (Krewski et al., 2000, Part II,
21 Table 31). Associations of mortality with long-term mean concentrations of criteria gaseous
22 co-pollutants were generally non-significant except for SO2 (Krewski et al., 2000, Part II, Tables
23 32, 34-38), which was highly significant, and for cardiopulmonary disease with warm-season
24 ozone. However, the regional association of SO2 with SO4 and SO2 with PM2 5 was very high;
25 and the effects of the separate pollutants could not be distinguished. Krewski et al. (2000,
26 p. 234) concluded that, "Collectively, our reanalyses suggest that mortality may be associated
27 with more than one component of the complex mix of ambient air pollutants in urban areas of
28 the United States." In the most recent extension of the ACS study, Pope et al. (2002) confirmed
29 the strong association with SO2 but found little evidence of effects for long-term exposures to
30 other gaseous pollutants.
June 2003 8-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 (3) The extensive temporal data on air pollution concentrations over time in the Six City
2 Study allowed the Reanalysis Team to evaluate time scales for mortality for long-term exposure
3 to a much greater extent than reported in Dockery et al. (1993). The first approach was to
4 estimate the log-hazard ratio as a function of follow up time using a flexible spline-function
5 model (Krewski et al., 2000, Part II, Figures 2 and 3). The results for both SO/2 and PM2 5
6 suggest very similar relationships, with larger risk after initial exposure decreasing to 0 after
7 about 4 or 5 years, and a large increase in risk at about 10 years follow-up time.
8 The analyses of the ACS Study proceeded somewhat differently, with less temporal data
9 but many more cities. Flexible spline regression models for PM2 5 and sulfate as function of
10 estimated cumulative exposure (not defined) were very nonlinear and showed quite different
11 relationships (Krewski et al., 2000, Part II, Figures 10 and 11). The PM25 relationship shows the
12 mortality log-hazard ratio increasing up to about 15 |ig/m3 and relatively flat above about
13 22 |ig/m3, then increasing again. The sulfate relationship is almost piecewise linear, with a low
14 near- zero slope below about 11 |ig/m3 and a steep increase above that concentration.
15 A third approach evaluated several time-dependent PM25 exposure indicators in the
16 Six City Study: (a) constant (at the mean) over the entire follow-up period; (b) annual mean
17 within each of the 13 years of the study; (c) city-specific mean concentration for the earliest
18 years of the study (i.e., very long-term effect); (d) exposure estimate in 2 years preceding death;
19 (e) exposure estimate in 3 to 5 years preceding death; and (f) exposure estimate > 5 years
20 preceding death. The time-dependent estimates (a-e) for mortality risk are generally similar and
21 statistically significant (Krewski et al., 2000, Part II, Table 53), with RR of 1.14 to 1.19 per
22 24.5 |ig/m3 being much lower than the risk of 1.31 estimated for exposure at the constant mean
23 for the period. Thus, it is highly likely the duration and time patterns of long-term exposure
24 affect the risk of mortality; and further study of this question (along with that of mortality
25 displacement from short-term exposures) would improve estimates of life-years lost from PM
26 exposure.
27 (4) The Reanalysis Study also advanced our understanding of the shape of the relationship
28 between mortality and PM. Again using flexible spline modeling, Krewski et al. (2000, Part II,
29 Figure 6) found a visually near-linear relationship between all-cause and cardiopulmonary
30 mortality residuals and mean sulfate concentrations, near-linear between cardiopulmonary
31 mortality and mean PM2 5, but a somewhat nonlinear relationship between all-cause mortality
June 2003 8-117 DRAFT-DO NOT QUOTE OR CITE
-------
1 residuals and mean PM2 5 concentrations that flattens above about 20 |ig/m3. The confidence
2 bands around the fitted curves are very wide, however, neither requiring a linear relationship nor
3 precluding a nonlinear relationship if suggested by reanalyses. An investigation of the mortality
4 relationship for other indicators may be useful in identifying a threshold, if one exists, for
5 chronic PM exposures.
6 (5) With regard to the role of various PM constituents in the PM-mortality association,
7 past cross-sectional studies have generally found the fine particle component, as indicated either
8 by PM2 5 or sulfates, to be the PM constituent most consistently associated with mortality. While
9 relative measurement errors of various PM indicators must be further evaluated as a possible
10 source of bias in these estimate comparisons, the Six-Cities and AHSMOG prospective
11 semi-individual studies both indicate that the fine mass components of PM are more strongly
12 associated with mortality effects of chronic PM exposure than are coarse fraction indicators.
13
14
15 8.3 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE
16 This effects of ambient PM on morbidity endpoints are assessed below in several
17 subsections: (a) cardiovascular morbidity effects of acute ambient PM exposure; (b) effects of
18 short-term PM exposure on the incidence of respiratory and other medical visits and hospital
19 admissions; and (c) short- and long-term PM exposure effects on lung function and respiratory
20 symptoms in asthmatics and non-asthmatics.
21
22 8.3.1 Cardiovascular Effects Associated with Acute Ambient Particulate
23 Matter Exposure
24 8.3.1.1 Introduction
25 Very little information specifically addressing cardiovascular morbidity effects of acute
26 PM exposure existed at the time of the 1996 PM AQCD. Since that time, a significantly
27 expanded body of literature has emerged, both on the ecologic relationship between ambient
28 particles and cardiovascular hospital admissions and associations of PM exposures with changes
29 in various physiological and/or biochemical measures. The latter studies are particularly
30 important in that they are suggestive of possible mechanisms underlying PM cardiovascular
31 effects.
June 2003 8-118 DRAFT-DO NOT QUOTE OR CITE
-------
1 This section begins with a brief summary of key findings from the 1996 PM AQCD
2 regarding acute cardiovascular effects of PM. Next, key new studies are reviewed in the two
3 categories noted above, i.e., ecologic time-series studies and individual-level studies of
4 physiological measures of cardiac function and/or biochemical measures in blood as they relate
5 to ambient pollution. This is followed by discussion of several issues of importance for
6 interpreting the available data, including identification of potentially susceptible sub-
7 populations, roles of environmental co-factors such as weather and other air pollutants, temporal
8 lags in the relationship between exposure and outcome, and the relative importance of various
9 size-classified PM components (e.g., PM2 5, PM10, PM10_2 5).
10
11 8.3.1.2 Summary of Key Findings on Cardiovascular Morbidity from the 1996
12 Particulate Matter Air quality Criteria Document
13 Just two studies were available for review in the 1996 PM AQCD that provided results for
14 acute cardiovascular (CVD) morbidity outcomes (Schwartz and Morris, 1995; Burnett et al.,
15 1995). Both studies were of ecologic time-series design and used standard statistical methods.
16 Analyzing four years of data on the > 65 year old Medicare population in Detroit, MI, Schwartz
17 and Morris (1995) reported significant associations between ischemic heart disease admissions
18 and PM10, controlling for environmental covariates. Based on an analysis of admissions data
19 from 168 hospitals throughout Ontario, Canada, Burnett et al. (1995) reported significant
20 associations between fine particle sulfate concentrations, as well as other air pollutants, and daily
21 cardiovascular admissions. The relative risk due to sulfate particles was slightly larger for
22 respiratory than for cardiovascular hospital admissions. The 1996 PM AQCD concluded on the
23 basis of these studies that: "There is a suggestion of a relationship to heart disease, but the
24 results are based on only two studies, and the estimated effects are smaller than those for other
25 endpoints" (U.S. Environmental Protection Agency, 1996a, p. 12-100). The PM AQCD also
26 stated that acute effects on CVD admissions had been demonstrated for elderly populations (i.e.,
27 > 65), but that insufficient data existed to assess relative effects on younger populations.
28 When viewed alongside the more extensive literature on acute CVD mortality that was
29 available at the time, the evidence from ecologic time-series studies reviewed in the 1996 PM
30 AQCD was consistent with acute health risks of PM being larger for cardiovascular and
31 respiratory causes than for other causes. Given the tendency for end-stage disease states to
32 include both respiratory and cardiovascular impairment, and the associated diagnostic overlap
June 2003 8-119 DRAFT-DO NOT QUOTE OR CITE
-------
1 that often exists, it was not possible on the basis of these studies alone to determine which of the
2 two organ systems, if either, was more critically effected.
3
4 8.3.1.3 New Particulate Matter-Cardiovascular Morbidity Studies
5 8.3.1.3.1 Acute Hospital Admission Studies
6 Salient methodological features and results of newly available studies that examine
7 associations between daily measures of ambient PM and daily hospital admissions for
8 cardiovascular disease are summarized in Table 8B-1 (see Appendix 8B). As discussed earlier
9 in Sections 8.1.4 and 8.2.2, many studies since 1996 used GAM with default convergence
10 criteria. Several of those studies have been reanalyzed by original investigators using GAM with
11 more stringent convergence criteria and GLM with parametric smooths, such as natural splines
12 (NS) or penalized splines (PN). Again, since the extent of possible bias in PM effect-size
13 estimates caused by the default criteria setting in the GAM models is difficult to estimate for
14 individual studies, the discussion here focuses mainly on the studies that either did not use GAM
15 Poisson models or those GAM studies which have been reanalyzed using more stringent
16 convergence criteria and/or alternative approaches. Newly available U.S. and Canadian studies
17 on relationships between short-term PM exposure and hospital admissions or emergency visits
18 that meet these criteria are summarized in Table 8-16, along with a few non-North American
19 studies. Reanalyses studies are indicated in Table 8-16 by indentation of the reference citation to
20 the pertinent short communication in the HEI Special Report (HEI, 2003). The table is
21 organized by first summarizing single-pollutant (PM only) analyses and then multi-pollutant
22 (PM + one or more copollutant) analyses for U.S. and non-U.S. studies.
23 Of particular importance is the NMMAPS multi-city study (Samet et al., 2000a,b;
24 Zanobetti et al., 2000a), as reanalyzed (Zanobetti and Schwartz, 2003b), which provides
25 evidence for significant PM effects on cardiovascular-related hospital admissions and visits,
26 using a variety of statistical models. These results are supported by another multi-city study
27 (Schwartz, 1999) which, however, has not been reanalyzed with alternative statistical models.
28 Numerous other studies, carried out by individual investigators in a variety of locales, present a
29 more varied picture, especially when gaseous co-pollutants have been analyzed in multipollutant
30 models. Most CVD hospital admissions studies reported to date have used PM10 as the main
31 particle measure due to the wide availability of ambient PM10 monitoring data. However, results
June 2003 8-120 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-16. SUMMARY OF STUDIES OF PM10, PM10 2 s, OR PM2 s
TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY
EFFECTS ON
VISITS
Reference
citation,
location, etc.
Outcome
measure
Mean PM
levels (IQR)
in ug/m3
Co-pollutants
analyzed with Lag
PM structure
Method
Effect measures
standardized to 50 (ig/m3
PM1(1 or 25 ng/m3 PM25*,
m**
10-2.5
U.S. Results Without Co-pollutants
Samet et al.
(2000a,b)
14 Cities
Total CVD
admissions
> 65 yrs
Zanobetti and Schwartz,
(2003b)
14 Cities
Lippmann et al.,
2000
Detroit (Wayne
County), MI
Ischemic heart
disease
> 65 yrs
Ito 2003
Detroit (Wayne County), MI
Lippmann etal.,
2000 Detroit
(Wayne
County), MI
Dysrhythmias
> 65 yrs
Ito 2003
Detroit (Wayne County), MI
Lippmann etal.,
2000
Detroit (Wayne
County), MI
Heart Failure
> 65 yrs
Ito 2003
Detroit (Wayne County), MI
Morris and
Naumova
(1998)
Chicago, IL
Congestive heart
failure
> 65 yrs
PM10 Means:
24.4-45.3
PM10 Means:
24.4-45.3
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 1 3 (7)
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 41 (23)
none 0 day
0-1 day
none 2 day
none 1 day
1 day*
0 day**
none 0 day
1 day*
0 day**
none 0 day
Default GAM
Default GAM
Strict GAM
GLMNS
GLMPS
Default GAM
Default GAM
Default GAM
Strict GAM
GLMNS
Strict GAM
GLMNS
Strict GAM
GLMNS
Default GAM
Default GAM
Default GAM
Strict GAM
GLMNS
Strict GAM
GLMNS
Strict GAM
GLMNS
Default GAM
Default GAM
Default GAM
Strict GAM
GLMNS
Strict GAM
GLMNS
Strict GAM
GLMNS
GAM not used
5.5% (4.7, 6.2)
5.9% (5. 1-6.7)
4.95% (3.95-5.95)
4.8% (3.55-6.0)
5.0% (4.0-5.95)
8.9% (0.5-18.0)
4.3% (-1.4-10.4)*
10. 5% (2.75-18.9)**
8.0% (-0.3-17.1)
6.2% (-2.0-15.0)
3.65% (-2.05-9.7)*
3.0% (-2.7-9.0)*
10.2% (2.4-18.6)**
8.1% (0.4-16.4)**
2.9% (-10. 8-18. 8)
3.2% (-6.5-14.0)*
0.2% (-12.2-14.4)**
2.8% (-10.9-18.7)
2.0% (-11. 7-17.7)
3.2% (-6.6-14.0)*
2.6% (-7.1-13.3)*
0.1% (-12.4-14.4)**
0.0% (-12. 5-14.3)**
9.7% (0.15-20.2)
9.1% (2.4- 16.2)*
5.2% (-3.25-14.4)**
9.2% (-0.3-19.6)
8.4% (-1.0-18.7)
8.0% (1.4-15.0)*
6.8% (0.3-13. 8)*
4.4% (-4.0-13. 5)**
4.9% (-3. 55-14.1)**
3.9% (1.0-6.9)
June 2003
8-121
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-16 (cont'd). SUMMARY OF STUDIES OF PM10, PM10 2 s, OR PM2 s EFFECTS
ON TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY VISITS
Reference
citation, Outcome
location, etc. measure
Mean PM
levels (IQR)
in ug/m3
Co-pollutants
analyzed with Lag
PM structure Method
Effect measures
standardized to 50 (ig/m3
PM10 or 25 ng/m3 PM25*,
PM **
rlvl!0-2.5
U.S. Results Without Co-pollutants (cont'd)
Linn et al. Total CVD
(2000) admissions
Los Angeles, > 30 yrs
CA
Moolgavkar Total CVD
(2000b) admissions
Cook County, > 65 yrs
IL
Moolgavkar (2003)
Cook County, IL
Moolgavkar Total CVD
(2000b) admissions
Los Angeles > 65 yrs
County, CA
Moolgavkar (2003)
Los Angeles County, CA
Tolbert et al., Total CVD
(2000a) emerg. dept.
Atlanta, GA visits, > 16 yrs
1993-1998
Tolbert et al., Total CVD
(2000a) emerg. dept.
Atlanta, GA visits, > 16 yrs
1998-1999
PM10: 45 (18)
PM10: 35* (22)
PM10: 44* (26)
PM25: 22* (16)
PM10: 44* (26)
PM25: 22* (16)
Period 1
PM10:
30.1, 12.4
Period 2
PM10: 29.1,
12.0
PM25: 19.4,9.4
PM10.25: 9.4,4.5
none 0 day GAM not used
none 0 day Default GAM
Strict
GAM100(lf
GLMNS10Mf
none 0 day Default GAM
Default GAM
Strict GAM3Mf
Strict
GAM100(lf
GLMNS10Mf
Strict GAM3Mf
Strict
GAM100(lf
GLM
nsplineloodf
none 0-2 day GAM not used
avg.
none 0-2 day GAM not used
avg.
3.25% (2.04, 4.47)
4. 2% (3.0, 5.5)
4.05% (2.9-5.2)
4.25% (3.0-5.5)
3.2% (1.2, 5.3)
4.3% (2.5, 6.1)*
3.35% (1.2-5. 5)
2.7% (0.6-4.8)
2.75% (0.1-5.4)
3.95% (2.2-5.7)*
2.9% (1.2-4.6)*
3. 15% (1.1-5.2)*
-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
Lippmann et al., Ischemic heart
2000 disease
Detroit (Wayne > 65 yrs
County), MI
Lippmann etal., Dysrhythmias
2000 > 65 yrs
Detroit (Wayne
County), MI
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.25: 13(7)
CO 2 day Default GAM
Default GAM
Default GAM
CO 1 day Default GAM
1 day Default GAM
0 day Default GAM
8.5% (-0.45-1 8.3)
3.7% (-2.4-10.3)*
10.1% (2.25-18.6)**
-1.3% (-15. 5-15.4)
0.55% (-9.7-12.0)*
-1.0% (-13.4-13.05)**
June 2003
8-122
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-16 (cont'd). SUMMARY OF STUDIES OF PM10, PM10 2 s, OR PM2 s EFFECTS
ON TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY VISITS
Reference
citation, Outcome
location, etc. measure
Mean PM
levels (IQR)
in ug/m3
Co-pollutants
analyzed with Lag
PM structure
Method
Effect measures
standardized to 50 (ig/m3
PM1(1 or 25 ng/m3 PM25*,
m**
10-2.5
U.S. Results With Co-pollutants (cont'd)
Lippmannetal., Heart Failure
2000 > 65 yrs
Detroit (Wayne
County), MI
Morris and Congestive heart
Naumova failure
(1998) > 65 yrs
Chicago, IL
Moolgavkar Total CVD
(2000b) admissions
Cook County, > 65 yrs
IL
Moolgavkar (2003)
Cook County, IL
Moolgavkar Total CVD
(2000b) admissions
Los Angeles > 65 yrs
County, CA
Moolgavkar (2003)
Los Angeles County, CA
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 41,23
PM10: 35,22
PM10: 35,22
PM10: 44* (26)
PM25: 22* (16)
PM10
PM2,
CO 0 day
1 day
Oday
CO, NO2, SO2, 0 day
03
N02 0 day
CO
CO 0 day
Default GAM
Default GAM
Default GAM
GAM not used
Default GAM
Strict
GAMloodf
GLMNS10Mf
Default GAM
Default GAM
Strict
GAMloodf
GLMNS10Mf
Strict
GAMloodf
GLMNS10Mf
7.5% (-2.6-18.7)
8.9% (2.2- 16.1)*
3.9% (-4.7-13.2)**
2% (-1-6)
1.8% (0.4, 3.2)
2.95% (1.7-4.2)
3.1% (1.8-4.4)
-1.8% (-4.4, 0.9)
0.8% (-1.3, 2.9)*
-1.3% (-3. 8-1. 2)
-1.1% (-4.2-2.0)
1.0% (-1.1-3.3)*
1.45% (-1.1-4.0)*
Non-U.S. Results Without Co-pollutants
Burnett et al., Total CVD
(1997a) admissions
Toronto, Canada all ages
Stieb et al. Total CVD
(2000) emerg. dept.
Saint John, visits, all ages
Canada
Atkinson et al. Total emerg.
(1999b) CVD
Greater London, admissions
England > 65 yrs
Prescottetal. Total CVD
(1998) admissions
Edinburgh, > 65 yrs
Scotland
Wong et al. Total emerg.
(1999a) CVD
Hong Kong admissions
> 65 yrs
PM10: 28,22
PM25: 17, 15
PM10.,5: 12,7
PM10: 14.0,9.0
PM25: 8.5, 5.9
PM10: 28.5,
90-10 %tile
range: 30.7
PM10: 20.7, 8.4
PM10: 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.
GAM not used
GAM not used
GAM not used
GAM not used
GAM not used
12.1% (1.4, 23.8)
7.2% (-0.6, 15.6)*
20. 5% (8.2, 34.1)**
29.3% (p=0.003)
14.4% (p = 0.055)*
2.5% (-0.2, 5.3)
12.4% (4.6, 20.9)
4.1% (1.3, 6.9)
June 2003
8-123
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-16 (cont'd). SUMMARY OF STUDIES OF PM10, PM10 2 s, OR PM2 s EFFECTS
ON TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY VISITS
Reference
citation,
location, etc.
Outcome
Measure
Mean PM
levels (IQR) in
ug/m3
Co-pollutants
Analyzed with Lag
PM Structure
Effect measures
standardized to 50 (ig/m3
PM10 or 25 ng/m3 PM25*,
Method PM1(V25**
Non-U.S. Results With Co-pollutants
Burnett et al.,
(1997a)
Toronto, Canada
Stieb et al.
(2000)
Saint John,
Canada
Atkinson et al.
(1999b)
Greater London,
England
Prescottet al.
(1998)
Edinburgh,
Scotland
Wong etal.
(1999a)
Hong Kong
Total CVD
admissions
all ages
Total CVD
emerg. dept.
visits, all ages
Total emerg.
CVD
admissions
> 65 yrs
Total CVD
admissions
> 65 yrs
Total emerg.
CVD
admissions
> 65 yrs
PM10: 28,
IQR 22
PM25: 17, 15
PM10.,5: 12,7
PM10: 14.0,9.0
PM10: 28.5,90-
10 %tile range:
30.7
PM10: 20.7,8.4
PM10: Median
45.0, IQR 34. 8
O3, NO2, SO2, 1-4 day
CO avg.
CO, H2S, N02, 1-3 day
O3, SO2, total avg.
reduced sulfur
N02, 03, S02, 0 day
CO
S02, N02, 03, 1-3 day
CO avg.
N02, 03, S02 0-2 day
avg.
GAM not used - 1 .4% (- 12.5, 1 1 .2)
-1.6% (-10. 5, 8.2)*
12.1% (-1.9, 28.2)**
GAM not used PM10 not significant;
no quantitative results
presented
GAM not used PM10 not significant;
no quantitative results
presented
GAM not used PM10 effect robust;
no quantitative results
presented
GAM not used PM10 effect robust;
no quantitative results
presented
*PM25 entries, **PM10.25. All others relate to PM10; 'Median.
1 from these studies may also be relevant to an assessment of PM2 5 health effects because PM2 5 is
2 known to represent 50% or more of PM10 in most locations, especially in urban areas typically
3 studied epidemiologically.
4 A substantial body of new results has emerged from analyses of daily CVD hospital
5 admissions in persons 65 and older in relation to PM10 in 14 cities from the NMMAPS multi-city
6 study (Samet et al., 2000a,b). The cities studied included Birmingham, AL; Boulder, CO;
7 Canton, OH; Chicago, IL; Colorado Springs, CO; Detroit, MI; Minneapolis/ St. Paul, MN;
8 Nashville, TN; New Haven, CT; Pittsburgh, PA; Provo/Orem, UT; Seattle, WA; Spokane, WA;
9 and Youngstown, OH. The range of years studied encompassed 1985-1994, although this varied
10 by city. Covariates included SO2, NO2, O3, and CO; however these were not analyzed directly as
11 regression covariates. Individual cities were analyzed first by Poisson regression methods on
June 2003
8-124
DRAFT-DO NOT QUOTE OR CITE
-------
1 PM10 for lags from 0 to 5 days. An overall PM10 risk estimate was then computed by taking the
2 inverse-variance weighted mean of the city-specific risk estimates. The city-specific risk
3 estimates for PM10 were also examined for correlations with omitted covariates, including other
4 pollutants. No relationship was observed between city-specific risk estimates and measures of
5 socioeconomic status, including percent living in poverty, percent non-white, and percent with
6 college educations. The overall weighted mean risk estimate for PM10 was greatest for lag 0 and
7 for the mean of lags 0-1. For example, the mean risk estimate for the mean of lags 0-1 was a
8 5.9% increase in CVD admissions per 50 |ig/m3 PM10 (95% CI: 5.1 - 6.7). The mean risk was
9 larger in a subgroup of data where PM10 was less than 50 |ig/m3, suggesting the lack of a
10 threshold. A weakness of this study was its failure to report multipollutant results. The authors
11 argued that confounding by co-pollutants was not present because the city-specific risk estimates
12 did not correlate with city-specific regressions of PM10 on co-pollutant levels. However, the
13 validity of this method for identifying meaningful confounding by co-pollutants at the daily
14 time-series level has not been demonstrated. Thus, it is not possible to conclude from these
15 results alone that the observed PM10 associations were independent of co-pollutants.
16 The Samet et al. (2000a,b) reports used GAM LOESS smoothing to control for time and
17 weather covariates. Data from the 14 city NMMAPs analysis of CVD hospital admissions were
18 reanalyzed recently (Zanobetti and Schwartz, 2003b) using three alternative control methods.
19 A small decrease in overall effects was observed as compared with the original study results.
20 Whereas the original 14 city pooled analysis yielded a 5.9% increase in CVD admissions per
21 50 |ig/m3 increase in mean lags 0 and 1 day PM10 (95% CI: 5.1-6.7%), the reanalysis reported
22 4.95% (3.95-5.95%), 4.8% (3.55-6.0%), and 5.0 (4.0-5.95%) when reanalyzed by GAM with
23 stringent convergence criteria, GLM with natural spline, and GLM with penalized spline,
24 respectively. On the basis of these results, no change is warranted with regard to the overall
25 conclusions for the original published study.
26 Zanobetti et al. (2000a) reanalyzed a subset of 10 cities from among the 14 evaluated by
27 Samet et al. (2000a,b). The same basic pattern of results obtained by Samet et al. (2000a,b) were
28 found, with strongest PM10 associations on lag 0 day, smaller effects on lag 1 and 2, and none at
29 longer lags. The cross-city weighted mean estimate at 0 day lag was excess risk = 5.6% (95%
30 CI 4.7, 6.4) per 50 |ig/m3 PM10 increment. The 0-1 day lag average excess CVD risk = 6.2%
31 (95% CI 5.4, 7.0) per 50 |ig/m3 PM10 increment. Effect-size estimates increased when data were
June 2003 8-125 DRAFT-DO NOT QUOTE OR CITE
-------
1 restricted to days with PM10 < 50 |ig/m3. As before, no evidence of gaseous (CO, O3, SO2)
2 co-pollutant modification of PM effects was seen in the second stage analyses. Again, however,
3 co-pollutants were not tested as independent explanatory variables in the regression analysis.
4 Like the larger NMMAPS morbidity analyses reported by Samet et al. (2000a,b), this sub-study
5 utilized the GAM function in SPlus. These 10 cities were among the 14 cities that Zanobetti and
6 Schwartz (2003b) recently reanalyzed using alternative statistical methods, and the results
7 discussed above would thus apply in general here.
8 Janssen et al. (2002), in further analyses of the data set examined above by Samet et al.
9 (2000a,b), evaluated whether differences in prevalence in air conditioning (AC) and/or the
10 contribution of different sources to total PM10 emissions could partially explain the observed
11 variability in exposure-effect relations in the 14 cities. Cities were characterized and analyzed as
12 either winter or nonwinter peaking for the AC analyses. Data on the prevalence of AC from the
13 1993 American Housing Survey of the United States Census Bureau (1995) were used to
14 calculate the percentage of homes with central AC for each metropolitan area. Data on PM10
15 emissions by source category were obtained by county from the U.S. EPA emissions and air
16 quality data web site (U.S. Environmental Protection Agency, 2000a). In an analysis of all
17 14 cities, central AC was not strongly associated with PM10 coefficients. However, separate
18 analysis for nonwinter-peaking and winter-peaking PM10 cities yielded coefficients for CVD-
19 related hospital admissions that decreased significantly with increased percentage of central AC
20 for both groups of cities. There were also significant positive relationships between CVD effects
21 and PM10 percent emissions from highways or from diesel vehicles, suggesting that mobile
22 source particles may have more potent cardiovascular effects than other particle types. For both
23 analyses, similar though weaker, patterns were found for hospitalization for COPD and
24 pneumonia. The authors note that the stronger relationship for hospital admission rates for CVD
25 over COPD and pneumonia may relate to the 10 times higher CVD hospital admissions rate
26 (which would result in a more precise estimate). However, no co-pollutant analyses were
27 reported. The ecologic nature and limited sample size also indicate the need for further study.
28 Because Janssen et al.'s analysis utilized the GAM function in SPlus, Zanobetti et al. (2003b)
29 reanalyzed the main findings from this study using alternative methods for controlling time and
30 weather covariates. While the main conclusions of the study were not significantly altered, some
31 changes in results are worth noting. The effect of air conditioning remained significant for the
June 2003 8-126 DRAFT-DO NOT QUOTE OR CITE
-------
1 non-winter PM10-peaking cities. The significance of highway vehicles and diesels on PM10
2 effect sizes remained significant, as did oil combustion. However, the effect of air conditioning
3 use on PM10 effect estimates was less pronounced and no longer statistically significant at p <
4 0.05 for the winter PM10-peaking cities using natural splines or penalized splines, in comparison
5 to the original Janssen et al. GAM analysis.
6 Schwartz (1999) extended the analytical approach he had used in Tucson (described below)
7 to eight more U.S. metropolitan areas, limiting analyses to a single county in each location to
8 enhance the representativeness of the air pollution data. The locations analyzed were Chicago,
9 IL; Colorado Springs, CO; New Haven, CT; Minneapolis, MN; St. Paul, MN; Seattle, WA;
10 Spokane, WA; and Tacoma, WA. Again, the analyses focused on total cardiovascular (CVD)
11 hospital admissions among persons > 65 years old. In univariate regressions, remarkably
12 consistent PM10 associations with CVD admissions were found across the eight locations, with a
13 50 |ig/m3 increase in PM10 associated with 3.6 to 8.6% increases in admissions. The univariate
14 eight-county pooled PM10 effect was 5.0% (CI 3.7-6.4), similar to the 6.1 % effect per 50 |ig/m3
15 observed in the previous Tucson analysis. In a bivariate model that included CO, the pooled
16 PM10 effect size diminished somewhat to 3.8% (CI 2.0-5.5) and the CO association with CVD
17 admissions was generally robust to inclusion of PM10 in the model. The Schwartz 1999 paper
18 used GAM LOESS smoothing with default convergence criteria to control for time and weather
19 covariates. To date, no revised results have been reported using alternative statistical methods.
20 Turning to some examples of independent single-city analyses, PM10 associations with
21 CVD hospitalizations were also examined in a study by Schwartz (1997), which analyzed three
22 years of daily data for Tucson, AZ linking total CVD hospital admissions for persons >65 years
23 old with PM10, CO, O3, and NO2. As was the above case in Chicago, only one site monitored
24 daily PM10, whereas multiple sites did so for gaseous pollutants (O3, NO2, CO). Both PM10 and
25 CO were independently (i.e., robustly) associated with CVD-related admissions; but O3 and NO2
26 were not. The percent effect of a 50 |ig/m3 increase in PM10 changed only slightly from
27 6.07 (CI 1.12-11.27) to 5.22 (CI 0.17 - 10.54) when CO was included in the model along with
28 PM10. The Schwartz 1997 paper utilized GAM smoothing to control for time and weather
29 covariates. To date, no revised results have been reported using alternative statistical methods.
30 Morris and Naumova (1998) reported results for PM10, as well as for O3, NO2, and SO2, in
31 an analysis of four years of congestive heart failure data among people > 65 years old in
June 2003 8-127 DRAFT-DO NOT QUOTE OR CITE
-------
1 Chicago, IL. As many as eight monitoring sites were available for calculating daily gaseous
2 pollutant concentrations; however, only one site in Chicago monitored daily PM10. Only same-
3 day results were presented, based on an initial exploratory analysis showing strongest effects for
4 same-day pollution exposure (i.e., lag 0). Associations between hospitalizations and PM10 were
5 observed in univariate regressions (3.9% [1.0, 6.9] per 50 |ig/m3 PM10 increase), but these
6 diminished somewhat in a multi-pollutant model (2.0%, [-1.4, 5.4]). Strong, robust associations
7 were seen between CO and congestive heart failure admissions. These results seem to suggest a
8 more robust association with CO than with PM10. However, the observed differences might also
9 be due in part to differential exposure misclassification for PM10 (monitored at one site) as
10 compared with CO (eight sites). This study did not use GAM functions to control for time and
11 weather covariates.
12 In a study designed to compare the effects of multiple PM indices, Lippmann et al. (2000)
13 analyzed associations between PM10, PM2 5, or PM10_2 5 and various categories of CVD hospital
14 admissions among the elderly (65+ yr) in Detroit on 344 days in the period 1992-1994. While
15 no consistent differences were observed in the relative risks for the alternative PM indices, many
16 of the associations involving PM were significant: (a) ischemic heart disease (IHD) in relation to
17 PM indices (i.e., 8.9% [0.5, 18.0] per 50 |ig PM10); 10.5% (2.8, 18.9) per 25 |ig/m3 PM10.25; and
18 4.3% (-1.4, 10.4) per 25 |ig/m3 PM25 (all at lag 2d); and (b) heart failure (i.e., 9.7% [0.2, 20.2]
19 per 50 |ig/m3 PM10); 5.2% (-3.3, 14.4) per 25 |ig/m3 PM10.25; and 9.1% (2.4, 16.2) per 25 |ig/m3
20 PM2 5 (the first two at lag 0 d and the latter at lag 1 d). No associations with dysrythmias were
21 seen however. The PM effects generally were robust when co-pollutants were added to the
22 model. Results for 2-pollutant models involving CO are given in Table 8-16 above.
23 As discussed earlier with regard to the Lippmann et al. (2000) mortality findings, it is difficult to
24 discern whether the observed associations with coarse fraction particles (PM10_2 5) are
25 independently due to such particles or may possibly be attributed to the moderately correlated
26 fine particle (PM2 5) fraction in Detroit. In addition, power was limited by the small sample size.
27 Because GAM was used in the analyses reported in Lippmann et al. (2000), Ito (2003) has
28 recently reported reanalyses results for the Detroit study using GAM with more stringent
29 convergence criteria and GLM with natural splines. PM effect sizes diminished somewhat (up to
30 30%) and sometimes lost significance. However, these changes tended to affect all PM metrics
31 in a similar fashion. Thus, there was no change in basic conclusions for the original Lippmann
June 2003 8-128 DRAFT-DO NOT QUOTE OR CITE
-------
1 et al. (2000) study, i.e., that there was no evidence for stronger effects for one size fraction
2 versus others. Ito (2003) also noted that study results were more sensitive to alternative weather
3 models and degree of smoothing (degrees of freedom used for the smoothing function) than to
4 whether or not GAM, with strict convergence criteria, was used.
5 As part of the ARIES Study, Tolbert et al. (2000a) initially reported preliminary results for
6 multiple PM indices as they relate to daily hospital emergency department (ED) visits for
7 dysrhythmias (DYS) and all CVD categories for persons aged 16 yrs or older, based on analyses
8 of data from 18 of 33 participating hospitals in Atlanta, GA. During Period 1 of the study (1993-
9 1998), PM10 from the EPA AIRS database was reported to be negatively associated with CVD
10 visits. In a subsequent one-year period (Aug. 1998-Aug. 1999), when data became available
11 from the Atlanta PM supersite, positive but non-significant associations were seen between CVD
12 and PM10 (RR of 5.1% per 50 |ig/m3 PM10) and PM2 5 (RR of 6.1% per 25 |ig/m3 PM25); and
13 significant positive associations were seen with certain fine particle components, i.e., elemental
14 carbon (p < 0.005) and organic carbon (p < 0.02), and CO (p < 0.005). No multi-pollutant
15 results were reported. Study power was limited due to the short data record in Period 2. More
16 complete analyses for January 1993 to August 2000 data from all participating hospitals have
17 recently been reported (Metzger et al., 2003) to show that, using an a priori 3-day morning
18 average in single-pollutant GLM analyses, CVD visits were associated with PM25, organic
19 carbon, elemental carbon, oxygenated hydrocarbons, CO, and NO2 (but not with O3 or SO2).
20 Secondary analyses suggested that these associations were strongest for same day air pollutant
21 levels.
22 In an analysis of 1992-1995 Los Angeles data, Linn et al. (2000) also found that PM10, CO,
23 and NO2 were all significantly associated with increased CVD admissions in single-pollutant
24 models among persons aged 30 yr and older. Associations generally appeared to be stronger for
25 CO than for PM10. No PM10 results were presented with co-pollutants in the model. Neither
26 Tolbert et al. nor Linn et al. reported any key findings based on GAM analyses.
27 Lastly, Moolgavkar (2000b) analyzed PM10, CO, NO2, O3, SO2 and limited PM2 5 data in
28 relation to daily total cardiovascular (CVD) and total cerebrovascular (CrD) admissions for
29 persons aged >65 from three urban counties (Cook, IL; Los Angeles, CA; Maricopa, AZ) in the
30 period 1987-1995. Of particular note was the availability of PM25 data in LA, though only every
31 sixth day. Consistent with most studies, in univariate regressions, PM10 (and PM2 5 in LA) were
June 2003 8-129 DRAFT-DO NOT QUOTE OR CITE
-------
1 associated at some lags with CVD admissions in Cook and LA counties, but not in Maricopa
2 county. However, in two-pollutant models in Cook and LA counties, the PM risk estimates
3 diminished substantially and/or were rendered non-significant, whereas co-pollutant (CO or
4 NO2) risk estimates were less affected. These results suggest that gaseous pollutants, with the
5 exception of O3, may have been more strongly associated with CVD hospitalizations than was
6 PM. These findings were based on an analysis that used GAM functions for time and weather
7 controls. Moolgavkar (2003) reported results of a reanalysis using improved GAM convergence
8 criteria and GLM with natural splines (nspline) and a range of degrees of freedom (30 versus
9 100) for the smooth function of time. Results were not very sensitive to the use of default versus
10 improved GAM or splines (Table 8-16) but did appear to be more sensitive to degrees of
11 freedom. The nspline results were given only with 100 degrees of freedom. This is an unusually
12 large number, especially for PM2 5, where data were available only every sixth day over a nine
13 year period.
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
16 increased ED visits or hospitalizations. The Lippmann results appear to implicate both PM2 5
17 and PM10_25 in increased 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,c, 1999). A variety of locations, outcomes,
22 PM exposure metrics, and analytical approaches were used, which hinders somewhat the ability
23 to draw broad conclusions across the full group of studies. The first study (Burnett et al., 1995),
24 reviewed briefly in the 1996 PM AQCD, analyzed six years of data from 168 hospitals in
25 Ontario, CN. Respiratory and CVD hospital admissions were analyzed in relation to sulfate and
26 O3 concentrations. Sulfate lagged one day was associated with CVD admissions, with an effect
27 of 2.8% (CI 1.8-3.8) increase per 13 |ig/m3 SO4'2 without O3 in the model and 3.3% (CI 1.7-4.8)
28 with O3 included. When CVD admissions were split out into sub-categories, larger associations
29 were seen between sulfates and coronary artery disease and heart failure than for cardiac
30 dysrhythmias. Sulfate associations with total admissions were larger for the elderly > 65 yr old
June 2003 8-130 DRAFT-DO NOT QUOTE OR CITE
-------
1 (3.5% per 13 |ig/m3) than for those < 65 yr old (2.5% per 13 |ig/m3). There was little evidence
2 for seasonal differences in sulfate associations.
3 Burnett et al. (1997c) analyzed daily congestive heart failure hospitalizations in relation to
4 CO and other air pollutants (O3, NO2, SO2, CoH) in ten large Canadian cities as a replication of
5 an earlier U.S. study by Morris et al. (1995). The Burnett Canadian study expanded upon the
6 previous work both by its size (11 years of data for each of 10 large cities) and by including a
7 measure of PM air pollution (coefficient of haze, CoH); whereas no PM data were included in
8 the earlier Morris et al. study. The Burnett study was restricted to the population > 65 years old.
9 The authors noted that all pollutants except O3 were correlated, making it difficult to separate
10 them statistically. CoH, CO, and NO2 measured on the same day as admission (i.e., lag 0) were
11 all strongly associated with congestive heart failure admissions in univariate models. In multi-
12 pollutant models, CO remained a strong predictor, but CoH did not (no gravimetric PM
13 measures were used).
14 The roles played by size-selected gravimetric and chemically-speciated particle metrics as
15 predictors of CVD hospitalizations were explored in analyses of data from metropolitan Toronto
16 for the summers of 1992-1994 (Burnett et al., 1997a). The analyses used dichotomous sampler
17 (PM2 5, PM10, and PM10_25), hydrogen ion, and sulfate data collected at a central site as well as
18 O3, 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-related admissions. Model specification
21 with respect to pollution lags was completely data-driven, with all lags and averaging times out
22 to 4 days prior to admission evaluated in exploratory analyses and "best" metrics chosen on the
23 basis of maximal t-statistics. The relative risks of CVD admissions were positive and generally
24 statistically significant for all pollutants analyzed in univariate regressions, but especially so for
25 O3, NO2, CoH, and PM10_2 5 (i.e., regression t-statistics > 3). Associations for gaseous pollutants
26 were generally robust to inclusion of PM covariates, whereas the PM indices (aside from CoH)
27 were 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 |ig/m3 increase in PM25
29 was associated with a 7.2% increase (t = 1.8) in CVD admissions in a univariate model, the
30 effect was reduced to -1.6% (t = 0.3) in a model that included O3, NO2, and SO2. CoH, like CO
31 and NO2, is generally thought of as a measure of primary motor-vehicle emissions during the
June 2003 8-131 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
9 of the peripheral circulation; the latter categories were included because they should show PM
10 associations if one mechanism of PM action is related to increased plasma viscosity, as
11 suggested by Peters et al. (1997a). The PM metrics analyzed were PM25, PM10, and PM10_25
12 estimated from daily TSP and TSP sulfate data, based on a regression analysis for dichotomous
13 sampling data that were available every sixth day during an eight-year subset of the full study
14 period. This use of estimated rather than measured PM components limits interpretation of the
15 reported PM results, i.e., in general, use of estimated PM exposure metrics should tend to
16 increase exposure measurement error and thereby tend to decrease effects estimates. Model
17 specification for lags was again data-driven, based on maximal t-statistics. Although some
18 statistically significant associations with one or another PM metric were found in univariate
19 models, there were no significant PM associations with any of the three CVD hospitalization
20 outcomes in multi-pollutant models. For example, whereas an 25 |ig/m3 increase in estimated
21 PM2 5 was associated with a 8.05% increase (t-statistic = 6.08) in ischemic heart disease
22 admissions in a univariate analysis, the PM2 5 association was reduced to 2.25% (n.s.) when NO2
23 and SO2 were included in the model. The gaseous pollutants dominated most regressions. There
24 also were no associations 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
27 various analyses confuses the picture. 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 PM25 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
June 2003 8-132 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 influence of primary motor vehicle emissions. This contrasts, however,
6 with CoH's lack of robustness in the 10-city analysis (Burnett et al., 1997 c).
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 The APHEAII (Le Tertre et al., 2002) project examined the association between PM10 and
13 hospital admissions for cardiac causes in eight European cities. They found a significant effect
14 of PM10 (0.5%; 0.2, 0.8) on admission for cardiac causes (all ages) and cardiac causes (0.7%;
15 0.4, 1.0) and ischemic heart disease (0.8%; 0.3, 1.2) for people over 65 years, with the effect of
16 PM10 per unit of pollution being half that found in the United States. PM10 did not seem to be
17 confounded by O3 or SO2. The PM10 effect was reduced when CO was incorporated in the
18 regression model and eliminated when controlling for NO2. In contrast to PM10, black smoke
19 was robustly associated with CVD hospital admissions when co-pollutants were introduced into
20 the model. This led the authors to suggest that diesel PM may be especially important. GAM
21 functions were used in the original analysis. In a recent reanalysis using GAM with stringent
22 convergence criteria and GLM with either natural or penalized splines, no marked changes from
23 original results were observed (Le Tertre et al., 2003).
24 Several additional non-U.S. studies, mainly in the U.K., have also been published since the
25 1996 PM AQCD. Most of these studies evaluated co-pollutant effects along with those of PM.
26 Interpretation is hindered somewhat, however, by the failure to report quantitative results for
27 PM10 in the presence of co-pollutants. In univariate models, Atkinson et al. (1999b) reported PM
28 associations for persons aged < 65 yr and for persons aged > 65 yr. Significant associations
29 were reported for both ambient PM10 and black smoke (BS), as well as all other co-pollutants,
30 with daily admissions for total cardiovascular disease and ischemic heart disease for 1992-1994
31 in London, UK, using standard time-series regression methods. In two-pollutant models, the
June 2003 8-133 DRAFT-DO NOT QUOTE OR CITE
-------
1 associations with PM10, NO2, SO2, and CO were moderated by the presence of BS in the model,
2 but the BS association was robust to co-pollutants. Interpretation is hampered somewhat by the
3 lack of quantitative results for two-pollutant models.
4 In another U.K. study, associations with PM10, and to a lesser extent BS, SO2, and CO,
5 were reported for analyses of daily emergency hospital admissions for cardiovascular diseases
6 from 1992-1995 for Edinburgh, UK (Prescott et al., 1998). No associations were observed for
7 NO2 and O3. Significant PM10 associations for CVD admissions were present only in persons
8 < 65 yrs old. The authors reported that the PM10 associations were unaffected by inclusion of
9 other pollutants; however, results were not shown. On the other hand, no associations between
10 PM10 and daily ischemic heart disease admissions were observed by Wordley and colleagues
11 (1997) in an analysis of two years of daily data from Birmingham, UK. However, PM10 was
12 associated with respiratory admissions and cardiovascular mortality during the same study
13 period. This inconsistency of results across causes and outcomes is difficult to interpret, but may
14 relate in part to the relatively short time-series analyzed. The authors stated that gaseous
15 pollutants did not have significant associations with health outcomes independent of PM, but no
16 results were presented for models involving gaseous pollutants.
17 A study in Hong Kong by Wong et al. (1999a) found associations between CVD
18 admissions and PM10, SO2, NO2, and O3 in univariate models, but did not examine multi-
19 pollutant models. In models including PM10 and dichotomous variables for gaseous pollutants
20 (high versus low concentration), the PM10 effects remained relatively stable. Ye and colleagues
21 analyzed a 16 year record of daily emergency hospital visits for July and August in Tokyo
22 among persons age 65 and older (Ye et al., 2001). In addition to PM10, the study included NO2,
23 O3, SO2, and CO. Models were built using an objective significance criterion for variable
24 inclusion. NO2 was the only pollutant significantly associated with angina, cardiac
25 insufficiency, and myocardial infarction hospital visits.
26
27 8.3.1.3.3 Summary of Salient Findings for Acute PM Exposure Effects on CVD Hospital
28 Admissions
29 The ecologic time-series studies reviewed here add substantially to the body of evidence on
30 acute CVD morbidity effects of PM and co-pollutants. Two U.S. multi-city studies offer the
31 strongest current evidence for effects of PM10 on acute CVD hospital admissions, but
32 uncertainties regarding the possible role of co-pollutants in the larger of the two studies hinders
June 2003 8-134 DRAFT-DO NOT QUOTE OR CITE
-------
1 interpretation with respect to independent PM10 effects. Among single-city studies carried out in
2 the U.S. and elsewhere by a variety of investigators (see Table 8-16), less consistent evidence for
3 PM effects is seen. Of particular importance is the possible roles of co-pollutants (e.g., CO) as
4 confounders of the PM effect. Among 13 independent studies that included gravimetrically-
5 measured PM10 and co-pollutants, three reported PM effects that appeared to be independent of
6 co-pollutants (Schwartz, 1997; Lippmann et al., 2000; Prescott et al., 1998); eight reported no
7 significant PM10 effects after inclusion of co-pollutants (Morris and Naumova, 1998;
8 Moolgavkar, 2000b; Tolbert et al., 2000a; Burnett et al., 1997a; Steib et al., 2000; Atkinson
9 et al., 1999b; Wordley et al. (1997); Morgan et al., 1998; Ye et al., 2001); and two studies were
10 unclear regarding independent PM effects (Linn et al., 2000; Wong et al., 1999a). In a recent
11 quantitative review of published results from 12 studies on airborne particles and hospital
12 admissions for cardiovascular disease, Morris (2001) noted that adjustment for co-pollutants
13 consistently reduced the PM10 effect, with reductions ranging from 10 to 320% across studies.
14 Thus, although several studies do appear to provide evidence for PM effects on CVD hospital
15 admissions independent of co-pollutant effects, a number of other studies examining
16 co-pollutants did not find results indicative of independent PM10 effects on CVD hospital
17 admissions
18 With respect to particle size, only a handful of studies have examined the relative effects of
19 different particle indicators (Lippmann et al., 2000; Burnett et al., 1997a; Tolbert et al., 2000a;
20 Steib et al., 2000; Moolgavkar, 2000b). Perhaps due to statistical power issues, no clear picture
21 has emerged as to particle-size fraction(s) most associated with acute CVD effects.
22 As discussed above, several studies originally based on statistical analyses involving the
23 SPlus GAM function have reported new results using alternative statistical methods. The
24 reanalyses yielded some slightly reduced effect estimates and/or increased confidence intervals
25 or little or no change resulted in other cases. Thus, based on these new results, the overall
26 conclusions from the cardiovascular hospitalization studies remain the same.
27 Because hospitalization can be viewed as likely reflecting some of the same
28 pathophysiologic mechanisms that may be responsible for acute mortality following PM
29 exposure, it is of interest to assess the coherence between the morbidity results reviewed here
30 and the mortality results reviewed in Section 8.2.2 (Borja-Aburto et al., 1997, 1998; Braga et al.,
31 2001; Goldberg et al., 2000; Gouveia and Fletcher, 2000; Hoek et al., 2001; Kwon et al., 2001;
June 2003 8-135 DRAFT-DO NOT QUOTE OR CITE
-------
1 Michelozzi et al., 1998; Morgan et al., 1998; Ponka et al., 1998; Schwartz et al., 1996a; Simpson
2 et al., 1997; Wordley et al., 1997; Zeghnoun et al., 2001; Zmirou et al., 1998). The mortality
3 studies reported significant associations between acute CVD mortality and measures of ambient
4 PM, though the PM metrics used and the relative risk estimates obtained varied across studies.
5 The PM measurement methods included gravimetrically analyzed filter samples (TSP, PM10,
6 PM2 5, PM10_2 5), beta gauge (particle attenuation of beta radiation), nephelometry (light
7 scattering), and black smoke (filter reflectance). Where tested, PM associations with acute CVD
8 mortality appeared to be generally more robust to inclusion of gaseous covariates than was the
9 case for acute hospitalization studies (Borja-Aburto et al., 1997, 1998; Morgan et al., 1998;
10 Wordley et al., 1997; Zmirou et al., 1998). One study (Goldberg et al., 2000) which examined
11 multiple alternative PM metrics, reported strongest associations with PM2 5 and no associations
12 for PM10_2 5 and hydrogen ion. Three studies (Braga et al., 2001; Goldberg et al., 2000; Hoek
13 et al., 2001), as noted in Section 8.2.2, provide data indicating that some specific CVD causes of
14 mortality (such as heart failure) were more strongly associated with air pollution than total CVD
15 mortality; but it was noted that ischemic heart disease (which contributes about half of all CVD
16 deaths) was the strongest contributor to the association between air pollution and cardiovascular
17 mortality. Checkoway et al. (2000) evaluated the possible association between the occurrence of
18 out-of-hospital sudden cardiac arrest (SCA) for cases free of prior clinically-recognized heart
19 disease or major life-threatening co-morbidity and daily PM levels in Seattle (PM10 mean =
20 31.9 |ig/m3) and reported an estimated relative risk at a one day lag of 0.87 (95% CI: 0.74,
21 1.01). The above-noted results for acute CVD mortality are qualitatively consistent with those
22 reviewed earlier in this section for hospital admissions.
23 Figure 8-12 illustrates PM10 excess risk estimates for single-pollutant models derived from
24 selected U.S. studies of PM10 exposure and total CVD hospital admissions, standardized to a
25 50 |ig/m3 exposure to PM10 as shown in Table 8-16. Results are shown both for studies yielding
26 pooled outcomes for multiple U.S. cities and for studies of single U.S. cities. The Zanobetti and
27 Schwartz (2003b) and Samet et al. (2000a) pooled cross-city results for 14 U.S. cities provide
28 the most precise estimate for relationships of U.S. ambient PM10 exposure to increased risk for
29 CVD hospitalization. That estimate, and those derived from most other studies depicted in
30 Figure 8-12, generally appear to confirm likely excess risk of CVD-related hospital admissions
31 for U.S. cities in the range of 3-9% per 50 |ig/m3 PM10, especially among the elderly (> 65 yr).
June 2003 8-136 DRAFT-DO NOT QUOTE OR CITE
-------
Zannobetti and Schwartz (2003b)
14 US cities
Moolgavkar (2003)
LA,CA
Moolgavkar (2003)
Cook County
Linn et al. (2000)
LA.CA
Tolbert et al. (2000a)
Atlanta
Morris &Naumova
(1998) Chicago
Ito (2003)
Detroit
Total CVD
CHF
HF
IHD —
-10
10
15
20
25
Reconstructed Excess Risk Percentage
50 iig/m3 Increase in PMio
Figure 8-12. Acute cardiovascular hospitalizations and particulate matter exposure excess
risk estimates derived from selected U.S. PM10 studies based on single-
pollutant models. Both multi-pollutant models and PM2 5 and PM10_2 5 results
are shown in Table 8-16. CVD = cardiovascular disease. CHF = congestive
heart failure, HF = heart failure.
1
2
3
4
5
6
7
8
9
10
11
Other individual-city results (see Table 8-16) from Detroit are also indicative of excess risk for
ischemic heart disease in the range of approximately 3.0 and 8.1% per 25 |ig/m3 of PM25 or
PMio-2.5, respectively, and for heart failure of 6.8% and 4.9% excess risk per 25 |ig/m3 of PM25
and PM10_2 5, respectively. However, the extent to which PM affects CVD-hospitalization risks
independently of, or together with other co-pollutants (such as CO), remains to be further
resolved.
8.3.1.3.4 Individual-Level Studies of Cardiovascular Physiology
Several new studies have evaluated longitudinal associations between ambient PM and
physiologic measures of cardiovascular function or biochemical changes in the blood that may
be associated with cardiac risks. In contrast to the ecologic time-series studies discussed above,
June 2003
8-137
DRAFT-DO NOT QUOTE OR CITE
-------
1 these studies measure outcomes and most covariates at the individual level, making it possible to
2 draw conclusions regarding individual risks, as well as to explore mechanistic hypotheses.
3 Heterogeneity of responses across individuals, and across subgroups defined on the basis of age,
4 sex, pre-existing health status, etc., also can be assessed, in principle. While exposure
5 assessment remains largely ecologic (i.e., the entire population is usually assigned the same
6 exposure value on a given day), exposure is generally well characterized in the small, spatially-
7 clustered study populations. The recent studies fall into two broad classes: (1) those addressing
8 cardiac rhythm or adverse events and (2) those addressing blood characteristics. While
9 significant uncertainty still exists regarding the interpretation of results from these new studies,
10 the varied responses that have been reported to be associated with ambient PM and co-pollutants
11 are of much interest in regard to mechanistic hypotheses concerning pathophysiologic processes
12 potentially underlying CVD-related mortality/morbidity effects discussed in preceding sections.
13
14 Cardiac Physiology and Adverse Cardiac Events
15 Alterations in heart rate and/or rhythm have been hypothesized as possible mechanisms by
16 which ambient PM exposures may exert acute effects on human health. Decreased heart rate
17 variability, in particular, has been identified as a predictor of increased cardiovascular morbidity
18 and mortality. Several independent studies have recently reported temporal associations
19 between PM exposures and various measures of heart beat rhythm in panels of elderly subjects
20 (Liao et al., 1999; Pope et al., 1999a,b,c; Dockery et al., 1999; Peters et al., 1999a, 2000a; Gold
21 et al. 2000; Creason et al., 2001). Changes in blood pressure may also reflect increases in CVD
22 risks (Linn et al., 1999; Ibald-Mulli et al., 2001). Finally, one important new study (Peters et al.,
23 200la) has linked acute (2- and 24-h) ambient PM2 5 and PM10 concentrations with increased risk
24 of myocardial infarction in subsequent hours and days.
25 Liao et al. (1999) studied 26 elderly subjects (age 65-89 years; 73% female) over three
26 consecutive weeks at a retirement center in metropolitan Baltimore, 18 of whom were classified
27 as "compromised" based on previous cardiovascular conditions (e.g., hypertension). Daily six-
28 minute resting electrocardiogram (ECG) data were collected, and time intervals between
29 sequential R-R intervals recorded. A Fourier transform was applied to the R-R interval data to
30 separate its variance into two major components: low frequency (LF, 0.04-0.15 Hz) and high
31 frequency (HF, 0.15-0.40 Hz). The standard deviation of all normal-to-normal (N-N; also
June 2003 8-138 DRAFT-DO NOT QUOTE OR CITE
-------
1 designated R-R) heartbeat intervals (SDNN) was computed as a time-domain outcome variable.
2 PM2 5 was monitored indoors by TEOM and outdoors by dichotomous sampler. Outdoor PM2 5
3 levels ranged from 8.0 to 32.2 |ig/m3 (mean = 16.1 |ig/m3). Regression analyses controlled for
4 inter-subject differences in average variability, allowing each subject to serve as his/her own
5 control. Consistent associations were seen between decreases in all three outcome variables (LF,
6 HF, SDNN) and increases in PM2 5 levels (both indoors and outdoors), with associations being
7 stronger for the 18 "compromised" subjects. No analyses of heart rate were reported.
8 Creason et al. (2001) reported results of a subsequent study using similar methods among
9 56 elderly residents of a retirement center in Baltimore County, MD. The 11 men and 45 women
10 ranged in age from 72 to 97 years and were all Caucasian. Associations between decreased
11 FIRV and ambient PM2 5 were again seen, though not significant at p < 0.05 level and smaller
12 than in the previous Baltimore study. When two episodic PM2 5 days with rainfall were excluded
13 from the 24-day data set, the PM25 associations increased in magnitude and became statistically
14 significant. There was no evidence of larger effects among subsets of subjects with
15 compromised health status. No results were presented for other pollutants besides PM25.
16 Pope and colleagues (1999c) reported similar findings in a panel of six elderly subjects
17 (69-89 years, 5/6 male) with histories of cardiopulmonary disease, and one 23-year old male
18 subject suffering from Crohn's disease and arrhythmias. Subjects carried Hotter monitors for up
19 to 48 hours during different weeks that varied in ambient PM10 concentrations. N-N heartbeat
20 intervals were recorded to calculate several measures of heart rate variability in the time domain:
21 the standard deviation of N-N intervals (SDNN), a broad measure of both high and low
22 frequency variations; the standard deviation of the averages of N-N intervals in all five minute
23 segments (SDANN), a measure of ultra-low frequency variations; and the root mean squared
24 differences between adjacent N-N intervals (r-MSSD), a measure of high frequency variations.
25 Daily gravimetric PM10 data obtained from three sites in the study area ranged from circa
26 10 |ig/m3 to 130 |ig/m3 during the study. A simple step function in PM concentration was
27 observed, with high levels occurring only during the first half of the 1.5 month study period.
28 Regression analysis with subject-specific intercepts was performed, with and without control for
29 daily barometric pressure and mean heart rate. Same-day, previous-day, and the two-day mean
30 of PM10 were considered. SDNN and SDANN were negatively associated with both same-day
31 and previous-day ambient PM10, and results were unaffected by inclusion of covariates. Heart
June 2003 8-139 DRAFT-DO NOT QUOTE OR CITE
-------
1 rate, as well as r-MSSD, were both positively, but less strongly, associated with PM10. No co-
2 pollutants were studied.
3 The Pope et al. (1999c) study discussed above was nested within a larger cohort of
4 90 subjects who participated in a study of heart rate and oxygen saturation in the Utah Valley
5 (Dockery et al., 1999; Pope et al., 1999b). The investigators hypothesized that decreases in
6 oxygen saturation might occur as a result of PM exposure, and that this could be a risk factor for
7 adverse cardiac outcomes. The study was carried out in winter months (mid-November through
8 mid-March), when frequent inversions lead to fine particle episodes. PM10 levels at the three
9 nearest sites averaged from 35 to 43 |ig/m3 during the study, and daily 24-h levels ranged from
10 5 to 147 |ig/m3. Two populations were studied: 52 retired Brigham Young University
11 faculty/staff and their spouses, and 38 retirement home residents. Oxygen saturation (SpO2) and
12 heart rate (HR) were measured once or twice daily by an optical sensor applied to a finger.
13 In regression analyses controlling for inter-individual differences in mean levels, SpO2 was not
14 associated with PM10, but was highly associated with barometric pressure. In contrast, HR
15 association with PM10 significantly increased but significantly decreased with barometric
16 pressure in joint regressions. Including CO in the regressions did not change these basic
17 findings. This was the first study of this type to examine the interrelationships among
18 physiologic measures (i.e., SpO2 and HR), barometric pressure, and PM10. The profound
19 physiological effects of barometric pressure noted here highlight the importance of carefully
20 controlling for barometric pressure effects in studies of cardiac physiology.
21 Gold and colleagues (2000) obtained somewhat different results in a study of heart rate
22 variability among 21 active elderly subjects, aged 53-87 yr, in a Boston residential community.
23 Resting, standing, exercising, and recovering ECG measurements were performed weekly using
24 a standardized protocol on each subject, which involved 25 min/week of continuous Hotter ECG
25 monitoring. Two time-domain measures were extracted: SDNN and r-MSSD (see above for
26 definitions). Heart rate also was analyzed as an outcome. Continuous PM10 and PM2 5
27 monitoring was conducted by TEOM at a site 6 km from the study site and PM data were
28 corrected for the loss of semivolatile mass. Data on CO, O3, NO2, SO2, temperature and relative
29 humidity were available from nearby sites. Outcomes were regressed on PM2 5 levels in the
30 0-24 hour period prior to ECG testing, with and without control for HR and temperature. As for
31 the other studies discussed above, declines in SDNN were associated with PM2 5 levels, in this
June 2003 8-140 DRAFT-DO NOT QUOTE OR CITE
-------
1 case averaged over 4 hours. These associations reached statistical significance at the
2 p < 0.05 level only when all testing periods (i.e., resting, standing, exercise) were combined.
3 In contrast to the above studies, both HR and r-MSSD here were negatively associated with
4 PM2 5 levels (i.e., lower HR and r-MSSD) when PM2 5 was elevated. These associations were
5 statistically significant overall, as well as for several of the individual testing periods, and were
6 unaffected by covariate control. Gold et al. (2003) has recently reported revised results that
7 involve analyzing temperature with either a GAM function with stringent convergence criteria or
8 a GLM with natural splines, with no substantial changes being reported.
9 Further evidence for decreased HRV in response to PM2 5 exposures comes from several
10 recent studies. Significant decreases in SDNN of 1.4% (95% CI = 2.1 to -0.6) per 100 ug/m3
11 3-hour mean PM25 were found in a group of young healthy boilermakers in the Boston area who
12 were studied during non-work periods (Magari et al., 2001). Use of estimated PM25 based on
13 light scattering precludes a firm quantitative interpretation of exposure levels in terms of
14 gravimetric PM25 concentrations. A previous study of 40 boilermakers (including the 20 studied
15 above) analyzed data collected during both work and non-work time periods (Magari et al.,
16 2002). That study reported a significant 2.7% decrease in SDNN and a 1.0% increase in HR, for
17 every 100 |ig/m3 increase in 4-hour moving average estimated PM2 5. The larger effect size for
18 the non-work PM exposure study may reflect differing health effects of ambient versus
19 occupational PM composition. These studies are important in showing HRV effects in young
20 healthy adults.
21 Peters et al. (1999a) reported HR results from a retrospective analysis of data collected as
22 part of the MONICA study (monitoring of trends and determinants in cardiovascular disease)
23 carried out in Augsburg, Germany. Analyses focused on 2,681 men and women aged 25-64
24 years who had valid ECG measurements taken in winter 1984-1985 and again in winter 1987-
25 1988. Ambient pollution variables included TSP, SO2, and CO. The earlier winter included a
26 10-day episode with unusually high levels of SO2 and TSP, but not of CO. Pollution effects
27 were analyzed in two ways: dichotomously comparing the episode and non-episode periods, and
28 continuously using regression analysis. However, it is unclear from the report as to what extent
29 the analyses reflect between-subject versus within-subject effects. A statistically significant
30 increase in mean heart rate was seen during the episode period versus other periods, controlling
June 2003 8-141 DRAFT-DO NOT QUOTE OR CITE
-------
1 for cardiovascular risk factors and meteorology. Larger effects were observed in women.
2 In single-pollutant regression analyses, all three pollutants were associated with increased HR.
3 In another retrospective study, Peters and colleagues (2000a) examined incidence of
4 cardiac arrhythmias among 100 patients (mean age 62.2 yr; 79% male) with implanted
5 cardioverter defibrillators followed over a three year period. PM2 5 and PM10 were measured in
6 South Boston by the TEOM method, along with black carbon, O3, CO, temperature and relative
7 humidity; SO2 and NO2 data were obtained from another site. The 5th percentile, mean, and 95th
8 percentiles of PM10 levels were 7.8, 19.3, and 37.0 |ig/m3, respectively. The corresponding PM25
9 values were 4.6, 12.7, and 26.6 |ig/m3. Logistic regression was used to analyze events in relation
10 to pollution variables, controlling for between-person differences, seasons, day-of-week, and
11 meteorology in two subgroups: 33 subjects with at least one arrhythmia event and 6 subjects
12 with 10 or more such events. In the larger subgroup, only NO2 on the previous day, and the
13 mean NO2 over five days, were significantly associated with arrhythmia incidence. In patients
14 with 10 or more events, the NO2 associations were stronger. Also, some of the PM2 5 and CO
15 lags became significant in this subgroup.
16 Linn et al. (1999) reported associations between both diastolic and systolic blood pressure
17 and PM10 in a panel study of 30 Los Angeles residents with severe COPD. Recently, Ibald-Mulli
18 et al. (2001) reported similar findings from a study of blood pressure among 2607 men and
19 women aged 25-64 years in the MONICA study in Augsburg, Germany. Systolic blood pressure
20 increased on average during an episode of elevated TSP and SO2, but the effect disappeared after
21 controlling for meteorological parameters that included temperature and barometric pressure.
22 However, when TSP and SO2 were analyzed as continuous variables, both were associated with
23 elevated systolic blood pressure, controlling for meteorological variables. In two-pollutant
24 models, TSP was more robust than SO2. Further, the TSP association was greater in the
25 subgroups of subjects with elevated blood viscosity and heart rates.
26 An exploratory study of a panel of COPD patients (Brauer et al., 2001) examined several
27 PM indicators in relation to CVD and respiratory health effects. The very low levels of ambient
28 particles (PM10 mean =19 |ig/m3) and low variability in these levels plus the sample size of
29 16 limit the conclusions that can be drawn. Nevertheless, for cardiovascular endpoints, single-
30 pollutant models indicated that both systolic and diastolic BP decreased with increasing
31 exposure, but this is not statistically significant. The size of the ambient PM10 effect estimate for
June 2003 8-142 DRAFT-DO NOT QUOTE OR CITE
-------
1 AFEVj was larger than the effect estimate for ambient PM2 5 and personal PM2 5 but not
2 statistically significant. This initial effort indicates that ambient PM10 consistently had the
3 largest effect estimates while models using personal exposure measurements did not show larger
4 or more consistently positive effect estimates relative to those using ambient exposure metrics.
5 An important study by Peters et al. (200la) reported associations between onset of
6 myocardial infarction (MI) and ambient PM (either PM10 or PM2 5) as studied in a cohort of
7 772 MI patients in Boston, MA. Precise information on the timing of the MI, obtained from
8 patient interviews, was linked with concurrent air quality data measured at a single Boston site.
9 A case crossover design enabled each subject to serve as his/her own control. One strength of
10 this study was its analysis of multiple PM indices and co-pollutants, including real-time PM25,
11 PM10, the PM10.2 5 difference, black carbon, O3, CO, NO2, and SO2. Only PM2 5 and PM10 were
12 significantly associated with MI risk in models adjusting for season, meteorological parameters,
13 and day of week. Both the mean PM25 concentration in the previous two hours and in the 24
14 hours lagged one day were independently associated with MI, with odds ratios of 1.48 (1.09-
15 2.02) for 25 ug/m3 and 1.62 (1.13-2.34) for 20 ug/m3, respectively. PM10 associations were
16 similar. The non-significant findings for other pollution metrics should be interpreted in the
17 context of potentially differing exposure misclassification errors associated with the single
18 monitoring site.
19 The above studies present a range of findings suggesting possible effects of PM25 on
20 cardiac rhythm and adverse events. Numerous studies reported decreases in HR variability
21 associated with PM in elderly subjects with preexisting cardiopulmonary disease, although
22 r-MSSD (a measure of high-frequency HR variability) showed elevations with PM in one study
23 (Pope et al., 1999a). Recent studies also reported effects in healthy elderly and young adult
24 populations. All of the studies which examined HR also found an association with PM; most
25 reported positive associations, but one (Gold et al., 2000) reported a negative relationship.
26 However, variations in methods and results across the studies argue for caution in drawing
27 strong conclusions regarding PM effects from them.
28
29 Viscosity and Other Blood Characteristics
30 Peters et al. (1997a) state that plasma viscosity, a risk factor for ischemic heart disease, is
31 affected by fibrinogen and other large asymmetrical plasma proteins, e.g., immunoglobulin M
June 2003 8-143 DRAFT-DO NOT QUOTE OR CITE
-------
1 and «2-macrogl°bulin. They note that, in a cohort study of elderly men and women, fibrinogen
2 levels were strongly related to inflammatory markers (e.g., neutrophil count and acute-phase
3 proteins, [C-reactive protein and 90th
24 percentile of pre-episode levels = 5.7 mg/L) tripled; and associated increases in TSP levels of
25 26 |ig/m3 (5-day averages) were associated with an odds ratio of 1.37 (95% CI 1.08-1.73) for
26 C-reactive protein levels exceeding the 90th percentile levels in two pollutant models that
27 included SO2 levels. The estimated odds ratio for a 30 |ig/m3 increase in the 5-day mean for SO2
28 was 1.12 (95% CI 0.92 = 1.47).
29 Two other recent studies also examined blood indices in relation to PM pollution (Seaton
30 et al., 1999; Prescott et al., 1999). Seaton and colleagues collected sequential blood samples
31 (up to 12) over an 18 month period in 112 subjects (all over age 60) in Belfast and Edinburgh,
June 2003 8-144 DRAFT-DO NOT QUOTE OR CITE
-------
1 UK. Blood samples were analyzed for hemoglobin, packed cell volumes, fibrinogen, blood
2 counts, factor VII, interleuken 6, and C-reactive protein. In a subset of 60 subjects, plasma
3 albumin also was measured. PM10 data monitored by TEOM were collected from ambient sites
4 in each city. Personal exposure estimates for three days preceding each blood draw were derived
5 from ambient data adjusted by time-activity patterns and I/O penetration factors.
6 No co-pollutants were analyzed. Data were analyzed by analysis of covariance, controlling for
7 city, seasons, temperature, and between-subject differences. Significant changes in several
8 blood indices were associated with either ambient or estimated personal PM10 levels. All
9 changes were negative, except for C reactive protein in relation to ambient PM10. Prescott et al.
10 (1999) also investigated factors that might increase susceptibility to PM exposure cardiovascular
11 events for a cohort of 1,592 subjects aged 55-74 in Edinburgh, UK, baseline measurements of
12 blood fibrinogen and blood and plasma viscosity were examined as modifiers of PM effects
13 (indexed by BS) on the incidence of fatal and non-fatal myocardial infarction or stroke. All
14 three blood indices were strong predictors of increased cardiac event risk; but there was no clear
15 evidence of either a main effect of BS, nor interactions between BS and blood indices.
16 Two more new studies examined air pollution associations with plasma fibrinogen. One by
17 Pekkanen and colleagues (2000) analyzed plasma fibrinogen data from a cross-sectional survey
18 of 4,982 male and 2,223 female office workers in relation to same-day and previous three-day
19 PM10, black smoke, NO2, CO, SO2, and O3 concentrations. In the full analysis, NO2 and CO
20 were significantly associated with fibrinogen levels. When the analysis was restricted to the
21 summer season, NO2 and CO, as well as PM10 and black smoke, showed significant univariate
22 associations. In another, Schwartz (2001) later reported not only significant associations
23 between PM10 exposures and plasma fibrinogen levels in a subset of the NHANES III cohort, but
24 also PM10 associations with platelet and white cell counts, the PM10 associations being robust
25 when O3, NO2, or SO2 were included. CO was not analyzed.
26 The above findings add support for intriguing hypotheses about possible mechanisms by
27 which PM exposure may be linked to adverse cardiac outcomes. They are interesting in
28 implicating both increased blood viscosity and C-reactive protein, a biological marker of
29 inflammatory responses thought to be predictive of increased risk for serious cardiac events.
30
31
June 2003 8-145 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3.1.4 Issues in the Interpretation of Acute Cardiovascular Effects Studies
2 Susceptible subpopulations. Because they lack extensive data on individual subject
3 characteristics, hospital admissions studies provide only limited information on susceptibility
4 factors based on stratified analyses. The relative effect sizes for PM-cardiovascular associations
5 (and respiratory) admissions reported in ecologic time-series studies are generally somewhat
6 higher than those for total admissions. This provides some limited support for hypothesizing
7 that acute PM effects operate via cardiopulmonary pathways or that persons with pre-existing
8 cardiopulmonary disease have greater susceptibility to PM, or both. Although there is some data
9 from ecologic time-series studies showing larger PM effects on cardiovascular admissions in
10 adults aged > 65 yr versus younger populations, the differences are neither striking nor
11 consistent. One recent study reported larger CVD hospitalization among persons with current
12 respiratory infections. The individual-level studies of cardiophysiologic function assessed above
13 generally suggest that elderly persons with pre-existing cardiopulmonary disease are susceptible
14 to subtle changes in heart rate variability in association with PM exposures. Because younger
15 and healthier populations have not yet been much studied, it is not yet possible to say whether
16 the elderly clearly have especially increased susceptibility.
17
18 Role of other environmental factors. The time-series studies published since 1996 have
19 all controlled adequately for weather influences. Thus, it is deemed unlikely that residual
20 confounding by weather accounts for the observed PM associations. With one possible
21 exception (Pope et al., 1999a), the roles of meteorological factors have not been analyzed
22 extensively as yet in the individual-level studies of cardiac function. Thus, the possibility of
23 confounding in such studies cannot yet be fully discounted. Co-pollutants have been analyzed
24 extensively in many recent time-series studies of PM and hospital admissions. In some studies,
25 PM clearly has an independent association after controlling for gaseous co-pollutants. In others,
26 the PM effects are reduced once co-pollutants are added to the model; but this may be in part due
27 to colinearity between PM10 and co-pollutants and/or gaseous pollutants (e.g., CO) having
28 independent effects on cardiovascular function.
29
June 2003 8-146 DRAFT-DO NOT QUOTE OR CITE
-------
1 Temporal patterns of responses following PM exposure. The evidence from recent time-
2 series studies of CVD admissions suggests rather strongly that PM effects tend to be maximal at
3 lag 0, with some carryover to lag 1, with little evidence for important effects beyond lag 1.
4
5 Relationship of CVD effects to PM size and chemical composition attributes. Insufficient
6 data exist from the time-series CVD admissions studies or the emerging individual-level studies
7 to provide clear guidance as to which ambient PM components, defined on the basis of size or
8 composition, determine ambient PM CVD effect potency. The epidemiologic studies have been
9 constrained by limited availability of multiple PM metrics. Where multiple metrics exist, they
10 often are highly correlated or are of differential quality due to differences in numbers of
11 monitoring sites and monitoring frequency.
12
13 PM effects on blood characteristics related to CVD events. Interesting, though limited,
14 new evidence has also been derived which is highly suggestive of associations between ambient
15 PM and increased blood viscosity, increased serum C-reactive protein, and fibrinogen (both
16 related to increased risks of serious cardiac events). The biologic plausibility of these findings is
17 supported by a study showing that ultrafine particles are rapidly distributed into the systemic
18 circulation following inhalation exposure (Nemmar et al., 2002).
19
20 8.3.2 Effects of Short-Term Particulate Matter Exposure on the Incidence of
21 Respiratory-Related Hospital Admissions and Medical Visits
22 8.3.2.1 Introduction
23 Although hospital admissions represent one severe morbidity measure evaluated in regard
24 to PM exposure, hospital emergency department (ED) visits are a notable related outcome.
25 Doctors' visits also represent another related health measure that, although less studied, is still
26 very relevant to assessing air pollution public health impacts. This category of pollution-
27 affected persons can represent a large population, yet one largely unevaluated due to the usual
28 lack of centralized data records for doctors' visits in the United States.
29 This section evaluates information on epidemiologic associations of ambient PM exposure
30 with both respiratory hospital admissions and medical visits. It intercompares various studies
31 examining size-related PM mass exposure measures (e.g., for PM10, PM2 5, etc.) or various PM
32 chemical components vis-a-vis their associations with such health endpoints, and discusses their
June 2003 8-147 DRAFT-DO NOT QUOTE OR CITE
-------
1 respective extents of coherence with PM associations across related health effects measures.
2 In the following discussion, the main focus for quantitative intercomparisons is on studies
3 considering PM metrics that measure mass or a specific mass constituent, i.e., PM10, PM10_2 5,
4 PM2 5, or sulfates (SO4"2). Study results for other related PM metrics (e.g., BS) are also
5 considered, but only qualitatively, primarily with respect to their relative coherence with studies
6 using mass or composition metrics measured in North America. In order to consider potentially
7 confounding effects of other co-existing pollutants, study results for various PM metrics are
8 presented both for (1) when the PM metric is the only pollutant in the model and (2) the case
9 where a second pollutant (e.g., O3) is also included. Results from models with more than two
10 pollutants included simultaneously, however, are not used for quantitative estimates of effect
11 size or statistical strength, because of increased likelihood of bias and variance inflation due to
12 multi-collinearity of various pollutants (e.g., see Harris, 1975).
13
14 8.3.2.2 Summary of Key Respiratory Hospital Admissions Findings from the 1996
15 Particulate Matter Air Quality Criteria Document
16 In the 1996 PM AQCD, both COPD and pneumonia hospitalization studies were found to
17 show moderate, but statistically significant, relative risks in the range of 1.06 to 1.25 (or 6 to
18 25% excess risk increment) per 50 |ig/m3 PM10 increase or its equivalent. Whereas many
19 hospitalizations for respiratory illnesses occur in those > 65 years of age, there were also
20 increased hospitalizations for those < 65 years of age. Several hospitalization studies restricted
21 their analysis by age group, but did not explicitly examine younger age groups. One exception
22 noted was Pope (1991), who reported increased hospitalization for Utah Valley children (0 to
23 5 yrs) for monthly numbers of admissions in relation to PM10 monthly averages, as opposed to
24 daily admissions in relation to daily PM levels used in other studies. Studies examining acute
25 associations between indicators of components of fine particles (e.g., BS; sulfates, SO4=; and
26 acidic aerosols, H+) and hospital admissions were reported, too, as showing significant
27 relationships. While sulfates were especially predictive of respiratory health effects, it was not
28 clear whether the sulfate-related effects were attributable to their acidity, to the broader effects
29 of associated combustion-related fine particles, or to other factors.
30
31
June 2003 8-148 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3.2.3 New Respiratory-Related Hospital Admissions Studies
2 New studies appearing since the 1996 PM AQCD have examined various admissions
3 categories, including: total respiratory admissions for all ages and by age; asthma for all ages
4 and by age; chronic obstructive pulmonary disease (COPD) admissions (usually for patients
5 > 64 yrs.), and pneumonia admissions (for patients > 64 yrs.). Table 8B-2, Appendix 8B
6 summarizes salient details regarding the study area, study period, study population, PM indices
7 considered and their concentrations, methods employed, study results, and "bottom-line" PM
8 index percent excess risks per standard PM increment (e.g., 50 |ig/m3 for PM10) for the newer
9 studies.
10 The percent excess risk (ER) estimates presented in Table 8B-2 are based upon the relative
11 risks (RR's) provided by the authors, but converted into percent increments per standardized
12 increments used by the U.S. EPA to facilitate direct intercomparisons of results across studies
13 (as discussed in Section 8.1). The ER's shown in the table are for the most positively significant
14 pollutant coefficient; and the maximum lag model is used to provide estimates of potential
15 pollutant-health effects associations.
16 Based on information from Dominici et al. (2002) indicating that the default convergence
17 criteria used in the S-Plus function GAM may not guarantee convergence to the best unbiased
18 estimate (as discussed earlier), only those studies that used other statistical algorithms or which
19 have reported reanalyzed S-Plus GAM results are assessed in the text below. However, given
20 the modest effects of this reanalysis on most study results (i.e., while effect estimates are
21 modified somewhat, the study conclusions remain largely unchanged), Table 8B-2 includes all
22 studies and notes those that originally used the S-Plus GAM algorithm, as well as which of those
23 studies have since been reanalyzed with more appropriate methods.
24 Of most pertinence here are those newly available studies that evaluate associations
25 between one or another ambient PM metric and respiratory hospital admissions in U.S. or
26 Canadian cities, as for PM10 mass concentrations are summarized in Table 8-17.
27 Among numerous new epidemiologic studies of PM10 morbidity, many evaluated relatively
28 high PM10 levels. However, some did evaluate associations with PM10 concentrations ranging to
29 rather low levels. Of note is the fact that associations have been reported by several
30 investigators between acute PM10 exposures and total respiratory-related hospital admissions for
31 numerous U.S. cities with annual mean PM10 concentrations extending to below 50 |ig/m3.
June 2003 8-149 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-17. SUMMARY OF UNITED STATES PM10 RESPIRATORY-RELATED
HOSPITAL ADMISSION STUDIES
Outcome Mean Levels
Reference Measures (ug/m3)
Schwartz et al. Respiratory PM10 = 43
(1996b)
Samet et al. COPD PM10 = 33
(2000a,b)*
Reanalysis by Zanobetti and
Schwartz (2003b)
Lippmann et al. COPD PM10 = 3 1
(2000)*
Reanalysis by Ito (2003)
Moolgavkar COPD PM10 = 35,
(2000c)* (> 64 yrs) Chicago
(median) PM10 = 44, LA
PM10 = 41,
Phoenix
PM10 = 44, LA
Reanalysis by COPD Chicago
Moolgavkar (> 64 yrs)
(2003)
Reanalysis by COPD Los Angeles
Moolgavkar (all ages)
(2003)
Samet et al. Pneumonia PM10 = 33
(2000a,b)*
Reanalysis by Zanobetti and
Schwartz (2003b)
Lippmann et al. Pneumonia PM10 = 31
(2000)
Reanalysis by Pneumonia
Ito (2003)
Jacobs etal. (1997) Asthma PM10 = 34
Nauenberg and Asthma PM10 = 45
Basu(1999)
Tolbert et al. Asthma PM10 = 39
(2000b)
Sheppard et al. Asthma PM10 = 31
(1999)*
Reanalysis by Sheppard
(2003)
Co-Pollutants Day
Measured Lag
S03 —
SO2'O3'NO2'CO 0
1
0-1
0-1
0-1
0-1
S02> 03' N02- 3
CO'H+ 3
3
— 0
— 2
— 0
CO 2
0
2
2
2
S02, 03, N02, 0
CO 1
0-1
0-1
0-1
0-1
S02, 03, N02, 1
CO, H+ 1
1
1
1
03, CO —
03 0
O3, NOX 1
CO, O3, SO2 1
Method
Poisson GLM
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
PS GLM
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
Default GAM: 30df
Default GAM: 30df
Default GAM: 30df
Default GAM: 30df
Strict GAM: lOOdf
Strict GAM: 30df
Strict GAM: lOOdf
NSGLM: lOOdf
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
PS GLM
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
Poisson GLM
Poisson GLM
GEE
Default GAM
NSGLM
Strict GAM
Effect Estimate (95% CL)
(% increase per 50 ug/m3)
5.8(0.5, 11.4)
7.4(5.1,9.8)
7.5(5.3,9.8)
9.4 (5.9, 12.9)
8.8(4.8, 13.0)
6.8(2.8, 10.8)
8.0(4.3, 11.9)
No Co Poll: 9.6 (-5.3,
Co Poll: 1.0 (-15, 20)
No Co Poll: 9.6 (-5.3,
No Co Poll: 6.5 (-7.8,
No Co Poll: 4.6 (-9.4,
2.4 (-0.2, 5.11)
6.1(1.1,11.3)
6.9 (-4.1, 19.3)
0.6 (-5. 1,6.7)
(two poll, model)
3.24 (.031, 6.24)
7.78(4.32-10.51)
5.52(2.53-8.59)
5.00(1.22,8.91)
8.1(6.5,9.7)
6.7(5.3,8.2)
9.9 (7.4, 12.4)
8.8(5.9, 11.8)
2.9 (0.2, 5.6)
6.3 (2.5, 10.3)
No Co Poll: 21.4(8.2,
Co Poll: 24(8.2,43)
No Co Poll: 21.5(8.3,
No Co-Poll: 18.1 (5.3,
No Co-Poll: 18.6(5.6,
6.11 (CI not reported)
16.2 (2.0, 30)
13.2(1.2,26.7)
13.2(5.5,22.6)
10.9(2.8, 19.6)
8.1 (0.1, 16.7)
26.8)
26.8)
23.0)
20.8)
36.3)
36)
32.5)
33.1
NS = Natural Spline General Linear Model; PS = Penalized Spline General Additive Model
June 2003
8-150
DRAFT-DO NOT QUOTE OR CITE
-------
1 On this account, the results of the NMMAPS multi-city study (Samet et al., 2000a,b) of PM10
2 levels and hospital admissions by persons > 65 in 14 U.S. cities are of particular interest.
3 As noted in Table 8-18, this study indicates PM10 effects similar to other cities, but with
4 narrower confidence bands, due to its greater power derived by combining multiple cities in the
5 same analysis. This allows significant associations to be identified, despite the fact that many of
6 the cities considered have relatively small populations and that each had mean PM10 below
7 50 |ig/m3. The cities considered and their respective annual mean/daily maximum PM10
8 concentrations (in |ig/m3) are Birmingham (34.8/124.8); Boulder (24.4/125.0); Canton
9 (28.4/94.8); Chicago (36.4/144.7); Colorado Springs (26.9/147.2); Detroit (36.8/133.6);
10 Minneapolis/St Paul (36.8/133.6); Nashville (31.6/128.0); New Haven (29.3/95.4); Pittsburgh
11 (36.0/139.3); Provo/Orem (38.9/241.0); Seattle (31.0/145.9); Spokane (45.3/605.8); and
12 Youngstown (33.1/104.0).
13 Table 8-18 also shows results of reanalyzing a number of the models considered in original
14 research with the use of models using more stringent convergence requirements than the original
15 default option. These results show that the effect estimates decline somewhat, but that the basic
16 direction of effect and conclusions about the significance of the PM effect on hospital
17 admissions remained unchanged.
18 Zanobetti and Schwartz (2003b), in their reanalyses, also considered spline models that are
19 thought to better estimate confidence intervals around pollutant effect estimates than the original
20 GAM analyses. With the spline models, confidence intervals usually increased over the original
21 GAM model and the coefficients also decreased somewhat (similar to GAM with more stringent
22 convergence criteria). As for possible co-pollutant confounding, it was reported that "In our
23 previous studies we did not find confounding due to other pollutants. These results are
24 confirmed in this reanalysis by the meta-regression analyses." Overall, the authors concluded
25 that "the general result is that the association of PM10 with hospital admissions remains and in
26 most cases is little changed."
27 Janssen et al. (2002) did further analyses for the Samet et al. (2000a,b) 14-city data set
28 examining associations for variable prevalence in air-conditioning (AC) and/or contributions of
29 different sources to total PM10. For COPD and pneumonia, the associations were less
30 significant, but the pattern of association was similar to that for CVD. The Zanobetti and
June 2003 8-151 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-18. PERCENT INCREASE IN HOSPITAL ADMISSIONS PER lO-jig/m3
INCREASE IN PM10 IN 14 U.S. CITIES (ORIGINAL AND REANALYZED RESULTS)
Constrained lag models
(Fixed Effect Estimates)
Original One day mean
(lagO)
Original Previous day mean
Original Two day mean
(for lag 0 and 1)
Reanalyzed Two day
mean (for lag 0 and 1)
Original PM10 < 50 ug/m3
(two day mean)
Reanalyzed PM10
< 50 ug/m3 (two day
mean)
Original Quadratic
distributed lag
Reanalyzed Quadratic
distributed lag
%
Increase
1.07
0.68
1.17
0.99
1.47
1.32
1.18
1.09
CVD
(95% CI)
(0.93, 1.22)
(0.54, 0.81)
(1.01,1.33)
(0.79,1.19)
(1.18, 1.76)
(0.77, 1.87)
(0.96, 1.39)
(0.81, 1.38)
%
Increase
1.44
1.46
1.98
1.71
2.63
2.21
2.49
2.53
COPD
(95% CI)
(1.00, 1.89)
(1.03, 1.88)
(1.49, 2.47)
(0.95, 2.48)
(1.71,3.55)
(1.02,3.41)
(1.78, 3.20)
(1.20, 3.88)
%
Increase
1.57
1.31
1.98
1.98
2.84
1.06
1.68
1.47
Pneumonia
(95% CI)
(1.27, 1.87)
(1.03, 1.58)
(1.65,2.31)
(1.65,2.31)
(2.21, 3.48)
(0.06, 2.07)
(1.25,2.11)
(0.86, 2.09)
Unconstrained distributed lag
Fixed effects estimate
Original Random effects
estimate
Reanalyzed Random
effects estimate
1.19
1.07
1.12
(0.97, 1.41)
(0.67, 1.46)
(0.84, 1.40)
2.45
2.88
2.53
(1.75,3.17)
(0.19,5.64)
(1.21,3.87)
1.90
2.07
2.07
(1.46, 2.34)
(0.94, 3.22)
(0.94, 3.22)
Source: Samet et al. (2000a,b) and Zanobetti and Schwartz (2003b) reanalyses.
1 Schwartz (2003b) reanalyses also examined these results, and they stated that "We still found a
2 decreased PM10 effect with increasing percentage of home with central AC."
3 Moolgavkar (2003) also reanalyzed his earlier GAM analyses of hospital admissions for
4 chronic obstructive pulmonary disease (Moolgavkar, 2000c) Los Angeles (Los Angeles County)
5 and Chicago (Cook County). In his original publication, Moolgavkar found ca. 5.0% excess risk
6 for COPD hospital admissions among the elderly (64+ yr) in Los Angeles to be significantly
7 related to both PM2 5 and PM10_2 5 in one pollutant models; but the magnitudes of the risk
8 estimates dropped by more than half to non-statistically significant levels in two-pollutant
9 models including CO. However, unlike the meta-regression approach to the multiple pollutant
June 2003 8-152 DRAFT-DO NOT QUOTE OR CITE
-------
1 issue used by Zanobetti and Schwartz (2003b), simultaneous regression of moderately to highly
2 correlated pollutants can lead to biased pollutant coefficients and commonly results in
3 diminished effect estimates for some or all of the pollutants considered. In the same study,
4 similar magnitudes of excess risk (i.e., in the range of ca. 4 to 7%) were found in one-pollutant
5 models to be associated with PM2 5 or PM10_25 for other age groups (0-19 yr; 20-64 yr) in Los
6 Angeles, as well.
7 In his reanalyses of these GAM results using the more stringent convergence criteria,
8 Moolgavkar (2003) combined all three Los Angeles age groups into one analysis, providing
9 greater power, but also complicating before/after comparisons as to the actual effect of using the
10 more stringent convergence criteria on the results. In the Cook County analyses, the author
11 changed other model parameters (i.e., the number of degrees of freedom in the model smooths)
12 at the same time as implementing more stringent convergence criteria; so direct before/after
13 comparisons are not possible for Moolgavkar's (2003) Chicago analyses. Moolgavkar noted that
14 "changes in the convergence criteria and the use of GLM instead of GAM can, but does not
15 always, have substantial impact on the results of the analyses and their interpretation." He also
16 concluded: "Given that different analytic strategies can make substantial differences to the
17 estimates of effects of individual pollutants I do not believe that these numerical estimates are
18 too meaningful. Patterns of association appear to be robust, however. For example, in Los
19 Angeles, with the exception of COPD admissions with which NO2 appears to show the most
20 robust association, it is clear that CO is the best single index of air pollution associations with
21 health end points, far better than the mass concentration of either PM10 or of PM2 5. In Cook
22 County the results are not so clear-cut, however, any one of the gases is at least as good an index
23 of air pollution effects on human health as is PM10."
24 Tolbert et al. (2000b) used generalized estimating equations (GEE), logistic regression, and
25 Baysian models to evaluate associations between emergency department visits for asthma (by
26 those < 17 yrs old) in Atlanta during the summers of 1993 - 1995 (~ 6000 visits for asthma out
27 of- 130,000 total visits) and several air pollution variables (PM10, O3, total oxides of nitrogen).
28 Logistic regression models controlling for temporal and demographic variables gave statistically
29 significant (p < 0.05) lag 1 day relative risk estimates of 1.04 per 15 |ig/m3 24-h PM10 increment
30 and 1.04 per 20 ppb increase in maximum 8-h O3 levels. In multipollutant models including
31 both PM10 and O3, the terms for each became non-significant due to high collinearity of the two
June 2003 8-153 DRAFT-DO NOT QUOTE OR CITE
-------
1 variables (r2 = 0.75). The authors interpreted their findings as suggesting positive associations
2 between pediatric asthma visits and both PM10 and O3. The PM10 effects appeared to be stronger
3 for concentrations > 20 |ig/m3 than below that 24-h value.
4 Other U.S. studies finding associations of respiratory-related hospital admissions or
5 medical visits with PM10 levels extending below 50 |ig/m3 include: Schwartz (1994) in
6 Minneapolis-St. Paul, Minnesota; Schwartz et al. (1996b) in Cleveland; Sheppard et al. (1999)
7 in Seattle; Linn et al. (2000) in Los Angeles; and Nauenberg and Basu (1999) in Los Angeles;
8 in Minneapolis-St. Paul, MN, but not in Birmingham, AL. The excess risk estimates most
9 consistently fall in the range of 5 to 25% per 50 |ig/m3 PM10 increment, with those for asthma
10 visits and hospital admissions often being higher than those for COPD and pneumonia
11 admissions.
12 Similar associations between increased respiratory related hospital admissions/medical
13 visits and low short-term PM10 levels were also reported by various investigators for several
14 non-U.S. cities. Wordley et al. (1997), for example, reported positive and significant
15 associations between PM10 (mean = 25.6 |ig/m3, max. = 131 |ig/m3) and respiratory admissions
16 in Birmingham, UK using multivariate linear regression methods; and Atkinson et al. (1999b),
17 using Poisson modeling, reported significant increases in hospital admissions for respiratory
18 disease to be associated with PM10 (mean = 28.5 |ig/m3) in London, UK. Hagen et al. (2000) and
19 Prescott et al. (1998) also found positive but non-significant associations of hospital admissions
20 and, PM10 levels in Drammen, Norway (mean = 16.8 |ig/m3) and Edinburgh, Scotland (mean =
21 20.7 |ig/m3). Admissions in Drammen considered relatively small populations, limiting
22 statistical power in this study. Petroeschevsky et al. (2001) examined associations between
23 outdoor air pollution and hospital admissions in Brisbane, Australia during 1987-1994 using a
24 light scattering index (BSP) for fine PM. The levels of PM are quite low in this city, relative to
25 most U.S. cities, but BSP was positively and significantly associated with total respiratory
26 admissions, but not for asthma.
27 If day-to-day increases in air pollution cause increases in hospital admissions, as shown by
28 time-series studies, then short-term removal of pollution should lower admissions. It is rarely
29 possible to test this hypothesis by examining a situation when pollution sources are abruptly
30 "turned off and then "turned on" again. One such opportunity did arise when a steel mill strike
31 resulted in concomitant reductions in both PM and respiratory admissions that were experienced
June 2003 8-154 DRAFT-DO NOT QUOTE OR CITE
-------
1 in Utah Valley, but not in surrounding valleys without the steel mill, as documented by Pope
2 (1991). A perhaps more broadly relevant case where this hypothesis was similarly tested was a
3 study of air quality improvements during the Atlanta Summer Olympics of 1996 (Friedman
4 et al., 2001). Potential associations between air quality improvements and changes in children's
5 hospital admissions, while weather and other "natural" influences on admissions remained
6 unchanged from normal, were evaluated by Friedman et al. Interestingly, compared to a baseline
7 period, traffic related pollution declined, as did PM10 levels by 16% and O3 by 28% as a result of
8 the alternative mass transportation strategy used to reduce road traffic during the Games. At the
9 same time, SO2, not related to traffic, actually increased during the Games. Both PM and O3
10 concentrations also rose noticeably after the Olympics. A significant reduction in asthma events
11 was associated with O3 concentrations, but the PM10 association was not statistically significant.
12 While the high correlation between PM and O3 limit the ability to determine which pollutant
13 may account for the reduction in asthma events, this study supports the hypothesis that
14 reductions of acute air pollution can provide immediate health improvements.
15
16 8.3.2.3.1 Particulate Matter Mass Fractions and Composition Comparisons
17 While PM10 mass has generally been the metric most often used as the particle pollution
18 index in the U.S. and Canada, some new studies have examined the relative roles of various
19 PM10 mass fractions (e.g., PM25 and PM10_25) and chemical constituents (such as SO4"2)
20 contributing to PM-respiratory hospital admissions associations. Several new studies (from
21 among those summarized in Tables 8-19 and 8-20, respectively) report significant associations
22 of increased respiratory-cause medical visits and/or hospital admissions with ambient PM25
23 and/or PM10_2 5 ranging to quite low concentrations. These include the Lippmann et al. (2000)
24 study in Detroit, where all PM metrics (PM10, PM2 5, PM10_2 5, FT) were positively related to
25 pneumonia and COPD admissions among the elderly (aged 65+ yr) in single pollutant models,
26 with their RR values for pneumonia generally remaining little changed (but with broader
27 confidence intervals) in multipollutant models including one or more gaseous pollutant (e.g.,
28 CO, O3, NO2, SO2). However, for COPD admissions, the effect estimates were reduced and
29 became non-significant in multipollutant models including gaseous copollutants. Excess risks
30 for pneumonia admissions in the one pollutant model using default GAM were 13% (3.7, 22)
June 2003 8-155 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-19. SUMMARY OF UNITED STATES PM25 RESPIRATORY-RELATED
HOSPITAL ADMISSION STUDIES
Reference
Lippmannetal.
(2000)
Reanalysis by
Ito (2003)
Moolgavkar
(2000c)*
Reanalysis by
Moolgavkar
(2003)
Lippmannetal.
(2000)
Reanalysis by
Ito (2003)
Sheppardetal.
(1999)*
Reanalysis by
Sheppard (2003)
Freidman et al.
(2001)
Outcome Mean Levels
Measures ug/m3
COPD PM25=18
COPD
COPD PM25=22, LA
(> 64 yrs) PM25 = 22, LA
(median)
COPD
(all ages)
Pneumonia PM25=18
Pneumonia
Asthma PM25 = 16.7
Asthma PM25 = (36.7-
30.8 decrease)
Co-Pollutants
Measured Lag
S02, 03, N02, 3
CO, H+ 3
— 2
CO 2
2
2
2
S02, 03, N02, 1
CO, FT 1
CO, 03, S02 1
CO
O3 3d.
cum
Method
Default GAM
Default GAM
Default GAM Strict
GAM
NSGLM
Default GAM
Default GAM
Strict GAM: 30df
Strict GAM: lOOdf
NSGLM: lOOdf
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
Default GAM
Default GAM
Strict GAM
NSGLM
Strict GAM
NSGLM
Poisson GEE
Effect Estimate (95% CL)
(% increase per 25 ug/m3)
No Co Poll: 5.5 (-4.7, 16.8)
Co Poll: 2. 8 (-9.2, 16)
No Co Poll: 5.5 (-4.7, 16.8)
No Co Poll: 3.0(-6.9, 13.9)
No Co Poll: 0.3(-9.3, 10.9)
5.1 (0.9, 9.4)
2.0 (-2.9, 7.1)
Two poll, model
4.69 (2.06, 7.38)
2.87 (0.53, 5.27)
2.59 (-0.29, 5.56)
No Co-Poll: 12.5(3.7,22.1)
Co Poll: 12(1.7,23)
No Co-Poll: 12.5(3.7,22.1)
No Co-Poll: 10.5(1.8, 19.8)
No Co-Poll: 10.1 (1.5, 19.5)
8.7(3.3, 14.3)
No Co-Poll: 8.7 (3.3, 14.3)
No Co-Poll: 8.7(3.2,14.4)
No Co-Poll: 6.5(1.1,12.0)
With Co-poll: 6.5(2.1, 10.9)
With Co-poll: 6.5(2.1, 10.9)
1.4(0.80-2.48)
NS = Natural Spline General Linear Model; PS = Penalized Spline General Additive Model.
1
2
3
4
5
and 12% (-0.6, 24) per 25 |ig/m3 of PM25 and PM10_25, respectively; those for COPD admissions
were 5.5% (-4.7, 17) and 9.3% (-4.2, 25) per 25 |ig/m3 PM25 and PM10.25, respectively.
Lippmann et al. (2000) reported weaker associations with sulfate and acidic components of
PM2 5 than with PM2 5 mass overall, but the acidity levels during this study were very low, being
below detection on most study days. In contrast, past studies of sulfates and aerosol acidity
associations with respiratory hospital admissions have found stronger sulfate associations when
the acidity of those aerosols was higher (e.g., Thurston et al, 1994). As noted by Lippman et al.
June 2003
8-156
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-20. SUMMARY OF UNITED STATES PM1025 RESPIRATORY-RELATED
HOSPITAL ADMISSION STUDIES
Outcome
Reference Measures
Moolgavkar COPD
(2000c )*
Lippmann etal. COPD
(2000)*
Reanalysis by Ito (2003)
Lippmann etal. Pneumonia
(2000)*
Reanalysis by Ito (2003)
Mean Levels Co-Pollutants
ug/m3 Measured Lag
— 3
PM10.25 = 12 SO2, 03,NO2, 3
CO, H+ 3
PM10.25 = 12 S02, 03,N02, 1
CO, H+ 1
1
1
1
Method
Default GAM
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
Default GAM
Default GAM
Default GAM
Strict GAM
NSGLM
Effect Estimates (95% CL)
(% increase per 25 ug/m3)
5.1% (-0.4, 10.9)
No Co-Poll: 9.3 (-4.2, 24.7)
Co-Poll: 0.3 (-14, 18)
No Co-Poll: 9.3 (-4.2, 24.7)
No Co-Poll: 8.7 (-4.8, 24.0)
No Co-Poll: 10. 8 (-3. 1,26.5)
No Co-Poll: 11. 9 (-0.6, 24.4)
Co-Poll: 13.9 (0.0, 29.6)
No Co-Poll: 11. 9 (-0.6, 24.4)
No Co-Poll: 9.9 (-0.1, 22.0)
No Co-Poll: 11. 2 (-0.02, 23.6)
Sheppardetal.
(1999)*
Asthma
PM10.25=16.2 CO, 03, S02
Reanalysis by Sheppard
(2003)
Default GAM
Strict GAM
NSGLM
11.1 (2.8,20.1)
5.5 (-2.7 11.1)
5.5 (0, 14.0)
NS = Natural Spline General Linear Model; PS = Penalized Spline General Additive Model.
1
2
3
4
5
9
10
11
12
13
(2000), "a notable difference between the data of Thurston and colleagues from Toronto and our
data is the H+ levels: the H+ levels in Toronto were 21.4, 12.6, and 52.3 nmol/m3 for the
summers of 1986, 1987, and 1988, respectively, whereas in our study, the H+ level averaged only
8.8 nmol/m3." Thus, these results are consistent with past studies and biological plausibility, in
that sulfates and its associated PM should be less toxic when in a less strongly acidic form, as
indeed found in this study.
In order to evaluate the potential influence of the Generalized Additive Model (GAM)
convergence specification on the results of the original Detroit data analysis, Ito (2003)
re-examined associations between PM components and daily mortality/morbidity by using more
stringent GAM convergence criteria, and by applying a Generalized Linear Models (GLM) that
approximated the original GAM models. The reanalysis of GAM Poisson models used more
stringent convergence criteria, as suggested by Dominici et al. (2002): the convergence precision
(epsilon) was set to 10-14 and maximum iteration was set to 1000, for both the local scoring and
June 2003
8-157
DRAFT-DO NOT QUOTE OR CITE
-------
1 back-fitting algorithms. The GLM model specification approximated the original GAM models.
2 Natural splines were used for smoothing terms. To model time trend, the same degrees of
3 freedom as the smoothing splines in the GAM models were used, with the default placement of
4 knots. For weather models, to approximate LOESS smoothing with a span of 0.5 in the GAM
5 model, natural splines with degrees of freedom were used. Generally, the GAM models with
6 stringent convergence criteria and GLM models resulted in somewhat smaller estimated relative
7 risks than those reported in the original study, e.g., for respiratory admissions in Table 8-21.
8 It was found that the reductions in the estimated relative risks were not differential across the
9 PM indices. Thus, conclusions of the original study about the relative roles of PM components
10 by size and chemical characteristics remained unaffected.
11
12
TABLE 8-21. INTERCOM?ARISON OF DETROIT PNEUMONIA HOSPITAL
ADMISSION RELATIVE RISKS (± 95% CI below) OF PM INDICES (per 5th-to-95th
percentile pollutant increment) FOR VARIOUS MODEL SPECIFICATIONS.*
Original GAM (default) GAM (stringent)
PM25 (1)
PM10.2, (1)
PM10 (1)
H+ (3)
S04= (1)
1.185
(1.053,
1.114
(1.006,
1.219
(1.084,
1.060
(1.005,
1.156
(1.050,
1.332)
1.233)
1.372)
1.118)
1.273)
1.154
(1.027,
1.095
(0.990,
1.185
(1.054,
1.049
(0.994,
1.128
(1.025,
1.298)
1.211)
1.332)
1.107)
1.242)
GLM
1.149
(1.022,
1.107
(1.00, 1
1.190
(1.057,
1.049
(0.994,
1.123
(1.020,
1.292)
.226)
1.338)
1.107)
1.235)
*The selected lag is indicated in parenthesis next to the pollutant name.
Source: Ito (2003).
1 Lumley and Heagerty (1999) illustrate the effect of reliable variance estimation on data
2 from hospital admissions for respiratory disease on King County, WA for eight years (1987-94),
3 together with air pollution and weather information, using estimating equations and weighted
4 empirical variance estimators. However, their weather controls were relatively crude (i.e.,
June 2003 8-158 DRAFT-DO NOT QUOTE OR CITE
-------
1 seasonal dummy variables and linear temperature terms). This study is notable for having
2 compared sub-micron PM (PMLO) versus coarse PM1(M 0 and for finding significant hospital
3 admission associations only with PMX 0. This may suggest that the PM2 5 versus PM10 separation
4 may not always be sufficient to differentiate submicron fine particle versus coarse-particle
5 toxicities.
6 Asthma hospital admission studies in various U.S. communities provide additional
7 important new data. Of particular note is a study by Sheppard et al. (1999) which evaluated
8 relationships between measured ambient pollutants (PM10, PM2 5, PM10_2 5, SO2, O3, and CO) and
9 non-elderly adult (< 65 years of age) hospital admissions for asthma in Seattle, WA. PM and
10 CO were found to be jointly associated with asthma admissions. An estimated 4 to 5% increase
11 in the rate of asthma hospital admissions (lagged 1 day) was reported to be associated with
12 interquartile range changes in PM indices (19 |ig/m3 for PM10, 11.8 |ig/m3 for PM25, and
13 9.3 |ig/m3 for PM10_25), equivalent to excess risk rates as follows: 13% (CI = 05-23) per
14 50 |ig/m3 for PM10; 9% (CI = 3-14) per 25 |ig/m3 PM2 5; 11% (CI = 3-20) per 25 |ig/m3 PM10.2 5.
15 Also of note for the same region by the same research team using similar methods is the Norris
16 et al. (1999) study showing associations of low levels of PM25 (mean = 12 |ig/m3) with markedly
17 increased asthma ED, i.e., excess risk = 44.5% (CI = 21.1-11 A) per 25 |ig/m3 PM25.
18 Sheppard (2003) recently conducted a reanalysis of their nonelderly hospital admissions
19 data for asthma in Seattle, WA, to evaluate the effect of the fitting procedure on their previously
20 published analyses. As shown in Figure 8-13, the effect estimates were slightly smaller when
21 more stringent convergence criteria were used with GAM, and there was an additional small
22 reduction in the estimates when GLM with natural splines were used instead. Confidence
23 intervals were slightly wider for the GLM model fit. Sheppard concluded that, "Overall the
24 results did not change meaningfully. There were small reductions in estimates using the
25 alternate fitting procedures. I also found that the effect of single imputation (i.e., not adjusting
26 for replacing missing exposure data with an estimate of its expected value) was to bias the effect
27 estimates slightly upward. In this data set this bias is of the same order as the bias from using
28 too liberal convergence criteria in the generalized additive model."
29 Moolgavkar (2003) also conducted reanalyses of respiratory-related hospital admissions,
30 but for COPD data for all ages in Los Angeles. Using GAM with strict convergence criteria and
31 30 degrees of freedom (df), an excess risk estimate of 4.7% (CI = 2.1 - 7.4) was obtained per
June 2003 8-159 DRAFT-DO NOT QUOTE OR CITE
-------
0)
tfl
0)
0 0_
C CM
a:
Q.
+j
re o
a: T--
c
0
en
E
T3
C i— i
0>
O
C
re
O
CD
DJ
« o
c V
(U
0
0)
CL
1
*
^
1
PM10
Lag 1
4
»,
^
j
PM2.5
Lag 1
t,
*
j
Coarse
Mass
Lag 1
t
4
i
i
i
i
i
i
i
No or Multiple imputation
.......... single Imputation
»
Carbon
Monoxide
Lag 3
• ,
•
Sulfur
Dioxide
Lag 0
4
,
>
Ozone
Lag 2
Figure 8-13. Percent change in hospital admission rates and 95% CIs for an IQR increase
in pollutants from single-pollutant models for asthma. Poisson regression
models are adjusted for time trends (64-df spline), day-of-week, and
temperature (4-df spline). The IQR for each pollutant equals: 19 ug/m3 for
PM10,11.8 ug/m3 for PM25, 9.3 ug/m3 for coarse PM, 20 ppb for O3, 4.9 ppb
for SO2, and 924 ppb for CO. Triplets of estimates for each pollutant are for
the original GAM analysis using smoothing splines, the revised GAM
analysis with stricter convergence criteria, and the GLM analysis with
natural splines. For pollutants that required imputation (i.e., estimation of
missing value) estimates ignoring (single imputation) or adjusting for
(multiple imputation) the imputation are shown.
Source: Sheppard (2003).
1
2
3
25 |ig/m3 PM25 increment. The notable effect of increasing degrees of freedom on modeling
results is well illustrated by the excess risk estimate dropping to 2.9% (CI = 0.5 - 5.3) with strict
GAM and 100 df or 2.6% (CI = -0.3, 5.6) with NS GLM 100 df.
June 2003
8-160
DRAFT-DO NOT QUOTE OR CITE
-------
1 Burnett et al. (1997a) evaluated the role that the ambient air pollution mix, comprised of
2 gaseous pollutants and PM indexed by various physical and chemical measures, plays in
3 exacerbating daily admissions to hospitals for cardiac diseases and for respiratory diseases
4 (tracheobronchitis, chronic obstructive lung disease, asthma, and pneumonia). They employed
5 daily measures of PM25 and PM10_2 5, aerosol chemistry (sulfates and H+), and gaseous pollutants
6 (O3, NO2, SO2, CO) collected in Toronto, Ontario, Canada, during the summers of 1992, 1993,
7 and 1994. Positive associations were observed for all ambient air pollutants for both respiratory
8 and cardiac diseases. Ozone was the most consistently significant pollutant and least sensitive to
9 adjustment for other gaseous and particulate measures. The PM associations with respiratory
10 hospital admissions were significant for: PM10 (RR = 1.11 for 50 |ig/m3; CI = 1.05-1.17); PM2 5
11 (fine) mass (RR = 1.09 for 25 |ig/m3; CI = 1.03-1.14);PM10.2 5 (coarse) mass (RR = 1.13 for
12 25 |ig/m3;CI =1.05-1.20); sulfate levels (RR= 1.11 for 155 nmoles/m3 = 15 |ig/m3; CI = 1.06-
13 1.17);andH+(RR= 1.40 for 75 nmoles/m3 = 3.6 |ig/m3, as H2SO4; CI = 1.15-1.70). After
14 inclusion of O3 in the model, the associations with the respiratory hospital admissions remained
15 significant for: PM10(RR= 1.10, CI = 1.04-1.16); fine mass (RR= 1.06; CI = 1.01-1.12); coarse
16 mass(RR= 1.11; CI= 1.04-1.19); sulfate levels (RR = 1.06; CI = 1.0-1.12); andH+(RR= 1.25;
17 CI = 1.03-1.53), using the same increments. Of the PM metrics considered here, H+ yielded the
18 highest RR estimate. Regression models that included all recorded pollutant simultaneously
19 (with high intercorrelations among the pollutants) were also presented.
20 There have also been numerous new time-series studies examining associations between
21 air pollution and respiratory-related hospital admissions in Europe, as summarized in Appendix
22 8B, Table 8B-2, but most of these studies relied primarily on black smoke (BS) as their PM
23 metric. BS is a particle reflectance measure that provides an indicator of PM blackness and is
24 highly correlated with airborne carbonaceous particle concentrations (Bailey and Clayton, 1982).
25 In the U.S., Coefficient of Haze (CoH) is a metric of particle transmittance that similarly most
26 directly represents a metric of particle blackness and ambient elemental carbon levels (Wolff
27 et al., 1983) and has been found to be highly correlated with BS (r = 0.9; Lee et al., 1972).
28 However, the relationship between airborne carbon and total mass of overall aerosol (PM)
29 composition varies over time and from locality to locality, so the BS-mass ratio is less reliable
30 than the BS-carbon relationship (Bailey and Clayton, 1982). This means that the BS-mass
31 relationship is likely to be very different between Europe and the U.S., largely due to differences
June 2003 8-161 DRAFT-DO NOT QUOTE OR CITE
-------
1 in local PM source characteristics (e.g., percentages of diesel powered motor vehicles).
2 Therefore, while these European BS-health effects studies may be of qualitative interest for
3 evaluating the PM-health effects associations, they are not as useful for quantitative assessment
4 of PM effects relevant to the U. S.
5 Probably the most extensive and useful recent European air pollution health effects
6 analyses have been conducted as part of the APHEA multi-city study, which evaluated
7 15 European cities from 10 different countries with a total population of over 25 million.
8 All studies used a standardized data collection and analysis approach, which included
9 consideration of the same suite of air pollutants (BS, SO2, NO2, SO2, and O3) and the use of time-
10 series regression addressing seasonal and other long-term patterns; influenza epidemics; day of
11 the week; holidays; weather; and autocorrelation (Katsouyanni et al., 1996). The general
12 coherence of the APHEA results with other results gained under different conditions strengthens
13 the argument for causality in the air pollution-health effects association. In earlier studies, the
14 general use of the less comparable suspended particle (SPM) measures and BS as PM indicators
15 in some of the APHEA locations and analyses lessens the quantitative usefulness of such
16 analyses in evaluating associations between PM and health effects most pertinent to the U.S.
17 situation. However, Atkinson et al. (2001) report results of PM10 analyses in a study of eight
18 APHEA cities.
19 As for other single-city European studies of potential interest here, Hagan et al. (2000)
20 compared the association of PM10 and co-pollutants with hospital admissions for respiratory
21 causes in Drammen, Norway during 1994-1997. Respiratory admissions averaged only 2.2 per
22 day; so, the power of this analysis is weaker than studies looking at larger populations and longer
23 time periods. The HEI IB Multi-city Report modeling approach was employed. While a
24 significant association was found for PM10 as a single pollutant, it became non-significant in
25 multiple pollutant models. In two pollutant models, the associations and effect size of pollutants
26 were generally diminished, and when all eight pollutants were considered in the model, all
27 pollutants became non-significant. These results are typical of the problems of analyzing and
28 interpreting the coefficients of multiple pollutant models when the pollutants are even
29 moderately inter-correlated over time. A unique aspect of this work was that benzene was
30 considered in this community strongly affected by traffic pollution. In two pollutant models,
31 benzene was most consistently still associated. The authors conclude that PM is mainly an
June 2003 8-162 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
indicator of air pollution in this city and emissions from vehicles seem most important for health
effects. Thompson et al. (2001) report a similar result in Belfast, Northern Ireland, where, after
adjusting for multiple pollutants, only the benzene level was independently associated with
asthma emergency department (ED) admissions.
8.3.2.4 Key New Respiratory Medical Visits Studies
As discussed above, medical visits include both hospital ED visits and doctors' office
visits. As in the past PM AQCD's, most available morbidity studies in Table 8B-3,
Appendix 8B and in Table 8-22 below are of ED visits and their associations with air pollution.
These studies collectively confirm the results provided in the previous AQCD, indicating a
positive and generally statistically significant association between ambient PM levels and
increased respiratory-related hospital visits.
TABLE 8-22. SUMMARY OF UNITED STATES PM10, PM25, AND PM1025 ASTHMA
MEDICAL VISIT STUDIES
Reference
PM10
Choudhury et al. (1997)
Lipsett et al.
Tolbert et al.
Tolbert et al.
(1997)
(2000b)
(2000a)*
Outcome
Measures
Asthma
Asthma
Asthma
Asthma
Mean
Levels Co-Pollutants
(jig/m3) Measured
41.5 Not considered
61.2 NO2, 03
38.9 O3
29.1 NO2, O3, CO, SO2
Lag Method
0 GLM
2 GLM
1 GEE
0-2 GLM
Effect Estimate
(95% CL)
20.9(11.8,30.8)
34.7 (16, 56.
20 °C
SP 13.2 (1.2
SP 8.8 (-8.7
5) at
, 26.7)
, 54.4)
PMiS
Tolbert et al. (2000a)* Asthma 19.4 NO2,03, CO, SO2 0-2 GLM SP 2.3 (-14.8,22.7)
PM
rlV110-2.5
Tolbert et al. (2000a)*
Asthma
9.39 NO2, O3, CO, SO2 0-2 GLM SP 21.1 (-18.2, 79.3)
NS = Natural Spline General Linear Model; PS = Penalized Spline General Additive Model; SP = Single Pollutant
Model; MP = Multipollutant Model
*Preliminary results based on emergency department visit data from 18 of 33 participating hospitals.
June 2003
8-163
DRAFT-DO NOT QUOTE OR CITE
-------
1 Of the medical visit and hospital admissions studies since the 1996 PM AQCD, among the
2 most informative are those that evaluate health effects at levels below previously well-implicated
3 PM concentrations. As for U.S. studies, Tolbert et al. (2000b) reported a significant PM10
4 association with pediatric ED visits in Atlanta where mean PM10 = 39 |ig/m3 and maximum PM10
5 = 105 |ig/m3. The Lipsett et al. (1997) study of winter air pollution and asthma emergency visits
6 in Santa Clara Co, CA, may provide insight where one of the principal sources of PM10 is
7 residential wood combustion (RWC). Their results demonstrate an association between PM
8 levels and asthma. Also of interest, Delfino et al. (1997) found significant PM10 and PM2 5
9 associations for respiratory ED visits among older adults in Montreal when mean PM10 =
10 21.7 |ig/m3 and mean PM25 = 12.2 |ig/m3. Hajat et al. (1999) also reported significant PM10
11 associations with asthma doctor's visits for children and young adults in London when mean
12 PM10 = 28.2 |ig/m3 and the PM10 90th percentile was only 46.4 |ig/m3. Overall, then, several new
13 medical visits studies indicate PM-health effects associations at lower PM2 5 and PM10 levels
14 than demonstrated previously for this health outcome.
15
16 8.3.2.4.1 Scope of Medical Visit Morbidity Effects
17 Several newer medical visit studies consider a new endpoint for comparison with ED
18 visits: visits in the primary care setting. In particular, key studies showing PM associations for
19 this health outcome include: the study by Hajat et al. (1999) that evaluated the relationship
20 between air pollution in London, UK; and daily General Practice (GP) doctor consultations for
21 asthma and other lower respiratory disease (LRD); the study by Choudhury et al. (1997) of
22 private asthma medical visits in Anchorage, Alaska; and the study by Ostro et al. (1999b) of
23 daily visits by young children to primary care health clinics in Santiago, Chile for upper or lower
24 respiratory symptoms.
25 While limited in number, the above studies collectively provide new insight into the fact
26 that there is a broader scope of severe morbidity associated with PM air pollution exposure than
27 previously documented. As the authors of the London study note: "There is less information
28 about the effects of air pollution in general practice consultations but, if they do exist, the public
29 health impact could be considerable because of their large numbers." Indeed, the London study
30 of doctors' GP office visits indicates that the effects of air pollution, including PM, can affect
31 many more people than indicated by hospital admissions alone.
June 2003 8-164 DRAFT-DO NOT QUOTE OR CITE
-------
1 These new studies also provide indications as to the quantitative nature of medical visits
2 effects, relative to those for hospital admissions. In the London case, comparing the number of
3 admissions from the authors' earlier study (Anderson et al., 1996) with those for GP visits in the
4 1999 study (Hajat et al., 1999) indicates that there are circa 24 asthma GP visits for every asthma
5 hospital admission in that city. Also, comparing the PM10 coefficients indicates that the all-ages
6 asthma effect size for the GP visits (although not statistically different) was about 30% larger
7 than that for hospital admissions. Thus, these new studies suggest that looking at only hospital
8 admissions and emergency hospital visit effects may greatly underestimate the overall numbers
9 of respiratory morbidity events due to acute ambient PM exposure.
10
11 8.3.2.4.2 Factors Potentially Affecting Respiratory Medical Visit Study Outcomes
12 Some newly available studies have examined certain factors that might extraneously affect
13 the outcomes of PM-medical visit studies. Stieb et al. (1998a) examined the occurrence of bias
14 and random variability in diagnostic classification of air pollution and daily cardiac or
15 respiratory ED visits, such as for asthma, COPD, respiratory infection, etc. They concluded that
16 there was no evidence of diagnostic bias in relation to daily air pollution levels. Also, Stieb et al.
17 (1998b) reported that for a population of adults visiting an emergency department with cardiac
18 respiratory disease, fixed site sulfate monitors appear to accurately reflect daily variability in
19 average personal exposure to particulate sulfate, whereas acid exposure was not as well
20 represented by fixed site monitors. Another study investigated possible confounding of
21 respiratory visit effects due to pollens (Steib et al, 2000). Pollen levels did not influence the
22 results, similar to asthma panel studies described below in Section 8.3.3. In London, Atkinson
23 et al. (1999b) studied the association between the number of daily ED visits to for respiratory
24 complaints and measures of outdoor air pollution for PM10, NO2, SO2 and CO. They examined
25 different age groups and reported strongest associations for children for visits for asthma, but
26 were unable to separate PM10 and SO2 effects.
27
28 8.3.2.5 Identification of Potential Susceptible Subpopulations
29 Associations between ambient PM measures and respiratory admissions have been found
30 for all age groups, but older adults and children generally have been indicated by hospital
31 admissions studies to exhibit the most consistent PM-health effects associations. As reported in
June 2003 8-165 DRAFT-DO NOT QUOTE OR CITE
-------
1 previous PM AQCDs, numerous studies of older adults (e.g., those 65+ years of age) have
2 related acute PM exposure with an increased incidence of hospital admissions (e.g., see
3 Anderson et al, 1998). However, only a limited number have specifically studied children as a
4 subgroup. Burnett et al. (1994) examined the differences in air pollution-hospital admissions
5 associations as a function of age in Ontario, reporting that the largest percentage increase in
6 admissions was found among infants (neonatal and post-neonatal, one year or less in age).
7 Further efforts have aimed at identifying and quantifying air pollution effects among
8 potentially especially susceptible sub-populations of the general public. Some new studies have
9 further investigated the hypothesis that the elderly are especially affected by air pollution.
10 Zanobetti et al. (2000a) examined PM10 associations with hospital admissions for heart and lung
11 disease in ten U.S. cities, finding an overall association for COPD, pneumonia, and CVD. They
12 found that these results were not significantly modified by poverty rate or minority status in this
13 population of Medicare patients. Ye et al. (2001) examined emergency transports to the hospital.
14 Both PM10 and N02 levels were significantly associated with daily hospital transports for angina,
15 cardiac insufficiency, myocardial infarction, acute and chronic bronchitis, and pneumonia. The
16 pollutant effect sizes were generally found to be greater in men than in women, except those for
17 angina and acute bronchitis, which were the same across genders. Thus, in these various studies,
18 cardiopulmonary hospital visits and admissions among the elderly were seen to be consistently
19 associated with PM levels across numerous locales in the U.S. and abroad, generally without
20 regard to race or income; but sex was sometimes an effect modifier.
21 Several new studies of children's morbidity also support the indication of air pollution
22 effects among children. Pless-Mulloli et al. (2000) evaluated children's respiratory health and
23 air pollution near opencast coal mining sites in a cohort of nearly 5,000 children aged 1-11 in
24 England. Mean PM levels were not high (mean < 20 |ig/m3 PM10), but statistically significant
25 PM10 associations were found with respiratory symptoms. A roughly 5 percent increase of
26 General Practitioner medical visits was also noted, but was not significant. Ilabaca et al. (1999)
27 also found an association between levels of fine PM and ED visits for pneumonia and other
28 respiratory illnesses among children < 15 years in Santiago, Chile, where the levels of PM25
29 were very high (mean = 71.3 |ig/m3) during 1995-1996. The authors found it difficult to separate
30 out the effects of various pollutants, but concluded that PM (especially the fine component) is
31 associated with the risk of these respiratory illnesses. Overall, these new studies support past
June 2003 8-166 DRAFT-DO NOT QUOTE OR CITE
-------
1 assertions that children, and especially neo-natal infants, are especially susceptible to the health
2 effects of air pollution.
3 The respiratory-related hospital admissions studies summarized in Appendix 8B reveal that
4 the PM RR's for all children (e.g., 0-18) are not often notably larger than those for adults, but
5 such comparisons of RR's must adjust for differences in baseline risks for each group. For
6 example, if hospital admissions per 100,000 per day for young children are double the rate for
7 adults, then they will have a pollution relative risk (RR) per |ig/m3 that is half that of the adults
8 given the exact same impact on admissions/100,000/|ig/m3/day. Thus, it is important to adjust
9 RR's or Excess Risks (ER's) for each different age groups' baseline, but this information is
10 usually not available (especially regarding the population catchment for each age group in each
11 study). One of the few indications that is notable when comparing children with other age group
12 effect estimates in Table 8B-2 is the higher excess risk estimate for infants (i.e., the group < 1 yr.
13 of age) in the Gouveia and Fletcher (2000) study, an age group that has estimated risk estimate
14 roughly twice as large as for other children or adults.
15
16 8.3.2.6 Summary of Salient Findings on Acute Particulate Matter Exposure and
17 Respiratory-Related Hospital Admissions and Medical Visits
18 The results of new studies discussed above are generally consistent with and supportive of
19 findings presented in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a),
20 with regard to ambient PM associations of short-term exposures with respiratory-related hospital
21 admissions/medical visits. Figure 8-14 summarizes results for maximum excess risk of
22 respiratory-related hospital admission and visits per 50 |ig/m3 PM10 based on single-pollutant
23 models for selected U.S. cities. The excess risk estimates fall most consistently in the range of
24 5 to 20% per 50 |ig/m3 PM10 increments, with those for asthma visits and hospital admissions
25 generally somewhat higher than for COPD and pneumonia hospital admissions. More limited
26 new evidence both (a) substantiates increased risk of respiratory-related hospital admissions due
27 to ambient fine particles (PM2 5, PMLO, etc.) and also (b) points towards such admissions being
28 associated with ambient coarse particles (PM10_2.5). Excess risk estimates tend to fall in the range
29 of ca. 5.0 to 15.0% per 25 |ig/m3 PM25 or PM10_25 for overall respiratory admissions or for COPD
30 admissions, whereas larger estimates are found for asthma admissions.
31
June 2003 8-167 DRAFT-DO NOT QUOTE OR CITE
-------
Tolbert et al. (2000)
Atlanta
Choudhury et al. (1997),
Anchorage
Sheppard (2003)
Seattle
Nauenberg and Basu (1999)
LA.CA
Zanobetti and Schwartz (2003)
14 US Cities
Moolgavkar (2003)
Chicago
Moolgavkar (2003),
LA
Ito (20Q3)>
Detroit
Zanobetti and Schwartz (2003)
14 US Cities
Ito (2003),
Detroit
Asthma Visits
i
-10
Asthma Hospital Admission
COPD Hospital Admission
Pneumonia Hospital Admission
-50 5 10 15 20 25 30 35
Figure 8-14. Maximum excess risk of respiratory-related hospital admissions and visits
per 50 ug/m3 PM10 increment in selected studies of U.S. cities based on single-
pollutant models.
1 Various new medical visits studies (including non-hospital physician visits) indicate that
2 the use of hospital admissions alone can greatly understate the total clinical morbidity effects of
3 air pollution. Thus, these results support the hypothesis that considering only hospital
4 admissions and ED visit effects may greatly underestimate the numbers of medical visits
5 occurring in a population as a result of acute ambient PM exposure. Those groups identified in
6 these morbidity studies as most strongly affected by PM air pollution are older adults and the
7 very young.
8
9 8.3.3 Effects of Particulate Matter Exposure on Lung Function and
10 Respiratory Symptoms
11 In the 1996 PM AQCD, the available respiratory studies used a wide variety of designs
12 examining pulmonary function and respiratory symptoms in relation to ambient concentrations
13 of PM10. The populations studied included several different subgroups (e.g., children, asthmatics,
June 2003
8-168
DRAFT-DO NOT QUOTE OR CITE
-------
1 etc.); and the models used for analysis varied, but did not include GAM use. The pulmonary
2 function studies were suggestive of short-term effects resulting from ambient PM exposure.
3 Peak expiratory flow rates showed decreases in the range of 2 to 5 1/min per 50 |ig/m3 increase in
4 24-h PM10 or its equivalent, with somewhat larger effects in symptomatic groups, e.g.,
5 asthmatics. Studies using FEVj or FVC as endpoints showed less consistent effects. The
6 chronic pulmonary function studies, less numerous than the acute studies, had were inconclusive
7 results.
8
9 8.3.3.1 Effects of Short-Term Particulate Matter Exposure on Lung Function and
10 Respiratory Symptoms
11 The available acute respiratory symptom studies discussed in the 1996 PM AQCD included
12 several different endpoints, but typically presented results for upper respiratory symptoms, lower
13 respiratory symptoms, or cough. These respiratory symptom endpoints had similar general
14 patterns of results. The odds ratios were generally positive, the 95% confidence intervals for
15 about half of the studies being statistically significant (i.e., the lower bound exceeded 1.0).
16 The earlier studies of morbidity health outcomes of PM exposure on asthmatics were
17 limited in terms of conclusions that could be drawn because of the few available studies on
18 asthmatic subjects. Lebowitz et al. (1987) reported a relationship with TSP exposure and
19 productive cough in a panel of 22 asthmatics but not for peak flow or wheeze. Pope et al. (1991)
20 reported on respiratory symptoms in two panels of Utah Valley asthmatics. The 34 asthmatic
21 school children panel yielded estimated odd ratios of 1.28 (1.06, 1.56) for lower respiratory
22 illness (LRI) and the second panel of 21 subjects aged 8 to 72 for LRI of 1.01 (0.81, 1.27) for
23 exposure to PM10. Ostro et al. (1991) reported no association for PM25 exposure in a panel of
24 207 adult asthmatics in Denver; but, for a panel of 83 asthmatic children age 7 to 12 in central
25 Los Angeles, found a relationship of shortness of breath to O3 and PM10, but could not separate
26 effects of the two pollutants (Ostro et al., 1995). These few studies did not indicate a consistent
27 relationship for PM10 exposure and health outcome in asthmatics.
28 Numerous new studies of short-term PM exposure effects on lung function and respiratory
29 symptoms published since 1996 were identified by an ongoing Medline search. Most of these
30 followed a panel of subjects over one or more time periods and evaluated daily lung function
31 and/or respiratory symptom in relation to changes in ambient PM10, PM10_2 5, and/or PM2 5. Some
32 used other measures of airborne particles, e.g. ultrafine PM, TSP, BS, and sulfate fraction of
June 2003 8-169 DRAFT-DO NOT QUOTE OR CITE
-------
1 ambient PM. Lung function was usually measured daily, with most studies including forced
2 expiratory volume (FEV), forced vital capacity (FVC) and peak expiratory flow rate (PEF),
3 measured both in the morning and afternoon. Various respiratory symptoms were measured,
4 e.g., cough, phlegm, difficulty breathing, wheeze, and bronchodilator use. Detailed summaries
5 of these studies are presented in Appendix 8B. Data on physical and chemical aspects of
6 ambient PM levels (especially for PM10, PM10_2 5, PM2 5, and smaller size fractions) are of
7 particular interest, as are new studies examining health outcome effects and/or exposure
8 measures not much studied in the past.
9 Specific studies were selected for summarization based on the following criteria:
10 • Peak flow was used as the primary lung function measurement of interest.
11 • Cough, phlegm, difficulty breathing, wheeze, and bronchodilator use were summarized as
measures of respiratory symptoms when available.
12 • Quantitative relationships were estimated using PM10, PM2 5, PM10_2 5, and/or smaller PM as
independent variables.
13 • Analyses used in the study were done such that each individual served as their own control.
14
15 8.3.3.1.1 Lung Function and Respiratory Symptom Effects in Asthmatic Subjects
16 Appendix B Tables 8B-4 and 8B-5 summarize salient features of new studies of short-term
17 PM exposure effects on lung function and respiratory symptoms, respectively, in asthmatic
18 subjects; and key quantitative results are summarized in Table 8-23 for PM10 and Table 8-24 for
19 PM2 5. The peak flow analyses results for asthmatics tend to show small decrements for PM10
20 and PM25 as seen in studies by Gielen et al. (1997), Peters et al. (1997b), Romieu et al. (1997),
21 and Pekkanen et al. (1997).
22 The peak flow analyses results for asthmatics tend to show small decrements for both PM10
23 and PM2 5. For PM10, the available point estimates for morning PEF lagged one day showed
24 decreases, but the majority of the studies were not statistically significant (as per Table 8-23 and
25 as shown in Figure 8-15 as an example of PEF outcomes). Lag 1 may be more relevant for
26 morning measurement of asthma outcome from the previous day. The figure presents studies
27 which provided such data. The results were consistent for both AM and PM peak flow analyses.
28 Effects using two- to five-day lags averaged about the same as did the zero to one-day lags, but
29
June 2003 8-170 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-23. SUMMARY OF QUANTITATIVE PFT CHANGES IN ASTHMATICS PER 50 jig/m3 PM10 INCREMENT
to
o
o
OJ
oo
1
^
o
H
6
o
0
H
O
o
w
o
o
HH
H
W
Reference citation, location, etc.
Asthma Studies
Pekkanen et al. (1997)
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Peters etal. (1997a)
Peters etal. (1997c)
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Pekkanen et al. (1997)
Peters etal. (1996)
Peters etal. (1997a)
Peters etal. (1997c)
Timonen & Pekkanen (1997) Urban
Timonen & Pekkanen (1997) Suburban
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Segalaetal. (1998)
Pekkanen etal. (1997)
Outcome
Measure
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Morning PEFR
Evening PEFR
Mean Particulate Levels
(Range) |ig/m3
14(10,23)
30.5(16,60)
166.8(29,363)
(12, 126)
47 (29, 73)
55(7,71)
30.5(16,60)
166.8(29,363)
(12, 126)
30.5(16,60)
166.8(29,363)
(12, 126)
14(10,23)
112
47 (29, 73)
55(7,71)
18(7,60)
13(7,37)
30.5(16,60)
166.8(29,363)
(12, 126)
34.2 (9, 95)
14(10,23)
Co-pollutants
Measured
NO2
Ozone
Ozone
Ozone
SO2, sulfate, H+
SO2, sulfate, H+
Ozone
Ozone
Ozone
Ozone
Ozone
Ozone
NO2
SO2, sulfate, PSA
S02, sulfate, H+
S02, sulfate, H+
NO2, SO2
NO2, SO2
Ozone
Ozone
Ozone
SO2, NO2
N02
Lag
Structure
Oday
1 day
Iday
Iday
1 day
1 day
2 day
2 day
2 day
Oday
Oday
Oday
Oday
Oday
Oday
Oday
Oday
Oday
2 day
2 day
2 day
2 day
2 day
Effect measures standardized
to 50 ng/m3 PM10
-2.71 (-6.57, 1.15)
1.39 (-0.57, 3.35)
-4.70 (-7.65, -1.70)
-0.65 (-5.32, 3. 97)
-0.84 (-1.62, -0.06)
-1.30 (-2. 36, -0.24)
0.34 (-1.78, 2.46)
-4.90 (-8.40, -1.50)
2.47 (-1.75, 6. 75)
-0.30 (-2.24, 1.64)
-4.80 (-8.00, -1.70)
-1.32 (-6.82, 4.17)
-0.35 (-4. 31, 3. 61)
-1.03 (-1.98, -0.08)
-0.92 (-1.96, 0.12)
-0.37 (-1.82, 1.08)
-1.10 (-5.20, 3.00)
-1.66 (-8.26, 4. 94)
-2.32 (-5.36, 0.72)
-3.65 (-7.20, 0.03)
-0.04 (-4.29, 4.21)
-0.62 (-1.52, 0.28)
0.14 (-6.97, 7.25)
-------
c
3
TABLE 8-23 (cont'd). SUMMARY OF QUANTITATIVE PFT CHANGES IN ASTHMATICS
o
o
LtJ
oo
1
^
to
Reference citation, location, etc.
Asthma Studies (cont'd)
Peters etal. (1997c)
Timonen & Pekkanen (1997) Urban
Timonen & Pekkanen (1997) Suburban
Peters etal. (1996)
Peters etal. (1997a)
Timonen & Pekkanen (1997) Urban
Timonen & Pekkanen (1997) Suburban
Hiltermann etal. (1998)
Hiltermann etal. (1998)
Hiltermann etal. (1998)
Vedal etal. (1998)
Outcome
Measure
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Ave. AM & PM
Ave. AM & PM
Ave. AM & PM
Ave. AM & PM
TABLE 8-24. SUMMARY OF PFT
Mean Particulate Levels
(Range) ug/m3
55(7,71)
18(7,60)
13(7,37)
112
47 (29, 73)
18(7,60)
13(7,37)
39.7(16,98)
39.7(16,98)
39.7(16,98)
19.1(1,159)
Co-pollutants
Measured
S02, sulfate, H+
N02, S02
N02, S02
SO2, sulfate, PSA
S02, sulfate, H+
N02, S02
N02, S02
Ozone, NO2, SO2
Ozone, NO2, SO2
Ozone, NO2, SO2
None
CHANGES IN ASTHMATICS PER 25
Outcome Mean Particulate Levels
O
l>
H
6
o
z;
0
H
O
O
w
o
o
H
W
Reference citation, location, etc.
Romieu etal. (1996)
Peters etal. (1997c)
Romieu etal. (1996)
Peters etal. (1997c)
Romieu etal. (1996)
Peters etal. (1997c)
Romieu etal. (1996)
Peters etal. (1997c)
Measure
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
(Range) ug/m3
85.7(23,177)
50.8 (9, 347)
85.7(23,177)
50.8 (9, 347)
85.7(23,177)
50.8 (9, 347)
85.7(23,177)
50.8 (9, 347)
Co-pollutants
Measured
Ozone
S02, sulfate, H+
Ozone
SO2, sulfate, H+
Ozone
S02, sulfate, H+
Ozone
SO2, sulfate, H+
Lag
Structure
2 day
2 day
2 day
5 day
1-5 day
1-4 day
1-4 day
1 day
2 day
1-7 day
1-4 day
ug/m3 PM2 5
Lag
Structure
1 day
1 day
2 day
1-5 day
Oday
Oday
2 day
1-5 day
Effect measures standardized
to 50 ug/m3 PM10
-2.31 (-4.53, -0.10)
-1.13 (-4.75, 2. 52)
0.38 (-6.37, 7.13)
-1.12 (-2. 13, -0.10)
-1.34 (-2.83, 0.15)
-0.73 (-7.90, 6.44)
-4.18 (-20.94, 12.58)
-0.90 (-3.84, 2.04)
-0.50 (-4.22, 3.22)
-2.20 (-10.43, 6.03)
-1.35 (-2.70, -.05)
INCREMENT
Effect measures standardized
to 25 ug/m3 PM25
-3. 65 (-8.25, 1.90)
-0.71 (-1.30, 0.12)
-3. 68 (-9. 37, 2.00)
-1.19 (-1.18,0.57)
-4.27 (-7.12, -0.85)
-0.75 (-1.66, 0.17)
-2.55 (-7.84, 2.740
-1.79 (-2.64, -0.95)
-------
Romieu et al. (1997)
(Mexico)
Pekkanen et al. (1997)
(Finland)
Romieu et al. (1996)
(Mexico)
Gielenetal. (1997)
(Netherlands)
-10.0
5.0 0.0 5.0
Change in Pulmonary Function, L/min
10.0
Figure 8-15. Selected acute pulmonary function change studies of asthmatic children.
Effect of 50 ug/m3 PM10 on morning Peak flow lagged one-day.
1
2
3
4
5
9
10
11
12
13
14
15
16
17
had wider confidence limits. Similar results were found for the fewer PM25 studies. Of these,
Pekkanen et al. (1997) and Romieu et al. (1996) found similar results for PM2 5 and PM10, while
the study of Peters et al. (1997c) found slightly larger effects for PM2 5.
Pekkanen et al. (1997) also reported changes in peak flow to be related to several sizes of
PM withPN 0.032-0.10 -0.970 (0.502) l(cm3) andPM10.32 -0.901 (0.536) and PM10 -1.13
(0.478) for morning PEF lag 2. Peters et al. (1997c) report that the strongest effects on peak
flow were found with ultrafme particles: PMMC001_01: -1.21 (-2.13, -0.30); PMMC001_25:
-1.01 (-1.92, -0.11); andPM10, -1.30 (-2.36, -0.24). Penttinen et al. (2001) using biweekly
spirometry over 6 months on a group of 54 adult asthmatics found that FVC, FEVb and
spirometric PEFR were inversely, but mostly nonsignificantly-associated with ultra fine particle
concentrations. Compared to the effect estimates for self-monitored PEFR, the effect estimates
for spirometric PEFR tended to be larger. The strongest associations were observed in the size
range of 0.1 to 1 |im. In a further study, von Klot et al. (2002) evaluated 53 adult asthmatics in
Erfurt, Germany in the winter of 1996-1997. Relationships were estimated from generalized
estimating equations, adjusting for autocorrelation. Asthma symptoms were related to small
particles (MC 0.1-0.5, MC 0.01-2.5) and PM25.10. The strongest relations were for 14 day mean
PM levels, especially for the smaller particles (MC 0.01-2.5).
June 2003
8-173
DRAFT-DO NOT QUOTE OR CITE
-------
1 Overall, then, PM10 and PM2 5 both appear to affect lung function in asthmatics, but there is
2 only limited evidence for a stronger effect of fine versus coarse fraction particles; nor do
3 ultrafine particles appear to have any notably stronger effect than other larger-diameter fine
4 particles. Also, of the studies provided, few if any analyses were able to clearly separate out the
5 effects of PM10 and PM25 from other pollutants.
6 The effects of PM10 on respiratory symptoms in asthmatics tended to be positive, although
7 they are somewhat less consistent than PM10 effects on lung function. Most studies showed
8 increases in cough, phlegm, difficulty breathing, and bronchodilator use, although these
9 increases were generally not statistically significant for PM10 (see Tables 8-25, 8-26, 8-27, and
10 8-28; and, for cough as an example, see Figure 8-16). Vedal et al. (1998) reported that
11 (a) increases in PM10 were associated with increased reporting of cough, phlegm production, and
12 sore throat and (b) children with diagnosed asthma are more susceptible to the effects than are
13 other children. Similarly, in the Gielen et al. (1997) study of a panel of children, most of whom
14 had asthma, low levels of PM increased symptoms and medication use. The Peters et al. (1997c)
15 study of asthmatics examined particle effects by size and found that fine particles were
16 associated with increases in cough, of which MC 0.01-2.5 was the best predictor.
17 Delfino et al. (1998) used an asthma symptom score to evaluate the effects of acute air
18 pollutant exposures. The 1- and 8-hr PM10 maximum concentrations had larger effects than the
19 24-hr mean. Subgroup analyses showed effects of current day PM maxima to be strongest in the
20 10 more frequently symptomatic children; the odds ratios for adverse symptoms from 90th
21 percentile increases were 2.24 (1.46, 3.46), for 1-hr PM10; 1.82 (1.18, 2.8), for 8-hr PM10, and
22 1.50 (0.80-2.80) for 24-hr PM10. Analyses suggested that effects of O3 and PM10 were largely
23 independent. Delfino et al. (2002) also studied 22 asthmatic children aged 9-19 years in March
24 and April 1996. Relationships were evaluated by use of generalized estimating equations,
25 adjusting for autocorrelation. The endpoint was symptoms interfering with daily activities. This
26 endpoint was associated with PM10, NO2, and ozone. There was a positive interaction effect of
27 PM10 and NO2 jointly.
28 Romieu et al. (1996) found children with mild asthma to be more strongly affected by high
29 ambient levels of PM (mean PM10 = 166.8 |ig/m3) observed in northern Mexico City than in a
30 study (Romieu et al., 1997) conducted in a nearby area with lower PM10 levels (mean PM10 =
31 54.2 |ig/m3). Yu et al. (2000) reported estimates of odds ratios for asthma symptoms and
June 2003 8-174 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-25. SUMMARY OF ASTHMA PM1(1 COUGH STUDIES
10
to
o
o
OJ
oo
^
o
£j
H
6
o
0
H
O
o
o
o
H
W
Reference citation,
location, etc.
Asthma Studies
Vedal et al. (1998)
Gielenetal. (1997)
Hiltermann etal. (1998)
Peters etal. (1997c)
Peters etal. (1997b)
Romieuetal. (1997)
Romieuetal. (1996)
Vedal et al. (1998)
Gielenetal. (1997)
Segala etal. (1998)
Neukirch et al. (1998)
Romieuetal. (1996)
Romieuetal. (1997)
Ostro etal. (2001)
Hiltermann et al. (1998)
Peters etal. (1997c)
Peters etal. (1997b)
Ostro etal. (2001)
Outcome
Measure
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR nocturnal cough
OR nocturnal cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
Mean Paniculate Levels
(Range) ug/m3
19.1(1,159)
30.5 (16, 60)
39.7 (16, 98)
55(7,71)
47 (29, 73)
(12, 126)
166.8 (29, 363)
19.1(1,159)
30.5 (16, 60)
34.2 (9, 95)
34.2 (9, 95)
166.8 (29, 363)
(12, 126)
47(11, 1 19)24 hr
39.7 (16, 98)
55(7,71)
47 (29, 73)
102 (47, 360) 1 hr max
Co-pollutants
Measured
None
Ozone
Ozone, NO2, SO2
SO2, sulfate, H+
SO2, sulfate, H+
Ozone
Ozone
None
Ozone
SO2, NO2
SO2, NO2
Ozone
Ozone
Ozone, NO2
Ozone, NO2, SO2
SO2, sulfate, H+
SO2, sulfate, H+
ozone, NO2
Lag
Structure
Oday
Oday
Oday
Oday
Oday
Oday
Oday
2 day
2 day
2 day
3 day
2 day
2 day
3 day
1-7 day
1-5 day
1-5 day
3 day
Effect measures standardized to
50 ug/m3 PM10
1.40 (1.04, 1.88)
2.19(0.77,6.20)
0.93 (0.83, 1.04)
1.32(1.16, 1.50)
1.01 (0.97, 1.07)
1.21 (1.10, 1.33)
1.27(1.16, 1.42)
1.40(1.13, 1.73)
2.19(0.47, 10.24)
(values not given because
not significant)
(values not given because
not significant)
1.27 (1.07, 1.50)
1.00(0.92, 1.10)
1.32(1.12, 1.55)
0.94 (0.82, 1.08)
1.30(1.09, 1.55)
1.10(1.04, 1.17)
1.05(1.02, 1.18)
-------
TABLE 8-26. SUMMARY OF ASTHMA PM1(1 PHLEGM STUDIES
10
to
o
o
OJ
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 Particulate Levels
(Range) ug/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 standardized
to 50 ug/m3 PM10
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)
oo
TABLE 8-27. SUMMARY OF ASTHMA PM10 LOWER RESPIRATORY ILLNESS (LRI) STUDIES
H
6
0
*
o
H
O
c
o
H
W
O
F>
O
HH
H
W
Reference citation,
location, etc.
Vedal etal. (1998)
Gielen etal. (1997)
Romieuetal. (1997)
Romieuetal. (1996)
Vedal etal. (1998)
Gielen etal. (1997)
Segala etal. (1998)
Romieuetal. (1997)
Romieuetal. (1996)
Delfino etal. (1998)
Outcome
Measure
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
Mean Particulate 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-h 43 (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 ug/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
*T • ^* • • • * • * u ^u. \J \J lYAlYAflAX A v^A rikj A AA_LYA_T^ A I.YAJQ A»AVV^.LI V>AAV^A^AA_^.T^ A V^AV tj kJAj kj A ^A^AA^LJ
^ Reference citation, Mean Paniculate Co-pollutants
o location, etc. Outcome Measure Levels (Range) ug/m3 Measured
Gielenetal. (1997) OR bronchodilator use 30.5(16,60) Ozone
Hiltermannetal. (1998) OR bronchodilator use 39.7(16,98) Ozone, NO2, SO2
Peters et al. (1997b) OR bronchodilator use 47(29,73) SO2, sulfate, H+
Gielenetal. (1997) OR bronchodilator use 30.5(16,60) Ozone
Hiltermann et al. (1998) OR bronchodilator use 39.7 (16, 98) Ozone, NO2, SO2
Peters et al. (1997b) OR bronchodilator use 47(29,73) SO2, sulfate, H+
Lag Effect measures standardized
Structure to 50 ug/m3 PM10
0 day 0.94 (0.59, 1.50)
Oday 1.03(0.93,1.15)
Oday 1.06(0.88,1.27)
2 day 2.90(1.81,4.66)
1-7 day 1.12(1.00,1.25)
1-5 day 1.23 (0.96, 1.58)
oo
O
H
6
0
o
H
0
0
H
W
0
O
HH
H
W
-------
Gielenetal. (1997)
(Netherlands)
Romieu et al. (1997)
(Mexico)
Peters etal. (1997b)
(Czech Republic)
H Vedal etal. (1998)
(Canada)
0.5
1.0 2.0
Odds Ratio for Cough
4.0
8.0
Figure 8-16. Odds ratios with 95% confidence interval for cough per 50-ug/m3 increase in
PM10 for selected asthmatic children studies at lag 0.
1 10 |ig/m3 increments in PM10 and PM10 values of 1.18(1.05, 1.33) and 1.09 (1.01, 1.18),
2 respectively. Multipollutant models with CO and SO2 yielded 1.06 (0.95, 1.19) for PM10, and
3 1.11 (0.98, 1.26) for PMj 0, thus showing a lower value for PM10 and a loss of significance for
4 both PM10 and PM10. The correlation between CO and PM10 and PM10 was 0.82 and 0.86. Ostro
5 et al. (2001) studied a panel of inner-city African American children using a GEE model with
6 several measures of PM, including PM10 (both 24-hour average and 1-hour max.) and PM25,
7 demonstrating positive associations with daily probability of shortness of breath, wheeze, and
8 cough.
9 Just et al. (2002) studied 82 asthmatic children for 3 months during spring and early
10 summer in Paris. Relationships were estimated from generalized estimating equations adjusting
11 for autocorrelation. No significant relationships were found between PM13 and lung function or
12 respiratory symptoms. Desqueyroux et al. (2002) studied 60 adult severe asthmatics from
13 November 1995 to November 1996. Relationships were estimated from generalized estimating
14 equations adjusting for autocorrelation. PM10 was not related to incident asthma attacks using
15 lags of 1 or 2 days; but PM10 associations for 3, 4, and 5 day lags were significant. PM10
16 remained significant even after adjusting for other pollutants including O3, SO2, and NO2.
June 2003
8-178
DRAFT-DO NOT QUOTE OR CITE
-------
1 For PM2 5 results, see Table 8-29. All showed positive associations (several being clearly
2 significant at p < 0.05) between PM2 5 and increased cough, phlegm, or LRI. Of studies that
3 included two indicators for PM (PM10, PM25) in their analyses, the study of Peters et al. (1997c)
4 found similar effects for the two PM measures, whereas the Romieu et al. (1996) study found
5 slightly larger effects for PM2 5.
TABLE 8-29. SUMMARY OF ASTHMA PM?, RESPIRATORY SYMPTOM STUDIES
-2.5
1
2
3
4
Reference citation,
location, etc.
Peters etal. (1997b)
Romieu et al.
(1996)
Tiittanen et al.
(1999)
Romieu et al.
(1996)
Tittanen et al.
(1999)
Ostro etal. (2001)
Peters etal. (1997b)
Romieu et al.
(1996)
Romieu et al.
(1996)
Romieu et al.
(1996)
Romieu et al.
(1996)
Outcome
Measure
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR
Phlegm
OR
Phlegm
OR LRI
OR LRI
Two asthma studies, both
well as by 24 hr averages. The
respiratory illness in
(Ostro etal. ,2001).
one study
Mean
Particulate
Levels (Range)
jig/m3
50.8 (9, 347)
85.7 (23, 177)
15 (3, 55)
85.7 (23, 177)
15 (3, 55)
40.8 (4, 208)
50.8 (9, 347)
85.7 (23, 177)
85.7 (23, 177)
85.7 (23, 177)
85.7 (23, 177)
in the United States,
Co-pollutants
Measured
SO2, sulfate, H+
Ozone
NO2, SO2, CO,
ozone
Ozone
NO2, SO2, CO,
ozone
Ozone, NO2
SO2, sulfate, H+
Ozone
Ozone
Ozone
Ozone
examined PM
PM10 1 hr outcome was larger than
(Delfmo et al., 1998)
but was lower
Lag
Structure
Oday
Oday
Oday
2 day
2 day
3 day
1-5 day
Oday
2 day
Oday
2 day
indicators by
Effect measures
standardized to
25 jig/m3 PM2 s
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)
1.02(0.90, 1.17)
1.21 (0.98, 1.48)
1.16(0.99, 1.39)
1.21 (1.05, 1.42)
1.16(1.05, 1.42)
1 hr averages as
the 24 hr outcome for lower
for cough in
the other study
June 2003
8-179
DRAFT-DO NOT QUOTE OR CITE
-------
1 Several of the studies reviewed above (Delfino et al., 1998, 2002; Ostro et al., 2001; Yu
2 et al., 2000; Mortimer et al., 2002; Vedal et al., 1998) that were conducted in the United States
3 and Canada found positive associations between various health endpoints for asthmatics and
4 ambient PM exposure (indexed by PM10, PM2 5, or PM10_2 5). The endpoints included PEF
5 decrements, various individual respiratory symptoms, and combinations of respiratory
6 symptoms. The various endpoints each represent effects on respiratory health.
7
8 8.3.3.1.2 Lung Function and Respiratory Symptom Effects in Nonasthmatic Subjects
9 Results for PM10 peak flow analyses in non-asthmatic studies (summarized in Appendix 8B
10 Table 8B-6) were inconsistent, with fewer studies reporting results in the same manner as for the
11 asthmatic studies. Many of the point estimates showed increases rather than decreases (see
12 Table 8-30). The effects on respiratory symptoms in non-asthmatics (see Appendix 8B Table
13 8B-7) were similar to those in asthmatics. Most studies showed that PM10 increases cough,
14 phlegm, difficulty breathing, and bronchodilator use, although these were generally not
15 statistically significant (Table 8-31). Vedal et al. (1998) reported no consistent evidence for
16 adverse health effects in a nonasthmatic control group.
17 Results of the PM2 5 peak flow and symptom analyses in non-asthmatic studies (see
18 Appendix 8B Table 8B-8, Table 8-32) were similar to PM10 results discussed above.
19 Three authors, Schwartz and Neas (2000), Tiittanen et al. (1999) and Neas et al. (1999),
20 used PM10_2 5 as a coarse fraction particulate measure (Table 8-33). Schwartz and Neas (2000)
21 found that PM10_2 5 was significantly related to cough. Tiittanen found that one day lag of
22 PM10_2 5 was related to morning PEF, but there was no effect on evening PEF. Neas et al. found
23 noeffectsofPM10.25onPEF.
24 The Schwartz and Neas (2000) reanalyses allows comparison of fine and coarse particle
25 effects on healthy school children using two pollutant models of fine and coarse PM. CM was
26 estimated by subtracting PM2 x from PM10 data. They report for cough for reanalysis of the
27 Harvard Six City Diary Study in the two PM pollutant model PM2 5 OR = 1.07 (0.90, 1.26; per
28 15 |ig/m3 increment) andPM10_25 OR 1.18 (1.04, 1.34; per 8 |ig/m3 increment) in contrast to
29 lower respiratory symptom results of PM25 OR 1.29 (1.06, 1.57) and PM10.25 1.05 (0.9, 1.23).
30 In the Uniontown reanalysis, peak flow for PM2 x for a 14 |ig/m3 increment was -0.91 1/m
31
June 2003 8-180 DRAFT-DO NOT QUOTE OR CITE
-------
•— ^
3
to
o
o
OJ
oo
1
oo
o
H
6
O
0
H
O
o
o
o
H
W
TABLE 8-30. SUMMARY OF NON-AS
Reference citation, location, etc.
Gold etal. (1999)
Tittanenetal. (1999)
Neas etal. (1999)
Tittanenetal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Neas etal. (1999)
Harre etal. (1997)
Neas etal. (1999)
Schwartz & Neas (2000) Uniontown
Schwartz & Neas (2000) State
College
Tittanenetal. (1999)
Tittanenetal. (1999)
Gold etal. (1999)
Neas etal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Van der Zee etal. (1999)
Van der Zee etal. (1999)
Van der Zee et al. (1999)
Harre etal. (1997)
THMA PM10 PFT ST
Outcome Mean Particulate Co-pollutants
Measure Levels (Range) ug/m3 Measured
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
OR > 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
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)
Ozone
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
N02, S02
N02, S02
NO2, SO2
Ozone
N02, S02, CO
Ozone
Sulfate fraction
Sulfate fraction
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Ozone
Ozone
N02, S02
NO2, SO2
NO2, SO2
NO2, SO2, sulfate
NO2, SO2, sulfate
NO2, SO2, sulfate
NO2, SO2, CO
UDIES
Lag
Structure
1 day
Oday
1-5 day
1-4 day
1 day
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 ug/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
to
o
o
OJ
oo
oo
to
o
£5
H
6
O
0
H
O
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) ug/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)
O
HH
H
W
-------
TABLE 8-32. SUMMARY OF NON-ASTHMA PM2< RESPIRATORY OUTCOME STUDIES
to
o
o
OJ
oo
oo
OJ
o
H
6
O
0
H
O
o
o
o
H
W
Reference citation, Outcome
location, etc. Measure
Gold et al. ( 1 999) Morning PEFR
Tittanen et al. ( 1 999) Morning PEFR
Tittanen et al. ( 1 999) Morning PEFR
Neas et al. ( 1 999) Morning PEFR
Schwartz & Neas (2000) Evening PEFR
Uniontown
Schwartz & Neas (2000) Evening PEFR
State College
Tittanen et al. ( 1 999) Evening PEFR
Tittanen et al. ( 1 999) Evening PEFR
Gold et al. ( 1 999) Evening PEFR
Neas et al. ( 1 999) Evening PEFR
Tittanen et al. ( 1 999) OR cough
Tittanen et al. ( 1 999) OR cough
Schwartz & Neas (2000) OR LRS
Mean Paniculate Co-pollutants Lag Effect measures standardized
Levels (Range) ug/m3 Measured Structure to 25 ug/m3 PM2 5
30.3 (9, 69) Ozone 1 day -0.22 (-0.46, 0.01)
NO2, SO2, CO, ozone Oday 1.11 (-0.64, 2.86)
NO2, SO2, CO, ozone 1-4 day -1.93 (-7.00, 3.15)
24.5(7,88) Ozone 1-5 day 2.64 (-6.56, 11.83)
(not given) Sulfate fraction Oday -1.52 (-2.80, -0.24)
(not given) Sulfate fraction 0 day -0.93 (- 1.88, 0.01)
NO2, SO2, CO, ozone Oday 0.70 (-0.81, 2.20)
NO2, SO2, CO, ozone Oday 1.52 (-3.91, 6.94)
30.3(9,69) Ozone Oday -0.10 (-0.43, 0.22)
24.5(7,88) Ozone 1-5 day 1.47 (-7.31, 10.22)
15(3,55) NO2, SO2, CO, ozone Oday 1.04(0.86,1.20)
15(3,55) NO2, SO2, CO, ozone 2 day 1.24(1.02,1.51)
(not given) Sulfate fraction Oday 1.61(1.19,2.14)
-------
TABLE 8-33. SUMMARY OF NON-ASTHMA COARSE FRACTION STUDIES OF RESPIRATORY ENDPOINTS
to
o
o
OJ
oo
1
oo
o
H
6
O
0
H
O
o
o
o
H
W
Reference citation,
location, etc.
Tittanenetal. (1999)
Neasetal. (1999)
Tittanenetal. (1999)
Tittanenetal. (1999)
Neasetal. (1999)
Tittanenetal. (1999)
Neasetal. (1999)
Tittanenetal. (1999)
Tittanenetal. (1999)
Neasetal. (1999)
Tittanenetal. (1999)
Tittanenetal. (1999)
Tittanenetal. (1999)
Schwartz & Neas (2000)
Schwartz & Neas (2000)
Outcome
Measure
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
OR cough
OR cough
OR cough
OR cough without
other symptoms
ORLRS
Mean Paniculate
Levels (Range) ug/m3
8 (.2, 67)
8.3
8 (.2, 67)
8 (.2, 67)
8.3
8 (.2, 67)
8.3
8 (.2, 67)
8 (.2, 67)
8.3
8 (.2, 67)
8 (.2, 67)
8 (.2, 67)
(not given)
(not given)
Co-pollutants
Measured
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Sulfate fraction
Sulfate fraction
Lag
Structure
Iday
Iday
2 day
1-4 day
1-5 day
Oday
Iday
2 day
1-4 day
1-5 day
Oday
2 day
1-4 day
Oday
Oday
Effect measures standardized
to 25 ug/m3 PM10.2 5
-1.26 (-2.71, 0.18)
-4.31 (-11.43, 2.75)
0.51 (-0.77,2.16)
-0.57 (-1.96, 0.81)
-6.37 (-21. 19, 8.44)
0.66 (-0.33, 1.81)
1.88 (-4.75, 8.44)
0.03 (-1.41, 1.47)
2.37 (-1.69, 4.96)
5.94(-7.00, 18.94)
0.99(0.87, 1.12)
1.23 (1.06, 1.42)
1.31(0.81,2.11)
1.77(1.24,2.55)
1.51(0.94,4.87)
-------
1 (-1.14, -1.68) andPM10.21 for 15 |ig/m3 +1.04 1/m (-1.32, +3.4); for State College PM21 -0.56
2 (-1.13,+0.01) and PM10.21 -0.17 (-2.07,+1.72).
3 Coull et al. (2001) reanalyzed data from the Pope et al. (1991) study of PM effects on
4 pulmonary function of children in the Utah Valley, using additive mixed models which allow for
5 assessment of heterogeneity of response or the source of heterogeneity. These additive models
6 describe complex covariate effects on each child's peak expiratory flow while allowing for
7 unexplained population heterogeneity and serial correlation among repeated measurements. The
8 analyses indicate heterogeneity among that population with regard to PM10 (i.e., specifically that
9 there are three subjects in the Utah Valley study who exhibited a particularly acute response to
10 PM10). However the limited demographic data available in the Utah Valley Study does not
11 explain the heterogeneity in PM sensitivity among the school children population.
12 Two studies examined multipollutant models. The Jalaludin et al. (2000) analyses used a
13 multipollutant model that evaluated PM10, O3, and NO2. They found in metropolitan Sydney that
14 ambient PM10 and O3 concentrations are poorly correlated (r = 0.13). For PEFR the P (SE) for
15 PM10 only was 0.0045 (0.0125), p = 0.72; and for PM10 and O3, 0.0051 (0.0124), p = 0.68.
16 Ozone was also unchanged in the one- and two-pollutant models. Gold et al. (1999) attempted to
17 study the interaction of PM25 and O3 on PEF in Mexico City children (age = 8 to 12 yrs). The
18 authors found independent effects of the two pollutants, but the joint effect was slightly less than
19 the sum of the independent effects.
20
21 8.3.3.2 Long-Term Particulate Matter Exposure Effects on Lung Function and
22 Respiratory Symptoms
23 8.3.3.2.1 Summary of 1996 Particulate Matter Air Quality Criteria Document Key Findings
24 In the 1996 PM AQCD, the available long-term PM exposure-respiratory disease studies
25 were limited in terms of conclusions that could be drawn. At that time, three studies based on a
26 similar type of respiratory symptom questionnaire administered at three different times as part of
27 the Harvard Six-City and 24-City Studies provided data on the relationship of chronic respiratory
28 disease to PM. All three studies suggest a long-term PM exposure effect on chronic respiratory
29 disease. The analysis of chronic cough, chest illness and bronchitis tended to be significantly
30 positive for the earlier surveys described by Ware et al. (1986) and Dockery et al. (1989). Using
31 a design similar to the earlier one, Dockery et al. (1996) expanded the analyses to include
32 24 communities in the United States and Canada. Bronchitis was found to be higher (odds ratio
June 2003 8-185 DRAFT-DO NOT QUOTE OR CITE
-------
1 =1.66) in the community with the highest particle strong acidity when compared with the least
2 polluted community. Fine particulate sulfate was also associated with higher reporting of
3 bronchitis (OR = 1.65, 95% CI 1.12, 2.42).
4 Interpretation of such studies requires caution in light of the usual difficulties ascribed to
5 cross-sectional studies. That is, evaluation of PM effects is based on variations in exposure
6 determined by a different number of locations. In the first two studies, there were six locations
7 and, in the third, twenty-four. The results seen in all studies were consistent with a PM gradient,
8 but it was not readily possible to separate out clear effects of PM from other factors or pollutants
9 having the same gradient.
10 Chronic pulmonary function studies by Ware et al. (1986), Dockery et al. (1989), and Neas
11 et al. (1994) had good monitoring data and well-conducted standardized pulmonary function
12 testing over many years, but showed no effect for children from airborne particle pollution
13 indexed by TSP, PM15, PM25 or sulfates. In contrast, the Raizenne et al. (1996) study of U.S.
14 and Canadian children found significant associations between FEVj and FVC and acidic
15 particles (FT). Overall, the available studies provided only limited evidence suggestive of
16 pulmonary lung function decrements being associated with chronic exposure to PM indexed by
17 various measures (TSP, PM10, sulfates, etc.). However, it was noted that cross-sectional studies
18 require very large sample sizes to detect differences because they cannot eliminate person to
19 person variation, which is much larger than the within person variation.
20
21 8.3.3.2.2 New Studies of Respiratory Effects of Long-Term Particulate Matter Exposure
22 Several studies published since 1996 evaluated effects of long-term PM exposure on lung
23 function and respiratory illness (see Appendix 8B, Table 8B-8). The new studies examining
24 PM10 and PM25 in the United States include McConnell et al. (1999), Abbey et al. (1998),
25 Berglund et al. (1999), Peters et al. (1999a,b), and Avol et al. (2001), all of which examined
26 effects in California cohorts but produced variable results. McConnell et al. (1999) noted that,
27 as PM10 increased across communities, the bronchitis risk per interquartile range also increased,
28 results consistent with those reported by Dockery et al. (1996). However, the high correlation of
29 PM10, acid, and NO2 precludes clear attribution of the McConnell et al. bronchitis effects
30 specifically to PM alone. Avol et al. (2001) reported that, for 110 children that moved to other
31 locations as a group, subjects who moved to areas of lower PM10 showed increased growth in
June 2003 8-186 DRAFT-DO NOT QUOTE OR CITE
-------
1 lung function and subjects who moved to communities with higher PM10 showed slowed lung
2 function growth.
3 Gauderman et al. (2000, 2002) presented results from a study that is both a cohort and a
4 cross-sectional study. This unique design followed two cohorts of southern California children
5 who were fourth graders in 1993 and 1996 respectively. The cohorts, located in 12 communities,
6 were followed for 4 years. A three stage model which allowed for individual slopes, within
7 community covariates, and community-wide air pollution averages, was fitted using SAS Proc
8 MIXED. Pulmonary function measurements included FVC, FEV1, MMEF, and PEFR, all of
9 which gave similar results for both PM2 5 and PM10. In the first cohort, PM10 showed a
10 significant 1.3% decrease in annual growth rates for a 51.5 |ig/m3 difference in PM10. This
11 difference was only 0.4% in the second cohort; however, the two were not significantly different
12 from each other. The effect for PM25 was slightly less for a difference of 22.2 jig/m3. Peters
13 et al. (1999b) studied the prevalence of respiratory symptoms in 12 southern California
14 communities in 1993. To estimate the relationship between symptoms and pollutants a two-
15 stage regression approach was used. The first stage estimated community-specific rates adjusted
16 for individual covariates. The second stage regressed these rates on pollutant averages from
17 1986 to 1990, finding no significant relationships between respiratory symptoms and average
18 PM10 levels.
19 In a non-U.S. PM10 study, Horak et al. (2002) conducted a combined cohort and cross-
20 sectional study similar in design to that of Gauderman et al. (2000). The cohorts were taken
21 from 975 school children in 8 communities in lower Austria between 1994-1997. Relationships
22 were estimated from generalized estimating equations adjusting for autocorrelation.
23 Adjustments were made for sex, atopy, ETS, baseline lung function, height, and site. Growth in
24 FVC and MEF were significantly related to winter PM10 levels.
25 Gehring et al. (2002) enrolled 1,756 newborn children in the Munich area. Individual
26 PM2 5 and NO2 levels were estimated from actual measurements at 40 sites combined with a GIS
27 predictor model. PM2 5 levels ranged from 11.9 to 21.9 |ig/m3. The incidence (in the first two
28 years of life) of cough without infection and dry cough at night were related to PM25 levels.
29 Wheeze, bronchitis, respiratory infections, and runny nose were not related to PM2 5 levels.
30 Other non-U.S. studies examined PM measures such as TSP and BS in European countries.
31 In Germany, Heinrich et al. (2000) reported a cross-sectional survey of children, conducted
June 2003 8-187 DRAFT-DO NOT QUOTE OR CITE
-------
1 twice (with the same 971 children included in both surveys). TSP levels decreased between
2 surveys as did the prevalence of all respiratory symptoms (including bronchitis). Also, Kramer
3 et al. (1999) reported a study in six East and West Germany communities, which found
4 decreasing yearly TSP levels to be related to ever-diagnosed bronchitis from 1991-1995. Lastly,
5 Jedrychowski et al. (1999) reported an association between both BS and SO2 levels in various
6 areas of Krakow, Poland, and slowed lung function growth (FVC and FEVj).
7 Leonard! et al. (2000) studied a different health outcome measure as part of the Central
8 European Air Quality and Respiratory Health (CESAR) study. Blood and serum samples were
9 collected from school children ages 9-11 yrs. in each of 17 communities in Central Europe
10 (N =10 to 61 per city). Numbers of lymphocytes increased as PM concentrations increased
11 across the cities. Regression slopes, adjusted for confounder effects, were largest and
12 statistically significant for PM2 5, but small and non-significant for PM10_2 5. A similar positive
13 relationship was found between IgG concentration in serum and PM2 5 gradient, but not for PM10
14 or PM10_2 5. These results tend to suggest a PM effect on immune function more strongly due to
15 ambient fine particle than coarse particle exposure.
16
17 8.3.3.2.3 Summary of Long-Term Particulate Matter Exposure Respiratory Effects
18 The methodology used in the long-term studies varies much more than the methodology in
19 the short-term studies. Some studies reported highly significant results (related to PM) while
20 others reported no significant results. The cross-sectional studies are often confounded, in part,
21 by unexplained differences between geographic regions. The studies that looked for a time trend
22 are also confounded by other conditions that were changing over time. The newer studies that
23 combine the features of cross-sectional and cohort studies provide the best evidence for chronic
24 effects. These studies include Peters et al. (1999b), Gauderman et al. (2000), and Gauderman
25 et al. (2002). The Gauderman studies found significant decreases in lung function growth among
26 So. California school children to be related to PM10 levels. However, Peters et al. (1999b) found
27 no relationship between respiratory symptoms and annual average PM10 levels in 12 So.
28 California communities.
29 The cross-sectional studies by Dockery et al. (1996) and Raizenne et al. (1996), assessed
30 before in the previous 1996 PM AQCD, found differences in peak flow and bronchitis rates
31 associated with fine particle acidity.
June 2003 8-188 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4 DISCUSSION OF EPIDEMIOLOGIC STUDIES OF HEALTH
2 EFFECTS OF AMBIENT PARTICULATE MATTER
3 8.4.1 Introduction
4 Numerous PM epidemiology studies assessed in the 1996 PM AQCD implicated ambient
5 PM as a likely contributor to mortality and morbidity effects associated with ambient air
6 pollution exposures. Since preparation of the 1996 PM AQCD, the epidemiologic evidence
7 concerning ambient PM-related health effects has vastly expanded. Past regulatory decisions
8 have been important in the selection of PM indices and evolution of PM epidemiologic literature.
9 That is, the adoption of PM10 standards in 1987 and of PM25 standards in 1997 have generated
10 ambient air concentration databases that have made it possible for research to address many
11 previously unresolved issues regarding possible linkages between airborne PM and human
12 health; and the newly authorized nationwide network of speciation samplers holds promise for
13 further advances regarding identification of the most influential specific components of the
14 ambient air pollution mixture and their sources.
15 As was discussed in Sections 8.2 and 8.3, numerous new PM epidemiology studies, both of
16 short-term and long-term PM exposure, have yielded findings indicating that statistically
17 significant excess risks for various mortality and/or morbidity endpoints in many U.S. cities and
18 elsewhere are associated with ambient PM indexed by a variety of ambient community
19 monitoring methods.
20 Still, several uncertainties discussed in the 1996 PM AQCD continue to be important in
21 assessing and interpreting the overall PM epidemiology database and its implications for
22 estimating risks associated with exposure to ambient PM concentrations in the United States:
23 (1) potential confounding of PM effects by co-pollutants (especially major gaseous pollutants
24 such as O3, CO, NO2, SO2); (2) the attribution of PM effects to specific PM components (e.g.,
25 PM10, PM10_2 5, PM2 5, ultrafmes, sulfates, metals, etc.) or source-oriented indicators (motor
26 vehicle emissions, vegetative burning, etc.); (3) the temporal relationship between exposure and
27 effect (lags, mortality displacement, etc.); (4) the general shape of exposure-response
28 relationship(s) between PM and/or other pollutants and observed health effects (e.g., potential
29 indications of thresholds for PM effects); and (5) the consequences of measurement error. All of
30 these modeling issues are of much importance and interest in selection of appropriate statistical
31 models for characterizing and interpreting ambient PM-health effects associations.
June 2003 8-189 DRAFT-DO NOT QUOTE OR CITE
-------
1 Assessing the above uncertainties in relation to the PM epidemiology data base remains a
2 challenge. The basic issue is that there are an extremely large number of possible models, any of
3 which may turn out to give the best statistical "fit" of a given set of data, and only some of which
4 can be dismissed a priori as biologically or physically illogical or impossible, except that
5 putative cause clearly cannot follow effect in time. Most of the models for daily time-series
6 studies are fitted by adjusting for changes over long time intervals and across season, by day of
7 week, weather, and climate. Many of the temporal and weather variable models have been fitted
8 to data using semi-parametric methods such as spline functions or local regression smoothers
9 (LOESS). The goodness of fit of these base models has been evaluated by criteria suitable for
10 generalized linear models (GLM) with Poisson or hyper-Poisson responses (number of events)
11 with a log link function, particularly the Akaike Information Criterion (AIC) and the more
12 conservative Bayes information criterion (BIC), which adjust for the number of parameters
13 estimated from the data. The Poisson over-dispersion index and the auto-correlation of residuals
14 are also often used. It is often assumed, but rarely proven, that the best-fitting models with PM
15 would be models with the largest and most significant PM indices. However, if high correlations
16 between PM and one or more gaseous pollutants emitted from a common source (e.g., motor
17 vehicles) exist in a given area, then disentangling their relative individual partial contributions to
18 observed health effects associations becomes very difficult. There have been very few attempts
19 at broad, systematic investigations of the model selection issue and little reporting of goodness-
20 of-fit criteria among competing models that represent one approach by which to assess or
21 compare models.
22 Substantial prior knowledge to guide model fitting now exists and an informed modeling
23 strategy can yield a useful set of models as one type of sensitivity analysis. To illustrate, a
24 systemic evaluation of model choice has been carried out by Clyde et al. (2000), using Bayesian
25 Model Averaging for the same Birmingham, AL, data as analyzed by Smith et al. (2000).
26 Several different calibrated information criterion priors were tried in which models with large
27 numbers of parameters are penalized to various degrees. After taking out a baseline trend
28 (estimated using a GLM estimate with a 30-knot thin-plate smoothing spline), 7,860 models
29 were selected for use in model averaging. These included lags 0-3 days of a daily monitor PM10,
30 an area-wide average PM10 value with the same lags, temperature (daily extremes and average)
31 lagged 0-2 days, humidity (dewpoint, relative humidity min and max, average specific humidity)
June 2003 8-190 DRAFT-DO NOT QUOTE OR CITE
-------
1 lagged 0-2 days, and atmospheric pressure, lagged 0-2 days. The model choice is sensitive to the
2 specification of calibrated information criterion priors, in particular disagreeing as to whether
3 different PM10 variables should be included or not. For example, one or another PM10 variable is
4 included in all the top 25 Akaike Information Criterion (AIC) models, but only in about 1/3 of
5 the top Bayes Information Criterion (BIC) models. Both approaches give a relative risk estimate
6 of about 1.05, with credibility intervals of (0.94, 1.17) for the AIC prior and (0.99, 1.11) for the
7 BIC prior. A validation study in which randomly selected data were predicted using the
8 different priors favored Bayesian model averaging with BIC prior over model selection (picking
9 the best model) with BIC or any approach with AIC. This type of modeling may represent
10 another type of multi-pollutant modeling approach in addition to more typical hypotheses-driven
11 model construction and interpretation that draws more on external information (e.g., exposure,
12 dosimetric, toxicologic relationships) in specifying models and interpreting their results.
13 The possibility that an observed effect is "real" (i.e., likely to be found in an independent
14 replication of the study) or merely a statistical artifact is usually characterized by its confidence
15 interval or by its estimated significance level. In most of this document, confidence intervals, or
16 credible intervals for Bayesian analyses, are reported in order to emphasize that the effect size is
17 not known with certainty, but some values are more nearly consistent with the data than effect
18 size values outside the interval. P-values or t-values are implicitly associated with a null
19 hypothesis of no effect. A nominal significance level of p < 0.05 or 5% (i.e., a 95% confidence
20 interval) is usually used as a guide for the reader, but P-values should not be used as a rigid
21 decision-making tool. If the observed confidence intervals were arrived at by a number of prior
22 model specification searches, eliminating some worse fitting models, the true interval may well
23 be wider.
24 Given the now extremely large number of published epidemiologic studies of ambient PM
25 associations with health effects in human populations and the considerably wide diversity in
26 applications of even similar statistical approaches (e.g., "time-series analyses" for short-term PM
27 exposure effects), it is neither feasible nor useful here to try to evaluate the methodological
28 soundness of every individual study. Rather, a three-pronged approach is likely to yield useful
29 evaluative information: (1) an overall characterization of evident general commonalities (and/or
30 notable marked differences) among findings from across the body of studies dealing with
31 particular PM exposure indices and types of health outcomes, looking for convergence of
June 2003 8-191 DRAFT-DO NOT QUOTE OR CITE
-------
1 evidence regarding types of effects and effect-sizes attributable to ambient PM indices across
2 various methodologically acceptable analyses; (2) thorough, critical assessment of newly
3 published multi-city analyses of PM effects, assuming that greater scientific weight is generally
4 ascribable to their results than those of smaller-sized studies (often of individual cities) yielding
5 presumably less precise effect size estimates; and (3) evaluation of coherence of the findings
6 among different types of effects and across various geographic locations, as well as with other
7 types of pertinent biological information (e.g., exposure, dosimetry, toxicity, etc.).
8 In the sections that follow, issues noted above are critically discussed. In addition, given
9 that both the newer multi-city study results and those of newer single-city analyses tend to show
10 evidence of somewhat greater geographical heterogeneity in estimated PM risks across cities and
11 regions than had been seen in studies assessed in the 1996 PM AQCD, the issue of geographical
12 heterogeneity in PM effect estimates is further evaluated here.
13 First follows a discussion of the GAM issue and a summary of some key findings emerging
14 from the short communications and peer-review commentary recently published by HEI (2003).
15
16 8.4.2 GAM Issue and Reanalyses Studies
17 As discussed earlier, Dominici et al. (2002) reported that the default convergence criteria
18 used in the S-Plus function GAM may not guarantee convergence to the best unbiased estimate
19 in all cases. The actual importance of this effect has only recently begun to be quantified, the
20 results of recent reanalyses of many key studies being especially helpful in this regard; those
21 reanalyses are described in short communicatons published in the HEI (2003b) Special Report.
22 As for the net outcome of these reanalyses efforts, HEI (2003b) summarizes it well, as follows:
23
24 Overall, the revised analyses using GAM with more stringent convergence criteria and
25 iterations and GLM-natural splines resulted in lower estimates, but largely confirmed the
26 effect of exposure to particulate matter on mortality (Burnett and Goldberg, 2003; Dominici
27 et al., 2003; Katsouyanni et al., 2003; Samoli et al. ,2003; Schwartz, 2003b; Zanobetti and
28 Schwartz, 2003a) and morbidity, especially for hospitalizations for cardiovascular and
29 respiratory diseases (Atkinson et al., 2003; Fairley, 2003; Gold et al., 2003; Hoek, 2003; Ito,
30 2003; Le Tertre et al., 2003; Ostro et al., 2003; Schwartz, 2003a; Sheppard, 2003; Zanobetti
31 and Schwartz, 2003b). As in earlier analyses, the effect was more pronounced among
32 individuals 65 years of age and older (Fairley; Gold et al.; Goldberg and Burnett; Ito; Le
33 Tertre et al.; Mar et al.; Mooigavkar; Schwartz a). The impact of various sensitivity analyses,
June 2003 8-192 DRAFT-DO NOT QUOTE OR CITE
-------
1 when these were performed, differed across the studies. No significant impacts were seen in
2 some (Ostro et al.), whereas in others, alternative modeling of time (Klemm and Mason;
3 Moolgavkar) and weather factors (Goldberg and Burnett; Ito) resulted in substantial changes.
4
5 The following discussion evaluates in more detail the nature and extent of potential
6 problems in the various studies that have used the GAM default algorithm, but which have also
7 had their analyses redone using alternative methods unaffected by this convergence issue.
8
9 8.4.2.1 Impact of Using the More Stringent GAM Model on PM Effect Estimates
10 for Mortality
11 Many of the reanalysis studies analyzed associations between PM10 and mortality, allowing
12 an examination of the impact of GAM convergence problem on this PM index. Table 8-34 and
13 Figure 8-17 shows the percent excess total non-accidental mortality (unless noted otherwise) risk
14 estimates per 50 |ig/m3 increase in PM10 derived from the reanalysis studies for (1) GAM with
15 default convergence criteria; (2) GAM with stringent convergence criteria; and, (3) GLM with
16 natural splines that approximate the original GAM model. The figure shows results only from
17 the studies that used all of the three alternative models for PM10. It can be seen that most, but
18 not all, reanalyses resulted in reductions in PM10 risk estimates when more stringent convergence
19 criteria were used in GAM models. Using GLM with natural splines resulted in additional
20 reduction in PM10 risk estimates for most, but not all, cases. The extent of reductions in PM10
21 risk estimates in GAM with more stringent convergence criteria or GLM with natural splines
22 was in most cases less than 1% excess deaths per 50 |ig/m3 increase in PM10. Obviously, the
23 relative reduction is greater for the studies that had smaller PM10 risk estimates in the original
24 analyses (e.g., NMMAPS U.S. 90 cities analyses). It can also be seen from Figure 8-17 that the
25 extent of reduction in PM10 risk estimates is smaller compared to the variability of PM10 risk
26 estimates across the studies. Thus, the effect of the GAM convergence problem does not appear,
27 in most cases, to be substantial. Potential factors affecting the heterogeneity of PM10 risk
28 estimates across studies are discussed in later sections. Several of the reanalysis reports also
29 analyzed PM25 and PM10_25. Generally, the pattern and extent of reductions in mortality risk
30 estimates were similar to those for PM10. The results and a comparison of PM25 and PM10_25
31 mortality risk estimates are presented in a later section.
32
June 2003 8-193 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-34. PM10 EXCESS RISK ESTIMATES FROM REANALYSIS STUDIES
FOR TOTAL NON-ACCIDENTAL MORTALITY PER 50 jig/m3 INCREASE IN PM10
Study
NMMAPS 90-cities; Dominici et al. (2002)
Harvard 6-cities; Klemm and Mason (2003)
US 10 cities; Schwartz (2003b)
8 Canadian cities; Burnett and Goldberg (2003)
APHEA2; Katsouyanni et al. (2003)
Santa Clara Co.; Fairley (2003)
Coachella Valley; Ostro et al. (2003)*
Los Angeles Co.; Moolgavkar (2003)
Cook Co.; Moolgavkar (2003)
Phoenix, AZ; Mar et al. (2003)*
Detroit, '85-'90; Ito (2003)
Detroit, '92-'94; Ito (2003)
The Netherlands; Hoek (2003)
Erfurt, Germany; Stolzel et al. (2003)
GAM-default
2.1(1.6,2.6)
4.1(2.8,5.4)
3.4(2.7,4.1)
4.5 (2.2, 6.7)
3.5(2.9,4.1)
8.0
(no interval given)
5.6 (1.7, 9.6)
2.4 (0.5, 4.4)
2.4(1.3,3.5)
9.9 (1.9, 18.4)
1.7 (0.2, 3.2)
4.4 (-1.0, 10.1)
0.9(0.1, 1.7)
6.4 (0.3, 12.9)
GAM-stringent
1.4 (0.9, 1.9)
3.6(2.1,5.0)
3.3 (2.6,4.1)
3.6(1.4,5.8)
3.3 (2.8, 3.9)
7.8(2.8, 13.1)
5.5(1.6,9.5)
2.4 (0.5, 4.3)
2.6(1.6,3.6)
9.7(1.7, 18.3)
0.9 (-0.5, 2.4)
3.3 (-2.0, 8.9)
0.9 (0.2, 1.7)
6.2(0.1, 12.7)
GLM
1.1 (0.5, 1.7)
2.0 (0.3, 3.8)
2.8 (2.0, 3.6)
2.7 (-0.1, 5.5)
2.1(1.5,2.8)
8.3 (2.9, 13.9)
5.1(1.2,9.1)
2.3 (0.1,4.5)
2.6(1.5,3.7)
9.5 (0.6, 19.3)
0.7 (-0.8, 2.1)
3.1 (-2.2,8.7)
0.9(0.1, 1.7)
5.3 (-1.8, 12.9)
* Cardiovascular Mortality
1 Dominici et al. (2002) also illustrated that GAM models, even with stringent convergence
2 criteria, still result in biased (downward) standard errors of regression coefficients. This was the
3 main reason for the use of GLM with natural splines in the reanalysis studies. As can be seen
4 from Figure 8-17, the 95% confidence bands are somewhat wider for GLM results than for GAM
5 results in some, but not all cases. However, the extent of wider confidence bands is not
6 substantial in most cases (the bias ranged from a few percent to -15% in most cases). It should
7 be noted that, while a GLM model with natural splines provides correct standard error of
8 regression coefficient, it is not equivalently as flexible as LOESS or smoothing splines. Unlike
9 LOESS or smoothing splines, natural splines fit linearly at both ends of the data span. Natural
10 splines therefore may not be an ideal model option for temperature effects, for which the slopes
11 are likely non-linear (especially at the higher end). Goldberg and Burnett (2003), in their
12 reanalysis of Montreal data, discussed related issues. In their reanalysis, the originally reported
June 2003
8-194
DRAFT-DO NOT QUOTE OR CITE
-------
% excess deaths per 50 |jg/m3 increase in PM-|Q
-2 0 2 4 6 8 10
US 90 cities (1)
Dominic! et al. (2002)
Harvard 6 cities (01)
Klemm and Mason (2003)
US 10 cities (01)
Schwartz (2003b)
Canadian 8 cities (1)
Burnett and Goldberg (2003)
APHEA: 21 cities (01)
Katsouyanni et al. (2003)
Santa Clara Co., CA (0)
Fairley (2003)
Coachella Valley, CA (0)*
Ostro et al. (2003)
Los Angeles Co., CA (2)
Moolgavkar (2003)
Cook Co., IL (0)
Moolgavkar (2003)
Phoenix, AZ (0)*
Mar et al. (2003)
Detriot, Ml: '85 -'90 (1)
Ito (2003)
Detriot, Ml: '92 -'94 (1)
Ito (2003)
The Netherlands (1)
Hoek (2003)
Erfurt, Germany (0)
Stolzel et al. (2003)
Multi - city studies
Single - city studies
-e-
-e-
Figure 8-17. PM10 excess risk estimates for total non-accidental mortality for numerous
locations (and for cardiovascular mortality!*] for Coachella Valley, CA and
Phoenix, AZ), using: (1) GAM with default convergence criteria (white
circle); (2) GAM with stringent convergence criteria (black circle); and,
(3) GLM/natural splines (x) that approximate the original GAM model from
the GAM reanalysis studies. The numbers in parenthesis indicate lag days
used ("01" is average of 0 and 1 day lags).
June 2003
8-195
DRAFT-DO NOT QUOTE OR CITE
-------
1 risk estimates of PM indices (CoH, extinction coefficient, predicted PM2 5, and sulfate) were
2 greatly attenuated in the GLM model with natural splines. One of the alternative explanations
3 for these results was that the natural spline does not fit the possibly non-linear (threshold) effect
4 of temperature as well as non-parametric smoothers. Hoek (2003), in his reanalysis of the
5 Netherlands data, also showed that, compared to GAM models, GLM/natural spline models
6 resulted in larger deviance, indicating poorer fits. Thus, there are remaining issues regarding the
7 trade-off between GAM/non-parametric smoothers and GLM/parametric smoothers. The
8 GLM/natural splines may produce correct standard errors but cannot guarantee "correct" model
9 specifications. More recently, Dominici et al. (2003) developed and published a GAM routine
10 for SPlus that gives correct standard errors, but it was not developed in time to be used for the
11 GAM reanalysis effects reported on in HEI (2003b).
12 Three reanalysis reports applied alternative smoothing approaches (e.g., penalized splines)
13 that, as with GLM/natural splines, did not have the problem of biased standard error. These
14 studies were: reanalyses of Harvard six cities data by Schwartz (2003a); reanalysis of 10 US
15 cities data by Schwartz (2003b); and reanalysis of APHEA2 by Katsouyanni et al. (2003).
16 Generally, as with GLM/natural splines, the use of alternative smoothing approaches resulted in
17 smaller PM risk estimates than GAM with stringent convergence criteria. In the re analysis of
18 APHEA2 study, the PM10 risk estimates from penalized splines were smaller than those from
19 GAM model, but larger than those from natural splines. Three alternative smoothing approaches
20 (B-splines, penalized splines, and thin-plate splines) used in the reanalysis of Harvard six cities
21 PM2 5 data resulted in generally smaller risk estimates than those from natural splines. As was
22 expected, all of these alternative smoothing approaches resulted in standard errors that were
23 comparable to those from natural splines but larger than those from GAM models.
24 Several of the GAM reanalysis reports included additional sensitivity analyses which
25 provided useful information. These sensitivity analyses included examinations of the effect of
26 changing degrees of freedom for smoothing of temporal trends and weather variables (Dominici
27 et al. [2002]; Ito [2003]; Klemm and Mason [2003]; Moolgavkar [2003]; and Burnett and
28 Goldberg [2003]). In these analyses, changing the degrees of freedom for smoothing of
29 temporal trends or weather effects often resulted in change of PM coefficients to a similar or
30 even greater extent than those caused by the GAM convergence problem. A distinctly less well
31 investigated issue is the effect of the use of different weather model specifications (i.e., how
June 2003 8-196 DRAFT-DO NOT QUOTE OR CITE
-------
1 many weather variables and their lags are included). In a limited examination of this issue in the
2 reanalysis of Detroit data (Ito, 2003), a weather model specification similar to that used in the
3 US 90 cities consistently resulted in smaller PM10 risk estimates than a weather model similar to
4 that used in Harvard six cities study.
5 In summary, the results from the GAM reanalysis studies indicate that PM risk estimates
6 from GAM models were often, but not always, reduced when more stringent convergence
7 criteria were used. However, the extent of the reduction was not substantial inmost cases. The
8 variability of PM risk estimates due to the model specification, including the number of weather
9 terms and extent of smoothing, is likely larger than the effect of the GAM convergence problem.
10 The extent of downward bias in standard error reported in these data (a few percent to -15%)
11 also appears not to be very substantial, especially when compared to the range of standard errors
12 across studies due to differences in population size and numbers of days available. Still, the
13 discussions in this chapter focus mainly on the reanalyzed studies or the studies that did not use
14 GAM with default convergence criteria, because the extent of the effect of this problem is not
15 always predictable in each individual study.
16
17 8.4.2.2 Impact of Using the More Stringent GAM Model on PM Effect Estimates for
18 Respiratory Hospital Admissions
19 The NMMAPS multi-city study (Samet et al., 2000a,b) of PM10 concentrations and hospital
20 admissions used the default GAM model specification with multiple smooths. To be
21 quantitative in terms of the change that results from the more stringent GAM criteria,
22 Figure 8-18 shows a plot of the respiratory models for which Zanobetti and Schwartz (2003b)
23 provided reanalyses. These results indicate that there was only about a 14% decline in the effect
24 estimates associated with use of the more appropriate stringent convergence requirement.
25 Moreover, it is clear that the two estimates are well within the 95% confidence interval of each
26 other, indicating that the two models are not statistically significantly different from one another.
27 To examine the potential influence of the GAM convergence specification on the results of
28 the original Detroit data analysis by Lippmann et al. (2000), the associations between PM
29 components and daily mortality/morbidity were re-examined by Ito using more stringent
30 convergence criteria, as well as by applying a GLM that approximated the original GAM models
31 (Ito, 2003). Generally, the GAM models with stringent convergence criteria and GLM models
June 2003 8-197 DRAFT-DO NOT QUOTE OR CITE
-------
S * 5j
S ^
en o
a:
4-
3-
2-
1-
0
0
= 0.86x
= 0.932
Original GAM % Increase
per 10 |jg/ni3
Figure 8-18. Comparison of GAM results for original (default) convergence case versus
those from reanalyses with a more stringent convergence criterion (10e-15)
for constrained lag respiratory model cases. Note very high overall
correlation (r = 0.932) of original default GAM values with reanalysis
stringent GAM results and slightly greater divergence from r2 = 1.0 (dotted
line) as excess risk values per 10 ug/m3 PM10 increase.
Source: Derived from Zanobetti and Schwartz (2003b).
1 resulted in somewhat smaller estimated relative risks than those reported in the original study,
2 but the reduction is quite small (averaging 17% less for the stringent GAM case versus default).
3 For COPD, the decrease associated with the more stringent convergence criteria is larger
4 (averaging 30%). Overall, for all types of hospital admissions (including pneumonia, COPD and
5 ischemic heart disease) the effect of the change to the more stringent GAM gave an average
June 2003
8-198
DRAFT-DO NOT QUOTE OR CITE
-------
1 decrease of 20 percent, while a switch to the GLM model specification gave an average 29%
2 decrease in estimated PM effect size.
3 As discussed earlier, Sheppard (2003) recently conducted a reanalysis of their non-elderly
4 hospital admissions data for asthma in Seattle, WA, in order to evaluate the effect of the fitting
5 procedure on their previously published analyses. A lag of 1 day was used for all PM models.
6 As shown in Table 8-35, the results were provided in the manuscript to only one significant
7 figure (to the nearest whole percent), making the calculation of percent changes between models
8 problematic, since the rounding of the effect estimates are nearly of the order of the size of the
9 effect estimate changes. However, it can be seen that the pattern of changes in effects estimates
10 and 95% CI values is similar to that seen in other studies.
11
12
TABLE 8-35. COMPARISON OF MAXIMUM SINGLE DAY LAG EFFECT
ESTIMATES FOR PM2 5, PM2 510, and PM10 FOR SEATTLE ASTHMA HOSPITAL
ADMISSIONS BASED ON ORIGINAL GAM ANALYSES USING DEFAULT
CONVERGENCE CRITERIA VERSUS REANALYSES USING GAM WITH MORE
STRINGENT CONVERGENCE CRITERIA AND GLM
PM25
PM2.5.10
PM10
Original Default GAM
Model* % Increase/IQR
(95% CI)
4 (2, 7)
4(1,7)
5 (2, 8)
Reanalysis Stringent GAM
% Increase/IQR
(95% CI)
4(1,6)
2 (0, 5)
4(1,7)
Reanalysis GLM (Natural
Spline) % Increase/IQR
(95% CI)
3 (1, 6)
2 (-1,4)
3 (0, 6)
*PM25 IQR=11.8 ug/m3; PM25.10 IQR = 9.3 ug/m3; PM10IQR = 19 ug/m3.
Source: Derived from Sheppard (2003).
1 Further evidence of the relatively small effect of the default convergence criteria issue in
2 most applications is the recent work by Moolgavkar (2003), in which he reanalyzed his earlier
3 GAM analyses of hospital admissions for COPD (Moolgavkar, 2000c) for the cities of Los
4 Angeles (Los Angeles County) and Chicago (Cook County). In his original publication,
5 Moolgavkar found ca. 5.0% excess risk for COPD hospital admissions among the elderly (64+
6 yr) in Los Angeles to be significantly related to both PM2 5 and PM10_2 5 in one pollutant models.
7 In the same study, similar magnitudes of excess risk (i.e., in the range of ca. 4 to 7%) were found
June 2003 8-199 DRAFT-DO NOT QUOTE OR CITE
-------
1 in one-pollutant models to be associated with PM25 or PM10_2 5 for other age groups (0-19 yr; 20-
2 64 yr) in Los Angeles, as well. In his reanalyses of these GAM results using the more stringent
3 convergence criteria, however, Moolgavkar (2003) combined all three Los Angeles age groups
4 into one analysis, providing greater power, but also complicating before/after comparisons as to
5 the actual effect of using the more stringent convergence criteria on the results. In the case of
6 the Cook County analyses, the author changed other model parameters (i.e., the number of
7 degrees of freedom in the model smooths) at the same time as implementing the more stringent
8 convergence criteria, so direct before/after comparisons were not possible for Moolgavkar's
9 Chicago reanalyses.
10 Therefore, in order to provide a one-to-one comparison for Los Angeles, the original age-
11 specific GAM analyses have been pooled using inverse variance weighting and are presented
12 along with Moolgavkar's (2003) reanalyses results (in terms of a % increase per 10 |ig/m3 mass
13 increase for both PM2 5 and PM10) in Table 8-36. As shown in that table, the Moolgavkar Los
14 Angeles results for all-age COPD admissions for the original and the more stringent convergence
15 criteria GAM cases (using the same degrees of freedom) are very similar, with the effects
16 estimate either decreasing (for PM25) or increasing (for PM10) very slightly. In those cases
17 where a much larger number of degrees of freedom were used with either the more stringent
18 GAM model or a natural spline GLM model, larger reductions in effects estimates were obtained
19 as compared to the original GAM model. For the same number of degrees of freedom, the
20 natural spline model resulted in either a slightly larger (for PM2 5) or a slightly smaller (for PM10)
21 effects estimate than the stringent GAM model. Thus, these reanalysis results indicate that the
22 use of the more stringent GAM convergence criteria results in minimal changes to the size of the
23 PM effect estimates in this case, as compared to those obtained using the default GAM model,
24 whereas the number of degrees of freedom used with either GAM or GLM models can result in
25 much larger changes in the size of the PM effects estimates. More specifically, use of the much
26 larger number of degrees of freedom results in a much less efficient estimate of the pollutant
27 effect.
28 These various reanalyses results therefore confirm that the PM effect estimates generally
29 do decline somewhat when using the more stringent convergence criteria, as compared to the
30 default GAM, with the new estimates being well within the confidence interval of the original
31 estimates. In addition, the effect of using a more stringent convergence criteria was indicated to
June 2003 8-200 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-36. COMPARISON OF LOS ANGELES COPD HOSPITAL ADMISSIONS
MAXIMUM SINGLE DAY LAG EFFECT ESTIMATES FOR PM2 s and PM10
FROM THE ORIGINAL GAM ANALYSES USING DEFAULT CONVERGENCE
CRITERIA VERSUS FOR REANALYSES USING MORE STRINGENT
CONVERGENCE CRITERIA AND FOR MODELS SMOOTHED WITH
MORE DEGREES OF FREEDOM
Original Default Reanalysis Stringent Reanalysis Stringent Reanalysis Natural
GAM Model* (30df) GAM (30df) GAM(lOOdf) Spline (lOOdf)
% Increase/10 ug/m3 % Increase/ 10ug/m3 % Increase/ 10ug/m3 % Increase/10 ug/m3
(95% CI) (95% CI) (95% CI) (95% CI)
PM25 1.90(0.97-2.84)** 1.85(0.82-2.89)** 1.38(0.51-2.25)*** 1.49(0.41-2.58)***
PM10 1.43(0.85-2.02)** 1.51(0.85-2.18)** 1.08(0.50-1.66)** 0.98(0.24-1.72)**
*Original GAM estimates derived for "all ages" from original analyses by age subgroups using inverse variance
weights.
**For (maximum) lag case = 2 days.
***For (maximum) lag case = 0 days.
Source: Derived from Moolgavkar (2000c) and Moolgavkar (2003).
1 have less influence on the effect estimate than potential investigator-to-investigator variations in
2 model specifications (e.g., extent of smoothing) can have. Overall, the absolute effect was
3 relatively small, and the basic direction of effect and conclusions regarding the significance of
4 the PM effect on hospital admissions remained unchanged in these analyses when the GAM
5 convergence requirement was made more stringent.
6
7 8.4.2.3 HEI Commentaries
8 The HEI Special Report (2003a,b) presents the HEI Special Panels' reviews of both the
9 Revised Analyses of the National Morbidity, Mortality, and Air Pollution Study, Part II
10 (NMMAPS) and the Revised Analyses of Selected Time-Series Studies, which includes short
11 communication reports presenting results from other revised analyses of original articles and
12 reports. Beyond looking at the results of reanalyses designed specifically to address problems
13 associated with the use of default convergence criteria in the S-Plus GAM function, the reviews
14 also identified issues associated with the sensitivity of study findings to the use of alternative
15 modeling approaches that some investigators employed in their reanalyses. In general, the
16 Special Panels concluded that the original PM effects estimates were more sensitive to the
June 2003 8-201 DRAFT-DO NOT QUOTE OR CITE
-------
1 modeling approach used to account for temporal effects and weather variables than to the
2 convergence criteria used in the GAM model.
3 A modeling issue of particular importance highlighted by HEI (2003b) is the sensitivity of
4 all models (e.g., GAM, GLM-natural splines, GLM-penalized splines) to the degrees of freedom
5 allotted to potentially confounding weather variables and time. The commentary discusses the
6 trade-off involved in selecting the number of degrees of freedom for time and weather variables,
7 while recognizing that there remains no altogether satisfactory way to choose the most
8 appropriate degrees of freedom. For example, in considering the effect of temperature, if the
9 degrees of freedom in the smoothing function for temperature are overly restricted, some actual
10 nonlinear effects of temperature would be falsely ascribed to the pollution variable. To avoid
11 this, the analyst is tempted to afford many degrees of freedom to temperature or other potentially
12 confounding variables. However, if more degrees of freedom are allotted than needed, such that
13 the temperature smooth function is more "wiggly" than the true dose response function, then the
14 result will be a much less efficient estimate of the pollutant effect. This would have the effect of
15 incorrectly ascribing part of the true pollution effect to the temperature variable, which would
16 compromise our ability to detect a true but small pollution effect. The commentary notes that
17 the empirical data cannot determine the optimal trade-off between these conflicting needs, and it
18 is difficult to use an a priori biological or meteorologic knowledge to determine the optimal
19 trade-off. Thus, the Special Panel generally recommends further exploration of the sensitivity of
20 these studies to a wider range of alternative degrees of smoothing and to alternative
21 specifications of weather variables in time-series models.
22 More specifically, the Specials Panels offered the following conclusions and
23 recommendations:
24
25 NMMAPS Revised Analyses
26 Dominici et al. (2002) conducted a range of revised analyses, applying alternative methods
27 to correct shortcomings in the S-Plus GAM programming. HEI's Special Panel review (HEI,
28 2003a) of this revised analyses yielded the following conclusions:
29 • While estimates of effect are quantitatively smaller than those in the original studies, a
statistically significant overall effect of PM10 on mortality remains, and the qualitative
conclusions that were initially drawn from NMMAPS remain unchanged.
June 2003 8-202 DRAFT-DO NOT QUOTE OR CITE
-------
1 • While the alternative approaches used to model temporal effects in the revised NMMAPS
analyses addressed the problems of obtaining incorrect effect estimates and standard errors
when using the preprogrammed GAMs software, no models can be recommended at this
time as being strongly preferred over another for use in this context.
2 • While formal tests of PM effect across cities did not indicate evidence of heterogeneity
because of the generally large individual-city effect standard errors, the power to assess the
presence of heterogeneity was low. The possibility of heterogeneity still exists.
3 • The appropriate degree of control for time in these time-series analyses has not been
determined. Thus, the impact of more aggressive control for time should continue to be
explored and studies to evaluate bias related to the analytic approach to smoothing and the
degree of smoothing should be encouraged.
4 • Weather continues to be a potential confounder of concern, such that further work should
be done on modeling weather-related factors.
5
6 Revised Analyses for Other Short Communications
7 Based on its review, the HEI Special Panel (HEI, 2003b) reached the following
8 conclusions:
9 • As was the case with the findings of the original studies, the revised findings will continue to
help inform regulatory decisions regarding PM.
10 • The PM effect persisted in the majority of studies, however, the number of studies showing
an adverse effect of PM was slightly smaller.
11 • In some of the large number of studies in which the PM effect persisted, the estimates of PM
effect were substantially reduced.
12 • In the few studies in which further sensitivity analyses were performed, some showed
marked sensitivity of the PM effect estimate to the degree of smoothing and/or the
specification of weather.
13 • The use of more appropriate convergence criteria on the estimates of PM effect in the
revised analyses produced varied effects across the studies. In some studies, stricter
convergence criteria had little impact, and in a few the impact was substantial. No study's
June 2003 8-203 DRAFT-DO NOT QUOTE OR CITE
-------
conclusions changed in a meaningful way by the use of stricter criteria compared to the
original analyses.
14 • In most studies, parametric smoothing approaches used to obtain correct standard errors of
the PM effect estimates produced slightly larger standard errors than the GAM. However,
the impact of these larger standard errors on level of statistical significance of the PM effect
was minor.
15 • For the most part, the original PM effect estimates were more sensitive to the method used to
account for temporal effects than to changing the convergence criteria.
16 • Even though the alternative approaches used to model temporal effects in the revised
analyses addressed the problems of obtaining incorrect effect estimates and standard errors
when using the GAMs software, none can be recommended at this time as being strongly
preferred over another for use in this context.
17 • Neither the appropriate degree of control for time nor the appropriate specification of the
effects of weather in these time-series analyses has been determined. This awareness
introduces a degree of uncertainty that has not been widely appreciated previously, such that
the sensitivity of these studies to a wider range of alternative degrees of smoothing and
alternative specifications of weather variables in time-series models should continue to be
explored.
18
19 8.4.3 Assessment of Confounding by Co-Pollutants
20 8.4.3.1 Introduction
21 Airborne particles are found among a complex mixture of atmospheric pollutants, some of
22 which are well measured (such as gaseous criteria co-pollutants O3, CO, NO2, SO2) and others
23 which are not routinely measured. The basic question here is one of determining the extent to
24 which observed health effects can be attributed to airborne particles acting alone or in
25 combination with other air pollutants. Many of the pollutants are closely correlated due to
26 emissions by common sources and dispersion by common meteorological factors, so that it may
27 be difficult to disentangle their effects (as noted in Section 8.1.1), because some are in the
28 pathway of formation of other pollutants (e.g., NO —> NO2 —^NO3"1 —> Particle Mass).
June 2003 8-204 DRAFT-DO NOT QUOTE OR CITE
-------
1 It is widely accepted that some PM metrics are associated with health effects, and that PM
2 has effects independent of the gaseous co-pollutants. The extent to which ambient gaseous
3 co-pollutants may have health effects independent of PM is important in considering the extent
4 to which health effects attributed to PM may actually be due in part to co-pollutants or to some
5 other environmental factors, and vice versa. EPA produces Air Quality Criteria Documents for
6 four gaseous pollutants: CO, NO2, SO2, and O3 (U.S. Environmental Protection Agency, 1982,
7 1996b, 2000b). The possible health effects of the gaseous pollutants exerted independently from
8 PM, and in some cases jointly with PM, are discussed in those documents. They are also
9 considered to some extent in this section and elsewhere in this document because they may
10 affect quantitative assessments of the effects of various PM metrics when these other pollutants
11 are also present in the atmosphere. The gaseous pollutants may also be of interest as PM effect
12 modifiers, or through interactions with PM.
13 Co-pollutant models have received a great deal of attention in the last few years because
14 there now exist improved statistical methods for estimating PM effects by analyses of daily time-
15 series of mortality (Schwartz and Marcus, 1990; Schwartz, 1991) or hospital admissions
16 (Schwartz, 1994) and/or in prospective cohort studies (Dockery et al., 1993). A number of
17 studies using the new methods have not only found significant positive relationships between
18 mortality and one or more PM indicators, but also with one or another of the four gaseous
19 criteria pollutants (O3, NO2, CO, SO2) in daily time-series studies, and between SO2 and
20 mortality in the reanalyses of two large prospective cohort studies (Krewski et al., 2000). In the
21 daily time-series studies, the estimated PM effect is relatively stable when the co-pollutant is
22 included in the model in some cities, whereas the estimated PM effect in other cities changes
23 substantially when certain co-pollutants are included. In the Krewski et al. (2000) analyses, the
24 estimated effect of SO4= is greatly decreased when SO2 is also included as a predictor in a
25 proportional hazards model. A number of the analyses presented below also discuss models in
26 which multiple particle metrics are present, either with or without the gaseous criteria pollutants.
27 These mixtures are encountered in urban air. Included among the studies evaluating both fine
28 and coarse particles are the following ones: Burnett et al. (2000), Chock et al. (2000), Clyde
29 et al. (2000), Fairley (1999), Lippmann et al. (2000), Mar et al.(2000), Cifuentes et al. (2000),
30 and Castillejos et al. (2000).
June 2003 8-205 DRAFT-DO NOT QUOTE OR CITE
-------
1 Carbon monoxide, NO2, and SO2 may be acting as indicators of distinct emission sources
2 (e.g., motor vehicle exhaust coal- or oil-burning electric power plants, etc.) and/or as indicators
3 of PM from these sources (primary particles and secondary nitrate particles). Concentrations of
4 such gaseous co-pollutants may therefore be correlated with total PM mass, and they may be
5 even more strongly correlated with specific PM constituents due to their emission from a
6 common source. Thus, one or another specific gaseous co-pollutant may serve as an indicator of
7 the day-to-day variation in the contribution of a distinct emission source and to the varying
8 composition of airborne PM. In a model with total PM mass, then, a gaseous co-pollutant may
9 well actually serve as a surrogate for the source-apportioned contribution to ambient air PM.
10 It would be interesting to evaluate models that include both source-relevant particle components
11 and gaseous pollutants derived from common sources (e.g., those attributable to motor vehicles,
12 coal combustion, oil combustion, etc.). The closest approach so far has been Model II in Burnett
13 et al. (2000), a default GAM analyses.
14 The role of gaseous pollutants as surrogates for source-apportioned PM may be distinct
15 from confounding. The true health effect may be independently associated with a particular
16 ambient PM constituent that may be more or less toxic than the particle mix as a whole. Thus,
17 a gaseous co-pollutant may give rise to the appearance of confounding in a regression model.
18 By serving as an indicator of the more toxic particles, the gaseous co-pollutant could greatly
19 diminish the coefficient for total particle mass. In such a model, the coefficient for total particle
20 mass would most properly be interpreted an indicator of the other, less-toxic particles.
21
22 8.4.3.2 Conceptual Issues in Assessing Confounding
23 Two main conceptual issues are encountered in evaluating potential confounding:
24 (a) biological plausibility and (b) exposure plausibility. These concerns overlap two of Hill's
25 (1965) suggested criteria for causal inference.
26 (a) Biological plausibility: It is generally accepted that O3, NO2, and SO2 are associated
27 with diminished pulmonary function and increased respiratory symptoms as well as more serious
28 consequences, and CO exposure has been associated with cardiovascular effects. While one may
29 question whether adverse health effects occur in most healthy people at current exposure to
30 ambient concentrations, there may be susceptible sub-populations for whom one or more
31 ambient gaseous pollutants could perhaps cause health effects at currently encountered ambient
June 2003 8-206 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposure levels. Thus, one should not necessarily assume, a priori, that the gaseous
2 co-pollutants at current ambient levels are not associated with respiratory and cardiovascular
3 health effects in susceptible subpopulations. Nor should the converse be assumed without
4 further evaluation.
5 Ambient gaseous co-pollutants can be potential confounders of ambient PM only if:
6 (a) both the gas and PM are able to cause the same health effects; (b) if personal exposure is
7 correlated with ambient concentrations for both particles and gases respectively; (c) if the
8 personal exposure to gases and to particles are correlated, and; (d) if the ambient concentrations
9 of particles and gases are correlated.
10
11 (b) Exposure plausibility: While most Americans spend most of their time in indoor
12 microenvironments, there is still sufficient personal exposure to O3 to cause notable respiratory
13 symptoms among sensitive children or adults exercising outdoors when ambient O3
14 concentrations are high (hence the declaration of "ozone alert" days). It is also likely that some
15 fraction of ambient CO can contribute to indoor air pollution and total personal CO exposure.
16 Nitrogen dioxide, while reactive, also penetrates indoors; and an ambient pollution component of
17 total personal exposure to NO2 can be identified among individuals without indoor NO2 sources
18 but living close to strong outdoor sources such as highways. While there may be some, perhaps
19 many, individuals exposed to elevated concentrations of gaseous criteria pollutants, in order for
20 them to contribute to health effects shown to be associated with ambient concentrations of
21 another given co-pollutant (e.g., PM), the ambient gaseous pollutants must be significantly and
22 positively correlated with the exposure to that co-pollutant.
23
24 8.4.3.3 Statistical Issues in the Use of Multi-Pollutant Models
25 Multi-pollutant models may be useful tools for assessing whether the gaseous co-pollutants
26 may ^potential confounders of PM effects, but cannot determine if in fact they are. Variance
27 inflation and effect size instability can occur in non-confounded multipollutant models as well as
28 in confounded models. Our usual regression diagnostic tools can only determine whether there
29 is a potential for confounding. In PM epidemiology studies, the gaseous pollutants, except
30 ozone, frequently have a high degree of positive linear correlation with PM metrics, a condition
31 known as multi-collinearity; therefore, although multi-colinearity leading to effect size estimate
June 2003 8-207 DRAFT-DO NOT QUOTE OR CITE
-------
1 instability and variance inflation are necessary conditions for confounding, they are not
2 sufficient in and of themselves to determine whether confounding exists.
3 The most commonly used methods include multi-pollutant models in which both the
4 putative causal agent (PM) and one or more putative co-pollutants are used to estimate the health
5 effect of interest. If the effect size estimate for PM is "stable," then it is often assumed that the
6 effects of confounding are minimal. "Stable" is usually interpreted as meaning that the
7 magnitude of the estimated effect is similar in models with PM alone and in models with PM and
8 one or more co-pollutants, and the statistical significance or width of the confidence interval for
9 the PM effect is similar for all models, with or without co-pollutants. These criteria (usually
10 unquantified) diagnose confounding in a narrow sense, interpreted as synonymous with multi-
11 collinearity, not as a failure of the study design or other forms of model mis-specification.
12 Beyond the conceptual issues discussed above that arise in assessing confounding, there
13 are a number of technical issues that arise in the use of statistical models. Those issues are
14 discussed below.
15
16 (a^ Model mis-specification assumes many forms. The omission of predictive regressors
17 ("underfitting", defined by Chen et al., 2000) may produce biased estimates of the effects of
18 truly predictive regressors that are included in the model. Inclusion of unnecessary or non-
19 predictive regressors along with all truly predictive regressors ("over-fitting") will produce
20 unbiased estimates of effect, but may increase the estimated standard error of the estimated
21 effect if it is correlated with other predictors. Omitting a truly predictive regressor while
22 including a correlated but non-causal variable ("mis-fitting") will attribute the effect of the
23 causal regressor to the non-causal regressor. Interaction terms are candidates for omitted
24 regressor variables. It is important to avoid the "mis-fitting" scenario. Assuming that there is a
25 linear relationship when the true concentration-response function is non-linear will produce a
26 biased estimate of the effect size, high or low at different concentrations. One of the most
27 common forms of model mis-specification is to use the wrong set of multi-day lags, which could
28 produce any of the consequences described as "under-fitting" (e.g., using single-day lags when a
29 multi-day or distributed lag model is needed), "over-fitting" (e.g., including a longer span of
30 days than is needed), or "mis-fitting" (e.g., using a limited set of lags while the effects are in fact
31 associated with different set of lags). Different PM metrics and gaseous pollutants may have
June 2003 8-208 DRAFT-DO NOT QUOTE OR CITE
-------
1 different lag structures, so that in a multi-pollutant model, forcing both PM and gases to have the
2 same lag structure is likely to yield "mis-fitting." Finally, classical exposure measurement errors
3 (from use of proxy variables) attenuates (biases) effect size estimates under most assumptions
4 about correlations among the regressors and among their measurement errors (Zeger et al.,
5 2000).
6
7 (b) Bias: All of the mis-specifications listed in (c) can bias the effect size estimate except
8 for "over-fitting" and measurement error of Berkson type. The estimates of the standard error of
9 the effect size estimate under "over-fitting" or Berkson error cases are inflated, however; and
10 result in broader confidence intervals than would otherwise occur with a more appropriately
11 specified model and/or one with less Berkson type measurement error.
12
13 (c) Estimates of effect size standard errors are usually sensitive to model mis-
14 specification. When all truly predictive regressors are added to an "underfit" model, the
15 uncertainty will almost always be reduced sufficiently that the standard errors of estimated effect
16 size are reduced ("variance deflation"). Adding correlated non-causal variables to "over-fitted"
17 or "mis-fitted" models will further increase the estimated standard errors ("variance inflation").
18 Variance inflation can occur whenever a covariate is highly correlated with the regressor
19 variable that is presumably the surrogate for the exposure of interest. Confounding with the
20 regressor variable can occur only when the covariate is correlated (a) with the regressor variable
21 proxy for the exposure of interest and (b) with the outcome of interest in the absence of the
22 exposure of interest.
23
24 (d^ Mis-specification errors may compound each other. If the concentration-response
25 function is nonlinear but there is measurement error in the exposures, then different sub-
26 populations will have greater or smaller risk than assigned by a linear model. Consider the
27 hypothetical case of a "hockey-stick" model with a threshold. If there were no exposure
28 measurement error, then the part of the population with measured concentrations above the
29 threshold would have excess risk, whereas those below would not. If exposures were measured
30 with error, even if the measured concentration were above the threshold, some people would
31 actually have exposures below the threshold and no excess risk. Conversely, if the measured
June 2003 8-209 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentration was below the threshold, some people would actually have concentrations above
2 the threshold and would have excess risk. The flattening of a non-linear concentration-response
3 curve by measurement error is a well known phenomenon that may be detected by standard
4 methods (Cakmak et al., 1999).
5
6 (e) The question of whether effect size estimates and their standard errors are really
7 significantly different among models is usually not addressed quantitatively. Some authors
8 report various goodness-of-fit criteria such as AIC, BIC, deviance, or over-dispersion index, e.g.,
9 (Chock et al., 2000; Clyde et al., 2000), but the practice is not yet so wide-spread as to assist in
10 analyses of secondary data for use in this document. Variance inflation may also happen with a
11 correctly specified model when both pollutants are causal and highly correlated, compared to a
12 model in which only one pollutant is causal and the non-causal pollutant is omitted. The
13 situation where the variance or standard error decreases when an additional variable is added
14 (variance deflation) suggests that the model with the covariate is more nearly correct and that the
15 standard errors of all covariates may decrease. Statistical significance is a concept of limited
16 usefulness in assessing or comparing results of many models from the same data set. Still, it is a
17 familiar criterion, and one addressed here by using a nominal two-sided 5% significance level
18 for all tests and 95% confidence intervals for all estimates, acknowledging their limitations.
19 There is at present no consensus on what clearly constitutes "stability" of a model estimate effect
20 size, e.g., effect sizes that differ by no more than 20% (or some other arbitrary number) from the
21 single-pollutant models. Simple comparison of the overlap of the confidence intervals of the
22 models is not used because the model estimates use the same data, and the confidence intervals
23 for effect size in different models are more-or-less correlated. In analyses with missing days of
24 data for different pollutants, comparisons must also incorporate differences in sample size or
25 degrees of freedom.
26 In any case, statistical comparisons alone cannot fully resolve questions about either
27 conceptual or statistical issues in confounding via considerations about statistical significance.
28 If the model is mis-specified in any of the numerous ways described above, then effect size
29 estimates and/or their estimated standard errors are likely biased. Statistical assessments alone
30 can determine if the PM metric is too closely correlated with other pollutants to allow for a
31 reasonably accurate quantitative effect size estimate (which is, of course, useful information
June 2003 8-210 DRAFT-DO NOT QUOTE OR CITE
-------
1 even if it is concluded that it is not feasible to estimate the separate effects of PM and/or the
2 gaseous co-pollutants). However, no matter what the statistical situation, confounding cannot
3 occur if the gaseous co-pollutant(s) cannot produce the health outcome, or if there is no personal
4 exposure to the gaseous co-pollutant(s), or if that personal exposure is not correlated with their
5 ambient concentrations.
6 The most commonly used approach to diagnose potential confounding is fitting multi-
7 pollutant models and evaluating the stability of the estimated particle effect sizes against
8 inclusion of co-pollutants. If an additional covariate is added to a baseline model (e.g., with PM
9 alone) and the model predicts the outcome better with the covariate, then the reduction in
10 variance (or deviance for generalized linear or additive models [GLM or GAM]) outweighs the
11 loss of degrees of freedom for variability. Although not always true, it is reasonable to expect a
12 decrease in the estimated asymptotic standard error of the effect size estimate ("variance
13 deflation"), but improved goodness-of-fit may not reduce the standard errors of all parameters in
14 equal proportion because introducing the new covariate modifies the covariate variance-
15 covariance matrix. The weighted inverse covariance matrix provides an exact estimate for
16 standard errors in ordinary linear regression models, and approximately so in GLM or GAM.
17 The effects on other parameter estimates are rarely reported.
18 "Variance inflation" may occur under several circumstances, including "under-fitting" and
19 "mis-fitting" in which a truly predictive covariate is omitted or replaced by a correlated proxy,
20 and "over-fitting" in which a non-predictive covariate correlated with the PM metric is also
21 included in the model. The potential for over-fitting can be diagnosed by evaluating the
22 eigenvalues of the correlation matrix of the predictors, with very small values identifying near-
23 collinearity. However, the complete covariate correlation matrix is almost never reported,
24 including all weather variables and nonlinear functions entered separately as covariates.
25 Nonetheless, even a correlation matrix among all pollutants would be informative. Furthermore,
26 composite correlation matrices in multi-city studies may conceal important differences among
27 the correlation matrices.
28 Multi-pollutant models may be sensitive to multi-colinearity (high correlations among
29 particle and gaseous pollutant concentrations) and to so-called "measurement errors", possibly
30 associated with spatial variability. Combining multi-pollutant models across several cities may
31 not improve the precision of the mean PM effect size estimate combined, if the differences
June 2003 8-211 DRAFT-DO NOT QUOTE OR CITE
-------
1 among the cities are as large or larger in the multi-pollutant models as in the single-pollutant PM
2 model. Second-stage regressions have been useful in identifying effect modifiers in the
3 NMMAPS and APHEA 2 studies, but may not, in general, provide a solution to the problem that
4 confounding of effects is a within-city phenomenon. Furthermore, the correlations among
5 pollutants may change from season to season and from place to place, suggesting that
6 confounding as indicated by co-linearity is not always the same.
7 Three promising alternative approaches versus simple reliance on multi-pollutant modeling
8 have begun to be used to evaluate more fully and definitively the likelihood that exposures to
9 gaseous co-pollutants can account for the ambient PM-health effects associations now having
10 been reported in hundreds of published epidemiology studies. The first is based on evaluation of
11 personal exposures to particles and gases as was done for three panels of participants in
12 Baltimore, MD (Sarnat et al., 2000, 2001). This study (discussed in detail in Chapter 5) directly
13 addresses the premise that if individuals are not exposed to a potential confounder, then it cannot
14 really be a confounder of the presumed causal effect. The results in this paper support the
15 conclusion that personal exposure to sulfates, fine particles, and PM10 are well correlated with
16 their corresponding fixed site ambient concentrations, but the correlations are much lower for
17 PM10_2 5, O3, and NO2. There is however a great deal of variation from one of three two-week
18 panels from one season to the next. The sample size is small (N = 56), but did detect marginally
19 significant associations between personal and ambient NO2 for the personal-ambient correlation,
20 although much lower than for particles. There were, however, a number of residences in which
21 personal and ambient NO2 were highly correlated. This has been known to happen in other
22 studies when the residences are close to a major road, which was the case for several members in
23 each of the three studied cohorts (i.e, health elderly adults, adults with COPD, and children 9-13
24 years).
25 An other promising approach is the use of principal component or factor analysis to
26 determine which combinations of gaseous criteria pollutants and PM size fractions or chemical
27 constituents together cannot be easily disentangled, and which pollutants are substantially
28 independent of the linear combinations of the others. For example, the source-oriented factor
29 analysis study of Mar et al. (2000) produced evidence suggesting independent effects of regional
30 sulfate, motor vehicle-related particles, particles from vegetive burning, and PM10_25 for
31 cardiovascular mortality in Phoenix (as discussed in Section 8.2.2.4.3).
June 2003 8-212 DRAFT-DO NOT QUOTE OR CITE
-------
1 There are also now available some recent examples of a third promising approach, i.e., the
2 use of so-called "intervention studies." Particularly interesting evidence for independent effects
3 of ambient PM beginning to emerge from such studies, which relate changes (decreases in health
4 risk outcomes) to decreases in airborne particles due to deliberate reductions in emissions from
5 sources that ordinarily contribute to elevated ambient PM levels in a given locale. As described
6 in the next subsection (8.4.3.4), the PM-health outcome changes occurred in the presence of low
7 concentrations of ambient gaseous co-pollutants or little change in at least some of the co-
8 pollutants in the presence of the reduced concentrations of PM mass or constituents.
9
10 8.4.3.4 Epidemiologic Studies of Ambient Air Pollution Interventions
11 To date, investigations of health risk in epidemiologic studies of ambient air pollutants,
12 including PM, have relied largely on studies that focus on increases in exposure and that
13 evaluate whether health risk changes occur in relation to such increases. Such studies are used to
14 support qualitative and quantitative inferences as to whether decreases in exposure will bring
15 about reductions in health risk, or improvement in health status.
16 Ambient criteria air pollutants are rarely, if ever, the only etiology of the health disorders
17 with which exposures to these pollutants are associated. For example, numerous reports have
18 implicated ambient air pollution exposure with exacerbations of pre-existing asthma. These
19 reports justify the expectation that further reduction in ambient air pollution exposure would
20 reduce the public health burden of asthma exacerbations. However, many other factors,
21 including allergens, passive smoking, exercise, cold, and stress are also associated with such
22 exacerbations. Asthmatics would continue to be exposed to these factors even with further
23 reduction in ambient air pollution exposure. Thus, reduction of ambient air pollution exposure,
24 even to zero concentration, would not bring about zero risk of the health disorders with which
25 such exposure is associated. Also, it is likely that at least some non-pollution risk factors would
26 behave differently in the absence of ambient air pollution exposure as in its presence. That is, in
27 the real world, risk factors probably do not behave in discrete, additive fashion.
28 Direct quantitative characterization of effects of reduction in air pollution concentrations
29 and exposures requires the study of situations in which such reductions actually occur. In such
30 studies, it is important to measure both exposure and health status before and after exposure is
June 2003 8-213 DRAFT-DO NOT QUOTE OR CITE
-------
1 reduced. It is also highly desirable to identify risk factors other than ambient air pollution, and
2 to ascertain their effects before and after air pollution exposure reduction.
3 In his classic monograph (The Environment and Disease: Association or Causation?), Hill
4 (1965) addressed the topic of preventive action and its consequences under Aspect 8, stating:
5
6 "Experiment: Occasionally it is possible to appeal to experimental, or semi-experimental,
7 evidence. For example, because of an observed association some preventive action is taken.
8 Does it in fact prevent? The dust in the workshop is reduced, lubricating oils are changed,
9 persons stop smoking cigarettes. Is the frequency of the associated events affected? Here the
10 strongest support for the causation hypothesis may be revealed."
11
12 The available epidemiologic literature on ambient air pollution generally offers only
13 limited evidence related to this aspect. A few pertinent studies have evaluated situations where
14 air pollution concentrations have been temporarily or permanently reduced through regulatory
15 action, industrial shutdown, or other intervening factor(s).
16 In the U.S., the most thoroughly studied example of such ambient air pollution reduction
17 occurred in the Utah Valley, UT, during the 1980s. The Valley's largest stationary source of
18 PM, a steel mill, was closed due a labor dispute for 13 months from autumn 1986 until autumn
19 1987. This offered the opportunity to study health effects not only of the closure-related
20 reduction in ambient PM concentrations, but also of the increases in PM that occurred after the
21 re-opening of the mill. Pope et al. have reported extensively on such health effects. The
22 relevant reports having been addressed in detail in the 1996 PM AQCD. Briefly, these
23 investigators observed reduction in frequency of a variety of health disorders during the period
24 in which the mill was closed. These included daily mortality (Pope et al., 1992), respiratory
25 hospital admissions (Pope, 1989), bronchitis and asthma admissions for preschool children
26 (Pope, 1991), reductions in lung function (Pope et al., 1991), and elementary school absences
27 (Ransom and Pope, 1992). Changes in these endpoints were reflected by differing strength of
28 positive associations between measures of these health endpoints and PM mass measurements
29 from filters collected before, during, and after the steel mill shut down.
30 As discussed in Chapter 7 of this document, several experimental studies investigated
31 effects of aqueous extracts of ambient Utah Valley particulate filters employing filter extracts
32 from January through March 1986 (mill open), 1987 (mill closed), and 1988 (mill open)
June 2003 8-214 DRAFT-DO NOT QUOTE OR CITE
-------
1 (Frampton et al., 1999; Dye et al., 2001; Soukup et al., 2000; Wu et al., 2001; and Ohio and
2 Devlin, 2001). In all of these studies, investigators observed less intense in vivo or in vitro
3 effects when treating with the 1987 (mill closed) extracts than when treating with (mill open)
4 extracts from 1986 and/or 1988. The methodology descriptions provided across the above five
5 papers are somewhat unclear as to the degree of comparability of source filters among these five
6 studies (some being from TSP and others from PM10 filters); and there is some uncertainty as to
7 the within-study comparability of filters from year to year, particularly in the studies that
8 employed 34 filters per year. Furthermore, some proportion of the extracted material may have
9 been derived from filter matrix, not ambient PM; and about 10 years elapsed between collection
10 and extraction of the filter samples.
11 Even so, the combined results of these five experimental studies provide support and
12 corroboration for the epidemiologic observations of reduced frequency and severity of health
13 disorders during the period of steel mill closure during which PM10 (and to some extent SO2)
14 levels were notably reduced, but already relatively low CO, NO2, and O3 were much less
15 changed. The experimental studies also provide support for hypotheses regarding potential
16 biological mechanisms underlying some of the observed effects. Perhaps the strongest of these
17 hypotheses is that PM-associated metals were etiologically related to some of the observed
18 disorders, and that reduction in ambient concentrations of these metals was at least partially
19 responsible for the health benefits observed during steel mill closure. In any event, these
20 experimental studies underscore the importance of particle composition in production or
21 promotion of harmful health effects (Beckett, 2001).
22 Another study (Avol et al., 2001) investigated effects of reductions and increases in
23 ambient air pollution concentrations on longitudinal lung function growth in a subsample of
24 participants in the Children's Health Study conducted by the University of Southern California.
25 Follow-up lung function tests were administered to 110 children who had moved away from the
26 study area after the baseline lung function test, which was administered while the children lived
27 within the study area. Lung function growth rates were analyzed against differences between the
28 children's original and new communities in annual average concentrations of PM10, NO2, and O3.
29 Analytical models were adjusted for anthropometric variables and other relevant covariates.
30 No multi-pollutant analyses were reported. Moving to a community with lower ambient PM10
31 concentration was associated with increased growth rates of FVC, FEV1, MMEF and PEFR;
June 2003 8-215 DRAFT-DO NOT QUOTE OR CITE
-------
1 whereas moving to a community with higher PM10 concentrations was associated with decreased
2 growth of these metrics. These associations were statistically significant for MMEF and PEFR,
3 and appear to have been marginally significant for FVC and FEV1. Moving to a community
4 with lower ambient NO2 or O3 concentration was also generally associated with increased lung
5 function growth, and vice versa; however, the associations of change in lung function growth
6 with change in community levels of NO2 and O3 were not statistically significant. This study
7 suggests, most clearly, that reduction in long-term ambient PM10 levels is indeed associated with
8 improvement of children's lung growth, and that increase in these levels is associated with
9 retardation of lung growth.
10 In yet another U.S. study, Friedman et al. (2001) investigated the influence of temporary
11 changes in transportation behaviors (instituted to reduce downtown traffic congestion during the
12 1996 Summer Olympic Games in Atlanta, GA) on ambient air quality and acute care visits and
13 hospitalizations for asthma in children residing in Atlanta. Ambient air quality and childhood
14 asthma during the 17 days of the Games were compared to those during a baseline period
15 consisting of the four weeks before and the four weeks after the Games. During the Games,
16 concentrations of PM10 (24-h average), O3 (daily peak 1-h average), CO (8-h average), and NO2
17 (daily peak 1-h average) were, respectively, 16.1%, 27.9%, 18.5%, and 6.8% lower than during
18 the baseline period. Twenty-four hour average concentrations of SO2 were 22.1% higher during
19 the Games than during the baseline period. Reductions in O3, PM10, and CO were statistically
20 significant at alpha = 0.05 (p = 0.01, p < 0.001, and p = 0.02, respectively). Ambient mold
21 counts during the Games did not differ significantly from those during the baseline period. Four
22 sources of asthma frequency data were examined: (1) the Georgia Medicaid claims file; (2) files
23 of a health maintenance organization; (3) emergency department records for two of Atlanta's
24 three pediatric hospitals; and (4) the Georgia Hospital Discharge Database. For all four sources,
25 asthma-related unadjusted and adjusted relative risks during the Games were less than 1 (as
26 compared to RR = 1 during the baseline period). Relative risks from the Medicaid database were
27 statistically significant (p < 0.005), and those from the HMO approached significance (p < 0.10).
28 These findings suggest strongly that, in Atlanta in summer 1996, temporary improvement in
29 ambient air quality contributed to temporary reduction in severity of pre-existing asthma. This
30 reduction could not be attributed specifically to any individual air pollutant, but reductions in
31 PM and O3 would seem to be among the most likely contributors to the observed effect on
June 2003 8-216 DRAFT-DO NOT QUOTE OR CITE
-------
1 asthma visits. In the opinion of Friedman et al., reductions in morning rush-hour traffic played
2 an important role in reduction of asthma-related visits and hospitalizations.
3 Heinrich et al. (2000) studied the effects of long-term air pollution reduction in the former
4 East Germany on prevalence of respiratory illnesses and symptoms in 5 to 14 year-old children.
5 Cross-sectional surveys were conducted in 1992-1993 and 1995-1996 in three areas, all of which
6 experienced reductions in annual mean ambient SO2 and TSP concentrations in the time interval
7 between the surveys. Percentage reductions in SO2 and TSP were substantial, ranging from
8 about 40%-60% and about 20%-35%, respectively, in the three areas. Longitudinal changes
9 were not measured for size-specific PM metrics. After adjustment for relevant covariates,
10 statistically significant temporal decreases in prevalences of bronchitis, otitis media, frequent
11 colds, and febrile infections were observed.
12 In Hong Kong, a regulation prohibiting the use of fuel oil containing more than 0.5% sulfur
13 by weight went into effect in July 1990. Investigators from the University of Hong Kong studied
14 respiratory health in children and non-smoking women before and after the regulation was
15 implemented. In a relatively polluted district (District A), the regulation resulted in rapid and
16 substantial reduction in the ambient SO2 concentration and in appreciable, but less marked,
17 reduction in the concentration of sulfate ion in "respirable suspended particulates" (RSP, thought
18 to be equivalent to PM10). Percentage reductions in these sulfur-containing pollutants were
19 considerably smaller in a less polluted district (District B). The regulation was not accompanied
20 by appreciable reductions in levels of PM metrics (TSP and RSP) in either district.
21 Tarn et al. (1994) reported that the prevalence of bronchial hyperreactivity (BHR) in
22 children (as defined by a > 20% drop in FEV1 in response to histamine challenge) was higher in
23 District A than in District B, even after exclusion of children with wheeze and asthma. Wong
24 et al. (1998) measured BHR prevalence rates in these districts in 1991 and 1992, and compared
25 these to rates before the regulation was implemented. In both districts, BHR prevalence was
26 statistically significantly lower in 1991 than before the intervention. In 1992, the pre- to post-
27 intervention decrease in BHR prevalence was significantly larger in District A than in
28 District B. Peters et al. reported that before the intervention, prevalences of children's respiratory
29 symptoms (e.g., cough, sore throat, wheeze) were statistically significantly higher in District A
30 than in District B. About one year after the intervention, there were greater pre- to post-
31 intervention declines in prevalences of cough or sore throat, phlegm, and wheezing in District A
June 2003 8-217 DRAFT-DO NOT QUOTE OR CITE
-------
1 than in District B. Wong et al. reported that before the intervention, the prevalence of poor
2 respiratory health in non-smoking women was significantly higher in District A than in District
3 B. Also, effects of passive smoking on the women's respiratory health were stronger in District
4 A than in District B, but not significantly so. About one year after the intervention, declines in
5 frequency of poor respiratory health were observed, but these declines did not differ significantly
6 between districts. Taken together, these Hong Kong studies suggest that reduction of sulfur in
7 fuel oil brought about appreciable improvement in children's respiratory health, and discernible
8 but lesser improvement in non-smoking women's respiratory health. These studies also suggest
9 that these benefits were associated with reduction in sulfur-containing ambient air pollutants, but
10 not necessarily with reduction in TSP or RSP per se.
11 Taken together, these epidemiologic intervention studies tend to support the conclusion
12 that reductions in ambient air pollution (especially PM) exposures resulted in decreased
13 respiratory and cardiovascular health effects. The available studies also give reason to expect
14 that further reductions in both particulate and gaseous air pollutants would benefit health. On
15 balance, these studies suggest that selective reduction in ambient PM concentrations might well
16 bring about greater benefit than would selective reduction in concentrations of other ambient
17 criteria air pollutants. Furthermore, the experimental studies of Utah Valley filter extracts point
18 to PM-associated metals as a likely cause or promoter of at least some of the health effects
19 associated with ambient PM. Beyond this, available epidemiologic intervention studies do not
20 yet give direct, quantitative evidence as to the relative health benefits that would result from
21 selective reduction of specific PM size fractions. Also, these studies do not yet provide firm
22 grounds for quantitative prediction of the relative health benefits of single-pollutant reduction
23 strategies versus multi-pollutant reduction strategies. Even in an almost ideal "natural
24 experiment" such as Utah Valley, potentially confounding factors other than ambient PM
25 concentrations may have also changed during the steel mill closure. These included changes in
26 concentrations of at least one other pollutants (i.e., SO2) and possible changes in population due
27 to out- and in-migration influenced by the closing and re-opening of the steel mill. While
28 changes in ambient PM concentrations undoubtedly played a role, other factors may also have
29 modified the size of the changes in health effects.
30
31
June 2003 8-218 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.4 Role of Participate Matter Components
2 In the 1996 PM AQCD, extensive epidemiologic evidence substantiated very well positive
3 associations between ambient PM10 concentrations and various health indicators, e.g., mortality,
4 hospital admissions, respiratory symptoms, pulmonary function decrements, etc. Some studies
5 were also then available which mortality and morbidity associations with various fine particle
6 indicators (e.g., PM2 5, sulfate, H+, etc.). One mortality study, the Harvard Six Cities analysis by
7 Schwartz et al. (1996a), evaluated relative contributions of the fine (PM25) versus the coarse
8 (PM10_2 5) fraction of PM10, and found, overall, that PM25 appeared to be associated more strongly
9 with mortality effects than PM10_2 5. A few studies seemed to be indicative of possible coarse
10 particle effects, e.g., increased asthma risks associated with quite high PM10 concentrations in a
11 few locations where coarse particles strongly dominated the ambient PM10 mix.
12
13 8.4.4.1 Fine-and Coarse-Particle Effects on Mortality
14 A rapidly growing number of new studies published since the 1996 PM AQCD provide an
15 expanded evidence base examining associations of ambient PM with increased human mortality
16 and morbidity risks. As was indicated in Table 8-1, most newly reported analyses, with a few
17 exceptions, continue to show statistically significant associations between short-term (24-h) PM
18 concentrations and increases in daily mortality in many U.S. and Canadian cities (as well as
19 elsewhere). Also, the reanalyses of Harvard Six City and ACS study data substantiate the
20 original investigator's findings of long-term PM exposure associations with increased mortality
21 as well.
22
23 8.4.4.1.1 Total Mortality Effects
24 The effects estimates from the newly reported studies are generally consistent with those
25 derived from the earlier 1996 PM AQCD assessment, which reported risk estimates for excess
26 total (nonaccidental) deaths associated with short-term PM exposures as generally falling within
27 the range of ca. 1 to 8% per 50 |ig/m3 PM10 (24-h) increment and ca. 2 to 6% increase per
28 25 |ig/m3 PM2 5 (24-h) increment.
29 Several new PM epidemiology studies which conducted time-series analyses in multiple
30 cities were noted to be of particular interest, in that they provide evidence of effects across
31 various geographic locations (using standardized methodologies) and more precise pooled effect
June 2003 8-219 DRAFT-DO NOT QUOTE OR CITE
-------
1 size estimates with narrow confidence bounds, reflecting the typically much stronger power of
2 such multi-city studies over individual-city analyses to estimate a mean effect. Based on pooled
3 analyses across multiple cities, using GAM stringent convergence criteria, the percent total
4 (non-accidental) excess deaths per 50 |ig/m3 PM10 (24-h) increment were estimated in different
5 multi-city analyses to be: (a) 1.4% in the 90 largest U.S. cities; (b) 3.4% in 10 large U.S. cities;
6 (c) 3.6% in the 8 largest Canadian cities; and (d) 3.0% in European cities.
7 Many new individual-city studies found positive associations (most statistically significant
8 at p < 0.05) for the PM25 fraction, with effect size estimates for U.S. and Canadian cities
9 typically ranging from ca. 2.0 to ca. 8% per 25 |ig/m3 PM25 (although one estimate for
10 cardiovascular mortality ranged up to about 19%). Of the 10 or so new analyses that not only
11 evaluated PM10 effects but also compared fine versus coarse fraction contributions to total
12 mortality, only two are multi-city analyses yielding pooled effects estimates: (a) the Klemm and
13 Mason (2000) and Klemm and Mason (2003) recomputation analyses for Harvard Six Cities
14 data, confirming the original findings published by Schwartz et al. (1996a); and (b) the Burnett
15 et al. (2000) and Burnett and Goldberg (2003) studies of the 8 largest Canadian cities. These
16 studies found roughly comparable, statistically significant excess risk estimates for PM2 5 (i.e.,
17 approximately 2% increased total mortality risk per 25 |ig/m3 PM2 5 increment).
18 As for possible coarse particle short-term exposure effects on mortality, in those new
19 studies which evaluated PM10_25 effects as well as PM25 effects, the coarse particle (PM10_25)
20 fraction was also consistently positively associated with increased total mortality, albeit the
21 coarse fraction effect size estimates were generally less precise than those for PM2 5 and
22 statistically significant at p < 0.05 in only a few studies (as can be seen in Figure 8-6). Still, the
23 overall picture tends to suggest that excess total mortality risks may well reflect actual coarse
24 fraction particle effects, in at least some locations. This may be most consistently the case in
25 arid areas, e.g., in the Phoenix area (as shown in Mar et al., 2000 and Mar et al., 2003) or in
26 Mexico City and Santiago, Chile. On the other hand, elevations in coarse PM-related total
27 mortality risks have also been detected for Steubenville, OH (an eastern U.S. urban area in the
28 Harvard Six City Study), as shown by Schwartz et al. (1996a); Klemm et al. (2000), Klemm and
29 Mason (2003). These results may reflect contamination of later-resuspended coarse PM by
30 metals in fine PM emitted from smelters (Phoenix) or steel mills (Steubenville) that was earlier
31 deposited on nearby soils. Excess total mortality risks associated with short-term (24-h)
June 2003 8-220 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposures to coarse fraction particles capable of depositing in the lower respiratory tract
2 generally fall in the range of 0.2 to 6.0% per 25 |ig/m3 PM10_25 increment for U.S. and Canadian
3 cities.
4 Three new papers provide particularly interesting new information on relationships
5 between short-term coarse particle exposures and total elderly mortality (age 65 and older),
6 using exposure TEOM data from the EPA ORD NERL monitoring site in Phoenix, AZ. Each
7 used quite different models but each reported statistically significant relationships between
8 mortality and coarse PM, specifically PM10_2 5, an indicator for the thoracic fraction of coarse-
9 mode PM.
10 Smith et al. (2000), using a three-day running average as the exposure metric, performed
11 linear regression of the square root of daily mortality on the long-term trend, meteorological and
12 PM-based variables. Two mortality variables were used, total (non-accidental) deaths for the
13 city of Phoenix and the same for a larger, regional area. Using a linear analysis, effects based on
14 coarse PM were statistically significant for both regions, whereas effects based on fine PM
15 (PM2 5) were not. However, when the possibility of a nonlinear response was taken into account,
16 no evidence was found for a nonlinear effect for coarse PM; but fine PM was found to have a
17 statistically significant effect for concentration thresholds of 20 and 25 |ig/m3. There was no
18 evidence of confounding between fine and coarse PM, suggesting that fine and coarse PM are
19 "essentially separate pollutants having distinct effects". Smith et al. (2000) also observed a
20 seasonal effect for coarse PM, the effect being statistically significant only during spring and
21 summer. Based on a principal component analysis of elemental concentrations, crustal elements
22 are highest in spring and summer and anthropogenic elements lowest, but Smith et al. (2000) felt
23 that the implication that crustal, rather than anthropogenic elements, were responsible for the PM
24 mortality was counterintuitive.
25 Clyde et al. (2000) used a more conventional model, a Poisson regression of log deaths on
26 linear PM variables; but they employed Bayesian model averaging to consider a wide variety of
27 variations in the basic model. They considered three regions: the Phoenix metropolitan area;
28 a small subset of zip code to give a region presumably with uniform PM25; and a still smaller zip
29 code region surrounding the monitoring site (thought to be uniform as to PM10 concentrations).
30 The models considered lags of 0, 1, 2, or 3 days but only for single day PM variables (no running
June 2003 8-221 DRAFT-DO NOT QUOTE OR CITE
-------
1 averages as used by Smith et al., 2000). A PM effect with a reasonable probability was found
2 only in the uniform PM2 5 region and only for coarse PM.
3 Mar et al. (2000, 2003) used conventional Poisson regression methods and limited their
4 analyses to the smallest area (called "Uniform PM10" by Clyde et al., 2000). They reported
5 modeling data for lag days 0 to 4. Coarse fraction PM was marginally significant on lag day 0.
6 No direct fine particle measures were statistically significant on day 0. A regional sulfate factor
7 determined from source apportionment, however, was statistically significant. No correlations
8 were reported for the source apportionment factors, but the correlation coefficient between sulfur
9 (S) in PM2 5 (as measured by XRF) with coarse fraction PM was only 0.13, suggesting separate
10 and distinct effects for regional sulfate and coarse fraction PM.
11 The above three studies of PM- total mortality relationships in Phoenix tend to suggest a
12 statistical association of coarse fraction PM with total elderly mortality in addition to and
13 different from any relationship with fine PM, fine PM components, or source factors for fine
14 PM.
15 With regard to long-term PM exposure effects on total (non-accidental) mortality, the
16 newly available evidence from the HEI Reanalyses of Harvard Six Cities and ACS data (and
17 extensions, thereof), substantiate well associations attributable to chronic exposures to inhalable
18 thoracic particles (indexed by PM15 or PM10) and the fine fraction of such particles (indexed by
19 PM2 5 and/or sulfates). Statistically significant excess risk for total mortality was shown by the
20 reanalyses to fall in the range of 4-18% per 20 |ig/m3 PM15/10 increment and 14-28% per
21 10 |ig/m3 PM25 increase.
22
23 Source-Oriented Analyses of Particle Component Contributions
24 Other recent studies on the relation of mortality to particle composition and source (Laden
25 et al., 2000; Mar et al., 2000; Ozkaynak et al., 1996; Tsai et al., 2000) suggest that particles from
26 certain sources may have much higher potential for adverse health effects than others, as shown
27 by source-oriented evaluations involving factor analyses. For example, Laden et al. (2000)
28 conducted factor analyses of the elemental composition of PM25 for Harvard Six Cities study
29 data for 1979-1988. For all six cities combined, the excess risk for daily mortality was estimated
30 to be 9.3% (95% CI; 4.0, 14.9) per 25 |ig/m3 PM2 5 (average of 0 and 1 day lags) increment in a
31 mobile source factor; 2.0% (95% CI; -0.3, 4.4) for a coal source factor, and -5.1% (95% CI;
June 2003 8-222 DRAFT-DO NOT QUOTE OR CITE
-------
1 -13.9, 4.6) for a crustal factor. There was large variation among the cities and suggestion of an
2 association (not statistically significant) with a fuel oil factor identified by V or Mn.
3 Mar et al. (2000) applied factor analysis to evaluate mortality in relation to 1995-1997 fine
4 particle elemental components and gaseous pollutants (CO, NO2, SO2) in an area of Phoenix,
5 AZ, close to the air pollution monitors. The PM2 5 constituents included sulfur, Zn, Pb, soil-
6 corrected potassium, organic and elemental carbon, and a soil component estimated from oxides
7 of Al, Si, and Fe. Based on models fitted using one pollutant at a time, statistically significant
8 associations were found between total mortality and PM10, CO (lags 0 and 1), NO2 (lags 0, 1,3,
9 4), S (negative), and soil (negative). Statistically significant associations were also found
10 between cardiovascular mortality and CO (lags 0 to 4), NO2 (lags 1 and 4), SO2 (lags 3 and 4),
11 PM2 5 (lags 1,3,4), PM10 (lag 0), PM10_2 5 (lag 0), and elemental, organic, or total carbon.
12 Cardiovascular mortality was significantly related to a vegetative burning factor (high loadings
13 on organic carbon and soil-corrected potassium), motor vehicle exhaust/resuspended road dust
14 factor (with high loadings on Mn, Fe, Zn, Pb, OC, EC, CO, and NO2), and a regional sulfate
15 factor (with a high loading on S). However, total mortality was negatively associated with a soil
16 factor (high loadings on Al, Fe, Si) and a local SO2 source factor, but was positively associated
17 with the regional sulfate factor.
18 Tsai et al. (2000) analyzed daily time-series of total and cardiorespiratory deaths, using
19 short periods of 1981-1983 data for Newark, Elizabeth, and Camden, NJ. In addition to
20 inhalable particle mass (PM15) and fine particle mass (PM2 5), the study evaluated data for metals
21 (Pb, Mn, Fe, Cd, V, Ni, Zn, Cu) and for three fractions of extractable organic matter. Factor
22 analyses were carried out using the metals, CO, and sulfates. The most significant sources or
23 factors identified as predictors of daily mortality were oil burning (targets V, Ni), Zn and Cd
24 processing, and sulfates. Other factors (dust, motor vehicles targeted by Pb and CO, industrial
25 Cu or Fe processing) were not significant predictors. In Newark, oil burning sources and
26 sulfates were positive predictors, and Zn/Cd a negative predictor for total mortality. In Camden
27 oil burning and motor vehicle emissions predicted total mortality, but copper showed a marginal
28 negative association. Oil burning, motor vehicle emissions, and sulfates were predictors of
29 cardiorespiratory mortality in Camden. In Elizabeth, resuspended dust indexed by Fe and Mn
30 showed marginal negative associations with mortality, as did industrial sources traced by Cu.
June 2003 8-223 DRAFT-DO NOT QUOTE OR CITE
-------
1 The set of results from the above factor analyses studies do not yet allow one to identify
2 with great certainty a clear set of specific high-risk chemical components of PM. Nevertheless,
3 some commonalities across the studies seem to highlight the likely importance of mobile source
4 and other fuel combustion emissions (and apparent lesser importance of crustal particles) as
5 contributing to increased total or cardiorespiratory mortality.
6
7 8.4.4.1.2 Cause-Specific Mortality Effects
8 Cardiovascular- and Respiratory-Related Mortality
9 Numerous new studies have evaluated PM-related effects on cause-specific mortality.
10 Most all report positive, often statistically significant (at p < 0.05), short-term (24-h) PM
11 exposure associations with CVD- and respiratory-related deaths. Cause-specific effects
12 estimates appear to mainly fall in the range of 3.0 to 7.0% per 25 |ig/m3 24-h PM25 for
13 cardiovascular or combined cardiorespiratory mortality and 2.0 to 7.0% per 25 |ig/m3 24-h PM25
14 for respiratory mortality in U.S. cities. Effect size estimates for the coarse fraction (PM10_25) for
15 cause-specific mortality generally fall in the range of ca. 3.0 to 8.0% for cardiovascular and ca.
16 3.0 to 16.0% for respiratory causes per 25 |ig/m3 increase in PM10_25.
17 Also of particular interest, the above noted study by Mar et al. examined the associations of
18 a variety of PM indicators with cardiovascular mortality (for age >65), again in the zip code area
19 near the Phoenix monitoring site. For this end point, coarse PM was statistically significant on
20 lag day 0 but not on subsequent lag days. PM25 and a number of fine PM indicators were
21 statistically significant on lag day 1 but not on lag day 0. This suggests a distinct and separate
22 relationship of PM25 and PM10_2 5. As in the case of total mortality, the only fine PM indicator
23 found to be statistically significant on lag day 0 was regional sulfate. However, the low
24 correlation coefficient between S in PM2 5 and PM10_2 5 (r = 0.13) suggests that the two
25 relationships represent different sets of deaths. Thus, there is some evidence suggesting that the
26 risk of cardiovascular mortality, as well as that of total mortality, may be statistically associated
27 with PM10_2 5 - possibly independent of any relationships with fine particle indicators.
28
29 Long-Term PM Exposure and Lung Cancer
30 Of particular interest with regard to PM-related cause-specific mortality is growing
31 evidence linking long-term PM exposure with increased risk of lung cancer. Historical evidence
June 2003 8-224 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
includes studies of lung cancer trends, studies of occupational groups, comparisons of urban and
rural populations, and case-control and cohort studies using diverse exposure metrics (Cohen and
Pope, 1995). Numerous past ecological and case-control studies of PM and lung cancer have
generally indicated a lung cancer RR greater than 1.0 to be associated with living in areas having
higher PM exposures despite possible problems with respect to potential exposure and other risk
factor measurement errors. Table 8-37 provides a partial listing of such studies.
TABLE 8-37. SUMMARY OF PAST ECOLOGIC AND CASE-CONTROL
EPIDEMIOLOGIC STUDIES OF OUTDOOR AIR AND LUNG CANCER
Study Type Authors
Ecologic Henderson etal.,
1975
Buffleretal.,
1988
Archer, 1990
Case-Control Pike et al, 1979
Vena, 1982
Jedrychowski,
etal., 1990
Katsouyanni,
etal., 1990
Barbone et al.,
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)
1.1 @ soot up to 400
ug/m3
(CI: N/A)
1.4 @ > 0.3 g/m2/day
(CI: 1.1-1.8)
1.3
(CI: 0.9-1.9)
Source: Derived from Cohen (2000).
1 Prospective cohort studies offer a potentially more powerful approach to evaluation of
2 apparent associations between PM exposures and development of lung cancer. The 1996 PM
3 AQCD (U.S. Environmental Protection Agency, 1996a) summarized three of these more
4 elaborate studies that carefully evaluated PM air pollution exposure effects on lung cancer using
June 2003
8-225
DRAFT-DO NOT QUOTE OR CITE
-------
1 the prospective cohort design. In the AHSMOG Study, Abbey et al. (1991) followed a cohort of
2 Seventh Day Adventists, whose extremely low prevalence of smoking and uniform, relatively
3 healthy dietary patterns reduce the potential for confounding by these factors. Excess lung
4 cancer incidence was observed in females in relation to both particle (TSP) and O3 exposure after
5 6 years follow-up time. Dockery et al. (1993) reported the results of a 14- to 16-year prospective
6 follow-up of 8,111 adults living in six U.S. cities that evaluated associations between air
7 pollution and mortality. After controlling for individual differences in age, sex, cigarette
8 smoking, BMI, education, and occupational exposure, Dockery et al. (1993) found an elevated
9 but non-significant risk for lung cancer (RR = 1.37; 95% CI = 0.81 to 2.31) for a difference in
10 PM2 5 pollution equal to that of the most polluted versus the least polluted city. Pope et al.
11 (1995) similarly analyzed PM2 5 and sulfate (SO4=) air pollution as predictors of mortality in a
12 prospective study of 7-year survival data (1982 to 1989) for about 550,000 adult volunteers
13 obtained by the American Cancer Society (ACS).
14 Both the ACS and Harvard studies have been subjected to much scrutiny, including an
15 extensive independent audit and reanalysis of the original data (Krewski et al., 2000) that
16 confirmed the originally published results. The ACS study controlled for individual differences
17 in age, sex, race, cigarette smoking, pipe and cigar smoking, exposure to passive cigarette
18 smoke, occupational exposure, education, BMI, and alcohol use. Lung cancer mortality was
19 significantly associated with particulate air pollution when SO4= was used as the index,, but not
20 when PM25 mass was used as the index for a smaller subset of the study population that resided
21 in metropolitan areas where PM2 5 data were available from the Inhalable Particle (IP) Network.
22 Thus, while these prospective cohort studies have also indicated that long-term PM exposure is
23 associated with an increased cancer risk, the effect estimates were generally not statistically
24 significant, quite possibly due to inadequate statistical power by these studies at that time (e.g.,
25 due to inadequate population size and/or follow-up time for long-latency cancers).
26 The AHSMOG investigators have re-examined the association between long-term PM
27 exposure and increased risk of both lung cancer incidence and lung cancer mortality in
28 nonsmokers using longer-term follow-up of this cohort and improved analytical approaches.
29 Beeson et al. (1998) considered this cohort of some 6,338 nonsmoking, non-Hispanic, white
30 Californian adults, ages 27-95, that was followed from 1977 to 1992 for newly diagnosed
31 cancers. Incident lung cancer in males was positively and significantly associated with
June 2003 8-226 DRAFT-DO NOT QUOTE OR CITE
-------
1 interquartile range (IQR) increases for mean concentrations of PM10 (RR = 5.21; 95% CI = 1.94-
2 13.99). For females in the cohort, incident lung cancer was positively associated with IQR
3 increases for SO2 (RR = 2.14; CI, 1.36-3.37) and IQR increases for PM10 exceedance frequencies
4 of 50 |ig/m3 (RR =1.21; 95% CI = 0.55-2.66) and 60 ug/m3 (RR= 1.25; 95% CI = 0.57-2.71).
5 Thus, increased risks of incident lung cancer were deemed by the authors to be associated with
6 elevated long-term ambient concentrations of PM10 and SO2 in both genders. The higher PM10
7 risk effect estimate for cancer in males appeared to be partially due to gender differences in
8 long-term air pollution exposures. Abbey et al. (1999) also related long-term ambient
9 concentrations of PM10, SO/2, SO2, O3, and NO2 to 1977-1992 mortality in the AHSMOG
10 cohort. After adjusting for a wide array of potentially confounding factors, including
11 occupational and indoor sources of air pollutants, PM10 showed a strong association with lung
12 cancer deaths in males (PM10 IQRRR=2.38; 95% CI: 1.42 - 3.97). In this cohort, males spent
13 more time outdoors than females, thus having higher estimated air pollution exposures than the
14 cohort females. Ozone showed an even stronger association with lung cancer mortality for
15 males, and SO2 showed strong associations with lung cancer mortality for both sexes. The
16 authors reported that other pollutants showed weak or no association with mortality. Therefore,
17 increases in both lung cancer incidence and lung cancer mortality in the extended follow-up
18 analysis of the AHSMOG study were found to be most consistently associated with elevated
19 long-term ambient concentrations of PM10 and SO2, especially among males.
20 A recent follow-up analysis of the major ACS study by Pope et al. (2002) responds to a
21 number of criticisms previously noted for the earlier ACS analysis (Pope et al., 1995) in the
22 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a). Most notably, the new study
23 examined other pollutants, had better occupational indices and diet information, and also
24 addressed possible spatial auto-correlations due to regional location. The recent extension of the
25 ACS study included -500,000 adult men and women drawn from ACS-CPS-II enrollment and
26 follow-up during 1982-1998. This new analysis of the ACS cohort substantially expands the
27 prior analysis, including: (1) more than doubling of the follow-up time to 16 years (and more
28 than tripling of the number of deaths in the analysis); (2) substantially expanded exposure data,
29 including gaseous co-pollutant data and new PM2 5 data collected in 1999-2001; (3) improved
30 control of occupational exposures; (4) incorporation of dietary variables that account for total fat
31 consumption, as well as that of vegetables, citrus and high-fiber grains; and (5) utilization of
June 2003 8-227 DRAFT-DO NOT QUOTE OR CITE
-------
1 recent advances in statistical modeling, including incorporation of random effects and non-
2 parametric spatial smoothing components in the Cox proportional hazards model.
3 In the extended ACS analysis, long-term exposure to air pollution, and especially to PM25,
4 was found to be associated with increased annual risk of mortality. With the longer 15-year
5 follow-up period and improved PM2 5 exposure metrics, this study detected for the first time, a
6 statistically significant association between living in a city with higher PM2 5 and increased risk
7 of dying of lung cancer. Each 10 ug/m3 increment in annual average fine PM was associated
8 with a 13 percent (95% CI=4%-23%) increase in lung cancer mortality. Coarse particles and
9 gaseous pollutants were generally not significantly associated with excess lung cancer mortality.
10 SO4"2 was significantly associated with mortality and lung cancer deaths in this extended data
11 set, yielding RR's consistent with (i.e., not significantly different from) the SO4"2 RR's reported
12 in the previously published 7-year follow-up (Pope et al, 1995). However, while PM25 was
13 specific to the causes most biologically plausible to be influenced by air pollution in this analysis
14 (i.e., cardiopulmonary and cancer), SO4"2 was significantly associated with every mortality
15 category in this new analysis, including that for "all-other causes". This suggests that the PM25
16 associations found are more biologically plausible than the less specific SO4"2 associations found.
17 The PM25 cancer risk appears greatest for non-smokers and among those with lower socio-
18 economic status (as indicated by lower educational attainment).
19 Overall, these new cohort studies confirm and strengthen the published older ecological
20 and case-control evidence indicating that living in an area that has experienced higher PM
21 exposures can cause a significant increase in the RR of lung cancer incidence and associated
22 mortality. In particular, the new ACS cohort analysis more clearly indicates that living in a city
23 with higher PM2 5 levels is associated with an elevated risk of lung cancer amounting to an
24 increase of some 10 to 15% above the lung cancer risk in a cleaner city.
25 With regard to specific ambient fine particle constituents that may significantly contribute
26 to the observed ambient PM-related increases in lung cancer, PM components of diesel engine
27 exhaust represent one class of likely important contributors. Diesel emission PM typically
28 comprises a noticeable fraction of ambient fine particles in many urban areas, having been
29 estimated to comprise from approximately 5 to 35% of ambient PM25 in some U.S. urban areas
30 (see Chapter 3). In addition, as discussed in a separate Health Effects Assessment of Diesel
31 Engine Exhaust (U.S. Environmental Protection Agency, 2002), extensive epidemiologic and
June 2003 8-228 DRAFT-DO NOT QUOTE OR CITE
-------
1 toxicologic evidence links diesel emissions (including fine PM components) to increased risk of
2 lung cancer.
3
4 8.4.4.2 PM10, PM2 5 (Fine), and PM10_2 5 (Coarse) Particulate Matter Effects on Morbidity
5 A body of new studies published since the 1996 PM AQCD provides further evidence
6 examining ambient PM association with increased human morbidity. At the time of the 1996
7 PM AQCD, fine particle morbidity studies were mostly limited to Schwartz et al. (1994) , Neas
8 et al. (1994, 1995); Koenig et al. (1993); Dockery et al. (1996); and Raizenne et al. (1996); and
9 discussion of coarse particles morbidity effects was also limited to only a few studies (Gordian
10 et al., 1996; Hefflin et al., 1994). Since the 1996 PM AQCD, several new studies have been
11 published in which newly available size-fractionated PM data allowed investigation of the
12 effects of both fine (PM25) and coarse fraction (PM10_25) particles. PM10, fine (FP) and coarse
13 fraction (CP) particle results are noted below for studies by morbidity outcome areas, as follows:
14 cardiovascular disease (CVD) hospital admissions (HA's); respiratory medical visits and
15 hospital admissions; and respiratory symptoms and pulmonary function changes.
16 As discussed in Section 8.3.1 (on cardiovascular effects associated with acute ambient PM
17 exposure), a substantial body of new results has emerged since the 1996 PM AQCD that
18 evaluates PM10 effects on cardiovascular-related hospital admissions and visits. Especially
19 notable new evidence has been provided by multi-city studies (Samet et al., 2000a,b; Zanobetti
20 and Schwartz, 2003b) that yield pooled estimates of PM-CVD effects across numerous U.S.
21 cities and regions. This study found not only significant PM associations, but also associations
22 with other gaseous pollutants as well, thus hinting at likely independent effects of certain gases
23 (O3, CO, NO2, SO2) and/or interactive effects with PM. These and other individual-city studies
24 generally appear to confirm likely excess risk of CVD-related hospital admission for U.S. cities
25 in the range of 2-9% per 50 |ig/m3 PM10, especially among the elderly (> 65 yr).
26 In addition to the PM10 studies, several new U.S. and Canadian studies evaluated fine-mode
27 PM effects on cardiovascular outcomes. Lippmann et al. (2000) and Ito (2003) report a positive
28 but not a significant association with PM2 5; and Moolgavkar (2003) reported PM2 5 to be
29 significantly associated with CVD HA for lag 0 and 1 in Los Angeles. Burnett et al. (1997a)
30 reported that fine particles were significantly associated with CVD HA in a single pollutant
31 model, but not when gases were included in multipollutant models for the 8 largest Canadian
June 2003 8-229 DRAFT-DO NOT QUOTE OR CITE
-------
1 city data. Stieb et al. (2000) reported both PM10 and PM2 5 to be associated with CVD
2 emergency department (ED) visits in single pollutant, but not multipollutant models. Similarly,
3 Morgan et al. (1998) reported that PM25 measured by nepholonetry was associated with CVD
4 HA for all ages and 65+ yr, but not in the multipollutant model. Tolbert et al. (2000a) reported
5 that coarse particles were significantly associated with dysrhythmias, whereas PM2 5 was not.
6 Other studies (e.g., Liao et al., 1999; Creason et al., 2001; Pope et al., 1999b,c) reported
7 associations between increases in PM2 5 and several measures of decreased heart rate variability,
8 but Gold et al. (2000) reported a negative association of PM25 with heart rate and decreased
9 variability in r-MSSD (one heart rate variability measure). A study by Peters and colleagues
10 (200la) reported significant temporal associations between acute (2-h or 24-h) measures of PM2 5
11 and myocardial infarction. Overall, these new studies collectively appear to implicate fine
12 particles, as well as possibly some gaseous co-pollutants, in cardiovascular morbidity, but the
13 relative contributions of fine particles acting alone or in combination with gases such as O3, CO,
14 NO2 or SO2 remain to be more clearly delineated and quantified. The most difficult issue relates
15 to interpretation of reduced PM effect size and /or statistical significance when co-pollutants
16 derived from the same source(s) as PM are included in multipollutant models.
17 Section 8.3.1 also discussed U.S. and Canadian studies that present analyses of coarse
18 fraction particles (CP) relationships to CVD outcomes. Lippmann et al. (2000) and Ito (2003)
19 found significant positive associations of PM10_2 5 with ischemic heart disease hospital
20 admissions in Detroit (RR = 1.08, CI 1.04, 1.16). Tolbert et al. (2000a) reported significant
21 positive associations of heart dysrhythmias with CP (p = 0.04) as well as for elemental carbon
22 (p = 0.004), but these preliminary results must be interpreted with caution until more complete
23 analyses are carried out and reported. Burnett et al. (1997b) noted that CP was the most robust
24 of the particle metrics examined to inclusion of gaseous covariates for cardiovascular
25 hospitalization, but concluded that particle mass and chemistry could not be identified as an
26 independent risk factor for exacerbation of cardiorespiratory disease in this study. Based on
27 another Canadian study, Burnett et al. (1999), reported statistically significant associations for
28 CP in univariate models but not in multipollutant models; but the use of estimated rather than
29 measured PM exposures indices limits the interpretation of the PM results reported.
June 2003 8-230 DRAFT-DO NOT QUOTE OR CITE
-------
1 The collective evidence reviewed above, in general, appears to suggest excess risks for
2 CVD-related hospital admissions of approximately 1 to 10% per 25 |ig/m3 PM25 or PM10_25
3 increment.
4 Section 8.3.2 also discussed new studies of effects of short-term PM10, PM25, and PM10_25
5 exposure on the incidence of respiratory hospital admissions and medical visits. Several new
6 U.S. and Canadian studies have yielded particularly interesting results that are also suggestive of
7 roles of both fine and coarse particles in respiratory-related hospital admissions. In an analysis
8 of Detroit data, Lippmann et al. (2000) and Ito (2003) found comparable effect size estimates for
9 PM2 5 and PM10_2 5. That is, the excess risk for pneumonia hospital admissions (in no co-pollutant
10 model) was 18.6% (CI 5.6, 33.1) per 50 |ig/m3 PM10, 10% (CI 1.5, 19.5) per 25 |ig/m3 PM25 and
11 11.2% (CI -0.02, 23.6) per 25 |ig/m3 PM10.25. Because PM25 and PM10.25 were not highly
12 correlated, the observed association between coarse particles and health outcomes were possibly
13 not confounded by smaller particles. Despite the greater measurement error associated with
14 PM10_2 5 than with either PM25 and PM10, this indicator of the coarse particles within the thoracic
15 fraction was associated with some of the outcome measures. The interesting result is that
16 PM10_2 5 appeared to be a separate factor from other PM metrics. Burnett et al. (1997b) also
17 reported PM (PM10, PM25, and PM10_2 5) associations with respiratory hospital admissions, even
18 with O3 in the model. Notably, the PM10_25 association was significant (RR = 1.13 for 25 |ig/m3;
19 CI = 1.05 - 1.20); and inclusion of ozone still yielded a significant coarse mass RR =1.11 (CI =
20 1.04 - 1.19). Moolgavkar (2000a) and Moolgavkar (2003) reported that, in Los Angeles, both
21 PM10 and PM25 yielded both positive and negative associations at different lags for single
22 pollutant models but not in two pollutant models. Delfino et al. (1997) reported that both PM2 5
23 and PM10 are positively associated with ED visits for respiratory disease. Morgan et al. (1998)
24 reported that PM2 5 estimated from nephelometry yielded a PM2 5 association with COPD
25 hospital admissions for 1-hr max PM that was more positive than 24-h average PM25.
26 A new study examines PM associations with asthma-related hospital admissions.
27 Sheppard et al. (1999) and Sheppard (2003) studied relationships between PM metrics that
28 included PM10_2 5 and non-elderly adult hospital admissions for asthma in the greater Seattle area
29 and reported significant relative risks for PM10, PM2 5 and PM10_2 5 (lagged 1 day). For PM10_2 5,
30 the relative risk was 1.05 (95% CI 1.0, 1.14) and for PM2.5, the relative risk 1.07(1.02, 1.11).
June 2003 8-231 DRAFT-DO NOT QUOTE OR CITE
-------
1 For a 16% decrease in PM10 levels, Friedman et al. (2001) reported decreased hospital
2 admissions for asthmatics during the Olympics in Atlanta.
3 Thus, although PM10 mass has most often been implicated as the PM pollution index
4 affecting respiratory hospital admissions, the overall collection of new studies reviewed in
5 Section 8.3.2 appear to suggest relative roles for PM10 and for both fine and coarse PM mass
6 fractions, such as PM2 5 and PM10_2 5.
7 Section 8.3.3 assessed relationships between PM exposure on lung function and respiratory
8 symptoms. While most data examine PM10 effects, several studies also examined fine and
9 coarse fraction particle effects. Schwartz and Neas (2000) report that cough was the only
10 response in which coarse fraction particles appeared to provide an independent contribution to
11 explaining the increased incidence. The correlation between CP and PM25 was moderate (0.41).
12 Coarse fraction particles had little association with evening peak flow. Tiittanen et al. (1999)
13 also reported a significant effect of PM10_25 for cough. Thus, cough may be an appropriate
14 outcome related to coarse fraction particle effects. However, the limited data base suggests that
15 further study is appropriate. The report by Zhang, et al. (2000) of an association between coarse
16 fraction particles and the indicator "runny nose" is noted also.
17 Published epidemiologic studies have collectively indicated that exposure to PM air
18 pollution can be associated with adverse human health effects, and that asthmatics represent a
19 population that can be especially affected by acute exposures to air pollution (e.g., see Koren and
20 Utell, 1997). In particular, prospective epidemiologic studies of panels of individuals confirm
21 the air pollution-asthma exacerbation association.
22 For respiratory symptoms and PFT changes, several new asthma studies report associations
23 with ambient PM measures. The peak flow analyses results for asthmatics tend to show small
24 decrements for both PM10 and PM2 5. Several studies included PM2 5 and PM10 independently in
25 their analyses of peak flow. Of these, Pekkanen et al. (1997) and Romieu et al. (1996) found
26 comparable results for PM2 5 and PM10 and the study of Peters et al. (1997c) found slightly larger
27 effects for PM25. Of studies that included both PM10 and PM25 in their analyses of respiratory
28 symptoms, the studies of Peters et al. (1997c) and found similar effects for the two PM
29 measures. Only the Romieu et al. (1996) study found slightly larger effects for PM25. While the
30 PM associations with adverse health effects among asthmatics and others are well documented,
31 the type/source(s) of those particles most associated with adverse health effects among
June 2003 8-232 DRAFT-DO NOT QUOTE OR CITE
-------
1 asthmatics are not known at this time. Indeed, the makeup of PM varies greatly from place to
2 place and over time, depending upon factors such as the sources that contribute to the pollution
3 and the prevailing atmospheric conditions, affecting particle formation, coagulation,
4 transformation, and transport. One suspected causal PM agent is the fine particle component of
5 diesel combustion exhaust.
6 Two studies (Delfmo et al., 1998; Ostro et al., 2001) examined PM effects on asthmatics
7 using one hour maximum exposure measures by TEOM, and both studies indicate a relationship
8 with measures of respiratory symptoms. Further research is needed at these shorter exposure
9 times for different PM size fractions.
10 For non-asthmatics, several studies evaluated PM2 5 effects. Naeher et al. (1999) reported
11 similar AM PEF decrements for both PM25 and PM10. Neas et al. (1996) reported a
12 nonsignificant negative association for PEF and PM213 and Neas et al. (1999) also reported
13 negative but nonsignificant PEF results. Schwartz and Neas (2000) reported a significantly PM
14 PEF association with PM25, and Tiittanen et al. (1999) also reported negative but nonsignificant
15 association for PEF andPM25. Gold et al. (1999) reported significantly PEF results. Schwartz
16 and Neas (2000) reported significant PM2 5 effects relative to lower respiratory symptoms.
17 Tiittanen et al. (1999) showed significant effects for cough and PM25 for a 4-day average.
18 The best evidence for chronic effects are found in the newer studies that combine the
19 features of cross-sectional and cohort studies. These studies include Peters et al. (1999b,c),
20 Gauderman et al. (2000), and Gauderman et al. (2002). The Gauderman studies found
21 significant decreases in lung function growth related to PM10 levels. However, Peters et al.
22 (1999) found no relationship between symptoms and PM10 levels. The cross-sectional studies by
23 Dockery et al. (1996) and Raizenne et al. (1996), reported in the previous 1996 PM AQCD,
24 found differences in peak flow and bronchitis rates associated with fine particle acidity.
25 The above new studies offer much more information than was available in 1996. Effects
26 were noted for several morbidity endpoints: cardiovascular hospital admissions, respiratory
27 hospital admissions and cough. Still insufficient data exists from these relatively limited studies
28 to allow strong conclusions at this time as to which size-related ambient PM components may be
29 most strongly related to one or another morbidity endpoints. Very preliminarily, however, fine
30 particles appear to be more strongly implicated in cardiovascular outcomes than are coarse
31 fraction particles, whereas both seem to impact respiratory endpoints.
June 2003 8-233 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.5 The Question of Lags
2 The effect of selecting lags on the resulting model for PM health effects is an important
3 issue in model selection. Using simulated data with parameters similar to a Seattle PM10_2 5 data
4 series, Lumley and Sheppard (2000) showed that the bias resulting from the selection is shown
5 to be similar in size to the relative risk estimates from the measured data. More precisely, the
6 log relative risk from the measured Seattle data is about twice the mean bias in the simulated
7 control data, and the published estimate of relative risk is only at the 90th percentile of the bias
8 distribution in these control analysis. The selection rule used was to choose the lag (between 0
9 and 6 day) with the largest estimated relative risk. In comparisons to real data from Seattle for
10 other years and from Portland, OR (with similar weather patterns to Seattle), similar bias issues
11 became evident.
12 In most of the past air pollution health effects time-series studies, after the basic model (the
13 best model with weather and seasonal cycles as covariates) was developed, several pollution lags
14 (usually 0 to 3 or 4 days) were individually introduced and the most significant lag(s) chosen for
15 the RR calculation. While this practice may bias the chance of finding a significant association,
16 without a firm biological reason to establish a fixed pre-determined lag, it appears reasonable.
17 Due to likely individual variability in response to air pollution, the apparent lags of effects
18 observed for aggregated population counts are expected to be "distributed" (i.e., symmetric or
19 skewed bell-shape). The "most significant lag" in such distributed lags is also expected to
20 fluctuate statistically. The "vote-counting" of the most significant lags reported in the past
21 PM-mortality studies shows that 0 and 1 day lags are, in that order, the most frequently reported
22 "optimal" lags, but such estimates may be biased because these lags are also likely the most
23 frequently examined ones. Thus, a more systematic approach across different data sets was
24 needed to investigate this issue.
25 The Samet et al. (2000b) analysis, and the reanalysis by Dominici et al. (2002), of the
26 90 largest U.S. cities provides particularly useful information on this matter. Figure 8-19 depicts
27 the Dominici et al. (2002) overall pooled results, showing the posterior distribution of PM10
28 effects for the 90 cities for lag 0, 1, and 2 days. It can be seen that the effect size estimate for lag
29 Iday is about twice that for lag 0 or lag 2 days, although their distributions overlap. The pattern
30 of lagged effects pooled for each of the seven regions (see Figure 8-5) in the 90 cities study also
31 shows that the lag with the largest effect was at 1 day, with the exception of Upper Midwest
June 2003 8-234 DRAFT-DO NOT QUOTE OR CITE
-------
Re-analyzed Pooled Estimates
0.4 0.6
%/10m(jg/m3
0.8
1.0
Figure 8-19. Marginal posterior distribution for effects of PM10 on all cause mortality at
lag 0,1, and 2 for the 90 cities. From Dominici et al. (2002a). The numbers
in the upper right legend are posterior probabilities that overall effects are
greater than 0.
Source: Dominici et al. (2002).
1 where the estimated PM10 effect was about the same for lag 0 and 1 days. However, the studies
2 that examined PM-mortality associations in individual cities sometimes show the "most
3 significant lags" at other lags. For example, in Moolgavkar's analysis of Los Angeles data (2000
4 and reanalysis 2003), both total non-accidental mortality and cardiovascular mortality showed
5 the strongest associations with PM10 at lag 2 days.
6 A review of current studies on the short-term adverse health effects of air pollution
7 indicates that there are essentially three different approaches to deal with temporal structure:
8 (1) assume all sites have the same lag (e.g., 1 day, for a given effect); (2) use the lag or moving
9 average giving the largest or most significant effect and for each pollutant and endpoint; and
June 2003
8-235
DRAFT-DO NOT QUOTE OR CITE
-------
1 (3) use a flexible distributed lag model, with parameters adjusted to each site. The NMMAPS
2 mortality analyses used the first approach. This approach introduces a consistent response
3 model across all locations. However, since the cardiovascular, respiratory, or other causes of
4 acute mortality usually associated with PM are not at all specific, there is little a priori reason to
5 believe that they must have the same relation to current or previous PM exposures at different
6 sites. The obvious advantage of the first approach in dealing with multi-city data is its
7 consistency in summarizing the point estimate. The major factor that makes it difficult to
8 conduct a meta-analysis of existing PM health effects studies is the lack of consistency in the
9 way lag structures were modeled across the studies.
10 The approach used in most of PM time-series studies is to use the model that maximizes
11 some global model goodness-of-fit criterion. This leads to selection of different models at
12 different sites, as might be expected. However, the best-fitting model (for lags, for example) is
13 often the model with the largest or most significant PM10 coefficient (i.e., the approach
14 [2] above). All models for the pollutant(s) of interest are usually compared among themselves
15 only after a preliminary baseline model has been fitted. The baseline model takes into account
16 most of the other variables with which PM10 could be plausibly associated, so that the remaining
17 variation in morbidity or mortality that can be explained by including PM10 indicators with
18 different temporal structures is nearly "orthogonal" or independent of the baseline model. The
19 restriction to the same lag day at all sites certainly increases the precision of that estimate, but
20 possibly at the cost of obscuring different relationships between time of exposure and health
21 effect at other sites.
22 An additional complication in assessing the shape of a distributed lag is that the apparent
23 spread of the distributed lag may depend on the pattern of persistence of air pollution (i.e.,
24 episodes may persist for a few days), which may vary from city to city and from pollutant to
25 pollutant. If this is the case, fixing the lag across cities or across pollutants may not be ideal, and
26 may tend to obscure important nuances of lag structures that may provide important clues to
27 possible different lags between PM exposures and different cause-specific effects.
28 It should also be noted that if one chooses the most significant single lag day only, and if
29 more than one lag day shows positive (significant or otherwise) associations with mortality, then
30 reporting a RR for only one lag would also underestimate the pollution effects. Schwartz
31 (2000b; reanalysis 2003b) investigated this issue, using the 10 U.S. cities data where daily PM10
June 2003 8-236 DRAFT-DO NOT QUOTE OR CITE
-------
1 values were available for 1986-1993. Daily total (non-accidental) deaths of persons 65 years of
2 age and older were analyzed. For each city, a GAM Poisson model (with stringent convergence
3 criteria) and penalized splines adjusting for temperature, dewpoint, barometric pressure, day-of-
4 week, season, and time were fitted. Effects of distributed lag were examined using two models:
5 second-degree distributed lag model using lags 0 through 5 days; and unconstrained distributed
6 lag model using lags 0 through 5 days. The inverse variance weighted averages of the ten cities'
7 estimates were used to combine results. The results indicated that the effect size estimates for
8 the quadratic distributed model and unconstrained distributed lag model using GAM were
9 similar: 6.3% (95% CI: 4.9-7.8) per 50 |ig/m3 increase for the quadratic distributed lag model,
10 and 5.8% (95% CI: 4.4-7.3). These risk estimates are about twice as large as the two-day
11 average (lag 0 and 1 day) estimate (3.4%; 95% CI: 2.6-4.1) obtained in the reanalysis of the
12 original 10 cities study (Schwartz, 2003b). There are indications that such distributed lag
13 estimates are even larger when more specific cause of deaths are examined (see US 10 cities
14 study description in section 8.2.2.3).
15 Mis-specification of the lag structure may cause important modeling biases. Most of the
16 published literature for the U.S. evaluates only single-day models, a choice dictated by the
17 every-sixth-day sampling schedule used for PM10 in many U.S. cities. When this occurs, it is not
18 possible to evaluate multi-day models with greater biological plausibility, such as moving
19 average models and distributed lag models. It should also be noted that, with the every-sixth-day
20 PM data, a different set of days of mortality series were evaluated at each lag. An every-other-
21 day sampling schedule was used in the Harvard Six City Study, for which the PM data on a
22 given day has been used as though it were a two-day moving, alternately concurrent with
23 mortality on half the days and lagging mortality by one day on the other days. While the most
24 commonly used lags in PM time-series models are zero or one day, some studies have found PM
25 effects with longer lags (e.g., Wichmann et al. (2000) and reanalysis by Stolzel et al. (2003);
26 Lippmann et al. (2000) and reanalysis by Ito (2003). It is plausible that mortality or hospital
27 admissions from PM may arise from different responses or PM-associated diseases with
28 different characteristic lags, for example, that cardiovascular responses may arise almost
29 immediately after exposure, within zero or one days or even within two hours (Peter et al.,
30 2001a, for myocardial infarction). One would then expect to see different best-fitting lags for
31 different cause-specific mortality or hospital admissions.
June 2003 8-237 DRAFT-DO NOT QUOTE OR CITE
-------
1 In summary, the largest time-series study to date (90 cities study) indicated that, of the 0, 1,
2 and 2 day PM10 lags examined, lag 1 day showed the strongest mortality associations. However,
3 other lags are reported for various mortality and morbidity outcomes from studies that examined
4 individual cities' data. Examinations of lag structures are often limited by the prevailing every-
5 6th-day sampling schedule for PM in the U.S., but a limited number of studies that examined
6 daily PM data using distributed lag model suggest that multi-day effects are larger than the
7 single-day effects. Thus, it is possible that current PM risk estimates, most frequently computed
8 for a single day or for two-day averages, may be underestimating these multi-day effects.
9
10 8.4.6 Concentration-Response Relationships for Ambient PM
11 In the 1996 PM AQCD, the limitations of identifying 'threshold' in the concentration-
12 response relationships in observational studies were discussed including the low data density in
13 the lower PM concentration range, the small number of quantile indicators often used, and the
14 possible influence of measurement error. Also, a threshold for a population, as opposed to a
15 threshold for an individual, has some conceptual issues that need to be noted. For example,
16 Schwartz (1999) discussed that, since individual thresholds would vary from person to person
17 due to individual differences in genetic level susceptibility and pre-existing disease conditions,
18 it would be almost mathematically impossible for a threshold to exist in the population. This
19 argument holds only if the most sensitive members of a population are sensitive to very low
20 concentrations, which may not be the case. The person-to-person difference in the relationship
21 between personal exposure and the concentration observed at a monitor would also add to the
22 variability. Because one cannot directly measure but can only compute or estimate a population
23 threshold, it would be difficult to interpret an observed threshold, if any, biologically. Despite
24 these issues, several studies have attempted to address the question of threshold by analyzing
25 large databases, or by conducting simulations.
26 Daniels et al. (2000; reanalysis by Dominici et al., 2003) examined the presence of
27 threshold using the largest 20 U.S. cities for 1987-1994. In the original analysis, the authors
28 compared three log-linear GAM regression models: (1) using a linear PM10 term; (2) using a
29 natural cubic spline of PM10 with knots at 30 and 60 |ig/m3 (corresponding approximately to
30 25 and 75 percentile of the distribution); and, (3) using a threshold model with a grid search in
31 the range between 5 and 200 |ig/m3 with 5 |ig/m3 increment. The covariates included in these
June 2003 8-23 8 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
models are similar to those used by the same research group previously (Kelsall et al., 1997;
Samet et al., 2000a,b), including the smoothing function of time, temperature and dewpoint, and
day-of-week indicators. In the reanalysis, the covariate adjustments were made using natural
splines in GLM models. Total, cardiorespiratory, and other mortality series were analyzed.
These models were fit for each city separately, and for model (1) and (2) the combined estimates
across cities were obtained by using inverse variance weighting if there was no heterogeneity
across cities, or by using a two-level hierarchical model if there was heterogeneity. The best fit
among the models, within each city and over all cities, were also determined using the Akaike's
Information Criterion (AIC). The results using the natural spline model showed that, for total
and cardiorespiratory mortality, the spline curves were roughly linear, consistent with the lack of
a threshold (see Figure 8-20). For mortality from other causes, however, the curve did not
increase until PM10 concentrations exceeded 50 |ig/m3. The hypothesis of linearity was
examined by comparing the AIC values across models. The results suggested that the linear
model was preferred over the spline and the threshold models. Thus, these results suggest that
linear models without a threshold may well be appropriate for estimating the effects of PM10 on
the types of mortality of main interest.
SIS
Pi/m
PS/in
pgftn
Figure 8-20. Particulate matter < 10 jam in aerodynamic diameter (PM10)-total mortality
concentration-response curves for total (TOTAL) mortality, cardiovascular
and respiratory (CVDRESP) mortality, and other causes (OTHERS)
mortality, 20 largest US cities, 1987-1994. The concentration-response
curves for the mean lag, current day, and previous day PM10 are denoted by
solid lines, squared points, and triangle points, respectively.
Source: Dominici et al. (2003).
June 2003
8-239
DRAFT-DO NOT QUOTE OR CITE
-------
1 Cakmak et al. (1999) investigated methods to detect and estimate threshold levels in time-
2 series studies. Based on the realistic range of error observed from actual Toronto pollution data
3 (average site-to-site correlation: 0.90 for O3; 0.76 for CoH; 0.69 for TSP; 0.59 for SO2; 0.58 for
4 NO2; and 0.44 for CO), pollution levels were generated with multiplicative error for six levels of
5 exposure error (1.0, 0.9, 0.8, 0.72, 0.6, 0.4, site-to-site correlation). Mortality series were
6 generated with three PM10 threshold levels (12.8 |ig/m3, 24.6 |ig/m3, and 34.4 |ig/m3). LOESS
7 with a 60% span was used to observe the exposure-response curves for these 18 combinations of
8 exposure-response relationships with error. A parameter threshold model was also fit using non-
9 linear least squares. Both mortality and PM10 data were pre-filtered for the influence of seasonal
10 cycles using LOESS smooth function. The threshold regression models were then fit to the
11 pre-filtered data. Graphical presentations indicate that LOESS adequately detects threshold
12 under no error, but the thresholds were "smoothed out" under the extreme error scenario. Use of
13 a parametric threshold model was adequate to give "nearly unbiased" estimates of threshold
14 concentrations even under the conditions of extreme measurement error, but the uncertainty in
15 the threshold estimates increased with the degree of error. They concluded, "if threshold exists,
16 it is highly likely that standard statistical analysis can detect it."
17 The Smith et al. (2000) study of associations between daily total mortality and PM25 and
18 PM10_25 in Phoenix, AZ (during 1995-1997) also investigated the possibility of a threshold.
19 In the linear model, the authors found that mortality was significantly associated with PM10_25,
20 but not with PM2 5. In modeling possible thresholds, they applied: (1) a piecewise linear model
21 in which several possible thresholds were specified; and (2) a B-spline (spline with cubic
22 polynomials) model with 4 knots. Using the piecewise model, there was no indication that there
23 was a threshold for PM10_2 5 However, for PM2 5, the piecewise model resulted in suggestive
24 evidence for a threshold, around 20 to 25 |ig/m3. The B-spline results also showed no evidence
25 of threshold for PM10_2 5, but for PM2 5, a non-linear curve showed a change in the slope around
26 20 |ig/m3. A further Bayesian analysis for threshold selection suggested a clear peak in the
27 posterior density of PM2 5 effects around 22 |ig/m3. These results, if they in fact reflect reality,
28 make it difficult to evaluate the relative roles of different PM components (in this case, PM2 5
29 versus PM10_2 5). However, the concentration-response curve for PM2 5 presented in this
30 publication suggests more of a U- or V-shaped relationship than the usual "hockey stick"
31 relationship. Such a relationship is, unlike the temperature-mortality relationship, difficult to
June 2003 8-240 DRAFT-DO NOT QUOTE OR CITE
-------
1 interpret biologically. Because the sample size of this data (3 years) is relatively small, further
2 investigation of this issue using similar methods but a larger data set is warranted. Other studies
3 evaluate non-linear relationships using a multi-city meta-smoothing approach based on non- or
4 semi-parametric smoothers rather than on linear parametric models.
5 Smith et al. (1999) analyzed PM10-mortality association in Birmingham, AL and Cook
6 County, IL. Temperature was modeled using piece-wise linear term with a change point. PM10
7 were modeled at lag 0 through 3 and 3-day averages at these lags. In addition to the linear
8 model, they also investigated the existence of a threshold using B-splines and a parametric
9 threshold model with the profile log likelihood evaluated at changing threshold points. B-splines
10 results suggest that an increasing effect above 80|ig/m3 for Birmingham, and above 100 |ig/m3
11 for Chicago. The threshold model through examination of log likelihood across the range of
12 threshold levels also suggested similar change points, but not to the extent that could achieve
13 statistical distinctions.
14 In summary, the results from large multi-city studies suggest that there is no strong
15 evidence for a threshold mortality effect of PM. Some single city studies suggest a hint of a
16 threshold, but not in a statistically clear manner. More data may need to be examined with
17 alternative approaches (e.g., Smith et al.'s parametric model), but meanwhile, the use of linear
18 PM effect model appears to be appropriate.
19
20 8.4.7 Heterogeneity of Particulate Matter Effects Estimates
21 Approximately 35 then-available acute PM exposure community epidemiologic studies
22 were assessed in the 1996 PM AQCD as collectively demonstrating increased risks of mortality
23 being associated with short-term (24-h) PM exposures indexed by various ambient PM
24 measurement indices (e.g., PM10, PM25, BS, CoH, sulfates, etc.) in many different cities in the
25 United States and internationally. Much homogeneity appeared to exist across various
26 geographic locations, with many studies suggesting, for example, increased relative risk (RR)
27 estimates for total nonaccidental mortality on the order of 1.025 to 1.05 (or 2.5 to 5.0% excess
28 deaths) per 50 |ig/m3 increase in 24-h PM10, with statistically significant results extending more
29 broadly in the range of 1.5 to 8.0%. The elderly > 65 yrs. old and those with preexisting
30 cardiopulmonary conditions had somewhat higher excess risks. One study, the Harvard Six City
June 2003 8-241 DRAFT-DO NOT QUOTE OR CITE
-------
1 Study, also provided estimates of increased RR for total mortality falling in the range of 1.02 to
2 1.056 (2.0 to 5.6% excess deaths) per 25 |ig/m3 24-h PM2 5 increment.
3 Now, more than 80 new time-series PM-mortality studies assessed earlier in this chapter
4 provide extensive additional evidence which, qualitatively, largely substantiates significant
5 ambient PM-mortality relationships, again based on 24-h exposures indexed by a wide variety of
6 PM metrics in many different cities of the United States, in Canada, in Mexico, and elsewhere
7 (in South America, Europe, Asia, etc.). The newly available effect size estimates from such
8 studies are reasonably consistent with the ranges derived from the earlier studies reviewed in the
9 1996 PM AQCD. For example, newly estimated PM10 effects generally fall in the range of 1.0 to
10 8.0% excess deaths per 50 |ig/m3 PM10 increment in 24-h concentration; and new PM2 5 excess
11 estimates for short-term exposures generally fall in the range of 2 to 8% per 25 |ig/m3 increment
12 in 24-h PM2 5 concentration.
13 However, somewhat greater spatial heterogeneity appears to exist across newly reported
14 study results, both with regard to PM-mortality and morbidity effects. The newly apparent
15 heterogeneity of findings across locations is perhaps most notable in relation to reports based on
16 multiple-city studies in which investigators used the same analytical strategies and models
17 adjusted for the same or similar co-pollutants and meteorological conditions, raising the
18 possibility of different findings reflecting real location-specific differences in exposure-response
19 relationships rather than potential differences in models used, pollutants measured and included
20 in the models, etc. Some examples of newly reported and well-conducted multiple-city studies
21 include: the NMMAPS analyses of mortality and morbidity in 20 and 90 U.S. cities (Samet
22 et al., 2000a,b; Dominici et al., 2000a); the Schwartz (2000b,c) analyses of 10 U.S. cities; the
23 study of eight largest Canadian cities (Burnett et al., 2000); the study of hospital admissions in
24 eight U.S. counties (Schwartz, 1999); and the APHEA studies of mortality and morbidity in
25 several European cities (Katsouyanni et al., 1997; Zmirou et al., 1998). The recently completed
26 large NMMAPS studies of morbidity and mortality in U.S. cities add especially useful and
27 important information about potential U.S. within- and between-region heterogeneity.
28 HEI (2003a) concluded that after examining the NMMAPS GAM reanalyses by Dominici
29 et al. (2002) that while formal tests of PM effects across cities did not indicate evidence of
30 heterogeneity because of the individual-city effects standard error being generally large that the
June 2003 8-242 DRAFT-DO NOT QUOTE OR CITE
-------
1 power to assess the presence of heterogeneity was low and, as such, the possibility of
2 heterogeneity still exists.
3
4 8.4.7.1 Evaluation of Heterogeneity of Particulate Matter Mortality Effect Estimates
5 In all of the U.S. multi-city analyses, the heterogeneity in the PM estimates across cities
6 was not explained by city-specific characteristics in the 2nd stage model. The heterogeneity of
7 effects estimates across cities in the multi-city analyses may be due to chance alone, to mis-
8 specification of covariate effects in small cities, or to real differences from location to location in
9 effects of different location-specific ambient PM mixes, for which no mechanistic explanations
10 are yet known. Or, the apparent heterogeneity may simply reflect imprecise PM effect estimates
11 derived from smaller-sized analyses of less extensive available air pollution data or numbers of
12 deaths in some cities tending to obscure more precise effects estimates from larger-size analyses
13 for other locations, which tend to be consistently more positive and statistically significant.
14 Some of these possibilities can be evaluated by using data from the NMMAPS study
15 (Samet et al., 2000b). Data in Figure 8-3 for excess risk and 95% confidence intervals were
16 plotted against the total number of effective observations, measured by the number of days of
17 PM10 data times the mean number of daily deaths in the community. This provides a useful
18 measure of the weight that might be assigned to the results, since the uncertainty of the RR
19 estimate based on a Poisson mean is roughly inversely proportional to this product. That is, the
20 expected pattern typically shows less spread of estimated excess risk with increasing death-days
21 of data. A more refined weight index would also include the spread in the distribution of PM
22 concentrations. The results are plotted in Figure 8-21 for all cities and Figure 8-22 for each of
23 the 7 regions.
24 Figure 8-21 for all cities suggests some relationship between precision of the effects
25 estimates and study weight, overall. That is, the more the mortality-days observations, the
26 narrower the 95% confidence intervals and the more precise the effects estimates (with nearly all
27 these for cities with > log 9 mortality-days being positive and many statistically significant at
28 p < 0.05).
29 The Figure 8-22 depiction for each of the 7 regions is also informative. In the Northeast,
30 there is considerable homogeneity (not heterogeneity) of effect size for larger study-size cities,
31 even with moderately wide confidence intervals for those with log mortality-days = 8 to 9, and
June 2003 8-243 DRAFT-DO NOT QUOTE OR CITE
-------
11 -
-11
7 8 9 10 11
Natural Log of Mortality - Days
12
13
Figure 8-21. An EPA-derived plot showing relationship of PM10 total mortality effects
estimates and 95% confidence intervals for all cities in the Dominici et al.
(2000a; 2003) NMMAPS 90-cities analyses in relation to study size (i.e., the
natural logarithm or numbers of deaths times days of PM observations).
Note the generally narrower confidence intervals for more homogeneously
positive effects estimates as study size increases beyond about In (mortality-
days) = 9.0 (i.e., beyond about 8,000 deaths-days of observation). The dashed
line depicts the overall nationwide effect estimate (grand mean) of
approximately 0.28% per 10 ug/m3 PM10 for models with no co-pollutants.
June 2003
8-244
DRAFT-DO NOT QUOTE OR CITE
-------
s
Q.
CO
O)
1>
Q.
-------
1 all clearly exceed the overall nationwide grand mean indicated by the dashed line. On the other
2 hand, the smaller study-size Northeast cities (with much wider confidence intervals at log
3 < 8) show much greater heterogeneity of effects estimates and less precision. Also, most of the
4 estimates for larger study-size (log > 9) cities in the industrial midwest are positive and several
5 statistically significant, so that an overall significant regional risk is plausible there as well.
6 There may even be some tendency for relatively large risks for some cities with small study sizes
7 and wide confidence intervals in the industrial midwest, and further investigation of that would
8 be of interest. The plot for Southern California in Figure 8-22 clearly shows a rather consistent
9 estimate of effect size and width of the confidence intervals across cities of varying study-size.
10 All risk estimates are positive and most are significant at p < 0.05 or nearly so for the Southern
11 California cities. For Northwestern cities plotted in Figure 8-22, the value for Oakland, CA
12 (at ca. log 9.5) is notable (it being very positive and significant), whereas many but not all of the
13 other cities have positive effect estimates not too far off the nationwide grand mean, but with
14 sufficiently wide confidence intervals so as not to be statistically significant at p < 0.05. The
15 Southwestern cities, too, mostly appear to have effect sizes near the nationwide mean, but with
16 confidence intervals too wide to be significant at p < 0.05. The "Other" (non-industrial or
17 "Upper," as per NMMAPS) Midwest cities and the Southeastern cities in Figure 8-22 show more
18 heterogeneity, although most of the larger study size cities (log > 9.0) tend to be positive and not
19 far off the nationwide mean (even though not significant at p < 0.05). Given the wide range of
20 effects estimates and confidence intervals seen for Southeastern cities, further splitting of the
21 region might be informative.
22 In fact, closer reexamination of results for each of the regions may reveal interesting new
23 insights into what factors may account for any apparent disparities among the cities within a
24 given region or across regions. Several possibilities readily come to mind. First, cursory
25 inspection of the mean PM10 levels shown for each city in (Samet et al., 2000b; Appendix A)
26 suggests that many of the cities showing low effects estimates and wide confidence intervals
27 tend to be among those having the lowest mean PM10 levels and, therefore, likely the smallest
28 range of PM10 values across which to distinguish any PM-related effect, if present. It may also
29 be possible that those areas with higher PM2 5 proportions of PM10 mass (i.e., larger percentages
30 of fine particles) may show higher effects estimates (e.g., in Northeastern cities) than those with
31 higher coarse-mode fractions (e.g., as would be more typical of Southwestern cities). Also, more
June 2003 8-246 DRAFT-DO NOT QUOTE OR CITE
-------
1 industrialized cities with greater fine-particle emissions from coal combustion (e.g., in the
2 industrial Midwest) and/or those with high fine-particle emissions from heavy motor vehicle
3 emissions (e.g., typical of Southern California cities) may show larger PM10 effects estimates
4 than other cities. Lastly, the extent of air-conditioning use may also account for some of the
5 differences, with greater use in many Southeastern and Southwestern cities perhaps decreasing
6 actual human exposure to ambient particles present versus higher personal exposure to ambient
7 PM (including indoors) in those areas where less air-conditioning is used (e.g., the Northeast and
8 industrial Midwest).
9
10 8.4.7.2 Comparison of Spatial Relationships in the NMMAPS and Cohort Reanalyses
11 Studies
12 Both the NMMAPS and HEI Cohort Reanalyses studies had a sufficiently large number of
13 U.S. cities to allow considerable resolution of regional PM effects within the "lower 48" states,
14 but an attempt was made to take this approach to a much more detailed level in the Cohort
15 Reanalysis studies than in NMMAPS. There were: 88 cities with PM10 effect size estimates in
16 NMMAPS; 50 cities with PM25 and 151 cities with sulfates in the original Pope et al. (1995)
17 ACS analyses and in the HEI reanalyses using the original data; and 63 cities with PM2 5 data
18 and 144 cities with sulfate data in the additional analyses done by the HEI Cohort Reanalysis
19 team. The relatively large number of data points utilized in the HEL reanalyses effort and
20 additional analyses allowed estimation of surfaces for elevated long-term concentrations of
21 PM25, sulfates, and SO2 with resolution on a scale of a few tens to hundreds of kilometers.
22 The patterns for PM2 5 and sulfates are similar, but not identical. In particular, the modeled
23 PM25 surface (Krewski et al., 2000; Figure 18) had peak levels around Chicago - Gary, in the
24 eastern Kentucky - Cleveland region, and around Birmingham AL, with elevated but lower PM2 5
25 almost everywhere east of the Mississippi, as well as southern California. This is similar to the
26 modeled sulfate surface (Krewski et al., 2000; Figure 16), with the absence of a peak in
27 Birmingham and an emerging sulfate peak in Atlanta. The only area with markedly elevated
28 SO2 concentrations was the Cleveland - Pittsburgh region. Secondary sulfates in particles
29 derived from local SO2 appeared more likely to be important in the industrial midwest, south
30 from the Chicago - Gary region into Ohio, northeastern Kentucky, West Virginia, and southwest
31 Pennsylvania, possibly related to combustion of high-sulfur fuels.
June 2003 8-247 DRAFT-DO NOT QUOTE OR CITE
-------
1 The overlay of mortality with air pollution patterns is also of much interest. The spatial
2 overlay of long-term PM25 and mortality (Krewski et al., 2000; Figure 21) was highest from
3 southern Ohio to northeastern Kentucky/West Virginia, but also included a significant
4 association over most of the industrial midwest. This was reflected, in diminished form, by the
5 sulfates and SO2 maps (Krewski et al., 2000; Figures 19 and 20), where there appeared to be a
6 somewhat tighter focus of elevated risk in the upper Ohio River Valley area. This suggests that,
7 while SO2 was an important precursor of sulfates in this region, there may also be some other
8 (non-sulfur) contributors to associations between PM2 5 and long-term mortality, encompassing a
9 wide area of the North Central Midwest and non-coastal Mid-Atlantic region.
10 The apparent differences in PM10 and/or PM2 5 effect sizes across different regions should
11 not be attributed merely to possible variations in measurement error or other statistical
12 artifact(s). Some of these differences may reflect: real regional differences in particle
13 composition or co-pollutant mix; differences in relative human exposures to ambient particles or
14 other gaseous pollutants; sociodemographic differences (e.g., percent of infants or elderly in
15 regional population); or other important, as of yet unidentified PM effect modifiers.
16 In their reanalyses of daily mortality in eight Canadian cities, Burnett and Goldberg (2003)
17 report positive estimates of heterogeneity of particulate effects across cities using LOESS,
18 whereas negative estimates of heterogeneity were obtained using natural splines. They stated
19 that this finding was due to the reduction in effect estimate using natural splines that resulted in
20 smaller observed variation in effect estimates across cities in addition to the increased within-
21 city estimate error compared to models using LOESS for time and weather. However, Burnett
22 and Goldberg (2003) ultimately concluded that evidence from their study is insufficient to
23 conclude that the PM association with mortality varies across Canadian cities.
24
25 8.4.8 New Assessments of Measurement Error Consequences
26 8.4.8.1 Theoretical Framework for Assessment of Measurement Error
27 Since the 1996 PM AQCD, advances have been made in conceptual framework
28 development to investigate effects of measurement error on PM health effects estimated in time-
29 series studies. Several new studies evaluate the extent of bias caused by measurement errors
30 under scenarios with varying extent of error variance and covariance structure between co-
31 pollutants.
June 2003 8-248 DRAFT-DO NOT QUOTE OR CITE
-------
1 Zidek et al. (1996) investigated, through simulation, the joint effects of multi-collinearity
2 and measurement error in Poisson regression model, with two covariates with varying extent of
3 relative errors and correlation. Their error model was of classical error form (W = X + U, where
4 W and X are surrogate and true measurements, respectively, and the error U is normally
5 distributed). The results illustrated the transfer of effects from the "causal" variable to the
6 confounder. However, for the confounder to have larger coefficients than the true predictor, the
7 correlation between the two covariates had to be large (r = 0.9), with moderate error (a > 0.5) for
8 the true predictor, and no error for the confounder in their scenarios. The transfer-of-causality
9 effect was mitigated when the confounder also became subject to error. Another interesting
10 finding that Zidek et al. reported is the behavior of the standard errors of these coefficients:
11 when the correlation between the covariates was high (r = 0.9) and both covariates had no error,
12 the standard errors for both coefficients were inflated by factor of 2; however, this phenomenon
13 disappeared when the confounder had error. Thus, multi-collinearity influences the significance
14 of the coefficient of the causal variable only when the confounder is accurately measured.
15 Marcus and Chapman (1998) also conducted a mathematical analysis of PM mortality
16 effects in ordinary least square model (OLS) with the classical error model, under varying extent
17 of error variance and correlation between two predictor variables. The error described here was
18 analytical error (e.g., discrepancy between the co-located monitors). In general, they found that
19 positive regression coefficients are only attenuated; and null predictors (zero coefficient) or
20 weak predictors are only able to appear stronger than true positive predictors under unusual
21 conditions: (1) true predictors must have very large positive or negative correlation (i.e.,
22 r| > 0.9); (2) measurement error must be substantial (i.e., error variance ~ signal variance); and
23 (3) measurement errors must have a large negative correlation. They concluded that estimated
24 FP health effects are likely underestimated, although the magnitude of bias due to the analytical
25 measurement error is not very large.
26 Zeger et al. (2000) illustrated the implication of the classical error model and the Berkson
27 error model (i.e., X = W + U) in the context of time-series study design. Their simulation of the
28 classical error model with two predictors, with various combinations of error variance and
29 correlation between the predictors/error terms, showed results similar to those reported by Zidek
30 et al. (1996). Most notably, for the transfer of the effects of one variable to the other (i.e., error-
31 induced confounding) to be large, the two predictors or their errors must to be substantially
June 2003 8-249 DRAFT-DO NOT QUOTE OR CITE
-------
1 correlated. Also, for the spurious association of a null predictor to be more significant than the
2 true predictor, their measurement errors have to be extremely negatively correlated—a condition
3 not yet seen in actual air pollution data sets.
4 Zeger et al. (2000) also laid out a comprehensive framework for evaluating effects of
5 exposure measurement error on estimates of air pollution mortality relative risks in time-series
6 studies. The error, i.e., the difference between personal exposure and a central station's
7 measurement of ambient pollutant concentration, was decomposed into three components:
8 (1) the error due to having aggregate rather than individual exposure; (2) the difference between
9 the average personal exposure and the true ambient concentration level; and, (3) the difference
10 between the true and measured ambient concentration level. By aggregating individual risks to
11 obtain expected number of deaths, they showed that the first component of error (the aggregate
12 rather than individual) is a Berkson error, and, therefore is not a significant contributor to bias in
13 the estimated risk. The second error component is a classical error and can introduce bias if
14 there are short-term associations between indoor source contributions and ambient concentration
15 levels. Recent analysis, however, both using experimental data (Mage et al., 1999; Wilson et al.,
16 2000) and theoretical interpretations and models (Ott et al., 2000) indicate that there is no
17 relationship between the ambient concentration and the nonambient components of personal
18 exposure to PM. Still, a bias could arise due to the difference between the personal exposure to
19 ambient PM (indoors plus outdoors) and the ambient concentration. The third error component
20 is the difference between the true and the measured ambient concentration. According to Zeger
21 et al. the final term is largely of the Berkson type if the average of the available monitors is an
22 unbiased estimate of the true spatially averaged ambient level.
23 Using this framework, Zeger et al. (2000) then used PTEAM Riverside, CA data to
24 estimate the second error component and its influence on estimated risks. The correlation
25 coefficient between the error (the average population PM10 total exposure minus the ambient
26 PM10 concentration) and the ambient PM10 concentration was estimated to be -0.63. Since this
y*.
27 correlation is negative, the/?,, (the estimated value of the pollution-mortality relative risk in the
28 regression of mortality on zt, the daily ambient concentration) will tend to underestimate the
y*.
29 coefficient/^ that would be obtained in the regression of mortality on Xt, the daily average total
30 personal exposure, in a single-pollutant analysis. Zeger et al. (2000) then proceeded to assess
31 the size of the bias that will result from this exposure misclassification, using daily ambient
June 2003 8-250 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3 a
4
concentration, zt. As shown in Equation 9, the daily average total personal exposure, Xt, can be
separated into a variable component, 6j zt, dependent on the daily ambient concentration, zt, and
0 constant component, 60, independent of the ambient concentration:
(8-5)
7 where et is an error term.
8 If the nonambient component of the total personal exposure is independent of the ambient
9 concentration, as appears to be the case, Equation 9 from Zeger et al. (2000) becomes the
10 regression analysis equation familiar to exposure analysts (Dockery and Spengler, 1981; Ott
11 et al., 2000; Wilson et al., 2000). In this case, 60 gives the average nonambient component of the
12 total personal exposure and 6j gives the ratio of the ambient component of personal exposure to
13 the ambient concentration. (The ambient component of personal exposure includes exposure to
14 ambient PM while outdoors and, while indoors, exposure to ambient PM that has infiltrated
15 indoors.) In this well-known approach to adjust for exposure measurement error, called
yv yv y».
16 regression calibration (Carroll et al., 1995), the estimate of /3Xhas the simple form ^X = ^Z/9V.
17 Thus, for the regression calibration, the value of /3X (based on the total personal exposure) does
18 not depend on the total personal exposure but is given by /?z, based on the ambient concentration,
19 times 9l5 the ratio of the ambient component of personal exposure to the ambient concentration.
20 A regression analysis of the PTEAM data gave an estimate Ql = 0.60.
ys.
21 Zeger et al. (2000) used Equation 9, with 90 = 59.95 and Ql = 0.60, estimated from the
22 PTEAM data, to simulate values of daily average personal exposure, x*t, from the ambient
23 concentrations, zft for PM10 in Riverside, CA, 1987-1994. They then compared the mean of the
24 simulated fix s, obtained by the series of log-linear regressions of mortality on the simulated x*t,
y*.
25 with the normal approximation of the likelihood function for the coefficient Pz from the
26 log-linear regression of mortality directly on zt. The resulting ffz I fix = 0.59 is very close to
yv
27 01 = 0.60. Dominici et al. (2000b) provide a more complete analysis of the bias in/?z as an
28 estimate of fix using the PTEAM Study and four other data sets and a more complete statistical
29 model. Their findings were qualitatively similar in that was close to /Q^ Thus, it appears that
30 the bias is very close to 6b which depends not on the total personal exposure but only on the
31 ratio of the ambient component of personal exposure to the ambient concentration.
June 2003 8-251 DRAFT-DO NOT QUOTE OR CITE
-------
1 Zeger et al. (2000), in the analyses described above, also suggested that the error due to the
2 difference between the average personal exposure and the ambient level (the second error type
3 described above) is likely the largest source of bias in estimated relative risk. This suggestion at
4 least partly comes from the comparison of PTEAM data and site-to-site correlation (the third
5 type of error described above) for PM10 and O3 in 8 US cities. While PM10 and O3 both showed
6 relatively high site-to-site correlation (» 0.6-0.9), a similar extent of site-to-site correlation for
7 other pollutants is not necessarily expected. Ito et al. (2000) estimated site-to-site correlations
8 (after adjusting for seasonal cycles) for PM10, O3, SO2, NO2, CO, temperature, dewpoint
9 temperature, and relative humidity, using multiple stations' data from seven central and eastern
10 states (IL, IN, MI, OH, PA, WV, WI), and found that, in a geographic scale of less 100 miles,
11 these variables could be categorized into three groups in terms of the extent of correlation:
12 weather variables (r> 0.9); O3,PM10,NO2(r: 0.6-0.8); CO and SO2 (r< 0.5). These results
13 suggest that the contribution from the third component of error, as described in Zeger et al.
14 (2000), would vary among pollution and weather variables. Furthermore, the contribution from
15 the second component of error would also vary among pollutants; i.e., the ratio of ambient
16 exposure to ambient concentration, called the attenuation coefficient, is expected to be different
17 for each pollutant. Some of the ongoing studies are expected to shed some light on this issue.
18 However, more information is needed on attenuation coefficients for a variety of pollutants.
19 With regard to the PM exposure, longitudinal studies (Wallace, 2000; Mage et al., 1999),
20 show reasonably good correlation (r = 0.6 to 0.9) between ambient PM concentrations and
21 average population PM exposure, lending support for the use of ambient data as a surrogate for
22 personal exposure to ambient PM in time-series mortality or morbidity studies. Furthermore,
23 fine particles are expected to show even better site-to-site correlation than PM10. Wilson and
24 Suh (1997) examined site-to-site correlation of PM10, PM25, and PM10_25 in Philadelphia and
25 St. Louis, and found that site-to-site correlations were high (r ~ 0.9) for PM2 5 but low for PM10_2 5
26 (r ~ 0.4), indicating that fine particles have smaller errors in representing community-wide
27 exposures. This finding supports Lipfert and Wyzga's (1997) speculation that the stronger
28 mortality associations for fine particles than coarse particles found in the Schwartz et al. (1996a)
29 study may be due in part to larger measurement error for coarse particles.
30 However, as Lipfert and Wyzga (1997) suggested, the issue is not whether the fine particle
31 association with mortality is a "false positive", but rather, whether the weaker mortality
June 2003 8-252 DRAFT-DO NOT QUOTE OR CITE
-------
1 association with coarse particles is a "false negative." Carrothers and Evans (2000) also
2 investigated the joint effects of correlation and relative error, but they specifically addressed the
3 issue of fine (FP) versus coarse particle (CP) effect, by assuming three levels of relative toxicity
4 of fine versus coarse particles (PFP / pcp =1,3, and 10) and, then, evaluating the bias, (B =
5 |E[PP]/ E[pc,]} / (PF / pc}, as a function of FP-CP correlation and relative error associated with
6 FP and CP. Their results indicate: (1) if the FP and CP have the same toxicity, there is no bias
7 (i.e., B=l) as long as FP and CP are measured with equal precision, but, if, for example, FP is
8 measured more precisely than CP, then FP will appear to be more toxic than CP (i.e., B > 1);
9 (2) when FP is more toxic than CP (i.e., PFP/PCp = 3 and 10), however, the equal precision of FP
10 and CP results in downward bias of FP (B < 1), implying a relative overestimation of the less
11 toxic CP. That is, to achieve non-bias, FP must be measured more precisely than CP, even more
12 so as the correlation between FP and CP increases. They also applied this model to real data
13 from the Harvard Six Cities Study, in particular, the data from Boston and Knoxville.
14 Estimation of spatial variability for Boston was based on external data and a range of spatial
15 variability for Knoxville (since there was no spatial data available for this city). For Boston,
16 where the estimated FP-CP correlation was low (r = 0.28), estimated error was smaller for FP
17 than for CP (0.85 versus 0.65, as correlation between true versus error-added series), and the
18 observed FP to CP coefficient ratio was high (11), the calculated FP to CP coefficient ratio was
19 even larger (26)-thus providing evidence against the hypothesis that FP is absorbing some of the
20 coefficient of CP. For Knoxville, where FP-CP correlation was moderate (0.54), the error for FP
21 was smaller than for CP (0.9 versus 0.75), and the observed FP to CP coefficient ratio was 1.4,
22 the calculated true FP to CP coefficient ratio was smaller (0.9) than the observed value,
23 indicating that the coefficient was overestimated for the better-measured FP, while the
24 coefficient was underestimated for the worse-measured CP. Since the amount (and the
25 direction) of bias depended on several variables (i.e., correlation between FP and CP; the relative
26 error for FP and CP; and, the underlying true ratio of the FP toxicity to CP toxicity), the authors
27 concluded "...for instance, it is inadequate to state that differences in measurement error among
28 fine and coarse particles will lead to false negative findings for coarse particles".
29 Fung and Krewski (1999) conducted a simulation study of measurement error adjustment
30 methods for Poisson models, using scenarios similar to those used in the simulation studies that
31 investigated implication of joint effects of correlated covariates with measurement error. The
June 2003 8-253 DRAFT-DO NOT QUOTE OR CITE
-------
1 measurement error adjustment methods employed were the Regression Calibration (RCAL)
2 method (Carroll et al., 1995) and the Simulation Extrapolation (SIMEX) method (Cook and
3 Stefanski, 1994). Briefly, RCAL algorithm consists of: (1) estimation of the regression of X on
4 W (observed version of X, with error) and Z (covariate without error); (2) replacement of X by
5 its estimate from (1), and conducting the standard analysis (i.e., regression); and (3) adjustment
6 of the resulting standard error of coefficient to account for the calibration modeling. SIMEX
7 algorithm consists of: (1) addition of successively larger amount of error to the original data;
8 (2) obtaining naive regression coefficients for each of the error added data sets; and, (3) back
9 extrapolation of the obtained coefficients to the error-free case using a quadratic or other
10 function. Fung and Krewski examined the cases for: (1) px = 0.25; pz = 0.25; (2) px = 0.0;
11 pz = 0.25; (3) px = 0.25; pz = O.O., all with varying level of correlation (-0.8 to 0.8) with and
12 without classical additive error, and also considering Berkson type error. The behaviors of naive
13 estimates were essentially similar to other simulation studies. In most cases with the classical
14 error, RCAL performed better than SIMEX (which performed comparably when X-Z correlation
15 was small), recovering underlying coefficients. In the presence of Berkson type error, however,
16 even RCAL did not recover the underlying coefficients when X-Z correlation was large (> 0.5).
17 This is the first study to examine the performance of available error adjustment methods that can
18 be applied to time-series Poisson regression. The authors recommend RCAL over SIMEX.
19 Possible reasons why RCAL performed better than SIMEX in these scenarios were not
20 discussed, nor are they clear from the information given in the publication. There has not been a
21 study to apply these error adjustment methods in real time-series health effects studies. These
22 methodologies require either replicate measurements or some knowledge on the nature of error
23 (i.e., distributional properties, correlation, etc.). Since the information regarding the nature of
24 error is still being collected at this time, it may take some time before applications of these
25 methods become practical.
26 Another issue that measurement error may affect is the detection of threshold in time-series
27 studies. Lipfert and Wyzga (1996) suggested that measurement error may obscure the true shape
28 of the exposure-response curve, and that such error could make the exposure-response curve to
29 appear linear even when a threshold may exist. However, based on a simulation with realistic
30 range of exposure error (due to site-to-site correlation), Cakmak et al. (1999) illustrated that the
June 2003 8-254 DRAFT-DO NOT QUOTE OR CITE
-------
1 modern smoothing approach, LOESS, can adequately detect threshold levels (12.8 |ig/m3,
2 24.6 |ig/m3, and 34.4 |ig/m3) even with the presence of exposure error.
3 Other issues related to exposure error that have not been investigated include potential
4 differential error among subpopulations. If the exposure errors are different between susceptible
5 population groups (e.g., people with COPD) and the rest of the population, the estimation of bias
6 may need to take such differences into account. Also, the exposure errors may vary from season
7 to season, due to seasonal differences in the use of indoor emission sources and air exchange
8 rates due to air conditioning and heating. This may possibly explain reported season-specific
9 effects of PM and other pollutants. Such season-specific contributions of errors from indoor and
10 outdoor sources are also expected to be different from pollutant to pollutant.
11 In summary, the studies that examined joint effects of correlation and error suggest that
12 PM effects are likely underestimated, and that spurious PM effects (i.e., qualitative bias such as
13 change in the sign of coefficient) due to transferring of effects from other covariates require
14 extreme conditions and are, therefore, unlikely. Also, one simulation study suggests that, under
15 the likely range of error for PM, it is unlikely that a threshold is ignored by common smoothing
16 methods. More data are needed to examine the exposure errors for other pollutants, since their
17 relative error contributions will influence their relative significance in relative risk estimates.
18
19 8.4.8.2 Spatial Measurement Error Issues That May Affect the Interpretation of
20 Multi-Pollutant Models with Gaseous Co-Pollutants
21 The measurement error framework put forth in Dominici et al. (2000) and Zeger et al.
22 (2000) explicitly assumes that one of the error components has a Berkson error structure.
23 As summarized in (Zeger et al., 2000, p. 421): "This Berkson model is appropriate when z
24 represents a measurable factor [e.g., measured PM or another pollutant] that is shared by a group
25 of participants whose individual [true] exposures x might vary because of time-activity patterns.
26 For example, z might be the spatially averaged ambient level of a pollutant without major indoor
27 sources and x might be the personal exposures that, when averaged across people, match the
28 ambient level." This assumption is likely accurate for sulfates, less so for fine particles and for
29 PM10, and almost certainly incorrect for gases such as CO and NO2 that may vary substantially
30 on an intra-urban spatial scale with widely distributed local sources.
31 The usual characterization of longitudinal or temporal pollutant correlation may not
32 adequately characterize the spatial variation that is the more important aspect of association in
June 2003 8-255 DRAFT-DO NOT QUOTE OR CITE
-------
1 evaluating possible Berkson errors. Temporal correlation coefficients, even across large
2 distances (e.g., Ito et al., 2001) may be a consequence of large-scale weather patterns affecting
3 the concentrations of many pollutants. Local concentrations for some pollutants with strong
4 local sources and low regional dispersion (especially for CO and NO2, and PM10_2 5 to a lesser
5 extent) may have somewhat smaller temporal correlations and much greater relative spatial
6 variations than PM. Thus, individuals in a large metropolitan area may have roughly similar
7 levels of PM exposure x on any given day for which the ambient average PM concentration z is
8 an adequate surrogate, whatever their space-time activity patterns, residence, or non-residential
9 micro-environments, while the same individuals may be exposed to systematically higher or
10 lower concentrations of a co-pollutant than the spatial average of the co-pollutant. This violates
11 the basic assumption of the Berkson error model that within each stratum of the measured
12 (spatially averaged) level z, the average value of the true concentration x is equal to z, i.e.,
13
14 E{ x | z } = z, (8-6)
15
16 where E{.} is the average or expected value over the population.
17 There are empirical reasons to believe that if the strata are chosen to be locations within a
18 metropolitan area, some individuals far from local sources have consistently less exposure than
19 the average ambient concentration (denoted p) for co-pollutants with local sources such as CO
20 and NO2, and PM25, whose true exposure (denoted q) depends on the location of the person's
21 residence or other micro-environment where most exposure occurs. For this group,
22
23 E{q p}p. (8-8)
29
30 There is a substantial and growing body of evidence that adverse health effects are
31 associated with proximity to a major road or highway (Wjst et al., 1993; Monn et al., 2001;
June 2003 8-256 DRAFT-DO NOT QUOTE OR CITE
-------
1 Roemer and Van Wijnen, 2001). As shown below, there is good reason to believe that intra-city
2 variation (even in PM2 5) is substantial within some U.S. cities. If we assume for the sake of
3 argument that concentrations of PM10 or PM2 5 are relatively uniformly distributed, then
4 associations of adverse health effects with proximity to a source cannot be readily attributed to a
5 pollutant such as PM with a uniform spatial distribution. NO2 is a pollutant often used to
6 illustrate the spatial non-uniformity of the gaseous co-pollutants. Figure 8-23 from Monn et al.
7 (1997) compares the concentrations of NO2 and PM10 as a function of curbside distance in a
8 moderately busy urban street in Zurich. The PM10 levels decrease only slightly with increasing
9 distance, the decrease more likely being due to decreasing coarse particle than decreasing fine
10 particle concentrations. The NO2 concentrations show a much stronger seasonal dependence,
11 decreasing rapidly with increasing distance in the summer and showing little decrease with
12 distance in the winter. However, the belief that PM2 5 is spatially uniform should also not be
13 accepted uncritically, as recent analyses for 27 U.S. cities shown in Chapter 3 and Appendix 3 A
14 of this document demonstrate.
15 The 90th Percentile differences (P90) between a pair of sites may provide a useful guide to
16 the differences between monitor pairs (and by implication, personal exposure to fine particles)
17 that might be reasonably expected within a metropolitan area. Shown below in Table 8-38 are
18 the maximum, median, and minimum differences between monitor pairs, the monitor pairs at
19 which the largest 90th percentile difference occurs (by reference to tables in Appendix 3 A).
20 Based on these differences, Table 8-39 shows cities to be "relatively homogeneous" (with
21 P90 < 10 |ig/m3) and "relatively heterogeneous" (if P90 > 10 |ig/m3). The results in
22 Appendix 3 A and Table 8-38 show a variety of spatial patterns of association of PM2 5 within a
23 Metroplitan Statistical Area (MSA). There may be some discernable regional differences; but,
24 because many major population centers are not represented in Appendix 3A, further
25 investigation is likely warranted.
26 The results shown here provide clear evidence that fine particle concentrations may be less
27 homogenous in at least some MSAs than has been previously assumed. This provides support
28 for earlier studies using TSP and PM10 cited below. As noted in Chapter 3, these differences
29 may not be strictly related to the distance between monitors, especially where topography and
30 sources of primary PM play a role. In many eastern sites, however, particle distribution may be
31 more substantially governed by regional rather than by local sources.
June 2003 8-257 DRAFT-DO NOT QUOTE OR CITE
-------
Concentration of PM10 and NO2 vs. Distance
50
40
o
"ro
"c
o
O
30
20
NO, Summer
PM,n Summer
0 20 40 60 80
Distance from Road (m)
Figure 8-23. Concentration of PM10 and NO2 versus distance.
Source: Monn et al. (2000).
1 Several recent studies have examined the role of spatial siting of monitors on the
2 estimation of PM effects. Ito et al. (1995) examined the ability of single-site versus multi-site
3 averages to best estimate total mortality versus PM10 in Cook County (Chicago), IL and
4 Los Angeles County, CA. In order to have a sufficiently large sample size to detect effects, Ito
5 et al. used six PM10 sites in Cook County (Chicago), IL and four sites in Los Angeles County,
6 CA. A sinusoidal model was used to account for temporal components, although spline or
7 LOESS methods would now be used. Only one Cook County site had every-day PM samples,
8 and the others as well as the Los Angeles sites had a one-in-six-day sampling schedule. The
9 monitor sites were located in urban and suburban settings, according to the State's objectives.
10 Three of the Los Angeles sites were located in residential areas and one was located in an area
June 2003
8-258
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-38. MAXIMUM, MEAN, 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
City
Pittsburgh, PA
Los Angeles, CA
Seattle, WA
Riverside-San Bernardino, CA
Birmingham, AL
St. Louis, MO
Cleveland, OH
Detroit, MI
Atlanta, GA
Salt Lake City, UT
Gary, IN
Chicago, IL
San Diego, CA
Steubenville, OH
Washington, DC
Boise, ID
Philadelphia, PA
Kansas City, MO
Portland, OR
Grand Rapids, MI
Louisville, KY
Dallas, TX
Milwaukee, WI
Tampa, FL
Norfolk, VA
Columbia, SC
Baton Rouge, LA
N Sites
11
6
5
4 (w/o A) *
5
5
11
8
10
7
6 (w/o G) *
6
4
11
4
5
6
5 (w/o F)
4
7
6
4
4
4
7
8
4
5
3
3
Maximum (Pair)
21.0 (CJ)
18.2 (CF)
17.9 (AE)
8.5 (CE)
17.8 (BC)
15.2 (AE)
15.2 (AH)
14.3 (BG)
13.8 (DI)
13.2 (EG)
10.8 (CF)
11.4(CF)
11.3(BC)
11.3(EJ)
11.0 (CD)
10.0 (BE)
9.1 (DF)
7.7 (AE)
8.8 (BD)
7.5 (BC)
6.5 (CF)
6.5 (AB)
6.1 (BC)
6.0 (AC)
5.5 (EG)
5.0 (FH)
5.0 (BD)
5.0 (AC)
3.3 (AB)
2.9 (AC)
Mean
8.4
13.1
9.8
6.8
12.3
10.6
6.7
8.6
8.1
9.4
8.1
7.5
7.8
6.8
9.1
7.9
6.6
5.8
5.3
6.7
4.2
4.8
4.8
5.2
3.4
3.7
4.1
3.6
3.1
2.7
Minimum
4.2
6.2
3.6
3.6
6.6
6.7
2.8
3.3
5.0
5.3
5.3
4.4
4.2
3.5
6.3
6.2
3.5
3.5
3.8
3.3
1.9
4.1
3.1
3.8
1.9
2.8
3.1
2.6
2.8
2.5
* Without one site > 100 km from the others.
Source: Based on Chapter 3 and Appendix 3A analyses.
June 2003
8-259
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-39. SUMMARY OF WITHIN-CITY HETEROGENEITY BY REGION
Relative Heterogeneity Among Pairs of Monitors
Relatively
East
Atlanta, GA
Birmingham, AL
Chicago, IL
Cleveland, OH
Detroit, MI
Gary, IN
Pittsburgh, PA
St. Louis, MO
Steubenville, OH
Heterogenous
West
Los Angeles, CA
Riverside, CA
Salt Lake City, UT
San Diego, CA
Seattle, WA (with A)
Relatively Homogeneous
East West
Baton Rouge, LA Boise, ID
Columbia, SC Portland, OR
Dallas, TX
Grand Rapids, MI
Kansas City, KS-MO
Milwaukee, WI
Norfolk, VA
Louisville, KY
Philadelphia, PA
Tampa, FL
Washington, DC
Seattle, WA (w/o A)
1 zoned for commercial use. One of the Cook County sites was classified as residential, two as
2 commercial, and three as industrial. One of the Chicago sites was intended to monitor
3 population exposure, three to monitor maximum concentrations, and two to monitor both
4 maximum concentrations and personal exposure. There was considerable variation among the
5 distribution of PM10 in Cook County (Chicago), IL sites, and among Los Angeles County, CA
6 sites, especially at the upper end of the distribution. The sites were temporally correlated, 0.83
7 to 0.63 in Cook County, 0.9 to 0.7 in Los Angeles (except for one site pair), across distances of 4
8 to 26 miles. The Cook County mortality estimates were better estimated by some single-site
9 estimates (Site 2 with everyday data, N = 1251) than by an average using all available data with
10 missing values estimated from non-missing data (N = 1357). The every-six-day subsamples
11 from Site 1 (N = 281) and Site 2 (lag 0, N = 246) were better predictors, and from Site 4 (N =
12 243) and Site 6 (N = 292) about as good predictors of mortality as the corresponding every-six-
13 day averages (N = 351). In Los Angeles, only Site 4 (N = 349) was about as predictive as the
14 spatial averages (N = 405).
June 2003
8-260
DRAFT-DO NOT QUOTE OR CITE
-------
1 Lipfert et al. (2000a) examined the relationship between the area in which mortality
2 occurred among residents and the locations of monitoring sites or averages over monitoring sites
3 for several particle size components and particle metrics. The mortality data were located for
4 Philadelphia, PA, for three additional suburban Philadelphia counties, for Camden, NJ and other
5 New Jersey counties in the Philadelphia - Camden MSA. A single site was used for fine and
6 coarse particles from the Harvard School of Public Health monitors. Additional PA and NJ
7 thoracic particle data were available for 2 to 4 stations and results averaged for at least two
8 stations reporting data. The authors conclude that mortality in any part of the region may be
9 associated with air pollution concentrations or average concentrations in any other part of the
10 region, whether particles or gases. The authors suggest two interpretations: (a) the associations
11 of mortality with pollution were random (from carrying out multiple significance tests) and not
12 causal, or (b) both particles and gaseous pollutants have a broad regional distribution. The
13 authors note that interpretation (b) may lead to large uncertainties in identifying which pollutant
14 exposures for the population are primarily responsible for the observed effects. These data could
15 be studied further to evaluate smaller-scale spatial relationships among health effects and gases.
16 Lippmann et al. (2000) evaluated the effects of monitor siting choice using 14 TSP
17 monitoring stations in Detroit, MI, and nearby Windsor, ON, Canada. The stations operated
18 from 1981-1987 with almost complete data. When a standard log-linear link Poisson regression
19 model for mortality was fitted to TSP data for each of the 14 sites, the relative risk estimates
20 were similar for within-site increments of 5th to 95th percentiles, generally highest and positive at
21 lag day 1, but not statistically significant except for site "w" (site 12, south of the urban center of
22 Wayne County) and nearly significant at sites "f' (west of the city of Detroit), "g" (south of the
23 city) and "v" (suburban site in northwestern Wayne County, MI, generally "upwind" of the
24 urban center). However, as the authors note, all of the reported relative risks are for site-specific
25 increments, which vary by a factor of about 2.5 over the Wayne County - Windsor area. When
26 converted to a common increment of 100 |ig/m3 TSP, the largest excess risks are found when the
27 monitor used in the model is "f' (4.5%), "v" (4.2%), or "w" (3.8%), which also show the most
28 significant effects among the 14 monitors. As the authors note, ". . . the distributional
29 increments [used] to calculate relative risk tend to standardize the scale of relative risks. This
30 actually makes sense in that if there is a concentration gradient of TSP within a city, and if the
31 various TSP concentrations fluctuate together, then using a site with a low mean TSP for time-
June 2003 8-261 DRAFT-DO NOT QUOTE OR CITE
-------
1 series analysis would result in a larger coefficient. This result does warn against extrapolating
2 the effects from one city to an other using a raw regression coefficient [excess relative risk]"
3 Other recent studies also point out other aspects of intra-urban spatial variation in PM
4 concentrations. Kinney et al. (2000) note that, in a study of personal and ambient PM25 and
5 diesel exhaust particle (DEP) exposure in a dense urban area of New York City, PM2 5
6 concentrations showed only a moderate site-to-site variation (37 to 47 |ig/m3), probably due to
7 broader regional sources of PM2 5, whereas elemental carbon concentrations (EC) showed a four-
8 fold range of site-to-site variations, reflecting the greater local variation in EC from DEP.
9 Several PM health studies for Seattle (King County), WA (e.g., Levy et al., 2001a, for out-
10 of-hospital primary cardiac arrests) found few statistically significant relationships, attributed by
11 the authors in part to the fact that Seattle has topographically diverse terrain with local "hot
12 spots" of residential wood burning, especially in winter. Sheppard et al. (2001) explored reasons
13 for these findings, particularly focusing on adjustments for location by use of a "topographic
14 index" that includes "downstream" normal flow of wood smoke from higher elevations and
15 trapping of wood smoke in topographic bowls or basins even at higher elevations. They also
16 adjusted for weather using a "stagnation index" (the average number of hours per day with wind
17 speed less than the 25th percentile of wind speeds) and temperature, as well as interaction terms
18 for stagnation on hilltop sites and temperature at suburban wood-smoke-exposed valley sites.
19 The adjustments for exposure measurement error based on methods developed in Sheppard
20 and Damian (2000) and Sheppard et al. (2001) had little effect on effect size estimates for the
21 case-crossover study (Levy et al., 200la), but may be useful in other studies where localized
22 effects are believed to be important, particularly for the gaseous co-pollutants. Bateson and
23 Schwartz (2001) note that investigators should be careful when making assumptions about the
24 reference exposure distribution, in that the issue of comparability of the case and reference
25 groups is a general one for case-cross over analyses.
26 Daniels et al. (2001) evaluated relative sources of variability or heterogeneity in PM10
27 monitoring in Pittsburgh, PA in 1996. The area is data-rich, having 25 monitors in a -40 by
28 80 km rectangle. The authors found no isotropic spatial dependence after accounting for other
29 sources of variability, but an indication of heterogeneity in the variability of the small-scale
30 processes over time and space and heterogeneity in the mean values and covariate effects across
31 sites. Important covariates included temperature, precipitation, wind speed and direction. The
June 2003 8-262 DRAFT-DO NOT QUOTE OR CITE
-------
1 authors concluded that significant unmeasured processes might be in operation. These methods
2 should also be useful in evaluating spatial and temporal variations in gaseous co-pollutants,
3 where small-scale processes are important.
4
5 8.4.8.3 Measurement Error and the Assessment of Confounding by Co-Pollutants in
6 Multi-Pollutant Models.
7 The Zeger et al. (2000) discussion may be interpreted as addressing the extent to which the
8 apparent lack of a PM10_2 5 effect in models with both fine and coarse particles demonstrates a
9 "false negative" due to larger measurement error of coarse particle concentrations. However, a
10 more important question may involve the relative attenuation of estimated effects of PM2 5 and
11 gaseous co-pollutants, especially those such as CO that are known to be highly correlated with
12 PM2 5. Tables 1 and 2 in (Zeger et al., 2000) may be particularly relevant here. The evidence
13 discussed in this chapter supports the hypothesis that PM has adverse health effects, but leaves
14 open the question as to whether the co-pollutants have effects as well when their exposure is
15 measured much less accurately than that of the PM metric. If both the PM metric and the co-
16 pollutant have effects, Table 1 of Zeger et al. (2000) shows that the co-pollutant effect size
17 estimate may be greatly attenuated and the PM effect size estimate much less so, depending on
18 the magnitude of correlation between the true PM and gaseous pollutant exposures and the
19 correlation between their measurement errors. One would expect that PM2 5, CO, and NO2
20 would often have a high positive correlation and their "exposure measurement errors" would
21 also be positively correlated if PM and the gaseous pollutants were positively correlated due to
22 common activity patterns, weather, and source emissions. Thus, the line with corr(x1,x2) = 0.5,
23 var(6j) = 0.5, var(62) = 2, corr^, 62) = 0.7 seems appropriate. This implies that the estimated
24 effect of the more accurately measured pollutant is 64% of the true value, and that of the less
25 accurately measured pollutant is 14% of the true value. In view of the substantially greater
26 spatial heterogeneity of traffic-generated ambient pollutants such as CO and NO2, and the
27 relative (though not absolute) regional spatial uniformity of ambient PM2 5 in some cities, but not
28 in others, it is likely that effect size estimates in multi-pollutant models are attenuated downward
29 to a much greater extent for the gaseous co-pollutants than for the PM metric in some cities, but
30 not in others. This may explain part of the heterogeneity of findings for multi-pollutant models
31 in different cities. Low effect size estimates for the gaseous co-pollutants in a multi-pollutant
32 model should be interpreted cautiously. The representativeness of the monitoring sites for
June 2003 8-263 DRAFT-DO NOT QUOTE OR CITE
-------
1 population exposure of both the particle metrics and gaseous pollutants should be evaluated as
2 part of the interpretation of the analysis. Indices such as the maximum 90th percentile of the
3 absolute difference in concentrations between pairs of sites as well as the median
4 cross-correlation across sites may be useful for characterizing for spatially heterogeneity of
5 gaseous co-pollutants as well as for fine particles.
6
7 8.4.8.4 Air Pollution Exposure Proxies in Long-Term Mortality Studies
8 The AHSMOG Study of mortality (Abbey et al., 1999; McDonnell et al., 2000), the
9 Harvard 6-Cities Study of mortality (Dockery et al, 1993), the ACS Study (Pope et al., 1995),
10 and the VA/Washington Univ. Study (Lipfert et al., 2000b) together provided a major step
11 forward in the assessment of the long-term effects of air pollution. These cohort studies
12 responded to many of the major criticisms of the prior cross-sectional mortality studies, while
13 largely confirming the results of those prior studies. In particular, unlike the ecological cross-
14 sectional studies, these new cohort studies had individual-level information about the members
15 of the study cohort, allowing the analysis to more properly control for other major factors in
16 mortality, such as smoking and socio-economic factors.
17 While several of these studies made use of newly available fine particle (PM25) mass data
18 to derive useful estimates of health effects of PM25 well before it was routinely measured, these
19 studies utilized air pollution exposure information in a manner similar to past studies, i.e., the
20 studies used central site metropolitan area (MA) spatial and time averages of air pollution
21 exposures, rather than exposure information at the individual level. For this reason, the
22 AHSMOG, Harvard Six-Cities, ACS, and VA/Washington Univ. studies have been term
23 "semi-individual" cohort studies of air pollution.
24
25 The AHSMOG Study
26 Although this study covers a large number of years (1977-1992 in Abbey et al., 1999), it is
27 much more limited in the availability of actually-observed versus estimated particle metrics.
28 Prior to 1987, PM10 could only be estimated from TSP, not observed. Also, for more recent
29 years, McDonnell et al. (2000) used participants who lived near an airport, so that PM25, and
30 PM10_2 5 as the difference of PM10 and PM2 5, could be estimated from airport visibility data using
June 2003 8-264 DRAFT-DO NOT QUOTE OR CITE
-------
1 methods described earlier (Abbey et al., 1995b). All this adds potential measurement error to the
2 exposure estimates.
3
4 The Veterans' Administration/Washington University Study
5 The air pollution concentrations for participants' counties of residence at time of
6 enrollment were used in analyses, rather than concentrations at the 32 VA hospitals in the final
7 study. County-wide pollution variables for five particle metrics and three gaseous pollutants
8 were used in the study, although TSP was most often the particle metric observed for the earlier
9 years of the study (before 1975 up to!988), which are important in assessing pollution effects for
10 many years of exposure. However, IPMN data for fine particles and sulfates were available for
11 ca. 1979-1983, as in the ACS study. Effects on average mortality for the intervals 1976-1981,
12 1982-1988, and 1989-1996 were related to multi-year particle exposures for four long intervals:
13 < 1975, 1975-1981, 1982-1988, and 1989-1996. TSP was used in the first three exposure
14 intervals; PM10 in the most recent. This study examined "concurrent" exposures (same interval
15 as average mortality), "causal" prior exposures (exposure interval precedes mortality interval),
16 and "non-causal" PM versus mortality associations. The mortality associations were also
17 examined for PM25, PM15, and PM15.25 for 1979-1981 and 1982-1984. This study uses
18 essentially the same air pollution data as the ACS study, which should be adequate for
19 characterizing fixed-site air pollution concentrations in the place of residence at the time of
20 enrollment. However, if any participants moved away from the county where air pollution is
21 measured, but were retained in the study because they continued in follow-ups at the same clinic,
22 then use of initial residence location may not be an adequate proxy for actual exposure after
23 initial enrollment.
24
25 Harvard Six-Cities Air Pollution Exposure Data
26 In the Harvard Six Cities Study, ambient concentrations of fine particles (PM25), total
27 suspended particles (TSP), sulfur dioxide (SO2), ozone (O3), nitrogen dioxide (NO2), and sulfate
28 (SO4=) were measured at a centrally located air monitoring station within each of six
29 communities. Long-term mean concentrations for each pollutant were calculated for periods that
30 were consistent among the six cities, but not across pollutants. The original epidemiologic
31 analysis characterized ambient air quality as long-term mean concentrations of total particles
June 2003 8-265 DRAFT-DO NOT QUOTE OR CITE
-------
1 (TSP) (1977-1985), inhalable and fine particles (1979-1985), sulfate particles (1979-1984),
2 aerosol acidity (H+ ) (1985-1988), sulfur dioxide (1977-1985), nitrogen dioxide (1977-1985),
3 and ozone (1977-1985), as follows:
4
5 Particles: Mean PM concentrations were reported for four classifications of particles in each of
6 the six cities: TSP (particles with aerodynamic diameters up to 50 jim), inhalable particles, fine
7 particles, and sulfate particles. Values of mass for TSP and sulfate particles were determined
8 from 24-h high-volume samplers. Inhalable particle mass was calculated from coarse and fine
9 particle mass, which had been determined from 24-h sample pairs collected by dichotomous
10 samplers. In these, the fine particle channel collected particles smaller than about 2.5 jam and
11 the measurement was recorded directly as fine particle (FP) mass. The coarse particle channel
12 collected particles 2.5 jam to 10 or 15 jim in aerodynamic diameter (the upper bound
13 measurement depended on the inlet size used at the time).
14
15 Acidity: Aerosol acidity (FT) was measured for about one year in each city. However,
16 measurements were conducted in only two cities at a time. Thus, it was not possible to compare
17 acidity for a common time period. Furthermore, the acidity data were not linked with particle
18 data in the same city. Thus, intercity and inter-pollutant comparisons of FT in this study were
19 confounded by inter-annual variability.
20
21 Gases: The gases (SO2, NO2, and O3) were measured (in parts per billion) hourly by
22 conventional continuous monitors.
23
24 ACS Study Air Pollution Exposure Data
25 In the ACS Study (Pope et al., 1995), two measures of particulate air pollution, fine
26 particles, and sulfate, but no gaseous pollutants were considered. The mean concentration of
27 sulfate air pollution by metropolitan area (MA) during 1980 was estimated using data from the
28 EPA Aerometric Information Retrieval System (AIRS) database. These means were calculated
29 as the averages of annual arithmetic mean 24-h sulfate values for all monitoring sites in the 151
30 MA's considered. The median concentration of fine particles between 1979 and 1983 was
31 estimated from the EPA's dichotomous sampler network. These estimates of fine particle levels
June 2003 8-266 DRAFT-DO NOT QUOTE OR CITE
-------
1 had been used previously in a population-based cross-sectional mortality study of 50 MA's.
2 Gaseous co-pollutants were not considered in Pope et al's original ACS analysis.
3
4 Six-City Study and ACS Exposure Data Strengths and Weaknesses
5 In each of these studies, there was a single mean pollution concentration assigned for each
6 city for each pollutant for the entire follow-up period considered. Concentrations were not
7 broken into each year or sub-groups of years (e.g., 5 year averages), largely because data were
8 not available in this form. This may represent a potential weakness, as a single number could
9 not accurately account for the different exposures in different years of follow-up. It is possible,
10 however, that the simultaneous or immediately preceding years alone might not as well represent
11 the effects of long-term pollution exposure.
12 The ACS analysis also uses metropolitan area (MA) pollutant concentrations for air
13 pollution exposure estimates, rather than individual level measurements. Thus, spatial
14 variability in air pollution levels and potential effects of different housing infiltration rates were
15 not addressed as potential factors in exposure variability. However, individual exposure data
16 would be economically impractical for such large cohorts, and the use of more localized
17 measurements (e.g., by county) might well lead to more error, due to day-to day mobility
18 between counties by individuals (e.g., to work and back) and changes of specific residence
19 within an MA over time. Thus, the MA average may actually be the best metric that can be
20 developed in the absence of individual level exposure data.
21 Another notable weakness of the original ACS Study was that only two PM air pollution
22 metrics were considered. Thus, this study did not consider the potentially confounding
23 influences of gaseous air pollutants or other particle indicators.
24 These two studies' analyses assign the subjects' residence MA on the basis of where they
25 were enrolled, which can lead to exposure errors if the subjects moved to another MA during the
26 follow-up period. However, a recent reanalysis of the Six Cities Study cohort (Krewski et al.,
27 2000) indicates that mobility in these older populations is limited, with only 18.5% leaving the
28 original city of enrollment over subsequent decades.
29
30
June 2003 8-267 DRAFT-DO NOT QUOTE OR CITE
-------
1 The HEI Reanalysis of the ACS Study
2 The HEI Reanalysis of these two cohort studies (Krewski et al, 2000) confirmed the
3 databases used in these two studies, but also developed new exposure data for the ACS Study
4 cohort. In particular, data for the gaseous pollutants (for the year 1980) were added to the
5 analysis. Table 8-38 displays summary data for the most recent data available for the analysis of
6 the ACS cohort (Pope et al., 2002). The variables noted with the data source "HEI" were added
7 to the analysis during the HEI reanalysis. These HEI results largely confirmed the original ACS
8 analysis results for PM, but also indicated that SO2 was also correlated with U.S. mortality.
9
10 The 16-Year Follow-Up of the ACS Cohort
11 Table 8-40 also includes summaries of the pollutant data developed to provide exposure
12 estimates for the latest 16-year follow-up analysis of the ACS cohort (Pope et al, 2002). These
13 new data are similarly city-wide averages of all monitoring stations in the MA's considered, but
14 for the entire period of follow-up (1982-1998), when possible. In addition, this new analysis has
15 incorporated the new PM25 air monitoring data collected routinely from 1999 onward. As a
16 result, this new analysis has increased the analysis power both by extending the length of follow-
17 up, and by adding significant new multiple and multi-year air pollution exposure data to the
18 analysis.
19
20 8.4.9 Implications of Airborne Particle Mortality Effects
21 The public health burden of mortality associated with exposure to ambient PM depends not
22 only on the increased risk of death, but also on the amount of life shortening that is attributable
23 to those deaths. The 1996 PM AQCD concluded that confident quantitive determination of years
24 of life lost to ambient PM exposure was not yet possible and life shortening may range from
25 days to years (U.S. Environmental Protection Agency, 1996a). Now, some newly available
26 analyses provide further interesting insights with regard to potential life-shortening associated
27 with ambient PM exposures.
28
29 8.4.9.1 Short-Term Exposure and Mortality Displacement
30 A few studies have investigated the question of "harvesting," a phenomenon in which a
31 deficit in mortality occurs following days with (pollution-caused) elevated mortality, due to
June 2003 8-268 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-40. 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)
PM2 5 (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
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/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 depletion of the susceptible population pool. This issue is very important in interpreting the
2 public health implication of the reported short-term PM mortality effects. The 1996 PM AQCD
3 discussed suggestive evidence observed by Spix et al. (1993) during a period when air pollution
4 levels were relatively high. Recent studies, however, generally used data from areas with lower,
5 non-episodic pollution levels.
June 2003
8-269
DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz (2000c; reanalysis 2003) separated time-series air pollution, weather, and
2 mortality data from Boston, MA, into three components: (1) seasonal and longer fluctuations;
3 (2) "intermediate" fluctuations; (3) "short-term" fluctuations. By varying the cut-off between
4 the intermediate and short term, evidence of harvesting was sought. The idea is, for example, if
5 the extent of harvesting were a matter of a few days, associations between weekly average values
6 of mortality and air pollution (controlling for seasonal cycles) would not be seen. Schwartz's
7 reanalysis using natural splines reported reductions in COPD mortality PM2 5 risk estimates for
8 longer time scale, suggesting that most of the COPD mortality was only displaced by a few
9 weeks. However, for pneumonia, ischemic heart disease, and all cause mortality, the effect size
10 increased, as longer time scales were included. For example, the percent increase in non-
11 accidental deaths associated with a 25 |ig/m3 increase in PM2 5 increased from 5.8% (95% CI:
12 4.5, 7.3) for thel5-day window to 9.7% (95% CI: 8.2, 11.2) for the 60-day window. Note,
13 however, that the 60-day time scale window is in the range of influenza epidemics. Some
14 caution is therefore needed in interpreting risk estimates in this range.
15 Zanobetti et al. (2000b) used what they termed "generalized additive distributed lag
16 models" (penalized splines using algorithm that did not require back-fitting were used for all the
17 smoothing terms) to help quantify mortality displacement in Milan, Italy, 1980-1989. Non-
18 accidental total deaths were regressed on smooth functions of TSP distributed over the same day
19 and the previous 45 days using penalized splines for the smooth terms and seasonal cycles,
20 temperature, humidity, day-of-week, holidays, and influenza epidemics. The mortality
21 displacement was modeled as the initial positive increase, negative rebound (due to depletion),
22 followed by another positive coefficients period, and the sum of the three phases were
23 considered as the total cumulative effect. TSP was positively associated with mortality up to
24 13 days, followed by nearly zero coefficients between 14 and 20 days, and then followed by
25 smaller but positive coefficients up to the 45 th day (maximum examined). The sum of these
26 coefficients was over three times larger than that for the single-day estimate.
27 Zanobetti et al. (2001; reanalysis by Zanobetti and Schwartz, 2003) also applied the same
28 concept described above (up to 41 lag days) to 10 cities from APHEA2 to estimate distributed
29 lag PM10 mortality risks. They applied the covariate adjustment in a GAM model used in
30 APHEA2 (Katsouyanni et al., 2001); and in reanalysis (Zanobetti and Schwartz, 2003), they also
31 used penalized splines in addition to the GAM model with stringent convergence criteria. The
June 2003 8-270 DRAFT-DO NOT QUOTE OR CITE
-------
1 resulting city specific coefficients were pooled in the second-stage model taking into account
2 heterogeneity across cities. The estimated shape of the distributed lag pooled across 10 cities
3 showed a similar pattern to that from Milan data described above, with the second "hump" of
4 smaller but positive coefficients between approximately 20 to 35 days. The results indicated
5 that, compared to PM10 risk estimates obtained for the average of lag 0 and 1 days, the
6 distributed lag estimates up to 40 days were about twice larger in both GAM and penalized
7 splines models. For example, the combined distributed lag estimates for the 10 cities using
8 penalized splines was 5.6% (95% CI: 1.5, 9.8), as compared to 2.9% (95% CI: 1.4, 4.4).
9 It should be noted, however, that the results for individual cities varied. For example, the
10 estimates for average of lag 0 and 1 days and the distributed lag model were comparable in Tel
11 Aviv, whereas it was nearly seven times bigger for distributed lag model in Lodz. Thus, while
12 these results do support the lack of mortality displacement up to 40-45 day period, the pattern of
13 lagged associations may vary from city to city.
14 Smith et al. (1999), as part of their analysis of PM10-mortality association in Birmingham,
15 AL and Cook County, IL, also examined the existence of mortality displacement. Their model
16 attempted to estimate the size of the frail population and the number of migrants into the frail
17 population. PM10 was modeled to affect both the entry into the frail population and death. The
18 latent variable structure was fitted through Bayesian techniques using Monte Carlo sampling.
19 The resulting posterior mean for the frail population in Chicago was 765 (posterior s.d. = 189).
20 The mean numbers of days lost as a result of 10 |ig/m3 increase in PM10 was estimated to be
21 0.079 day (posterior s.d. = 0.032). These results indicate that the frail population is small and
22 therefore has short lifetime (less than 10 days) in that state. Consequently, the impact of PM
23 (life shortening) had to be small. These results are not consistent with those suggested by
24 Zanobetti or Schwartz studies described above.
25 Murray and Nelson (2000) used Kalman filtering to estimate hazard function of TSP in a
26 state space model in the Philadelphia mortality data during 1973-1990. The model framework,
27 which assumes harvesting effect, allows estimation of at-risk population and the effect of
28 changes in air quality on the life expectancy of the at-risk population. The model was first
29 verified by simulation. Combinations of TSP, linear temperature, squared temperature, and
30 interaction of TSP and temperature were considered in six models. The size of at-risk (or frail)
31 population estimated was about 500 people, with its life expectancy between 11.8 to 14.3 days,
June 2003 8-271 DRAFT-DO NOT QUOTE OR CITE
-------
1 suggesting that the hazard causing agent making the difference of 2.5 days in the at-risk
2 population. These results are, taking into account the difference in population size between
3 Philadelphia and Cook County, comparable with those obtained by Smith et al. described above.
4 In both cases, the size of the frail population is small with short lifetime such that life-shortening
5 by PM or any external stress for the frail population could not be long (more than a few days).
6 These results are, again, in contrast to the results from the Zanobetti or Schwartz studies above
7 or a frequency domain approach described below.
8 Zeger et al. (1999) first illustrated, through simulation, the implication of harvesting for
9 PM regression coefficients (i.e., mortality relative risk) as observed in frequency domain. Three
10 levels of harvesting (3 days, 30 days, and 300 days) were simulated. As expected, the shorter the
11 harvesting, the larger the PM coefficient in the higher frequency range. However, in the analysis
12 (and reanalysis by Dominici et al., 2003) of real data from Philadelphia, regression coefficients
13 increased toward the lower frequency range, suggesting that the extent of harvesting, if it exists,
14 is not in the short-term range. Zeger suggested that "harvesting-resistant" regression coefficients
15 could be obtained by excluding coefficients in the very high frequency range (to eliminate short-
16 term harvesting) and in the very low frequency range (to eliminate seasonal confounding). Since
17 the observed frequency domain coefficients in the very high frequency range were smaller than
18 those in the mid frequency range, eliminating the "short-term harvesting" effects would only
19 increase the average of those coefficients in the rest of the frequency range.
20 Frequency domain analyses are rarely performed in air pollution health effects studies,
21 except perhaps the spectral analysis (variance decomposition by frequency) to identify seasonal
22 cycles. Examinations of the correlation by frequency {coherence) and the regression coefficients
23 by frequency (gain) may be useful in evaluating the potentially frequency-dependent
24 relationships among multiple time series. A few past examples in air pollution health effects
25 studies include: (1) Shumway etal.'s (1983) analysis of London mortality analysis, in which
26 they observed that significant coherence occurred beyond two week periodicity (they interpreted
27 this as "pollution has to persist to affect mortality"); (2) Shumway et al.'s (1988) analysis of Los
28 Angeles mortality data, in which they also found larger coherence in the lower frequency; (3)
29 Ito's (1990) analysis of London mortality data in which he observed relatively constant gain
30 (regression coefficient) for pollutants across the frequency range, except the annual cycle. These
June 2003 8-272 DRAFT-DO NOT QUOTE OR CITE
-------
1 results also suggest that associations and effect size, at least, are not concentrated in the very
2 high frequency range.
3 Schwartz (2000c), Zanobetti et al. (2000b), Zanobetti et al., (2001; reanalysis by Zanobetti
4 and Schwartz, 2003) and Zeger et al.'s analysis (1999; reanalysis by Dominici et al., 2003) all
5 suggest that the extent of harvesting, if any, is not a matter of only a few days. Other past
6 studies that used frequency domain analyses are also at least qualitatively in agreement with the
7 evidence against the short-term only harvesting. Since long wave cycles (> 6 months) need to be
8 controlled in time-series analyses to avoid seasonal confounding, the extent of harvesting beyond
9 6 months periodicity is not possible in time-series study design. Also, influenza epidemics can
10 possibly confound the PM-mortality associations in the 1 to 3 month periodicity ranges.
11 Therefore, interpreting PM risk estimates in these "intermediate" time scale also requires
12 cautions. In contrast to Zanobetti, Schwartz and Zeger et al. studies, Smith et al. and Murray and
13 Nelson studies suggest that the frail population is very small and its lifetime short, such that PM
14 or any external stress cannot have more than a few days of life-shortening impacts. This may be
15 an inherent limitation of the model itself. Thus, there appears to be consistency in results within
16 the similar models but not across different types of models. Clearly, more research is needed in
17 this area both in terms of development of conceptual framework that can be tested with real data,
18 and applications of these models to more data sets. However, at least in the models that extend
19 the common time-series modeling, there appears to be no strong evidence to suggest that PM is
20 shortening life by a few days.
21
22 8.4.9.2 Life-Shortening Estimates Based on Semi-Individual Cohort Study Results
23 Brunekreef (1997) reviewed the available evidence of the mortality effects of long-term
24 exposure to PM air pollution and, using life table methods, derived an estimate of the reduction
25 in life expectancy implied by those effect estimates. Based on the results of Pope et al. (1995)
26 and Dockery et al. (1993), a relative risk of 1.1 per 10 |ig/m3 exposure over 15 years was
27 assumed for the effect of PM air pollution on men 25-75 years of age. A 1992 life table for men
28 in the Netherlands was developed for 10 successive five-year categories that make up the
29 25-75 year old age range. Life expectancy of a 25 year old was then calculated for this base case
30 and compared with the calculated life expectancy for the PM-exposed case, in which the death
31 rates were increased in each age group by a factor of 1.1. A difference of 1.11 years was found
June 2003 8-273 DRAFT-DO NOT QUOTE OR CITE
-------
1 between the "exposed" and "clean air" cohorts' overall life expectancy at age 25. Looked at
2 another way, this implies that the expectation of the lifespan for persons who actually died from
3 air pollution was reduced by more than 10 years, because they represent a small percentage of
4 the entire cohort population. A similar calculation by the authors for the 1969-71 life table for
5 U.S. white males yielded an even larger reduction of 1.31 years for the entire population's life
6 expectancy at age 25. Thus, these calculations imply that relatively small differences in long-
7 term exposure to ambient PM can substantially affects on life expectancy.
8
9 8.4.9.3 Potential Effects of Infant Mortality on Life-Shortening Estimates
10 Deaths among children can logically have the greatest influence on a population's overall
11 life expectancy, but the Brunekreef (1997) life table calculations did not consider any possible
12 long-term air pollution exposure effects on the population aged < 25 years. As discussed above,
13 some of the older cross-sectional studies and the more recent studies by Bobak and Leon (1992),
14 Woodruff et al. (1997), Bobak and Leon (1999), and Loomis et al. (1999) suggest that infants
15 may be among the sub-populations notably affected by long-term PM exposure. Thus, although
16 it is difficult to quantify, any premature mortality that does occur among children due to long-
17 term PM exposure (as suggested by these new studies) would significantly increase the overall
18 population life shortening over and above that estimated by Brunekreef (1997) for long-term PM
19 exposure of adults aged 25 years and older.
20
21
22 8.5 SUMMARY OF KEY FINDINGS AND CONCLUSIONS DERIVED
23 FROM PARTICULATE MATTER EPIDEMIOLOGY STUDIES
24 The most important types of additions to the database beyond that assessed in the 1996 PM
25 AQCD, as evaluated above in this chapter, are:
26 (1) New multi-city studies on a variety of endpoints which provide more precise estimates of
the average PM effect sizes than most smaller-scale individual city studies;
27 (2) More studies of various health endpoints using ambient PM10 and/or closely related mass
concentration indices (e.g., PM13 and PM7), which substantially lessen the need to rely on
non-gravimetric indices (e.g., BS or CoH);
June 2003 8-274 DRAFT-DO NOT QUOTE OR CITE
-------
1 (3) New studies evaluating relationships of a variety of health endpoints to the ambient PM
coarse fraction (PM10_2 5), the ambient fine-particle fraction (PM2 5), and even ambient
ultrafine particles measures (PMai and smaller), using direct mass measurements and/or
estimated from site-specific calibrations;
2 (4) A few new studies in which the relationship of some health endpoints to ambient particle
number concentrations were evaluated;
3 (5) Many new studies which evaluated the sensitivity of estimated PM effects to the
inclusion of gaseous co-pollutants in the model;
4 (6) Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures
including PM components, based on empirical combinations (e.g., factor analysis or
source profiles;
5 (7) Numerous new studies of cardiovascular endpoints, with particular emphasis on
assessment of cardiovascular risk factors as well as symptoms;
6 (8) Additional new studies on asthma and other respiratory conditions potentially
exacerbated by PM exposure;
7 (9) New analyses of lung cancer associations with long-term exposures to ambient PM;
8 (10) New studies of infants and children as a potentially susceptible population.
9 It is not possible to assign any absolute measure of certainty to conclusions based on the
10 findings of the epidemiology studies discussed in this chapter. However, these observational
11 study findings would be further enhanced by supportive findings of causal studies from other
12 scientific disciplines (dosimetry, toxicology, etc.), in which other factors could be eliminated or
13 controlled, as discussed in Chapters 6 and 7. The epidemiology studies discussed in this chapter
14 demonstrate biologically-plausible responses in humans exposed at ambient concentrations. The
15 most salient conclusions derived from the PM epidemiology studies include:
16 (1) A large and reasonably convincing body of epidemiology evidence confirms earlier
associations between short- and long-term ambient PM10 exposures (inferred from
stationary air monitor measures) and mortality/morbidity effects and suggest that PM10
(or one or more PM10 components) is a probable contributing cause of adverse human
health effects.
June 2003 8-275 DRAFT-DO NOT QUOTE OR CITE
-------
(2) There appears to be some spatial heterogeneity in city-specific excess risk estimates for
the relationships between short-term ambient PM10 concentrations and acute health
effects. The reasons for such variation in effects estimates are not well understood at
this time, but do not negate ambient PM's likely causative contribution to observed PM-
mortality and/or morbidity associations in many locations. Possible factors contributing
to the apparent heterogeneity include geographic differences in air pollution mixtures,
composition of PM components, and personal and sociodemographic factors affecting
PM exposure (such as use of air conditioners, education, and so on).
(3) A growing body of epidemiology studies confirm associations between short- and long-
term ambient PM2 5 exposures (inferred from stationary air monitor measures) and
adverse health effects and suggest that PM2 5 (or one or more PM2 5 components) is a
probable contributing cause of observed PM-associated health effects. Some new
epidemiology findings also suggest that health effects are associated with mass or
number concentrations of ultrafine (nuclei-mode) particles, but not necessarily more so
than for other ambient fine PM components.
(4) A smaller body of evidence appears to support an association between short-term
ambient thoracic coarse fraction (PM10_2 5) exposures (inferred from stationary air
monitor measures) and short-term health effects in epidemiology studies. This suggests
that PM10_2 5, or some constituent component(s) of PM10_25, may be a contributory cause
of adverse health effects in some locations. Reasons for differences among findings on
coarse-particle health effects reported for different cities are still poorly understood, but
several of the locations where significant PM10_25 effects have been observed (e.g.,
Phoenix, Mexico City, Santiago) tend to be in drier climates and may have
contributions to observed effects due to higher levels of organic particles from biogenic
processes (endotoxins, molds, etc.) during warm months. Other studies suggest that
particles of crustal origin are generally unlikely to exert notable health effects under
most ambient exposure conditions, (however, see Item 14, below). Also, in some
western U.S. cities where PM10_25 is a large part of PM10, the relationship between
June 2003 8-276 DRAFT-DO NOT QUOTE OR CITE
-------
hospital admissions and PM10 may be an indicator of response to coarse thoracic
particles from wood burning.
(5) Long-term PM exposure durations, on the order of months to years, as well as on the
order of a few days, are statistically associated with serious human health effects
(indexed by mortality, hospital admissions/medical visits, etc.). More chronic PM
exposures, on the order of years or decades, appear to be associated with life shortening
well beyond that accounted for by the simple accumulation of the more acute effects of
short-term PM exposures (on the order of a few days). Some uncertainties remain
regarding the magnitude of and mechanisms underlying chronic health effects of long-
term PM exposures and the relationship between chronic exposure and acute responses
to short-term exposure.
(6) Recent investigations of the public health implications of such chronic PM exposure-
mortality effect estimates were also reviewed. Life table calculations by Brunekreef
(1997) found that relatively small differences in long-term exposure to airborne PM of
ambient origin can have substantial effects on life expectancy. For example, a
calculation for the 1969-71 life table for U.S. white males indicated that a chronic
exposure increase of 10 |ig/m3 PM was associated with a reduction of 1.31 years for the
entire population's life expectancy at age 25. Also, new evidence of associations of PM
exposure with infant mortality (Bobak and Leon, 1992, 1999; Woodruff et al., 1997;
Loomis et al., 1999) and/or intrauterine growth retardation (Dejmek et al., 1999) and
consequent increase risk for many serious health conditions associated with low birth
weight, if further substantiated, would imply that life shortening in the entire population
from long-term PM exposure could well be significantly larger than that estimated by
Brunekreef (1997).
(7) Considerable coherence exists among effect size estimates for ambient PM health
effects. For example, results derived from several multi-city studies, based on pooled
analyses of data combined across multiple cities (thought to yield the most precise
estimates of mean effect size), show the percent excess total (non-accidental) deaths
estimated per 50 |ig/m3 increase in 24-h PM10 to be: 1.4% in the 90 largest U.S. cities
with the estimate for the Northeast being the largest (approximately twice the
nationwide estimate); 3.4% in 10 large U.S. cities; 3.6% in the 8 largest Canadian cities;
June 2003 8-277 DRAFT-DO NOT QUOTE OR CITE
-------
and 3.0% in western European cities (using PM10 = TSP*0.55). These combined
estimates are consistent with the range of PM10 estimates previously reported in the
1996 PM AQCD. These and excess risk estimates from many other individual-city
studies, generally falling in the range of ca. 1.5 to 8.0% per 50 |ig/m3 24-h PM10
increment, also comport well with numerous new studies confirming increased cause-
specific cardiovascular- and respiratory-related mortality. They are also coherent with
larger effect sizes reported for cardiovascular and respiratory hospital admissions and
visits, as would be expected for these morbidity endpoints versus those for PM10-related
mortality.
(8) Several independent panel studies (but not all) that evaluated temporal associations
between PM exposures and measures of heart beat rhythm in elderly subjects provide
generally consistent indications of decreased heart rate variability (HRV) being
associated with ambient PM exposure (decreased HRV being an indicator of increased
risk for serious cardiovascular outcomes, e.g., heart attacks). Other studies point
toward changes in blood characteristics (e.g., C-reactive protein levels) related to
increased risk of ischemic heart disease also being associated with ambient PM
exposures. However, these heart rhythm and blood characteristics findings should
currently be viewed as providing only limited or preliminary support for PM-related
cardiovascular effects.
(9) Notable new evidence now exists which substantiates positive associations between
ambient PM concentrations and increased respiratory-related hospital admissions,
emergency department, and other medical visits, particularly in relation to PM10 levels.
Of much interest are new findings tending to implicate not only fine particle
components but also coarse thoracic (e.g., PM10_2 5) particles as likely contributing to
exacerbation of asthma conditions. Also of much interest are emerging new findings
indicative of likely increased occurrence of chronic bronchitis in association with
(especially chronic) PM exposure. Also of particular interest are reanalyses or
extensions of earlier prospective cohort studies of long-term ambient PM exposure
effects which demonstrate substantial evidence for association of increased lung cancer
risk with such PM exposures, especially exposure to fine PM or its subcomponents.
June 2003 8-278 DRAFT-DO NOT QUOTE OR CITE
-------
(10) One major methodological issue affecting epidemiology studies of both short-term and
long-term PM exposure effects is that ambient PM of varying size ranges is typically
found in association with other air pollutants, including gaseous criteria pollutants (e.g,
O3, NO2, SO2, CO), air toxics, and/or bioaerosols. Available statistical methods for
assessing potential confounding arising from these associations may not yet be fully
adequate. The inclusion of multiple pollutants often produces statistically unstable
estimates. Omission of other pollutants may incorrectly attribute their independent
effects to PM. Second-stage regression methods may have certain pitfalls that have not
yet been fully evaluated. Much progress in sorting out relative contributions of ambient
PM components versus other co-pollutants is nevertheless being made and, overall,
tends to substantiate that observed PM effects are at least partly due to ambient PM
acting alone or in the presence of other covarying gaseous pollutants. However, the
statistical association of health effects with PM acting alone or with other pollutants
should not be taken as an indicator of a lack of effect of the other pollutants. Indeed,
the effects of the other pollutants may at times be greater or less than the effects
attributed to PM and may vary from place to place or from time to time.
(11) It is possible that differences in observed health effects will be found to depend on site-
specific differences in chemical and physical composition characteristics of ambient
particles and on factors affecting exposure (such as air conditioning) as well as on
differences in PM mass concentration. For example, the Utah Valley study (Dockery
et al., 1999; Pope et al., 1991, 1999b) showed that PM10 particles, known to be richer in
metals during exposure periods while the steel mill was operating, were more highly
associated with adverse health effects than was PM10 during the PM exposure reduction
while the steel mill was closed. In contrast, PM10 or PM2 5 was relatively higher in
crustal particles during windblown dust episodes in Spokane and in three central Utah
sites than at other times, but was not associated with higher total mortality. These
differences require more research that may become more feasible as the PM2 5 sampling
network produces air quality data related to speciated samples.
(12) The above reasons suggest it is inadvisable to pool epidemiology studies at different
locations, different time periods, with different population sub-groups, or different
health endpoints, without assessing potential causes and the consequences of these
June 2003 8-279 DRAFT-DO NOT QUOTE OR CITE
-------
differences. Published multi-city analyses using common data bases, measurement
devices, analytical strategies, and extensive independent external review, as carried out
in the APHEA and NMMAPS studies are likely to be useful. Pooled analyses of more
diverse collections of independent studies of different cities, using varying
methodology and/or data quality or representativeness, are likely less credible and
should not, in general, be used without careful assessment of their underlying scientific
comparability.
(13) It may be possible that different PM size components or particles with different
composition or sources produce effects by different mechanisms manifested at different
lags, or that different preexisting conditions may lead to different delays between
exposure and effect. Thus, although maximum effect sizes for PM effects have often
been reported for 0-1 day lags, evidence is also beginning to suggest that more
consideration should be given to lags of several days. Also, if it is considered that all
health effects occurring at different lag days are all real effects, so that the risks for each
lag day should be additive, then higher overall risks may exist that are higher than
implied by maximum estimates for any particular single or two-day lags. In that case,
multi-day averages or distributed lag models should be used.
(14) Certain classes of ambient particles may be distinctly less toxic than others and may not
exert human health effects at typical ambient exposure concentrations or only under
special circumstances. Coarse thoracic particles of crustal origin, for example, may be
relatively non-toxic under most circumstances compared to those of combustion origin
such as wood burning. However, crustal particles may be sufficiently toxic to cause
human health effects under some conditions; resuspended crustal particles, for example,
may carry toxic trace elements and other components from previously deposited fine
PM, e.g., metals from smelters (Phoenix) or steel mills (Steubenville, Utah Valley),
PAH's from automobile exhaust, or pesticides from administration to agricultural lands.
Likewise, fine particles from different sources have different effect sizes. More
research is needed to identify conditions under which one or another class of particles
may cause little or no adverse health effects, as well as conditions under which particles
may cause notable effects.
June 2003 8-280 DRAFT-DO NOT QUOTE OR CITE
-------
(15) Certain epidemiology evidence suggests that reducing ambient PM10 concentrations
may reduce a variety of health effects on a time scale from a few days to a few months.
This has been found in epidemiology studies of "natural experiments" such as in the
Utah Valley, and by supporting toxicology studies using the particles from ambient
community sampling filters from the Utah Valley. Recent studies in Germany and in
the Czech Republic also tend to support a hypothesis that reductions in air pollution are
associated with reductions in the incidence of adverse health effects.
(16) Studies that combine the features of cross-sectional and cohort studies provide the best
evidence for chronic effects of PM exposure. Gauderman et al. (2000; 2002) have
found significant decreases in lung function growth related to PM10 levels using these
techniques.
(17) Adverse health effects in children are emerging as a more important area of concern
than in the 1996 PM AQCD. Unfortunately, relatively little is known about the
relationship of PM to the most serious health endpoints (low birth weight, preterm birth,
neonatal and infant mortality, emergency hospital admissions and mortality in older
children).
(18) Little is yet known about involvement of PM exposure in the progression from less
serious childhood conditions, such as asthma and respiratory symptoms, to more serious
disease endpoints later in life. This is an important health issue because childhood
illness or death may cost a very large number of productive life-years. Lastly, new
epidemiologic studies of ambient PM associations with increased non-hospital medical
visits (physician visits) and asthma effects suggest likely much larger health impacts
and costs to society due to ambient PM than just those indexed by mortality and/or
hospital admissions/visits.
June 2003 8-281 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
3 suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
4 Adventists. Environ. Health Perspect. 94: 43-50.
5 Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
6 ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
7 nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
8 paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
9 7: 19-34.
10 Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
11 2.5 microns in aerodynamic diameter (PM2 5) from airport visibility data in California. J. Exposure Anal.
12 Environ. Epidemiol. 5: 161-180.
13 Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
14 paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med. 158:
15 289-298.
16 Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
17 Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
18 Care Med. 159: 373-382.
19 Ackermann-Liebrich, U.; Leuenberger, P.; Schwartz, J.; Schindler, C.; Monn, C.; Bolognini, B.; Bongard, J. P.;
20 Brandli, O.; Domenighetti, G.; Elsasser, S.; Grize, L.; Karrer, W.; Keller, R.; Keller-Wossidlo, H.; Ktinzli, N.;
21 Martin, B. W.; Medici, T. C.; Perruchoud, A. P.; Schoni, M. H.; Tschopp, J. M.; Villiger, B.; Wuthrich, B.;
22 Zellweger, J. P.; Zemp, E. (1997) Lung function and long term exposure to air pollutants in Switzerland. Am.
23 J. Respir. Crit. Care Med. 155: 122-129.
24 Agocs, M. M.; White, M. C.; Ursicz, G.; Olson, D. R.; Vamos, A. (1997) A longitudinal study of ambient air
25 pollutants and the lung peak expiratory flow rates among asthmatic children in Hungary. Int. J. Epidemiol.
26 26: 1272-1280.
27 Alberdi Odriozola, J. C.; Diaz Jimenez, J.; Montero Rubio, J. C.; MironPerez, I. J.; Pajares Ortiz, M. S.; Ribera
28 Rodrigues, P. (1998) Air pollution and mortality in Madrid, Spain: a time-series analysis. Int. Arch. Occup.
29 Environ. Health 71: 543-549.
30 Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Strachan, D. P. (1996) Air pollution and daily
31 mortality in London: 1987-92. Br. Med. J. 312: 665-669.
32 Anderson, H. R.; Spix, C.; Medina, S.; Schouten, J. P.; Castellsague, J.; Rossi, G.; Zmirou, D.; Touloumi, G.;
33 Wojtyniak, B.; Ponka, A.; Bacharova, L.; Schwartz, J.; Katsouyanni, K. (1997) Air pollution and daily
34 admissions for chronic obstructive pulmonary disease in 6 European cities: results from the APHEA project.
35 Eur. Respir. J. 10: 1064-1071.
36 Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Emberlin, J.; Strachen, D. P. (1998) Air pollution,
3 7 pollens, and daily admissions for asthma in London 1987-92. Thorax 53: 842-848.
38 Anderson, H. R.; Bremner, S. A.; Atkinson, R. W.; Harrison, R. M.; Walters, S. (2001) Paniculate matter and daily
3 9 mortality and hospital admissions in the west midlands conurbation of the United Kingdom: associations with
40 fine and coarse particles, black smoke and sulphate. Occup. Environ. Med. 58: 504-510.
41 Archer, V. E. (1990) Air pollution and fatal lung disease in three Utah counties. Arch. Environ. Health 45: 325-334.
42 Atkinson, R. W.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Bremner, S. A.; Ponce de Leon, A. (1999a)
43 Short-term associations between outdoor air pollution and visits to accident and emergency departments in
44 London for respiratory complaints. Eur. Respir. J. 13: 257-265.
45 Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Ponce de Leon, A. (1999b)
46 Short-term associations between emergency hospital admissions for respiratory and cardiovascular disease
47 and outdoor air pollution in London. Arch. Environ. Health 54: 398-411.
48 Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
49 Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2001) Acute effects of paniculate air pollution on
50 respiratory admissions: results from APHEA 2 project. Am. J. Respir. Crit. Care Med. 164: 1860-1866.
51 Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
52 Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2003) Acute effects of paniculate air pollution on
53 respiratory admissions. In: Revised analyses of time-series studies of air pollution and health. Special report.
54 Boston, MA: Health Effects Institute; pp. 81-84. Available: http://www.healtheffects.org/news.htm [16 May,
55 2003].
June 2003 8-282 DRAFT-DO NOT QUOTE OR CITE
-------
1 Avol, E. L.; Gauderman, W. I; Tan, S. M; London, S. I; Peters, J. M. (2001) Respiratory effects of relocating to
2 areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
3 Awasthi, S.; Click, H. A.; Fletcher, R. H.; Ahmed, N. (1996) Ambient air pollution & respiratory symptoms
4 complex in preschool children. Indian J. Med. Res. 104: 257-262.
5 Ayres, J. G. (2002) Chronic effects of air pollution [editorial]. Occup. Environ. Med. 59: 147-148.
6 Bailey, D. L. R.; Clayton, P. (1982) The measurement of suspended particle and total carbon concentrations in the
7 atmosphere using standard smoke shade methods. Atmos. Environ. 16: 2683-2690.
8 Baldi, L; Tessier, J. F.; Kauffmann, F.; Jacqmin-Gadda, H.; Nejjari, C.; Salamon, R. (1999) Prevalence of asthma
9 and mean levels of air pollution: results from the French PAARC survey. Eur. Respir. J. 14: 132-138.
10 Barbone, F.; Bovenzi, M.; Cavalleri, F.; Stanta, G. (1995) Air pollution and lung cancer in Trieste, Italy. Am. J.
11 Epidemiol. 141: 1161-1169.
12 Basu, R.; Samet, J. M. (1999) A review of the epidemiological evidence on health effects of nitrogen dioxide
13 exposure from gas stoves. J. Environ. Med. 1: 173-187.
14 Bateson, T. F.; Schwartz, J. (1999) Control for seasonal variation and time trend in case-crossover studies of acute
15 effects of environmental exposures. Epidemiology 10: 539-544.
16 Bateson, T. F.; Schwartz, J. (2001) Selection bias and confounding in case-crossover analyses of environmental
17 time-series data. Epidemiology 12: 654-661.
18 Beckett, W. S. (2001) The air pollution detectives. Am. J. Respir. Crit. Care Med. 164: 515-516.
19 Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
20 lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
21 Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
22 and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
23 Beyer, U.; Franke, K.; Cyrys, J.; Peters, A.; Heinrich, J.; Wichmann, H. E.; Brunekreef, B. (1998) Air pollution and
24 respiratory health of children: the PEACE panel study in Hettstedt and Zerbst, Eastern Germany. Eur. Respir.
25 Rev. 8: 61-69.
26 Bobak, M.; Leon, D. A. (1992) Air pollution and infant mortality in the Czech Republic, 1986-1988.
27 Lancet (8826): 1010-1014.
28 Bobak, M.; Leon, D. (1998) Air pollution and infant mortality: the effects are specific for respiratory causes in
29 postneonatal period. Epidemiology 9: S58.
30 Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
31 Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
32 Bobak, M.; Roberts, A. (1997) Heterogeneity of air pollution effects is related to average temperature [letter].
33 Br. Med. J. 315: 1161.
34 Boezen, M.; Schouten, J.; Rijcken, B.; Vonk, J.; Gerritsen, J.; VanDer Zee, S.; Hoek, G.; Brunekreef, B.;
3 5 Postma, D. (1998) Peak expiratory flow variability, bronchial responsiveness, and susceptibility to ambient
36 air pollution in adults. Am. J. Respir. Crit. Care Med. 158: 1848-1854.
37 Boezen, H. M.; Van Der Zee, S. C.; Postma, D. S.; Vonk, J. M.; Gerritsen, J.; Hoek, G.; Brunekreef, B.; Rijcken, B.;
3 8 Schouten, J. P. (1999) Effects of ambient air pollution on upper and lower respiratory symptoms and peak
39 expiratory flow in children. Lancet 353: 874-878.
40 Borja-Aburto, V H.; Loomis, D. P.; Bangdiwala, S. I.; Shy, C. M.; Rascon-Pacheco, R. A. (1997) Ozone, suspended
41 particulates, and daily mortality in Mexico City. Am. J. Epidemiol. 145: 258-268.
42 Borja-Aburto, V H.; Castillejos, M.; Gold, D. R.; Bierzwinski, S.; Loomis, D. (1998) Mortality and ambient fine
43 particles in southwest Mexico City, 1993-1995. Environ. Health Perspect. 106: 849-855.
44 Braga, A. L. F.; Conceicao, G. M. S.; Pereira, L. A. A.; Kishi, H. S.; Pereira, J. C. R.; Andrade, M. F.; Goncalves,
45 F. L. T.; Saldiva, P. H. N.; Latorre, M. R. D. O. (1999) Air pollution and pediatric respiratory hospital
46 admissions in Sao Paulo, Brazil. J. Environ. Med. 1: 95-102.
47 Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2000) Do respiratory epidemics confound the association between air
48 pollution and daily deaths? Eur. Respir. J. 16: 723-728.
49 Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2001) The lag structure between paniculate air pollution and respiratory
50 and cardiovascular deaths in ten U.S. cities. J. Occup. Environ. Med. 43: 927-933.
51 Brauer, M.; Ebelt, S. T.; Fisher, T. V.; Brumm, J.; Petkau, A. J.; Vedal, S. (2001) Exposure of chronic obstructive
52 pulmonary disease patients to particles: respiratory and cardiovascular health effects. J. Exposure Anal.
53 Environ. Epidemiol. 11: 490-500.
54 Braun-Fahrlander, C.; Vuille, J. C.; Sennhauser, F. H.; Neu, U.; Ktinzle, T.; Grize, L.; Gassner, M.; Minder, C.;
55 Schindler, C.; Varonier, H. S.; Wuthrich, B.; SCARPOL team. (1997) Respiratory health and long-term
56 exposure to air pollutants in Swiss schoolchildren. Am. J. Respir. Crit. Care Med. 155: 1042-1049.
June 2003 8-283 DRAFT-DO NOT QUOTE OR CITE
-------
1 Bremner, S. A.; Anderson, H. R.; Atkinson, R. W.; McMichael, A. J.; Strachan, D. P.; Bland, J. M; Bower, J. S.
2 (1999) Short term associations between outdoor air pollution and mortality in London 1992-4. Occup.
3 Environ. Med. 56: 237-244.
4 Brook, J. R.; Wiebe, A. H.; Woodhouse, S. A.; Audette, C. V.; Dann, T. F.; Callaghan, S.; Piechowski, M.;
5 Dabek-Zlotorzynska, E.; Dloughy, J. F. (1997) Temporal and spatial relationships in fine particle strong
6 acidity, sulphate, PM10, and PM2 5 across multiple Canadian locations. Atmos. Environ. 31: 4223-4236.
7 Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
8 Buffler, P. A.; Cooper, S. P.; Stinnett, S.; Contant, C.; Shirts, S.; Hardy, R. J.; Agu, V.; Gehan, B.; Burau, K. (1988)
9 Air pollution and lung cancer mortality in Harris County, Texas, 1979-1981. Am. J. Epidemiol. 128: 683-699.
10 Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian
11 cities. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
12 Health Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
13 Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M.;
14 Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of
15 respiratory admissions to Ontario hospitals. Environ. Res. 65: 172-194.
16 Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
17 paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
18 142: 15-22.
19 Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997a) The role of paniculate size and chemistry in the
20 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
21 Environ. Health Perspect. 105: 614-620.
22 Burnett, R. T.; Brook, J. R.; Yung, W. T.; Dales, R. E.; Krewski, D. (1997b) Association between ozone and
23 hospitalization for respiratory diseases in 16 Canadian cities. Environ. Res. 72: 24-31.
24 Burnett, R. T.; Dales, R. E.; Brook, J. R.; Raizenne, M. E.; Krewski, D. (1997c) Association between ambient
25 carbon monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities.
26 Epidemiology 8: 162-167.
27 Burnett, R. T.; Cakmak, S.; Brook, J. R. (1998a) The effect of the urban ambient air pollution mix on daily mortality
28 rates in 11 Canadian cities. Can. J. Public Health 89: 152-156.
29 Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
30 Ozkaynak, H. (1998b) The association between ambient carbon monoxide levels and daily mortality in
31 Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
32 Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
33 pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54:130-139.
34 Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
3 5 (2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
36 in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
37 12(suppl. 4): 15-39.
38 Burnett, R.; Ma, R.; Jerrett, M.; Goldberg, M. S.; Cakmak, S.; Pope, C. A., Ill; Krewski, D. (2001a) The spatial
3 9 association between community air pollution and mortality: a new method of analyzing correlated geographic
40 cohort data. Environ. Health Perspect. 109(suppl. 3): 375-380.
41 Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Raizenne, M. E.; Brook, J. R.; Dales, R. E.; Leech, J. A.; Cakmak, S.;
42 Krewski, D. (200 Ib) Association between ozone and hospitalization for acute respiratory diseases in children
43 less than 2 years of age. Am. J. Epidemiol. 153: 444-452.
44 Cakmak, S.; Burnett, R. T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
45 concentrations for air pollution-related mortality with exposure measurement error. Risk Anal. 19: 487-496.
46 Calderon-Garciduenas, L.; Mora-Tiscareno, A.; Chung, C. J.; Valencia, G.; Fordham, L. A.; Garcia, R.; Osnaya, N.;
47 Romero, L.; Acuna, H.; Villarreal-Calderon, A.; Devlin, R. B.; Koren, H. S. (2000) Exposure to air pollution
48 is associated with lung hyperinflation in healthy children and adolescents in southwest Mexico City: a pilot
49 study. Inhalation Toxicol. 12: 537-561.
50 Carroll, R. J.; Ruppert, D.; Stefanski, L. A. (1995) Measurement error in nonlinear models. London, United
51 Kingdom: Chapman & Hall. (Cox, D. R.; Hinkley, D. V.; Keiding, N.; Reid, N.; Rubin, D. B.; Silverman,
52 B. W., eds. Monographs on statistics and applied probability: v. 63).
53 Carrothers, T. J.; Evans, J. S. (2000) Assessing the impact of differential measurement error on estimates of fine
54 particle mortality. J. Air Waste Manage. Assoc. 50: 65-74.
55 Cassino, C.; Ito, K.; Bader, L; Ciotoli, C.; Thurston, G.; Reibman, J. (1999) Cigarette smoking and ozone-associated
56 emergency department use for asthma by adults in New York City. Am. J. Respir. Crit. Care Med.
57 159: 1773-1779.
June 2003 8-284 DRAFT-DO NOT QUOTE OR CITE
-------
1 Castillejos, M; Borja-Aburto, V. H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
2 mortality. In: Phalen, R. F., ed. Inhalation toxicology: proceedings of the third colloquium on paniculate air
3 pollution and human health (first special issue); June, 1999; Durham, NC. Inhalation Toxicol.
4 12(suppl. 1): 61-72.
5 Chay, K. Y.; Greenstone, M. (2001a) Air quality, infant mortality, and the Clean Air Act of 1970. Berkeley, CA:
6 University of California, Berkeley, Department of Economics. Available:
7 http://elsa.berkeley.edu/users/kenchay/ [30 May, 2003].
8 Chay, K. Y.; Greenstone, M. (200 Ib) The impact of air pollution on infant mortality: evidence from geographic
9 variation in pollution shocks induced by a recession. Berkeley, CA: University of California, Berkeley,
10 Department of Economics. Available: http://elsa.berkeley.edu/users/kenchay/ [30 May, 2003].
11 Checkoway, H.; Levy, D.; Sheppard, L.; Kaufman, I; Koenig, I; Siscovick, D. (2000) A case-crossover analysis of
12 fine paniculate matter air pollution and out-of-hospital sudden cardiac arrest. Cambridge, MA: Health Effects
13 Institute; research report no. 99. Available: http://www.healtheffects.org/pubs-recent.htm [19 March, 2001].
14 Chen, C.; Chock, D. P.; Winkler, S. L. (1999) A simulation study of confounding in generalized linear models for air
15 pollution epidemiology. Environ. Health Perspect. 107: 217-222.
16 Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
17 chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
18 Chen, L.; Yang, W.; Jennison, B. L.; Goodrich, A.; Omaye, S. T. (2002) Air pollution and birth weight in northern
19 Nevada, 1991-1999. Inhalation Toxicol. 14: 141-157.
20 Chew, F. T.; Goh, D. Y. T.; Ooi, B. C.; Saharom, R.; Hui, J. K. S.; Lee, B. W. (1999) Association of ambient
21 air-pollution levels with acute asthma exacerbation among children in Singapore. Allergy (Copenhagen)
22 54:320-329.
23 Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
24 pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
25 Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
26 pollution. Arch. Environ. Health 52: 113-117.
27 Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
28 coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
29 50: 1287-1298.
30 Clench-Aas, J.; Bartonova, A.; Skjensberg, O. H.; Leegaard, J.; Hagen, L. O.; Gisver, P.; Moseng, J.; Roemer, W.
31 (1998) Air pollution and respiratory health of children: the PEACE study in Oslo, Norway. Eur. Respir. Rev.
32 8:36-43.
33 Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
34 Arizona. Seattle, WA: University of Washington, National Research Center for Statistics and the
35 Environment; NRCSE technical report series, NRCSE-TRS no. 040. Available:
36 http://www.nrcse.washington.edu/research/reports.html [30 May, 2003].
37 Cohen, A. J. (2000) Outdoor air pollution and lung cancer. Environ. Health Perspect. 108(suppl. 4): 743-750.
3 8 Cohen, A. J.; Pope, C. A., III. (1995) Lung cancer and air pollution. Environ. Health Perspect.
39 103(suppl. 8): 219-224.
40 Conceicao, G. M. S.; Miraglia, S. G. E. K.; Kishi, H. S.; Saldiva, P. H. N.; Singer, J. M. (2001) Air pollution and
41 child mortality: a time-series study in Sao Paulo, Brazil. Environ. Health Perspect. 109(suppl. 3): 347-350.
42 Cook, J. R.; Stefanski, L. A. (1994) Simulation-extrapolation estimation in parametric measurement error models.
43 J. Am. Stat. Assoc. 89: 1314-1328.
44 Coull, B. A.; Schwartz, J.; Wand, M. P. (2001) Respiratory health and air pollution: additive mixed model analyses.
45 Biostatistics 2: 337-349.
46 Coutant, R. W. (1977) Effect of environmental variables on collection of atmospheric sulfate. Environ. Sci. Technol.
47 11:873-878.
48 Creason, J.; Neas, L.; Walsh, D.; Williams, R.; Sheldon, L.; Liao, D.; Shy, C. (2001) Paniculate matter and heart
49 rate variability among elderly retirees: the Baltimore 1998 PM study. J. Exposure Anal. Environ. Epidemiol.
50 11: 116-122.
51 Cropper, M. L.; Simon, N. B.; Alberini, A.; Arora, S.; Sharma, P. K. (1997) The health benefits of air pollution
52 control in Delhi. Am. J. Agric. Econ. 79: 1625-1629.
53 Cuijpers, C. E. J.; Swaen, G. M. H.; Wesseling, G.; Wouters, E. F. M. (1994) Acute respiratory effects of summer
54 smog in primary school children. Toxicol. Lett. 72: 227-235.
55
June 2003 8-285 DRAFT-DO NOT QUOTE OR CITE
-------
1 Dab, W.; Medina, S.; Quenel, P.; Le Moullec, Y.; Le Tertre, A.; Thelot, B.; Monteil, C.; Lameloise, P.; Pirard, P.;
2 Momas, I.; Ferry, R.; Festy, B. (1996) Short term respiratory health effects of ambient air pollution: results of
3 the APHEA project in Paris. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on
4 health: a European approach using epidemiological time series data. J. Epidemiol. Community Health
5 50(suppl. 1): S42-S46.
6 Damia, A. D.; Fabregas, M. L.; Tordera, M. P.; Torrero, L. C. (1999) Effects of air pollution and weather conditions
7 on asthma exacerbation. Respiration 66: 52-58.
8 Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
9 curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
10 152:397-406.
11 Daniels, M. J.; Lee, Y.-D.; Kaiser, M. (2001) Assessing sources of variability in measurement of ambient paniculate
12 matter. Environmetrics 12: 547-558.
13 Dejmek, J.; Selevan, S. G.; Benes, L; Solansky, L; Sram, R. J. (1999) Fetal growth and maternal exposure to
14 paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
15 Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
16 relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. CareMed. 154: 633-641.
17 Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
18 on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
19 155: 568-576.
20 Delfino, R. J.; Murphy-Moulton, A. M.; Becklake, M. R. (1998) Emergency room visits for respiratory illnesses
21 among the elderly in Montreal: association with low level ozone exposure. Environ. Res. 76: 61-11.
22 Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; McLaren, C. E. (2002) Association of asthma symptoms
23 with peak paniculate air pollution and effect modification by anti-inflammatory medication use. Environ.
24 Health Perspect. 110: A607-A617.
25 Desqueyroux, H.; Pujet, J.-C.; Prosper, M.; Squinazi, F.; Momas, I. (2002) Short-term effects of low-level air
26 pollution on respiratory health of adults suffering from moderate to severe asthma. Environ. Res. A 89: 29-37.
27 Dewanji, A.; Moolgavkar, S. H. (2000) A Poisson process approach for recurrent event data with environmental
28 covariates. Environmetrics 11: 665-673.
29 Diaz, J.; Garcia, R.; Ribera, P.; Alberdi, J. C.; Hernandez, E.; Pajares, M. S.; Otero, A. (1999) Modeling of air
30 pollution and its relationship with mortality and morbidity in Madrid, Spain. Int. Arch. Occup. Environ.
31 Health 72: 366-376.
32 Dockery, D. W.; Spengler, J. D. (1981) Personal exposure to respirable particulates and sulfates. J. AirPollut.
33 Control Assoc. 31: 153-159.
34 Dockery, D. W.; Spengler, J. D.; Neas, L. M.; Speizer, F. E.; Ferris, B. G., Jr.; Ware, J. H.; Brunekreef, B. (1989)
35 An epidemiologic study of respiratory health status and indicators of indoor air pollution from combustion
36 sources. In: Harper, J. P., ed. Combustion processes and the quality of the indoor environment: transactions of
37 an international specialty conference; September 1988; Niagara Falls, NY. Pittsburgh, PA: Air & Waste
38 Management Association; pp. 262-271. (A&WMA transactions series: TR-15).
39 Dockery, D. W.; Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily mortality: associations with particulates
40 and acid aerosols. Environ. Res. 59: 362-373.
41 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.
42 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 329: 1753-1759.
43 Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
44 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
45 symptoms. Environ. Health Perspect. 104: 500-505.
46 Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
47 saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
48 Health Effects Institute; research report no. 83.
49 Dominici, F.; Samet, J. M.; Zeger, S. (2000a) Combining evidence on air pollution and daily mortality from the
50 largest 20 US cities: a hierarchical modeling strategy. J. R. Stat. Soc. A 163: 263-302.
51 Dominici, F.; Zeger, S. L.; Samet, J. (2000b) A measurement error model for time-series studies of air pollution and
52 mortality. Biostatistics 1: 157-175.
53 Dominici, F.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2002) Air pollution and mortality: estimating regional and
54 national dose-response relationships. J. Am. Stat. Assoc. 97: 100-111.
55
June 2003 8-286 DRAFT-DO NOT QUOTE OR CITE
-------
1 Dominici, F.; Daniels, M.; McDermott, A.; Zeger, S. L.; Samet, J. M. (2003) Shape of the exposure-response
2 relation and mortality displacement in the NMMAPS database. In: Revised analyses of time-series studies of
3 air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 91-96. Available:
4 http://www.healtheffects.org/news.htm [16 May, 2003].
5 Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. L; Ohio, A. J.; Costa, D. L.
6 (2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
7 studies in Utah Valley residents. Environ. Health Perspect. 109(suppl. 3): 395-403.
8 Fairley, D. (1990) The relationship of daily mortality to suspended particulates in Santa Clara county, 1980-86.
9 Environ. Health Perspect. 89: 159-168.
10 Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
11 Perspect. 107: 637-641.
12 Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
13 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp.
14 97-106. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
15 Forsberg, B.; Segerstedt, B.; Stjernberg, N.; Roemer, W. (1998) Air pollution and respiratory health of children:
16 the PEACE panel study in Umea, Sweden. Eur. Respir. Rev. 8: 12-19.
17 Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
18 extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol.
19 277: L960-L967.
20 Friedman, M. S.; Powell, K. E.; Hutwagner, L.; Graham, L. M.; Teague, W. G. (2001) Impact of changes in
21 transportation and commuting behaviors during the 1996 summer Olympic games in Atlanta on air quality and
22 childhood asthma. JAMA J. Am. Med. Assoc. 285: 897-905.
23 Frischer, T.; Studnicka, M.; Gartner, C.; Tauber, E.; Horak, F.; Veiter, A.; Spengler, J.; Ktihr, J.; Urbanek, R. (1999)
24 Lung function growth and ambient ozone: a three-year population study in school children. Am. J. Respir.
25 Crit. Care Med. 160: 390-396.
26 Fung, K. Y.; Krewski, D. (1999) On measurement error adjustment methods in Poisson regression. Environmetrics
27 10:213-224.
28 Fusco, D.; Forastiere, F.; Michelozzi, P.; Spadea, T.; Ostro, B.; Area, M.; Perucci, C. A. (2001) Air pollution and
29 hospital admissions for respiratory conditions in Rome, Italy. Eur. Respir. J. 17: 1143-1150.
30 Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
31 1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
32 June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98-MP26.03.
33 Garcia-Aymerich, J.; Tobias, A.; Anto, J. M.; Sunyer, J. (2000) Air pollution and mortality in a cohort of patients
34 with chronic obstructive pulmonary disease: a time series analysis. J. Epidemiol. Community Health
35 54:73-74.
36 Garry, B. Z.; Kosman, E.; Ganor, E.; Berger, V; Garty, L.; Wietzen, T.; Waisman, Y.; Mimouni, M.; Waisel, Y.
37 (1998) Emergency room visits of asthmatic children, relation to air pollution, weather, and airborne allergens.
38 Ann. Allergy Asthma Immunol. 81: 563-570.
39 Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
40 Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
41 function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
42 Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
43 Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
44 lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
45 Med. 166: 76-84.
46 Gauvin, S.; Zmirou, D.; Pin, I.; Quentin, J.; Balducci, F.; Boudet, C.; Poizeau, D.; Brambilla, C. (1999) Short-term
47 effect of exposure to suspended paniculate matter (PM10) on the respiratory function of urban asthmatic and
48 control adults. J. Environ. Med. 1: 71-79.
49 Gehring, U.; Cyrys, J.; Sedlmeir, G.; Brunekreef, B.; Bellander, T.; Fischer, P.; Bauer, C. P.; Reinhardt, D.;
50 Wichmann, H. E.; Heinrich, J. (2002) Traffic-related air pollution and respiratory health during the first 2 yrs.
51 of life. Eur. Respir. J. 19: 690-698.
52 Gherghinova, M.; Kostyanve, S.; Ivanova, M. (1989) Theoretic values of body plethysmography indices in healthy
53 children between 7 and 14 years of age. Pediatria 28: 52-58.
54 Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
55 Am. J. Respir. Crit. Care Med. 164: 704-708.
June 2003 8-287 DRAFT-DO NOT QUOTE OR CITE
-------
1 Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. I; Brunekreef, B. (1997) Acute effects of
2 summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
3 155:2105-2108.
4 Gold, D. R.; Damokosh, A. I.; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
5 Castillejos, M. (1999) Paniculate and ozone pollutant effects on the respiratory function of children in
6 southwest Mexico City. Epidemiology 10: 8-16.
7 Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
8 Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
9 Gold, D. R.; Schwartz, J.; Litonjua, A.; Verrier, R.; Zanobetti, A. (2003) Ambient pollution and reduced heart rate
10 variability. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
11 Health Effects Institute; pp. 107-112. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
12 Goldberg, M. S.; Burnett, R. T. (2003) Revised analysis of the Montreal time-series study. In: Revised analyses of
13 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
14 pp. 113-132. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
15 Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
16 Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
17 short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
18 Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm
19 [15 February, 2001].
20 Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Brook, J.; Bonvalot, Y.; Tamblyn, R.; Singh, R.; Valois, M.-F.
21 (200 la) The association between daily mortality and ambient air particle pollution in Montreal, Quebec. 1.
22 Nonaccidental mortality. Environ. Res. 86: 12-25.
23 Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Brook, J.; Bonvalot, Y.; Tamblyn, R.; Singh, R.; Valois, M.-F.;
24 Vincent, R. (200 Ib) The association between daily mortality and ambient air particle pollution in Montreal,
25 Quebec. 2. Cause-specific mortality. Environ. Res. 86: 26-36.
26 Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Tamblyn, R.; Ernst, P.; Flegel, K.; Brook, J.; Bonvalot, Y.;
27 Singh, R.; Valois, M.-F.; Vincent, R. (200 Ic) Identification of persons with cardiorespiratory conditions who
28 are at risk of dying from the acute effects of ambient air particles. Environ. Health Perspect.
29 109(suppl. 4): 487-494.
30 Goldberg, M. S.; Burnett, R. T.; Brook, J.; Bailar, J. C., Ill; Valois, M.-F.; Vincent, R. (2001d) Associations between
31 daily cause-specific mortality and concentrations of ground-level ozone in Montreal, Quebec. Am. J.
32 Epidemiol. 154: 817-826.
33 Goldberg, M. S.; Burnett, R. T.; Valois, M.-F.; Flegel, K.; Bailar, J. C., Ill; Brook, J.; Vincent, R.; Radon, K. (2003)
34 Associations between ambient air pollution and daily mortality among persons with congestive heart failure.
3 5 Environ. Res. 91: 8-20.
36 Gordian, M. E.; Ozkaynak, H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Paniculate air pollution and respiratory
37 disease in Anchorage, Alaska. Environ. Health Perspect. 104: 290-297.
3 8 Gouveia, N.; Fletcher, T. (2000) Respiratory diseases in children and outdoor air pollution in Sao Paulo, Brazil:
39 a time series analysis. Occup. Environ. Med. 57: 477-483.
40 Grievink, L.; VanderZee, S. C.; Hoek, G.; Boezen, H. M.; Van't Veer, P.; Brunekreef, B. (1999) Modulation of the
41 acute respiratory effects of winter air pollution by serum and dietary antioxidants: a panel study. Eur. Respir.
42 J. 13: 1439-1446.
43 Gtintzel, O.; Bollag, U.; Helfenstein, U. (1996) Asthma and exacerbation of chronic bronchitis: sentinel and
44 environmental data in a time series analysis. Zentralbl. Hyg. Umweltmed. 198: 383-393.
45 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,
46 D. W. (1999) Climate, traffic-related air pollutants, and asthma prevalence in middle-school children in
47 Taiwan. Environ. Health Perspect. 107: 1001-1006.
48 Gwynn, R. C.; Thurston, G. D. (2001) The burden of air pollution: impacts among racial minorities. Environ. Health
49 Perspect. 109(suppl. 4): 501-506.
50 Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
51 mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
52 Hagen, J. A.; Nafstad, P.; Skrondal, A.; Bjerkly, S.; Magnus, P. (2000) Associations between outdoor air pollutants
53 and hospitalization for respiratory diseases. Epidemiology 11: 136-140.
54 Hajat, S.; Haines, A.; Goubet, S. A.; Atkinson, R. W.; Anderson, H. R. (1999) Association of air pollution with daily
55 GP consultations for asthma and other lower respiratory conditions in London. Thorax 54: 597-605.
June 2003 8-288 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hajat, S.; Haines, A.; Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Emberlin, J. (2001) Association between
2 air pollution and daily consultations with general practitioners for allergic rhinitis in London, United
3 Kingdom. Am. J. Epidemiol. 153: 704-714.
4 Haluszka, J.; Pisiewicz, K.; Miczynski, J.; Roemer, W.; Tomalak, W. (1998) Air pollution and respiratory health in
5 children: the PEACE panel study in Krakow, Poland. Eur. Respir. Rev. 8: 94-100.
6 Harre, E. S. M; Price, P. D.; Ayrey, R. B.; Toop, L. J.; Martin, I. R.; Town, G. I. (1997) Respiratory effects of air
7 pollution in chronic obstructive pulmonary disease: a three month prospective study. Thorax 52: 1040-1044.
8 Harris, R. J. (1975) A primer of multivariate statistics. New York, NY: Academic Press; pp. 155-224.
9 Health Effects Institute. (2002) Reanalysis of National Morbidity, Mortality, and Air Pollution Study (NMMAPS)
10 [letter with attachments]. Boston, MA: May 30.
11 Health Effects Institute. (2003a) Revised analyses of the National Morbidity, Mortality, and Air Pollution Study
12 (NMMAPS), part II. In: Revised analyses of time-series studies of air pollution and health. Special report.
13 Boston, MA: Health Effects Institute; pp. 9-72. Available: http://www.healtheffects.org/news.htm [16 May,
14 2003].
15 Health Effects Institute. (2003b) Commentary on revised analyses of selected studies. In: Revised analyses of
16 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
17 pp. 255-290. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
18 Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
19 dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
20 Heinrich, J.; Hoelscher, B.; Wjst, M.; Ritz, B.; Cyrys, J.; Wichmann, H.-E. (1999) Respiratory diseases and allergies
21 in two polluted areas in East Germany. Environ. Health Perspect. 107: 53-62.
22 Heinrich, J.; Hoelscher, B.; Wichmann, H. E. (2000) Decline of ambient air pollution and respiratory symptoms in
23 children. Am. J. Respir. Crit. Care Med. 161: 1930-1936.
24 Heinrich, J.; Hoelscher, B.; Frye, C.; Meyer, L; Pitz, M.; Cyrys, J.; Wjst, M.; Neas, L.; Wichmann, H.-E. (2002)
25 Improved air quality in reunified Germany and decreases in respiratory symptoms. Epidemiology
26 13:394-401.
27 Henderson, B. E.; Gordon, R. J.; Menck, H.; SooHoo, J.; Martin, S. P.; Pike, M. C. (1975) Lung cancer and air
28 pollution in southcentral Los Angeles County. Am. J. Epidemiol. 101: 477-488.
29 Hickey, R. J.; Boyce, D. E.; Harner, E. B.; Clelland, R. C. (1970) Ecological statistical studies on environmental
30 pollution and chronic disease in metropolitan areas of the United States. Philadelphia, PA: Regional Science
31 Research Institute. (RSRI discussion paper series no. 35).
32 Hill, A. B. (1965) The environment and disease: association or causation? Proc. R. Soc. Med. 58: 295-300.
33 Hiltermann, T. J. N.; de Bruijne, C. R.; Stolk, J.; Zwinderman, A. H.; Spieksma, F. Th. M.; Roemer, W.;
34 Steerenberg, P. A.; Fischer, P. H.; van Bree, L.; Hiemstra, P. S. (1997) Effects of photochemical air pollution
3 5 and allergen exposure on upper respiratory tract inflammation in asthmatics. Am. J. Respir. Crit. Care Med.
36 156: 1765-1772.
37 Hiltermann, T. J. N.; Stolk, J.; Van der Zee, S. C.; Brunekreef, B.; De Bruijne, C. R.; Fischer, P. H.; Ameling, C. B.;
38 Sterk, P. J.; Hiemstra, P. S.; VanBree, L. (1998) Asthma severity and susceptibility to air pollution. Eur.
39 Respir. J. 11:686-693.
40 Hoek, G. (2003) Daily mortality and air pollution in The Netherlands. In: Revised analyses of time-series studies of
41 air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 133-142. Available:
42 http://www.healtheffects.org/news.htm [16 May, 2003].
43 Hoek, G.; Schwartz, J. D.; Groot, B.; Eilers, P. (1997) Effects of ambient paniculate matter and ozone on daily
44 mortality in Rotterdam, the Netherlands. Arch. Environ. Health 52: 455-463.
45 Hoek, G.; Dockery, D. W.; Pope, A.; Neas, L.; Roemer, W.; Brunekreef, B. (1998) Association between PM10 and
46 decrements in peak expiratory flow rates in children: reanalysis of data from five panel studies. Eur. Respir. J.
47 11: 1307-1311.
48 Hoek, G.; Brunekreef, B.; Verhoeff, A.; van, Wijnen, J.; Fischer, P. (2000) Daily mortality and air pollution in the
49 Netherlands. J. Air Waste Manage. Assoc. 50: 1380-1389.
50 Hoek, G.; Brunekreef, B.; Fischer, P.; Van Wijnen, J. (2001) The association between air pollution and heart failure,
51 arrhythmia, embolism, thrombosis, and other cardiovascular causes of death in a time series study.
52 Epidemiology 12: 355-357.
53 Hodgkin, J. E.; Abbey, D. E.; Euler, G. L.; Magie, A. R. (1984) COPD prevalence in nonsmokers in high and low
54 photochemical air pollution areas. Chest 86: 830-838.
55 Hong, Y.-C.; Leem, J.-H.; Ha, E.-H.; Christiani, D. C. (1999) PM10 exposure, gaseous pollutants, and daily mortality
56 in Inchon, South Korea. Environ. Health Perspect. 107: 873-878.
June 2003 8-289 DRAFT-DO NOT QUOTE OR CITE
-------
1 Horak, F., Jr.; Studnicka, M.; Gartner, C.; Spengler, J. D.; Tauber, E.; Urbanek, R.; Veiter, A.; Frischer, T. (2002)
2 Paniculate matter and lung function growth in children: a 3-yr follow-up study in Austrian schoolchildren.
3 Eur. Respir. J. 19: 838-845.
4 Howel, D.; Darnell, R.; Pless-Mulloli, T. (2001) Children's respiratory health and daily paniculate levels in
5 10 nonurban communities. Environ. Res. 87: 1-9.
6 Huang, Y.-L.; Batterman, S. (2000) Residence location as a measure of environmental exposure: a review of air
7 pollution epidemiology studies. J. Exposure Anal. Environ. Epidemiol. 10: 66-85.
8 Ibald-Mulli, A.; Stieber, J.; Wichmann, H. -E.; Koenig, W.; Peters, A. (2001) Effects of air pollution on blood
9 pressure: a population-based approach. Am. J. Public Health 91: 571-577.
10 Ilabaca, M.; Olaeta, I.; Campos, E.; Villaire, J.; Tellez-Rojo, M. M.; Romieu, I. (1999) Association between levels
11 of fine paniculate and emergency visits for pneumonia and other respiratory illnesses among children in
12 Santiago, Chile. J. Air Waste Manage. Assoc. 49: 154-163.
13 Ito, K. (1990) An examination of the role of aerosol acidity in historical London, England daily mortality
14 [dissertation]. Syracuse, NY: New York University. Available from: University Microfilms International,
15 Ann Arbor, MI; AAD91-13012.
16 Ito, K. (2003) Associations of paniculate matter components with daily mortality and morbidity in Detroit,
17 Michigan. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
18 Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
19 Ito, K.; Kinney, P.; Thurston, G. D. (1995) Variations inPM-10 concentrations within two metropolitan areas and
20 their implications for health effects analyses. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the
21 colloquium on paniculate air pollution and human mortality and morbidity, part II; January 1994; Irvine, CA.
22 Inhalation Toxicol. 7: 735-745.
23 Ito, K.; Thurston, G. D.; Nadas, A.; Lippmann, M. (2001) Monitor-to-monitor temporal correlation of air pollution
24 and weather variables in the North-Central U.S. J. Exposure Anal. Environ. Epidemiol. 11: 21-32.
25 Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
26 1983-1992. Environ. HealthPerspect. 105: 980-985.
27 Jalaludin, B. B.; Chey, T.; O'Toole, B. I.; Smith, W. T.; Capon, A. G.; Leeder, S. R. (2000) Acute effects of low
28 levels of ambient ozone on peak expiratory flow rate in a cohort of Australian children. Int. J. Epidemiol.
29 29: 549-557.
30 Jamason, P. F.; Kalkstein, L. S.; Gergen, P. J. (1997) A synoptic evaluation of asthma hospital admissions in
31 New York City. Am. J. Respir. Crit. Care Med. 156: 1781-1788.
32 Janssen, N. A. H.; Schwartz, J.; Zanobetti, A.; Suh, H. H. (2002) Air conditioning and source-specific particles as
3 3 modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ. Health Perspect.
34 110:43-49.
35 Jedrychowski, W.; Becher, H.; Wahrendorf, J.; Basa-Cierpialek, Z. (1990) A case-control study of lung cancer with
36 special erence to the effect of air pollution in Poland. J. Epidemiol. Commun. Health 44: 114-120.
37 Jedrychowski, W.; Flak, E. (1998) Separate and combined effects of the outdoor and indoor air quality on chronic
3 8 respiratory symptoms adjusted for allergy among preadolescent children. Int. J. Occup. Med. Environ. Health
39 11: 19-35.
40 Jedrychowski, W.; Flak, E.; Mroz, E. (1999) The adverse effect of low levels of ambient air pollutants on lung
41 function growth in preadolescent children. Environ. Health Perspect. 107: 669-674.
42 Just, J.; Segala, C.; Sahraoui, F.; Priol, G.; Grimfeld, A.; Neukirch, F. (2002) Short-term health effects of paniculate
43 and photochemical air pollution in asthmatic children. Eur. Respir. J. 20: 899-906.
44 Katsouyanni, K.; Touloumi, G. (1998) Causes of regional differences in air pollution effects are being studied
45 further [letter]. Br. Med. J. 316: 1982.
46 Katsouyanni, K.; Karakatsani, A.; Messari, L; Touloumi, G.; Hatzakis, A.; Kalandidi, A.; Trichopoulos, D. (1990)
47 Air pollution and cause specific mortality in Athens. J. Epidemiol. Community Health 44: 321-324.
48 Katsouyanni, K.; Schwartz, J.; Spix, C.; Touloumi, G.; Zmirou, D.; Zanobetti, A.; Wojtyniak, B.; Vonk, J. M.;
49 Tobias, A.; Ponka, A.; Medina, S.; Bacharova, L.; Andersen, H. R. (1996) Short term effects of air pollution
50 on health: a European approach using epidemiology time series data: the APHEA protocol. In: St Leger, S.,
51 ed. The APHEA project. Short term effects of air pollution on health: a European approach using
52 epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S12-S18.
53 Katsouyanni, K.; Touloumi, G.; Spix, C.; Schwartz, J.; Balducci, F.; Medina, S.; Rossi, G.; Wojtyniak, B.;
54 Sunyer, J.; Bacharova, L.; Schouten, J. P.; Ponka, A.; Anderson, H. R. (1997) Short term effects of ambient
55 sulphur dioxide and paniculate matter on mortality in 12 European cities: results from time series data from
56 the APHEA project. Br. Med. J. 314: 1658-1663.
June 2003 8-290 DRAFT-DO NOT QUOTE OR CITE
-------
1 Katsouyanni, K.; Touloumi, G.; Samoli, E.; Gryparis, A.; Le Tertre, A.; Monopolis, Y.; Rossi, G.; Zmirou, D.;
2 Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
3 Schindler, C.; Schwartz, J. (2001) Confounding and effect modification in the short-term effects of ambient
4 particles on total mortality: results from 29 European cities within the APHEA2 project. Epidemiology
5 12:521-531.
6 Katsouyanni, K.; Touloumi, G.; Samoli, E.; Petasakis, Y.; Analitis, A.; Le Tertre, A.; Rossi, G.; Zmirou, D.;
7 Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
8 Schindler, C.; Schwartz, J. (2003) Sensitivity analysis of various models of short-term effects of ambient
9 particles on total mortality in 29 cities in APHEA2. In: Revised analyses of time-series studies of air pollution
10 and health. Special report. Boston, MA: Health Effects Institute; pp. 157-164. Available:
11 http://www.healtheffects.org/news.htm [16 May, 2003].
12 Keatinge, W. R.; Donaldson, G. C. (2001) Mortality related to cold and air pollution in London after allowance for
13 effects of associated weather patterns. Environ. Res. A 86: 209-216.
14 Keles, N.; Ilicali, O. C.; Deger, K. (1999) Impact of air pollution on prevalence of rhinitis in Istanbul. Arch. Environ.
15 Health 54: 48-51.
16 Kelsall, J. E.; Samet, J. M.; Zeger, S. L.; Xu, J. (1997) Air pollution and mortality in Philadelphia, 1974-1988.
17 Am. J. Epidemiol. 146: 750-762.
18 Kinney, P. L.; Aggarwal, M.; Northridge, M. E.; Janssen, N. A. H.; Shepard, P. (2000) Airborne concentrations of
19 PM2 5 and diesel exhaust particles on Harlem sidewalks: a community-based pilot study. Environ. Health
20 Perspect. 108: 213-218.
21 Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
22 and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
23 Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
24 of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
25 pp. 165-172. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
26 Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
27 specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
28 50: 1215-1222.
29 Koenig, J. Q.; Larson, T. V.; Hanley, Q. S.; Rebolledo, V.; Dumler, K.; Checkoway, H.; Wang, S.-Z.; Lin, D.;
30 Pierson, W. E. (1993) Pulmonary function changes in children associated with fine paniculate matter.
31 Environ. Res. 63: 26-38.
32 Kontos, A. S.; Fassois, S. D.; Deli, M. F. (1999) Short-term effects of air pollution on childhood respiratory illness
33 in Piraeus, Greece, 1987-1992: nonparametric stochastic dynamic analysis. Environ. Res. 81: 275-296.
34 Koren, H. S.; Utell, M. J. (1997) Asthma and the environment. Environ. Health Perspect. 105: 534-537.
35 Korrick, S. A.; Neas, L. M.; Dockery, D. W.; Gold, D. R.; Allen, G. A.; Hill, L. B.; Kimball, K. D.; Rosner, B. A.;
36 Speizer, F. E. (1998) Effects of ozone and other pollutants on the pulmonary function of adult hikers.
37 Environ. Health Perspect. 106: 93-99.
38 Kostianev, S.; Gerginova, M.; Ivanova, M. (1994) erence values of lung function parameters in Bulgarian girls aged
39 7 to 14 years. Pediatria 33: 30-33.
40 Kotesovec, F.; Vitnerova, N.; Leixner, M.; Benes, I.; Skorkovsky, J.; Roemer, W. (1998) Air pollution and
41 respiratory health of children: the PEACE panel study in Teplice, Czech Republic. Eur. Respir. Rev. 8: 70-77.
42 Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Peters, A.; Heinrich, J. (2000) Daily mortality and air pollution in
43 northern Bohemia; different effects for men and women. Cent. Eur. J. Public Health 8: 120-127.
44 Kramer, U.; Behrendt, H.; Dolgner, R.; Ranft, U.; Ring, J.; Wilier, H.; Schlipkoter, H.-W. (1999) Airway diseases
45 and allergies in East and West German children during the first 5 years after reunification: time trends and the
46 impact of sulphur dioxide and total suspended particles. Int. J. Epidemiol. 28: 865-873.
47 Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
48 W. H. (2000) Reanallysis of the Harvard Six Cities study and the American Cancer Society study of
49 paniculate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
50 Project. Cambridge, MA: Health Effects Institute.
51 Ktinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared
52 to ecologic studies. Environ. Health Perspect. 105: 1078-1083.
53 Ktinzli, N.; Ackermann-Liebrich, U.; Brandli, O.; Tschopp, J. M.; Schindler, C.; Leuenberger, P.; SAPALDIA
54 Team. (2000) Clinically "small" effects of air pollution on FVC have a large public health impact. Eur.
55 Respir. J. 15: 131-136.
56 Kwon, H.-J.; Cho, S.-H.; Nyberg, F.; Pershagen, G. (2001) Effects of ambient air pollution on daily mortality in a
57 cohort of patients with congestive heart failure. Epidemiology 12: 413-419.
June 2003 8-291 DRAFT-DO NOT QUOTE OR CITE
-------
1 Laden, F.; Neas, L. M; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
2 sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
3 Lave, L. B.; Seskin, E. P. (1977) Air pollution and human health. Baltimore, MD: The Johns Hopkins University
4 Press.
5 Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
6 Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2002) Short term effects of paniculate
7 air pollution on cardiovascular diseases in eight European cities. J. Epidemiol. Community Health
8 56:773-779.
9 Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
10 Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2003) Short-term effects of paniculate
11 air pollution on cardiovascular diseases in eight European cities. In: Revised analyses of time-series studies of
12 air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 173-176. Available:
13 http://www.healtheffects.org/news.htm [16 May, 2003].
14 Lebowitz, M. D. (1971) A stimulus response method for the analysis of environmental and health events related in
15 time. Environ. Lett. 2: 23-24.
16 Lebowitz, M. D.; Collins, L.; Holberg, C. J. (1987) Time series analyses of respiratory responses to indoor and
17 outdoor environmental phenomena. Environ. Res.43:332-341.
18 Lee, J.-T.; Schwartz, J. (1999) Reanalysis of the effects of air pollution on daily mortality in Seoul, Korea:
19 a case-crossover design. Environ. Health Perspect. 107: 633-636.
20 Lee, J.-T.; Shy, C. M. (1999) Respiratory function as measured by peak expiratory flow rate and PM10:
21 six communities study. J. Exposure Anal. Environ. Epidemiol. 9: 293-299.
22 Lee, R. E., Jr.; Caldwell, J. S.; Morgan, G. B. (1972) The evaluation of methods for measuring suspended
23 particulates in air. Atmos. Environ. 6: 593-622.
24 Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
25 Perspect. 107: 149-154.
26 Lee, J.-T.; Kim, H.; Hong, Y.-C.; Kwon, H.-J.; Schwartz, J.; Christiani, D. C. (2000) Air pollution and daily
27 mortality in seven major cities of Korea, 1991-1997. Environ. Res. 84: 247-254.
28 Leonardi, G. S.; Houthuijs, D.; Steerenberg, P. A.; Fletcher, T.; Armstrong, B.; Antova, T. (2000) Immune
29 biomarkers in relation to exposure to paniculate matter: a cross-sectional survey in 17 cities of central
30 Europe. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 1-14.
31 Levy, D. (1998) Fine paniculate air pollution and out-of-hospital mortality in King County, Washington. In: Vostal,
32 J. J., ed. Health effects of paniculate matter in ambient air. Proceedings of an international conference; 1997;
33 Prague, Czech Republic. Pittsburgh, PA: Air & Waste Management Association; pp. 262-271. (A&WMA
34 publication VIP-80).
35 Levy, J. I.; Hammitt, J. K.; Spengler, J. D. (2000) Estimating the mortality impacts of paniculate matter: what can be
36 learned from between-study variability? Environ. Health Perspect. 108: 109-117.
37 Levy, D.; Sheppard, L.; Checkoway, H.; Kaufman, J.; Lumley, T.; Koenig, J.; Siscovick, D. (2001a) A
3 8 case-crossover analysis of paniculate matter air pollution and out-of-hospital primary cardiac arrest.
39 Epidemiology 12: 193-199.
40 Levy, D.; Lumley, T.; Sheppard, L.; Kaufman, J.; Checkoway, H. (200Ib) Referent selection in case-crossover
41 analyses of acute health effects of air pollution. Epidemiology 12: 186-192.
42 Lewis, P. R.; Hensley, M. J.; Wlodarczyk, J.; Toneguzzi, R. C.; Westley-Wise, V. J.; Dunn, T.; Calvert, D. (1998)
43 Outdoor air pollution and children's respiratory symptoms in the steel cities of New South Wales. Med. J.
44 Aust. 169: 459-463.
45 Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
46 pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
47 Lin, C. A.; Martins, M. A.; Farhat, S. C. L.; Pope, C. A., Ill; Conceicao, G. M. S.; Anastacio, V. M.; Hatanaka, M.;
48 Andrade, W. C.; Hamaue, W. R.; Bohm, G. M.; Saldiva, P. H. N. (1999) Air pollution and respiratory illness
49 of children in Sao Paulo, Brazil. Paediatr. Perinat. Epidemiol. 13: 475-488.
50 Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D.; Gong, H., Jr. (1996)
51 Short-term air pollution exposures and responses in Los Angeles area schoolchildren. J. Exposure Anal.
52 Environ. Epidemiol. 6: 449-472.
53 Linn, W. S.; Gong, H., Jr.; Clark, K. W.; Anderson, K. R. (1999) Day-to-day paniculate exposures and health
54 changes in Los Angeles area residents with severe lung disease. J. Air Waste Manage. Assoc.
55 49:PM108-PM115.
56 Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
57 admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
June 2003 8-292 DRAFT-DO NOT QUOTE OR CITE
-------
1 Lipfert, F. W. (2000) The use and misuse of surrogate variables in environmental epidemiology. J. Environ. Med.
2 1:267-278.
3 Lipfert, F. W.; Morris, S. C. (2002) Temporal and spacial relationships between age-specific mortality and ambient
4 air quality in the United States: preliminary results for counties, 1960-97. Occup. Environ. Med. 59: 156-174.
5 Lipfert, F. W.; Wyzga, R. E. (1996) The effects of exposure error on environmental epidemiology. In: Phalen, R. F.;
6 Mannix, R. C.; Tonini, M. C., eds. The second colloquium on paniculate air pollution & human mortality &
7 morbidity: report to the California Air Resources Board; May; Park City, UT. Sacramento, CA: State of
8 California, Air Resources Board; pp. 4-295--4-302; ARE contract no. 95-323. (University of California Air
9 Pollution Health Effects Laboratory report no. 96-02). Available:
10 http://www.arb.ca.gov/research/abstracts/95-323.htm#Ch95-323 (24 September 2002).
11 Lipfert, F. W.; Wyzga, R. E. (1997) Air pollution and mortality: the implications of uncertainties in regression
12 modeling and exposure measurement. J. Air Waste Manage. Assoc. 47: 517-523.
13 Lipfert, F. W.; Wyzga, R. E. (1999) Statistical considerations in determining the health significance of constituents
14 of airborne paniculate matter. J. Air Waste Manage. Assoc. 49: PM-182-191.
15 Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
16 size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
17 Lipfert, F. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) The Washington
18 University-EPRI veterans' cohort mortality study: preliminary results. In: Grant, L. D., ed. PM2000:
19 paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 41-73.
20 Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000c) Infant mortality and air pollution: a comprehensive analysis of U.S.
21 data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
22 Lippmann, M.; Thurston, G. D. (1996) Sulfate concentrations as an indicator of ambient paniculate matter air
23 pollution for health risk evaluations. J. Exposure Anal. Environ. Epidemiol. 6: 123-146.
24 Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
25 mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report no.
26 95.
27 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
28 California. Environ. Health Perspect. 105: 216-222.
29 Long, W.; Tate, R. B.; Neuman, M.; Manfreda, J.; Becker, A. B.; Anthonisen, N. R. (1998) Respiratory symptoms in
30 a susceptible population due to burning of agricultural residue. Chest 113: 351-357.
31 Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
32 mortality in Mexico City. Epidemiology 10: 118-123.
33 Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
34 J. R. Stat. Soc. B 61(part 2): 459-477.
3 5 Lumley, T.; Levy, D. (2000) Bias in the case-crossover design: implications for studies of air pollution.
36 Environmetrics 11: 689-704.
37 Lumley, T.; Sheppard, L. (2000) Assessing seasonal confounding and model selection bias in air pollution
3 8 epidemiology using positive and negative control analyses. Environmetrics 11: 705-717.
39 Magari, S. R.; Hauser, R.; Schwartz, J.; Williams, P. L.; Smith, T. J.; Christiani, D. C. (2001) Association of heart
40 rate variability with occupational and environmental exposure to paniculate air pollution. Circulation
41 104:986-991.
42 Magari, S. R.; Schwartz, J.; Williams, P. L.; Hauser, R.; Smith, T. J.; Christiani, D. C. (2002) The association
43 between personal measurements of environmental exposure to particulates and heart rate variability.
44 Epidemiology 13: 305-310.
45 Mage, D.; Wilson, W.; Hasselblad, V.; Grant, L. (1999) Assessment of human exposure to ambient paniculate
46 matter. J. Air Waste Manage. Assoc. 49: 174-185.
47 Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
48 Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
49 Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003) Air pollution and cardiovascular
5 0 mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
51 Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
52 http://www.healtheffects.org/news.htm [16 May, 2003].
53 Marcus, A. H.; Chapman, R. (1998) Estimating the health effects of fine particles from epidemiology studies: how
54 serious are problems of measurement error, correlation, and confounding? In: Chow, J.; Koutrakis, P., eds.
55 PM2 5: a fine particle standard. Volume II: proceedings of an international specialty conference.; January;
56 Long Beach, CA. Pittsburgh, PA: Air & Waste Management Association; pp. 899-919.
June 2003 8-293 DRAFT-DO NOT QUOTE OR CITE
-------
1 McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. I; Margolis, H. G.;
2 Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
3 California children with asthma. Environ. Health Perspect. 107: 757-760.
4 McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Islam, T.; Gauderman, W. J.; Avol, E.; Margolis, H. G.;
5 Peters, J. M. (2002) Asthma in exercising children exposed to ozone: a cohort study. Lancet 359: 386-391.
6 McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of
7 mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers.
8 J. Exposure Anal. Environ. Epidemiol. 10: 427-436.
9 McGregor, G. R.; Walters, S.; Wordley, J. (1999) Daily hospital respiratory admissions and winter air mass types,
10 Birmingham, UK. Int. J. Biometeorol. 43: 21-30.
11 McMahon, B.; Pugh, T. F. (1970) Epidemiology: principles and methods. Boston, MA: Little, Brown and Company.
12 Medina, S.; Le Tertre, A.; Quenel, P.; Le Moullec, Y.; Lameloise, P.; Guzzo, J. C.; Festy, B.; Ferry, R.; Dab, W.
13 (1997) Air pollution and doctors' house calls: results from the ERPURS system for monitoring the effects of
14 air pollution on public health in greater Paris, France, 1991-1995. Environ. Res. 75: 73-84.
15 Mendelsohn, R.; Orcutt, G. (1979) An empirical analysis of air pollution dose-response curves. J. Environ. Econ.
16 Manage. 6: 85-106.
17 Metzger, K. B.; Tolbert, P. E.; Klein, M.; Peel, J. L.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
18 Frumkin, H. (2003) Ambient air pollution and cardiovascular emergency department visits in Atlanta,
19 Georgia, 1993-2000. Epidemiology: in press.
20 Michelozzi, P.; Forastiere, F.; Fusco, D.; Perucci, C. A.; Ostro, B.; Ancona, C.; Pallotti, G. (1998) Air pollution and
21 daily mortality in Rome, Italy. Occup. Environ. Med. 55: 605-610.
22 Miller, J. P.; Perry, H. M., Jr.; Rossiter, J. E.; Baty, J. D.; Carmody, S. E.; Sambhi, M. P. (1994) Regional
23 differences in mortality during 15-year follow-up of 11,936 hypertensive veterans. Hypertension 23: 431-438.
24 Mills, P. K.; Abbey, D.; Beeson, W. L.; Petersen, F. (1991) Ambient air pollution and cancer in California
25 Seventh-day Adventists. Arch. Environ. Health 46: 271-280.
26 Monn, C. (2001) Exposure assessment of air pollutants: a review on spatial heterogeneity and
27 indoor/outdoor/personal exposure to suspended paniculate matter, nitrogen dioxide and ozone. Atmos.
28 Environ. 35: 1-32.
29 Monn, Ch.; Carabias, V.; Junker, M.; Waeber, R.; Karrer, M.; Wanner, H. U. (1997) Small-scale spatial variability
30 of paniculate matter <10 |im (PM10) and nitrogen dioxide. Atmos. Environ. 31: 2243-2247.
31 Moolgavkar, S. H. (2000a) Air Pollution and Mortality in Three U.S. Counties. Environ. Health Perspect.
32 108:777-784.
33 Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
34 metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
3 5 Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
36 metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
37 Toxicol. 12(suppl. 4): 75-90.
3 8 Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
39 In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
40 Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
41 Moolgavkar, S. H.; Luebeck, E. G. (1996) A critical review of the evidence on paniculate air pollution and
42 mortality. Epidemiology 7: 420-428.
43 Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
44 causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
45 Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and admissions
46 for chronic respiratory disease in King County, Washington. In: Inhalation toxicology: proceedings of the
47 third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC. Inhalation Toxicol.
48 12(suppl. 1): 157-171.
49 Morgan, G.; Corbett, S.; Wlodarczyk, J.; Lewis, P. (1998) Air pollution and daily mortality in Sydney, Australia,
50 1989 through 1993. Am. J. Public Health 88: 759-764.
51 Morris, R. D. (2001) Airborne particulates and hospital admissions for cardiovascular disease: a quantitative review
52 of the evidence. Environ. Health Perspect. 109(suppl. 4): 495-500.
53 Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
54 evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
55 Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
56 heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
June 2003 8-294 DRAFT-DO NOT QUOTE OR CITE
-------
1 Mortimer, K. M; Neas, L. M; Dockery, D. W.; Redline, S.; Tager, I. B. (2002) The effect of air pollution on
2 inner-city children with asthma. Eur. Respir. J. 19: 699-705.
3 Murray, C. J.; Nelson, C. R. (2000) State-space modeling of the relationship between air quality and mortality. J. Air
4 Waste Manage. Assoc. 50: 1075-1080.
5 Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
6 Healthy women's PEF variations with ambient summer concentrations of PM10, PM2 5, SO42", H+, and O3.
7 Am. J. Respir. Crit. Care Med. 160: 117-125.
8 National Research Council. (2001) Research priorities for airborne paniculate matter. III. Early research progress.
9 Washington, DC: National Academy Press. Available: http://www.nap.edu/catalog/10065.html [30 April,
10 2002].
11 Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
12 and exposure to air pollution. Public Health Rep. 114: 135-148.
13 Navidi, W. (1998) Bidirectional case-crossover designs for exposures with time trends. Biometrics 54: 596-605.
14 Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
15 indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
16 139: 1088-1099.
17 Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
18 pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
19 141: 111-122.
20 Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
21 determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
22 Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
23 Philadelphia. Environ. Health Perspect. 107: 629-631.
24 Nemmar, A.; Hoet, P. H. M.; Vanquickenborne, B.; Dinsdale, D.; Thomeer, M.; Hoylaerts, M. F.; Vanbilloen, H.;
25 Mortelmans, L.; Nemery, B. (2002) Passage of inhaled particles into the blood circulation in humans.
26 Circulation 105: 411-414.
27 Neukirch, F.; Segala, C.; Le Moullec, Y.; Korobaeff, M.; Aubier, M. (1998) Short-term effects of low-level winter
28 pollution on respiratory health of asthmatic adults. Arch. Environ. Health 53: 320-328.
29 Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
3 0 between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
31 Perspect. 107: 489-493.
32 Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
33 stagnant air. Thorax 55: 466-470.
34 Nyberg, F.; Gustavsson, P.; Jarup, L.; Bellander, T.; Berglind, N.; Jakobsson, R.; Pershagen, G. (2000) Urban air
3 5 pollution and lung cancer in Stockholm. Epidemiology 11: 487-495.
36 Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
37 70:98-104.
38 Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
39 Am. J. Public Health 81: 694-702.
40 Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
41 exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
42 Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
43 January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
44 Ostro, B.; Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1998) The impact of fine paniculate matter in Bangkok,
45 Thailand. In: Chow, J.; Koutrakis, P., eds. PM2.5: a fine particle standard. Volume II: proceedings of an
46 international specialty conference; January; Long Beach, CA. Pittsburgh, PA: Air & Waste Management
47 Association; pp. 939-949. (A&WMA publication VIP-81).
48 Ostro, B. D.; Hurley, S.; Lipsett, M. J. (1999a) Air pollution and daily mortality in the Coachella Valley, California:
49 a study of PM10 dominated by coarse particles. Environ. Res. 81: 231-238.
50 Ostro, B. D.; Eskeland, G. S.; Sanchez, J. M.; Feyzioglu, T. (1999b) Air pollution and health effects: a study of
51 medical visits among children in Santiago, Chile. Environ. Health Perspect. 107: 69-73.
52 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
53 Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol.: 10: 412-419.
54 Ostro, B.; Lipsett, M.; Mann, J.; Braxtron-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
55 African-American children in Los Angeles. Epidemiology 12: 200-208.
June 2003 8-295 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coachella Valley,
2 California. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
3 Health Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
4 Osunsanya, T.; Prescott, G.; Seaton, A. (2001) Acute respiratory effects of particles: mass or number? Occup.
5 Environ. Med. 58: 154-159.
6 Ott, W.; Wallace, L.; Mage, D. (2000) Predicting paniculate (PM10) personal exposure distributions using a random
7 component superposition statistical model. J. Air Waste Manage. Assoc. 50: 1390-1406.
8 Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
9 pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
10 Environmental Health; March 25.
11 Pantazopoulou, A.; Katsouyanni, K.; Kourea-Kremastinou, J.; Trichopoulos, D. (1995) Short-term effects of air
12 pollution on hospital emergency outpatient visits and admissions in the greater Athens, Greece area. Environ.
13 Res. 69:31-36.
14 Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
15 in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
16 Pekkanen, J.; Brunner, E. J.; Anderson, H. R.; Tiittanen, P.; Atkinson, R. W. (2000) Daily concentrations of air
17 pollution and plasma fibrinogen in London. Occup. Environ. Med. 57: 818-822.
18 Penttinen, P.; Timonen, K. L.; Tiittanen, P.; Mirme, A.; Ruuskanen, J.; Pekkanen, J. (2001) Number concentration
19 and size of particles in urban air: effects on spirometric lung function in adult asthmatic subjects. Environ.
20 HealthPerspect. 109: 319-323.
21 Pereira, L. A. A.; Loomis, D.; Conceigao, G. M. S.; Braga, A. L. F.; Areas, R. M.; Kishi, H. S.; Singer, J. M.; Bohm,
22 G. M.; Saldiva, P. H. N. (1998) Association between air pollution and intrauterine mortality in Sao Paulo,
23 Brazil. Environ. Health Perspect. 106: 325-329.
24 Perry, H. M., Jr.; Schnaper, H. W.; Meyer, G.; Swatzell, R. (1982) Clinical program for screening and treatment of
25 hypertension in veterans. J. Natl. Med. Assoc. 74: 433-444.
26 Peters, A.; Goldstein, I. F.; Beyer, U.; Franke, K.; Heinrich, J.; Dockery, D. W.; Spengler, J. D.; Wichmann, H.-E.
27 (1996) Acute health effects of exposure to high levels of air pollution in eastern Europe. Am. J. Epidemiol.
28 144:570-581.
29 Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
30 episode: a link to mortality? Lancet 349: 1582-1587.
31 Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997b) Short-term effects of paniculate air pollution on
32 respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
33 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997c) Respiratory effects are associated with the
34 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
35 Peters, A.; Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Heinrich, J. (1999a) Akute Auswirkung der
3 6 Schwebstaubkonzentrationen in der Aulknluft auf die Mortalitat - Vergleichsstudie Nordost-Bayern /
37 Nordbohmen. AbschluBbericht [Acute effects of suspended particle concentrations in the atmosphere on
3 8 mortality - a study comparing northeast Bavaria and north Bohemia. Final report]. Bavaria, Federal Republic
39 of Germany: Institut fur Epidemiologie; report no. GSF-EP S 1/99.
40 Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
41 Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
42 differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
43 Med. 159: 760-767.
44 Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
45 Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999c) A study of twelve southern California communities with
46 differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
47 159:768-775.
48 Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
49 Monahan, K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology
50 11: 11-17.
51 Peters, A.; Skorkovsky, J.; Kotesovec, F.; Brynda, J.; Spix, C.; Wichmann, H. E.; Heinrich, J. (2000b) Associations
52 between mortality and air pollution in central Europe. Environ. Health Perspect. 108: 283-287.
53 Peters, A.; Dockery, D. W.; Muller, J. E.; Mittleman, M. A. (2001a) Increased paniculate air pollution and the
54 triggering of myocardial infarction. Circulation 103: 2810-2815.
55 Peters, A.; Frohlich, M.; Doring, A.; Immervoll, T.; Wichmann, H.-E.; Hutchinson, W. L.; Pepys, M. B.; Koenig, W.
56 (200 Ib) Paniculate air pollution is associated with an acute phase response in men: results from the
57 MONICA-Augsburg Study. Eur. Heart J. 22: 1198-1204.
June 2003 8-296 DRAFT-DO NOT QUOTE OR CITE
-------
1 Petroeschevsky, A.; Simpson, R. W.; Thalib, L.; Rutherford, S. (2001) Associations between outdoor air pollution
2 and hospital admissions in Brisbane, Australia. Arch. Environ. Health 56: 37-52.
3 Pike, M. C.; Jing, J. S.; Rosario, I. P.; Henderson, B. E.; Menck, H. R. (1979) Occupation: "explanation" of an
4 apparent air pollution related localized excess of lung cancer in Los Angeles County. In: Breslow, N. E.;
5 Whitemore, A. S., eds. Energy and Health: proceedings of a conference sponsored by Siam Institute for
6 Mathematics and Society; 1978; Alta, Utah. [SIAM-SIMS conference series 6].
7 Pless-Mulloli, T.; Howel, D.; King, A.; Stone, L; Merefield, J.; Bessell, J.; Darnell, R. (2000) Living near opencast
8 coal mining sites and children's respiratory health. Occup. Environ. Med. 57: 145-151.
9 Ponce de Leon, A.; Anderson, H. R.; Bland, J. M.; Strachan, D. P.; Bower, J. (1996) Effects of air pollution on daily
10 hospital admissions for respiratory disease in London between 1987-88 and 1991-92. In: St Leger, S., ed. The
11 APHEA project. Short term effects of air pollution on health: a European approach using epidemiological
12 time series data. J. Epidemiol. Community Health 50(suppl. 1): S63-S70.
13 Ponka, A.; Savela, M.; Virtanen, M. (1998) Mortality and air pollution in Helsinki. Arch. Environ. Health
14 53:281-286.
15 Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
16 Am. J. Public Health 79: 623-628.
17 Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and
18 Cache Valleys. Arch. Environ. Health 46: 90-97.
19 Pope, C. A., Ill; Kalkstein, L S. (1996) Synoptic weather modeling and estimates of the exposure-response
20 relationship between daily mortality and paniculate air pollution. Environ. Health Perspect. 104: 414-420.
21 Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM10 pollution:
22 a daily time series analysis. Am. Rev. Respir. Dis. 144: 668-674.
23 Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah valley.
24 Arch. Environ. Health 47: 211-217.
25 Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
26 (1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J. Respir.
27 Crit. Care Med. 151: 669-674.
28 Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
29 Front. Environ. Health Perspect. 107: 567-573.
30 Pope, C. A., Ill; Dockery, D. W.; Kanner, R. E.; Vollegas, G. M.; Schwartz, J. (1999b) Oxygen saturation, pulse
31 rate, and paniculate air pollution: a daily time-series panel study. Am. J. Respir. Crit. Care Med.
32 159: 365-372.
33 Pope, C. A., Ill; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
34 G. M.; Gold, D. R.; Dockery, D. W. (1999c) Heart rate variability associated with paniculate air pollution.
35 Am. Heart J. 138: 890-899.
36 Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
37 cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
38 Assoc. 287: 1132-1141.
39 Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
40 cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
41 Prescott, E.; Lange, P.; Vestbo, J.; Copenhagen City Heart Study Group. (1999) Socioeconomic status, lung function
42 and admission to hospital for COPD: results from the Copenhagen City Heart Study. Eur. Respir. J.
43 13: 1109-1114.
44 Qian, Z. Chapman, R. S.; Tian, Q.; Chen, Y.; Lioy, P. J.; Zhang, J. (2000) Effects of air pollution on children's
45 respiratory health in three Chinese cities. Arch. Environ. Health 55: 126-133.
46 Rahlenbeck, S. I.; Kahl, H. (1996) Air pollution and mortality in East Berlin during the winters of 1981-1989. Int. J.
47 Epidemiol. 25: 1220-1226.
48 Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
49 F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
50 Perspect. 104: 506-514.
51 Ransom, M. R.; Pope, C. A., III. (1992) Elementary school absences and PM10 pollution in Utah Valley. Environ.
52 Res. 58: 204-219.
53 Roemer, W. H.; Van Wijnen, J. H . (2001) Daily mortality and air pollution along busy streets in Amsterdam,
54 1987-1998. Epidemiology 12: 649-653.
55 Roemer, W.; Hoek, G.; Brunekreef, B. (1993) Effect of ambient winter air pollution on respiratory health of children
56 with chronic respiratory symptoms. Am. Rev. Respir. Dis. 147: 118-124.
June 2003 8-297 DRAFT-DO NOT QUOTE OR CITE
-------
1 Roemer, W.; Hoek, G.; Brunekreef, B.; Haluszka, J.; Kalandidi, A.; Pekkanen, J. (1998) Daily variations in air
2 pollution and respiratory health in a multicentre study: the PEACE project. Eur. Respir. J. 12: 1354-1361.
3 Roemer, W.; Hoek, G.; Brunekreef, B.; Clench-Aas, J.; Forsberg, B.; Pekkanen, J.; Schutz, A. (2000) PM10
4 elemental composition and acute respiratory health effects in European children (PEACE project). Eur.
5 Respir. J. 15: 553-559.
6 Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
7 on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
8 154:300-307.
9 Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
10 intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
11 Mexico City. Arch. Environ. Health 52: 368-376.
12 Rooney, C.; McMichael, A. J.; Kovats, R. S.; Coleman, M. P. (1998) Excess mortality in England and Wales, and in
13 greater London, during the 1995 heatwave. J. Epidemiol. Community Health 52: 482-486.
14 Rosas, I.; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, J.; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis
15 of the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
16 emergency admissions to a hospital in Mexico City. Allergy (Copenhagen) 53: 394-401.
17 Rossi, G.; Vigotti, M. A.; Zanobetti, A.; Repetto, F.; Gianelle, V.; and Schwartz, J. (1999) Air pollution and
18 cause-specific mortality in Milan, Italy, 1980-1989. Arch. Environ. Health 54: 158-164.
19 Rothman, K. J.; Greenland, S., eds. (1998) Modern epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven
20 Publishers.
21 Rutherford, S.; Clark, E.; McTainsh, G.; Simpson, R.; Mitchell, C. (1999) Characteristics of rural dust events shown
22 to impact on asthma severity in Brisbane, Australia. Int. J. Biometeorol. 42: 217-225.
23 Samet, J. M. (2000) What properties of paniculate matter are responsible for health effects? In: Inhalation
24 Toxicology: proceedings of the third colloquium on paniculate air pollution and human health; June, 1999;
25 Durham, NC. Inhalation Toxicol. 12(suppl. 1): 19-21.
26 Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1996) Weather, air pollution and mortality in
27 Philadelphia, 1973-1980, report to the Health Effects Institute on phase IB, Particle Epidemiology Evaluation
28 Project. Cambridge, MA: Health Effects Institute; review draft.
29 Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
30 (2000a) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
31 pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
32 Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000b) National morbidity, mortality, and
33 air pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute;
34 research report no. 94.
35 Samet, J. M.; Dominici, F.; Curriero, F. C.; Coursac, L; Zeger, S. L. (2000c) Fine paniculate air pollution and
36 mortality in 20 U.S. cities, 1987-1994. N. Engl. J. Med. 343: 1742-1749.
37 Samoli, E.; Schwartz, J.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Balducci, F.; Medina, S.; Rossi, G.; Sunyer, J.;
38 Bacharova, L.; Anderson, H. R.; Katsouyanni, K. (2001) Investigating regional differences in short-term
3 9 effects of air pollution on daily mortality in the APHEA project: a sensitivity analysis for controlling
40 long-term trends and seasonality. Environ. Health Perspect. 109: 349-353.
41 Samoli, E.; Schwartz, J.; Analitis, A.; Petasakis, Y.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Balducci, F.;
42 Medina, S.; Rossi, G.; Sunyer, J.; Anderson, H. R.; Katsouyanni, K. (2003) Sensitivity analyses of regional
43 differences in short-term effects of air pollution on daily mortality in APHEA cities. In: Revised analyses of
44 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
45 pp. 205-210. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
46 Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
47 exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
48 Sarnat, J. A.; Schwartz, J.; Catalano, P. J.; Suh, H. H. (2001) Gaseous pollutants in paniculate matter epidemiology:
49 confounders or surrogates? Environ. Health Perspect. 109: 1053-1061.
50 Scarlett, J. F.; Abbott, K. J.; Peacock, J. L.; Strachan, D. P.; Anderson, H. R. (1996) Acute effects of summer air
51 pollution on respiratory function in primary school children in southern England. Thorax 51:1109-1114.
52 Schouten, J. P.; Vonk, J. M.; de Graaf, A. (1996) Short term effects of air pollution on emergency hospital
53 admissions for respiratory disease: results of the APHEA project in two major cities in The Netherlands,
54 1977-89. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on health: a European
55 approach using epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S22-S29.
56 Schwartz, J. (1991) Paniculate air pollution and daily mortality in Detroit. Environ. Res. 56: 204-213.
June 2003 8-298 DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137: 1136-1147.
2 Schwartz, J. (1994) PM10, ozone, and hospital admissions for the elderly in Minneapolis, MN. Arch. Environ. Health
3 49: 366-374.
4 Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
5 8:371-377.
6 Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
7 10: 17-22.
8 Schwartz, J. (2000a) Assessing confounding, effect modification, and thresholds in the association between ambient
9 particles and daily deaths. Environ. Health Perspect. 108: 563-568.
10 Schwartz, J. (2000b) The distributed lag between air pollution and daily deaths. Epidemiology 11: 320-326.
11 Schwartz, J. (2000c) Harvesting and long term exposure effects in the relation between air pollution and mortality.
12 Am. J. Epidemiol. 151: 440-448.
13 Schwartz, J. (2000d) Daily deaths are associated with combustion particles rather than SO2 in Philadelphia.
14 Occup. Environ. Med. 57: 692-697.
15 Schwartz, J. (2001) Air pollution and blood markers of cardiovascular risk. Environ. Health Perspect.
16 109(suppl. 3): 405-409.
17 Schwartz, J. (2003a) Daily deaths associated with air pollution in six US cities and short-term mortality
18 displacement in Boston. In: Revised analyses of time-series studies of air pollution and health. Special report.
19 Boston, MA: Health Effects Institute; pp. 219-226. Available: http://www.healtheffects.org/news.htm
20 [16 May, 2003].
21 Schwartz, J. (2003b) Airborne particles and daily deaths in 10 US cities. In: Revised analyses of time-series studies
22 of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 211-218. Available:
23 http://www.healtheffects.org/news.htm [16 May, 2003].
24 Schwartz, J.; Marcus, A. (1990) Mortality and air pollution in London: a time series analysis. Am. J. Epidemiol.
25 131: 185-194.
26 Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
27 Michigan. Am. J. Epidemiol. 142: 23-35.
28 Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
29 respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
30 Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
31 with application to air pollution and daily death. Epidemiology 11: 666-672.
32 Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
33 Ferris, B. G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
34 Am. J. Respir. Crit. Care Med. 150: 1234-1242.
35 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles?
36 J. Air Waste Manage. Assoc. 46: 927-939.
37 Schwartz, J.; Spix, C.; Touloumi, G.; Bacharova, L.; Barumamdzadeh, T.; le Tertre, A.; Piekarksi, T.; Ponce de
38 Leon, A.; Ponka, A.; Rossi, G.; Saez, M.; Schouten, J. P. (1996b) Methodological issues in studies of air
39 pollution and daily counts of deaths or hospital admissions. In: St Leger, S., ed. The APHEA project. Short
40 term effects of air pollution on health: a European approach using epidemiological time series data.
41 J. Epidemiol. Community Health 50(suppl. 1): S3-S11.
42 Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
43 concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
44 Seaton, A.; Soutar, A.; Crawford, V.; Elton, R.; McNerlan, S.; Cherrie, J.; Watt, M.; Agius, R.; Stout, R. (1999)
45 Paniculate air pollution and the blood. Thorax 54: 1027-1032.
46 Segala, C.; Fauroux, B.; Just, J.; Pascual, L.; Grimfeld, A.; Neukirch, F. (1998) Short-term effect of winter air
47 pollution on respiratory health of asthmatic children in Paris. Eur. Respir. J. 11: 677-685.
48 Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
49 1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
50 MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
51 Sheppard, L.; Damian, D. (2000) Estimating short-term PM effects accounting for surrogate exposure measurements
52 from ambient monitors. Environmetrics 11: 675-687.
53 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
54 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
55 Sheppard, L.; Levy, D.; Checkoway, H. (2001) Correcting for the effects of location and atmospheric conditions on
56 air pollution exposures in a case-crossover study. J. Exposure Anal. Environ. Epidemiol. 11: 86-96.
June 2003 8-299 DRAFT-DO NOT QUOTE OR CITE
-------
1 Shumway, R. H.; Tai, R. Y.; Tai, L. P.; Pawitan, Y. (1983) Statistical analysis of daily London mortality and
2 associated weather and pollution effects. Sacramento, CA: California Air Resources Board; contract no.
3 Al-154-33.
4 Shumway, R. H.; Azari, A. S.; Pawitan, Y. (1988) Modeling mortality fluctuations in Los Angeles as functions of
5 pollution and weather effects. Environ. Res. 45: 224-241.
6 Simpson, R. W.; Williams, G.; Petroeschevsky, A.; Morgan, G.; Rutherford, S. (1997) Associations between outdoor
7 air pollution and daily mortality in Brisbane, Australia. Arch. Environ. Health 52: 442-454.
8 Smith, M. A.; Jalaludin, B.; Byles, J. E.; Lim, L.; Leeder, S. R. (1996) Asthma presentations to emergency
9 departments in western Sydney during the January 2194 bushfires. Int. J. Epidemiol. 25: 1227-1236.
10 Smith, R. L.; Davis, J. M.; Speckman, P. (1999) Assessing the human health risk of atmospheric particles. In:
11 Environmental Statistics: Analysing Data for Environmental Policy: proceedings of a Novartis Foundation
12 Symposium; May 1998; London, England. New York, NY: Wiley. [Novartis Foundation Symposium 220,
13 pp. 59-79].
14 Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
15 coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
16 Soukup, J. M.; Ohio, A. J.; Becker, S. (2000) Soluble components of Utah Valley paniculate pollution alter alveolar
17 macrophage function in vivo and in vitro. Inhalation Toxicol. 12:401-414.
18 Spix, C.; Heinrich, J.; Dockery, D.; Schwartz, J.; Volksch, G.; Schwinkowski, K.; Collen, C.; Wichmann, H. E.
19 (1993) Air pollution and daily mortality in Erfurt, East Germany, 1980-1989. Environ. Health Perspect.
20 101:518-526.
21 Spix, C.; Anderson, H. R.; Schwartz, J.; Vigotti, M. A.; LeTertre, A.; Vonk, J. M.; Touloumi, G.; Balducci, F.;
22 Piekarski, T.; Bacharova, L.; Tobias, A.; Ponka, A.; Katsouyanni, K. (1998) Short-term effects of air
23 pollution on hospital admissions of respiratory diseases in Europe: a quantitative summary of APHEA study
24 results. Arch. Environ. Health 53: 54-64.
25 Stieb, D. M.; Burnett, R. T.; Beveridge, R. C.; Brook, J. R. (1996) Association between ozone and asthma
26 emergency department visits in Saint John, New Brunswick, Canada. Environ. Health Perspect.
27 104: 1354-1360.
28 Stieb, D. M.; Beveridge, R. C.; Rowe, B. H.; Walter, S. D.; Judek, S. (1998a) Assessing diagnostic classification in
29 an emergency department: implications for daily time series studies of air pollution. Am. J. Epidemiol.
30 148:666-670.
31 Stieb, D. M.; Brook, J. R.; Broder, I.; Judek, S.; Burnett, R. T.; Beveridge, R. C. (1998b) Personal exposure of adults
32 with cardiorespiratory disease to paniculate acid and sulfate in Saint John, New Brunswick, Canada. Appl.
33 Occup. Environ. Hyg. 13: 461-468.
34 Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
3 5 Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
36 John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
37 Stolzel, M.; Peters, A.; Wichmann, H.-E. (2003) Daily mortality and fine and ultrafine particles in Erfurt, Germany.
3 8 In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
39 Effects Institute; pp. 231-240. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
40 Sunyer, J.; Basagana, X. (2001) Particles, and not gases, are associated with the risk of death in patients with chronic
41 obstructive pulmonary disease. Int. J. Epidemiol. 30: 1138-1140.
42 Sunyer, J.; Spix, C.; Quenel, P.; Ponce-de-Leon, A.; Ponka, A.; Barumandzadeh, T.; Touloumi, G.; Bacharova, L.;
43 Wojtyniak, B.; Vonk, J.; Bisanti, L.; Schwartz, J.; Katsouyanni, K. (1997) Urban air pollution and emergency
44 admissions for asthma in four European cities: the APHEA project. Thorax 52: 760-765.
45 Sunyer, J.; Schwartz, J.; Tobias, A.; Macfarlane, D.; Garcia, J.; Anto, J. M. (2000) Patients with chronic obstructive
46 pulmonary disease are at increased risk of death associated with urban particle air pollution: a case-crossover
47 analysis. Am. J. Epidemiol. 151: 50-56.
48 Taggart, S. C. O.; Custovic, A.; Francis, H. C.; Faragher, E. B.; Yates, C. J.; Higgins, B. G.; Woodcock, A. (1996)
49 Asthmatic bronchial hyperresponsiveness varies with ambient levels of summertime air pollution. Eur.
50 Respir.J. 9: 1146-1154.
51 Tarn, A. Y. C.; Wong, C. M.; Lam, T. H.; Ong, S. G.; Peters, J.; Hedley, A. J. (1994) Bronchial responsiveness in
52 children exposed to atmospheric pollution in Hong Kong. Chest 106: 1056-1060.
53 Tan, W. C.; Qiu, D.; Liam, B. L.; Ng, T. P.; Lee, S. H.; VanEeden, S. F.; D'Yachkova, Y.; Hogg, J. C. (2000)
54 The human bone marrow response to acute air pollution caused by forest fires. Am. J. Respir. Crit. Care Med.
55 161: 1213-1217.
56 Tanaka, H.; Honma, S.; Nishi, M.; Igarashi, T.; Teramoto, S.; Nishio, F.; Abe, S. (1998) Acid fog and hospital visits
57 for asthma: an epidemiological study. Eur. Respir. J. 11: 1301-1306.
June 2003 8-300 DRAFT-DO NOT QUOTE OR CITE
-------
1 Tellez-Rojo, M. M; Romieu, I.; Ruiz-Velasco, S.; Lezana, M.-A.; Hernandez-Avila, M. M. (2000) Daily respiratory
2 mortality and PM10 pollution in Mexico City: importance of considering place of death. Eur. Respir. J.
3 16:391-396.
4 Tenias, J. M.; Ballester, F.; Rivera, M. L. (1998) Association between hospital emergency visits for asthma and air
5 pollution in Valencia, Spain. Occup. Environ. Med. 55: 541-547.
6 Thompson, A. J.; Shields, M. D.; Patterson, C. C. (2001) Acute asthma exacerbations and air pollutants in children
7 living in Belfast, Northern Ireland. Arch. Environ. Health 56: 234-241.
8 Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
9 summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
10 65:271-290.
11 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
12 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
13 Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
14 resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
15 Timonen, K. L.; Pekkanen, J. (1997) Air pollution and respiratory health among children with asthmatic or cough
16 symptoms. Am. J. Respir. Crit. Care Med. 156: 546-552.
17 Tobias, A.; Campbell, M. J. (1999) Modelling influenza epidemics in the relation between black smoke and total
18 mortality. A sensitivity analysis. J. Epidemiol. Community Health 53: 583-584.
19 Tolbert, P. E.; Klein, M.; Metzger, K. B.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.; Frumkin, H.
20 (2000a) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure Anal.
21 Environ. Epidemiol. 10: 446-460.
22 Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
23 Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000b) Air quality and
24 pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
25 Touloumi, G.; Katsouyanni, K.; Zmirou, D.; Schwartz, J.; Spix, C.; Ponce de Leon, A.; Tobias, A.; Quennel, P.;
26 Rabczenko, D.; Bacharova, L.; Bisanti, L.; Vonk, J. M.; Ponka, A. (1997) Short-term effects of ambient
27 oxidant exposure on mortality: a combined analysis within the APHEA project. Am. J. Epidemiol.
28 146: 177-185.
29 Tsai, F. C.; Apte, M. G.; Daisey, J. M. (1999) An exploratory analysis of the relationship between mortality and the
30 chemical composition of airborne paniculate matter. Berkeley, CA: Lawrence Berkeley National Laboratory,
31 Environmental Energy Technologies Division; report no. LBNL-43583.
32 Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
33 chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
34 Turnovska, T.; Kostianev, S. (1999) Effects of reduced air pollution on children's pulmonary function. Cent. Eur. J.
35 Public Health 7: 77-79.
36 U.S. Census Bureau. (1995) American housing survey for the United States in 1993. Washington, DC: U.S.
37 Department of Commerce; current housing reports H150/93. Available:
38 http://www.census.gov/hhes/www/housing/ans/nationaldata.html [29 April, 2002].
39 U.S. Environmental Protection Agency. (1982) Air quality criteria for particulate matter and sulfur oxides. Research
40 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
41 Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
42 U.S. Environmental Protection Agency. (1986) Second addendum to air quality criteria for particulate matter and
43 sulfur oxides (1982): assessment of newly available health effects information. Research Triangle Park, NC:
44 Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report
45 no. EPA-600/8-86-020F. Available from: NTIS, Springfield, VA; PB87-176574.
46 U.S. Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park,
47 NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
48 nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525, and
49 PB95-124517.
50 U.S. Environmental Protection Agency. (1996a) Air quality criteria for particulate matter. Research Triangle Park,
51 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
52 U.S. Environmental Protection Agency. (1996b) Air quality criteria for ozone and related photochemical oxidants.
53 Research Triangle Park, NC: Office of Research and Development; report nos. EP A/600/AP-93/004aF-cF. 3v.
54 Available from: NTIS, Springfield, VA; PB96-185582, PB96-185590, and PB96-185608. Available online at:
55 www.epa.gov/ncea/ozone.htm.
56 U.S. Environmental Protection Agency. (2000a) Emissions and air quality data. Washington, DC: Technology
57 Transfer Network. Available: http://www.epa.gov/ttn/naaqs/ozone/areas/ [20 May, 2003].
June 2003 8-301 DRAFT-DO NOT QUOTE OR CITE
-------
1 U.S. Environmental Protection Agency. (2000b) Air quality criteria for carbon monoxide. Research Triangle Park,
2 NC: National Center for Environmental Assessment; report no. EPA/600/P-99/00 IF. Available:
3 www.epa.gov/ncea/co/ [2000, October 6].
4 U.S. Environmental Protection Agency. (2002) Health assessment document for diesel engine exhaust. Washington,
5 DC: Office of Research and Development, National Center for Environmental Assessment; report no.
6 EPA/600/8-90/057F. Available: http://cfpub.epa.gov/ncea/ [22 May, 2003].
7 Van Der Zee, S. C.; Hoek, G.; Boezen, H. M; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (1999) Acute
8 effects of urban air pollution on respiratory health of children with and without chronic respiratory symptoms.
9 Occup. Environ. Med. 56: 802-813.
10 Van Der Zee, S. C.; Hoek, G.; Boezen, M. H.; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (2000) Acute
11 effects of air pollution on respiratory health of 50-70 yr old adults. Eur. Respir. J. 15: 700-709.
12 Vasan, R. S.; Larson, M. G.; Leip, E. P.; Kannel, W. B.; Levy, D. (2001) Assessment of frequency of progression to
13 hypertension in non-hypertensive participants in the Framingham Heart Study: a cohort study. Lancet
14 358: 1682-1686.
15 Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
16 nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
17 Vena, J. E. (1982) Air pollution as a risk factor in lung cancer. Am. J. Epidemiol. 116: 42-56.
18 Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1967) Effects of treatment on
19 morbidity in hypertension. Results in patients with diastolic blood pressures averaging 115 through 129 mm
20 Hg. JAMA J. Am. Med. Assoc. 202: 1028-1034.
21 Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1970) Effects of treatment on
22 morbidity in hypertension. II. Results in patients with diastolic blood pressure averaging 90 through 114 mm
23 Hg. JAMA J. Am. Med. Assoc. 213: 1143-1152.
24 Vigotti, M. A.; Rossi, G.; Bisanti, L.; Zanobetti, A.; Schwartz, J. (1996) Short term effects of urban air pollution on
25 respiratory health in Milan, Italy, 1980-89. In: Leger, S., ed. The APHEA project. Short term effects of air
26 pollution on health: a European approach using epidemiological time series data. J. Epidemiol. Community
27 Health 50(suppl. 1): S71-S75.
28 Villeneuve, P. J.; Goldberg, M. S.; Krewski, D.; Burnett, R. T.; Chen, Y. (2002) Fine paniculate air pollution and
29 all-cause mortality within the Harvard six-cities study: variations in risk by period of exposure. Ann.
30 Epidemiol. 12: 568-576.
31 VonKlot, S.; Wolke, G.; Tuch, T.; Heinrich, J.; Dockery, D. W.; Schwartz, J.; Kreyling, W. G.; Wichmann, H. E.;
32 Peters, A. (2002) Increased asthma medication use in association with ambient fine and ultrafine particles.
33 Eur. Respir. J. 20: 691-702.
34 Wallace, L. (1996) Indoor particles: a review. J. Air Waste Manage. Assoc. 46: 98-126.
35 Wallace, L. (2000) Real-time monitoring of particles, PAH, and CO in an occupied townhouse. Appl. Occup.
36 Environ. Hyg. 15: 39-47.
37 Wang, B.; Peng, Z.; Zhang, X.; Xu, Y.; Wang, H.; Allen, G.; Wang, L.; Xu, X. (1999) Paniculate matter, sulfur
3 8 dioxide, and pulmonary function in never-smoking adults in Chongqing, China. Int. J. Occup. Environ. Health
39 5: 14-19.
40 Ward, D. J.; Miller, M. R.; Walters, S.; Harrison, R. M.; Ayres, J. G. (2000) Impact of correcting peak flow for
41 nonlinear errors on air pollutant effect estimates from a panel study. Eur. Respir. J. 15: 137-140.
42 Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
43 ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
44 Respir. Dis. 133: 834-842.
45 Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
46 mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
47 Cambridge, MA: Health Effects Institute; Research Report no. 98.
48 Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
49 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
50 Wilson, W. E.; Mage, D. T.; Grant, L. D. (2000) Estimating separately personal exposure to ambient and
51 nonambient paniculate matter for epidemiology and risk assessment: why and how. J. Air Waste Manage.
52 Assoc. 50: 1167-1183.
53 Wjst, M.; Reitmeir, P.; Dold, S.; Wulff, A.; Nicolai, T.; VonLoeffelholz-Colberg, E. F.; VonMutius, E. (1993)
54 Road traffic and adverse effects on respiratory health in children. Br. Med. J. 307: 596-600.
55 Wolff, G. T.; Stroup, C. M.; Stroup, D. P. (1983) The coefficient of haze as a measure of paniculate elemental
56 carbon. J. AirPollut. Control Assoc. 33: 746-750.
June 2003 8-302 DRAFT-DO NOT QUOTE OR CITE
-------
1 Wong, C. M; Lam, T. H.; Peters, I; Hedley, A. I; Ong, S. G.; Tarn, A. Y. C; Liu, I; Spiegelhalter, D. J. (1998)
2 Comparison between two districts of the effects of an air pollution intervention on bronchial responsiveness
3 in primary school children in Hong Kong. J. Epidemiol. Community Health 52: 571-578.
4 Wong, T. W.; Lau, T. S.; Yu, T. S.; Neller, A.; Wong, S. L.; Tarn, W.; Pang, S. W. (1999a) Air pollution and
5 hospital admissions for respiratory and cardiovascular diseases in Hong Kong. Occup. Environ. Med.
6 56:679-683.
7 Wong, C. M.; Hu, Z. G.; Lam, T. H.; Hedley, A. J.; Peters, J. (1999b) Effects of ambient air pollution and
8 environmental tobacco smoke on respiratory health of non-smoking women in Hong Kong. Int. J. Epidemiol.
9 28: 859-864.
10 Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal infant
11 mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
12 Wordley, J.; Walters, S.; Ayres, J. G. (1997) Short term variations in hospital admissions and mortality and
13 paniculate air pollution. Occup. Environ. Med. 54: 108-116.
14 Wu, W.; Samet, J. M.; Ohio, A. J.; Devlin, R. B. (2001) Activation of the EOF receptor signaling pathway in human
15 airway epithelial cells exposed to Utah Valley PM. Am. J. Physiol. 281: L483 -L489.
16 Xu, Z.; Yu, D.; Jing, L.; Xu, X. (2000) Air pollution and daily mortality in Shenyang, China. Arch. Environ. Health
17 55: 115-120.
18 Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
19 Inhalation Toxicol. 9: 15-29.
20 Ye, F.; Piver, W. T.; Ando, M.; Portier, C. J. (2001) Effects of temperature and air pollutants on cardiovascular and
21 respiratory diseases for males and females older than 65 years of age in Tokyo, July and August 1980-1995.
22 Environ. Health Perspect. 109: 355-359.
23 Yu, O.; Sheppard, L.; Lumley, T.; Koenig, J. Q.; Shapiro, G. G. (2000) Effects of ambient air pollution on symptoms
24 of asthma in Seattle-area children enrolled in the CAMP study. Environ. Health Perspect. 108: 1209-1214.
25 Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
26 J. Occup. Environ. Med. 42: 469-474.
27 Zanobetti, A.; Schwartz, J. (2001) Are diabetics more susceptible to the health effects of airborne particles? Am. J.
28 Respir. Crit. Care Med. 164: 831-833.
29 Zanobetti, A.; Schwartz, J. (2003a) Multicity assessment of mortality displacement within the APHEA2 project. In:
3 0 Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
31 Institute; pp. 249-254. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
32 Zanobetti, A.; Schwartz, J. (2003b) Airborne particles and hospital admissions for heart and lung disease. In:
3 3 Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
34 Institute; pp. 241-248. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
35 Zanobetti, A.; Schwartz, J.; Dockery, D. W. (2000a) Airborne particles are a risk factor for hospital admissions for
36 heart and lung disease. Environ. Health Perspect. 108: 1071-1077.
37 Zanobetti, A.; Wand, M. P.; Schwartz, J.; Ryan, L. M. (2000b) Generalized additive distributed lag models:
38 quantifying mortality displacement. Biostatistics 1: 279-292.
39 Zeger, S. L.; Dominici, F.; Samet, J. (1999) Harvesting-resistant estimates of air pollution effects on mortality.
40 Epidemiology 10: 171-175.
41 Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
42 measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
43 Perspect. 108: 419-426.
44 Zeghnoun, A.; Beaudeau, P.; Carrat, F.; Delmas, V.; Boudhabhay, O.; Gayon, F.; Guincetre, D.; Czernichow, P.
45 (1999) Air pollution and respiratory drug sales in the city of Le Havre, France, 1993-1996. Environ. Res.
46 81:224-230.
47 Zeghnoun, A.; Czernichow, P.; Beaudeau, P.; Hautemaniere, A.; Froment, L.; Le Tertre, A.; Quenel, P. (2001)
48 Short-term effects of air pollution on mortality in the cities of Rouen and Le Havre, France, 1990-1995. Arch.
49 Environ. Health 56: 327-335.
50 Zemp, E.; Elsasser, S.; Schindler, C.; Ktinzli, N.; Perruchoud, A. P.; Domenighetti, G.; Medici, T.;
51 Ackermann-Liebrich, U.; Leuenberger, P.; Monn, C.; Bolognini, G.; Bongard, J.-P.; Brandli, O.; Karrer, W.;
52 Keller, R.; Schoni, M. H.; Tschopp, J.-M.; Villiger, B.; Zellweger, J.-P.; SAPALDIA Team. (1999)
53 Long-term ambient air pollution and respiratory symptoms in adults (SAPALDIA study). Am. J. Respir. Crit.
54 Care Med. 159: 1257-1266.
55 Zhang, J.; Qian, Z.; Kong, L.; Zhou, L.; Yan, L.; Chapman, R. S. (1999) Effects of air pollution on respiratory health
56 of adults in three Chinese cities. Arch. Environ. Health 54: 373-381.
June 2003 8-303 DRAFT-DO NOT QUOTE OR CITE
-------
1 Zhang, H.; Triche, E.; Leaderer, B. (2000) Model for the analysis of binary time series of respiratory symptoms.
2 Am. J. Epidemiol. 151: 1206-1215.
3 Zidek, J. V.; Wong, H.; Le, N. D.; Burnett, R. (1996) Causality, measurement error and multicollinearity in
4 epidemiology. Environmetrics 7: 441-451.
5 Zmirou, D.; Schwartz, J.; Saez, M; Zanobetti, A.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Ponce de Leon, A.;
6 Le Moullec, Y.; Bacharova, L.; Schouten, J.; Ponka, A.; Katsouyanni, K. (1998) Time-series analysis of air
7 pollution and cause-specific mortality. Epidemiology 9: 495-503.
June 2003 8-304 DRAFT-DO NOT QUOTE OR CITE
-------
APPENDIX 8A
SHORT-TERM PM EXPOSURE-MORTALITY STUDIES:
SUMMARY TABLE
June 2003 8A-1 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8A-1. SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
United States
Sametetal. (2000a,b).*
90 largest U.S. cities.
1987-1994.
PM10 mean ranged from
15.3 (Honolulu) to
52.0 (Riverside).
Dominici et al. (2002).
Re-analysis of above study.
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.
Illustration of the issues related to GAM convergence
criteria using simulation; and re-analysis of above study
using stringent convergence criteria as well as comparable
GLM model with natural splines.
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 |ig/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).
The overall estimate was reduced but major findings of the study
were not changed. Sensitivity analysis using alternative degrees
of freedom for temporal trends and weather terms showed that
PM10 risk estimates were larger when smaller number of degrees
of freedom were used.
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 |ig/m3 PM10 increment.
Posterior mean estimates and 95%
credible intervals for total mortality
excess deaths per 50 ug/m3 increase in
PM10 at lag 1 day: 1.4% (0.9, 1.9) using
GAM with stringent convergence criteria
and 1.1 (0.5, 1.7) using GLM with
natural splines. Northeast still has the
largest PM10 risk estimate.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
United States (cont'd)
Dominici et al. (2000a).
+20 largest U.S. cities.
1987-1994. PM10 mean
ranged from 23.8 ug/m3
(San Antonio) to 52.0 ug/m3
(Riverside).
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.
This study examined the shape of concentration-response
curve. Three log-linear GAM regression models were
compared: (1) using a linear PM10 term; (2) using a natural
cubic spline of PM10 with knots at 30 and 60 ug/m3
(corresponding approximately to 25 and 75 percentile of the
distribution); and, (3) using a threshold model with a grid
search in the range between 5 and 200 ug/m3 with 5 ng/m3
increment. Covariates included the smoothing function of
time, temperature and dewpoint, and day-of-week
indicators. These models were fit for each city separately,
and for model (1) and (2) the combined estimates across
cities were obtained by using inverse variance weighting if
there was no heterogeneity across cities, or by using a two-
level hierarchical model if there was heterogeneity.
Dominici et al. (2003). Re-analysis of above model using GLM/natural splines.
Re-analysis of above study.
Daniels et al. (2000).*
The largest U.S. 20 cities,
1987-1994.
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 ng/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.
For total and cardiorespiratory mortality, the spline curves were
roughly linear, consistent with the lack of a threshold. For
mortality from other causes, however, the curve did not increase
until PM10 concentrations exceeded 50 ug/m3. The hypothesis of
linearity was examined by comparing the AIC values across
models. The results suggested that the linear model was
preferred over the spline and the threshold models.
The shapes of concentration-response curves were similar to the
original analysis.
Total mortality excess deaths per
50 ug/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
50ng/m3PM10: 3.4(1.0,5.9).
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
United States (cont'd)
Klemm et al. (2000).
Replication study of the
Harvard Six Cities
time-series analysis by
Schwartz et al. (1996).
Reconstruction and replication study of the Harvard
Six Cities time-series study. The original investigators
provided PM data; Klemm et al. reconstructed daily
mortality and weather data from public records. Data
analytical design (GAM Poisson model) was the same
as that from the original study.
The combined PM effect estimates were essentially equivalent to
the original results.
Total mortality percent excess risks:
PM10715: 4.1(2.8, 5.4) per 50ug/m3
PM25: 3.3(2.3, 4.3)per 25 ug/m3
PM10.25: 1.0(-0.4, 2.4)per 25 ug/m3
OO
>
H
6
o
o
H
O
o
H
W
O
Klemm and Mason (2003).
Re-analysis of the above
study.
Schwartz (2003a). Re-
analysis of the Harvard Six
Cities time-series analysis.
Zegeretal. (1999).
Philadelphia, 1974-1988.
Dominici et al. (2003).
Re-analysis of above study.
Re-analysis of the above study using GAM with
stringent convergence criteria and GLM/natural splines.
Sensitivity of results to alternative degrees of freedom were
also examined.
PM2 5 data were re-analyzed using GAM with stringent
convergence criteria, GLM/natural splines, B-splines,
penalized splines, and thin-plate splines.
The implication of harvesting for PM regression
coefficients, as observed in frequency domain, was
illustrated using simulation. Three levels of harvesting,
3 days, 30 days, and 300 days were simulated. Real data
from Philadelphia was then analyzed.
Re-analysis of above model using GLM/natural splines.
When GAM with stringent convergence criteria were applied,
PM effect estimates were reduced by 10 to 15%. GLM/natural
splines, and increasing the degrees of freedom for temporal
trends resulted in further reductions in PM coefficients.
When GAM with stringent convergence criteria were applied,
PM2 5 effect estimates were reduced by —5%. GLM/natural
splines, B-splines, penalized splines, and thin-plate splines each
resulted in further reductions in PM,, excess risk estimates.
In the simulation results, as expected, the shorter the harvesting,
the larger the PM coefficient in the higher frequency range.
However, in the Philadelphia data, the regression coefficients
increased toward the lower frequency range, suggesting that the
extent of harvesting, if it exists, is not in the short-term range.
Results were essentially unchanged.
Total mortality percent excess risks
using GAM stringent convergence
criteria:
PM10/15: 3.5(2.0, 5.1) per 50ug/m3
PM25: 3.0(2.1, 4.0) per 25 ug/m3
PM10.25: 0.8(-0.5, 2.0)per 25 ug/m3
Using GLM/natural splines:
PM10/15: 2.0(0.3, 3.8) per 50ug/m3
PM25: 2.0(0.9, 3.2)per 25 ug/m3
PM10.25: 0.3(-1.2, 1.8)per 25 ug/m
Total mortality percent excess risks
using per 25 ug/m3 PM2 5:
GAM (default):3.7(2.7, 4.7)
GAM (stringent): 3.5(2.5, 4.5)
Natural splines: 3.3(2.2, 4.3)
B-splines: 3.0(2.0, 4.0)
Penalized splines: 2.9(1.8, 4.)
Thin-plate splines: 2.6(1.5, 3.8)
O
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
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 ug/m3,
respectively in these cities.
Potential confounding caused by respiratory epidemics on
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.
When respiratory epidemics were adjusted for, small decreases
in the PM10 effect were observed (9% in Chicago, 11% in
Detroit, 3% in Minneapolis, 5% in Pittsburgh, and 15% in
Seattle).
The overall estimated percent excess
deaths per 50 |ig/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.
OO
>
Braga etal. (2001).*
Ten U.S. cities.
Same as Schwartz (2000b).
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.
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.
In the 7-day unconstrained distributed
lag model, the estimated percent excess
deaths per 50 ug/m3 PM10 were
14.2%(7.8, 21.1), 8.8%(0.6, 17.7),
5.1%(3.0, 7.2), and 3.0%(0.0, 6.2) for
pneumonia, COPD, all cardiovascular,
and myocardial infarction mortality,
respectively.
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
Schwartz (2003b).
Re-analysis of above study.
Re-analysis of above study using stringent convergence
criteria as well as penalized splines.
Small changes in PM risk estimates. Original findings
unchanged.
Above estimates using stringent
convergence criteria were: 16.5%(8.3,
25.3), 9.9%(0.6, 20.0), 5.1%(2.8, 7.5),
and 3.5%(-0.7, 8.0). Corresponding
numbers for penalized splines were:
11.5%(3.1, 20.6), 7.2%(-2.6, 18.0),
4.6%(2.0, 7.2), and 2.5%(-2.2, 7.5).
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
United States (cont'd)
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.
Schwartz (2003b).
of above study.
Re-analysis
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.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. The case for in vs. out of
hospital deaths and days PM10 < 50 ug/m3 were not re-
analyzed.
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 g/m3.
The total mortality RR estimates
combined across cities per 50 ug/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).
The total mortality RR estimates
combined across cities per 50 ug/m3
increase of mean of lag 0- and 1-days
PM10: overall 3.3 (2.6, 4.1); summer 3.4
(2.5, 4.4); winter 3.1 (2.0, 4.1); with SO2
3.2 (1.7, 4.8); with CO 4.5 (2.7, 6.4);
with 03 3.5 (2.2, 4.8).
Corresponding values for natural splines
are: overall 2.8 (2.0, 3.6); summer 2.6
(1.6, 3.7); winter 2.9 (1.8, 4.1); with SO2
2.8 (1.0, 4.6); with CO 3.7 (1.6, 5.8);
with O3 3.0 (1.6, 4.4).
o
H
O
c
o
H
W
O
V
O
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
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 (2003b). Re-analysis
of above study.
Schwartz and Zanobetti
(2000). + Ten U.S. cities.
Same as above.
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.
Re-analysis of above study using stringent convergence
criteria as well as penalized splines. Only quadratic
distributed lag and unconstrained distributed lag models
were re-analyzed.
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 |ig/m3 increments between 5.5 |ig/m3 and 69.5 ug/m3 of
PM10 levels. Then, the predicted values were combined
across cities using inverse-variance weighting.
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.
PM risk estimates were reduced but not substantially.
Original findings unchanged.
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.
Total mortality percent increase
estimates combined across cities per
50 ug/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.
Total mortality percent increase
estimates combined across cities per
50 ug/m3 increase in PM10: 6.3 (4.9, 7.8)
for quadratic distributed lag; and 5.8
(4.4, 7.3) for unconstrained distributed
lag using stringent convergence criteria.
Corresponding numbers for penalized
splines were: 5.3%(4.2, 6.5) and
5.3%(3.9).
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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
>
OO
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
United States (cont'd)
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.
Moolgavkar (2000a)*
Cook County, Illinois
Los Angeles County, CA
Maricopa County, AZ
1987-1995
PM10, CO, 03, N02, S02 in
all three locations.
PM2 5 in Los Angeles
County.
Cook Co:
PM10 Median = 47 ug/m3.
Maricopa Co:
PM10 Median = 41.
Los Angeles Co:
PM10 Median = 44;
PM,, Median = 22.
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.
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.
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.
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.
The total mortality RR estimates
combined across cities per 50 |ig/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 > 12y 3.6 (1.0, 6.3).
In single pollutant models, estimated
daily total mortality % excess deaths per
50 ug/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 ug/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/m3 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 PM25, 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 ug/m3 PM2 5, LA
3.6 (-0.6, 7.9) lag 3.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
United States (cont'd)
Moolgavkar (2003).
Re-analysis of above study,
but Maricopa Co. data were
not analyzed.
Ostro et al. (1999a).+
Coachella Valley, CA.
1989-1992. PM10
(beta-attenuation)
Mean = 56.8 ug/m3.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. Cerebrovascular deaths
data were not analyzed. Ozone was not analyzed.
In addition to the 30 degrees of freedom used for smoothing
splines for temporal trends in the original analysis, results
for 100 degrees of freedom were also presented. Two-
pollutant model results were not reported for Cook county.
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.
The sensitivity of results to the degrees of freedom was often
greater than that to the GAM convergence criteria. The main
conclusion of the original study remained the same.
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.
Maximum estimated non-accidental
deaths % excess deaths (95% CI) per
50 ug/m3 PM10: Cook Co. 2.4 (1.3,3.5),
lag 0; LA 2.4 (0.5, 4.4) Iag2; with CO, -
1.6(-3.7, 0.6); per 25 ug/m3 PM25 in LA
1.5 (0, 3.0).
Maximum estimated COPD % excess
deaths (95% CI) per 50 ug/m3 PM10:
Cook Co. 5.5 (0.3,11.0), lag 2; LA 4.4 (-
3.1,12.6) lag I;per25 ng/m3PM25in
LA 1.9 (-10.0, 15.4).
CVD % per 50 ug/m3 PM10:
Cook 2.2 (0.3, 4.1) lag 3; LA 4.5 (1.6,
7.5) lag 2; Percent per 25 ug/m3 PM25,
LA 2.6 (0.4, 4.9)lagl.
All the estimates above are for 30
degrees of freedom cases.
Total mortality percent excess deaths per
50 ug/m3 PM10 at 2-day lag = 4.6 (0.6,
Cardiac deaths:
8.33(2.14, 14.9)
Respiratory deaths:
13.9(3.25,25.6)
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
>
H
6
o
o
H
O
o
HH
H
W
United States (cont'd)
Ostro et al. (2000).*
Coachella Valley, CA.
1989-1998.
PM25 = 16.8;
PM10.25 = 25.8inIndio;
PM25 = 12.7;
PM10.2 5 = 17.9 in Palm
Springs.
Ostro et al. (2003).
Re-analysis of above study.
A follow-up study of the Coachella Valley data, with PM25
and PM10_25 data in the last 2.5 years. Both PM25 and
PM10_2 5 were estimated for the remaining years to increase
power of analyses. However, only PM10_2 5 could be reliably
estimated. Therefore, predicted PM2 5 data were not used for
mortality analysis. Thus, the incomparable sample size
make it difficult to directly assess the relative importance of
PM2 5 and PM10_2 5 in this data set.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. Only cardiovascular
mortality data were analyzed. Additional sensitivity
analyses were conducted.
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 PM2 5 data).
The PM risk estimates were slightly reduced with stringent
convergence criteria and GLM. Sensitivity analysis showed that
results were not sensitive to alternative degrees of freedom for
temporal trends and temperature. Multi-day averages for PM
increased risk estimates.
Total percent excess deaths:
PM10 (lag 0 or 2) = 2.0 (-1.0, 5.1) per
50 ug/m3
PM25 (lag 4) = 11.5 (0.2, 24.1) per
25 ug/m3
PM10.2 5 (lag 0 or 2) = 1.3 (-0.6, 3.5) per
25 ug/m3
Cardio deaths:
PM10(lagO) = 6.1 (2.0, 10.3) per
50 ug/m3
PM25 (lag 4) = 8.6 (-6.4, 25.8) per
25 ug/m3
PM10.25(lagO) = 2.6(0.7,4.5)per
25 ug/m3
Respiratory deaths:
PM10(lag3) = -2.0(-11.4, 8.4) per
50 ug/m3
PM25 (lag 1) = 13.3 (-43.1, 32.1) per
25 ug/m3
PM10.25(lag3) = -1.3(-6.2,4.0)per
25 ug/m3
Cardio deaths (GAM with stringent
convergence criteria):
PM10 (lag 0) = 5.5 (1.6, 9.5) per
50 ug/m3
PM25 (lag 4) = 10.2 (-5.3, 28.3) per
25 |ig/m3
PM10.2.5 (lag 0) = 2.9 (0.7, 5.2) per
25 ug/m3
Cardio deaths (GLM/natural splines):
PM10 (lag 0) = 5.1(1.2, 9.1) per
50 |ig/m3
PM2 5 (only 0-2 day lags reported)
PM10.25(lagO) = 2.7(0.5, 5.1) per
25 ug/m3
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
to
O
o
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
United States (cont'd)
oo
>
H
6
o
o
H
O
o
HH
H
W
Fairley(1999).*
Santa Clara County, CA
1989-1996.
PM25(13);PM10(34);
PM10.25(11);COH(0.5
unit);
NO3(3.0); SO4(1.8)
Fairley (2003). Re-analysis
of above study.
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).
Re-analysis of above study using stringent convergence criteria
as well as natural splines.
PM2 5 and nitrate were most significantly associated with
mortality, but all the pollutants (except PM10_25) 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
PM25. The 1980-1986 results were similar, except that COH
was very significantly associated with mortality.
PM coefficients were either unchanged, slightly decreased, or
slightly increased. Original findings, including the pattern in
two-pollutant models unchanged.
Total mortality per 25 ug/m3 PM2 5 at 0 d
lag: 8% in one pollutant model; 9-12%
in 2 pollutant model except with
NO3(~0). Also, 8% per 50 ug/m3 PM10
in one pollutant model and 2% per
25 ug/m3 PM10.2.5.
Cardiovascular mortality:
PM10 = 9% per 50 ug/m3
PM25 = 13%per 25 ug/m3
PM10.2 5 = 3% per 25 ug/m3
Respiratory mortality:
PM10 = 11% per 50 ug/m3
PM2 5 = 7% per 25 ug/m3
PM10_25 = 16% per 25 ug/m3
Percent excess mortality for GAM
(stringent) and GLM/natural splines,
respectively per 50 ug/m3 for PM10 and
25 ug/m3 for PM2 5 and PM10_2 5.
Total mortality:
PM10 =7.8(2.8,13.1); 8.3(2.9, 13.9)
PM25 = 8.2(1.6, 15.2); 7.1(1.4, 13.1)
PM10.25 = 4.5(-7.6, 18.1); 3.3(-5.3, 12.7)
Cardiovascular mortality:
PM10 = 8.5(0.6, 17.0); 8.9(1.3, 17.0)
PM25 = 6.4(-4.1, 18.1);6.8(-2.5, 16.9)
PM10.25 = 5.1(-13.4, 27.4); (no GLM)
Respiratory mortality:
PM10 = 10.7(-3.7, 27.2); 10.8(-3.4, 27.1)
PM25 = 11.8(-9.9, 38.7); 13.6(-3.7, 34.1)
PM10.25 = 32.2(-12.1, 98.6); (no GLM)
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
H
6
o
o
H
O
o
HH
H
W
United States (cont'd)
Schwartz et al. (1999).
Spokane, WA
1989-1995
PM10: "control" days:
42 ug/m3;
dust-storm days: 263
Popeetal. (1999a).
+ Ogden, Salt Lake City,
and Provo/Orem, UT
1985-1995
PM10 (32 for Ogden;
41forSLC;38forP/0)
Schwartz and Zanobetti
(2000) +Chicago 1988-
1993.
PM10. Median = 36 ug/m3.
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.
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.
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).
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.
0% (-4.5, 4.7) for dust storm days at 0
day lag (50 ng/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 (0-4 d) = 6.5% (2.2, 11.0)
Resp. (0-4 d) = 8.2 (2.4, 15.2)
Provo/Ovem PM10
Total (0 d) = 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 PM10.2 5 all
per 25 ug/m3; all PM10 % per 50 ug/m3.
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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
United States (cont'd)
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-S04-)
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, PM25, PM10_25, 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.
PM10, PM25, and PM10_25 were more significantly associated with
mortality outcomes than sulfate or H+. PM coefficients were
generally not sensitive to inclusion of gaseous pollutants. PM10,
PM25, and PM10_25 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.
Percent excess mortality per 50 ug/m3
for PM10 and 25 ug/m3 for PM2 5 and
PM10.2.5:
Total mortality:
PM10(ld) = 4.4(-l.0,10.1)
PM25(3d) = 23.1(-0.6, 7.0)
PM10.2.5(ld) = 4.0(-1.2,9.4)
Circulatory mortality:
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, 16.2)
Respiratory mortality:
PM10(Od) = 7.8(-10.2, 29.5)
PM25(Od) = 2.3 (-10.3, 16.6)
PM10.2 5 (2 d) = 7.4(-9.1,26.9)
H
O
O
o
H
O
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
United States (cont'd)
Ito (2003). Re-analysis of
above study.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. Additional sensitivity
analysis examined alternative weather models and influence
of the degrees of freedom in a limited data sets.
PM coefficients were often reduced (but sometimes unchanged
or increased) somewhat when GAM with stringent convergence
criteria or GLM/natural splines were used. The reductions in
coefficients were not differential across PM components; the
original conclusion regarding the relative importance of PM
components remained the same.
Percent excess mortality for GAM
(stringent) and GLM/natural splines,
respectively per 50 ug/m3 for PM10 and
25 ug/m3 for PM25 and PM10.25:
Total mortality:
PM10 (1 d) = 3.3(-2.0, 8.9); 3.1(-2.2, 8.7)
PM25 (3 d) = 1.9 (-1.8,5.7); 2.0(-1.7, 5.8)
PM10 25 (1 d) = 3.2(-1.9, 8.6); 2.8(-2.2,
8.1)
oo
>
Circulatory mortality:
PM10 (1 d) = 5.4(-2.6, 14.0); 4.9(-3.0,
13.5)
PM25 (1 d) = 2.2 (-3.2, 7.9); 2.0(-3.4,
7.7)'
PM10.25 (1 d) = 6.7 (-1.0, 15.0); 6.0(-1.6,
14.3)
H
6
o
Respiratory mortality:
PM10 (0 d) = 7.5(-10.5, 29.2); 7.9(-10.2,
29.7)
PM25 (0 d) = 2.3 (-10.4, 16.7); 3.1(-9.7,
17.7)
PM10.25 (2 d) = 7.0(-9.5, 26.5); 6.4(-10.0,
25.7) '
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
O
H
O
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
>
H
6
o
o
H
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.
PM,, mean = 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 GLM regression used,
including filtering of data based on cubic B-spline basis
functionsas adjustments for seasonal trends. Day-of-week
effects, temperature was modeled as a V-shape terms.
Single- and multi-pollutant models run for 0, 1,2, and 3 day
lags. PM2.5/PM10 0.67.
Reported "interim" results for 1 yr period of observations
regarding total mortality in Atlanta, GA during 1998-1999.
Poisson GLM model with natural splines used to assess
effects of PM25 vs PM10_25, 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
PM,n.,,.
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:
PM25 = 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)
PM10.25 = 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 |ig/m3 increase in PM25:
5.8(4.5, 7.2) for 15-day window
fluctuations; 9.6 (8.2, ll.l)forthe 60
day window.
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
United States (cont'd)
Schwartz (2003a).
Re-analysis of above study.
Reanalysis of above study using GLM/natural splines.
PM risk estimates at different time scales changed only slightly
(more often increased). Increase in standard error of PM
coefficients was also small (<3%). Original findings unchanged.
Total mortality percent increase per
25 |ig/m3 increase in PM25:
5.8 (4.5, 7.3) for 15-day window; 9.7
(8.2, 11.2) for the 60 day window.
OO
H
6
o
o
H
O
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
PM25: Si, S, Cl, K, Ca, V,
Mn, Al, Ni, Zn, Se, Br, Pb,
Cu, and Fe.
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 19-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.
Significant associations were found for a wide variety gaseous
and particulate 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.
The fractional Philadelphia mortality
risk attributed to the pollutant levels:
"average risk" was 0.0423 for 25 ug/m3
PM25; 0.0517 for 25 ug/m3 PM10.25;
0.0609 for 50 ug/m3 PM10, using the
Harvard PM indices at avg. of 0 and 1 d
lags.
Percent excess total mortality per
25|ig/m3 increase in PM2.5 from source
types:
Crustal: -5.6(-13.6, 3.1)
Traffic: 8.9(4.2, 13.8)
Coal: 2.8(0.8, 4.8)
Residual oil: 6.3(0.4, 12.5)
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
o
H
O
o
HH
H
W
United States (cont'd)
Schwartz (2003a).
Re-analysis of above study.
Levy (1998).
King County, WA.
1990-1994.
PM10 Nephelometer (30);
(0.59 bsp unit)
Maretal. (2000).*
Phoenix, AZ. 1995-1997.
PM10, PM2 5, and PM10.2 5
(TEOM), with means =
46.5, 13.0, and 33.5,
respectively; and PM2 5
(DFPSS), mean = 12.0.
Mar et al. (2003).
Re-analysis of above study.
Re-analysis of above study using penalized splines.
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 m fine particles), SO2, and CO, adjusting for day-of-
week, month of the year, temperature and dewpoint, using
Poisson GLM regression.
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.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. Only cardiovascular
mortality was re-analyzed.
The change in risk estimates for each source-apportioned PM2 5
in each city were either positive or negative, but the combined
estimates across cities increased for traffic factor and decreased
for coal factor and residual oil factor.
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.
Total mortality was significantly associated with CO and NO2
and weakly associated with SO2, PM10, PM10_25, 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.
Reductions on PM risk estimates for PM mass concentration
indices in the GAM/stringent convergence criteria or
GLM/natural splines were small. The change in coefficient for
source factors varied: moderate reductions for motor vehicle
factor, but slight increase for regional sulfate factor. EC and OC
coefficients were also slightly reduced.
Percent excess total mortality per
25ug/m3 increase in PM2.5 from source
types:
Crustal: -5.1(-13.9, 4.6)
Traffic: 9.3(4.0, 14.9)
Coal: 2.0(-0.3, 4.4)
Residual oil: 5.9(-0.9, 13.2)
Total mortality percent excess:
5.6% (-2.4, 14.3) per 50 ug/m3 PM10 at
avg. of 2 to 4 d lag; 7.2% (-6.3, 22.8)
withSO2CO. 1.8% (-3.5, 7.3) per
25 ug/m3 PMI; -1.0 (-8.7,. 7.7) with SO2
and CO.
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 ug/m3 at lag 0 d; 3.0 (-0.7, 6.9) for
PM25 (TEOM) 25 ug/m3 at lag 0 d.
Cardiovascular mortality RRs: 9.9 (1.9,
18.4) for PM10 (TEOM) 50 ug/m3 at lag
0 d; 18.7 (5.7, 33.2) for PM25 (TEOM)
25 ug/m3 at lag 1 d; and 6.4 (1.4, 11.7)
PM10 (TEOM) 25 ug/m3 PM10.25 at lag 0
d.
Percent excess cardiovascular mortality
per 50 ug/m3 PM10; 25 ug/m3 for PM2 5
and PM10_2 5: GAM with stringent
convergence criteria and GLM/natural
splines, respectively:
PM10 (0 d): 9.7(1.7, 18.3); 9.5(0.6, 19.3)
PM25 (1 d): 18.0(4.9, 32.6); 19.1(3.9,
36.4)
PM10.25 (0 d): 6.4(1.3, 11.7); 6.2(0.8,
12.0)
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
o
HH
H
W
United States (cont'd)
Clyde et al. (2000).
Phoenix, AZ. 1995-1998.
PM10, and PM25, (from
TEOM), with means = 45.4,
and 13.8. PM10_25 computed
asPM10-PM25.
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), and parametric terms for
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.
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.
Posterior mean RRs and 90% probability
intervals per changes of 25 ug/m3 in all
lags of fine and coarse PM for elderly
mortality for uniform PM10 region: 1.06
(1+, 1.11).
oo
oo
H
6
o
o
H
O
Smith et al. (2000). Study evaluated effects of daily and 2- to 5-day average
Phoenix, AZ. coarse (PM10_2 5) and fine (PM2 5) particles from an
1995-1997 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. Mortality was
square-root transformed. Initial model selected to represent
long-term trends (using B-splines) 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.
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_2 5. A seasonal interaction in the PM10_2 5 effect was also
reported: the effect being highest in spring and summer when
anthropogenic concentration of PM10_25 is lowest.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
VO
H
6
o
o
H
O
United States (cont'd)
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.
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)
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, PM25
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.
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 GLM 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.
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 ug/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.
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.
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
PM2 5; 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.
-3.6% (-12.7, 6.6) per 50 ug/m3 PM10 at 0
lag (other lags also reported to have no
associations)
Percent excess deaths per 25 ug/m3 of
estimated PM2 5, 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.
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
O
H
6
o
o
H
O
United States (cont'd)
Moolgavkar and Luebeck
(1996). Philadelphia, PA.
1973-1988. TSP(68)
Murray and Nelson (2000).
Philadelphia, PA,
1973-1990.
Smith et al. (1999).
Birmingham, AL 1985-
1988; Chicago (Cook Co.),
IL, 1986-1990. PM10
median = 45 ug/m3 for
Birmingham and
37.5 ug/m3 for Chicago.
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.
(Only one non-parametric smoothing terms in GAM
models)
Kalman filtering used to estimate hazard function in a state
space model. The model framework, which assumes
harvesting effect, allows estimation of at-risk population
and the effect of changes in air quality on the life
expectancy of the at-risk population. The model was first
verified by simulation. Combinations of TSP, linear
temperature, squared temperature, and interaction of TSP
and temperature were considered in six models.
Study evaluated associations between lagged/averaged PM10
and non-accidental mortality in two cities. Mortality was
square root-transformed in Birmingham data, and log-
transformed in Chicago data. Seasonal cycles were modeled
using B-splines. Temperature was modeled using piece-
wise linear terms with a change point. PM10 data were
included in the models at lag 0 through 3 and 3-day averages
at these lags. Also, to examine the possible existence of a
threshold, PM10 was modeled using a B-spline
representation, and also using parametric threshold model,
with the profile log likelihood evaluated at changing
threshold points. In addition, the possibility of mortality
displacement was examined with a model that attempts to
estimate the frail population size through Bayesian
techniques using Monte Carlo sampling.
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.
Both TSP and the product of TSP and average temperature are
significant, but not together. The size of at-risk population
estimated was about 500 people, with its life expectancy
between 11.8 to 14.3 days, suggesting that the hazard causing
agent making the difference of 2.5 days in the at-risk population.
The authors reported that, while significantly positive
associations were found in both cities, the results were sensitive
to the choice of lags. The PM10-mortality associations were
more stable in Chicago (perhaps in part due to sample size). The
non-linear estimates of relative risk using B-splines suggest that
an increasing effect above 80ug/m3 for Birmingham, and above
100 ug/m3 for Chicago. The threshold model through
examination of log likelihood at various possible threshold
levels also suggested similar change points, but not to the extent
that could achieve statistical distinctions. The mortality
displacement model in Chicago data suggested that the size of
the frail population was very small (mean —765), and the mean
lifetime within the frail population short (< 10 days).
Total mortality excess risk: ranged
0 (winter) to 4% (summer) per
100 ug/m3 TSP at 1 day lag.
The coefficients obtained in the models
cannot be directly compared to the
relative risk per ug/m3 PM obtained in
other time-series models.
Birmingham: 4.8%(t=1.98) per 50ug/m3
change in 1 through 3 day lag average of
PM10. Chicago: 3.7% (t=3.17) per
50|ig/m3 change in 0 through 2 day lag
average of PM10.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
H
6
o
o
H
O
United States (cont'd)
Neas et. al. (1999).
Philadelphia. 1973-1980.
TSP mean = 77.2.
Schwartz (2000d).
+Philadelphia. 1974-1988.
TSP. Mean = 70 ug/m3 for
warm season (April through
August) and 64 ug/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 uug/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 (> 80°F), 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 degreedays. Several of the
studies included were those that used GAM with multiple
non-parametric smoothing terms.
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 PM2 5/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 ug/m3 TSP.
The pooled estimate from 19 U.S. cities
was 0.70% (0.54, 0.84) per 10 ug/m3
increase in PM10.
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
Canada
Burnett etal.(1998a).+
11 Canadian cities.
1980-1991.
No PM index data available
on consistent daily basis.
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.
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.
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.
OO
to
to
H
6
o
o
H
O
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 and Goldberg
(2003).
Re-analysis of above study.
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.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. In the main model of the
original analysis, both dependent and independent variables
were pre-filtered, but in the re-analysis, co-adjustment (i.e.,
more common simultaneous regression) approach was used.
Additional sensitivity analysis included alternative fitting
criteria and changing the extent of smoothing for temporal
trends. Only PM10, PM25 and PM10_25 were analyzed.
No multiple pollutant models.
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. PM2 5 was a stronger predictor of mortality than PM10_2 5.
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.
In the GAM model (stringent convergence criteria), inclusion of
day-of-week variable made moderate increase in PM
coefficients (up to 30%). Alternative fitting criteria and degrees
of freedom for temporal trends also changed PM coefficients.
Generally, larger the degrees of freedom for temporal trends,
smaller the PM coefficients. PM10_25 were more sensitive to
alternative models than PM2 5.
Percentage increase in daily filtered non-
accidental deaths associated with
increases of 50 ug/m3 PM10 and
25 ug/m3 PM2 5 or PM10_2 5 at lag 1 day:
3.5 (1.0, 6.0)forPM10; 3.0 (1.1, 5.0) for
PM25; and 1.8 (-0.7, 4.4) for PM10.25.
In the multiple pollutant model with
PM25, PM10_25, and the 4 gaseous
pollutants, 1.9 (0.6, 3.2) for PM25; and
1.2 (-1.3, 3.8)forPM10.25.
Excess total mortality in the
GLM/natural splines with knot/2months,
and using AIC and White-noise test
fitting criteria at 1-day lag:
PM10: 2.7(-0.1, 5.5) per 50 ug/m3
PM25: 2.2(0.1, 4.2) per 25 ug/m3
PM10.25: 1.8(-0.6, 4.4)per 25 ug/m3
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
H
6
o
o
H
O
o
HH
H
W
Canada (cont'd)
Burnett et al. (1998b). +
Toronto, 1980-1994.
TSP (60); COH (0.42);
SO4= (9.2 ug/m3);
PM10 (30, estimated);
PM2 5 (18, estimated)
Goldberg et al. (2000)*
Montreal, Quebec
1984-95 Mean
TSP = 53.1
(14.6- 211.l)ug/m3
PM10 = 32.2
(6.5 - 120.5) ug/m3
PM25 = 3.3 (0.0 - 30.0)
ug/m3
Goldberg et al. (2001b)*
Montreal, Quebec.
1984-1993. Predicted PM2 5
mean= 17.6. CoH (1000ft)
mean = 0.24, sulfate mean
= 3.3.
Goldberg et al. (2001).
Data same as above.
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.
Study aimed to shed light on population subgroups that my
be susceptible to PM effects. Linked data on daily deaths
with other health data from the Quebec Health Insurance
Plan (QHIP) (physician visits, pharmaceutical Rx, 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.).
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.
Significant associations found for all-cause (total non-
accidental) and cause-specific (cancer, CAD, respiratory disease,
diabetes) with PM measures. Results reported for PM2 5, 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 (CHE).
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 PM2 5 or total sulfates).
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 ug/m3 PM2 5. 0 day lag for
TSP and PM10; Avg. of 0 and 1 day for
PM,,.
Percent excess mortality per 25 ug/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)
CHE (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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
Canada (cont'd)
Goldberg and Burnett
(2003). Re-analysis of
above studies by Goldberg
etal.
Re-analysis of above study using stringent convergence
criteria as well as natural splines. Cause-specific mortality
was not re-analyzed; re-analysis was focused only on the
sub-groups defined using the QHIP data that showed
associations with particles in the original study. Sensitivity
analyses included alternative weather models and using
different degrees of freedom for temporal trends.
The PM coefficients were not very sensitive to the extent of
temporal smoothing but were sensitive to the functional form of
weather models. Most of the originally reported associations
except for congestive heart failure were highly attenuated when
natural splines were used for weather model.
The percent excess deaths estimates for
non-accidental deaths per IQR (average
of 0-2 day lags) for CoH, predicted
PM2 5, and sulfate for GAM(stringent
convergence criteria) and GLM/natural
splines, respectively, were: CoH: 1.38,
0.85; Predicted PM25: 1.57, 0.55; sulfate:
1.03,0.27. Confidence bands were not
given but the GAM results for predicted
PM2 5 and sulfate were indicated as
significant at 0.05 level.
OO
>
to
Ozkaynaketal. (1996).
Toronto, 1970-1991.
TSP (80); COH (0.42
/1000ft).
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.
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.
Total mortality excess risk: 2.8% per 100
ug/m3 TSP at 0 day lag.
H
6
o
o
H
O
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location,
Years, PM Index, Mean
or Median, IQR in 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
H
6
o
o
H
O
Europe
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.
Samoli et al. (2001). *
APHEA 1 cities (see
Katsouyanni (1997). At least
five years between 1980-
1992. The PM levels are the
same as those in Katsouyanni
etal. (1997).
Samoli et al. (2003).
Re-analysis of above study.
Total daily deaths regressed on BS or SO2 using Poisson
GLM 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 re-analyzed 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.
Re-analysis of above study using stringent convergence
criteria as well as natural splines.
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 < 150 ug/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.
BS risk estimates using GAM were reduced by ~ 10% when
stringent convergence criteria were applied. Use of
GLM/natural splines resulted in further and greater reductions.
Total mortality excess deaths per
25 |ig/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 ug/m3 BS.
Total mortality RRs per 50 ug/m3 BS for
all cities, western cities, and central-
eastern cities using the GAM approach
were: 2.5% (2.1, 2.9); 3.1% (2.3, 3.8);
and, 2.3% (1.7, 2.9), 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.
Results corresponding to above using the
GAM with stringent convergence criteria
were: 2.3%(1.9, 2.7); 2.7% (2.0, 3.4);
and, 2.1% (1.5, 2.7), respectively.
Corresponding GLM/natural splines
results were: 1.2%(0.7, 1.7); 1.6%(0.8,
2.4); and, 1.0%(0.3, 1.7).
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
Europe (cont'd)
Katsouyanni et al. (2001).*
1990-1997 (variable from city
to city). 29 European cities.
Median PM10 ranged from 14
(Stockholm) to 66 (Prague).
Median BS ranged from 10
(Dublin) to 64 (Athens).
Katsouyanni et al. (2003).
Re-analysis of above study.
The 2ntl 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.
Re-analysis of above study using stringent convergence
criteria as well as natural splines and penalized splines.
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 pooled estimate (random effects estimate) was reduced by
4% when stringent convergence criteria in GAM were used, by
34% when natural splines were used, and by 11% when
penalized splines were used. The pattern of effect modification
originally reported remained the same. The original findings
were unchanged.
Total mortality excess risk per 50ug/m3
increase in PM10:
Fixed effects model: 3.5(2.9, 4.1)
Random effects model: 3.1(2.1, 4.2)
Total mortality excess risk per 50ug/m3
increase in PM10 using GAM (stringent
convergence criteria): 3.3(2.7, 3.9) and
3.0(2.0, 4.1) for fixed effects and random
effects models, respectively.
Corresponding estimates for
GLM/natural splines are: 2.1(1.5, 2.8)
and 2.1(1.2, 3.0). Using penalized
splines, the estimates are 2.9(2.3, 3.6)
and 2.8(1.8, 3.8).
H
6
o
o
H
O
o
HH
H
W
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.
Zanobetti and Schwartz
(2003a). Re-analysis of
above study.
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 GLM 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.
Re-analysis of above study using stringent convergence
criteria as well as natural splines and penalized splines.
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.
The pooled PM10 (average of 0 and 1 day) mortality risk
estimate was reduced by 4% when stringent convergence criteria
in GAM were used, by 18% when penalized splines were used.
For the 4th degree polynomial distributed lag model,
corresponding reductions were 10% and 26%.
NO2 and/or O3 estimates only.
Combined total mortality excess risk per
50ug/m3 increase in the average of 0 and
1 day lag PM10 was 3.4(2.0, 4.8) using
GAM with stringent convergence
criteria. For 4th degree polynomial
distributed lag model, it was 7.5(4.4,
10.7). Corresponding reductions using
penalized splines were 2.9(1.4, 4.4) and
5.6(1.5,9.8)
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
H
6
o
o
H
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).
Prescott et al. (1998).
Edinburgh, UK, 1981-1995.
PM10 (21, by TEOM only for
1992-1995); BS (8.7).
Rooney etal. (1998).
England and Wales, and
Greater London, UK
PM10 (56, during the worst
heat wave; 39, July-August
mean)
Cardiovascular, respiratory, and digestive mortality series
inlO 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 GLM
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 GLM 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 GLM 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).
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.
All effect size estimates (except O3) were positive for total deaths
(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.
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).
1.9% (0.0, 3.8) per 25 ug/m3 BS at lag 1
day; 1.3% (-1.0, 3.6) per 50 ug/m3 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 ug/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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
Europe (cont'd)
Wordley et al. (1997).
Birmingham, UK,
1992-1994.
PM10 (apparently
beta-attenuation, 26)
Hoek et al. (2000).
The Netherlands,
1986-1994.
PM10 (median 34);
BS (median 10).
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.
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.
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.
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.
5.6% (0.5, 11.0) per 50 ug/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)
Total mortality excess risk estimate per
50 ug/m3 PM10 (average of 0-6 days):
1.2(0.2, 2.2); 0.9(-0.8, 2.7) for CVD;
5.9(0.9, 11.2) for COPD; and 10.1(3.6,
17.1) for pneumonia.
OO
to
OO
H
6
o
o
H
O
o
HH
H
W
Hoek (2003). Re-analysis of
above study.
Hoeketal. (2001).*
The Netherlands.
1986-1994. PM10 (median
34); BS (median 10).
Re-analysis of above study using stringent convergence
criteria and natural splines.
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.
Very little change in PM risk coefficients (often slightly
increased) whether GAM with stringent convergence criteria or
GLM./natural splines were used.
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).
Total mortality excess risk estimate per
50 |ig/m3 PM10 (average of 0-6 days)
using GAM with stringent convergence
criteria: 1.4(0.3, 2.6); 0.9(-0.8, 2.7) for
CVD; 6.1(1.0, 11.4) for COPD; and
10.3(3.7, 17.2) for pneumonia.
Corresponding numbers using
GLM/natural splines are: 1.2(-0.1, 2.5);
1.6(-0.3, 3.5); 6.0(0.4, 11.8); 10.7(3.5,
18.3).
For PM10 (7-day mean), RRs for total
CVD, MI/IHD, arrhythmia, heart failure,
cerebrovascular, and thrombocytic
mortality per 50 ug/m3 increase
were:0.9(-0.8, 2.7), 0.3(-2.3, 3.0), 2.5(-
4.3, 9.9), 2.2(-2.5, 7.2), 1.9(-1.8, 5.8),
and 0.6(-6.8, 8.7)
, respectively. The RRs for BS were
larger and more significant than those for
PM10.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
VO
H
6
o
o
H
O
Europe (cont'd)
Hoek (2003). Re-analysis of
above study.
Ponkaetal. (1998).
Helsinki, Finland,
1987-1993.
TSP (median 64);
PM10 (median 28)
Peters et al. (2000b).
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).
Re-analysis of above study using stringent convergence
criteria and natural splines.
Total and cardiovascular deaths, for age groups < 65 and 65
+, were related to PM10, TSP, SO2, NO2, and O3, using
Poisson GLM 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 GLM model
with sine/cosine, temperature as a quadratic function,
relative humidity, influenza, day-of-week as covariates), as
well as GLM with natural splines for temporal trends and
weather terms 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.
Very little change in PM risk coefficients (often slightly
increased) whether GAM with stringent convergence criteria or
GLM./natural splines were used.
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 PM2 5 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.
For PM10 (7-day mean), RRs for total
CVD, MI/IHD, arrhythmia, heart failure,
cerebrovascular, and thrombocytic
mortality per 50 ug/m3 increase using
GAM with stringent convergence criteria
were:0.9(-0.8, 2.7), 0.4(-2.2, 3.0), 2.7(-
4.2, 10.1), 2.4(-2.3, 7.4), 2.0(-1.7, 5.9),
and 0.7(-6.8, 8.8), respectively. The
RRs for BS were larger and more
significant than those for PM10.
18.8% (5.6, 33.2) per 50 ug/m3 PM10 4
day lag (other lags negative or zero).
Total mortality excess deaths per 100
ug/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 PM25.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
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).
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.
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.
5.5 (1.1, 9.9) per 100 ug/m3 TSP at 1 day
lag.
Total mortality percent excess deaths per
100 ug/m3 increase in TSP at 2 day lag
was 3.4 (0.5, 6.4).
OO
H
6
o
o
H
O
Zanobetti et al. (2000a).
Milan, Italy. 1980-1989.
TSP mean = 142.
Anderson et al. (1996).
London, UK, 1987-1992.
BS(15)
The focus of this study was to quantify mortality
displacement using what they termed "GAM distributed lag
models", (smoothing term was filed with Penalized Plines)
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 GLM model.
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.
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 ug/m3 BS at 1-d
lag for total deaths.
CVD(1 d)= 1.0 (-1.1, 3.1).
Resp. (1 d)= 1.1 (-2.7,5.0).
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
H
6
o
o
H
O
o
HH
H
W
Europe (cont'd)
Michelozzi et al. (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 et al. (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 GLM
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 (OLS) 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 ug/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 ug/m3 PM13 at 0
day lag.
Total mortality percent increase per
25 ug/m3 increase in avg. of 0-3 day lags
ofBS: 2.76(1.31, 4.23) in general
population, and 1.14 (-4.4, 6.98) in the
COPD cohort.
6.1% per 100 ug/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 ng/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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
o
HH
H
W
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
for all causes. In the two-pollutant models, the PM10-mortality
associations were not diminished, whereas those with gaseous
pollutants were.
Odds ratio for all cause mortality per 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 N02, 03, and CO,
respectively.
OO
to
H
6
o
o
H
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 et al. (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 GLM 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
ug/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
Iday.
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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
H
6
o
o
H
O
Europe (cont'd)
Wichmannetal., (2000)
* Erfurt, Germany.
1995-1998.
Number counts (NC) &
mass concentrations (MC)
of ultrafine particles in three
size classes, 0.01 to 0.1 m,
and fine particles in three
size classes from 0.1 to 2.5
m 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).
Stolzel et al. (2003).
Re-analysis of above study.
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).
Re-analysis of above study using GAM with stringent
convergence criteria as well as GLM/natural splines. The
polynomial distributed lag model was not re-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 ultrafine
particles can coagulate into larger aggregates in a few hours,
ultrafine particle size and numbers can increase into the fine
particle category, resulting in some ambiguity. Significant
associations were found between mortality and ultrafine particle
number concentration (NC), ultrafine 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 ultrafine 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.
Very little change in PM risk coefficients when GAM models
with stringent convergence criteria were used. When
GLM./natural splines were used, many of the coefficients for
number concentrations slightly increased, but the coefficients for
mass concentrations decreased slightly.
Total mortality excess deaths:
Filter PM10 (0-4 d lag) = 6.6 (0.7, 12.8)
per 50 ug/m3. Filter PM25 (0-1 d) = 3.0
(-1.7, 7.9). MC for PM001.25 6.2% (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 ultrafine 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)
Best single-day lag:
PM0.01.0.lPer25 ug/m3: 3.6(-0.4, 7.7)
PM001.25per25 ug/m3: 3.9(0.0, 8.0)
PM25per25 ug/m3: -4.0(-7.9, 0)
PM10per25 ug/m3: 6.4(0.3, 12.9)
Best single-day lag using GAM
(stringent):
PM0.01.01per25 ug/m3: 3.6(-0.4, 7.7)
PM0.01.2.5per25ng/m3:3.8(-0.1,7.8)
PM25per25 ug/m3: -4.0(-7.8, -0.1)
PM10per25 ug/m3: 6.2(0.1, 12.7)
Best single-day lag using GLM/natural
splines:
PM0.01.0.lPer25ng/m3:3.1(-1.6,7.9)
PM001.25per25 ug/m3: 3.7(-0.9, 8.4)
PM25per25 ug/m3: -3.4(-7.9, 1.4)
PM10per25 ug/m3: 5.3(-1.8, 12.9)
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model, GEE = Generalized
Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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
H
6
o
o
H
O
Europe (cont'd)
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.
Roemer and Van Wijnen
(2001). + Amsterdam.
1987-1998.
BS and PM10 means in
"background" = 10 and 39;
BS mean in "traffic" area
= 21. (NoPM10
measurements available at
traffic sites)
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.
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.
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 (OLS) 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).
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.
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.
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)
Percent excess mortality for PM10, PM2 5,
and PM10_2 5 (avg. lag 0 and 1 days) were
0.2% (-1.8, 2.2) per 24.4 uug/m3 PM10,
0.6% (-1.5, 2.7) per 17.7 uug/m3 PM25,
and -0.6% (-4.2, 2.3) per 11.3 uug/m3
PM10_2 5 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).
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
Latin America
Cifuentes et al. (2000).+
Santiago, Chile.
1988-1996.
PM25(64.0), andPM1025
(47.3).
Castillejos et al. (2000).
Mexico City.
1992-1995.
PM10 (44.6), PM25 (27.4),
andPM10.25(17.2).
OO
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. They also
reported results for comparable GLM model.
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 model (only one non-parametric
smoothing term), 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.
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_2 5. 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 PM2 5 and PM10_2 5, the effect size for
PM10_2 5 remained about the same, but the effect size for PM2 5
became negligible.
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
PM25 and 2.3 (1.4, 3.2) for PM10.25 in
single pollutant GAM models. In GLM
models (whole year only), 1.4 (0.6, 2.1)
for PM25 and 1.6 (0.2, 3.0) for PM10.25
Total mortality percent increase
estimates per increase for average of
previous 5 days: 9.5 (5.0, 14.2) for
50 ug/m3 PM10; 3.7 (0, 7.6) for 25 ug/m3
PM25; and 10.5 (6.4, 14.8) for 25 ug/m3
PM10.2,.
H
6
o
o
H
O
Loomis et al. (1999).
Mexico-City, 1993-1995.
PM25 (mean: 27.4 ug/m3)
Borja-Aburto et al. (1998).
Mexico-City,
1993-1995.
PM25(mean: 27)
Borja-Aburto et al. (1997).
Mexico-City,
1990-1992.
TSP (median: 204)
Infant mortality (avg. 3/day) related to PM2 5, O3, and NO2,
adjusting for temperature and smoothed time, using Poisson
GAM model (same model as above, with only one non-
parametric smoothing term)
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
smoothing splines for temporal trend, using Poisson GAM
model (only one non-parametric smoothing term).
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 GLM models. The final
models were estimated using the iteratively weighted and
filtered least squares method to account for overdispersion
and autocorrelation.
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.
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). PM2 5 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.
Infant mortality excess risk: 18.2% (6.4,
30.7) per 25 ug/m3 PM2 5 at avg. 3-5 lag
days.
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).
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in |ig/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.
Latin America (cont'd)
Tellez-Rojo et al. (2000).
Mexico City. 1994.
PM10 mean = 75.1.
One year of daily total respiratory and COPD mortality
series were analyzed for their associations with PM10 and O3
using Poisson GLM model adjusting for cold or warm
months, and 1-day lagged minimum temperature. The data
were stratified by the place of deaths.
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.
Percent excess for total respiratory and
COPD mortality were 2.9% (0.9, 4.9) and
4.1% (1.3, 6.9) per 10 uug/m3 increase in
3-day lag PM10,
OO
H
6
o
o
H
O
Pereiraetal. (1998).
Sao Paulo, Brazil,
1991-1992.
PM10 (beta-attenuation, 65)
Gouveia and Fletcher
(2000). Sao Paulo, Brazil.
1991-1993. PM10
mean = 64.3.
Concei9ao et al. (2001)
+Sao Paulo, Brazil.
1994-1997. PM10
mean = 66.2
Intrauterine mortality associations with PM10, NO2, SO2,
CO, and O3 investigated using Poisson GLM 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 GLM
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 uug/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.
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
Australia
Morgan etal. (1998).
Sydney, 1989-1993.
Nephelometer (0.30
bscat/104m).
Site-specific conversion:
PM,.9;PM,0 18
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 model to adjust for
autocorrelation.
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.
4.7% (1.6, 8.0) per 25 ug/m3 estimated
PM2 5 or 50 ug/m3 estimated PM10 at avg.
of 0 and 1 day lags.
(Note: converted from nephelometry
data)
OO
Simpson et al. (1997).
Brisbane, 1987-1993.
PM10 (27, not used in
analysis). Nephelometer
(0.26 bscat/104m,
size range: 0.01-2 m).
Asia
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 model to adjust for autocorrelation. Season-
specific (warm and cold) analyses were also conducted.
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.
3.4% (0.4, 6.4) per 25 ug/m3 1-h PM2 5
increment at 0 d lag; and 7.8% (2.5, 13.2)
per 25 ug/m3 24-h PM25 increment.
H
6
o
o
H
O
Hong et al. (1999)
+Inchon, South Korea,
1995-1996 (20 months).
PM10 mean = 71.2.
Lee et al. (1999).
Seoul and Ulsan, Korea,
1991-1995. TSP(beta
attenuation, 93 for Seoul
and 72 for Ulsan)
Lee and Schwartz (1999).
Seoul, Korea. 1991-1995.
TSP mean = 9,,.
o
HH
H
W
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.
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 1-hr max O3 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.
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.
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.
Percent excess deaths (t-ratio) per 50
ug/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 100 ug/m3 TSP at
avg. of 0, 1, and 2 day lags.
OR for non-accidental mortality per 100
ug/m3 increase in 3-day average TSP was
1.010(0.988, 1.032).
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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
OO
H
6
o
o
H
O
Asia (cont'd)
Xu et al. (2000).
Shenyang, China, 1992.
TSP (430).
Ostro et al. (1998).
Bangkok, Thailand,
1992-1995
PM10 (beta attenuation, 65)
Cropper et al. (1997).
Delhi, India, 1991-1994
TSP (375)
Kwonetal. (2001)
+Seoul, South Korea,
1994-1998.
PM10 mean = 68.7.
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. While GAM models were used in
the process, the risk estimates presented were for a fully
parametric model (i.e., GLM).
Total (non-accidental), cardiovascular, respiratory deaths
examined for associations with PM10 (separate
measurements showed 50% of PM10 was PM25),using
Poisson GAM model (only one non-parametric smoothing
term in the model) adjusting for seasonal cycles, day-of-
week, temp., humidity.
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
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.
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.
Percent total excess deaths per 100
ug/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
S02= 1.07 (-1.05, 3.23).
Other deaths TSP = 3.52 (0.82, 6.30);
with S02 = 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 (3 d ave.) = 8.3 (3.1, 13.8)
Resp. (3 d ave.) = 3.0 (-8.4, 15.9)
2.3% (significant at 0.05, but SE of
estimate not reported) per 100 ug/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 42.1 ng/m3.
Corresponding ORs using the case-
crossover approach were 0.1% (-0.9,
1.2) and 7.4% (-2.2, 17.9),
respectively.
o
HH
H
W
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
to
O
o
Reference, Location, Years,
PM Index, Mean or Median,
IQR in 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.
Asia (cont'd)
Lee et al. (2000)
+Seven major cities, Korea.
1991-1997.
TSP mean = 77.9.
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.
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.
Percent excess deaths for all non-
accidental deaths was 1.7% (0.8,
2.6) per 100 ug/m3 2-day moving
average TSP.
OO
VO
H
6
o
o
H
O
+ = Used GAM with multiple non-parametric smooths, but have not yet re-analyzed. * = Used S-Plus Default GAM, and have re-analyzed results; GAM = Generalized Additive Model,
GEE = Generalized Estimation Equations, GLM = Generalized Linear Model.
o
HH
H
W
-------
APPENDIX 8B
PARTICULATE MATTER-MORBIDITY STUDIES:
SUMMARY TABLES
June 2003 8B-1 DRAFT-DO NOT QUOTE OR CITE
-------
Appendix 8B.1: PM-Cardiovascular Admissions Studies
June 2003 8B-2 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8B-1. ACUTE PARTICULATE MATTER EXPOSURE AND CARDIOVASCULAR HOSPITAL ADMISSIONS
to
O
o
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
H
6
o
o
H
O
o
HH
H
W
United States
oo
td
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: 29.3h, 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
Zanobetti and Schwartz (2003b)
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.
Statistical reanalysis using GAM with improved
convergence criterion (New GAM), GLM with natural
splines (GLM NS), and GLM with penalized splines
(GLM PS). Lag structure: average of lags 0 and 1.
City-specific risk estimates for a 10 ug/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 ug/m3: 0- 1 d lag.
7.6% (6.0, 9.1)
Default GAM: 5.9% (5.1-6.7)
New GAM: 4.95% (3.95-5.95)
GLM NS: 4.8% (3.55-6.0)
GLM PS: 5.0% (4.0-5.95)
-------
to
O
o
oo
td
H
6
o
o
H
O
o
HH
H
W
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
United States (cont'd)
Janssen et al. (2002)
14 U.S. cities studied in Samet et al.
(2000a,b) above
PM,n (ug/m3)
Birmingham
Boulder*
Canton
Chicago
Colorado Springs*
Detroit
Minneapolis
Nashville
New Haven
Pittsburgh
Seattle*
Spokane*
Provo-Urem*
Youngstown
Mean
Summer/Winter
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
Ratio
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.)
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 mile 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.
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]
Zanobetti and Schwartz (2003b)
Statistical reanalysis of Janssen et al., 2002 findings
using GLM with natural splines (GLM NS), and GLM
with penalized splines (GLM PS). Lag structure:
average of lags 0 and 1.
Zanobetti and Schwartz (2003b) reanalyzed the
main findings from this study using alternative
methods for controlling time and weather
covariates. While the main conclusions of the
study were not significantly altered, some
changes in results are worth noting. The effect of
air conditioning use on PM10 effect estimates
was less pronounced and no longer statistically
significant for the winter PMlO-peaking cities
using natural splines or penalized splines in
comparison to the original Janssen et al. GAM
analysis. The effect of air conditioning remained
significant for the non-winter PMlO-peaking
cities. The significance of highway vehicles and
diesels on PM10 effect sizes remained
significant, as did oil combustion.
Homes with AC
PCVD
% change (SE)
All cities
GLMNS: -13.55(14.9)
GLM PS:-12.0 (14.1)
Nonwinter peaking cities
GLMNS: -44.1" (20.15)
GLM PS:-38.4" (17.8)
Winter peaking cities
GLMNS: -6.1 (40.3)
GLM PS:-41.5 (39.6)
Source PM10 from
highway vehicles
% change (SE) P CVD
GLMNS: 51.1" (14.7)
GLM PS: 35.1" (14.3)
[**p <0.05]
-------
to
O
o
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
United States (cont'd)
oo
>
Zanobetti et al. (2000b)
10 US cities
1986-1994
PM10 (ug/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
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/day in the 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.
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.
This study used the old GAM model. Results
have not been explicitly reanalyzed, but note that
the 14 cities noted above in Zanobetti and
Schwartz (2003b) include these 10 cities.
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 ug/m3: 0- 1 d.
7.8 (6.2, 9.4)
H
6
o
o
H
O
o
H
W
O
O
HH
H
W
Schwartz (1999)
8 US metropolitan counties
1988-1990
median, IQR for PM10 (ug/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
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.
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.
This study used the old GAM model. No
reanalysis has been reported.
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
-------
to
O
o
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
United States (cont'd)
oo
td
i
a\
Linn et al. (2000)
Los Angeles
1992-1995
mean, SD:
PM10eat(ng/m3):45, 18
Hospital admissions for total cardiovascular diseases
(CVD), congestive heart failure (CHF), 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 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.
This study did not use the GAM model in
developing its main findings.
% increase with PM10 change of 50 ug/m3:
CVD ages 30+
3.25% (2.04, 4.47)
MI ages 30+
3.04% (0.06, 6.12)
CHF ages 30+
2.02% (-0.94, 5.06)
CA ages 30+
1.01% (-1.93, 4.02)
H
6
o
o
H
O
Morris and Naumova (1998)
Chicago, IL
1986-1989
mean, median, IQR, 75th percentile:
PM10(ug/m3): 41,38,23,51
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.
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.
This study did not use the GAM model.
Percent Excess Risk (95% CI)
per 50 |ig/m3 change in PM10.
PM10: Od.
3.92% (1.02, 6.90)
1.96% (-1.4, 5.4) with
4 gaseous pollutants
o
HH
H
W
-------
to
O
o
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
United States (cont'd)
oo
td
H
6
o
o
H
O
o
HH
H
W
Schwartz (1997)
Tucson, AZ
1988-1990
mean, median, IQR:
PM10 (ug/m3): 42, 39, 23
Gwynn et al (2000)
Buffalo, NY
mn/max
PM10 = 24.1/90.8ng/m3
S04' = 2.4/3.9
H+ = 36.4/38.2 nmol/m3
CoH = 0.2/0.9 10'3 ft
Lippmann et al. (2000)
Detroit (Wayne County), MI
1992-1994
mean, median, IQR:
PM25(ug/m3): 18, 15, 11
PM10(ug/m3):31,28, 19
PM10.25 (ng/m3): 13, 12, 9
Daily hospital admissions for total cardiovascular
diseases (ICD9 codes 390-429) among persons over
65 years. Meanhospitalizations: 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.
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.
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, PM25, 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).
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.
This study used the old GAM model. No
reanalysis has been reported.
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.
This study used the old GAM model. No
reanalysis has been reported.
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 Risk (95% CI)
per 50 ug/m3 change in PM10.
PM10: Od.
6.07% (1.12, 1.27)
5.22% (0.17, 10.54) w. CO
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)
Percent excess CVD HA risks (95% CI) per
50 ug/m3 PM10, 25 ug/m3 PM25 and
PM,,
PM25 (lag 2) 4.3 (- 1.4, 10.4)
PM10 (lag 2) 8.9 (0.5, 18.0)
PM10.2.5(lag2)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:
PM2 5 (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(lagO) 1.8 (-5.3, 9.4)
PM10(lag 1)4.8 (-5.5, 16.2)
PM10.2.5 (lag 1)4.9 (-4.7, 15.5)
-------
to
O
o
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
i
OO
H
6
o
o
H
O
United States (cont'd)
Ito 2003
Detroit (Wayne County), MI
Statistical reanalysis using GAM with improved
convergence criterion (New GAM), and GLM with
natural splines (GLM NS). Same model structure as
before.
IHD:
New GAM: 8.0% (-0.3-17.1)
GLM NS: 6.2% (-2.0-15.0)
New GAM: 3.65% (-2.05-9.7)*
GLM NS: 3.0% (-2.7-9.0)*
New GAM: 10.2% (2.4-18.6)**
GLMNS: 8.1% (0.4-16.4)**
Dysrhythmias:
New GAM: 2.8% (-10.9-18.7)
GLMNS: 2.0% (-11.7-17.7)
New GAM: 3.2% (-6.6-14.0)*
GLMNS: 2.6% (-7.1-13.3)*
New GAM: 0.1% (-12.4-14.4)**
GLMNS: 0.0% (-12.5-14.3)**
Heart Failure:
New GAM: 9.2% (-0.3-19.6)
GLMNS: 8.4% (-1.0-18.7)
New GAM: 8.0% (1.4-15.0)*
GLM NS: 6.8% (0.3-13.8)*
New GAM: 4.4% (-4.0-13.5)**
GLMNS: 4.9% (-3.55-14.1)**
o
HH
H
W
-------
to
O
o
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
H
6
o
o
H
O
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. PM25 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/PM2 5 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)
PM2 5, 0 d.
4.3(2.5,6.1)
PM2 5, 0 d. 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)
PM2 „ 0 d.
3.5(1.8,5.3)
PM25, 0 d., w/CO)
2.3 (-0.2, 4.8)
Maricopa:
PM10, 0 d.
-2.4 (-6.9, 2.3)
o
HH
H
W
-------
to
O
o
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)
Moolgavkar (2003)
Zanobetti et al. (2000a)
Cook County, IL
1985-1994
Median, IQR:
PM10 (ng/m3): 33, 23
Statistical reanalysis using GAM with improved
convergence criterion (New GAM), and GLM with
natural splines (GLM NS). New analyses were run
with variable and in some cases more extensive control
of time than in original analysis.
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.
Evidence seen for increased CVD effects among
persons with concurrent respiratory infections or
with previous admissions for conduction
disorders.
Cook County, IL:
New GAMlOOdf: 4.05% (2.9-5.2)
GLMNSlOOdf: 4.25% (3.0-5.5)
Los Angeles County, CA:
New GAM30df: 3.35% (1.2-5.5)
New GAMlOOdf: 2.7% (0.6-4.8)
GLMNSlOOdf: 2.75% (0.1-5.4)
New GAM30df: 3.95% (2.2-5.7)*
New GAMlOOdf: 2.9% (1.2-4.6)*
GLM nsplinelOOdf: 3.15% (1.1-5.2)*
Percent Excess CVD Risk (95% CI)
Effects computed for 50 ug/m3
PM10, 0-1D. AVG.
CVD: 6.6 (4.9-8.3)
fe
H
6
o
o
H
O
o
HH
H
W
-------
to
O
o
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
Tolbert et al. (2000a)
Atlanta
Period 1: 1/1/93-7/31/98
Mean, median, SD:
PM10(ng/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 (ug/m3): 19.4, 17.5, 9.35
CP (ug/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
PM2 5 soluble metals (ug/m3): 0.0327,
0.0226, 0.0306
PM25 Sulfates (ug/m3): 5.59, 4.67, 3.6
PM25 Acidity (ng/m3): 0.0181, 0.0112,
0.0219
PM25 organic PM (ug/m3): 6.30, 5.90, 3.16
PM25 elemental carbon (|ig/m3): 2.25, 1.88,
1.74
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.
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 ng/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 Risk (p-value):
Effects computed for 50 ug/m3 change in
PM10; 25 ug/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
HH
H
W
-------
to
O
o
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
Canada
oo
td
i
to
fe
H
6
o
o
H
O
Burnett et al. (1995)
Ontario, Canada
1983-1988
Sulfate
Mean: 4.37 ug/m3
Median: 3.07 ug/m3
95thpercentile: 13 ug/m3
Burnett et al. (1997a)
Canada's 10 largest cities
1981-1994
COH daily maximum
Mean: 0.7 103 In feet
Median: 0.6 103 In feet
95thpercentile: 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 /«CO 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, N02, SO2, O3:
-2.26 (-6.5, 2.2)
o
HH
H
W
-------
to
O
o
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
Canada (cont'd)
oo
td
fe
H
6
o
o
H
O
Burnett etal. (1997b)
Metro-Toronto, Canada
1992-1994
Pollutant: mean, median, IQR:
COH(103lnft): 0.8,0.8,0.6
H+ (nmol/m3): 5, 1, 6
SO4 (nmol/m3): 57, 33, 57
PM10(ng/m3): 28,25,22
PM25 (ug/m3): 17, 14, 15
PM,,
(Hg/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 ug/m3
PM10, 25 jig/m3 PM25 and PM10_25, and IQR
for other indicators.
COH: 0-4d.
6.2 (4.0, 8.4)
5.9(2.8, 9.1) w. gases
H+: 2-4d.
2.4(0.4,4.5)
0.5 (-1.6, 2.7) w. gases
S04:2-4d.
1.7 (-0.4, 3.9)
-1.6 (-4.4, 1.3) w. gases
PM10: 1-4 d.
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
HH
H
W
-------
to
O
o
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
o
HH
H
W
Canada (cont'd)
oo
td
fe
H
6
o
o
H
O
Burnett et al. (1999)
Metro-Toronto, Canada
1980-1994
Pollutant: mean, median, IQR:
FPest (ug/m3): 18, 16, 10
CPest (ug/m3): 12,10,8
PM10est(ng/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 ug/m3 PM10.2.5.
All cardiac HA (lags 2-5 d):
PM25 l-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)
PM10est: (Od): 8.41(2.89, 14.2)
HF:
FPest(0-2d): 6.59(2.50, 10.8)
CPest(0-2d): 7.9(2.28, 13)
PM10est(0-2d): 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)
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
to
O
o
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
Canada (cont'd)
oo
td
fe
H
6
o
o
H
O
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(ng/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 ng/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+: 4-9d. 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
HH
H
W
-------
to
O
o
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
Europe
oo
td
fe
H
6
o
o
H
O
o
HH
H
W
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 et al. (1999b)
Greater London, England
1992-1994
Pollutant: mean, median, 90-10 percentile
range:
PM10 (ug/m3): 28.5, 24.8, 30.7
Black Smoke (ug/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;forIHD: 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 the 5% level.
Effects computed for 50 ug/m3 PM10 and
25 ug/m3 BS
PM10 0 d.
All ages:
CVD: 3.2(0.9, 5.5)
0-64yr:
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)
0-64yr:
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)
-------
to
O
o
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
Europe (cont'd)
oo
td
fe
H
6
o
Prescott et al. (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
Wordley et al. (1997)
Birmingham, UK
4/1/92-3/31/94
mean, min, max:
PM10(ng/m3): 26,3, 131
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
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.
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 showed 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).
Percent Excess Risk (95% CI):
Effects computed for 50 ug/m3 change in
PM10 and 25 ug/m3 change in BS.
Long series:
BS, l-3d. avg.
<65: -0.5 (-5.4, 4.6)
65+: -0.5 (-3.8, 2.9)
Short series:
BS, 1-3d. avg.
<65: -9.5 (-24.6, 8.0)
65+: 5.8 (-4.9, 17.8)
PM10, 1-3 d. avg.
<65: 2.0 (-12.5, 19.0)
65+: 12.4(4.6,20.9)
% change (95% CI) per
50 |ig/m3 change PM10
IHD admissions:
PM10 0-dlag:
1.4% (-4.4, 7.2)
PM10 1-dlag:
-1.3% (-7.1, 4.4)
o
H
O
o
HH
H
W
Diaz et al. (1999)
Madrid, Spain
1994-1996
TSP by beta attenuation
Summary statistics not given.
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.
No significant effects of TSP on CVD reported. No quantitative results presented for PM.
-------
to
O
o
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
Australia
oo
td
i
oo
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 1-hr bscat/104m: 0.76, 0.57, 60,
1.23
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 PM2 5.
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.
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.
Percent Excess Risk (95% CI):
Effects computed for 25 ug/m3 PM2 5
(converted from bscat).
24-hr avg. PM2 5 0 d.
<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: 0.19 (-1.6, 2.0)
65+: 1.8(0.5,3.2)
All: 1.3 (0.3, 2.3)
Asia
fe
H
6
o
o
H
O
Wong et al. (1999a)
Hong Kong
1994-1995
median, IQR for PM10 (ug/m3): 45.0, 34.8
Daily emergency hospital admissions for
cardiovascular diseases, CVD (ICD9 codes 410-417,
420-438, 440-444), heart failure, HE (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, 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 HE 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 50 ug/m3 change in
PM10.
PM10, 0-2d. avg.
CVD:
5-64: 2.5 (-1.5, 6.7)
65+: 4.1(1.3,6.9)
All: 3.0 (0.8, 5.4)
HE (PM10, 0-3 d ave.):
All: 26.4(17.1,36.4)
o
HH
H
W
IHD (PM10, 0-3 d ave.):
All: 3.5 (-0.5, 7.7)
-------
Appendix 8B.2. PM-Respiratory Hospitalization Studies
June 2003 8B-19 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 8B-2. ACUTE PARTICULATE 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
fe
H
6
o
o
H
O
o
HH
H
W
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 ug/m3
PM10IQR = NR
Hospital admissions for adults 65+ yrs. for CVD
(mean=22. I/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.
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.
PM10 = 50 ug/m3
COPD HA's for Adults 65+ yrs.
Lag 0 ER = 7.4% (CI: 5.1,9.8)
Lag 1 ER = 7.5% (CI: 5.3, 9.8)
2 day mean (lagO,lagl) ER = 10.3%
(CI: 7.7, 13)
Pneumonia HA's for Adults 65+ vrs.
LagOER=8.1%(CI: 6.5,9.7)
Lag 1 ER = 6.7% (CI: 5.3, 8.2)
2 day mean (lagO, lagl) = 10.3% (CI: 8.5, 12.1)
OO
td
i
to
o
Reanalysis of Samet et al (2000) by
Zanobetti and Schwartz (2003b)
Re-analyses of Samet et al. (2000) with more
stringent GAM convergence criteria and
alternative models.
Results differ somewhat from original analyses,
especially for pneumonia. Results indicate that
the stricter convergence criteria results in about
a 14% lower GAM effect than in the originally
published analyses method. Authors
recommend the penalized spline model results.
COPD 2 day mean (lag 0, lagl):
Default GAM ER=9.4 (5.9, 12.9)
Strict GAM ER = 8.8 (4.8, 13.0)
NS GLM ER=6.8 (2.8, 10.8)
PSGLMER=8.0 (4.3, 11.9)
Pneumonia 2 day mean (lag 0, lagl):
Default GAM ER=9.9 (7.4, 12.4)
Strict GAMER =8.8(5.9, 11.8)
NS GLM ER=2.9 (0.2,5.6)
PS GLM ER = 6.3 (2.5,10.3)
-------
to
O
o
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
i
to
fe
H
6
o
o
H
O
United States (cont'd)
Zanobetti et al. (2000b)+
10 U.S. Cities
Jamason et al. (1997)
New York City, NY (82 - 92)
Population = NR
PM10 mean = 38.6 ug/m3
Chen et al. (2000)+
Reno-Sparks, NV (90 - 94)
Population = 307,000
B-Gauge PM10 mean=36.5 ug/m3
PM10IQR = 18.3-44.9 ug/m3
PM10 maximum = 201.3 ug/m3
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 ug/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.
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.
Only one LOESS used with GAM, so the default
convergence criteria may be satisfactory in this
case. No co-pollutants considered.
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. Suggests
association between PM10 and total respiratory
hospital admissions among the elderly.
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 ug/m3.
Percent excess respiratory risk (95% CI) per 50 ug/m3
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)
NR
COPD All age Admissions
50 ug/m3 IQR PM10 (single pollutant):
ER = 9.4%(CI: 2.2, 17.1)
o
HH
H
W
-------
to
O
o
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
i
to
to
United States (cont'd)
Gwynn et al. (2000)+
Buffalo, NY (5/88-10/90)
PM10 mn./max. = 24.1/90.8 ug/m3
PM10 IQR =14.8-29.2 ug/m3
SO4" mn./max. = 2.4/3.9 ug/m3
SO4- IQR = 23.5-7.5 ug/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 and Thurston (2001)+
New York City, NY
1988, 89, 90
PM10 37.4 ug/m3 mean
Jacobs et al. (1997)
Butte County, CA (83 - 92)
Population = 182,000
PM10 mean = 34.3 ug/m3
PM10 min/max = 6.6 / 636 ug/m3
CoH mean = 2.36 per 1000 lin. ft.
CoH min/max = 0 /16.5
Air pollutant-health effect associations with
total, respiratory, and circulatory hospital
admissions and mortality examined using
Poisson methods controlling for weather,
seasonally, long-wave effects, day of week, and
holidays using GAM with LOESS terms.
Respiratory hospital admissions, race specific for
PM10, H+, O3, SO4- LOESS GAM regression
model used to model daily variation in
respiratory hospital admissions, day-week,
seasonal, and weather aspects addressed in
modeling.
Association between daily asthma HA's (mean =
0.65/day) and rice burning using Poisson GLM
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.
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.
Increases in rice straw burn 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.
Respiratory Hospital Admissions(all ages) PM Index
(using standardized cone, increment)
-Single Pollutant Models
For PM10 = 50 ug/m3; SO4 = 15 ug/m3;
H+ = 75nmoles/m3;COH = 0.5 units/lOOOft
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)
Asthma HA's (all ages)
For an increase of 50 ug/m3 PM10:
ER = 6.11% (not statistically significant)
fe
H
6
o
o
H
O
Linn et al. (2000)
Los Angeles, CA (92 - 95)
Population = NR
PM10 mean = 45.5 ug/m3
PM10 Min/Max = 5/132 ug/m3
Pulmonary hospital admissions (HA's)
(mean=74/day) related to CO, NO2, PM10, and O3
in Los Angeles using GLM Poisson model with
long-wave spline, day of week, holidays, and
weather controls.
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).
Pulmonary HA's (>29 yrs.)
PM10 = 50 ug/m3
(Lag 0)ER = 3.3% (CI: 1.7, 5)
o
HH
H
W
-------
to
O
o
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)
Moolgavkar et al. (1997)+
Minneapolis-St. Paul 86-91
Populations NR
Birmingham, AL '86-'91
Population. = NR
PM10 mean = 34 ug/m3 (M-SP)
PM10 IQR =22-41 ug/m3 (M-SP)
PM10 mean =43.4 ug/m3(Birm)
PM10 IQR =26-56 ng/m3(Birm)
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).
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.
COPD + Pneumonia Admissions (>64yrs.)
In M-SP, For PM10 = 50 ug/m3 (max Ig)
ER(lg 1) = 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 ug/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)
OO
td
i
to
fe
H
6
o
o
H
O
Nauenberg and Basu (1999)
Los Angeles (91 -94)
Wet Season = 11/1-3/1
Dry Season = 5/1-8/15
Population .= 2.36 Million
PM10 Mean = 44.81 ug/m3
PM10 SE = 17.23 ug/m3
Schwartz et al. (1996b)
Cleveland (Cayahoga County), Ohio
(88 - 90)
PM10 mean = 43 ug/m3
PM10 IQR = 26-56 ug/m3
The effect of insurance status on the association
between asthma-related hospital admissions and
exposure to PM10 and O3 analyzed, using GLM
Poisson 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.
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 PM,n.
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.
All Age Asthma HA's
PM10 = 50 ug/m3, no co-pollutant, during wet season
(Jan. 1 - Mar. 1):
All Asthma Hospital Admissions
0-d lag PM10ER= 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-d avg. PM10 ER = 6.2 (-3.6, 16.1)
Respiratory HA's for persons 65+ years
50 ug/m3 PM10
ER = 5.8%(CI: 0.5, 11.4)
o
HH
H
W
-------
to
O
o
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)
Zanobetti, et al. (2000a)+
Study Period: 86-94
Chicago (Cook Count), IL
Population = 633,000 aged 65+
PM10 mean = 33.6 ug/m3
PM10 range = 2.2, 157.3 ug/m3
Analyzed HA's for older adults (65 + yr) for
COPD (mean = 7.8/d), pneumonia (mean =
25.5/d), and CVD, using GLM 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.
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.
PM10 = 50 ug/m3(average of lags 0,1)
COPD (adults 65+ yrs.)
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+ yrs.)
W/o pr. Asthma ER =11% (CI: 7.7, 14.3)
With pr. 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)
OO
td
i
to
fe
H
6
o
o
H
O
o
HH
H
W
-------
to
O
o
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
i
to
fe
H
b
O
o
H
O
United States (cont'd)
Lippmann et al. (2000)*
Detroit, MI ('92-'94)
Population = 2.1 million
PM10 Mean = 31 ug/m3
(IQR= 19, 38 ug/m3; max=105 ug/m3)
PM25Mean= 18 ug/m3
(IQR= 10, 21 ug/m3; max=86 ug/m3)
PM10_2 5 Mean =12 ug/m3
(IQR= 8, 17 ug/m3; max=50 ug/m3)
SO4"Mean = 5 ug/m3
(IQR=1.8, 6.3 ug/m3;
max=34.5 ug/m3)
H+ Mean = 8.8 nmol/m3 = 0.4 ug/m3
(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 using LOESS
smooths. 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
ug/m3 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 ug/m3) Id lag
ER = 22%(CI: 8.3,36)
PM25(25 ug/m3) Id lag:
ER=13%(CI: 3.7,22)
PM2.5.10 (25 ug/m3) Id lag:
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 ug/m3) Id lag,
ER = 24%(CI: 8.2,43)
PM25(25 ug/m3) Id lag:
ER=12%(CI: 1.7,23)
PM25.10(25ug/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 ug/m3) 3d lag
ER = 9.6%(CI: -5.1,27)
PM25(25 ug/m3) 3d lag:
ER = 5.5%(CI: -4.7, 17)
PM25.10 (25 ug/m3) 3d lag:
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 ug/m3) 3d lag,
ER= 1.0% (-15, 20)
PM25(25 ug/m3) 3d lag:
ER = 2.8%(CI: -9.2, 16)
PM25.10(25ug/m3)3dlag:
ER = 0.3%(CI: -14, 18)
H+ (75 nmol/m3) 3d lag:
ER= 13% (CI: -0.6,28)
o
HH
H
W
-------
to
O
o
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
i
to
fe
H
6
o
o
H
O
United States (cont'd)
Reanalysis by
Ito (2003)
Re-analyses of Lippmann et al. (2000) with more
stringent GAM convergence criteria and
alternative models.
Lumley and Heagerty (1999)
Seattle (King Cty.), WA (87-94)
Population = NR
PM; daily mean = NR
PMj.n, daily mean = NR
From Sheppard et al, 1999:
PM10 mean = 31.5 ug/m3
PM10IQR = 19-39 ug/m3
PM25 mean = 16.7 ug/m3
PM25 IQR = 8-21 ug/m3
Estimating equations based on marginal
generalized linear models (GLM) 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.
More stringent GAM generally, but not always,
resuled in reduced RR estimates, but effect sizes
not significantly different from originals. Extent
fo reuction independent of risk estimate size.
The reductions were not differential across PM
components, so study conclusions unchanged.
PMj at lag 1 day associated with respiratory
HA's in children and younger adults (<65), but
not PM10-!, suggesting a dominant role by the
submicron particles in PM2 5-asthma HA
associations reported by Sheppard et al. (1999).
0-day lag PM1 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).
Pneumonia (PM10= 50 ug/m3, LAG= ID, No Co
Poll):
Default GAM: ER= 21.5 (8.3, 36)
Strict GAM: ER=18.1 (5.3, 32.5)
NSGLM: ER=18.6 (5.6, 33.1)
COPD (PM10= 50 ug/m3, LAG= 3D, No Co Poll):
Default GAM: ER= 9.6 (-5.3, 26.8)
Strict GAM: ER=6.5 (-7.8, 23.0)
NS GLM: ER=4.6 (-9.4, 20.8)
COPD (PM25,=25 ug/m3, Lag=lD, No Co Poll):
Default GAM: ER=5.5 (-4.7, 16.8)
Strict GAM: ER=3.0(-6.9, 13.9)
NS GLM: ER=0.3(-9.3, 10.9)
Pneumonia (PM25,=25 ug/m3, LAG= lD,No Co
Poll):
Default GAM: ER = 12.5 (3.7, 22.1)
Strict GAM:ER = 10.5 (1.8, 19.8)
NSGLM: 10.1 (1.5, 19.5)
Respiratory HA's for persons <65 vrs. old
PMj = 25 ug/m3, no co-pollutant:
l-dlagER =
, 11.0)
o
HH
H
W
-------
to
O
o
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)
OO
td
i
to
fe
H
6
o
o
H
O
Moolgavkar et al. (2000)+
King County, WA (87 - 95)
Population = NR
PM10 mean = 30.0 ug/m3
PM10IQR =18.9-37.3 ug/m3
PM2 5 mean =18.1 ug/m3
PM25 IQR =10-23 ug/m3
Moolgavkar (2000a)*
Study Period: 1987-1995
Chicago (Cook County). IL
Population = NR
PM10 median = 35 ug/m3
PM10 IQR = 25-47 ug/m3
Los Angeles (LA County), CA
Population = NR
PM10 median = 44 ug/m3
PM10 IQR = 33-59 ug/m3
PM2 5 median = 22 ug/m3
PM25IQR= 15-31 ug/m3
Phoenix (Maricopa County), AZ
Population = NR
PM10 median = 41 ug/m3
PM10 IQR = 32-51 ug/m3
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. PM2 5 only had one monitoring
site versus multiple sites averaged for other
pollutants.
Investigated associations between air pollution
(PM10, 03, S02, N02, 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. Analyses employed 30df
to fit long wave. 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.
Of the PM metrics, PM10 showed the most
consistent associations across lags (0-4 d).
PM2 5 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.
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
COPD HA's all ages (no co-pollutant)
PM10 (50 ug/m3, lag 2)
ER = 5.1%(CI: 0, 10.4)
PM2.5 (25 ug/m3, lag 3)
ER = 6.4%(CI: 0.9, 12.1)
COPD HA's all ages (CO as co-pollutant)
PM10 (50 ug/m3, lag 2)
ER = 2.5%(CI: -2.5,7.8)
PM2.5 (25 ug/m3, lag 3)
ER = 5.6%(CI: 0.2, 11.3)
Most Significant Positive ER
Single Pollutant Models:
COPD HA's (>64vrs.) (50 ug/m3 PM10):
Chicago: Lag 0 ER =2.4% (CI: -0.2, 4.3)
LA: Lag2ER = 6.1%(CI: 1.1, 11.3)
Phoenix: Lag 0 ER = 6.9% (CI: -4.1,19.3)
LA COPD HA's
(50 ug/m3 PM10, 25 ug/m3 PM2 5 or PM2 5.10)
(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.): PM(25.10) 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.): PM25.10 lg2=9%(CI: 3,15.3)
(>64yrs): PM10 Ig2 = 6.1% (1.1, 11.3)
(> 64 yrs): PM25 Ig2 = 5.1% (0.9, 9.4)
(>64yrs.): PM25.10 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
HH
H
W
-------
to
O
o
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
i
to
OO
fe
H
6
o
o
H
O
United States (cont'd)
Reanalysis by Moolgavkar (2003)
Sheppard et al. (1999)*
Seattle, WA, Pop. = NR
1987-1994
PM10 mean = 31.5 ug/m3
PM10IQR = 19-39 ug/m3
PM25 mean = 16.7 ug/m3
PMi5 IQR = 8-21 ug/m3
PM25.10 mean = 16.2 ug/m3
PMi5.10 IQR = 9-21 ug/m3
Re-analyses of Moolgavkar (2000a) with more
stringent GAM convergence criteria and
alternative models.
Daily asthma hospital admissions (HA's) for
residents aged <65 (mean=2.7/day) regressed on
PM10, PM25, PM25.10, 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 for PM.
GAM effect estimates virtually unchanged from
originals using when GAM stringent criteria
applied in LA (direct comparisons not possible
in Chicago). In LA, changes in spline degrees
of freedom had much more influence on effect
size than the change in convergence criteria,
especially for PM10. In Chicago, small
insignificant association of PM10 in the original
work actually increased and became significant
with the lOOdf model. Authors conclude the
"basic qualitative conclusions unchanged".
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.
LA COPD (all ages), LAG= 2D, PM10 =50ug/m3
Default GAM:30df** ER= 7.36% (CI:4.32-11.39)
Strict GAM:30df ER= 7.78% (CI:4.32-10.51)
Strict GAM: lOOdf ER = 7.78% (CI:4.32-10.51)
NS GLM: lOOdf ER=5.00% (CL1.22, 8.91)
LA COPD (all ages), LAG=2D, PM2 5 =25 ug/m3
Default GAM:30df** ER=4.82% (CL2.44, 7.25)
Strict GAM:30df ER=4.69% (CL2.06, 7.38)
Strict GAM: lOOdf ER=2.87% (CL0.53, 5.27)
NS GLM: lOOdf ER=2.59% (CL-0.29, 5.56)
Chicago COPD (>64yrs) LAG= OD, PM10=50ug/m3
Default GAM (30df) ER =2.4% (CI: -0.2, 4.3)
Default GAM (lOOdf) not provided for comparison
Strict GAM (lOOdf) ER=3.24% (CL0.031-6.24)
Asthma Admissions (ages 0-64)
PM10 (lag=lday); 50 ug/m3
ER = 13.7% (CI: 5.5%, 22.6)
PM25 (lag=lday); 25 ug/m3
ER = 8.7%(CI: 3.3%, 14.3)
PM25.10(lag=lday);25ng/m3
ER=11.1%(CI: 2.8%, 20.1)
o
HH
H
W
-------
to
O
o
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
i
to
VO
United States (cont'd)
Reanalysis by Sheppard (2003)
Freidman et al. (2001)
Atlanta, GA
Summer 1996/control vs. Olympics
PM10 decrease for 36.7 ug/m3 to
30.8 ug/m3
Re-analyses of Sheppard et al. (1999) with more
stringent GAM convergence criteria and
alternative models.
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 after
the Olympics. GEE Poisson regression with
autoregressive terms employed.
The author notes that "While the biases from
computational details of the fitting were small,
they are not completely trivial given the small
effects of interest." She concludes that:
"Overall the results did not change
meaningfully".
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.
Asthma (ages 0-64) LAG=lday, PM10=50 ug/m3
No Co-Poll:
Default GAM: ER = 13.7% (CI: 5.5%, 22.6)
Strict GAM: ER= 8.1 (0.1, 16.7)
NS GLM : ER=10.9 (2.8, 19.6)
Asthma (all ages) LAG=lday, PM25=25 ug/m3
No Co-Poll:
Default GAM : ER= 8.7% (3.3, 14.3)
Strict GAM: ER=6.5% (1.1,12.0)
NSGLM: ER= 8.7% (3.3,14.4)
With Co-poll:
Strict GAM: ER=6.5 (2.1, 10.9)
NS GLM: ER=6.5 (2.1, 10.9)
3 day cumulative exposure PM10
per 10 |ig/m3
1.0(0.80-2.48)
fe
H
O
O
o
H
O
Zanobetti and Schwartz (2001)+
Cook County, Illinois
1988-1994
PM10: 33 ng/m3 median
Janssen et al. (2002)+
14 U.S. cities
1985-1994
see Samet et al. (2000a,b)
Respiratory admissions for lung disease in
persons with or without diabetes as a
co-morbidity related to PM10 measures. The
generalized additive model used nonparametric
LOESS functions to estimate the relation
between the outcome and each predictor. The
covariates examined were temperature, prior
day's temperature, relative humidity, barometric
pressure, and day of week.
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).
Weak evidence that diabetes modified the risks COPD
of PM10 induced respiratory hospital admissions PM10
while diabetes modified the risk of PM10 10 ug/m3
induced COPD admissions in older people. with diabetes
Found a significant interaction with hospital 2.29 (-0.76-5.44)
admissions for heart disease and PM with more without diabetes
than twice the risk in diabetics as in persons 1.50 (0.42-2.60)
without diabetes.
Regression coefficients of the relation between
ambient PM10 and hospital admissions for
COPD decreased with increasing percentage of
homes with central AC.
o
HH
H
W
-------
to
O
o
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
Canada (cont'd)
Burnett etal. (1997b)
Toronto, Canada (1992-1994),
Pop. = 4 mill.
PM25 mean = 16.8 ug/m3
PM2 5IQR = 8-23 ug/m3
j mean =11.6 ug/m
PM2
PM25.10IQR = 7-14ng/m3
PM10 mean = 28.4 ug/m3
PM10 IQR = 16-38 ug/m3
CoH mean = 0.8 (per 103 lin. ft.)
CoH IQR = 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:
FPe!t (ng/m3): 18, 16, 10
CPea (ng/m3): 12, 10, 8
PM10 e!t (ng/m3): 30,27, 15
Burnett etal. (1997c)
16 Canadian Cities('81-91)
Population=12.6 MM
CoH mean=0.64(per 103 lin. ft)
CoH IQR=0.3-0.8(per 103 lin ft)
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). Both linear
prefiltering Poisson regression and LOESS
GAM models applied. Daily particle measures:
PM2 5, coarse particulate mass(PM10_2 5), 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. Applied with 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.
Air pollution data were compared to respiratory
hospital admissions (mean=1.46/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 using a linear prefiltered
random effects relative risk regression model.
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, PM25.10, 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.
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.
Respiratory HA's all ages(no co-pollutant)
PM10 (50 ug/m3, 4d avg. lag 0)
ER= 10.6% (CI: 4.5- 17.1)
PM25 (25 ug/m3, 4d avg. lag 1)
ER = 8.5%(CI: 3.4, 13.8)
PM25.10(25ng/m3, Sdavg. lag 0)
ER=12.5%(CI: 5.2,20.0)
Respiratory HA's all ages(O, co-pollutant)
PM10 (50 ug/m3, 4d avg. lag 0)
ER = 9.6%(CI: 3.5, 15.9)
PM25 (25 ug/m3, 4d avg., lag 1)
ER = 6.2%(1.0, 11.8)
PM2.5.10 (25 ug/m3, 5d avg. lag 0)
ER= 10.8% (CI: 3.7, 18.1)
Percent excess risk (95% CI) per 50 ug/m3 PM1(
25 ug/m3 PM2 5 and PM(10.2 5):
Asthma
PM25 (0-1-2 d): 6.4(2.5, 10.6)
PM10(0-ld): 8.9(3.7, 14.4)
PM10.25(2-3-4d): 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)
Respiratory HA's all ages (with O,,CO)
CoH IQR = 0.5, lag 0:
CoHER = 3.1%(CI: 1.0-4.6%)
o
HH
H
W
-------
to
O
o
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
Canada (cont'd)
Burnett et al. (2001b)+
Toronto, Canada
1980-1994
PM25: 18 ug/m3
PM10.2.5: 16.2 ug/m3
(both estimated values)
Respiratory admissions in children aged <2 years Summertime urban air pollution, especially
PM2 5 lag 0
relates to mean pollution levels. O3, NO2, SO2, ozone, increases the risk that children less than 2 15.8% (t=3.29)
and CO
(ICD-9: 493 asthma; 466 acute bronchitis; 464.4
croup or pneumonia, 480-486). Time-series
analysis adjusted with LOESS.
years of age will be hospitalized for respiratory
disease.
PM2 5 lag 0
withOj 1.4% (0.24)
PM10.25lagl
18.3%(t=3.29)
with O3 4.5% (0.72)
oo
td
fe
H
6
o
o
H
O
Europe
Atkinson et al. (1999a)
London (92 - 94)
Population = 7.2 MM
PM10 Mean = 28.5
lO'-go"1 IQR = 15.8-46.5 ug/m3
BS mean =12.7 ug/m3
lO'-gO"1 IQR = 5.5-21.6 ug/m3
Wordley et al. (1997)
Study Period: 4/92-3/94
Birmingham, UK
Population = NR
PM10 daily values:
Mean = 25.6 ug/m3
range = 2.8, 130.9 ug/m3
PM10 3 day running, mean:
Mean = 25.5 ug/m3
range = 7.3, 104.7 ug/m3
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. I/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 GLM regression used,
controlling for season, day-of-week,
meteorology, autocorrelation, overdispersion,
and influenza epidemics.
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 log-
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.
Positive associations found between respiratory-
related emergency hospital admissions and PM10
and SO2, but not for O3 or BS. 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 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
ug/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".
PM10 (50 ug/m3), no co-pollutant.
All Respiratory Admissions:
All age (lag Id) ER = 4.9% (CI: 1.8, 8.1)
0-14y(lagld)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)
0-14 y (lag 3d) 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+vrs.)
(lag 3d) ER = 8.6% (CI: 2.6, 15)
Lower Respiratory Admissions (65+ yrs.)
(lag 3d) ER = 7.6% (CI: 0.9, 14.8)
50 ug/m3 in PM10
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)
o
HH
H
W
-------
to
O
o
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)
Prescott et al. (1998)
Edinburgh (10/92-6/95)
Population = 0.45 MM
PM10 mean. =20.7 ug/m3
PM10 min/max=5/72 ug/m3
PM10 90*% - 10th% = 20 ug/m3
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.
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.
Single Pollutant Models
PM10 = 50 ug/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)
OO
td
i
OJ
to
fe
H
6
o
o
H
O
McGregor et al. (1999)
Birmingham, UK.
Population = NR
Mean PM10 = 30.0 ug/m3
Hagen et al. (2000)+
Drammen, Sweden(l 1/94-12/97)
Population = 110,000
PM10 mean = 16.8 ug/m3
PM10IQR = 9.8-20.9 ug/m3
Dabetal. (1996)
Paris, France (87 - 92)
Population = 6.1 MM
PM13 mean = 50.8 ug/m3
PM13 5'h-95'h range = 19.0-137.3
BSmean = 31.9ng/m3
BS 5
-------
to
O
o
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
Europe (cont'd)
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
Diaz et al. (1999)
Madrid (94 - 96)
Population = NR
TSP mean 40 ug/m3
Spix et al. (1998)
London (L) (87-91)
Pop. =7.2 Million (MM)
BS Mean= 13 ug/m3
Amsterdam (A) (77 - 89)
Pop. =0.7 MM
BS Mean = 6 ug/m3
TSP mean = 41 ug/m3
Rotterdam (R) (77 - 89)
Pop. =0.6MM
BS Mean = 22 ug/m3
TSP mean = 41 ug/m3
Paris (P) (87 - 92),
Pop.= 6.14MM
BS Mean = 26 ug/m3
Milano (M) (80 - 89)
Pop. = 1.5 MM
TSP Mean =120 (ug/m3)
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), l.l(R).
Poisson GLM 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.
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 GLM 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.
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.
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."
BS (25 ug/m3) Id lag, no co-pollutant:
All Age COPD Hospital Admissions
ER= 1.7% (0.5, 2.97)
TSP (100 ug/m3) Id lag, no co-pollutant:
All Age COPD Hospital Admissions
ER = 4.45%(CI: -0.53,9.67)
N/A
Respiratory Admissions (BS = 25 ug/m3)
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) (100 ug/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 ug/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 ug/m3)
15-64 yrs: 2,0% (0.8, 3.2)
65+yrs: 0% (-2.2, 2.3)
TSP (A, R, M): Warm (100 ug/m3)
15-64yrs: 6.1%(0.1, 12.5)
65+yrs: 2.0% (-3.9, 8.3)
TSP (A, R, M): Cold (100 ug/m3)
15-64 yrs: -5.9% (-14.2, 3.2)
65+yrs: 4.0% (-0.9, 9.2)
o
HH
H
W
-------
to
O
o
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)
o
H
o
o
HH
H
W
Vigottietal. (1996)
Study Period.: 80-89
Milan, IT
Population =1.5 MM
TSP mean = 139.0 ug/m3
TSPIQR = 82.0, 175.7 ug/m3
Association between adult respiratory HA's
(15-64 yr mean =11.3/day, and 65 +yrmean
=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
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.
Young Adult (15-64 yrs.) Resp. HA's
100 ug/m3 increase in TSP
Lag 2 ER = 5% (CI: 0, 10)
Older Adult (65+ vrs.) Resp. HA's
100 |ig/m3 increase in TSP
Lag 1 ER = 5% (CI: -1, 10)
OO
td
Anderson et al. (1998)
London (87 - 92)
Population = 7.2 MM
BS daily mean = 14.6 ug/m3
BS 25-75"1 IQR = 24-38
Poisson GLM log-linear 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).
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.
Asthma Admissions. BS=25 ug/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 ng/m3, 2d lag & co-pollutant:
Older Adult (>64 vrs.) Asthma Visits:
BS alone: ER= 14.6% (2.7, 27.8)
&O3: ER = 20.0% (3.0, 39.8)
&NO2: ER = 7.4% (-8.7, 26.5)
S02: ER=11.8% (-2.2, 27.8)
fe
H
6
o
Kontos etal. (1999)
Piraeus, Athens GR (87 - 92)
Population = NR
BS mean =46.5 ng/m3
BS max =200 ug/m3
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.
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.
NR
-------
to
O
o
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
Europe (cont'd)
Ponce de Leon et al. (1996)
London (4/87-2/92)
Population = 7.3 million
BS mean. =14.6 ug/m3
BS 564 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.
Respiratory HA's (all ages)
Single Pollutant Models
For Oct-Mar. BS = 25 ug/m3
Lag 1 ER = 0.2% (-1.9, 2.3)
For Apr-Sep. BS = 25 ug/m3
Lag 1 ER = -2.7% (-6.0, 0.8)
Respiratory HA's (>65)
Single Pollutant Models
For Oct-Mar. BS = 25 ug/m3
Lag2ER= 1.2% (-2.1, 4.5)
For Apr-Sep. BS = 25 ug/m3
Lag 2 ER = 4.5% (-1.0, 10.4)
Single Pollutant Models
For BS=25 ug/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)
o
HH
H
W
-------
to
O
o
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
Europe (cont'd)
Sunyer et al. (1997)
Barcelona (86 - 92)
Population = NR
BS Median: 40 ug/m3
BS Range: 11-258 (B
Helsinki (86 - 92)
Population = NR
BS Median: -
BS Range: -
Paris (86 - 92)
Population = NR
BS Median: 28 ug/m3
BS Range: 4-186 ug/m3
London (86 - 92)
Population = NR
BS Median: 13 ug/m3
BS Range: 3-95 ug/m3
Teniasetal(1998)
Study Period.: 94-95
Valencia, Spain
Hosp. Cachment Pop. =200,000
BS mean = 57.7 ug/m3
BS IQR = 25.6-47.7 ug/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
Time-series GLM regression, controlling for
temperature and RH, viral epidemics, day of
week effects, and seasonal and secular trends
applied using the APHEA study approach.
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 GLM 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) with
BS. 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 ug/m3 BS (24 h Average)
Asthma Admissions/Visits:
<15yrs.:
London ER = 1.5% (Ig Od)
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% (Ig Od)
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)
Paris ER = 2.9% (Ig 2d)
Total ER= 1.8% (-0.6, 4.3)
<15yrs, (BS&S02):
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%(lg Id)
Total ER = -0.5% (-5.1, 4.4)
Adult Asthma HA's, BS = 25 ug/m3
For 1993-1995:
Lag 0 ER = 10.6% (0.9, 21.1)
For 1994-1995:
Lag 0 ER = 6.4% (-4.8, 18.8)
o
HH
H
W
-------
to
O
o
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
O
O
o
H
O
Europe (cont'd)
Anderson et al. (2001)
West Midland, England
(October 1994-December 1996)
Population = 2.3 million
PM10 mean = 23.3 ug/m3
PM25 mean =14.5 ug/m3
PM10.2.5 = 9.0 ug/m3
(by subtraction)
Respiratory hospital admissions (mean = 66/day)
related to PM10, PM25, PM10.25, BS, SO4", NO2,
O3, SO2, CO. GLM regression with quasi-
likelihood 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_2 5
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 ug/m3)
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(4.1to 15.2)
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)
Age 65: -1.8 (-6.9 to 3.5)
PM,,
Age 65: -3.9 (-9.0 to 1.6)
PMlO-2.5.
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)
PM,.
Ages 0-14: 6.0 (-0.9 to 13.4)
Ages 15-64: -8.4 (-16.4 to 0.3)
PMlO-2.5
Ages 0-14: 7.1 (-2.1 to 17.2)
Ages 15-64: -10.7 (-19.9 to-0.5)
o
HH
H
W
-------
to
O
o
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
i
OJ
OO
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 21.5 ug/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 ug/m3 (11.3, 130.8)
Paris 1/92 - 9/96
PM1020.1 ug/m3 (5.8, 80.9)
Rome - No PM10
Stockholm 3/94 - 12/96
PM10 13.6 ug/m3 (4.3, 43.3)
Thompson et al. (2001) Belfast,
Northern Ireland 1/1/93 - 12/31/95.
PM10 |ig/m3 mean (SD)
May - October 24.9 (13.7)
November-April 31.9 (24.3)
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)
regressed using GAM, controlling for
environmental factors and temporal patterns.
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 GLM Poisson regression.
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.
For 10 ug/m3 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)
fe
H
6
o
o
H
O
Fusco et al. (2001)+ Rome, Italy
1995-1997
PM - suspended particles measured
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 GAM
models controlling for mean temperature,
influenza, epidermics, and other factors using
spline smooths.
No effect was found for PM. Total respiratory
admission 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.
o
HH
H
W
-------
to
O
o
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
i
OJ
VO
fe
H
6
o
o
H
O
Lflrin America
Bragaetal. (1999)
Sao Paulo, Brazil (92 - 93)
Population = NR
PM10 mean = 66.3 ug/m3
PM10 Std. Deviation = 26.1
PM,n 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 ug/m3
PM10IQR = 42.9-75.5 ug/m3
PM1010/90'h%=98.1 ug/m3
PM1095th%= 131.6ng/m3
Rosas etal. (1998)
SW Mexico City (1991)
Population = NR
PM10 mean. =77 ug/m3
PM10 min/max= 25/183 ug/m3
Pediatric (<13 yrs.) hospital admissions
(mean=67.6/day) to public hospitals serving 40%
of the population were regressed (using both
GLM and GAM) 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, and NO2.
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.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.
PM10 and O3 were the two pollutants found to
exhibit the most robust associations with
respiratory HA's. SO2 showed no correlation at
any lag. Simultaneous regression of respiratory
HA's on PM10, O3, and CO decreased effect
estimates and their significance, suggesting that
"there may not be a predominance of any one
pollutant over the others". Associations
ascribed primarily to auto emissions by the
authors.
Children's HA's for total respiratory and
pneumonia positively associated with O3, NO2,
and PM10. Effects for pneumonia greater than
for all respiratory diseases. Effects on infants
(<1 yr. old) gave higher estimates. Similar
results for asthma, but estimates higher than for
other causes. Results noted to agree with other
reports, but smaller RR's. This may be due to
higher baseline admission rates in this poor sub-
population vs. other studies, but this was not
intercompared by the authors.
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.
PM10 (50 ug/m3), no-co-pollutant
Respiratory Hospital Admissions (<13 yr.)
GLM Model:
(0-5day Ig avg.) ER = 8.9% (CI: 4.6, 13.4)
GAM Model
(0-5day Ig avg.) ER = 8.3% (CI: 4.1, 12.7)
PM10 = 50 ug/m3:
All Respiratory HA's for children < 5vrs.
ER = 2.0% (-0.8, 4.9)
Pneumonia HA's for children <5 yrs.
ER = 2.5% (-0.8, 6.0)
Asthma HA's for children <5 vrs.
ER = 2.6% (-4.0, 9.7)
Pneumonia HA's for children <1 yrs.
ER = 4.7% (0.7, 8.8)
NR
o
HH
H
W
-------
to
O
o
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
Australia
Morgan etal. (1998)
Sydney, AU (90 - 94)
Population = NR
PM2 5 24 h mean = 9.6 ug/m3
PM25 10th-90th% = 3.6-18 ug/m3
PM25 max-1 h mean = 22.8 ug/m3
PM25 lO'-gO^/o = 7.5-44.4 ug/m3
A Poisson analysis, controlled for overdispersion
and autocorrelation via generalized estimating
equations (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.
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 PM2 5. 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.
Asthma HA's
Single Pollutant Model:
For 24 hr PM2 5 = 25 ug/m3
1-14 yrs.(lagl) ER = -1.5% (CI: -7.8, 5.3)
15-64 yrs.(lagO) ER = 2.3% (CI: -4, 9)
For Ih PM25 =25 ug/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:
For 24h PM2 5 = 25 ug/m3
1-14 yrs.(lagl) ER = -0.6% (CI: -7.4, 6.7)
COPD (65+yrs.)
Single Pollutant Model:
For24hPM25 = 25 ug/m3
(lagO)ER=4.2%(CI: -1.5,10.3)
For lhPM25 = 25 ug/m3
(lag 0) ER = 2% (CI: -0.3,4.4)
Multiple Pollutant Model:
For Ih PM25 = 25 ug/m3
(lag 0) ER = 1.5% (CI: -0.9, 4)
Asia
fe
H
6
o
o
H
O
Tanakaetal. (1998)
StdyPd.: 1/92-12/93
Kushiro, Japan
Pop. = 102 adult asthmatics
PM10 mean = 24.0 ug/m3
PM10IQR = NR
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% CI's
calculated between high and low days for each
environmental variable. Poisson GLM
regression was performed for the same
dichototomized variables.
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-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.
For same-day (lag=0) PM10
Adult Asthma HA's
OR for <30 vs. >30 ug/m3 PM10:
Non-atopic OR = 0.77 (CI: 0.61,0.98)
Atopic OR = 0.87 (CI: 0.75, 1.02)
Poisson Coefficient for PM10 > 30 ug/m3
Non-atopic = -0.01 (SE = 0.15)
Atopic = -0.002 (SE = 0.09)
o
HH
H
W
-------
to
O
o
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
Asia (cont'd)
Wong et al. (1999a)
Study Period.: 94-95
Hong Kong
Population = NR
PM10 mean = 50.1 ug/m3
PM10 median = 45.0 ug/m3
PM10IQR = 30.7, 65.5 ug/m3
Poisson GLM regression analyses were applied
to assess association of daily NO2, SO2, O3, and
PM10 with emergency HA's for all respiratory
(median = 13 I/day) and COPD (median =
10 I/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.
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
PM10 = 50 ug/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: 5b.l, 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)
OO
td
fe
H
o
o
o
H
O
+ = Used GAM with multiple smooths, but have not yet reanalyzed. * = Used S-Plus Default GAM, and have reanalyzed results.
GAM=Generalized Additive Model, GLM=Generalized Linear Model; NS= Natural Spline, PS=Penalized Spline.
o
HH
H
W
-------
Appendix 8B.3: PM-Respiratory Visits Studies
June 2003 8B-42 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 8B-3. ACUTE PARTICULATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
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
OO
td
fe
H
6
o
o
H
O
Choudhury et al. (1997)
Anchorage, Alaska (90 - 92)
Population = 240,000
PM10 mean = 41.5 ug/m3
PM10 (SD) = 40.87
PM10 maximum=565 ug/m3
Lipsettetal. (1997)
Santa Clara County, CA
Population = NR
(Winters 88 - 92)
PM10 mean = 61.2 ug/m3
PM10 Min/Max = 9/165 ug/m3
Norris et al. (1999)+
Seattle, WA (9/95-12/96)
Pop. Of Children <18= 107,816
PM10 mean. =21.7 ug/m3
PM10IQR=11.6ng/m3
sp mean = 0.4m 1/104
( 12.0 ug/m3 PM25)
!p IQR = 0.3 m 1/10 4
(= 9.5 ug/m3 PM2.5)
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 GLM 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 GLM
model with long-wave, day of week, holiday, and
weather controls (analysis stratified by minimum
T). Analyses using GAM also run for
comparison. 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 (sp), CO, SO2, and
NO2 using a semiparametric S-Plus Poisson
regression model with spline smooths for season
and weather variables, 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. For example, .GAM
model gave simi.lar, though sometimes less
significant, results. 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/lOK population indicates a higher PM
attributable risk (AR) in the inner city.
Asthma Medical Visits (all ages):
For mean = 50 ug/m3 PM10 (single poll.)
Lag = 0 days
ER = 20.9%(CI: 11.8,30.8)
Asthma ED Visits (all ages)
PM10 = 50 ug/m3 (2 day lag):
GLM Results:
At 20 F, ER = 34.7% (CI: 16,56.5)
At 30 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 ug/m3
Lagl ER = 75.9% (25.1, 147.4)
For 25 ug/m3 PM2 5
Lagl ER = 44.5% (CI: 21.7, 71.4)
Multiple Pollutant Models:
24h PM10 =50 ug/m3
Lagl ER = 75.9% (CI: 16.3, 166)
For 25 ug/m3 PM2 5
Lagl ER = 51.2% (CI: 23.4, 85.2)
o
HH
H
W
-------
fe
H
6
o
o
H
O
o
HH
H
W
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
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)
Norris et al. (2000)+
Spokane, WA (1/95 - 3/97)
Population = 300,000
PM10 mean. = 27.9 ug/m3
PM10 Min/Max =4.7/186.4 ug/m3
PM10 IQR = 21.4 ug/m3
Seattle, WA (9/95 - 12/96)
Pop. Of Children <18 = 107,816
PM10 mean. = 21.5 ug/m3
PM10 Min/Max = 8/69.3 ug/m3
PM10IQR= 11.7ng/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=1.8/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.
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.
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 ug/m3
Lag 3 ER = 56.2% (95 CI: 10.4,121.1)
OO
td
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) ug/m3
PM10 Range = 9, 105 ug/m3
Pediatric (<17 yrs. of age) ED visits (mean =
467/day) related to air pollution (PM10, O3, NOX,
pollen and mold) using GEE and logistic
regression and Bayesian models.
Autocorrelation, day of week, long-term trend
terms, and linear temperature controls included.
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).
Pediatric (<17 yrs. of age) ED Visits
PM10 = 50 ug/m3
Lag 1 day ER = 13.2% (CI: 1.2, 26.7)
With 03 8.2 (-7.1,26.1)
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
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
b
O
o
H
O
United States (cont'd)
Tolbert et al. (2000a)
Atlanta
Period 1: 1/1/93-7/31/98
Mean, median, SD:
PM10(ug/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 (ug/m3): 19.4, 17.5, 9.35
CP (ug/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 (ug/m3): 0.0327,
0.0226, 0.0306
PM25 Sulfates (ug/m3): 5.59, 4.67, 3.6
PM25 Acidity (ug/m3): 0.0181, 0.0112,
0.0219
PM25 organic PM (ug/m3): 6.30,5.90,
3.16
PM2 5 elemental carbon (ug/m3): 2.25,
1.88, 1.74
Yang et al (1997)
Study Period: 92 - 94
Reno-Sparks, Nevada
Population = 298,000
PM10 mean = 33.6 ug/m3
PM10 range = 2.2, 157.3 ug/m3
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 GLM regression analyses
were conducted with cubic splines for time,
temperature and dewpoint. Day-of-week and
hospital entry/exit indicators also included.
Pollutants
Association between asthma ER visits (mean =
1.75/d, SD=1.53/d) and PM10, CO and O3
assessed using linear WLS and ARIMA GLM
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.
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 PM25.
These preliminary results should be interpreted
with caution given the incomplete and variable
nature of the databases analyzed.
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.
Period 1:
PM10(0-2d):
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)
NR
o
HH
H
W
-------
to
O
o
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
Reference/Citation, Location, Duration,
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
Canada
o
HH
H
W
Delfmo et al. (1997)
Montreal, Canada
Population= 3 million
6-9/92, 6-9/93
1993 Means (SD):
PM10=21.7ug/m3(10.2)
PM25= 12.2 ug/m3 (7.1)
S04'= 34.8 nmol/m3 (33.1)
H+= 4 nmol/m3 (5.2)
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 GLM regression
with controls for temporal trends, auto-
correlation, and weather. Five age sub-groups
considered.
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 O3.
Respiratory ED Visits
Adults >64: (pollutant lags = 1 day)
50 ug/m3 PM10ER = 36.6% (10.0, 63.2)
25 ug/m3 PM25 ER = 23.9% (4.9, 42.8)
OO
td
fe
H
6
o
o
H
O
Delfmo et al. (1998)
Montreal, Canada
6-8/89,6-8/90
MeanPM10= 18.6 ug/m3
(SD=9.3, 90th% = 30.0 ug/m3)
Stiebetal. (1996)
St. John, New Brunswick, Canada
Population = 75,000
May-Sept. 84 - 92
SO42'Mean = 5.5 ug/m3
Range: 1-23, 95th% =14 ug/m3
TSP Mean = 36.7 ug/m3
Range:5-108, 95*% =70 ug/m3
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 GLM
regressions.
Asthma ED visits (mean=1.6/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 GLM
regressed on pollution and weather variables for
the same day and the 3 previous days.
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.
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".
Older Adults(>64 yr) Respiratory ED Visits
Estimated PM25 = 25 ug/m3
Single Pollutant:
(lag 1 PM25) ER = 13.2 (-0.2, 26.6)
With Ozone (lag 1 PM25):
Est. PM25 (lagl) ER = 0.8% (CI: -14.4, 15.8
Emergency Department Visits (all ages)
Single Pollutant Model
100 ug/m3 TSP = 10.7% (-66.4, 87.8)
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
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
o
o
o
H
O
Canada (cont'd)
Stieb et al. (2000)+
Saint John, New Brunswick, 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
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
Europe
Atkinson et al. (1999b)
London (92 - 94)
Population = NR
PM10 Mean = 28.5 ug/m3
lO'-go"1 IQR = 15.8-46.5 ug/m3
BS mean =12.7 ug/m3
lO'-gO"1 IQR = 5.5-21.6 ug/m3
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.
All-age Respiratory (mean=90/day), Asthma
(25.9/day), and Other Respiratory (64. I/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 GLM regression used, controlling for
season, day of week, meteorology,
autocorrelation, overdispersion, and influenza
epidemics.
In single-pollutant models, significant positive
associations were observed between all
respiratory ED visits and PM10, PM2 5, 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.
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).
PM25, (lag 3) 15.1 (-0.2,32.8)
PM10, (lag 3) 32.5 (10.2, 59.3)
PM10 (50 ug/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 (Ig Id) ER = 13% (CI: 4.6, 22.1)
PM10(50 |ig/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)
&SO2: ER = 8.1% (CI: -1.1, 18.2)
&CO: ER = 12.1% (CI: 3.2,21.7)
o
HH
H
W
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
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
-k
OO
fe
H
6
o
o
H
O
Europe (cont'd)
Hajatetal. (1999)
London, England (92 - 94)
Population = 282,000
PM10 mean = 28.2 ug/m3
PM10 10'-90'h%=16.3-46.4 ug/m3
BS mean = 10.1 |ig/m3
BS 10'-90'h%=4.5-15.9ug/m3
Hajatetal. (2001)+
London (1992-1994)
44,406-49,596 registered patients <1 to
14 years
PM10 mean 28.5 (13.9)
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.
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.
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.
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.
Asthma Doctor's Visits:
50 ug/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)
15-64yrs.(lgO): 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/03: ER = 5.5%(-2.1, 13.8)
W/SO2: ER = 3.2%(CI: -6.4,13.7)
Other Lower Resp. Pis. Doctor's Visits:
50 ug/m3 PM10
-Year-round, Single Pollutant:
All ages (Ig 2): ER = 3.5%(CI: 0,7.1)
0-14 yrs.(lg 1): ER = 4.2% (CI: -1.2, 9.9)
15-64yrs.(lg2): ER=3.7%(CI: 0.0,7.6)
>64yrs.(lg 2): ER = 6.2% (CI: 0.5, 12.9)
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)
o
HH
H
W
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
Reference/Citation, Location, Duration,
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
o
HH
H
W
Europe (cont'd)
Medina etal. (1997)+
Greater Paris 91 - 95
Populations 6.5 MM
Mean PM13 = 25 ug/m3
PM13 mill/max = 6/95 ug/m3
MeanBS = 21 ug/m3
BS min/max = 3/130 ug/m3
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.
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.
Doctor's Asthma House Visits:
50 ug/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)
15-64yrs.(lg2): ER = 6.3%(CI: -4.6,18.5)
OO
td
Damiaetal. (1999)
Valencia, Spain (3/94-3/95)
Population = NR
BS mean =101 ug/m3
BS range = 34-213 ug/m3
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 GLM 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.
Both BS and SO2 correlated with ED
admissions for asthma (SO2: r=0.32; BS:
r=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.
Asthma ED Visits (all ages):
BS = 40 ug/m3 (single pollutant)
BS as a lag 0 weekly average:
ER = 41.5% (CI = 39.1,43.9)
fe
H
6
o
o
H
O
Pantazopoulou et al. (1995)
Athens, GR (1988)
Population = NR
Winter (1/88-3/88,9/88-12/88)
BS mean. =75 ug/m3
BS S'-gS"1 %=26 - 161 ug/m3
Summer (3/22/88-3/88,9/21/88)
BS mean. =55 ug/m3
BS S'-gS"1 %=19 - 90 ug/m3
Examined effects of air pollution on daily
emergency outpatient visits and admissions for
cardiac and respiratory causes. Air pollutants
included: BS, CO, and NO2. Multiple linear
GLM 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.
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.
Single Pollutant Models
For Winter (BS = 25 ug/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 ug/m3)
Outpatient Hospital Visits
ER = 0.6% (-4.7, 6.0)
Respiratory HA's
ER = 5.5% (-3.6, 14.7)
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
Reference/Citation, Location, Duration,
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
o
HH
H
W
Europe (cont'd)
Gartyetal. (1998)
PM10 mean 45 ug/m3
Tel Aviv, Israel (1993)
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.
No PM10 associations found with ED visits. N/A
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.
oo
td
fe
H
6
o
o
H
O
Latin America
Ilabacaetal. (1999)
Santiago, Chile
February 1995-August 1996
PM10: warm: 80.3 ug/m3
cold: 123.9 ug/m3
PM25: warm: 34.3 ug/m3
cold: 71.3 ug/m3
Number of daily respiratory emergency visits
(REVs) related to PM by Poisson GLM model
with longer- and short-term trend terms. SO2,
N02, 03.
Stronger coefficients for models including
PM2 5 than for models including PM10 or
PM10_25. 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
PM2 5, lag 2
OR: 1.027 (1.01 to 1.04) for a 45 ug/m3 increment
PM10, lag 2
OR: 1.02 (1.01 to 1.04) for a 76 ug/m3 increment
PM2.5,lag2
OR: 1.01 (1.00* to 1.03) for a 32 ug/m3 increment
Pneumonia, lag 2
PM10: 1.05 (1.00* to 1.10)
64 |ig/m3 increment
PM25: 1.04(1.00*to 1.09)
45 ug/m3 increment
PM10.25: 10.5 (1.00* to 1.10)
32 |ig/m3 increment
'decimals <1.00
-------
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
to
O
o
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
Latin America (cont'd)
Lin et al. (1999)
Sao Paulo, BR (91-93)
Population=NR
PM10 mean =65 ug/m3
PM10 SD=27 ug/m3
PM10range=15-193 ug/m3
Ostro et al. (1999b)+
Santiago, CI (7/92—12/93)
<2 yrs. Population 20,800
3-14 yrs. Population 128,000
PM10 mean. =108.6 ug/m3
PM10 Min/Max=18.5/380 ug/m3
PM10IQR = 70.3 - 135.5 ug/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 various
GLM models: 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
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 PM10-LRS associations.
50 ug/m3 PM10 (0-5-day lag mean)
Respiratory ED Visits (<13 vrs.)
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)
Lower Resp. Symptoms Clinic Visits
PM10 = 50 ug/m3
Single Pollutant Models:
-Children<2 years
Lag 3 ER = 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)
fe
H
6
o
o
H
O
o
HH
H
W
-------
to
O
o
TABLE 8B-3 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY MEDICAL VISITS
Reference/Citation, Location, Duration,
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
Australia
OO
td
i
(^
to
fe
H
6
o
o
H
O
Smith et al. (1996)
StdyPd.: 12/92-1/93,12/93-1/94
West Sydney, AU
Population = 907,000
-Period 1 (12/92-1/93)
Bscatt median = 0.25 10 4/m
BscattIQR = 0.18-0.39 10 4/m
Bscas 95th% = 0.86 10 4/m
-Period 2 (12/93-1/94)
Bscatt median = 0.19 10 4/m
B!cattIQR = 0.1-0.38 10 4/m
Bscatt 95th5% = 3.26 10 4/m PM10 median
= 18 ug/m3
PM10IQR = 11.5-28.8 ug/m3
PM10 95*% = 92.5 ug/m3
Asia
Yeetal. (2001)
Tokyo, Japan
Summer months
July-August, 1980-1995
PM10 46.0 mean
Chew etal. (1999)
Singapore (90 - 94)
Population = NR
TSP mean = 51.2 ug/m3
TSP SD = 20.3 ug/m3
TSP range = 13-184 ug/m3
Study evaluated whether asthma visits to
emergency departments (ED) in western Sydney
(meanlO/day) increased as result of bushfire-
generated PM ( Bscatt from nephelometry) in Jan.,
1994 (period 2). Air pollution data included
nephelometry (Bsciltt), PM10, SO2, and NO2. Data
analyzed using two methods: (1) calculation of
the difference in proportion of all asthma ED
visits between the time periods, and; (2) Poisson
GLM regression analyses. Control variables
included T, RH, BP, WS, and rainfall.
Hospital emergency transports for respiratory
disease for >65 years of age were related to
pollutant levels NO2, O3, PM10, SO2, and CO.
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 GLM
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, mean
HA=3.0/day).
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. Bsciltt 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.
For chronic bronchitis PM10 with a lag time of
2 days was the most statistically significant
model covariate.
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.
ED Asthma Visits (all ages)
Percent change between bushfire and non bushfire
weeks:
PM10 = 50 ug/m3
ER = 2.1%(CI: -0.2,4.5)
Asthma (ICD-9-493)
Coefficienct estimate (SE)
0.003 (0.001)
TSP(100 ug/m3) No co-pollutant:
Child (3-13 yrs.)Asthma ED visits
Lag Id ER = 541% (CI: 198.4, 1276.8)
+ = Used GAM with multiple smooths, but have not yet reanalyzed.
= Used S-Plus Default GAM, and have reanalyzed results
o
HH
H
W
-------
Appendix 8B.4: Pulmonary Function Studies
June 2003 8B-53 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 8B-4. SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ug/m3
PM10 (25 ug/m3 PMM). 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
United States
Thurston et al. (1997)
Summers 1991-1993.
O3, H+, sulfate
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.
The Oj-APEFR relationship was seen as
the strongest.
Canada
oo
td
Vedaletal. (1998)
Port Alberni, BC
PM10 via a Sierra-Anderson dichotomous sampler. PM1(
ranged from 1 to 159 ug/m3.
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.
Ozone, SO2 and sulfate levels low due to
low vehicle emissions. PM10 associated
with change in peak flow.
Lag 0, PM 10 average
PEE = 0.27 (-0.54, -0.01) per 10 ug/m3
increment
Europe
fe
H
6
o
o
H
O
Gielen et al. (1997)
Amsterdam, NL
Mean PM10 level: 30.5 ug/m3 (16, 60.3).
Mean maximum 8 hr O3: 67 ug/m3.
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.
The strongest relationships were found
with ozone, although some significant
relationships found with PM10.
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)
o
HH
H
W
Hiltermann et al. (1998)
Leiden, NL
July-Oct, 1995
O3, NO2, SO2, BS, and PM10 ranged from 16.4 to
97.9 ug/m3)
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.
No relationship between ozone or PM10
and PET was found
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
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 PMM). 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
Europe (cont'd)
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
ultrafine 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 urn.
Mean PM10 level: 55 ug/m3 (max 71). Mean SO2:
100 ug/m3 (max 383).
PM was measured using a Harvard impactor. Particle
size distributions were estimated using a conduction
particle counter.
Peters et al. (1997c)
Sckolov, Czech Republic
Winter 1991-1992
PM10, SO2, TSP, sulfate, and particle strong acid.
Median PM10 level: 47 ug/m3 (29, 73).
Median SO2: 46 ug/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.
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.
Five day average SO2 was associated with
decreased PEF. Changes in PEF were not
associated with PM levels.
Strongest effects on peak flow found with
ultrafine 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)
Lag 0, 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)
LagO,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
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 PMM). 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
Europe (cont'd)
Timonen and Pekkanen (1997)
Kupio, Finland
PM10, BS, N02, and SO2.
The intequartile range on PM10 was 8 to 23.
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 ug/m3.
PM2 5 ranged from 2.4 to 38.3 ug/m3.
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 PM1: PM25, PM10, particle counts, CO,
NO, and
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 urn. No associations were
found between particulate pollution and
respiratory symptoms.
AM PEF = -.115 (-.448, .218) PM25 lag one
day
AM PEF = -.001 (-.334, .332) PM25 lag two
days
fe
H
O
O
o
H
O
O
H
W
O
Pekkanen etal. (1997)
Kuopio, Finland
PM fractions measured over range of sizes from
ultrafine to fine, including PM10.
Mean PM10 level: 18 ug/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 ug/m3 (range 4.4, 83.8).
Mean NO2 level: 56.9 ug/m3 (range 23.8, 121.9).
PM was measured by p-radiometry.
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.
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,.
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)
O
-------
rt>
to
o
o
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 PMM). 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
Europe (cont'd)
Gauvin et al. (1999)
Grenoble, France
Summer 1996, Winter 1997
Mean (SD) ug/m3
PM10 Summer 23 (6.7)
PM10 Winter 38 (17.3)
Sunday 15.55(5.12)
Weekday 24.03 (7.2)
Agocs et al. (1997)
Budapest, Hungary
SO2 and TSP were measured. TSP was measured by
beta reactive absorption methods.
Australia
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 FEVj and
PEE. Bronchial reactivity was compared
Sunday vs. weekday. Temperature and RH
controlled.
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
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.
For a 10 ug/m3 increase in PM10
Summer
FEVj
-1.25% (-0.58 to-1.92)
PEE
-0.87% (-0.1 to-1.63)
No significant TSP-PEFR relationships found.
fe
H
6
o
o
H
O
O
H
W
O
Jaulaludin et al. (2000)
Sydney, Austrlia
1 February 1994 to 31 December 1994
Six PM10 TEOM monitors
PM10 Mean - 22.8 ±13.9 ug/m3
(max 122.8 ug/m3)
Rutherford et al. (1999)
Brisbane, Australia
PM10, TSP, and particle diameter.
PM10 ranged form 11.4 to 158.6 ug/m3. Particle sizing
was done by a Coulter Multisizer.
Population regression and GEE models used a
cohort of 125 children (mean age of 9.6 years)
in three groups; two with doctor's diagnoses of
asthma. This study was designed to examine
effects of ambient O3 and peak flow while
controlling for PM10.
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.
In Syndey, O3 and PM10 poorly correlated
(0.13). For PM10 with O3, 0.0051
(0.0124) p-0.68 peak flow
The paired t-tests were stat. significant for
some days, but not others. No general
conclusions could be drawn.
PM10 only
B(SE) = 0.0045 (0.0125)
p-0.72 peak flow
o
-------
to
o
o
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 PMM). 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
i
u\
oo
fe
H
O
O
o
H
O
O
H
W
O
Latin America
Romieu et al. (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 ug/m3, SD 72.8 ug/m3).
For 53 percent of study days, PM10 levels exceeded
150 ug/m3. PM10 was measured by a Harvard impactor.
Romieu et al. (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 ug/m3.
PM10 was measured by a Harvard impactor.
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.
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.
Lag 0, PM10:
Evening PEE = -4.80 (-8.00, -1.70)
Lag 2, PM10:
Evening PEE = -3.65 (-7.20, 0.03)
Lag 0, 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)
O
-------
Appendix 8B.5: Short-Term PM Exposure Effects
On Symptoms in Asthmatic Individuals
June 2003 8B-59 DRAFT-DO NOT QUOTE OR CITE
-------
rt>
to
O
O
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 |ig/m3
PM10 (25 |ig/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
United States
Delfmo et al. (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 ug/m3
with a mean of 25.
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 PM2 5.
Pollen not associated with asthma
symptom scores. 12-hr personal O3 but
not ambient O3 related to symptoms.
No significant relationships with PM10.
OO
td
Delfmo et al. (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.
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
Although PM10 never exceeded 51
ug/m3, bronchodilator use was
significantly associated with PM10(0.76
[0.027, 0.27]) puffs per 50 ug/m3.
Fungal spores were associated with all
respiratory outcomes.
fe
H
6
o
o
H
O
Delfmo et al. (1998)
So. California community
Aug. - Oct. 1995
Highest 24-hour PM10 mean: 54 ug/m3.
PM10 and ozone PM was measured using a tapered-
element oscillating microbalance. PM10 ranged from
6 to 51 |ig/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. PMj „ ranged
from 2 to 62 ug/m3 with a mean of 10.4. PM10 9 to
86 |ig/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.
PM10 ranged from 21 to 119 ug/m3 with a mean of
51.8.
O
HH
H
W
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 fungi 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.
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 SO2 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.
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
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
United States (cont'd)
Delfmo et al. (2002)
PM10, ozone, NO2, fungi, pollen, temperature, relative
humidity
Mortimer et al. (2002)
Eight U.S. urban areas
Daily PM10 were collected in Chicago, Cleveland, and
Detroit with an average intra-diary range of 53 ng/m3
from the Aerometric Information Retrieval System of
EPA.
Thurston et al. (1997)
Summers 1991-1993.
O3, H+, sulfate, pollen, daily max temp, measured.
Canada
22 asthmatic children aged 9-19 were followed
March through April of 1996. Study used an asthma
symptom score.
Study of 846 asthmatic children in the eight urban
area National Cooperative Inner City Asthma study.
Peak flow and diary symptom data are the outcome
measures. Morning symptoms consist of cough,
chest tightness, and wheeze. Mixed linear and GEE
models were used.
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.
No relationship between PM1(
symptom score was found
and
In the three cities with PM10 data, a
stronger association was seen for PM10
than ozone for respiratory symptoms.
Ozone related to respiratory symptoms
No relationship between symptoms and
other pollutants.
LagO
Score OR =1.17 (0.53, 2.59)
3 Day moving average
Score OR = 1.49(0.71,2.59)
all for 50 ug/m3 increase in PM10
Morning symptoms
PM10 - 2day ave.
OR= 1.26(1.0-1.59)
fe
H
b
o
o
H
O
o
HH
H
W
Vedaletal. (1998)
PM10 measured by Sierra- Anderson dichotomous
sampler
PM10 range: - 1 to 159 ug/m3
Port Alberrni
British, Columbia
Europe
Gielen et al. (1997)
Amsterdam, NL
PM10 and ozone.
PM10 was measured using a Sierra- Anderson
dichotomous sampler. PM10 ranged from 15 to
60 ug/m3.
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.
PM10 associated with respiratory
symptoms.
Strongest relationships found with O3,
although some significant relationships
found with PM10.
LagO
Cough OR = 1.08(1.00, 1.16) per 10 ug/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)
-------
to
O
o
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
ON
to
fe
H
6
o
o
H
O
o
HH
H
W
Europe (cont'd)
Hiltermann et al. (1998)
Leiden, NL
July-Oct 1995.
Ozone, PM10, NO2, SO2, BS
PM10 ranged from 16 to 98 |ig/m3 with a mean of 40.
Hiltermann et al. (1997)
The Netherlands
Ozone and PM10
PM10 averaged 40 ug/m3,
Peters et al. (1997a)
Erfurt, Germany
PM fractions measured over range of sizes from
ultrafine 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. (1997b)
Sokolov, Czech Republic
Winter 1991-1992
PM10, SO2, TSP, sulfate, and particle strong acid.
Median PM10: 47 ug/m3 (29, 73).
Median SO2: 46 ug/m3 (22, 88).
PM was measured using a Harvard impactor. Particle
size distributions were estimated using a conduction
particle counter.
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 O3 were evaluated.
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.
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.
Weak associations found with 5 day
mean sulfates and respiratory
symptoms.
Significant relationships found between
TSP and sulfate with both phlegm and
runny nose.
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)
Lag 0, PM10:
Cough OR = 1.32(1.16, 1.50)
Feeling ill OR =1.2
., .
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, 1.8
LaO PM:
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)
-------
to
O
o
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
Europe (cont'd)
Peters et al. (1997c)
Sokolov, Czech Republic
PM10 one central site. SO4 reported.
Mean PM10: 55 |ig/m3, max 177 ug/m3.
SO4-fme: mean 8.8 |ig/m3, max 23.8 ug/m3. PM was
measured using a Harvard impactor. Particle size
distributions were estimated using a conduction
particle counter.
Neukirch et al. (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.
Segalaetal. (1998)
Paris, France
SO2, NO2, PM13 (instead of PM10), and BS.
PM was measured by p-radiometry.
Giintzeletal. (1996)
Switzerland
S02, N02, TSP
Taggartetal. (1996)
Northern England
SO2, NO2 and BS.
Just et al. (2002)
PM13, S02, N02, 03
o
HH
H
W
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.
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 ARIMAtime
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.
82 medically diagnosed asthmatic children living in
Paris, followed for 3 months. Study measured
asthma attacks and nocturnal cough, symptoms, and
PEF
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 urn, 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.
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.
PM13 was only associated with eye
irritation.
Cough 1.16 (1.00, 1.34) 6.5 ug/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.
Lag 2, Symptoms:
Short. Breath OR = 1.22 (0.83, 1.81)
Resp. Infect. OR = 1.66 (0.84, 3.30)
LagO
Asthma episodes OR = 1.34 (0.08, 20.52) for
50 ug/m3 PM13.
-------
to
O
o
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
HH
H
W
Von Klot et al. (2002)
PM25.10, PM10, NO2, SO2, CO, temperature
Desqueyroux et al. (2002)
PM10, 03, S02, andN02
Latin America
Romieu et al. (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.
Romieu et al. (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 ug/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 ug/m3 (mean
85.7 ug/m3)
PM was measured by a Harvard impactor.
53 adult asthmatics in Erfurt, Germany in the winter
1996/1997. Study measured inhaled medication use,
wheezing, shortness of breath, phlegm and cough
60 severe asthmatic adults in Paris were followed for
13 months. Study measured incident asthma attacks
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.
Medication use and wheezing were
associated with PM, ,.,„
Attacks were associated with PM10 for
lags 4 and 5 but not for lags 1, 2, and 3
Strongest relationships found between
O3 and respiratory symptoms.
Cough and LRI were associated with
increased O3 and PM10 levels.
5 Day mean
Corticosteroid use OR= 1.12 (1.04-1.20) for
12 ug/m3 PM2.5.10.
Wheezing OR = 1.06(0.98, 1.15) for 12
ug/m3 PM2 5.10.
Lagl
Attack OR = 0.50 (0.18, 1.34)
Lag 2
Attack OR = 0.67 (0.33, 1.47)
Lag 3
Attack OR =1.69 (0.90, 3.18)
Lag 4
Attack OR = 2.19 (1.16, 4.16)
Lag5
Attack OR = 2.10 (1.05, 4.32)
all for 50 ug/m3 increase in PM10
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 ug/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
June 2003 8B-65 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 8B-6. SHORT-TERM PARTICULATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
OO
td
ON
a\
fe
H
6
o
o
H
O
United States
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
Neasetal. (1996)
State College, PA
PM2 p 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 PM21
An autoregressive linear regression model was used. The
regression was weighted by reciprocal number of children of each
reporting period. Fungus spore cone., temp., O3 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 ug/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
HH
H
W
-------
to
O
o
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
o
HH
H
W
United States (cont'd)
oo
td
Neasetal. (1999)
Philadelphia, PA
Median PM10 level: 31.6 in SW camps,
27.8 in NE camps (IQR ranges of about 18).
Median PM25 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.
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 that included sulfate
fraction and O3 separately also
found relationship to decreased
flow. No analyses reported for
multiple pollutant models.
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., PM25
Morning PEE = 3.18 (-2.64, 9.02)
Evening PEE = 0.95 (-4.69, 6.57)
fe
H
6
o
Schwartz and Neas (2000)
Eastern U.S.
PM25 and CM (PM10_25) measured.
Summary levels not given.
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.
Sulfate fraction was highly
correlated with PM2 5 (0.94), and,
not surprisingly, gave similar
Uniontown Lag 0,PM2 5 :
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.
O Linn et al. (1996)
H So. California
(iQ NO2 ozone, and PM5 measured.
C^ PM5 was measured using a Marple low volume
O sampler PM5 ranged from 1-145 ug/m3 with a mean
H of 24.
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.
Morning FVC was significantly
decreased as a function of PM5 and
NO,
-------
to
O
o
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
Canada
OO
td
ON
OO
fe
H
6
o
o
H
O
Vedaletal. (1998)
Port Alberni, BC
PM10 via a Sierra-Anderson dichotomous sampler.
PM10 ranged from 1 to 159 ng/m3.
Europe
Boezenetal. (1999)
Netherlands
PM10, BS, SO2, and NO2 measured, but methods were
not given. PM10 ranged from 4.8 to 145 ug/m3 with
site means ranging from 26 to 54 ug/m3.
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.
Grievink et al. (1999)
Netherlands
PM10 and BS.
PM10 ranged from 12 to 123 ug/m3 with a mean of 44.
o
HH
H
W
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.
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 PEF decrease was related to pollutants
for children with bronchial hyperresponsiveness and high serum
Ige levels.
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 ±0.7 yr. 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.
No consistent evidence for adverse
health effects was seen in the
nonasthmatic control group.
No consistent pattern of effects
observed with any of the pollutants
for 0, 1, and 2 day lags.
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.
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
-------
to
O
o
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
o
HH
H
W
Europe (cont'd)
Kiinzli et al. (2000)
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.
The results were for two
hypothetical communities, A and
B.
oo
td
ON
vo
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 et al. (1996)
PM10, O3, and NO2 measured.
Combined results from 1208 children divided among 17 panels
studied. Separate results reported by endpoints included symptoms
as reported in a dairy and PEE. Individual panels were analyzed
using multiple linear regression analysis on deviations from mean
PEE 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.
Daily concentrations of most
elements were not associated with
the health effects.
PM10 was related to changes in
FEV and FVC
PM10 analyses not focus of this
paper.
fe
H
6
o
o
H
O
van der Zee et al. (1999)
Netherlands
PM10 averages ranged 20 to 48 ug/m3.
BS, sulfate fraction, SO2, and NO2 also measured.
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.
In children with symptoms,
significant associations found
between PM10, BS and sulfate
fraction and the health endpoints.
No multiple pollutant models
analyses reported.
Lag 0, PM10, Urban areas
Evening PEE OR = 1.15 (1.02,
1.29)
Lag 2, PM10, Urban areas
Evening PEE OR = 1.07 (0.96,
1.19)
5 day ave, PM10, Urban areas
Evening PEE = 1.13 (0.96, 1.32)
-------
to
O
o
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
OO
td
fe
H
6
o
o
H
O
Europe (cont'd)
van der Zee et al. (2000)
Netherlands
PM10 averages ranged 24 to 53 ug/m3.
BS, sulfate fraction, SO2, and NO2 also measured.
PM10 was measured using a Sierra Anderson 241
dichotomous sampler.
Tiittanen et al. (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 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.
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
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.
LagO,PM10:
Morning PEF
Evening PEF
4 day ave, PMj
Morning PEF
3.33)
Evening PEF
Lag 0, PM2 5
Morning PEF
Evening PEF
4 day ave., PM
Morning PEF
3.15)
Evening PEF
= 1.21 (-0.43, 2.85)
= 0.72 (-0.63, 1.26)
= -1.26 (-5.86,
= 2.33 (-2.62, 7.28)
= 1.11 (-0.64,2.86)
= 0.70 (-0.81, 2.20)
2.5
= -1.93 (-7.00,
= 1.52 (-3.91, 6.94)
o
HH
H
W
-------
to
O
o
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
Europe (cont'd)
oo
td
fe
H
6
o
o
H
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 ug/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 ug/m3.
Latin America
Gold etal. (1999)
Mexico City, Mexico
Mean 24 hO3 levels: 52ppb.
Mean PM25: 30 ng/m3.
Mean PM10: 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
ultrafine particles. A change of
PMIO from 10 to 20 ug/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
HH
H
W
-------
to
O
o
TABLE 8B-6 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF NONASTHMATICS
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
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
oo
td
i
-------
Appendix 8B.7: Short-Term PM Exposure Effects
On Symptoms in Nonasthmatics
June 2003 8B-73 DRAFT-DO NOT QUOTE OR CITE
-------
rt>
to
o
o
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 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
United States
OO
td
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.
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 PM2 5 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.
Sulfate fraction was highly correlated
with PM2 5 (0.94), and not surprisingly
gave similar answers.
Of all pollutants considered, only the
level of coarse particles as calculated
(PM10-PM25) independently related
to incidence of new episode of runny
noses.
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.
Canada
fe
H
6
o
o
H
O
o
HH
H
W
Vedaletal. (1998)
Port Alberni, BC
PM10 via a Sierra-Anderson dichotomous sampler.
PM10 ranged from 1 to 159 |ig/m3.
Longetal. (1998)
Winnepeg, CN
PM10, TSP, and VOC measured.
Methods for PM monitoring not given. Ranges of
values also not given.
Europe
Boezenetal. (1998)
Amsterdam, NL
PM10, SO2, and NO2 measured.
PM10 ranged from 7.9 to 242.2 ug/m3 with a median of
43.
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 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.
No consistent evidence for adverse
health effects was seen in the
nonasthmatic control group.
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
-------
to
O
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
OO
td
fe
H
6
o
o
H
O
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 ug/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 |ig/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 PM10 was not a focus of this
paper.
o
HH
H
W
-------
to
O
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
OO
td
fe
H
6
o
o
H
O
Europe (cont'd)
van der Zee et al. (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 et al. (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.
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 GLIMMIX). 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.
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.
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)
o
HH
H
W
Keles et al. (1999)
Istanbul, Turkey
Nov. 1996 to Jan. 1997.
TSP levels ranged from annual mean of 22
ug/m3 in unpolluted area to 148.8 |ig/m3 in
polluted area.
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.
No difference found for atopic status in
children living in area with different air
pollution levels.
-------
to
O
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
New Zealand
Harreetal. (1997)
Christchurch, NZ
SO2, NO2, PM10, and CO measured.
Details on monitoring methods and pollutant ranges
were not given.
Asia
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.
NO2 was associated with increased
bronchodilator use.
PM10 was associated with increased nighttime
chest symptoms.
OO
td
Awasthi et al. (1996)
India
Suspended particulate matter, SO2, nitrates, coal,
wood, PM and kerosene measured. SPM was
measured using a high-volume sampler.
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.
A significant regression coefficient
between PM and symptoms was found
fe
H
6
o
o
H
O
o
HH
H
W
-------
Appendix 8B.8: Long-Term PM Exposure Effects On
Respiratory Health Indicators, Symptoms, and Lung Function
June 2003 8B-78 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 8B-8. LONG-TERM PARTICULATE 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/m3 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 ug/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
Berglund et al.. (1999)
California communities
Peters et al. (1999a,b)
12 southern California communities
5 year exposure to PM10, ozone, NO2, acid levels.
PM10 annual averages ranged from 13 to 70
ug/m3.
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 |ig/m3.
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 ug/m3
across the communities.
o
HH
H
W
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.
Asthma, bronchitis, cough and wheeze rates were
adjusted for individual covariates. Community rates
were then regressed on pollutant averages for 1986-
1990.
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.
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.
Significant risk factors identified: childhood
respiratory illness, reported ETS exposure, age, sex
and parental history.
Wheeze was associated with NO2 and acid levels.
No symptoms were associated with PM10 levels.
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.
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 FEVj 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.
For PM10 > 100ug/m3, 42 d/yr:
RR = -1.09 CT (0.92, 1.30) for
obstructive disease determined by
pulmonary function tests.
OR for PM10 (per 25 ug/m3):
Asthma 1.09 ( 0.86, 1.37)
Bronchitis 0.94 (0.74, 1.19)
Cough 1.06(0.93, 1.21)
Wheeze 1.05 (0.89, 1.25)
PM10 24 hr average
PERF ml/s per 10 ug/m3
mean = -34.9
95% CI
-59.8,-10.1
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% for NO2
(p = 0.019); and -0.73% for inorganic
acid vapor
(p = 0.042).
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
United States (cont'd)
oo
td
i
oo
o
Gauderman et al. (2002)
Follow-up on 12 southern California
communities
5 year exposure to PM10, ozone, NO2, acid levels.
PM10 annual averages ranged from 5 to 27 ug/m3.
McConnell et al. (1999)
12 Southern California communities
1994 air monitoring data.
PM10 (mean 34.8; range 13.0 - 70.7 ug/m3). PM25
(yearly mean 2 week averaged mean 15.3 ug/m3;
range 6.7 - 31.5 ug/m3).
Linear regression analysis was used to estimate the
individual lung function growth adjusted for height,
weight, body mass index, and smoking. Growth rates
were then adjusted for individual covariates to obtain
community adjusted growth rates. These rates were
then related to pollutant averages for 1996-1999.
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 function growth was related to total acid.
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 of FEV1 were:
PM10
ozone
NO2
PM25
total acid
-0.21 (-1.04,0.64),
-0.55 (-1.27, 0.16),
-0.48 (-1.12, 0.17),
-0.39 (-1.06, 0.28),
-0.63 (-1.21, 0.17)
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)
fe
H
6
o
o
H
O
o
HH
H
W
McConnell et al. (2002)
12 Southern California communities
1994-1997
4-year mean cone. PM10 ng/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 ug/m3 with a mean
of 14.5.
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.
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.
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.
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
United States (cont'd)
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, PM11; particle strong
acidity, O3, NO2, and SO2. PM was measured
with a Harvard impactor. For pollutant ranges,
see Dockery et al. (1996).
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.
PM measures (e.g., particle strong acidity)
associated with FEV and FVC decrement.
Europe
oo
td
i
oo
fe
H
6
o
o
H
O
o
HH
H
W
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 ug/m3 with a mean of
37.
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 |ig/m3.
Zemp et al. (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 |ig/m3 with a mean of
21.
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 FEVj 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, NO2 and SO2.
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, PM,n and NO,.
Estimated regression coefficient for
PM10 versus FVC = -0.035 (95% CI
-0.041, -0.028). Corresponding
value for FEVJ -0.016 (95% CI
-0.023 to-0.01). Thus, 10 ug/m3
PM10 increase estimated to lead to
estimated 3.4 percent decrease in
FVC and 1.6 percent decrease in
FEVj.
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.
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
OO
td
i
OO
to
fe
H
6
o
o
H
O
o
HH
H
W
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 |ig/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 ug/m3 in the three areas.
PM was measured with a Harvard impactor.
Heinrich et al. (2002)
Surveyed children aged 5-14 in 1992-3, 1995-6,
1998-9. Annual TSP levels ranged from 25-79
ug/m3. Smallparticles (NC0 01.2 5 per 103cm"3)
remained relatively constant.
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; 2,335
children surveyed in first round, and 2,536 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.
A two-stage logistic regression model was used to
analyze the data which adjusted for age, gender,
educational level of parents, and indoor factors. The
model included fixed area effects, random deviations,
and errors from the adjustments. Parameters were
estimated using GEE methods.
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.
The study found bronchitis and frequency of colds
were significantly related to TSP.
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 overtime.
An increment of 50 ug/m3 TSP was
associated with an odds ratio for
bronchitis of 3.02 (1.72-5.29) and an
odds ratio of 1.90 ( 1.17-3.09) for
frequency of colds.
Bronchitis ever diagnosed
TSP per 50 ug/m3
OR 1.63 CI (1.37- 1.93)
Halle (East) %
TSP ug/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
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
Europe (cont'd)
Baldi et al. (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 |ig/m3. TSP was measured by the
gravimetric method.
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.
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.
For a 50 ug/m3 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
OO
td
i
OO
fe
H
6
o
o
H
O
Zeghnoun et al. (1999)
La Havre, France during 1993 and 1996. Daily
mean BS levels measured in three stations ranged
12 - 14 ug/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 ug/m3 for PM10, from 29 to 67 ug/m3 for
PM25, and from 12 to 38 ug/m3 for PM 10.25.
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 10 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.
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 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. forPM10_25. Positive
relationship found between concentration of IgG in
serum and PM2 5 but not for PM10 or PM10_2 5. Two
other models produced similar outcomes: a multi-
level linear regression model and an ordinal logistic
regression model.
Adjusted
Regression slope
PM,,
CD4+
80% 95% CI (34; 143)
p< 0.001
Total IgG
24%
95% CI (2; 52)
p 0.034
o
HH
H
W
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
OO
td
i
OO
fe
H
6
o
o
H
O
Europe (cont'd)
Turnovska and Kostiranev (1999)
Dimitrovgrad, Bulgaria, May 1996
Total suspended particulate matter (TSPM) mean
levels were 520 ± I 61 ug/m3 in 1986 and 187 ±
9 ug/m3 in 1996. SO2, H2S, and NO2 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 ug/m ± 53.98 in high area and
33.23 ± 35.99 in low area.
o
HH
H
W
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 ug/m3 mean compared to
low areas at 33.2 ug/m3.
Horak et al. (2002)
Frischer et al. (1999)
Eight communities in lower Austria between
1994-1997. PM10 mean summer value of 17.36
ug/m3 and winter value of 21.03 ug/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 FEVj 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.
Lung function assessed in 975 school children in grade
2-3. A several step analysis included GEE and
sensitivity analyses.
Vital capacity and FEVj 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 FEVj) 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.
Concluded that long term exposure to PM10 had a
significant negative effect on lung function with
additional evidence for a further effect for O3 and
NO,.
Boys
SLFG (FVC)
OR = 2.15 (CI 1.25- 3.69)
SLFG (FEVO
OR= 1.90 (CI 1.12- 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.
After adjusting for confounders an
increase in PM10 by 10 ug/m3 was
associated with a decrease in FEVj
growth at 84 mL/yr and 329 mL/5 yr
for MEF,,_7,.
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
Europe (cont'd)
Gehring et al. (2002)
In Munich, Germany
December 1997 - January 1999
Annual PM2 5 levels determined by 40 sites and a
GIS predictor for model.
Mean PM25 annual average of 13.4 ng/m3 with
range of 11.90 to 21.90 ug/m3
Effect of traffic-related air pollutants. PM2 5 and NO2
on respiratory health outcomes wheeze, cough,
bronchitis, respiratory infections, and runny nose were
evaluated using multiple logistic regression analyses of
1, 756 children during the first and second year of life
adjusting for potential confounding factors.
There was some indication of an association
between PM2 5 and symptoms of cough but not
other outcomes. In the second year of life most
effects were attenuated.
Latin America
oo
td
i
oo
Calderon-Garciduenas et al. (2000)
Southwest Metropolitan Mexico City (SWMMC)
winter of 1997 and summer of 1998.
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.
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.
fe
H
6
o
o
H
O
Australia
Lewis et al. (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
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.
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)
o
HH
H
W
Wongetal. (1999b)
Hong Kong, 1989 to 1991
Sulfate concentrations in respirable particles fell
by 38% after implementing legislation reducing
fuel sulfur levels.
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.
Comparison was by district; no PM measurements
reported. Results suggest control regulation may
have had some (but not statistically significant)
impact.
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
OO
td
i
OO
fe
H
6
o
o
H
O
Asia (cont'd)
Wangetal. (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 ug/m3)
TSP ranged from 112.81 to 237.82 ug/m3
(median = 181.00). CO, NO2, SO2, hydrocarbons
and O3 also measured.
Guoetal. (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 ug/m3 with a mean
of 69.
Wangetal. (1999)
Chongquing, China
April to July 1995
Dichot samplers used to measure PM2 5.
Mean PM2 5 level high in both urban (143 ug/m3)
and suburban (139 ug/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 PET 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 FEVj, FVC, and FEV/FVC% 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
(S02, PM10).
Mean SO2 concentration in the urban and suburban
area highly statistically significant different (213
and 103 ug/m3 respectfully). PM25 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
SE 28.17
t - 4.25
p<0.01
FVC
B- 57.89
SE30.80
t- 1.88
p<0.05
o
HH
H
W
-------
to
O
o
TABLE 8B-8 (cont'd). LONG-TERM PARTICIPATE MATTER EXPOSURE RESPIRATORY HEALTH INDICATORS:
RESPIRATORY SYMPTOM, LUNG FUNCTION
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
Effect estimates as reported by study
authors. Negative coefficients for
lung function and ORs greater than 1
for other endpoints suggest effects of
PM
OO
td
i
OO
Asia (cont'd)
Zhang et al. (1999)
4 areas of 3 Chinese Cities (1985 - 1988)
TSP levels ranged from an annual arithmetic
mean 137 ug/m3 to 1250 ug/m3 using gravimetric
methods.
Qian et al. (2000)
4 China cities
The 4 year average TSP means were 191, 296,
406, and 1067 ng/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.
fe
H
6
o
o
H
O
o
HH
H
W
-------
i 9. INTEGRATIVE SYNTHESIS
2
3
4 9.1 INTRODUCTION
5 This chapter focuses on integration of key information drawn from the preceding detailed
6 chapters, to provide a coherent framework for assessment of human health risks posed by
7 ambient particulate matter (PM) in the United States. As such, the chapter updates the integrated
8 assessment of available scientific information regarding ambient PM sources, exposures, and
9 health risks as they pertain to the United States that was provided in the 1996 Particulate Matter
10 Air Quality Criteria Document (1996 PM AQCD; U.S. Environmental Protection Agency,
11 1996a). It also highlights key findings on environmental effects of airborne PM.
12
13 9.1.1 Legislative Requirements and Past NAAQS Reviews
14 As indicated in U.S. Code (1991), the U.S. Clean Air Act (CAA), Sections 108 and 109
15 (42 U.S.C. Sections 7408 and 7409) govern the establishment, review, and revision of National
16 Ambient Air Quality Standards (NAAQS). Section 108(a) directs the EPA Administrator to list
17 pollutants, which, in the Administrator's judgement, cause or contribute to air pollution which
18 may reasonably be anticipated to endanger either public health or welfare and to issue air quality
19 criteria for them. The air quality criteria are to reflect the latest scientific information useful in
20 indicating the kind and extent of all identifiable effects on public health and welfare that may be
21 expected from the presence of the pollutant in ambient air. Section 109 directs the EPA
22 Administrator to propose and promulgate "primary" and "secondary" NAAQS for pollutants
23 identified under Section 108. Section 109(b)(l) defines a primary standard as a level of air
24 quality, the attainment and maintenance of which, in the judgement of the Administrator, based
25 on the criteria and allowing for an adequate margin of safety, is requisite to protect the public
26 health. Section 109(b)(2) defines a secondary standard as one which, in the judgement of the
27 Administrator, based on the criteria, is requisite to protect public welfare from any known or
28 anticipated adverse effects associated with the presence of such pollutants. Welfare effects
29 include, but are not limited to, effects on soils, water, crops, vegetation, man-made materials,
30 animals, wildlife, weather, visibility and climate, damage to and deterioration of property, and
31 hazards to transportation, as well as effects on economic values and personal comfort and
June 2003 9-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 well-being. Section 109(d) requires periodic review and, as appropriate, revision of existing
2 criteria and standards. It also requires an independent committee of non-EPA experts, the Clean
3 Air Scientific Advisory Committee (CASAC), to provide advice and recommendations to the
4 EPA Administrator regarding the scientific soundness and appropriateness of criteria and
5 NAAQS for PM and other "criteria air pollutants" (i.e., O3, NO2, SO2, CO, and Pb) regulated
6 under CAA Sections 108-109.
7 EPA first promulgated primary and secondary NAAQS for PM on April 30, 1971 (Federal
8 Register, 1971). These standards measured PM as "total suspended particulate" (TSP), which
9 refers to ambient PM up to a nominal size of 25 to 45 micrometers (|im). The primary standards
10 for PM (measured as TSP) were 260 |ig/m3 (24-h average), not to be exceeded more than once
11 per year, and 75 |ig/m3 (annual geometric mean). The secondary standard (measured as TSP)
12 was 150 |ig/m3 (24-h average), not to be exceeded more than once per year. In July 1987, EPA
13 revised the 1971 standards to protect against adverse health effects of inhalable airborne particles
14 which can be deposited in the lower (thoracic) regions of the human respiratory tract, with
15 "PM10" as the indicator, i.e., those particles collected by a sampler with a specified penetration
16 curve yielding an upper 50% cut-point of 10-|im aerodynamic diameter (Federal Register, 1987).
17 EPA established identical primary and secondary PM10 standards for two averaging times:
18 150 |ig/m3 (24-h average), with no more than one expected exceedance per year and 50 |ig/m3
19 (expected annual arithmetic mean), averaged over three years.
20 Taking into account information and assessments presented in the 1996 PM AQCD and
21 associated 1996 PM Staff Paper (SP), advice and recommendations of CASAC, and public
22 comments received on proposed revisions to the PM NAAQS (Federal Register, 1996), the EPA
23 Administrator promulgated significant revisions to the PM NAAQS in July 1997 (Federal
24 Register, 1997). In that decision, although it was determined that the PM NAAQS should
25 continue to focus on particles less than or equal to 10 jim in diameter, it was also determined that
26 the fine and coarse fractions of PM10 should be considered separately. New standards were
27 added, using PM2 5 as the indicator for fine particles, and PM10 standards were retained for the
28 purpose of regulating coarse-fraction particles. Two new PM25 standards were set: an annual
29 standard of 15 |ig/m3, based on the 3-year average of annual arithmetic mean PM25
30 concentrations from single or multiple community-oriented monitors; and a 24-hour standard of
31 65 |ig/m3, based on the 3-year average of the 98th percentile of 24-hour PM2 5 concentrations at
June 2003 9-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 each population-oriented monitor within an area. To continue to address coarse-fraction
2 particles, the annual PM10 standard was retained, and the form, but not the level, of the 24-hour
3 PM10 standard was revised to be based on the 99th percentile of 24-hour PM10 concentrations at
4 each monitor in an area. The secondary standards were revised by making them identical in all
5 respects to the primary standards.
6 Following 1997 promulgation of the revised PM NAAQS, legal challenges were filed by
7 many parties, addressing a broad range of issues. In May 1998, the U.S. Court of Appeals for
8 the District of Columbia Circuit issued an initial opinion upholding EPA's decision to establish
9 fine particle standards, finding that such standards were amply justified by the growing body of
10 empirical evidence showing a relationship between fine particle pollution and adverse health
11 effects. Further, the court found "ample support" for EPA's decision to regulate coarse fraction
12 particles, although it vacated the revisions to the 1987 PM10 standards on the basis of PM10 being
13 a "poorly matched indicator for coarse particulate pollution" because PM10 includes fine
14 particles. As a result of this aspect of the court's ruling, which EPA did not appeal, the 1987
15 PM10 standards remain in effect. In addition, the U.S. Court of Appeals initially broadly held
16 that EPA's approach to establishing the level of the standards in its 1997 decisions on both the
17 PM and ozone NAAQS (which were promulgated on the same day and considered together by
18 the court in this aspect of its opinion) effected "an unconstitutional delegation of legislative
19 authority." EPA appealed this aspect of the court's ruling to the U.S. Supreme Court. In
20 February 2001, the U.S. Supreme Court unanimously reversed the Court of Appeals' ruling on
21 the constitutional issue, and sent the case back to the Court of Appeals for resolution of any
22 remaining issues not addressed in that court's earlier rulings. In March 2002, the Court of
23 Appeals rejected all remaining challenges to the standards, finding that the 1997 PM25 standards
24 were reasonably supported by the record and were not "arbitrary or capricious." American
25 Trucking Associations v. EPA. 283 F. 3d 355, 369-72 (D.C. Cir. 2002). Thus, the 1997 PM25
26 standards are in effect.
27 This updated revision of the PM AQCD, then, focuses on assessment of extensive newly
28 available (since the 1996 PM AQCD) information pertinent to consideration of (a) possible
29 retention or revision of the PM2 5 NAAQS set to protect mainly against health effects related to
30 exposures to ambient (outdoor) concentrations of airborne fine-mode particles now experienced
31 in the United States; (b) the possible setting of new primary standards to protect against thoracic
June 2003 9-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 coarse fraction (PM10_2.5) health effects; and (c) possible revisions to PM secondary standards to
2 protect against PM-related welfare effects.
3
4 9.1.2 Organization of the Chapter
5 Unlike the other criteria pollutants (O3, CO, NO2, SO2, and Pb), PM is not a specific
6 chemical entity but is a mixture of particles of different sizes, compositions, and properties. This
7 chapter first provides background information on key features of atmospheric particles,
8 highlighting important distinctions between fine and coarse particles with regard to size,
9 chemical composition, sources, atmospheric behavior, and potential human exposure
10 relationships — distinctions that collectively continue to suggest that fine and coarse particles
11 should be treated as two distinct subclasses of air pollutants. Recent data for the concentrations
12 of different ambient PM size and composition fractions (e.g., PM10, PM2 5, and PM10_2 5) and
13 ranges of variability seen in selected U.S. urban airsheds are also summarized to place the
14 ensuing human exposure and health effects discussions in perspective. After discussing human
15 exposure aspects, the chapter next summarizes key points regarding respiratory tract dosimetry,
16 followed by a discussion of the extensive PM health database that has expanded greatly during
17 recent years.
18 The latter includes numerous new epidemiologic studies of populations throughout the
19 world published since the 1996 PM AQCD that provide further evidence that notable health
20 effects (mortality, exacerbation of chronic disease, increased hospital admissions, etc.) are
21 associated with exposures to ambient levels of PM found in contemporary U.S. urban air sheds.
22 Epidemiologic findings related to specific PM components (by size, chemical composition) and
23 source contributions are also noted. Evaluations of other possible explanations for the reported
24 PM epidemiology results (e.g., other co-pollutants, choice of models, etc.) also are discussed,
25 ultimately leading to the conclusion that the reported associations of PM exposure and effects
26 are valid. Quantitative evidence is also discussed that (a) further substantiates associations of
27 such serious health effects with U.S. ambient PM10 levels, (b) also more strongly establishes fine
28 particles (as indexed by various indicators, e.g., PM2 5) as likely being important contributors to
29 the observed human health effects, and (c) now provides additional information on associations
30 between thoracic coarse particles (as indexed by PM10_2 5) and adverse health impacts. The
31 overall coherence of the newer epidemiologic database also is discussed.
June 2003 9-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 New toxicologic evidence (derived from controlled exposure studies of humans and
2 laboratory animals) is also highlighted, which elucidates findings on mechanisms of action and
3 other information that greatly enhances the plausibility of the epidemiologic findings in
4 comparison to 1996. The nature of the observed effects and the biological mechanisms that
5 might underlie such effects then are discussed, including with regard to effects seen in
6 compromised laboratory animal models meant to mimic features thought to contribute to
7 increased risk for susceptible human subpopulations. The increased, but still limited, availability
8 of new experimental evidence necessary to evaluate or directly substantiate the viability of
9 hypothesized mechanisms is noted. Information concerning possible contributions of particular
10 classes of specific ambient PM constituents also is summarized.
11 The chapter also provides information on the identification of susceptible human
12 population groups at special risk for ambient PM effects and factors placing them at increased
13 risk, which need to be considered in generating risk estimates for the possible occurrence of
14 PM-related health events in the United States. In addition, the chapter also makes note of new
15 information related to estimation of potential life-shortening attributable to PM effects.
16 As such, the overall sequencing of topics covered in the chapter is basically organized to
17 follow the risk assessment framework shown in Figure 9-1, along with some additional
18 information being provided by which to place current findings in perspective in relation to some
19 potential public health implications for U.S. population groups. The information presented here
20 and overall in this revised PM AQCD will provide key inputs to development of a PM Staff
21 Paper and associated exposure and risk analyses being developed by EPA's Office of Air Quality
22 Planning and Standards (OAQPS) to support consideration of options for possible retention or
23 revision of the primary PM NAAQS. In addition, information highlighted at the end of this
24 chapter and discussed in more detail in Chapter 4 with regard to environmental effects of
25 ambient PM will provide inputs to OAQPS analyses supporting considerations related to
26 secondary PM NAAQS.
27
28
29
June 2003 9-5 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
particulate matter
Deposition,
clearance, retention
and disposition of
particulate 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 9.2 BACKGROUND
2 9.2.1 Basic Concepts
3 Atmospheric particles originate from a variety of sources and possess a range of
4 morphological, chemical, physical, and thermodynamic properties. Sources include combustion,
5 photochemical oxidation of precursors, and soil dust. Atmospheric particles contain inorganic
6 ions, metallic compounds, elemental carbon, organic compounds, and crustal compounds. Some
7 atmospheric particles are hygroscopic and contain particle-bound water. The organic fraction is
8 especially complex, containing hundreds of organic compounds. Individual particles may be
9 composed of any number of the above and other components.
10
11 9.2.2 Particle Size Distributions
12 As discussed in Chapter 2, the distribution of particles with respect to size is an important
13 physical parameter governing their behavior. Atmospheric particles vary in density and often
14 are not spherical. Therefore, their diameters are often described by an "equivalent" diameter
15 (i.e., that of a unit density sphere that would have the same physical behavior). Diffusion and
June 2003
9-6
DRAFT-DO NOT QUOTE OR CITE
-------
1 gravitational settling are important physical behaviors for particle transport, collection, and
2 removal processes, including deposition in the respiratory tract. Different equivalent diameters
3 are used depending on which process is more important. For smaller particles diffusion is more
4 important and the Stokes diameter, Dp, is often used. For a smooth, spherically shaped particle,
5 Dp exactly equals the physical diameter of the particle. For irregularly shaped particles, Dp is the
6 diameter of an equivalent sphere that would have the same aerodynamic resistance. For larger
7 particles gravitational setting is more important and the aerodynamic diameter, Da, is often used.
8 Da depends on the density of the particle and is defined as the diameter of a spherical particle
9 with a density of 1 g/cm3 but with a settling velocity equal to that of the particle in question. The
10 atmospheric deposition rates of particles, and therefore, their residence times in the atmosphere,
11 are a strong function of their diameters. The diameter also influences deposition patterns of
12 particles within the lung. The effects of atmospheric particles on visibility, radiative balance,
13 and climate, will also be influenced by the size distribution of the particles. Atmospheric
14 particles cover several orders of magnitude in particle size. Therefore, size distributions often
15 are expressed in terms of the logarithm of the particle diameter on the X-axis and the measured
16 differential concentration on the Y-axis. If the differential concentration is plotted on a linear
17 scale, the number of particles (per cm3 of air), or the surface area, the volume, or the mass of
18 particles (per m3 of air) having diameters in the size range from log D to log(D + AD), will be
19 proportional to the area under that part of the size distribution curve.
20 Averaged atmospheric size distributions are shown in Figures 9-2. Figure 9-2a shows the
21 number distributions of particles, on a logarithmic scale, as a function of particle diameter for
22 several aerosols. The particle volume distributions for two of these are shown in Figure 9-2b.
23 These distributions show that most of the particles are quite small, below 0.1 jam; whereas most
24 of the particle volume (and therefore most of the mass) is found in particles larger than 0.1 jim.
25
26 9.2.3 Definitions of Particle Size Fractions
27 Aerosol scientists use three different approaches or conventions in the classification of
28 particles by size: (1) modes, based on the observed size distributions and formation
29 mechanisms; (2) cut point, usually based on the 50% cut point of the specific sampling device,
30 including legally specified, regulatory sizes for air quality standards; and (3) dosimetry or
June 2003 9-7 DRAFT-DO NOT QUOTE OR CITE
-------
1,000,000 -
o
0)
H—«
CD
CD
Q
_CD
O
CD
Q_
O)
2 0.0001 -
0.000001 -
10,000 -
100 -
1 -
0.01 -
Clean Rural
Urban Influenced Rural
Average Urban
Urban + Freeway
0.01 0.1 1 10 100
Particle Diameter (pm)
"E
o
0)
E
ro
b
ro
Q_
O
"O
70
65 -
60 -
55 -
50 -
45 -
40 -
35 -
30 -
25 -
20 -
^5 -
10 -
5 -
Average Urban
Urban + Freeway
0.01 0.1 1 10
Particle Diameter (|jm)
100
Figure 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.
Source: Whitby and Sverdrup (1980).
1
2
3
4
5
6
7
occupational health sizes, based on the entrance into various compartments of the respiratory
system.
Modal. The modal classification, first proposed by Whitby (1978), is shown in Figure 9-3.
New modes introduced since 1978 are shown in Figure 9-4. The nucleation and Aitkin modes
are best observed in the number distribution. The observed modal structure is frequently
approximated by several log-normal distributions. Terms used in the modal description of
particle size distributions are defined as follows. Nucleation Mode: Freshly formed particles
with diameters below 10 nm, observed during active nucleation events. The lower limit, where
June 2003
9-8
DRAFT-DO NOT QUOTE OR CITE
-------
o
7 -
5 -
4 -
D)
I 3
I
2 -
1 -
0.002
Mechanically
Generated
Nuclei Mode
0.1 1
Particle Diameter, Dp(|jm)
Accumulation Mode
Fine Particles
Coarse Mode
Coarse Particles
100
Figure 9-3. Volume size distribution, measured in traffic, showing fine and coarse
particles and the nuclei and accumulation modes within the fine particles.
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 particles and large molecules overlap, is uncertain. Current techniques limit measurements to
2 particles 3 nm or greater. AitkinMode: Larger particles with diameters between 10 and 100 nm.
3 The Aitken mode may result from growth of smaller particles or nucleation from higher
4 concentrations of precursors. Nucleation and Aitkin nuclei modes are normally observed in the
5 number distribution. Accumulation Mode: Particles with diameters from about 0.1 jim to just
6 above the minimum in the mass or volume distributions which usually occurs between 1 and
7 3 |im. Accumulation-mode particles normally do not grow into the coarse mode.
June 2003
9-9
DRAFT-DO NOT QUOTE OR CITE
-------
^,UUU"
"g 1,500-
(U
o
••e
g_ 1,000-
Q.
Q
O)
O
5 500-
"Z.
T3
0 „
Nucleation
Mode
p
1.
—
1 n-T
p
—
-
Aitken Mode
-i Accumulation
Mode
~U
Rr
Thn
10 20 50 100
Particle Diameter (nm)
200
500
1,000
Figure 9-4. Submicron number size distribution observed in a boreal forest in Finland
showing the tri-modal structure of fine particles. The total particle number
concentration was 1011 particles/cm3 (10 minute average).
Source: Makelaetal. (1997).
1 Nucleation-mode and Aitkin-mode particles grow by coagulation (two particles combining to
2 form one) or by condensation (low-equilibrium vapor pressure gas molecules condensing on a
3 particle) and "accumulate" in this size range. Coarse Mode or Coarse Particles: Particles with
4 diameters mostly greater than the minimum in the particle mass or volume distributions, which
5 generally occurs between 1 and 3 jim. These particles are usually formed by mechanical
6 breakup of larger particles or bulk material. Fine Particles: Fine particles include the
7 nucleation, Aitkin, and accumulation modes, i.e., particles from the lowest measurable size,
8 currently about 3 nm, to just above the minimum in the mass or volume distribution which
9 generally occurs between 1 and 3 jim. These particles are generated during combustion or
10 formed from gases. Ultrafine Particles: That portion of fine particles with diameters below
11 about 0.1 |im (100 nm), i.e., the Aitkin and nucleation modes.
12 Modes are defined primarily in terms of their formation mechanisms but also differ in
13 terms of sources, composition, age, and size. The major processes that influence the formation
14 and growth of particles are also shown in Figure 9-3. New particles may be formed by
15 nucleation from gas phase material. Particles may grow by condensation as gas phase material
June 2003
9-10
DRAFT-DO NOT QUOTE OR CITE
-------
1 condenses on existing particles. Particles also may grow by coagulation as two particles
2 combine to form one. Nucleation mode applies to newly formed particles which have had little
3 chance to grow by condensation or coagulation. Aitkin mode particles are also recently formed
4 particles that are still actively undergoing coagulation. However, because of higher
5 concentrations of precursors or more time for condensation and coagulation, the particles have
6 grown to larger sizes. Accumulation mode applies to the final stage as particles, originally
7 formed as nuclei, grow to a point where growth slows down. Gas phase material condenses
8 preferentially on smaller particles and the rate constant for coagulation of two particles decreases
9 as the particle size increases. Therefore, nucleation-mode particles grow into the Aitkin mode
10 and further into the accumulation mode, but accumulation-mode particles do not normally grow
11 into the coarse mode. The nucleation, Aitkin, and accumulation modes, which together are
12 called fine particles, are formed primarily by combustion or chemical reactions of gases yielding
13 products with low saturated vapor pressures. Fine particles include metals and elemental and
14 organic carbon (primary PM) and sulfate, nitrate, ammonium ions, and organic compounds
15 (secondary PM).
16 The coarse mode refers to particles formed by mechanical breakdown of minerals, crustal
17 material, and organic debris. The composition includes primary minerals and organic material.
18 The accumulation mode and the coarse mode overlap in the region between 1 and 3 jim (and
19 occasionally over an even larger range). In this region, chemical composition of individual
20 particles can usually, but not always, allow identification of a source or formation mechanism
21 and so permit identification of a particle as belonging to the accumulation or coarse mode.
22 Over the years, the terms fine and coarse, as applied to particle sizes, have lost the precise
23 meaning given in Whitby's (1978) definition. In any given article, therefore, the meaning of fine
24 and coarse, unless defined, must be inferred from the author's usage. In particular, PM25 and
25 fine particles are not equivalent because PM25 includes some particles between about 1 and 2.5
26 |im Da from the small-size tail of the coarse mode.
27
28 Sampler Cut Point. Another set of definitions of particle size fractions arises from
29 considerations of size-selective sampling. Size-selective sampling refers to the collection of
30 particles below or within a specified aerodynamic size range. Size fractions are usually specified
31 by the 50% cut point size; e.g., PM2 5 refers to particles collected by a sampling device that
June 2003 9-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 collects 50% of 2.5 jim particles and rejects 50% of 2.5 jim particles. However, size fractions
2 are defined, not merely by the 50% cut point, but by the entire penetration curve. Examples of
3 penetration curves are given in Figure 9-5. Thus, as shown by Figure 9-5, a PM25 sampler, as
4 defined by the Federal Reference Method, rejects 94% of 3 jim particles, 50% of 2.5 jim
5 particles, and 16% of 2 jim. Samplers with the same 50% cut point but differently shaped
6 penetration curves would collect different fractions of PM. Size-selective sampling has arisen in
7 an effort to measure particle size fractions with some special significance (e.g., health, visibility,
8 source apportionment, etc.), to measure mass size distributions, or to collect size-segregated
9 particles for chemical analysis. Dichotomous samplers split the particles into smaller and larger
10 fractions that may be collected on separate filters. However, some fine particles (=10%) are
11 collected with the coarse particle fraction. Cascade impactors use multiple size cuts to obtain a
12 distribution of size cuts for mass or chemical composition measurements. One-filter samplers
13 with a variety of upper size cuts are also used, e.g., PM25, PM10.
14 Regulatory size cuts are a specific example of size-selective sampling. As noted earlier,
15 the NAAQS for PM were revised in 1987 to use PM10, rather than total suspended particulate
16 matter (TSP), as the indicator for the PM NAAQS (Federal Register, 1987). The use of PM10 as
17 an indicator is an example of size-selective sampling based on a regulatory size cut (Federal
18 Register, 1987). The selection of PM10 as an indicator was based on health considerations and
19 was intended to focus regulatory concern on those particles small enough to enter the thoracic
20 region of the human respiratory tract. The PM25 standard set in 1997 is also an example of size-
21 selective sampling based on a regulatory size cut (Federal Register, 1997). The PM25 standard
22 was based primarily on epidemiologic studies using concentrations measured with PM2 5
23 samplers as an exposure index. However, the PM2 5 sampler was not designed to collect
24 respirable particles. It was designed to collect fine particles. EPA is currently considering the
25 possibility of a thoracic coarse particle standard with PM10_2 5 as an indicator. Examples of
26 regulatory size cuts are shown in Figure 9-6. Note also that, in the range of particle aerodynamic
27 diameter (Da) between 1.0 and 2.5 jim, there is overlap between fine and coarse particles. The
28 degree of overlap depends on prevailing conditions of humidity and the amount of soil dust in
29 the atmosphere.
30
June 2003 9-12 DRAFT-DO NOT QUOTE OR CITE
-------
100
0
APM10
• IPM
• TPM
O RPM
VPM25
4 10 20
Aerodynamic Diameter ((jm)
100
Figure 9-5. 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 (200la), PM10
(2001b). PM25 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 Occupational Health Size Fractions. The occupational health community has defined size
2 fractions for use in the protection of human health. This convention classifies particles into
3 inhalable, thoracic, and respirable particles according to their upper size cuts (also shown in
4 Figure 9-4). However, these size fractions may also be characterized in terms of their entrance
5 into various compartments of the respiratory system. Thus, inhalable particles enter the
6 respiratory tract, including the head airways. Thoracic particles travel past the larynx and reach
June 2003
9-13
DRAFT-DO NOT QUOTE OR CITE
-------
to
70
60 -
50
Q" 40
O)
o
30 -
20 -
10 -
Accumulation Mode
0.1 0,2
0,5 1 2 5 10 20
Particle Dp(pm)
Total Suspended Particles (TSP) '
PM«
10
PM
2.5
w«_ PM
10-2.5
50 100
Figure 9-6. An idealized distribution of ambient particulate matter showing the
accumulation mode and the coarse mode and the size 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).
1
2
3
4
the lung airways and the gas-exchange regions of the lung. Respirable particles are a subset of
thoracic particles that are more likely to reach the gas-exchange region of the lung.
June 2003
9-14
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.3 CHARACTERIZATION OF PM SOURCES
2 The linkages between airborne PM and its sources are not as well defined as they are for
3 many other pollutants. In large part this is because PM is not a well defined chemical entity but
4 represents a complex mixture of primary and secondary components. PM is called "primary" if
5 it is in the same chemical form in which it was emitted into the atmosphere. PM is called
6 "secondary" if it is formed by chemical reactions in the atmosphere. Primary coarse particles are
7 usually formed by mechanical processes, such as the abrasion of surfaces or by the suspension of
8 soil or biological material. This includes material emitted in particulate form, such as wind-
9 blown dust, sea salt, road dust, and combustion-generated particles such as fly ash and soot.
10 PM10_25 is mainly primary in origin. Primary fine particles are emitted from sources either
11 directly as particles or as vapors that rapidly condense to form ultrafme or nuclei-mode particles.
12 Secondary PM is formed by chemical reactions of free, adsorbed, or dissolved gases. Most
13 secondary fine PM is formed from condensable vapors generated by chemical reactions of
14 gas-phase precursors. Secondary formation processes can result in either the formation of new
15 particles or the addition of condensable vapor to preexisting particles. Most of the sulfate and
16 nitrate and a portion of the organic compounds in atmospheric particles are formed by chemical
17 reactions in the atmosphere. Because precursor gases undergo mixing during transport from
18 their sources, it is difficult to identify individual sources of secondary constituents of PM.
19 Table 9-1 summarizes anthropogenic and natural sources for the major primary and
20 secondary aerosol constituents of fine and coarse particles. Anthropogenic sources can be
21 further divided into stationary and mobile sources. Stationary sources include fuel combustion
22 for electrical utilities, residential space heating and industrial processes; construction and
23 demolition; metals, minerals, and petrochemicals; wood products processing; mills and elevators
24 used in agriculture; erosion from tilled lands; waste disposal and recycling; and fugitive dust
25 from paved and unpaved roads. Mobile, or transportation-related, sources include direct
26 emissions of primary PM and secondary PM precursors from highway and off-highway vehicles
27 and nonroad sources. In addition to fossil fuel combustion, biomass in the form of wood is
28 burned for fuel. Vegetation is burned to clear new land for agriculture and for building
29 construction, to dispose of agricultural and domestic waste, to control the growth of animal or
30 plant pests, and to manage forest resources (prescribed burning). Also shown are sources for
31 precursor gases whose oxidation forms secondary particulate matter.
June 2003 9-15 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-1. CONSTITUENTS OF ATMOSPHERIC PARTICLES AND THEIR MAJOR SOURCES1
to
O
o
Sources
Primary (PM<2.5 um)
Aerosol
species Natural Anthropogenic
SO4= Sea spray Fossil fuel combustion
Sulfate
N03- — —
Nitrate
Primary (PM >2.5 um) Secondary PM Precursors (PM <2.5 um)
Natural Anthropogenic Natural
Sea spray — Oxidation of reduced sulfur
gases emitted by the oceans and
wetlands and SO2 and H2S
emitted by volcanism and forest
fires
— — Oxidation of NOX produced by
soils, forest fires, and lighting
Anthropogenic
Oxidation of SO2 emitted
from fossil fuel combustion
Oxidation of NOX emitted
from fossil fuel combustion
Minerals
and in motor vehicle
exhaust
Erosion and
re-entrainment
Fugitive dust paved and
unpaved roads,
agriculture, and
forestry
Erosion and re-
entrainment
Fugitive dust, paved
and unpaved road dust,
agriculture, and
forestry
NH4+ —
Ammonium
Organic Wild fires
carbon (OC)
Elemental Wild fires
carbon
(EC)
Metals Volcanic
activity
Bioaerosols Viruses and
bacteria
Prescribed burning,
wood burning, motor
vehicle exhaust, and
cooking
Motor vehicle exhaust,
wood burning, and
cooking
Fossil fuel combustion,
smelting, and brake
wear
— Tire and asphalt wear
and paved road dust
— Tire and asphalt wear
and paved road dust
Erosion, re-entrainment, —
and organic debris
Plant and insect —
fragments, pollen, fungal
spores, and bacterial
agglomerates
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
'Dash (-) indicates either very minor source or no known source of component.
-------
1 In general, the sources of fine PM are very different from those for coarse PM. Some of
2 the mass in the fine size fraction has been formed during combustion from material that
3 volatilized in combustion chambers and then recondensed before emission into the atmosphere.
4 By and large, however, most ambient PM2 5 is secondary, having been formed in the atmosphere
5 from photochemical reactions involving precursor gases. Transport and transformations of
6 precursors can occur over distances of hundreds of kilometers. The coarse PM constituents have
7 shorter lifetimes in the atmosphere, so their effects tend to be more localized. Only major
8 sources for each constituent within each broad category shown at the top of Table 9-1 are listed.
9 Not all sources are equal in magnitude. Chemical characterizations of primary particulate
10 emissions for a wide variety of natural and anthropogenic sources (as shown in Table 9-1) were
11 given in Chapter 5 of the 1996 PM AQCD. Summary tables of the composition of source
12 emissions presented in the 1996 PM AQCD and updates to that information are provided in
13 Appendix 3D of Chapter 3 in this document. The profiles of source composition are based
14 largely on results of various studies that collected signatures for use in source apportionment
15 studies.
16 Natural sources of primary PM include windblown dust from undisturbed land, sea spray,
17 and plant and insect debris. The oxidation of a fraction of terpenes emitted by vegetation and
18 reduced sulfur species from anaerobic environments leads to secondary PM formation.
19 Ammonium (NH4+) ions, which play a major role in regulating the pH of particles, are derived
20 from emissions of ammonia (NH3) gas. Source categories for NH3 have been divided into
21 emissions from undisturbed soils (natural) and emissions that are related to human activities
22 (e.g., fertilized lands, domestic and farm animal waste). There is ongoing debate about
23 characterizing emissions from wild fires (i.e., unwanted fire) as either natural or anthropogenic.
24 Wildfires have been listed in Table 9-1 as natural in origin, but land management practices and
25 other human actions affect the occurrence and scope of wildfires. For example, fire suppression
26 practices allow the buildup of fire fuels and increase the susceptibility of forests to more severe
27 and infrequent fires from whatever cause, including lightning strikes. Similarly, prescribed
28 burning is listed as anthropogenic, but can viewed as a substitute for wildfires that would
29 otherwise eventually occur on the same land.
30 The precursors to secondary PM have natural and anthropogenic sources, just as primary
31 PM has natural and anthropogenic sources. Whereas the major atmospheric chemical
June 2003 9-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 transformations leading to the formation of particulate nitrate and sulfate have been relatively
2 well studied, those involving the formation of secondary aerosol organic carbon are still under
3 active investigation. A large number of organic precursors are involved, many of the kinetic
4 details still need to be determined, and many of the actual products of the oxidation of
5 hydrocarbons have yet to be identified.
6 However, over the past decade, a significant amount of research has been carried out to
7 improve the understanding of the atmospheric chemistry of secondary organic PM (SOPM)
8 formation. Although additional sources of SOPM might still be identified, there appears to be a
9 general consensus that biogenic compounds (monoterpenes, sesquiterpenes) and aromatic
10 compounds (toluene, ethylbenzene) are the most significant SOPM precursors. A large number
11 of compounds have been detected in biogenic and aromatic SOPM, although the chemical
12 composition of these two categories has not been fully established, especially for aromatic
13 SOPM. Transformations that occur during the aging of particles are still not adequately
14 understood. There are still large gaps in current understanding of a number of key processes
15 relating to the partitioning of semivolatile compounds between the gas phase and ambient
16 particles containing organic compounds, liquid water, inorganic salts, and acids. In addition,
17 there is a general lack of reliable analytical methods for measuring multifunctional oxygenated
18 compounds in the gas and aerosol phases.
19 The relative strengths of the different sources shown in Table 9-1 can be estimated either
20 on the basis of ambient measurements using source apportionment techniques or on the basis of
21 chemistry-transport models using emissions inventories. For most practical purposes, the
22 relative contributions of sources affecting different sites are determined by source apportionment
23 models. The major approaches to source apportionment modeling have been reviewed in
24 Section 3-3 of this document and in greater detail in Section 5-5 of the 1996 PM AQCD. These
25 methods are capable of supplying errors in the apportionments; however, there is some
26 subjectivity in the assignment of the input errors. The results of source-apportionment modeling
27 studies conducted throughout the United States indicate that the combustion of fossil and
28 biomass fuels is the major source of measured ambient PM25. Fugitive dust constitutes a major
29 fraction of PM10_25 and can contribute extensively to PM25, especially in arid western regions.
30 Primary biologic particles can contribute substantially to both the PM25 and PM10_25 size ranges.
31 However data for their concentrations are sparse.
June 2003 9-18 DRAFT-DO NOT QUOTE OR CITE
-------
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
8 during these events. Uncontrolled biomass burning in central America and Mexico may have
9 contributed to elevated PM levels that exceeded the daily NAAQS level for PM in Texas; and
10 wildfires throughout the United States, Canada, Mexico, and Central America all contribute to
11 PM background concentrations in the United States.
12
13
14 9.4 AMBIENT CONCENTRATIONS
15 9.4.1 Measurement of Particulate Matter
16 It is possible to measure a variety of PM indicators with high precision. However, the
17 absolute accuracy of a PM monitoring techniques cannot be established because no standard
18 reference calibration material or procedure has been developed for suspended, atmospheric PM.
19 Therefore, accuracy is defined as the degree of agreement between a field PM sampler and a
20 collocated PM reference method audit sampler. Intercomparison studies, therefore, are very
21 important for establishing the reliability of PM measurements.
22 One important measurement problem arises from the presence of semivolatile components
23 (i.e., species that exist in the atmosphere in dynamic equilibrium between the condensed phase
24 and gas phase) in atmospheric PM. Important examples include ammonium nitrate, semivolatile
25 organic compounds, and particle-bound water. Most filter-weighing techniques for PM,
26 including the U.S. Federal Reference Methods (FRM), require equilibration of collected material
27 at fixed, near-room temperature (25 °C) and moderate relative humidity (40%) to reduce
28 particle-bound water. However, as shown in Figure 9-7, this also causes the loss of an unknown,
29 but possibly significant fraction, of ammonium nitrate and semivolatile organic compounds.
30 Some modest amount of particle-bound water may be present at the 40 % relative humidity at
31 which filter samples are equilibrated. However, in the case of continuous measurement
June 2003 9-19 DRAFT-DO NOT QUOTE OR CITE
-------
Q.
Q
D)
O
03
E
T3
Should be
retained
0.1
(NH4)XS04
X= 0 to 2
elemental carbon
Mineral/Metal
1.0 2.5
Aerodynamic Diameter (|jm)
;;;;; Semivolatile components subject to evaporation during or after sampling
Figure 9-7. 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 techniques, particle-bound water must be reduced in situ in order to avoid measurement of large
2 amounts of particle-bound water that would be present at higher relative humidities,. One
3 technique is to stabilize PM at a specified temperature high enough to remove all, or almost all,
4 particle-bound water. This results in loss of much of the semivolatile PM. Examples include the
5 tapered element oscillating microbalance (TEOM) operated at 50 °C and beta gauge monitors
6 with heated inlets. Another technique is the use of a diffusion denuder to remove water vapor
7 without heating. Examples include the Brigham Young absorptive sampler and Harvard
8 pressure drop monitor. The three approaches give different mass concentrations, especially in
9 air sheds with high nitrate, wood smoke, or secondary organic aerosols. Current PM standards
10 are based on health effects studies mainly using filter techniques. However, the need to provide
11 new real time information to the public and the economic pressure to replace filter samplers with
June 2003
9-20
DRAFT-DO NOT QUOTE OR CITE
-------
1 continuous monitors will require a better understanding of the physics and chemistry of the
2 semivolatile components of PM and studies of the potential health effects of these components.
3
4 9.4.2 Mass Concentrations
5 Data for ambient PM2 5 and PM10 concentrations are obtained routinely by networks
6 operated by various state and local agencies. Data are also collected as part of research efforts
7 by governmental, academic and industrial groups. Data from state and local agencies are stored
8 in the AIRS (Aerometric Information Retrieval System) data base, maintained by the U.S.
9 Environmental Protection Agency. Concentrations of PM10_2 5 based on FRM PM10 and PM2 5
10 monitors are estimated by taking the difference between these two measurements. The spatial
11 coverage and frequency of sampling depends on the resources of the agency carrying out the
12 monitoring. Thus, the amount of data collected in a given urban area varies across the United
13 States.
14 The median PM2 5 concentration was 13 |ig/m3 in the United States on a county basis, for
15 1999 to 2001. The corresponding median PM10_25 concentration was about 10 |ig/m3 for the
16 same period. However, there was a good deal of variability in the annual means in different
17 environments in the United States. The mean PM2 5 concentration was below 7 |ig/m3 in 5% and
18 below 17 |ig/m3 in 95% of counties that met minimum AIRS data completeness criteria for
19 calculation of an annual mean concentration (at least 11 days data for each calendar quarter).
20 The mean PM10_2 5 concentration was below 4 |ig/m3 in 5% and below 21 |ig/m3 in 95% of
21 counties meeting the criteria given above. Mean PM2 5 and PM10_2 5 concentrations reported by
22 the IMPROVE network were considerably lower than the lowest 5th percentile values reported by
23 state and local agencies.
24
25 9.4.3 Physical and Chemical Properties of Ambient PM
26 Physical and chemical properties of fine-mode and coarse-mode particles that are produced
27 by sources listed in Table 9-1 are summarized in Table 9-2. It can readily be seen that fine and
28 coarse particles show striking differences in the nature of their sources, their composition, and
29 hence, their chemical properties, and in their removal processes. Differences in sources and
30 removal processes for fine and coarse particles account for many differences in their behavior in
31 the atmosphere. The much shorter atmospheric lifetimes of coarse particles compared to fine
June 2003 9-21 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-2. COMPARISON OF AMBIENT PARTICLES, FINE (ultrafine plus
accumulation mode) AND COARSE
Fine
Coarse
Ultrafine
Accumulation
Formation
Processes:
Formed by:
Combustion, high-temperature
processes, and atmospheric reactions
Nucleation
Condensation
Coagulation
Composition:
Solubility:
Atmospheric
half-life:
Removal
Processes:
Travel
distance:
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 9-3. CONCENTRATIONS OF PM?<, PM,n,« AND SELECTED ELEMENTS
-2.51
l!0-2.5,
IN THE PM2 5 AND PM10 2 5 SIZE RANGE
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
ND*
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
ND
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
ND
2.0
14
52
0
ND
3.0
13
Source: Zweidinger et al. (1998); Pinto et al. (1995).
*ND = non-detectable level
June 2003
9-23
DRAFT-DO NOT QUOTE OR CITE
-------
1 different fine/coarse
relationships
2 PM2 5 from several sites
3 the eastern, interior,
and
4 not shown because their
in
theU.S
are given in Table 9-3. The major chemical components of
. Environmental
Protection Agency's speciation network in
western parts of the United States
are shown in Figure 9-8. Metals are
concentrations are much lower than the components shown.
5 Concentrations of ammonium, nitrate, and sulfate ions tend to be
6 and central United States compared to those in
the western
higher at sites in the eastern
United States (except for the
7 Riverside site). Concentrations of elemental and organic carbon
8 United States
9
10
32 -i
30-
28-
ffT 26-
Jl 24-
S* 22 -
"c* 20~
0 18-
"g 16-
'Ej 14-
O 12~
§ 10-
O 8-
N 6-
E 4-
2 -
n
E33
\
(although
Burlington, VT
Ea<
m
nn:
I
~*
>i
Philidelphia, PA
values are
m
I
>
w
3
rt-
B)
g
ra
__
nn:
I
highest at the Riverside
in
~~~~
Detroit, Ml
In
in
M
=
nn:
0 1
- i
(B WV
a flTf
I $$v^
terior
w
r~
o
c
S
O
w
1
i 1
c m
S- |
M
x y
i ^
site).
raimimm»^B;#
Minneapolis, MN
MI 1 1 1 1 1 1 1 1 ff^yi — Y//
are broadly similar across the
tit*~_*
Boulder, CO
BJ8HIIIIIIIIIIB^I=M
¥ »C3l „
i
I
i
w
i
nn:
H? __w Us
1
3!
i.
ai
Y////////A
1 VD 'xnopjqny - «
Crustal
Elemental
Carbon
Sulfate
Nitrate
Ammonium
Figure 9-8. Major chemical components of PM25 as determined in the U.S. Environmental
Protection Agency's national speciation network from October 2001 to
September 2002.
June 2003
9-24
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.5 EXPOSURE TO PARTICULATE MATTER AND CO-POLLUTANTS
2 For airborne particulate matter (PM), an individual's total personal exposure is ideally
3 based on measurements of the PM concentrations in the air in the individual's breathing zone as
4 the individual moves through space and time. Total personal exposure includes exposure to
5 ambient pollutants while outdoors, exposure while indoors to ambient pollutants that have
6 infiltrated indoors, exposure while indoors to indoor-generated pollutants, and exposure to
7 pollutants generated by an individual's personal activities (personal cloud) that are not recorded
8 by outdoor or indoor monitors. Epidemiological studies frequently use ambient PM
9 concentration as a surrogate for personal exposure to ambient PM. Therefore, an important issue
10 for exposure analysis is determination of the quantitative relationships between concentrations of
11 particulate matter and gaseous co-pollutants measured at stationary community air-monitoring
12 sites (ambient pollution) and the contributions of these concentrations to personal exposures.
13 It is useful to separate these relationships into two components: (a) the relationship between
14 central site concentrations and outdoor concentrations; and (b) the relationship between outdoor
15 concentrations and personal exposures to ambient PM.
16
17 9.5.1 Central Site to Outdoor Relationships
18 The first component to be examined is the relationship between ambient PM
19 concentrations measured by a central monitor, located at a site presumably representative of the
20 community (or the average of several such sites), and the outdoor ambient PM concentration just
21 outside an indoor microenvironment such as a home.
22
23 9.5.1.1 Exposure for Acute Epidemiology
24 In acute time-series studies, daily deaths (or other health effects) are regressed against the
25 daily ambient PM concentrations as measured at a single site (or the average of several sites) in a
26 city. Spatial variations in daily exposure can lead to errors in the estimated relative risk. Under
27 the assumption of a linear relationship between exposure and effect, analysis of exposure error
28 suggests that a key indicator of the effect on epidemiologic results of spatial variations in
29 exposure will be the strength of the daily site-to-site correlations of ambient PM concentrations.
30 However, if the relationship were nonlinear, spatial variability in concentration might be more
31 important. Chapter 3 presents analyses of spatial variability based on a substantial body of new
June 2003 9-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 monitoring data from AIRS. An adequate characterization of the PM concentrations found in
2 urban areas cannot be obtained by considering only annual average concentrations for the whole
3 urban area. There can be considerable spatial and temporal variability in the concentration
4 fields. Typically, annual mean concentrations are within 5 |ig/m3 of each other in urban areas
5 metropolitan statistical areas (MSAs). The spread in values can be much greater if Consolidated
6 MSAs (CMSAs) are considered. Even within some MSAs, concentrations measured at separate
7 sites on individual days can differ by over 100 |ig/m3.
8 Pairs of sites within MSAs are correlated with each other to varying degrees, depending on
9 the urban area. There are some very general regional patterns evident in the data base in which
10 sites tend to be more highly correlated with each other in the eastern United States and less well
11 correlated with each other in the western United States. Site-to-site correlations tend to be
12 higher for a site pair where both are dominated by regional PM than for a site pair where one is
13 more strongly influenced by local sources. Correlation coefficients are smaller for the PM10_25
14 data than for the PM2 5 data, indicating a higher degree of spatial variability for PM10_2 5.
15 However, it should be noted that at least some of this enhanced variability may be due to errors
16 generated by the difference technique that is used to calculate PM10_25 concentrations. The
17 exceptions to very general patterns are frequent enough to prevent extrapolation from one city to
18 another without first examining the data. Although sites may be highly correlated with each
19 other within an MSA, this does not mean that the concentration fields are uniform, as illustrated
20 by Figure 9-9 for three urban areas. Concentrations for the three site pairs chosen are all well
21 correlated with each other (r > 0.9), but the concentrations display different degrees of
22 uniformity. A range of correlations of PM2 5 concentrations were found between monitoring
23 sites in the cities chosen for analysis. PM10 and TSP sites were frequently chosen to monitor
24 specific local point or area sources. However, PM2 5 sites are chosen primarily to be
25 representative of community exposures. Still it would be wise to check the representativeness of
26 a site before choosing a site or group of sites to provide a representative community
27 concentration for exposure or epidemiologic studies.
28
29 9.5.1.2 Exposure for Chronic Epidemiology
30 In chronic studies, total or annual deaths in large cohorts in different cities are regressed
31 against long-term or annual average concentrations in the different cities. Few analyses of
June 2003 9-26 DRAFT-DO NOT QUOTE OR CITE
-------
Columbia SC 1999 & 2000
0>
o
2
o
o
O
i^
o
0)
1
z
120 -
100 -
80 -
60 -
40 -
20 -
g
45-079-0007 vs 45-079-001
r = 0.97
COD = 0.06
P90 = 2.7/N ^ & £ ^ fr
\ \" \ \ *V"^ O/
30 -,
Chicago IL 2000
17-031-2001 vs 17-031-4201
Z 25 -
0)
1 20-
o
o
0 15 -
•5
° 10 -
0)
.a
E 5 -
^ n
r = 0.94
COD = 0.14
P90 = 5.5 fig/m3
I.M .
Detroit Ml 2000
8 12-
€ 10 -
§ 8-
o
•B 6-
CD 4 ~
2 n
1
26-099-0009 vs 26-163-0033
r=0.93
COD = 0.22
1
I
P90= 12.7Ag/m3
II
I. .1 . 1
Concentration Difference (/(g/m3)
Figure 9-9. Occurrence of differences between pairs of sites in three MSAs. The absolute
differences in daily average VM2S 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 al. (2003).
June 2003
9-27
DRAFT-DO NOT QUOTE OR CITE
-------
1 exposure error have been performed for this case. However, the key consideration for chronic
2 studies might be differences in the annual (or seasonal) averages in different parts of a city.
3 Prior to 1998, there was little information on the variations of long-term PM concentration
4 averages within cities. Some information on the spatial variations in long-term (seasonal)
5 averages are reported in Chapter 3 of this document, based on data from AIRS.
6
7 9.5.2 Home Outdoor Concentrations Versus Concentrations of Ambient
8 PM Infiltrated Indoors
9 9.5.2.1 Mass Balance Model
10 It is useful to review some concepts derived from the equilibrium mass balance model,
11 discussed in detail in Chapter 5. The ratio of the ambient PM concentration outdoors, C, to the
12 concentration of ambient PM that has infiltrated indoors, C(AI), is given by the infiltration factor
13 where P is the particle penetration efficiency, a is the air exchange rate, and k is the deposition
14 rate.
15
16 C(AI)/C = Pa/(a+k) = Fmp (the infiltration factor) (9-1)
17
18 As will be discussed later, P and k are functions of the particle size, so FINP will also depend on
19 particle size. The mass balance equation may be modified to include particle removal by air
20 handling systems and to account for nonequilibrium behavior.
21 While indoors, a person will be exposed to a concentration of ambient pollution given by
22 C *FINF. However, while outdoors a person will be exposed to the full ambient concentration.
23 The infiltration factor and the fraction of time outdoors may be used with the ambient
24 concentration to estimate the ratio of the ambient PM exposure (while indoors and outdoors) to
25 the ambient PM concentration, where y = the fraction of time spent outdoors,
26
27 A/C =y + (l-y)FINP =y + (l-y)Pa/(a+k) = a (the attenuation factor). (9-2)
28
29 Since y and a may vary from day to day and person to person and P and k will vary with particle
30 size, cc will also be a variable.
31
June 2003 9-28 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.5.2.2 Separation of Total Personal Exposure into its Ambient and Nonambient
2 Components
3 A person's total exposure to PM or other pollutants includes a nonambient component,
4 usually divided into a component due to indoor-generated pollutants that are evenly distributed
5 through out the house and a component, sometimes called the personal cloud, due to activities of
6 the person that generate pollutants which influence that person more than other persons in the
7 same house. Thus, total personal exposure, T, equals the sum of ambient exposure, A, and
8 nonambient exposure, N:
9
10 T = A+N (9-3)
11
12 A key variable of interest is A, the ambient exposure, i.e., the contributions of paniculate matter
13 and gaseous co-pollutants measured at stationary outdoor air-monitoring sites to actual personal
14 exposures, not T, the total personal exposures due to ambient and indoor-generated pollutants.
15 However, it is not possible to measure A or N directly. Only T and C can be measured directly.
16 The infiltration factor, used to estimate the concentration of ambient PM concentration indoors,
17 \C(AI) = C *FINP\, and the attenuation factor, used to estimate the ambient exposure, [A = C •
18 a], are important because these factors may be estimated from exposure measurements and used
19 to estimate A, the ambient component of total personal exposure.
20 In recent years, the need to separate personal exposure into ambient and nonambient
21 components has been recognized, techniques for separating total personal exposure into its
22 ambient and nonambient components have been recommended, several papers have reported
23 regressions which give average values of a and N, and one paper has reported individual, daily
24 values of A and the distribution of individual, daily values of a.
25
26 Average Values
27 As shown in Figure 9-10, regression of individual measurements of personal exposure on
28 the corresponding measurements of ambient concentrations yields two components of total
29 exposure, one dependent on concentration, one not (T= 00 + GjC). Exposure analysts associate
30 the component independent of concentration, 00, with cohort average nonambient exposure and
31 the component dependent on concentration, 9l3 with alpha, a, the ratio of ambient exposure to
June 2003 9-29 DRAFT-DO NOT QUOTE OR CITE
-------
CO
E
1
250
.« 200-
° 150-
o
§ 100 H
"ro
o
50-
0-
• T= 00 + 9, C or T = N + aC = N + A
o
50
100
150
200
250
Ambient PM Concentration, Ct (M9/m )
Figure 9-10. 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
2
3
4
5
6
1
8
9
10
ambient concentration (T = N+ aC = N + A; Dockery and Spengler, 1981; Ott et al., 2000;
Wilson et al., 2000). Most exposure studies report the correlation between ambient
concentrations and personal exposure, and many of these also report the slope of the
relationship. Since the slope may be interpreted as the average alpha there are a number of
studies from which estimates of the average alpha may be estimated. However, the slope may
not accurately reflect the average alpha unless the data has been examined for outliers. Several
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).
June 2003
9-30
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
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
(Ebelt et al., 2000; Sarnat et al., 2000). 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 9-11). 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, TV cannot be a confounder in a regression of
health effects on ambient concentration (Zeger et al., 2000). For the PTEAM data, the
correlation coefficient of TV with C was r = 0.05.
100
E 75H
75>
.
•2
50-
O oj
J«S
^ E 25
50
100
150
200
250
Ambient PM Concentration, C, (|jg/m3)
Figure 9-11. 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).
June 2003
9-31
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.5.3 Variability in the Relationship Between Outdoor Concentrations and
2 Personal Exposures
3 The values of the infiltration factor (FINF)and the attenuation factor (a) may vary from
4 person-to-person as shown by the distribution of the infiltration factor and attenuation factor in
5 the PTEAM study (Wilson et al., 2000). The average value of the air exchange rate, and
6 therefore the average value of the attenuation factor may vary from season-to-season and from
7 city-to-city due to differences in climate. The variation in average attenuation factor across
8 cities, as estimated by city-to-city air-conditioning use, can explain some of the variation in the
9 quantitative effects of particles on health across cities (Figure 9-12). For a given PM
10 component, the air exchange rate (a) is a major factor in determining the relationship between
11 outdoor and personal exposure. This has been shown in a study in which personal exposure data
12 were classified into three groups based on home ventilation status. High attenuation factor
13 values and high correlations were found for the well-ventilated homes, lower values for
14 moderately well-ventilated homes, and much lower values for poorly ventilated homes. The
15 attenuation factor, a, will be low for a home that is tightly closed for heating or air conditioning,
16 but high for a home with open windows. The air exchange rate also increases as the temperature
17 difference between indoors and outdoors increases. A temperature difference of 10 °C can
18 almost double the air exchange rate over no indoor/outdoor temperature difference for a tight
19 home with no windows open. Variations in wind speed, and direction appear to have a minimal
20 influence on the air exchange rate, especially in homes with tighter construction.
21 Information on the infiltration rate, FINP , as a function of particle size may be obtained as
22 follows. Indoor and outdoor measurements of PM concentrations as a function of particle size
23 are made during the night when it is assumed that there are no indoor activities occurring that
24 might generate indoor PM. Under this assumption the indoor concentration measurement is
25 C(AI) and C(AI)/C = FINF (Long et al., 2000). As can be seen in Figure 9-13, FINP is low for
26 ultrafme and coarse particles but high for accumulation mode particles. FINP also depends on the
27 air exchange rate; i.e., FINF increases when the air exchange rate (a) increases. The variation of
28 the particle penetration efficiency and deposition rate as a function of particle size can also be
29 determined by this technique (Long et al., 2000). There is little information on ambient
30 concentration - exposure relationships for specific chemical components, except sulfate, or for
31 specific source categories, other than what would be inferred from the size distributions.
June 2003 9-32 DRAFT-DO NOT QUOTE OR CITE
-------
0,0025
0,0020
§
"3
E
o
o
o
o
0.0010 -
0.0005 -
0.0000
Winter peaking cities
Non-winter peaking cities
'O.
O
10 20 30 40 50 60 70 80
Central Air Conditioning {%)
Figure 9-12. Percentage of homes with air conditioning versus the regression coefficient
for the relationship of cardiovascular-related hospital admissions 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., the lower attenuation factor, cc, and therefore a lower
regression coefficient (increase in risk per increment in PM10 exposure).
Source: Janssen et al. (2002).
1
2
3
4
5
9
10
11
Infiltration ratios are low for components like strong acidity (H+) that are neutralized by indoor-
generated ammonia or like ammonium nitrate (NH4NO3) that evaporate indoors.
9.5.4 Exposure Relations for Co-Pollutants
The key issue is whether the gaseous co-pollutants (CO, NO2, SO2, and O3) likely
contribute to the health effects attributed to PM or whether they are more likely to serve as
surrogates for PM. To the extent that the gaseous co-pollutants may contribute to the health
effects attributed to PM in a single pollutant, community time-series epidemiologic analysis,
they could confound the PM associations, and the health effects attributed to PM would be
overestimated. However, to the extent that the gaseous co-pollutants are more likely serving as
surrogates for PM, i.e., significantly correlated with PM but not contributing to the health effects
June 2003
9-33
DRAFT-DO NOT QUOTE OR CITE
-------
1.0 -
0.9 -
0.8 -
0.7 -
0.6 -
g 0.5
E 0.4 -
_c
0.3 -
0.2 -
0.1 -
u
(B
c
o
0.0
CO
p
p
CNI
p
o
9
CO
p
o
0.1
Summer Fall
8
d
oo •<-
p d
9 00
CD P
9 o
o
9
CN
p
LO
co
9
CNI
O
-T in
99
CO •*
Ci O
co ^r 10 co
C\J CO -4- If)
Particle Diameter (|jm)
Figure 9-13. 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).
1 attributed to PM in the analysis, in a multiple regression the surrogate would share some of the
2 health effect with the causal agent, especially if the surrogate were measured more accurately
3 than the causal agent. Thus, use of a surrogate in a multiple regression would result in an
4 underestimation of the health effects due to PM.
5 In community, time-series epidemiology, in which daily, community-average health effects
6 are regressed against daily ambient concentrations, ambient gaseous co-pollutant can be
7 potential confounders of ambient PM only if (1) both the gas and PM are able to cause the same
8 health effects; (2) personal exposure is correlated with ambient concentrations for both particles
9 and gases respectively; (3) the personal exposure to gases and to particles are correlated; and
10 (4) the ambient concentrations of particles and gases are correlated. Also, the gaseous
11 co-pollutant must not be in the formation pathway of the particles. For example, SO2 and NO2
June 2003
9-34
DRAFT-DO NOT QUOTE OR CITE
-------
1 are in the formation pathway for the sulfate and nitrate components of PM and O3 is a key
2 chemical reactant in the formation of the sulfate, nitrate, and organic compounds of PM.
3 Questions of particular concern from an exposure perspective include (1) How well are the
4 daily ambient concentrations of the gaseous co-pollutants correlated with the daily ambient
5 concentrations of PM (or specific PM components or indicators) and (2) are the daily ambient
6 concentrations of the gaseous co-pollutants correlated with the daily personal exposures to the
7 ambient? In order to answer these questions quantitatively, information would be needed on the
8 spatial variability of PM indicators and the gaseous co-pollutants and on the variability of the
9 factors which control the infiltration factors (penetration factor and deposition or removal rates).
10 Exposure relationships for gaseous co-pollutants were not reviewed in the exposure chapter
11 (Chapter 5) of this document. Although there have been many exposure studies of the gaseous
12 co-pollutants, there has been little analysis of the experimental data in terms relevant to
13 epidemiology. Qualitative information on exposure relationships that may be inferred from the
14 available literature is given in Table 9-4. The relationships are relative and should apply to
15 many but not necessarily all urban areas. Before assuming a level of spatial homogeneity, it is
16 necessary to check the representativeness of any individual site.
17 Based on the estimates in Table 9-4, it might be expected that the correlation between daily
18 ambient concentrations of PM25 and sulfate and personal exposure to PM25 and sulfate would be
19 high and statistically significant, but that this relationship would not be as significant for the
20 gaseous co-pollutants. Two recent studies (Sarnat et al., 2000, 2001) provide new information
21 relevant to possible contributions of gaseous co-pollutants to health effects attributed to PM.
22 Personal exposure measurements were made of NO2, O3, and sulfate (winter and summer) and of
23 SO2 and EC (winter only). Ambient measurements were made of these species (same seasons)
24 and of CO (both seasons). Personal exposures to ambient PM2 5 were estimated by using the
25 daily, individual ratios of personal exposure to sulfate to ambient concentrations of sulfate as an
26 estimate of the attenuation factor for PM25. Correlations among ambient concentrations, among
27 personal exposures, and between ambient concentrations and personal exposures were examined.
28 Daily personal exposures to NO2 and O3 were not significantly correlated with daily
29 ambient concentrations of those gaseous co-pollutants in either summer or winter. This suggests
30 that NO2 and O3 are unlikely to confound the health effects associations attributed to PM in an
31 epidemiologic analysis using daily ambient concentrations. In the winter, daily personal
June 2003 9-3 5 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-4. QUALITATIVE ESTIMATES OF EXPOSURE VARIABLES
Highest
High
Medium
Low
Lowest
Spatial Homogeneity1
SO4=, Secondary PM2 5
PM254
Primary PM2 5, PM10_2 5,
N02,03, S02,EC5
CO
UF6, trace metals
Infiltration Factor2
CO
PM2 5, SO4=, EC5
NO2
PM10.2.5
UF6, 03, S02
Stability of the
Infiltration Factor3
CO
PM2 5, SO4=, EC5
N02,PM10.2.5,UF6
O3, SO2
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. High in some cities, medium in others.
5. Elemental carbon.
6. Ultrafine particles.
Source: U.S. Environmental Protection Agency (1993, 1996b, 2000a); (Monn, 2001).
1 exposures to SO2 were negatively correlated with daily ambient concentrations of SO2. Personal
2 exposures to CO were not reported. During summer, O3 and NO2 were positively and
3 significantly associated with PM2 5; the association with CO was positive but not significant.
4 During winter, CO and NO2 were positively and significantly associated with PM2 5 while O3 was
5 negatively and significantly associated with PM2 5; the association with SO2 was negative but not
6 significant. Similar associations of gaseous co-pollutants were found with personal exposure to
7 PM2 5 except that the winter association with SO2 became significant. Also, the significant
8 associations were more significant with personal exposure to ambient PM2 5. This indicates that
9 daily ambient concentrations of CO, NO2, O3 and SO2 can be surrogates for daily ambient
10 concentrations of PM25 but that exposure and epidemiologic analyses including O3 and SO2 need
11 to examine relationships on a seasonal basis. These studies also suggest that, for the Baltimore
12 data set, daily ambient concentrations of PM25, CO, NO2, O3 and SO2 may serve as surrogates
13 for daily personal exposures to PM2 5 and may even be better surrogates for daily personal
14 exposures to ambient PM2 5. Thus, for similar urban situations, in a multiple regression using
June 2003 9-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 ambient PM2 5 concentrations and a gaseous co-pollutant, both variables would likely be
2 surrogates for personal exposure to ambient PM2 5.
3 Sarnat et al. (2001) point out that "it is inappropriate to treat one variable as a confounder
4 of another when both variables are actually surrogates of the same thing." While the exposure
5 results from these studies are based on a small number of non-randomly chosen subjects and
6 therefore cannot be extrapolated with assurance to other situations, they do indicate the value of
7 exposure analysis in identifying which of several collinear variables are likely to be causal. The
8 work also suggests that neither NO2, O3, nor SO2 are likely to confound the reported associations
9 of ambient PM with health effects. No information was found on the correlation of ambient CO
10 with personal exposure to CO in homes with no indoor CO sources. However, the low spatial
11 homogeneity of ambient CO concentrations suggests that the relationship would be weak.
12 Therefore, it seems likely, but not certain, that exposure relationships would also indicate that
13 CO is unlikely to confound the health effects associations attributed to PM. It is important to
14 understand that this does not indicate that these ambient pollutants do not cause health effects of
15 the type associated with PM in epidemiologic analyses.
16 Sarnat et al. (2001) also suggest that some of the gaseous co-pollutants may be acting as
17 surrogates for specific PM25 source categories or components. "For subjects with COPD,
18 ambient CO and NO2 were not significantly associated with total personal PM2 5, but were
19 significantly associated with personal exposure to PM25 of ambient origin and also to personal
20 elemental carbon (EC). These significant associations may be due to the fact that motor vehicles
21 are a major source of CO, NO2, EC, and, to a lesser degree, to PM2 5 of ambient origin.
22 Conversely, ambient CO and NO2 were not significantly associated with personal sulfate, a
23 pollutant not associated with motor vehicle emissions. O3, in contrast, was predominantly
24 associated with personal sulfate (positively in summer and negatively in winter) . . ." Thus, CO,
25 NO2, EC, and PM2 5 may be surrogates for personal exposure to pollutants from motor vehicles
26 and O3 may be a surrogate for regional sulfate. It should be noted that since PM2 5, CO, NO2,
27 EC, and PM associated with motor vehicles are all correlated with each other to some extent,
28 a community, time-series epidemiologic analysis, in one community for one time period, cannot
29 tell whether a variable is actually responsible for relationship between concentration and health
30 effects observed in the analysis, or whether the variable is a surrogate for the causal variable.
31 In order to more clearly differentiate between contributor and surrogate, it will be necessary to
June 2003 9-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 integrate information from toxicology and exposure analysis, as well as from epidemiologic
2 studies in different time periods and different communities.
3
4 9.5.5 Exposure Relationships for Susceptible Subpopulations
5 Children, the elderly, and people with pre-existing diseases such as diabetes, respiratory
6 disease, and cardiovascular disease appear to constitute susceptible subpopulations. A number
7 of studies of small cohorts drawn from these and other subpopulations have been conducted
8 recently by EPA and other organizations. Correlations between ambient concentrations and total
9 personal exposure have been presented for a few of these. However, most of the studies have
10 not yet been published, most of the studies have not reported the ambient exposure, and the
11 studies have not been analyzed to determine if there are indeed exposure differences between
12 susceptible groups and the general population.
13 An analysis of cohort exposure studies available in 1998 (Wallace, 2000) concluded that
14 the personal cloud component of nonambient exposure was less for subjects with COPD than for
15 the general population, healthy elderly subjects or children, presumably because of the higher
16 activity level of younger or healthier subjects. However, the relationship between ambient
17 concentrations and personal exposure for COPD patients was not better than that for other
18 cohorts. Wallace (2000) noted that the desirable correlation is that "between personal exposure
19 to particles originating outdoors and outdoor concentrations." However, at that time there was
20 no information on the ambient component of personal exposure. There is still no published
21 information that would suggest differences in exposure relationships for healthy versus
22 susceptible populations.
23
24 9.5.6 Air Pollutants Generated Indoors
25 Total personal exposure includes both ambient and nonambient sources. Important sources
26 of indoor PM are smoking, cooking, and cleaning. Because of the variation of Finfwhh particle
27 size, ambient-infiltrated PM tends to be primarily in the accumulation mode. However, indoor
28 PM is generated primarily in the ultrafme mode (smoking, other combustion sources, most
29 cooking) or the coarse mode (cleaning, sauteing). Another, possibly important indoor source, is
30 the reaction of ambient-infiltrated ozone with indoor emissions of terpenes from air fresheners or
31 cleaning agents, e.g., cleaning with Pine Sol. These particles are generated largely in the
June 2003 9-3 8 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
ultrafine mode, as is the case with analogous nucleation bursts that occur at times in ambient air
as the result of similar reactions with natural terpenes and other reactive hydrocarbons. Ambient
and indoor generated PM also differ somewhat in their chemical composition as shown in
Table 9-5.
TABLE 9-5. 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 9.6 DOSIMETRY: DEPOSITION AND FATE OF PARTICLES IN
2 THE RESPIRATORY TRACT
3 Knowledge of the dose, deposition patterns, and fate of particles delivered to a target site
4 or sites in the respiratory tract is important for understanding possible health effects associated
5 with human exposure to ambient PM and for extrapolating and interpreting data obtained from
6 studies of laboratory animals. The dosimetry of particles of different sizes are subject to large
7 differences in regional respiratory tract deposition, translocation, and clearance mechanisms and
8 pathways and, consequently, retention times. The following sections summarize the current
9 understanding of the physical characteristics of particles and the biological determinants that
10 affect particle dosimetry mechanisms and pathways, as discussed in Chapter 6.
11
12 9.6.1 Particle Deposition in the Respiratory Tract
13 For dosimetry purposes, the respiratory tract can be divided into three regions:
14 (1) extrathoracic (ET), (2) tracheobronchial (TB), and (3) alveolar (A). The ET region consists
June 2003
9-39
DRAFT-DO NOT QUOTE OR CITE
-------
1 of head airways (i.e., nasal and oral passages) through the larynx and represents the areas
2 through which inhaled air first passes. In humans, inhalation can occur through the nose or
3 mouth (or both, known as oronasal breathing). However, most laboratory animals commonly
4 used in respiratory toxicological studies are obligate nose breathers.
5 From the ET region, inspired air enters the TB region at the trachea. From the level of the
6 trachea, the conducting airways then undergo branching for a number of generations. The
7 terminal bronchiole is the most peripheral of the distal conducting airways and these lead,
8 in humans, to the respiratory bronchioles, alveolar ducts, alveolar sacs, and alveoli (all of which
9 comprise the A region). All of the conducting airways, except the trachea and portions of the
10 mainstem bronchi, are surrounded by parenchymal tissue. This is composed primarily of the
11 alveolated structures of the A region and associated blood and lymphatic vessels. It should be
12 noted that the respiratory tract regions are comprised of various cell types and that there are
13 distinct differences in the cells of airway surfaces in the ET, TB, and A regions.
14 Particles deposit in the respiratory tract by five mechanisms: (1) inertial impaction,
15 (2) sedimentation, (3) diffusion, (4) electrostatic precipitation, and (5) interception. Sudden
16 changes in airstream direction and velocity cause inhaled particles to impact onto airway
17 surfaces. The ET and upper TB airways are dominant sites of inertial impaction, a key
18 mechanism for particles with aerodynamic diameter (Da) >1 jim. Particles with Da > 0.5 jim
19 mostly are affected by sedimentation out of the airstream. Both sedimentation and inertial
20 impaction influence deposition of particles in the same size range and occur in the ET and TB
21 regions, with inertial impaction dominating in the upper airways and gravitational settling
22 (sedimentation) increasingly more dominant in lower conducting airways. Particles with actual
23 physical diameters < 1 |im are increasingly subjected to diffusive deposition due to random
24 bombardment by air molecules, resulting in contact with airway surfaces. Particles between
25 0.3 and 0.5 jim in size are small enough to be little influenced by impaction or sedimentation and
26 large enough to be minimally influenced by diffusion, and so, they undergo the least respiratory
27 tract deposition. The interception potential of any particle depends on its physical size; fibers
28 are of chief concern for interception, their aerodynamic size being determined mainly by their
29 diameter. Electrostatic precipitation is deposition related to particle charge; effects of charge on
30 deposition are inversely proportional to particle size and airflow rate. This type of deposition is
31 likely small compared to effects of other deposition mechanisms and is generally a minor
June 2003 9-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 contributor to overall particle deposition, but one recent study found it to be a significant TB
2 region deposition mechanism for ultrafine, and some fine, particles.
3 Deposition of inhaled PM depends primarily on exposure concentrations, physical
4 characteristics of the particles, lung size and structure, tidal volume, and breathing rate.
5 Computer models have proven to be important tools to analyze PM dosimetry. The overall
6 dosimetric model for the respiratory tract consists of several critical elements important for dose
7 calculations including detailed descriptions of morphometry, respiratory physiology, and
8 deposition processes. The morphometric element of the model describes the structure of the
9 respiratory tract and its dimensions. A description of respiratory physiology provides the rates
10 and volumes of inhaled and exhaled air which determines the amount of material that can be
11 deposited in the respiratory tract. Deposition characterizes the initial distribution of the inhaled
12 material within the different regions of the respiratory tract as a function of particle size. The
13 percent deposition as a function of particle size has been calculated with the ICRP model
14 (International Commission on Radiological Protection; ICRP, 1994) using the adult worker
15 default respiratory parameters (see Table 6-3 in Chapter 6). Results for the total percent
16 deposition in the respiratory tract (TOT) and the percent deposition in the ET, TB, and A regions
17 are shown in Figure 9-14. The ET regions filters out some of the particles in the nucleation-
18 mode size range (< 0.01 jim) and the coarse-mode size range (> 1.0 jim). Changing from nasal
19 breathing to mouth breathing results in less deposition of particles in these size regions in the ET
20 and more in the TB and A regions. However, there is little difference in percent deposition
21 between nasal and mouth breathing for particles in the Ailken-mode size range (0.01 to 0.1) and
22 the accumulation-mode size range (0.1 to 1.0). Nasal breathing removes almost all particles
23 > 10 jim. However, mouth breathing allows some particles > 10 jim to deposit in the TB
24 regions.
25 Hygroscopicity, the propensity of a material for taking up and retaining moisture, is a
26 property of some ambient particle species and affects respiratory tract deposition. Such particles
27 can increase in size in humid air in the respiratory tract and, when inhaled, deposit according to
28 their hydrated size rather than their initial size. Compared to nonhygroscopic particles of the
29 same initial size, deposition of hygroscopic aerosols in different regions varies, depending on
30 initial size: hygroscopicity generally increases total deposition for particles with initial sizes
June 2003 9-41 DRAFT-DO NOT QUOTE OR CITE
-------
0.001
0.01 0.1 1 2.5 5 10 25
Particle Diameter, |jm
100
100
C
_o
"-?
'w
o
Q.
0
o
0.001
0.01 0.1 1 2.5 5 10 25 100
Particle Diameter,
0.001
0.01
0.1 1 2.5 5 10 25 100
Diameter,
Figure 9-14. Percent deposition for total results of LUDEP model for an adult male
worker (default) showing total percent deposition in the respiratory tract
(TOT) and in the ET, TB, and A regions. Respiratory parameters given in
Table 6-3. (a) nasal breathing (NB), (b) mouth breathing (MB),
(c) comparison of nasal and mouth breathing for TB and A regions.
June 2003
9-42
DRAFT-DO NOT QUOTE OR CITE
-------
1 larger than =0.5 jam, but decreases deposition for particles between =0.01 and 0.5 and again
2 increases deposition for particles < 0.01 jim.
3 Enhanced particle retention occurs on carinal ridges in the trachea and throughout the
4 segmental bronchi; and deposition "hot spots" occur at airway bifurcations or branching points.
5 Peak deposition sites shift from distal to proximal sites as a function of particle size, with greater
6 surface dose in conducting airways than in the A region for all particle sizes. Surface number
7 dose (particles/cm2/day) is much higher for fine than for coarse particles, indicating much higher
8 numbers of fine particles depositing, with the fine fraction contributing upwards of 10,000 times
9 greater particle number per alveolar macrophage.
10 Ventilation rate, gender, age, and respiratory disease status are all factors that affect total
11 and regional respiratory tract particle deposition. In general, because of somewhat faster
12 breathing rates and likely smaller airway size, women have somewhat greater deposition of
13 inhaled particles than men in upper TB airways, but somewhat lower A region deposition than
14 for men. Children appear to show four effects: (1) greater total respiratory tract deposition than
15 adults (possibly as much as 50% greater for those < 14 years old than for adults > 14 years),
16 (2) distinctly enhanced ET region deposition (decreasing with age from 1 year), (3) enhanced TB
17 deposition for particles < 5 jim, and (4) enhanced A region deposition (also decreasing with
18 age). Overall, given that children have smaller lungs and higher minute volumes relative to lung
19 size, they likely receive greater doses of particles per lung surface area than adults for
20 comparable ambient PM exposures. This and the propensity for young children to generally
21 exhibit higher activity levels and associated higher breathing rates than adults likely contribute to
22 enhanced susceptibility to ambient particle effects resulting from particle dosimetry factors.
23 In contrast, limited available data on respiratory tract deposition across adult age groups (18 to
24 80 years) with normal lung function do not indicate age-dependent effects (e.g., enhanced
25 deposition in healthy elderly adults). Altered PM deposition patterns due to respiratory disease
26 status may put certain groups of adults (including some elderly) and children at greater risk for
27 PM effects.
28 Both information noted in the 1996 PM AQCD and newly published findings discussed in
29 this document indicate that respiratory disease status is an especially important determinant of
30 respiratory tract particle deposition. Importantly, the pathophysiologic characteristics of chronic
31 obstructive pulmonary disease (COPD) contribute to more heterogenous deposition patterns and
June 2003 9-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 differences in regional deposition. One study indicates that people with COPD tend to breath
2 faster and deeper than those with normal lungs (i.e., about 50% higher resting ventilation) and
3 had about 50% greater deposition than age-matched healthy adults under typical breathing
4 conditions, with average deposition rates 2.5 times higher under elevated ventilation rates.
5 Enhanced deposition appears to be associated more with the chronic bronchitic than the
6 emphysematous component of COPD. In this and other new studies, fine-particle deposition
7 increased markedly with increased degree of airway obstruction (ranging up to 100% greater
8 with severe COPD). With increasing airway obstruction and uneven airflow because of irregular
9 obstruction patterns, particles tend to penetrate more into remaining better ventilated lung areas,
10 leading to enhanced focal deposition at airway bifurcations and alveoli in those A region areas.
11 In contrast, TB deposition increases with increasingly more severe bronchoconstrictive states, as
12 occur with asthmatic conditions.
13 Differences between species in particle deposition patterns were summarized in the 1996
14 PM AQCD and more recently by Schlesinger et al. (1997), as discussed in Chapter 6 of this
15 document. These differences should be considered when relating biological responses obtained
16 in laboratory animal studies to effects in humans. Various species used in inhalation toxicology
17 studies serving as the basis for dose-response assessment may not receive identical doses in a
18 comparable respiratory tract region (i.e., ET, TB, A) when exposed to the same aerosol at the
19 same inhaled concentration. This is illustrated by mathematical modeling studies that evaluate
20 interspecies differences in respiratory tract deposition. For example, Hofmann et al. (1996)
21 found total deposition efficiencies for all particles (0.01, 1, and 10 jim) at upper and lower
22 airway bifurcations to be comparable for rats and humans, but when higher penetration
23 probabilities from preceding airways in the human lung were considered, bronchial deposition
24 fractions were mostly higher for humans. For all particle sizes, deposition at rat bronchial
25 bifurcations was less enhanced on the carinas than in human airways. Numerical simulations of
26 three-dimensional particle deposition patterns within selected (species-specific) bronchial
27 bifurcations indicated that interspecies differences in morphologic asymmetry is a major
28 determinant of local deposition patterns.
29 Models are useful for calculating percent deposition for different species. The percent
30 deposition can then be used with exposure concentrations and respiratory parameters (tidal
31 volume, breathing rate, lung size to calculate normalized deposition on a jig per g of lung, jig per
June 2003 9-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 surface area, number of particles per alveoli or other parameters. A comparison of human and
2 rat percent deposition, calculated using the Multiple Path Dosimetry Model (MPPD model)
3 developed by CUT (the Chemical Industry Institute of Technology, USA) and RIVM
4 (Directorate-General for Environmental Protection, The Netherlands), is described by Winter-
5 Sorkina and Cassee (2002). The percent deposition patterns and human/rat ratios will change
6 with changes in activity levels. However, the model can be used to predict the ratios for
7 different activity levels and to choose exposure and activity scenarios to give comparable
8 depositions in humans and animals for extrapolation or design of experimental studies.
9 In a histology study, Nikula et al. (2000) examined particle retention in rats (exposed to
10 diesel soot) and humans (exposed to coal dust). In both, the volume density of deposition
11 increased with increasing dose. In rats, diesel exhaust particles were found mainly in lumens of
12 the alveolar duct and alveoli, whereas in humans, retained dust was mainly in interstitial tissue.
13 Thus, in the two species, different lung cells appear to contact retained particles and may result
14 in different biological responses with chronic exposure.
15 The probability of any biological effect of PM in humans or animals depends on particle
16 dosimetry, and subsequent particle retention, as well as underlying dose-response relationships.
17 Interspecies dosimetric extrapolation must, therefore, consider differences in deposition,
18 clearance, translocation, and dose-response. Even similar deposition patterns may not result in
19 similar effects in different species, because dose also is affected by clearance mechanisms and
20 species sensitivity. Total number of particles deposited in the lung may not be the most relevant
21 dose metric by which to compare species; rather, the number of deposited particles per unit
22 surface area may determine response. Even if deposition is similar in rats and humans, there
23 would be a higher deposition density in the rat because of the smaller surface area of the rat lung.
24 Thus, species-specific differences in deposition density are important when attempting to
25 extrapolate health effects observed in laboratory animals to humans.
26
27 9.6.2 Particle Clearance and Translocation
28 Particles depositing on airway surfaces may be cleared from the respiratory tract
29 completely or translocated to other sites within this system by regionally specific clearance
30 mechanisms, as follow: ET region—mucocialiary transport, sneezing, nose wiping and blowing,
31 and dissolution and absorption into blood; TB region—mucociliary transport, endocytosis by
June 2003 9-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 macrophages and epithelial cells, coughing, and dissolution and absorption into blood and
2 lymph; A region— macrophages, epithelial cells, interstitial, and dissolution and absorption into
3 blood and lymph.
4 Regionally specific clearance defense mechanisms operate to clear deposited particles of
5 varying particle characteristics (size, solubility, etc.) from the ET, TB, and A regions and are
6 variously affected by different disease states. For example, particles are cleared from the ET
7 region by mucociliary transport to the nasopharynx area, dissolution and absorption into the
8 blood, or sneezing, wiping or blowing of the nose; but such clearance is slowed by chronic
9 sinusitis, bronchiectasis, rhinitis, and cystic fibrosis. Also, in the TB region, poorly soluble
10 particles are cleared mainly by upward mucociliary transport or by phagocytosis by airway
11 macrophages that move upward on the mucociliary blanket, followed by swallowing. Soluble
12 particles in the TB region are absorbed mostly into the blood and some by mucociliary transport.
13 Although TB clearance is generally fast and much material is cleared in <24 h, the slow
14 component of TB clearance (likely associated with bronchioles 24 h and clearance
16 half-times of about 50 days. Bronchial mucous transport is slowed by bronchial carcinoma,
17 chronic bronchitis, asthma, and various acute respiratory infections; these are disease conditions
18 that logically would be expected to increase retention of deposited particle material and, thereby,
19 increase the probability of toxic effects from inhaled ambient PM components reaching the TB
20 region. Also, spontaneous coughing, an important TB region clearance mechanism, does not
21 appear to fully compensate for impaired mucociliary clearance in small airways and may become
22 depressed with worsening airway disease, as seen in COPD.
23 Clearance of particles from the A region by alveolar macrophages and their mucociliary
24 transport is usually rapid (< 24 h). However, penetration of uningested particles into the
25 interstitium increases with increasing particle load and results in increased translocation to
26 lymph nodes. Soluble particles not absorbed quickly into the blood stream and translocated to
27 extrapulmonary organs (e.g., the heart) within minutes may also enter the lymphatic system, with
28 lymphatic translocation probably being increased as other clearance mechanisms (e.g., removal
29 by macrophages) are taxed or overwhelmed under "particle overload" conditions. Insoluble
30 particles < 2 jim clear to the lymphatic system at a rate independent of size; particles of this size,
31 more so than those > 5.0 jim, are deposited significantly in the A region. Translocation into the
June 2003 9-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 lymphatic system is quite slow, and elimination from lymph nodes even slower (half-times
2 estimated in decades). Focal accumulations of reservoirs of potentially toxic materials and their
3 slow release for years after initial ambient PM exposure may account partially for the
4 observation in epidemiologic studies that higher relative risks are associated with long-term
5 ambient PM exposure than can be accounted for by additive effects of acute PM exposures.
6 Alveolar region clearance rates are decreased in human COPD sufferers and slowed by acute
7 respiratory infections, and the viability and functioning of alveolar macrophages are reduced in
8 human asthmatics and in animals with viral lung infections. These observations suggest that
9 persons with asthma or acute lung infections are likely at increased risk for ambient PM
10 exposure effects.
11 Differences in regional and total clearance rates between some species reflect differences
12 in mechanical clearance processes. The importance of interspecies clearance differences is that
13 retention of deposited particles can differ between species and may result in differences in
14 response to similar PM exposures. Hsieh and Yu (1998) summarize existing data on pulmonary
15 clearance of inhaled, poorly soluble particles in the rat, mouse, guinea pig, dog, monkey, and
16 human. Two clearance phases, "fast" and "slow," in the A region are associated with
17 mechanical clearance along two pathways, the former with the mucociliary system and the latter
18 with lymph nodes. Rats and mice are fast clearers, compared to other species. Increasing initial
19 lung burden results in an increasing mass fraction of particles cleared by the slower phase.
20 As lung burden increases beyond 1 mg particles/g lung, the fraction cleared by the slow phase
21 increases to almost 100% for all species. The rate for the fast phase is similar in all species, not
22 changing with increasing lung burden, whereas the slow phase rate decreases with increasing
23 lung burden. At elevated burdens, the "overload" effect on clearance rate is greater in rats than
24 in humans.
25
26 9.6.3 Dosimetric Considerations in Comparing Dosages for Inhalation,
27 Instillation, and Exposure of Cultured Cells
28 There are three common experimental approaches for studying the biological effects of
29 particulate material: inhalation, instillation, and in vitro. Inhalation studies are the more
30 realistic physiologically, and thus the most applicable to risk assessment. However, because
31 they are expensive, time consuming and require specialized equipment and personnel, they must
32 be supplemented by other techniques. In vitro studies using live cells are cost-effective, provide
June 2003 9-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 for precise dose delivery, and permit investigators who do not have access to inhalation
2 techniques to perform mechanistic and comparative toxicity studies of particulate material.
3 Commonly, the initial information on likely mechanisms of action of particles is obtained
4 through in vitro techniques. For in vitro studies, dose selection is important because it is easy to
5 overwhelm normal defense mechanisms.
6 Instillation studies, in which particles suspended in a carrier such as physiological saline
7 are applied to the airways, have certain advantages over in vitro studies. The exposed cells have
8 normal attachments to basement membranes and adjacent cells, circulatory support, surrounding
9 cells and normal endocrine, exocrine and neuronal relationships. Thus, instillation experiments
10 can bridge between in vitro and inhalation studies as well as produce useful mechanistic and
11 comparative toxicity information. Although the tracheobronchial region is most heavily dosed,
12 alveolar regions can also be exposed via instillation techniques.
13 It is difficult to compare particle deposition and clearance among different inhalation and
14 instillation studies because of differences in experimental methods and in quantification of
15 particle deposition and clearance. Key points from a recent detailed evaluation (Driscoll et al.,
16 2000) of the role of instillation in respiratory tract dosimetry and toxicology studies are
17 informative. In brief, inhalation may result in deposition within the ET region, the extent of
18 which depends on the size of the particles used, but intratracheal instillation bypasses this
19 portion of the respiratory tract and delivers particles directly to the TB tree. Although some
20 studies indicate that short (0 to 2 days) and long (100 to 300 days postexposure) phases of
21 clearance of insoluble particles delivered either by inhalation or intratracheal instillation are
22 similar, others indicate that the percent retention of particles delivered by instillation is greater
23 than for inhalation, at least up to 30 days postexposure. Another salient finding is that inhalation
24 generally results in a fairly homogeneous distribution of particles throughout the lungs, but
25 instillation is typified by heterogeneous distribution (especially in the A region) and high levels
26 of focal particles. Most instilled material penetrates beyond the major tracheobronchial airways,
27 but the lung periphery is often virtually devoid of particles. This difference is reflected in
28 particle burdens within macrophages, those from animals inhaling particles being burdened more
29 homogeneously and those from animals with instilled particles showing some populations of
30 cells with no particles and others with heavy burdens, and is likely to impact clearance pathways,
June 2003 9-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 dose to cells and tissues, and systemic absorption. Exposure method, thus, clearly influences
2 dose distribution that argues for caution in interpreting results from instillation studies.
3 Dosimetric calculations must be performed to relate tracheobronchial cell exposures from
4 instillation in terms of particle concentrations (on a number of particles per unit surface area
5 basis) to those occurring in human environmental exposures. Such calculations require selecting
6 characteristics associated with the particles, the exposed subject and the environmental exposure
7 scenario. Hence each study can present a unique dosimetric analysis. In most cases, it will be
8 useful to know the relationship between the surface doses in instillation studies and realistic
9 local surface doses that could occur in vivo in human subpopulations receiving the maximum
10 potential dose. Although these subpopulations have not been completely defined some
11 characteristics of individuals do serve to enhance the local surface deposition doses to
12 respiratory tract cells. These characteristics include: exercise and mouth breathing non-uniform
13 inhaled air distribution such as occurs in chronic obstructive pulmonary disease and chronic
14 bronchitis, impaired particle clearance as occurs in some disease states and location near
15 pollutant sources. In addition, even normal subjects exposed by inhalation are expected to have
16 numerous sites of high local particle deposition (specifically at bifurcations) within the
17 tracheobronchial tree.
18 Consideration in Chapter 6 of all these factors that could enhance local surface doses in
19 humans led to the conclusion that an enhancement factor of 3,000 was appropriate to represent
20 the most heavily exposed human epithelial cells. Hence, an instillation of 150 jig in a rat might
21 be expected to represent the enhanced dose to small areas in the human TB region produced by a
22 24-hour exposure to 65 |ig/m3. Well-conducted instillation studies are valuable for examining
23 the relative toxicity of particulate materials and for providing mechanistic information that is
24 useful for interpreting in vitro and inhalation studies. However, because mechanisms of injury
25 may vary with the delivered dose, it would be useful if instillation studies designed to provide
26 information relevant to human risk assessment were accompanied by dosimetric calculations.
27
28 9.6.4 Inhaled Particles as Potential Carriers of Toxic Agents
29 It has been proposed that particles also may act as carriers to transport toxic gases into the
30 deep lung. Water-soluble gases, which would be removed by deposition to wet surfaces in the
31 upper respiratory system during inhalation, could dissolve in particle-bound water and be carried
June 2003 9-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 with the particles into the deep lung. Equilibrium calculations indicate that particles do not
2 increase vapor deposition in human airways. However, these calculations do show that soluble
3 gases are carried to higher generation airways (deeper into the lung) in the presence of particles
4 than in the absence of particles. In addition, species such as SO2 and formaldehyde react in
5 water, reducing the concentration of the dissolved gas-phase species and providing a kinetic
6 resistence to evaporation of the dissolved gas. Thus, the concentration of the dissolved species
7 may be greater than that predicted by the equilibrium calculations. Also, certain other toxic
8 species (e.g., nitric oxide [NO], nitrogen dioxide [NO2], benzene, polycyclic aromatic
9 hydrocarbons [PAH], nitro-PAH, a variety of allergens) may be absorbed onto solid particles and
10 carried into the lungs. Thus, ambient particles may play important roles not only in inducing
11 direct health impacts of their constituent components but also in facilitating delivery of toxic
12 gaseous pollutants or bioagents into the lung and may, thereby, serve as key mediators of health
13 effects caused by the overall air pollutant mix.
14
15
16 9.7 TOXICOLOGIC ASSESSMENT OF PARTICULATE-MATTER
17 PROPERTIES LINKED TO HEALTH EFFECTS
18 Ambient PM comprises a complex mix of constituents derived from many sources, both
19 natural and anthropogenic. Hence, the physicochemical composition of PM generally reflects
20 the major contributing sources locally and regionally. Within this framework of source or
21 origin, PM composition also varies significantly by the size-mode within which it is classified
22 (ultrafme, accumulation, or coarse). It should be clear that any given particle can differ
23 appreciably from another individual particle of similar size, but that the region of origin with all
24 of its contributing sources determines the general composition of the generic PM in that
25 classification mode. By its nature then, exposure to airborne ambient PM constitutes an
26 exposure to what is very clearly a mixture of different particles of differing composition and to
27 other gaseous co-pollutants that coexist in that air-shed.
28 The epidemiology information reviewed in the 1996 PM AQCD and updated in this
29 document convincingly shows that a positive correlation exists between the levels of ambient
30 PM pollution and mortality/morbidity. However, this correlation is based mainly on a mass
31 metric, which is somewhat counter-intuitive considering the complexities in composition of PM
32 and given the typically low ambient concentrations of most PM constituents, even when
June 2003 9-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 fractionated by PM size. What has evolved since the 1996 PM AQCD are notable advances in
2 our understanding that the linkages between PM exposure and health impacts appear to be most
3 strongly related to accumulation mode particles, with combustion-derived PM typically being
4 the most active of the source-based contributors. It is also now better appreciated that discovery
5 of a single "magic bullet" regarding PM physicochemical attributes is not likely to occur, and
6 perhaps the sources from which the PM derive may be the best linkage one can achieve.
7 Approaches to assessing likely "causation" and "biological plausibility" have attempted to
8 integrate the wealth of epidemiologic data with the growing body of toxicology information in
9 order to reveal coherence among the findings that support newly emerging sound hypotheses.
10 Thus, while it is often difficult to separate the physicochemical attributes of PM that may be of
11 health significance from the mechanisms by which individual factor(s) may function in the
12 response, a number of hypotheses have evolved espousing various PM characteristics as
13 potentially significant contributors to the observed health effects (reviewed by Dreher, 2000).
14 Each of the attribute-based hypotheses has a sufficient data base to merit consideration and
15 further investigation.
16 To date, toxicologic studies on PM have provided important, albeit still limited, evidence
17 for specific PM attributes being primarily or essentially responsible for the cardiopulmonary
18 effects linked to ambient PM. In most cases, however, exposure concentrations in laboratory
19 studies have been inordinately high compared to the exposures at which epidemiologic studies
20 have found effects. Reasons for this dosimetric discrepancy range from the limited numbers of
21 animals or human subjects that can be practically studied, the uncertainty and narrow range of
22 responsiveness of the study groups and especially the typically limited use of young, elderly,
23 unhealthy, or otherwise at-high-risk animals or humans, especially in light of poorly understood
24 risk factors. Thus, most of the toxicology data-base resides in the "hazard-identification"
25 compartment of the risk assessment paradigm. However, sufficient coherence in the
26 epidemiologic and toxicological data has provided a level of "plausibility" to the observational
27 studies and thus opened new avenues for investigation to link PM properties and constituents to
28 specific sources and to health outcomes. The primary PM properties thought to be related to
29 health effects are discussed below.
30
31
June 2003 9-51 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
9.7.1 Chemical Components and Source Categories Associated with Health
Effects in Epidemiologic Studies
Epidemiologic studies using either individual chemical species or classes or using source
category factors (SCF) derived from factor analysis have identified a variety of species whose
ambient concentrations are statistically associated with either total mortality or more specific
mortality groupings.
9.7.1.1 Toxicologically Important Components of PM
Inherent in the NRC research agenda (NRC, 1998) was the consideration that one, or
perhaps a few, characteristics of PM would be associated with toxicity, and exposure monitoring
could concentrate on these components. However, such narrowing of focus is not yet possible,
given the wide array of PM characteristics that have been found to be associated with toxicity
either through epidemiologic or toxicologic studies, as listed in Table 9-6.
TABLE 9-6. PARTICIPATE MATTER ASSOCIATED WITH MORTALITY IN
EPIDEMIOLOGIC STUDIES
PM Size Fractions
Ions/Elements
Carbon/Organic Fractions
Mass TSP
Mass PM10
Mass-thoracic coarse PM
[PM1(W.5orPM1(M]
Mass-fine PM
[PM25orPM10]
Mass-ultrafine PM
[PM01]
Particle number
Particle surface area
Sulfate (SO4=)
Nitrate (NO3)
Transition metals (e.g., Ni)
Other toxic metals (e.g., Pb)
Strong Acid (H+)
TC (Total Carbon)
EC (Elemental Carbon)
BC (Black Carbon)
COH (Coefficient of Haze)
OC (Organic Carbon)
CX (Cyclohexene-extractable
Carbon)
June 2003
9-52
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.7.1.2 Source Category Factors
2 A major goral of air pollution-health outcome studies is to relate health outcomes to
3 specific sources of air pollutants. A number of techniques have been developed that apportion
4 PM in ambient samples to its sources (see Section 3.3 of this document and Section 5.5 of the
5 1996 PM AQCD for descriptions of these techniques). These powerful techniques are limited by
6 their ability to resolve PM produced by sources having similar compositional profiles and by the
7 lack of data for the composition (especially the organic composition) of emissions from many
8 sources. This limitation may be mitigated in the future by further analytical developments in
9 analyzing the composition of PM samples. In the meantime, it is probably best to refer to source
10 categories, although the ambiguity is removed when there are unique sources in a given area
11 (e.g., Utah Valley steel mills). There are also three studies in which factor analysis has been
12 used to identify several specific source category factors. In two cases (Laden et al., 2000 and
13 Tsai et al., 2000), the source category factors (SCF) were then used in a multiple regression, the
14 nonsignificant factors were eliminated, and the multiple regression was rerun with only the
15 significant factors. In the third case (Mar et al., 2000), relative risk values are reported for
16 regression with SCF one at a time but the paper states that "Regression analysis with all of the
17 factors included in a multi-source model produced similar results." The similar results in single
18 and multiple regressions and the low correlation between SCF indicates that there is low
19 potential for confounding among the various SCFs.
20 Source categories that have been found to be significantly associated (p < 0.05) with total,
21 cardiovascular, or cardiovascular plus respiratory mortality in one or more cities are shown in
22 Table 9-7. A source category associated with motor vehicles was found in all four studies. The
23 epidemiologic studies do not provide sufficient information to determine whether the causal
24 factor is one or both of the gaseous co-pollutants (CO and NO2); soot particles from cars
25 (indexed by BS, COH, or EC); organic PM from vehicles, transition metals emitted by vehicle
26 (Mn, Fe, Zn); or other particles generated or resuspended by vehicular traffic.
27 The three studies that investigated multiple source categories also found a sulfate factor.
28 The factor reported by Laden et al. (2000) as "coal burning" contains high loadings of both
29 selenium and sulfur and could also have been called "regional sulfate." Mar et al. (2000) refer to
30 the factor with high sulfate specifically as "regional sulfate." They were able to make this
31 connection because they also had a factor with a high loading of SO2 which they called a "local
June 2003 9-53 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-7. 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 SO2" factor. The regression with the elemental S (assumed to be sulfate) was not significant, but
2 the regression with the regional sulfate factor was significant. This may be because the factor
3 analysis will tend to remove other more localized sulfate sources such as CaSO4 and Na2SO4,
4 leaving only acid sulfates ([NH4]2SO4, NH4HSO4, and H2SO4) for a regional sulfate factor. (In
5 Phoenix, there was a modest loading of S in the soil factor.) Therefore, all three sulfate factors
6 should be considered as regional sulfate.
7 The studies of specific chemical components and source categories are especially
8 important because they indicate the association of health effects with the three major
9 components of PM mass: sulfate, nitrate, and organic PM. Examination of PM25 and nitrate
10 effects, alone and in multiple regressions, indicates that PM2 5 and nitrate were not confounded
11 by NO2, CO or O3 in Santa Clara, CA (Fairley, 1999). Examination of the lag structure from the
12 Phoenix study reveals that neither the regional sulfate factor nor the vegetative burning factor
June 2003 9-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 was confounded by NO2, CO, SO2, or O3. The epidemiologic results suggest the need for
2 toxicologic studies of the sulfate, nitrate, and organic components of PM, including studies with
3 compromised or susceptible subjects.
4 All of the studies that investigated multiple source categories found a soil or crustal source
5 that was negatively associated with mortality. This suggests that the components of natural soil
6 may have minimal toxicity unless contaminated by anthropogenic sources, such transition metals
7 or polyaromatic hydrocarbons. In any event, the epidemiologic associations suggest additional
8 PM components that should be investigated in toxicologic studies. Although results such as
9 those presented above are illuminating, it should be noted that there can be ambiguity regarding
10 the identification of source categories as the marker elements used in many of the methods used
11 (e.g., specific rotation factor analysis) can have more than one source. As an example, before
12 lead was phased out of gasoline it was (and still is) produced by smelters and other industries
13 (see Appendix 3D of Chapter 3). Methods such as principal components analysis do not have
14 optimal weighting of the factors, thereby leading to distortion in the results and although newer
15 methods such as positive matrix factorization overcome many of these difficulties, the results are
16 still subject to some degree of rotational ambiguity. In addition, there can be substantial spatial
17 variability in source contributions across an urban area leading to the potential for exposure
18 characterization error.
19
20 9.7.2 Specific Properties of Ambient PM Linked to Health Effects
21 9.7.2.1 Physical Properties
22 Fine and Thoracic Coarse Particles: In contrast to ultrafme particles, the respective roles of
23 PM2 5 (indicator for fine PM) and PM10_2 5 (indicator for thoracic coarse PM) in defining health
24 outcomes have garnered considerable research attention because they are the most frequently
25 measured size-fractions of ambient PM and for which most health effects data exist. The fine
26 fraction comprises most of the combustion-related constituents discussed below under chemical
27 properties. The fine fraction has greater surface area than the thoracic coarse fraction, but much
28 less surface area and particle number than the ultrafme fraction. To the extent that inhaled PM
29 may carry chemicals or reactive species on their surfaces, these smaller size fractions may have
30 an additional dimension to their toxicity (in terms of surface chemical bioavailablilty) that is not
31 found with coarse PM. For example, acute exposure to sulfate-coated carbon black was found to
June 2003 9-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 impair alveolar macrophage phagocytosis and intrapulmonary bactericidal activity in mice
2 (Jakab et al., 1996; Clarke et al., 2000). On the other hand, coarse PM usually is of mineral
3 (earthen) or biologic (discussed below) origin and, thus, has a less complex bioavailable
4 chemical matrix than the finer PM mode. The relative toxicity of most earthen-derived PM has
5 been observed to be less than that of the finer combustion-derived or surrogate ultrafme
6 particles. However, because ambient coarse PM would tend to impact on the airways of humans,
7 it is thought this fraction may be adverse to those with airways sensitivities or disease (e.g.,
8 asthma).
9
10 Ultrafme Particles (Size. Surface Area. Number): The physical attributes of PM - size,
11 surface area and number - are intimately interrelated. These properties influence lung
12 deposition, penetrance and persistence in lung tissues, and systemic transport, and, in several
13 studies, apparently the inherent toxicity of the particle itself. While a few epidemiologic studies
14 (Wichmann et al., 2000) show correlations between health outcomes and ultrafme (<100 nm)
15 ambient PM, the bulk of the information regarding its toxic potential, and the role of surface
16 area, has derived from studies of surrogate insoluble particles, such as mineral oxides (e.g.,
17 TiO2) and carbon black (Oberdorster et al., 1994; Osier and Oberdorster, 1997; Li et al., 1997,
18 1999). These studies have shown that on an equivalent mass exposure-dose metric, ultrafme PM
19 can induce more acute lung injury than fine PM. Similarly, surrogate PM with high surface
20 areas induced more toxicity than those of like composition, but having smaller surface areas
21 (Lison et al., 1997). On the other hand, studies have shown that composition also matters; for
22 example MgO ultrafmes produce less injury than ZnO (Kuschner et al., 1997), as did sparked
23 carbon versus similarly generated metal oxides (Elder et al., 2000).
24 As with acid aerosols, studies of ultrafme particles have focused largely on effects in the
25 lung, but inhaled ultrafme particles may also have the potential to be distributed systemically and
26 have effects that are independent of lung effects. Recent epidemiologic studies evaluating blood
27 viscosity as a biologic correlate of ultrafme exposures, have reported slight increases that raise
28 the prospect of potential cardiovascular implications (Wichmann et al., 2000).
29
30
June 2003 9-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.7.2.2 Chemical Properties
2 Acid Aerosols: There is relatively little new information on the effects of acid aerosols,
3 and the basic conclusions of the the 1996 PM AQCD remain unchanged. It previously was
4 concluded that acid aerosols cause little or no change in pulmonary function in healthy subjects,
5 but asthmatics may experience small decrements in pulmonary function. Long-term exposures
6 of animals to acid aerosols, on the other hand, have been shown to alter airway morphology with
7 epithelial cell desquamation and an increase in secretory cells, but these changes have been
8 considered relatively minor. The conclusions about the acute health effects, however, are
9 supported by a study by Linn and colleagues (1997), in which healthy children (and children
10 with allergy or asthma) were exposed to sulfuric acid aerosol (100 |ig/m3) for 4 hours. While
11 there were no significant effects on symptoms or pulmonary function when the entire group was
12 analyzed, the allergy group did have significant acid-related increases in symptoms, although the
13 acid concentrations were distinctly higher than typical ambient concentrations. These findings
14 were consistent with those reported for adolescent asthmatics exposed to acid aerosols in earlier
15 studies reported in the 1996 PM AQCD.
16 Although pulmonary effects of acid aerosols have been the subject of extensive research,
17 the cardiovascular effects of acid aerosols have received little attention. One example which
18 raises the issue is a study of acetic acid fumes where reflex mediated increases in blood pressure
19 were found in normal and spontaneously hypertensive rats (Zhang et al., 1997). Similarly, acidic
20 residual oil fly ash (ROFA) PM (which also contains a considerable amount of metal sulfates)
21 was found to alter ecocardiogram (ECG) patterns in the same strain of rats at high air
22 concentrations (Kodavanti et al., 2000). Thus, acidic components should not be entirely
23 dismissed as possible mediators of ambient PM health effects, since so little is known about
24 potential cardiovascular impacts or impacts in compromised subjects.
25
26 Transition Metals: The 1996 PM AQCD relied on data from occupational exposures to
27 initially evaluate the potential toxicity of metals in PM air pollution. Since that time, in vivo and
28 in vitro studies using ROFA or soluble transition metals have contributed substantial new
29 information on the health effects of PM-associated soluble metals. The metals of most interest,
30 notably the transition metals of iron, vanadium, copper, nickel, chromium, cadmium, arsenic, are
31 ubiquitous constituents of PM-derived from anthropogenic fossil fuel emissions. Exposure
June 2003 9-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 seems to be widespread with studies in autopsy specimens (1980's) showing dramatic increases
2 in the content of the first row transition metals in lung tissues of Mexico City residents since the
3 1950's, consistent with industrialization and pollution (Fortoul et al., 1996). Similar studies in
4 North America show metals in the lung tissues of urban dwellers. Although there remain
5 uncertainties about the differential effects of one transition metal versus another, water-soluble
6 or bioavailable metals leached from ROFA or bulk ambient PM cause a variety of biological
7 effects. Many studies show that the action of instilled ROFA and constituent metals are
8 pro-inflammatory (cells, mediators, and molecular signaling processes - in vivo and in vitro),
9 and recently, they have been shown to induce cardiac arrhythmias in animal models (both
10 healthy and diseased). In studies in which various ambient and emission source PM were
11 instilled into rats, the soluble metal content appeared to be the primary determinant of lung
12 injury (Costa and Dreher, 1999). However, these and the related findings on metal toxicity
13 generally have derived from relatively high dose instillation or inhalation exposures, lending
14 them to criticism as to their relevancy for ambient PM that is typically relatively low in metal
15 content.
16 Nevertheless, a series of studies associated with the closing of a metal smelter in Utah
17 Valley, where ambient PM extracts (containing metals and other soluble constituents) were
18 instilled into the lungs of humans (Ohio and Devlin, 2001) and animals (Dye et al., 2001), as
19 well as tested in vitro (Frampton et al., 1999), showed remarkable coherence with epidemiologic
20 studies of hospitalization and mortality (Pope, 1989; Pope et al., 1999b) in the same area and at
21 the same times of the PM samples used in the laboratory studies. The response patterns in each
22 study paralleled the metal content. Furthermore, recent application of novel statistical
23 approaches to the study of source-associated constituents (often metals are the elemental
24 markers) have shown promise in linking sources with their associated emission profiles
25 (including metals) to health outcomes in both humans (Laden et al., 2000) and animals (Clarke
26 et al., 2000). Thus, while metals appear to be one component involved in PM associated health
27 effects, the full story is incomplete.
28
29 Other Inorganic Constituents: The inorganic constituents of ambient PM comprise a
30 number of compounds and elements that derive from either natural or combustion sources. The
31 earthen or natural constituents of PM are typically silicates that contain surface and matrix
June 2003 9-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 bound metals such as calcium, magnesium, aluminum, and iron. As noted above, most of these
2 silicates do not appear to contribute much toxicity to ambient PM, as considered in this
3 document. Sulfate and nitrate anions derived from combustion or photochemical processes
4 usually complex with other constituents in PM - often more water-soluble ammonium ions or
5 organic acids, as well as elemental cations, such as metals. The intrinsic, independent toxicities
6 of sulfates (as per above) and nitrates appear to be rather low, but they may influence the toxicity
7 or bioavailability of other PM components. Of the cations, metals represent a potential class of
8 causal constituents for PM-associated health effects that have received considerable attention
9 (discussed in more detail below). Sulfate, nitrate, ammonium, and metals make up a substantial
10 part of the mass of ambient PM, often with a silicate or carbonaceous (see below) core, layering,
11 or matrix. The majority of PM-associated metals in fine PM are derived from stationary or
12 mobile combustion sources whereas particle sulfate, nitrate and ammonium originate from
13 secondary atmospheric transformation reactions of involving SO2, NOX and biomass ammonia
14 emissions. Organic PM has both primary and secondary sources.
15
16 Organic Constituents: Published research on the acute effects of PM-associated organic
17 carbon constituents is conspicuous by its relative absence, except for diesel exhaust particles
18 (DEP). Like metals, organics are common constituents of combustion-generated PM and are
19 found in ambient PM samples over a wide geographical range. Organic carbon constituents
20 comprise a substantial portion of the mass of ambient PM (10 to 60% of the total dry mass
21 [Turpin, 1999]). Although the organic fraction of PM is a poorly characterized heterogeneous
22 mixture of a widely varying number of different compounds, strategies have been proposed for
23 examining the health effects of potentially important organic constituents (Turpin, 1999).
24 In contrast, the mutagenic effects of ambient PM and evidence of DNA-adducts have had more
25 extensive study and have been linked to specific organic fractions (Binkova et al., 1999; Chorqzy
26 et al., 1994; Izzotti et al., 1996). The extent to which organic constituents of ambient PM
27 contribute to adverse health effects identified by current epidemiology studies is not known.
28 Nevertheless, organic constituents remain of concern regarding PM health effects due in large
29 part to the contribution of DEP to the fine PM fraction and the health effects associated with
30 exposure to these particles.
31
June 2003 9-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 Diesel Exhaust Particles (PEP): There is growing toxicological evidence that DEP
2 exacerbates the allergic response to inhaled antigens. The organic fraction of diesel exhaust has
3 been linked to eosinophil degranulation and induction of cytokine production suggesting that the
4 organic constituents of DEP are responsible for the immune effects. It is known that the
5 adjuvant-like activity of DEP is not unique, and that certain metals have analogous adjuvant
6 effects (Lambert et al., 2000). It is important to compare the immune effects of other source-
7 specific emissions, as well as concentrated ambient PM, to DEP to determine the extent to which
8 exposure to diesel exhaust may contribute to the incidence and severity of allergic rhinitis and
9 asthma. Other types of noncancer and carcinogenic (especially lung cancer) effects are of
10 concern with regard to DEP exposures, as discussed in a separate EPA Health Assessment
11 Document for Diesel Exhaust (U.S. Environmental Protection Agency, 2002).
12
13 Biological Constituents: Recent studies support the conclusion of the 1996 PM AQCD that
14 bioaerosols (e.g., fungal spores, plant and insert fragments, airborne bacteria and viruses) at the
15 concentrations present in the ambient environment, are unlikely to account for the health effects
16 of ambient PM. Dose-response inhalation studies in healthy volunteers exposed to 0.55 and
17 50 jig endotoxin showed the threshold for pulmonary and systemic effects for endotoxin to be
18 between 0.5 and 5.0 jig (Michel et al., 1997). Urban ambient air PM contains variable amounts
19 of endotoxin, but the levels typically are several orders of magnitude less. The in vitro
20 toxicological studies that have shown endotoxin associated with ambient PM to be
21 pro-inflammatory, inducing cytokine expression in human and rat alveolar macrophages, appear
22 to relate to the endotoxin dose to cell ratio (Becker et al., 1996; Dong et al., 1996). However,
23 endotoxin content does appear to vary by size-mode. Monn and Becker (1999) demonstrated
24 cytokine induction by human monocytes, characteristic of endotoxin activity, in the coarse size
25 fraction of outdoor PM, but not in the fine fraction. Interestingly, while studies in animals
26 models also require more endotoxin than typically found in ambient PM to induce inflammation,
27 recent studies suggest endotoxin may have a priming effect on PM-induced inflammatory
28 processes (Imrich et al., 1999). Thus, the role of biogenic material like endotoxin may have a
29 subtle role that is poorly understood.
30
31
June 2003 9-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.7.2.3 Summary
2 Toxicological studies have provided considerable supportive evidence that certain
3 physicochemical particle attributes can provide elements of "causality" to observed health
4 effects of ambient PM. A primary causative attribute may not exist but rather many attributes
5 may contribute to a complex mechanism driven by the nature of a given PM and its contributing
6 sources. The multiple interactions that may occur in eliciting a response in a host may make the
7 identification of any single causal component difficult and may account for the fact that mass as
8 the most basic metric shows the relationships to health outcomes that it does.
9
10 9.7.3 Mechanisms of Action Underlying PM Cardiovascular Effects
11 Numerous epidemiologic studies have shown statistically significant associations between
12 ambient PM levels and a variety of human health endpoints, including mortality, hospital
13 admissions, emergency department visits, respiratory illness, and symptoms measured in
14 community surveys. These associations have been observed with both short and long-term PM
15 exposure. There was little information available in the 1996 PM AQCD to provide biologically
16 plausible mechanisms to support the epidemiologic observations. However, in the intervening
17 years significant progress has been made in identifying pathophysiological effects in humans and
18 animals exposed to various PM that can provide insight into the mechanisms by which PM may
19 exert its effects. Potential mechanisms include neural mechanisms affecting the autonomic
20 nervous system (ANS) via direct pulmonary reflexes or through pulmonary inflammatory
21 processes, direct effects of PM or its components on ion channel function in myocardial cells,
22 ischemic responses of the myocardium, or systemic responses including inflammation that can
23 trigger endothelial cell dysfunction, and thrombosis via alterations in the coagulation cascade.
24 The interactions between these pathways which may lead to sudden cardiac death is shown in the
25 Figure 9-15. However, it must be noted that PM is a complex mixture of many different
26 components and it is possible that different components may stimulate different mechanistic
27 pathways. Thus exposure to PM may result in one or more pathways being activated, depending
28 on the chemical and physical makeup of the PM.
29 There is now ample evidence that inhaled particles can affect the heart through the ANS.
30 Direct input from the lungs to the ANS via pulmonary afferent fibers can affect both heart rate
31 (HR) and heart rate variability (HRV). The heart is under the constant influence of both
June 2003 9-61 DRAFT-DO NOT QUOTE OR CITE
-------
Pulm onary Reflexes
Ayfonora Ic Nervoys
S ^ste ni
Conduction/Repolarizafion
\
He»jt Rate
Ca rdiac R hy thm
1
B radycardia
Tachycardia
Ventricular Fibrillation
\
Pulmonary Inflmwi wt atlors
Pl«t«l*t
1 Activation
/
Sudden Cardiac
Death
Viscosity
Figure 9-15. Schematic representation of potential pathophysiological pathways and
mechanisms by which ambient PM may increase risk of cardiovascular
morbidity and/or mortality.
1 sympathetic and parasympathetic innervation from the ANS; and monitoring changes in HR and
2 HRV can provide insight into the balance between those two ANS subdivisions. During recent
3 decades a large clinical database has developed describing a significant relationship between
4 autonomic dysfunction and sudden cardiac death. One measure of this dysfunction, low HRV,
5 has been implicated as a predictor of increased cardiovascular morbidity and mortality. Several
6 independent epidemiologic panel studies of elderly volunteers (some having cardiovascular or
7 pulmonary disease) have reported associations between PM concentrations and various measures
8 of HR and HRV. Although there are some differences among the studies, in general they report
9 an association between PM levels and a reduction in the standard deviation of normal to normal
10 beat intervals (SDNN), a time-domain variable of which the reduction was associated in the
11 Framingham Heart Study with a higher risk of death. Some studies also reported an association
12 between PM and decreased HRV in the high frequency (HF) range, which is a reflection of
13 parasympathetic modulation of the heart. Other studies have reported a positive association
June 2003
9-62
DRAFT-DO NOT QUOTE OR CITE
-------
1 between PM and HR; elevated HR has been associated with hypertension, coronary heart
2 disease, and death. Thus taken as a whole, evidence from panel studies indicates that PM can
3 directly affect the ANS in such as way as to alter heart rate and heart rate variability. However,
4 it should be noted that lowered HRV has primarily been used as a predictor of subsequent
5 increased mortality and morbidity. It is not clear whether a single reversible acute change in
6 HRV places a person more at risk for an immediate adverse cardiac event. Whether changes in
7 HRV associated with exposure to PM represent an independent risk or is just a marker of
8 exposure is not yet known.
9 PM as also been shown to induce changes in conductance and repolarization of the heart as
10 well. Repolarization duration and morphology may reflect subtle changes in myocardial
11 substrate and vulnerability governed by changes in ion channel function. There is considerable
12 evidence linking changes in T wave morphology, QT and T wave variability, T wave Alternans,
13 and changes in ST segment height, to the risk of sudden death. In some studies, rodent models
14 of susceptibility (monocrotaline injected, spontaneously hypertensive) exposed to ROFA showed
15 exacerbated ST segment depression, a factor reflecting T wave morphology during
16 repolarization and which as been useful in diagnosing patients with ischemic heart disease.
17 Healthy dogs exposed to CAPS also showed changes in ST segment elevation; this was
18 exacerbated in dogs with coronary artery occlusion.
19 While PM-induced changes in HRV and HR, as well as changes associated with
20 repolarization and conductance, have the potential to progress to malignant arrhythmias, there is
21 now evidence from both human and animal studies that PM exposure may be linked with severe
22 events directly associated with sudden cardiac death. A recent epidemiology study of patients
23 with implanted cardiac defibrillators reported associations between PM and increased
24 defibrillator discharges. Presumably, some of these patients would have suffered a fatal event
25 had they not had an implanted defibrillator. A second study reported that the risk for myocardial
26 infarction (MI) onset increased in association with PM levels in the 2 hours preceding the MI.
27 PM exposure has also been linked with malignant arrhythmia in some toxicology studies.
28 Healthy rodents exposed to ROFA demonstrated an increase in serious arrhythmic events,
29 including bradycardia. Rats treated with monocrotaline had significantly exacerbated
30 arrhythmias, and some animals even died within 24 hours following exposure. Older rats,
31 exposed to both ROFA and PM collected from Ottowa, also experienced increased arrhythmias.
June 2003 9-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 Dogs exposed to CAPS experienced a slight bradycardia following exposure. Some of these
2 studies involved instillation of a specific PM component (ROFA) at high concentrations, making
3 it uncertain that these observations would hold true using ambient PM at more realistic
4 concentrations. Nevertheless, at least one study used ambient particles collected from Ottowa,
5 and other studies exposed animals by inhalation to CAPS. Taken as a whole, these studies
6 provide convincing evidence that exposure of animals to high levels of PM can affect
7 conductance and repolarization, potentially leading to fatal arrhythmias. However, it remains to
8 be seen if these mechanisms, that can potentially explain acute mortality associated with PM
9 exposure, operate at the lower concentrations of ambient PM to which most people are exposed.
10 Particulate matter could potentially affect the ANS by direct interaction with nerve ending
11 in the lung, or indirectly through the production of inflammatory mediators. Numerous studies
12 have documented that exposure of rodents to ROFA results in substantial lung inflammation and
13 injury. However, due to the levels of ROFA used in many of these studies and the fact that
14 ROFA only makes up a small portion of most airsheds, studies with ambient air particles may be
15 more relevant. There are several studies in which humans, dogs, or rodents have been exposed
16 to CAPS and mild pulmonary inflammation observed. Other studies have shown similar effects
17 when ambient PM collected on filters was used. However, the level of inflammation was quite
18 low in most of these studies, certainly lower than reported in humans or animals exposed to
19 ozone, and it is not yet clear whether lung inflammation plays a role in PM-induced changes in
20 the ANS.
21 In addition to affecting the ANS via the lung, it is also possible that PM or its components
22 could directly attack the myocardium. There is substantial evidence that chronic exposure to
23 fibers encountered in the workplace (e.g., asbestos) result in deposition of fibers in organs other
24 than the lung. Some recent studies have suggested that ultrafine PM may exit the lung and
25 deposit in other organs, including the liver and heart. So far these studies have used sources of
26 particles not naturally found in the air (e.g., silver colloid, latex) so it is not yet clear to what
27 extent PM actually leaves the lung or, if it does, how it interacts directly with the heart.
28 However, there is some evidence of direct changes in the myocardium following PM exposure.
29 For example, rats exposed to ROFA, which is made up mostly of soluble transition metals, have
30 increased pro-inflammatory cytokine expression in the left ventricle. In another study, dogs
31 living in highly-polluted Mexico City had histopathology changes in heart tissue compared with
June 2003 9-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 dogs living in areas with low air pollution. Substantial deposits of particulate matter could be
2 seen throughout the myocardium in the Mexico City dogs. Though preliminary, these
3 observations point to a need for additional work to better define PM-induced changes in
4 myocardial tissue.
5 Acute coronary events frequently occur as a result of thrombus formation in the site of a
6 ruptured atherosclerotic plaque. Increased levels of clotting and coagulation factors, platelet
7 aggregability, and blood viscosity, together with reduced fibrinolytic activity and endothelial cell
8 dysfunction can promote a pro-coagulant state which could potentially contribute to thrombus
9 formation. C reactive protein, a marker of systemic inflammation which correlates with some
10 cardiac events, is positively associated with PM in several panel studies. Some of these studies
11 also report associations between PM and enhanced blood viscosity or increased fibrinogen, a
12 known risk factor for ischemic heart disease. Controlled human and animal exposure studies
13 have also reported that exposure to CAPS (in humans) or ROFA (in animals) results in increased
14 levels of blood fibrinogen. These studies suggest that PM may alter the coagulation pathways in
15 such a way as to trigger cardiovascular events in susceptible individuals.
16 Panel studies have also reported associations between PM and changes in white blood
17 cells, although these findings are not easy to interpret since some studies report positive
18 associations while others report negative associations. Animal studies are similarly unclear, with
19 some studies (rodents exposed to CAPS) reporting increased numbers of blood platelets and
20 white blood cells and others (rodents exposed to ROFA) reporting decreased numbers of white
21 blood cells. In one study, rabbits instilled with colloidal carbon had an increase in neutrophils
22 released from the bone marrow. The same research group found an association between PM and
23 elevated band neutrophil counts (a marker for bone marrow precursor release) in humans
24 exposed to high levels of carbon from biomass burning during the 1997 Southeast Asian smoke-
25 haze episodes.
26 Endothelial cell dysfunction may contribute to myocardial ischemia in some susceptible
27 populations. The vascular endothelium secretes multiple factors that control vascular tone,
28 modulate platelet activity, and influence thrombogenesis. A recent study has reported
29 endothelial cell dysfunction in humans exposed to CAPS, as measured by dilation of the brachial
30 artery. This vasoconstriction could be caused by an increase in circulating endothelin-1, which
31 has been described in rats exposed to PM.
June 2003 9-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 Taken as a whole, these studies are difficult to interpret but clearly indicate that PM can
2 affect the circulatory system. However, a complete understanding of the pathways by which
3 very small concentrations of inhaled ambient PM can produce vascular changes that can
4 contribute to increased mortality/morbidity remains to be more fully elucidated.
5
6
7 9.8 HEALTH EFFECTS OF AMBIENT PARTICULATE MATTER
8 OBSERVED IN HUMAN POPULATION STUDIES
9 9.8.1 Introduction
10 This section assesses available scientific evidence regarding the health effects of exposure
11 to ambient PM as observed in epidemiologic (human population) studies. The main objectives
12 of this evaluation are (1) to summarize and evaluate strengths and limitations of available
13 epidemiologic findings; (2) to summarize quantitative relationships between ambient PM
14 exposures and increased human health risks; (3) to assess the biomedical coherence of findings
15 across studied endpoints; and (4) to note the increased biologic plausibility of the available
16 epidemiologic evidence in light of (a) linkages between specific PM components and health
17 effects and (b) various dosimetric, mechanistic, and pathophysiologic considerations discussed
18 earlier in this chapter.
19 Numerous epidemiologic studies have shown statistically significant associations of
20 ambient PM levels with a variety of human health endpoints, including mortality, hospital
21 admissions, emergency department visits, other medical visits, respiratory illness and symptoms
22 measured in community surveys, and physiologic changes in pulmonary function. Associations
23 have been consistently observed between both short- and long-term PM exposure and these
24 endpoints. The general internal consistency of the epidemiologic database and available findings
25 demonstrate well that notable human health effects are associated with exposures to ambient PM
26 at concentrations currently found in many geographic locations across the United States.
27 However, many challenges still exist with regard to delineating the magnitudes and variabilities
28 of risk estimates for ambient PM, the ability to attribute observed health effects to specific PM
29 constituents, the time intervals over which PM health effects are manifested, the extent to which
30 findings in one location can be generalized to other locations, and the nature and magnitude of
31 the overall public health risk imposed by ambient PM exposure.
June 2003 9-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 The etiology of most air pollution-related health outcomes is highly multifactorial, and the
2 impact of ambient air pollution exposure on these outcomes is often small in comparison to that
3 of other etiologic factors (e.g., smoking). Also, ambient PM exposure usually is accompanied by
4 exposure to many other pollutants, and PM itself is composed of numerous physical/chemical
5 components. Assessment of the health effects attributable to PM and its constituents within an
6 already-subtle total air pollution effect is therefore very challenging, even with well-designed
7 studies. Indeed, statistical partitioning of separate pollutant effects may not characterize fully
8 the etiology of effects that actually depend on simultaneous exposure to multiple air pollutants.
9 In this regard, several viewpoints existed at the time of the 1996 PM AQCD regarding how best
10 to interpret the epidemiology data: one saw the PM exposure indicators as surrogate measures of
11 complex ambient air pollution mixtures and the reported PM-related effects as representative of
12 those of the overall mixture; another held that reported PM-related effects are attributable to PM
13 components (per se) of the air pollution mixture and reflect independent PM effects; and a third
14 viewpoint held that PM can be viewed both as a surrogate indicator, as well as a specific cause
15 of health effects.
16 Several other key issues must be considered when attempting to interpret the data reviewed
17 in this document. For example, although the epidemiology data provide strong support for the
18 associations mentioned above, questions remain regarding potential underlying mechanisms.
19 Even considering the progress made toward identification of anatomic sites at which particles
20 trigger specific health effects and elucidation of biological mechanisms that underlie induction
21 of such effects, this area of scientific inquiry is still at an early stage. Still, compared to the lack
22 of such evidence available in the 1996 PM AQCD, there now is a stronger basis for assessing
23 biologic plausibility of the epidemiologic observations, given notable improvement in the
24 conceptual formulation of reasonable mechanistic hypotheses and the generation of research
25 evidence bearing on such hypotheses. New evidence related to several hypotheses was noted in
26 the prior section (Section 9.7) with regard to possible mechanisms by which ambient PM may
27 exert human health effects, which tends to support the likelihood of a causal relationship
28 between low ambient concentrations of PM and increased mortality or morbidity risks observed
29 in human population studies. Much still remains to be done, however, to identify more
30 confidently specific causal agents among typical ambient PM constituents.
June 2003 9-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 In recent years, epidemiologic studies showing associations of ambient air pollution
2 exposure with mortality, exacerbation of preexisting illness, and pathophysiologic changes have
3 increased concern about the extent to which exposure to ambient air pollution exacerbates or
4 causes harmful health outcomes at pollutant concentrations now experienced in the United
5 States. The PM epidemiology studies assessed in the 1996 PM AQCD implicated ambient PM
6 as a likely key contributor to mortality and morbidity effects observed epidemiologically to be
7 associated with ambient air pollution exposures. New studies appearing since the 1996 PM
8 AQCD are important in extending earlier results to many more cities and in confirming earlier
9 findings.
10 In epidemiologic studies of ambient air pollution, small positive estimates of air pollutant
11 health effects have been observed quite consistently, frequently being statistically significant at
12 p < 0.05. If ambient air pollution promotes or produces harmful health effects, relatively small
13 effect estimates from current PM concentrations in the United States and many other countries
14 would generally be expected on biological and epidemiologic grounds. Also, magnitudes and
15 significance levels of observed air pollution-related effects estimates would be expected to vary
16 somewhat from place to place, if the observed epidemiologic associations denote actual effects,
17 because (a) not only would the complex mixture of PM vary from place to place, but also
18 (b) affected populations may differ in characteristics that could affect susceptibility to air
19 pollution health effects. Such characteristics include sociodemographic factors, underlying
20 health status, indoor-outdoor activities, diet, medical care access, exposure to risk factors other
21 than ambient air pollution (such as extreme weather conditions), and variations in factors (e.g.,
22 air-conditioning) affecting human exposures to ambient-generated PM.
23 As noted above, small health relative risk estimates for health effects have generally been
24 observed for ambient air pollutants, as would be expected on biological and epidemiologic
25 grounds. In contrast to effect estimates for mortality derived for the 1952 London smog episode,
26 i.e., relative risk (RR) exceeding 4.0 (i.e., 400% increase over baseline) for extremely high
27 (>2 mg/m3) ambient PM levels, effects estimates in most current epidemiology studies at
28 distinctly lower PM concentrations (often < 100 |ig/m3) are relatively small. The statistical
29 estimates are more often subject to small (but proportionately large) differences in estimated
30 effects of PM and other pollutants; may be sensitive to a variety of methodological choices; and
June 2003 9-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 sometimes may not be statistically significant, reflecting low statistical power of the study
2 design to detect a relatively small but real effect.
3 The ambient atmosphere contains numerous air pollutants, and it is important to continue
4 to recognize that health effects associated statistically with any single pollutant may actually be
5 mediated by multiple components of the complex ambient mix. Specific attribution of effects to
6 any single pollutant may therefore be overly simplistic. Particulate matter is one of many air
7 pollutants derived from combustion sources, including mobile sources. These pollutants include
8 PM, CO, SO2, NO2, and O3, all of which have been considered in various epidemiologic studies.
9 Many volatile organic compounds (VOCs) or semivolatile compounds (SVOCs) are also emitted
10 by combustion sources or formed in the atmosphere but have not yet been systematically
11 considered in relation to noncancer health outcomes most usually associated with exposure to
12 criteria air pollutants. In many newly available epidemiologic studies, harmful health outcomes
13 are often associated with multiple combustion-related or mobile-source-related air pollutants,
14 and some investigators have raised the possibility that PM may be a key surrogate or marker for
15 a larger subset of the overall ambient air pollution mix. This possibility takes on added potential
16 significance to the extent that ambient aerosols indeed may not only exert health effects directly
17 attributable to their constituent components, per se, but also serve as carriers for more efficient
18 delivery of water soluble toxic gases (e.g., O3, NO2, SO2) deeper into lung tissue, as noted earlier
19 in Section 9.6.4. This suggests that airborne particle effects may be enhanced by the presence of
20 other toxic agents or mistakenly attributed to them if their respective concentrations are highly
21 correlated temporally. Thus, although associations of PM with harmful effects continue to be
22 observed consistently across most new studies, the newer findings do not fully resolve issues
23 concerning relative contributions to the observed epidemiologic associations of (a) PM acting
24 alone, (b) PM acting in combination with gaseous co-pollutants, (c) the gaseous pollutants per
25 se, and (d) the overall ambient pollutant mix.
26 It is possible that, for pollutants whose ambient concentrations are not highly correlated,
27 effects estimates in multipollutant models could be more biologically and epidemiologically
28 sound than those in single-pollutant models, although single-pollutant models could also be
29 credible if independent biological plausibility evidence supported designation of PM or some
30 other single pollutant as likely being the key toxicant in the ambient pollutant mix evaluated.
31 Because neither of these possibilities have been definitively demonstrated and there is not yet
June 2003 9-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 full scientific consensus as to optimal interpretation of modeling outcomes for time series-air
2 pollution studies, the choice of appropriate effects estimates to employ in risk assessments for
3 ambient PM effects remains a difficult issue. Issues related to confounding by co-pollutants,
4 along with issues related to time scales of exposure and response and concentration-response
5 function, still apply to new epidemiologic studies relating concentrations of PM or correlated
6 ambient air pollutants to hospital admissions, exacerbation of respiratory symptoms, asthma in
7 children, reduced pulmonary function in children and adults, and to changes in heart rate and
8 heart rate variability in adults. However, with considerable new experimental evidence now in
9 hand, it is possible to hypothesize various ways in which ambient exposure to PM acting alone
10 or in combination with other co-pollutants can plausibly be involved in the complex chain of
11 biological events leading to harmful health effects in the human population. This newer
12 experimental evidence, coupled with new exposure analyses results, adds much support for
13 interpreting the epidemiologic findings discussed here as likely being indicative of causal
14 relationships between exposures to ambient PM (or specific size or chemical components) and
15 consequent associated increased mortality and morbidity effects.
16
17 9.8.1.2 GAM Convergence Issue
18 In the spring of 2002, the original investigators of a key newly available multi-city study
19 (the National Mortality and Morbidity Air Pollution Study; NMMAPS) cosponsored by the
20 Health Effects Institute (HEI) reported that use of the default convergence criteria setting used in
21 the GAM routine of certain widely-used statistical software (Splus) could result in biased
22 estimates of air pollution effects when at least two non-parametric smoothers are included in the
23 model (Health Effects Institute letter, May 2002). The NMMAPS investigators also reported
24 (Dominici et al., 2002), as determined through simulation, that such bias was larger when the
25 size of the risk estimate was smaller and when the correlation between the PM and the covariates
26 (i.e., smooth terms for temporal trend and weather) was higher. While the NMMAPS
27 investigators reported that reanalysis (using stringent convergence criteria) of the 90 cities air
28 pollution-mortality data did not qualitatively change their original findings (i.e., the positive
29 association between PM10 and mortality; lack of confounding by gaseous pollutants; regional
30 heterogeneity of PM, etc.), the reduction in the PM10 risk estimate was apparently not negligible
31 (dropping, upon reanalysis, from -2.1% to 1.4% excess deaths per 50 |ig/m3 increase in PM10).
June 2003 9-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 Issues surrounding potential bias in PM risk estimates from time-series studies using GAM
2 analyses and default convergence criteria were raised by EPA and discussed in July 2002 at the
3 CAS AC review of the Third External Review Draft of this PM AQCD. In keeping with a follow
4 up consultation with CASAC in August 2002, EPA encouraged investigators for a number of
5 important published studies to reanalyze their data by using GAM with more stringent
6 convergence criteria, as well as by using Generalized Linear Model (GLM) analyses with
7 parametric smoothers that approximated the original GAM model. EPA, working closely with
8 HEI, also arranged for (a) the resulting reanalyses first to be discussed at an EPA-sponsored
9 open Workshop on GAM-Related Statistical Issues in PM Epidemiology held in November
10 2002; (b) then for any revamping of the preliminary analyses in light of the workshop
11 discussions; before (c) submittal by the investigators of short communications describing the
12 reanalyses approaches and results to EPA and HEI for peer-review by a special panel assembled
13 by HEI; and (d) the publication of the short communications on the reanalyses, along with
14 commentary by the HEI peer-review panel, in an HEI Special Report (2003a). Some of the
15 short-communications included in the HEI Special Report (2003) included discussion of
16 reanalyses of data from more than one original publication because the same data were used to
17 examine different issues of PM-mortality associations (e.g., concentration/response function,
18 harvesting, etc.). In total, reanalyses were reported for more than 35 originally published
19 studies.
20
21 9.8.1.3 Ambient PM Increments Used to Report Risk Estimates
22 The effect of mortality from exposure to PM or other pollutants is usually expressed in this
23 document as a relative risk or risk rate (RR) relative to a baseline mortality or morbidity rate.
24 The pollutant concentration increments utilized here to report Relative Risks (RR's) or Odds
25 Ratio for various health effects are as follow for short-term (< 24 h) exposure studies: 50 |ig/m3
26 for PM10; 25 |ig/m3 for PM25 and PM10.25; 155 nmoles/m3 (15 |ig/m3) for SO4'2; and
27 75 nmoles/m3 (3.6 |ig/m3, if as H2SO4) for H+. The increments for short-term studies are the
28 same as were used in the 1996 PM AQCD, a choice now driven by more current data. In the
29 1996 PM AQCD, the same increments were used for the long- and short-term exposure studies.
30 However, for long-term exposure studies, 20 |ig/m3 is the increment used here for PM10 and
June 2003 9-71 DRAFT-DO NOT QUOTE OR CITE
-------
1 10 |ig/m3 for PM2 5 and PM10_2 5. These latter increments, derived from new 1999-2001 data, are
2 smaller than those used in the 1996 PM AQCD for long-term exposure studies.
3
4 9.8.2 Short-Term Particulate Matter Exposure Effects on Mortality
5 This section focuses primarily on discussion of short-term PM exposure effects on
6 mortality, but also highlights some morbidity effects in relation to the mortality findings.
7 Morbidity effects of short-term ambient PM exposures are discussed more fully in subsection
8 (9.8.3). Subsequent sections include discussion of mortality and morbidity effects of long-term
9 PM exposures.
10
11 Summary of Previous Findings on Short-Term Particulate Matter Exposure-Mortality Effects
12 Time-series mortality studies reviewed in the 1996 PM AQCD provided strong evidence
13 that ambient PM air pollution is associated with increased daily mortality. The 1996 PM AQCD
14 summarized about 35 PM-mortality time series studies published between 1988 and 1996. The
15 available information from those studies was consistent with the hypothesis that PM is a causal
16 agent in the mortality impacts of air pollution. The PM10 relative risk estimates derived from the
17 PM10 studies reviewed in the 1996 PM AQCD suggested that an increase of 50 |ig/m3 in the 24-h
18 average of PM10 is associated with an increased risk of premature total mortality (total deaths
19 minus accidents and injuries) mainly on the order of RR = 1.025 to 1.05 (i.e., 2.5 to 5.0% excess
20 risk) in the general population, with statistically significant increases being reported more
21 broadly across the range of 1.5 to 8.5% per 50 |ig/m3 PM10. Higher relative risks were indicated
22 for the elderly and for those with preexisting respiratory conditions. Also, based on the then
23 recently published Schwartz et al. (1996) analysis of Harvard Six City data, the 1996 PM AQCD
24 found the RR for excess total mortality in relation to 24-h fine-particle concentrations to be in
25 the range of RR = 1.026 to 1.055 per 25 |ig/m3 PM25 (i.e., 2.6 to 5.5% excess risk per 25 |ig/m3
26 PM2 5). Relative risk estimates for morbidity and mortality effects associated with standard
27 increments in ambient PM10 concentrations and for fine-particle indicators (e.g., PM2 5, sulfates,
28 etc.) were presented in Chapters 12 and 13 of the 1996 PM AQCD (see Appendix 9A); and those
29 effect estimates are updated below in light of the extensive newly available evidence discussed
30 in Chapter 8 of this document.
June 2003 9-72 DRAFT-DO NOT QUOTE OR CITE
-------
1 Although numerous studies reported PM-mortality associations, several important issues
2 needed to be addressed in interpreting those relative risks. The 1996 PM AQCD extensively
3 discussed the following critical issues: (1) seasonal confounding and effect modification,
4 (2) confounding by weather, (3) confounding by co-pollutants, (4) measurement error,
5 (5) functional form and threshold, (6) harvesting and life shortening; and (7) the roles of specific
6 PM components.
7 Season-specific analyses are often not feasible because of small magnitudes of expected
8 effect size or small sample sizes (low power) available for some studies. Some earlier studies
9 had suggested possible season-specific variations in PM coefficients, but it was not clear if these
10 were caused by peak variations in PM effects from season to season, varying extent of PM
11 correlations with other co-pollutants, or weather factors during different seasons. The likelihood
12 of PM effects being accounted for mainly by weather factors was addressed by various methods
13 that controlled for weather variables in most studies (including some involving sophisticated
14 synoptic weather pattern evaluations); and that possibility was found to be very unlikely.
15 Many early PM studies considered at least one co-pollutant in the mortality regression, and
16 an increasing number have examined multiple pollutants. At times, when PM indices were
17 significant in single-pollutant models, addition of a co-pollutant diminished the PM effect size
18 somewhat, but did not eliminate PM associations. In multiple-pollutant models performed by
19 season, the PM coefficients became less stable, again possibly because of varying correlations of
20 PM with co-pollutants among seasonal or smaller sample sizes. However, in many studies, PM
21 indices showed the highest significance in both single- and multiple-pollutant models. Thus,
22 PM-mortality associations did not appear to be seriously distorted by co-pollutants.
23 Interpretation of the relative significance of each pollutant in mortality regression in
24 relation to its relative causal strength was difficult, however, because of lack of quantitative
25 information on pertinent exposure measurement errors among the air pollutants. Measurement
26 errors can influence the size and significance of air pollution coefficients in time series
27 regression analyses, an issue also important in assessing confounding among multiple pollutants,
28 because the varying extent of such errors among pollutants may influence corresponding relative
29 significance. The 1996 PM AQCD discussed several types of exposure measurement and
30 characterization errors, including site-to-site variability and site-to-person variability. These
June 2003 9-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 errors are thought to bias the estimated PM coefficients downward in most cases, but there was
2 insufficient quantitative information available at the time to allow estimation of such bias.
3 The 1996 PM AQCD also reviewed evidence for threshold and various other functional
4 forms of short-term PM mortality associations. Some studies indicated that associations were
5 seen at levels even below then-existing PM standards. It was considered difficult, however, to
6 statistically evaluate the possibility of a threshold from available data because of low data
7 density at lower ambient PM concentrations, potential influence of measurement error, and
8 adjustments for other covariates. Thus, use of relative risk (rate ratio) derived from log-linear
9 Poisson models was deemed adequate.
10 The extent of prematurity of death, i.e., mortality displacement (or harvesting) in observed
11 PM-mortality associations has important public health policy implications. At the time of the
12 1996 PM AQCD review, only a few studies had investigated this issue. Although one of the
13 studies suggested that the extent of such prematurity might be only a few days, this may not be
14 generalized because this estimate was obtained for identifiable PM episodes. Insufficient
15 evidence then existed to suggest the extent of prematurity for nonepisodic periods, from which
16 most of the recent PM relative risks were derived.
17 Only a few PM-mortality studies had analyzed fine particles and chemically specific
18 components of PM. Using Harvard Six Cities Study data, Schwartz et al. (1996) analyzed size-
19 fractionated PM (PM25, PM10/15, and PM10/15.25) and PM chemical components (sulfates and H+).
20 The results suggested that PM2 5 was associated most significantly with mortality among the PM
21 components. Although H+ was not significantly associated with mortality in this and earlier
22 analyses, the smaller sample size for H+ than for other PM components made direct comparison
23 difficult. Also, certain respiratory morbidity studies showed associations between hospital
24 admissions and visits with components of PM in the fine-particle range. Thus, the 1996 PM
25 AQCD concluded that there was adequate evidence to suggest that fine particles play an
26 especially important role in observed PM mortality effects.
27 Overall, then, the outcome of assessment of the above key issues in the 1996 PM AQCD
28 can be thusly summarized: (1) observed PM effects are not likely seriously biased by inadequate
29 statistical modeling (e.g., control for seasonality); (2) observed PM effects are not likely
30 significantly confounded by weather; (3) observed PM effects may be confounded or modified to
31 some extent by co-pollutants, and such extent may vary from season to season; (4) determining
June 2003 9-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 the extent of confounding and effect modification by co-pollutants requires knowledge of
2 relative exposure measurement/characterization error among pollutants (there was not sufficient
3 information on this); (5) no clear evidence for any threshold for PM-mortality associations was
4 reported (statistically identifying a threshold from existing data also was considered difficult, if
5 not impossible); (6) some limited evidence for harvesting, a few days of life-shortening, was
6 reported for episodic periods (no study was conducted to investigate harvesting in nonepisodic
7 U.S. data); and (7) only a relatively limited number of studies suggested a causal role of fine
8 particles in PM-mortality associations, but in light of historical data, biological plausibility, and
9 results from morbidity studies, a greater role for fine particles than coarse particles was
10 suggested as being likely.
11
12 Updated Epidemiologic Findings for Short-Term Ambient Particulate Matter
13 Exposure Effects on Mortality
14 With regard to updating the assessment of PM effects in light of new epidemiologic
15 information published since the 1996 PM AQCD, the most salient key points on relationships
16 between short-term PM exposure and mortality (drawn from Chapter 8 discussions in this
17 document) can be summarized as follows.
18 Since the 1996 PM AQCD, there have been more than 80 new time-series PM-mortality
19 analyses, several of which investigated multiple cities using consistent data analytical
20 approaches. With only a few exceptions, the estimated mortality RR's in these studies are
21 generally positive, many are statistically significant, and they generally comport well with
22 previously reported PM-mortality effects estimates delineated in the 1996 PM AQCD. There are
23 also now numerous additional studies demonstrating associations between short-term (24-h) PM
24 exposures and various morbidity endpoints.
25 Several new studies conducted time series analyses in multiple cities. The major
26 advantage of these studies over meta-analyses for multiple "independent" studies is the
27 consistency in data handling and model specifications, thus eliminating variation in results
28 attributable to study design. Also, many of the cities included in these studies were ones for
29 which no earlier time series analyses had been conducted. Therefore, unlike regular meta-
30 analysis, they likely do not suffer from omission of negative studies caused by publication bias.
31 Furthermore, any spatial or geographic variability of air pollution effects can be systematically
32 evaluated in such multi-city analyses.
June 2003 9-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 PM10 Effect Size Estimates
2 Table 9-8 provides a summary of effect size estimates per variable 24-h PM10, PM2 5, and
3 PM10_2 5 increments for total and cause-specific (cardiovascular; respiratory) mortality derived
4 from epidemiological studies of U.S. and Canadian cities. These include GAM results mainly
5 derived from newly published studies and/or their reanalyses using stringent convergence criteria
6 (GAM strict) or other acceptable alternate methods (e.g., GLM). Also included in the table are
7 results for some key studies assessed in the 1996 PM AQCD that did not use GAM (default)
8 analyses. Emphasis is placed in Table 9-8 (and ensuing analogous tables) on the presentation of
9 percent excess risk increases per designated increment in a given PM indicator (e.g., PM10,
10 PM25, etc.), as derived from single-pollutant PM models of the type indicated.
11 The NMMAPS (Samet et al., 2000a,b) analysis of the 90 largest U.S. cities using default
12 GAM convergence criteria found a combined nationwide RR estimate of-2.3% increase in total
13 mortality per 50-|ig/m3 increase in PM10. The NMMAPS effect size estimates did vary
14 somewhat by U.S. region, with the largest estimate being for the Northeast (4.5% for a 1-day lag,
15 the lag typically showing maximum effect size for most U.S. regions). Reanalyses of the same
16 NMMAPS data reported by Dominici et al. (2002; 2003), using other more appropriate
17 alternative analyses, found smaller effect size estimates, the overall nationwide combined
18 estimate being -1.4% excess total deaths per 50 |ig/m3 PM10 increment based on GAM analyses
19 with stringent convergence criteria (the effect size of the Northeast region being about twice the
20 nationwide estimate). Reanalyses for various other U.S. multi-city studies, as well as single-city
21 analyses, obtained PM10 effect sizes mainly in the range of 2.5 to 5.0% per 50-|ig/m3 increase in
22 PM10. There is some evidence that, if the effects over multiple days are considered, the effect
23 size may be larger. What heterogeneity existed for the estimated PM10 risks across NMMAPS
24 cities could not be explained with the city-specific explanatory variables (e.g., as the mean levels
25 of pollution and weather), mortality rate, sociodemographic variables (e.g., median household
26 income), urbanization, or variables related to measurement error.
27 Original results reported for the multi-city APHEA study showed generally consistent
28 associations between mortality and both SO2 and PM indices in western European cities, but not
29 for central and eastern European cities. More recent studies from APHEA II analyses, however,
30 found analogous increased risks to be associated with PM exposures in central and eastern
31 Europe as in western European cities; and these findings were substantiated by reanalyses
June 2003 9-76 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-8. ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY EFFECT SIZES PER
INCREMENTS IN 24-h CONCENTRATIONS OF PM,n, PM?, AND PM,n,, FROM U.S. AND CANADIAN STUDIES
to
o
o
OJ
VO
i
^J
>
H
6
o
o
H
O
w
0
r\
i — i
H
W
• 11% .1^1'jlTll'jll i
Original study*
Reanalysis study
Study location
MORTALITY: Total
Ito and Thurston (1996)
Chicago, IL
Sty er etal. (1995)
Chicago, IL
Kinney etal. (1995)
Los Angeles, CA
Pope etal. (1992)
Utah Valley, UT
Schwartz (1993)
Birmingham, AL
Schwartz etal. (1996)
Schwartz (2003a)
Boston, MA
Schwartz et al. (1996)
Schwartz (2003a)
Knoxville, TN
Schwartz etal. (1996)
Schwartz (2003a)
St. Louis, MO
Schwartz et al. (1996)
Schwartz (2003a)
Steubenville, OH
Schwartz etal. (1996)
Schwartz (2003a)
Portage, WI
k_7 Ai-i ^-r-ii ^^ \_7-Li v^imi i ixr^ i iv^n k} v^i1 i IT
Analysis % increase (95% CI) per
Comments** 50 ug/m3 PM10 Increase
(nonaccidental) Mortality
GAM not 2.47(1.26,3.69)
used
GAM not 4.08(0.08,8.24)
used
GAM not 2.47 (-0.17, 5.18)
used
GAM not 7.63(4.41,10.95)
used
GAMnot 5.36(1.16,9.73)
used
GAM Strict
GLMNS
GLMBS
GMLPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
MO' A 1TA2.5 J^ms A 1TA10-2.5
% increase (95% CI) per
25 ug/m3 PM25 Increase
—
—
—
—
5.3(3.5,7.1)
5.7(3.7,7.6)
5.0(3.1,7.0)
4.5(2.5,6.5)
3.1(0.0,6.2)
3.0 (-0.3, 6.6)
2. 8 (-0.5, 6.3)
2.6 (-0.8, 6.1)
2.6(0.9,4.3)
2.4(0.6,4.1)
2.6(0.9,4.4)
2.3(0.6,4.1)
2.4 (-0.4, 5.3)
1.7 (-1.34. 8)
1.5 (-1.5, 4.6)
1.8 (-1.2, 4. 9)
2.6 (-1.2, 6.6)
0.8 (-3. 3, 5.1)
1.5 (-2.7, 5.8)
1.1 (-3.1,5.4)
PM10, PM2 5 and PM10_2 5
% increase (95% CI) per 25 Mean (Range) Levels
ug/m3 PM10_2 5 Increase Reported***
— PM10 38 (max 128)
— PM10 37 (4, 365)
— PM1058(15, 177)
— PM1047(11,297)
— PM1048(21,80)
PM1024.5(SD12.8)
PM25 15.7 (SD 9.2)
PM10.258.8(SD7.0)
0.7 (-1.9, 3.4)
PM10 32.0 (SD 14.5)
PM2520.8(SD9.6)
PM10.2511.2(SD7.4)
1.7 (-2.7, 6. 3)
PM10 30.6 (SD 16.2)
PM2518.7(SD10.5)
PM10.2511.9(SD8.5)
0.3 (-2.1, 2.7)
PM10 45.6 (SD 32.3)
PM25 29.6(SD21.9)
PM10.25 16.1 (SD 13.0)
5.2(0.0,10.7)
PM10 17.8 (SD1 1.7)
PM2511.2(SD7.8)
PM10.256.6(SD6.8)
0.7 (-4.0, 5.6)
-------
to
O
o
VO
-!j
oo
TABLE 9-8 (cont'd). ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY
EFFECT SIZES PER INCREMENTS IN 24-h CONCENTRATIONS OF PM10, PM2 5 AND PM10 2 5
FROM U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis
Comments**
% increase (95% CI) per
50 ug/m3 PM10 Increase
% increase (95% CI) per
25 ug/m3 PM25 Increase
% increase (95% CI) per 25
ug/m3 PM10_2 5 Increase
PM10,PM25andPM10_25
Mean (Range) Levels
Reported***
MORTALITY: Total (nonaccidental) Mortality (cont'd)
Schwartz et al. (1996)
Schwartz (2003a)
Topeka, KS
Schwartz etal. (1996)
Schwartz (2003a)
6 Cities, Overall
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis-St. Louis
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis-
Steubenville
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis-Topeka
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis -
Knoxville
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis - Boston
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis - Madison
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
2.0(0.0,4.1)
1.0 (-1.5, 3.6)
2.5 (-1.7, 7.0)
1.5 (-1.7, 4.9)
-3. 5 (-11. 6, 5.4)
-5.4 (-14.3, 4.4)
6.1(1.5,11.0)
5.1 (-0.2, 10.7)
6.1(3.6,8.8)
5.6(2.8,8.5)
1.0 (-4.6, 7.0)
-1.5 (-7.7, 5.1)
1.6 (-5.3, 9.0)
2.7 (-5.0, 10.9)
1.3 (-6.2, 9.3)
1.4 (-6.3, 9.6)
3.5(2.5,4.5)
3.3(2.2,4.3)
3.0(2.0,4.0)
2.9(1.8,4.0)
2.0(0.5,3.5)
1.3 (-0.5, 3.0)
1.5 (-1.6, 4.7)
0.5 (-2.7, 3.8)
1.5 (-6. 5, 10.2)
-0.5 (-9.5, 9.4)
4.3(0.9,7.8)
3.8 (-0.1, 7.8)
5.1 (3.3,6.9)
4.0(1.9,6.2)
1.5 (-2.7, 5.9)
-1.2 (-5.7, 3.5)
-3.0(-8.1,2.3)
0.0 (-2.2, 2.3)
-0.5 (-3.0, 2.0)
4.6 (-0.7, 10.1)
4.0 (-1.6, 10.0)
-3.7(-9.2,2.1)
-4.7(-10..8, 1.8)
3.5 (-1.0, 8.2)
3.0 (-1.9, 8.2)
1.3(-1. 1,3.7)
1.8(-1.0,4.6)
0.0 (-4.8, 5.0)
-1.0 (-6.2, 4. 5)
PM10 26.7 (SD 16.1)
PM2512.2(SD7.4)
PM10.25 14.5 (SD 12.2)
PM10 means 17.8-45.6
PM25means 11.2-29.6
PM10_25 means 6.6-16.1
PM10 30.6 (SD 16.2)
PM25 18.7(SD10.5)
PM10.2511.9(SD8.5)
PM10 45.6 (SD 32.3)
PM2529.6(SD21.9)
PM10.25 16.1 (SD 13.0)
PM10 26.7 (SD 16.1)
PM2512.2(SD7.4)
PM10.25 14.5 (SD 12.2)
PM10 32.0 (SD 14.5)
PM2520.8(SD9.6)
PM10.2511.2(SD7.4)
PM1024.5(SD12.8)
PM25 15.7 (SD 9.2)
PM10.25 8.8 (SD 7.0)
PM10 17.8 (SD1 1.7)
PM2511.2(SD7.8)
PM10.256.6(SD6.8)
-------
3
TABLE 9-8 (cont'd). ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY
EFFECT SIZES PER INCREMENTS IN 24-h CONCENTRATIONS OF PM,n, PM?, AND PM,
to
o
o
FROM U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis
Comments**
% increase (95% CI) per
50 ug/m3 PM10 Increase
% increase (95% CI) per
25 ug/m3 PM25 Increase
PM10,PM25andPM10_25
% increase (95% CI) per 25 Mean (Range) Levels
ug/m3 PM10_2 5 Increase Reported***
MORTALITY: Total (nonaccidental) Mortality (cont'd)
i
VO
O
H
6
o
0
H
O
o
H
W
O
HH
s
Klemm et al. (2000)
Klemm and Mason (2003)
Six City reanalysis - overall
Samet et al. (2000a,b)
Dominici et al. (2002, 2003)
90 Largest U.S. Cities
Schwartz (2000a)
Schwartz (2003b)
10 U.S. cities
Burnett et al. (2000)
Burnett and Goldberg (2003)
8 Canadian Cities
Chock et al. (2000)
Pittsburgh, PA
Clyde et al. (2000)
Phoenix, AZ
Fairley(1999)
Fairley (2003)
Santa Clara County, CA
Gamble (1998)
Dallas, TX
Goldberg et al. (2000)
Goldberg and Burnett (2003)
Montreal, CAN
Klemm and Mason (2000)
Atlanta, GA
Levy (1998)
King Co., WA
GAM Strict
GLMNS
GAM strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS (6
knots/yr)
GAM not
used
GAM not
used
GAM Strict
GLMNS
GAM not
used
GAM Strict
GLMNS
GAM not
used
GAM not
used
3.5(2.0,5.1)
2.5(0.8,4.3)
1.4(0.9,1.9)
1.1(0.5,1.7)
3.4(2.6,4.1)
2.8(2.0,3.6)
3.2(1.1,5.5)
2.7(-0.1,5.5)
6(>0, 11)
7.8(2.8,13.1)
8.3(2.9,13.9)
-3.56 (-12.73, 6.58)
—
7.2 (-6.3, 22. 8)
3.0(2.0,4.1)
2.0 (0.9, 3.2)
—
2.8(1.2,4.4)
2.1(0.1,4.2)
< 75 years 2.6 (2.0, 7.3)
> 75 years 1.5 (-3.0, 6.3)
—
8.1(1.6,15.0)
7.0(1.4,13.0)
—
4.2 (p< 0.05)
1.5(p>0.05)
4. 8 (-3.2, 13.4)
1.76 (-3. 53, 7.34)
0.8 (-0.6, 2.1) PM10 means 17.8-45. 6
0.5(-1.0,2.0) PM25 means 11.2-29.6
PM10_25 means 6.6-16.1
— PM10mean range 15.3-52.0
— PM10 mean range
27.1-40.6
1.9 (-0.1, 3. 9) PM1025.9(maxl21)
1.8 (-0.6, 4.4) PM2 5 13. 3 (max 86)
PM10.25 12.9 (max 99)
< 75 years 0.7 (-1.7, 3.7) NR
> 75 years 1.3 (-1.3, 3.8)
— PM10 mean 45. 4
4.5 (-7.6, 18.1) PM10 34 (6, 165)
3.3 (-5.3, 12.6) PM2513(2, 105)
PM10.2.511(0,45)
— PM1024.5(11,86)
— PM25 17.6 (4.6, 71.7)
1.4 (-11.3, 15.9) PM25 19.9 (1.0, 54.8)
PM10.25 10. 1(0.2, 39.5)
— PM1029.8(6.0, 123.0)
PM; 28. 7 (16. 3, 92.2)
-------
3
^w'
s
TABLE 9-8 (cont'd). ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY
EFFECT SIZES PER INCREMENTS IN 24-h CONCENTRATIONS OF PM,n, PM,S AND PM,
to
o
o
OJ
VO
oo
o
O
H
6
o
o
H
O
FROM U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
MORTALITY: Total
Lipfert et al. (2000a)
Philadelphia, PA
Lippmann et al. (2000)
Ito (2003)
Detroit, MI
Moolgavkar (2000a)
Moolgavkar (2003)
Los Angeles, CA
Moolgavkar (2000a)
Moolgavkar (2003)
Cook Co., IL
Analysis
Comments**
% increase (95% CI) per
50 ug/m3 PM10 Increase
% increase (95% CI) per
25 ug/m3 PM25 Increase
PM10, PM2 5 and PM10_2 5
% increase (95% CI) per 25 Mean (Range) Levels
ug/m3 PM10_2 5 Increase Reported* * *
(nonaccidental) Mortality (cont'd)
GAM not
used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
5. 99 (p> 0.055)
3. 3 (-2.0, 8.9)
3.1 (-2.2, 8.7)
2.4(0.5,4.2)
2.3(0.5,4.1)
2.4(1.4,3.5)
2.6(1.6,3.6)
Ostro(1995) GAM not —
San Bemadino and Riverside used
Counties, CA
Schwartz (2000b) GLMNS —
Schwartz (2003a)
Boston, MA
4.21 (p< 0.055)
1.9 (-1.8, 5.7)
2.0 (-1.7, 5.8)
1.5(0,3.0)
1.4 (-0.4, 3.2)
0.28 (-0.61, 1.17)
5. 8 (4. 5, 73) (15-day)
9.7(8.2, 11. 2) (60-day)
5.07 (p > 0.055) PM10 32.20 (7.0, 95.0)
PM25 17.28 (-0.6, 72.6)
PM10.25 6.80 (-20.0, 28.3)
3.2 (-1.9, 8.6) PM1031(12, 105)
2.8 (-2.2, 8.1) PM2518(6, 86)
PM10.2513(4,50)
mean (5%, 95%)
— PM10 median 44 (7, 166)
PM25 22 (4, 86)
— PM10 median 35 (3, 365)
— PM2532.5(9.3, 190.1)
(estimated from visibility)
— PM25 15.6 (±9.2)
Laden et al. (2000)
Schwartz (2003a)
Six City source-oriented
analysis
GLMPS
-5.1(-13.9,4.6)crustal
9.3 (4.0, 14.9) traffic
2.0 (-0.3, 4.4) coal
PM2 5 same as Six City
O
H H
H
W
Tsai et al. (2000)
Newark, NJ
Tsai et al. (2000)
Camden, NJ
GAM not
used
GAM not
used
5.65 (4.62, 6.70)
11.07(0.70,22.51)
4.34 (2.82, 5.89)
5.65(0.11,11.51)
— PM1555(SD6.5)
PM2542.1(SD22.0)
— PM15 47.0 (SD 20.9)
PM25 39.9 (SD 18.0)
-------
TABLE 9-8 (cont'd). ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY
EFFECT SIZES PER INCREMENTS IN 24-h CONCENTRATIONS OF PM,n, PM?, AND PM,
to
o
o
Original study*
Reanalysis study
Study location
Analysis
Comments**
FROM U.S. AND
% increase (95% CI) per
50 ug/m3 PM10 Increase
CANADIAN STUDIES
% increase (95% CI) per % increase (95% CI) per 25
25 ug/m3 PM25 Increase ug/m3 PM10_25 Increase
^ ^'^lU-2.5
PM10,PM25andPM10_25
Mean (Range) Levels
Reported***
MORTALITY: Total (nonaccidental) Mortality (cont'd)
VO
oo
o
H
6
o
o
H
O
o
s
0
o
HH
H
W
Tsai et al. (2000)
Elizabeth, NJ
Cardio respiratory
Mortality:
Tsai et al. (2000)
Newark, NJ
Tsai et al. (2000)
Camden, NJ
Tsai et al. (2000)
Elizabeth, NJ
Total Cardiovascular Mortality
Ito and Thurston (1996)
Chicago, IL
Pope etal. (1992)
Utah Valley, UT
Fairley(1999)
Fairley (2003)
Santa Clara County, CA
Goldberg et al. (2000)
Goldberg and Burnett (2003)
Montreal, CAN
Lipfert et al. (2000a)
Philadelphia, PA (7-county
area)
GAM not
used
GAM not
used
GAM not
used
GAM not
used
GAM not
used
GAM not
used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not
used
-4.88 (-17.88, 10.19)
7.79(3.65,12.10)
15.03(4.29,26.87)
3.05 (-11.04, 19.36)
1.49 (-0.72, 3.74)
9.36(1.91,17.36)
8.5(0.6,17.0)
8.9(1.3,17.0)
6.92 (p< 0.055)
1.77 (-5.44,9.53) —
5.13(3.09,7.21) —
6.18(0.61,12.06) —
2.28 (-4.97, 10.07) —
— —
6. 3 (-4.1. 17.9) (GAM strict)
6.7 (-2.5, 16.7) 5.0 (-13.3,27.3)
3.48 (-0.16, 7.26) —
10.26 (p< 0.055) 7.57 (p > 0.055)
PM15 47.5 (SD 18. 8)
PM2 5 37. 1(80 19. 8)
PM15 55 (80 6.5)
PM2 5 42.1(80 22.0)
PM15 47.0(80 20.9)
PM2 5 39.9(8018.0)
PM1547.5 (80 18.8)
PM2 5 37.1(80 19.8)
PM10 38 (max 128)
PM1047(11,297)
PM10 34 (6, 165)
PM2513(2, 105)
PM10.2.511(0,45)
PM25 17.6 (4.6, 71.7)
PM10 32.20 (7.0, 95.0)
PM25 17.28 (-0.6, 72.6)
PM10.256.80(-20.0,28.3)
-------
to
O
o
VO
oo
to
TABLE 9-8 (cont'd). ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY
EFFECT SIZES PER INCREMENTS IN 24-h CONCENTRATIONS OF PM10, PM2 5 AND PM10 2 5
FROM U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Total Cardiovascular Mortality
Lippmann et al. (2000)
Ito (2003)
Detroit, MI
Mar et al. (2000)
Mar et al. (2003)
Phoenix, AZ
Moolgavkar (2000a)
Moolgavkar (2003)
Los Angeles, CA
Moolgavkar (2000a)
Moolgavkar (2003)
Cook Co., IL
Ostro et al. (2000)
Ostro et al. (2003)
Coachella Valley, CA
Ostro (1995)
San Bemadino and Riverside
Counties, CA
Total Respiratory
Mortality:
Ito and Thurston (1996)
Chicago, IL
Pope etal. (1992)
Utah Valley, UT
Fairley(1999)
Fairley (2003)
Santa Clara County, CA
Analysis
Comments**
• (cont'd)
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not
used
GAM not
used
GAM not
used
GAM Strict
GLMNS
% increase (95% CI) per
50 ug/m3 PM10 Increase
5.4 (-2.6, 14.0)
4.9 (-3. 0,13.5)
9.7(1.7,18.3)
9.5(0.6,19.3)
4.5(1.6,7.5)
3.9(0.6,7.4)
2.2(0.3,4.1)
1.2 (-0.8, 3.1)
5.5(1.6,9.5)
5.1(1.2,9.1)
6.77(1.97,11.79)
19.78(3.51,38.61)
10.7 (-3.7, 27.2)
10.8 (-3.4, 27.1)
% increase (95% CI) per % increase (95% CI) per
25 ug/m3 PM25 Increase 25 ug/m3 PM10_25 Increase
2.2 (-3.2, 7.9) 6.7 (-1.0, 15.0)
2.0 (-3.4, 7.7) 6.0 (-1.6, 14.3)
18.0(4.9,32.6) 6.4(1.3,11.7)
19.1(3.9,36.4) 6.2(0.8,12.0)
2.6(0.4,4.9) —
1.7 (-0.8, 4. 3)
9. 8 (-5.7, 27.9) 2.9(0.7,5.2)
10.2 (-5.3, 28.3) 2.7(0.4,5.1)
0.69 (-0.34, 1.74) —
— —
— —
11. 7 (-9. 8, 38.3) (GAM strict)
13.5 (-3.6, 33.7) 32.1 (-9.1, 92.2)
PM10, PM2 5 and PM10_2 5
Mean (Range) Levels
Reported***
PM1031(12, 105)
PM2518(6, 86)
PM10.2513(4,50)
mean (10%, 90%)
18.0(4.9,32.6)
19.1(3.9,36.4)
PM10 median 44 (7, 166)
PM2 5 median 22 (4, 86)
PM10 median 35 (3, 365)
PM10 47.4(3,417)
PM2516.8(5,48)
PM10.25 17.9 (0, 149)
PM25 32.5 (9.3, 190.1)
(estimated from visibility)
PM10 38 (max 128)
PM1047(11,297)
PM10 34 (6, 165)
PM2513(2, 105)
PM10.2.511(0,45)
-------
3
TABLE 9-8 (cont'd). ESTIMATED TOTAL, CARDIOVASCULAR AND RESPIRATORY MORTALITY
EFFECT SIZES PER INCREMENTS IN 24-h CONCENTRATIONS OF PM,n, PM,S AND PM,
to
o
o
OJ
FROM U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
PM10, PM2 5 and PM10_2 5
Analysis % increase (95% CI) per % increase (95% CI) per % increase (95% CI) per 25 Mean (Range) Levels
Comments** 50 ug/m3 PM10 Increase 25 ug/m3 PM25 Increase ug/m3 PM10_25 Increase Reported***
Total Respiratory Mortality (cont'd)
VO
oo
OJ
o
H
6
O
o
H
O
Goldberg et al. (2000)
Goldberg and Burnett (2003)
Montreal, CAN
Lippmann et al. (2000)
Ito (2003)
Detroit, MI
Ostro(1995)
San Bemadino and Riverside
Counties, CA
COPD Mortality:
Moolgavkar (2000a)
Moolgavkar (2003)
Cook Co., IL
Moolgavkar (2000a)
Mookgavkar (2003)
Los Angeles, CA
GAM Strict —
GLMNS
GAM Strict 7.5 (- 10.5, 29.2)
GLMNS 7.9 (-10.2, 29.7)
GAM not —
used
GAM Strict 5.5(0.2,11.0)
GLMNS 4.5 (-1.6, 11.0)
GAM Strict 4.4 (-3.2, 12.6)
GLMNS 6.2 (-3.4, 16.7)
* Both original published studies and recent reanalyses reported in HEI
et al. (1996) were assessed in 1996 PM AQCD.
21.6(13.0,31.0) — PM25 17.6 (4.6, 71.7)
2.3 (-10.4, 16.7) 7.0 (-9.5, 26. 5) PM10 31 (12, 105)
3.1 (-9.7, 17.7) 6.4 (-10.0, 25.7) PM2518(6, 86)
PM10.2513(4,50)
mean (10%, 90%)
2.08 (-0.35, 4.51) — PM25 32.5 (9.3, 190.1)
(estimated from visibility)
— — PM10 median 35 (3, 365)
1.0 (-5.1, 7.4) — PM10 median 44 (7, 166)
0.5 (-6.8, 8.4) PM2522(4, 86)
(2003) Special Report for many cited here. Original studies published before 1996 and Schwartz
^w'
s
** Where GAM not used in original analysis cited, original results are reported here. Otherwise reanalyses results are reported here if GAM (default) was used in original
analysis. GAM strict = GAM with stringent criteria; GLM = general linear model; NS = natural splines; BS = B splines; PS = penalized splines.
*** Mean (minimum, maximum) 24-h PM level in parentheses unless otherwise noted.
O
HH
H
W
-------
1 reported in HEI (2003). The pooled estimate of PM10-mortality relative risks for European cities
2 comport well with estimates derived from U.S. data.
3 Certain other individual-city studies using similar methodology in analyses for each city
4 (but not generating combined overall pooled effect estimates) also report variations in PM effect
5 size estimates between cities and in their robustness to inclusion of gaseous co-pollutants in
6 multi-pollutant models. Thus, one cannot entirely rule out that real differences may exist in
7 excess risk levels associated with varying size distributions, number, or mass of the chemical
8 constituents of ambient PM; the combined influences of varying co-pollutants present in the
9 ambient air pollution mix from location to location or season to season; or to variations in the
10 relationship between exposure and ambient PM concentration.
11 Nevertheless, there still appears to be reasonably good consistency among the results
12 derived from reanalyses (HEI, 2003) of several new multi-city studies providing pooled analyses
13 of data combined across multiple cities (thought to yield the most precise effect size estimates).
14 Such reanalyses produced an overall U.S. nationwide effects estimate for percent excess total
15 (nonaccidental) deaths per 50 |ig/m3 increase in 24-h PM10 of 1.4% at 1 day lag (1.1% using
16 GLM) in the 90 largest U.S. cities (about twice that in the Northeast region); 3.4% using GAM
17 (2.8% GLM) for average of 0 and 1 day lags in 10 U.S. cities; 3.6% using GAM (2.7% GLM)
18 for 1-day lag in the eight largest Canadian cities; and 3.0% using GAM (2.1% GLM) in
19 APHEA2 for average of and 1-day lags for 29 European cities during 1990-1997. These
20 combined estimates are reasonably consistent with the range of PM10 estimates previously
21 reported in the 1996 PM AQCD (i.e., 1.5 to 8.5% per 50 |ig/m3 PM10). These and other excess
22 risk estimates from many other individual-city studies comport well with a number of new
23 studies confirming increased cause-specific cardiovascular- and respiratory-related mortality,
24 and those noted below as showing ambient PM associations with increased cardiovascular and
25 respiratory hospital admissions and medical visits.
26
27 Fine and Coarse Particle Effect Size Estimates. Table 9-6 also summarizes effects
28 estimates (RR values) for increased mortality and/or morbidity associated with variable
29 increments in short-term (24-h) exposures to PM10, ambient fine particles indexed by various
30 fine PM indicators (PM2 5, sulfates, H+, etc.) and for inhalable thoracic fraction coarse particles
31 (i.e., PM10_2 5) in U.S. and Canadian cities. The table includes studies that were highlighted in
June 2003 9-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 comparable tables in the 1996 PM AQCD which did not use GAM analyses with default
2 convergence criteria; or for those few that did and have since been reanalyzed by more
3 appropriate alternative methods, the results of the reanalyses are presented as reported in HEI
4 (2003). For purposes of comparison across studies, results of single-pollutant models are
5 presented in these tables; co-pollutant model results were summarized and/or discussed in more
6 detail in Chapter 8 Appendix tables and/or main text.
7 The effect size estimates derived for PM2 5 as an ambient fine particle indicator
8 (especially those based on directly measured versus estimated PM2 5 levels) generally appear to
9 fall in the range of 2.0 to 6.0% increase in total (nonaccidental) deaths per 25-|ig/m3 increment
10 in 24-h PM2 5 for U.S. and Canadian cities. Cause-specific effects estimates appear to fall mainly
11 in the range of 2.0 to 10.0% per 25 |ig/m3 24-h PM25 for cardiovascular or combined
12 cardiorespiratory mortality (although one estimate for cardiovascular mortality ranged up to
13 about 19%) and 2.0 to 14.0% per 25 |ig/m3 24-h PM2 5 for respiratory mortality in U.S. cities.
14 As noted earlier, there was only one study in the 1996 PM AQCD, the Harvard Six Cities
15 study (Schwartz et al., 1996), in which the relative importance of fine and coarse particles was
16 examined. That study suggested that fine particles, but not coarse particles, were associated with
17 daily mortality. Both Schwartz (2003a) and Klemm and Mason (2003) have carried out
18 reanalyses of the same Harvard Six-Cities data set using GAM (stringent convergence criteria)
19 and/or other alternate approaches and have essentially replicated the original findings, abeit
20 finding slightly smaller effect size estimates than obtained in the original GAM (default)
21 analyses reported by Schwartz et al. (1996). In addition, several more studies have analyzed
22 both PM25 and PM10_25 for their associations with mortality (see Figure 9-16). Although some of
23 these studies (e.g., the Santa Clara County, CA, analysis and the eight largest Canadian cities
24 analysis) suggest that PM2 5 is more important than PM10_2 5 in predicting mortality fluctuations,
25 several others (e.g., the Phoenix, AZ, and Santiago, Chile studies) seem to suggest that PM10_2 5
26 may be as important as PM2 5 in certain locations (some shown to date being drier, more arid
27 areas). Seasonal dependence of PM components' associations observed in some of the locations
28 (e.g., higher coarse [PM10_2 5] fraction estimates for summer than winter in Santiago, Chile) hint
29 at possible contributions of biogenic materials (e.g., molds, endotoxins, etc.) to the observed
30 coarse particle effects in at least some locations. Overall, for U.S. and Canadian cities, effect
31 size estimates for the coarse fraction (PM10_2 5) generally appear to fall mainly in the range of
June 2003 9-85 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
Percent excess death (total unless otherwise noted) per
25 ug/m3 increase in PM2.5 (•) or PMio-2.5 (O)
VO
oo
o
H
6
o
o
H
O
o
0
o
I-H
H
W
Schwartz (2003a)
Harvard 6 Cities
Klemm and Mason (2003)*
Harvard 6 Cities
(recomputed)
Burnett and Goldberg (2003)*
8 Canadian Cities
Chock et al. (2000)
Pittsburgh, PA
Klemm and Mason (2000)
Atlanta, GA ~
Lipfert et al. (2000a)
Philadelphia, PA
Ito (2003)*
Detroit, Ml
Mar et al (2003)*
Phoenix, AZ
Fairley (2003)*
Cifuentes et al. (2000) _
Santiago, Chile
Figure 9-16. 1
e
(
-20 2 4 6 8 10 12 14 16
i I i i i i i i i i
Lag 0-*-1 day A
Lag 1 day p,*
w
_
Lag 0 day O J aOe less than 75
Lag 0 day
r^i
_
Lag 1 day °
i 1 j
Lag 0 day u
i n j «
Lag 0 day •
Lag 1-*-2 day (|S
18
i
'ercent excess risks estimated per 25-ug/m3 increase in PM2 5 or PM10_2 5 from new
evaluating both PM2 5 and PM10_2 5 data for multiple years. All lags = 1 day, unless
)therwise.
studies
indicated
-------
1 2.0 to 6.0% excess total (nonaccidental) deaths per 25 |ig/m3 of 24-h PM10_25. Respective
2 increases for cause-specific mortality mainly range from -3.0 to 7.0% for cardiovascular and
3 from ~3.0 to 6.0% for respiratory causes per 25-|ig/m3 increase in 24-h PM10_2 5.
4
5 Chemical Components of Particulate Matter. Several new studies examined the role of
6 specific chemical components of PM in relation to mortality risks. Studies of U.S. and Canadian
7 cities showed mortality associations with one or more of several specific fine particle
8 components of PM, including H+, sulfate, nitrate, as well as COH; but their relative importance
9 varied from city to city, likely depending, in part, on their concentrations (e.g., no clear
10 associations in those cities where H+ and sulfate levels were very low [i.e., circa nondetection
11 limits]). Figure 9-17 depicts relatively consistent estimates of total mortality excess risk
12 resulting from a 5-|ig/m3 increase in sulfate, possibly reflecting impacts of sulfate per se or
13 perhaps sulfate serving as a surrogate for fine particles in general. Sulfate effect size estimates
14 generally fall in the range of 0.5 to 4% excess total mortality per 5-|ig/m3 increase for U.S. and
15 Canadian cities.
16 A significant factor in some western cities is the occasional occurrence of high levels of
17 windblown crustal particles that constitute much of the coarse PM fraction. The small-size tail
18 of windblown crustal particles extends into the PM2 ^ (intermodal) size range at times
19 constituting a substantial fraction of PM25. Claiborn et al. (2000) report that in Spokane, WA,
20 PM25 constitutes about 30% of PM10 on dust event days, but 48% on days preceding the dust
21 event. The intermodal fraction represents about 51% of PM25 during windblown dust events,
22 about 28% on preceding days. However, PMl in Spokane often shows little change during dust
23 events, when coarse particles (presumably crustal particles) are transported into the region. The
24 lack of increased mortality during time periods with high wind speeds and presumably high
25 crustal material concentrations was shown by Schwartz et al. (1999) for Spokane, and by Pope
26 et al. (1999a) for three cities in the Wasatch front region of Utah. Other recent studies suggest
27 that coarse particles, as well as fine particles, may be associated with excess mortality in certain
28 U.S. locations e.g., in Phoenix, AZ (Smith et al., 2000; Clyde et al., 2000; Mar et al., 2000) the
29 Coachella Valley of California (Ostro et al., 2000), Mexico City (Castillejos et al., 2000) or
30 Santiago, Chile (Cifuentes et al., 2000). However, the coarse particle association with mortality
does not appear to be caused by the crustal components. An important advantage of using
June 2003 9-87 DRAFT-DO NOT QUOTE OR CITE
-------
Klemm and Mason (2003)* -
Harvard 6 cities (recomputed)
Fairley (2003)* -
Santa Clara Co.
Klemm et al. (2000)
Atlanta, GA
Lipfert et al. (2000a) -
Philadelphia, PA
lto(2003)* -
Detroit, Ml
Tsai et al.(2000) -
3 NJ cities
Percent excess death (total non-accidental mortality)
per 5 Mg/m3 increase in sulfate
-2 0 2 4 6 8 10
Newark
Camden
Elizabeth
Figure 9-17. Relative risks estimated per 5-ug/m3 increase in sulfate from U.S. and
Canadian studies in which both PM2 5 and PM10_2 5 data were available.
1 source profiles for PM2 5 in western cities is that it allows separation of crustal PM from
2 accumulation-mode PM derived from anthropogenic origins.
3 Several new studies highlighted in Chapter 8 conducted source-category-oriented
4 evaluations of PM components using factor analysis (see Table 9-9). The results of these studies
5 (Laden et al., 2000; Mar et al., 2000; Tsai et al., 2000; Ozkaynak et al., 1996) generally suggest
6 that a number of combustion-related source-categories are associated with excess mortality risk,
7 including: regional sulfate; automobile emissions; coal combustion; oil burning; and vegetative
8 (biomass) burning. In contrast, the crustal factor from fine particles was generally not positively
9 associated with total mortality, with Mar et al. (2000) reporting a negative association between
10 the crustal component of PM25 and cardiovascular mortality.
11 However, these source-category-oriented evaluation results are derived from relatively
12 limited underlying analytic bases for resolving source categories and the identification of source
June 2003
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-9. SUMMARY OF SOURCE-ORIENTED EVALUATIONS OF PM
COMPONENTS IN RECENT STUDIES
Author, City
Source types identified (or suggested)
and associated variables
Source types associated with mortality
(Comments)
Laden et al.,
(2000);
Schwartz (2003)*
Harvard Six Cities.
1979-1988.
Mar et al.
(2000, 2003)*
Phoenix, AZ.
1995-1997.
5*0/7 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
PM2S (fromDFPSS) trace elements:
Motor vehicle emissions and re-suspended
road dust: Mn, Fe, Zn, Pb, OC, EC, CO,
andNO2
5*0/7: Al, Si, and Fe
Vegetative burning: OC, and Ks
(soil-corrected potassium)
Local SO2 sources: SO2
Regional sulfate: S
Strongest increase in daily mortality was
associated with the mobile source factor.
Coal combustion factor was also positively
associated with mortality. Crustal factor
from fine particles not associated (negative
but not significant) with mortality. Coal
and mobile sources account for the majority
of fine particles in each city.
PM2 5 factors results: Motor vehicle factor
(1 day lag), vegetative burning factor (3 day
lag), and regional sulfate factor (0 day lag)
were significantly positively associated
with cardiovascular mortality.
Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ.
1981-1983.
Ozkaynak et al.
(1996).
Toronto, Canada.
PM10_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
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
Motor vehicle emissions: CO, CoH, and
NO,
Factors from dichot PM10_2 5 trace elements
not analyzed for their associations with
mortality because of the small sample size
(every 3rd-day samples from June 1996).
Oil burning, industry, secondary aerosol,
and motor vehicle factors were associated
with mortality.
Motor vehicle factor was a significant
predictor for total, cancer, cardiovascular,
respiratory, and pneumonia deaths.
*Note: The study was originally analyzed using GAM models only with default convergence criteria using at
least two non-parametric smoothing terms, but was later reanalyzed using more stringent convergence criteria
and/or other approaches.
1 categories must be viewed with caution at this time. Nevertheless, although somewhat limited at
2 this time, the new factor analysis results appear to implicate ambient PM derived from fossil fuel
3 (oil, coal) combustion and vegetative burning, as well as secondarily formed sulfates, as
4 important contributors to observed mortality effects, but not crustal particles.
June 2003
9-89
DRAFT-DO NOT QUOTE OR CITE
-------
1 In summary, the new evidence suggests that exposure to particles from several different
2 source categories, and of different composition and size, may have independent associations
3 with health outcomes. The excess risks from different types of combustion sources (coal, oil,
4 gasoline, wood, and vegetation) may vary from place to place and from time to time, so that
5 some intra-regional and inter-regional heterogeneity would be expected. Likewise, although
6 earlier evaluations in the 1996 PM AQCD seemed to indicate coarse particles and intermodal
7 particles of crustal composition as not likely being associated with adverse health effects, there
8 are now some reasonably credible studies suggesting that coarse particles (although not
9 necessarily those of crustal composition) may be associated with excess mortality in at least
10 some locations. These notably include areas where past deposition of fine PM metals from
11 smelter (Phoenix) or steel mills (Steubenville) onto surrounding soils may result in enhanced
12 toxicity of later resuspended coarse (PM10_25) particles.
13
14 Relationships of Ambient Particulate Matter Concentrations to Morbidity Outcomes
15 New epidemiology studies add greatly to the overall database relating morbidity outcomes
16 to ambient PM levels. These include much additional evidence for cardiovascular and
17 respiratory diseases being related to ambient PM. The newer epidemiology studies expand the
18 evidence on cardiovascular (CVD) disease and are discussed first below, followed by discussion
19 of respiratory disease effects with particular emphasis on newly enhanced evidence for
20 PM-asthma relationships. Table 9-10 summarizes cardiovascular and respiratory-related
21 morbidity effect size estimates for variable increments in PM10, PM2 5, and PM10_25
22 concentrations for studies of U.S. and Canadian cities.
23
24 Cardiovascular Effects of Ambient Particulate Matter Exposures
25 Cardiovascular Hospital Admissions. Just two studies were available for review in the 1996
26 PM AQCD that provided data on acute cardiovascular morbidity outcomes (Schwartz and
27 Morris, 1995; Burnett et al., 1995). Both studies were of ecologic time series design using
28 standard statistical methods. Analyzing 4 years of data on the > 65-year-old Medicare
29 population in Detroit, MI, Schwartz and Morris (1995) reported significant associations between
30 ischemic heart disease admissions and PM10, controlling for environmental covariates. Based on
31 an analysis of admissions data from 168 hospitals throughout Ontario, Canada, Burnett and
June 2003 9-90 DRAFT-DO NOT QUOTE OR CITE
-------
to
O
o
TABLE 9-10. CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT SIZE ESTIMATES PER
INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5 IN U.S. AND CANADIAN STUDIES
H
6
o
o
H
O
o
HH
H
W
Original study*
Reanalysis study
Study location
Analysis
Comments**
% increase (95% CI) per % increase (95% CI) per
50 ug/m3 PM10 Increase 25 ug/m3 PM2 5 Increase
% increase (95% CI) per PM10, PM2 s and PM10_2.5
25 ug/m3 PM10_2 5 Mean (Range) Levels
Increase Reported***
CARDIOVASCULAR MORBIDITY
Total Cardiovascular Hospital Admissions:
Samet et al. (2000a,b)
14 U.S. Cities
(> 65 years)
Zanobetti and Schwartz
(2003)
Linn et al. (2000)
Los Angeles, CA
(> 29 years)
Moolgavkar (2000b)
Moolgavkar (2003)
Cook Co., IL
(> 65 years)
Moolgavkar (2000b)
Moolgavkar (2003)
Los Angeles, CA
(> 65 years)
Morris and Naumova
(1998)
Chicago, IL (> 65
years)
Tolbertetal., (2000a)
Atlanta, GA 1993-1998
Tolbert et al. (2000a)
Atlanta, GA (all ages)
Stieb et al. (2000)
St. John, CAN (all ages)
strict GAM
GLMNS
GLMPS
GAM not used
strict
GAMloodf
GLM NSloodf
GAM30df
GAMloodf
GLMNSloodf
GAM not used
GAM not used
GAM not used
GAM not used
4.95% (3.95-5.95) —
4.8% (3. 55-6.0)
5.0% (4.0-5.95)
3. 25% (2.04, 4.47) —
4.05% (2.9-5.2) —
4.25% (3.0-5.5)
3.35% (1.2-5.5) 3.95% (2.2-5.7)
2.7% (0.6-4.8) 2.9% (1.2-4.6)
2.75% (0.1-5.4) 3. 15% (1.1-5.2)
3.92 (1.02, 6.90) —
- 8.2% (p = 0.002) —
5.1 (-7.9, 19.9) 6.1 (-3.1, 16.2)
39.2 (5.0, 84.4) 15.11 (0.61, 11.03)#
— PM10 means 24.4-45.3
— PM10 45.5 (5, 132)
— PM10 median 35 (3, 365)
— PM10 median 44 (7, 166)
PM2 5 median 22 (4, 86)
— PM1041(6, 117)
— Period 1PM10 30.1
(SD 12.4)
17.6 (-4.6, 45.0) PM10 29.1 (SD 12.0)
PM2519.4(SD9.35)
PM10.25 9.39 (SD 4.52)
— summer 93
PM10 14.0 (max 70.3)
PM25 8.5 (max 53.2)
-------
to
O
o
VO
to
H
6
o
o
H
O
o
HH
H
W
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study Analysis
Study location Comments**
Total Cardiovascular Hospital Admissions:
Burnett et al. (1997) GAM not used
Toronto, CAN (all ages)
% increase (95% CI) per % increase (95% CI) per
50 ug/m3 PM10 Increase 25 ug/m3 PM2 5 Increase
(cont'd)
12.07 (1.43, 23.81)# 7.18 (-0.61, 15.60)#
% increase (95% CI) per
25 ug/m3 PM10_2.5
Increase
20.46 (8.24, 34.06)#
PM10,PM25andPM10_25
Mean (Range) Levels
Reported***
PM10 28.4 (4, 102)
PM25 16.8(1,66)
PM10.25 11.6 (1,56)
Ischemic Heart Disease Hospital Admissions:
Lippmann et al. (2000) Strict GAM
Detroit, MI (> 65 years) GLM NS
Ito 2003
Dysrhythmias Hospital Admissions:
Tolbert et al. (2000a) GAM not used
Atlanta, GA (all ages)
Lippmann et al. (2000) Strict GAM
Detroit, MI (> 65 years) GLM NS
Ito (2003)
Heart Failure Hospital Admissions:
Linn et al. (2000) GAM not used
Los Angeles, CA
(> 29 years)
Lippmann et al. (2000) Strict GAM
Ito (2003) GLM NS
Detroit, MI (> 65 years)
Myocardial Infarction Hospital
Admissions:
Linn et al. (2000) GAM not used
Los Angeles, CA
(> 29 years)
8.0% (-0.3-17.1) 3.65% (-2.05-9.7)
6.2% (-2.0-15.0) 3.0% (-2.7-9.0)
13.41 (-14.08,48.99) 6.11 (-12.63, 28.86)
2.8% (- 10.9-18.7) 3.2% (-6.6-14.0)
2.0% (-11.7-17.7) 2.6% (-7.1-13. 3)
2.02 (-0.94, 5.06) —
9.2% (-0.3-19.6) 8.0% (1.4-15.0)
8.4% (- 1.0-18.7) 6.8% (0.3-13.8)
3.04(0.06,6.12) —
10.2% (2.4-18.6)
8.1% (0.4-16.4)
53.16(2.07, 129.81)
0.1% (-12.4-14.4)
0.0% (-12.5-14.3)
4.4% (-4.0-13.5)
4.9% (-3. 55-14.1)
"
PM1031(maxl05)
PM2518(6, 86)
PM10.25 13 (4, 50)
PM25 19.4 (SD 9.35)
PM10.25 9.39 (SD 4.52)
PM1031(maxl05)
PM25 18(6,86)
PM10.25 13 (4, 50)
PM10 45.5 (5, 132)
PM1031(maxl05)
PM2518(6, 86)
PM10.25 13 (4, 50)
PM10 45.5 (5, 132)
-------
to
O
o
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis % increase (95% CI) per % increase (95% CI) per
Comments** 50 ug/m3 PM10 Increase 25 ug/m3 PM25 Increase
% increase (95% CI) per PM10, PM25 and PM10_2.5
25 ug/m3 PM10_2,
Increase
Mean (Range) Levels
Reported***
Cardiac arrhythmia Hospital
Admissions:
Linn et al. (2000)
Los Angeles, CA
(> 29 years)
GAM not used 1.01 (-1.93, 4.02)
Cerebrovascular Hospital Admissions:
Linn et al. (2000) GAMnotused 0.30 (-2.13, 2.79)
Los Angeles, CA
(> 29 years)
Stroke Hospital Admissions:
Linn et al. (2000)
Los Angeles, CA
(> 29 years)
GAM not used 6.72 (3.64, 9.90)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
£
H
6
o
o
H
O
O
s
0
o
HH
H
W
RESPIRATORY MORBIDITY
Total Respiratory Hospital Admissions:
Thurston et al. (1994) GAM not used 23.26 (2.03, 44.49) 15.00 (1.97, 28.03)
Toronto, Canada
Linn et al. (2000) GAMnotused 2.89(1.09,4.72) —
Los Angeles, CA
(> 29 years)
Schwartz etal. (1996) GAMnotused 5.8(0.5,11.4) —
Cleveland, OH (> 65
years)
Lumley and Heagerty GAMnotused — 5.91(1.10,10.97)
(1999)
King County, WA (all
ages)
22.25 (-9.53, 54.03) PM10 29.5-38.8 (max 96.
PM2 5 15.8-22.3 (max 66
PM10.25 12.7-16.5 (max
33.0)
— PM10 45.5 (5, 132)
— PM10 43
— PMjNR
0)
.0)
-------
to
O
o
VO
H
6
o
o
H
O
o
HH
H
W
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis
Comments**
% increase (95% CI) per PM10, PM2 5 and PM10_2.5
% increase (95% CI) per % increase (95% CI) per 25 ug/m3 PM10_2.5 Mean (Range) Levels
50 ug/m3 PM10 Increase 25 ug/m3 PM25 Increase Increase Reported***
Total Respiratory Hospital Admissions:
(cont'd)
Burnett etal. (1997)
Toronto, CAN (all ages)
Delfino etal. (1997)
Montreal, CAN (> 64
years)
Delfino etal. (1998)
Montreal, CAN (> 64
years)
Stieb et al. (2000)
St. John, CAN (all ages)
GAM not used
GAM not used
GAM not used
GAM not used
10.93 (4.53, 17.72) 8.61 (3.39, 14.08) 12.71 (5.33, 20.74)
36.62 (10.02, 63.21) 23.88 (4.94, 42.83) —
— 13. 17 (-0.22, 26.57) —
8.8(1.8,16.4) 5.69(0.61,11.03) —
PM1028.1(4, 102)
PM25 16.8(1,66)
PM10.25 11.6 (1,56)
summer 93
PM1021.7(max51)
PM25 12.2 (max 31)
PM25 18.6 (SD 9.3)
summer 93
PM1014.0 (max 70.3)
PM25 8.5 (max 53.2)
Pneumonia Hospital Admissions:
Samet et al. (2000a,b)
14 U.S. Cities (> 65
years)
Zanobetti and Schwartz
(2003)
Lippmann et al. (2000)
Detroit, MI (> 65 years)
Ito (2003)
Strict GAM
GLMNS
GLMPS
Strict GAM
GLMNS
8.8(5.9,11.8) — —
2.9 (0.2, 5.6)
6.3 (2.5, 10.3)
18.1(5.3,32.5) 10.5(1.8,19.8) 9.9 (-0.1, 22.0)
18.6(5.6,33.1) 10.1(1.5,19.5) 11.2 (-0.02, 23.6)
PM10 means 24.4-45.3
PM1031(maxl05)
PM25 18(6,86)
PM10.25 13 (4, 50)
COPD Hospital Admissions:
Samet et al. (2000a,b)
14 U.S. Cities
(> 65 years)
Zanobetti and Schwartz
(2003)
Strict GAM
GLMNS
GLMPS
8.8(4.8,13.0) — —
6.8 (2.8, 10.8)
8.0(4.3, 11.9)
PM10 means 24.4-45.3
-------
to
O
o
VO
H
6
o
o
H
O
o
HH
H
W
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis
Comments**
COPD Hospital Admissions (cont'd)
Linn et al. (2000) GAM not used
Los Angeles, CA
(> 29 years)
Tolbert et al. (2000a)
Atlanta, GA (all ages)
Lippmann et al. (2000)
Detroit, MI (> 65 years)
Ito (2003)
Moolgavkar (2000c)
Cook Co., IL
(> 65 years)
Moolgavkar 2003
Moolgavkar (2000c)
Los Angeles, CA
(> 65 years)
Moolgavkar 2003
GAM not used
Strict GAM
GLMNS
Strict GAM:
lOOdf
Strict GAM:
lOOdf
GLMNS:
lOOdf
% increase (95% CI) per PM10, PM2 5 and PM10_2.5
% increase (95% CI) per % increase (95% CI) per 25 ug/m3 PM10_2.5 Mean (Range) Levels
50 ug/m3 PM10 Increase 25 ug/m3 PM25 Increase Increase Reported***
1.5 (-0.5, 3.5)
-3.5(33.0, -29.9)
6.5 (-7.8, 23.0)
4.6 (-9.4, 20.8)
3.24 (.031, 6.24)
5.52 (2.53-8.59)
5.00 (1.22, 8.91)
— — PM10 45.5 (5, 132)
12.44 (-7.89, 37.24) -23.03 (-50.69, 20.15) PM10 29.1 (SD 12.0)
PM25 19.4 (SD 9.35)
PM10.25 9.39 (SD 4.52)
3.0(-6.9, 13.9) 8.7 (-4.8, 24.0) PM10 31 (max 105)
0.3(-9.3, 10.9) 10.8 (-3. 1,26.5) PM2518(6, 86)
PM10.25 13 (4, 50)
— — PM10 median 35 (3, 365)
2.87 (0.53, 5.27) PM10 median 44 (7, 166)
PM2 5 median 224, 86)
2.59 (-0.29, 5.56) PM10.25NR
Asthma Hospital Admissions:
Choudbury etal. (1997)
Anchorage, AK
Medical Visits (all ages)
Jacobs etal. (1997)
Butte County, CA (all
ages)
Linn et al. (2000)
Los Angeles, CA
(> 29 years)
GAM not used
GAM not used
GAM not used
20.9(11.8,30.8)
6.11 (p>0.05)
1.5 (-2.4, 5.6)
— — PM10 42.5 (1, 565)
— — PM10 34.3 (6.6, 636)
— — PM10 45.5 (5, 132)
-------
to
O
o
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis % increase (95% CI) per
Comments** 50 ug/m3 PM10 Increase
% increase (95% CI) per
25 ug/m3 PM2 5 Increase
% increase (95% CI) per
25 ug/m3 PM10_2 5
Increase
PM10, PM2 5 and PM10_2.5
Mean (Range) Levels
Reported***
o
HH
H
W
Asthma Hospital Admissions: (cont'd)
Lipsettetal. (1997)
Santa Clara Co., CA
(all ages)
Nauenberg and Basu
(1999)
Los Angeles, CA
(all ages)
Tolbert et al. (2000b)
Atlanta, GA (< 17
years)
Tolbert et al. (2000a)
Atlanta, GA (all ages)
GAM not used 34.7 (16, 56.5)
(at 20° F)
GAM not used 20.0(5.3,35)
GAM not used 13.2 (1.2, 26.7)
GAM not used 18.8 (-8.7, 54.4)
2.3 (-14.8,22.7)
21.1 (-18.2,79.3)
PM1061.2(9, 165)
44.8 (SE 17.23)
PM10 38.9 (9, 105)
PM1029.1(SD 12.0)
PM2519.4(SD9.35)
PM10.25 9.39 (SD 4.52)
5
H
6
o
o
H
O
Sheppard et al. (1999) Strict GAM
Seattle, WA (< 65 GLM NS
years)
Respiratory Symptoms
10.9 (2.8, 19.6)
8.1(0.1, 16.7)
Odds Ratio (95% CI) for
50 ug/m3 increase in
PM10
8.7 (3.2, 14.4)
6.5(1.1,12.0)
Odds Ratio (95% CI) for
25 ug/m3 increase in
PM25
5.5 (0, 14.0)
5. 5 (-2.7, 11.1)
Odds Ratio (95% CI) for
25 ug/m3 increase in
PM10.2.5
PM1031.5(90%55)
PM2 5 16.7 (90% 32)
PM10.25 16.2 (90% 29)
PM10_25 Mean (Range)
Levels Reported**
Schwartz etal. (1994)
6 U.S. cities
(children, cough)
Schwartz etal. (1994)
6 U.S. cities
(children, lower
respiratory symptoms)
GAM not used 1.39 (1.05, 1.85)
GAM not used 2.03 (1.36, 3.04)
1.24(1.00, 1.54)
1.58(1.18,2.10)
PM10 median 30.0 (max
117)
PM25 median 18.0 (max
86)
PM10 median 30.0 (max
117)
PM25 median 18.0 (max
86)
-------
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
to
o
o
VO
OD
^
O
>
H
6
o
o
H
O
O
w
0
o
HH
H
W
ai^H; i^aiiivi/viiLa riLis. ii> ^ISJLIVIILI> i in zi-ii ^ui"N^iLi"N i HA i «ji> a \JF nvi10, nvi25, /vn
IN U.S. AND CANADIAN STUDIES
Original study* % increase (95% CI) per
Reanalysis study Analysis % increase (95% CI) per % increase (95% CI) per 25 ug/m3 PM10_2 5
Study location Comments** 50 ug/m3 PM10 Increase 25 ug/m3 PM25 Increase Increase
Respiratory Symptoms (cont'd)
Neas etal. (1995) GAMnotused — 2.45(1.29,4.64) —
Uniontown, PA
(children, cough)
Ostroetal. (1991) GAMnotused 1.09(0.57,2.10) — —
Denver, CO
(adults, cough)
Pope etal. (1991) GAMnotused 1.28(1.06,1.56) — —
Utah Valley, UT
(lower respiratory
symptoms,
schoolchildren)
Pope etal. (1991) GAMnotused 1.01(0.81,1.27) — —
Utah Valley, UT
(lower respiratory
symptoms, asthmatic
patients)
Neas etal. (1996) GAMnotused NR 1.48 (1.17, 1.88) (1-d) —
State College, PA
(children, cough)
Neas et al. (1996) GAM not used NR 1.59 (0.93, 2.70) (1-d) —
State College, PA
(children, wheeze)
Neas etal. (1996) GAMnotused NR 1.61 (1.21, 2.17) (0-d) —
State College, PA
(children, cold)
Ostroetal. (1995) GAMnotused 1.05(0.64,1.73) — —
Los Angeles, CA
(children, asthma episode)
" ™AlO-2.5
PM10, PM2 5 and PM10_2.5
Mean (Range) Levels
Reported***
PM2 5 24.5 (max 88.1)
PM10 22 (0.5, 73)
PM1044(11, 195)
PM1044(11, 195)
PM1031.9(max82.7)
PM2123.5(max85.8)
PM1031.9(max82.7)
PM2123.5(max85.8)
PM1031.9(max82.7)
PM2123.5(max85.8)
PM10 55.87 (19.63, 101.42)
-------
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
to
o
o
OJ
VO
OD
oo
O
>
H
6
o
o
o
o
w
o
o
H
w
ai^H; i^aiiivi/viiLa riLis. in ^ JSJLIVIILI"N i m zt-ii ^ui"N^iLi"N i HA i «ji> a ur nvi10, nvi25, /vn
IN U.S. AND CANADIAN STUDIES
Original study* % increase (95% CI) per
Reanalysis study Analysis % increase (95% CI) per % increase (95% CI) per 25 ug/m3 PM10_2 5
Study location Comments** 50 ug/m3 PM10 Increase 25 ug/m3 PM25 Increase Increase
Respiratory Symptoms (cont'd)
Ostroetal. (1995) GAMnotused 1.51(1.04,2.17) — —
Los Angeles, CA
(children, shortness of
breath)
Schwartz and Neas (2000) GAMnotused 1.28(0.98,1.67) 1.77(1.23,2.54)
Six Cities reanalysis
(children, cough)
Schwartz and Neas (2000) GAMnotused — 1.61(1.20,2.16) 1.51(0.66,3.43)
Six Cities reanalysis
(children, lower
respiratory symptoms)
Vedal etal. (1998) GAMnotused 1.40(1.14,1.73) — —
Port Alberni, CAN
(children, cough)
Vedal etal. (1998) GAMnotused 1.40(1.03,1.90) — —
Port Alberni, CAN
(children, phlegm)
Vedal etal. (1998) GAMnotused 1.22(1.00,1.47) — —
Port Alberni, CAN
(children, nose
symptoms)
Vedal etal. (1998) GAMnotused 1.34(1.06,1.69) — —
Port Alberni, CAN
(children, sore throat)
Vedal etal. (1998) GAMnotused 1.16(0.82,1.63) — —
Port Alberni, CAN
(children, wheeze)
" ™AlO-2.5
PM10, PM2 5 and PM10_2.5
Mean (Range) Levels
Reported***
PM10 55.87 (19.63, 101.42)
PM2 5 (same as Six Cities)
PM10.25 NR
PM2 5 (same as Six Cities)
PM10.25 NR
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
-------
to
O
o
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis
Comments**
% increase (95% CI) per
50 ug/m3 PM10 Increase
% increase (95% CI) per
25 ug/m3 PM2 5 Increase
% increase (95% CI)
per25ug/m3PM10_25
Increase
PM10, PM2 5 and PM10_2.5
Mean (Range) Levels
Reported***
VO
H
6
o
o
H
O
o
HH
H
W
Respiratory Symptoms (cont'd)
Vedaletal. (1998)
Port Alberni, CAN
(children, chest tightness)
Vedaletal. (1998)
Port Alberni, CAN
(children, dyspnea)
Vedaletal. (1998)
Port Alberni, CAN
(children, any symptom)
GAM not used 1.34 (0.86, 2.09)
GAM not used 1.05 (0.74, 1.49)
GAM not used 1.16 (1.00, 1.34)
PM10 median 22.1(0.2,
159.0) (north site)
PM10 median 22.1(0.2,
159.0) (north site)
PM10 median 22.1(0.2,
159.0) (north site)
Lung Function Changes
Lung Function change
(L/min) (95% CI) for
50 ug/m3 increase in
PM10
Lung Function change
(L/min) (95% CI) for
25 ug/m3 increase in
PM25
Lung Function change
(L/min) (95% CI) for 25
ug/m3 increase in PM10_2 5
PM10_25 Mean (Range)
Levels Reported**
Neasetal. (1995)
Uniontown, PA
(children)
Thurstonetal. (1997)
Connecticut summer
camp
(children)
Naeheretal. (1999)
Southwest VA
(adult women)
Neasetal. (1996)
State College, PA
(children)
GAM not used
GAM not used
GAM not used
GAM not used
amPEFR-3.65 (-6.79,
-0.51)
pmPEFR-1.8 (-5.03,
1.43)
-2.58 (-5.33,+0.35)
PEFR-5.4 (-12.3, 1.5)
(15 ug/m3 S04=)
am PEFR-1.83 (-3.44,
-0.21)
pm PEFR-1.05 (-2.77,
0.67)
pm PEFR-0.64 (-1.73,
0.44)
PM2 5 24.5 (max 88.1)
SO4=7.0 (1.1, 26.7)
am PEFR -6.33 (-12.50, PM10 27.07 (4.89, 69.07)
-0.15)
pm PEFR-2.4 (-8.48,
3.68)
PM25 21.62 (3.48, 59.65)
PM10.25 5.72 (0.00, 19.78)
PM25 23.5 (max 85.8)
-------
to
O
o
TABLE 9-10 (cont'd). CARDIOVASCULAR AND RESPIRATORY-RELATED MORBIDITY EFFECT
SIZE ESTIMATES PER INCREMENT IN 24-h CONCENTRATIONS OF PM10, PM2 5, AND PM10 2 5
IN U.S. AND CANADIAN STUDIES
Original study*
Reanalysis study
Study location
Analysis % increase (95% CI) per
Comments** 50 ug/m3 PM10 Increase
% increase (95% CI) per
25 ug/m3 PM2 5 Increase
% increase (95% CI) per PM10, PM2 5 and PM10_2.5
25 ug/m3 PM10_2 5 Mean (Range) Levels
Increase Reported***
o
o
fe
H
6
o
o
H
O
Lung Function Changes (cont'd)
Neasetal. (1999)
Philadelphia, PA
(children)
GAMnotused amPEFR-8.17 (-14.81,
-1.56)
pmPEFR-1.44 (-7.33,
4.44)
Schwartz and Neas (2000) GAM not used
Uniontown, PA
(reanalysis)
(children)
Schwartz and Neas (2000) GAM not used
State College PA
(reanalysis)
(children)
am PEFR-3.29 (-6.64,
0.07)
pm PEFR-0.91 (-4.04,
2.21)
pm PEFR-1.52, (-2.80,
-0.24)
pm PEFR -0.93 (-1.8
0.01)
am PEFR -4.31 (-11.44, PM25 22.2 (IQR 16.2)
2.75)
pm PEFR 1.88 (-4.75,
8.44)
pm PEFR+1.73 (-2.2,
5.67)
PM10.259.5(IQR5.1)
PM2 5 24.5 (max 88.1)
PM10.25 NR
Vedaletal. (1998)
Port Alberni, CAN
(children)
GAM not used PEF -1.35 (-2.7, -0.05)
pmPEFR -0.28 (-3.45, PM25 23.5 (max 85.8)
2.87) PM10.2.5 NR
PM10 median 22.1(0.2,
159.0) (north site)
* Both original published studies and recent reanalyses reported in HEI (2003) Special Report for many cited here. Original studies published before 1996
and Schwartz et al. (1996) were assessed in 1996 PM AQCD.
** Where GAM not used in original analysis cited, original results reported here. Otherwise reanalyses results reported here if GAM (default) used in original
analysis. GAM strict = GAM with stringent criteria. GLM = general linear model; NS = natural splines; BS = B splines; PS = penalized splines.
*** Mean (minimum, maximum) 24-h PM level in parentheses unless otherwise noted.
o
HH
H
W
-------
1 colleagues (1995) reported significant associations between particle sulfate concentrations, as
2 well as other air pollutants, and daily cardiovascular admissions. The relative risk because of
3 sulfate particles was slightly larger for respiratory than for cardiovascular hospital admissions.
4 The 1996 PM AQCD concluded on the basis of these studies that, "There is a suggestion of a
5 relationship to heart disease, but the results are based on only two studies and the estimated
6 effects are smaller than those for other endpoints." The PM AQCD went on to state that acute
7 impacts on CVD admissions had been demonstrated for elderly populations (i.e., > 65), but that
8 insufficient data existed to assess relative impacts on younger populations.
9 Although the literature still remains relatively sparse, an important new body of data now
10 exists that both extends the available quantitative information on relationships between ambient
11 PM pollution and hospital CVD admissions, and that, more intriguingly, illuminates some of the
12 physiological changes that may occur on the mechanistic pathway leading from PM exposure to
13 adverse cardiac outcomes. Results of these new findings (including from GAM reanalyses) are
14 summarized in Table 9-10; and Figure 9-18 depicts excess risk estimates derived from several
15 studies of acute PM10 exposure effects on CVD admissions in U.S. cities. Although new studies
16 depicted in Figure 9-17 have reported generally consistent associations between daily
17 hospitalizations for cardiovascular disease and measures of PM, the data not only implicate PM,
18 but also CO and NO2 as well, possibly because of covarying of PM and these other gaseous
19 pollutants derived from common emission sources (e.g., motor vehicles). Taken as a whole, this
20 body of evidence suggests that PM is likely an important risk factor for cardiovascular
21 hospitalizations in the United States.
22 The NMMAPS study of PM10 concentrations and hospital admissions by persons 65 and
23 older in 14 U.S. cities provides particularly important findings of positive and significant
24 associations, even when concentrations are below 50 |ig/m3 (Samet et al., 2000a,b; and
25 reanalyses by Zanobetti and Schwartz, 2003b). As noted in Table 9-10 and Figure 9-18, this
26 study indicates PM10 CVD hospitalization effects similar to other cities, but with narrower
27 confidence bands, because of its greater power derived by combining multiple cities in the same
28 analysis. This allows significant associations to be identified, despite the fact that many of the
29 cities considered have relatively small populations and that each of the 14 cities had mean PM10
30 below 50 |ig/m3.
June 2003 9-101 DRAFT-DO NOT QUOTE OR CITE
-------
Zannobetti and Schwartz (2003)
14 US cities
Moolgavkar (2003)
LA,CA
Moolgavkar (2003)
Cook County
Linn et al. (2000)
LA.CA
Tolbert et al. (2000s)
Atlanta
Morris &Naumova
(1998) Chicago
Ito (2003)
Detroit
-10
Total CVD
CHF
HF
IHD —
10
15
20
25
Reconstructed Excess Risk Percentage
50 [jg/m3 Increase in PM10
Figure 9-18. Acute cardiovascular hospitalizations and PM exposure excess risk estimates
derived from selected U.S. PM10
CHF = congestive heart failure.
derived from selected U.S. PM10 studies. CVD = cardiovascular disease and
1
2
3
4
5
9
10
11
Several new studies have evaluated fine and coarse fraction particle effects on CVD
hospital admissions, with mixed overall results. That is, most all of the studies found positive
associations between PM2 5 or PM10_2 5 and increased CVD hospitalizations (Moolgavkar, 2000b;
reanalysis Moolgavkar, 2003; Tolbert et al., 2000a; Lippmann et al., 2000; reanalysis Ito, 2003;
Burnett et al., 1997; Stieb et al., 2000). Excess risks generally fell in the range of-3.0 to 8.0%
per 25 jig PM25 (24-h) increment; however, only a few were statistically significant at p < 0.05.
The PM10_2 5 CVD admissions results all showed positive associations as well, but the RR values
spanned a much wider range, from ~ 0.0% on up to -20% per 25 |ig/m3; and several were
statistically significant at p < 0.05. Thus, no clear evidence emerged for stronger associations
with fine versus coarse fraction short-term PM exposures.
June 2003
9-102
DRAFT-DO NOT QUOTE OR CITE
-------
1 Physiologic Measures of Cardiac Function. Several studies by independent groups of
2 investigators have also reported longitudinal associations between ambient PM concentrations
3 and physiologic measures of cardiovascular function. These studies measure outcomes and most
4 covariates at the individual level, making it possible to draw conclusions about individual risks,
5 as well as to explore mechanistic hypotheses. Such studies, for example, have reported temporal
6 associations between PM exposures and various electrocardiogram (ECG) measures of heart beat
7 or rhythm in panels of elderly subj ects. Reduced HR variability is a predictor of increased
8 cardiovascular morbidity and mortality risks. Three independent studies reported decreases in
9 HR variability associated with PM in elderly cohorts, although r-MSSD (one measure of
10 high-frequency HR variability) only showed elevations with PM in one study. Differences in
11 methods used and results across the studies argue for caution in drawing any strong conclusions
12 regarding PM effects from them, especially in light of the complex intercorrelations that exist
13 among measures of cardiac physiology, meteorology, and air pollution (Dockery et al., 1999).
14 Still, the new heart rhythm results, in general, comport well with other findings of cardiovascular
15 mortality and morbidity endpoints being associated with ambient PM. Chapter 5 discusses
16 available exposure studies of elderly subjects with CVD, such as the Sarnat et al. (2000)
17 Baltimore study. Less active groups tend to have lower exposure to nonambient PM because of
18 reduced personal activity. However, Williams et al. (2000a,b,c) report a very high pooled
19 correlation coefficient between PM25 personal exposure and outdoor concentrations. These
20 exposure studies tend to enhance the plausibility of panel study findings of impacts on HR
21 variability being caused by exposure to ambient-generated PM.
22
23 Changes in Blood Characteristics. Additional epidemiologic findings (Peters et al., 1997a)
24 also provide new evidence for ambient PM exposure effects on blood characteristics (e.g.,
25 increased c-reactive protein in blood) thought to be associated with increased risk of serious
26 cardiac outcomes (e.g., heart attacks).
27
28 Key Conclusions Regarding PM-CVD Morbidity
29 Overall, the newly available studies of PM-CVD relationships appear to support the
30 following conclusions regarding several key issues:
31
June 2003 9-103 DRAFT-DO NOT QUOTE OR CITE
-------
1 (1) Temporal Patterns of Response. The evidence from recent time series studies of CVD
admissions suggests rather strongly that PM effects are likely maximal at lag 0, with
some carryover to lag 1.
2 (2) Physical and Chemical Attributes Related to Particulate Matter Health Effects. The
characterization of ambient PM attributes associated with acute CVD is incomplete.
Insufficient data exist from the time series CVD hospital admissions literature or from
the emerging individual-level studies to provide clear guidance as to which PM
attributes, defined either on the basis of size or composition, determine potency. The
epidemiologic studies published to date have been constrained by the limited availability
of multiple PM metrics. Where multiple PM metrics exist, they often are of differential
quality because of differences in numbers of monitoring sites and in monitoring
frequency. Until more extensive and consistent data become available for epidemiologic
research, the question of PM size and composition, as they relate to acute CVD impacts,
will remain open.
3 (3) Susceptible Subpopulations. Because they lack data on individual subject
characteristics, ecologic time series studies provide only limited information on
susceptibility factors based on stratified analyses. The relative impact of PM on
cardiovascular (and respiratory) admissions reported in ecologic time series studies is
generally somewhat higher than those reported for total admissions. This provides some
support for the hypothesis that acute effects of PM operate via cardiopulmonary
pathways or that persons with preexisting cardiopulmonary disease have greater
susceptibility to PM, or both. Although there is some data from the ecologic time series
studies showing larger relative impacts of PM on cardiovascular admissions in adults
65 and over as compared with younger populations, the differences are neither striking
nor consistent. Some individual-level studies of cardiophysiologic function suggest that
elderly persons with preexisting cardiopulmonary disease are susceptible to subtle
changes in heart rate variability (HRV) in association with PM exposures. However,
because younger and healthier populations have not yet been assessed, it is not possible
to say at present whether the elderly have clearly increased susceptibility compared to
other groups, as indexed by cardiac pathophysiological indices such as HRV.
June 2003 9-104 DRAFT-DO NOT QUOTE OR CITE
-------
1 (4) Role of Other Environmental Factors. The ecologic time series morbidity studies
published since 1996 generally have controlled adequately for weather influences. Thus,
it is unlikely that residual confounding by weather accounts for the PM associations
observed. With one possible exception (Pope et al., 1999a), the roles of meteorological
factors have not been analyzed extensively as yet in the individual-level studies of
cardiac physiologic function. Thus, the possibility of confounding in such studies as yet
cannot be discounted totally or readily. Co-pollutants have been analyzed rather
extensively in many of the recent time series studies of hospital admissions and PM.
In some studies, PM clearly carries an independent association after controlling for
gaseous co-pollutants. In others, the "PM effects" are reduced markedly once
co-pollutants are added to the model. Among the gaseous criteria pollutants, CO has
emerged as the most consistently associated with cardiovascular (CVD) hospitalizations.
The CO effects are generally robust in the multi-pollutant model, sometimes as much so
as PM effects. However, the typically low levels of ambient CO concentrations in most
such studies and minimal expected consequent impacts on carboxyhemoglobin levels
and associated hypoxic effects thought to underlie CO CVD effects argue for the
likelihood that CO may be serving as a general surrogate for combustion products (e.g.,
PM) in the ambient pollution mix. See the most recent EPA CO Criteria Document
(U.S. Environmental Protection Agency, 2000a).
2
3 Respiratory Effects of Ambient Particulate Matter Exposures
4 The number of studies examining hospitalization and emergency department visits for
5 respiratory-related causes and other respiratory morbidity endpoints has increased markedly
6 since the 1996 PM AQCD. In addition to evaluating statistical relationships for PM10, quite a
7 few new studies also evaluated other PM metrics. Those providing estimates of increased risk in
8 U.S. and Canadian cities for respiratory-related morbidity measures (hospitalizations, respiratory
9 symptoms, etc.) in relation to 24-h increments in ambient fine particles (PM2 5) or coarse fraction
10 (PM10_25) of inhalable thoracic particles are included in Table 9-10.
11
12 Respiratory-Related Hospital Admission/Visits. Hospital admissions/ visit studies that
13 evaluated excess risks in relation to PM10 measures are still quite informative. Maximum excess
June 2003 9-105 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
risk estimates for PM10 associations with respiratory-related hospital admissions and visits in
U.S. cities are shown in Figure 9-19. Nearly all the studies showed positive, statistically
significant relationships between ambient PM10 and increased risk for respiratory-related
doctors' visits and hospital admissions. Overall, the results substantiate well ambient PM10
impacts on respiratory-related hospital admissions/visits. The excess risk estimates fall most
consistently in the range of 5 to 25.0% per 50 |ig/m3 PM10 increment, with those for asthma
hospital admissions and doctor's visits being higher than for COPD and pneumonia
hospitalization. Other, more limited, new evidence (not depicted in Figure 9-19) shows excess
risk estimates for overall respiratory-related or COPD hospital admissions falling mainly in the
range of-3.0 to 24% per 24-h 25 |ig/m3 increment in PM25 or PM10_25. Analogous estimates
were found for asthma admissions or physician visits, ranging up to ca. From -2.0 to 22.0% per
25 |ig/m3 24-h PM2 5 or PM10_2 5 increment.
Tolbert et al. (2000b)
Atlanta "
Choudhuryetal. (1997).
Anchorage
Sheppard (2003)
Seattle
Nauenberg and Basu (1999)
LA, CA
Zanobetti and Schwartz (2003)
14 U& Gities
Moolgavkar (2003)_
Chicago
Moolgavkar (2003).
LA
Ito f^OO1?!-
Detroit
Zanobetti and Schwartz (2003)
14 US Cities"
Ito (2003)
Detroit
1 Asthma Visits
Asthma Hospital Admission
I » i
COPD Hospital Admission
Pneumonia Hospital Admission
-10
-5
5 10
15 20 25 30 35
Figure 9-19. Maximum excess risk in selected studies of U.S. cities relating PM10 estimate
of exposure (50 ug/m3) to respiratory-related hospital admissions and visits.
June 2003
9-106
DRAFT-DO NOT QUOTE OR CITE
-------
1 Of particular note in Figure 9-19 are the large effect size estimates now being reported for
2 asthma hospitalizations and visits. Very importantly, these hospital admission/visit studies and
3 other new studies on respiratory symptoms and lung function decrements in asthmatics are
4 emerging as possibly indicative of ambient PM likely being a notable contributor to exacerbation
5 of asthma.
6
7 Pulmonary Function Changes and Respiratory Symptoms. Additional evidence for
8 PM-asthma effects is also emerging from panel studies of lung function and respiratory
9 symptoms. New panel studies of lung function and respiratory symptoms in asthmatic subjects
10 have been conducted by more than 10 research teams in various locations world-wide. As a
11 group, the studies examine health outcome effects that are similar, such as pulmonary peak flow
12 rate (PEFR); and the studies typically characterize the clinical-symptomatic aspects in a sample
13 of mild to moderate asthmatics (mainly children aged 5 to 16 yrs) observed in their natural
14 setting. Their asthma typically is being treated to keep them symptom free (with "normal"
15 pulmonary function rates, and activity levels) and to prevent recurrent exacerbations of asthma.
16 Severity of their asthma is characterized by symptom, pulmonary function, and medication use
17 and would be classified to include mild intermittent to mild persistent asthma suffers (National
18 Institutes of Health, 1997). As a group, they may thusly differ from asthmatics examined in
19 studies of hospitalization or doctor visits for acute asthmatic episodes, who may have more
20 severe asthma.
21 Most studies reported ambient PM10 results, but PM25 was examined in two studies. Other
22 ambient PM measures (BS and SO4) also were used. For these studies, mean PM10 levels range
23 from a low of 13 |ig/m3 in Finland to a high of 167 |ig/m3 in Mexico City. The Mexico City
24 level is over three times more than each of the other levels and is unique compared to the others.
25 Related 95% CI for these means or ranges show 1-day maximums above 100 |ig/m3 in four
26 studies, with two of these above 150 |ig/m3. Hence, these studies mainly evaluated different PM
27 metrics indexing PM concentrations in the range found in U.S. cities (see Chapter 3). All the
28 studies controlled for temperature, and several controlled for relative humidity.
29 Many panel studies are analyzed using a design that takes advantage of the repeated
30 measures on the same subject. Study subject number (N) varied from 12 to 164, with most
31 having N > 50; and all gathered adequate subject-day data to provide sufficient power for their
June 2003 9-107 DRAFT-DO NOT QUOTE OR CITE
-------
1 analyses. Linear models often are used for lung function and logistic models for dichotomous
2 outcomes. Meteorological variables are used as covariates; and medication use is also
3 sometimes evaluated as a dependent variable or treated as an important potential confounder.
4 However, perhaps the most critical choice in the model is selection of the lag for the pollution
5 variable. Presenting lag periods with only the strongest associations introduces potential bias,
6 because the biological basis for lag structure may be related to effect. No biological bases for
7 pertinent lag periods are known, but some hypotheses can be proposed. Acute asthmatic
8 reactions can occur 4 to 6 h after exposure and, thus, 0-day lag may be more appropriate than 1-
9 day lags for that acute reaction. Lag 1 may be more relevant for morning measurement of
10 asthma outcome from PM exposure the day before, and longer term lags (i.e., 2 to 5 days) may
11 represent the outcome of a more prolonged inflammatory mechanism; but too little information
12 is now available to predetermine appropriate lag(s).
13 Chapter 8 noted that people with asthma tend to have greater TB deposition than do
14 healthy people, but this data was not derived from the younger age group studied in most asthma
15 panel studies. The Peters et al. (1997b) study is unique for two reasons: (1) they studied the size
16 distribution of the particles in the range 0.01 to 2.5 |im and (2) examined the number of particles.
17 They reported that asthma-related health effects of 5-day means of the number of ultrafme
18 particles were larger than those of the mass of the fine particles. In contrast, Pekkanen et al.
19 (1997) also examined a range of PM sizes, but PM10 was more consistently associated with PEF.
20 Delfino et al. (1998) is unique in that they report larger effects for 1- and 8-h maximum PM10
21 than for the 24-h mean.
22 The results for the asthma panels of the peak flow analysis consistently show small
23 decrements for both PM10 and PM2 5. The effects using 2- to 5-day lags averaged about the same
24 as did the 0 to 1 day lags. Stronger relationships often were found with ozone. The analyses
25 were not able to clearly separate co-pollutant effects. The effects on respiratory symptoms in
26 asthmatics also tended to be positive. Most studies showed increases in cough, phlegm,
27 difficulty breathing, and bronchodilator use. The only endpoint more strongly related to longer
28 lag times was bronchodilator use, which was observed in three studies. The peak flow
29 decrements and respiratory symptoms are indicators for asthma episodes.
30 For PM10, nearly all of the point estimates for PEF showed decreases, but most were not
31 statistically significant, as shown in Figure 9-20 as an example of PEF outcomes. Lag 1 may be
June 2003 9-108 DRAFT-DO NOT QUOTE OR CITE
-------
Romieu etal. (1996)
(Mexico)
Pekkannen etal. (1997)
(Finland)
Gielen etal. (1997)
(Netherlands)
Romieu etal. (1997)
(Mexico)
-10 -505
Change in Pulmonary Function, L/min
Figure 9-20. Selected acute pulmonary function change studies of asthmatic children.
Effect of 50 ug/m3 PM10 on morning peak flow lagged 1 day.
1 more relevant for morning measurement of asthma outcome from the previous day. The figure
2 presents studies that provided this data. The results were consistent for both AM and PM peak
3 flow analyses. Similar results were found for the PM2 5 studies, although there were fewer
4 studies. Several studies included PM25 and PM10 independently in their analyses of peak flow.
5 Of these, Gold et al. (1999), Naeher et al. (1999), Tiittanen et al. (1999), Pekkanen et al. (1997),
6 and Romieu et al. (1996) all found similar results for PM2 5 and PM10. The study of Peters et al.
7 (1997b) found slightly larger effects for PM2 5. The study of Schwartz and Neas (2000) found
8 larger effects for PM2 5 than for PM10_2 5. Naeher et al. (1999) found that H+ was related
9 significantly to a decrease in morning PEF. Thus, there is no evidence here for a stronger effect
10 of PM2 5 when compared to PM10. Also, of studies that provided analyses that attempted to
11 separate out effects of PM10 and PM25 from other pollutants, Gold et al. (1999) studied possible
12 interactive effects of PM2 5 and ozone on PEF; they found independent effects of the two
13 pollutants, but the joint effect was slightly less than the sum of the independent effects.
14 The effects on respiratory symptoms in asthmatics also tended to be positive, although
15 much less consistent than the lung function effects. Most studies showed increases in cough,
June 2003
9-109
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
phlegm, difficulty breathing, and bronchodilator use (although generally not statistically
significant), as shown in Figure 9-21 for cough as an example. Three studies included both PM10
and PM2 5 in their analyses. The studies of Peters et al. (1997c) and Tiittanen et al. (1999) found
comparable effects for the two measures. Only the Romieu et al. (1996) found slightly larger
effects for PM25. These studies also give no good evidence for a stronger effect of PM25 when
compared to PM10.
Vedaletal. (1998)
(Canada)
Romieu etal. (1997)
(Mexico)
Gielenetal. (1997)
(Netherlands)
Peters et al. (1997c)
(Czech Republic)
2345
Odds Ratios for Cough
Figure 9-21. Odds ratios for cough for a 50-ug/m3 increase in PM10 for selected asthmatic
children studies, with lag 0 with 95% CL
1 Two asthma studies, both in the United States, examined PM indicators by 1 hr averages as
2 well as by 24 hr averages. The PM10 1 hr outcome was larger than the 24 hr outcome for lower
3 respiratory illness in one study (Delfino et al., 1998) but was lower for cough in the other study
4 (Ostro et al., 2001). Several of the studies reviewed above (Delfino et al., 1998, 2002; Ostro
5 et al., 2001; Yu et al., 2000; Mortimer et al., 2002; Vedal et al., 1998) that were conducted in the
6 United States and Canada found positive associations between various health endpoints for
7 asthmatics and ambient PM exposure (indexed by PM10, PM2 5, or PM10_2 5). The endpoints
June 2003
9-110
DRAFT-DO NOT QUOTE OR CITE
-------
1 included PEF decrements, various individual respiratory symptoms, and combinations of
2 respiratory symptoms. The various endpoints each represent effects on respiratory health.
3 The results of PM10 peak flow analyses for nonasthmatic populations were inconsistent.
4 Fewer studies reported results in the same manner as the asthmatic studies. Many of the point
5 estimates showed increases rather than decreases. PM2 5 studies found similar results. The
6 effects on respiratory symptoms in nonasthmatics were similar to those in asthmatics: most
7 studies showed that PM10 increases cough, phlegm, and difficulty breathing, but these increases
8 were generally not statistically significant. Schwartz and Neas (2000) found that PM10_2 5 was
9 significantly related to cough. Tiittanen et al. (1999) found that 1-day lag of PM10_2 5 was related
10 to morning PEF, but not evening PEF. Neas et al. (1999) found no association of PM10_2 5 with
11 PEF in non-asthmatic subjects.
12 The Schwartz and Neas (2000) reanalyses allows comparison of fine and coarse particle
13 effects on healthy school children using two pollutant models of fine and coarse PM. Coarse PM
14 was estimated by subtracting PM2 x from PM10 data. They report for cough (based on reanalysis
15 of the Harvard Six City Diary Study in the two PM pollutant model) PM25 OR = 1.07 (0.90,
16 1.26; per 15 |ig/m3 increment) and PM10_25 OR 1.18 (1.04, 1.34; per 8 |ig/m3 increment) in
17 contrast to lower respiratory symptom results of PM25 OR 1.29 (1.06, 1.57) and PM10_25 1.05
18 (0.9, 1.23). In the Uniontown reanalysis, peak flow for PM2 x for a 14 |ig/m3 increment was
19 -0.91 1/m (-1.14, -1.68) andPM10.21 for 15 |ig/m3 +1.04 1/m (-1.32, +3.4); for State College
20 PM21 -0.56 (-1.13, +0.01) and PM10.21 -0.17 (-2.07, +1.72).
21
22 9.8.2.1 Methodological Issues for Short-Term Exposure Studies
23 Chapter 8 discussed several still important methodological issues related to assessment of
24 the overall PM epidemiologic database. These include, especially, issues related to model
25 specifications and consequent adequacy of control for potentially confounding of PM effects by
26 co-pollutants, evaluations of possible source relationships to pollutant effects that may be useful
27 in better sorting out effects attributable to PM versus other co-pollutants or both, and other issues
28 such as lag structure. Key points are discussed concisely below.
29
30
31
June 2003 9-111 DRAFT-DO NOT QUOTE OR CITE
-------
1 Time Series Studies: Confounding by Co-Pollutants in Individual Cities
2 The co-pollutant issue was discussed at length in the 1996 document and still remains an
3 important issue. It must be recognized that there are large differences in concentrations of
4 measured gaseous co-pollutants (and presumably unmeasured pollutants as well) in different
5 parts of the United States, as well as the rest of the world; and the concentrations are often
6 correlated with concentrations of PM and its components because of commonality in source
7 emissions, wind speed and direction, atmospheric processes, and other human activities and
8 meteorological conditions. Large sources in the United States include motor vehicle emissions
9 (gasoline combustion, diesel fuel combustion, evaporation, particles generated by tire wear, etc.),
10 coal combustion, fuel oil combustion, industrial processes, residential wood burning, solid waste
11 combustion, and so on. Thus, one might reasonably expect some large correlations among PM
12 and co-pollutants, but possibly with substantial differences in relation by season in different
13 cities or regions. Statistical theory suggests that PM and co-pollutant effect size estimates will
14 be highly unstable and often insignificant in multi-pollutant models when collinearity exists.
15 Many recent studies demonstrate this effect, for both hospital admissions (Moolgavkar, 2000b)
16 and mortality (Moolgavkar, 2000a; Chock et al., 2000). Because the problem seems largely
17 insoluble in studies in single cities, the new multi-city studies (Samet et al., 2000a,b; Schwartz,
18 1999; Schwartz and Zanobetti, 2000) have provided important new insights. See discussions of
19 NMMAPS analysis in Chapter 8 and below for discussion of issues related to control for
20 co-pollutant effects. Overall, although such issues may warrant further evaluation, it now
21 appears unlikely that such confounding accounts for the vast array of effects attributed to
22 ambient PM based on the rapidly expanding PM epidemiology database.
23 Numerous new studies have reported associations not only between PM, but also gaseous
24 pollutants (O3, SO2, NO2, and CO), and mortality. In many of these studies, simultaneous
25 inclusion of one or more gaseous pollutants in regression models did not markedly affect PM
26 effect size estimates, as was generally the case in the NMMAPS analyses for 90 cities (see
27 Figure 9-22). On the other hand, some studies reporting positive and statistically significant
28 effects for gaseous co-pollutants (e.g., O3, NO2, SO2, CO) found varying degrees of robustness of
29 their effects estimates or those of PM in multi-pollutant models as discussed in Chapter 8
30 (Section 8.4). Thus, although it is likely that there are independent health effects of PM and
31 gaseous pollutants, there is not yet sufficient evidence by which to confidently separate out fully
June 2003 9-112 DRAFT-DO NOT QUOTE OR CITE
-------
I
PM10
PM10 + 03
PM10 + 03
PM10
PM10
N02
03+S02
03 + CO
— — — .
_ . _ . .
_.._.._
H nn
I .UU
.UU
1.00
1.00
1.00
r
0.0 0.2 0.4 0.6 0.8
% Change in Mortality per 10 |jg/m3 Increase in PM10
1.0
Figure 9-22. Marginal posterior distributions for effect of PM10 on total mortality at lag 1,
with and without control for other pollutants, for the NMMAPS 90 cities.
The numbers in the upper right legend are the posterior probabilities that
the overall effects are greater than 0.
Source: Dominici et al. (2003).
1 the relative contributions of PM versus those of other gaseous pollutants or by which to
2 quantitate modifications of PM effects by other co-pollutants, including possible synergistic
3 interactions that may vary seasonally or from location to location. Overall, it appears, however,
4 that ambient PM and O3 can be most clearly separated out as likely having independent effects,
5 their concentrations often not being highly correlated. More difficulty is encountered, at times,
6 in sorting out whether NO2, CO, or SO2 are exerting independent effects in cities where they tend
7 to be highly correlated with ambient PM concentrations, possibly because of derivation of
8 important PM constituents from the same source (e.g., NO2, CO, and PM from mobile sources)
June 2003
9-113
DRAFT-DO NOT QUOTE OR CITE
-------
1 and/or a gaseous pollutant (e.g., SO2) serving as a precursor for a significant PM component
2 (e.g., sulfate).
3 Other information discussed in Section 8.4 on conceptual frameworks for evaluating
4 possible confounding makes it clear that diagnostic evaluations of inflation or deflation of PM
5 effect size estimates by addition of gaseous co-pollutants into multiple pollutant models, at best,
6 may indicate potential confounding of PM effects in a given analysis. Other independently-
7 derived exposure analyses, i.e., Sarnat et al. (2000, 2001), however, strongly suggest a very low
8 probability of observed PM effects being due to confounding with gaseous criteria pollutants
9 (CO, NO2, SO2, O3) having high correlations with important PM constituents from the same
10 source (e.g., NO2, CO, and PM from mobile sources) or for gaseous pollutants (e.g., SO2)
11 serving as a precursor for a significant PM component (e.g., sulfate).
12
13 Time Series Studies: Model Selection for Lags, Moving Averages, and Distributed Lags
14 A number of different approaches have been used to evaluate the temporal dependence of
15 mortality or morbidity on time-lagged PM concentrations, including unweighted moving
16 averages of PM concentrations over one or more days, general weighted moving averages, and
17 polynomial distributed moving averages. Unless there are nearly complete daily data, each
18 different lag will be using a different set of mortality data corresponding to spaced PM
19 measurement; for example, for lag 0 with every-sixth-day PM measurements, the mortality data
20 are on the same day as the PM data, for lag 1 the mortality data are on the next day after the PM
21 data, and so on. Although this effect is likely to be small, it should nonetheless be kept in mind.
22 The distributed lag models used in the NMMAPS II morbidity studies are a noteworthy
23 methodological advance. The fitted distributed lag models showed significant heterogeneity
24 across cities for COPD and pneumonia, however (see Table 15 therein), again raising the
25 question of how heterogeneous effects can best be combined so as not to obscure potentially real
26 city-specific or region-specific differences.
27 Only three cities with nearly complete daily PM10 data were used to evaluate more general
28 multi-day lag models (Chicago, Minneapolis/St. Paul, Pittsburgh), and these show somewhat
29 different patterns of effect, with lag 0 < lag 1 and lag 1 » lag 2 for Chicago, lag 0 = lag 1 > lag
30 2 for Minneapolis, and lag 0 < lag 1 = lag 2 for Pittsburgh. The 7-day distributed lag model is
June 2003 9-114 DRAFT-DO NOT QUOTE OR CITE
-------
1 significant for Pittsburgh, but less so in the other cities. The remaining data are limited
2 intrinsically in what they can reveal about temporal structure.
3
4 Time Series Studies: Model Selection for Concentration-Response Functions
5 Given the number of analyses that needed to be performed, it is not surprising that most of
6 the NMMAPS studies focused on linear concentration-response models. More recent studies
7 (Daniels et al., 2000) for the 20 largest U.S. cities have found posterior mean effects of 2 to 2.7%
8 excess risk of total daily mortality per 50 |ig/m3 24-h PM10 at lags 0, 1, 0+1 days; 2.4 to 3.5%
9 excess risk of cardiovascular and respiratory mortality; and 1.2 to 1.7% for other causes of
10 mortality. The posterior 95% credible regions are all significantly greater than 0. However, the
11 threshold models gave distinctly different estimates of 95% credible regions for the threshold for
12 total mortality (15 |ig/m3 at lag 1, range 10 to 20), cardiovascular and respiratory mortality
13 (15 |ig/m3 at lag 0+1, range 0 to 20), and other causes of mortality (65 |ig/m3 at lag 0+1, range
14 50 to 75 |ig/m3).
15 Another problem is that the shape of the relationship between mortality and PM10 may
16 depend, to some extent, on the associations of PM10 with gaseous co-pollutants. The association
17 is not necessarily linear, and is indeed likely to have both seasonal and secular components that
18 depend on the city location. Thus, further elaborations of these models is desirable.
19
20 Effects of Exposure Error in Daily Time Series Epidemiology
21 There has been considerable controversy over how to deal with the nonambient component
22 of personal exposure. Recent biostatistical analyses of exposure error have indicated that the
23 nonambient component will not bias the statistically calculated risk in community time-series
24 epidemiology, provided that the nonambient component of personal exposure is independent of
25 the ambient concentration. Consideration of the random nature of nonambient sources and
26 recent studies, in which estimates of cc, ambient-generated PM divided by ambient PM
27 concentrations, have been used to estimate separately the ambient-generated and nonambient
28 components of personal exposure, support the assumption that the nonambient exposure is
29 independent of the ambient concentration. Therefore, it is reasonable to conclude that
30 community time series epidemiology describes statistical associations between health effects and
June 2003 9-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposure to ambient-generated PM, but does not provide any information on possible health
2 effects resulting from exposure to nonambient PM (e.g., indoor-generated PM).
3 From the point of view of exposure error, it is also significant to note that, although
4 ambient concentrations of a number of gaseous pollutants (O3, NO2, SO2) often are found to be
5 highly correlated with various PM parameters, personal exposures to these gases are not
6 correlated highly with personal exposure to PM indicators. The correlations of the ambient
7 concentrations of these gases also are not correlated highly with the personal exposure to these
8 gases. Therefore, when significant statistical associations are found between these gases and
9 health effects, it could be that these gases may, at times, be serving as surrogates for PM rather
10 than being causal themselves. Pertinent information on CO has not been reported.
11 The attenuation factor, a, is a useful variable. For relatively constant a, the risk because of
12 a personal exposure to 10 |ig/m3 of ambient PM is equal I/a times the risk from a concentration
13 of 10 |ig/m3 of ambient PM, where a varies from a low of 0.1 to 0.2 to a maximum of 1.0. (The
14 health risk for an interquartile change in ambient concentration of PM is the same as that for an
15 interquartile change in exposure to ambient PM). Differences in a among cities, reflecting
16 differences in air-exchange rates (e.g., because of variation in seasonal temperatures and in
17 extent of use of air conditioners) and differences in indoor/outdoor time ratios, may, in part,
18 account for any differences in risk estimates based on statical associations between ambient
19 concentrations and health effects for different cities or regions. If a were 0.3 in city A, but 0.6 in
20 city B, and the risks for an increase in personal exposure of 10 |ig/m3 were identical, then a
21 regression of health effects on ambient concentrations would yield a health risk for city B that
22 would be twice that obtained for city A.
23 A number of exposure analysts have discussed the PM exposure paradox (i.e., that
24 epidemiology yields statistically significant associations between ambient concentrations and
25 health effects even though there is a near zero correlation between ambient concentrations and
26 personal exposure in many studies). Several explanations have been advanced to resolve this
27 paradox. First, personal exposure contains both an ambient-generated and a nonambient
28 component. Community time series epidemiology yields information only on the ambient-
29 generated component of exposure. Therefore, the appropriate correlation to investigate is the
30 correlation between ambient concentration and personal exposure to ambient-generated PM, not
31 between ambient concentrations and total personal exposure (i.e., the sum of ambient-generated
June 2003 9-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 and nonambient PM). Second, biostatistical analysis of exposure error indicates that if the risk
2 function is linear in the PM indicator, the average of the sum of the individual risks (risk
3 function times individual exposure) may be replaced by the risk function times the community
4 average exposure. Thus, the appropriate correlation (of ambient concentrations and ambient-
5 generated exposure) is not the pooled correlation of different days and different people but the
6 correlation between the daily ambient concentrations and the community average daily personal
7 exposure to ambient-generated PM. Because the nonambient component is not a function of the
8 ambient concentration, its average will tend to be similar each day. Therefore, the correlation
9 coefficient will depend on a but not on the nonambient exposure. These types of correlation
10 yield high correlation coefficients.
11 A few studies have conducted simulation analyses of effects of measurement errors on the
12 estimated PM mortality effects. These studies suggest that ambient PM excess risk effects are
13 more likely underestimated than overestimated, and that spurious PM effects (i.e., qualitative
14 bias such as change in the sign of the coefficient) because of transferring of effects from other
15 covariates require extreme conditions and are therefore very unlikely. The error because the
16 difference between the average personal exposure and the ambient concentration is likely the
17 major source of bias in the estimated relative risk. One study also suggested that apparent linear
18 exposure-response curves are unlikely to be artifacts of measurement error.
19 In conclusion, for time-series epidemiology, ambient concentration is a useful surrogate for
20 personal exposure to ambient-generated PM, although the risk per unit ambient PM
21 concentration is biased low by the factor a, compared to the risk per unit exposure to ambient-
22 generated PM. Epidemiologic studies of statistical associations between long-term effects and
23 long term ambient concentrations compare health outcome rates across cities with different
24 ambient concentrations. Ordinarily, PM exposure measurement errors are not expected to
25 influence the interpretation of findings from either the community time-series or long-term
26 epidemiologic studies that have used ambient concentration data if they include sufficient
27 adjustments for seasonality and key personal and geographic confounders. When individual
28 level health outcomes are measured in small cohorts, to reduce exposure misclassification errors,
29 it is essential that better real-time exposure monitoring techniques be used and that further
30 speciation of indoor-generated, ambient, and personal PM mass be accomplished. This should
31 enable measurement (or estimation) of both ambient and nonambient components of personal
June 2003 9-117 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposure and evaluation of the extent to which personal exposure to ambient-generated PM,
2 personal exposure to nonambient PM, or total personal exposure (to ambient-generated plus
3 nonambient PM) contribute to observed health effects.
4
5 9.8.3 Health Effects of Long-Term Exposures to Particulate Matter
6 The health effects of long-term ambient PM exposures have been epidemiologically
7 studied in recent years mainly by prospective cohort studies that offer advantages over purely
8 ecological analyses. Prospective cohort studies of ambient air pollutants are methodologically
9 similar to typical epidemiologic studies of occupational cohorts and, in some respects, to
10 experimental trials. Subjects are enrolled, characterized as to their exposures and other relevant
11 health factors, and followed over time as they experience adverse health outcomes.
12 Methodological issues regarding the loss of subjects to follow-up, the movement of subjects
13 between exposure groups or levels, and the characterization of exposure are well-understood and
14 are adequately handled by standard epidemiologic methods.
15 The assignment of exposure in both environmental and occupational studies is generally
16 based on area rather than personal sampling and any consequential exposure misclassification
17 will generally bias effect estimates towards the null. With appropriate individual-level
18 assessment and analysis of other risk factors, the assignment of a common exposure to a group
19 does not give raise to an ecological fallacy (Kunzli and Tager, 1997). This PM AQCD has
20 avoided a reliance on purely ecological analyses of county-level data that lack individual-level
21 data on non-environmental determinants of mortality.
22
23 Updated Epidemiologic Findings for Long-Term Particulate Matter Exposure
24 Effects on Mortality
25 The 1996 PM AQCD indicated that past epidemiologic studies of chronic PM exposures
26 collectively indicate increases in mortality to be associated with long-term exposure to airborne
27 particles of ambient origins (see appendix Table 9A-3). The PM effect size estimates for total
28 mortality from these studies also indicated that a substantial portion of these deaths reflected
29 cumulative PM impacts above and beyond those exerted by acute exposure events. Table 9-11
30 shows long-term exposure effects estimates (RR values) per variable increments in ambient PM
31 indicators in U.S. and Canadian cities, including results from newer analyses since the 1996 PM
32 AQCD.
June 2003 9-118 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-11. EFFECT ESTIMATES PER INCREMENTS* IN LONG-TERM MEAN
LEVELS OF FINE AND COARSE FRACTION PARTICLE INDICATORS FROM
U.S. AND CANADIAN STUDIES
Type of Health Effect
Study and Location
Increased Total Mortality
in Adults
Six CityA
ACS Study8
(151 U.S. SMSA)
Six City Reanalysis0
ACS Study Reanalysis0
ACS Study Extended
Analyses0
Southern CaliforniaE
Veterans CohortF
Increased
Cardiopulmonary
Mortality in Adults
Six CityA
Six City Reanalysis0
ACS Study8
(151 U.S. SMSA)
ACS Study Reanalysis0
Southern CaliforniaE
Indicator
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
SOJ (15 Ug/m3)
PM25 (10 ug/m3)
SOI (15 ug/m3)
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3)
(dichot)
PM25 (10 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM10 (cutoff =
30 days/year
>100 ug/m3)
PM10 (20 ug/m3)
PM10 (cutoff =
30 days/year
>100 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3)
(dichot)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
Change in Health Indicator per
Increment in PM
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)
1.14(1.05-1.23)
1.04(1.01, 1.07)
1.07(1.04-1.10)
1.04(1.01-1.08)
1.091 (0.985-1.212; males)
1.082 (1.008-1. 162; males)
0.950 (0.873-1.033; females)
0.958 (0.899-1.021; females)
0.90 (0.85, 0.954; males)
Relative Risk (95% CI)
***
1.18(1.06, 1.32)
1.20(1.29, 1.41)
1.19(1.07, 1.33)
1.12(1.07-1.17)
1.07(1.03, 1.12)
1.12(1.07-1.17)
1.01(0.92, 1.10)
Range of City
PM Levels**
Means (jig/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)
5.6-42.3
18-47
11-30
18.2-46.5
11.0-29.6
9-34
58.7(34-101)
9.0-33.4
51 (±17)
June 2003
9-119
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-11 (cont'd).
MEAN LEVELS OF
EFFECT ESTIMATES PER INCREMENTS* IN LONG-TERM
FINE AND COARSE FRACTION PARTICLE INDICATORS
FROM U.S. AND CANADIAN STUDIES
Type of Health Effect
Study and Location
Increased Bronchitis in
Children
Six City0
24 City11
24 CityH
24 City11
Southern California1
12 Southern California
communities1
(all children)
12 Southern California
communitiesK
(children with asthma)
Increased Cough in
Children
12 Southern California
communities1
(all children)
12 Southern California
communitiesK
(children with asthma)
10 Canadian CommunitiesL
Increased Wheeze in
Children
10 Canadian
CommunitiesM
Increased Airway
Obstruction in Adults
Southern CaliforniaM
Decreased Lung Function
in Children
Six City0
24 CityN
24 CityN
24 CityN
Indicator
PM15/10 (50 ug/m3)
SOI (15 ug/m3)
PM2 ! (10 ug/m3)
PM10 (20 ug/m3)
SOI (15 ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
PM15/10 (50 ug/m3)
PM2 , (10 ug/m3)
SOI (7 ug/m3)
PM10 (20 ug/m3)
Change in Health Indicator per
Increment in PM
Odds Ratio (95% CI)
3.26(1.13, 10.28)
3.02 (1.28, 7.03)
1.31(0.94, 1.84)
1.60 (0.92, 2.78)
1.39 (0.99, 1.92)
0.95(0.79, 1.15)
1.4(1.1, 1.8)
1.3 (0.9, 1.7)
Odds Ratio (95% CI)
1.05(0.94, 1.16)
1.1 (0.7, 1.8)
1.2 (0.8, 1.8)
1.19(1.04, 1.35)
Odds Ratio (95% CI)
1.35(1.10, 1.64)
Odds Ratio (95% CI)
1.19(0.84, 1.68)
Odds Ratio (95% CI)
NS Changes
-2.15% (-3. 34, -0.95) FVC
-3.06% (-4.50, -1.60) FVC
-2.80% (-4.97, -0.59) FVC
Range of City
PM Levels**
Means (jig/m3)
20-59
18.1-67.3
9.1-17.3
22.0-28.6
—
28.0-84.9
13.0-70.7
6.7-31.5
28.0-84.9
13.0-70.7
6.7-31.5
13-23
13-23
NR
20-59
18.1-67.3
9.1-17.3
22.0-28.6
June 2003
9-120
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-11 (cont'd).
MEAN LEVELS OF
Type of Health Effect
Study and Location
EFFECT ESTIMATES PER INCREMENTS* IN LONG-TERM
FINE AND COARSE FRACTION PARTICLE INDICATORS
FROM U.S. AND CANADIAN STUDIES
Range of City
PM Levels**
Indicator Increment in PM Means (jig/m3)
Change in Health Indicator per
Increment in PM
Decreased Lung Function
in Children (cont'd)
12 Southern California
,o
communities
(all children)
12 Southern California
communities1
(all children)
o
12 Southern California
communities1"
(4th grade cohort)
12 Southern California
communities1"
(4th grade cohort)
12 Southern California
communities'3
(4th grade cohort follow-up)
12 Southern California
communities'3
(4th grade cohort follow-up)
12 Southern California
communities'3
(4th grade cohort follow-up)
Southern CaliforniaR
Southern CaliforniaR
Southern CaliforniaR
PM10 (20 ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
PM25(10ug/m3)
PM10.25 (10 ug/m3)
PM10 (20 ug/m3)
PM25(10ug/m3)
PM10.25 (10 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
-19.9 (-37.8, -2.6) FVC 28.0-84.9
-25.6(-47.1,-5.1)MMEF 28.0-84.9
-0.23 (-0.44, -0.01) FVC % growth NR
-0.18 (-0.36, 0.0) FVC % growth
-0.22 (-0.47, 0.02) FVC % growth
-0.51 (-0.94, -0.08) MMEF % growth NR
-0.4 (-0.75, -0.04) MMEF % growth
-0.54 (-1.0, -0.06) MMEF % growth
-0.23 (-0.46,-0.0) FVC % growth NR
-0.19 (-0.39, 0.01) FVC % growth
-0.55 (-1.0, -0.08) MMEF % growth NR
-0.42 (-0.85, 0.01) MMEF % growth
-0.49 (-0.84, -0.14) PEFR % growth NR
-0.37 (-0.70, -0.04) PEFR % growth
-3.6 (-18, 11) FVC growth 15.0-66.2
-33 (-64, -2.2) MMEF growth 15.0-66.2
-70 (-120, -20) PEFR growth 15.0-66.2
Lung Function Changes
in Adults
Odds Ratio (95% CI)
Southern California8
(% predicted FEVb
females)
Southern California8
(% predicted FEVb males)
Southern California8
(% predicted FEVb males
whose parents had asthma,
bronchitis, emphysema)
PM10 (cutoff of
54.2 days/year
>100 ug/m3)
PM10 (cutoff of
54.2 days/year
>100 ug/m3)
PM10 (cutoff of
54.2 days/year
>100 ug/m3)
+0.9 % (-0.8, 2.5) FEVj
+0.3 % (-2.2, 2.8) FEVj
-7.2 % (-11.5,-2.7) FEVj
52.7 (21.3, 80.6)
54.1(20.0,80.6)
54.1(20.0,80.6)
June 2003
9-121
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-11 (cont'd). EFFECT ESTIMATES PER INCREMENTS* IN LONG-TERM
MEAN LEVELS OF FINE AND COARSE FRACTION PARTICLE INDICATORS
FROM U.S. AND CANADIAN STUDIES
Range of City
Type of Health Effect Change in Health Indicator per PM Levels **
Study and Location Indicator Increment in PM Means (jig/m3)
Lung Function Changes
in Adults (cont'd)
Southern California8
(% predicted FEVb males)
S0l(1.6ug/m3)
-1.5% (-2.9, -O.rtFEVj 7.3(2.0, 10.1)
* Results calculated using PM increment between the high and low levels in cities, or other PM increments
given in parentheses; NS Changes = No significant changes.
** Range of mean PM levels given unless, as indicated, studies reported overall study mean (min, max), or
mean (±SD); NR=not reported.
*** Results only for smoking category subgroups.
References:
ADockery et al. (1993) KMcConnell et al. (1999)
BPope et al. (1995) LHowel et al. (2001)
GKrewski et al. (2000) MBerglund et al. (1999)
DPope et al. (2002) NRaizenne et al. (1996)
EAbbey et al. (1999) °Peters et al. (1999c)
FLipfert et al. (2000b) pGauderman et al. (2000)
HDockery et al. (1989) QGauderman et al. (2002)
HDockery et al. (1996) RAvol et al. (2001)
'Abbey et al. (1995a,b,c) sAbbey et al. (1998)
JPetersetal. (1999b)
1 Several advances have been made in terms of further analyses and/or reanalyses of several
2 studies of long-term PM exposure effects on total, cardiopulmonary, or lung cancer mortality.
3 These include reanalyses by Krewski et al. (2000) of the Harvard Six-Cities Study originally
4 reported by Dockery et al. (1993); reanalyses by Krewski et al. (2000) of America Cancer
5 Society (ACS) Study data and analyses originally reported on by Pope et al. (1995); extended
6 analyses of ACS data covering 16 more years of follow-up (Pope et al., 2000), new analysis of
7 extended years of data from the Adventist Health Study of SMOG (AHSMOG) reported by
8 Abbey et al. (1999) and McConnell et al. (2000); for Southern California residents; and a newly
9 available Veterans Administration (VA) study of U.S. veterans published by Lipfert et al.
10 (2000b). Table 9-11 includes key results (excess relative risks for total, cardiopulmonary, and
11 lung cancer mortality associated with long-term ambient PM exposure) from these studies.
12 Two of these survival studies were national in scope, the Harvard Six-Cities Study
13 (Dockery et al., 1993) and the American Cancer Society (ACS) Study (Pope et al., 1995), and
June 2003 9-122 DRAFT-DO NOT QUOTE OR CITE
-------
1 one focused solely on California, the Adventist Health Study of Smog or AHSMOG (Abbey et
2 al., 1991). The ACS was a secondary analysis of a very extensive cohort of 552,138 subjects in
3 151 cities whose exposures were characterized by routinely collected air quality data and who
4 were followed for seven years. The Harvard Six-Cities Study enrolled 8,111 subjects in six
5 cities, characterized their exposures with investigator-conducted measurements of
6 size-fractionated paniculate matter, and followed these subjects for 14 to 16 years. The
7 AHSMOG Study enrolled 6,340 non-smoking subjects, grouped into three major urban areas and
8 the remainder of California, whose exposures were characterized by routinely collected air
9 quality data, and who were followed for an average of 10 years. The VA cohort study (Lipfert
10 et al., 2000b) enrolled -70,000 veterans who, at the time of enrollment, had high blood pressure
11 and followed them health-wise for twenty years. Air quality data for their place of residence at
12 time of enrollment was tracked from time periods prior to and during the study and related to
13 mortality events.
14 One of the most important advances since the 1996 PM AQCD is the substantial
15 verification and extension of the findings of the Harvard Six City prospective cohort study
16 (Dockery et al., 1993) and the cohort study relating American Cancer Society (ACS) health data
17 to fine-particle data from 50 cities and sulfate data from 151 cities (Pope et al., 1995). The
18 reanalyses, sponsored by the Health Effects Institute (HEI), included a data audit, replication of
19 the original investigators' findings, and additional analyses to explore the sensitivity of the
20 original findings to other model specifications. The investigators of the HEI Reanalysis Project
21 (Krewski et al., 2000) first performed a data audit, using random samples to verify the accuracy
22 of the data sets used in the original Six City analyses, including death certificate data, air
23 pollution data, and socioeconomic data. In general, the air pollution data were reproducible and
24 correlated highly with the original aerometric data in Pope et al. (1995).
25 The reanalyses substantially verified the findings of the original investigators, with PM2 5
26 or sulfate relative risk (RR) estimates for total mortality and for cardiopulmonary mortality
27 differing at most by ±0.02 (±2% excess risk) from the least polluted to the most polluted cities in
28 the study. A larger difference was noted for the PM2 5 lung cancer relative risk in the Six Cities
29 study, 1.37 originally and 1.43 in the reanalysis, neither estimate being statistically significant.
30 The sensitivity analyses for the Six Cities study found generally similar results with other
31 individual covariates included. The time-dependent covariate model for total mortality (taking
June 2003 9-123 DRAFT-DO NOT QUOTE OR CITE
-------
1 into account higher postexposures in early years of the study and changes over time to the last
2 years of the study) had a substantially lower RR than the model without time-dependent
3 covariates. Educational level made a large difference, with individuals having less than a high
4 school education at much greater risk for mortality than those with any postsecondary education.
5 Among the ecological covariates, sulfates adjusted for artifact had little effect on the risk
6 estimates for total mortality compared to that without adjustment, but, in the ACS study, the
7 filter adjustment actually increased the relative risk for all causes and cardiopulmonary
8 mortality, while substantially reducing the estimated sulfate effect on lung cancer. Inclusion of
9 SO2 as an additional ecological covariate greatly reduced the estimated PM2 5 and sulfate effects
10 in the ACS study, whereas a spatial model including SO2 effects caused only a modest reduction
11 of the estimated PM2 5 and sulfate effects. However, the SO2 effects were reduced greatly when
12 sulfates were included in the model. Sulfur dioxide and sulfates often are highly correlated,
13 because of the formation of secondary sulfates.
14 Many model selection issues in the prospective cohort studies are analogous to those in the
15 time series analyses. One issue of particular concern is whether the exposure indices used in the
16 analyses adequately characterize the exposure of the participants in the study during the months
17 or years preceding death. This question is particularly conspicuous in regard to the Pope et al.
18 (1995) study, in which PM25 and sulfate data were collected in the 1979 to 1982 period from the
19 EPA AIRS database and the Inhalable Particle Network, largely preceding the collection of the
20 ACS cohort data by only a few years, and so possibly not adequately reflecting exposure to
21 presumably much higher PM concentrations occurring long before the cohort was recruited, nor
22 exposure to presumably lower concentrations during the study. This issue was raised in the 1996
23 PM AQCD. However, the Six Cities Study did have air pollution data and repeated survey data
24 over time, with PM2 5 and sulfate data measured every other day and sometimes daily, and so the
25 new investigators were able to use the information about time-dependent cumulative PM
26 concentrations during the course of the study. Changes in smoking status and body mass index
27 over the 10 to!2 years of the study had little effect on risk estimates, but taking into account the
28 decrease in particle concentrations from the earlier years to the later years reduced the effect size
29 estimate substantially, although it remained statistically significant. Nevertheless, overall, the
30 reanalyses of the ACS and Harvard Six-Cities studies (Krewski et al., 2000) "replicated the
31 original results, and tested those results against alternative risk models and analytic approaches
June 2003 9-124 DRAFT-DO NOT QUOTE OR CITE
-------
1 without substantively altering the original findings of an association between indicators of
2 participate matter air pollution and mortality."
3 The shape of the relationship of concentration to mortality also was explored. Preliminary
4 findings suggest some possible nonlinearity, but further study is needed. Among the most
5 important new findings of the study are spatial relationships between mortality and air pollution,
6 discussed later below.
7 Recently reported extension of the ACS analyses (Pope et al., 2002) to include additional
8 years of data provides further substantiation of originally reported findings for total, respiratory,
9 and cardiovascular mortality. Also of great importance, these new analyses provide much
10 stronger evidence substantiating links between long-term ambient fine PM exposures and lung
11 cancer. This is consistent with findings of increased lung cancer risk being associated with
12 exposure with diesel exhaust particles, an important constituent of PM25 in many U.S. urban
13 areas.
14 With regard to the role of various PM constituents in the PM-mortality association, past
15 cross-sectional studies generally have found that the fine particle component, as indicated either
16 by PM2 5 or sulfates, was the PM constituent most consistently associated with chronic PM
17 exposure-mortality. Although the relative measurement errors of the various PM constituents
18 must be further evaluated as a possible source of bias in these estimate comparisons, the Harvard
19 Six-Cities study and the latest reported AHSMOG prospective semi-individual study results
20 (Abbey, et al., 1999; McDonnell et al., 2000) are both indicative of the fine mass components of
21 PM likely being associated more strongly with the mortality effects of PM than coarse PM
22 components. The ACS study, its reanalyses, and its recent extension all further substantiate
23 ambient fine particle effects, including increased risk not only of cardiopulmonary-related
24 mortality but lung cancer mortality as well.
25 The Harvard Six Cities analyses (as confirmed by the HEI reanalyses) and the recent
26 extension of the ACS study by Pope et al. (2002) probably provide the most credible and precise
27 estimates of excess mortality risk associated with long-term PM exposures in the United States.
28 Of particular interest are their statistically significant effects estimates for PM2 5, falling in a
29 range of 4.0 to 14.0% total mortality per 10 |ig/m3 annual average increment ,and the 10-46%
30 excess risk per 15 |ig/m3 increase in long-term sulfate concentrations.
June 2003 9-125 DRAFT-DO NOT QUOTE OR CITE
-------
1 Several other new studies report epidemiologic evidence indicating that: (a) PM exposure
2 early in pregnancy (during the first month) may be associated with slowed intrauterine growth
3 leading to low birth weight events (Dejmek et al., 1999); and (b) early postnatal PM exposures
4 may lead to increased infant mortality (Woodruff et al., 1997; Boback and Leon, 1999; Loomis
5 et al., 1999; Lipfert et al., 2000b).
6
7 Long-Term Particulate Matter Exposure Effects on Lung Function and
8 Respiratory Symptoms
9 In the 1996 PM AQCD, the available respiratory disease studies were limited in terms of
10 conclusions that could be drawn. At that time, three studies based on a similar type of
11 questionnaire administered at three different times as part of the Harvard Six-City and 24-City
12 Studies provided data on the relationship of chronic respiratory disease to PM. All three studies
13 suggest a chronic PM exposure effect on respiratory disease. The analysis of chronic cough,
14 chest illness, and bronchitis tended to be significantly positive for the earlier surveys described
15 by Ware et al. (1986) and Dockery et al. (1989). Using a design similar to the earlier one,
16 Dockery et al. (1996) expanded the analyses to include 24 communities in the United States and
17 Canada. Bronchitis was found to be higher (odds ratio = 1.66) in the community with highest
18 exposure of strongly acidic particles when compared with the least polluted community. Fine
19 ZPM sulfate was also associated with higher reporting of bronchitis (OR = 1.65, 95% CI 1.12,
20 2.42).
21 The studies by Ware et al. (1986), Dockery et al. (1989), and Neas et al. (1994) all had
22 good monitoring data and well-conducted standardized pulmonary function testing over many
23 years, but showed no effect on children of PM pollution indexed by TSP, PM15, PM25, or
24 sulfates. In contrast, the later 24-city analyses reported by Raizenne et al. (1996) found
25 significant associations of effects on FEVj or FVC in U.S. and Canadian children with both
26 acidic particles and other PM indicators. Overall, the available studies provided limited
27 evidence suggestive of pulmonary lung function decrements being associated with chronic
28 exposure to PM indexed by various measures (TSP, PM10, sulfates, etc.).
29 A number of studies have been published since 1996 which evaluate the effects of
30 long-term PM exposure on lung function and respiratory symptoms, as presented in Chapter 8.
31 The methodology in the long-term studies varies much more than the methodology in the short-
32 term studies. Some studies reported highly significant results (related to PM), whereas others
June 2003 9-126 DRAFT-DO NOT QUOTE OR CITE
-------
1 reported no significant results. Of particular note are several studies reporting associations
2 between long-term PM exposures (indexed by various measures) or changes in such exposures
3 over time and chronic bronchitis rates, consistent with the findings on bronchitis from the
4 Dockery et al. (1996) study noted above.
5 Unfortunately, the cross-sectional studies often are potentially confounded, in part, by
6 unexplained differences in geographic regions; and it is difficult to separate out results consistent
7 with a PM gradient from any other pollutants or factors having the same gradient. The studies
8 that looked for a time trend also are confounded by other conditions that changed over time. The
9 most credible cross-sectional study remains that described by Dockery et al. (1996) and
10 Raizenne et al. (1996). Whereas most studies include two to six communities, this study
11 included 24 communities and is considered to provide the most credible estimates of long-term
12 PM exposure effects on lung function and respiratory symptoms.
13 Thus, the relative risk estimates for these three survival cohorts have converged in the
14 range of 7 to 13 percent increase in the non-external mortality rate associated with a 10 |ig/m3
15 increment in a long-term average of PM2 5. Methodological criticisms of these studies have been
16 largely resolved in favor of the validity of their original findings of a strong association between
17 long-term exposures to particulate matter and decreased survival (Bates, 2000).
18
19 9.8.4 Coherence of Reported Epidemiologic Findings
20 Interrelationships Between Health Endpoints. Considerable coherence exists across
21 newly available epidemiologic study findings. For example, it was earlier noted that effects
22 estimates for total (nonaccidental) mortality generally fall in the range of 2.5 to 5.0% excess
23 deaths per 50 |ig/m3 24-h PM10 increment. These estimates comport well with those found for
24 cause-specific cardiovascular- and respiratory-related mortality.
25 Furthermore, larger effect sizes for cardiovascular (in the range of 3 to 6% per 25 |ig/m3
26 24-h PM10 increment) and respiratory (in the range of 5 to 25% per 25 |ig/m3 24-h PM10) hospital
27 admissions and visits are found, as would be expected versus those for PM10-related mortality.
28 Also, several independent panel studies, evaluating temporal associations between PM exposures
29 and measures of heart beat rhythm in elderly subjects, provide generally consistent indications of
30 decreased heart rate (HR) variability being associated with ambient PM exposure (decreased HR
31 variability being an indicator of increased risk for serious cardiovascular outcomes, e.g., heart
June 2003 9-127 DRAFT-DO NOT QUOTE OR CITE
-------
1 attacks). Other studies point toward changes in blood characteristics (e.g., increased C-reactive
2 protein levels) related to increased risk of ischemic heart disease as also being associated with
3 ambient PM exposures. In addition, new evidence exists for ambient PM associations with
4 reductions in pulmonary function and/or increased respiratory symptoms, especially of note in
5 relation to asthmatic or other chronic lung disease individuals. All these CVD and respiratory
6 morbidity effects add to the coherence of the overall evidence substantiating short-term PM
7 exposure effects on susceptible population groups.
8 The overall body of controlled human and/or laboratory animal exposure studies discussed
9 earlier also add coherence to the evidence for ambient PM-related health impacts. A number
10 provide evidence that supports one or another hypothesis with regard to (a) PM components (by
11 size, chemical composition, etc.) and/or (b) mechanisms likely contributing to PM effects on
12 various cardiovascular or respiratory endpoints. The results of instillation studies, using filter
13 extracts from community monitoring stations in the Utah Valley before, during, and after
14 temporary shut down of a steel mill there are particularly compelling on two accounts: (1) the
15 evidence of greater lung inflammation from instilled extracts from periods of mill operation
16 parallel epidemiologic findings of increased cardiorespiratory hospitalizations during such
17 periods; and (2) dosimetric calculations indicate that concentrations of particulate extract
18 materials likely delivered to affected lung tissue with the instillation would probably be
19 reasonably comparable to those likely experienced in connection with inhalation exposures to
20 PM10 concentrations in the Utah Valley PM mixture.
21
22 Spatial Interrelationships. Both the NMMAPS and Cohort Reanalyses studies had a
23 sufficiently large number of cities to allow considerable resolution of regional PM effects within
24 the "lower 48" states, but this approach was taken much farther in the Cohort Reanalysis studies
25 than in NMMAPS. There were 88 cities with PM10 effect size estimates in NMMAPS; 50 cities
26 with PM25 and 151 cities with sulfates in Pope et al. (1995) and in the reanalyses using the
27 original data; and, in the additional analyses by the cohort study reanalysis team, 63 cities with
28 PM2 5 data and 144 cities with sulfate data. The relatively large number of data points allowed
29 estimation of surfaces for elevated long-term concentrations of PM2 5, sulfates, and SO2 with
30 resolution on a scale of a few tens to hundreds of kilometers. Information drawn from the maps
31 presented in Figures 16-21 in Krewski et al. (2000) is summarized below.
June 2003 9-128 DRAFT-DO NOT QUOTE OR CITE
-------
1 The patterns are similar, but not identical. In particular, the modeled PM25 surface
2 (Krewski, Figure 18) has peak levels in the industrial midwest, including the Chicago and
3 Cleveland areas, the upper Ohio River Valley, and around Birmingham, AL. Lower, but
4 elevated, PM25 is found almost everywhere else east of the Mississippi, as well as in southern
5 California. This was rather similar to the modeled sulfate surface (Krewski, Figure 16), with the
6 absence of a peak in Birmingham and an emerging sulfate peak in Atlanta. The only region with
7 elevated SO2 concentrations was the Cleveland-Pittsburgh area. A preliminary evaluation is that
8 secondary sulfates in particles derived from local SO2 were more likely to be important in the
9 industrial midwest, south from the Chicago-Gary region and along the upper Ohio River region.
10 The overlay of mortality and air pollution is also of interest. The spatial overlay of long-
11 term PM2 5 and mortality (Krewksi, Figure 21) was highest for the upper Ohio River region, but
12 also includes a significant association over most of the industrial midwest. This was reflected, in
13 diminished form, by the sulfates map (Krewski, Figure 19) where the peak sulfate-mortality
14 associations occur somewhat east of the peak PM2 5-mortality associations. The SO2 map
15 (Krewski, Figure 20) shows peak associations similar to, but slightly east of, the peak sulfate
16 associations. This suggests that, although SO2 may be an important precursor of sulfates in this
17 region, there may be other considerations (e.g., metals) in the association between PM2 5 and
18 long-term mortality, embracing a wide area of the midwest and northeast (especially noncoastal
19 areas).
20
21
22 9.9 SUSCEPTIBLE SUBPOPULATIONS AND IMPLICATIONS OF
23 EFFECTS OF AMBIENT PM EXPOSURE ON HUMAN HEALTH
24 9.9.1 Introduction
25 The 1996 PM AQCD identified several population groups potentially being at increased
26 risk for experiencing health impacts of ambient PM exposure. Elderly individuals (> 65 years)
27 were most clearly identified, along with those having preexisting cardiovascular or respiratory
28 disease conditions. Smokers and ex-smokers likely comprise a large percentage of individuals
29 with cardiovascular and respiratory disease, e.g., chronic obstructive pulmonary disease
30 (COPD). Individuals with asthma, especially children, also were identified as a potential
31 susceptible population group. The studies appearing since the 1996 PM AQCD provide
June 2003 9-129 DRAFT-DO NOT QUOTE OR CITE
-------
1 additional evidence to substantiate the above named groups as likely being at increased risk for
2 ambient PM-related morbidity or mortality effects. There is even evidence, though quite limited
3 at this time, of prenatal effects on cardiac development and potential mortality impacts on
4 infants in the first two years of life.
5 While the identification of susceptible population groups is a critical element of the risk
6 paradigm, characterizing risk factors that underlie susceptibility and that may be common to
7 multiple groups would better substantiate risk estimates and provide better predictability to PM
8 responsiveness. Information relating to these factors, as gleaned from recent epidemiology and
9 toxicology studies, suggests contributing host attributes that may be useful in gaining perspective
10 on their relative public health impact.
11
12 9.9.2 Preexisting Disease as a Risk Factor for Particulate Matter
13 Health Effects
14 The information reviewed in the 1996 PM AQCD is now augmented by numerous new
15 studies which substantiate the finding that preexisting disease conditions represents an important
16 risk factor for ambient PM health effects. Cardiovascular and respiratory diseases continue to
17 appear to be of greatest concern in relation to increasing risk for PM mortality and morbidity.
18 Indeed, the fact that these disease 'entities' often involve both organ systems, albeit to varying
19 degrees, might argue for their compilation under a broader classification of 'cardiopulmonary'
20 disease. Nevertheless, as they are diagnosed and reported separately, Table 9-12 shows the 1996
21 numbers of U.S. cases reported for COPD, asthma, heart disease, and hypertension.
22
23 9.9.2.1 Ambient PM Exacerbation of Cardiovascular Disease Conditions
24 Exacerbation of cardiovascular disease (CVD) has been associated epidemiologically, not
25 only with ambient PM, but also with other combustion-related ambient pollutants such as CO.
26 Thus, while leaving little doubt that ambient PM exposures importantly affect CVD mortality
27 and morbidity, the quantitation of the proportion of risk for such exacerbation specifically
28 attributable to ambient PM exposure is difficult. Recent studies (e.g., concentrated ambient
29 particle studies [CAPS]) have demonstrated cardiovascular effects in response to ambient
30 particle exposures, and studies utilizing animals and other approaches also have produced results
31 suggesting plausible mechanisms leading to cardiovascular effects. However, much remains to
32 be resolved with regard to delineation of dose-response relationships for the induction and
June 2003 9-130 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-12. INCIDENCE OF SELECTED CARDIORESPIRATORY DISORDERS BY AGE AND
l->*
o
o
OJ
VO
OJ
O
H
6
o
0
H
O
o
H
W
O
O
w
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 et al. (1999).
-------
1 extrapolation of such effects to estimate appropriate and effective human equivalent PM (or
2 specific constituent/s) exposures.
3 The recent appreciation for underlying cardiovascular dysfunction as a risk factor for PM
4 health effects derives from a growing and diverse body of literature. While many time-series
5 studies have revealed stronger associations between PM exposures and mortality when a
6 subpopulation was segregated for pre-existent cardiac disease, no direct and plausible evidence
7 had previously been available. However, recent panel studies of human subjects with CVD
8 (Peters et al., 2000) have shown correlations between air pollution levels, notably PM, and
9 intervention discharge frequency of implanted cardiac defribrillators. Analogously, Pope and
10 colleagues (2001) have noted altered autonomic control of cardiac electrocardiograms (in terms
11 of heart rate variability) over a wide age- range of ostensibly healthy subjects when they were
12 introduced into a room with active smokers. Evidence of vascular narrowing with exposure to
13 concentrated ambient PM (CAPS) has likewise been reported suggesting parallel cardiovascular
14 responses (Brook et al., 2002). Collectively, these and previous studies that have shown ambient
15 PM-induced alterations in cardiac physiology (Pope et al, 1999a,b; Liao et al., 1999; Peters et al.,
16 1999; Gold et al., 2000) in human subjects, complemented with animal studies (Godleski et al.,
17 1996; Watkinson et al., 1998, 2001; Kodavanti et al., 2000), reinforce the notion of significant
18 cardiac responses to PM. Moreover, indications of changes in plasma viscosity (Peters et al.,
19 1997a) and other factors involved in clotting function (Ohio et al., 2000) provide a plausible
20 cascade of events that could culminate in a sudden cardiac event in some individuals.
21 To the extent that the observed associations between ambient PM and heart disease
22 exacerbation are causal and specific, the impact on public health could be dramatic. In 1997,
23 there were about 4,188,000 U.S. hospital discharges with heart disease as the first-listed
24 diagnosis (Lawrence and Hall, 1999). Among these, about 2,090,000 (50%) were for ischemic
25 heart disease, 756,000 (18%) for myocardial infarction or heart attack (a subcategory of ischemic
26 heart disease), 957,000 (23%) for congestive heart failure, and 635,000 (15%) for cardiac
27 dysrhythmias. Also, there were 726,974 deaths from heart disease (Hoyert et al., 1999). Thus,
28 even a small percentage reduction in PM-associated admissions or deaths from heart disease
29 would predict a large number of avoided cases.
30
31
June 2003 9-132 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.9.2.2 Ambient PM Exacerbation of Respiratory Disease Conditions
2 Many time-series studies have shown that pre-existent chronic lung diseases as a group
3 (but especially chronic obstructive pulmonary disease - COPD) constitutes a risk factor for
4 mortality with PM exposure. Studies with humans that might reveal more specific data have
5 been limited both ethically, as well as by the absence of good biomarkers of response (such as
6 ECG's serve cardiac disease). Measures of blood-gas saturation and lung function appear not to
7 be 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
25 the elderly with cardiopulmonary disease appears to result in complications of underlying
26 cardiac disorders when PM exposure is involved (Zanobetti et al., 2000), and likewise is linked
27 to subsequent hospitalization. Animal studies with surrogate PM, however, show varied impact
28 on the induction of infection, but in general can alter lung phagocyte functions, which might
29 worsen the condition. Thus, while there appears to be a strong likelihood that infections may be
30 worsened by exposure to PM, general statements regarding interaction of PM with response to
June 2003 9-133 DRAFT-DO NOT QUOTE OR CITE
-------
1 infectious agents are difficult given the unique attributes of various infectious agents and the
2 immune status 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
30
June 2003 9-134 DRAFT-DO NOT QUOTE OR CITE
-------
1 9.9.3 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
6 in studies where age is a factor in the analysis, mortality risk increases above the age of 45 and
7 continues to increase significantly throughout the remainder of life. Cardiopulmonary diseases
8 more common to the elderly play into the risk within older age groups, but panel studies of
9 morbidity focusing on generally healthy people in retirement homes or elderly volunteers
10 exposed to concentrated ambient PM in chambers show subtle alterations of autonomic control
11 of cardiac function (i.e., slight depression of heart rate variability) and blood factors concordant
12 with a putative response to ambient PM levels. Though small, these changes are considered
13 clinically significant based on studies of risk in cardiac patients and general population studies of
14 cardiac disease progression. Moreover, these changes are in contrast to the lack of similar
15 physiologic changes in healthy young people. Over the long term, innate differences in
16 metabolism or other mechanisms may impact the likelihood of chronic outcomes, e.g., COPD or
17 lung cancer. To what extent progression occurs with repeated PM exposures and how much
18 disease or other risk factors add 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
June 2003 9-135 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-13. NUMBER OF ACUTE RESPIRATORY CONDITIONS PER
100 PERSONS PER YEAR, BY AGE: UNITED STATES, 1996
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
45
Total
53.3
16.1
7.0
23.3
3.8
*2.0
*1.1
Years and Over
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 et al. (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
June 2003
9-136
DRAFT-DO NOT QUOTE OR CITE
-------
0.6
0.5
p
b>
E 0.4
-ffi
CO
0.3
_co
CO
_c
— 0.2
0.1
10
I
20
I
30
I
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
13 deposition (either in terms of dose and/or intrapulmonary distribution) or other biologic aspects
June 2003
9-137
DRAFT-DO NOT QUOTE OR CITE
-------
1 of the cardiopulmonary system or disease thereof. The role of innate attributes of risk grounded
2 in one's genetic code is largely unknown but potentially of great importance. Animal models
3 have been used to show clear differences in response to PM and other pollutants, and the critical
4 involvement of varied genes in the induction of asthma, emphysema, and many other ailments is
5 widely accepted, but poorly understood.
6
7 9.9.4 Impact on Life-Expectancy
8 The increased rate of non-external mortality found in the three prospective cohort studies
9 (Harvard Six Cities; ACC, AHSMOG) is greater than the mere accumulation of the adverse
10 effects of short-term exposures for a few days. Conceptually, ambient PM exposures may be
11 associated with both the long-term development of underlying health problems ("frailty") and
12 with the short-term variations in timing of mortality among a susceptible population with some
13 underlying health condition (Kunzli et al. 2001). Epidemiologic studies of the mortality effects
14 of short-term exposure to particulate matter using time-series studies can only capture PM's
15 association with short-term variations in mortality and, therefore, must systematically
16 underestimate the proportion of total mortality attributable to PM. A recent time-series study
17 that examined the contribution of daily PM levels over an extended lag period (42 days) could
18 only partially bridge the gap between the effects of short-term and long-term exposures to
19 particulate matter (Zanobetti et al., 2002).
20 Recent investigations of the public health implications of effect estimates for long-term
21 PM exposures also were reviewed in Chapter 8. Life table calculations by Brunekreef (1997)
22 found that relatively small differences in long-term exposure to ambient airborne PM can have
23 substantial effects on life expectancy. For example, a calculation for the 1969 to 1971 life table
24 for U.S. white males indicated that a chronic exposure increase of 10 |ig/m3 PM was associated
25 with a reduction of- 1.3 years for the entire population's life expectancy at age 25. The new
26 evidence noted above of infant mortality associations with PM exposure suggests that life
27 shortening in the entire population from long-term PM exposure could well be significantly
28 larger than estimated by Brunekreef (1997).
29 The increase in non-external mortality cannot be explained by increases in chronic
30 respiratory diseases since chronic non-malignant lower respiratory disease accounts for only
31 5.6 percent and lung cancer for only another 6.9 percent of all deaths over age 24 years due to
June 2003 9-138 DRAFT-DO NOT QUOTE OR CITE
-------
1 non-external causes. Cardiovascular diseases, which account for 43 percent of non-external
2 mortality, must play the leading role in the decreased survival associated with exposure to
3 ambient PM. It is nevertheless useful to highlight the newer results of the extension of the ACS
4 study analyses (that include more years of participant follow-up and address previous criticisms
5 of the earlier ACS analyses), which provide the strongest evidence to date that long-term
6 ambient PM exposures are associated with increased risk of lung cancer. That increased risk
7 appears to be in about the same range as that seen for a non-smoker residing with a smoker and,
8 therefore, passively exposed chronically to tobacco smoke, with any consequent life-shortening
9 impacts due to lung cancer.
10
11
12 9.10 INTEGRATIVE SYNTHESIS OF KEY FINDINGS FOR
13 ENVIRONMENTAL EFFECTS OF AMBIENT AIRBORNE PM
14 9.10.1 Introduction
15 The 1997 EPA revisions to the U.S. PM NAAQS, discussed in Chapter 1 (Introduction),
16 included establishment of PM25 secondary standards identical to the primary PM25 NAAQS set
17 at that time. The 1997 FR notice promulgating these standards noted "The new secondary
18 standards, in conjunction with a regional haze program, will provide appropriate protection
19 against PM-related public welfare effects including soiling, material damage, and visibility
20 impairment." This section of Chapter 9 concisely highlights salient information expected to
21 provide inputs to EPA decision making on secondary National Ambient Air Quality Standards
22 (NAAQS) aimed at protecting against welfare effects of ambient airborne particulate matter
23 (PM). More specifically, it discusses effects of atmospheric PM on the environment, including:
24 (a) direct and indirect effects on vegetation and natural ecosystem integrity; (b) effects on
25 visibility; and (c) effects on man-made materials, as well as (d) relationships of atmospheric PM
26 to climate change processes.
27
28 9.10.2 Effects of Ambient Airborne PM on Vegetation and Natural
29 Ecosystems
30 The effects of airborne particles are manifested via direct physical and chemical effects
31 exerted at the individual plant level and/or indirectly via deposition on soils and/or waterways.
June 2003 9-139 DRAFT-DO NOT QUOTE OR CITE
-------
1 However, plants are key members of ecosystems, structurally complex communities comprised
2 of populations of plants, animals (including humans), insects, and microorganisms that interact
3 with one another and with their non-living (abiotic) chemical and physical environment in which
4 they exist (Odum, 1989; U.S. Environmental Protection Agency, 1993). All life on Earth is
5 dependent on chemical energy in the form of carbon compounds to sustain their life processes.
6 Terrestrial vegetation, via the process of photosynthesis, provides approximately half of the
7 carbon that annually cycles between the Earth and the atmosphere (Chapin and Ruess, 2001).
8 Ecosystems respond to stresses through their constituent organisms. The responses of
9 plant species and populations to environmental perturbations (such as those caused by
10 atmospheric PM) depend on their genetic constitution (genotype), their life cycles, and the
11 microhabitats in which they are growing. Stresses that produce changes in their physical and
12 chemical environment apply selection pressures on individual organisms (Treshow, 1980). The
13 changes that occur within populations and plant communities reflect these new and different
14 pressures. A common response in a community under stress is the elimination of the more
15 sensitive populations and an increase in abundance of species that tolerate or are favored by
16 stress (Woodwell, 1970, Guderian et al., 1985).
17 The present section is organized to discuss: (1) factors affecting deposition of airborne PM
18 on plants and ecosystems and then (2) the effects of PM deposition on individual plants, plant
19 populations, forest trees, and terrestrial and aquatic ecosystems. As such, the section is
20 organized to follow, in rough outline, the Framework for Assessing and Reporting on Ecological
21 Condition recommended in a report by the Ecological Processes and Effects Committee (EPEC)
22 of EPA's Science Advisory Board (Science Advisory Board, 2002), which states "The purpose
23 of this report is to provide the Agency with a sample framework that may serve as a guide for
24 designing a system to assess, and then report on, ecological condition at local, regional, or
25 national scale. The sample framework is intended as an organizing tool that may help the
26 Agency decide what ecological attributes to measure and how to aggregate those measurements
27 into an understandable picture of ecological integrity." This framework is not actually a risk
28 assessment per se, but it can be used to "construct a report of ecological condition" that
29 characterizes the ecological integrity of an ecosystem based on "the relationship between
30 common anthropogenic stressors and one or more of the six Essential Ecological Attributes."
June 2003 9-140 DRAFT-DO NOT QUOTE OR CITE
-------
1 It nevertheless does provide a useful approach for organizing discussions of stressor effects on
2 ecosystem components at successive levels of complexity.
3
4 9.10.2.1 Ecological Attributes
5 The EPEC Framework provides a checklist of generic ecological attributes that should be
6 considered when evaluating the integrity of ecological systems (see Table 9-14). The six generic
7 ecological attributes, termed Essential Ecological Attributes (EEA), represent groups of related
8 ecological characteristics (Science Advisory Board, 2002; Harwell et al., 1999) and include:
9 Chemical and Physical Characteristics; Biotic Conditions; Landscape Conditions; Ecological
10 Processes; Hydrology and Geomorphology; and Natural Disturbance Regimes. All of the EEAs
11 are interrelated (i.e., changes in one EEA may directly or indirectly affect other EEAs).
12 The first three ecological attributes listed in Table 9-14 are primarily "patterns," whereas the last
13 three are "processes." Ecological science has used "patterns" and "processes" as terms to
14 describe features of ecological systems for many years (e.g., Bormann and Likens, 1979).
15 Of main concern in this chapter are relationships between a certain class of diverse airborne
16 stressors from anthropogenic sources, termed particulate matter (PM), and one or more of the
17 EEAs. Changes in patterns resulting from responses of vegetation and ecosystems to the effects
18 of fine and coarse PM deposition, along with known or possible effects on ecological processes
19 associated with changes in the patterns, are discussed in the subsections that follow.
20 The reader is also referred to several other sources for more detailed discussions of several
21 topics only briefly alluded to or addressed here. For example, an extensive discussion of various
22 types of effects of acidic deposition is presented in the U.S. National Acid Precipitation
23 Assessment Program (NAPAP) Biennial Report to Congress: An Integrated Assessment
24 Program (National Scientific and Technology Council, 1998). Additionally, ecological effects
25 of acidic precipitation and nitrate deposition on aquatic systems are discussed in the EPA Air
26 Quality Criteria Document for Nitrogen Oxides (U.S. Environmental Protection Agency, 1993);
27 and sulfate deposition and effects, as related to wetlands and aquatic habitats, are discussed in
28 U.S. Environmental Protection Agency (1982). Effects of lead on crops, vegetation, and
29 ecosystems are assessed in the EPA document, Air Quality Criteria for Lead (U.S.
30 Environmental Protection Agency, 1986). Lastly, effects of "certain pesticides, metal
31 compounds, chlorinated organic compounds, and nitrogen compounds" are discussed in
June 2003 9-141 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-14. ESSENTIAL ECOLOGICAL ATTRIBUTES AND
REPORTING CATEGORIES
Landscape Condition
• Extent of Ecological System/Habitat Types
• Landscape Composition
• Landscape Pattern and Structure
Biotic Condition
• Ecosystems and Communities
- Community Extent
- Community Composition
- Trophic Structure
- Community Dynamics
- Physical Structure
• Species and Populations
- Population Size
- Genetic Diversity
- Population Structure
- Population Dynamics
- Habitat Suitability
• Organism Condition
- Physiological Status
- Symptoms of Disease or Trauma
- Signs of Disease
Chemical and Physical Characteristics
(Water, Air, Soil, and Sediment)
• Nutrient Concentrations
- Nitrogen
- Phosphorus
- Other Nutrients
• Trace Inorganic and Organic Chemicals
- Metals
- Other Trace Elements
- Organic Compounds
• Other Chemical Parameters
-pH
- Dissolved Oxygen
- Salinity
- Organic Matter
- Other
• Physical Parameters
Ecological Processes
• Energy Flow
- Primary Production
- Net Ecosystem Production
- Growth Efficiency
• Material Flow
- Organic Carbon Cycling
- Nitrogen and Phosphorus Cycling
- Other Nutrient Cycling
Hydrology and Geomorphology
• Surface and Groundwater flows
- Pattern of Surface flows
- Hydrodynamics
- Pattern of Groundwater flows
- Salinity Patterns
- Water Storage
• Dynamic Structural Characteristics
- Channel/Shoreline Morphology, Complexity
- Extent/Distribution of Connected Floodplain
- Aquatic Physical Habitat Complexity
• Sediment and Material Transport
- Sediment Supply/Movement
- Particle Size Distribution Patterns
- Other Material Flux
Natural Disturbance Regimes
• Frequency
• Intensity
• Extent
• Duration
Source: Science Advisory Board (2002).
1 Deposition of Air Pollutants to the Great Waters, Third Report to Congress (U. S. Environmental
2 Protection Agency, 2000b).
June 2003
9-142
DRAFT-DO NOT QUOTE OR CITE
-------
1 9.10.2.2 Ecosystem Exposures - Particle Deposition
2 Airborne particles, their precursors, and their transformation products are removed from
3 the atmosphere by wet and dry deposition processes. This atmospheric cleansing process
4 fortunately lowers the long-term buildup of lethal concentrations of these pollutants in the air
5 and moderates the potential for direct human health effects caused by their inhalation.
6 Unfortunately, these deposition processes also mediate the transfer of PM pollutants to other
7 environmental media where they can and do alter the structure, function, diversity, and
8 sustainability of complex ecosystems.
9 The potential effects of PM deposition on vegetation and ecosystems encompass the full
10 range, scales, and properties of biological organization listed under Biotic Condition. Exposure
11 to a given mass concentration of airborne PM, however, may lead to widely differing responses,
12 depending on the particular mix of deposited particles. Particulate matter is not a single
13 pollutant, but rather a heterogeneous mixture of particles differing in size, origin, and chemical
14 composition. This heterogeneity exists across individual particles within samples from
15 individual sites and, to an even greater extent, between samples from different sites. Thus far,
16 atmospheric PM has been defined, for regulatory purposes, mainly by size fractions and less
17 clearly so in terms of chemical nature, structure, or source. While size is related to the mode and
18 magnitude of deposition to vegetated landscapes and may be a useful surrogate for chemical
19 constitution, PM size classes do not necessarily have specific differential relevance for
20 vegetation effects (Whitby, 1978; U.S. Environmental Protection Agency, 1996a); that is, both
21 fine- and coarse-mode particles may affect plants. Much of the burden of sulfates, nitrates,
22 ammonium salts, and hydrogen ions resides in the atmosphere either dissolved in fog water or as
23 liquid or solid aerosols. Therefore, assessment of atmospheric PM deposition and effects on
24 vegetation unavoidably include discussion of nitrates and sulfates and associated compounds
25 involved in acidic and acidifying deposition. Other important issues relate to trace elements and
26 heavy metals often found in ambient airborne PM.
27
28 9.10.2.3 Direct and Indirect Effects on Ecosystems
29 The deposition of PM onto vegetation and soil, depending on its chemical composition
30 (acid/base, trace metal, or nutrients, e.g., nitrates or sulfates), can produce direct or indirect
31 responses within an ecosystem. Direct effects are chiefly physical. The effects of toxic particles
June 2003 9-143 DRAFT-DO NOT QUOTE OR CITE
-------
1 are both chemical and physical. Direct ecosystem effects have been observed largely in the
2 neighborhood of point sources such as limestone quarries; cement kilns; and iron and lead
3 smelting factories. The nitrates and sulfates whose indirect effects occur through the soil
4 environment are considered to be the stressors of greatest environmental significance. Upon
5 entering the soil environment, they can alter the ecological processes of energy flow and nutrient
6 cycling, inhibit nutrient uptake, change ecosystem structure, and affect ecosystem biodiversity.
7 The soil environment is one of the most dynamic sites of biological interaction in nature.
8 Bacterial communities are essential participants in the nitrogen and sulfur cycles that make these
9 elements available for plant uptake. Fungi in association with plant roots form mycorrhizae,
10 a mutualistic symbiotic relationship that is integral in mediating plant uptake of mineral
11 nutrients. Changes in the soil environment that influence the role of the bacteria in nutrient
12 cycling and fungi in nutrient uptake determine plant and ultimately ecosystem response..
13 Ecosystem response to pollutant deposition is a direct function of the level of sensitivity of
14 the ecosystem and its ability to ameliorate resulting change. The Essential Ecological Attributes
15 (EEA's) provide a hierarchical framework for determining ecosystem status associated wiht the
16 last three EEA's (Table 9-14). The first three are considered to be "patterns" and the last three
17 "processes". The ecological processes create and maintain the ecosystem elements in the
18 patterns of the first three EEA's. The patterns in turn affect how the ecosystem processes are
19 expressed. Patterns at the higher level of biological organization emerge from the interactions
20 and selection processes at localized levels. Changes in patterns or processes result in changes in
21 the status and functioning of an ecosystem. The relationships among the EEAs are complex
22 because all are interrelated (i.e., changes in one EEA may affect, directly or indirectly, every
23 other EEA). The functioning of the ecological processes associated with the Ecological Process
24 EEAs must be scaled in both time and space and propagated to the more complex levels of
25 community interaction to produce observable ecosystem changes.
26 Both ecosystem structure (Biotic condition) and functions (Ecological Processes) are
27 important in providing products and services essential to human existence on planet Earth.
28 Ecosystem processes maintain clean water, clean air, a vegetated earth, and a balance of
29 organisms. Also included in the benefits are absorption and breakdown of pollutants, cycling of
30 nutrients, binding of the soil, degradation of organic waste, maintenance of a balance of gases in
31 the air, regulation of radiation balance, climate, and fixation of solar energy. Concern has arisen
June 2003 9-144 DRAFT-DO NOT QUOTE OR CITE
-------
1 in recent years regarding biodiversity and the integrity of ecosystems. Human-induced changes
2 in biotic diversity and alterations in EEA patterns and the functioning of EEA processes are the
3 two most dramatic ecological trends of the past century. Biodiversity is of major importance in
4 the functioning of ecosystems.
5 Nitrogen in nature may be divided into two groups: nonreactive (N2) and reactive (Nr).
6 Although nitrogen as molecular nitrogen (N2) is the most abundant element in the atmosphere, it
7 is not available to more than 99% of living organisms. It only becomes available after it is
8 converted into reactive (Nr) forms. Reactive Nr includes all biologically, photochemically, and
9 radioactively active nitrogen compounds in the earth's atmosphere and biosphere. Among those
10 included are: the inorganic reduced forms of nitrogen (e.g., ammonia [NH3] and ammonium
11 [NH4+]), inorganic oxidized forms (e.g., nitrogen oxide [NOJ, nitric acid [HNO3], nitrous oxide
12 [N2O], and nitrate [NO3"]) , and organic compounds (e.g., urea, amine, proteins, and nucleic
13 acids)]).
14 The overall increase in global Nr is the result of three main causes: (1) widespread
15 cultivation of legumes, rice and other crops that promote conversion of N2 to organic nitrogen
16 through biological nitrogen fixation; (2) combustion of fossil fuels, which converts both
17 atmospheric N2 and fossil nitrogen to reactive NOX; and (3) the Haber-Bosch process, which
18 converts nonreactive NH3 to sustain food production and some industrial activities. The
19 deposition of nitrogen in the United States from human activity has doubled between 1961 and
20 1997 due mainly to the use of inorganic nitrogen fertilizers and the emissions of nitrogen oxides
21 (NOX) from fossil fuel emissions with the largest increase occurring in the 1960s and 1970s.
22 As a result, Nr is accumulating in various environmental reservoirs, e.g., the atmosphere, soils
23 and waters. The accumulation of Nr in the terrestrial environment results in major changes in
24 the nitrogen cycle, as it moves thru various environmental reservoirs depicted in Figure 9-24.
25 The results of increased Nr in the global system and the wide variety changes in the
26 nitrogen cycle are both beneficial and detrimental to the health and welfare of humans and
27 ecosystems. The synthetic fertilizers used in cultivation and the cultivation-induced bacterial
28 nitrogen fertilization (BNF) sustain a large portion of the world's population.
29 Reactive nitrogen can be widely dispersed and accumulate in the environment when the
30 rates of its formation exceed the rates of removal via denitrification. Nr creation and
31 accumulation is projected to increase as per capita use of resources by human populations
June 2003 9-145 DRAFT-DO NOT QUOTE OR CITE
-------
NOX
Energy
Production
|
Ozi
Effi
A
Atmosphere
Food
Production
People
(Food; Fiber)
Human Activities
The Nitrogen
Cascade
Indicates cienitrification potential
Figure 9-24. Illustration of the nitrogen cascade showing the movement of the human-
produced reactive nitrogen (Nr) as it cycles through the various
environmental reservoirs in the atmosphere, terrestrial ecosystems, and
acquatic ecosystems.
Modified from Galloway and Cowling (2002).
1 increases. The cascade of environmental effects resulting from increases in Nr include the
2 following: (1) production of tropospheric ozone and aerosols that induce human health and
3 environmental problems; (2) increases in the productivity in forests and grasslands followed by
4 decreases wherever deposition increases significantly and exceeds critical thresholds; Nr
5 additions probably also decrease biodiversity in many natural habitats; (3) in association with
6 sulfur is responsible for acidification and loss of biodiversity in lakes and streams in many
7 regions of the world; (4) eutrophication, hypoxia, loss of biodiversity, and habitat degradation in
8 coastal ecosystems. [Eutrophication is now considered the biggest pollution problem in coastal
June 2003
9-146
DRAFT-DO NOT QUOTE OR CITE
-------
1 waters.] (5) contributes to global climate change and stratospheric ozone depletion, which can in
2 turn affect ecosystems and human health (Figure 9-24).
3 Direct effects of Nr on human health and the environment include (1) increased yields and
4 nutritional quality of food needed to meet dietary requirements and food preferences for growing
5 populations; (2) respiratory and cardiac disease induced by exposure to high ozone and fine PM
6 concentrations; (3) decreased growth and yields of certain sensitive plant species; (4) nitrate and
7 nitrite contamination of drinking water leading to the "blue baby syndrome" and certain types of
8 cancer; and (5) blooms of toxic algae, with resultant injury to humans and to fish and other
9 aquatic life.
10 Indirect effects on societal values include: (1) regional hazes that decrease visibility at
11 scenic vistas and airports; (2) depletion of stratospheric ozone by N2O emissions; (3) global
12 climate change induced by emissions of N2O and formation of tropospheric ozone; (4) formation
13 of acidic deposition. The magnitude of Nr flux often determines whether effects are beneficial
14 or detrimental (Table 9-15).
15 Among the most important effects of chronic nitrogen deposition are changes in the
16 composition of plant communities, disruptions in nutrient cycling, increased emissions of
17 nitrogenous greenhouse gases from soil and accumulation of nitrogen compounds resulting in
18 the enhanced availability of nitrate or ammonium, the soil-mediated effects of acidification, and
19 increased susceptibility to stress factors. A major concern is "nitrogen saturation," the result of
20 the atmospheric deposition of large amounts of particulate nitrates, often as a consequence of
21 slow deposition over long time periods. Nitrogen saturation results when additions to soil
22 background nitrogen (nitrogen loading) exceeds the capacity of plants and soil microorganisms
23 to utilize and retain nitrogen. Under these circumstances, disruptions of ecosystem functioning
24 may result.
25 Although soils of most North American forest ecosystems are nitrogen limited, there are
26 some that exhibit severe symptoms of nitrogen saturation. Increases in soil nitrogen play a
27 selective role in ecosystems. Plant succession patterns and biodiversity are affected significantly
28 by chronic nitrogen additions in some North American ecosystems. Plants adapted to living in
29 an environment of low nitrogen availability will be replaced by nitrophilic plants capable of
30 using increased nitrogen because they have a competitive advantage when nitrogen becomes
31 more readily available. Long-term nitrogen fertilization studies in both New England and
June 2003 9-147 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9-15. EFFECTS OF REACTIVE NITROGEN
Direct effects ofNr on ecosystems include:
• Increased productivity of Nr-limited natural ecosystems.
• Ozone-induced injury to crop, forest, and natural ecosystems and predisposition to attack by pathogens
and insects.
• Acidification and eutrophication effects on forests, soils, and freshwater aquatic ecosystems.
• Eutrophication and hypoxia in coastal ecosystems.
• N saturation of soils in forests and other natural ecosystems.
• Biodiversity losses in terrestrial and aquatic ecosystems and invasions by N-loving weeds.
• Changes in abundance of beneficial soil organisms that alter ecosystem functions.
Indirect effects ofNr on other societal values include:
• Increased wealth and well being of human populations in many parts of the world.
• Significant changes in patterns of land use.
• Regional hazes that decrease visibility at scenic vistas and airports.
• Depletion of stratospheric ozone by N2O emissions.
• Global climate change induced by emissions of N2O and formation of tropospheric ozone.
• Damage to useful materials and cultural artifacts by ozone, other oxidants, and acid deposition.
• Long-distance transport ofNr which causes harmful effects in countries distant from emission sources and/or
increased background concentrations of zone and fine paniculate matter.
In addition to these effects, it is important to recognize that:
• The magnitude of Nr flux often determines whether effects are beneficial or detrimental.
• All of these effects are linked by biogeochemical circulation pathways of Nr.
• Nr is easily transformed among reduced and oxidized forms in many systems. Nr is easily distributed by
hydrologic and atmospheric transport processes.
1 Europe suggest that some forests receiving chronic inputs of nitrogen may decline in
2 productivity and experience greater mortality. Declining coniferous forest stands with slow
3 nitrogen cycling may be replaced by deciduous fast-growing forests that cycle nitrogen more
4 rapidly.
5 Linked to the nitrogen cascade (see Figure 9-24) is the deposition ofNr and sulfates and
6 the associated hydrogen ion in acidic precipitation, a critical environmental stress that affects
7 forest landscapes and aquatic ecosystems in North America, Europe, and Asia. Composed of
8 ions, gases, and particles derived from gaseous emissions of sulfur dioxide (SO2), nitrogen
June 2003 9-148 DRAFT-DO NOT QUOTE OR CITE
-------
1 oxides (NOX), ammonia (NH3) and participate emissions of acidifying and neutralizing
2 compounds, acidic precipitation is highly variable across time and space. Its deposition and the
3 resulting soil acidity can lead to plant nutrient deficiencies and to high aluminum-to-nutrient
4 ratios that limit plant uptake of calcium and magnesium and create a nutrient deficiency.
5 Aluminum accumulation in root tissue can reduce calcium uptake and causes Ca2+deficiencies.
6 Tree species can be adversely affected if altered Ca/Al ratios impair calcium or magnesium
7 uptake. Calcium is essential in the formation of wood and the maintenance of the primary plant
8 tissues necessary for tree growth.
9 Notable impacts of excess nitrogen deposition also have been observed with regard to
10 aquatic systems. For example, atmospheric nitrogen deposition into soils in watershed areas
11 feeding into estuarine sound complexes (e.g., the Pamlico Sound of North Carolina) appear to
12 contribute to excess nitrogen flows in runoff (especially during and after heavy rainfall events
13 such as hurricanes) from agricultural practices or other uses (e.g., fertilization of lawns or
14 gardens), massive influxes of such nitrogen into watersheds and sounds can lead to dramatic
15 decreases in water oxygen and increases in algae blooms that can cause extensive fish kills and
16 damage to commercial fish and sea food harvesting.
17 An important characteristic of fine particles is their ability to affect flux of solar radiation
18 passing through the atmosphere directly, by scattering and absorbing solar radiation, and
19 indirectly, by acting as cloud condensation nuclei that, in turn, influence the optical properties of
20 clouds. Regional haze has been estimated to diminish surface solar visible radiation by
21 approximately 8%. Crop yield have been reported as being sensitive to the amount of sunlight
22 receive, and crop losses have been attributed to increased airborne particle levels in some areas
23 of the world.
24
25 9.10.3 Visibility Effects of Airborne Particles
26 Visibility is defined as the degree to which the atmosphere is transparent to visible light
27 and the clarity and color fidelity of the atmosphere. Visual range is the farthest distance a black
28 object can be distinguished against the horizontal sky. Visibility impairment is any humanly
29 perceptible change in visibility. For regulatory purposes, visibility impairment, characterized by
30 light extinction, visual range, contrast, and coloration, is classified into two principal forms:
31 (1) "reasonably attributable" impairment, attributable to a single source or small group of
June 2003 9-149 DRAFT-DO NOT QUOTE OR CITE
-------
1 sources, and (2) regional haze, any perceivable change in visibility caused by a combination of
2 many sources over a wide geographical area.
3 Visibility is measured by human observation, light scattering by particles, the light
4 extinction-coefficient, and parameters related to the light-extinction coefficient (visual range and
5 deciview scale), and fine PM mass concentrations.
6 The air quality within a sight path will affect the illumination of the sight path by scattering
7 or absorbing solar radiation before it reaches the Earth's surface. The rate of energy loss with
8 distance from a beam of light is the light extinction coefficient. The light extinction coefficient
9 is the sum of the coefficients for light absorption by gases (oag), light scattering by gases (osg),
10 light absorption by particles (oap), and light scattering by particles (osp). Corresponding
11 coefficients for light scattering and absorption by fine and coarse particles are osfp and oafp and
12 oscp and oacp, respectively. Visibility within a sight path longer than approximately 100 km
13 (60 mi) is affected by the change in the optical properties of the atmosphere over the length of
14 the sight path.
15 Visual range was developed for and continues to be used as an aid in military operations
16 and to a lesser degree in transportation safety. Visual range is commonly taken to be the greatest
17 distance a dark object can be seen against the background sky. The deciview is an index of
18 haziness. A change of 1 or 2 deciviews is seen as a noticeable change in the appearance of a
19 scene.
20 Under certain conditions, fine particle mass concentrations may be used as a visibility
21 indicator. However, the relationship may differ between locations and for different times of the
22 year. Also, measurement should be made under dry conditions.
23 Visibility impairment is associated with airborne particle properties, including size
24 distributions (i.e., fine particles in the 0.1- to 1.0-|im size range) and aerosol chemical
25 composition, and with relative humidity. With increasing relative humidity, the amount of
26 moisture available for absorption by particles increases, thus causing the particles to increase in
27 both size and volume. As the particles increase in size and volume, the light scattering potential
28 of the particles also generally increases. Visibility impairment is greatest in the eastern United
29 States and Southern California. In the eastern United States, visibility impairment is caused
30 primarily by light scattering by sulfate aerosols and, to a lesser extent, by nitrate particles and
31 organic aerosols, carbon soot, and crustal dust. Up to 86% of the haziness in the eastern United
June 2003 9-150 DRAFT-DO NOT QUOTE OR CITE
-------
1 States is caused by atmospheric sulfate. Further West, scattering contributions to visibility
2 impairment decrease to from 25 to 50%. Light scattering by nitrate aerosols is the major cause
3 of visibility impairment in southern California. Nitrates contribute about 45% to the total light
4 extinction in the West and up to 17% of the total extinction in the East. Organic particles are the
5 second largest contributors to light extinction in most U.S. areas. Organic carbon is the greatest
6 cause of light extinction in the West, accounting for up to 40% of the total extinction and up to
7 18% of the visibility impairment in the East. Coarse mass and soil, primarily considered
8 "natural extinction," is responsible for some of the visibility impairment in the West, accounting
9 for up to 25% of the light extinction.
10
11 9.10.4 Materials Damage Related to Airborne Participate Matter
12 Building materials (metals, stones, cements, and paints) undergo natural weathering
13 processes from exposure to environmental elements (wind, moisture, temperature fluctuations,
14 sun light, etc.). Metals form a protective film of oxidized metal (e.g., rust) that slows
15 environmentally induced corrosion. On the other hand, the natural process of metal corrosion
16 from exposure to natural environmental elements is enhanced by exposure to anthropogenic
17 pollutants, in particular SO2, that render the protective film less effective.
18 Dry deposition of SO2 enhances the effects of environmental elements on calcereous stones
19 (limestone, marble, and cement) by converting calcium carbonate (calcite) to calcium sulfate
20 dihydrate (gypsum). The rate of deterioration is determined by the SO2 concentration, the
21 stone's permeability and moisture content, and the deposition rate; however, the extent of the
22 damage to stones produced by the pollutant species apart from the natural weathering processes
23 is uncertain. Sulfur dioxide also has been found to limit the life expectancy of paints by causing
24 discoloration and loss of gloss and thickness of the paint film layer.
25 A significant detrimental effect of particle pollution is the soiling of painted surfaces and
26 other building materials. Soiling changes the reflectance of an opaque material and reduces the
27 transmission of light through transparent materials. Soiling is a degradation process that requires
28 remediation by cleaning or washing, and, depending on the soiled surface, repainting. Available
29 data on pollution exposure indicates that particles can result in increased cleaning frequency of
30 the exposed surface and may reduce the usefulness of the soiled material. Attempts have been
31 made to quantify the pollutants exposure levels at which materials damage and soiling have been
June 2003 9-151 DRAFT-DO NOT QUOTE OR CITE
-------
1 perceived. However, to date, insufficient data are available to advance our knowledge regarding
2 perception thresholds with respect to pollutant concentration, particle size, and chemical
3 composition.
4
5 9.10.5 Atmospheric Particle Effects on Global Warming Processes and
6 Transmission of Solar Ultraviolet Radiation
7 The physical processes (i.e., scattering and absorption) responsible for airborne particle
8 effects on transmission of solar visible and ultraviolet radiation are the same as those responsible
9 for visibility degradation. Scattering of solar radiation back to space and absorption of solar
10 radiation determine the effects of an aerosol layer on solar radiation.
11 Atmospheric particles greatly complicate projections of future trends in global warming
12 processes because of emissions of greenhouse gases; consequent increases in global mean
13 temperature; resulting changes in regional and local weather patterns; and mainly deleterious
14 (but some beneficial) location-specific human health and environmental effects. The body of
15 available evidence, ranging from satellite to in situ measurements of aerosol effects on radiation
16 receipts and cloud properties, is strongly indicative of an important role in climate for aerosols.
17 This role, however, is poorly quantified. No significant advances have been made in reducing
18 the uncertainties assigned to forcing estimates provided by the IPCC for aerosol-related forcing,
19 especially for black carbon-containing aerosol. The IPCC characterizes the scientific
20 understanding of greenhouse gas-related forcing as "high" in contrast to that for aerosol, which it
21 describes as "low" to "very low."
22 Quantification of the effect of anthropogenic aerosol on hydrological cycles requires more
23 information than is presently available regarding ecosystems responses to reduced solar radiation
24 and other changes occurring in the climate system. However, several global scale studies
25 indicate that aerosol cooling alone can slow down the hydrological cycle, while cooling plus the
26 nucleation of additional cloud droplets can dramatically reduce precipitation rates.
27 In addition to direct climate effects through the scattering and absorption of solar radiation,
28 particles also exert indirect effects on climate by serving as cloud condensation nuclei, thus
29 affecting the abundance and vertical distribution of clouds. The direct and indirect effects of
30 particles appear to have significantly offset global warming effects caused by the buildup of
31 greenhouse gases on a globally-averaged basis. However, because the lifetime of particles is
32 much shorter than that required for complete mixing within the Northern Hemisphere, the
June 2003 9-152 DRAFT-DO NOT QUOTE OR CITE
-------
1 climate effects of particles generally are felt much less homogeneously than are the effects of
2 long-lived greenhouse gases.
3 Any effort to model the impacts of local alterations in particle concentrations on projected
4 global climate change or consequent local and regional weather patterns would be subject to
5 considerable uncertainty.
6 Atmospheric particles also complicate estimation of potential future impacts on human
7 health and the environment projected as possible to occur because of increased transmission of
8 solar ultraviolet radiation (UV-B) through the Earth's atmosphere, secondary to stratospheric
9 ozone depletion due to anthropogenic emissions of chlorofluorcarbons (CFCs), halons, and
10 certain other gases. The transmission of solar UV-B radiation is affected strongly by
11 atmospheric particles. Measured attenuations of UV-B under hazy conditions range up to 37%
12 of the incoming solar radiation. Measurements relating variations in PM mass directly to UV-B
13 transmission are lacking. Particles also can affect the rates of photochemical reactions occurring
14 in the atmosphere, e.g., those involved in catalyzing tropospheric ozone formation. Depending
15 on the amount of absorbing substances in the particles, photolysis rates either can be increased or
16 decreased. Thus, atmospheric particle effects on UV-B radiation, which vary depending on size
17 and composition of particles, can differ substantially over different geographic areas and from
18 season to season over the same area. Any projection of effects of location-specific airborne PM
19 alterations on increased atmospheric transmission of solar UV radiation (and associated potential
20 human health or environmental effects) due to stratospheric ozone-depletion would, therefore,
21 also be subject to considerable uncertainty.
22
June 2003 9-153 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
3 suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
4 Adventists. Environ. Health Perspect. 94: 43-50.
5 Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
6 ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
7 nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
8 paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
9 7: 19-34.
10 Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
11 2.5 microns in aerodynamic diameter (PM2.5) from airport visibility data in California. J. Exposure Anal.
12 Environ. Epidemiol. 5: 161-180.
13 Abbey, D. E.; Ostro, B. E.; Petersen, F.; Burchette, R. J. (1995c) Chronic respiratory symptoms associated with
14 estimated long-term ambient concentrations of fine particulates less than 2.5 microns in aerodynamic
15 diameter (PM2 5) and other air pollutants. J. Exposure Anal. Environ. Epidemiol. 5: 137-159.
16 Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
17 paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
18 158:289-298.
19 Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
20 Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
21 Care Med. 159: 373-382.
22 Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
23 Survey, 1996. Vital Health Stat. 10(200).
24 American Conference of Governmental Industrial Hygienists. (1994) Appendix D: particle size-selective sampling
25 criteria for airborne paniculate matter. In: 1994-1995 Threshold limit values for chemical substances and
26 physical agents and biological exposure indices. Cincinnati, OH: American Conference of Governmental
27 Industrial Hygienists; pp. 43-46.
28 Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Ponce de Leon, A. (1999)
29 Short-term associations between emergency hospital admissions for respiratory and cardiovascular disease
3 0 and outdoor air pollution in London. Arch. Environ. Health 54: 398-411.
31 Avol, E. L.; Gauderman, W. J.; Tan, S. M.; London, S. J.; Peters, J. M. (2001) Respiratory effects of relocating to
32 areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
3 3 Bates, D. V. (2000) Lines that connect: assessing the causality inference in the case of paniculate pollution.
34 Environ. Health Perspect. 108: 91-92.
35 Becker, S.; Soukup, J. M.; Gilmour, M. I.; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
36 by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
37 Pharmacol. 141:637-648.
38 Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
39 and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
40 Binkova, B.; Vesely, D.; Vesela, D.; Jelinek, R.; Sram, R. J. (1999) Genetoxicity and embryotoxicity of urban air
41 paniculate matter collected during winter and summer pertod in two different districts of the Czech Republic.
42 Mutat. Res. 440: 45-58.
43 Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
44 Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
45 Bormann, F. H.; Likens, G. E. (1979) Catastrophic disturbance and the steady state in northern hardwood forests.
46 Am. Sci. 67: 660-669.
47 Brook, R. D.; Brook, J. R.; Urch, B.; Vincent, R.; Rajagopalan, S.; Silverman, F. (2002) Inhalation of fine paniculate
48 air pollution and ozone causes acute arterial vasoconstriction in health adults. Circulation 105: 1534-1536.
49 Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
50 Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian
51 cities. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
52 Health Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
53 Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M.; Dann,
54 T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of respiratory
55 admissions to Ontario hospitals. Environ. Res. 65: 172-194.
June 2003 9-154 DRAFT-DO NOT QUOTE OR CITE
-------
1 Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
2 paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
3 142: 15-22.
4 Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
5 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
6 Environ. Health Perspect. 105: 614-620.
7 Burnett, R. T.; Smith-Doiron, M; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
8 pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
9 Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
10 (2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
11 in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12 12(suppl. 4): 15-39.
13 Castillejos, M.; Borja-Aburto, V H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
14 mortality. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the third colloquium on paniculate air
15 pollution and human health (first special issue); June, 1999; Durham, NC. Inhalation Toxicol.
16 12(suppl. 1): 61-72.
17 Chapin, F. S., Ill; Ruess, R. W. (2001) The roots of the matter. Nature (London) 411: 749, 751-752.
18 Chapman, R. S.; Watkinson, W. P.; Dreher, K. L.; Costa, D. L. (1997) Ambient paniculate matter and respiratory
19 and cardiovascular illness in adults: particle-borne transition metals and the heart-lung axis. Environ. Toxicol.
20 Pharmacol. 4:331-338.
21 Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
22 pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
23 Chorqzy, M.; Szeliga, J.; Strozyk, M.; Cimander, B. (1994) Ambient air pollutants in Upper Silesia: partial chemical
24 composition and biological activity. In: Symposium of risk assessment of urban air: emissions, exposure, risk
25 identification, and risk quantitation; May-June 1992; Stockholm, Sweden. Environ. Health Perspect.
26 102(suppl4):61-66.
27 Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
28 pollution. Arch. Environ. Health 52: 113-117.
29 Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
30 coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
31 50:1287-1298.
32 Claiborn, C. S.; Finn, D.; Larson, T. V.; Koenig, J. Q. (2000) Windblown dust contributes to high PM2 s
33 concentrations. J. Air Waste Manage. Assoc. 50: 1440-1445.
34 Clarke, R. W.; Antonini, J. M.; Hemenway, D. R.; Frank, R.; Kleeberger, S. R.; Jakab, G. J. (2000) Inhaled
3 5 particle-bound sulfate: effects on pulmonary inflammatory responses and alveolar macrophage function.
36 Inhalation Toxicol. 12: 169-186.
37 Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
3 8 Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
3 9 Code of Federal Regulations. (200 la) Appendix J to Part 50-reference method for the determination of paniculate
40 matter as PM10 in the atmosphere. C. F. R. 40: §50.
41 Code of Federal Regulations. (200 Ib) Appendix L to Part 50-reference method for the determination of fine
42 paniculate matter as PM2.5 in the atmosphere. C. F. R. 40: §50.
43 Costa, D. L.; Dreher, K. L. (1999) What do we need to know about airborne particles to make effective risk
44 management decisions? A toxicology perspective. Hum. Ecol. Risk Assess. 5: 481-491.
45 Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
46 curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
47 152:397-406.
48 Dejmek, J.; Selevan, S. G.; Benes, I.; Solansky, I.; Sram, R. J. (1999) Fetal growth and maternal exposure to
49 paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
50 Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
51 on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
52 155: 568-576.
53 Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998) Symptoms in pediatric asthmatics and air pollution:
54 differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging time.
55 Environ. Health Perspect. 106: 751-761.
56 Dockery, D. W.; Spengler, J. D. (1981) Personal exposure to respirable particulates and sulfates. J. AirPollut.
57 Control Assoc. 31: 153-159.
June 2003 9-155 DRAFT-DO NOT QUOTE OR CITE
-------
1 Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
2 inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
3 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.
4 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 329: 1753-1759.
5 Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
6 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
7 symptoms. Environ. Health Perspect. 104: 500-505.
8 Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
9 saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
10 Health Effects Institute; research report no. 83.
11 Dominici, F.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2002) Air pollution and mortality: estimating regional and
12 national dose-response relationships. J. Am. Stat. Assoc. 97: 100-111.
13 Dominici, F.; Daniels, M.; McDermott, A.; Zeger, S. L.; Samet, J. M. (2003) Shape of the exposure-response
14 relation and mortality displacement in the NMMAPS database. In: Revised analyses of time-series studies of
15 air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 91-96. Available:
16 http://www.healtheffects.org/news.htm [16 May, 2003].
17 Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor "alpha" expression from
18 alveolar macrophages treated with urban air particles. Exp. Lung Res. 22: 577-592.
19 Dreher, K. L. (2000) Paniculate matter physicochemistry and toxicology: in search of causality—a critical
20 perspective. Inhalation Toxicol. 12(suppl. 3): 45-57.
21 Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
22 Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
23 limitations. Toxicol. Sci. 55: 24-35.
24 Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. J.; Costa, D. L.
25 (2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
26 studies in Utah Valley residents. Environ. Health Perspect. 109(suppl. 3): 395-403.
27 Ebelt, S. T.; Petkau, A. J.; Vedal, S.; Fisher, T. V.; Brauer, M. (2000) Exposure of chronic obstructive pulmonary
28 disease patients to paniculate matter: relationships between personal and ambient air concentrations. J. Air
29 Waste Manage. Assoc. 50: 1081-1094.
30 Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000) Endotoxin priming affects the lung
31 response to ultrafine particles and ozone in young and old rats. In: Inhalation Toxicology: proceedings of the
32 third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC. Inhalation
33 Toxicology 12(suppl. 1): 85-98.
34 Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
35 Perspect. 107: 637-641.
36 Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
3 7 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp.
38 97-106. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
39 Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B. K.; Van Winkle, L. S.; Evans, M. J.; Plopper, C. G. (2000)
40 Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
41 rhesus monkeys. Am. J. Respir. Crit. Care Med. 161: A615.
42 Federal Register. (1971) National primary and secondary ambient air quality standards. F. R. (April 30)
43 36:8186-8201.
44 Federal Register. (1987) Revisions to the national ambient air quality standards for paniculate matter. F. R. (July 1)
45 52:24,634-24,669.
46 Federal Register. (1996) National ambient air quality standards for paniculate matter; proposed rule. F. R.
47 (December 13) 61: 65,638-65,713.
48 Federal Register. (1997) National ambient air quality standards for paniculate matter; final rule. F. R. (July 18)
49 62: 38,652-38,752.
50 Fortoul, T. L; Osorio, L. S.; Tovar, A. T.; Salazar, D.; Castilla, M. E.; Olaiz-Fernandez, G. (1996) Metals in lung
51 tissue from autopsy cases in Mexico City residents: comparison of cases from the 1950s and the 1980s.
52 Environ. Health Perspect. 104: 630-632.
53 Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
54 extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol.
55 277: L960-L967.
56 Galloway, J. N.; Cowling, E. B. (2002) Reactive nitrogen and the world: 200 years of change. Ambio 31: 64-71.
June 2003 9-156 DRAFT-DO NOT QUOTE OR CITE
-------
1 Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
2 1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
3 June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98-MP26.03.
4 Garner, J. H. B. (1994) Nitrogen oxides, plant metabolism, and forest ecosystem response. In: Alscher, R. G.;
5 Wellburn, A. R., eds. Plant responses to the gaseous environment: molecular, metabolic and physiological
6 aspects, [3rd international symposium on air pollutants and plant metabolism]; June 1992; Blacksburg, VA.
7 London, United Kingdom: Chapman & Hall; pp. 301-314.
8 Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
9 Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
10 function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
11 Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
12 Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
13 lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
14 Med. 166: 76-84.
15 Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
16 Am. J. Respir. Crit. Care Med. 164: 704-708.
17 Ohio, A. J.; Kim, C.; Devlin, R. B. (2000) Concentrated ambient air particles induce mild pulmonary inflammation
18 in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
19 Ohio, A. J.; Gilbey, J. G.; Roggli, V. L.; Richards, J. H.; McGee, J. K.; Carson, J. L.; Devlin, R. B.; Cascio, W. E.
20 (2001) Diffuse alveolar damage after exposure to an oil fly ash. Am. J. Respir. Crit. Care Med. 164:
21 1514-1518.
22 Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
23 summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
24 155:2105-2108.
25 Godleski, J. J.; Sioutas, C.; Katler, M.; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
26 particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
27 Gold, D. R.; Damokosh, A. I.; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
28 Castillejos, M. (1999) Paniculate and ozone pollutant effects on the respiratory function of children in
29 southwest Mexico City. Epidemiology 10: 8-16.
30 Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
31 Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
32 Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
33 Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
34 short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
3 5 Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm
36 [15 February, 2001].
37 Guderian, R.; Tingey, D. T.; Rabe, R. (1985) Effects of photochemical oxidants on plants. In: Guderian, R., ed. Air
3 8 pollution by photochemical oxidants: formation, transport, control, and effects on plants. Berlin, Federal
39 Republic of Germany: Springer-Verlag; pp. 127-333. (Billings, W. D.; Golley, F.; Lange, O. L.; Olson, J. S.;
40 Remmert, H., eds. Ecological studies: analysis and synthesis, v. 52).
41 Harwell, M. A.; Myers, V; Young, T.; Bartuska, A.; Gassman, N.; Gentile, J. H.; Harwell, C. C.; Appelbaum, S.;
42 Barko, J.; Causey, B.; Johnson, C.; McLean, A.; Smola, R.; Templet, P.; Tosini, S. (1999) A framework for
43 an ecosystem integrity report card. BioScience 49: 543-556.
44 Health Effects Institute. (2002) Reanalysis of National Morbidity, Mortality, and Air Pollution Study (NMMAPS)
45 [letter with attachments]. Boston, MA: May 30.
46 Health Effects Institute. (2003) Revised analyses of time-series studies of air pollution and health. Boston, MA:
47 Health Effects Institute; special report.
48 Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
49 bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
50 Howel, D.; Darnell, R.; Pless-Mulloli, T. (2001) Children's respiratory health and daily paniculate levels in 10
51 nonurban communities. Environ. Res. 87: 1-9.
52 Hoyert, D. L.; Kochanek, K. D.; Murphy, S. L. (1999) Deaths: final data for 1997. Atlanta, GA: U.S. Centers for
53 Disease Control and Prevention, National Center for Health Statistics. (National vital statistics reports: v. 47,
54 no. 19).
55 Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
56 Inhalation Toxicol. 10: 121-130.
June 2003 9-157 DRAFT-DO NOT QUOTE OR CITE
-------
1 Imrich, A.; Ning, Y. Y.; Koziel, H.; Coull, B.; Kobzik, L. (1999) Lipopolysaccharide priming amplifies lung
2 macrophage tumor necrosis factor production in response to air particles. Toxicol. Appl. Pnarmacol.
3 159: 117-124.
4 International Commission on Radiological Protection. (1994) Human respiratory tract model for radiological
5 protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
6 United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP: v. 24, pp. 1-482).
7 Ito, K. (2003) Associations of paniculate matter components with daily mortality and morbidity in Detroit,
8 Michigan. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
9 Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
10 Ito, K.; Thurston, G. D. (1996) Daily PMj/mortality associations: an investigation of at-risk subpopulations.
11 J. Exposure Anal. Environ. Epidemiol. 6: 79-95.
12 Izzotti, A.; Camoirano, A.; DAgostini, F.; Sciacca, S.; De Naro Papa, F.; Cesarone, C. F.; De Flora, S. (1996)
13 Biomarker alterations produced in rat lung by intratracheal instillations of air paniculate extracts and
14 chemoprevention with oral N-acetylcysteine. Cancer Res. 56: 1533-1538.
15 Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
16 1983-1992. Environ. HealthPerspect. 105: 980-985.
17 Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
18 acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
19 Janssen, N. A. H.; Schwartz, J.; Zanobetti, A.; Suh, H. H. (2002) Air conditioning and source-specific particles as
20 modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ. Health Perspect.
21 110:43-49.
22 Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM10 associations in Los Angeles.
23 In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air pollution and human
24 mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 59-69.
25 Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
26 and daily mortality statistical modeling-interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
27 Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
28 of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp.
29 165-172. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
30 Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
31 specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
32 50: 1215-1222.
33 Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Watkinson, W. P.; Campen, M. J.; Winsett, D. W.;
34 Richards, J. R.; Crissman, K. M.; Hatch, G. E.; Costa, D. L. (2000) The spontaneously hypertensive rat as a
3 5 model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress
36 from inhaled emission particulate matter. Toxicol. Appl. Pharmacol. 164: 250-263.
37 Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
38 W. H.. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of
39 particulate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
40 Project. Cambridge, MA: Health Effects Institute.
41 Ktinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared
42 to ecologic studies. Environ. Health Perspect. 105: 1078-1083.
43 Ktinzli, N.; Medina, S.; Kaiser, R.; Quenel, P.; Horak, F., Jr.; Studnicka, M. (2001) Assessment of deaths
44 attributable to air pollution: should we use risk estimates based on time series or on cohort studies? Am. J.
45 Epidemiol. 153: 1050-1055.
46 Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
47 experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
48 Perspect. 105: 1234-1237.
49 Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine particulate matter from different
50 sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
51 Lambert, A. L.; Dong, W.; Selgrade, M. K.; Gilmour, M. I. (2000) Enhanced allergic sensitizationby residual oil fly
52 ash particles is mediated by soluble metal constituents. Toxicol. Appl. Pharmacol. 165: 84-93.
53 Landis, M. S.; Norris, G. A.; Williams, R. W.; Weinstein, J. P. (2001) Personal exposures to PM2 5 mass and trace
54 elements in Baltimore, MD, USA. Atmos. Environ. 35: 6511-6524.
55 Lawrence, L.; Hall, M. J. (1999) 1997 summary: national hospital discharge survey. Hyattsville, MD: U.S.
56 Department of Health & Human Resources, Centers for Disease Control and Prevention, National Center for
57 Health Statistics. (Advance data from vital and health statistics: no. 308).
June 2003 9-158 DRAFT-DO NOT QUOTE OR CITE
-------
1 Layton, D. W. (1993) Metabolically consistent breathing rates for use in dose assessments. Health Phys. 64: 23-36.
2 Levy, D. (1998) Fine paniculate air pollution and out-of-hospital mortality in King County, Washington. In: Vostal,
3 J. J., ed. Health effects of paniculate matter in ambient air. Proceedings of an international conference; 1997;
4 Prague, Czech Republic. Pittsburgh, PA: Air & Waste Management Association; pp. 262-271. (A&WMA
5 publication VIP-80).
6 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
7 paniculate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international
8 meeting on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake
9 Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
10 Li, X. Y.; Brown, D.; Smith, S.; MacNee, W.; Donaldson, K. (1999) Short-term inflammatory responses following
11 intratracheal instillation of fine and ultrafine carbon black in rats. Inhalation Toxicol. 11: 709-731.
12 Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
13 pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
14 Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
15 mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
16 Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
17 admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
18 Lioy, P. J. (1990) Assessing total human exposure to contaminants: a multidisciplinary approach. Environ. Sci.
19 Technol. 24: 938-945.
20 Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
21 size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
22 Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000b) Infant mortality and air pollution: a comprehensive analysis of U.S.
23 data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
24 Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
25 mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report
26 no. 95.
27 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
28 California. Environ. Health Perspect. 105: 216-222.
29 Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
30 insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
31 Long, C. M.; Suh, H. H.; Koutrakis, P. (2000) Characterization of indoor particle sources using continuous mass and
32 size monitors. J. Air Waste Manage. Assoc. 50: 1236-1250.
33 Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
34 mortality in Mexico City. Epidemiology 10: 118-123.
35 Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
36 J. R. Stat. Soc. B 61(part 2): 459-477.
37 Lundgren, D. A.; Burton, R. M. (1995) Effect of particle size distribution on the cut point between fine and coarse
38 ambient mass fractions. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air
39 pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 131-148.
40 Makela, J. M.; Aalto, P.; Jokinen, V.; Pohja, T.; Nissinen, A.; Palmroth, S.; Markkanen, T.; Seitsonen, K.;
41 Lihavainen, H.; Kulmala, M. (1997) Observations of ultrafine aerosol particle formation and growth in boreal
42 forest. Geophys. Res. Lett. 24: 1219-1222.
43 Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
44 Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
45 Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003) Air pollution and cardiovascular
46 mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
47 Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
48 http://www.healtheffects.org/news.htm [16 May, 2003].
49 McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
50 Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
51 California children with asthma. Environ. Health Perspect. 107: 757-760.
52 McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of
53 mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers.
54 J. Exposure Anal. Environ. Epidemiol. 10: 427-436.
55 Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
56 relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
June 2003 9-159 DRAFT-DO NOT QUOTE OR CITE
-------
1 Monn, C. (2001) Exposure assessment of air pollutants: a review on spatial heterogeneity and
2 indoor/outdoor/personal exposure to suspended paniculate matter, nitrogen dioxide and ozone. Atmos.
3 Environ. 35: 1-32.
4 Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
5 exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
6 155:245-252.
7 Moolgavkar, S. H. (2000a) Air pollution and daily mortality in three U.S. counties. Environ. Health Perspect.
8 108:777-784.
9 Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
10 metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
11 Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
12 metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
13 Toxicol. 12(suppl. 4): 75-90.
14 Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
15 In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
16 Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
17 Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
18 evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
19 Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
20 Healthy women's PEF variations with ambient summer concentrations of PM10, PN2 5, SO42_, H+, and O3.
21 Am. J. Respir. Crit. Care Med. 160: 117-125.
22 National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
23 Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
24 publication no. 97-4051.
25 National Research Council. (1983) Risk assessment in the federal government: managing the process. Washington,
26 DC: National Academy Press. Available: http://books.nap.edu/books/PODl 15/html/index.html (3 June 2003).
27 National Research Council. (1994) Science and judgment in risk assessment. Washington, DC: National Academy
28 Press. Available: http://www.nap.edu/books/030904894X/html/ (4 June 2003).
29 National Research Council. (1998) Research priorities for airborne paniculate matter. I. Immediate priorities and a
30 long-range research portfolio. Washington, DC: National Academy Press.
31 National Research Council. (2001) Research priorities for airborne paniculate matter. III. Early research progress.
32 Washington, DC: National Academy Press. Available: http://www.nap.edu/books/0309073375/html/ (4 June
33 2003).
34 National Science and Technology Council. (1998) National acid precipitation assessment program biennial report to
35 Congress: an integrated assessment; executive summary. Washington, DC: National Science and Technology
36 Council, Committee on Environment and Natural Resources. Available:
37 www.nnic.noaa.gov/CENR/NAPAP/NAPAP_96.htm [2003 June 25].
3 8 Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
39 and exposure to air pollution. Public Health Rep. 114: 135-148.
40 Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
41 indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
42 139: 1088-1099.
43 Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
44 pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
45 141: 111-122.
46 Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
47 determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
48 Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
49 Philadelphia. Environ. Health Perspect. 107: 629-631.
5 0 Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
51 paniculate material in rat and human lungs. Presented at: PM2000: paniculate matter and health—the
52 scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
53 Pittsburgh, PA: Air & Waste Management Association.
54 Oberdorster, G.; Ferin, J.; Lehnert, B. E. (1994) Correlation between particle size, in vivo particle persistence, and
55 lung injury. Environ. Health Perspect. 102(suppl. 5): 173-179.
56 Odum, E. P. (1989) Ecology and our endangered life-support systems. Sunderland, MA: Sinauer Associates, Inc.;
57 pp. 58-66, 108-118.
June 2003 9-160 DRAFT-DO NOT QUOTE OR CITE
-------
1 Osier, M; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
2 Fundam. Appl. Toxicol. 40: 220-227.
3 Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
4 70:98-104.
5 Ostro, B. D.; Lipsett, M. I; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
6 Am. J. Public Health 81: 694-702.
7 Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Krupnick, A.; Harrington, W. (1993) Air pollution and respiratory
8 morbidity among adults in Southern California. Am. J. Epidemiol. 137: 691-700.
9 Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
10 exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
11 Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
12 January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
13 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
14 Valley, California: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10: 412-419.
15 Ott, W.; Wallace, L.; Mage, D. (2000) Predicting paniculate (PM10) personal exposure distributions using a random
16 component superposition statistical model. J. Air Waste Manage. Assoc. 50: 1390-1406.
17 Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
18 pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
19 Environmental Health; March 45.
20 Patterson, E.; Eatough, D. J. (2000) Indoor/outdoor relationships for ambient PM25 and associated pollutants:
21 epidemiological implications in Lindon, Utah. J. Air Waste Manage. Assoc. 50: 103-110.
22 Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
23 in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
24 Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
25 episode: a link to mortality? Lancet 349: 1582-1587.
26 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b) Respiratory effects are associated with the
27 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
28 Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c) Short-term effects of paniculate air pollution on
29 respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
30 Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
31 Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities with
32 differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
33 Med. 159: 760-767.
34 Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
35 Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
3 6 differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
37 159:768-775.
38 Peters, A.; Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Heinnch, J. (1999c) Akute Auswirkung der
3 9 Schwebstaubkonzentrationen in der Aulknluft auf die Mortalitat - Vergleichsstudie Nordost-Bayern /
40 Nordbohmen. AbschluBbericht [Acute effects of suspended particle concentrations in the atmosphere on
41 mortality - a study comparing northeast Bavaria and north Bohemia. Final report]. Bavaria, Federal Republic
42 of Germany: Institut fur Epidemiologie; report no. GSF-EP S 1/99.
43 Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
44 Monahan, K.; Dockery, D. W. (2000) Air pollution and incidence of cardiac arrhythmia. Epidemiology
45 11: 11-17.
46 Pinto, J. P.; Pleasant, M.; Kellogg, R.; Torok, Sz.; Stiles, D.; Willis, R. D.; Koutrakis, P.; Allen, G. A. (1995)
47 Chemical characterization of the ambient aerosol in Philadelphia. Presented at: Paniculate matter: health and
48 regulatory issues: an A&WMA international specialty conference; April; Pittsburgh, PA. Pittsburgh, PA: Air
49 & Waste Management Association; submitted abstract.
50 Pinto, J. P.; Lefohn, A. S.; Shadwick, D. S. (2003) Aspects of the spatial variability of PM2.5 concentrations within
51 urban areas of the United States. J. Air Waste Manage. Assoc.: submitted.
52 Plopper, C. G.; Fanucchi, M. V. (2000) Do urban environmental pollutants exacerbate childhood lung diseases?
53 Environ. Health Perspect. 108: A252-A253.
54 Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
55 Am. J. Public Health 79: 623-628.
56 Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM10 pollution:
57 a daily time series analysis. Am. Rev. Respir. Dis. 144: 668-674.
June 2003 9-161 DRAFT-DO NOT QUOTE OR CITE
-------
1 Pope, C. A., Ill; Schwartz, I; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah valley. Arch.
2 Environ. Health 47: 211-217.
3 Pope, C. A., Ill; Thun, M. I; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
4 (1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J. Respir.
5 Crit. Care Med. 151: 669-674.
6 Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
7 Front. Environ. Health Perspect: 107: 567-573.
8 Pope, C. A.; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
9 G. M.; Gold, D. R.; Dockery, D. W. (1999b) Heart rate variability associated with paniculate air pollution.
10 Am. Heart J. 138: 890-899.
11 Pope, C. A., Ill; Eatough, D. J.; Gold, D. R.; Pang, Y.; Nielsen, K. R.; Nath, P.; Verrier, R. L.; Kanner, R. E. (2001)
12 Acute exposure to environmental tobacco smoke and heart rate variability. Environ. Health Perspect.
13 109:711-716.
14 Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
15 cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
16 Assoc. 287: 1132-1141.
17 Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
18 cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
19 Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
20 F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
21 Perspect. 104: 506-514.
22 Ramadan, Z.; Song, X.-H.; Hopke, P. K. (2000) Identification of sources of Phoenix aerosol by positive matrix
23 factorization. J. Air Waste Manage. Assoc. 50: 1308-1320.
24 Ritz, B.; Yu, F.; Fruin, S.; Chapa, G.; Shaw, G. M.; Harris, J. A. (2002) Ambient air pollution and risk of birth
25 defects in southern California. Am. J. Epidemiol. 155: 17-25.
26 Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
27 on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
28 154:300-307.
29 Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
30 intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
31 Mexico City. Arch. Environ. Health 52: 368-376.
32 Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and air
33 pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute; research
34 report no. 94.
35 Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
36 (2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
37 pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
38 Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
39 exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
40 Sarnat, J. A.; Schwartz, J.; Catalano, P. J.; Suh, H. H. (2001) Gaseous pollutants in paniculate matter epidemiology:
41 confounders or surrogates? Environ. Health Perspect. 109: 1053-1061.
42 Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
43 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
44 Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137: 1136-1147.
45 Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
46 10: 17-22.
47 Schwartz, J. (2000a) The distributed lag between air pollution and daily deaths. Epidemiology 11: 320-326.
48 Schwartz, J. (2000b) Harvesting and long term exposure effects in the relation between air pollution and mortality.
49 Am. J. Epidemiol. 151: 440-448.
50 Schwartz, J. (2003) Daily deaths associated with air pollution in six US cities and short-term mortality displacement
51 in Boston. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
52 Health Effects Institute; pp. 219-226. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
53 Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
54 Michigan. Am. J. Epidemiol. 142: 23-35.
55 Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
56 respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
June 2003 9-162 DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
2 with application to air pollution and daily death. Epidemiology 11: 666-672.
3 Schwartz, J.; Dockery, D. W.; Neas, L. M; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
4 Ferris, E.G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
5 Am. J. Respir. Crit. Care Med. 150: 1234-1242.
6 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles? J. Air
7 Waste Manage. Assoc. 46: 927-939.
8 Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
9 concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
10 Science Advisory Board. (2002) A framework for assessing and reporting on ecological condition: an SAB report.
11 Young, T. F.; Sanzone, S., eds. Washington, DC: U.S. Environmental Protection Agency, Ecological
12 Processes and Effects Committee; report no. EPA-SAB-EPEC-02-009. Available at:
13 http://www.epa.gov/sciencel/index.html (27 January 2003).
14 Sexton, K.; Selevan, S. G.; Wagener, D. K.; Lybarger, J. A. (1992) Estimating human exposures to environmental
15 pollutants: availability and utility of existing databases. Arch. Environ. Health 47: 398-407.
16 Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
17 1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
18 MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
19 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
20 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
21 Smiley-Jewell, S. M.; Liu, F. J.; Weir, A. J.; Plopper, C. G. (2000) Acute injury to differentiating Clara cells in
22 neonatal rabbits results in age-related failure of bronchiolar epithelial repair. Toxicol. Pathol. 28: 267-276.
23 Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
24 coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
25 Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
26 Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
27 John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
28 Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne paniculate matter on daily
29 death counts. Environ. Health Perspect. 103: 490-497.
30 Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
31 hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers.
32 J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
33 Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
34 summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
35 65:271-290.
36 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
37 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
38 Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
39 resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
40 Tolbert, P. E.; Klein, M.; Metzger, K. B.; Peel, J.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
41 Frumkin, H. (2000a) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure
42 Anal. Environ. Epidemiol. 10: 446-460.
43 Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
44 Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000b) Air quality and
45 pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
46 Treshow, M. (1980) Pollution effects on plant distribution. Environ. Conserv. 7: 279-286.
47 Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
48 chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
49 Turpin, B. J. (1999) Options for characterizing organic paniculate matter. Environ. Sci. Technol. 33: 76A-79A.
50 U.S. Code. (1991) Clean Air Act, §108, air quality criteria and control techniques, §109, national ambient air quality
51 standards. U. S. C. 42: §§7408-7409.
52 U.S. Court of Appeals for the District of Columbia. (2002) American Trucking Associations, Inc. vs. U.S.
53 Environmental Protection Agency. 283 F.3d 355, 378-79 (D.C. Cir. 2002).
54 U.S. Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
55 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
56 Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
June 2003 9-163 DRAFT-DO NOT QUOTE OR CITE
-------
1 U.S. Environmental Protection Agency. (1986) Air quality criteria for lead. Research Triangle Park, NC: Office of
2 Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report no.
3 EPA-600/8-83/028aF-dF. 4v. Available from: NTIS, Springfield, VA; PB87-142378.
4 U.S. Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park,
5 NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
6 nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525, and
7 PB95-124517.
8 U.S. Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park,
9 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
10 U.S. Environmental Protection Agency. (1996b) Air quality criteria for ozone and related photochemical oxidants.
11 Research Triangle Park, NC: Office of Research and Development; report nos. EP A/600/ AP-93/004aF-cF. 3v.
12 Available from: NTIS, Springfield, VA; PB96-185582, PB96-185590, and PB96-185608. Available online at:
13 www.epa.gov/ncea/ozone.htm.
14 U.S. Environmental Protection Agency. (2000a) Air quality criteria for carbon monoxide. Research Triangle Park,
15 NC: National Center for Environmental Assessment; report no. EPA/600/P-99/00 IF. Available:
16 www.epa.gov/ncea/co/ [2000, October 6].
17 U.S. Environmental Protection Agency. (2000b) Deposition of air pollutants to the great waters. Third report to
18 Congress. [Executive Summary]. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office
19 of Air Quality Planning and Standards; report no. EPA-453/R-00-005.
20 U.S. Environmental Protection Agency. (2002) Health assessment document for diesel engine exhaust. Washington,
21 DC: Office of Research and Development, National Center for Environmental Assessment; report no.
22 EPA/600/8-90/057F. Available: http://cfpub.epa.gov/ncea/ [22 May, 2003].
23 Undem, B. I; Carr, M. J. (2002) The role of nerves in asthma. Curr. Allergy Asthma Rep. 2: 159-165.
24 Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
25 nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
26 Wallace, L. (2000) Correlations of personal exposure to particles with outdoor air measurements: a review of recent
27 studies. Aerosol Sci. Technol. 32: 15-25.
28 Wallace, L. A.; Emmerich, S. J.; Howard-Reed, C. (2002) Continuous measurements of air change rates in an
29 occupied house for 1 year: the effect of temperature, wind, fans, and windows. J. Exposure Anal. Environ.
30 Epidemiol. 12: 296-306.
31 Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
32 ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
33 Respir. Dis. 133:834-842.
34 Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil fly
35 ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
36 Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Costa, D. L. (2001) Cardiovascular and systemic responses to
3 7 inhaled pollutants in rodents: effects of ozone and paniculate matter. Environ. Health Perspect.
38 109(suppl. 4): 539-546.
39 Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
40 Whitby, K. T.; Sverdrup, G. M. (1980) California aerosols: their physical and chemical characteristics. In: Hidy,
41 G. M.; Mueller, P. K.; Grosjean, D.; Appel, B. R.; Wesolowski, J. J., eds. The character and origins of smog
42 aerosols: a digest of results from the California Aerosol Characterization Experiment (ACHEX). New York,
43 NY: John Wiley & Sons, Inc.; pp. 477-517. (Advances in environmental science and technology: v. 9).
44 Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
45 mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
46 Cambridge, MA: Health Effects Institute; Research Report no. 98.
47 Williams, R.; Suggs, J.; Zweidinger, R.; Evans, G.; Creason, J.; Kwok, R.; Rodes, C.; Lawless, P.; Sheldon, L.
48 (2000a) The 1998 Baltimore Particulate Matter Epidemiology-Exposure Study: part 1. Comparison of
49 ambient, residential outdoor, indoor and apartment particulate matter monitoring. J. Exposure Anal. Environ.
50 Epidemiol. 10: 518-532.
51 Williams, R.; Suggs, J.; Creason, J.; Rodes, C.; Lawless, P.; Kwok, R.; Zweidinger, R.; Sheldon, L. (2000b) The
52 1998 Baltimore Particulate Matter Epidemiology-Exposure Study: part 2. Personal exposure assessment
53 associated with an elderly study population. J. Exposure Anal. Environ. Epidemiol. 10: 533-543.
54 Williams, R. W.; Creason, J.; Zweidinger, R.; Watts, R. R.; Shy, C. (2000c) Indoor, outdoor, and personal exposure
5 5 monitoring of particulate air pollution: the Baltimore elderly epidemiology-exposure pilot study.
56 Atmos. Environ. 34: 4193-4204.
June 2003 9-164 DRAFT-DO NOT QUOTE OR CITE
-------
1 Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
2 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
3 Wilson, W. E.; Mage, D. T.; Grant, L. D. (2000) Estimating separately personal exposure to ambient and
4 nonambient paniculate matter for epidemiology and risk assessment: why and how. J. Air Waste Manage.
5 Assoc. 50: 1167-1183.
6 Winter-Sorkina, R. de; Cassee, F. R. (2002) From concentration to dose: factors influencing airborne paniculate
7 matter deposition in humans and rats. Bilthoven, The Netherlands: National Institute of Public Health and the
8 Environment (RIVM); report no. 650010031/2002. Available:
9 http://www.rivm.nl/bibliotheek/rapporten/650010031.html (13 June 2003).
10 Wong, T. W.; Lau, T. S.; Yu, T. S.; Neller, A.; Wong, S. L.; Tarn, W.; Pang, S. W. (1999) Air pollution and hospital
11 admissions for respiratory and cardiovascular diseases in Hong Kong. Occup. Environ. Med. 56: 679-683.
12 Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal infant
13 mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
14 Woodwell, G. M. (1970) Effects of pollution on the structure and physiology of ecosystems: changes in natural
15 ecosystems caused by many different types of disturbances are similar and predictable. Science (Washington,
16 DC) 168: 429-433.
17 Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
18 J. Occup. Environ. Med. 42: 469-474.
19 Zanobetti, A.; Schwartz, J. (2001) Are diabetics more susceptible to the health effects of airborne particles? Am. J.
20 Respir. Crit. Care Med. 164: 831-833.
21 Zanobetti, A.; Schwartz, J. (2003) Airborne particles and hospital admissions for heart and lung disease. In: Revised
22 analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
23 Institute; pp. 241-248. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
24 Zanobetti, A.; Schwartz, J.; Gold, D. (2000) Are there sensitive subgroups forthe effects of airborne particles?
25 Environ. Health Perspect. 108: 841-845.
26 Zanobetti, A.; Schwartz, J.; Samoli, E.; Gryparis, A.; Touloumi, G.; Atkinson, R.; Le Tertre, A.; Bobros, J.;
27 Celko, M.; Goren, A.; Forsberg, B.; Michelozzi, P.; Rabczenko, D.; Ruiz, E. A.; Katsouyanni, K. (2002)
28 The temporal pattern of mortality responses to air pollution: a multicity assessment of mortality displacement.
29 Epidemiology 13: 87-93.
30 Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
31 measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
32 Perspect. 108: 419-426.
3 3 Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
34 normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
35 Zweidinger, R.; Purdue, L.; Fitzgerald, K.; Carmichael, L.; Kellog, R. (1998) NERL PM research monitoring
36 platforms: Baltimore, Fresno and Phoenix, data report. Research Triangle Park, NC: U.S. Environmental
37 Protection Agency, National Exposure Research Laboratory; report no. EPA/600/R-98/138. Available from:
3 8 NTIS, Springfield, VA; PB99-105660.
39
June 2003 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
June 2003 9A-1 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-1. EFFECT ESTIMATES PER SO-jig/m3 INCREASE
IN 24-HOUR PM™ CONCENTRATIONS FROM U.S. AND CANADIAN STUDIES
Study Location
RR (±CI)
OnlyPM
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, WA
New Haven, CF
Cleveland, OHk
Spokane, WA1
COPD
Minneapolis, MN11
Birmingham, ALm
Spokane, WA1
Detroit, MI°
1.04 (0.98,
1.06(1.04,
0.98 (0.90,
1.03 (1.00,
1.05 (1.00,
1.05 (1.00,
1.08(1.01,
1.09 (0.94,
1.04(1.00,
1.03 (1.02,
1.08(1.05,
1.05(1.01,
1.03 (1.00,
> 65 years)
1.23 (1.02,
1.10(1.03,
1.06(1.00,
1.06(1.00,
1.08(1.04,
1.25(1.10,
1.13(1.04,
1.17(1.08,
1.10(1.02,
1.09)
1.09)
1.05)
1.05)
1.09)
1.08)
1.12)
1.25)
1.08)
1.04)
1.11)
1.10)
1.055)
1.43)*
1.17)
1.13)
1.11)
1.14)
1.44)
1.22)
1.27)
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)
June 2003
9A-2
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-1 (cont'd). EFFECT ESTIMATES PER SO-jig/m3 INCREASE
IN 24-HOUR PM™ CONCENTRATIONS FROM U.S. AND CANADIAN STUDIES
Study Location
Pneumonia
Minneapolis, MM11
Birmingham, ALm
Spokane, WA1
Detroit, MI°
Ischemic HD
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)^
— 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 Cities'
Utah Valley, UT
Utah Valley, UTS
Cough
Denver, COX
Six Cities'
Utah Valley, UTS
Decrease in Lung Function
Utah Valley, UT
Utah Valley, UTS
Utah Valley, UT1*
2.03 (1.36, 3.04)
1.28 (1.06, 1.56)T
1.01 (0.81, 1.27)"
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.
cDockeryetal. (1992)/O3.
dSchwartz (1993).
10 and reported.
glto and Thurston (1996)/O3.
•Kinney et al. (1995)/O3, CO.
hStyeretal. (1995).
'Thurstonetal. (1994)/O3.
JSchwartz (1995)/SO2.
kSchwartz et al. (1996b).
years.
'Schwartz (1996).
mSchwartz (1994a).
"Schwartz (1994b).
"Schwartz (1994c).
xOstro etal. (1991).
tMin/Max 24-h PM10 in parentheses unless
noted otherwise as standard deviation
90 percentile (10, 90). NR = not(±SD),
pSchwartz and Morris (1995)/O3, CO, SO2.
"Children.
'Schwartzetal. (1994).
Tope etal. (1991).
Tope andDockery (1992).
'Schwartz (1994d).
"Pope and Kanner (1993).
11 Asthmatic children and adults.
'Means of several cities.
"PEFR decrease in mL/s.
"TEVj decrease.
*RR refers to total population, not just >65
June 2003
9A-2
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 9A-2. EFFECT ESTIMATES PER VARIABLE INCREMENTS IN 24-HOUR
CONCENTRATIONS OF FINE PARTICLE INDICATORS (PM25, SO^, H+)
FROM U.S. AND CANADIAN STUDIES
Acute Mortality
Six Citya
Portage, WI
Topeka, KS
Boston, MA
St. Louis, MO
Kingston/Knoxville, TN
Steubenville, OH
Indicator
PM25
PM25
PM25
PM25
PM25
PM,,
RR (±CI) per 25 ug/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)
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)
Increased Hospitalization
Ontario, Canadab
Ontario, Canada0
NYC/Buffalo, NYd
so;
so;
03
so;
Toronto* H+ (Nmol/m3)
so;
PM,,
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)
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)
Increased Respiratory Symptoms
Southern Californiaf
Six Cities8
(Cough)
Six Cities8
(Lower Resp. Symp.)
so;
PM25
PM2 5 Sulfur
H+
PM25
PM2 5 Sulfur
H+
1.48(1.14, 1.91)
1.19(1.01, 1.42)"
1.23 (0.95, 1.59)"
1.06 (0.87, 1.29)"
1.44(1.15-1.82)"
1.82 (1.28-2.59)"
1.05 (0.25-1.30)"
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)"*
Decreased Lung Function
Uniontown, PAe
PM25
PEFR23.1 (-0.3, 36.9) (per 25 ug/m3)
25/88 (NR/88)
References:
"Schwartz etal. (1996a).
bBurnettetal. (1994).
cBurnettetal. (1995) O3.
dThurston et al. (1992, 1994).
dNeas etal. (1995).
fOstro etal. (1993).
BSchwartzetal. (1994).
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 ug/m3 for PM2 5; per 5 ug/m3 for PM2 5 sulfur;
per 25 nmoles/m3 for H+.
***50th percentile value (10,90 percentile).
June 2003
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
Indicator
Increased Total Chronic Mortality in Adults
Six Cityb
ACS Study0
(151 U.S. SMSA)
Increased Bronchitis
Six Cityd
Six City6
24 Cityf
24 Cityf
24 Cityf
24 Cityf
Southern California8
PM15/10
PM25
so;
PM25
so;
in Children
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 (ug/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 Cityd'h
Six City6
24 Citylj
24 City1
24 City1
24 City1
PM15/10
TSP
H+ (52 nmoles/m3)
PM2 , (15 ug/m3)
SO; (7 ug/m3)
PM10 (17 ug/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-ug/m3 increase for TSP; a 50-ug/m3
increase for PM10 and PM15; a 25-ug/m3 increase for PM2 5; and a 15-ug/m3 increase for SO;, except where
noted otherwise; a 100-nmole/m3 increase for H+.
bDockery et al. (1993).
Tope etal. (1995).
dDockery et al. (1989).
eWareetal. (1986).
TJockery et al. (1996).
gAbbeyetal. (1995).
hNS Changes = No significant changes.
'Raizenne et al. (1996).
JPollutant data same as for Dockery et al. (1996).
June 2003
9A-5
DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Abbey, D. E.; Ostro, B. E.; Petersen, F.; Burchette, R. J. (1995) Chronic respiratory symptoms associated with
3 estimated long-term ambient concentrations of fine particulates less than 2.5 microns in aerodynamic
4 diameter (PM2 5) and other air pollutants. J. Exposure Anal. Environ. Epidemiol. 5: 137-159.
5 Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
6 paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
7 158:289-298.
8 Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
9 Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
10 Care Med. 159: 373-382.
11 Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
12 Survey, 1996. Vital Health Stat. 10(200).
13 Bigger, J. T., Jr.; Fleiss, J. L.; Steinman, R. C.; Rolnitzky, L. M.; Kleiger, R. E.; Rottman, J. N. (1992) Frequency
14 domain measures of heart period variability and mortality after myocardial infarction. Circulation
15 85: 164-171.
16 Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
17 CzechRepublic 1986-8. Occup. Environ. Med. 56: 539-543.
18 Broday, D. M.; Georgopoulos, P. G. (2000) Growth and deposition of hygroscopic paniculate matter in the human
19 lungs. Aerosol Sci. Technol.: submitted.
20 Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
21 Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M.;
22 Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of respiratory
23 admissions to Ontario hospitals. Environ. Res. 65: 172-194.
24 Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
25 paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
26 142: 15-22.
27 Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
28 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
29 Environ. Health Perspect. 105: 614-620.
30 Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
31 Ozkaynak, H. (1998) The association between ambient carbon monoxide levels and daily mortality in
32 Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
33 Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
34 pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
35 Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
3 6 (2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
37 in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
38 12(suppl. 4): 15-39.
39 Castillejos, M.; Borja-Aburto, V H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
40 mortality. In: Inhalation Toxicology: proceedings of the third colloquium on paniculate air pollution and
41 human health; June, 1999; Durham, NC. Inhalation Toxicology 12(suppl. 1): 61-72.
42 Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
43 chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
44 Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
45 pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
46 Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
47 pollution. Arch. Environ. Health 52: 113-117.
48 Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
49 coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
50 50: 1287-1298.
51 Claiborn, C. S.; Finn, D.; Larson, T. V.; Koenig, J. Q. (2000) Windblown dust contributes to high PM2 s
52 concentrations. J. Air Waste Manage. Assoc. 50: 1440-1445.
53 Clark, W. E.; Whitby, K. T. (1975) Measurements of aerosols produced by the photochemical oxidation of SO2 in
54 air. J. Colloid Interface Sci. 51: 477-490.
55 Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
56 Arizona. J. Exposure Anal. Environ. Epidemiol.: submitted.
June 2003 9A-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 Code of Federal Regulations. (1991a) Appendix J to Part 50-reference method for the determination of paniculate
2 matter as PM10 in the atmosphere. C. F. R. 40: §50.
3 Code of Federal Regulations. (1991b) Ambient air monitoring reference and equivalent methods. C. F. R. 40: §53.
4 Costa, D. L.; Dreher, K. L. (1999) What do we need to know about airborne particles to make effective risk
5 management decisions? A toxicology perspective. Hum. Ecol. Risk Assess. 5: 481-491.
6 Daniels, M; Dominici, F.; Samet, J. M; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
7 curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
8 152:397-406.
9 Dejmek, J.; Selevan, S. G.; Benes, L; Solansky, L; Sram, R. J. (1999) Fetal growth and maternal exposure to
10 paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
11 Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
12 on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
13 155: 568-576.
14 Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998) Symptoms in pediatric asthmatics and air pollution:
15 differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging time.
16 Environ. Health Perspect. 106: 751-761.
17 Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
18 inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
19 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.
20 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 329: 1753-1759.
21 Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
22 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
23 symptoms. Environ. Health Perspect. 104: 500-505.
24 Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
25 saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
26 Health Effects Institute; research report no. 83.
27 Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
28 Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
29 limitations. Toxicol. Sci. 55: 24-35.
30 Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000) Endotoxin priming affects the lung
31 response to ultrafine particles and ozone in young and old rats. In: Inhalation Toxicology: proceedings of the
32 third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC. Inhalation
33 Toxicology 12(suppl. 1): 85-98.
34 Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
35 Perspect. 107: 637-641.
36 Fanucchi, M. V.; Plopper, C. G. (1997) Pulmonary developmental responses to toxicants. In: Roth, R. A., ed.
37 Toxicology of the Respiratory System, v. 8 of Comprehensive Toxicology. New York, NY: Pergamon,
38 p. 203-220.
39 Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B. K.; Van Winkle, L. S.; Evans, M. J.; Plopper, C. G. (2000)
40 Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
41 rhesus monkeys. Am. J. Respir. Crit. Care Med. 161: A615.
42 Federal Register. (1987) Revisions to the national ambient air quality standards for paniculate matter. F. R. (July 1)
43 52:24,634-24,669.
44 Fitz-Simons, T. S.; Mathias, M.; Rizzo, M. (2000) Analyses of 1999 PM data for the PM NAAQS review.
45 U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards; September 29.
46 Gardner, S. Y.; Lehmann, J. R.; Costa, D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in rats.
47 Toxicol. Sci. 56: 175-180.
48 Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
49 Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
50 function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
51 Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
52 summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
53 155:2105-2108.
54 Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
55 effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
56 53:817-822.
June 2003 9A-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 Godleski, J. I; Sioutas, C.; Katler, M; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
2 particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
3 Godleski, J. J.; et al. (1998) Increased cardiac vulnerability during exposure to inhaled environmental particles
4 [abstract]. Presented at: Health Effects Institute annual meeting; May; Boston, MA. Boston, MA: Health
5 Effects Institute.
6 Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms of morbidity and mortality from
7 exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
8 Gold, A.; Litonjua, J.; Schwartz, M.; Verrier, R.; Milstein, A.; Larson, E.; Lovett, B. (1998) Cardiovascular
9 vulnerability to paniculate pollution. Presented at: 1998 International conference [American Thoracic
10 Society]; April; Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A261.
11 Gold, D. R.; Damokosh, A. L; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
12 Castillejos, M. (1999) Paniculate and ozone pollutant effects on the respiratory function of children in
13 southwest Mexico City. Epidemiology 10: 8-16.
14 Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
15 Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
16 Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
17 Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
18 short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
19 Health Effects Institute; research report 97. Available: http://www.healtheffects.org/pubs-research.htm
20 [15 February, 2001].
21 Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
22 (1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
23 Inhalation Toxicol. 11: 981-998.
24 Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
25 hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
26 Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
27 mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
28 Hayano, J.; Sakakibara, Y.; Yamada, M.; Ohte, N.; Fujinami, T.; Yokoyama, K.; Watanabe, Y.; Takata, K. (1990)
29 Decreased magnitude of heart rate spectral components in coronary artery disease. Its relation to angiographic
30 severity. Circulation 81: 1217-1224.
31 Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
32 bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
33 Hoyert, D. L.; Kochanek, K. D.; Murphy, S. L. (1999) Deaths: final data for 1997. Atlanta, GA: U.S. Centers for
34 Disease Control and Prevention, National Center for Health Statistics. (National vital statistics reports: v. 47,
35 no. 19).
36 Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
37 Inhalation Toxicol. 10: 121-130.
38 Kamens, R.; Jang, M.; Chien, C.-J.; Leach, K. (1999) Aerosol formation from the reaction of a-pinene and ozone
39 using a gas-phase kinetics-aerosol partitioning model. Environ. Sci. Technol. 33: 1430-1438.
40 Keywood, M. D.; Ayers, G. P.; Gras, J. L.; Gillett, R. W.; Cohen, D. D. (1999) Relationships between size
41 segregated mass concentration data and ultrafine particle number concentrations in urban areas.
42 Atmos. Environ. 33: 2907-2913.
43 Kim, Y. J.; Boatman, J. F.; Gunter, R. L.; Wellman, D. L.; Wilkison, S. W. (1993) Vertical distribution of
44 atmospheric aerosol size distribution over south-central New Mexico. Atmos. Environ. Part A 27: 1351-1362.
45 Kittelson, D. B. (1998) Engines and nanoparticles: a review. J. Aerosol Sci. 29: 575-588.
46 Kleiger, R. E.; Miller, J. P.; Bigger J. T. Jr.; Moss, A. J.; Multicenter Post-infarction Research Group. (1987)
47 Decreased heart rate variability and its association with increased mortality after acute myocardial infarction.
48 Am. J. Cardiol. 59: 256-262.
49 Kleinman, M. T.; Leaf, D. A.; Kelly, E.; Caiozzo, V; Osann, K; O'Niell, T. (1998) Urban angina in the mountains:
50 effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
51 Health 53: 388-397.
52 Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
53 and daily mortality statistical modeling-interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
54 Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
55 specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
56 50: 1215-1222.
June 2003 9A-8 DRAFT-DO NOT QUOTE OR CITE
-------
1 Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
2 W. H.. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of
3 paniculate air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis
4 Project. Cambridge, MA: Health Effects Institute.
5 Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
6 experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
7 Perspect. 105: 1234-1237.
8 Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
9 sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
10 Lawrence, L.; Hall, M. J. (1999) 1997 summary: national hospital discharge survey. Hyattsville, MD: U.S.
11 Department of Health & Human Resources, Centers for Disease Control and Prevention, National Center for
12 Health Statistics. (Advance data from vital and health statistics: no. 308).
13 Layton, D. W. (1993) Metabolically consistent breathing rates for use in dose assessments. Health Phys. 64: 23-36.
14 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
15 paniculate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international
16 meeting on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake
17 Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
18 Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
19 pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
20 Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
21 mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
22 Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
23 admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
24 Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
25 size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
26 Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000b) Infant mortality and air pollution: a comprehensive analysis of U.S.
27 data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
28 Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
29 mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report
30 no. 95.
31 Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
32 mortality in Mexico City. Epidemiology 10: 118-123.
33 Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
34 J. R. Stat. Soc. B 61(part 2): 459-477.
35 Lundgren, D. A.; Burton, R. M. (1995) Effect of particle size distribution on the cut point between fine and coarse
36 ambient mass fractions. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air
37 pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 131-148.
38 Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
39 Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
40 Martin, G. J.; Magid, N. M.; Myers, G.; Barnett, P. S.; Schaad, J. W.; Weiss, J. S.; Lesch, M.; Singer, D. H. (1987)
41 Heart rate variability and sudden death secondary to coronary artery disease during ambulatory
42 electrocardiographic monitoring. Am. J. Cardiol. 60: 86-89.
43 McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
44 Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
45 California children with asthma. Environ. Health Perspect. 107: 757-760.
46 Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
47 relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
48 Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
49 exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
50 155:245-252.
51 Moolgavkar, S. H. (2000a) Air pollution and daily mortality in three U.S. counties. Environ. Health Perspect.
52 108:777-784.
53 Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
54 metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
55 Toxicol. 12(suppl. 4): 75-90.
56 Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
57 metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
June 2003 9A-9 DRAFT-DO NOT QUOTE OR CITE
-------
1 Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
2 causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
3 Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and admissions
4 for chronic respiratory disease in King County, Washington. In: Inhalation Toxicology: proceedings of the
5 third colloquium on paniculate air pollution and human health; June, 1999; Durham, NC. Inhalation
6 Toxicology 12(suppl. 1): 157-171.
7 Morawska, L.; Thomas, S.; Bofinger, N; Wainwright, D.; Neale, D. (1998) Comprehensive characterization of
8 aerosols in a subtropical urban atmosphere: particle size distribution and correlation with gaseous pollutants.
9 Atmos. Environ. 32: 2467-2478.
10 Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
11 evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
12 Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
13 heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
14 Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
15 Healthy women's PEF variations with ambient summer concentrations of PM10, PN2 5, SO42_, H+, and O3.
16 Am. J. Respir. Crit. Care Med. 160: 117-125.
17 Nakamura, T.; Hayashida, Y. (1992) Autonomic cardiovascular responses to smoke exposure in conscious rats. Am.
18 J. Physiol. 262(5 pt. 2): R73 8-745.
19 National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
20 Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
21 publication no. 97-4051.
22 Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
23 and exposure to air pollution. Public Health Rep. 114: 135-148.
24 Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
25 indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol.
26 139: 1088-1099.
27 Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
28 pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
29 141: 111-122.
30 Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
31 determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
32 Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
33 Philadelphia. Environ. Health Perspect. 107: 629-631.
3 4 Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Halm, F. F. (2000) Influence of dose on the distribution of retained
3 5 paniculate material in rat and human lungs. Presented at: PM2000: paniculate matter and health—the
36 scientific basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC.
37 Pittsburgh, PA: Air & Waste Management Association.
38 Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
3 9 between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
40 Perspect. 107: 489-493.
41 Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
42 stagnant air. Thorax 55: 466-470.
43 Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Branson, K.; Jakab, G. J.; Kleeberger, S. R. (2000) Interstrain variation in
44 murine susceptibility to inhaled acid-coated particles. Am. J. Physiol. L469-L476.
45 Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
46 Am. J. Public Health 81: 694-702.
47 Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Krapnick, A.; Harrington, W. (1993) Air pollution and respiratory
48 morbidity among adults in Southern California. Am. J. Epidemiol. 137: 691-700.
49 Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
50 exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
51 Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
52 January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
53 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
54 Valley, California: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10: 412-419.
55 Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
56 pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
5 7 Environmental Health; March 45.
June 2003 9A-10 DRAFT-DO NOT QUOTE OR CITE
-------
1 Pekkanen, I; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
2 in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
3 Perry, K. D.; Cahill, T. A.; Eldred, R. A.; Dutcher, D. D.; Gill, T. E. (1997) Long-range transport of North African
4 dust to the eastern United States. J. Geophys. Res. [Atmos] 102: 11,225-11,238.
5 Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
6 episode: a link to mortality? Lancet 349: 1582-1587.
7 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b) Respiratory effects are associated with the
8 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
9 Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c) Short-term effects of paniculate air pollution on
10 respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
11 Peters, J. M; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
12 Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities with
13 differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
14 Med. 159: 760-767.
15 Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
16 Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
17 differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
18 159:768-775.
19 Plopper, C. G.; Fanucchi, M. V. (2000) Do urban environmental pollutants exacerbate childhood lung diseases?
20 Environ. Health Perspect. 108: A252-A253.
21 Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
22 (1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J. Respir.
23 Crit. Care Med. 151: 669-674.
24 Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
25 Front. Environ. Health Perspect.: 107: 567-573.
26 Pope, C. A.; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
27 G. M.; Gold, D. R.; Dockery, D. W. (1999b) Heart rate variability associated with paniculate air pollution.
28 Am. Heart J. 138: 890-899.
29 Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
30 F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
31 Perspect. 104: 506-514.
32 Ramadan, Z.; Song, X.-H.; Hopke, P. K. (2000) Identification of sources of Phoenix aerosol by positive matrix
33 factorization. J. Air Waste Manage. Assoc. 50: 1308-1320.
34 Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
35 on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
36 154:300-307.
37 Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
3 8 intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
39 Mexico City. Arch. Environ. Health 52: 368-376.
40 Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (1999) Acute inflammatory
41 responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
42 volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.
43 Samet, J. M.; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000a) National morbidity, mortality, and air
44 pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute; research
45 report no. 94.
46 Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
47 (2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
48 pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
49 Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
50 exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
51 Schlesinger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
52 emissions. In: Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
53 health. Washington, DC: National Academy Press; pp. 239-298.
54 Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
55 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
56 Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
57 8:371-377.
June 2003 9A-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
2 10: 17-22.
3 Schwartz, J. (2000) Harvesting and long term exposure effects in the relation between air pollution and mortality.
4 Am. J. Epidemiol. 151: 440-448.
5 Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
6 Michigan. Am. J. Epidemiol. 142: 23-35.
7 Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
8 respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
9 Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
10 with application to air pollution and daily death. Epidemiology 11: 666-672.
11 Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
12 Ferris, E.G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
13 Am. J. Respir. Crit. Care Med. 150: 1234-1242.
14 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles? J. Air
15 Waste Manage. Assoc. 46: 927-939.
16 Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
17 concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
18 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
19 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
20 Singer, D. H.; Martin, G. J.; Magid, N.; Weiss, J. S.; Schaad, J. W.; Kehoe, R.; Zheutlin, T.; Fintel, D. J.; Hsieh,
21 A. M.; Lesch, M. (1988) Low heart rate variability and sudden cardiac death. J. Electrocardiol.
22 21(suppl.): S46-S55.
23 Smiley-Jewell, S. M.; Liu, F. J.; Weir, A. J.; Plopper, C. G. (2000) Acute injury to differentiating Clara cells in
24 neonatal rabbits results in age-related failure of bronchiolar epithelial repair. Toxicol. Pathol. 28: 267-276.
25 Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
26 coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
27 Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
28 Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
29 John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
3 0 Task Force of the European Society of Cardiology and the North American Society of Pacing and
31 Electrophysiology. (1996) Heart rate variability: standards of measurement, physiological interpretation and
32 clinical use. Circulation 93: 1043-1065.
33 Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; VanEeden, S. F. (1997) Phagocytosis of small carbon particles
34 (PM10) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
35 Am. J. Respir. Crit. Care Med. 155: 1441-1447.
36 Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
37 hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers.
38 J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
39 Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
40 summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
41 65:271-290.
42 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
43 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
44 Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
45 resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
46 Tolbert, P. E.; Klein, M.; Metzger, K. B.; Peel, J.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
47 Frumkin, H. (2000) Interim results of the study of participates and health in Atlanta (SOPHIA). J. Exposure
48 Anal. Environ. Epidemiol. 10: 446-460.
49 Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
50 chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
51 Turpin, B. J. (1999) Options for characterizing organic paniculate matter. Environ. Sci. Technol. 33: 76A-79A.
52 U.S. Environmental Protection Agency. (1996) Air quality cntena for paniculate matter. Research Triangle Park,
53 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
54 U.S. Environmental Protection Agency. (2000a) Air quality criteria for carbon monoxide. Research Triangle Park,
55 NC: National Center for Environmental Assessment; report no. EPA/600/P-99/00 IF. Available:
56 www.epa.gov/ncea/co/ [2000, October 6].
June 2003 9A-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 U.S. Environmental Protection Agency. (2000b) Health assessment document for diesel exhaust [external review
2 draft]. Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria
3 and Assessment Office; report nos. EPA/600/8-90/057E.
4 Vedal, S.; Petkau, I; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
5 nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
6 Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of
7 ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev.
8 Respir. Dis. 133:834-842.
9 Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil fly
10 ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
11 Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U.P.; Highfill, J. W.; Costa, D. L. (2000)
12 Thermoregulatory effects following exposure to paniculate matter in healthy and
13 cardiopulmonary-compromised rats. J. Therm. Biol. 25: 131-137.
14 Weschler, C. J.; Shields, H. C. (1999) Indoor ozone/terpene reactions as a source of indoor particles. Atmos.
15 Environ. 33:2301-2312.
16 Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
17 Whitby, K. T.; Sverdrup, G. M. (1980) California aerosols: their physical and chemical characteristics. In: Hidy,
18 G. M.; Mueller, P. K.; Grosjean, D.; Appel, B. R.; Wesolowski, J. J., eds. The character and origins of smog
19 aerosols: a digest of results from the California Aerosol Characterization Experiment (ACHEX). New York,
20 NY: John Wiley & Sons, Inc.; pp. 477-517. (Advances in environmental science and technology: v. 9).
21 Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
22 mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
23 Cambridge, MA: Health Effects Institute; Research Report no. 98.
24 Williams, R.; Suggs, J.; Zweidinger, R.; Evans, G.; Creason, J.; Kwok, R.; Rodes, C.; Lawless, P.; Sheldon, L.
25 (2000a) The 1998 Baltimore Paniculate Matter Epidemiology-Exposure Study: part 1. Comparison of
26 ambient, residential outdoor, indoor and apartment paniculate matter monitoring. J. Exposure Anal. Environ.
27 Epidemiol. 10: 518-532.
28 Williams, R.; Suggs, J.; Creason, J.; Rodes, C.; Lawless, P.; Kwok, R.; Zweidinger, R.; Sheldon, L. (2000b) The
29 1998 Baltimore Paniculate Matter Epidemiology-Exposure Study: part 2. Personal exposure assessment
30 associated with an elderly study population. J. Exposure Anal. Environ. Epidemiol. 10: 533-543.
31 Williams, R. W.; Creason, J.; Zweidinger, R.; Watts, R. R.; Shy, C. (2000c) Indoor, outdoor, and personal exposure
32 monitoring of paniculate air pollution: the Baltimore elderly epidemiology-exposure pilot study.
33 Atmos. Environ. 34: 4193-4204.
34 Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
35 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
36 Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal infant
37 mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
38 Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
39 J. Occup. Environ. Med. 42: 469-474.
40 Zhang, T.; Huang, C.; Johns, E. J. (1997) Neural regulation of kidney function by the somatosensory system in
41 normotensive and hypertensive rats. Am. J. Physiol. 273: R1749-R1757.
42 Zhang, H.; Triche, E.; Leaderer, B. (2000) Model for the analysis of binary time series of respiratory symptoms. Am.
43 J. Epidemiol. 151: 1206-1215.
44
45
June 2003 9A-13 DRAFT-DO NOT QUOTE OR CITE
------- |