&EPA
United States nth ?nn4
Proleclio" EPAKOO/P-°99/002bF
Air Quality Criteria for
Particulate Matter
Volume II of II
-------
EPA/600/P-99/002bF
October 2004
Air Quality Criteria for Particulate Matter
Volume II
National Center for Environmental Assessment-RTF Office
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
-------
DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
PREFACE
National Ambient Air Quality Standards (NAAQS) are promulgated by the United States
Environmental Protection Agency (EPA) to meet requirements set forth in Sections 108 and 109
of the U.S. Clean Air Act (CAA). Sections 108 and 109 require the EPA Administrator (1) to
list widespread air pollutants that reasonably may be expected to endanger public health or
welfare; (2) to issue air quality criteria for them that assess the latest available scientific
information on nature and effects of ambient exposure to them; (3) to set "primary" NAAQS to
protect human health with adequate margin of safety; (4) to set "secondary" NAAQS to protect
against welfare effects (e.g., effects on vegetation, ecosystems, visibility, climate, manmade
materials, etc.); and (5) to periodically (every 5 years) review and revise, as appropriate, the
criteria and NAAQS for a given listed pollutant or class of pollutants.
The original NAAQS for particulate matter (PM), issued in 1971 as "total suspended
particulate" (TSP) standards, were revised in 1987 to focus on protecting against human health
effects associated with exposure to ambient PM less than 10 microns (< 10 jim) that are capable
of being deposited in thoracic (tracheobronchial and alveolar) portions of the lower respiratory
tract. Later periodic reevaluation of newly available scientific information, as presented in the
last previous version of this "Air Quality Criteria for Paniculate Matter" document published in
1996, provided key scientific bases for PM NAAQS decisions published in July 1997. More
specifically, the PM10 NAAQS set in 1987 (150 |ig/m3, 24-h; 50 |ig/m3, annual average) were
retained in modified form and new standards (65 |ig/m3, 24-h; 15 |ig/m3, annual average) for
particles < 2.5 jim (PM25) were promulgated in July 1997.
This final version of revised Air Quality Criteria for P articulate Matter assesses new
scientific information that has become available (published or accepted for publication) mainly
Il-ii
-------
between early 1996 through April 2002, although a few important new studies published through
2003 are also considered. Several previous successive drafts of this document were released for
public comment and review by the Clean Air Scientific Advisory Committee (CASAC), to
obtain comments on the organization and structure of the document, the issues addressed, the
approaches employed in assessing and interpreting the newly available information on PM
exposures and effects, and the key findings and conclusions arrived at by this assessment. Public
comments and CASAC review recommendations were taken into account in making revisions to
this document for incorporation into this final draft. Evaluations contained in this document will
be drawn on to provide inputs to associated PM Staff Paper analyses prepared by EPA's Office
of Air Quality Planning and Standards (OAQPS) to pose alternatives for consideration by the
EPA Administrator with regard to proposal and, ultimately, promulgation of decisions on
potential retention or revision of the current PM NAAQS.
The document describes the nature, sources, distribution, measurement, and concentrations
of PM in outdoor (ambient) environments. It also evaluates the latest data on human exposures
to ambient PM and consequent health effects in exposed human populations, to support decision
making regarding primary, health-related PM NAAQS. The document also evaluates ambient
PM environmental effects on vegetation and ecosystems, visibility, and man-made materials, as
well as atmospheric PM effects on climate change processes, to support decision making on
secondary PM NAAQS.
Preparation of this document was coordinated by EPA's National Center for
Environmental Assessment in Research Triangle Park (NCEA-RTP). NCEA-RTP scientific
staff, together with experts from other EPA/ORD laboratories and academia, contributed to
writing of document chapters. The NCEA of EPA acknowledges the contributions provided by
authors, contributors, and reviewers and the diligence of its staff and contractors in the
preparation of this document.
-------
Air Quality Criteria for Particulate Matter
VOLUME I
1. INTRODUCTION 1-1
2. PHYSICS, CHEMISTRY, AND MEASUREMENT OF
OF PARTICULATE MATTER 2-1
APPENDIX 2A: Techniques for Measuring 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: Composition of Particulate Matter Source Emissions .... 3A-1
APPENDIX 3B: Organic Composition of Particulate Matter 3B-1
APPENDIX 3C: Aerosol Composition Data from the Speciation
Network 3C-1
APPENDIX 3D: Spatial and Temporal Variability of the Nationwide
AIRS PM25 and PM10.25 Data Sets 3D-1
APPENDIX 3E: Characterization of PM25, PM10, and PM10.25
Concentrations at IMPROVE Sites 3E-1
4. ENVIRONMENTAL EFFECTS OF AIRBORNE PARTICULATE
MATTER 4-1
APPENDIX 4A: Common and Latin Names 4A-1
5. HUMAN EXPOSURE TO PARTICULATE MATTER AND
ITS CONSTITUENTS 5-1
Il-iv
-------
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
APPENDIX 7A: Rat-to-Human Dose Extrapolation 7A-1
APPENDIX 7B: Ambient Bioaerosols 7B-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
II-v
-------
Table of Contents
Page
List of Tables II-xvii
List of Figures II-xxiv
Authors, Contributors, and Reviewers II-xxxi
U.S. Environmental Protection Agency Project Team for Development of Air
Quality Criteria for Particulate Matter II-xl
U.S. Environmental Protection Agency Science Advisory Board (SAB)
Staff Office Clean Air Scientific Advisory Committee (CASAC)
Particulate Matter Review Panel II-xliii
Abbreviations and Acronyms II-xlvi
6. DOSIMETRY OF PARTICULATE MATTER 6-1
6.1 INTRODUCTION 6-1
6.1.1 Size Characterization of Inhaled Particles 6-2
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 6-22
6.2.3 Biological Factors Modulating Deposition 6-24
6.2.3.1 Gender 6-24
6.2.3.2 Age 6-26
6.2.3.3 Respiratory Tract Disease 6-31
6.2.3.4 Anatomical Variability 6-34
6.2.3.5 Inhaled Irritants 6-36
6.2.4 Interspecies Patterns of Deposition 6-36
6.3 PARTICLE CLEARANCE 6-43
6.3.1 Mechanisms and Pathways of Clearance 6-43
6.3.1.1 Extrathoracic Region 6-45
6.3.1.2 Tracheobronchial Region 6-46
6.3.1.3 Alveolar Region 6-46
6.3.2 Clearance Kinetics 6-48
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 6-56
6.3.4 Factors Modulating Clearance 6-57
Il-vi
-------
Table of Contents
(cont'd)
Page
6.3.4.1 Age 6-57
6.3.4.2 Gender 6-58
6.3.4.3 Physical Activity 6-58
6.3.4.4 Respiratory Tract Disease 6-58
6.3.4.5 Inhaled Irritants 6-60
6.4 PARTICLE OVERLOAD 6-60
6.5 COMPARISON OF DEPOSITION AND CLEARANCE PATTERNS
OF PARTICLES ADMINISTERED BY INHALATION AND
INTRATRACHEAL INSTILLATION 6-62
6.6 MODELING THE DEPOSITION AND DISPOSITION OF
PARTICLES IN THE RESPIRATORY TRACT 6-68
6.6.1 Modeling Deposition, Clearance, and Retention 6-68
6.6.2 Models To Estimate Retained Dose 6-75
6.6.3 Fluid Dynamics Models for Deposition Calculations 6-78
6.6.4 Modeling Results Obtained with Models Available to the Public .. 6-84
6.6.4.1 International Commission on Radiological
Protection (ICRP) 6-85
6.6.4.2 Multiple Path Particle Dosimetry Model (MPPD) 6-90
6.6.43 Comparisons of Deposition in Humans and Rats 6-99
6.7 SUMMARY AND CONCLUSIONS 6-105
6.7.1 Particle Dosimetry 6-105
6.7.2 Host Factors 6-106
6.7.3 Laboratory Animal Studies 6-107
6.7.4 Mathematical Models 6-108
6.7.5 Key Points 6-108
7. TOXICOLOGY OF PARTICIPATE MATTER IN HUMANS AND
LABORATORY ANIMALS 7-1
7.1 INTRODUCTION 7-1
7.1.1 Methodological Considerations 7-2
7.1.2 Organization of the Chapter 7-6
7.2 CARDIOVASCULAR AND SYSTEMIC EFFECTS OF IN VIVO PM
EXPOSURES IN HUMANS AND LABORATORY ANIMALS 7-8
7.2.1 Introduction 7-8
7.2.2 Ambient Particulate Matter Cardiovascular Effects 7-23
7.2.3 ROFA and Other Combustion Source-Related Particles 7-27
7.2.4 Summary of Cardiovascular/Systemic Effects 7-35
7.3 RESPIRATORY EFFECTS OF CONTROLLED IN VIVO PM
EXPOSURES OF HUMANS AND LABORATORY ANIMALS 7-36
7.3.1 Ambient Particulate Matter 7-37
7.3.1.1 Ambient Particle Inhalation Exposures 7-38
-------
Table of Contents
(cont'd)
Page
7.3.1.2 Intratracheal and Intrabronchial Instillation of
Ambient Particulate Matter 7-44
7.3.2 ROFA and Other Combustion Source-Related Particles 7-48
7.3.3 Metals 7-60
7.3.4 Acid Aerosols 7-64
7.3.5 Diesel Particulate Matter 7-69
7.3.5.1 Salient Findings from U.S. EPA 2002
Diesel Document 7-70
7.3.6 Ambient Bioaerosols 7-78
7.3.7 Summary of Respiratory Effects 7-85
7.4 CARDIOVASCULAR AND RESPIRATORY PATHOPHYSIOLOGY
AND TOXICITY: IN VITRO PM EXPOSURES 7-86
7.4.1 Introduction 7-86
7.4.2 Ambient Particles Effects 7-97
7.4.3 Comparison of Ambient and Combustion Source-Related
Particles 7-102
7.4.4 Potential Cellular and Molecular Mechanisms 7-106
7.4.4.1 Reactive Oxygen Species 7-107
7.4.4.2 Intracellular Signaling Mechanisms 7-112
7.4.4.3 Particle Charge and Stimulation of Sensory
Nerve Receptors 7-117
7.4.4.4 Other Potential Cellular and Molecular
Mechanisms 7-120
7.4.5 Specific Particle Size and Surface Area Effects 7-120
7.5 FACTORS AFFECTING SUSCEPTIBILITY TO PARTICULATE
MATTER EXPOSURE EFFECTS 7-125
7.5.1 Pulmonary Effects of Particulate Matter in
Compromised Hosts 7-125
7.5.2 Genetic Susceptibility to Inhaled Particles and
Their Constituents 7-131
7.5.3 Particulate Matter Effects on Allergic Hosts 7-133
7.5.4 Resistance to Infectious Disease 7-140
7.6 RESPONSES TO PARTICULATE MATTER AND GASEOUS
POLLUTANT MIXTURES 7-142
7.7 QUANTITATIVE COMPARISONS OF EXPERIMENTAL PM
EFFECTS ON CARDIOVASCULAR/RESPIRATORY ENDPOINTS
IN HUMANS AND LABORATORY ANIMALS 7-151
7.7.1 Introduction 7-151
7.7.1.1 Cardiovascular and Systemic Effects of Inhaled
Particulate Matter 7-152
7.7.1.2 Cardiovascular and Systemic Effects of Instilled
Particulate Matter 7-154
-------
Table of Contents
(cont'd)
Page
7.7.1.3 Respiratory Effects of Inhaled Particulate Matter ....7-156
7.7.1.4 Respiratory Effects of Instilled Parti culate Matter .... 7-158
7.7.1.5 In Vitro Effects of Particulate Matter on ng/Cell
Dose Basis 7-160
7.7.2 Interspecies Comparisons of Experimental Results 7-162
7.7.2.1 Introduction 7-162
7.7.2.2 Dosimetric Intercomparison for PMN Influx as a
Marker for Lung Inflammation 7-164
7.7.2.3 Inhibition of Phagocytosis by PM Exposure 7-166
7.8 MUTAGENICITY/GENOTOXICITY EFFECTS 7-170
7.8.1 Ambient Particulate Matter Effects 7-170
7.8.2 Wood and Coal Combustion-Source Effects 7-181
7.8.2.1 Biomass/Wood Burning 7-181
7.8.2.2 Coal Combustion 7-184
7.8.3 Mobile Combustion-Source Effects 7-186
7.8.3.1 Diesel 7-186
7.8.3.2 Gasoline 7-193
7.8.4 Summary of Mutagenic/Genotoxic Effects 7-196
7.9 INHALED PARTICLES AS POTENTIAL CARRIERS OF
TOXIC AGENTS 7-198
7.10 INTERPRETIVE SUMMARY OF PM TOXICOLOGY FINDINGS 7-204
7.10.1 Particulate Matter Health Effects and Potential Mechanisms
of Action 7-205
7.10.1.1 Direct Pulmonary Effects 7-206
7.10.1.2 Cardiovascular and Other Systemic Effects
Secondary to Lung Injury 7-209
7.10.1.3 Direct Effects on the Heart 7-212
7.10.1.4 Mutagenic/Genotoxic Effects of PM 7-214
7.10.2 Links Between Specific Particulate Matter Components and
Health Effects 7-215
7.10.2.1 Ambient Particle Studies 7-215
7.10.2.2 Acid Aerosols 7-217
7.10.2.3 Metals 7-217
7.10.2.4 Diesel Exhaust Particles 7-219
7.10.2.5 Organic Compounds 7-219
7.10.2.6 Ultrafme Particles 7-220
7.10.2.7 Bioaerosols 7-220
7.10.3 PM Interactions with Gaseous Co-Pollutants 7-221
7.10.4 Susceptibility 7-222
REFERENCES 7-225
Il-ix
-------
Table of Contents
(cont'd)
Pat
APPENDIX 7A. RAT-TO-HUMAN DOSE EXTRAPOLATION 7A-1
7A.1 INTRODUCTION 7A-1
7A.2 QUANTITATIVE INTERSPECIES EXTRAPOLATION 7A-2
7A.3 THE MULTIPLE PATH PARTICLE DOSIMETRY MODEL (MPPD) .. 7A-6
7A.4 RAT AND HUMAN DOSIMETRY: INTERSPECIES
DIFFERENCES 7A-10
7A.4.1 Anatomy 7A-10
7A.4.2 Exposure Scenarios 7A-11
7A.4.2.1 Exertion Level 7A-11
7A.4.2.2 Size Distribution 7A-12
7A.4.3 Quantities Calculated by Dosimetric Models 7A-15
7A.4.3.1 Deposition Fraction (DF) 7A-15
7A.4.3.2 Clearance 7A-16
7A.4.3.3 Retention 7A-16
7A.4.3.4 Long-Term Burden from Chronic Exposure ... 7A-21
7A.4.4 Dose Metrics 7A-23
7A.4.5 Normalizing Factors and Other Differences Between
Humans and Rats 7A-24
7A.5 DOSIMETRIC CALCULATION FOR EXTRAPOLATION
MODELING: COMPARING RATS TO HUMANS 7A-27
7A.5.1 General Exposure Scenarios 7A-27
7A.5.1.1 Acute Exposures 7A-27
7A.5.1.2 Rat and Human Each Exposed to One
Mode of the Atmospheric Particle
Size Distribution 7A-30
7A.5.1.3 Exposure to Resuspended Combustion
Particles 7A-33
7A.5.1.4 Rat Exposed to One Fraction, Human
Exposed to All Three Modes of the
Atmospheric Particle Size Distribution 7A-36
7A.5.1.5 Discussion 7A-36
7A.5.1.6 Rat-to-Human Extrapolation of Long-Term
PM Burden in the Alveolar Region 7A-39
7A.5.1.7 Caveats 7A-43
7A.6 HEALTH STATUS: A NON-DOSIMETRIC CONSIDERATION 7A-44
7A.7 COMPARATIVE DOSIMETRY FOR SPECIFIC
PUBLISHED STUDY EXAMPLES 7A-44
7A.7.1 Utah Valley Dust 7A-45
7A.7.2 Concentrated Ambient Particles (CAPs) 7A-51
7A.7.3 Clearance Overload in Rats 7A-53
7A.8 SUMMARY 7A-57
II-x
-------
Table of Contents
(cont'd)
Page
7A.9 CONCLUSIONS 7A-62
REFERENCES 7A-64
APPENDIX 7B. AMBIENT BIOAEROSOLS 7B-1
7B. 1 INTRODUCTION AND BACKGROUND INFORMATION ON
AMBIENT BIOAEROSOLS 7B-1
7B.1.1 Plant Aerosols 7B-2
7B.1.2 Animal Aerosols 7B-3
7B.1.3 Fungal Aerosols 7B-4
7B.1.4 Bacterial Aerosols 7B-6
7B.1.5 Viral Aerosols 7B-7
7B.2 NEWLY AVAILABLE BIOAEROSOLS RESEARCH 7B-8
7B.2.1 Atmospheric Levels of Cellulose/Other Plant
Debris Markers 7B-12
7B.2.2 Pollen 7B-14
7B.2.3 Fungi and Their Byproducts 7B-20
7B.2.4 Endotoxins 7B-24
7B.2.5 (1 - 3)-p-D-Glucan 7B-29
7B.3 SUMMARY 7B-31
REFERENCES 7B-33
EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS ASSOCIATED WITH
AMBIENT PARTICULATE MATTER 8-1
8.1 INTRODUCTION 8-1
8.1.1 Approaches for Identifying, Presenting, and Assessing Studies .... 8-2
8.1.2 Types of Epidemiologic Studies Reviewed 8-5
8.1.3 Overview of Key Methodological Issues 8-8
8.1.3.1 Issues Related to Use of Generalized Additive
Models (GAM) in PM Epidemiology 8-8
8.1.3.2 Confounding and Effect Modification 8-10
8.1.4 Approach to Assessing Epidemiologic Evidence 8-15
8.2 MORTALITY EFFECTS ASSOCIATED WITH AIRBORNE
PARTICULATE MATTER EXPO SURE 8-18
8.2.1 Introduction 8-18
8.2.2 Mortality Effects of Short-Term Particulate Matter Exposure 8-18
8.2.2.1 Summary of 1996 Particulate Matter Criteria
Document Findings and Key Issues 8-18
8.2.2.2 Newly Available Information on Short-Term
Mortality Effects 8-23
8.2.2.3 New Multicity Studies 8-30
8.2.2.4 U.S. Single-City Studies 8-50
8.2.2.5 The Role of Particulate Matter Components 8-56
Il-xi
-------
Table of Contents
(cont'd)
Page
8.2.2.6 New Assessments of Cause-Specific Mortality 8-77
8.2.2.7 Salient Points Derived from Assessment of Studies
of Short-Term Particulate Matter Exposure Effects
on Mortality 8-83
8.2.3 Mortality Effects of Long-Term Exposure to Ambient
Particulate Matter 8-86
8.2.3.1 Studies Published Prior to the 1996 Particulate
Matter Criteria Document 8-86
8.2.3.2 New Prospective Cohort Analyses of Mortality
Related to Chronic Particulate Matter Exposures 8-90
8.2.3.3 Studies by Particulate Matter Size-Fraction
and Composition 8-121
8.2.3.4 PM-Mortality Intervention Studies 8-131
8.2.3.5 Salient Points Derived from Analyses of Chronic
Particulate Matter Exposure Mortality Effects 8-135
.3 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE .... 8-139
8.3.1 Cardiovascular Morbidity Effects Associated with Acute
Ambient Particulate Matter Exposure 8-139
8.3.1.1 Introduction 8-139
8.3.1.2 Summary of Key Findings on Cardiovascular
Morbidity from the 1996 Particulate Matter Air
Quality Criteria Document 8-140
8.3.1.3 New Particulate Matter-Cardiovascular
Morbidity Studies 8-140
8.3.1.4 Issues in the Interpretation of Acute
Cardiovascular Effects Studies 8-169
8.3.2 Effects of Short-Term Particulate Matter Exposure on the
Incidence of Respiratory-Related Hospital Admissions and
Medical Visits 8-171
8.3.2.1 Introduction 8-171
8.3.2.2 Summary of Key Respiratory Hospital Admissions
Findings from the 1996 Particulate Matter Air
Quality Criteria Document 8-172
8.3.2.3 New Respiratory-Related Hospital
Admissions Studies 8-172
8.3.2.4 Key New Respiratory Medical Visits Studies 8-188
8.3.2.5 Identification of Potential Susceptible
Subpopulations 8-191
8.3.2.6 Summary of Salient Findings on Acute Particulate
Matter Exposure and Respiratory-Related Hospital
Admissions and Medical Visits 8-192
-------
Table of Contents
(cont'd)
Page
8.3.3 Effects of Particulate Matter Exposure on Lung Function and
Respiratory Symptoms 8-194
8.3.3.1 Effects of Short-Term Particulate Matter Exposure
on Lung Function and Respiratory Symptoms 8-194
8.3.3.2 Long-Term Particulate Matter Exposure Effects
on Lung Function and Respiratory Symptoms 8-211
8.3.4 Ambient PM Impacts on Fetal and/or Early Postnatal
Development/Mortality 8-215
8.3.4.1 PM Effects on Intrauterine Fetal
Morbidity/Mortality 8-216
8.3.4.2 PM Effects on Postneonatal Infant Mortality 8-219
8.3.4.3 Summary of Saliant Points on PM Effects on Fetal
and/or Early Postnatal Development/Mortality 8-221
.4 INTERPRETIVE ASSESSMENT OF THE EPIDEMIOLOGIC
EVIDENCE 8-222
8.4.1 Introduction 8-222
8.4.2 GAM Issue and Reanalyses Studies 8-227
8.4.2.1 Impact of Using the More Stringent GAM Model
on PM Effect Estimates for Mortality 8-227
8.4.2.2 Impact of Using the More Stringent GAM Model
on PM Effect Estimates for Respiratory
Hospital Admissions 8-232
8.4.2.3 HEI Commentaries 8-236
8.4.3 Assessment of Confounding by Co-Pollutants and Adjustments
for Meteorological Variables 8-238
8.4.3.1 Introduction to Assessment of Confounding
by Co-Pollutants 8-238
8.4.3.2 Statistical Issues in the Use of Multipollutant
Models 8-240
8.4.3.3 Multipollutant Modeling Outcomes 8-246
8.4.3.4 Bioaerosols as Possible Confounders or Effect
Modifiers in PM Epidemiologic Studies 8-255
8.4.3.5 Adjustments for Meteorological Variables 8-256
8.4.4 The Question of Lags 8-269
8.4.5 Measurement Error: Concepts and Consequences 8-282
8.4.5.1 Theoretical Framework for Assessment of
Measurement Error 8-282
8.4.5.2 Measurement Error Issues Related to Divergence
Between Monitors and to Monitoring Frequency 8-289
8.4.5.3 Measurement Error and the Assessment of
Confounding by Co-Pollutants in
Multipollutant Models 8-300
-------
Table of Contents
(cont'd)
Page
8.4.6 Role of Particulate Matter Components 8-301
8.4.6.1 Thoracic Particle (PM10) Mortality/Morbidity
Effects 8-301
8.4.6.2 Fine and Coarse Fraction Particle Effects
on Mortality 8-302
8.4.6.3 Source-Oriented Analyses of PM and Mortality 8-307
8.4.6.4 Fine and Coarse Fraction Particle Effects
on Morbidity 8-309
8.4.7 Concentration-Response Relationships for Ambient PM 8-318
8.4.8 The Question of Heterogeneity of Particulate Matter
Effects Estimates 8-323
8.4.8.1 Evaluation of Heterogeneity in Time-Series
Studies 8-323
8.4.8.2 Comparison of Spatial Relationships in the
NMMAPS and Cohort Reanalyses Studies 8-326
8.4.9 Age-Related Differences in PM Effect Estimates 8-327
8.4.10 Implications of Airborne Particle Mortality Effects 8-328
8.4.10.1 Short-Term Exposure and Mortality
Displacement 8-329
8.4.10.2 Life-Shortening Estimates Based on Prospective
Cohort Study Results 8-334
8.4.10.3 Potential Effects of Infant Mortality on
Life-Shortening Estimates 8-335
8.5 SUMMARY OF KEY FINDINGS AND CONCLUSIONS DERIVED
FROM PARTICULATE MATTER EPIDEMIOLOGY STUDIES 8-335
REFERENCES 8-348
APPENDIX 8A: SHORT-TERM PM EXPOSURE—MORTALITY
STUDIES: SUMMARY TABLE 8A-1
APPENDIX 8B: PARTICULATE MATTER-MORBIDITY STUDIES:
SUMMARY TABLES 8B-1
8B.1 PM-Cardiovascular Admissions Studies 8B-2
8B.2 PM-Respiratory Hospitalization Studies 8B-19
8B.3 PM-Respiratory Visits Studies 8B-42
8B.4 Pulmonary Function Studies 8B-53
8B.5 Short-Term PM Exposure Effects on Symptoms in
Asthmatic Individuals 8B-59
8B.6 Short-Term PM Exposure Effects on Pulmonary Function
in Nonasthmatics 8B-65
8B.7 Short-Term PM Exposure Effects on Symptoms in Nonasthmatics 8B-73
8B.8 Long-Term PM Exposure Effects on Respiratory Health Indicators,
Symptoms, and Lung Function 8B-78
-------
Table of Contents
(cont'd)
Page
9. INTEGRATIVE SYNTHESIS 9-1
9.1 INTRODUCTION 9-1
9.1.1 Chapter Organization 9-1
9.1.2 Trends in United States PM Air Quality 9-2
9.2 SYNTHESIS OF AVAILABLE INFORMATION ON PM-RELATED
HEALTH EFFECTS 9-5
9.2.1 Fine and Coarse Particles as Separate Subclasses of
PM Pollution 9-7
9.2.1.1 Physics and Chemistry Considerations 9-10
9.2.1.2 Exposure-Related Considerations 9-15
9.2.1.3 Dosimetric Considerations 9-17
9.2.1.4 Summary and Conclusions 9-20
9.2.2 Assessment of Epidemiologic Evidence 9-22
9.2.2.1 Strength of Epidemiologic Associations 9-24
9.2.2.2 Robustness of Epidemiologic Associations 9-34
9.2.2.3 Consistency of Findings Across Epidemiologic
Studies 9-39
9.2.2.4 Temporality and the Question of Lags 9-42
9.2.2.5 Concentration-Response Relationships 9-43
9.2.2.6 Natural Experiment Studies 9-45
9.2.2.7 Summary and Conclusions 9-46
9.2.3 Integration of Experimental and Epidemiologic Evidence 9-48
9.2.3.1 Background on Cross-Cutting Issues 9-51
9.2.3.2 Biological Plausibility and Coherence of Evidence
for Different Health Endpoint Categories 9-63
9.2.3.3 Summary and Conclusions 9-77
9.2.4 Potentially Susceptible and Vulnerable Subpopulations 9-81
9.2.4.1 Preexisting Disease as a Risk Factor 9-82
9.2.4.2 Age-Related At-Risk Population Groups:
the Elderly, Infants, and Children 9-83
9.2.4.3 Genetic Susceptibility 9-85
9.2.4.4 Gender 9-85
9.2.4.5 Factors Related to Enhanced Vulnerability 9-86
9.2.4.6 Summary and Conclusions 9-88
9.2.5 Potential Public Health Impacts in the United States 9-88
9.2.5.1 Magnitude of Susceptible Groups 9-89
9.2.5.2 Impact on Life Expectancy 9-93
9.3 SYNTHESIS OF AVAILABLE INFORMATION ON PM-RELATED
WELFARE EFFECTS 9-94
9.3.1 Airborne Particle Effects on Visibility 9-95
9.3.1.1 Relationships Between Ambient PM and Visibility .... 9-95
II-xv
-------
Table of Contents
(cont'd)
Page
9.3.1.2 Public Perception and Valuation of Visibility
Improvements 9-98
9.3.1.3 Summary and Conclusions 9-99
9.3.2 Effects of Ambient PM on Vegetation and Ecosystems 9-100
9.3.2.1 Direct and Indirect Effects of Ambient PM 9-100
9.3.2.2 Major Ecosystem Stressors 9-101
9.3.2.3 Characterization of PM-Related Ecosystem
Stressors 9-109
9.3.2.4 Summary and Conclusions 9-110
9.3.3 Relationships Between Atmospheric PM and Climate
Change Processes 9-111
9.3.4 Effects of Ambient PM on Man-Made Materials 9-113
REFERENCES 9-115
-------
List of Tables
Number Page
6-1 Effects of Age on Particle Deposition in Respiratory Tract 6-27
6-2 Overview of Respiratory Tract Particle Clearance and Translocation
Mechanisms 6-44
6-3 Respiratory Parameters Used in LUDEP Model 6-86
6-4 Breathing Patterns for Comparison of ICRP and MPPD Models 6-92
6-5 Ratio of MPPD to ICRP Deposition Fraction for Several
Size Distributions 6-95
6-6 Levels of Physical Exertion for Adult, Corresponding Representative
Activities, and Breathing Parameters 6-97
6-7 Respiratory Parameters for Humans and Rats 6-99
6-8 Surface Area Values for Lung Mass and of Tracheobronchial and
Alveolar Regions for Humans and Rats 6-102
7-1 Cardiovascular and Systemic Effects of Inhaled Ambient
Particulate Matter 7-15
7-2 Cardiovascular and Systemic Effects of Inhaled ROFA and Other
Combustion-Related Particulate Matter 7-17
7-3 Cardiovascular and Systemic Effects of Instilled ROFA and Other
Particulate Matter 7-19
7-4 Respiratory Effects of Inhaled Ambient Particulate Matter in Controlled
Exposure Studies of Human Subjects and Laboratory Animals 7-39
7-5 Respiratory Effects of Instilled Ambient Particulate Matter in Laboratory
Animals and Human Subjects 7-45
7-6 Respiratory Effects of Intratracheally Instilled ROFA and Other
Combustion Source-Related Particulate Matter in Healthy
Laboratory Animals 7-49
7-7 Respiratory Effects of Inhaled and Instilled ROFA and Other
Combustion Source-Related Particulate Matter in Compromised
Laboratory Animal Models 7-53
-------
List of Tables
(cont'd)
Number Page
7-8 Respiratory Effects of Inhaled and Instilled Metal Particles in Human
Subjects and Laboratory Animals 7-62
7-9 Respiratory Effects of Acid Aerosols in Humans and Laboratory Animals . . . 7-65
7A-1 Human and Rat Breathing Patterns Used in Dosimetric Calculations 7A-12
7A-2 Particle Characteristics Used by EPA in MPPD Model Calculations
and Some Examples of Regional Deposition Fractions 7 A-15
7A-3 Exposure Scenarios for Accumulation of Long-term Burden used by
EPA in MPPD Model Calculations 7A-25
7A-4 Parameters Used to Define a Dose Metric 7A-25
7A-5 Characteristics of Human and Rat Lungs 7A-26
7A-6 Dosimetric Differences Between Rats and Humans 7A-28
7A-7a Predicted Particle Mass Dose Metrics for Human and Rat Each Exposed
to One Atmospheric Mode: Equivalent Exposure Ratio (EqER) for PM
Dose After 6-hour Exposure for Several Breathing Patterns 7A-31
7A-7b Predicted Particle Surface Area and Number Dose Metrics for Human
and Rat Each Exposed to One Atmospheric Mode: Equivalent Exposure
Ratio (EqER) for PM Dose After 6-hour Exposure for Several
Breathing Patterns 7A-32
7A-8a Predicted Particle Mass Dose Metrics for Rat Exposed to Resuspended
PM (e.g., ROFA), Human Exposed to All Three Atmospheric Modes:
Equivalent Exposure Ratio (EqER) for PM Dose After a 6-hour Exposure
for Several Breathing Patterns 7A-34
7A-8b Predicted Particle Surface Area and Number Dose Metrics for Rat
Exposed to Resuspended PM (e.g., ROFA), Human Exposed to All
Three Atmospheric Modes: Equivalent Exposure Ratio (EqER) for
PM Dose After a 6-hour Exposure for Several Breathing Patterns 7A-35
7A-9a Predicted Particle Mass Dose Metrics for Rat Exposed to One Mode at
a Time, Human Exposed to All Three Atmospheric Modes: Equivalent
Exposure Ratio (EqER) for PM Dose After a 6-hour Exposure for
Several Breathing Patterns 7A-37
-------
List of Tables
(cont'd)
Number Page
7A-9b Predicted Particle Surface Area and Number Dose Metrics for Rat
Exposed to One Mode at a Time, Human Exposed to All Three
Atmospheric Modes: Equivalent Exposure Ratio, EqER, for PM
Dose After a 6-hour Exposure for Several Breathing Patterns 7A-38
7A-10a Utah Valley Dust: Exposure Scenario 7A-46
7A-10b Utah Valley Dust: Human Instillation Study 7A-47
7A-10c(l) Utah Valley Dust: Rat Instillation Study 7A-50
7A-10c(2) Utah Valley Dust: Rat Instillation Study Exposure Scenarios
Achieving Instilled Dose 7A-50
7A-1 la CAPs: Human Inhalation Study (Ohio et al., 2000) 7A-51
7A-1 Ib CAPs: Rat Inhalation Study (Kodavanti et al., 2000) 7A-52
7A-1 Ic CAPs: Rat Inhalation Study (Clarke et al., 1999) 7A-53
7A-12 Estimated Exposure Concentrations (mg/m3) Leading to Varied
Levels of Alveolar Loading as a Function of Particle Size and
Exposure Duration 7A-56
7B-1 Examples of Major Sources, Types of Particles, and Disease Agents
Associated with Bioaerosols 7B-1
7B-2 Respiratory Effects of Pollen/Fungi and PM Exposures 7B-9
7B-3 Respiratory Effects of Inhaled Endotoxin-Laden Ambient Bioaerosols 7B-11
8-1 Recent U.S. and Canadian Time-Series Studies of PM-Related
Daily Mortality 8-25
8-2 Synopsis of Short-Term Mortality Studies that Examined Relative
Importance of PM25 and PM10_25 8-58
8-3 Newly Available Studies of Mortality Relationships to PM
Chemical Components 8-70
8-4 Summary of Source-Oriented Evaluations of PM Components in
Recent Studies 8-74
-------
List of Tables
(cont'd)
Number Page
8-5 Comparison of Six Cities and American Cancer Society Study Findings
from Original Investigators and Health Effects Institute Reanalysis 8-92
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 8-96
8-7 Summary of Results from the Extended ACS Study 8-99
8-8 Relative Risk of Mortality From all Nonexternal Causes, by Sex and Air
Pollutant, for an Alternative Covariate Model in the AHSMOG Study 8-106
8-9 Relative Risk of Mortality From Cardiopulmonary Causes, by Sex and Air
Pollutant, for an Alternative Covariate Model in the AHSMOG Study 8-107
8-10 Relative Risk of Mortality from Lung Cancer by Air Pollutant and by
Gender for an Alternative Covariate Model 8-108
8-11 Particulate Matter Effects on Mortality by Exposure and Mortality
Period With Ecological Variables for the Veterans Cohort Study
Expressed as Excess Mortality 8-113
8-12 Comparison of Excess Relative Risks of Long-Term Mortality in the
Harvard Six Cities, ACS, AHSMOG, and VA Studies 8-117
8-13 Comparison of Estimated Relative Risks for All-Cause Mortality in
Six U.S. Cities Associated with the Reported Intercity Range of
Concentrations of Various Particulate Matter Metrics 8-122
8-14 Comparison of Reported SO42 and PM2 5 Relative Risks for Various
Mortality Causes in the American Cancer Society Study 8-123
8-15 Comparison of Total Mortality Relative Risk Estimates and t-Statistics
for Particulate Matter Components in Three Prospective Cohort Studies ... 8-124
8-16 Comparison of Cardiopulmonary Mortality Relative Risk Estimates and
t-Statistics for Particulate Matter Components in Three Prospective
Cohort Studies 8-125
8-17 Percent Attributable Risk of Mortality (from Lipfert and Morris, 2000)
and Risk Estimates Calculated per 10 |ig/m3 PM2 5 8-128
II-xx
-------
List of Tables
(cont'd)
Number Page
8-18 Summary of Studies of PM10, PM10.2 5, or PM2 5 Effects on Total CVD
Hospital Admissions and Emergency Visits 8-142
8-19 Summary of United States PM10 Respiratory-Related Hospital
Admission Studies 8-174
8-20 Percent Increase in Hospital Admissions per 10-|ig/m3 Increase in PM10
in 14 U.S. Cities (original and reanalyzed results) 8-176
8-21 Summary of United States PM25 Respiratory-Related Hospital
Admission Studies 8-180
8-22 Summary of United States PM10_25 Respiratory-Related Hospital
Admission Studies 8-181
8-23 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 8-182
8-24 Summary of United States PM10, PM2 5, and PM10_25 Asthma Medical
Visit Studies 8-189
8-25 Summary of Quantitative PFT Changes in Asthmatics per
50 |ig/m3 PM10 Increment 8-197
8-26 Summary of PFT Changes in Asthmatics per 25 |ig/m3 PM25 Increment.... 8-198
8-27 Summary of Asthma PM10 Cough Studies 8-200
8-28 Summary of Asthma PM10 Phlegm Studies 8-201
8-29 Summary of Asthma PM10 Lower Respiratory Illness Studies 8-201
8-30 Summary of Asthma PM10 Bronchodilator Use Studies 8-202
8-31 Summary of Asthma PM25 Respiratory Symptom Studies 8-205
8-32 Summary of Non-Asthma PM10 PFT Studies 8-207
8-33 Summary of Non-Asthma PM10 Respiratory Symptom Studies 8-208
-------
List of Tables
(cont'd)
Number Page
8-34 Summary of Non-Asthma PM25 Respiratory Outcome Studies 8-209
8-35 Summary of Non-Asthma Coarse Fraction Studies of
Respiratory Endpoints 8-210
8-36 PM10 Excess Risk Estimates From Reanalysis Studies for Total
Nonaccidental Mortality per 50 |ig/m3 Increase in PM10 8-228
8-37 Comparison of Maximum Single Day Lag Effect Estimates for PM2 5,
PM10_2 5, 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 8-233
8-38 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 Effect
Estimates Derived from Reanalyses Using More Stringent Convergence
Criteria and for Models Smoothed with More Degrees of Freedom 8-235
8-39 Effects of Different Models for Weather and Time Trends on Mortality
in Utah Valley Study 8-261
8-40 Summary Statistics Showing Mean Site-Pair Pearson Correlation
Coefficients, Annual Mean PM2 5 Concentrations (|ig/m3), the Range
in Annual Mean Concentrations (|ig/m3), Mean of 90th Percentile
Differences in Concentrations Between all Site Pairs (|ig/m3), and
Coefficients of Divergence for MS As Meeting Selection Criteria
Given in Appendix 3 A 8-292
8-41 Summary of Relative Homogeneity/Heterogeneity Characteristics for
MSAs Given in Table 8-40 8-294
8-42 Summary of Past Ecologic and Case-Control Epidemiologic Studies
of Outdoor Air and Lung Cancer 8-315
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
-------
List of Tables
(cont'd)
Number Page
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 Comparison of Ambient Particles, Fine Particles (ultrafine plus
accumulation-mode) and Coarse Particles 9-14
9-2 Exposure-Related Relationships for Particle Size Fractions 9-17
9-3 Particulate Matter Characteristics, Components, or Source Categories
Shown to be Associated with Mortality in U.S., Canadian, or European
Epidemiologic Studies 9-31
9-4 Prevalence of Selected Cardiorespiratory Disorders by Age Group and
by Geographic Region, 2000 (reported as percent or numbers of cases
in millions) 9-90
9-5 Number of Acute Respiratory Conditions per 100 Persons per Year,
by Age: United States, 1996 9-91
-------
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 (as percentage of the amount inhaled) in
humans as a function of particle size 6-16
6-6 Alveolar deposition (as percentage 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 jim MMAD; Panel B) and coarse (dp = 3 and
5 |im MMAD; Panels C and D) 6-21
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 (a) total respiratory tract, (b) extrathoracic region, (c) tracheobronchial
region, and (d) alveolar region 6-40
6-11 Major clearance pathways for particles deposited in the extrathoracic region
and tracheobronchial tree 6-44
6-12 Known and suspected (?) clearance pathways for poorly soluble particles
depositing in the alveolar region 6-45
II-xxiv
-------
List of Figures
(cont'd)
Number Page
6-13 Deposition fraction for total results of LUDEP model for an adult male
worker (ICRP default breathing parameters as shown in Table 6-3) showing
total percent deposition in the respiratory tract (TOT) and in the ET, TB,
and A regions: (a) nasal breathing (NB), (b) mouth breathing (MB),
(c) comparison of nasal and mouth breathing for TB and A regions 6-87
6-14 Deposition fraction for total results of LUDEP model for a young adult
showing total percent deposition in the respiratory tract (TOT) and in the
ET, TB, and A regions: (a) nasal breathing (NB), (b) mouth breathing
(MB), (c) comparison of nasal and mouth breathing for TB and A regions . . . 6-88
6-15 Comparison of deposition fraction in the TB and A regions for a worker
(WK; light exercise, ICRP default) and a young adult (YA; resting):
(a) nasal breathing and (b) mouth breathing 6-89
6-16 Comparison of regional deposition results from the ICRP (LUDEP) and
the MPPD models for a resting breathing pattern: (a) and (b), nose
breathing; (c) and (d), mouth breathing 6-93
6-17 Comparison of regional deposition results from the ICRP (LUDEP) and
the MPPD models for a light exercise breathing pattern: (a) and (b),
nose breathing; (b) and (c), mouth breathing 6-94
6-18 Dependency of aerosol deposition in human adults on physical exertion
expressed as minute ventilation for different particle sizes 6-98
6-19 Comparison of fractional deposition for rats (nasal breathing, at rest) and
humans (nasal and mouth breathing, light exercise) and the ratio of human
to rat for nasal and mouth breathing humans for the (a) ET, (b) TB, and
(c) A regions of the respiratory tract 6-101
6-20 Normalized deposition patterns for rats (nasal breathing) and humans
(nasal breathing and mouth breathing) and the ratio of human to rat 6-103
7-1 Schematic illustration of hypothesized pathways/mechanisms potentially
underlying the cardiovascular effects of PM 7-9
7-2 Simplified overview of blood coagulation system 7-12
II-XXV
-------
List of Figures
(cont'd)
Number Page
7A-la,b,c Size distributions of the Aitken, accumulation, and coarse modes of the
average urban aerosol (as reported by Whitby [1978]) and a resuspended
PM mode: (a) mass distribution, (b) surface area distribution, and
(c) number distribution 7A-14
7A-2a,b,c The ratio of predicted the deposition fractions for human relative to rat
at rest, DFH/DFR, (a) the head region, (b) the TB region, and (c) the A
region for nasal breathing corrected for particle inhalability 7A-17
7A-3a,b,c The ratio of the predicted deposition fractions for human relative to rat
at rest, DFH/DFR, (a) the head region, (b) the TB region, and (c) the A
region for oral breathing corrected for particle inhalability 7A-18
7A-4 Inhalability curves for human and rat showing the fraction of PM that
enters the nose (based on empirical fit to experimental data given in
Table 1 from Menache et al., 1995) 7A-19
7A-5 Predicted clearance curves for the TB region for poorly soluble particles
for (a) human and (b) rat 7A-20
7A-6 Alveolar region clearance curves for measured poorly soluble particles
in several different species 7A-21
7A-7a,b Burden in the (a) TB and (b) A region of the lung normalized to the
total particle mass predicted to deposit in these respective regions
during a 3-day (6 h/day) exposure 7A-22
7A-8 Mass burden of poorly soluble PM predicted for the A region of
(a) human and (b) rat 7A-24
7A-9 Mass of poorly soluble PM predicted to be retained in the alveolar (A)
region of the rat lung as a fraction of the total mass deposited in the
A region during a 6-month exposure 7A-40
7A-10 Mass of poorly soluble PM predicted to be retained in the alveolar (A)
lung region normalized to A surface area 7A-41
7A-11 Mass of poorly soluble PM predicted to be retained in the alveolar (A)
lung region normalized to A surface area 7A-42
7A-12 Mass of poorly soluble PM predicted to be retained in the alveolar (A)
lung region normalized to A surface area 7A-43
II-xxvi
-------
List of Figures
(cont'd)
Number Page
8-1 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
8-2 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
8-3 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 8-2 [[[ 8-34
8-4 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 [[[ 8-36
8-5 Percent excess risks estimated per 25 |ig/m3 increase in PM2 5 or PM
10_2 5
from new studies that evaluated both PM2 5 and PM10_2 5, based on single
pollutant (PM only) models ........................................ 8-61
8-6 Excess risks estimated per 5 |ig/m3 increase in sulfate, based on U.S.
studies for which both PM2 5 and PM10_2 5 data were available ............. 8-72
8-7 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 ....................... 8-99
8-8 Relative risk of total and cause-specific mortality per 10 |ig/m3 PM25,
derived for means of 1979-1983 PM2 5 data for various cities, using
alternative statistical models ...................................... 8-100
8-9 Relative risk of total and cause-specific mortality for particle metrics
and gaseous pollutants over different averaging periods (years 1979-2000
in parentheses) ................................................. 8-101
8-10 Acute cardiovascular hospitalizations and PM exposure excess risk
estimates derived from U.S. PM10 studies based on single-pollutant
models from GAM strict covergence criteria reanalyses (2003 studies)
or alternative (non-GAM) original analyses .......................... 8-159
8-11 Percent change in hospital admission rates and 95% CIs for an IQR
-------
List of Figures
(cont'd)
Number Page
8-12 Maximum excess risk of respiratory-related hospital admissions and
visits per 50 |ig/m3 PM10 increment in studies of U.S. cities based on
single-pollutant models 8-193
8-13 Illustrative acute pulmonary function change studies of
asthmatic children 8-199
8-14 Odds ratios with 95% confidence interval for cough per 50-|ig/m3
increase in PM10 for illustrative asthmatic children studies at lag 0 8-203
8-15 PM10 excess risk estimates for total nonaccidental mortality for
numerous locations (and for cardiovascular mortality for Coachella
Valley, CA and Phoenix, AZ), using: (1) GAM with default
convergence criteria; (2) GAM with stringent convergence criteria;
and, (3) GLM/natural splines that approximate the original GAM
model from the GAM reanalysis studies 8-230
8-16 Excess risk estimates for total nonaccidental mortality in single-pollutant
(PM only) and multipollutant models 8-248
8-17 Excess risk estimates for cardiovascular-related effects, including
mortality, hospital admissions, and changes in biomarkers (e.g., increases
in blood parameters or decreases in heart rate variability measures) in
single-pollutant (PM only) and multipollutant models 8-249
8-18 Excess risk estimates for respiratory-related effects, including mortality,
hospital admissions and medical visits in single-pollutant (PM only)
and multipollutant models 8-250
8-19 Excess risk estimates for increases in respiratory symptoms or decreases in
lung function measures in single-pollutant (PM only) and multipollutant
models 8-251
8-20 PM10 (lag 1 day) coefficient (P) for total mortality, for 1992-1994, as a
function of alternative weather models and varying degrees of freedom
for fitting temporal trends using natural splines 8-265
8-21 PM10 (lag 1 day) coefficient (P) for hospital admissions for pneumonia
among the elderly, for 1992-1994, as a function of alternative weather
models and varying degrees of freedom for fitting temporal trends
using natural splines 8-266
-------
List of Figures
(cont'd)
Number Page
8-22 Odds ratios (and 95% confidence intervals) for associations between
onset of myocardial infarction and 25 |ig/m3 increase in hourly or
daily 24-h average PM25 concentrations 8-271
8-23 Marginal posterior distribution for effects of PM10 on all-cause mortality
at lag 0, 1, and 2 for the 90 cities 8-272
8-24 Excess risk estimates for associations between various health outcomes
and PM10 (50 |ig/m3 increment) from different studies conducted in
Cook County, IL 8-274
8-25 Excess risk estimates for associations between various health outcomes
and PM10 (50 |ig/m3 increment) from studies conducted in Los Angeles
County, CA 8-275
8-26 Excess risk estimates for associations between various health
outcomes and PM10 (50 |ig/m3 increment) from studies conducted in
Pittsburgh, PA 8-276
8-27 Excess risk estimates for associations between various health outcomes
and PM10 (50 |ig/m3 increment) from studies conducted in Detroit, MI 8-277
8-28 Excess risk estimates for associations between various health outcomes
and PM10 (50 |ig/m3 increment) from studies conducted in Seattle or
King County, WA 8-278
8-29 Relative risk estimates and 95% confidence intervals for total mortality
per 100 |ig/m3 increase in PM10, adjusting for ozone, temperature,
seasonal cycles, day of week, and linear trend for 1985-1990 in
Cook County, IL 8-297
8-30 Concentration-response curves for PM10 mortality relationships in
20 largest U.S. cities (1987-1994), for total mortality, cardiovascular
and respiratory mortality, and other-causes mortality 8-321
8-31 Posterior probabilities of thresholds for each cause-specific mortality and
for mean PM10, 20 largest U.S. cities, 1987-1994 8-321
9-1 An idealized size distribution, as might be observed in traffic, showing
fine and coarse particles and the nucleation, Aitken, and accumulation
modes that comprise fine particles 9-11
II-xxix
-------
List of Figures
(cont'd)
Number Page
9-2 Geometric mean infiltration factor (indoor/outdoor ratio) for hourly
nighttime, nonsource data for two seasons 9-16
9-3 Deposition fraction as a function of particle size for nasal and
oral breathing during rest and exercise: (a) extrathoracic (ET),
(b) tracheobronchial (TB), and (c) alveolar (A) regions 9-19
9-4 Excess risk estimates for total nonaccidental, cardiovascular, and
respiratory mortality in single-pollutant models for U.S. and Canadian
studies, including aggregate results from two multicity studies 9-26
9-5 Excess risk estimates for hospital admissions and emergency department
visits for cardiovascular and respiratory diseases in single-pollutant
models from U.S. and Canadian studies, including aggregate results
from one multicity study 9-27
9-6 Illustration of the nitrogen cascade showing the movement of the
human-produced reactive nitrogen (Nr) as it cycles through the
various environment reservoirs in the atmosphere, terrestrial
ecosystems, and aquatic ecosystems 9-103
II-XXX
-------
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 Brown—Health Scientist, National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Wilson—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributing Authors
Dr. Lester D. Grant—National Center for Environmental Assessment (B243-01),
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. James McGrath—Chapel Hill, NC 27517
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 (B143-02),
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 (B 143-02),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Mark Frampton—University of Rochester, 601 Elmwood Avenue, Box 692, Rochester, NY
14642
II-xxxi
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont1 d)
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
Dr. Judith Graham—American Chemistry Council, 1300 Wilson Boulevard, Arlington, VA
22207
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 (B143-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
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. John Stanek—National Center for Environmental Assessment (B243-01),
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. John J. Vandenberg—National Center for Environmental Assessment (8601 D), U.S.
Environmental Protection Agency, 1200 Pennsylvania Avenue, NW, Washington, DC 20460
Dr. William Watkinson—National Health and Environmental Effects Research Laboratory
(B 143-02), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
II-xxxii
-------
Authors, Contributors, and Reviewers
(cont1 d)
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
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
Dr. Lester D. Grant—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lori White—National Center for Environmental Assessment (B243-01), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Dr. James Brown—Health Scientist, National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William E. Wilson—National Center for Environmental Assessment (B143-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributing Authors
Dr. Dan Costa—National Health and Environmental Effects Research Laboratory (B 143-02),
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
Mr. James Raub—National Center for Environmental Assessment (B 143-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
-------
Authors, Contributors, and Reviewers
(cont'd)
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. Kevin Dreher—National Health and Environmental Effects Research Laboratory (B143-02),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Janice Dye—National Health and Environmental Effects Research Laboratory (B 143-02),
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. Stephen Gavett—National Health and Environmental Effects Research Laboratory
(B 143-02), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Andrew Ohio—National Health and Environmental Effects Research Laboratory (MD-58D),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Ian Gilmour—National Health and Environmental Effects Research Laboratory (B 143-04),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
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
(B 143-02), 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
II-xxxiv
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
Dr. Ted Martonen—National Health and Environmental Effects Research Laboratory (B143-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Kent Pinkerton—University of California, ITEH, One Shields Avenue, Davis, CA 95616
Mr. William 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
Dr. John J. Vandenberg—National Center for Environmental Assessment (8601D), U.S.
Environmental Protection Agency, 1200 Pennsylvania Avenue, NW, Washington, DC 20460
Dr. Bellina Veronesi—National Health and Environmental Effects Research Laboratory
(B105-06), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Watkinson—National Health and Environmental Effects Research Laboratory
(B 143-02), 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. Kazuhiko Ito—New York University Medical Center, Institute of Environmental Medicine,
Long Meadow Road, Tuxedo, NY 10987
Dr. George Thurston—New York University Medical Center, Institute of Environmental
Medicine, Long Meadow Road, Tuxedo, NY 10987
II-XXXV
-------
Authors, Contributors, and Reviewers
(cont'd)
Principal Authors
(cont'd)
Dr. Patrick Kinney—Columbia University, 60 Haven Avenue, B-l, Room 119, New York, NY
10032
Dr. Vic Hasselblad—Durham, NC 27713
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 (B243-02),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributing Authors
Dr. David Svendsgaard —Statistician, National Caucus and Center on Black Aged, Inc., Senior
Environmental Employment Program, Washington, DC 20005
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
Dr. Raymond Carroll—Texas A & M University, Department of Statistics, College Station, TX
77843-3143
Dr. Wayne E. Casio—University of North Carolina School of Medicine, Chapel Hill, NC 27517
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
II-xxxvi
-------
Authors, Contributors, and Reviewers
(cont'd)
Contributors and Reviewers
(cont'd)
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
Dr. John Vandenberg—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Wilson—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
II-xxxvii
-------
Authors, Contributors, and Reviewers
(cont1 d)
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
Contributing Authors
Dr. James Brown—Health Scientist, National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Beverly Comfort—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. J.H.B. Garner—National Center for Environmental Assessment (B243-01),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Brooke Hemming—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
Dr. Lori White—National Center for Environmental Assessment (B243-01), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
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
II-xxxviii
-------
Authors, Contributors, and Reviewers
(cont1 d)
Contributors and Reviewers
(cont'd)
Dr. Mary Ross—Office of Air Quality Planning and Standards (C539-01), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Mr. William Russo—National Health and Environmental Effects Research Laboratory
(B305-02), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. John J. Vandenberg—National Center for Environmental Assessment (8601 D), U.S.
Environmental Protection Agency, 1200 Pennsylvania Avenue, NW, Washington, DC 20460
II-xxxix
-------
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
Dr. James Brown—Health 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. Brooke L. Hemming—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
Dr. Lori White—National Center for Environmental Assessment (B243-01), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
II-xl
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
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
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
Ms. Carolyn T. Perry—Manager/Word Processor, Computer Sciences Corporation, 2803 Slater
Road, Suite 220, Morrisville, NC 27560
Dr. Barbara Liljequist—Technical Editor, Computer Sciences Corporation, 2803 Slater Road,
Suite 220, Morrisville, NC 27560
Dr. Carol A. Seagle—Technical Editor, 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
Mr. Matthew Kirk—Graphic Artist, Computer Sciences Corporation, 2803 Slater Road,
Suite 220, Morrisville, NC 27560
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Document Production Staff
(cont'd)
Mr. John A. Bennett—Technical Information Specialist, Library Associates of Maryland,
11820 Parklawn Drive, Suite 400, Rockville, MD 20852
Ms. Sandra L. Hughey—Technical Information Specialist, Library Associates of Maryland,
11820 Parklawn Drive, Suite 400, Rockville, MD 20852
Ms. Rebecca Caffey—Records Management Technician, Reference Retrieval and Database
Entry Clerk, InfoPro, Inc., 8200 Greensboro Drive, Suite 1450, McLean, VA 22102
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD (SAB) STAFF OFFICE
CLEAN AIR SCIENTIFIC ADVISORY COMMITTEE (CASAC)
PARTICULATE MATTER REVIEW PANEL*
CASAC Chair
Dr. Philip Hopke—Bayard D. Clarkson Distinguished Professor, Department of Chemical
Engineering, Clarkson University, Box 5708, Potsdam, NY 13699-5708
CASAC Members
Dr. Ellis Cowling—University Distinguished Professor At-Large, North Carolina State
University, Colleges of Natural Resources and Agriculture and Life Sciences, North Carolina
State University, 1509 Varsity Drive, Raleigh, NC 27695-7632
Dr. James D. Crapo—Chairman, Department of Medicine, National Jewish Medical and
Research Center, 1400 Jackson Street, Denver, CO, 80206, and Chief Executive Officer (CEO)
of Aeolus Pharmaceuticals, Inc.
Dr. Frederick J. Miller—Vice President for Research, CUT Centers for Health Research, 6 Davis
Drive, P.O. Box 12137, Research Triangle Park, NC 27709
Mr. Richard L. Poirot—Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Bldg. 3 South, 103 South
Main Street, Waterbury, VT 05671-0402
Dr. Frank Speizer—Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, 181 Longwood Avenue, Boston, MA 02115-5804
Dr. Barbara Zielinska—Research Professor , Division of Atmospheric Science, Desert Research
Institute, 2215 Raggio Parkway, Reno, NV 89512-1095
* Members of this CASAC Panel consist of:
a. CASAC Members: Experts appointed to the statutory Clean Air Scientific Advisory Committee by
the EPA Administrator; and
b. CASAC Consultants: Experts appointed by the SAB Staff Director to serve on one of the
CASAC's National Ambient Air Quality Standards (NAAQS) Panels for a particular criteria air pollutant.
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD (SAB) STAFF OFFICE
CLEAN AIR SCIENTIFIC ADVISORY COMMITTEE (CASAC)
PARTICULATE MATTER REVIEW PANEL*
(cont'd)
CASAC Consultants
Dr. Jane Q. Koenig—Professor, Department of Environmental Health, School of Public Health
and Community Medicine, University of Washington, Box 357234, Seattle, WA 98195-7234
Dr. Petros Koutrakis—Professor of Environmental Science, Environmental Health , School of
Public Health, Harvard University, HSPH, 401 Park Dr., Room 410 West, Boston, MA 02215
Dr. Allan Legge—President, Biosphere Solutions, 1601 11th Avenue NW, Calgary, Alberta,
CANADA T2N 1H1
Dr. Paul J. Lioy—Associate Director and Professor, Environmental and Occupational Health
Sciences Institute, UMDNJ - Robert Wood Johnson Medical School, 170 Frelinghuysen Road,
Rm 301, Piscataway, NJ 08854
Dr. Morton Lippmann—Professor, Nelson Institute of Environmental Medicine, New York
University School of Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. Joe Mauderly—Vice President, Senior Scientist, and Director, National Environmental
Respiratory Center, Lovelace Respiratory Research Institute, 2425 Ridgecrest Drive SE,
Albuquerque, NM, 87108
Dr. Roger O. McClellan—Consultant, 1370 Quaking Aspen Pine, Albuquerque, NM, 87111
Dr. Gunter Oberdorster—Professor of Toxicology, Department of Environmental Medicine,
School of Medicine and Dentistry, University of Rochester, 575 Elmwood Avenue, Box 850,
Rochester, NY 14642
Dr. RobertD. Rowe—President, Stratus Consulting, Inc.,, PO Box 4059, Boulder, CO
80306-4059
Dr. Jonathan M. Samet—Professor and Chair, Department of Epidemiology, Bloomberg School
of Public Health, Johns Hopkins University, 615 N. Wolfe Street, Suite W6041, Baltimore, MD
21205-2179
Dr. Sverre Vedal—Professor of Medicine, School of Public Health and Community Medicine
Department of Environmental and Occupational Health Sciences, University of Washington,
4225 Roosevelt Way NE, Suite 100, Seattle, WA 98105-6099
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD (SAB) STAFF OFFICE
CLEAN AIR SCIENTIFIC ADVISORY COMMITTEE (CASAC)
PARTICULATE MATTER REVIEW PANEL*
(cont'd)
CASAC Consultants
(cont'd)
Mr. Ronald White—Research Scientist, Epidemiology, Bloomberg School of Public Health,
Room W6035, The Johns Hopkins University, 615 N. Wolfe St., Rm W6035, Baltimore, MD
21205
Dr. Warren H. White—Visiting Professor, Crocker Nuclear Laboratory, University of
California-Davis,, Davis, CA 95616-8569
Dr. George T. Wolff—Principal Scientist, General Motors Corporation, 300 Renaissance Center
(MC482-C27-B76), Detroit, MI 48265-3000
EPA Science Advisory Board Staff
Dr. Vanessa Vu—SAB Staff Office Director, EPA Science Advisory Board Staff Office
(Mail Code 1400F), 1200 Pennsylvania Avenue, N.W., Washington DC, 20460
Mr. Fred Butterfield—CASAC Designated Federal Officer, EPA Science Advisory Board Staff
Office (Mail Code 1400F), 1200 Pennsylvania Avenue, N.W., Washington, DC 20460
II-xlv
-------
Abbreviations and Acronyms
A
AC
ACE
ACS
ADP
ADS
AED
AHSMOG
AI
AIC
AM
AOR
APHEA
AQCD
ASOS
BAD
BAL
BALF
BaP
BAUS
BB
bb
BIC
BMI
BN
BNF
alveolar
air conditioning
angiotensin-converting enzyme
American Cancer Society
platelet aggregation
anatomic dead space
aerodynamic equivalent diameter
Adventist Health Study on Smog
alveolar-interstitial region
Akaike Information Criterion
alveolar macrophage
adjusted odds ratio
Air Pollution and Health: a European Approach
Air Quality Criteria Document
Automated Surface Observing System
brachial artery diameter
bronchoalveolar lavage
brochoalveglar lavage fluid
benzo[a]pyrene
brachial artery ultrasonography
bronchial region
bronchiolar region
Bayes Information Criterion
body mass index
Brown Norway
biological nitrogen fixation
-------
BP
BrdU
BS
BW
CAPs
CARS
CAT
CB
CB
CC
CESAR
CF
CFA
CFC
CFD
CHF
CHO
CI
CL
CMD
CMP
COD
CoH, COH
COPD
CP
CPC
CPZ
CRC
blood pressure
5'-bromo-2'deoxyuridine
black smoke
bronchial wash
concentrated ambient particles
California Air Resources Board
computer-aided tomography
carbon black
chronic bronchitis
conventional combustion
Central European Air Quality and Respiratory Health
cystic fibrosis
coal fly ash
chl orofluorocarb on
computational fluid dynamics
congestive heart failure
Chinese hamster ovary
confidence interval
chemiluminescence
count mean diameter
copper smelter dust
coefficient of determination
coefficient of haze
chronic obstructive pulmonary disease
coarse particle
condensation particle counter
capsazepine
contributing respiratory causes
-------
CrD
CRV
CVD
CVDRESP
CVM
DBF
DCFH
DCM
DE
DE
DEF
DEP
df
DF
DFPSS
DHR
DMTU
DOFA
DPL
DPM
DRG
DTPA
DYS
EC
ECG
ED
EOF
ELF
cerebrovascular disease
cerebrovascular disease
cardiovascular disease
cardiorespiratory
cardiovascular mortality
diastolic blood pressure
dichlorofluorescin
dichloromethane
diesel exhaust
deposition efficiency
deferoxamine
diesel exhaust particles
degrees of freedom
deposition fraction
dual fine particle sequential sampler
dihy drorhodamine-123
dimethylthiourea
domestic oil fly ash
dipalmitoyl lecithin
diesel particulate matter; diesel soot
dorsal root ganglia
techetium-diethylenetriamine-pentaacetic acid
dysrhythmias
elemental carbon
el ectrocardi ogram; el ectrocardi ographi c
emergency department
epidermal growth factor
epithelial lining fluid
-------
EOC
EqER
ER
ERK
ESR
ET
ET
EU
FA
FBC
FEF
FEVj
FMD
FP
FRC
FVC
GAM
GEE
GJIC
GLM
GM
GMCSF
GMPD
GP
GSF
GSH
HA
HC
equivalent organic carbon
equivalent exposure ratio
excess risk
extracellular receptor kinase; extracellular signal-regulated kinase
electron spin resonance
endothelin
extrathoracic
endotoxin units
filtered air
fluidized-bed combustion
forced expiratory flow
forced expiratory volume in 1 second
flow-mediated dilation
fine particle
functional reserve capacity
forced vital capacity
Generalized Additive Model
generalized estimating equations
gap-junctional intercellular communications
Generalized Linear Model
gestation month
granulocyte macrophage colony stimulating factor
geometric mean particle diameter
general practice
Gessellschaft fur Strahlenforschung
glutathione
hospital admission
hydrocarbon
-------
HD
HDM
HEI
HF
HPLC
HR
HRV
HULIS
ICAM-1
ICD9
ICRP
IDF
IgE
IgG
IHD
IL
IMF
IMPROVE
iNOS
IP
IP
IPM
IPN
IQR
IT
IUGR
IV
heavy duty
house dust mite
Health Effects Institute
high frequency
high-pressure liquid chromatography
heart rate
heart rate variability
humic-like substances
inhibitory kappa B alpha
intercellular adhesion molecule-1
International Classification of Disease
International Commission on Radiological Protection
intrathoracic deposition fraction
immunoglobin E
immunoglobin G
ischemic heart disease
interleukin
induced mutant fraction
Interagency Monitoring of Protected Visual Environments
inducible nitric oxide synthase
intraperitoneal
inhalable particle
inhalable particle mass
inhalable particle network
interquartile range
intratracheal
intrauterine growth retardation
intravenous
II-l
-------
JNK
KS
LEW
LCL
LDH
LF
LFA-1
LN
LOEL
LOESS
LPS
LRD
LRI
LUDEP
MAPK
MAS
MC
MCh
MCT
MEK
MI
MIP
MMAD
MMD
MMPs
MO
MONICA
MPL
c-jun N-terminal kinase
soil-corrected potassium
low birth weight
lower 95th% confidence limit
lactate dehydrogenase
low frequency
leukocyte function-associated antigen-1
lymph nodes
lowest observed effect level
local regression smoothers
lipopolysaccharide
lower respiratory disease
lower respiratory illness
Lung Dose Evaluation Program
mitogen-activated protein kinase
Mobile Aerosol Spectrometer
mass concentration
methylcholine
monocrotaline
mitogen-activated protein kinase/ERK kinase
myocardial infarction
macrophage inflammatory protein
mean median aerodynamic diameter
mass median diameter
matrix metalloproteinases
monocyte
monitoring of trends and determinants in cardiovascular disease
multipath lung
Il-li
-------
MPO myeloperoxidase
MPPD multiple path particle dosimetry
MSA Metropolitan Statistical Area
MSH Mount St. Helens
MVRD motor vehicle and resuspended dust
n, N number
NAAQS National Ambient Air Quality Standards
NAC N-acetylcysteine (antioxidant)
NAL nasal lavage fluid
NC number concentration
NCEA-RTP National Center for Environmental Assessment-Research Triangle Park
NCRP National Council on Radiation Protection and Measurement
NF nuclear factor
NF-KB nuclear factor kappa B
NHANES National Health and Nutrition Examination Surveys
NHBE normal human bronchial epithelial
NIST National Institutes of Standards Technology
NLCS Netherlands Cohort Study on Diet and Cancer
NMD nitroglycerine-mediated dilation
NMMAPS National Morbidity, Mortality, and Air Pollution Study
NMRI Naval Medical Research Institute
N-N normal-to-normal
NOX nitrogen oxide
NOEL no observed effect level
NOPL nasa-oro-pharyngo-laryngeal
NS natural splines
n.s. statistically nonsignificant
nspline natural splines
-------
OAA
OAQPS
OC
OR
OLS
OVA
P90
P
PAC
PAH
PB
PBW
PDGF
PDL
PEF
PEFR
PFT
PHS-2
PM
PMN
PN
P°
poly I:C
post-Mi
PTEAM
PTFE
PVCs
Q
Ottawa ambient air
Office of Air Quality Planning and Standards
organic carbon
odds ratio
ordinary least squares
ovalbumin
90th percentile value
pulmonary
polyaromatic compound
polycyclic aromatic hydrocarbon
polymyxin-B
particle-bound water
platelet-derived growth factor
polynomial distributed lag
peak expiratory flow
peak expiratory flow rate
pulmonary function tests
prostaglandin H synthase-2
particulate matter
polymorphonuclear leukocyte
penalized splines
equilibrium vapor pressure
polyinosinic-polycytidilic acid
post-myocardial infarction
Particle Total Exposure Assessment Methodology
polytetrafluoroethylene (Teflon)
premature ventricular complexes
respiratory flow rate
-------
QHIP
RBCs
rENP
r-MSSD
RCAL
RIVM
RME
ROFA
ROI
ROS
RR
RTD
RTE
RWC
SAD
SBP
SCO
SCE
SD
SDANN
SDNN
SH
SIMEX
SL
SOD
SP-A
SPM
Sp02
Quebec Health Insurance Plan
red blood cells
recombinant endotoxin-neutralizing protein
root mean squared differences between adjacent normal-to-normal
heartbeat intervals
Regression Calibration
Dutch National Institute of Public Health and the Environment
rapeseed oil methyl ester
residual oil fly ash
reactive oxygen intermediates
reactive oxygen species
relative risk
road tunnel dust
rat tracheal epithelial
residential wood combustion
small airway disease
systolic blood pressure
sudden cardiac death
sister chromatid exchanges
Sprague-Dawley
standard deviation of the average of normal-to-normal heartbeat intervals
standard deviation of normal-to-normal heartbeat intervals
spontaneously hypertensive
Simulation Extrapolation
stochastic lung
superoxide dismutase
surfactant protein A
synthetic polymer microspheres
oxygen saturation
Il-liv
-------
SoCARB South Coast Air Basin
SSC spatial synoptic category
SVOC semivolatile organic compound
TC total carbon
TC tungstan carbide
T[CO] core temperature
TB tracheabronchial
TDF total deposition fraction
TEOM tapered element oscillating microbalance
TIMP tissue inhibitor of metalloproteinase
TK thymidine kinase
TLC total lung capacity
TLR Toll-like receptors
TNF tumor necrosis factor
TPA tissue plasminogen activator
TSI Temporal Synoptic Index
TSP total suspended paniculate
TV tidal volume
UAP urban air particles
UCL upper 95th% confidence limit
UDS unscheduled DNA synthesis
ufCB, UCB ultrafme carbon black
UFP ultrafme fluorospheres
URT upper respiratory tract
UV-B ultraviolet-B radiation
UVD Utah Valley dust
VA Veterans' Administration
VAPS Versatile Air Pollution Samplers
II-lv
-------
VCAM-1 vascular cell adhesion molecule-1
VLBW very low body weight
V, tidal volume
WBC white blood cell
WINS Well Impactor Ninety-Six
WIS Wistar
WKY Wistar-Kyoto
XRF X-ray fluororescence
-------
6. DOSIMETRY OF PARTICULATE MATTER
6.1 INTRODUCTION
The proximal cause of a biological response to particulate matter (PM) is due to the dose
deposited at the target site rather than the external exposure. Characterization of the exposure-
dose-response continuum for PM requires an understanding of the mechanistic determinants of
inhaled particle dose. Furthermore, dosimetric information is critical for extrapolating human
health effects based on animal toxicological studies and for comparing results from controlled
clinical studies involving healthy human subjects or those with preexisting disease.
Dose to a target tissue depends on the initial deposition and subsequent retention of
particles within the respiratory tract. Once particles have deposited onto the surfaces of the
respiratory tract, they are subsequently subjected to either absorptive or nonabsorptive
particulate removal processes, which may result in their removal or translocation from airway
surfaces, as well as their removal from the respiratory tract itself. Clearance of deposited
particles depends upon the initial site of deposition and upon the physicochemical properties of
the particles, both of which affect specific translocation pathways. Retained particle burdens are
determined by the dynamic relationship between deposition and clearance rates.
This chapter discusses particle dosimetry, the study of the deposition, translocation,
clearance, and retention of particles within the respiratory tract and extrapulmonary tissues.
It summarizes basic concepts as presented in Chapter 10 of the 1996 EPA document, Air Quality
Criteria for Particulate Matter or (1996 PM AQCD) (U.S. Environmental Protection Agency,
1996a); and it updates the state of the science based upon new literature appearing since
publication of the 1996 PM AQCD. Although our understanding of the basic mechanisms
governing deposition and clearance of inhaled particles has not changed, there has been
significant additional information on the role of certain biological determinants of the
deposition/clearance processes, such as gender, age and lung disease. Additionally, the
understanding of regional dosimetry within the respiratory tract and the particle size range over
which this has been evaluated has been expanded.
The dose of inhaled particles to the respiratory tract is governed by a number of factors.
These include exposure concentration, exposure duration, respiratory tract anatomy, ventilatory
6-1
-------
parameters, and particle properties (e.g., particle size, hygroscopicity, and solubility in airway
fluids and cellular components). The basic characteristics of particles as they relate to deposition
and retention, as well as anatomical and physiological factors influencing particle deposition and
retention, were discussed in depth in the 1996 PM AQCD. Thus, in this chapter, only an
overview of basic information related to one critical factor in deposition, namely particle size, is
provided (Section 6.1.1) to allow the reader to understand the different terms used in the
remainder of this and subsequent chapters dealing with health effects. Section 6.1.2 provides a
basic overview of respiratory tract structure as it relates to particle deposition and clearance.
The ensuing major sections of this chapter provide updated information on particle deposition,
clearance, and retention in the respiratory tract of humans, as well as laboratory animals, that are
useful in evaluating of PM health effects. Issues related to the phenomenon of particle overload
as it may apply to human exposure and the use of instillation of particle suspensions as an
exposure technique to evaluate PM health effects also are discussed. The final sections of the
chapter deal with mathematical models of particle deposition and clearance in the respiratory
tract.
It must be emphasized that any dissection into discrete topics or factors that control dose
from inhaled particles tends to mask the dynamic and interdependent nature of the intact
respiratory system. For example, although deposition is discussed separately from clearance
mechanisms, retention (i.e., the actual amount of particles found in component regions of the
respiratory tract at any point in time) is, as noted previously, determined by the relative rates of
both deposition and clearance. Thus, an overall dosimetric assessment requires integration of
these various components of the overall process. In summarizing the literature on particle
dosimetry, when applicable, changes from control are described if they were statistically
significant at a p-value of less than 0.05 (i.e., p < 0.05). When trends are described, actual p
values given in the published reports will be provided if possible.
6.1.1 Size Characterization of Inhaled Particles
Particle size is an important determinant of the fraction of inhaled particles deposited in the
various regions of the respiratory tract. Particle attributes, as well as some general definitions
important in understanding particle fate within the respiratory tract, are described in Chapter 2.
6-2
-------
Most aerosols present in natural and work environments are polydisperse. This means that
the constituent particles within an aerosol have a range of sizes and are more appropriately
described in terms of size distribution parameters. The lognormal distribution (i.e., the situation
in which the logarithms of particle diameter [dp] are distributed normally) can be used for
describing size distributions of most aerosols. The geometric mean is the median of the
distribution, and the metric of variability around this central tendency is the geometric standard
deviation (og). The og, a dimensionless term, is the ratio of the 84th (or 16th) percentile particle
size to the 50th percentile size. Thus, the only two parameters needed to describe a lognormal
distribution of particle sizes for a specific aerosol are the median diameter and the geometric
standard deviation. However, the actual size distribution may be obtained in various ways.
When a distribution is described by counting particles, the median is called the count median
diameter (CMD). On the other hand, the median of a distribution based on particle mass in an
aerosol is the mass median diameter (MMD). When using aerodynamic diameters, a term that is
encountered frequently is mass median aerodynamic diameter (MMAD), which is the median of
the distribution of mass with respect to aerodynamic equivalent diameter (AED). Most of the
present discussion will focus on MMAD because it is the most commonly used measure of
aerosol distribution. However, alternative descriptions should be used for particles with actual
physical sizes below «0.5 jim because, for these, aerodynamic properties become less important.
One such metric is thermodynamic-equivalent size, i.e., the diameter of a spherical particle that
has the same diffusion coefficient in air as the particle of interest.
6.1.2 Structure of the Respiratory Tract
A detailed discussion of respiratory tract structure was provided in the 1996 PM AQCD
(U.S. Environmental Protection Agency, 1996a), and only a brief synopsis is presented here.
For dosimetry purposes, the respiratory tract can be divided into three regions (Figure 6-1):
(1) extrathoracic (ET), (2) tracheobronchial (TB), and (3) alveolar (A). The ET region consists
of airways within the head (i.e., nasal and oral passages) through the larynx and represents the
areas through which inhaled air first passes. In humans, inhalation can occur through the nose or
mouth (or both, known as oronasal breathing). However, most laboratory animals commonly
used in respiratory toxicological studies are obligate nose breathers.
-------
Extrathoracic
Region
Posterior
Nasal Passage
Nasal Part
Pharynx-4 ~ ,,-, .
' I Oral Part
Trachea
Tracheobronchial
Region
Main Bronchi
Bronchi
Bronchioles
Alveolar
Region
Bronchiolar Region
Alveolar Interstitial
Bronchioles
Terminal Bronchioles
Respiratory Bronchioles
Alveolar Duct +
Alveoli
Figure 6-1. Diagrammatic representation of respiratory tract regions in humans.
Source: Based on International Commission on Radiological Protection (1994) and U.S. Environmental
Protection Agency (1996a).
From the ET region, inspired air enters the TB region at the trachea. From the level of the
trachea, the conducting airways then undergo dichotomous branching for a number of
generations. The terminal bronchiole is the most peripheral of the distal conducting airways and,
in humans, leads to the gas-exchange region, which consists of respiratory bronchioles, alveolar
6-4
-------
ducts, alveolar sacs, and alveoli (all of which comprise the A region). All of the conducting
airways, except the trachea and portions of the mainstem bronchi, are surrounded by
parenchymal tissue composed primarily of the alveolated structures of the A region and
associated blood and lymphatic vessels. It should be noted that the respiratory tract regions are
comprised of various cell types and that there are distinct differences in the distribution of cells
lining the airway surfaces in the ET, TB, and A regions. Although a discussion of cellular
structure of the respiratory tract is beyond the scope of this section, details may be found in a
number of sources (e.g., Crystal et al., 1997).
6.2 PARTICLE DEPOSITION
This section discusses the deposition of particles in the respiratory tract. It begins with an
overview of the basic physical mechanisms that govern deposition. This is followed by an
update on both total respiratory tract and regional deposition patterns in humans. Some critical
biological factors that may modulate deposition are then presented. The section ends with a
discussion of issues related to interspecies patterns of particle deposition.
6.2.1 Mechanisms of Deposition
Particles may deposit within the respiratory tract by five mechanisms: (1) inertial
impaction, (2) sedimentation, (3) diffusion, (4) interception, and (5) electrostatic precipitation.
Sudden changes in airstream direction and velocity may cause some particles to fail to
follow the streamlines of airflow. As a consequence, the particles contact, or impact, airway
surfaces. The ET and upper TB airways are characterized by high air velocities and sharp
directional changes and, thus, dominate as sites of inertial impaction. Impaction is a significant
deposition mechanism for particles larger than 2 jim AED.
All aerosol particles are continuously influenced by gravity, but particles with an
AED > 1 |im are affected to the greatest extent. A particle will acquire a terminal settling
velocity when a balance is achieved between the acceleration of gravity acting on the particle
and the viscous resistance of the air, and this settling out of the airstream may bring it into
contact with airway surfaces. Both sedimentation and inertial impaction can influence the
deposition of particles within the same size range. These deposition processes act together in the
6-5
-------
ET and TB regions: inertial impact!on dominates in the upper airways, and gravitational settling
becomes increasingly dominant in the smaller conducting airways.
Particles having actual physical diameters < 1 |im are increasingly subjected to diffusive
deposition because of random bombardment by air molecules, resulting in contact with airway
surfaces. The root mean square displacement that a particle experiences in a unit of time along a
given cartesian coordinate is a measure of its diffusivity. The density of a particle is unimportant
in determining its diffusivity. Thus, instead of having an aerodynamic equivalent size, diffusive
particles of different shapes can be related to the diffusivity of a thermodynamic equivalent size
based on spherical particles.
The particle size range around 0.2 to 1.0 jim is frequently described as consisting of
particles that are small enough to be minimally influenced by impaction or sedimentation and
large enough to be minimally influenced by diffusion. Such particles are the most persistent in
inhaled air and undergo the lowest degree of deposition in the respiratory tract.
Interception is deposition by physical contact with airway surfaces. The interception
potential of any particle depends on its physical size. Fibers are of chief concern in relation to
the interception process. Their aerodynamic size is determined predominantly by their diameter,
but their length is the factor that influences probability of interception deposition.
Electrostatic precipitation is deposition related to particle charge. The minimum charge an
aerosol particle can have is zero. This condition rarely is achieved because of the random
charging of aerosol particles by ions in the air. Particles acquire charges by collisions with air
ions because of their random thermal motion. Many laboratory-generated aerosols are highly
charged, but methods such as passage of the particle-containing airstream through a Kr-85
charge neutralizer can neutralize charge. In addition, these charged aerosols will generally lose
their initial charge as they attract oppositely charged ions, and an equilibrium state is eventually
achieved. This Boltzmann equilibrium represents the charge distribution of an aerosol in charge
equilibrium with bipolar ions. The minimum amount of charge is very small: there is a
statistical probability that some particles within the aerosol will have no charge and that others
will have one or more positive and negative charges.
The electrical charge on some particles will result in an enhanced deposition over what
would be expected from size alone. This increase in deposition is thought to result from image
charges induced on the surface of the airway by charged particles and possibly space-charge
6-6
-------
effects whereby repulsion of particles with like charges results in increased migration toward the
airway wall. The effect of charge on deposition is inversely proportional to particle size and
airflow rate. This type of deposition is often small compared to the effects of turbulence and
other deposition mechanisms, and it generally has been considered to be a minor contributor to
overall particle deposition. However, a study by Cohen et al. (1998), employing hollow airway
casts of the human tracheobronchial tree to assess deposition of ultrafine (0.02 jim) and fine
(0.125 |im) particles, found the deposition of singly charged particles to be 5 to 6 times that of
particles having no charge and 2 to 3 times that of particles at Boltzmann equilibrium. This
suggests that electrostatic precipitation may, in certain situations such as workplace exposures or
indoor tobacco smoke, be a significant deposition mechanism for ultrafine and some fine
particles within the TB region. However, the influence of charge on the deposition of urban
aerosols should be minimal.
6.2.2 Deposition Patterns in the Human Respiratory Tract
Knowledge of sites where particles of different sizes deposit in the respiratory tract and the
amount of deposition therein is necessary for understanding and interpreting the health effects
associated with exposure to particles. Particles deposited in the various respiratory tract regions
are subjected to large differences in clearance mechanisms and pathways and, consequently,
retention times. This section summarizes concepts of particle deposition in humans and
laboratory animals as reported in the 1996 PM AQCD (U.S. Environmental Protection Agency,
1996a) and provides additional information based on studies published since that earlier
document.
Ambient air often contains particles too massive to be inhaled. The term "inhalability" is
used to denote the overall spectrum of particle sizes that are potentially capable of entering the
respiratory tract. Inhalability is defined as the ratio of the number concentration of particles of a
certain aerodynamic diameter that are inspired through the nose or mouth to the number
concentration of the same diameter particle present in ambient air (International Commission on
Radiological Protection, 1994). In general, for humans, unit density particles > 100 jim diameter
have a low probability of entering the mouth or nose in still air, but there is no sharp cutoff to
zero probability. Additionally, there is no lower limit to inhalability, so long as the particle
6-7
-------
exceeds a critical size where the aggregation of atomic or molecular units is stable enough to
endow it with "particulate" properties in contrast to those of free ions or gas molecules.
6.2.2.1 Total Respiratory Tract Deposition
Total human respiratory tract deposition, as a function of particle size, is depicted in
Figure 6-2. These data were obtained by various investigators using different sizes of spherical
test particles in healthy male adults under different ventilation conditions; the large standard
deviations reflect inter-individual variability in airway dimensions and airway branching and
breathing-pattern related variability of deposition efficiencies. Deposition in the ET region with
nose breathing is generally higher than that with mouth breathing because the superior filtration
capabilities of the nasal passages results in somewhat higher total deposition with nasal
breathing for particles > 1 |im AED. For particles greater than 1 jim AED, deposition is
governed by impaction and sedimentation, and it increases with increasing AED. For AED
> 10 jim, almost all inhaled particles are deposited. As the particle size decreases from «0.5 jim,
diffusional deposition becomes dominant and total deposition depends more on the actual
physical diameter of the particle. Decreasing particle diameter below 0.1 jim leads to an
increase in total deposition. Total deposition shows a minimum for particle diameters in the
range of 0.2 to 1.0 |im where, as noted above, neither sedimentation, impaction, or diffusion
deposition are very effective. Deposition never reaches zero because of mixing between
particle-rich tidal air and nearly particle-free residual lung air. The particles in the tidal air
remaining in the deep lung are gradually deposited.
Besides particle size, breathing pattern (tidal volume, breathing frequency, route of
breathing) is the most important factor affecting lung deposition. Kim (2000) reported total lung
deposition values in healthy adults for a wide range of breathing patterns: tidal volumes (375 to
1500 mL), flow rates (150 to 1000 mL/s), and respiratory times (2 to 12 s). Total lung
deposition increased with increasing tidal volume at a given flow rate and with increasing flow
rate at a given respiratory time. Various deposition values were correlated with a single
composite parameter consisting of particle size, flow rate, and tidal volume.
Ultrafine particles (dp < 0.1 jim) are being specifically evaluated for determination of their
potential toxicity. There is, however, little information on total respiratory tract deposition of
such particles. Frampton et al. (2000) exposed healthy adult human males and females,
6-8
-------
i uu -
90-
80-
g 70-
.1 60~
\^
0 50-
Q.
0
Q 40-
m
,2 30-
20-
10-
n
A T
^ A
L
A Human
A Human
i
A
tf
I
f
1
L
A
J
i
^
k
X T-
I
A
1
1
A^
*
X -i-
(oral inhalation)
(nasal inhalation)
i
i
i
4
y,
i
A
: T
4ffi
HIT
A
I .
A\
1 A\ &
U
k
-A-
P-
r lfAi
Lv.
T A
Si A
TM
k A
1
\
1
A1!^^
0.01 0.1 1.0
Particle Diameter (urn)
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 Ł 0.5 urn and geometric (or diffusion equivalent) for those
< 0.5 jim.
Source: Modified from Schlesinger (1989).
via mouthpiece, to 0.0267 jim diameter carbon particles (at 10 |ig/m3) for 2 h at rest. The
inspired and expired particle number concentration and size distributions were evaluated. Total
respiratory tract deposition fraction was determined for six particle size fractions ranging from
0.0075 to 0.1334 jim. They found an overall total lung deposition fraction of 0.66 (by particle
number) or 0.58 (by particle mass), indicating that exhaled mean particle diameter was slightly
larger than inhaled diameter. There was no gender difference. The deposition fraction
decreased with increasing particle size within the ultrafine range, from 0.76 at the smallest size
to 0.47 at the largest.
6-9
-------
Jaques and Kim (2000) measured total deposition fraction (TDF) of ultrafine particles
(number median diameter CMD = 0.04 to 0.1 jim and og = 1.3) in 22 healthy adults (men and
women in equal number) under a variety of breathing conditions. The study was designed to
obtain a rigorous data set for ultrafine particles that could be applied to health risk assessment.
TDF was measured for six different breathing patterns: tidal volume (Vt) of 500 mL at
respiratory flow rates (Q) of 150 and 250 mL/s; V, = 750 mL at Q of 250 and 375 mL/s; V, = 1 L
at Q of 250 and 500 mL/s. Aerosols were monitored continuously by a modified condensation
nuclei counter during mouthpiece inhalation with the prescribed breathing patterns. For a given
breathing pattern, TDF increased as particle size decreased, regardless of the breathing pattern
used. For example, at V, = 500 mL and Q = 250 mL/s, TDF was 0.26, 0.30, 0.35, and 0.44 for
CMD = 0.10, 0.08, 0.06, and 0.04 |im, respectively (see Figure 6-3). For a given particle size,
TDF increased with an increase in V, and a decrease in Q, indicating the importance of breathing
pattern in assessing respiratory dose. The study also found that TDF was somewhat greater for
women than men at CMD = 0.04 jim and 0.06 jim with all breathing patterns used, but the
difference was smaller or negligible for larger-sized ultrafine particles. The results clearly
demonstrate that the TDF of ultrafine particles increases with a decrease of particle size and with
breathing patterns of longer respiratory time, a pattern that is consistent with deposition by
diffusion. These data are the only systematic human experimental data for ultrafine particles
reported since the 1996 PM AQCD.
A property of some ambient particulate species that affects deposition is hygroscopicity,
the propensity of a material for taking up and retaining moisture under certain conditions of
humidity and temperature. Ambient fine particles (sulfate, nitrate, and possibly organics) tend to
be hygroscopic (see Chapter 2). Such particles can increase in size in the humid air within the
respiratory tract and, when inhaled, will deposit according to their hydrated size rather than their
initial size. The implications of hygroscopic growth on deposition have been reviewed
extensively by Morrow (1986) and Hiller (1991) and the difficulties of studying lung deposition
of hygroscopic aerosols have been reviewed by Kim (2000). In general, compared to
nonhygroscopic particles of the same initial size, the deposition of hygroscopic aerosols in
different regions of the lung may be higher or lower, depending on the initial size. For particles
with initial sizes larger than «0.5 |im, the influence of hygroscopicity would be to increase total
deposition with a shift from peripheral to central or ET region; whereas for smaller ones total
6-10
-------
1 .U
OQ
.y
OP
c
o
o
ro
LL n p
c
.0
~ 0.5
>
O
Q.
O
IS n-a.
f
OH
.1
n n
T
JL
T
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
^
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
^
[1
: =
• =
. =
' ~
; =
1 =
1
Vt Q T f
I | 10UUml_, 25Uml_/s, bs /.bbpm
P^^ 1000mL, 500ml_/s, 4s, 15 bpm
kx^H 750ml, 250ml_/s, 6s, 10 bpm
^OT 750ml, 375ml_/s, 4s, 15 bpm
I I 500ml, 150ml_/s, 6.7s, 9 bpm
fTTTTTTTTI 500ml, 250 ml_/s, 4s, 15 bpm
-TT - T
1
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
,_L-
X
X
Xrr
x
X
X
X
\
X
X
X
X
X
X
X
X
X
X
\
[I
• =
. —
• =
• =
• =
JL
T
I
/
XfT
x
X
X
X
X
X
X
X
X
X
X
X
T T , _
; =
' ~
• =
• =
L\\\\\\\\\\\\N-H
x -
IX.
xE
X
X
X
X
X
X
X
X
X
X
X
1
L =
' ~
; =
\ •=.
0.02
0.04
0.06 0.08
Particle Diameter (|jm)
0.10
0.12
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).
-------
deposition would tend to be decreased. See Chapter 2 for a detailed description of particle
hygroscopicity.
6.2.2.2 Deposition in the Extrathoracic Region
The fraction of inhaled particles depositing in the ET region is quite variable and
dependent on particle size, flow rate, breathing frequency, whether breathing is through the nose
or the mouth (Figure 6-4), and the cross-sectional area of the flow path. Mouth breathing
bypasses much of the filtration capabilities of the nasal airways and leads to increased deposition
in the lungs (TB and A regions). The ET region is clearly the site of first contact with particles
in the inhaled air and essentially acts as a "prefilter" for the lungs.
Since release of the 1996 PM AQCD, a number of studies have explored ET deposition
with in vivo studies, as well as in both physical and mathematical model systems. In one study,
the relative distribution of particle deposition between the oral and nasal passages was assessed
during "inhalation" by use of a physical model (silicone rubber) of the human upper respiratory
tract, which extended from the nostrils and mouth through the main bronchi (Lennon et al.,
1998). Monodisperse particles ranging in size from 0.3 to 2.5 jim were evaluated at flow rates
ranging from 15 to 50 L/min. Regional deposition in the oral passages, the lower
oropharynx-trachea, nasal passages, and nasopharynx-trachea, as well as total deposition in the
model, were assessed. Deposition within the nasal passages was found to agree with available
data obtained from a human inhalation study (Heyder and Rudolf, 1977), being proportional to
particle size, density, and inspiratory flow rate. It also was found that for oral inhalation, the
relative distribution of particle deposition between the oral cavity and the oropharynx-trachea
was similar; whereas for nasal inhalation, the nasal passages contained most of the particles
deposited in the model, with only about 10% deposited in the nasopharynx-trachea region.
Furthermore, the deposition efficiency of the nasopharynx-trachea region was greater than that
of the oropharynx-trachea region. For simulated oronasal breathing, deposition in the ET region
depended primarily on particle size rather than flow rate. For all flows and for all breathing
modes, total deposition in the ET region increased as particle diameter increased. Such
information on deposition patterns in the ET region is useful in refining empirical deposition
models.
6-12
-------
i uu -
90 -
C? 80 -
0*"
0 70 -
1 60-
Q.
d>
O 50 -
o
"o
ro 40 -
o
^™
Ijj 30 -
2
UJ 20-
10 -
0 -
0.
31
A Human (oral inhalation)
A Human (nasal inhalation)
* \
III I I I I I I | I I I
0.1
I
A.
A
A
T 1*
1 i ^ ^
1
A
k
A
-L
A
A
t
k
k
A
k
^
L
T
• \
14
1 1 1 1 II
1.0
k
A
^
1
1
1
1 1 1 1
K
Particle Diameter (|jm)
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 Ł 0.5 jim and geometric (or diffusion equivalent) for those < 0.5 um.
Source: Modified from Schlesinger (1989).
Deposition within the nasal passages was further evaluated by Kesavanathan and Swift
(1998), who examined the deposition of 1- to 10-jim particles in the nasal passages of normal
adults under an inhalation regime in which the particles were drawn through the nose and out
through the mouth at flow rates ranging from 15 to 35 L/min. At any particle size, deposition
increased with increasing flow rate; whereas deposition increased with increasing particle size at
any flow rate. In addition, as shown experimentally by Lennon et al. (1998) under oronasal
breathing conditions, deposition of 0.3- to 2.5-|im particles within the nasal passages was
significantly greater than within the oral passages, and nasal inhalation resulted in greater total
deposition in the model than did oral inhalation. These results are consistent with other studies
6-13
-------
discussed in the 1996 PM AQCD and with the known dominance of deposition by impact!on
within the ET region.
Rasmussen et al. (2000) measured deposition of 0.7 jim particles consisting of sodium
chloride and radioactively-labeled technetium-diethylenetriamine-pentaacetic acid (DTPA) in
the nasal cavity of normal adult humans. Each subject inhaled one liter for each inspiration at
flow rates ranging from 10 to 30 L/min. They found that the deposition fraction in the nasal
passages increased as flow rate increased and that an estimate of maximum linear air velocity
was the best single predictor of nasal deposition fraction.
For ultrafine particles (dp < 0.1 |im), deposition in the ET region is controlled by diffusion,
which depends only on the particle's geometric diameter. Prior to 1996, ET deposition for this
particle size range had not been studied extensively in humans, and this remains the case. In the
1996 PM AQCD, the only data available for ET deposition of ultrafine particles were from
hollow airway cast studies. More recently, deposition in the ET region has been examined using
mathematical modeling. Three-dimensional numerical simulations of flow and particle diffusion
in the human upper respiratory tract, which included the nasal region, oral region, larynx, and
first two generations of bronchi, were performed by Yu et al. (1998). Deposition of 0.001- and
0.01-|im particles in these different regions was calculated under inspiratory and expiratory flow
conditions. Deposition efficiencies in the total model were lower on expiration than inspiration,
although values for the former were quite high. During inspiration, about 75% of the 0.001-|im
particles were deposited compared to only 31% of the 0.01-jim particles. Deposition in the nose
accounted for 74 to 81% of total deposition in the model system during nasal inspiration. With
oral inhalation, deposition in the mouth was 60 to 67% of total deposition in the model (Yu
etal., 1998).
Swift and Strong (1996) examined the deposition of ultrafine particles, ranging in size
from 0.00053 to 0.00062 jim (0.53 to 0.62 nm), in the nasal passages of normal adults during
constant inspiratory flows of 6 to 22 L/min. In this case, deposition ranged from 94 to 99% (of
amount inhaled). These results are consistent with results noted in studies above, namely that
the nasal passages are highly efficient collectors for ultrafine particles. Only a weak dependence
of deposition on flow rate was found, which contrasts with results noted above (i.e., Lennon
et al., 1998) for particles > 0.3 jim, but is consistent with diffusion being the main deposition
mechanism. This report has important implications for assessing the toxicity of PM because the
6-14
-------
filtration efficiency of the nasal passages will lessen the deposition probability of ultrafme
particles, particularly smaller-size ultrafines, in the lungs.
Cheng et al. (1997) examined oral airway deposition in a replicate cast of the human nasal
cavity, oral cavity, and laryngeal-tracheal sections. For particle sizes of 0.005 to 0.150 jim,
using constant inspiratory and expiratory flow rates of 7.5 to 30 L/min, they noted that the
deposition fractions within the oral cavity were essentially the same as that in the
laryngeal-tracheal sections for all particle sizes and flow rates. They ascribed this to the balance
between flow turbulence and residence time in these two regions. Svartengren et al. (1995)
examined the effect of changes in external resistance on oropharyngeal deposition of 3.6-|im
particles in asthmatics. Under controlled mouthpiece breathing conditions (flow rate = 0.5 L/s),
the median deposition as a percentage of inhaled particles in the mouth and throat was 20%
(mean = 33%; range 12 to 84%). Although the mean deposition fell to 22% with added
resistance, the median value remained at 20% (range = 13 to 47%). Fiberoptic examination of
the larynx revealed a trend for increased mouth and throat deposition which was associated with
laryngeal narrowing. On the basis of mathematical model calculations, Katz et al. (1999) found
that turbulence plays a key role in enhancing particle deposition in the larynx and trachea.
The results of all of the above studies support the previously known ability of the ET
region, especially the nasal passages, to act as an efficient filter for small ultrafme particles
(< 0.01 |im) as well as for larger ones (> 2 jim), potentially reducing the amount of particles
within a wide size range that are available for deposition in the TB and A regions (for nasal
breathing, head deposition would be about 20% for 0.01 jim particles).
6.2.2.3 Deposition in the Tracheobronchial and Alveolar Regions
Particles that do not deposit in the ET region of the respiratory tract enter the lungs;
however, their regional deposition within the lungs cannot be precisely measured. Much of the
available deposition data for the TB and A regions have been obtained from experiments with
radioactively labeled, poorly soluble particles (Figures 6-5 and 6-6, respectively). These have
been described previously (U.S. Environmental Protection Agency, 1996a).
Since the publication of that document, a novel serial bolus delivery method has been
introduced. Using this bolus technique, regional deposition has been estimated for fine and
coarse aerosols (Kim et al., 1996; Kim and Hu, 1998) and for ultrafme aerosols (Kim and Jaques,
6-15
-------
60
50 -
O 40
Q.
0)
Q
re 30 -
IE
o
2 20
O
0)
O 10 -
A Oral Inhalation
1.0
Particle Diameter (|jm)
0.1
1.0
Figure 6-5.
Tracheobronchial deposition
(as percentage 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).
Source: Modified from Schlesinger (1989).
10
/u —
60 -
I50'
O
+J
M 40 -
O
Q.
O
Q 30 -
§ 20-
^
10 -
0 -
A Oral Inhalation T
A
A Nasal Inhalation
I
A L
T T
I I T *f A A
T ^r
1 I
A A
[I
^
1
|
A
A
Figure 6-6.
Alveolar deposition (as percentage 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 um and
geometric (or diffusion equivalent)
for those < 0.5 urn.
Source: Modified from Schlesinger (1989).
10
Particle Diameter (|jm)
6-16
-------
2000). The serial bolus method uses nonradioactive aerosols and can estimate regional
deposition in a virtually unlimited number of lung compartments. Because of experimental
limitations of the technique, the investigators estimated regional lung deposition in ten serial,
50-mL increments from the mouth to the end of a typical 500-mL tidal volume. Deposition
estimates in the TB and A regions were obtained for both men and women for particles ranging
from 0.04 to 5.0 jim in diameter. It should be noted that particle deposition in the TB and
A regions was based on volumetric compartments of 50 to 150 mL and > 150 mL, respectively.
Deposition in the ET region was based on the 0 to 50 mL compartment. However, their
estimates of TB deposition may artificially increase with decreasing anatomical dead space, i.e.,
the TB region of 50 to 150 mL includes more anatomically distal airways in the smaller lungs of
women relative to men. Lung deposition fractions are shown in Figure 6-7. In men, 24 to 32%
of total particle deposition (0.04-, 0.06-, 0.08-, and 0.10-|im particles) was deposited in the TB
region and 67 to 76% was deposited in the A region. In women, compared to men, the
deposition of these particles was consistently greater in the TB region (21-48%; p < 0.05 for
0.04 and 0.06 jim), but was comparable or slightly smaller in the A region. As a result, total
lung deposition of ultrafine particles was slightly greater (-5-14%) in women than men,
particularly for 0.04 and 0.06 |im (p < 0.05). For 1-, 3-, and 5-|im particles in men, 16 to 37% of
total particle deposition was in the TB region and 57 to 83% was in the A region. Deposition of
these size particles, in women was consistently greater in the TB region, particularly for 3 and
5 |im (56 to 68%, p < 0.05), but was comparable or slightly smaller in the A region as compared
to men. As a result, total lung deposition was slightly greater (-20%) in women than men for
3 and 5 jim (p < 0.5). Thus, deposition of ultrafine and coarse particles in the TB region was
somewhat greater for women than men, but there were no differences in either total or regional
deposition of 0.08, 0.10, or 1.0 jim particles in men vs. women. The potential biological
significance of the relatively small absolute magnitude (versus percentage) differences, if any,
remains to be elucidated.
Fine particles that penetrate to the gas exchange airways are deposited on airway
bifurcations at higher concentrations. The deposition diminishes rapidly with airway generation,
consistent with the concentration of streamlines near the bifurcations and the penetration depth
of convective tidal flow.
6-17
-------
c
o
"•5
o
(0
c
o
"•5
'w
o
Q.
0.04
0.06
0.08
0.10
Particle Diameter (|jm)
Male Vt = 500 ml_
Female
Q = 250 rnL/s
* = p < 0.05
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 68% and 50% greater for 3 and 5 um respectively (both p < 0.05),
whereas A deposition was comparable. For ultrafine particles of 0.04 to
0.1 jim 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 in women
was 27% and 48% greater for 0.04 and 0.06 um respectively (both p < 0.05),
whereas A deposition was comparable. There was no difference for 0.08, 0.1,
and 1 jim between men and women. 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).
6-18
-------
6.2.2.4 Local Distribution of Deposition
Airway structure and its associated air flow patterns are exceedingly complex, and
ventilation distribution of air in different parts of the lung is uneven. Thus, it is expected that
particle deposition patterns within the ET, TB, and A regions would be highly nonuniform, with
some sites exhibiting deposition that is much greater than average levels within these regions.
This was discussed in detail in the 1996 PM AQCD. Basically, using deposition data from living
subjects as well as from mathematical and physical models, enhanced deposition has been shown
to occur in the nasal passages and trachea and at branching points in the TB and A regions (see
Chapter 10 of U.S. Environmental Protection Agency, 1996a). Churg and Vedal (1996)
examined retention of particles on carinal ridges and tubular sections of airways from lungs
obtained at necropsy. Results indicated significant enhancement of particle retention on carinal
ridges through the segmental bronchi; the ratios were similar in all airway generations examined.
Kim and Fisher (1999) studied local deposition efficiencies and deposition patterns of
aerosol particles (2.9 to 6.7 jim) in sequential double-bifurcation tube-models with two different
branching geometries: one with in-plane (Model-A) and another with out-of-plane
(Model-B) bifurcation. The deposition efficiencies (DE) in each bifurcation increased with
increasing Stokes number (ratio of the stopping distance of a particle to a characteristic
dimension of an obstacle). The Stokes number is used to characterize the ability of a particle to
follow a streamline in curvilinear motion. As the Stokes number increases, particles tend to
become less able to follow a streamline around an obstacle and more likely to impact the
obstacle (Hinds, 1999). With symmetric flow conditions, DE was somewhat smaller in the
second than the first bifurcation in both models. DE was greater in the second bifurcation in
Model-B than in Model-A. With asymmetric flows, DE was greater in the low-flow side
compared to the high-flow side; this was consistent in both models. Deposition pattern analysis
showed highly localized deposition on and in the immediate vicinity of each bifurcation ridge,
regardless of branching and flow patterns.
Comer et al. (2000) used a three-dimensional computer simulation technique to investigate
local deposition patterns in sequentially bifurcating airway models that were previously used in
experiments by Kim and Fisher (1999). The simulation was for 3-, 5-, and 7-|im particles and
assumed steady, laminar, constant air flow with symmetry about the first bifurcation. The
overall trend of the particle deposition efficiency, i.e., an exponential increase with Stokes
6-19
-------
number, was similar for all bifurcations; and deposition efficiencies in the bifurcation regions
agreed very well with experimental data. Local deposition patterns consistently showed that the
majority of the deposition occurred within the carinal region.
Deposition "hot spots" at airway bifurcations have undergone additional analyses using
mathematical modeling techniques. Using calculated deposition sites, a strong correlation has
been demonstrated between secondary flow patterns and deposition sites and density both for
large (10 jim) particles and for ultrafine (0.01 jim) particles (Heistracher and Hofmann, 1997;
Hofmann et al., 1996). This supports experimental work, noted in U.S. Environmental
Protection Agency (1996a), indicating that, like larger particles, ultrafine particles show
enhanced deposition at airway branch points — even in the upper tracheobronchial tree.
The pattern of particle distribution on a more regional scale was evaluated by Kim et al.
(1996) and Kim and Hu (1998). Deposition patterns were measured in situ in nonsmoking
healthy young adult males using an aerosol bolus technique that delivered 1-, 3-, or 5-|im
particles into specific volumetric depths within the lungs. The distribution of particle deposition
shifted from distal to proximal regions of the lungs with increasing particle size (Figure 6-8).
Furthermore, the surface dose was found to be greater in the conducting airways than in the
alveolar region for all of the particle sizes evaluated. Within the conducting airways, the largest
airway regions (i.e., 50 to 100 mL volume distal to the larynx) received the greatest surface
doses.
Bennett et al. (1998) studied the effect of variable anatomic dead space (ADS) on particle
deposition using an aerosol bolus technique in healthy subjects inhaling radiolabeled (99mTc) iron
oxide particles (3.5 jim MMAD). The subjects inhaled 40 mL aerosol boluses to a volumetric
front depth of 70 mL into the lung at a lung volume of 70% total lung capacity end-inhalation
and estimated the fraction of the inhaled boluses deposited in intrathoracic airways. ADS was
also measured from 70% total lung capacity. The intrathoracic deposition fraction (IDF) varied
from 0.04 to 0.43 and increased with decreasing ADS. The IDF was lower in subjects with large
ADS (> 250 mL). Hence, women had twice the IDF due to their smaller ADS and smaller
airspace dimensions. They observed significantly greater deposition in the left (L) versus
right (R) lung; mean L/R (ratio of deposition in L lung to R lung, normalized to ratio of L-to-R
lung volume) was 1.58 ± 0.42. Retention of deposited particles at 2 h was independent of ADS
or IDF. There was significant retention of particles in the whole lung at 24 h post deposition and
6-20
-------
0.14
o
'+J
o
c
0
O
0.
0)
O
"re
o
o
0.06-
0.04-
0.02-
dp =
-O- 150mL/s
250mL/s
500mL/s
dp = 3pm
-O- 150mL/s
-n- 250mL/s
-A- 500mL/s
dp = 5\im
-O- 150mL/s
250mL/s
500mL/s
0 100 200 300 400 500
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 um (Panel A) to
fine (dp = 1.0 um MMAD; Panel B) and coarse (dp = 3 and 5 um MMAD;
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 in in.).
Source: Kim et al. (1996); Kim and Jaques (2000).
6-21
-------
slow clearance of these particles continued through 48 h post deposition. There was significant
retention of insoluble particles in large bronchial airways at 24 h post deposition (i.e., 24 h
central-to-peripheral ratio of 1.40 and 1.82 in the R and L lung, respectively).
Kim and Jaques (2000) used the respiratory bolus technique to estimate the deposition
distribution of ultrafme particles (0.04, 0.06, 0.08, and 0.1 jim) in young adults. Under normal
breathing conditions (Vt, tidal volume 500 mL; Q, 250 mL/s), bolus aerosols were delivered
sequentially to a lung depth ranging from 50 to 500 mL in 50-mL increments. The results
indicate that regional deposition of ultrafme particles (0.04 to 0.10 jim) varies widely along the
depth of the lung (Figure 6-8). Regional deposition of these particles is approximately bounded
by those of larger particles (1.0 to 5.0 jim). The variability with depth is small for 1 jim particles
but large for 5 jim particles. (Note the difference in y-axis values.) The deposition patterns for
ultrafme particles, especially for very small ultrafme particles, were similar to those for coarse
particles. Peak deposition occurred in the lung regions situated between 150 and 200 mL from
the mouth, and sites of peak deposition shifted proximally with decreasing particle size.
Deposition dose per unit average surface area was greatest in the proximal lung regions and
decreased rapidly with increased lung depth. Peak surface dose was 5 to 7 times greater than
average lung dose. These results indicate that local enhancement of dose occurs in healthy
lungs, which could be an important factor in eliciting pathophysiological effects.
6.2.2.5 Deposition of Specific Size Modes of Ambient Aerosol
Several recent modeling studies provide estimates of the deposition profiles for "real
world" particle size fractions. One such study using a lung-anatomical model (Venkataraman
and Kao, 1999) examined the contribution of two specific size modes of the PM10 ambient
aerosol, namely the fine mode (defined as particles with diameters up to 2.5 jim) and the thoracic
fraction of the coarse mode (defined as particles with diameters between 2.5 and 10 jim) to total
lung and regional lung doses (i.e., a daily dose expressed as jig/day, and a surface dose
expressed as |ig/cm2/day) resulting from a 24-h exposure to a particle concentration of
150 |ig/m3. The study also evaluated deposition in terms of two metrics, namely mass dose and
number dose. Deposition was calculated using a mathematical model for a healthy human lung
under both simulated moderate exertion (V, = 1 L at 15 breaths/min) and vigorous exertion
(V, = 1.5 L at 15 breaths/min) and for a compromised lung (V, 0.5 L at 30 breaths/min).
6-22
-------
Regional deposition values were obtained for the ET, TB, and A regions. Because the exposure
scenario was is quite unrealistic, only general trends should be inferred from this study rather
than actual deposition values. These estimates would also be highly uncertain for the
compromised lung.
The daily modeled mass dose for the fine particles peaked in the A region for all breathing
patterns, whereas the mass dose for coarse particles was comparable in the TB and A regions.
The mass per unit surface area of various airways from the fine and coarse fractions was larger
in the trachea and first few generations of bronchi. Venkataraman and Kao (1999) suggested
that these large surface doses may be related to aggravation of upper respiratory tract illness.
The modeled daily number dose was different for fine and coarse fractions in all lung airways:
the dose from the fine fraction was higher by about 100 times in the ET and about 10s times in
internal lung airways. The surface number dose (particles/cm2/day) was 103 to 10s times higher
for fine than for coarse particles in all lung airways, indicating the larger number of fine particles
depositing. Particle number doses did not follow trends in mass doses and were much higher for
fine than coarse particles. It also was concluded that the fine fraction contributes 10,000 times
greater particle number per alveolar macrophage than the coarse fraction particles. As noted,
these results must be viewed with caution because they were obtained using a pure mathematical
model that remains to be validated in terms of realistic physiologic conditions.
Another evaluation of deposition that included consideration of size mode of the ambient
aerosol was that of Broday and Georgopoulos (2001). In this study, a mathematical model was
used to account for particle hygroscopic growth, transport, and deposition. It was concluded that
different rates of particle growth in the inspired air resulted in a change in the aerosol size
distribution such that the initially inspired ultrafine particles (< 0.1 jim) grew into the size range
between 0.1 to 1 |im. Due to their growth, particles deposited to a lesser extent than expected for
ultrafine particles due to a decrease in diffusive deposition. On the other hand, particles that
were originally in the 0.1- to l-|im size range when inhaled will undergo enhanced deposition
because as they increase in size due to hygroscopic growth. Hence, the initial size distribution of
the inhaled polydisperse aerosol affects the evolution of size distribution once inhaled and, thus,
its deposition profile in the respiratory tract. Hygroscopicity of respirable particles must be
considered for accurate predictions of deposition. Because different size fractions likely have
6-23
-------
different chemical composition, such changes in deposition patterns could affect biological
responses.
6.2.3 Biological Factors Modulating Deposition
Experimental deposition data in humans have been commonly derived using healthy adult
Caucasian males. Various factors can act to alter deposition patterns from those obtained in this
group. Evaluation of these factors is important to help understand potentially susceptible
subpopulations because differences in biological response following pollutant exposure may be
caused by dosimetry differences as well as by differences in innate sensitivity. The effects of
different biological factors on deposition were discussed in the 1996 PM AQCD (U.S.
Environmental Protection Agency, 1996a) and are summarized below together with additional
information obtained from more recent studies.
6.2.3.1 Gender
Males and females differ in body size, conductive airway size, and ventilatory parameter
distributions; therefore, gender differences in deposition should be expected. In some of the
controlled studies, however, the men and women were constrained to breathe at the same tidal
volume and frequency. Since women are generally smaller than the men, the increased minute
ventilation compared to their normal ventilation could affect deposition patterns. This may help
to explain some of the differing results discussed below.
Using particles in the 2.5- to 7.5-|im size range, Pritchard et al. (1986) indicated that, for
comparable particle sizes and inspiratory flow rates, females had higher ET and TB deposition
and smaller A deposition than did males. The ratio of A deposition to total thoracic deposition
in females also was found to be smaller. These differences were attributed to gender differences
in airway size.
In another study (Bennett et al., 1996), the total respiratory tract deposition of 2-|im
particles was examined in adult males and females aged 18 to 80 years who breathed with a
normal resting pattern. Deposition was assessed in terms of a deposition fraction, the difference
between the amount of particles inhaled and exhaled during oral breathing. Although there was
a tendency for a greater deposition fraction in females compared to males, because males had
6-24
-------
greater minute ventilation, the deposition rate (i.e., deposition per unit time) was greater in males
than in females.
Kim and Hu (1998) assessed the regional deposition patterns of 1-, 3-, and 5-|im MMAD
particles in healthy adult males and females using an aerosol bolus technique and controlled
breathing. The total fractional deposition in the lungs was similar for both genders with the
l-|im particle size, but was greater in women for the 3- and 5-jim particles regardless of the
inhalation flow rate used; this difference ranged from 9 to 31% (p < 0.05), with higher values
associated with higher flow rates. The pattern of deposition was similar for both genders,
although females showed enhanced deposition peaks for all three particle sizes. The volumetric
depth location of these peaks was found to shift from peripheral (i.e., increased volumetric
depth) to proximal (i.e., shallow volumetric depth) regions of the lung with increasing particle
size, with the shift being greater in females than in males. Thus, deposition appeared to be more
localized in the lungs of females compared to those of males. These differences were attributed
to the smaller size of the upper airways, particularly of the laryngeal structure, in females. Local
deposition of l-|im particles was somewhat flow dependent but, for larger (5-jim) particles, was
largely independent of flow (flows did not include those that would be typical of exercise).
In a related study, Kim (2000) evaluated differences in deposition between males and
females under varying breathing patterns (simulating breathing conditions of sleep, resting, and
mild exercise). Using particles at the same size noted above and a number of breathing
conditions, total fractional lung deposition was comparable between men and women for l-jim
particles, but was slightly greater in women than men for 3- and 5-|im particles with all
breathing patterns. The gender difference was about 15% at rest and variable during exercise
depending on particle size. However, total lung deposition rate (i.e., deposition per unit time)
was found to be 3 to 4 times greater during moderate exercise than at rest for all particle sizes.
Thus, it was concluded that exercise may increase the health risk from particles because of
increased large airway deposition and that women may be more susceptible to these exercise-
induced changes.
Jaques and Kim (2000) and Kim and Jaques (2000) expanded the evaluation of deposition
in males and females to particles < 1 jim. They measured total fractional lung deposition in
healthy adults using sizes in the ultrafine mode (0.04 to 0.1 jim). Total fractional lung
deposition was greater in females than in males for 0.04- and 0.06-|im particles (p < 0.05). The
6-25
-------
difference was negligible for 0.08- and O.l-jim particles. Therefore, the gender effect was
particle-size dependent, showing a greater fractional deposition in females for very small
ultrafme and large coarse particles, but not for particles ranging from 0.08 to 0.1 jim. A local
deposition fraction was determined in each volumetric compartment of the lung to which
particles were injected based on the inhalation procedure (Kim and Jaques, 2000). The
fractional deposition was found to increase with increasing lung depth from the mouth, reach a
peak value, and then decrease with further increase in lung volumetric depth. Figure 6-8 shows
these ultrafme data along with fine and coarse particle data from Kim et al. (1996). The height
of the peak and its depth varied with particle size and breathing pattern. Peak fractional
deposition for the 5-|im particles was more proximal than that for the l-|im particles, whereas
that for the ultrafme particles occurred between these two peaks. For the ultrafme particles, the
peak fractional deposition became more proximal as particle size decreased. Although this
pattern of deposition distribution was similar for both men and women, the region of peak
fractional deposition was shifted closer to the mouth and peak height was slightly greater for
women than for men for all exposure conditions.
6.2.3.2 Age
Airway structure and respiratory conditions vary with age, and these variations may alter
the deposition pattern of inhaled particles (Table 6-1). The limited experimental studies reported
in the 1996 PM AQCD (U. S. Environmental Protection Agency, 1996a) indicated results
ranging from no clear dependence of total deposition on age to slightly higher deposition in
children than adults. However, children have a different resting ventilation than do adults.
Experimental studies must adjust for different breathing patterns and the higher minute
ventilation per unit body weight in children when comparing deposition results to those obtained
in adults.
Inhaled Deposition Patterns
Bennett et al. (1997a) analyzed the regional deposition of poorly soluble 4.5-|im particles
inhaled via mouthpiece. The subjects were children and adults with mild cystic fibrosis (CF),
but who likely had normal upper airway anatomy such that intra- and ET deposition would be
similar to that in healthy people. The mean age of the children was 13.8 years and 29.1 years for
6-26
-------
TABLE 6-1. EFFECTS OF AGE ON PARTICLE DEPOSITION IN RESPIRATORY TRACT
to
Type study
Inhalation
Inhalation
Inhalation
Inhalation
Airway models
Nasal casts
Model
Model
Model
Model
Particles
(MMAD) Summary
2 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.
2 um Examined deposition of particles in subjects aged 18-80 yrs. Fractional deposition not found to
be age-related but more dependant 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
decreased as 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.
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)
CF = Cystic fibrosis.
-------
the adults. Extrathoracic deposition of the 4.5-|im particles, as a percentage of total respiratory
tract deposition, was higher by about 50% in children compared to adults. There was an age
dependence of ET deposition for the 4.5-jim particles in the children in that the percentage ET
deposition tended to be higher at a younger age (p = 0.08). The younger group (< 14 years;
p = 0.05) had almost twice the percentage ET deposition of the older group (> 14 years).
Additional analyses showed an inverse correlation of ET deposition with body height. These
results are consistent with the predicted increase in head deposition of particle greater than 2 jim
with decreasing age reported by Xu and Yu (1986).
Becquemin et al. (1991) compared nasal filtering efficiency in children and adults; two
groups of children (12 children, aged 5.5 to 11.5 years; 8 children, aged 12 to 15 years) were
studied along with 10 adults. The deposition of polystyrene beads (1, 2.05, and 2.8 |im MMAD)
was measured for both nose and mouth breathing. Ventilation was controlled to scale breathing
patterns appropriate for each age either at rest or during moderate exercise. Anterior nasal
resistance and standard lung function were measured for each subject. For the same inhalation
flow rate, children had much higher nasal resistances than adults. Individually, nasal deposition
increased with particle size, ventilation flow rate, and nasal resistance from rest to exercise. The
average for these particle sizes were better correlated with inspiratory airflow rate than with
resistances or pressure drops at rest and during moderate exercise. Although the nasal airways
of children are narrower, they are also shorter and the inhalation flow rate is reduced. The
authors conclude that while the nasal deposition percentages were lower in children than in
adults at rest, these differences were even greater during exercise. This would mean that the
thoracic airways of children are protected to a lesser degree than those of adults.
Bennett and Zeman (1998) measured the deposition of monodisperse 2-|im (MMAD)
particles in children (aged 7 to 14 years) and adolescents (aged 14 to 18 years) for comparison to
that in adults (19 to 35 years). Each subject inhaled the particles by following their previously
determined individual spontaneous resting breathing pattern. Deposition was assessed by
measuring the amount of particles inhaled and exhaled. There was no age-related difference in
deposition within the children group. There was also no significant difference in deposition
between the children and adolescents, between the children and adults, nor between the
adolescents and adults. However, the investigators noted that, because the children had smaller
6-28
-------
lungs and higher minute ventilation relative to lung size, they likely would receive greater doses
of particles per lung surface area compared to adults. Furthermore, breath-to-breath fractional
deposition in children varied with tidal volume, increasing with increasing tidal volume. The
rate of deposition normalized to lung surface area tended (p = 0.07) to be greater (35%) in
children when compared to the combined group of adolescents and adults. These additional
studies still do not provide unequivocal evidence for significant differences in deposition
between adults and children, even when considering differences in lung surface area. However,
it should be noted that differences in levels of activity between adults and children are likely to
play a fairly large role in age-related differences in deposition patterns of ambient particles.
Children generally have higher activity levels during the day and higher associated minute
ventilation per lung size, which can contribute to a greater size-specific dose of particles.
Activity levels in relationship to exposure are discussed more fully in Chapter 5.
Another subpopulation of potential concern related to susceptibility to inhaled particles is
the elderly. In the study of Bennett et al. (1996), the total respiratory tract deposition of 2-|im
particles was examined in people aged 18 to 80 years, the deposition fraction in the lungs of
people with normal lung function was found to be independent of age, however, depending
solely on breathing pattern and airway resistance.
Modeled Deposition Patterns
Differences in regional deposition between children and adults have been assessed to a
greater extent using mathematical models than experimentally. These models indicate that, if the
entire respiratory tract and a complete breathing cycle at normal rate are considered, then ET
deposition in children would be generally higher than in adults. However, TB and A regional
deposition in children may be either higher or lower than that in adults, depending on particle
size (Xu and Yu, 1986). There is enhanced TB deposition in children for particles < 5 jim (Xu
and Yu, 1986; Hofmann et al., 1989a).
An age-dependent theoretical model to predict regional particle deposition in children's
lungs that incorporates breathing parameters and morphology of the growing lung was developed
by Musante and Martonen (1999). The model was used to compare deposition of monodisperse
aerosols, ranging from 0.25 to 5 jim, in the lungs of children (ages 7, 22, 48, and 98 months) at
6-29
-------
rest to that in adults (age 30 years) at rest. Compared to adults, the fractional deposition was
highest in the 48- and 98-month subjects for all particle sizes. Total fractional lung deposition
(i.e., TB + A) was generally higher in children than adults, with children of all ages showing
similar total deposition fractions. The TB fractional deposition was reported to monotonically
decrease as a function of age for all sizes. This apparent linear relationship may arise due to the
way the investigators scaled the anatomical data for the adult down to the lung size for children.
Still, the key point is that children may have greater deposition of mass per unit area than adults,
even if this relationship may not be a linear function of age.
The model was later used by Musante and Martonen (2000a) to evaluate the deposition of a
residual oil fly ash (ROFA) having an MMAD of 1.95 |im, a og of 2.19, and a CMD of 0.53
(assuming a particle density of 0.34 g/cm2). Deposition was evaluated under resting breathing
conditions. The mass-based deposition fraction of the particles was found to decrease with age
from 7 months to adulthood, and the mass deposition per unit surface area in the lungs of
children could be significantly greater than in adults.
Phalen and Oldham (2001) calculated the respiratory deposition of particles with sizes
ranging from 0.1 to 10 nm in diameter for 20 year-old adults and 2 year-old children. Total
fractional lung deposition was comparable between adults and children for all particle sizes
tested; however, TB deposition fraction was much greater in children than in adults (from 13 to
81%, depending on particle size). Particle deposition fraction in the A region was significantly
reduced in children.
Cheng et al. (1995) examined the deposition of ultrafine particles in replica casts of the
nasal airways of children aged 1.5 to 4 years. Particle sizes ranged from 0.0046 to 0.2 jim, and
both inspiratory and expiratory flow rates were used (3 to 16 L/min). Deposition efficiency was
found to decrease with increasing age for a given particle size and flow rate.
Oldham et al. (1997) examined the deposition of monodisperse particles having diameters
of 1, 5, 10, and 15 jim in hollow airway models that were designed to represent the trachea and
the first few bronchial airway generations of an adult, a 7-year-old child, and a 4-year-old child.
They noted that, in most cases, the total deposition efficiency was greater in the child-size
models than in the adult model.
6-30
-------
6.2.3.3 Respiratory Tract Disease
The presence of respiratory tract disease can affect airway structure and ventilatory
parameters, thus altering deposition compared to that occurring in healthy individuals. The
effect of airway diseases on deposition has been studied extensively, as described in the 1996
PM AQCD (U.S. Environmental Protection Agency, 1996a). Studies described therein showed
that people with chronic obstructive pulmonary disease (COPD) had very heterogeneous
deposition patterns and differences in regional deposition compared to healthy individuals.
People with obstructive pulmonary diseases tended to have greater deposition in the TB region
than did healthy people. Furthermore, there tended to be an inverse relationship between
bronchoconstriction and the extent of deposition in the A region, whereas total respiratory tract
deposition generally increased with increasing degrees of airway obstruction. The described
studies were performed during controlled breathing, i.e., all subjects breathed with the same tidal
volume and respiratory rate. However, although resting tidal volume is similar or elevated in
people with COPD compared to healthy individuals, the former tend to breathe at a faster rate,
resulting in higher than normal tidal peak flow and resting minute ventilation. Thus, given that
breathing patterns differ between healthy and obstructed individuals, particle deposition data for
controlled breathing may not be appropriate for estimating respiratory doses from ambient PM
exposures. Although the extent to which lung deposition may change with respect to particle
size, breathing pattern, and disease status in people with COPD is still unclear, some recent
studies have attempted to provide additional insight into this issue.
Bennett et al. (1997b) measured the fractional deposition of insoluble 2-jim particles in
people with severe to moderate COPD (mix of emphysema and chronic bronchitis, mean age
62 years) and compared this to healthy older adults (mean age 67 years) under conditions where
the subjects breathed using their individual resting breathing pattern as well as a controlled
breathing pattern. People with COPD tended to have an elevated tidal volume and a faster
breathing rate than people with healthy lungs, resulting in about 50% higher resting minute
ventilation. Total respiratory tract deposition was assessed in terms of deposition fraction
(determined from measures of the amount of aerosol inhaled and exhaled) and deposition rate
(the amount of particulate deposited per unit time). Under typical breathing conditions, people
with COPD had about 50% greater deposition fraction than did age-matched healthy adults.
Because of the elevation in minute ventilation, people with COPD had average deposition rates
6-31
-------
about 2.5 times that of healthy adults. Similar to previously reviewed studies (U.S.
Environmental Protection Agency, 1996a), these investigators observed an increase in deposition
with an increase in airway resistance, suggesting that, at rest, COPD resulted in increased
deposition of fine particles in proportion to the severity of airway disease. The investigators also
reported a decrease in deposition with increasing mean effective airspace diameter; this
suggested that the enhanced deposition was associated more with the chronic bronchitic
component of COPD than with the emphysematous component. Greater deposition was noted
with natural breathing compared to the fixed pattern.
Brown et al. (2002) measured the deposition of an ultrafine aerosol (CMD = 0.033 jim) in
10 patients with moderate-to-severe COPD (mean age 61 years) and 9 healthy adults (mean age
53 years). The COPD group consisted of 7 patients with chronic bronchitis and 3 patients with
emphysema. All subjects respired aerosol at their individual resting breathing pattern, which had
been previously measured. The aerosol deposition fraction in the bronchitic patients (0.67) was
significantly (p < 0.02) greater than in either the patients with emphysema (0.48) or the healthy
subjects (0.54). Minute ventilation increased with disease severity (healthy, 5.8 L/min; chronic
bronchitic, 6.9 L/min; emphysema, 11 L/min). For an aerosol exposure of 10 |ig/m3, the dose
rates for the healthy, bronchitic, and emphysemic subjects were 1.9 |ig/h, 2.8 |ig/h (different
from healthy, p < 0.05), and 3.2 |ig/h, respectively. Hence, relative to the healthy subjects, the
average dose rate was significantly (p < 0.05) increased by 54% in the COPD patients, whereas
the deposition fraction only tended to be increased by 15%. Consistent with Bennett et al.
(1997b), these data demonstrate the need to consider dose rates (which depend on minute
ventilation) rather than just deposition fractions when evaluating the effect of respiratory disease
on particle deposition and dose.
Kim and Kang (1997) measured lung deposition of l-|im particles inhaled via the mouth
by healthy adults (mean age 27 years) and by those with various degrees of airway obstruction,
namely smokers (mean age 27 years), smokers with small airway disease (SAD; mean age
37 years), asthmatics (mean age 48 years), and patients with COPD (mean age 61 years)
breathing under the same controlled pattern. Deposition fraction was obtained by measuring the
number of particles inhaled and exhaled, breath by breath. There was a marked increase in
deposition in people with COPD. Deposition was 16%, 49%, 59%, and 103% greater in
smokers, smokers with SAD, asthmatics, and people with COPD, respectively, than in healthy
6-32
-------
adults. Deposition in COPD patients was significantly greater than that associated with either
SAD or asthma; there was no significant difference in deposition between people with SAD and
asthma. Deposition fraction was found to be correlated with percent predicted forced expiratory
volume (FEVj) and forced expiratory flow (FEF25.750/0). Airway resistance was not correlated
strongly with total lung deposition. Kohlhaufl et al. (1999) showed increased deposition of fine
particles (0.9 jim) in women with bronchial hyperresponsiveness.
Brown et al. (2001) examined the relationship between regional lung deposition for coarse
particles (5 jim) and ventilation distribution in healthy adults and in patients with CF. They
found that deposition in the TB region was positively associated with regional ventilation in
healthy subjects, but negatively associated in CF patients. The relationships were reversed for
deposition in the A region. These data suggest that significant coarse particle deposition may
occur in the TB region of poorly ventilated lungs, as occurs in CF; whereas TB deposition
follows ventilation in healthy subjects. This study demonstrated that there can be large
differences in regional particle deposition within the diseased lung and that these differences are
at least partially due to ventilation distribution.
Segal et al. (2000a) developed a mathematical model for airflow and particle motion in the
lung that was used to evaluate how lung cancer affects deposition patterns in the lungs of
children. It was noted that the presence of airway tumors could affect deposition by increasing
the probability of inertial deposition and diffusion. The former would occur on upstream
surfaces of tumors and the latter on downstream surfaces. It was concluded that particle
deposition is affected by the presence of airway disease, that effects may be systematic and
predictable, and that, therefore, they could be incorporated into dosimetry models. Segal et al.
(2002) used a computer model to calculate the deposition fractions of PM within the lungs of
COPD patients. The original model was for a healthy lung with a total volume of 4800 mL. The
chronic bronchitis component of COPD was modeled by reducing airway diameters based on
airway resistance measurements in vivo. The emphysema component was modeled by
increasing alveolar volumes by 10 to 30%. The calculated results were compared with
experimental data obtained from COPD patients for controlled breathing trials (tidal volume of
500 mL, respiratory time of 1 s) with a particle size of 1 |im. The model successfully depicts
PM deposition patterns and their dependence on the severity of disease and indicates that airway
obstructions are the main cause for increased deposition in the COPD lung.
6-33
-------
Thus, the database related to particle deposition and lung disease suggests that total lung
deposition generally is increased with obstructed airways, regardless of deposition distribution
between the TB and A regions. Airflow distribution is very uneven in diseased lungs because of
the irregular pattern of obstruction, and there can be closure of small airways. In this situation, a
part of the lung is inaccessible, and particles can penetrate deeper into other, better ventilated
regions. Thus, deposition can be enhanced locally in regions of active ventilation, particularly in
the A region.
6.2.3.4 Anatomical Variability
As indicated above, variations in anatomical parameters between genders and between
healthy people and those with obstructive lung disease can affect deposition patterns. However,
previous analyses generally have overlooked the effect on deposition of normal inter-individual
variability in airway structure in healthy individuals. This is an important consideration in
dosimetry modeling, which often is based on a single idealized structure. Studies that have
become available since the 1996 PM AQCD have attempted to assess the influence of such
variation in respiratory tract structure on deposition patterns.
The ET region is the first to contact inhaled particles and, therefore, deposition within this
region would reduce the amount of particles available for deposition in the lungs. Variations in
relative deposition within the ET region will, therefore, propagate through the rest of the
respiratory tract, creating differences in calculated doses from individual to individual.
A number of studies have examined the influence of variations in airway geometry on deposition
in the ET region.
Cheng et al. (1996) examined nasal airway deposition in healthy adults using particles
ranging in size from 0.004 to 0.15 jim and at two constant inspiratory flow rates, 167 and
333 mL/s. Deposition was evaluated in relation to measures of nasal geometry as determined by
magnetic resonance imaging and acoustic rhinometry. They noted that inter-individual
variability in deposition was correlated with the wide variation of nasal dimensions, in that
greater surface area, smaller cross-sectional area, and increasing complexity of airway shape
were all associated with enhanced deposition.
Using a regression analysis of data on nasal airway deposition derived from Cheng et al.
(1996), Guilmette et al. (1997) noted that the deposition efficiency within this region was highly
6-34
-------
correlated with both nasal airway surface area and volume. This indicated that airway size and
shape factors were important in explaining intra-individual variability observed in experimental
studies of human nasal airway aerosol deposition. Thus, much of the variability in measured
deposition among people arose from differences in the size and shape of specific airway regions.
As described in Section 6.2.2.4, Bennett et al. (1998) investigated the effects of ADS on
particle deposition and retention in bronchial airways, using an aerosol bolus technique. They
found that the fractional deposition was dependant on the subject's ADS and that a significant
number of particles was retained beyond 24 h. This finding of prolonged retention of insoluble
particles in the airways is consistent with the findings of Scheuch et al. (1995) and Stahlhofen
et al. (1986a) and with the predictions of asymmetric stochastic human lung models (Asgharian
et al., 2001). Bennett et al. (1999) also found a lung volume-dependent asymmetric distribution
of particles between the left and right lung; the leftright ratio was increased at increased
percentage of total lung capacity (e.g., at 70% TLC, L:R was 1.60).
From the analysis of detailed deposition patterns measured by a serial-bolus mouth-
delivery method, Kim and Hu (1998) and Kim and Jaques (2000) found a marked enhancement
in deposition in the very shallow region (lung penetration depth < 150 mL) of the lungs in
females. The enhanced local deposition for both ultrafme and coarse particles was attributed to a
smaller size of the upper airways, particularly of the laryngeal area.
Kesavanathan and Swift (1998) also evaluated the influence of geometry in affecting
deposition in the nasal passages of normal adults from two ethnic groups. Mathematical
modeling of the results indicated that the shape of the nostril affected particle deposition in the
nasal passages, but that there still remained large inter-subject variations in deposition when this
was accounted for, and which was likely caused by geometric variability in the mid and posterior
regions of the nasal passages.
Hofmann et al. (2000) examined the role of heterogeneity of airway structure in the rat
acinar region in affecting deposition patterns within this area of the lungs. Using different
morphometric models, they showed a substantial variability in predicted particle deposition and
concluded that the heterogeneity of acinar airway structure is primarily responsible for the
heterogeneity of acinar particle deposition.
6-35
-------
6.2.3.5 Inhaled Irritants
As noted above (Section 6.2.3.3), the narrowing of bronchial airways due to certain chronic
disease states (e.g., bronchoconstriction associated with asthma) tends to increase TB region
deposition, but decreases A region deposition. Bronchoconstriction induced by inhaled irritants,
as discussed by Schlesinger (1995), would analogously be expected to increase TB deposition.
Bronchoconstrictive effects are among the more notable effects associated with acute exposures
to both SO2 and O3 as discussed in detail in respective EPA criteria documents for those
pollutants (U.S. Environmental Protection Agency, 1994, 1996b). Thus, co-exposures to either
SO2 or O3 sufficient to induce bronchoconstriction could potentially enhance particle deposition
in the TB region, but reduce in the deposition A region.
6.2.4 Interspecies Patterns of Deposition
The primary purpose of this document is to assess the health effects of particles in humans.
As such, human dosimetry studies have been stressed in this chapter. Such studies avoid the
uncertainties associated with the extrapolation of dosimetric data from laboratory animals to
humans. However, animal models have been and continue to be used in evaluating PM health
effects because of ethical limits on the types of studies that can be performed with human
subjects. Thus, there is a considerable need to understand dosimetry in animals and dosimetric
differences between animals and humans. In this regard, there are a number of newly published
studies that assess particle dosimetry in commonly used animals and attempt to relate this to
dosimetry in humans.
The various species used in inhalation toxicologic studies that serve as the basis for
dose-response assessment may not receive identical doses in a comparable respiratory tract
region (i.e., ET, TB, or A) when exposed to the same aerosol at the same inhaled concentration.
Such interspecies differences are important because toxic effects are potentially related to the
quantitative pattern of deposition within the respiratory tract as well as to the exposure
concentration. The pattern of deposition determines not only the initial respiratory tract tissue
dose, but also the specific pathways by which deposited material is cleared and redistributed
(Schlesinger, 1985). Differences in patterns of deposition between humans and animals were
assessed previously in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) and
6-36
-------
by others (Schlesinger et al., 1997). Such differences in initial deposition must be considered
when relating biological responses obtained in laboratory animal studies to effects in humans.
It is difficult to systematically compare interspecies deposition patterns obtained from
various reported studies because of variations in experimental protocols, measurement
techniques, definitions of specific respiratory tract regions, and so on. For example, tests with
humans are generally conducted under protocols that standardize the breathing pattern; whereas
those using laboratory animals involve a wider variation in respiratory exposure conditions (e.g.,
spontaneous breathing versus ventilated breathing or varying degrees of sedation). Much of the
variability in the reported data for individual species may be due to the lack of normalization for
specific respiratory parameters during exposure. In addition, the various studies have used
different exposure techniques, such as nasal mask, oral mask, oral tube, or tracheal intubation.
Regional deposition is affected by the exposure route and delivery technique employed.
Figure 6-9 shows the regional deposition data versus particle diameter in commonly used
laboratory animals obtained by various investigators as compiled by Schlesinger (1988, 1989).
The results are described in detail in the 1996 PM AQCD (U.S. Environmental Protection
Agency, 1996a). In general, there is much variability in the data; however, it is possible to make
some generalizations concerning comparative deposition patterns. The relationship between
total respiratory tract deposition and particle size is approximately the same in humans and most
of these animals: deposition increases on both sides of a minimum that occurs for particles of
0.2 to 1 jim. Interspecies differences in regional deposition occur due to anatomical and
physiological factors. In most laboratory animal species, deposition in the ET region is near
100% percent for particles greater than 5 jim AED (Raabe et al., 1988), indicating greater
efficiency than that seen in humans. In the TB region, there is a relatively constant, but lower,
deposition fraction for particles greater than 1 |im AED in all species compared to humans.
Finally, in the A region, deposition fraction peaks at a lower particle size (about 1 |im AED) in
laboratory animals than in humans.
One of the issues that must be considered in interspecies comparisons of hazards from
inhaled particles is inhalability of the aerosol in the atmosphere of concern. Inhalability is the
fraction of suspended PM in ambient air that actually enters the nose or mouth with the volume
of air inhaled and is a function of a particle's AED, inspiratory flow rate, wind speed, and wind
direction. Although inhalability may not be an issue for humans per se as far as exposure to
6-37
-------
100
c
O
'55
0)
Q
^ 80 -
O Rat
D Hamster
A Mouse
<> Guinea Pig
V Dog
0.01
0.1
0.01
0.1
60 -
O Rat
D Hamster
A Mouse
^ Guinea Pig
V Dog
1.0
0.01
0.1 1.0
Particle Diameter (pm)
40 -
20 -
n
n Rat
n Hamster
A Mouse
O Guinea Pig
V Dog
V <
t7^
I
t Ł C
\ A<>
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
Ł 0.5 jim and geometric (or diffusion equivalent) for those < 0.5 um.
Source: Schlesinger(1988).
6-38
-------
ambient particles is concerned, it can be important when attempting to extrapolate to humans the
results of studies using animal species commonly employed in inhalation toxicological studies
(Miller et al., 1995). For example, differences between rat and human become very pronounced
for particles > 5 jim AED, and some differences are also evident for particles as small as 1 |im
AED (Figure 6-10). Menache et al. (1995) developed equations that can be used to determine
the inhalability adjustments needed as a function of particle size to compare laboratory animal
and human studies.
A number of studies have addressed various aspects of interspecies differences in
respiratory tract deposition using mathematical modeling approaches. Hofmann et al. (1996)
compared deposition between rat and human lungs using three-dimensional asymmetric
bifurcation models and mathematical procedures for obtaining air flow and particle trajectories.
Deposition in segmental bronchi and terminal bronchioles was evaluated under both inspiration
and expiration at particle sizes of 0.01, 1.0, and 10 jim, which covers the range of deposition
mechanisms from diffusion to impaction. Total deposition efficiencies of all particles in the
upper and lower airway bifurcations were comparable in magnitude for both rat and human.
However, the investigators noted that penetration probabilities from preceding airways must be
considered. When considering the higher penetration probability in the human lung, the
resulting bronchial deposition fractions were generally higher than in the rat. For all particle
sizes, deposition at rat bronchial bifurcations was less enhanced on the carinas compared to that
found in human airways.
Hofmann et al. (1996) attempted to account for interspecies differences in branching
patterns in deposition analyses. Numerical simulations of three-dimensional particle deposition
patterns within selected (species-specific) bronchial bifurcations indicated that morphologic
asymmetry was a major determinant of the heterogeneity of local deposition patterns. They
noted that many interspecies deposition calculations used morphometry that was described by
deterministic lung models (i.e., the number of airways in each airway generation is constant, and
all airways in a given generation have identical lengths and diameters). Such models cannot
account for variability and branching asymmetry of airways in the lungs. Thus, their study
employed computations that used stochastic morphometric models of human and rat lungs
(Koblinger and Hofmann, 1985, 1988; Hofmann et al., 1989b) and evaluated regional and local
particle deposition. Stochastic models of lung structure describe, in mathematical terms, the
6-39
-------
a. Total Respiratory Tract
b. Extrathoracic Region
100
Human
. Oral Breathing
100
O
jE
0>
O
a.
o>
Q 0
Human
- Nasal Breathing
100 r Rat
Particle Diameter
100
Human
. Oral Breathing
100
O
jE
0>
O
a.
o>
Q 0
Human
- Nasal Breathing
100r
Particle Diameter (\im)
c. Tracheobronchial Region
100r Human
_ Oral Breathing
C
O
+J
'to
o
Q.
0)
Q
100r
Rat
Particle Diameter
d. Alveolar Region
100i- Human
. Oral Breathing
100
o
Q.
0)
Q 0
• Human
. Nasal Breathing
100r
Rat
Particle Diameter (\im)
Figure 6-10. Particle deposition efficiency in rats and humans as a function of particle size
for (a) total respiratory tract, (b) extrathoracic 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 um.
Source: Modified from Schlesinger (1989).
6-40
-------
inherent asymmetry and variability of the airway system, including diameter, length, and angle.
They are based on statistical analyses of actual morphometric analyses of lungs. The model also
incorporated breathing patterns for humans and rats.
In a later analysis (Hofmann and Bergmann, 1998), the dependence of deposition on
particle size was found to be qualitatively similar in both rats and humans. Deposition minima
were found for total deposition as well as deposition within the TB and A regions in the size
range of 0.1 to 1 |im. In addition, a deposition maximum occurred at about 0.02 to 0.03 jim and
between 3 and 5 jim in both species. The deposition decrease in the A region at the smallest and
largest sizes resulted from the filtering efficiency of upstream airways. Although deposition
patterns were qualitatively similar in rat and human, deposition in the human lung appeared to be
consistently higher than in the rat in all regions of the lung (TB and A) over the entire size range.
Both species showed a similar pattern of dependence of deposition on flow rate. In both human
and rat, deposition of 0.001-|im particles was highest in the upper bronchial airways; whereas
0.1- and l-|im particles showed higher deposition in more peripheral airways, namely the
bronchiolar airways in rat and the respiratory bronchioles in humans. Deposition was variable
within any branching generation because of differences in airway dimensions, and regional and
total deposition also exhibited intrasubject variations. Airway geometric differences between
rats and humans were reflected in deposition. Because of the greater branching asymmetry in
rats prior to about generation 12, each generation showed deposition maxima at two particle
sizes, reflecting deposition in major and minor daughters. These geometric differences became
reduced with depth into the lung; beyond generation 12, these two maxima were no longer seen.
Another comparison of deposition in lungs of humans and rats was performed by Musante
and Martonen (2000b). An interspecies mathematical dosimetry model was used to determine
the deposition of ROFA (MMAD, 1.95 jim; og, 2.19)in the lungs under sedentary and light
activity breathing patterns. This latter condition was mimicked in the rat by increasing the CO2
level in the exposure system. They noted that physiologically comparable respiratory intensity
levels did not necessarily correspond to comparable dose distribution in the lungs. Because of
this, the investigators speculated that the resting rat may not be a good model for the resting
human. The ratio of aerosol mass deposited in the TB region to that in the A region for the
human at rest was 0.961, indicating fairly uniform deposition throughout the lungs. On the other
hand, in the resting rat, the ratio was 2.24, indicating greater deposition in the TB region than in
6-41
-------
the A region. However, by mimicking light activity in the rat, the ratio was reduced to 0.97,
similar to the human. These data underscore the need for dose-response studies and for models
that are capable of adjusting for the dosimetric differences between species.
The relative distribution of particles deposited within the bronchial and alveolar regions of
the airways may differ in the lungs of animals and humans for the same total amount of
deposited matter because of structural differences. The effect of such structural differences
between rat and human airways on particle deposition patterns was examined by Hofmann et al.
(1999, 2000) in an attempt to find the most appropriate morphometric parameter to characterize
local particle deposition for extrapolation modeling purposes. Particle deposition patterns were
evaluated as functions of three morphometric parameters, namely (1) airway generation,
(2) airway diameter, and (3) cumulative path length. It was noted that airway diameter was a
more appropriate morphometric parameter for comparison of particle deposition patterns in
human and rat lungs than was airway generation.
The manner in which particle dose is expressed, that is, the specific dose metric, may affect
relative differences in deposition between humans and other animal species. For example,
although deposition when expressed on a mass per unit alveolar surface area basis may not be
different between rats and humans, dose metrics based on particle number per various
anatomical parameters (e.g., per alveolus or alveolar macrophage) can differ between rats and
humans, especially for particles around 0.1 to 0.3 |im (Miller et al., 1995). Furthermore, in
humans with lung disease (e.g., asthma or COPD), rat and human differences can be even more
pronounced.
The probability of any biological effect occurring in humans or animals depends on
deposition and retention of particles, as well as the underlying tissue sensitivity. Interspecies
dosimetric extrapolation must consider these differences in evaluating dose-response
relationships. Thus, even similar deposition patterns may not result in similar effects in different
species because dose is also affected by clearance mechanisms. In addition, the total number of
particles deposited in the lung may not be the most relevant dose metric for interspecies
comparisons. For example, it may be the number of deposited particles per unit surface area or
dose to a specific cell (e.g., alveolar macrophage) that determines response for specific regions.
More specifically, even if fractional deposition is similar in the rat and human, there would be
differences in deposition density because of the higher metabolic rate in the rat. Thus, species-
6-42
-------
specific differences in deposition density should be considered when health effects observed in
laboratory animals are being evaluated for potential effects occurring in humans.
6.3 PARTICLE CLEARANCE
This section discusses the clearance and translocation of particles that have deposited in the
respiratory tract. Here, clearance refers to the processes by which deposited particles are
removed from the surface of the respiratory tract. Translocation refers to specific clearance
processes emphasizing movement of particles from one specific location to another either within
the lung or to extrapulmonary organs. First, a basic overview of biological mechanisms and
pathways of clearance in the various region of the respiratory tract is presented. This is followed
by an update on regional kinetics of particle clearance. Interspecies patterns of clearance are
then addressed, followed by new information on biological factors that may modulate clearance.
6.3.1 Mechanisms and Pathways of Clearance
Particles that deposit on airway surfaces may be either cleared from the respiratory tract
completely or translocated to other sites within this system by various regionally distinct
processes. These clearance mechanisms, outlined in Table 6-2, can be categorized as either
absorptive (i.e., dissolution) or nonabsorptive (i.e., transport of intact particles) and may occur
simultaneously or with temporal variations. It should be mentioned that particle solubility in
terms of clearance refers to solubility within the respiratory tract fluids and cells. Thus, a poorly
soluble particle is considered to be one whose rate of clearance by dissolution is insignificant
compared to its rate of clearance as an intact particle. All deposited particles, therefore, are
subject to clearance by the same basic mechanisms, with their ultimate fate a function of
deposition site, physicochemical properties (including solubility and any toxicity), and
sometimes deposited mass or number concentration. Clearance routes from the various regions
of the respiratory tract have been discussed previously in detail (U.S. Environmental Protection
Agency, 1996a; Schlesinger et al., 1997). They are schematically shown in Figure 6-11 (for
extrathoracic and tracheobronchial regions) and in Figure 6-12 (for poorly soluble particle
clearance from the alveolar region) and are reviewed only briefly below.
6-43
-------
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 j|
C Blood J
Dissolution
Blood H C 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).
6-44
-------
Deposited Particle
P
T
(Phagocytosis by ^
Alveolar Macrophages ) V
1 ^
Movement within . Passac
Alveolar Lumen * Alveola
1
Bronchiolar/Bronchial ^ Inte
Lurnen •*
» Lymph a
Mucociliary Blanket
^ Lymf
Gl Tract
Endocyt
Epithelic
1 r
je Through
r Epithelium "*""
g
1
rstitium ^
osis I
Iveol
alCe
1
tic Channels ^
1
3V
ar
Is «™ L* I u u u ^
^ t
Passage through
Pulmonary Capillary
Endothelium
!'
—/^Phagocytosis by~\
* S Interstitial 1
Y Macrophages J
Figure 6-12. 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).
6.3.1.1 Extrathoracic Region
The clearance of poorly soluble particles deposited in the posterior portions of the nasal
passages occurs via mucociliary transport, with the general flow of mucus being towards the
nasopharynx. Mucus flow in the most anterior portion of the nasal passages is forward, clearing
deposited particles to the vestibular region where removal occurs by sneezing, wiping, or
blowing. Soluble material deposited on the nasal epithelium is accessible to underlying cells via
diffusion through the mucus. Dissolved substances may be translocated subsequently into the
bloodstream. The nasal passages have a rich vasculature, and uptake into the blood from this
region may occur rapidly.
Clearance of poorly soluble particles deposited in the oral passages is by coughing and
expectoration or by swallowing into the gastrointestinal tract. Soluble particles are likely to be
rapidly absorbed after deposition, but deposition depends on the rate of dissolution of the particle
and the molecular size of the solute.
6-45
-------
6.3.1.2 Tracheobronchial Region
Poorly soluble particles deposited within the TB region are cleared by mucociliary
transport towards the oropharynx, followed by swallowing. Poorly soluble particles also may
traverse the epithelium by endocytotic processes, entering the peribronchial region, where they
may be phagocytized by airway macrophages (which can then be cleared via the mucociliary
blanket), or enter the airway lumen from the bronchial or bronchiolar mucosa. Soluble particles
may be absorbed through the epithelium into the blood. It has been shown that blood flow
affects translocation from the TB region in that decreased bronchial blood flow is associated
with increased airway retention of soluble particles (Wagner and Foster, 2001). There is,
however, evidence that even soluble particles may be cleared by mucociliary transport (Bennett
and Ilowite, 1989; Matsui et al., 1998; Wagner and Foster, 2001).
6.3.1.3 Alveolar Region
Clearance from the A region occurs via a number of mechanisms and pathways. Particle
removal by macrophages is the main nonabsorptive clearance process in this region. Alveolar
macrophages, which reside on the epithelium, phagocytize and transport deposited material that
they contact by random motion or via directed migration under the influence of chemotactic
factors.
Although alveolar macrophages normally account for up to about 3 to 19% of the total
alveolar cells in healthy, nonsmoking humans and other mammals (Crapo et al., 1982), the actual
cell count may be altered by particle loading. The magnitude of any increase in cell number is
related to the number of deposited particles rather than to total deposition by weight. Thus,
equivalent masses of an identically deposited substance would not produce the same response if
particle sizes differed; and the deposition of smaller particles would tend to result in a greater
elevation in macrophage number than would deposition of larger particles.
Particle-laden macrophages may be cleared from the A region via a number of pathways.
As noted in Figure 6-11, this includes transport toward the pharynx via the mucociliary system
after the cells reach the distal terminus of the mucus blanket; movement within the interstitium
to a lymphatic channel; or possibly traversing of the alveolar-capillary endothelium and directly
entering the bloodstream. Note that the latter pathway of particle-laden macrophages entering
the bloodstream is speculative and not yet demonstrated. Particles within the lymphatic system
6-46
-------
may be translocated to TB lymph nodes, which can become reservoirs of retained material.
Particles subsequently reaching the postnodal lymphatic circulation will enter the blood. Once
in the systemic circulation, these particles can travel to extrapulmonary organs. Deposited
particles (especially those < 0.5 to 1.0 jim) that are not ingested by alveolar macrophages may
enter the interstitium and be phagocytized by resident interstitial macrophages, and may travel to
perivenous, peribronchiolar or subpleural sites where they become trapped, increasing particle
burden. The migration and grouping of particles and macrophages within the lungs can lead to
the redistribution of initially diffuse deposits into focal aggregates. Some particles or
components can bind to epithelial cell membranes, to macromolecules, or to other cell
components, delaying their clearance from the lungs.
Churg and Brauer (1997) examined lung autopsy tissue from 10 people who had never
smoked from Vancouver, Canada. They noted that the geometric mean particle diameter
(GMPD) in lung parenchymal tissue was 0.38 jim (og = 2.4). Ultrafine particles accounted for
less than 5% of the total retained particulate mass. Metal particles had a GMPD of 0.17 jim and
silicates, 0.49 jim. Ninety-six percent of retained PM had a GMPD less than 2.5 jim.
A subsequent study considered retention of ambient particles in the lungs. Brauer et al. (2001)
showed that small particles could undergo significant steady-state retention within the lungs.
Using lungs obtained at autopsy from long-term, nonsmoking residents of an area having high
levels of ambient PM (Mexico City, Mexico) and those from an area with relatively low PM
levels (Vancouver, Canada), the investigators measured the particle concentration per gram of
lung within the parenchyma. They found that living in the high PM region resulted in
significantly greater retention of both fine and ultrafine particles within the lungs: levels in the
lungs from residents of Mexico City contained over 7.4 times the concentration of these particles
as did lungs from residents of Vancouver. These results indicate a clear relationship between
ambient exposure concentration and retention in the A region.
Clearance by the absorptive mechanism involves dissolution in the alveolar surface fluid
followed by transport through the epithelium and into the interstitium, and then diffusion into the
lymph or blood. Solubility is influenced by the particle's surface to volume ratio and other
properties, such as hydrophilicity and lipophilicity (Mercer, 1967; Morrow, 1973; Patton, 1996).
6-47
-------
6.3.2 Clearance Kinetics
The kinetics of clearance have been reviewed in U.S. Environmental Protection Agency
(1996a) and in a number of monographs (e.g., Schlesinger et al., 1997) and are discussed only
briefly here. The actual time frame over which clearance occurs affects the cumulative dose
delivered to the respiratory tract, as well as the dose delivered to extrapulmonary organs.
6.3.2.1 Extrathoracic Region
Mucus flow rates in the posterior nasal passages are highly nonuniform, but the median
rate in a healthy adult human is about 5 mm/min, resulting in a mean anterior to posterior
transport time of about 10 to 20 min for poorly soluble particles (Rutland and Cole, 1981;
Stanley et al., 1985). Particles deposited in the anterior portion of the nasal passages are cleared
more slowly by mucus transport and are usually more effectively removed by sneezing, wiping,
or nose blowing (Fry and Black, 1973; Morrow, 1977).
6.3.2.2 Tracheobronchial Region
Mucus transport in the TB tree occurs at different rates in different local regions: the
velocity of movement is fastest in the trachea, and it becomes progressively slower in more
distal airways. In healthy nonsmoking humans, using noninvasive procedures and no anesthesia,
average tracheal mucus transport rates have been measured at 4.3 to 5.7 mm/min (Yeates et al.,
1975, 1981; Foster et al., 1980; Leikauf et al., 1981, 1984), whereas that in the main bronchi has
been measured at -2.4 mm/min (Foster et al., 1980). Estimates for human medium bronchi
range between 0.2 to 1.3 mm/min, while those in the most distal ciliated airways range down to
0.001 mm/min (Morrow et al., 1967; Cuddihy and Yeh, 1988; Yeates and Aspin, 1978).
The total duration of bronchial clearance or some other time parameter is often used as an
index of mucociliary kinetics. Although clearance from the TB region is generally rapid,
experimental evidence, discussed in U.S. Environmental Protection Agency (1996a), was shown
that a fraction of material deposited in the TB region is retained much longer than the 24 h
commonly considered the outer range of clearance time for particles within this region
(Stahlhofen et al., 1986a,b; Scheuch and Stahlhofen, 1988; Smaldone et al., 1988). A study by
Asgharian et al. (2001) showed that it is not necessary to have a slow- and fast-phase TB
clearance for particles to be retained longer than 24 h. Based upon asymmetric stochastic
6-48
-------
human-lung modeling-data, inter-subject variability in path length and the number of generations
to the alveoli, may result in some material reaching the alveoli even with shallow breathing, this
can explain experimental observations while still describing TB clearance as a single
compartment model. Other studies described below, however, do support the concept that TB
regional clearance consists of both a fast and a slow component.
Falk et al. (1997) studied clearance in healthy adults using monodisperse
polytetraflouethylene (PTFE; Teflon) particles (6.2 jim) inhaled at two flow rates. Each subject
inhaled twice at two flow rates (0.45 and 0.045 L/s). Theoretical calculations indicated that the
particles inhaled at 0.45 L/s should deposit mainly in large bronchi and in the alveolar region;
whereas the particles inhaled at 0.045 L/s should deposit mainly in small ciliated airways.
At twenty-four hours after inhalation about half of the particles inhaled by both modes of
inhalation had cleared. Clearance beyond twenty-four hours was biphasic. For the inhalation
rate of 0.45 L/s, 15% cleared with a half-time of 3.4 days and 85% with a half-time of 190 days.
For the inhalation rate of 0.045 L/s, 20% cleared with a half-time of 2.0 days and 80% with a
half-time of 50 days. The results indicate that a considerable fraction of particles deposited in
small ciliated airways had not cleared within 24 h, and that these particles cleared differently
from particles deposited in the alveolar region. The authors observed that the experimental data
agreed well with the theoretical predictions. Camner et al. (1997) also noted that clearance from
the TB region was incomplete by 24 h postexposure and suggested that this may be caused by
incomplete clearance from bronchioles. Healthy adults inhaled teflon particles (6, 8, and 10 jim)
under low flow rates to maximize deposition in the small ciliated airways. The investigators
noted a decrease in 24-h retention with increasing particle size, indicating a shift toward either a
smaller retained fraction, deposition more proximally in the respiratory tract, or both. They
calculated that a large fraction, perhaps as high as 75% of particles depositing in generations
12 through 16, was still retained at 24 h postexposure.
In a study to examine retention kinetics in the TB tree (Falk et al., 1999), nonsmoking
healthy adults inhaled radioactively tagged 6. l-jim particles at both a normal flow rate and a
slow flow rate designed to deposit particles preferentially in small ciliated airways. Lung
retention was measured from 24 h to 6 mo after exposure. Following normal flow rate
inhalation, 14% of the particles retained at 24 h cleared with a half-time of 3.7 days and 86%
with a half-time of 217 days. Following slow flow rate inhalation, 35% of the particles retained
6-49
-------
at 24 h cleared with a half-time of 3.6 days and 65% with a half-time of 170 days. Estimates
using a number of mathematical models indicated higher deposition in the bronchiolar region
(generations 9 through 15) with the slow rate inhalation compared to the normal rate. The
experimental data and predictions of the deposition modeling indicated that 40% of the particles
deposited in the conducting airways during the slow inhalation were retained after 24 h. The
particles that cleared with the shorter half-time were mainly deposited in the bronchiolar region,
but only about 25% of the particles deposited in this region cleared in this phase. This study
provided additional support for a phase of slow clearance from the bronchial tree.
The underlying sites and mechanisms of long-term TB retention in the smaller airways are
not known. Some proposals were presented in the 1996 PM AQCD (U.S. Environmental
Protection Agency, 1996a). This slow clearing TB compartment likely is associated with
bronchioles < 1 mm in diameter (Lay et al., 1995; Kreyling et al., 1999; Falk et al., 1999). In a
study in which an adrenergic agonist was used to stimulate mucus flow to examine the role of
mucociliary transport in the bronchioles, clearance from the smaller airways was not influenced
by the drug, suggesting to the investigators that mucociliary transport was not as an effective
clearance mechanism from this region as it is in larger airways (Svartengren et al., 1998, 1999).
Although slower or less effective mucus transport may result in longer retention times in small
airways, other factors may account for long-term TB retention. One possibility is that particles
are displaced into the gel phase by the surface tension forces of the liquid lining the small
airways (Gehr et al., 1990, 1991). The issue of long-term particle retention in the TB tree
certainly is not resolved.
Long-term TB retention patterns are not uniform. An enhancement at bifurcation regions
(Radford and Martell, 1977; Henshaw and Fews, 1984; Cohen et al., 1988), is likely the result of
both greater deposition and less effective mucus clearance within these areas. Thus, doses
calculated based on uniform surface retention density may be misleading, especially if the
material is lexicologically slow acting.
6.3.2.3 Alveolar Region
Particles deposited in the A region generally are retained longer than are those deposited in
airways cleared by mucociliary transport. There are limited data on alveolar clearance rates in
humans. Within any species, reported clearance rates vary widely because, in part, of different
6-50
-------
properties of the particles used in the various studies. Furthermore, some chronic experimental
studies have employed high concentrations of poorly soluble particles that may have interfered
with normal clearance mechanisms, resulting in clearance rates different from those that would
typically occur at lower exposure levels. Prolonged exposure to high particle concentrations is
associated with what is termed particle "overload." This is discussed in greater detail in
Section 6.4.
There are numerous pathways of A-region clearance, and the utilization of these may
depend on the nature of the particles being cleared. Little is known concerning relative rates
along specific pathways. Thus, generalizations about clearance kinetics are difficult to make.
Nevertheless, A-region clearance is usually described as a multiphasic process, with each phase
representing removal by a different mechanism or pathway and often characterized by increased
retention half-times following toxicant exposure.
The initial uptake of deposited particles by alveolar macrophages is very rapid and
generally occurs within 24 h of deposition (Lehnert and Morrow, 1985; Naumann and
Schlesinger, 1986; Lay et al., 1998). The time for clearance of particle-laden alveolar
macrophages via the mucociliary system depends on the site of uptake relative to the distal
terminus of the mucus blanket at the bronchiolar level. Furthermore, clearance pathways and
subsequent kinetics may depend to some extent on particle size. For example, some smaller
ultrafine particles (< 0.02 jim) may be less effectively phagocytosed than larger ones
(Oberdorster, 1993).
Nonphagocytized particles may penetrate into the interstitium within a few hours following
deposition. This transepithelial passage seems to increase as particle loading increases,
especially to that level above which macrophage numbers increase (Ferin, 1977; Ferin et al.,
1992; Adamson and Bowden, 1981). It also may be particle-size dependent because insoluble
ultrafine particles (< 0.1 jim diameter) of low intrinsic toxicity show increased access to the
interstitum and greater lymphatic uptake than do larger particles of the same material
(Oberdorster et al., 1992; Ferin et al., 1992). However, ultrafine particles of different materials
may not enter the interstitium to the same extent. Similarly, a depression of phagocytic activity,
a reduction in macrophage ability to migrate to sites of deposition (Madl et al., 1998), or the
deposition of large numbers of ultrafine particles may increase the number of free particles in the
alveoli, perhaps enhancing removal by other routes. In any case, free particles may reach the
6-51
-------
lymph nodes perhaps within a few days after deposition (Lehnert et al., 1988; Harmsen et al.,
1985) although this route is not definitive and may be species dependent.
Kreyling et al. (2002) studied the translocation of insoluble ultrafme 192Ir radiolabeled
particles (15 and 80 nm CMD) inhaled by healthy, young adult, male rats ventilated for 1 h via
an endotracheal tube. At time points ranging from 6 h to 7 d, rats were sacrificed; and a
complete balance of 192Ir activity retained in the body and cleared by excretion was determined.
Thoracic deposition fractions of inhaled 15 and 80 nm particles were 0.49 and 0.28, respectively.
One week after inhalation, particles were predominantly cleared from the lungs into the
gastrointestinal tract and eliminated in feces. Minute particle translocation of < 1% of the
deposited particles into secondary organs such as liver, spleen, heart, and brain was measured
after systemic uptake from the lungs. The translocated fraction of the 80-nm particles was about
an order of magnitude less than that of 15-nm particles. In further investigations, the biokinetics
of ultrafme particles and soluble 192Ir was studied after administration by either gavage or
intratracheal instillation or intravenous injection. These studies confirmed the low solubility of
the 192Ir particles and showed that (1) particles were neither dissolved nor absorbed from the gut,
(2) systemically circulating particles were rapidly and quantitatively accumulated and retained in
the liver and spleen, and (3) soluble 192Ir instilled in the lungs was rapidly excreted via urine with
little retention in the lungs and other organs. This study indicates that only a rather small
fraction of ultrafme 192Ir particles are translocated from peripheral lungs to systemic circulation
and extrapulmonary organs following short-term exposures.
Oberdorster et al. (2002) exposed Fisher rats for 6 h to 13C-labeled ultrafme carbon
particles (CMDs = 20 to 29 nm) at concentrations of 80 or 180 |ig/m3 in compartmentalized
whole-body inhalation exposure chambers. Animals were sacrificed at 0.5, 18, and 24 h
postexposure; and 13C levels were determined in lung, liver, heart, kidney, olfactory bulb, and
brain. Interestingly, the 13C retained in lung at 0.5 h postexposure was -70% lower than
predicted for ultrafme particles in the rat lung by the MPPD model described below
(Section 6.6). Also of much interest, significant increases over control levels of 13C in lung and
liver (but not the other organs) were observed at all postexposure time points following
exposures to 180 |ig/m3 but only after 18- and 24-h for the 80 |ig/m3 exposures. This implies
translocation of the insoluble 13C ultrafme particles to liver in < 18-24 h postexposure at the
lower 80 |ig/m3 exposure and, possibly, even more rapid translocation to liver (by 0.5 h) at
6-52
-------
sufficiently higher concentrations. The relative increase due to translocation from lung to liver is
difficult to estimate, given potential contributions to liver 13C levels of carbon particles
translocated from the GI tract as derived from ingestion of particles by the rats while grooming
during and after the whole-body exposures.
The extent of lymphatic uptake of particles may depend on the effectiveness of other
clearance pathways in that lymphatic translocation likely increases when the phagocytic activity
of alveolar macrophages decreases. This may be a factor in lung overload. However, it seems
that the deposited mass or number of particles must exceed some threshold below which
increases in loading do not affect translocation rate to the lymph nodes (Ferin and Feldstein,
1978; LaBelle and Brieger, 1961). In addition, the rate of translocation to the lymphatic system
may be somewhat particle-size dependent. Although no human data are available, translocation
of latex particles to the lymph nodes of rats was greater for 0.5- to 2-|im particles than for 5- and
9-|im particles (Takahashi et al., 1992). Translocation of 3 jim particles has also been reported
(Snipes and Clem, 1981). On the other hand, translocation to the lymph nodes was similar for
both 0.4-|im barium sulfate or 0.02-jim gold colloid particles (Takahashi et al., 1987). It seems
that particles < 2 jim clear to the lymphatic system at a rate independent of size; and it is
particles of this size, rather than those > 5 jim, that would have significant deposition within the
A region following inhalation. In any case, the normal rate of translocation to the lymphatic
system is quite slow; and elimination from the lymph nodes is even slower, with half-times
estimated in tens of years (Roy, 1989).
Soluble particles depositing in the A region may be cleared rapidly via absorption through
the epithelial surface into the blood. Actual rates depend on the size of the particle (i.e., solute
size), with smaller molecular weight solutes clearing faster than larger ones. Absorption may be
considered as a two-stage process: first, deposited particles are dissolved; second, this dissolved
material moves into the blood circulation. Each of these stages may be time dependent. The rate
of dissolution depends on a number of factors, including particle surface area and chemical
structure. A portion of the dissolved material may be absorbed more slowly because of binding
to respiratory tract components. Accordingly, there is a very wide range for absorption rates,
depending on the physicochemical properties of the material deposited.
As indicated in both the toxicology and epidemiology chapters (Chapters 7 and 8) of this
document, concern exists about how ambient PM affects the cardiovascular system. Thus, an
6-53
-------
important dosimetric issue involves the pathways by which inhaled and deposited particles in the
lungs could affect extrapulmonary systems. Several studies (Huchon et al., 1987; Peterson et al.,
1989; Morrison et al., 1998) had earlier investigated lung clearance of labeled macromolecule
solutes with widely varying molecular weight and labeled albumin as well as albumin ultrafine
aggregates. Clearance rates found by these studies were much slower than for some more recent
studies described below which suggests the possibility of a fast clearing pathway for solid
ultrafine particles. Several newer studies have also evaluated possible pathways by which PM,
constituents of PM, or cytokines released by the respiratory tract in response to PM could affect
systems distal to the respiratory tract.
For example, Takenaka et al. (2001) exposed rats by inhalation to 0.015 jim particles of
elemental silver (Ag) and found elevated Ag levels in various extrapulmonary organs up to
7 days postexposure. They found that the amount of Ag in the lungs decreased rapidly with
time; by day 7, only about 4% of the initial lung burden remained. On the exposure day, Ag was
already found in the blood. By 1 day postexposure, Ag had been distributed to the liver, kidney,
heart, and brain. The Ag concentration was highest in the kidney, followed by the liver, and then
the heart. A similar clearance pattern was found after intratracheal instillation of AgNO3
solution. Therefore, the investigators hypothesized that the rapid clearance of ultrafine silver
particles was due to rapid dissolution into the lung fluid and subsequent diffusion into the
bloodstream, although a possibility of direct translocation of solid particles into the bloodstream
was not excluded. The investigators also instilled an aqueous suspension of elemental Ag
particles (100+ nm) into some animals. In this case, there was more retention in the lungs,
which was ascribed to phagocytic accumulation of agglomerated particles in alveolar
macrophages and slow dissolution of particles in cells. Thus, this study also suggested that
particle size and the tendency of particles to aggregate can affect the translocation pathway from
the lungs.
In another study, Nemmar et al. (2001) evaluated the movement of radiolabeled (99mTc)
ultrafine particles out of the lungs of hamsters receiving a single IT instillation of albumin
colloid particles (<80 nm) and killed after 5, 15, 30, and 60 min. Blood radioactivity, at 5, 15,
30, and 60 min, respectively, expressed as percentage of total body radioactivity per gram blood,
was 2.88 ± 0.80%, 1.30 ± 0.17%, 1.52 ± 0.46%, and 0.21 ± 0.06%. Liver radioactivity, at 5, 15,
30, and 60 min, respectively, expressed as a percentage of total radioactivity per organ, was
6-54
-------
0.10 ± 0.07%, 0.23 ± 0.06%, 1.24 ± 0.27%, and 0.06 ± 0.02%. Lower values were observed in
the heart, spleen, kidneys, and brain. Dose dependence was assessed at 30 min after instillation
of 10 jig and 1 jig 99mTc-albumin per animal (n = 3 at each dose), and values of the same relative
magnitudes as after instillation of 100 jig were obtained. The authors concluded that a
significant fraction of ultrafine 99mTc-albumin diffuses rapidly from the lungs into the systemic
circulation.
Nemmar et al. (2002) investigated the extent to which inhaled particles entered into the
systemic circulation in 5 healthy volunteers who inhaled "Technegas," an aerosol consisting
mainly of ultrafine 99mTc -labeled carbon particles (< 100 nm). Radioactivity detected in blood
at 1 minute, reached a maximum between 10 and 20 minutes, and remained at this level for up to
60 minutes. Thin layer chromatography of blood showed that in addition to a species
corresponding to oxidized 99mTc (i.e., pertechnetate) there was also a species corresponding to
particle-bound 99mTc. Gamma camera images showed substantial radioactivity over the liver and
other areas of the body. These workers concluded that inhaled 99mTc-labeled ultrafine carbon
particles pass rapidly into the systemic circulation. This appears to suggest that ultrafine
particles can rapidly diffuse from the lungs into the systemic circulation, thus providing a
pathway by which ambient PM may rapidly affect the heart and other extrapulmonary organs.
However, the stability of the 99mTc label of ultrafine particles could pose a serious problem in the
interpretation of the study. If the 99mTc label is leached from the particles, it can quickly spread
to other organs.
Results in marked contrast to Nemmar et al. (2001, 2002) have been reported by Brown
et al. (2002). The deposition and clearance of a 99mTc-labeled ultrafine aerosol (CMD,
33 ± 2 nm; AMD, 61 ± 4nm) were studied by Brown et al. (2002) in 9 healthy human adult
volunteers (aged 40 to 67 yrs) and 10 COPD patients (45 to 70 yrs) with moderate to severe
airway obstruction. No differences in clearance rates were detected between healthy and COPD
patients; nor was any rapid accumulation of radiolabeled particles found in the liver, based on
analyses at 10-min increments up to two hours postexposure. Brown et al. (2002) noted that
during Technegas generation by the method employed by Nemmar et al. (2002), the presence of
minimal oxygen (0.1 to 0.2%) can cause the formation of Pertechnegas. Unlike Technegas,
which is generally stable in the lung, Pertechnegas is rapidly ionized into pertechnetate and, with
Pertechnegas, the radiolabel quickly dissociates from the ultrafine particles following deposition
6-55
-------
in the lung. Highly soluble in normal saline, pertechnetate clears rapidly from the lung with a
half-time of-10 minutes and accumulates most notably in the bladder, stomach, thyroid, and
salivary glands. Brown et al. (2002) contend that the findings reported by Nemmar et al. (2002)
are consistent with pertechnetate clearance, but not insoluble ultrafme particles. Burch (2002)
also suggested that the Nemmar et al. (2002) findings were due to a Pertechnegas-like aerosol
and not Technegas (which shows < 3% lung clearance by 24-h post inhalation). Thus, the
dosimetric pathways by which inhaled particles may rapidly exert acute cardiovascular or other
systemic effects remain to be delineated.
6.3.3 Interspecies Patterns of Clearance
The inability to study the retention of certain materials in humans for direct risk assessment
requires use of laboratory animals. Because dosimetry depends on clearance rates and routes,
adequate toxicological assessment necessitates that clearance kinetics in such animals be related
to those in humans. The basic mechanisms and overall patterns of clearance from the respiratory
tract are similar in humans and most other mammals. However, regional clearance rates can
show substantial variation between species, even for similar particles deposited under
comparable exposure conditions, as extensively reviewed elsewhere (U.S. Environmental
Protection Agency, 1996a; Schlesinger et al., 1997; Snipes et al., 1989).
In general, there are species-dependent rate constants for various clearance pathways.
Differences in regional and total clearance rates between some species are a reflection of
differences in mechanical clearance processes. For example, the relative proportion of particles
cleared from the A region in the short- and longer-term phases differs between laboratory
rodents and larger mammals, with a greater percentage cleared in the faster phase in rodents.
A recent study (Oberdorster et al., 1997) showed inter-strain differences in mice and rats in the
handling of particles by alveolar macrophages. Macrophages of B6C3F1 mice could not
phagocytize 10-|im particles, but those of C57 black/61 mice could. In addition, the
nonphagocytized 10-|im particles were efficiently eliminated from the alveolar region; whereas
previous work in rats found that these large particles were retained persistently after uptake by
macrophages (Snipes and Clem, 1981; Oberdorster et al., 1992). The ultimate implication of
interspecies differences in clearance that need to be considered in assessing particle dosimetry is
6-56
-------
that the retention of deposited particles can differ between species and may result in differences
in response to similar PM exposure atmospheres.
Hsieh and Yu (1998) summarized the existing data on pulmonary clearance of inhaled,
poorly soluble particles in the rat, mouse, guinea pig, dog, monkey, and human. Clearance at
different initial lung burdens, ranging from 0.001 to 10 mg particles/g lung, was analyzed using
a two-phase exponential decay function. Two clearance phases in the alveolar region, namely
fast and slow, were associated with mechanical clearance along two pathways, the former with
the mucociliary system and the latter with the lymph nodes. Rats and mice were fast clearers in
comparison to the other species. Increasing the initial lung burden resulted in an increasing mass
fraction of particles cleared by the slower phase. As lung burden increased beyond 1 mg
particles/g lung, the fraction cleared by the slow phase increased to almost 100% for all species.
However, the rate for the fast phase was similar in all species and did not change with increasing
lung burden of particles; whereas the rate for the slow phase decreased with increasing lung
burden. At elevated burdens, the effect on clearance rate was greater in rats than in humans, an
observation consistent with previous findings (Snipes, 1989).
6.3.4 Factors Modulating Clearance
A number of factors have previously been assessed in terms of modulation of normal
clearance patterns, including age, gender, workload, disease, and irritant inhalation. Such factors
have been discussed in detail previously (U.S. Environmental Protection Agency, 1996a).
6.3.4.1 Age
Studies described in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a)
indicated that there appeared to be no clear evidence for any age-related differences in clearance
from the lung or total respiratory tract, either from child to adult, or young adult to elderly.
Studies of mucociliary function have shown either no changes or some slowing in mucus
clearance function with age after maturity, but at a rate that would be unlikely to significantly
affect overall clearance kinetics.
6-57
-------
6.3.4.2 Gender
Previously reviewed studies (U.S. Environmental Protection Agency, 1996a) indicated no
gender-related differences in nasal mucociliary clearance rates in children (Passali and Bianchini
Ciampoli, 1985) nor in tracheal transport rates in adults (Yeates et al., 1975).
6.3.4.3 Physical Activity
The effect of increased physical activity on mucociliary clearance is unresolved:
previously discussed studies (U.S. Environmental Protection Agency, 1996a) indicate either no
effect or an increased clearance rate with exercise. There are only limited data concerning
changes in A region clearance with increased activity levels (Sweeney et al., 1990). Breathing
with an increased tidal volume was noted to increase the rate of particle clearance from the
A region, and this was suggested to result from distension-related evacuation of surfactant into
proximal airways leading to a facilitated movement of particle-laden macrophages or uningested
particles because of the accelerated motion of the alveolar fluid film (John et al., 1994).
6.3.4.4 Respiratory Tract Disease
Various respiratory tract diseases are associated with clearance alterations. Evaluation of
clearance in individuals with lung disease requires careful interpretation of results because
differences in deposition of particles used to assess clearance function may occur between
normal individuals and those with disease; this would directly affect the measured clearance
rates, especially in the tracheobronchial tree. Studies reported in the 1996 PM AQCD (U.S.
Environmental Protection Agency, 1996a) noted findings of (a) slower nasal mucociliary
clearance in humans with chronic sinusitis, bronchiectasis, rhinitis, or cystic fibrosis and
(b) slowed bronchial mucus transport associated with bronchial carcinoma, chronic bronchitis,
asthma, and various acute respiratory infections. However, a study by Svartengren et al. (1996a)
concluded, based on deposition and clearance patterns, that particles cleared equally effectively
from the small ciliated airways of healthy humans and those with mild to moderate asthma; but,
this similarity was attributed to effective asthma therapy.
In another study, Svartengren et al. (1996b) examined clearance from the TB region in
adults with chronic bronchitis who inhaled 6-|im Teflon particles. Based on calculations,
particle deposition was assumed to be in small ciliated airways at low flow and in larger airways
6-58
-------
at higher flow. The results were compared to those obtained in healthy subjects from other
studies. At low flow, a larger fraction of particles was retained over 72 h in people with chronic
bronchitis compared to healthy subjects, indicating that clearance resulting from spontaneous
cough could not fully compensate for impaired mucociliary transport in small airways. For
larger airways, patients with chronic bronchitis cleared a larger fraction of the deposited particles
over 72 h than did healthy subjects, but this was reportedly because of differences in deposition
resulting from airway obstruction.
Cough is an important mechanism of clearance from the TB region, under some
circumstances. Although cough can be a reaction to an inhaled stimulus, in most individuals
with respiratory infections and disease, spontaneous coughing also serves to clear the upper
bronchial airways by dislodging mucus from the airway surface. Recent studies confirm that this
mechanism likely plays a significant role in clearance for people with mucus hypersecretion, at
least for the upper bronchial tree, and for a wide range of deposited particle sizes (0.5 to 5 jim)
(Toms et al., 1997; Groth et al., 1997). There appears to be a general trend towards an
association between the extent (i.e., number) of spontaneous coughs and the rate of particle
clearance; faster clearance is associated with a greater number of coughs (Groth et al., 1997).
Thus, recent evidence continues to support cough as an adjunct to mucociliary movement in the
removal of particles from the lungs of individuals with COPD. However, some recent evidence
suggests that, like mucociliary function, cough-induced clearance may become depressed with
worsening airway disease. Noone et al. (1999) found that the efficacy of clearance via cough in
patients with primary ciliary dyskinesia (who rely on coughing for clearance because of
immotile cilia) correlated with lung function in that decreased cough clearance was associated
with decreased percentage of predicted FEVj.
Earlier studies (U.S. Environmental Protection Agency, 1996a) indicated that rates of
A region particle clearance were reduced in humans with COPD and in laboratory animals with
viral infections, whereas the viability and functional activity of macrophages were impaired in
human asthmatics and in animals with viral-induced lung infections. However, any modification
of the functional properties of macrophages appears to be injury-specific in that they reflect the
nature and anatomic pattern of disease.
One factor that may affect clearance of particles is the integrity of the epithelial surface
lining of the lungs. Damage or injury to the epithelium may result from disease or from the
6-59
-------
inhalation of chemical irritants or cigarette smoke. Earlier studies performed with particle
instillation showed that alveolar epithelial damage in mice at the time of deposition resulted in
increased translocation of inert carbon to pulmonary interstitial macrophages (Adamson and
Hedgecock, 1995). A similar response was observed in a more recent assessment (Adamson and
Prieditis, 1998), in which silica (< 0.3 jim) was instilled into a lung having alveolar epithelial
damage (as evidenced by increased permeability) and particles were noted to reach the
interstitium and lymph nodes.
6.3.4.5 Inhaled Irritants
Inhaled irritants of various kinds can affect clearance functions in both humans and
laboratory animals (Wolff, 1986; Schlesinger, 1990). As previously reviewed in the 1996 PM
AQCD, single exposures to certain materials may increase or decrease the overall rate of TB
clearance, often depending on the irritant exposure concentration, with alterations in clearance
rates generally being transient (i.e., lasting < 24 hrs). Repeated exposures, however, may result
in increased intra-individual variability in clearance rates and persistently slowed clearance.
Alveolar region clearance can also be altered by acute and chronic exposures to inhaled irritants,
as noted in the 1996 PM AQCD, including acceleration or slowing of clearance depending on the
specific irritant inhaled and/or exposure duration. Of particular interest are studies noted in the
1996 PM AQCD which (a) show increased numbers of macrophages recovered by
bronchoalveolar lavage in smoke-exposed humans and animals and (b) retardation of particle
clearance from alveolar regions of the lung in cigarette smokers, in part, due to impaired alveolar
macrophage-mediated clearance.
6.4 PARTICLE OVERLOAD
Experimental studies using some laboratory rodents have employed high exposure
concentrations of relatively nontoxic, poorly soluble particles. These particle loads interfered
with normal clearance mechanisms and produced clearance rates different from those that would
occur at lower exposure levels. Prolonged exposure to high particle concentrations is associated
with a phenomenon that has been termed particle "overload," defined as the overwhelming of
macrophage-mediated clearance by the deposition of particles at a rate that exceeds the capacity
6-60
-------
of that clearance pathway. It has been suggested that volumetric overloading will begin when
particle retention approaches 1 mg particles/g lung tissue (Morrow, 1988) and that, in the rat,
overload is more dependent upon total particle surface area (Tran et al., 2000). The importance
of surface area to inflammation and the tumorogenic response is detailed in an analysis
performed by Driscoll (1995). He observed a positive tumor response associated with
pulmonary inflammation and epithelial cell proliferation in the rat. Moreover, there was a
significant relationship between lung particle dose, expressed as particle surface area/lung, and
the lung tumor response. There was a positive correlation between the surface area
characteristics of various chemically distinct particulate materials and tumorogenic activity.
Overload is a nonspecific effect noted in experimental studies using many different kinds of
poorly soluble particles and results in A region clearance slowing or stasis, with an associated
chronic inflammation and aggregation of macrophages in the lungs and increased translocation
of particles into the interstitium.
The relevance of lung overload to humans exposed to poorly soluble, nonfibrous particles
remains unclear. Although it is likely to be of little relevance for most "real world" ambient
exposures, it may be of concern in interpreting some long-term experimental exposure data and,
perhaps, also for occupational exposures. For example, it has been suggested that a condition
called progressive massive fibrosis, which is unique to humans, has features indicating that dust
overload is a factor in its pathogenesis (Green, 2000). This condition is associated with
cumulative dust exposure and impaired clearance and can occur following high exposure
concentrations associated with occupational situations. In addition, any relevance to humans is
clouded by the suggestion that macrophage-mediated clearance is normally slower, and perhaps
of less relative importance in overall clearance, in humans than in rats (Morrow, 1994) and that
there can be significant differences in macrophage loading between species. On the other hand,
overload may be a factor in individuals with compromised lungs even under normal exposure
conditions. Thus, it has been hypothesized (Miller et al., 1995) that localized overload of
particle clearance mechanisms in people with compromised lung status may occur whereby
clearance is overwhelmed and results in morbidity or mortality from particle exposure.
6-61
-------
6.5 COMPARISON OF DEPOSITION AND CLEARANCE PATTERNS
OF PARTICLES ADMINISTERED BY INHALATION AND
INTRATRACHEAL INSTILLATION
The most relevant exposure route by which to evaluate the toxicity of particulate matter is
inhalation. However, many toxicological studies deliver particles by intratracheal instillation.
This latter technique has been used because it is easy to perform; requires significantly less
effort, cost, and amount of test material than does inhalation; and can deliver a known, exact
dose of a toxicant to the lungs. It is also an extremely useful technique for mechanistic studies.
Because particle disposition is a determinant of dose, it is important to compare deposition and
clearance of particles delivered by these two routes in order to evaluate the relevance of studies
using instillation. However, in most instillation studies, the effect of this route of administration
on particle deposition and clearance per se was not examined. Although these parameters were
evaluated in some studies, it has been very difficult to compare particle deposition/clearance
between different inhalation and instillation studies because of differences in experimental
procedures and in the manner by which particle deposition/clearance was quantitated. Thus,
while instillation studies are valuable in providing mechanistic insights, inhalation studies are
more appropriate for risk assessment. A recent paper provides a detailed evaluation of the role
of instillation in respiratory tract dosimetry and toxicology studies (Driscoll et al., 2000).
A short summary derived from this paper is provided below in this section.
The pattern of initial regional deposition is strongly influenced by the exposure technique
used. Furthermore, the patterns within specific respiratory tract regions also are influenced in
this regard. Depending on particle size, inhalation results in varying degrees of deposition
within the ET airways, a region that is completely bypassed by instillation. Thus, differences in
amount of particles deposited in the lower airways will occur between the two procedures,
especially for those particles in the coarse mode. This is important if inhaled particles in
ambient air affect the upper respiratory tract and such responses are then involved in the
evaluation of health outcomes.
Exposure technique also influences the intrapulmonary distribution of particles, which
potentially would affect routes and rates of ultimate clearance from the lungs and dose delivered
to specific sites within the respiratory tract or to extrapulmonary organs. Intratracheal
instillation tends to disperse particles fairly evenly within the TB region but can result in
heterogeneous distribution in the A region, whereas inhalation tends to produce a more
6-62
-------
homogeneous distribution throughout the major conducting airways as well as the A region for
the same particles. Thus, inhalation results in a randomized distribution of particles within the
lungs; whereas intratracheal instillation produces an heterogeneous distribution, in that the
periphery of the lung receives little particle load and most of the instilled particles are found in
regions that have a short path length from the major airways. Furthermore, inhalation results in
greater deposition in apical areas of the lungs and less in basal areas; whereas intratracheal
instillation results in less apical than basal deposition. Thus, toxicological effects from instilled
materials may not represent those which would occur following inhalation, due to differences in
sites of initial deposition following exposure. In addition, instillation studies generally deliver
high doses to the lungs, much higher than those which would occur with realistic inhalation
exposure. This would also clearly affect the initial dose delivered to target tissue and its
relevance to ambient exposure.
Comparison of the kinetics of clearance of particles administered by instillation or
inhalation have shown similarities, as well as differences, in rates for different clearance phases
depending on the exposure technique used (Oberdorster et al., 1997). However, some of the
differences in kinetics can be explained by differences in the initial sites of deposition. One of
the major pathways of clearance involves particle uptake and removal via pulmonary
macrophages. Domes and Valberg (1992) noted that inhalation resulted in a lower percentage of
particles recovered in lavaged cells and a more even distribution of particles among
macrophages. More individual cells received measurable amounts of particles via inhalation
than via intratracheal instillation; whereas with the latter, many cells received little or no
particles and others received very high burdens. Furthermore, with intratracheal instillation,
macrophages at the lung periphery contained few, if any, particles; whereas cells in the regions
of highest deposition were overloaded, reflecting the heterogeneity of particle distribution when
particles are administered via instillation. Additionally, both the relative number of particles
phagocytized by macrophages as well as the percentage of these cells involved in phagocytosis
is affected by the burden of administered particles, which is clearly different in instillation and
inhalation (Suarez et al., 2001). Thus, when guinea pigs were administered latex microspheres
(1.52 to 3.97 |im MMAD) by inhalation or instillation, the percentage of cells involved in
phagocytosis, as well as the amount of particles per cell, were both significantly higher with the
6-63
-------
latter route. The route of exposure, therefore, influences particle distribution in the macrophage
population and could, by assumption, influence clearance pathways and clearance kinetics.
In summary, inhalation may result in deposition within the ET region, and the extent of
deposition depends on the size of the particles used. Of course, intratracheal instillation
bypasses this portion of the respiratory tract and delivers particles directly to the
tracheobronchial tree. Although some studies indicate that short (0 to 2 days) and long (100 to
300 days) postexposure phases of clearance of insoluble particles delivered either by inhalation
or intratracheal instillation are similar, other studies indicate that the percentage retention of
particles delivered by instillation is greater than that for inhalation at least up to 30 days
postexposure. Thus, there is some inconsistency in this regard.
Perhaps the most consistent conclusion regarding differences between inhalation and
intratracheal instillation is related to the intrapulmonary distribution of particles. Inhalation
generally results in a fairly homogeneous distribution of particles throughout the lungs. On the
other hand, instillation results in a heterogeneous distribution, especially within the alveolar
region, and focally high concentrations of particles. The bulk of instilled material penetrates
beyond the major tracheobronchial airways, but the lung periphery is often virtually devoid of
particles. This difference is reflected in particle burdens within macrophages, with those from
animals inhaling particles having more homogeneous burdens and those from animals with
instilled particles showing groups of cells with no particles and others with heavy burdens. This
difference affects clearance pathways, dose to cells and tissues, and systemic absorption.
Exposure method, thus, clearly influences dose distribution.
Dosimetric Considerations in Comparing Dosages for Inhalation, Instillation, and
Exposure of Cultured Cells
There are three common experimental approaches for studying the biological effects of
particulate material: inhalation, instillation, and in vitro. Inhalation studies are the more
realistic physiologically, and thus the most applicable to risk assessment. However, because
they are expensive, time consuming, and require specialized equipment and personnel, they must
be supplemented by other techniques. In vitro studies using live cells are cost-effective, provide
for precise dose delivery, and permit investigators who do not have access to inhalation
techniques to perform mechanistic and comparative toxicity studies of parti culate material.
6-64
-------
Commonly, the initial information on likely mechanisms of action of particles is obtained
through in vitro techniques.
Instillation studies, in which particles suspended in a carrier such as physiological saline
are applied to the airways, have certain advantages over in vitro studies. The exposed cells have
normal attachments to basement membranes and adjacent cells; circulatory support; surrounding
cells; and normal endocrine, exocrine and neuronal relationships. Thus, instillation experiments
can bridge between in vitro and inhalation studies as well as produce useful mechanistic and
comparative toxicity information (Benson et al., 1986; Dorries and Valberg, 1992; Henderson
et al., 1995; Kodavanti et al., 2002; Leong et al., 1998; Oberdorster et al., 1997; Osier and
Oberdorster, 1997; Pritchard et al., 1985; Sabaitis et al., 1999; Suarez et al., 2001; Warheit et al.,
1991). Although the tracheobronchial region is most heavily dosed, alveolar regions can also be
exposed via instillation techniques (Kodavanti et al., 2002; Leong et al., 1998; Oberdorster et al.,
1997; Pritchard et al., 1985; Suarez et al., 2001; Warheit et al., 1991).
Selection of the doses of particles used in instillation studies is important because it is easy
to overwhelm normal defense mechanisms; but it is far from an exact process. If the goal is to
expose tracheobronchial tree cell populations to particle concentrations (on a number of particles
per unit surface area basis) that are similar to those occurring with ambient environmental
exposures of humans (or to a known multiple of such exposures), dosimetric calculations must
be performed. Such calculations require selecting characteristics associated with the particles,
the exposed subject, and the environmental exposure scenario. Hence, each study can present a
unique dosimetric analysis. In most cases, it will be useful to know the relationship between the
surface doses in instillation studies and realistic local surface doses that could occur in vivo
among those human subpopulations receiving maximum potential doses derived from
"real-world" ambient air exposures. Although these subpopulations have not been completely
defined (NRC, 2001), some characteristics of individuals do serve to enhance the local PM
surface doses to respiratory tract cells. These characteristics include: exercise and mouth
breathing (ICRP, 1994; NCRP, 1997); nonuniform inhaled air distribution such as occurs in
COPD and chronic bronchitis (Smaldone et al., 1993; Subramaniam et al., 2003; Sweeney et al.,
1995; Segal et al., 2002; Brown et al., 2002; Kim and Kang, 1997); impaired particle clearance
as occurs in some disease states (Pavia, 1987; Pavia et al., 1980; Smaldone, 1993) and location
near pollutant sources (Adgate et al., 2002; Zhu et al., 2002). In addition, even normal subjects
6-65
-------
exposed by inhalation are expected to have numerous sites of high local particle deposition
(specifically at bifurcations) within the tracheobronchial tree (Balashazy et al., 1999; Oldham
et al., 2000; Kaye et al., 2000).
It is difficult to provide precise estimates of dose. However, by considering the several
factors discussed above that enhance local surface doses, order of magnitude estimates can be
made. As an example, consider the scenario of a physically active nose breather with chronic
lung disease that lives near a PM source. The increase in minute ventilation during exercise, due
to an increase in breaths per minute and in tidal volume, results in an increase in the number of
particles inhaled per unit time. Even light exertion can double the minute ventilation, and heavy
exertion can produce a six-fold increase (Phalen et al., 1985). Exercise can also cause a shift
from nasal to oral breathing which bypasses the filtering efficiency of the nose (ICRP, 1994;
NCRP, 1997), leading to increased exposure of the TB and A regions in a particle size-
dependent fashion. As particle aerodynamic diameter increases from 1 to 10 jim, nasal region
deposition at rest increases from 17 to 71% (NCRP, 1997). It is reasonable to assume that oral
breathing can lead to a doubling of TB and A deposition of thoracic coarse particles (PM10_25) in
many individuals (see Figure 6-13). In disease states that produce uneven distribution of inhaled
air, available measurements and models indicate that an enhancement factor of 2 to 5 is realistic
for surface doses (Bennett et al., 1997b; Brown et al., 2002; Kim and Kang, 1997; Miller et al.,
1995; Segal et al., 2002).
The airflow patterns at airway bifurcations lead to high surface deposition doses of inhaled
particles in the TB region. An enhanced deposition of particles (for all sizes that have been
examined) is seen at bifurcations in the TB tree (Bell and Friedlander, 1973; Schlesinger et al.,
1982; Kim and Iglesias, 1989; Kim et al., 1994; Kim and Fisher, 1999; Balashazy et al., 1999;
Kaye et al., 2000; Oldham et al., 2000). The dose enhancement factor is dependent on both
inhaled particle diameter and size of the deposition region. In experimental studies using a
bifurcating airway structure, a majority of deposition (-90% of total) was found within a short
distance (one airway diameter) from the carina and the local deposition pattern was further
intensified when a local obstruction was imposed on the airway structure (Kim et al., 1994).
Using the computational fluid dynamic modeling in a physiologically realistic (human TB tree)
three-dimensional group of bifurcations, Balashazy et al. (1999) provided numerical
enhancement (over average airway surface deposition doses) factors. For the smallest region
6-66
-------
considered, which would comprise less than a few hundred epithelial cells, the enhancement
factors ranged from 52-fold for 0.01 jim diameter particles up to 113-fold for 10 jim diameter
particles. An enhancement factor of 81-fold was calculated for 1 jim diameter particles. The
local deposition enhancement factor could reach higher if factors such as increased ventilation,
oral vs. nasal breathing, and lung disease were considered. However, one must be cautious in
applying the enhancement factor, in that such a high enhancement factor is based on a very small
spot (i.e., a surface area of ~1 mm2). Thus, for the purposes of simulating the exposure of the
heavily dosed TB bifurcation cells to PM10 and/or PM25, an enhancement factor of as much as
80-fold could be reasonable. Taken together, the combined dose enhancing effects of increased
ventilation (2-fold), oral breathing (2-fold), lung disease (2-fold) and bifurcation effects
(80-fold), one could expect populations of epithelial cells to experience enhanced deposition
(over average surface deposition) of as much as 640-fold. Under these conditions, the average
deposition also increases. Considering that clearance impairment may also be a factor in
subpopulations with some disease states, the buildup of particles at such TB bifurcations could
further increase the maximum dose in relation to healthy individuals.
As a final consideration in this susceptibility scenario, the proximity of exposure to sources
of PM may be important. Although data are sparse in this regard, Zhu et al. (2002) have
measured time-averaged concentrations of black carbon and particle number at various distances
downwind from freeways in Los Angeles. In comparison to upwind concentrations,
concentrations at 30 m downwind were about 4-fold higher for black carbon, and about 3-fold
higher for particle number. A factor of 3 for increased dose over the average might be expected
for this subpopulation. By taking all of the above factors into account, it is reasonable to expect
that highly localized PM doses to groups of cells in potentially susceptible subpopulations could
be 3,000 to 4,000 times greater than the average TB surface exposures for the general
population. Other scenarios could be evaluated that lead to greater, or to lesser, local dose
estimates.
In conclusion, well-conducted instillation studies are valuable for examining the relative
toxicity of particulate materials and for providing mechanistic information that is useful for
interpreting in vitro and inhalation studies. However, because mechanisms of injury may vary
with the delivered dose, published instillation studies designed to provide information relevant to
6-67
-------
human risk assessment should report dosimetric calculations, taking into account the types of
dosimetric considerations just discussed.
6.6 MODELING THE DEPOSITION AND DISPOSITION OF
PARTICLES IN THE RESPIRATORY TRACT
6.6.1 Modeling Deposition, Clearance, and Retention
Over the years, mathematical models for predicting deposition, clearance and, ultimately,
retention of particles in the respiratory tract have been developed. Such models help interpret
experimental data and can be used to make dosimetry predictions for cases where data are not
available. In fact, model predictions described below are estimates based on the best available
models at the time of publication and, except where noted, have not been verified by
experimental data.
A review of various mathematical deposition models was given by Morrow and Yu (1993)
and in U.S. Environmental Protection Agency (1996a). There are three major elements involved
in mathematical modeling. First, a structural model of the airways must be specified in
mathematical terms. Second, deposition efficiency in each airway must be derived for each of
the various deposition mechanisms. Finally, a computational procedure must be developed to
account for the transport and deposition of the particles in the airways. As noted earlier, most
models are deterministic in that particle deposition probabilities are calculated using anatomical
and airflow information on an airway generation by airway generation basis. Other models are
stochastic, whereby modeling is performed using individual particle trajectories and finite
element simulations of airflow.
Recent reports involve modeling the deposition of ultrafine particles and deposition at
airway bifurcations. Zhang and Martonen (1997) used a mathematical model to simulate
diffusion deposition of ultrafine particles in the human upper tracheobronchial tree and
compared the results to those in a hollow cast obtained by Cohen et al. (1990). The model
results were in good agreement with experimental data. Zhang et al. (1997) studied the inertial
deposition of particles in symmetric three-dimensional models of airway bifurcations,
mathematically examining effects of geometry and flow. They developed equations for use in
predicting deposition based on Stokes numbers, Reynolds numbers (a dimensionless number that
6-68
-------
describes the tendency for a flowing fluid to change from laminar flow to turbulent flow), and
bifurcation angles for specific inflows.
Models for deposition, clearance, and dosimetry of the respiratory tract of humans have
been available for the past four decades. For example, the International Commission on
Radiological Protection (ICRP) has recommended three different mathematical models during
this time period (International Commission on Radiological Protection, 1960, 1979, 1994).
These models make it possible to calculate the mass deposition and retention in different parts of
the respiratory tract and provide, if needed, mathematical descriptions of the translocation of
portions of the deposited material to other organs and tissues beyond the respiratory tract.
A somewhat simplified variation of the 1994 ICRP dosimetry model was used by Snipes et al.
(1997) to predict average particle deposition in the ET, T, and A regions and retention patterns in
the A region under a repeated exposure situation for two characterized environmental aerosols
obtained from Philadelphia, PA and Phoenix, AZ. Both of these aerosols contained both fine
and coarse particles. They found similar retention for the fine particles in both aerosols, but
significantly different retention for the coarse-mode particles. Because the latter type dominated
the aerosol in the Phoenix sample, this type of evaluation can be used to improve the
understanding of the relationship between exposures to ambient PM and retention patterns that
affect health endpoints in residents of areas where the particle distributions and particle
chemistry may differ.
A morphological model based on laboratory data from planar gamma camera and single-
photon emission tomography images has been developed (Martonen et al., 2000). This model
defines the parenchymal wall in mathematical terms, divides the lung into distinct left and right
components, derives a set of branching angles from experimental measurements, and confines
the branching network within the left and right components (so there is no overlapping of
airways). The authors conclude that this more physiologically realistic model can be used to
calculate PM deposition patterns for risk assessment.
Musante and Martonen (2000c) developed an age-dependent theoretical model to predict
dosimetry in the lungs of children. The model includes the dimensions of individual airways and
the geometry of branching airway networks within developing lungs and breathing parameters as
a function of age. The model suggests that particle size, age, and activity level markedly affect
deposition patterns of inhaled particles. Simulations thus far predict a lung deposition fraction of
6-69
-------
38% in an adult and 73% (nearly twice as high) in a 7-mo-old for 2-|im particles inhaled during
heavy breathing. The authors conclude that this model will be useful for estimating dose
delivered to sensitive subpopulations such as children.
Martonen et al. (2001a) developed a three-dimensional (3D) physiologically realistic
computer model of the human upper-respiratory tract (URT). The URT morphological model
consists of the extrathoracic region (nasal, oral, pharyngeal, and laryngeal passages) and upper
airways (trachea and main bronchi) of the lung. The computer representation evolved from a
silicone rubber impression of a medical school teaching model of the human head and throat.
The final unified 3D computer model may have significant applications in inhalation toxicology
for evaluating lung injuries from PM inhalation.
Segal et al. (2000a) developed a computer model for airflow and particle motion in the
lungs of children to study how airway disease, specifically cancer, affects inhaled PM
deposition. The model considers how tumor characteristics (size and location) and ventilatory
parameters (breathing rates and tidal volumes) influence particle trajectories and deposition
patterns. The findings indicate that PM may be deposited on the upstream surfaces of tumors
because of enhanced efficiency of inertial impaction. Additionally, submicron particles and
larger particles may be deposited on the downstream surfaces of tumors because of enhanced
efficiency of diffusion and sedimentation, respectively. The mechanisms of diffusion and
sedimentation are functions of the particle residence times in airways. Eddies downstream of
tumors would trap particles and allow more time for deposition to occur by diffusion and
sedimentation. The authors conclude that particle deposition is complicated by the presence of
airway disease and that the effects are systematic and predictable.
Segal et al. (2000b) have used a traditional mathematical model based on Weibel's lung
morphology and calculated total lung deposition fraction of 1- to 5-|im diameter particles in
healthy adults. Airway dimensions were scaled by individual lung volume. Deposition
predictions were made with both plug flow and parabolic flow profiles in the airways. The
individualized airway dimensions improved the accuracy of the predicted values when compared
with experimental data. There were significant differences, however, between the model
predictions and experimental data depending on the flow profiles used, indicating that use of
more realistic parameters is essential to improving the accuracy of model predictions.
6-70
-------
Broday and Georgopoulos (2001) presented a model that solves a variant of the general
dynamic equation for size evolution of respirable particles within human tracheobronchial
airways. The model considers poly disperse aerosols of variable thermodynamic states and
chemical compositions. The aerosols have an initial bimodal lognormal size distribution that
evolves with time in response to condensation-evaporation and deposition processes.
Simulations reveal that submicron size particles grow rapidly and cause increased number and
mass fractions of the particle population to be found in the intermediate size range. Because
deposition by diffusion decreases with increasing size, hygroscopic ultrafme particles may
persist longer in the inspired air than nonhygroscopic particles of comparable initial size
distribution. In contrast, the enhanced deposition probability of hygroscopic particles initially
from the intermediate size range increases their fraction deposited in the airways. The model
demonstrates that the combined effect of growth and deposition tends to decrease the
nonuniformity of the persistent aerosol, forming an aerosol which is characterized by a size
distribution of smaller variance. These factors also alter the deposition profile along airways.
Lazaridis et al. (2001) developed a deposition model for humans that was designed to
better describe the dynamics of respirable particles within the airways. The model took into
account alterations in aerosol particle size and mass distribution that may result from processes
such as nucleation, condensation, coagulation, and gas-phase chemical reactions. The airway
geometry used was the regular dichotomous model of Weibel, and it incorporated the influences
of airway boundary layers on particle dynamics although simplified velocity profiles were used
so as to maintain a fairly uncomplicated description of respiratory physiology. Thus, this model
was considered to be an improvement over previous models which did not consider either the
effects of boundary layers on both the airborne and deposited particles or the effects of gas-phase
transport processes because it can account for the polydispersity, multimodality, and
heterogeneous composition of common ambient aerosols. The authors indicate that the model
predictions were both qualitatively and quantitatively consistent with experimental data for
particle deposition within the TB and A regions.
Another respiratory tract dosimetry model was developed concurrently with the ICRP
model by the National Council on Radiation Protection and Measurements (NCRP, 1997).
As with the ICRP model, the NCRP model addresses inhalability of particles, revised subregions
of the respiratory tract, dissolution-absorption as an important aspect of the model, body size,
6-71
-------
and age. The NCRP model defines the respiratory tract in terms of a
naso-oro-pharyngo-laryngeal (NOPL) region, which is equivalent to the ICRP (1994) model's
ET region, a tracheobronchial (TB) region, a pulmonary (P) region (equivalent to the ICRP
model AI region), and lung-associated lymph nodes (LN). Deposition and clearance are
calculated separately for each of these regions. As with the 1994 ICRP model, inhalability of
aerosol particles was considered, and deposition in the various regions of the respiratory tract
was modeled using methods that relate to mechanisms of inertial impaction, sedimentation, and
diffusion.
Fractional deposition in the NOPL region was developed from empirical relationships
between particle diameter and air flow rate. Deposition in the TB and P regions were based on
geometric or aerodynamic particle diameter (where appropriate) and physical deposition
mechanisms such as impaction, sedimentation, diffusion, and interception. Deposition in the TB
and P regions used the lung model of Yeh and Schum (1980) with a method of calculation
similar to that of Findeisen (1935) and Landahl (1950). This method was modified to
accomodate an adjustment of lung volume and substitution of realistic deposition equations.
These calculations were based on air flow information and idealized morphometry and used a
typical pathway model. Comparison of regional deposition fraction predictions between the
NCRP and ICRP models was provided in U.S. Environmental Protection Agency (1996a). The
definition of inhalability was that of the American Conference of Governmental Industrial
Hygenists (1985). Breathing frequency, tidal volume, and functional residual capacity were the
ventilatory factors used to model deposition. These were related to body weight and to three
levels of physical activity (low activity, light exertion, and heavy exertion).
Clearance from all regions of the respiratory tract was considered to result from
competitive mechanical and absorptive mechanisms. Mechanical clearance in the NOPL and TB
regions was considered to result from mucociliary transport. This was represented in the model
as a series of escalators moving towards the glottis and where each airway had an effective
clearance velocity. Clearance from the P region was represented by fractional daily clearance
rates to the TB region, the pulmonary LN region, and the blood. A fundamental assumption in
the model was that the rates for absorption into blood were the same in all regions of the
respiratory tract. The rates of dissolution-absorption of particles and their constituents were
derived from clearance data primarily from laboratory animals. The effect of body growth on
6-72
-------
particle deposition also was considered in the model, but particle clearance rates were assumed
to be independent of age. Some consideration for compromised individuals was incorporated
into the model by altering normal rates for the NOPL and TB regions.
Mathematical deposition models for a number of nonhuman species have been developed;
these were discussed in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a).
Despite difficulties, modeling studies in laboratory animals remain a useful step in extrapolating
exposure-dose-response relationships from laboratory animals to humans.
Respiratory tract clearance begins immediately upon deposition of inhaled particles. Given
sufficient time, the deposited particles may be removed completely by these clearance processes.
However, single inhalation exposures may be the exception rather than the rule. It generally is
accepted that repeated or chronic exposures are common for environmental aerosols. As a result
of such exposures, accumulation of particles may occur. Chronic exposures produce respiratory
tract burdens of inhaled particles that continue to increase with time until the rate of deposition is
balanced by the rate of clearance. This is defined as the "equilibrium respiratory tract burden."
It is important to evaluate these accumulation patterns, especially when assessing ambient
chronic exposures, because they dictate what the equilibrium respiratory tract burdens of inhaled
particles will be for a specified exposure atmosphere. Equivalent concentrations can be defined
as "species-dependent concentrations of airborne particles which, when chronically inhaled,
produce equal lung deposits of inhaled particles per gram of lung during a specified exposure
period" (Schlesinger et al., 1997). Other metrics are also possible, i.e., mass of PM, surface area
of particles, or number of particles deposited per cm2 of airway surface or per alveolus.
Available data and approaches with which to evaluate exposure atmospheres that produce
similar respiratory tract burdens in laboratory animals and humans were discussed in detail in the
1996 PM AQCD. Some examples are given in 6.6.4.
Several laboratory animal models have been developed to help interpret results from
specific studies that involved chronic inhalation exposures to nonradioactive particles (Wolff
et al., 1987; Strom et al., 1988; Stober et al., 1994). These models were adapted to data from
studies involving high level chronic inhalation exposures in which massive lung burdens of low
toxicity, poorly soluble particles were accumulated. Koch and Stober (2001) further adapted
clearance models for more relevant particle deposition in the pulmonary region. They published
a pulmonary retention model that accounts for dissolution and macrophage-mediated removal of
6-73
-------
deposited polydisperse aerosol particles. The model provides a mathematical solution for the
size distribution of particles in the surfactant layer of the alveolar surface and in the cell plasma
of alveolar macrophages and accounts for the different kinetics and biological effects in the two
compartments. It does not, however, account for particle penetration to the lung interstitium and
particle clearance by the lymph system.
Estimating regional particle deposition patterns is important for establishing the
comparability of animal models, for understanding interspecies differences in the expression of
chemical toxicities, and, ultimately, for the human risk assessment process. Different species
exposed to the same particle atmosphere may not receive identical initial doses in comparable
respiratory tract regions, and the selection of a certain species for toxicological evaluation of
inhaled particles may, thus, influence the estimated human lung or systemic dose, as well as its
relationship to potential adverse health effects. Asgharian et al. (1995) described a strategy for
summarizing published data on regional deposition of particles of different diameters and
calculating a deposited fraction for a specific particle size distribution. The authors constructed
nomograms for three species, namely the human, monkey, and rat, to allow estimation of
alveolar deposition fractions. They then developed a regression model to permit the calculation
of more exact deposition fractions. While this paper describes the procedure for one region of
the lungs, the authors maintain that the technique can be applied to other regions of the
respiratory tract or to the total system for which deposition data are available. The model is
somewhat constrained at present due to the limitations of the underlying deposition database.
Tran et al. (1999) used a mathematical model of clearance and retention in the A region of
rats lungs to determine the extent to which a sequence of clearance mechanisms and pathways
could explain experimental data obtained from inhalation studies using relatively insoluble
particles. These pathways were phagocytosis by macrophages with subsequent clearance,
transfer of particles into the interstitium and to lymph nodes, and overloading of defense
mechanisms. The model contained a description of the complete defense system in this region
using both clearance and transfer processes as represented by sets of equations. The authors
suggested that the model could be used to examine the consistency of various hypotheses
concerning the fate of inhaled particles and could be used for species other than the rat with
appropriate scaling.
6-74
-------
Hofmann et al. (2000) used three different morphometric models of the rat lung to compute
particle deposition in the acinar (alveolar) airways: the multipath lung (MPL) model with a
fixed airway geometry; the stochastic lung (SL) model with a randomly selected branching
structure; and a hybrid of the MPL and SL models. They calculated total and regional deposition
for a range of particle sizes during quiet and heavy breathing. Although the total bronchial and
acinar deposition fractions were similar for the three models, the SL and the hybrid models
predicted a substantial variation in particle deposition among different acini. Acinar deposition
variances in the MPL model were consistently smaller than in the SL and the hybrid lung
models. The authors conclude that the similarity of acinar deposition variations in the latter two
models and their independence of the breathing pattern suggest that the heterogeneity of the
acinar airway structure is primarily responsible for the heterogeneity of acinar particle
deposition.
The combination of MPL and SL models developed for the human lung takes into
consideration both intra- and inter-human variability in airway structure. The models also have
been developed to approximately the same level of complexity for laboratory animals and,
therefore, can be readily used for interspecies extrapolation (Asgharian et al., 1999). A variation
of these models will soon be developed for inclusion of the airway geometry of children.
By incorporating particle clearance in the TB region (Asgharian et al., 2001) and in the alveolar
region (Koch and Stober, 2001), this suite of models should prove to be very useful in better
predicting PM dosimetry in humans.
6.6.2 Models To Estimate Retained Dose
Models have been used routinely to express retained dose in terms of temporal patterns for
A region retention of acutely inhaled materials. Available information for a variety of
mammalian species, including humans, can be used to predict deposition patterns in the
respiratory tract for inhalable aerosols with reasonable degrees of accuracy. Additionally,
alveolar clearance data for non-human mammalian species commonly used in inhalation studies
are available from numerous experiments that involved inhaled radioactive particles.
An important factor in using models to predict retention patterns in laboratory animals or
humans is the dissolution-absorption rate of the inhaled material. Factors that affect the
dissolution of materials or the leaching of their constituents in physiological fluids and the
6-75
-------
subsequent absorption of these constituents are not fully understood. Solubility is known to be
influenced by the surface-to-volume ratio and other surface properties of particles (Mercer,
1967; Morrow, 1973). The rates at which dissolution and absorption processes occur are
influenced by factors that include the chemical composition of the material. Temperature history
of materials is also an important consideration for some metal oxides. For example, in
controlled laboratory environments, the solubility of oxides usually decreases when the oxides
are produced at high temperatures, which generally results in compact particles having small
surface-to-volume ratios. It is sometimes possible to accurately predict dissolution-absorption
characteristics of materials based on physical/chemical considerations, but predictions for in
vivo dissolution-absorption rates for most materials, especially if they contain multivalent
cations or anions, should be confirmed experimentally.
Phagocytic cells, primarily macrophages, clearly play a role in dissolution-absorption of
particles retained in the respiratory tract (Kreyling, 1992). Some particles dissolve within the
phagosomes because of the acidic milieu in those organelles (Lundborg et al., 1984, 1985), but
the dissolved material may remain associated with the phagosomes or other organelles in the
macrophage rather than diffuse out of the macrophage to be absorbed and transported elsewhere
(Cuddihy, 1984). This same phenomenon has been reported for organic materials. For example,
covalent binding of benzo[a]pyrene or metabolites to cellular macromolecules resulted in an
increased alveolar retention time for that compound after inhalation exposures of rats (Medinsky
and Kampcik, 1985). Understanding these phenomena and recognizing species similarities and
differences are important for evaluating alveolar retention and clearance processes and for
interpreting the results of inhalation studies.
Dissolution-absorption of materials in the respiratory tract is clearly dependent on the
chemical and physical attributes of the material. Although it is possible to predict rates of
dissolution-absorption, it is prudent to determine this important clearance parameter
experimentally. It is important to understand the effect of this clearance process for the lungs,
TB lymph nodes, and other body organs that might receive particles or their constituents that
enter the circulatory system from the lung.
Additional research must be done to provide the information needed to evaluate properly
retention of particles in conducting airways. However, a number of earlier studies, discussed in
the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) and in Section 6.3.2.2
6-76
-------
herein, noted that some particles were retained for relatively long times in the tracheobronchial
regions, effectively contradicting the general conclusion that almost all inhaled particles that
deposit in the TB region clear within hours or days. These studies have demonstrated that
variable portions of the particles that deposit in, or are cleared through, the TB region are
retained with half-times on the order of weeks or months. Long-term retention and clearance
patterns for particles that deposit in the ET and TB regions must continue to be thoroughly
evaluated because of the implications of this information for respiratory tract dosimetry and risk
assessment.
Model projections are possible for the A region using the cumulative information in the
scientific literature relevant to deposition, retention, and clearance of inhaled particles.
Clearance parameters for six laboratory animal species were summarized by the U.S.
Environmental Protection Agency (1996a). Nikula et al. (1997) evaluated results in rats and
monkeys exposed to high levels of either diesel soot or coal dust. Although the total amount of
retained material was similar in both species, the rats retained a greater portion in the lumens of
the alveolar ducts and alveoli than did monkeys; whereas the monkeys retained a greater portion
of the material in the interstitium. The investigators concluded that intrapulmonary retention
patterns in one species may not be predictive of those in another species at high levels of
exposure, but this may not be the case at lower levels of exposure.
The influence of exposure concentration on the pattern of particle retention in rats (exposed
to diesel soot) and humans (exposed to coal dust) was examined by Nikula et al. (2000) using
histological lung sections obtained from both species. The exposure concentrations for diesel
soot were 0.35, 3.5, or 7.0 |ig/m3; and exposure duration was 7 h/day, 5 days/week for 24 mo.
The human lung sections were obtained from nonsmoking nonminers, nonsmoking coal miners
exposed to levels < 2 mg dust/m3 for 3 to 20 years, or nonsmoking miners exposed to 2 to
10 mg/m3 for 33 to 50 years. In both species, the amount of retained material (using
morphometric techniques based on the volume density of deposition) increased with increasing
dose (which is related to exposure duration and concentration). In rats, the diesel exhaust
particles were found to be primarily in the lumens of the alveolar duct and alveoli; whereas in
humans, retained dust was found primarily in the interstitial tissue within the respiratory acini.
Dosimetric models may be used to adjust for differences in the exposure-dose relationship
in different species, thus allowing for comparison of lung responses at different doses. In a
6-77
-------
series of papers (Kuempel 2000, 200 la; Kuempel 200 Ib), Kuempel presents a biologically-
based human dosimetric lung model to describe the fate of respirable particles in the lungs of
humans. The model uses data from coal miners and assumptions about the overloading of
alveolar clearance from studies in rats. The form of the model that provides the best fit to the
lung-dust burden data in the coal miners includes a first-order interstitialization process and
either a no-dose-dependent decline in alveolar clearance or a much lower decline than expected
from the rodent studies. These findings were consistent with particle retention patterns observed
previously in the lungs of primates.
6.6.3 Fluid Dynamics Models for Deposition Calculations
The available models developed to simulate PM deposition in the lung are based on
simplifying assumptions about the morphometry of the lung and the fluid dynamics of inspired
air through a branching airway system. As new models are developed, they will better predict
particle deposition patterns in a more realistic airway geometry under realistic flow conditions
that can result in local nonuniformity of particle deposition and the formation of hot spots. One
example is the model of ventilation distribution in the human lung developed by Chang and Yu
(1999). This model was designed as an improvement over those that assumed uniform
ventilation in the lungs because it better simulated the effect of airway dynamics on the
distribution of ventilation under different conditions which may occur in the various lobes of the
lungs and under various inspiratory flow rates. The authors indicated that the results of the
model compared favorably with experimental data and that the model will be incorporated into a
particle deposition model which will allow for the evaluation of the nonuniformity of deposition
within the lungs resulting from the physiological situation of nonuniform distribution of
ventilation. Computational fluid dynamics (CFD) modeling adds another step to better model
development by providing increased ability to predict local airflow and particle deposition
patterns and provides a better representation of ET deposition in the human respiratory tract.
The CFD models developed to date, however, are limited in scope because they are unable to
simulate flow in the more complex gas exchange regions. Due to a lack of more realistic
simulations for the lower airways, they impose another "idealized" boundary condition at the
distal end of the human respiratory tract.
6-78
-------
Airflow patterns within the lung are determined by the interplay of structural and
ventilatory conditions. These flow patterns govern the deposition kinetics of entrained particles
in the inspired air. A number of CFD software programs are available to simulate airflow
patterns in the lung by numerically solving the Navier-Stokes equations (White, 1974). The
CFD modeling requires a computer reconstruction of the appropriate lung region and the
application of boundary conditions. The flow field resulting from the CFD modeling is
represented by velocity vectors in the grid points of a two- or three-dimensional mesh.
Numerical models of particle deposition patterns are computed by simulating the trajectories of
particles introduced into these flow streams after solving for the particles' equation of motion.
Such CFD models have been developed for different regions of the respiratory tract, including
the nasal cavity (Yu et al., 1998; Sarangapani and Wexler, 2000); larynx (Martonen et al. 1993;
Katz et al., 1997; Katz, 2001); major airway bifurcations (Gradoii and Orlicki, 1990; Balashazy
and Hofmann, 1993a,b, 1995, 2001; Heistracher and Hofmann, 1995; Lee et al., 1996; Zhang
et al., 1997, 2000, 2001, 2002; Comer et al., 2000, 2001a,b); and alveoli (Tsuda et al., 1994a,b;
Darquenne, 2001).
Kimbell (2001) has recently reviewed the literature on CFD models of the upper
respiratory tract (URT). Most of these models have focused on characterizing the airflow
patterns in the URT and have not included simulation of particulate dosimetry. Keyhani et al.
(1995) were the first to use computer-aided tomography (CAT) scans of the human nasal cavity
to construct an anatomically accurate three-dimensional airflow model of the human nose.
Subramaniam et al. (1998) used data from magnetic resonance imaging to extend these CFD
studies to include the nasopharynx. However, neither of these studies investigated particle
deposition in the upper respiratory tract.
Yu et al. (1998) developed a three-dimensional CFD model of the entire human URT,
including the nasal airway, oral airway, laryngeal airway, and the first two generations of the TB
airways. They have used this CFD model to investigate the effect of breathing pattern, i.e., nasal
breathing, oral breathing, and simultaneous nasal and oral breathing, on airflow and ultrafine
particle deposition. They concluded that the ultrafine particle deposition simulated using the
CFD model was in reasonable agreement with the corresponding experimental measurements.
In a study led by Sarangapani and Wexler (2000), an upper respiratory tract CFD model that
included the nasal cavity, nasopharynx, pharynx, and larynx was developed to study the
6-79
-------
deposition efficiency of hygroscopic and non-hygroscopic particles in this region. They used the
CFD model to simulate the temperature and water vapor conditions in the upper airways and
predicted high relative humidity conditions in this region. They also simulated particle
trajectories for 0.5 |im, 1 |im, and 5 jim particles under physiologically realistic flow rates. The
predictions of the CFD model indicated that high relative humidity conditions contribute to rapid
growth of hygroscopic particles and would dramatically alter the deposition characteristics of
ambient hygroscopic aerosols.
Stapleton et al. (2000) investigated deposition of a poly disperse aerosol (MMD = 4.8 jim
and GSD = 1.65) in a replica of a human mouth and throat using both experimental results and
three-dimensional CFD simulation. They found that CFD results were comparable with
experimental results for a laminar flow case, but were more than 200% greater for a turbulent
flow case. The results suggest that accurate predictions of particle deposition in a complex
airway geometry require a careful evaluation of geometric and fluid dynamic factors in
developing CFD models.
Due to the complex structural features and physiological conditions of the human laryngeal
region, only a limited number of modeling studies have been conducted to evaluate laryngeal
fluid dynamics and particle deposition. A high degree of inter-subject variability, a compliant
wall that presents challenges in setting appropriate boundary conditions, and a complex turbulent
flow field are some of the difficulties encountered in developing CFD models of the laryngeal
airways. Martonen et al. (1993) investigated laryngeal airflow using a two-dimensional CFD
model and concluded that laryngeal morphology exerts a pronounced influence on regional flow,
as well as fluid motion in the trachea and the main bronchi. In this study, the glottal aperture
(defined by the geometry of the vocal folds) was allowed to change in a prescribed manner with
the volume of inspiratory flow (Martonen and Lowe, 1983), and three flow rates corresponding
to different human activity were examined.
In a subsequent CFD analysis, a three-dimensional model of the larynx based on
measurements of human replica laryngeal casts (Martonen and Lowe, 1983; Katz and Martonen,
1996; Katz et al., 1997) simulated the flow field in the larynx and trachea under steady
inspiratory flow conditions at three flow rates. They observed that the complex geometry
produces jets, recirculation zones, and circumferential flow that may directly influence particle
deposition at select sites within the larynx and tracheobronchial airways. The primary
6-80
-------
characteristics of the simulated flow field were a central jet penetrating into the trachea created
by the ventricular and vocal folds, a recirculating zone downstream of the vocal folds, and a
circumferential secondary flow. Recently, a computational model for fluid dynamics and
particle motion for inspiratory flow through the human larynx and trachea has been described
(Katz, 2001). This model calculates the trajectory of single particles introduced at the entrance
to the larynx using a stochastic model for turbulent fluctuations incorporated into the particles'
equation of motion and time-averaged flow fields in the larynx and trachea. The effects of flow
rate and initial particle location on overall deposition were presented in the form of probability
density histograms of final particle deposition sites. At present, however, there are no
experimental data to validate results of such modeling.
A number of CFD models have been developed to study fluid flow and particle deposition
patterns in airway bifurcations. The bifurcation geometries that have been modeled include
two-dimensional (Li and Ahmadi, 1995); idealized three-dimensional using circular airways
(Kinsara et al., 1993) or square channels (Asgharian and Anjilvel, 1994); symmetric bifurcations
(Balashazy and Hofmann, 1993a,b); or physiologically realistic asymmetric single (Balashazy
and Hofmann, 1995; Heistracher and Hofmann, 1995) and multiple bifurcation models (Lee
et al., 1996; Heistracher and Hofmann, 1997; Comer et al., 2000, 2001a,b; Zhang et al., 2000,
2001, 2002) with anatomical irregularities such as cartilaginous rings (Martonen et al., 1994a)
and carinal ridge (Martonen et al., 1994b; Comer et al., 2001a) shapes incorporated. The CFD
flow simulations in the bifurcating geometry models show distinct asymmetry in the axial
(primary) and radial (secondary) velocity profile in the daughter and parent airway during
inspiration and expiration, respectively. In a systematic investigation of flow patterns in airway
bifurcations, numerical simulations were performed to study primary flow (Martonen et al.,
2001b), secondary currents (Martonen et al., 2001c), and localized flow conditions (Martonen
et al., 2001d) for different initial flow rates. The effects of inlet conditions, Reynolds numbers,
ratio of airway diameters, and branching angles with respect to intensity of primary flow, vortex
patterns of the secondary currents, and reverse flow in the parent-daughter transition region were
investigated. These simulated flow patterns match experimentally observed flow profiles in
airway bifurcations (Schroter and Sudlow, 1969).
Gradoii and Orlicki (1990) computed the local deposition flux of submicron size particles
in a three-dimensional bifurcation model for both inhalation and exhalation, and they found
6-81
-------
enhanced deposition in the carinal ridge region during inspiration and in the central zone of the
parent airway during expiration. Numerical models of particle deposition in symmetric three-
dimensional bifurcations were developed by Balashazy and Hofmann (1993a,b), and these were
subsequently extended to incorporate effects of asymmetry in airway branching (Balashazy and
Hofmann, 1995) and physiologically realistic shapes of the bifurcation transition zone and the
carinal ridge (Heistracher and Hofmann, 1995; Balashazy and Hofmann, 2001). In these
numerical models, three-dimensional airflow patterns were computed by finite difference or
finite volume methods; and the trajectories of particles entrained in the airstream were simulated
using Monte Carlo techniques considering the simultaneous effects of gravitational settling,
inertial impaction, diffusion, and interception. The spatial deposition pattern of inhaled particles
was examined for a range of particle sizes (0.01 to 10 jim) and flow rates (16 to 32 L/min) by
determining the intersection of particle trajectories with the surrounding surfaces. The overall
deposition rates derived using the CFD models correspond reasonably well with experimental
data (Kim and Iglesias, 1989). These simulations predict deposition hot spots at the inner side of
the daughter airway downstream of the carinal ridge during inspiration, corresponding to the
secondary fluid motion of the inhaled air stream. During exhalation, the CFD models predict
enhanced deposition at the top and bottom parts of the parent airway, consistent with secondary
motion in the exhaled air stream. These studies indicate that secondary flow patterns within the
bifurcating geometry play a dominant role in determining highly nonuniform local particle
deposition patterns.
Zhang et al. (1997) numerically simulated particle deposition in three-dimensional
bifurcating airways (having varying bifurcation angles) due to inertial impaction during
inspiration for a wide range of Reynolds numbers (100 to 1000). Inlet velocity profile, flow
Reynolds number, and bifurcation angle had substantial effects on particle deposition efficiency.
Based on the simulated results, equations were derived for particle deposition efficiency as a
function of nondimensional parameters (such as Stokes number and Reynolds number) and
bifurcation angle and were shown to compare favorably with available experimental results.
More recently, Comer et al. (2000) have estimated the deposition efficiency of 3-, 5-, and 7-|im
particles in a three-dimensional double bifurcating airway model for both in-plane and out-of-
plane configurations for a wide range of Reynolds numbers (500-2000). They demonstrated
6-82
-------
deposition in the first bifurcation to be higher than in the second bifurcation, with deposition
mostly concentrated near the carinal region. The nonuniform flow generated by the first
bifurcation had a dramatic effect on the deposition pattern in the second bifurcation. Based on
these results, they concluded that use of single bifurcation models is inadequate to capture the
complex fluid-particle interactions that occur in multigeneration airway systems.
Comer et al. (2001a) further investigated detailed characteristics of the axial and secondary
flow in a double bifurcation airway model using 3-D CFD simulation. Effects of carina shape
(sharp vs. rounded) and bifurcation plane (planar vs. non-planar) were examined. Particle
trajectories and deposition patterns were subsequently investigated in the same airway model
(Comer et al., 2001b). There was a highly localized deposition at and near the carina both in the
first and second bifurcation, and deposition efficiency was much lower in the second bifurcation
than in the first bifurcation as demonstrated in the earlier study (Comer et al., 2000). They found
that deposition patterns were not much different between the sharp versus rounded carina shape
at Stokes numbers of 0.04 and 0.12. However, deposition patterns were altered significantly for
these particles when the bifurcation plane was rotated, suggesting that a careful consideration of
realistic airway morphology is important for accurate prediction of particle deposition by CFD
modeling.
Zhang et al. (2000, 2001) extended the studies of Comer et al. described above and
investigated effects of angled inlet tube as well as asymmetric flow distribution between
daughter branches. The flow asymmetry caused uneven deposition between downstream
daughter branches. Also noted was that the absolute deposition amount was higher, but
deposition efficiency per se was lower in the high flow branch than in the low flow branch.
The intriguing relationship between flow asymmetry and deposition was in fact consistent with
experimental data of Kim and Fisher (1999), indicating that the CFD model could correctly
simulate complicated airflow and particle dynamics that may occur in the respiratory airways.
Most CFD models use constant inspiratory or expiratory flows for simplicity and practical
reasons. However, the respiratory airflow is cyclic, and such flow characteristics cannot be fully
described by constant flows. Recent studies of Zhang et al. (2002) investigated particle
deposition in a triple bifurcation airway model under cyclic flow conditions mimicking resting
and light activity breathing. Deposition dose was obtained for every mm2 area. They found that
deposition patterns were similar to those obtained with constant flows. However, deposition
6-83
-------
efficiencies were greater with the cyclic flows than constant flows, and the difference could be
as high as 50% for mean Stokes numbers between 0.02 and 0.12 during normal breathing. The
CFD results are qualitatively comparable to experimental data (Kim and Garcia, 1991) that
showed about 25% increase in deposition with cyclic flows. With further improvement of
airway morphology and computational scheme, CFD modeling could be a valuable tool for
exploring the microdosimetry in the airway structure.
Current CFD models of the acinar region are limited due to the complex and dynamic
nature of the gas exchange region. Flow simulation in a linearly increasing volume of a
spherical truncated two-dimensional alveolus model show distinct velocity maxima in the
alveolar ducts close to the entrance and exit points of the alveolus and a radial velocity profile in
the interior space of the alveolus (Tsuda et al., 1996). This is in contrast to simulations based on
a rigid alveolus (Tsuda, 1994a,b) and suggests that a realistic simulation of the flow pattern in
the acinar region should involve application of time-dependent methods with moving boundary
conditions. Nonuniform deposition patterns of micron size particles, with higher deposition near
the alveolar entrance ring, have been predicted using numerical models (Tsuda, 1994a,b, 1996).
Studies by Darquenne (2001) examined aerosol transport and deposition in 6-generation
alveolated ducts using two-dimensional computer simulation. Particle trajectories and
deposition patterns were obtained for one complete breathing cycle (2 s inspiration and
2 s expiration). There were large nonuniformities in deposition between generations, between
ducts of a given generation, and within each alveolated duct, suggesting that local deposition
dose can be much greater than the mean acinar dose.
6.6.4 Modeling Results Obtained with Models Available to the Public
Two relatively user-friendly computer models for calculating fractional deposition in
various compartments of the respiratory tract as a function of particle size are publicly available.
Several model runs have been done to demonstrate the outputs of the models. Published results
from one model are also presented. Model simulations were performed for particles of density
of 1 g/cm3 so aerodynamic, Stokes, and thermodynamic diameters are the same.
6-84
-------
6.6.4.1 International Commission on Radiological Protection (ICRP)
The LUDEP (Lung Dose Evaluation Program; National Radiologic Protection Board,
1994) model was developed concurrently with the ICRP (International Commission on
Radiological Protection, 1994) respiratory tract model mainly to help the ICRP Task Group
examine the model in detail by testing the predictions of deposition, clearance, and retention of
inhaled radionuclides against experimental data and by determining the model's implications for
doses to the respiratory tract (ICRP, 1994; NRPB, 1994). This model was designed to predict
the deposition of inhaled particles in the respiratory tract, the subsequent biokinetic behavior of
inhaled radionuclides, and the doses delivered to the respiratory tract. Although created for
calculating the internal dose of radionuclides, the model is useful for determining the deposition
of nonradioactive materials. In particular, the model has wide applicability for calculating the
regional deposition of particles in the respiratory tract based on particle size, body size (age),
breathing rate, activity patterns, and exposure environment. The overall dosimetric model for
the respiratory tract consists of several critical elements important for deposition and clearance
calculations including detailed descriptions of morphometry and respiratory physiology. The
morphometric element of the model describes the structure of the respiratory tract and its
dimensions. A description of respiratory physiology provides the rates and volumes of inhaled
and exhaled air which affects the amount of material deposited in the respiratory tract. The
ICRP model covers the particle size range from 0.001 to 100 jim. However, deposition in the
size region between 0.001 and 0.01 jam may not be correct due to neglect of axial diffusion of
particles and deposition of particles > 25 [im may be uncertain so only the 0.01 to 25 [im range
will be shown.
The ICRP model (see Figure 6-1) calculates deposition in five compartments:
- ET1 - the extrathoracic region comprising the anterior nose;
- ET2 - the extrathoracic region comprising the posterior nasal passages, larynx, pharynx
and mouth;
- BB - the bronchial region;
- bb - the bronchiolar region consisting of bronchioles and terminal bronchioles; and
- Al - the alveolar-interstitial region consisting of the respiratory bronchioles, the alveolar
ducts with their alveoli and the interstitial connective tissue.
6-85
-------
Two simulations were run to demonstrate some aspects of deposition as predicted by the
ICRP model. Respiratory parameters for a worker with a moderately high activity level (ICRP
default) and a young adult with a lower activity level are given in Table 6-3. Each simulation
was run for nasal breathing and mouth breathing. In the presentation of model results, ET1 and
ET2 are combined to give an ET (extrathoracic) region, BP and bb are combined to give a TB
(tracheobronchial) region, and Al gives the A (alveolar) region. Results are shown in
Figures 6-13 to 6-15. Figure 6-13 shows the total and regional deposition as a function of
particle size for the worker: nasal breathing (13a), mouth breathing (13b), and a comparison of
nasal and mouth breathing for the TB and A regions (13c). Figure 6-14 gives similar results for
the young adult. For both simulations, the deposition is a minimum between 0.1 and 1 |im
diameter (the accumulation mode size range) and increases for larger (coarse mode) and smaller
(ultrafme particle) size ranges. For ultrafine particles, A deposition peaks between 0.01 and
0.1 |im and TB deposition increases as particle size decreases below 0.1 jim.
TABLE 6-3. RESPIRATORY PARAMETERS USED IN LUDEP MODEL
Activity Related Physiological
Parameters
Activity Percent
Adult Male (ICRP default values)
Sleep
Sitting
Light Exercise
Heavy Exercise
0
50
38
12
Ventilation Rate
(m3/hr)
0.45
0.54
1.5
3
Frequency
(breaths/min)
12
12
20
26
Tidal Volume
(mL)
625
750
1250
1923
Young Adult
Sitting
100
0.45
15
500
6-86
-------
55 8°
O 60 "
~V) AC) -
0 40
Q.
Q9n -
n -
"••.
\
•.
*m
•" "'. •
• uf •
• •
• •
\.
x<^
^•<*,
.... TOT(NB
ET(NB)
* * * TB (NB)
... A(NB)
'. .'
•- v /y
'*'**' jf
^sr.:
)
•
•
/
^
.•*"••*
•* ^^
/ /^
: /
: i
7
f
..-..
A*A*Si^
a
•
\
\
^
•s..
0.01 0.1 1 2.5 5 10
Particle Diameter, pm
25
0-
0.01 0.1 1 2.5 5 10 25
Particle Diameter, \im
tn
o
Q.
0
o
0
TB (MB)
A (MB)
TB (NB)
A(NB)
0.01 0.1 1 2.5 5 10
Particle Diameter, pm
Figure 6-13. Deposition fraction for total results of LUDEP model for an adult male
worker (ICRP default breathing parameters as shown in Table 6-3) showing
total percent deposition in the respiratory tract (TOT) and in the ET, TB,
and A regions: (a) nasal breathing (NB), (b) mouth breathing (MB),
(c) comparison of nasal and mouth breathing for TB and A regions.
6-87
-------
100
80
c
o
100
80
D
TOT(NB)
ET(NB)
TB (NB)
A(NB)
0
0.01 0.1 1 2.5 5 10 25
Particle Diameter, pm
• ••• TOT (MB)
ET(MB)
•» A A TB (MB)
• •• A (MB)
0.01 0.1 1 2.5 5 10 25
Particle Diameter,
o
o
Q.
0)
Q
60
50
40
30
20
10
C
TB (MB)
A (MB)
TB (NB)
A(NB)
0.01 0.1 1 2.5 5 10 25
Particle Diameter, pm
Figure 6-14. Deposition fraction for total results of LUDEP model for a young adult
showing total percent deposition in the respiratory tract (TOT) and in the
ET, TB, and A regions: (a) nasal breathing (NB), (b) mouth breathing (MB),
(c) comparison of nasal and mouth breathing for TB and A regions.
Respiratory parameters given in Table 6-3.
-------
tfi
o
Q.
to
O
Q.
60
50
40
30
20
10
a
* *
10
0.01
TB(YA)
A(YA)
TB(WK)
'A(WK)
***«Ł. ^-"^
0.01 0.1 1 2.5 5 10 25
Diameter, |jm
TB(YA)
A(YA)
TB(WK)
A(WK)
0.1 1 2.5 5 10 25
Diameter, |jm
Figure 6-15. Comparison of deposition fraction in the TB and A regions for a worker
(WK; light exercise, ICRP default) and a young adult (YA; resting):
(a) nasal breathing and (b) mouth breathing.
The comparisons of nasal and mouth breathing in Figures 6-13c and 6-14c show almost no
difference in deposition for particles between 0.01 and 1 jim. Below 0.1 jim, more particles are
removed by diffusion in the extrathoracic (ET) region while above 1.0 more particles are
removed by impaction in the ET region. Due to increased deposition of particles in the head
during nasal breathing, switching to breathing through the mouth leads to greater TB and
A deposition of coarse mode particles (AED > 1 jim) and of the smaller ultrafine particles
6-89
-------
(dp < 0.01 jirn). The A deposition approaches zero as particle size increases to 10 jim. However,
TB deposition continues for larger particle sizes.
The TB and A deposition patterns of the worker under moderate activity and the young
adult under low activity are compared in Figure 6-15a and b. Increased activity in the worker
compared to the resting young adult results in lower A deposition of coarse particles for nasal
breathing (Figure 6-15a) and lower A and TB deposition of coarse particles for mouth breathing
(Figure 6-15b). It also shifts the maximum deposition for coarse particles to smaller sizes.
Increased activity increases A deposition of ultrafme particles and shifts the maximum
deposition to larger sizes. Increased activity also increases the A deposition of
accumulation-mode particles.
6.6.4.2 Multiple Path Particle Dosimetry Model (MPPD)
The MPPD model, developed by the CUT Centers for Health Research (formerly the
Chemical Industry Institute of Toxicology, USA) with support from the Dutch National Institute
of Public Health and the Environment, is described in a RIVM report (Winter-Sorkina and
Cassee, 2002). The MPPD model allows calculation of PM deposition and retention for humans
and rats. The model includes age-specific human lung models, but some parameters are
expected to be modified in a newer version of the model; so, age-specific results will not be
discussed here. The MPPD model covers the particle size range from 0.01 to 20 jim. With the
MPPD model, parameters such as the particle size distribution, inhalability, particle density, and
respiratory pause can be modeled and dose per airway surface area can be calculated. The model
may be used to improve understanding of the exposure-dose-response relationships observed in
environmental epidemiological studies and for extrapolation of studies in experimental animals
to humans. Factors resulting in increased susceptibility can also be studied.
The RIVM report (Winter-Sorkina and Cassee, 2002) describes the results of monodisperse
aerosol deposition calculations with the MPPD model and its sensitivity to various parameters.
The deposition fraction of inhaled PM depends primarily on physical characteristics of the
particles, lung morphometry, and breathing parameters, and is difficult to measure for the many
possible variations in parameters. Therefore, computer models such as the MPPD model have
proven to be important tools to analyze PM dosimetry. Dosimetric models, such as those
discussed above, use an explicit set of equations which describe real-life processes, based on
6-90
-------
theory or empirical data, and may be used to analyze effects of scenarios such as particulate
exposure control strategies. The age of the subject, the functional capacity of the lungs, and
breathing parameters, as well as the individual lung morphometry, are factors that significantly
affect particle deposition and can explain variations in susceptibility of subpopulations due to
differences in dose per unit exposure. First, the predictions of the MPPD model will be
compared to those of the LUDEP model. Next, results from the MPPD model depicting
deposition as a function of minute ventilation (a surrogate for exertion or exercise level) for
various respiratory tract regions will be shown.
6.6.4.2.1 Comparisons ofL UDEP and MPPD models
Predicted regional deposition patterns, calculated using the ICRP (LUDEP) model and the
MPPD (Yeh-Schum 5-lobe) model, for two breathing patterns (Table 6-4) are shown in Figures
6-16 and 6-17. The total deposition patterns are similar: low for the accumulation mode size
range (0.1 to 1.0 jim), with increasing deposition at smaller and larger sizes. However, the
MPPD model shows somewhat greater total deposition in the accumulation size range. The ET
deposition is also similar except for slightly higher depositions in the ultrafine (UF) region for
the MPPD model. In the A region, the UF deposition predicted by the MPPD model is lower for
particle sizes below about 0.05 |_im for resting and 0.03 |_im for exercising. In the coarse particle
size range, the A deposition peaks predicted by the MPPD model are shifted to larger particle
sizes. While the absolute values of the deposition fraction are low in the accumulation mode
size range for both models, the relative deposition is much higher for the MPPD model in this
size region where most of the fine particle mass is found. The TB differences become more
pronounced for higher exertion. The MPPD model predicts that TB deposition will decrease
during exercise relative to rest whereas the LUDEP model predicts a large increase in TB
deposition.
Both models are for nonhygroscopic particles. Hygroscopic particles, mostly found in the
fine particle size range (although sea salt may be present as larger size particles), will grow in
the high relative humidity of the respiratory system. As shown in Figures 6-16 and 6-17,
deposition fractions decrease with increasing particle size above about 0.01 to 0.02 jim until
around 0.3 to 0.5 jim. This suggests that as hygroscopic particles in these size ranges grow their
deposition fraction will decrease. Particles in the accumulation mode, however, will grow into a
6-91
-------
TABLE 6-4. BREATHING PATTERNS FOR COMPARISON OF ICRP AND
MPPD MODELS
Breathing Parameters
Activity
Resting
Light Exercise
Minute Ventilation
L/m
7.5
25
Breathing Frequency
min"1
12
20
Tidal Volume
mL
625
1250
size range with a larger deposition fraction. The amount of growth will depend on the specific
hygroscopic components and the fraction of the particle that is hygroscopic. The effects of
hygroscopicity on dosimetry, therefore, will depend on the size distribution and composition of
the particles and will be difficult to model. However, studies of the effects of hygroscopicity on
dosimetry have been reported. (Martonen, 1982; Martonen et al., 1985; Martonen et al., 1989;
Schroeter et al., 2001; Broday and Georgopoulos, 2001). The implications of hygroscopic
growth on deposition have been reviewed extensively by Morrow (1986) and Hiller (1991),
whereas the difficulties of studying lung deposition of hygroscopic aerosols have been reviewed
by Kim (2000). See Chapter 2 for a detailed description of particle hygroscopicity.
Figures 6-16 and 6-17 compare deposition fractions as a function of particle size for
monodisperse particles. In practice, most exposures are to polydisperse size distributions.
The comparison might be somewhat different for size distributions with a given MMD as
compared to monodisperse particles. Several examples are shown in Table 6-5 based on the
three modes in the urban average size distribution reported by Whitby (1978) and a distribution
with MMD = 2 |im and og = 2 as representative of resuspended particles used in laboratory
studies. The fractional deposition values are given for the specified size distributions and for
monodisperse particles with the same MMD as well as for the ratio of the two model predictions.
As can be seen, the differences between the two models are reduced in some cases and increased
in others.
6-92
-------
Nose
Breathing
Open Symbols MPPD yodel
Closed Symbols ICRP Model
and ^ Total
O and @ ET
and | TB
and J,, A
Breathing
.
10 0.01
0.01
0.1 1
Diameter, pro
10 0.01
Diameter, pm
Figure 6-16. Comparison of regional deposition results from the ICRP (LUDEP) and the MPPD models for a
resting breathing pattern: (a) and (b), nose breathing; (c) and (d), mouth breathing.
-------
Nose Breathing
Mouth Breathing
Open Symbols MPPD Model
Closed Symbols ICRP Model
and ^ Total
O and • ET
Q and • TB
A and A A
0.01
0.1 1
Diameter, |jm
Diameter,
Figure 6-17. Comparison of regional deposition results from the ICRP (LUDEP) and the MPPD models for a light
exercise breathing pattern: (a) and (b), nose breathing; (b) and (c), mouth breathing.
-------
TABLE 6-5. RATIO OF MPPD TO ICRP DEPOSITION FRACTION FOR SEVERAL SIZE DISTRIBUTIONS
Nose Breathing
MMD
5.7
Total
ET
TB
A
MMD
2
Total
ET
Oi
vb TB
A
MMD
0.31
Total
ET
TB
A
MMD
0.031
Total
ET
TB
A
ICRP
0.86
0.796
0.027
0.037
0.757
0.635
0.033
0.089
0.211
0.095
0.022
0.095
0.598
0.087
0.106
0.405
MPPD
Og = 2.15
0.905
0.844
0.018
0.044
Og = 2.00
0.741
0.631
0.033
0.077
Og = 2.03
0.264
0.112
0.055
0.097
0R=1.7
0.625
0.133
0.135
0.357
Ratio
(M/iy
1.05
1.06
0.67
1.2
0.98
0.99
1.01
0.87
1.25
1.18
2.5
1.03
1.05
1.53
1.27
0.88
I, Light Exercise
Ratio
(Mfl)a
1.04
1.04
0.57
1.87
0.95
0.98
0.84
0.84
1.53
1.55
3.16
1.12
1.05
1.49
1.26
0.92
ICRP
Og=1.00
0.95
0.901
0.032
0.017
0R=1.00
0.866
0.706
0.041
0.119
0R=1.00
0.129
0.042
0.017
0.069
0R=1.00
0.605
0.082
0.101
0.422
MPPD
0.99
0.94
0.018
0.032
0.825
0.692
0.034
0.099
0.197
0.065
0.054
0.078
0.637
0.122
0.128
0.387
MMD
5.7
Total
ET
TB
A
MMD
2
Total
ET
TB
A
MMD
0.31
Total
ET
TB
A
MMD
0.031
Total
ET
TB
A
Mouth Breathing, Light
ICRP
0.717
0.349
0.208
0.16
0.445
0.099
0.125
0.221
0.135
0.01
0.024
0.101
0.577
0.043
0.111
0.423
MPPD
Og = 2.15
0.719
0.413
0.143
0.163
Og = 2
0.371
0.09
0.08
0.2
Og = 2.03
0.176
0.01
0.06
0.103
0R=1.7
0.594
0.06
0.146
0.384
Ratio
(M/iy
i
1.18
0.69
1.02
0.83
0.88
0.67
0.91
1.3
1.35
2.5
1.02
1.03
1.49
1.32
0.91
Exercise
Ratio
(Mfl)a
0.9
0.96
0.49
1.57
0.74
0.67
0.76
0.75
1.54
1.48
3.14
1.15
1.04
1.46
1.3
0.94
ICRP
Og=1.00
0.883
0.351
0.345
0.187
0R=1.00
0.401
0.045
0.091
0.265
0R=1.00
0.096
0.008
0.018
0.071
0R=1.00
0.581
0.04
0.105
0.437
MPPD
0.797
0.336
0.168
0.293
0.297
0.03
0.069
0.198
0.148
0.012
0.055
0.081
0.606
0.059
0.136
0.411
"MPPD model prediction/ICPP model prediction.
-------
6.6.4.2.2 Deposition as a function of physical exertion
Prior studies show that PM deposition depends on the level of physical exertion. Taking
into account exertion levels and activity patterns is necessary in order to estimate the actual
exposure of a whole population.
Winter-Sorkina and Cassee (2002) used the MPPD with the five-lobe lung model of Yeh
and Schum (1980) to calculate aerosol deposition in the human adult at different levels of
physical exertion. Results are quoted directly from Winter-Sorkina and Cassee (2002). Note:
Table and figure numbers were changed in the quotation that follows in order to fit sequence in
this chapter:
Levels of physical exertion for adults, corresponding representative activities and
corresponding minute ventilation (CARB, 1987) used in the calculation are presented in
Table 6-6. The breathing frequency and tidal volume for different physical exertion levels
(Table 6-6) are calculated from minute ventilation keeping the ratio of breathing frequency
and tidal volume nearly constant. For normal augmenters, the switch to oronasal breathing
(combined nose and mouth breathing) is considered to occur at a minute ventilation of
35.3 L/min. Partitions of airflow between the nose and mouth as given by Niinimaa et al.
(1981) are used for the oronasal breathing. The partitioning flow is assumed to be the same
for inhaled and exhaled air. For minute ventilation lower than this value, breathing is only
through the nose, therefore, the calculations present a discontinuity at this point.
Calculations are performed for monodisperse aerosol particles with 10 different
aerodynamic diameters ranging from 0.01 um to 10 um and with a particle density of
1 g/cm3. The deposited mass rates were calculated for an aerosol concentration of
140 ug/m3.
Results on aerosol deposition as a function of physical exertion for different particle
sizes are shown in Figure 6-18. The head deposition fractions for 1.3 um, 2.5 um and
5 um particles increase from rest to light exercise. They decrease with a factor of
respectively 2.3, 1.8, and 1.5 and further stay about constant when breathing is changed
from nasal to oronasal at modest and heavy exercise with minute ventilation of 40 L/min
and higher. The head deposition fraction of ultrafme particles decreases slightly from rest
to light exercise. Tracheobronchial deposition fractions for ultrafme particles of 0.01 um,
0.02 um, and 0.04 um decrease from rest to light exercise, decrease slightly further to
heavy exercise for 0.01 um particles and stay constant for 0.04 um particles.
Tracheobronchial deposition fraction for coarse particles decreases slightly from rest
to light exercise and rises when breathing is changed from nasal to oronasal. It increases
from modest to heavy exercise especially for 5 um particles. Tracheobronchial deposition
fraction of ultrafine particles is larger than deposition fraction of coarse particles at rest,
light and modest exercise; however, at heavy exercise the deposition fraction of 5 um
particles is larger than that of ultrafine particles. Pulmonary or alveolar deposition fraction
of ultrafine particles increases from rest to light exercise, deposition fraction of coarse
2.5 um and 5 um particles decreases from rest to light exercise, rises when breathing is
changed from nasal to oronasal and decreases slightly from modest to heavy exercise.
Thoracic deposition fraction shows a [s]light increase for 0.01 um and 0.02 um particles
and a decrease for 2.5 um and 5 um particles from rest to light exercise. Deposited
6-96
-------
TABLE 6-6. 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)
Breathing
Frequency, min"1
10
12
16
19
22
24
26
28
34
37
40
44
Tidal Volume,
mL
500
625
813
1,000
1136
1,250
1,346
1,429
1735
1,946
2,125
2273
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
Source: California Air Resources Board (1987).
-------
a. Total Deposition
b. Head Deposition
0 20 40 60 80 100
Minute Ventilation (L/min)
c. Tracheobronchial (TB)Deposition
= 0.3
O
O
'a!
o
Q.
01
Q
0.2
0.0
0 20 40 60 80 100
Minute Ventilation (L/min)
e. Thoracic (TB+A) Deposition
0.6
c
° 0.5
Ł °4
| 0.3
'in
§. 0.2
0.1
0.0
-I—H—+—4-
40
60
80
20
Minute Ventilation (L/min)
100
o.o
0.5 -
U
re
0.3
0.2
O
Q.
d>
0.1
o.o
20
40
60
80
0 20 40 60 80 100
Minute Ventilation (L/min)
d. Pulmonary (A) Deposition
0 20 40 60 80 100
Minute Ventilation (L/min)
f. Thoracic Mass Rate
100
Minute Ventilation (L/min)
Figure 6-18. 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).
6-98
-------
thoracic mass rate increases with increasing physical exertion, faster for heavy exercise.
At light exercise with a minute ventilation of 25 L/min the deposited thoracic mass rate is
13 times larger than at rest awake (7.5 L/min) for 0.01 um particles and 4 times larger for
5 um particles. At modest exercise with minute ventilation of 40 L/min the deposited
thoracic mass rate is 36 times larger than at rest awake (7.5 L/min) for 0.01 um particles
and 44 times larger for 5 um particles.
6.6.4.3 Comparisons of Deposition in Humans and Rats
This section presents some results in which 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 for humans used the 5-lobe lung model (Yeh and Schum, 1980). Respiratory
parameters used in the model runs are shown in Table 6-7. Resting conditions were used for the
rat since rats are usually resting in exposure studies. However, humans are exposed during a
variety of conditions from sleep to heavy exercise. Light exercise, as specified in the ICRP
model, was chosen for the human. The resulting human to rat ratios would differ for different
breathing patterns. 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, and A deposition, respectively. Figure 6-19 also shows the ratios of percent deposition
for human to rat for mouth breathing and nasal breathing humans.
TABLE 6-7. RESPIRATORY PARAMETERS FOR HUMANS AND RATS:
Rat
Human
Breaths
min"1
102
20
Tidal Volume
mL
2.1
1250
FRCb
mL
4
3300
URTb
mL
0.42
50
a Parameters are for light exercise in humans and at rest in rats.
bFRC, functional residual capacity; URT, upper respiratory tract volume.
6-99
-------
ET deposition is shown in Figure 6-19a-l. Deposition of coarse mode particles in the ET
region increases significantly with particle size because of impaction. However, increased
inertia poses a limitation to the ability of particles to enter the ET region. This reduction in the
fraction of the aerosol that is actually inhaled is relevant for particle sizes larger than 3 to 4 jim
for rats and larger than about 8 jim for humans and is more significant for rats than for humans.
The inhalability adjustment (Menache et al., 1995) used in the MPPD model does not change
deposition results for humans significantly; the TB deposition fraction is reduced only 3.5% and
thoracic deposition fraction only 2.5% for 10 jim particles. However, in rats, inhalability
reduces the nasal deposition fraction by about 80% for 2.5 jim particles, 65% for 5 jim particles,
and 44% for 10 jim particles. As a result, TB and pulmonary deposition fractions for large
particles are also reduced by about the same fractions. For particle sizes above about 0.15 jim,
the ET fractional deposition for nose breathing is greater for humans than rats. This leads to a
peak in the human/rat ET deposition ratio for nose breathing at 1 |im (Figure 6-19, Panel a-2).
The ET deposition ratio is lower for mouth breathing up to about 8 jim.
The TB deposition fraction (Figure 6-19b-l) is lower for rats than humans in the
accumulation mode size range. However, between 1.5 and 5 jim, the fractional deposition for
the rat is greater than that for the nasal breathing human. Above about 2.5 jim, the fractional
deposition for the mouth breathing human increases rapidly relative to that of the rat.
For A deposition (Figure 6-19c-l), rats and humans have almost the same fractional deposition
in the accumulation mode size range. However, the fractional deposition for the nasal breathing
human and the rat fall off for particles above about 3 jim, with deposition in the rat decreasing
more rapidly than the human. These differences are borne out in the human/rat ratios which
become very high for particles above 3 jim.
The relationship between dose per given exposure concentration and time for humans and
rats can be understood better if the dose is normalized to a parameter such as lung mass, TB
surface area, or A surface area. Values for these parameters are given in Table 6-8. The volume
and the surface areas of the lung are not fixed but increase with inhalation and decrease with
exhalation. It seems reasonable, therefore, to choose the functional reserve capacity (FRC) at
rest as the appropriate lung size to use in normalizing rat and human deposition to lung surface
areas. Tracheobronchial and A surface areas were estimated from the human morphology given
6-100
-------
a-1.
a-2.
LU
0.01
b-1,
0.01
c-1.
0.01
0.1 1 2.5 5 10
Diameter, |jm
0.1 1 2.5 5 10
Diameter, |jm
0.1
Diameter, |jm
1 2.5 5 10
^ m
1 8
*
DŁ 6
c
E "
I 2
ET(MB)
- - - ET(NB)
0 -I
0.01
.2 3
13 ,
oi ^-
1 1
C '
3
X
0 -,
TB(MB)
TB(NB)
/
;'
;'
f
/
.--'
'.
\
\
\
^/
0.1 1 2.5 5 1(
Diameter, |jm
^r^
0.01
o ,
1
(0 0
c
n
E1
3
I
Q
A (MB)
---A(NB)
,X"\
\
\
I
v/
v-.''
0.1 1 2.5 5 1
Diameter, |jm
v
"V^
0.01
^
1
/ /
'.J
0.1 1 2.5 5 1
Diameter, |jm
Figure 6-19. Comparison of fractional deposition for rats (nasal breathing, at rest) and
humans (nasal and mouth breathing, light exercise) and the ratio of human
to rat for nasal and mouth breathing humans for the (a) ET, (b) TB, and
(c) A regions of the respiratory tract.
6-101
-------
in Yeh and Schum (1980) and the rat morphology given in Yeh et al. (1979 ) and scaling the
lung surface area to the FRC volume.
The fractional deposition values for human and rat, shown in Figure 6-19, can be used with
the parameters shown in Table 6-8 to normalize dose for a given exposure to lung mass, TB
surface area, or A surface area. Such normalization could be applied to the amount of material
deposited or retained in the lung. Retained dose was not considered in these simulations.
However, both TB and A clearance are more rapid in rats than in humans and would need to be
considered in any estimation of retained dose. While the MPPD model can estimate the
clearance of poorly soluble particles, many exposure durations and levels would need to be
presented here in order to illustrate the complexity of clearance on retained dose and therefore on
human to rat dose ratios.
TABLE 6-8. SURFACE AREA VALUES FOR LUNG MASS AND OF
TRACHEOBRONCHIAL AND ALVEOLAR REGIONS FOR HUMANS AND RATS
Human
Lung mass, g 1100
Surface Areas, m2
TB A
Values used in analyses .442* 57. 2 a
Other values .269 c 54°
150.3d
Rat Human/Rat Ratios
4.34 253
TB A TB A
.00235b .300 b 188 191
.55 e
"Based on morphology of Yeh and Schum (1980) scaled to FRC of 3300 cm3.
b Based on morphology of Yeh et al. (1979) scaled to FRC of 4 cm3.
CU.S. EPA (1996a) based on U.S. EPA 1994).
dGehr et al. (1978). (143 m2 alveolar + 7.3 m2 respiratory bronchioles).
eMauderly (1979).
6-102
-------
a-1. Thoracic Region
a-2. Thoracic Region
1 2.5 5 10 0.01
re
E
0.01
0.1 1
Diameter, |jm
2.5 5 10
0.01
12
c-1. Alveolar Region
0.1 1 2.5 5
Diameter, |jm
c-2. Alveolar Region
0)
u
I
(O
u
0.01
0.1 1 2.5 5
Diameter, |jm
0.1 1
Diameter, [im
Figure 6-20. Normalized deposition patterns for rats (nasal breathing) and humans
(nasal breathing [NB] and mouth breathing [MB]) and the ratio of human
to rat. Quantity of PM deposited based on 8-h exposure to 100 ug/m3.
a. Normalized deposition in the thoracic region (in terms of jig PM per g
of lung).
b. Normalized deposition in the TB region C (in terms of jig PM per cm2
of TB surface).
c. Normalized deposition in the A region (in terms of jig PM per m2 of
A surface).
6-103
-------
Figure 6-20a compares deposition of PM by size in humans and rats normalized to lung
mass for thoracic (TB + A) deposition. Thoracic deposition, in terms of jig of PM deposited per
gram of lung, is smaller for humans than rats for particles below about 2.5 jim for mouth
breathing humans and for particles below about 5 jim for nasal breathing humans. As can be
seen in 6-20a-2, the ratio of human to rat deposition, especially for mouth breathing, increases
to very high values for particles above about 2.5 jim.
Normalized results for surface areas are shown in Figure 6-20b,c. For TB deposition in
terms of jig of PM per cm2 of TB surface, shown in Figure 6-20b, the normalized deposition in
the human is greater than that in the rat for accumulation mode particles. However, for particles
between about 1 and 2.5 jim, the normalized deposition in the rat is greater than that in the nasal
breathing human. Again, the ratios increase rapidly, especially for mouth breathing and for
larger particles. The normalized A deposition (Figure 6-20c) for rats is greater than that for
humans from about 0.02 to 1 |im. From about 1 to 5 jim, particle deposition is lower in the rat
than the mouth breathing human but higher than the nose breathing human. Above about 5 jim,
deposition in rats decreases rapidly.
To use the deposited dose ratio plots in Figure 6-20, the following equation applies for
exposure (E) levels in humans (H) and rats (R) that yield equivalent doses for a specific particle
size and a given dose metric when both species are exposed for the same length of time:
EH = ER/X or ER = XEH
where X is the human to rat dose ratio for a specific particle size and given dose metric (Miller,
2000a,b). From the above comparison of rats and humans, it would appear that for inhalation
studies with accumulation mode aerosols, as might be done using concentrated air particles,
equivalent thoracic deposition in rats could be obtained with about 75% of the concentrations for
humans (for particles < 2.5 jim). However, for coarse particles the deposition ratios are very
sensitive to particle size. Thus, for coarse particles resuspended from bulk material particle size
distribution measurements would be needed and very high concentration ratio might be needed
for equivalent deposition on a per gram of lung basis.
The ratio would be changed if different tidal volumes or breathing rates were used.
For example, the ratio would be higher if a breathing pattern for more rigorous exercise were
6-104
-------
used for the humans. The ratio would be lower if a breathing pattern for a human at rest or
sleeping or a rat exercising was used. However, for the breathing patterns used, the human/rat
comparisons, whether normalized by lung mass or surface areas, indicate that for particles
< 2.5 |im normalized human and rat depositions differ by only a factor of about 2. However, for
particles between 2.5 and 5 jim, much higher exposures may be required for rats to obtain
equivalent normalized doses. As shown in Figure 20, panels a-2, b-2, and c-2, for particles
> 5 |im, the inhalation exposure level required to obtain a dose in a rat equivalent to the dose in a
human at a given inhalation concentration becomes almost boundless. Given the poor
inhalability of particles > 5 jim in rats, few inhaled coarse particles will reach the thorax no
matter how high the concentration. Therefore, in order to study the effects of coarse mode
particles, either larger animals (e.g., dog, pig, monkey) should be used or rats should be used
with an endotracheal inhalation system that allows 100% thoracic penetration.
6.7 SUMMARY AND CONCLUSIONS
6.7.1 Particle Dosimetry
Understanding the mechanisms of action and ultimate biological effects of inhaled
particulate matter (PM) requires knowledge of the dosimetry of such material. This is because
the proximal cause of the biological response is due to the dose of particles delivered to and
retained at the target site, rather than the exposure concentration. Deposition, clearance, and
retention comprise the essential elements of dosimetry. Properly characterizing the dosimetry of
inhaled particles is essential for extrapolating effects found in controlled exposure studies of
laboratory animals to those observed in human exposure studies and for relating effects in
healthy individuals to those in potentially susceptible persons.
The understanding of total and regional deposition as a function of particle size has
improved since publication of the 1996 PM AQCD. The extrathoracic (ET) region, especially
the nasal passages, is an efficient filter for small ultrafine (< 0.01 jim) and larger coarse
particles, but filtration is less efficient for larger ultrafine and fine particles. Accordingly,
particles removed in the ET region are not available for deposition in the tracheobronchial (TB)
and alveolar (A) regions of the respiratory tract. Within the thoracic region, the deposition
distribution of ultrafine particles (0.01 to 0.1 jim) is highly skewed towards the proximal airway
6-105
-------
regions and resembles that of coarse particles. Thus, the deposition patterns for ultrafme
particles are similar to those of coarse-mode particles, with significant fractional deposition in all
three regions. Particles in the accumulation mode size range (0.1 to 1.0 jim) have lower
fractional deposition in all three regions.
The dose information expressed by fractional deposition may be applied only to acute
exposure conditions. Retained dose at any given time is determined by the balance between
deposition and clearance. In this regard, a long-term retained dose can be much greater than an
acute exposure dose in individuals with impaired clearance mechanisms.
6.7.2 Host Factors
Certain host factors have a marked effect on particle dosimetry and can affect the fraction
of inhaled particles that are deposited and/or retained.
Gender
There are small but statistically significant gender differences in the homogeneity of
deposition as well as the deposition rate of particles. These differences arise from differences
between males and females in body size, conducting airway size, and ventilatory parameters.
At a fixed breathing pattern (i.e., tidal volume and breathing frequency), females have a greater
deposition of coarse mode particles in the ET and TB regions, and lower deposition in the
A region. This gender effect appears to be particle-size dependent, showing a greater fractional
deposition in females for very small ultrafme and large coarse particles. Specifically, total
fractional lung deposition for 0.04 and 0.06 jim particles is slightly greater in females than males
but only negligibly so for particles in the 0.8 to 1.0 |im size range. As the particle size increases
(3 to 5 |im), total fractional deposition increases more rapidly in females than in males. While
deposition is greater in certain particle size ranges and more localized in females than males at a
fixed breathing condition, the gender difference may not be evident unless breathing patterns are
controlled. In fact, the deposition rate can be greater in males during spontaneous breathing
because of a greater ventilation rate in males compared to women.
6-106
-------
Exercise
Exercise may also increase the potential health risks of inhaled particles because exercise
increases the rate of oxygen consumption and changes ventilatory parameters (airflow rate and
breathing patterns). The switch from nose breathing to mouth breathing, which occurs as
exercise intensity increases, leads to an increase in fractional deposition of coarse particles in the
TB and A regions. The higher breathing rate and larger tidal volume lead to a greater amount of
deposition. Total lung deposition rate may be 3 to 4 times greater during exercise. The more
rapid breathing of children also leads to a greater amount of deposition.
Age
Airway structure and physiological function vary with age and health status of the
respiratory tract. Such variations may alter the deposition patterns for inhaled particles.
Significant age differences have been predicted by mathematical models and observed in
experimental studies. Although reported data are insufficient for making a firm conclusion,
these studies generally indicate that ET and TB deposition is greater in children and that children
receive greater doses of particles per lung surface area than adults. Unfortunately, deposition
studies in another susceptible population, the elderly, are still limited.
Lung Disease
A number of studies have examined particle deposition in chronic lung disease. These
studies indicate that total lung deposition is generally increased with airways obstruction.
Airflow distribution is uneven in obstructive diseases, and deposition can be enhanced locally in
areas of active ventilation.
6.7.3 Laboratory Animal Studies
It is difficult to systematically compare deposition patterns in laboratory animals used in
dosimetric studies. Deposition patterns are generally similar between laboratory animals and
humans, but there are absolute differences in deposition fractions. In most laboratory animal
species, deposition in the ET region is near 100% for particles greater than 5 jim, indicating
greater nasal deposition efficiency than that seen in humans. Clearance processes are similar in
6-107
-------
animals and humans, but the clearance rate for particles is typically faster in small laboratory
animals.
Once particles are deposited on the surface of the airways, they are subsequently cleared
from the respiratory tract or translocated to other sites within the body by distinct regional
processes. Ultrafme 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
epidemiology studies (Chapter 8). However, there is a need for better laboratory animal models
of susceptible human populations.
6.7.4 Mathematical Models
There has been significant improvement in the mathematical and computational fluid
dynamic modeling of particle dosimetry in the respiratory tract. Although the models have
become more sophisticated and adaptable, models inevitably use simplistic lung morphology and
idealistic airflow patterns. Models use a number of assumptions in deposition processes and
employ different computational schemes. As such, model predictions can vary substantially
depending on approaches used. Validation of the models by experimental data is critical as new
experimental data become available.
6.7.5 Key Points
• 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.
• Dosimetric information is critical for 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.
• Based on anatomical features, the respiratory tract may be divided into three regions:
extrathoracic (ET), tracheobronchial (TB), and alveolar (A). Particle deposition and
clearance differ for these regions.
• Particles in the middle of accumulation mode size range (0.3 to 1.0 jim) have the lowest
deposition fraction in the ET and TB regions.
6-108
-------
The fractional deposition of ultrafine particles in the A region peaks between 0.02 and
0.03 |im and is greater than predicted for both accumulation and coarse mode particles.
For coarse particles, fractional deposition peaks between 4 and 6 jim for the TB region
and between 2.5 and 5 jim 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 or translocated to other sites within the system by distinct
regional processes.
Fractional deposition depends on particle size, lung size, tidal volume, and breathing rate.
Exercising subjects receive higher thoracic doses of particles per cm2 of lung surface than
subjects at rest. This occurs mainly due to an increase in tidal volume and minute
ventilation during exercise. Shifting from nasal to oronasal breathing is another factor for
increased thoracic deposition.
Airway structure and physiological function vary with age. Such variations may alter the
deposition patterns for inhaled particles. Airflow distribution can be very uneven in
obstructed 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 limited.
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 (e.g., the laboratory rat). Such calculations can be used to predict the exposures
needed to produce comparable doses for animal to human extrapolation.
Computational models have been improved in recent years, but experimental validation
of model predictions is still required.
6-109
-------
REFERENCES
Adamson, I. Y. R.; Bowden, D. H. (1981) Dose response of the pulmonary macrophagic system to various
particulates and its relationship to transepithelial passage of free particles. Exp. Lung Res. 2: 165-175.
Adamson, I. Y. R.; Hedgecock, C. (1995) Patterns of particle deposition and retention after instillation to mouse
lung during acute injury and fibrotic repair. Exp. Lung Res. 21: 695-709.
Adamson, I. Y. R.; Prieditis, H. (1998) Silica deposition in the lung during epithelial injury potentiates fibrosis and
increases particle translocation to lymph nodes. Exp. Lung Res. 24: 293-306.
Adgate, J. L.; Ramachandran, G.; Pratt, G. C.; Waller, L. A.; Sexton, K. (2002) Spatial and temporal variability in
outdoor, indoor, and personal PM25 exposure. Atmos. Environ. 36: 3255-3265.
American Conference of Governmental Industrial Hygienists (ACGIH). (1985) Particle size-selective sampling in
the workplace: report of the ACGIH technical committee on air sampling procedures. Cincinnati, OH:
American Conference of Governmental Industrial Hygienists.
Anjilvel, S.; Asgharian, B. (1995) A multiple-path model of particle deposition in the rat lung. Fundam. Appl.
Toxicol. 28: 41-50.
Asgharian, B.; Anjilvel, S. (1994) Inertia! and gravitational deposition of particles in a square cross section
bifurcating airway. Aerosol Sci. Technol. 20: 177-193.
Asgharian, B.; Wood, R.; Schlesinger, R. B. (1995) Empirical modeling of particle deposition in the alveolar region
of the lungs: abasis for interspecies extrapolation. Fundam. Appl. Toxicol. 27: 232-238.
Asgharian, B.; Miller, F. J.; Subramaniam, R. P. (1999) Dosimetry software to predict particle deposition in humans
and rats. CUT Activities 19(3). Available: http://libpc.ciit.org/Activities/1999/V19N3.pdf [30 September,
2004].
Asgharian, B.; Hofmann, W.; Miller, F. J. (2001) Mucociliary clearance of insoluble particles from the
tracheobronchial airways of the human lung. Aerosol Sci. 32: 817-832.
Balashazy, I.; Hofmann, W. (1993a) Particle deposition in airway bifurcations-I: inspiratory flow. J. Aerosol Sci.
24: 745-772.
Balashazy, I.; Hofmann, W. (1993b) Particle deposition in airway bifurcations-II: expiratory flow. J. Aerosol Sci.
24: 773-786.
Balashazy, I.; Hofmann, W. (1995) Deposition of aerosols in asymmetric airway bifurcations. J. Aerosol Sci.
26: 273-292.
Balashazy, I.; Hofmann, W. (2001) Fluid dynamics and related particle deposition patterns in human airway
bifurcations. In: Martonen, T. B., ed., Medical Applications of Computer Modeling: The Respiratory System.
WIT Press, Southampton, Boston, p. 84-108.
Balashazy, I.; Hofmann, W.; Heistracher, T. (1999) Computation of local enhancement factors for the quantification
of particle deposition patterns in airway bifurcations. J. Aerosol Sci. 30: 185-203.
Becquemin, M. H.; Swift, D.L.; Bouchikhi, A.; Roy, M.; Teillac, A. (1991) Particle deposition and resistance in the
noses of adults and children. Eur. Resp. J. 4: 694-702.
Bell, K. A.; Friedlander, S. K. (1973) Aerosol deposition in models of a human lung bifurcation. Staub Reinhalt.
Luft33: 178-182.
Bennett, W. D.; Ilowite, J. S. (1989) Dual pathway clearance of 99mTc-DTPA from the bronchial mucosa. Am. Rev.
Respir. Dis. 139: 1132-1138.
Bennett, W. D.; Zeman, K. L. (1998) Deposition of fine particles in children spontaneously breathing at rest.
Inhalation Toxicol. 10: 831-842.
Bennett, W. D.; Zeman, K. L.; Kim, C. (1996) Variability of fine particle deposition in healthy adults: effect of age
and gender. Am. J. Respir. Crit. Care Med. 153: 1641-1647.
Bennett, W. D.; Zeman, K. L.; Kang, C. W.; Schechter, M. S. (1997a) Extrathoracic deposition of inhaled, coarse
particles (4.5 um) in children vs. adults. In: Cherry, N.; Ogden, T., eds. Inhaled particles VIII: proceedings of
an international symposium on inhaled particles organized by the British Occupational Hygiene Society;
August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 497-502.
Bennett, W. D.; Zeman, K. L.; Kim, C.; Mascarella, J. (1997b) Enhanced deposition of fine particles in COPD
patients spontaneously breathing at rest. Inhalation Toxicol. 9: 1-14.
Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1998) Bronchial
airway deposition and retention of particles in inhaled boluses: effect of anatomic dead space. J. Appl.
Physiol. 85: 685-694.
6-110
-------
Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1999) Regional
deposition and retention of particles in shallow, inhaled boluses: effect of lung volume. J. Appl. Physiol.
86: 168-173.
Benson, J. M; Henderson, R. F.; McClellan, R. O.; Hanson, R. L.; Rebar, A. H. (1986) Comparative acute toxicity
of four nickel compounds to F344 rat lung. Fundam. Appl. Toxicol. 7: 340-347.
Brauer, M.; Avila-Casado, C.; Fortoul, T. L; Vedal, S.; Stevens, B.; Churg, A. (2001) Air pollution and retained
particles in the lung. Environ. Health Perspect. 109: 1039-1043.
Broday, D. M.; Georgopoulos, P. G. (2001) Growth and deposition of hygroscopic paniculate matter in the human
lungs. Aerosol Sci. Technol. 34: 144-159.
Brown, J. S.; Kirby, Z. L.; Bennett, W. D. (2001) Regional deposition of coarse particles and ventilation distribution
in healthy subjects and patients with cystic fibrosis. J. Aerosol Med. 14: 443-454.
Brown, J. S.; Zeman, K. L.; Bennett, W. D. (2002) Ultrafine particle deposition and clearance in the healthy and
obstructed lung. Am. J. Respir. Crit. Care Med. 166: 1240-1247.
Burch, W. M. (2002) Comment on "Passage of inhaled particles into the blood circulation in humans". Circulation
106: el41-e!42.
California Air Resources Board. (1987) Ambient air quality standard for ozone: health and welfare effects.
Sacramento, CA: California Air Resources Board.
Camner, P.; Anderson, M.; Philipson, K.; Bailey, A.; Hashish, A.; Jarvis, N.; Bailey, M.; Svartengren, M. (1997)
Human bronchiolar deposition and retention of 6-, 8-, and 10-u particles. Exp. Lung Res. 23: 517-535.
Chang, Y. H.; Yu, C. P. (1999) A model of ventilation distribution in the human lung. Aerosol Sci. Technol.
30: 309-319.
Cheng, Y.-S.; Smith, S. M.; Yeh, H.-C.; Kim, D.-B.; Cheng, K.-H.; Swift, D. L. (1995) Deposition of ultrafine
aerosols and thoron progeny in replicas of nasal airways of young children. Aerosol. Sci. Technol.
23: 541-552.
Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Guilmette, R. A.; Simpson, S. Q.; Yang, Y.-H.; Swift, D. L. (1996) /« vivo
measurements of nasal airway dimensions and ultrafine aerosol deposition in the human nasal and oral
airways. J. Aerosol Sci. 27: 785-801.
Cheng, K.-H.; Cheng, Y.-S.; Yeh, H.-C.; Swift, D. L. (1997) An experimental method for measuring aerosol
deposition efficiency in the human oral airway. Am. Ind. Hyg. Assoc. J. 58: 207-213.
Churg, A.; Brauer, M. (1997) Human lung parenchyma retains PM2 5. Am. J. Respir. Crit. Care Med.
155:2109-2111.
Churg, A.; Vedal, S. (1996) Carinal and tubular airway particle concentrations in the large airways of non-smokers
in the general population: evidence for high particle concentration at airway carinas. Occup. Environ. Med.
53: 553-558.
Cohen, B. S.; Harley, N. H.; Schlesinger, R. B.; Lippmann, M. (1988) Nonuniform particle deposition on
tracheobronchial airways: implications for lung dosimetry. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.;
Fisher, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung
dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32 (suppl. 1): 1045-1053.
Cohen, B. S.; Sussman, R. G.; Lippmann, M. (1990) Ultrafine particle deposition in a human tracheobronchial cast.
Aerosol Sci. Technol. 12: 1082-1091.
Cohen, B. S.; Xiong, J. Q.; Fang, C.-P.; Li, W. (1998) Deposition of charged particles on lung airways. Health Phys.
74: 554-560.
Comer, J. K.; Kleinstreuer, C.; Hyun, S.; Kim, C. S. (2000) Aerosol transport and deposition in sequentially
bifurcating airways. J. Biomech. Eng. 122: 152-158.
Comer, J. K.; Kleinstreuer, C.; Zhang, Z. (2001a) Flow structures and particle deposition patterns in double
bifurcation airway models. Part 1. Air flow fields. J. Fluid Mech. 435: 24-54.
Comer, J. K.; Kleinstreuer, C.; Kim, C. S. (200Ib) Flow structures and particle deposition patterns in double
bifurcation airway models. Part 2. Aerosol transport and deposition. J. Fluid Mech. 435: 55-80.
Crapo, J. D.; Barry, B. E.; Gehr, P.; Bachofen, M.; Weibel, E. R. (1982) Cell characteristics of the normal human
lung. Am. Rev. Respir. Dis. 125: 740-745.
Crystal, R. G., West, J. B.; Barnes, P. J.; Weibel, E. R., eds. (1997) The lung: scientific foundations. Volume 1.
Section III: major components. 2nd ed. Philadelphia, PA: Lippincott-Raven; chapters 30-67, pp. 445-991.
Cuddihy, R. G. (1984) Mathematical models for predicting clearance of inhaled radioactive materials. In: Smith, H.;
Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a meeting jointly
organized by the Commission of the European Communities and the National Radiological Protection Board;
March; Oxford, United Kingdom. Luxembourg: Commission of the European Communities; pp. 167-179;
report no. EUR 9384 EN.
6-111
-------
Cuddihy, R. G.; Yeh, H. C. (1988) Respiratory tract clearance of particles and substances dissociated from particles.
In: Mohr, U.; Dungworth, D.; Kimmerle, G.; Lewkowski, J.; McClellan, R.; Stober, W., eds. Inhalation
toxicology: the design and interpretation of inhalation studies and their use in risk assessment. New York,
NY: Springer-Verlag; pp. 169-193.
Darquenne, C. (2001) A realistic two-dimensional model of aerosol transport and deposition in the alveolar zone of
the human lung. J. Aerosol Sci. 32: 1161-1174.
Dorries, A. M; Valberg, P. A. (1992) Heterogeneity of phagocytosis for inhaled versus instilled material. Am. Rev.
Respir.Dis. 146:831-837.
Driscoll, K. E. (1995) The toxicology of crystalline silica studied in vitro. Appl. Occup. Environ. Hyg.
10: 1118-1125.
Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
limitations. Toxicol. Sci. 55: 24-35.
Falk, R.; Philipson, K.; Svartengren, M.; Jarvis, N.; Bailey, M.; Camner, P. (1997) Clearance of particles from small
ciliated airways. Exp. Lung Res. 23: 495-515.
Falk, R.; Philipson, K.; Svartengren, M.; Bergmann, R.; Hofmann, W.; Jarvis, N.; Bailey, M.; Camner, P. (1999)
Assessment of long-term bronchiolar clearance of particles from measurements of lung retention and
theoretical estimates of regional deposition. Exp. Lung Res. 25: 495-516.
Ferin, J. (1977) Effect of particle content of lung on clearance pathways. In: Sanders, C. L.; Schneider, R. P.; Dagle,
G. E.; Ragan, H. A., eds. Pulmonary macrophages and epithelial cells: proceedings of the sixteenth annual
Hanford biology symposium; September 1976; Richland, WA. Oak Ridge, TN: Energy Research and
Development Administration; pp. 414-423. Available from: NTIS, Springfield, VA; CONF-760927.
(ERDA symposium series 43).
Ferin, J.; Feldstein, M. L. (1978) Pulmonary clearance and hilar lymph node content in rats after particle exposure.
Environ. Res. 16: 342-352.
Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
Respir. CellMol. Biol. 6: 535-542.
Findeisen, W. (1935) Uber das Absetzen kleiner, in der Luft suspendierter Teilchen in der menschlichen Lunge bei
der Atmung [The deposition of small airborne particles in the human lung during respiration]. Pfluegers Arch.
Gesamte Physiol. Menschen Tiere 236: 367-379.
Foster, W. M.; Langenback, E.; Bergofsky, E. H. (1980) Measurement of trachea! and bronchial mucus velocities in
man: relation to lung clearance. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 48: 965-971.
Frampton, M. W.; Chalupa, D.; Morrow, P. E.; Gibb, F. R.; Oberdorster, G.; Speers, D. M.; Zareba, W.; Utell, M. J.
(2000) Deposition and effects of inhaled ultrafine carbon particles in healthy subjects at rest. Presented at:
PM2000: paniculate matter and health—the scientific basis for regulatory decision-making, specialty
conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.
Fry, F. A.; Black, A. (1973) Regional deposition and clearance of particles in the human nose. J. Aerosol Sci.
4: 113-124.
Gehr, P.; Bachofen, M.; Weibel, E. R. (1978) The normal human lung: ultrastructure and morphometric estimation
of diffusion capacity. Respir. Physiol. 32: 121-140.
Gehr, P.; Schurch, S.; Berthaiume, Y.; Im Hof, V.; Geiser, M. (1990) Particle retention in airways by surfactant.
J. Aerosol Med. 3:27-43.
Gehr, P.; Im Hof, V.; Geiser, M.; Schurch, S. (1991) The fate of particles deposited in the intrapulmonary
conducting airways. J. Aerosol Med. 4: 349-362.
Gradon, L.; Orlicki, D. (1990) Deposition and inhaled aerosol particles in a generation of the tracheobronchial tree.
J. Aerosol Sci. 21:3-19.
Green, F. H. Y. (2000) Pulmonary responses to inhaled poorly soluble paniculate in the human. In: Gardner, D. E.,
ed. ILSI Risk Science Institute Workshop: The Relevance of the Rat Lung Response to Particle Overload for
Human Risk Assessment; March, 1998. Inhalation Toxicol. 12: 59-95.
Groth, M. L.; Macri, K.; Foster, W. M. (1997) Cough and mucociliary transport of airway paniculate in chronic
obstructive lung disease. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an
international symposium on inhaled particles organised by the British Occupational Hygiene Society;
August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 515-521.
Guilmette, R. A.; Cheng, Y. S.; Griffith, W. C. (1997) Characterising the variability in adult human nasal airway
dimensions. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international symposium
on inhaled particles organised by the British Occupational Hygiene Society; August 1996; Cambridge, UK.
Ann. Occup. Hyg. 41(suppl. 1): 491-496.
6-112
-------
Harmsen, A. G.; Muggenburg, B. A.; Snipes, M. B.; Bice, D. E. (1985) The role of macrophages in particle
translocation from lungs to lymph nodes. Science (Washington, DC) 230: 1277-1280.
Heistracher, T.; Hofmann, W. (1995) Physiologically realistic models of bronchial airway bifurcation. J. Aerosol
Sci. 26: 497-509.
Heistracher, T.; Hofmann, W. (1997) Flow and deposition patterns in successive airway bifurcations. In: Cherry, N.;
Ogden, T., eds. Inhaled Particles VIII: proceedings of an international symposium on inhaled particles
organised by the British Occupational Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg.
41(suppl.): 537-542.
Henderson, R. F.; Driscoll, K. E.; Harkema, J. R.; Lindenschmidt, R. C.; Chang, I.-Y.; Maples, K. R.; Barr, E. B.
(1995) A comparison of the inflammatory response of the lung to inhaled versus instilled particles inF344
rats. Fundam. Appl. Toxicol. 24: 183-197.
Henshaw, D. L.; Fews, A. P. (1984) The microdistribution of alpha emitting particles in the human lung.
In: Smith, H.; Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a
meeting jointly organized by the Commission of the European Communities and the National Radiological
Protection Board; March; Oxford, United Kingdom. Luxembourg: Commission of the European
Communities; pp. 199-218; report no. EUR 9384 EN.
Heyder, J.; Rudolf, G. (1977) Deposition of aerosol particles in the human nose. In: Walton, W. H.; McGovern, B.,
eds. Inhaled particles IV: proceedings of an international symposium, part 1; September 1975; Edinburgh,
United Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 107-126.
Killer, F. C. (1991) Health implications of hygroscopic particle growth in the human respiratory tract. J. Aerosol
Med. 4: 1-23.
Hinds, W. C. (1999) Aerosol technology: properties, behavior, and measurement of airborne particles. 2nd ed.
New York, NY: John Wiley & Sons.
Hofmann, W.; Bergmann, R. (1998) Predictions of particle deposition patterns in human and rat airways. Inhalation
Toxicol. 10: 557-583.
Hofmann, W.; Martonen, T. B.; Graham, R. C. (1989a) Predicted deposition of nonhygroscopic aerosols in the
human lung as a function of subject age. J. Aerosol Med. 2: 49-68.
Hofmann, W.; Koblinger, L.; Martonen, T. B. (1989b) Structural differences between human and rat lungs:
implications for Monte Carlo modeling of aerosol deposition. In: Mahaffey, J. A., ed. 26thHanford life
sciences symposium, modeling for scaling to man: biology, dosimetry, and response. Health Phys.
57(suppl. 1): 41-47.
Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
Hofmann, W.; Bergmann, R.; Koblinger, L. (1999) Characterization of local particle deposition patterns in human
and rat lungs by different morphometric parameters. J. Aerosol Sci. 30: 651-667.
Hofmann, W.; Asgharian, B.; Bergmann, R.; Anjilvel, S.; Miller, F. J. (2000) The effect of heterogeneity of lung
structure on particle deposition in the rat lung. Toxicol. Sci. 53: 430-437.
Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammalian species.
Inhalation Toxicol. 10: 121-130.
Huchon, G. J.; Montgomery, A. B.; Lipavsky, A.; Hoeffel, J. M.; Murray, J. F. (1987) Respiratory clearance of
aerosolized radioactive solutes of varying molecular weight. J. Nucl. Med. 28: 894-902.
International Commission on Radiological Protection. (1960) Report of Committee II on permissible dose for
internal radiation (1959). Health Phys. 3.
International Commission on Radiological Protection. (1979) Limits for intakes of radionuclides by workers.
Oxford, United Kingdom: Pergamon Press; ICRP publication 30, part 1.
International Commission on Radiological Protection. (1994) Human respiratory tract model for radiological
protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP: v. 24, pp. 1-482).
Jaques, P. A.; Kim, C. S. (2000) Measurement of total lung deposition of inhaled ultrafine particles in healthy men
and women. Inhalation Toxicol. 12: 715-731.
John, J.; Wollmer, P.; Dahlback, M.; Luts, A.; Jonson, B. (1994) Tidal volume and alveolar clearance of insoluble
particles. J. Appl. Physiol. 76: 584-588.
Katz, I. M. (2001) Computer modeling of fluid dynamics and particle motion in the larynx and trachea. In:
Martonen, T. B., ed. Medical Applications of Computer Modeling: The Respiratory System. Boston, MA:
WIT Press, pp. 47-63.
Katz, I. M.; Martonen, T. B. (1996) Three-dimensional fluid particle trajectories in the human larynx and trachea.
J. Aerosol Med. 9:513-520.
6-113
-------
Katz, I. M, Martonen, T. B., and Flaa, W. (1997) Three-dimensional computational study of inspiratory aerosol flow
through the larynx: the effect of glottal aperture modulation. J. Aerosol Sci. 28: 1073-83.
Katz, I. M.; Davis, B. M.; Martonen, T. B. (1999) Anumerical study of particle motion within the human larynx and
trachea. J. Aerosol Sci. 30: 173-183.
Kaye, S. R.; Phillips, C. G.; Winlove, C. P. (2000) Measurement of non-uniform aerosol deposition patterns in the
conducting airways of the porcine lung. J. Aerosol Sci. 31: 849-866.
Kesavanathan, J.; Swift, D. L. (1998) Human nasal passage particle deposition: the effect of particle size, flow rate,
and anatomical factors. Aerosol Sci. Technol. 28: 457-463.
Keyhani, K.; Scherer, P. W.; Mozell, M. M. (1995) Numerical simulation of airflow in the human nasal cavity.
J. Biomech. Eng. 117: 429-41.
Kim, C. S. (2000) Methods of calculating lung delivery and deposition of aerosol particles. Respir. Care
45:695-711.
Kim, C. S.; Fisher, D. M. (1999) Deposition characteristics of aerosol particles in sequentially bifurcating airway
models. Aerosol Sci. Technol. 31:198-220.
Kim, C. S.; Garcia, L. (1991) Particle deposition in cyclic bifurcating tube flow. Aerosol Sci. Technol. 14: 302-315.
Kim, C. S.; Hu, S. C. (1998) Regional deposition of inhaled particles in human lungs: comparison between men and
women. J. Appl. Physiol. 84: 1834-1844.
Kim, C. S.; Iglesias, A. J. (1989) Deposition of inhaled particles in bifurcating airway models: I. inspiratory
deposition. J. Aerosol Med. 2: 1-14.
Kim, C. S.; Jaques, P. A. (2000) Respiratory dose of inhaled ultrafine particles in healthy adults. Phil. Trans. Roy.
Soc. London A 358: 2693-2705.
Kim, C. S.; Kang, T. C. (1997) Comparative measurement of lung deposition of inhaled fine particles in normal
subjects and patients with obstructive airway disease. Am. J. Respir. Crit. Care Med. 155: 899-905.
Kim, C. S.; Fisher, D. M.; Lutz, D. J.; Gerrity, T. R. (1994) Particle deposition in bifurcating airway models with
varying airway geometry. J. Aerosol Sci. 25: 567-581.
Kim, C. S.; Hu, S. C.; DeWitt, P.; Gerrity, T. R. (1996) Assessment of regional deposition of inhaled particles in
human lungs by serial bolus delivery method. J. Appl. Physiol. 81: 2203-2213.
Kimbell, J. S. (2001) Computational fluid dynamics of the extrathoracic airways. In: Martonen, T. B., ed. Medical
Applications of Computer Modeling: The Respiratory System. Boston, MA: WIT Press; pp. 11-45.
Kinsara, A. A.; Tompson, R. V.; Loyalka, S. K. (1993) Computational flow and aerosol concentration profiles in
lung bifurcations. Health Phys. 64: 13-22.
Koblinger, L.; Hofmann, W. (1985) Analysis of human lung morphometric data for stochastic aerosol deposition
calculations. Phys. Med. Biol. 30: 541-556.
Koblinger, L.; Hofmann, W. (1988) Stochastic morphological model of the rat lung. Anat. Rec. 221: 533-539.
Koch, W.; Stober, W. (2001) A simple pulmonary retention model accounting for dissolution and
macrophage-mediated removal of deposited polydisperse particles. Inhalation Toxicol. 13: 129-148.
Kodavanti, U. P.; Schladweiler, M. C. J.; Ledbetter, A. D.; Hauser, R.; Christiani, D. C.; Samet, J. M.; McGee, J.;
Richards, J. H.; Costa, D. L. (2002) Pulmonary and systemic effects of zinc-contining emission particles in
three rat strains: multiple exposure scenarios. Toxicol. Sci. 70: 73-85.
Kohlhaufl, M.; Brand, P.; Scheuch, G.; Meyer, T. S.; Schulz, H.; Haussinger, K.; Heyder, J. (1999) Increased fine
particle deposition in women with asymptomatic nonspecific airway hyperresponsiveness. Am. J. Respir.
Crit. Care Med. 159: 902-906.
Kreyling, W. G. (1992) Intracellular particle dissolution in alveolar macrophages. Environ. Health Perspect.
97:121-126.
Kreyling, W. G.; Blanchard, J. D.; Godleski, J. J.; Haeussermann, S.; Heyder, J.; Hutzler, P.; Schulz, H.; Sweeney,
T. D.; Takenaka, S.; Ziesenis, A. (1999) Anatomic localization of 24-and 96-h particle retention in canine
airways. J. Appl. Physiol. 87: 269-284.
Kreyling, W. G.; Semmler, M.; Erbe, F.; Mayer, P.; Takenaka, S.; Schulz, H.; Oberdorster, G.; Ziesenis, A. (2002)
Translocation of ultrafine insoluble iridium particles from lung epithelium to extrapulmonary organs is size
dependent but very low. J. Toxicol. Environ. Health Part A 65: 1513-1530.
Kuempel, E. D. (2000) Comparison of human and rodent lung dosimetry models for particle clearance and retention.
Drug Chem. Toxicol. 23: 203-222.
Kuempel, E. D.; O'Flaherty, E. J.; Stayner, L. T.; Smith, R. J.; Green, F. H. Y.; Vallyathan, V. (2001a)
A biomathematical model of particle clearance and retention in the lungs of coal miners. I. Model
development. Regul. Toxicol. Pharmacol. 34: 69-87.
6-114
-------
Kuempel, E. D.; Tran, C.-L.; Smith, R. J.; Bailer, A. J. (200Ib) A biomathematical model of particle clearance and
retention in the lungs of coal miners. II. Evaluation of variability and uncertainty. Regul. Toxicol. Pharmacol.
34: 88-101.
LaBelle, C. W.; Brieger, H. (1961) Patterns and mechanisms in the elimination of dust from the lung. In: Davies,
C. N., ed. Inhaled particles and vapours: proceedings of an international symposium; March-April 1960;
Oxford, United Kingdom. New York, NY: Pergamon Press; pp. 356-368.
Landahl, H. D. (1950) On the removal of air-borne droplets by the human respiratory tract: I. the lung. Bull. Math.
Biophys. 12: 43-56.
Lay, J. C.; Berry, C. R.; Kim, C. S.; Bennett, W. D. (1995) Retention of insoluble particles after local intrabronchial
deposition in dogs. J. Appl. Physiol. 79: 1921-1929.
Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular distribution
of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol. 18: 687-695.
Lazaridis, M.; Broday, D. M.; Hov, 0.; Georgopoulos, P. G. (2001) Integrated exposure and dose modeling and
analysis system. 3. Deposition of inhaled particles in the human respiratory tract. Environ. Sci. Technol.
35: 3727-3734.
Lee, J. W.; Goo, J. H.; Chung, M. K. (1996) Characteristics of inertia! deposition in a double bifurcation. J. Aerosol
Sci. 27: 119-138.
Lehnert, B. E.; Morrow, P. E. (1985) Association of 59iron oxide with alveolar macrophages during alveolar
clearance. Exp. Lung Res. 9: 1-16.
Lehnert, B. E.; Valdez, Y. E.; Bomalaski, S. H. (1988) Analyses of particles in the lung free cell, tracheobronchial
lymph nodal, and pleural space compartments following their deposition in the lung as related to lung
clearance mechanisms. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fisher, D. R., eds. Inhaled particles
VI: proceedings of an international symposium and workshop on lung dosimetry; September 1985;
Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 125-140.
Leikauf, G.; Yeates, D. B.; Wales, K. A.; Spektor, D.; Albert, R. E.; Lippmann, M. (1981) Effects of sulfuric acid
aerosol on respiratory mechanics and mucociliary particle clearance in healthy nonsmoking adults. Am. Ind.
Hyg. Assoc. J. 42: 273-282.
Leikauf, G. D.; Spektor, D. M.; Albert, R. E.; Lippmann, M. (1984) Dose-dependent effects of submicrometer
sulfuric acid aerosol on particle clearance from ciliated human lung airways. Am. Ind. Hyg. Assoc. J.
45: 285-292.
Lennon, S.; Zhang, Z.; Lessmann, R.; Webster, S. (1998) Experiments on particle deposition in the human upper
respiratory system. Aerosol Sci. Technol. 28: 464-474.
Leong, B. K. J.; Coombs, J. K.; Sabaitis, C. P.; Rop, D. A.; Aaron, C. S. (1998) Quantitative morphometric analysis
of pulmonary deposition of aerosol particles inhaled via intratracheal nebulization, intratracheal instillation or
nose-only inhalation in rats. J. Appl. Toxicol. 18: 149-160.
Li, A.; Ahmadi, G. (1995) Computer simulation of particle deposition in the upper tracheobronchial tree. Aerosol
Sci. Tech. 23:201-223.
Lundborg, M.; Lind, B.; Camner, P. (1984) Ability of rabbit alveolar macrophages to dissolve metals. Exp. Lung
Res. 7: 11-22.
Lundborg, M.; Eklund, A.; Lind, B.; Camner, P. (1985) Dissolution of metals by human and rabbit alveolar
macrophages. Br. J. Ind. Med. 42: 642-645.
Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
Pharmacol. 153: 28-38.
Martonen, T. B. (1982) Analytical model of hygroscopic particle behavior in human airways. Bull. Math. Biol.
44: 425-442.
Martonen, T. B.; Lowe, J. (1983) Assessment of aerosol deposition patterns in human respiratory tract casts.
In: Marple, V. A.; Liu, B. Y. H., eds. Aerosols in the Mining and Industrial Work Environments Vol. 1
Fundamentals and Status, p. 151-164. Ann Arbor, MI: Ann Arbor Science Publishers.
Martonen, T. B.; Barnett, A. E.; Miller, F. J. (1985) Ambient sulfate aerosol deposition in man: modeling the
influence of hygroscopicity. Environ. Health Perspect. 63: 11-24.
Martonen, T. B.; Hofmann, W.; Eisner, A. D.; Menache, M. G. (1989) The role of particle hygroscopicity in aerosol
therapy and inhalation toxicology. In: Crapo, J. D.; Miller, F. J.; Smolko, E. D.; Graham, J. A.; Hayes, A. W.,
eds. Extrapolation of dosimetric relationships for inhaled particles and gases. San Diego, CA: Academic
Press, Inc.; pp. 303-316.
Martonen, T. B.; Zhang, Z.; Lessmann, R. (1993) Fluid dynamics of the human larynx and upper tracheobronchial
airways. Aerosol Sci. Technol. 19: 133-156.
6-115
-------
Martonen, T. B.; Yang, Y.; Xue, Z. Q. (1994a) Influences of cartilagious rings on tracheobronchial fluid dynamics.
InhalationToxicol. 6: 185-203.
Martonen, T. B.; Yang, Y.; Xue, Z. Q. (1994b) Effects of carinal ridge shapes on lung airstreams. Aerosol Sci.
Technol. 21: 119-136.
Martonen, T. B.; Schroeter, J. D.; Hwang, D.; Fleming, J. S.; Conway, J. H. (2000) Human lung morphology models
for particle deposition studies. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12(suppl. 4): 109-121.
Martonen, T. B.; Zhang, Z; Yu, G.; Musante, C. J. (2001a) Three-dimensional computer modeling of the human
upper respiratory tract. CellBiochem. Biophys. 35: 255-261.
Martonen, T. B.; Guan, X.; Schreck, R. M. (200Ib) Fluid dynamics in airway bifurcations: I. Primary flows.
Inhalation Toxicol. 13: 261-79.
Martonen, T. B.; Guan, X.; Schreck, R. M. (2001c) Fluid dynamics in airway bifurcations: II. Secondary currents.
Inhalation Toxicol. 13: 281-9.
Martonen, T. B.; Guan, X.; Schreck, R. M. (2001d) Fluid dynamics in airway bifurcations: III. Localized flow
conditions. Inhalation Toxicol. 13: 291-305.
Matsui, H.; Randell, S. H.; Peretti, S. W.; Davis, C. W.; Boucher, R. C. (1998) Coordinated clearance of periciliary
liquid and mucus from airway surfaces. J. Clin. Invest. 102: 1125-1131.
Mauderly, J. L. (1979) Effect of age on pulmonary structure and function of immature and adult animals and man.
Fed. Proc. 38: 173-177.
Medinsky, M. A.; Kampcik, S. J. (1985) Pulmonary retention of [14C]benzo/o/pyrene in rats as influenced by the
amount instilled. Toxicology 35: 327-336.
Menache, M. G.; Miller, F. J.; Raabe, O. G. (1995) Particle inhalability curves for humans and small laboratory
animals. Ann. Occup. Hyg. 39: 317-328.
Mercer, T. T. (1967) On the role of particle size in the dissolution of lung burdens. Health Phys. 13: 1211-1221.
Miller, F. J. (2000a) Dosimetry of particles in laboratory animals and humans in relationship to issues surrounding
lung overload and human health risk assessment: a critical review. Inhalation Toxicol. 12: 19-57.
Miller, F. J. (2000b) Errata: F. J. Miller, 2000. Dosimetry of particles in laboratory animals and humans in
relationship to issues surrounding lung overload and human health risk assessment: a critical review. Inhal.
Toxicol. 12: 19-57. Inhalation Toxicol. 12: 1257-1259.
Miller, F. J.; Anjilvel, S.; Menache, M. G.; Asgharian, B.; Gerrity, T. R.. (1995) Dosimetric issues relating to
paniculate toxicity. Inhalation Toxicol. 7: 615-632.
Morrison, D.; Skwarski, D.; Millar, A. M.; Adams, W.; MacNee, W. (1998) A comparison of three methods of
measuring 99mTc-DTPA lung clearance and their repeatability. Eur. Respir. J. 11: 1141-1146.
Morrow, P. E. (1973) Alveolar clearance of aerosols. Arch. Intern. Med. 131: 101-108.
Morrow, P. E. (1977) Clearance kinetics of inhaled particles. In: Brain, J. D.; Proctor, D. F.; Reid, L. M., eds.
Respiratory defense mechanisms (in two parts), part II. New York, NY: Marcel Dekker, Inc.; pp. 491-543.
(Lenfant, C., ed. Lung biology in health and disease: v. 5).
Morrow, P. E. (1986) Factors determining hygroscopic aerosol deposition in airways. Physiol. Rev. 66: 330-376.
Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
10: 369-384.
Morrow, P. E. (1994) Mechanisms and significance of "particle overload." In: Mohr, U.; Dungworth, D. L.;
Mauderly, J. L.; Oberdorster, G., eds. Toxic and carcinogenic effects of solid particles in the respiratory tract:
[proceedings of the 4th international inhalation symposium]; March 1993; Hannover, Germany. Washington,
DC: International Life Sciences Institute Press; pp. 17-25.
Morrow, P. E.; Yu, C. P. (1993) Models of aerosol behavior in airways and alveoli. In: Moren, F.; Dolovich, M. B.;
Newhouse, M. T.; Newman, S. P., eds. Aerosols in medicine: principles, diagnosis and therapy. 2nd rev. ed.
Amsterdam, The Netherlands: Elsevier; pp. 157-193.
Morrow, P. E.; Gibb, F. R.; Gazioglu, K. M. (1967) A study of paniculate clearance from the human lungs.
Am. Rev. Respir. Dis. 96: 1209-1221.
Musante, C. J.; Martonen, T. B. (1999) Predicted deposition patterns of ambient paniculate air pollutants in
children's lungs under resting conditions. In: Proceedings of the third colloquium on paniculate air pollution
and human health; June; Durham, NC. Irvine, CA: University of California, Air Pollution Health Effects
Laboratory, p. 7-15 - 7-20.
Musante, C. J.; Martonen, T. B. (2000a) Computer simulations of particle deposition in the developing human lung.
J. Air Waste Manage. Assoc. 50: 1426-1432.
6-116
-------
Musante, C. I; Martonen, T. B. (2000b) An extrapolation model to aid in lexicological studies of paniculate air
pollutants. Presented at: PM2000: paniculate matter and health—the scientific basis for regulatory
decision-making, specialty conference & exhibition; January; Charleston, SC. Pittsburgh, PA: Air & Waste
Management Association.
Musante, C. I; Martonen, T. B. (2000c) Paniculate matter deposition in the lungs of children and adults: predictions
of an age-dependent computer model. Presented at: PM2000: paniculate matter and health—the scientific
basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC. Pittsburgh,
PA: Air & Waste Management Association.
National Council on Radiation Protection and Measurements (NCRP). (1997) Deposition, retention and dosimetry of
inhaled radioactive substances. Bethesda, MD: National Council on Radiation Protection and Measurements;
report no. 125.
National Radiologic Protection Board (NRPB). (1994) National Radiologic Protection Board Report R287,
implementation of the ICRP publication 66 respiratory tract model: LUDEP 2.0 (LUng Dose Evaluation
Program). Richland, WA: ACJ & Associates, Inc. Software information available: www.acj-associates.com
(13 June 2003).
National Research Council. (2001) Research priorities for airborne paniculate matter. III. Early research progress.
Washington, DC: National Academy Press. Available: http://www.nap.edu/books/0309073375/html/ (4 June
2003).
Naumann, B. D.; Schlesinger, R. B. (1986) Assessment of early alveolar particle clearance and macrophage function
following an acute inhalation of sulfuric acid mist. Exp. Lung Res. 11: 13-33.
Nemmar, A.; Vanbilloen, H.; Hoylaerts, M. F.; Hoet, P. H. M.; Verbruggen, A.; Nemery, B. (2001) Passage of
intratracheally instilled ultrafine particles from the lung into the systemic circulation in hamster. Am. J.
Respir. Crit. Care Med. 164: 1665-1668.
Nemmar, A.; Hoet, H. M.; Vanquickenborne, B.; Dinsdale, D.; Thomeer, M.; Hoylaerts, M. F.; Vanbilloen, H.;
Mortelmans, L.; Nemery, B. (2002) Passage of inhaled particles into the blood circulation in humans.
Circulation 105: 411-414.
Niinimaa, V.; Cole, P.; Mintz, S.; Shephard, R. J. (1981) Oronasal distribution of respiratory airflow. Respir.
Physiol. 43:69-75.
Nikula, K. J.; Avila, K. J.; Griffith, W. C.; Mauderly, J. L. (1997) Lung tissue responses and sites of particle
retention differ between rats and cynomolgus monkeys exposed chronically to diesel exhaust and coal dust.
Fundam. Appl. Toxicol. 37: 37-53.
Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of retained
paniculate material in rat and human lungs. Presented at: PM2000: paniculate matter and health-the scientific
basis for regulatory decision-making, specialty conference & exhibition; January; Charleston, SC. Pittsburgh,
PA: Air & Waste Management Association.
Noone, P. G.; Bennett, W. D.; Regnis, J. A.; Zeman, K. L.; Carson, J. L.; King, M.; Boucher, R. C.; Knowles, M. R.
(1999) Effect of aerosolized uridine-5'-triphosphate on airway clearance with cough in patients with primary
ciliary dyskinesia. Am. J. Respir. Crit. Care Med.: 160: 144-149.
Oberdorster, G. (1993) Lung dosimetry: pulmonary clearance of inhaled particles. Aerosol Sci. Technol.
18: 279-289.
Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer particles
for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
Oberdorster, G.; Sharp, Z.; Atudorei, V.; Elder, A.; Gelein, R.; Lunts, A.; Kreyling, W.; Cox, C. (2002)
Extrapulmonary translocation of ultrafine carbon particles following whole-body inhalation exposure of rats.
J. Toxicol. Environ. Health A 65: 1531-1543.
Oldham, M. J.; Mannix, R. C.; Phalen, R. F. (1997) Deposition of monodisperse particles in hollow models
representing adult and child-size tracheobronchial airways. Health Phys. 72: 827-834.
Oldham, M. J.; Phalen, R. F.; Heistracher, T. (2000) Computational fluid dynamic predictions and experimental
results for particle deposition in an airway model. Aerosol Sci. Technol. 32: 61-71.
Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
Fundam. Appl. Toxicol. 40: 220-227.
Passali, D.; Bianchini Ciampoli, M. (1985) Normal values of mucociliary transport time in young subjects. Int. J.
Pediatr. Otorhinolaryngol. 9: 151-156.
Patton, J. S. (1996) Mechanisms of macromolecule absorption by the lungs. Adv. Drug Delivery Rev. 19: 3-36.
Pavia, D. (1987) Acute respiratory infections and mucociliary clearance. Eur. J. Respir. Dis. 71: 219-226.
6-117
-------
Pavia, D.; Bateman, J. R. M; Clarke, S. W. (1980) Deposition and clearance of inhaled particles. Bull. Eur.
Physiopath. Respir. 16: 335-366.
Peterson, B. T.; Dickerson, K. D.; James, H. L.; Miller, E. J.; McLarty, J. W.; Holiday, D. B. (1989) Comparison of
three tracers for detecting lung epithelial injury in anesthetized sheep. J. Appl. Physiol. 66: 2374-2383.
Phalen, R. F.; Oldham, M. J. (2001) Methods for modeling particle deposition as a function of age. Respir. Physiol.
128: 119-130.
Phalen, R. F.; Oldham, M. J.; Beaucage, C. B.; Crocker, T. T.; Mortensen, J. D. (1985) Postnatal enlargement of
human tracheobronchial airways and implications for particle deposition. Anat. Rec. 212: 368-380.
Pritchard, J. N.; Holmes, A.; Evans, J. C.; Evans, N.; Evans, R. J.; Morgan, A. (1985) The distribution of dust in the
rat lung following administration by inhalation and by single intratracheal instillation. Environ. Res.
36: 268-297.
Pritchard, J. N.; Jefferies, S. J.; Black, A. (1986) Sex differences in the regional deposition of inhaled particles in the
2.5-7.5 umsize range. J. Aerosol Sci. 17: 385-389.
Raabe, O. G.; Al-Bayati, M. A.; Teague, S. V.; Rasolt, A. (1988) Regional deposition of inhaled monodisperse,
coarse, and fine aerosol particles in small laboratory animals. In: Dodgson, J.; McCallum, R. L; Bailey,
M. R.; Fischer, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop
on lung dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 53-63.
Radford, E. P.; Martell, E. A. (1977) Polonium-210: lead-210 ratios as an index of residence time of insoluble
particles from cigarette smoke in bronchial epithelium. In: Walton, W. H.; McGovern, B., eds. Inhaled
particles IV: proceedings of an international symposium, part 2; September 1975; Edinburgh, United
Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 567-581.
Rasmussen, T. R.; Andersen, A.; Pedersen, O. F. (2000) Particle deposition in the nose related to nasal cavity
geometry. Rhinology 38: 102-107.
Roy, M. (1989) Lung clearance modeling on the basis of physiological and biological parameters. Health Phys.
57(suppl. 1): 255-262.
Rutland, J.; Cole, P. J. (1981) Nasal mucociliary clearance and ciliary beat frequency in cystic fibrosis compared
with sinusitis and bronchiectasis. Thorax 36: 654-658.
Sabaitis, C. P.; Leong, B. K. J.; Rop, D. A.; Aaron, C. S. (1999) Validation of intratracheal instillation as an
alternative for aerosol inhalation toxicity testing. J. Appl. Toxicol. 19: 133-140.
Sarangapani, R.; Wexler, A. S. (2000) Modeling particle deposition in extrathoracic airways. Aerosol Sci. Technol.
32: 72-89.
Scheuch, G.; Stahlhofen, W. (1988) Particle deposition of inhaled aerosol boluses in the upper human airways.
J. Aerosol Med. 1:29-36.
Scheuch, G.; Philipson, K.; Falk, R.; Svartengren, M.; Stahlhofen, W.; Camner, P. (1995) Retention of particles
inhaled in boli with and without induced bronchoconstriction. Exp. Lung Res. 21: 901-916.
Schlesinger, R. B. (1985) Clearance from the respiratory tract. Fundam. Appl. Toxicol. 5: 435-450.
Schlesinger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
emissions. In: Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
health. Washington, DC: National Academy Press; pp. 239-298.
Schlesinger, R. B. (1989) Deposition and clearance of inhaled particles. In: McClellan, R. O.; Henderson, R. F., eds.
Concepts in inhalation toxicology. New York, NY: Hemisphere Publishing Corp.; pp. 163-192.
Schlesinger, R. B. (1990) The interaction of inhaled toxicants with respiratory tract clearance mechanisms. Crit.
Rev. Toxicol. 20: 257-286.
Schlesinger, R. B. (1995) Deposition and clearance of inhaled particles. In: McClellan, R. O.; Henderson, R. F., eds.
Concepts in inhalation toxicology. 2nd ed. Washington, DC: Taylor & Francis; pp. 191-224.
Schlesinger, R. B.; Gurman, J. L.; Lippmann, M. (1982) Particle deposition within bronchial airways: comparisons
using constant and cyclic inspiratory flows. Ann. Occup. Hyg. 26: 47-64.
Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
Schroeter, J. D.; Musante, C. J.; Hwang, D.; Burton, R.; Guilmette, R.; Martonen, T. B. (2001) Hygroscopic growth
and deposition of inhaled secondary cigarette smoke in human nasal pathways. Aerosol Sci. Technol.:
34: 137-143.
Schroter, R. C.; Sudlow, M. F. (1969) Flow patterns in models of the human bronchial airways. Respir. Physiol.
7: 341-55.
Segal, R. A.; Guan, X.; Shearer, M; Martonen, T. B. (2000a) Mathematical model of airflow in the lungs of children
I: effects of tumor sizes and locations. J. Theor. Med. 2:199-213.
6-118
-------
Segal, R. A.; Martonen, T. B.; Kim, C. S. (2000b) Comparison of computer simulations to total lung deposition to
human subject data in healthy test subjects. J. Air Waste Manage. Assoc. 50: 1262-1268.
Segal, R. A.; Martonen, T. B.; Kim, C. S.; Shearer, M. (2002) Computer simulations of particle deposition in the
lungs of chronic obstructive pulmonary disease patients. Inhalation Toxicol. 14: 705-720.
Smaldone, G. C.; Perry, R. J.; Bennett, W. D.; Messina, M. S.; Zwang, J.; Ilowite, J. (1988) Interpretation of
"24 hour lung retention" in studies of mucociliary clearance. J. Aerosol Med. 1: 11-20.
Smaldone, G. C.; Foster, W. M.; Oriordan, T. G.; Messina, M. S.; Perry, R. J.; Langenback, E. G. (1993) Regional
impairment of mucociliary clearance in chronic obstructive pulmonary disease. Chest 103: 1390-1396.
Snipes, M. B. (1989) Long-term retention and clearance of particles inhaled by mammalian species. CRC Crit. Rev.
Toxicol. 20: 175-211.
Snipes, M. B.; Clem, M. F. (1981) Retention of microspheres in the rat lung after intratracheal instillation.
Environ. Res. 24: 33-41.
Snipes, M. B.; McClellan, R. O.; Mauderly, J. L.; Wolff, R. K. (1989) Retention patterns for inhaled particles in the
lung: comparisons between laboratory animals and humans for chronic exposures. Health Phys.
57(suppl. 1): 69-78.
Snipes, M. B.; James, A. C.; Jarabek, A. M. (1997) The 1994 ICRP66 human respiratory tract dosimetry model as a
tool for predicting lung burdens from exposures to environmental aerosols. Appl. Occup. Environ. Hyg.
12: 547-554.
Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G.; Philipson, K. (1986a) Clearance from the human airways of
particles of different sizes deposited from inhaled aerosol boli. In: Aerosols: formation and reactivity,
proceedings of the second international aerosol conference; September; Berlin, Federal Republic of Germany.
Oxford, United Kingdom: Pergamon Press; pp. 192-196.
Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G. (1986b) Measurement of lung clearance with pulses of
radioactively-labelled aerosols. J. Aerosol Sci. 17: 333-336.
Stanley, P. J.; Wilson, R.; Greenstone, M. A.; Mackay, I. S.; Cole, P. J. (1985) Abnormal nasal mucociliary
clearance in patients with rhinitis and its relationship to concomitant chest disease. Br. J. Dis. Chest
79: 77-82.
Stapleton, K. W.; Guentsch, E.; Hoskinson, M. K.; Finlay, W. H. (2000) On the suitability of K-E turbulence
modeling for aerosol deposition in the mouth and throat: a comparison with experiment. J. Aerosol Sci.
31:739-749.
Stober, W.; Morrow, P. E.; Koch, W.; Morawietz, G. (1994) Alveolar clearance and retention of inhaled insoluble
particles in rats simulated by a model inferring macrophage particle load distributions. J. Aerosol Sci.
25: 975-1002.
Strom, K. A.; Chan, T. L.; Johnson, J. T. (1988) Pulmonary retention of inhaled submicron particles in rats: diesel
exhaust exposures and lung retention model. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.; Fischer, D. R.,
eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung dosimetry;
September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 645-657.
Suarez, S.; Kazantseva, M.; Bhat, M.; Costa, D.; Hickey, A. J. (2001) The influence of suspension nebulization or
instillation on particle uptake by guinez pig alveolar macrophages. Inhalation Toxicol. 13: 773-788.
Subramaniam, R. P.; Richardson, R. B.; Morgan, K. T.; Guilmette, R. A.; Kimbell, J. S. (1998) Computational fluid
dynamics simulations of inspiratory airflow in the human nose and nasopharynx. Inhalation Toxicol.
10: 91-120.
Subramaniam, R. P.; Asgharian, B.; Miller, F. J.; Anjilvel, S.; Freijer, J. I. (2003) Analysis of lobar differences in
particle deposition in the human lung. Inhalation Toxicol. 15: 1-21.
Svartengren, K.; Lindestad, P.; Svartengren, M.; Philipson, K.; Bylin, G.; Camner, P. (1995) Added external
resistance reduces oropharyngeal deposition and increases lung deposition of aerosol particles in asthmatics.
Am. J. Respir. Crit. Care Med. 152: 32-37.
Svartengren, K.; Philipson, K.; Svartengren, M.; Anderson, M.; Camner, P. (1996a) Tracheobronchial deposition and
clearance in small airways in asthmatic subjects. Eur. Respir. J. 9: 1123-1129.
Svartengren, K.; Ericsson, C. H.; Svartengren, M.; Mossberg, B.; Philipson, K.; Camner, P. (1996b) Deposition and
clearance in large and small airways in chronic bronchitis. Exp. Lung Res. 22: 555-576.
Svartengren, K.; Philipson, K.; Svartengren, M.; Camner, P. (1998) Effect of adrenergic stimulation on clearance
from small ciliated airways in healthy subjects. Exp. Lung Res. 24: 149-158.
Svartengren, M.; Svartengren, K.; Aghaie, F.; Philipson, K.; Camner, P. (1999) Lung deposition and extremely slow
inhalations of particles. Limited effect of induced airway obstruction. Exp. Lung Res. 25: 353-366.
Sweeney, T. D.; Tryka, A. F.; Brain, J. D. (1990) Effect of exercise on redistribution and clearance of inhaled
particles from hamster lungs. J. Appl. Physiol. 68: 967-972.
6-119
-------
Sweeney, T. D.; Skornik, W. A.; Brain, J. D.; Hatch, V.; Godleski, J. J. (1995) Chronic bronchitis alters the pattern
of aerosol deposition in the lung. Am. J. Respir. Crit. Care Med. 151: 482-488.
Swift, D. L.; Strong, J. C. (1996) Nasal deposition of ultrafine 218Po aerosols in human subjects. J. Aerosol Sci.
27: 1125-1132.
Takahashi, S.; Asaho, S.; Kubota, Y.; Sato, H.; Matsuoka, O. (1987) Distribution of 198Au and 133Ba in thoracic and
cervical lymph nodes of the rat following the intratracheal instillation of 198Au-colloid and 133BaSO4. J. Radiat.
Res. 28:227-231.
Takahashi, S.; Kubota, Y.; Hatsuno, H. (1992) Effect of size on the movement of latex particles in the respiratory
tract following local administration. Inhalation Toxicol. 4: 113-123.
Takenaka, S.; Karg, E.; Roth, C.; Schulz, H.; Ziesenis, A.; Heinzmann, U.; Schramel, P.; Heyder, J. (2001)
Pulmonary and systemic distribution of inhaled ultrafine silver particles in rats. Environ. Health Perspect.
109(suppl. 4): 547-551.
Toms, N.; Hasani, A.; Pavia, D.; Clarke, S. W.; Agnew, J. E. (1997) Effect of mucus hypersecretion on initial time
course of inert particle clearance from the lung. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII:
proceedings of an international symposium on inhaled particles organised by the British Occupational
Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 509-514.
Tran, C. L.; Jones, A. D.; Cullen, R. T.; Donaldson, K. (1999) Exploration of the mechanisms of retention and
clearance of low-toxicity particles in the rat lung using a mathematical model. Inhalation Toxicol.
11: 1077-1108.
Tran, C. L.; Buchanan, D.; Cullen, R. T.; Searl, A.; Jones, A. D.; Donaldson, K. (2000) Inhalation of poorly soluble
particles. II. Influence of particle surface area on inflammation and clearance. Inhalation Toxicol.
12: 1113-1126.
Tsuda, A.; Butler, J. P.; Fredberg, J. J. (1994a) Effects of alveolated duct structure on aerosol kinetics. I. Diffusional
deposition in the absence of gravity. J. Appl. Physiol. 76: 2497-2509.
Tsuda, A.; Butler, J. P.; Fredberg, J. J. (1994b) Effects of alveolated duct structure on aerosol kinetics. II.
Gravitational sedimentation and inertia! impaction. J. Appl. Physiol. 76: 2510-2516.
Tsuda, A.; Henry, F. S.; Otani, Y.; Haber, S.; Butler, J. P. (1996) Aerosol transport and deposition in the
rhythmically expanding pulmonary acinus. J. Aerosol. Med. 9: 389-408.
U.S. Environmental Protection Agency. (1994) Supplement to the second addendum (1986) to air quality criteria for
paniculate matter and sulfur oxides (1982): assessment of new findings on sulfur dioxide acute exposure
health effects in asthmatic individuals. Research Triangle Park, NC: Office of Health and Environmental
Assessment, Environmental Criteria and Assessment Office; report no. EPA-600/FP-93/002. Available from:
NTIS, Springfield, VA; PB98-132384.
U.S. Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
U.S. Environmental Protection Agency. (1996b) Air quality criteria for ozone and related photochemical oxidants, v.
I-III. Research Triangle Park, NC: Office of Research and Development; report no. EPA/600/P-93/004aF-cF.
Available from: NTIS, Springfield, VA; PB96-185582. PB96-185590, andPB96-185608. Available:
www.epa.gov/ncea/ozone.htm [25 February 2002].
Venkataraman, C.; Kao, A. S. (1999) Comparison of particle lung doses from the fine and coarse fractions of urban
PM-10 aerosols. Inhalation Toxicol. 11: 151-169.
Wagner, E. M; Foster, W. M. (2001) The role of the bronchial vasculature in soluble particle clearance. Environ.
Health Perspect. 109(suppl. 4): 563-565.
Warheit, D. B.; Carakostas, M. C.; Hartsky, M. A.; Hansen, J. F. (1991) Development of a short-term inhalation
bioassay to assess pulmonary toxicity of inhaled particles: comparisons of pulmonary responses to carbonyl
iron and silica. Toxicol. Appl. Pharmacol. 107: 350-368.
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
White, F. M. (1974) Viscous fluid flow. New York, NY: McGraw-Hill, pp. 65-72.
Winter-Sorkina, R. de; Cassee, F. R. (2002) From concentration to dose: factors influencing airborne paniculate
matter deposition in humans and rats. Bilthoven, The Netherlands: National Institute of Public Health and the
Environment (RIVM); report no. 650010031/2002. Available:
http://www.rivm.nl/bibliotheek/rapporten/650010031.html (13 June 2003).
Wolff, R. K. (1986) Effects of airborne pollutants on mucociliary clearance. Environ. Health Perspect. 66: 223-237.
Wolff, R. K.; Henderson, R. F.; Snipes, M. B.; Griffith, W. C.; Mauderly, J. L.; Cuddihy, R. G.; McClellan, R. O.
(1987) Alterations in particle accumulation and clearance in lungs of rats chronically exposed to diesel
exhaust. Fundam. Appl. Toxicol. 9: 154-166.
6-120
-------
Xu, G. B.; Yu, C. P. (1986) Effects of age on deposition of inhaled aerosols in the human lung. Aerosol Sci.
Technol. 5: 349-357.
Yeates, D. B.; Aspin, M. (1978) A mathematical description of the airways of the human lungs. Respir. Physiol.
32: 91-104.
Yeates, D. B.; Aspin, N.; Levison, H.; Jones, M. T.; Bryan, A. C. (1975) Mucociliary tracheal transport rates in man.
J. Appl. Physiol. 19: 487-495.
Yeates, D. B.; Pitt, B. R.; Spektor, D. M.; Karron, G. A.; Albert, R. E. (1981) Coordination of mucociliary transport
in human trachea and intrapulmonary airways. J. Appl. Physiol. 51: 1057-1064.
Yeh, H. -C.; Schum, G. M. (1980) Models of human lung airways and their application to inhaled particle deposition.
Bull. Math. Biol. 42: 461-480.
Yeh, H. C.; Schum, G. M.; Duggan, M. T. (1979) Anatomic models of the tracheobronchial and pulmonary regions
of the rat. Anat. Rec. 195: 483-492.
Yu, G.; Zhang, Z.; Lessmann, R. (1998) Fluid flow and particle diffusion in the human upper respiratory system.
Aerosol Sci. Technol. 28: 146-158.
Zhang, Z.; Martonen, T. (1997) Deposition of ultrafme aerosols in human tracheobronchial airways. Inhalation
Toxicol. 9:99-110.
Zhang, L.; Asgharian, B.; Anjilvel, S. (1997) Inertia! deposition of particles in the human upper airway bifurcations.
Aerosol Sci. Tech. 26: 97-110.
Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2000) Effects of asymmetric branch flow rates on aerosol deposition in
bifurcating airways. J. Med. Eng. Technol. 24: 192-202.
Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2001) Effects of curved inlet tube on air flow and particle deposition in
bifurcating lung models. J. Biomechanics 34: 659-669.
Zhang, Z.; Kleinstreuer, C.; Kim, C. S. (2002) Cyclic micron-size particle inhalation and deposition in a triple
bifurcation lung airway model. J. Aerosol Sci. 33: 257-281.
Zhu, Y.; Hinds, W. C.; Kim, S.; Shen, S.; Sioutas, C. (2002) Study of ultrafine particles near a major highway with
heavy-duty diesel traffic. Atmos. Environ. 36: 4323-4335.
6-121
-------
7. TOXICOLOGY OF PARTICULATE MATTER IN
HUMANS AND LABORATORY ANIMALS
7.1 INTRODUCTION
The 1997 U.S. Particulate Matter National Ambient Air Quality Standards (PM NAAQS)
revisions (Federal Register, 1997) were based, in large part, on new epidemiologic evidence
showing associations between (a) ambient PM measured at community monitoring stations and
(b) increased risks for mortality and morbidity (especially cardiorespiratory-related) among
human populations exposed to contemporary U.S. ambient concentrations. However, very little
experimental toxicology data from controlled human or laboratory animal exposure studies were
available that provided more direct evidence supporting the plausibility of the observed
PM-mortality/morbidity associations being causal at the relatively low ambient PM
concentrations studied epidemiologically. The then-limited PM toxicologic data was assessed in
Chapter 11 of the 1996 PM Air Quality Criteria Document or PM AQCD (U.S. Environmental
Protection Agency, 1996a), which provided scientific assessment inputs supporting the 1997 PM
NAAQS decisions.
Since the 1996 PM AQCD, numerous hypotheses have been advanced and extensive new
toxicologic evidence generated with regard to possible pathophysiological mechanisms by
which PM exposures at ambient or near ambient concentrations might induce increased
morbidity and/or mortality. The extensive new PM toxicological research during the past five
years or so has focused mainly on addressing several interrelated questions, such as: (1) what
types of pathophysiological effects are exerted by ambient PM or constituent substances and
what are the potential mechanisms underlying them; (2) what PM characteristics (size, chemical
composition, etc.) cause or contribute to health effects; (3) what susceptible subgroups are at
increased risk for PM health effects and what factors contribute to increased susceptibility;
(4) what types of interactive effects of particles and gaseous co-pollutants have been
demonstrated; and (5) are there toxicologic findings on PM-related mutagenic/genotoxic effects
that support the plausibility of ambient PM-lung cancer relationships observed epidemiologically
in U.S. populations?
7-1
-------
7.1.1 Methodological Considerations
Various research approaches have been and continue to be used to address the above
questions, including studies of human volunteers exposed to PM under controlled conditions;
in vivo studies of laboratory animals including nonhuman primates, dogs, and rodent species;
and in vitro studies of tissue, cellular, genetic, and biochemical systems. A wide variety of
exposure conditions have been employed, including: whole body, mouth-only, and nose-only
inhalation exposures to concentrated ambient particles (CAPs) or laboratory-generated particles;
intratracheal, intrapulmonary, and intranasal instillation; and in vitro exposures to test materials
in solution or suspension. These research approaches have been targeted mainly to test
hypotheses to provide improved understanding of the role of PM in producing those types of
health effects identified by PM-related epidemiologic studies. Thus, many of the new
toxicological studies have been designed to address the question of biologic plausibility of
epidemiologically-demonstrated effects, rather than being explicitly aimed at providing
quantification of dose-response relationships for experimentally-induced toxic effects.
Reflecting this, most of the toxicology studies assessed here have generally used exposure
concentrations or doses that are relatively high compared to concentrations commonly observed
in ambient air. An important consideration contributing to the use of relatively high
experimental exposure concentrations is the fact that healthy animals have most typically been
used in many controlled-exposure toxicology studies, whereas epidemiologic findings often
reflect ambient pollutant effects on compromised humans (e.g., those with one or another
chronic disease) or other susceptible groups rendered at increased risk due to other factors.
Implicit in the use of relatively high concentrations in experimental studies of healthy subjects is
the assumption that increasing the dose somehow makes up for compromised tissue/organ
functions that may contribute to observed ambient PM effects. However, this may not be the
case, unless the increased susceptibility of an "at risk" group is based on enhanced respiratory
tract PM deposition/retention per se. In light of this, there exists a great need for expanded
development and use of animal models that more closely mimic important characteristics
contributing to increased human susceptibility to ambient PM effects; and some notable progress
has been made in this regard, as reflected by the growing number of PM toxicologic studies of
compromised animal models published since the 1996 PM AQCD and assessed in this chapter.
7-2
-------
Given the relatively high concentrations used, much care should therefore be taken when
attempting to interpret and extrapolate effects seen in these studies to provide insight into the
biological plausibility and mechanisms of action underlying effects seen in humans under "real
world" exposure conditions. Some of the responses might only be seen at the higher
concentrations more typical of occupational and experimental laboratory exposures and not
necessarily at (usually much lower) ambient particle exposure concentrations. However, the
high concentration studies play important roles in generating hypotheses and identifying
mechanisms, which can then be tested with more relevant low doses. On the other hand, it is
possible that differences between humans and rodents with regard to the inhalability, deposition,
clearance, and retention profiles for PM (see Chapter 6 for details) could conceivably make
doses to some specific respiratory tract tissues from experimental exposures relatively analogous
to doses resulting from human ambient exposures. To help place the lexicologically relevant
concentrations/doses into context in relation to ambient conditions, EPA has carried out some
illustrative dosimetric/extrapolation modeling analyses to provide comparisons between the high
doses typically used in toxicological studies and doses typical of human exposures under
ambient conditions. Building upon advances in dosimetric modeling discussed in Chapter 6,
these analyses compare PM doses delivered to human or rat lung tissue from experimental
exposures and PM doses to the human lung from exposures during normal activities. These
analyses and their interpretation of results (described in Appendix 7A) provide context for the
exposure concentrations used and results obtained in toxicological studies assessed here.
Additionally, it is important to keep in mind that the responses observed in toxicological studies,
e.g., inflammatory cell influx, can range from being a normal adaptive or physiological response
to a toxic response that is deleterious to the organism.
The effects of controlled exposures to ambient PM have been increasingly investigated
since the 1996 PM AQCD by use of particles collected from ambient samplers (e.g., impactors,
diffusion denuders, etc.) and, more recently, by the use of aerosol concentrators (e.g., Sioutas
et al., 1995a,b, 2000; Gordon et al., 1998; Chang et al., 2000, Kim et al., 2000a,b). In the first
type of study, particles from ambient air samplers are first collected on filters or other media,
then stored, and later resuspended in an aqueous medium for use in inhalation, intratracheal
instillation, or in vitro studies. Some ambient PM has been standardized as a reference material
and compared to existing dust and soot standards, e.g., standard materials from the National
-------
Institutes of Standards and Technology (NIST). Both ambient PM extracts and concentrated
ambient particles (CAPs) have been used to evaluate effects in healthy and compromised
laboratory animals and humans. Particle concentrators provide a technique for exposing animals
or humans by inhalation to concentrated ambient particles (CAPs) at levels higher than typical
ambient PM concentrations.
The development of particle concentrators has permitted the study of ambient real-world
particles under controlled conditions. This strength is offset somewhat by the inability of CAPs
studies to precisely control the mass concentration and day-to-day variability in ambient particle
composition, and they often lack detailed characterization of variations in chemical composition
from one CAPs exposure to another. Because the composition of concentrated ambient PM
varies across both time and location, a thorough physical-chemical characterization is necessary
to compare results between studies or even among exposures within studies in order to link
particle composition to effects. Two other limitations that should be taken into account in
interpreting results from CAPs studies are: (1) concentrators in use at the time of many of the
studies assessed here could not efficiently concentrate ambient particles < 0.1 jim, and
(2) gaseous components of PM were not concentrated. Thus, it is likely that a large portion of
potentially important combustion-generated particles (e.g, from diesel, gasoline vehicle, wood
smoke, coal smoke, etc.) were present only at ambient (not higher concentrated) levels in most
or all of the CAPs studies assessed here. A new generation of concentrators is now available
that can capture both coarse and ultrafine CAPs, allowing more broadly relevant exposures to
be tested. However, the gaseous components are not captured or proportionally concentrated,
so potential interactions in the ambient atmosphere are not fully recapitulated.
Controlled human and laboratory animal exposures to particulate material obtained from
emission source bag house filters or other emission source collection devices have also been
used extensively in recent years to evaluate the in vitro and in vivo respiratory toxicity of
complex combustion-related PM. Residual oil fly ash (ROFA) collected from large industrial
sources (e.g., oil-fired power plants) has been extensively used, as well as, to a lesser extent,
domestic oil furnace ash (DOFA) or coal fly ash (CFA). The major disadvantage associated with
the use of such emission source materials derives from questions concerning the potential
relevance of results obtained for helping to understand and interpret current ambient PM
exposure effects. There is little doubt that in years past, before the extensive implementation of
7-4
-------
air pollution controls, U.S. ambient air PM contained mixtures of high concentrations of
chemical species analogous to those found in many of the emission-source samples used in
toxicologic studies during the past decade or so. It is rare, however, that high concentrations of
materials that typify such samples would be found in ambient air PM samples obtained at
community monitoring sites in the United States, Canada, and much of Western Europe, which
provided aerometric data collected during the past 20 to 30 years that were used to estimate PM
exposures in most of the PM epidemiology studies assessed in this document. For example, very
high concentrations of metals typify most ROFA samples (especially extremely high nickel and
vanadium levels), and experimental exposures to such materials have generally resulted in
exposures/doses that are orders of magnitude (100s of times) higher than would be associated
with exposures to much lower levels of such metals in ambient PM measured routinely since the
1970s at community monitoring sites across the U.S. (except perhaps at times very near some
sources without modern emission control devices or during temporary breakdowns of such).
Thus, significant issues arise concerning the extent to which effects of ROFA or other high
concentration emission source can be extrapolated to aid in interpretation of ambient air PM
exposure effects on humans.
Analogous issues arise in connection with evaluation of the toxicity of PM obtained as
emission products from mobile source combustion devices, e.g., diesel and gasoline vehicle
engines. The complex combustion-related mixtures in such mobile source emissions include
many different types of particles and gaseous compounds in high concentrations which are not
necessarily representative of ambient PM derived from such sources after passage through
particle traps, catalytic converters, exhaust pipes, etc. For example, ultrafine particles emitted
from gasoline and diesel engines are reduced in numbers and concentrations as they agglomerate
to form larger, accumulation-mode particles as they cool in passing through exhaust systems
and/or as they undergo further physical and chemical transformation as they "age" as ambient air
components. Further complicating evaluation of the toxicity of mobile source emission
components is: (1) the difficulty in separating out toxic effects attributable to particles versus
those of gaseous components in automotive exhausts; and (2) the changing nature of those
exhaust mixes as a function of variations in engine operating mode (e.g., cold start versus warm
start or "light" versus "heavy" load operation, etc.) and changes in engine technology (e.g., "old
diesels" versus "new diesels").
7-5
-------
The in vivo studies discussed first and the in vitro studies discussed later have almost
exclusively used PM10 or PM2 5 as particle size cutoffs for studying the adverse effects of
ambient PM. Collection of these size fractions has been made easier by widespread availability
of ambient sampling equipment for PM10 and PM25. However, the study of other important size
fractions, such as the coarse fraction (PM10_2 5) and PMX 0 has been largely ignored, and only
limited toxicology data are now available by which to assess effects of these potentially
important particle sizes. Similarly, although organic compounds often comprise 20 to 70% of
the dry fine particle mass of ambient PM (see Chapter 3), little research has addressed
mechanisms by which such compounds may contribute to ambient PM-related effects. One
exception to this has been the evaluation of contributions of diverse organic compounds to
mutagenic, genotoxic, and carcinogenic effects discussed later in Section 7.8.
7.1.2 Organization of the Chapter
Ambient PM, as noted above, is comprised of myriad physical and chemical species that
can vary widely from one geographic location to another or even from one time to another time
at a given location. It is not surprising that only a relatively few ambient air mixes from selected
urban areas or subsets or combinations of the diverse variety of physical/chemical species have
been investigated in controlled human or laboratory animal studies. However, a full discussion
of all types of ambient particles that have been identified (see Chapter 2) is beyond the scope of
this chapter. Thus, specific criteria were used to select topics for presentation. High priority was
placed on studies that (a) may contribute to enhanced understanding of ambient PM
epidemiologic study results and/or (b) elucidate mechanisms of health effects of ambient PM or
its major common constituents. Diesel particulate matter (DPM) generally fits the above
criteria; however, because it is discussed in great detail in other documents (Health Effects
Institute, 1995; U.S. Environmental Protection Agency, 2002), only some aspects are discussed
to a limited extent in this chapter. Individual particle species with high inherent toxicity that are
of concern mostly because of occupational exposure (e.g., silica) that are discussed in detail in
other documents and reports (e.g., U.S. Environmental Protection Agency, 1996b; Gift and
Faust, 1997 for silica) are also not assessed in detail in this chapter.
Because of the sparsity of toxicological data on ambient PM at the time of the 1996 PM
AQCD (U.S. Environmental Protection Agency, 1996a), the discussion of toxicologic effects
7-6
-------
of PM was organized there into specific chemical components of ambient PM or "surrogate"
particles (e.g., acid aerosols, metals, ultrafme particles, bioaerosols, "other particle matter").
Many of the newer toxicological studies evaluate potential toxic effects of combustion-related
particles. The main reason for this extensive current interest in combustion particles is that these
particles, along with materials adsorbed to such particles and secondary aerosols formed from
them, are typically among the most dominant components represented in the fine fraction of
ambient air PM found in most U.S. urban areas.
This chapter is organized as follows. First, cardiovascular and systemic effects of
in vivo PM exposure are discussed in Section 7.2. Next, Section 7.3 discusses respiratory effects
of ambient PM, specific components of ambient PM, combustion source-related particle mixes,
or other laboratory-generated particles delivered by controlled in vivo exposures of humans or
laboratory animals (note that the specific components discussion includes summary points drawn
from detailed discussion of ambient bioaerosols in Appendix 7B). In vitro exposure studies are
then next discussed (Section 7.4) and are useful in providing information on potential hazardous
constituents and mechanisms of PM injury. The next section (Section 7.5) focuses on studies of
PM effects in laboratory animal models meant to mimic human disease, as a means for providing
information useful in characterizing factors affecting susceptibility to PM cardiovascular and
respiratory system effects. Section 7.6 then assesses controlled-exposure studies evaluating
health effects of mixtures of ambient PM or specific PM constituents with gaseous pollutants.
Section 7.7 discusses exposure/dose-effects relationships for cardiovascular and respiratory
effects and comparisons for illustrative health endpoints of experimental exposures/data needed
to produce similar effects across species (rats, humans) and/or under ambient conditions
(drawing upon extrapolation modeling results presented in Appendix 7A). The ensuing Section
(7.8) discusses studies of PM-related mutagenic/genotoxic effects related to evaluation of the
relative carcinogenic potential of ambient PM and its constituents, as well as particulate
constituents in emissions from various types of combustion sources. Section 7.9 then discusses
important new cross-cutting information on airborne particles as carriers of other toxic agents.
This organization provides the underlying information used for interpretive summarization
(in Section 7.10) of the extensive new findings discussed in the earlier sections with regard to
PM-related effects, all of which may individually contribute to and/or combine through intricate
interlinkages to mediate ambient PM exposure effects.
7-7
-------
7.2 CARDIOVASCULAR AND SYSTEMIC EFFECTS OF IN VIVO PM
EXPOSURES IN HUMANS AND LABORATORY ANIMALS
7.2.1 Introduction
A growing number of epidemiology studies are finding (a) associations between
ambient PM and increases in cardiac-related deaths and/or morbidity indicators and (b) that the
risk of PM-related cardiac effects may be as great or greater than those attributed to respiratory
causes (see Chapter 8). Both acute and chronic PM exposures have been implicated in the
observed cardiovascular morbidity and mortality effects. These effects appear to be induced via
direct particle uptake into the blood and/or via mediation by the nervous system. Figure 7-1
schematically illustrates hypothesized mechanisms thought to be involved in cardiovascular
responses to PM exposure. Such effects may be especially deleterious to individuals
compromised by disease states such as ischemic heart disease, cardiac arrythmias, and COPD.
As shown in Figure 7-1, the heart receives both parasympathetic and sympathetic inputs,
which serve to decrease or increase heart rate, respectively. Vasoconstriction, possibly due to
release of endothelin elicited by PM, could cause increased blood pressure (which is detected by
baroreceptors). Parasympathetic neural input may then be increased to the heart, slowing heart
rate and decreasing cardiac output (which is sensed by aortic and carotid chemoreceptors).
These, in turn, may cause a sympathetic response, manifested by increased heart rate and
contractile force, thus increasing cardiac output. This arrhythmogenesis and altered cardiac
output in either direction can be life-threatening to susceptible individuals. Pathophysiological
changes in cardiac function can be detected by electrocardiographic recordings, with certain
ECG parameters (e.g., heart rate variability or HRV) recently gaining widening use as indicators
of PM-induced cardiac effects.
Heart rate variability (HRV), a measure of the beat-to-beat change in heart rate, is a
reflection of the overall autonomic control of the heart. HRV has been used for many years as a
research tool to study cardiovascular physiology and pharmacology. Its role as a clinical
predictor of outcome for populations with heart disease has been extensively studied. HRV can
be divided into time and frequency measures. Frequency measures of variability are more
commonly used for mechanistic studies because they resolve parasympathetic and sympathetic
influences on the heart better than do time domain measurements. It has been well established
that the frequency analysis of heart rate variability is a robust method for measuring the
-------
CMS
^
Problematic for
Disease States:
Cardiac Arrythmias,
COPD, etc.
——A
Sympathetic
^ I ANS
/* Parasympathefc
/. ANS
Chemoreceptors . i
/ 4(1
T C-Reactive Proteins
Cytokines
Figure 7-1. Schematic illustration of hypothesized pathways/mechanisms potentially
underlying the cardiovascular effects of PM.
7-9
-------
autonomic modulation of heart rate. Under certain circumstances, HRV provides insight into
sympathetic nervous activity, but more commonly it is a very good measurement of
parasympathetic modulation. For prognostication in heart disease, both the time and frequency
domain measures of heart rate variability seem equivalent in predicting events. Heart rate
variability can be used to judge the relative influences of sympathetic and parasympathetic
forces on the heart, as such short-term spectral parameters (i.e., measures averaged over five
minute intervals) can vary as much as 4-fold during the course of a 1-h period (Kleiger et al.,
1991). Despite the inherent variability of short-term HRV measures during routine daily
activity, long-term measures (i.e., measures averaged over 24 h) show excellent day-to-day
reproducibility. Given this inherent variability in the minute-to-minute spectral measurements,
great care is required in the experimental design of studies utilizing HRV techniques and
interpretation of HRV results. When appropriately designed and carefully interpreted, studies
utilizing measures of HRV provide insight into the relationship between the perturbation of the
internal or external environment and subsequent changes in the modulation of autonomic neural
input to the heart.
Heart rate variability has been studied in multiple settings, using different parameters (both
time and frequency domain) to determine prognosis in populations. This has been studied most
frequently in coronary artery disease populations, particularly in the post-myocardial infarction
(post-Mi) population. Most reports have dichotomized the study group by HRV parameters and
then compared outcomes. To summarize those results, lower time domain as well as frequency
domain variables are associated with an increase in cardiac and all-cause mortality. Those
variables most closely correlated with parasympathetic tone appear to have the strongest
predictive value in heart disease populations. Specifically, acute changes in RR-variability
temporally precede and are predictive of increased long-term risk for the occurrence of ischemic
sudden death and/or precipitating ventricular arrhythmias in individuals with established heart
disease (see for example La Rovere et al., 2003). However, acute changes in HRV parameters
do not necessarily occur immediately prior to sudden fatal ventricular arrhythmias (Waxman
et al., 1994). The heart rate variability itself is not the causative agent, nor has it been implied to
be a causative agent in any of the studies performed to date. Altered HRV, including changes in
HRV associated with exposure to PM, is simply a marker for enhanced risk.
7-10
-------
Another route by which PM could exert deleterious cardiovascular effects may involve
ambient PM effects on endothelial function. In particular, as hypothesized by Seaton et al.
(1995), PM exposure could affect blood coagulation, possibly through endothelial injury that
results in platelet activation. This then could initiate a cascade of effects, e.g., increased
fibrinogen and fibrin formation, leading to increased formation of clots. Figure 7-2 (from
Nadziejko, et al., 2002) nicely illustrates physiological events (and applicable timeframes)
involved in the blood clotting cascade, as well as denoting important substances released at
successive steps which, in turn, stimulate the next step in the clotting cascade and, ultimately,
trigger clot lysing events that normally terminate the cascade. Various studies have measured
such substances as a means to evaluate possible PM-induced effects on blood coagulation.
Another significant effect of PM exposure could be vascular inflammation, which induces
release of C-reactive proteins and cytokines. These cause further inflammatory responses that,
on a chronic basis, can lead to atherosclerosis. In narrowed coronary arteries, the clots formed in
the aforementioned cascade may block blood flow, resulting in acute myocardial infarction.
Nadziejko et al. (2002) further note that small prothrombotic changes in blood coagulation
parameters in a large population can have substantial effects on the incidence and prevalence of
cardiovascular disease events (Di Minno and Mancini, 1990; Braunwald, 1997; Lowe et al.,
1997). In particular, altered coagulation can increase heart attack risk through formation of clots
on atherosclerotic plaques in coronary arteries that cut off blood supply to the myocardium or
induce ischemic strokes via clots forming or lodging in the carotid arteries and blocking blood
flow to cerebral arteries and brain tissue. Also, Nadziejko et al. (2002) note that (a) evidence
exists for formation of small thrombi being common in persons with atherosclerosis (Meade
et al., 1993) and (b) whether such thrombi lead to more serious effects (heart attack, stroke)
depends in part on the balance between thrombogenic factors underlying blood clot formation
and fibrinolytic factors that lyse clots. Also, they note that effects of small changes in
coagulation on heart attack risk are reflected by the risk of sudden cardiac death being 70%
higher between 6:00 a.m. and 9:00 a.m. than the average risk for the rest of the day (Willich,
et al., 1987), likely due in part to the circadian rhythm of fibrinolytic factors that are at their
lowest levels in the early morning (Andrews et al., 1996). Also, as stated by Nadziejko et al.
(2002), sympathetic nervous system activity is increased by standing up after lying prone
(Tofler, et al., 1987; Andrews et al., 1996), and increased sympathetic activity causes
7-11
-------
Endothelial Injury
Platelet Activation
Binding of fibrinogen by platelets and formation of platelet plug
Fibrinogen
Factor VII
Fibrin
Reinforcement of the thrombus by fibrin strands
Downregulation of coagulation and initiation of clot lysis
Tissue Plasmmogen Activator
Plasminogen Activator Inhibitor
Seconds
Minutes
Hours
Event
Time Frame
Figure 7-2. Simplified overview of blood coagulation system. The coagulation parameters
often measured in the study of PM effects on blood coagulation are indicated
by bold type. The relations of these selected parameters with the rest of the
coagulation system are outlined.
Source: Nadziejko et al. (2002).
prothrombotic changes in blood coagulation parameters such that even small, homeostatic
modulations of coagulation within a normal range could translate into significant increased risk
for heart attack.
Thus, potentially dangerous alterations in cardiovascular functions due to PM exposures
could be signaled by even small PM-related (a) changes in blood coagulation cascade indicators,
e.g., increased blood platelet, fibrinogen, or Factor VII, or decreased tissue plasma activator
7-12
-------
(TPA) levels; (b) increased C-reactive protein, or cytokines contributing to increased
atherosclerosis plaque formation and/or blood coagulation; c) increased blood pressure; and/or
(d) certain alterations in heart rate, heart rate variability, or other ECG indicators indicative of
deleterious shifts in parasympathetic/sympathetic neural inputs to the heart or other underlying
cardiac pathophysiology.
Another cardiovascular-related effect of PM exposure could be plasma extravasation from
post-capillary venules. The mechanisms by which this occurs are thought to include the release
of peptides such as neurokinin A, substance P, and calcitonin-gene-related peptide from
unmyelinated sensory nerves, near to or on the blood vessels. These peptides bind to receptors
on the endothelial cells of vessels and create gaps, allowing leakage of plasma, which is one
component of neurogenic inflammation (Piedimonte et al., 1992; Baluk et al., 1992).
There were few studies assessed in the 1996 PM AQCD that evaluated cardiovascular
system effects of PM exposures. Since 1996, numerous studies have now become available that
evaluated cardiovascular effects of exposures (via inhalation or instillation) of ambient PM,
constituent components, complex mixtures from PM emission sources and/or exposures to single
PM substances or binary/ternary combinations of particles of varying chemical composition.
Also, whereas earlier studies tended to focus on healthy animals, more recent studies have, in
addition, begun to focus on evaluation of PM effects in animal models of disease states thought
to mimic aspects of pathophysiologic states experienced by compromised humans at increased
risk for PM effects.
The toxicological consequences of inhaled particles on the cardiovascular system had not
been extensively investigated prior to 1996. Since then, Costa and colleagues (e.g., Costa and
Dreher, 1997) have demonstrated that intratracheal instillation of high levels of ambient particles
can increase or accelerate death in an animal model of cardiorespiratory disease induced by
monocrotaline (MCT) administration in rats. These deaths did not occur with all types of
ambient particles tested. Some dusts, such as volcanic ash from Mount Saint Helens, were
relatively inert; whereas other ambient dusts, including those from urban sites, were toxic.
These early observations suggested that particle composition plays an important role in the
adverse health effects associated with episodic exposure to ambient PM, despite an apparent
"general particle" effect that seemed to be implied by somewhat similar epidemiologic
observations of ambient PM exposure associations with increased mortality and morbidity in
7-13
-------
many regions of the United States with varying particle composition. Studies evaluating
possible increased susceptibility to the adverse effects of PM in compromised animal models of
human pathophysiology provide a potentially important link to epidemiologic observations and
are among those discussed below.
Muggenburg et al. (2000a) has described several potential animal models of cardiac
disease (MCT-induced pulmonary hypertension, dilated cardiomyopathy, viral and mycoplasmal
myocarditis, and ischemic heart disease) and discussed advantages and disadvantages associated
with the use of animal models to study cardiac disease and air pollution. The first type of animal
model has probably been most extensively used in recent years. Pulmonary hypertension in
humans may result from airway and vascular effects due to COPD, asthma, and cystic fibrosis.
The MCT-induced vascular disease model exhibits common features of progressive pulmonary
hypertension in humans. The injury effects include selective pulmonary endothelial damage and
progressive pulmonary arterial muscularization. Pulmonary hypertension develops, the blood
flow is impeded, and compensatory right ventricular hypertrophy occurs. To produce pulmonary
hypertension, animals are injected subcutaneously with 50-60 mg/kg monocrotaline (MCT).
Within two weeks after treatment, experimental animals, primarily rats, develop pulmonary
hypertension (Kodavanti et al., 1998a).
A growing number of studies have used extracts of collected/stored ambient PM or real-
time generated concentrated ambient particles (CAPs) drawn from various airsheds (e.g., Boston,
New York City, etc.) to evaluate cardiovascular and other systemic effects of PM. Numerous
other new animal studies have also used metal-associated ROFA as one type of combustion
source particle mix; and others have used other types of combustion source materials, e.g.,
domestic oil fly ash (DOFA), coal fly ash (CFA), or diesel exhaust (DE). The following
discussion of cardiovascular/systemic effects of PM first focuses mainly on the ambient PM
studies and then discusses findings from the studies using ROFA and other types of particles.
Tables 7-1 and 7-2 summarize newly-available studies (since the 1996 PM AQCD) that
evaluated cardiovascular effects of ambient PM mixtures or other types of PM in response to
controlled inhalation exposures of humans or laboratory animals. Intratracheal instillation
studies are then summarized in Table 7-3, and in vitro exposure studies of cardiovascular effects
are discussed in Section 7.4.
7-14
-------
TABLE 7-1. CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INHALED AMBIENT PARTICIPATE MATTER
Species, Gender,
Strain Age, or
Body Weight
Humans, healthy
nonsmokers,
18-40 years old
n = 38
Humans, healthy
18-40 years old
n = 4
Humans, healthy;
19-41 years old
n = 4
Humans, healthy
(n = 12) and
asthmatic (n = 12)
18-45 years old,
nonsmoking
Dogs, female
mongrel,
14 to 17 kg
Dogs, mongrel;
Balloon-occluded
LAD coronary
artery in some,
n= 14
Particle"
CAPs
(Chapel Hill)
CAPs
(Toronto)
CAPs
(Los
Angeles)
CAPs
(Los
Angeles)
CAPs
(Boston)
CAPs
(Boston)
Exposure Mass
Technique Concentration
Inhalation 23.1 to
311.1 ug/m3
Inhalation 24 to
(face mask) 124 ug/m3
Inhalation 148 to
246 ug/m3
Inhalation 99-224 ug/m3
whole body (mean 174)
chamber
Inhalation via 3-360 ug/m3
tracheostomy
Inhalation via ~1 00- 1 000 ug/m3
tracheostomy
Exposure
Duration, PE"
Particle Size Time to Analysis
0.65 um 2 h, analysis
og = 2.35 at!8h
0.1 -2.5 um
PM25 2 h
80% 0.1 to 2 h with
2.5 um alternating
exercise/rest.
Analysis at 0,
4, and 22 h PE.
0.2 to 0.3 um 6 h/day for
3 days
0.23 to 6 h/day for
0.34 um 3 days
og = 0.2 to 2.9
Cardiovascular Effects
Increased blood fibrinogen with CAPs
exposure. PM concentration in chamber varied
with ambient air PM level. Estimated total dose
of 1200 ug.
Trend toward increased fibrinogen levels 2 h
post high CAPs (124 ug/m3) exposure, but stat.
sig. not specified. Also, no sig. ECG Holier
effects.
No significant changes in in arterial O2 saturation
or Holier ECGs observed, nor in lung function or
symptoms. The maximum steady stale fine
particle concenlralion in Ihe brealhing zone was
typically seven limes Ihe ambienl concenlralion.
CAPs-relaled decrease in Factor VII blood
levels, - bul no significanl changes in blood
fibrinogen or serum amyloid A wilh CAPs. Bolh
heallhy and aslhmalic subjecls had modesl
increases in HR variability and significanl
increases in HR during exercise. Some reported
cardiac symptoms (fainlness, dizziness, pain
related to heart, elc) during CAPs exposure.
Peripheral blood parameters were related to
specific particle consliluenls. Factor analysis
from paired and crossover experimenls showed
lhal hemalologic changes were nol associated
wilh increases in total CAP mass concenlralion.
Decreased heart and respiratory rale and
increased lavage fluid neulrophils in normal
dogs. Decreased lime to ST segmenl elevation
and increased magnilude in compromised
dogs. PM concenlralion varied depending on
ambienl PM level and concenlralor operation.
No dose-response relationship evident
Reference
Ohio el al.
(2000a)
Pelrovic el al.
(2000)
Gong el al.
(2000)
Gong el al.
(2003)
Clarke elal.
(2000a)
Godleski el al.
(2000)
-------
TABLE 7-1 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INHALED AMBIENT
PARTICIPATE MATTER
Species, Gender,
Strain Age, or
Body Weight
Rats
Rats, male,
F-344,
MCT-treated
Hamsters,
6-8 months old;
Bio TO-2
Rats, male, F344,
250-275 g
Rats, Wistar
Humans, male,
healthy, age
19-24 years
Exposure Mass
Particle" Technique Concentration
CAPs Inhalation 11 0-3 50 ug/m3
(Tuxedo, (nose-only)
NY)
CAPs Inhalation 132-919 ug/m3
(Manhattan)
CAPs Inhalation 95-341 ug/m3
(NYC) (nose-only)
Ott ambient Inhalation 48 mg/m3
(EHC-93) (nose only)
(ECH-93L) 49 mg/m3
Diesel soot
(DPM) 5 mg/m3
Carbon black
(CB) 5 mg/m3
PM10 from Inhalation -125 ug/m3
S.E. Asian (range = 47 to
Smoke Haze 216)
Exposure
Duration, PE"
Particle Size Time to Analysis
N/A 3h
0.2- 1.2 urn Single 3 h or
og = 0.2-3.9 3 daily 6 h
exposures
< 2. 5 urn 0,12, and 24 h
36, 56, 80, 4 h, Analyses at 2,
100, and 32, 36, 48 h PE
300 urn
N/A 4 weeks, analysis
at 3 and 5 weeks
PE
Cardiovascular Effects
Small but consistent increase in HR; increased
peripheral blood neutrophils and decreased
lymphocytes. No pulmonary injury found.
Concentration to chamber varied from 132 to
199 ug/m3.
No increase in cardiac arrhythmias;
inconsistent PM-associated increases in HR,
blood cell differential counts, and atrial
conduction time of rats. No adverse cardiac
or pulmonary effects in hamsters.
No consistent exposure-related effects on platelet
count fibrinogen level, factor VII activity,
thrombin-anti-thrombin complex, tissue
plasminogen activator, or plasminogen activator
inhibitor.
EHC-93 elevated blood pressure on day 2, ET-1
levels at 32 h, and ET-3 levels at 2, 32, and 48 h
postexposure. EHC-93 L had no effect on blood
pressure, transient effect on ET-1, -2, -3 levels at
2 h but not 32 h postexposure. DPM had no
effect on blood pressure, but elevated ET-3
levels at 36 h PE. CB had no effect.
Band cell counts were significantly increased
during the haze period
Reference
Gordon et al.
(1998)
Gordon etal.
(2000)
Nadziejko
et al. (2002)
Vincent etal.
(2001)
Tan etal. (2000)
a CAPs = concentrated ambient particles
UAP = urban ambient partciles
DPM = diesel particulate matter
Ott ambient = resuspended UAP from Ottawa, CA
b PE = postexposure
-------
TABLE 7-2. CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INHALED ROFA AND OTHER COMBUSTION-
RELATED PARTICULATE MATTER
Species, Gender,
Strain Age, or
Body Weight
Dogs, beagles,
10.5 years old,
healthy, n = 4
Rats, S-D,
MCT-treated,
250 g
Rats, S-D, SH rats,
WKY rats, healthy
and MCT-treated
Rats, male WKY
and SH, 12 to
13 weeks old
Rats, male, SH
and WKY; 12 to
13 weeks old
Rats, male, S-D,
WKY and SH
Exposure
Particle" Technique
ROFA Oral
(Boston) inhalation
ROFA Inhalation
(Boston)
ROFA Inhalation
(location not
given)
ROFA Nose-only
(Florida) inhalation
ROFA Inhalation
(Boston)
ROFA Inhalation
(Boston) (nose only)
Exposure
Mass Duration, PE b
Concentration Particle Size Time to Analysis
3 mg/m3 2.22 urn 3 h/day for
MMAD 3 days
og = 2.71
580 ± 1 10 ug/m3 2.06 urn 6 h/day for 3 days
MMAD
og=1.57
15 mg/m3 1.95 um 6 h/day for 3 days
MMAD
1 5 mg/m3 N/A 6 h/day for 3 days
15 mg/m3 1.5 um 6 h/day,
og = 1 .5 3 days/week for 1,
2, or 4 weeks
2, 5, 10 mg/m3 6 h/day for
4 consec. days
1 0 mg/m3 6 h/day
1 day/week,
4 or 16 weeks
Cardiovascular Effects
No consistent changes in ST segment, the form
or amplitude of the T wave, or arrhythmias;
slight bradycardia during exposure.
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.
Pulmonary hypertensive (MCT-treated S-D) and
spontaneously hypertensive (SH) rats exposed to
ROFA by inhalation showed similar effects, but
of diminished amplitude. There were no
lethalities by the inhalation route.
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.
One week exposure increased plasma fibrinogen
in SH rats only; longer (2 or 4 week) exposure
caused pulmonary injury but no changes in
fibrinogen.
No cardiovascular effects see in SD or SH rats
with acute or chronic exposure. Cardiac lesions
(active chronic inflammatory, multifocal
myocardial degeneration, fibrosis, decreases in
number of granulated mast cells) seen for WKY
rats with chronic (16 week) exposures.
Reference
Muggenburg
et al. (2000b)
Killingsworth
etal. (1997)
Watkinson etal.
(2000a,b)
Kodavanti et al.
(2000a)
Kodavanti et al.
(2002a)
Kodavanti et al.
(2003)
-------
oo
TABLE 7-2 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INHALED ROFA AND OTHER
COMBUSTION-RELATED PARTICULATE MATTER
Species, Gender,
Strain Age, or
Body Weight
Rats, male, S-D,
healthy and MI
Human, healthy,
nonsmoking males
Rats, SD, 60 days
old
Rats, SH
Exposure Mass
Particle" Technique Concentration
ROFA Inhalation 3 mg/m3
(Boston)
Carbon black
Ultrafme Mouthpiece 10 ug/m3
carbon exposure
particles
VSO4 Inhalation 0.3 -2.4 mg/m3
NiS04
DE Whole body 30,100, 300, or
1000 ug/m3
Exposure
Duration, PE"
Particle Size Time to Analysis
1.81 urn 1 h
0.95 urn
<100nm 2 h exposure;
assessment before
and immediately,
3. 5 h, and 21 h
after exposure
N/A 6h/day x 4 days
90% < 1 urn 6 h day,
7 day/week for
1 week
Cardiovascular Effects
In MI group, with thermocoagulation of left
coronary artery, 41% of rats had one or more
premature ventricular complexes (PVCs).
ROFA but not CB or room air, increased
arrhythmia frequency in those with PVCs and
decreased heart rate variability.
No effects on blood coagulability, circulating
leukocyte activation, leukocyte expression of
activation and adhesion molecules.
No effects with V at all doses. Ni caused
delayed bradycardia, hypothermia and
arrhythmogenesis at >1.2 mg/m3. V+Ni
produced delayed effects at 0.5 mg/m3 and
potentiated responses at 1 .3 mg/m3, suggesting
synergism of effect.
Elevated daytime HR; concentration-dependent
prolongation of PQ interval.
Reference
Wellenius et al.
(2002)
Frampton(2001)
Campen et al.
(2001)
Campen et al.
(2003)
"ROFA = Residual oil fly ash
NiSO2 = Nickel sulfate
Fe2(SO4)3 = Iron sulfate
DE = diesel exhaust
VSO4 = Vanadium sulfate
MCT = monocrotaline
MI = Myocardial infarction
bPE = Post Exposure
-------
TABLE 7-3. CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INSTILLED ROFA AND OTHER
PARTICIPATE MATTER
VO
Species, Gender,
Strain Age, or
Body Weight
Rats, male, S-D,
60 days old, healthy
and MCT-treated
Rats, male, S-D,
60 days old, healthy
and MCT-treated,
n = 64
Rats, male,
SD, 60 days old,
healthy and
MCT-treated
MCT-treated
Rats, male, S-D;
60 days old
Particle"
ROFA
DOFA
CFA
Ambient PM
(St. Louis
Dusseldorf
Ottawa,
Wash. DC)
ROFA
(location not
given)
Fe2(S04)3
NiS04
VS04
Fe2(S04)3
+VS04
Fe2(S04)3 +
NiS04
NiS04 + VS04
VS04 +
Fe2(S04)3 +
NiSO4
ROFA (Florida)
MSH Vol. Ash
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Dose
Total mass:
2.5 mg/rat
Total transition
metal:
46 ug/rat
0.0,0.25, 1.0,
and 2. 5 mg/rat
105 ug
263 ug
245 ug
105 ug
245 ug
105 ug
263 ug
263 ug
245 ug
245 ug
105 ug
263 ug
0.3, 1.7, or
8.3 mg/kg
PE"Time
Particle Size to Analysis
Emission PM: Analysis at
1.78-4.17um 24 and 96 h
following
instillation
Ambient PM:
3.27-4.09 um
1 .95 um Analysis at
96 h
postexposure
Analysis at
96 h
postexposure
1 .95 um Analysis at
og = 2.19 24 h
Cardiovascular Effects
ROFA alone induced some mild arrhythmias;
MCT-ROFA showed enhanced neutrophilic
inflammation.
MCT-ROFA animals showed more numerous
and severe arrhythmias including S-T segment
inversions and A-V block.
Dose-related hypothermia and bradycardia in
healthy rats, potentiated by compromised
models at 2.5 mg dose.
V caused bradycardia, arrhythmogenesis and
hypothermia immediately. Ni caused delayed
bradycardia, arrhythmogenesis and
hypothermia. Fe had little effect.
Ni exacerbated the immediate effects of V.
Fe attenuated them.
Increased plasma fibrinogen only at 8.3 mg/kg
ROFA.
Reference
Costa and Dreher
(1997)
Campen et al.
(2000)
Campen et al.
(2002)
Gardner et al.
(2000)
-------
TABLE 7-3 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INSTILLED ROFA AND OTHER
PARTICIPATE MATTER
to
o
Species, Gender,
Strain Age, or
Body Weight
Rats, male, SD,
60 days old
Rats, male, S-D,
MCT-treated
Rat, SD, 60 d old;
250-300 g healthy or
MCT-treated
Rats, male SH
andWKY; 12-13
weeks old
(l)Rats, S-D healthy
and MCT,
cold-stressed, and
ozone-treated
Particle"
ROFA (Boston)
classified by
soluble metals
(As, Be, Cd,
Co, Cr, Cu, Fe,
Mn, Ni, Pb, V,
Zn, and sulfate)
ROFA
(Florida)
ROFA (Florida)
ROFA
(Boston)
ROFA
(location not
given)
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
instillation
Dose Particle Size
0.83, 3.3 or <3.0\im
8.3 mg/kg MADD
0.25, 1.0, or 1.95 urn
2.5 mg in MMAD
0.3 mL saline og = 2.19
0.83 or 1.95 urn
3.33 mg/kg MMAD,
og = 2.19
1 and 5 mg/kg 1.5 urn
og=1.5
0.0,0.25,1.0, 1.95 urn
or 2.5 mg/rat og = 2.19
PE"Time
to Analysis
Anaylsis at
24 h
postexposure
Monitored for
96 h after
instillation of
ROFA
particles
Analysis at
24 and 96 h
postexposure
Analysis at 1,
2, and 4 days
Monitored for
96 h after
instillation
Cardiovascular Effects
Dose-dependent increase in BAL protein, LDH,
hemoglobin and NAG activity (only high dose
data shown). ROFA containing highest
concentration of water-leachable Fe, V, and Ni
or V and Ni caused largest increase. ROFA
with highest V content induced greatest
increase in BAL neutrophils. AM
chemiluminescence was greatest with ROFA
containing primarily soluble V and less with
Ni + V.
Dose-related increases in incidence and
duration of serious arrhythmic events in normal
rats. Incidence and severity of arrythmias
increased greatly in MCT rats. Changes
occurred at all doses ranging from modest
effects at the lowest to more serious
disturbances at the higher doses. Deaths seen at
each instillation level in MCT rats only
(6/12 died after MCT + ROFA).
Increases in BAL markers of lung injury and
inflammation; 58% of MCT rats exposed to
ROFA died by 96 h regardless of the dose.
ROFA increased plasma fibrinogen and
decreased peripheral lymphocytes in both SH
and WKY rats at 5.0 mg/kg dose.
(1) Healthy rats exposed IT to ROFA
demonstrated dose-related hypothermia,
bradycardia, and increased arrhythmias at
2.5 mg dose. Similar response pattern seen at
Reference
Kodavanti etal.
(1998a)
Watkinson
etal. (1998)
Kodavanti etal.
(1999)
Kodavanti
etal. (2002a)
Watkinson et al.
(2000a,b);
Watkinson et al.
(2001)
0.25 and 1.0 mg, but reduced in magnitude and
duration. Compromised rats showed
exaggerated hypothermia and cardiac responses
to IT ROFA at all doses. Mortality was seen
only in the MCT-treated rats exposed to ROFA
by IT.
-------
TABLE 7-3 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INSTILLED ROFA AND OTHER
PARTICIPATE MATTER
to
Species, Gender,
Strain Age, or
Body Weight
(2) Rats, SH,
1 5 months old
(3) Rats, S-D
MCT-treated
Rats, Wistar, male,
200-250g,
healthy and ozone-
treated
Rabbits, female,
New Zealand,
2.2 to 3.0 kg
Rabbits, female,
Watanabe heritable
hyperlipidemic
3.2 ±0.1 kg
Rabbits, female,
New Zealand White,
1.8 to 2.4 kg
Particle"
OTT
ROFA
MSH
Fe2(S04)3
VS04
NiSO2
Ottawa EHC-93
OTT PM10
(EHC-93)
OTT PM10
(EHC-93)
Colloidal
carbon
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Instillation
Intrapharyngeal
instillation
Intrapharyngeal
instillation
Instillation
Dose Particle Size
2.5mg
0.5 mg
2.5 mg
105 ug
245 ug
262.5 ug
0.5, 1.5 or 0.5 um
5 mg/rat in
0.3 mL saline
5 mg/dose 4-5 um mass
median
diameter
5 mg in 1 mL 0.8 ± 0.4 um
saline
2 mL of 1% < 1 um
colloidal carbon
(20 mg)
PE"Time
to Analysis
Analysis at
2,4, or 7 days
after exposure
5 mg 2 times
per week for
3 weeks
5 mg 2 times
per week for
4 weeks
Examined at
24 to 192 h
after
instillation
Cardiovascular Effects
(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 seen with MSH exposure.
(3) Ni and V showed the greatest toxicity;
Fe-exposed rats did not differ from controls.
At high doses, 20% increase in plasma
fibrinogen at 2 days PE correlated with
increases in ET-1 and iNOS mRNA and
decrease in ACE.
PM10 increased circulating band cells and
shortened transit time of PMN through
postmitotic pool in marrow. Increased bone
marrow pool of PNM, esp. in mitotic pool.
Increased circulating PMN band cell counts and
size of bone marrow mitotic pools of PMNs.
Progression of atherosclerotic lesions. Increase
in plaque cell turnover, extracellular lipid pools,
and total lipids in aortic lesions.
Colloidal carbon stimulated the release of
BRDU-labeled PMNs from bone marrow. The
supernatant of alveolar macrophages treated
with colloidal carbon in vitro also stimulated
release of PMNs from bone marrow, likely via
cytokines.
Reference
Watkinson et al.
(2000a,b);
Watkinson et al.
(2001)
Ulrichetal.
(2002)
Mukae et al.
(2001)
Suwaetal. (2002)
Terashima
etal. (1997)
-------
TABLE 7-3 (cont'd). CARDIOVASCULAR AND SYSTEMIC EFFECTS OF INSTILLED ROFA AND OTHER
PARTICIPATE MATTER
Species, Gender,
Strain Age, or
Body Weight
Hamsters,
100-150 g
Particle"
polystyrene
particles
(unmodified)
carboxylate-
modified
polystyrene
amine-
modified
polystyrene
Exposure
Technique
Intravenous (IV)
administration
and intratracheal
instillation
Dose Particle Size
5, 500, 60 nm
5000 ug/kg
50, 100,
500 ug/kg
5, 50, 500 ug/kg
PEbTime
to Analysis Cardiovascular Effects
IV doses of 5,500 or 5,000 ug/kg b.wt. of
unmodified polystyrene particles did not
affect thrombus formation.
IV doses of 100 and 500 ug/kg b.wt.
carboxylate-modified polystyrene particles
decreased (p < 0.05) thrombus formation
intensity.
IV doses of 100 and 500 ug/kg b.wt. of
amine-polystyrene particles increased
thrombus formation (p < 0.01).
Reference
Nemmar et al.
(2002)
to
to
Intratracheal instillation of 5,000 ug/kg of
amine-polystyrene particles significantly
(p < 0.05) increased thrombus formation
but not 5,000 ug/kg of unmodified- or
carboxylate-polystyrene particles.
Platelet aggregation (ADP-induced in vitro)
was not affected by unmodified-
polystyrene up to 100 ug/mL; and was
strongly increased by amine-polystyrene
particles.
Authors attributed observed prothrombic
activity of ultrafine paticles, at least, in
part, to platelet activation by positively
charged amine-modified polystyrene
particles.
aROFA = Residual oil fly ash
Fe2(SO4)3 = Iron sulfate
VSO4 = Vanadium sulfate
CFA = Coal fly ash
DOFA = Domestic oil burning furnace fly ash
OTT = Ottawa urban ambient particles
MSH = Mt. St. Helen's volcanic ash
NiSO2 = Nickel sulfate
3PE = Post Exposure
-------
7.2.2 Ambient Particulate Matter Cardiovascular Effects
Epidemiology studies discussed in Chapter 8 suggest that homeostatic changes in the
vascular system can occur after episodic exposure to ambient PM. However, very few controlled
human exposure studies of ambient PM effects on cardiovascular endpoints have been conducted
thus far. In one such study, Ohio et al. (2000a) reported that inhalation of concentrated ambient
particles (CAPs) in healthy nonsmokers increased blood fibrinogen levels. They exposed
38 volunteers exercising intermittently at moderate levels of exertion for 2 h to either filtered air
(FA) or particles concentrated from the air in Chapel Hill, NC. The CAP exposures ranged from
23 to 311 |ig/m3, reflecting variations in particles collected outside the facility. Dividing CAPs
exposures into three groups, blood fibrinogen levels were somewhat increased for each tritile of
exposure (mean = 47 |ig/m3 for lowest group) in comparison to FA-exposure fibrinogen levels
(the CAPs fibrinogen levels being significantly higher for all tritiles combined but not for any
one tritile in comparison to FA control values). However, differences between pre- and post-
CAPs exposure fibrinogen levels at 18 h postexposure were not significant. Other blood
parameters tested in this study (including numbers of RBCs, monocytes, lymphocytes, platelets,
or neutrophils) did not change significantly. The blood effects may be associated with mild
pulmonary inflammation found 18 h after exposure to such CAPs (see Section 7.2.3).
Two other human CAPs inhalation studies are limited by small numbers of subjects
studied. In one, Petrovic et al. (2000) exposed four healthy volunteers (aged 18 to 40 years)
under resting conditions to FA and low, mid, and high concentrations (23 to 124 |ig/m3) of CAPs
(0.1 to 2.5 |im) from downtown Toronto for 2 h using a face mask. The low CAP exposures
were reflective of typical ambient PM25 levels and the high ones of maximum PM25 levels seen
in Toronto. On each day prior to exposure, pulmonary function, nasal lavage, nasal acoustic
rhinometry, blood collection for plasma fibrinogen and clotting factor VII antigen, and a resting
ECG were taken. Pulmonary function measurements were taken every 30 min and ECG
readings recorded continuously during exposure, followed by ECG readings after 30 min of
postexposure exercise and by nasal lavage, sputum induction, and blood collection at about 2 h
and about 24 h postexposure. Review of the ECG data by a cardiologist showed no clinically
significant cardiac effects during exposure, the ensuing exercise period, or 24 h postexposure;
but 2 of 4 subjects showed 15-20% pre- to postexposure increases in blood fibrinogen levels
within 2 h post high CAPs exposure (124 |ig/m3) versus maximum 5-6% increases after FA
7-23
-------
exposures, although there were no statistically significant differences for mean fibrinogen
between CAP and FA exposures.
In another small pilot study, reported by Gong et al. (2000), four healthy adult volunteers
(2 male, 2 female; aged 19 to 40 years) were exposed for 2 h while at rest in a whole body
chamber to FA or to PM2 5 CAPs from Los Angeles air. The CAP exposures at mean 2 h
concentrations of 148 to 246 |ig/m3 (the latter approximating likely maximum exposure levels in
Los Angeles) resulted in "no meaningful changes" in heart-rate variability, or ECG ST voltages,
lung function, or respiratory symptoms, based on data collected during 2-h exposures or 10 or
22 h afterward. These results appear to be suggestive of Los Angeles ambient PM25 exposures
being unlikely to affect cardiorespiratory functions in healthy non-elderly adults. However, such
a conclusion must be tempered by several considerations: (a) the small number of subjects
tested and only while at rest (versus planned further studies to evaluate larger numbers of both
healthy and compromised volunteers with respiratory or vascular disease, presumably to include
exposures involving intermittent exercise); and (b) the Harvard model concentrator used did not
likely concentrate effectively ambient particles < 0.1 to 0.2 jim MMAD, thus not exposing
subjects to concentrated levels of potentially important combustion-derived (from
diesel/gasoline vehicles; wood smoke, etc.) ambient particles.
More recently, Gong et al. (2003) exposed 12 healthy and 12 asthmatic subjects (age range
18 to 45 years) to Los Angeles CAPs (PM25). An exposure to CAPs averaging 174 |ig/m3 (range
99 to 224) in a whole body chamber for 2 h was alternated with a filtered air (FA) exposure at
least 14 days apart. Subjects exercised for 15 min of each half hour at a ventilation rate of 15 to
20 L/min/m2 body surface. Tests were performed just before exposure (pre), just after
(immediately post), 3.5 to 4 h after (4 h), and the next day (day 2). No significant CAPS-related
changes in routine blood parameters were seen, except for some mediators of blood coagulation
and systemic inflammation, (e.g., Factor VII, which declined immediately postexposure and at
4 h later and then rebounded on day 2). However, there were no accompanying changes in
fibrinogen or serum amyloid A with CAPs exposure. Both groups exhibited modest increases in
HR variability, and both had significant differences in HR during exposure (FA: 76 at rest and
96 at exercise; CAPs: 72 at rest and 92 at exercise). There were no significant differences in
diastolic BP and HR-systolic product. Systolic BP increased in healthy subjects after CAP
exposure at immediate, 4 h, and 2 days postexposure. Systolic BP decreased in asthmatics
7-24
-------
immediately and at day 2 postexposure. Holter ECG data suggested to the authors a CAP-
induced increase in parasympathetic input to the heart, about which they are uncertain as to the
health significance. During CAPs exposure, some subjects also reported "cardiac symptoms,"
which included faintness, dizziness, irregular heartbeat, and pain related to the heart.
Godleski and colleagues (2000) have performed a series of experiments examining the
cardiopulmonary effects of inhaled CAPs on normal mongrel dogs and on dogs with coronary
artery occlusion. Dogs were exposed by inhalation via a tracheostomy tube to Boston CAPs for
6 h/day for 3 consecutive days. The investigators found little biologically-relevant evidence of
pulmonary inflammation or injury in normal dogs exposed to CAPs (daily range of mean
concentrations was -100 to 1000 |ig/m3). The only statistically significant effect was a doubling
of the percentage of neutrophils in lung lavage. Despite the absence of major pulmonary effects,
a significant increase in heart rate variability (an index of cardiac autonomic activity), a decrease
in heart rate, and a decrease in T alternans (an index of vulnerability to ventricular fibrillation)
were seen. Exposure assessment of particle composition yielded no indication of which specific
components of the CAPs were correlated with the day-to-day variability in response. The
significance of these effects is not yet clear, given that the effects did not occur on all exposure
days (e.g., changes in heart rate variability were observed on only 10 of the 23 exposure days).
Although the HRV increase and the decrease in t-wave alternans might suggest a reduction in
cardiovascular risk in response to inhaled concentrated ambient PM, the clinical significance of
this effect is unclear. However, the magnitude of the observed changes, while small, are clearly
not consistent with increased risk for cardiovascular events.
The most important finding of Godleski et al. (2000) was the observation of a potential
increase in ischemic stress of the cardiac tissue from repeated exposure to CAPs from Boston.
During coronary occlusion in four dogs exposed to PM, they observed (a) significantly more
rapid development of ST elevation of the ECG waveform; and (b) greater peak ST-segment
elevation after CAP exposure. Together, these changes are not internally consistent with those
noted above. That is, on one hand, the ST segment elevation timing suggests a lower ischemic
threshold and a higher risk for serious outcomes in the compromised dog model, but the HRV
and T-wave alternans changes in the normal dogs suggest lower cardiac risk. Clearly, much
further work in more dogs (and other species) will be necessary to confirm such findings and to
better understand their potential significance.
7-25
-------
In a series of studies, (Gordon et al., 2000) examined rodent cardiovascular system
responses to CAPs derived from New York City air. Particles of 0.2 to 2.5 jim diameter were
concentrated up to 10 times their levels in ambient air (-130 to 900 |ig/m3) to maximize possible
differences in effects between normal and cardiopulmonary-compromised laboratory animals.
No ECG changes were detected in normal Fischer 344 rats or hamsters exposed by inhalation to
the New York City CAPs for a single 3-h exposure or for 3 daily 6-h exposures. Similarly, no
deaths or ECG changes were seen in MCT rats or cardiomyopathic hamsters exposed to PM.
In contrast to the nonsignificant decrease in heart rate observed in dogs exposed to Boston CAPs
(Godleski et al., 2000), statistically significantly heart rate increases (-5%) were observed by
Gordon et al. (2000) in both the normal and MCT rats exposed to New York CAPs, but not on
all exposure days. Thus, extrapolation of the heart rate changes in these animal studies to human
health effects is difficult, although the observed increase in heart rate in rats is similar to that
observed in some human population CAPs studies.
Gordon and colleagues (1998) have also reported other cardiovascular effects in animals
exposed to inhaled CAPs. Increases in peripheral blood platelets and neutrophils were observed
in control and MCT rats at 3 h, but not 24 h, after a 3 h exposure to 150 to 400 |ig/m3 New York
City CAPs. This neutrophil effect did not appear to be dose-related and did not occur on all
exposure days, suggesting that day-to-day changes in particle composition may play an
important role in the systemic effects of inhaled particles. The number of studies reported was
small; and, it is therefore not possible to determine statistically if the day-to-day variability was
truly due to differences in particle composition or even to determine the size of this effect.
Nadziejko et al. (2002) exposed healthy rats to concentrated ambient particles (CAPs) from
New York City air at concentrations in the range of 95 to 341 |ig/m3 for 6 h and sampled blood
at 0, 12, and 24 h postexposure. They found no consistent differences in counts of platelets,
blood cells, or in levels of proteins in the blood coagulation system that included fibrinogen,
thrombin-anti-thrombin complex, tissue plasminogen activator, plasminogen activator inhibitor,
and factor VII. Nadziejko et al. (2002) present a thorough discussion of the blood coagulation
system, demonstrating its complexity, and further discuss limitations of the study that include
particle composition and size, the possible blunted response seen in rats compared to humans,
the healthy status of the animals compared to a cardiovascular compromised model, and the
endpoints chosen.
7-26
-------
Studies by Vincent et al. (2001) indicate that very high concentrations (48 mg/m3) of urban
ambient particles from Ottawa ambient air administered by nose-only inhalation for 4 h to
laboratory rats can cause a vasopressor response and affect blood levels of endothelin without
causing acute lung injury. The authors reported a MMAD of ~4 jim for the EHC93 Ottawa
urban ambient particles resuspended from samples from the air purification system of an office
building. In this study, they also found that exposure to water-leached Ottawa samples
(EHC93L) can modify the potency to influence hemodynamic changes by removal of polar
organic compounds and soluble elements from the Ottawa particles. Exposure to DPM
(5 mg/m3) had no effect on blood pressure, but caused elevated endothelin levels, whereas a
comparable exposure to 5 mg/m3 carbon black (CB) had no effects. The authors concluded that
their results suggest a novel mechanism by which inhaled particles can affect cardiovascular
function, i.e., by causing elevated endothelins, which are among the most potent vasoconstrictors
in the systemic circulation and which have been shown to correlate with severity of disease in
congestive heart failure and to predict cardiac death (Galatius-Jensen, et al., 1996), possibly due
to ET-1 being cardiotoxic by promoting infarct size (Omland et al., 1994). However, the very
high exposure concentrations used leave it unclear as to the extent that such effects may be
pertinent to ambient PM exposure conditions.
Another study (Ulrich et al., 2002) utilized Ottawa EHC93 and an exposure paradigm
consisting of saline control, 5 mg EHC93 only, or ozone (O3) pretreatment (8 h to 1600 |ig/m3)
on the day preceding instillation. Instillations were at concentrations of 0.5, 1.5, or 5 mg/rat
(-2.2, 6.7, or 22.2 mg/kg based on reported body weights). At the high doses, both with
(170 mg/dl) and without (160 mg/dl) O3-pretreatment, they observed a 20% increase in plasma
fibrinogen at 2 days post exposure compared to saline controls (140 mg/dl). These changes
correlated with increases in endothelin (ET)-l levels and iNOS mRNA and a decrease in
angiotensin-converting enzyme (ACE). The authors suggest that the hematological changes seen
in this study model heart failure in high-risk groups exposed to PM.
7.2.3 ROFA and Other Combustion Source-Related Particles
Turning to studies evaluating cardiovascular effects of controlled exposures to combustion
source-related materials, using the MCT model of cardiorespiratory disease, Killingsworth et al.
(1997) examined the effects of a combustion source-related irritant particle mix (residual fuel oil
7-27
-------
fly ash [ROFA] from the Boston area). They observed 42% mortality in MCT rats exposed to
-580 |ig/m3 ROFA for 6 h/day for 3 consecutive days but no deaths among MCT rats exposed to
filtered air or saline-treated healthy rats exposed to ROFA. The increase in MCT/ROFA group
deaths was accompanied by (a) increased neutrophils in lavage fluid and (b) increased
immunostaining of macrophage inflammatory protein (MIP-2), from among several
proinflammatory chemokines evaluated, in the lungs and hearts of the MCT/ROFA animals.
Cardiac immunohistochemical analysis indicated increased MIP-2 in cardiac macrophages. The
ROFA-induced deaths did not result from a change in pulmonary arterial pressure, and the cause
of death was not identified. The results suggest that MCT treatment and ROFA exposure can
produce significant lung inflammation and possible increases in proinflammatory signals in the
heart.
Using a similar experimental model, Watkinson et al. (1998) examined the effects of
intratracheally instilled Florida area power plant ROFA (0.0, 0.25, 1.0, 2.5 mg in 0.3 mL saline)
on ECG measurements in healthy control and MCT rats. They observed a dose-related increase
in the incidence and duration of arrhythmic events in control animals exposed to ROFA
particles, and these effects appeared to be exacerbated in the MCT animals (the strength of these
conclusions and determination of lowest observed effective dose levels being limited due to lack
of statistical analyses). Similar to the results of Killingsworth et al. (1997), healthy animals
treated with ROFA suffered no deaths, but there were 1,3, and 2 deaths in the low-, medium-,
and high-dose MCT groups, respectively. Further, given that the observed rhythm disturbances
were mimicked by infusion of acetylcholine, increased vagal (parasympathetic) input may have
contributed to the ROFA-induced increase in arrythmias. Thus, ROFA PM may be linked to
conductive and hypoxemic arrhythmias in rats having MCT-induced pulmonary hypertension.
However, the specific data and analyses in this study do not establish that relationship with
certainty. Such small sampling frequency as was used here does not allow any extrapolation in
terms of the total frequency of arrhythmia because of the inherent variability of arrhythmia
frequency. Also, since the increased arrhythmia reported by these investigators in this animal
model is almost entirely dropped beats, these findings have questionable bearing on the
mechanism of potential increased risk of cardiac mortality in humans exposed to PM. It is also
possible that the reported mortalities were simply related to the MCT-induced pulmonary
hypertension.
7-28
-------
In order to help gauge the biological relevance of intratracheal instillation of ROFA
particles, Kodavanti et al. (1999) exposed MCT rats to Florida area ROFA by either instillation
(0.83 or 3.33 mg/kg) or nose-only inhalation (15 mg/m3, 6 h/day for 3 consecutive days).
Similar to Watkinson et al. (1998), intratracheal instillation of ROFA in MCT rats caused about
50% mortality. However, very notably, no mortality occurred in MCT rats exposed to ROFA by
the inhalation route despite the high exposure concentration (15 mg/m3); nor was there any
mortality in healthy rats exposed to ROFA or in MCT rats exposed to clean air. Despite the fact
that mortality was not associated with ROFA inhalation exposure of MCT rats, exacerbation of
lung lesions and pulmonary inflammatory cytokine gene expression, as well as ECG
abnormalities, were evident.
Watkinson and colleagues further examined whether the effects of instilled ROFA would
be exacerbated in rodents already under increased stress by being previously exposed to O3 or
being housed in the cold (Watkinson et al., 2000a,b; Watkinson et al., 2001; Campen et al.,
2000). The effect of O3-induced pulmonary inflammation (preexposure to 1 ppm O3 for 6 h) or
housing in the cold (10 °C) on the response to instilled ROFA in rats were similar to that
produced with MCT. Bradycardia, arrhythmias, and hypothermic changes were consistently
enhanced in the O3-exposed and cold-stressed animals treated with ROFA (0.25, 1.0, or
2.5 mg/rat); but, unlike for the MCT animals, no deaths occurred. Thus, it appears that
preexisting cardiopulmonary disease or increased physiological stress may make rodents more
susceptible to cardiovascular changes induced by intratracheal instillation of > 0.25 mg of
ROFA. While studies of instilled ROFA demonstrated immediate and delayed responses,
consisting of bradycardia, hypothermia, and arrhythmogenesis in conscious, unrestrained rats
(Watkinson et al., 1998; Campen et al., 2000), further study of instilled ROFA-associated
transition metals showed that vanadium (V) induced the immediate responses, while nickel (Ni)
was responsible for the delayed effects (Campen et al., 2002). Moreover, Ni, when administered
concomitantly, potentiated the immediate effects caused by V.
In another study, Campen et al. (2001) examined responses to these metals in conscious
rats by whole-body inhalation exposure. The authors tried to ensure valid dosimetric
comparisons with the instillation studies, by using concentrations of V and Ni ranging from
0.3 to 2.4 mg/m3. The concentrations used incorporated estimates of total inhalation dose
derived using different ventilatory parameters. Heart rate (HR), core temperature (T[CO]), and
7-29
-------
electrocardiographic (ECG) data were measured continuously throughout the exposure. Animals
were exposed to aerosolized Ni, V, or Ni + V for 6 h per day for 4 days, after which serum and
bronchoalveolar lavage samples were taken. While Ni caused delayed bradycardia,
hypothermia, and arrhythmogenesis at concentrations > 1.2 mg/m3, V failed to induce any
significant change in HR or T[CO], even at the highest concentration. When combined, Ni and
V produced observable delayed bradycardia and hypothermia at 0.5 mg/m3 and potentiated these
responses at 1.3 mg/m3 to a greater degree than were produced by the highest concentration of
Ni (2.1 mg/m3) alone. The results are suggestive of a possible synergistic relationship between
inhaled Ni and V, albeit these studies were performed at metal concentrations orders of
magnitude greater than their typical ambient concentrations.
Watkinson et al. (2000a,b) also sought to examine the relative toxicity of different particles
on the cardiovascular system of spontaneously hypertensive (SH) rats. They instilled 2.5 mg of
representative particles from ambient (Ottawa) or natural (Mount Saint Helens volcanic ash)
sources and compared the response to 0.5 mg ROFA. Instilled particles were either mass
equivalent dose or adjusted to produce equivalent metal dose. They observed adverse changes in
ECG, heart rate, and arrhythmia incidence that were much greater in the Ottawa ambient PM-
and ROFA-treated rats than in the volcanic ash-treated rats. The cardiovascular changes
observed with the Ottawa particles were actually greater than with the ROFA particles. These
experiments indicate: (a) that instillation of ambient air particles, albeit at a very high
concentration, can produce cardiovascular effects; and (b) that exposures of equal mass dose to
particle mixes of differing composition did not produce the same cardiovascular effects,
suggesting that PM composition rather than just mass was responsible for the observed effects.
Kodavanti et al. (2000a) exposed (via nose-only inhalation) SH and normotensive Wistar-
Kyoto (WKY) rats to 15 mg/m3 ROFA (Florida area) particles for 6 h/day for 3 days. The high
exposure concentration (-1,000 times higher than current U.S. ambient PM levels) was selected
to produce a frank but nonlethal injury. Exposure to ROFA produced alterations in the ECG
waveform of SH but not the normotensive rats. Although the ST segment area of the ECG was
depressed in the SH rats exposed to air, further depressions in the ST segment were observed at
the end of the 6-h exposure to ROFA on days 1 and 2. The enhanced ST segment depression
was not observed on the third day of exposure, suggesting that adaptation to the response may
have occurred. Thus, exposure to a very high concentration of ROFA exacerbated an aberrant
7-30
-------
variation in the electroconductivity pattern of the heart in an animal model of hypertension.
However, this ROFA-induced alteration in the ECG waveform was not accompanied by an
enhancement in the monocytic cell infiltration and cardiomyopathy that also develop in SH rats.
Contrary to findings from Godleski's dog study, Muggenburg et al. (2000b) reported that
inhalation exposure to high concentrations of Boston area ROFA caused no consistent changes
in amplitude of the ST-segment, form of the T wave, or arrhythmias in healthy dogs. In their
studies, four beagle dogs were exposed to 3 mg/m3 ROFA particles for 3 h/day for 3 consecutive
days. They noted a slight but variable decrease in heart rate, but the changes were not
statistically or biologically significant. The transition metal content of the ROFA used by
Muggenburg was -15% by mass, a value on the order of a magnitude higher than that found in
current U.S. urban ambient PM samples. Although the study did not specifically address the
effect of metals, it suggests that inhalation of high concentrations of metals may have little effect
on the cardiovascular system of a healthy individual. In a second study using dogs with pre-
existing cardiovascular disease, Muggenburg et al. (2003) evaluated the effects of short-term
inhalation exposure (oral inhalation for 3 h on each of 3 successive days) to aerosols of transition
metals. Heart rate and the ECG readings were studied in conscious beagle dogs (selected for
having pre-existing cardiovascular disease) that inhaled respirable particles of oxide and sulfate
forms of transition metals (Mn, Ni, V, Fe, Cu oxides, and Ni and V sulfates) at concentrations of
0.05 mg/m3. Such concentrations are 2 to 4 orders of magnitude higher than for typical ambient
U.S. levels (usually < 0.1 to 1.0 |ig/m3 for such metals). No significant effects of exposure to the
transition metal aerosols were observed. The discrepancy between the results of Muggenburg
et al. and those of Godleski and colleagues leave open major questions about PM effects on the
cardiovascular system of the dog.
Wellenius et al. (2002) developed and tested a model for investigating the effects of
inhaled PM on arrhythmias and HRV in rats with acute MI. Left-ventricular MI was induced in
Sprague-Dawley rats by thermocoagulation of the left coronary artery, whereas control rats
underwent sham surgery. Diazepam-sedated rats were exposed (1 h) to ROFA (Boston area),
carbon black (CB), or room air at 12 to 18 h after surgery. Each exposure, at 3 mg/m3, was
immediately preceded and followed by a 1-h exposure to room air (baseline and recovery
periods, respectively). Lead-II electrocardiograms were recorded. In the MI group, 41% of rats
exhibited one or more premature ventricular complexes (PVCs) during the baseline period.
7-31
-------
Exposure to ROFA, but not to CB or room air, increased arrhythmia frequency in animals with
preexisting PVCs. Furthermore, MI rats exposed to ROFA, but not to CB or room air, had
decreased HRV, but there was no difference in arrhythmia frequency or HRV among sham-
operated animals. The limited statistical significance (one MI rat mainly exhibited the reported
changes) of the reported results call into question the biological relevance of the change
observed in arrhythmia frequency in this myocardial infarction model exposed to ROFA at
3 mg/m3.
Gardner et al. (2000) examined whether the instillation of particles would alter blood
coagulability factors in laboratory animals. Sprague-Dawley rats were instilled with 0.3, 1.7, or
8.3 mg/kg of ROFA (Florida) or 8.3 mg/kg Mount Saint Helens volcanic ash. Because
fibrinogen is a known risk factor for ischemic heart disease and stroke, the authors suggested
that PM-induced alterations in the blood fibrinogen or other coagulation pathway components
could take part in the triggering of cardiovascular events in susceptible individuals. Elevations
in plasma fibrinogen, however, were observed in healthy rats only at the highest treatment dose
(8.3 mg/kg); and no other changes in clotting function were noted. Because the lower treatment
doses are known to cause pulmonary injury and inflammation, albeit to a lesser extent, the
absence of plasma fibrinogen changes at the lower doses suggests that only high levels of
pulmonary injury are likely to produce an effect in healthy test animals.
To establish the temporal relationship between pulmonary injury, increased plasma
fibrinogen, and changes in peripheral lymphocytes, Kodavanti et al. (2002a) exposed SH and
WKY rats to Boston ROFA using both inhalation and intratracheal instillation exposure (acute
and long-term) scenarios. Increases in plasma fibrinogen and decreases in circulating white
blood cells were found for both strains in response to acute ROFA exposure (15 mg/m3; 6 h/day;
1 week) by inhalation and were temporally associated with acute (1 week post exposure), but not
longer-term (2 to 4 week) lung injury. A bolus intratracheal instillation of ROFA at 5 mg/kg
body weight increased plasma fibrinogen in both SH and WKY rats; whereas the increase was
evident only in SH rats after acute (1 week) ROFA inhalation. The increased fibrinogen in
SH rats was associated with greater pulmonary injury and inflammation than was found in the
WKY rats. The authors concluded that acute PM exposure can provoke an acute thrombogenic
response associated with pulmonary injury/inflammation and oxidative stress in cardiovascular-
compromised rats.
7-32
-------
Kodavanti et al. (2003) exposed male SD, WKY, and SH male rats to nose-only doses of
oil combustion-derived ROFA from Boston, which contained bioavailable zinc (Zn) at doses of
2, 5, or 10 mg/m3 for 6 h/day for 4 consecutive days. A second exposure paradigm used
exposure to 10 mg/m3 ROFA for 6 h/day, 1 day/week, for 4 or 16 consecutive weeks.
Cardiovascular effects were not seen in SD and SH rats with the acute or chronic exposure, but
WKY rats from the 16 week exposure group had cardiac lesions consisting of chronic-active
inflammation, multifocal myocardial degeneration, fibrosis, and decreased numbers of
granulated mast cells. These results suggest that myocardial injury in sensitive rats can be
caused by long-term inhalation of high concentrations of ROFA.
Perhaps of more direct relevance to evaluation of ambient PM effects, the effects of diesel
emissions (DE) on ECG and FIR were evaluated in SH rats, both male and female, exposed to
DE generated from a 2000 model diesel engine (Campen et al., 2003). Whole body exposures
included dilutions at concentrations of 30, 100, 300, and 1000 |ig/m3 for an exposure period of
6 h/day, 7 days/week, for 1 week. Exposed rats showed a significantly elevated daytime HR
(290 ± 7 versus 265 ± 5 for control male rats) throughout the exposure that was not
concentration dependent. Additionally, a concentration-dependent prolongation of the PQ
interval was observed in exposed rats. The authors suggested that these high level exposures to
DE may affect the pacemaking system of rats by means of ventricular arrhythmias. However,
the design of the study did not include testing of DPM (versus whole DE) so that one cannot
clearly attribute the reported effects to DPM versus associated gases or a combination of both.
Suwa et al. (2002) studied the effect of PM10 on the progression of atherosclerosis in
rabbits. They exposed Watanabe heritable hyperlipidemic rabbits (with naturally increased
susceptibility to atherosclerosis) to 5 mg PM10 in 1 mL saline administered by intrapharyngeal
instillation (2 times per week for 4 weeks) or to saline vehicle for 4 weeks, and then both
(a) measured bone marrow stimulation and (b) used quantitative histologic methods to determine
the morphologic features of the atherosclerotic lesions. Exposure to PM10 (99% < 3.0 jim) from
Ottawa, CN air caused an increase in circulating polymorphonuclear leukocytes (PMN) band cell
counts and an increase in the size of the bone marrow mitotic pool of PMNs. Exposure to PM10
also caused progression of atherosclerotic lesions toward a more advanced phenotype. The
volume fraction (vol/vol) of the coronary atherosclerotic lesions was increased by PM10
exposure. The vol/vol of atherosclerotic lesions correlated with the number of alveolar
7-33
-------
macrophages that phagocytosed PM10. Exposure to PM10 also caused an increase in plaque cell
turnover and extracellular lipid pools in coronary and aortic lesions, as well as in the total
amount of lipids in aortic lesions.
Terashima et al. (1997) also examined the effect of particles on circulating neutrophils.
They instilled rabbits with 20 mg colloidal carbon, a relatively inert particle (< 1 |im), and
observed a stimulation of the release of 5'-bromo-2'deoxyuridine (BrdU)-labeled PMNs from the
bone marrow at 2 to 3 days after instillation. Because the instilled supernatant from rabbit AMs
treated in vitro with colloidal carbon also stimulated the release of PMNs from the bone marrow,
the authors hypothesized that cytokines released from activated macrophages may be responsible
for this systemic effect. However, this group (Terashima et al., 2001) has determined that the
BAL procedure itself has the effect of producing an increase in peripheral blood neutrophils,
IL-6, and G-CSF. Bronchoscopy alone does not induce this acute phase response and bone
marrow stimulation. The same research group (Tan et al., 2000) looked for increased white
blood cell counts as a marker for bone marrow PMN precursor release in humans exposed to
very high levels of carbon from biomass burning during the 1997 Southeast Asian smoke-haze
episodes. They found a significant association between PM10 (1-day lag) and elevated band
neutrophil counts expressed as a percentage of total PMNs. The biological relevance of this
latter study to more usual urban PM exposure-induced systemic effects is unclear, however,
because of the high dose of carbon particles.
Frampton (2001) exposed healthy, nonsmoking subjects (18 to 55 years old) to 10 |ig/m3
ultrafine carbon while at rest via a mouthpiece for 2 h 1A , with a 10-min break between each
hour of exposure. The exposure concentration (10 |ig/m3) corresponded to 2 x 106 particles/cm3;
and respiratory symptoms, spirometry, blood pressure, pulse-oximetry, blood markers, and
exhaled NO were evaluated before, immediately following, and 3.5 and 21 h postexposure.
Blood markers focused on parameters related to acute response, i.e., blood coagulation,
circulating leukocyte activation (including complete blood leukocyte counts and differentials),
IL-6, fibrinogen, and clotting factor VII. Heart rate variability and repolarization phenomena
were evaluated by continuous 24-h ECG Hotter monitoring. Preliminary findings indicated no
particle-related effects, neither for cardiovascular nor respiratory-related endpoints.
Nemmar et al. (2002) studied effects of ultrafine (60 nm) polystyrene particles on
thrombus formation in a hamster model after intravenous (IV) administration of unmodified,
7-34
-------
carboxylate-polystyrene, or amine-polystyrene particles, which carry a negative or positive
charge, respectively. Unmodified particles did not affect thrombosis at IV doses up to
5000 |ig/kg; whereas carboxylate-polystyrene particles significantly inhibited thrombus
formation at 100 and 500 |ig/kg, but not at 50 |ig/kg body weight. Thrombosis was significantly
enhanced by amine-polystyrene particles at 50 and 500 |ig/kg, but not at 5 |ig/kg body weight.
Intratracheal instillation of 5 mg/kg of amine-polystyrene particles, but not unmodified or
carboxylate-modified particles, also increased thrombosis formation. Platelet aggregation (ADP-
induced in vitro) was also enhanced significantly by amine-modified polystyrene particles, but
not by unmodified or carboxylate-modified particles. Thus, only positively charged ultrafme
particles resulted in enhancement of thrombus formation. The authors concluded that (a) the
presence of ultrafme particles in the circulation may affect hemostasis and (b) this is dependent
on the surface charge of the particles, i.e., positive-charged particles induce prothrombotic
effects, at least partly via platelet activation.
7.2.4 Summary of Cardiovascular/Systemic Effects
In summary, experimental controlled exposure studies of cardiovascular-related effects in
healthy humans have yielded only very limited evidence for ambient PM effects on cardiac
function as indexed by ECG readings or on systemic endpoints (e.g., vasopressor control, blood
coagulation control, etc.) linked to more serious cardiovascular events. Probably of most note,
the controlled human exposure CAPs study by Ohio et al. (2000a) and another by Petrovic et al.
(2000) did find evidence indicating that ambient levels (ranging up to -125 to 300 |ig/m3) of
inhaled PM2 5 can produce some biochemical changes (increased fibrinogen) in blood suggestive
of PM-related increased risk for prothrombotic effects. Similarly, Ulrich et al. (2002) found a
20% increase in plasma fibrinogen in rats 2 days after instillation exposure to 6.7 or 22.2 mg/kg
of Ottawa EHC93 ambient PM extract. Also, decreased Factor VII levels were observed by
Gong et al. (2003) in humans (with 2-h CAPs exposure at -174 |ig/m3). The decreased Factor
VII levels may be due to that enzyme being consumed in an ongoing coagulation process. On
the other hand, the same and many other human and animal studies did not find changes in other
factors (e.g., increased platelets or their aggregation) related to blood coagulation control.
Overall, then, some available laboratory studies provide limited evidence suggesting that high
concentrations/doses of inhaled or instilled particles can exert cardiovascular-related systemic
7-35
-------
effects; but many of the studies provide conflicting evidence, especially with regard to heart rate,
HRV, or other ECG markers of cardiac function. Thus, although some of the reported changes
have been used as clinical "markers" for cardiovascular diseases, the causal relationship between
such PM-related changes and potential life-threatening alterations in cardiovascular function
remains to be better established.
Among the most salient hypotheses proposed to account for cardiovascular/systemic
effects of PM are: alterations in coagulability (Seaton et al., 1995; Sjogren, 1997); cytokine
effects on heart tissue (Killingsworth et al., 1997); perturbations in both conductive and
hypoxemic arrythmogenic mechanisms (Watkinson et al., 1998; Campen et al., 2000); altered
endothelin levels (Vincent et al., 2001); and activation of neural reflexes (Veronesi and
Oortgiesen, 2001). Only limited progress has been made in obtaining evidence bearing on such
hypotheses (as discussed in later sections of this chapter), and much future research using
controlled exposures to PM of laboratory animals and human subjects will clearly be needed to
test further such mechanistic hypotheses (as well as others proposed in the future) so as to more
fully understand pathways by which low concentrations of inhaled ambient PM may be able to
produce life-threatening systemic changes.
7.3 RESPIRATORY EFFECTS OF CONTROLLED IN VIVO PM
EXPOSURES OF HUMANS AND LABORATORY ANIMALS
This section assesses the respiratory effects of controlled in vivo exposures of laboratory
animals and humans to various types of PM. In vitro studies using animal or human respiratory
tract cells are discussed in Section 7.4. Biological responses occurring in the respiratory tract
following controlled PM inhalation include changes in pulmonary inflammation and systemic
effects that result from direct effects on lung tissue. The observed responses are dependent on
the physicochemical characteristics of the PM, exposure parameters (concentration, duration,
etc.), and health status of the host.
As noted earlier, data available in the 1996 PM AQCD were derived from studies that
evaluated respiratory effects of specific components of ambient PM or laboratory-generated
surrogate particles, e.g., metals or pure sulfuric acid droplets. Toxicological studies of various
"other" types of PM species were also discussed in the 1996 PM AQCD (U.S. Environmental
7-36
-------
Protection Agency, 1996a). These studies included exposures to fly ash, volcanic ash, coal dust,
carbon black, and miscellaneous other particles, either alone or in mixtures. Some of the
particles discussed were considered to be models of "respirable low toxicity particles" and were
used in instillation studies to delineate nonspecific particle effects from effects of known
toxicants.
A number of studies on "other PM" examined effects of up to 50,000 |ig/m3 (50 mg/m3) of
respirable particles with inherently low toxicity. Although there was no mortality, some mild
pulmonary function changes after exposure ranging from 1 h to 24 months to 5,000 to
10,000 |ig/m3 (5 to 10 mg/m3) of relatively inert particles were observed in rats and guinea pigs.
Lung morphology studies revealed focal inflammatory responses, some epithelial hyperplasia,
and fibrotic responses after chronic exposure (6 to 7 h/day 5 day/week for 20 to 24 months) of
rats to > 5,000 |ig/m3 of coal dust. Changes in macrophage clearance after exposure to
> 10,000 |ig/m3 to a variety of particles over various exposure periods (days to months) were
equivocal (no host defense effects). In studies of mixtures of particles and other pollutants,
observed effects varied, depending on the toxicity of the associated pollutant. For example, in
humans, co-exposure to carbon particles appeared to increase responses to formaldehyde but not
to acid aerosol. None of the "other" particles mentioned above are present in ambient air in
more than trace quantities.
Thus, it was concluded that the relevance of any of these studies to standard setting for
ambient PM may be extremely limited. Newer studies, on the other hand, do provide evidence
of likely greater relevance to understanding respiratory effects of ambient PM exposure and
underlying mechanisms, as discussed below.
7.3.1 Ambient Particulate Matter
Some new in vivo toxicology studies have employed inhalation exposures to evaluate the
respiratory effects of ambient particles in humans and laboratory animals, using either CAPs or
resuspended urban ambient PM from various U.S. and Canadian locations. Other new in vivo
exposure studies have mainly used intratracheal instillation techniques. The pros and cons of the
latter in comparison to inhalation are discussed in Chapter 6 (Section 6.5) and have also been
reviewed elsewhere (Driscoll et al., 2000; Oberdorster et al., 1997; Osier and Oberdorster, 1997).
In most of the instillation studies, ambient PM samples were first collected on filters; then after
7-37
-------
various storage times, PM materials were extracted from the filters and resuspended in a vehicle
(usually saline), followed by a small volume of the suspension medium then being instilled
intratracheally into the animals. It is important to note that physiochemical characteristics of the
collected PM may be altered by deposition and storage on a filter and resuspension in an
aqueous medium. Therefore, in terms of use in attempting to extrapolate experimentally
observed results to humans exposed under ambient exposure scenarios, greater importance
should be placed on inhalation study results. Instillation studies have been most valuable in
comparing effects of different types of PM and/or for investigating potential mechanisms by
which particles may cause inflammation and lung injury.
7.3.1.1 Ambient Particle Inhalation Exposures
Table 7-4 summarizes newly available studies in which various biological endpoints were
measured following inhalation exposures to CAPs or, in the case of one study, resuspended
urban ambient particles.
With regard to newly available experimental studies that most directly parallel aspects of
ambient PM inhalation exposures, Ohio et al. (2000a) exposed 38 nonsmoking healthy adult
volunteers (aged 18 to 40 years) exercising intermittently at moderate levels of exertion
(breathing rate = 25 L/min) for 2 h to either filtered air or PM2 5 concentrated 6- to 10-fold from
Chapel Hill, NC air at the inlet of the exposure chamber. Neither respiratory symptoms nor
decrements in pulmonary function (RAW, FVC, FEVj 0, PEF measurements) were found
immediately after exposure to CAPs. However, analysis of bronchoalveolar lavage (BAL) cells
and fluid obtained 18 h after Chapel Hill CAPs exposure (at 23 to 311 |ig/m2) showed a mild
increase in neutrophils in the bronchial and alveolar fractions of BAL in subjects exposed to the
highest quartile concentration of concentrated PM (mean of 206.7 |ig/m3). Lavage protein did
not increase, and there were no other indicators of pulmonary injury. The 38 human volunteers
reported on by Ohio et al. (2000a) were also examined for changes in host defense and immune
parameters in BAL and blood (Harder et al., 2001). There were no changes in the number of
lymphocytes or macrophages, subcategories of lymphocytes (according to surface marker
analysis by flow cytometry), cytokines IL-6 and IL-8, or macrophage phagocytosis. Similarly,
there was no effect of CAPs exposure on lymphocyte subsets in blood. Thus, a mild
inflammatory response to Chapel Hill CAPs was not accompanied by any evident effect on
7-38
-------
TABLE 7-4. RESPIRATORY EFFECTS OF INHALED AMBIENT PARTICIPATE MATTER IN CONTROLLED
EXPOSURE STUDIES OF HUMAN SUBJECTS AND LABORATORY ANIMALS
VO
Species, Gender,
Strain, Age, etc. Particle
Humans, healthy CAPs
nonsmokers; (Chapel Hill)
18 to 40 years old
n = 38
Humans, healthy CAPs
1 8-40 years old (Toronto)
n = 4
Humans, healthy; CAPs
19-41 years old (Los Angeles)
n = 4
Humans healthy (12) CAPs
and asthmatic (12) (Los Angeles)
18-45 years old,
nonsmoking
Mongrel dogs, some CAPs
with balloon (Boston)
occluded LAD
coronary artery
n= 14
Rats, male S-D CAPs
200-225 g, healthy- (Boston)
air, bronchitic-air,
healthy-CAPs,
bronchitic-CAPs
n = 48 (12 per group)
Exposure
Technique
Inhalation
Inhalation
(face mask)
Inhalation
Inhalation
whole body
chamber
Inhalation via
tracheostomy
Inhalation;
Harvard/EPA
fine particle
concentrator;
animals
restrained
in chamber
Exposure
Concentration'
23.1 to 31 1.1 ug/m3
24 to 124 ug/m3
148 to 246 ug/m3
99-224 ug/m3
(mean 174)
-100-1000 ug/m3
(variable from
day-to-day)
206, 733, and
607 ug/m3 for
Days 1-3,
respectively; 29 °C,
59% RH
Exposure
Particle Duration; Time
Size to PE" Analysis
0.65 um 2 h; analysis
og = 2.35 at!8h
0.1 - 2.5 2 h; analyses pre-
um & during
exposure and
about 2 to 24 h
postexposure
PM2 5 2 h
80% 0.1 to 2 h with
2.5 um alternating
exercise/rest.
Analysis at 0,
4, and 22 h PE
0.23 to 6 h/day x 3 days
0.34 um
og = up to
2.9
0.18um 5 h/day for
og = 2.9 3 days
Particle Effects/Comments
Dose-dependent increase in BAL neutrophils in both
bronchial and alveolar fractions. Increase noted at
all exposure levels. Particles were concentrated 6- to
10-fold at the inlet of the chamber.
Only stat. sig. effect on pulmonary function was
mean increase of 6.4% in thoracic gas volume after
high CAP exposure versus mean 5.6% after filtered
air exposure, but not in respiratory symptoms.
Trend towards increased nasal neutrophils, but no
respiratory inflammatory response.
No significant changes in lung function, symptoms,
SaO2,** or Holier ECGs observed. The maximum
steady state fine particle concentration in the
breathing zone was typically seven times the ambient
concentration.
CAPs-related decrease in columnar cells in induced
sputum at 0 h PE. No significant changes in SaO2,
FVC, or FEVj.
Decreased respiratory rate over time and modest
increase in lavage fluid neutrophils in normal dogs.
Study utilized Harvard ambient particle concentrator.
Ambient particles concentrated by approximately
30-fold.
Bronchitis induced by pre-treatment with 170 ppm
SO2 for 6 weeks at 5 h/day, 5 days/week. The CAPs-
exposed rats had significant increase in TV,
increased protein and percent neutrophils and
lymphocytes in lavage fluid after CAPs exposure.
Responses were greater in bronchitic than healthy
rats. Bronchitic CAPs-exposed rats showed
evidence of inflammation-related epithelial
permeability.
Reference
Ohio et al.
(2000a)
Petrovic
etal. (2000)
Gong et al.
(2000)
Gong et al.
(2003)
Godleski
etal. (2000)
Clarke etal.
(1999)
-------
TABLE 7-4 (cont'd). RESPIRATORY EFFECTS OF INHALED AMBIENT PARTICIPATE MATTER IN CONTROLLED
EXPOSURE STUDIES OF HUMAN SUBJECTS AND LABORATORY ANIMALS
Species, Gender,
Strain, Age, etc.
Rats, male S-D
200-250 g, healthy-
air, bronchitic-air,
healthy-CAPs,
bronchitic-CAPs
n = 259 (6 studies,
40-48 per study, 8-10
per group)
Rats, male, 90- to
100-day-old S-D,
with or without
SO2-induced
bronchitis
Rats, male F344
Hamsters, male,
8-month-old Bi TO-2
Rats, male F344
7-8 months
Particle
CAPs
(Boston)
CAPs
(RTF)
CAPs
(NYC)
CAPs
(NYC)
Exposure
Technique
Inhalation;
Harvard/EPA
fine particle
concentrator;
animals
restrained
in chamber
Inhalation
Inhalation
Inhalation
Exposure Particle
Concentration' Size
3-day mean CAPs 0.27 um
ranged from 1 87 to og = 2.3
481 ug/m3
29 °C, 47% RH
650 ug/m3
132 to 919 ug/m3 0.2 to
1.2 um
og = up to
3.9
100 to 350 ug/m3 0.4 um
(mean 225 ug/m3) og = 2.5
Exposure
Duration; Time
to PE" Analysis
5 h/day for
3 days
6 h/day x
2-3 days
1 x 3 h or
3x6h
3h
Particle Effects/Comments
Bronchitis induced by pre-treatment with an
average of 276 ppm, SO2 for 5 weeks at 5 h/day,
5 days/week.
Increase in neutrophils in both healthy and bronchitic
rats associated with CAPs exposure concentration.
Specific CAPs components associated with
neutrophil increase. In bronchitic rats, CAPs
components also associated with lymphocyte
increase. Histologic examination suggested
bronchial-alveolar junction as the site of greatest
inflammation response.
No significant changes in healthy rats. Increased
BAL protein and neutrophil influx in bronchitic rats
sacrificed immediately afer last CAPs exposure;
responses variable between exposure regimens.
No CAPs effects seen at 18 h postexposure.
No inflammatory responses, no cell damage
responses, no PFT changes. The PM mean
concentration factor (gravimetric) was 19.5 ± 18.6.
Basal levels of superoxide ('O2~) reduced by 90%
72 h postexposure; zymosan-stimulated O2~
Reference
Saldiva et al.
(2002)
Kodavanti
etal. (2000b)
Gordon etal.
(2000)
Zelikoff
etal. (2003)
Rats, male, F-344;
200-250 g
Ottawa ambient
Nose-only
Inhalation
40 mg/m3
4 to 5 um
MMAD
4h
formation increased by > 150% after 24 h; basal
level H2O2 production by PAM depressed 90% 3 h
following exposure and remained 60% below levels
at least 24 h; zymosan-stimulated H2O2 unaffected.
Concentrations tested represents a range over the 3 h
exposure period.
No acute lung injury; however, lung NO production
decreased and macrophage inflammatory protein-2
from lung lavage cells increased after exposure.
Increased plasma levels of endothelin-1.
Bouthillier
etal. (1998)
*Concentration = range of CAP concentrations at inlet of exposure chamber or in breathing zone of exposed subjects.
PE = Post Exposure
*SaO2 = arterial oxygen saturation.
-------
immune defenses, as determined by lymphocyte or macrophage effects. The increase in BAL
neutrophils may represent a normal physiological response of the lung to particles, although the
presence of activated neutrophils may release biochemical mediators which produce lung injury.
Whether this mild inflammatory increase in neutrophils, per se, constitutes a biologically
significant injury to the lung is an ongoing controversial issue.
In the small study by Petrovic et al. (2000) described earlier (Section 7.2.2), the four
healthy volunteer subjects (aged 18 to 40 years) exposed for 2 h to CAPs (23-124 |ig/m3 from
downtown Toronto were not only evaluated for CAP effects on cardiovascular endpoints, but
also several respiratory endpoints (nasal lavage, nasal acoustic rhinometry, pulmonary function).
No cellular signs of inflammation were observed in induced sputum samples collected at 2 or
24 h after exposure. The authors said there was a trend toward an increase in nasal lavage
neutrophils (but level of statistical significance was not specified). The only statistically
significant (p < 0.01) change in pulmonary function was a 6.4% decrease in thoracic gas volume
after high CAPs exposure to 124 |ig/m3 PM versus a 5.6% increase after filtered air exposure,
but no increase was seen in respiratory symptoms. These results, overall, suggest that acute
exposures (~2 h) to Toronto ambient PM are unlikely to exert untoward respiratory effects in
healthy adults at PM 2.5 levels below about 100 |ig/m3, but may begin to have some mild effects
on pulmonary function as PM 25 levels reach or exceed 125 |ig/m3. However, further evaluations
of these possibilities with larger numbers of healthy subjects are needed, as well as analogous
studies of compromised human subjects.
As also discussed earlier in Section 7.2.2, Gong et al. (2000) conducted a small pilot study
(n = 4 subjects, aged 19 to 41 years) in which healthy adult volunteers were exposed to
Los Angeles CAPs (148 to 246 |ig/m3) for 2 h and evaluated during or immediately after
exposure for possible respiratory effects or ECG changes. No significant effects were observed
for various lung function measures, respiratory symptoms, oxygen saturation, or in Hotter ECG
readings, even at PM2 5 concentrations (-246 |ig/m3) approximating likely maximum levels for
Los Angeles.
As also noted earlier, the effects of Los Angeles ambient air were studied further by Gong
et al. (2003), who exposed 12 healthy and 12 asthmatic subjects (age 18 to 45 years) to
Los Angeles CAPS (PM25). Exposures averaging 174 |ig/m3 (range 99-224) CAPs in a whole
body chamber for 2 h were alternated with filtered air exposure at least 14 days apart. Subjects
7-41
-------
exercised for 15 min of each half hour at a ventilation rate of 15 to 20 L/min/m2 body surface.
Tests were performed just before exposure, just after, 3.5 to 4 h after (4 h), and the next day.
Ventilation during CAPs exposure was significantly lower in both the healthy and asthmatic
groups, and both groups showed a CAPs-related decrease in columnar cells in induced sputum
immediately postexposure, but the authors were uncertain as to the health significance of this
effect. No significant differences in SaO2, FVC, FEVl5 or other respiratory parameters were
seen.
Godleski, et al. (2000), in another study noted previously (Section 7.2.2), exposed mongrel
dogs to Boston CAPs (ranging from -100 to 1000 |ig/m3) for 6 h/day for 3 days. The only two
respiratory effects reported were a decreased respiratory rate over time and a modest increase in
neutrophils in lavage fluid from the lungs of the normal dogs, even with exposures to ambient
Boston particles concentrated by 30-fold over ambient levels.
Saldiva et al. (2002) studied the effects of CAPs from Boston on rat lung. The study
was designed to: (1) determine whether short-term exposures to CAPs cause pulmonary
inflammation in normal rats and in compromised rats with chronic bronchitis (CB); (2) identify
the site within the lung parenchyma where CAPs-induced inflammation occurs; and
(3) characterize the component(s) of CAPs significantly associated with development of the
inflammatory reaction. Chronic bronchitis was induced by exposure to high doses of SO2 for
5 h/day, 5 days/week during 5 weeks prior to experimental exposures to filtered air or to the
Boston CAPs. Thus, four groups of animals were studied: (1) air pre-treated, filtered air-
exposed (air-sham); (2) sulfur dioxide pre-treated (CB), filtered air-exposed (CB-sham); (3) air
pre-treated, CAPs-exposed (air-CAPs); and (4) SO2 pre-treated, CAPs-exposed (CB-CAPs).
Normal and CB rats were exposed by inhalation either to filtered air or to CAPs during
3 consecutive days (5 h/day). The Boston CAPs concentrations varied considerably (73.5 to
733 |ig/m3), with 3-day mean CAPs mass = 126 to 481 |ig/m3. The average MMAD of the CAPs
particles was 0.27 jim ( og = 2.3). CAPs mass (as a binary exposure term) and CAPs mass (in
regression correlations) induced a significant increase in BAL neutrophils and in both normal
and CB animals. Numerical density of neutrophils in alveolar walls significantly increased with
CAPs in normal animals only, with greater neutrophils seen in central, compared to peripheral,
regions of the lung. Significant dose-dependent associations were observed between various
CAPs components and BAL neutrophils or lymphocytes; however, only vanadium and bromine
7-42
-------
concentrations were significantly associated with both BAL neutrophils and neutrophils in
CAPs-exposed groups analyzed together. The authors concluded that (a) short-term exposures to
CAPs from Boston induce a significant inflammatory reaction in rat lungs and (b) the reaction is
influenced by particle composition.
In another study of Boston CAPs, Clarke et al. (1999) exposed healthy normal rats and
(SO2-induced) bronchitic rats by inhalation via tracheostomy for 5 h/day for 3 days to filtered air
or concentrated Boston ambient PM2 5 particles averaging 206, 773, and 607 |ig/m3 on the three
different days. Significantly increased tidal volume (TV) was observed with CAPs exposures for
both the normal rats and, even greater, for the chronic bronchitic rats. Bronchiolar lavage
performed 24-h after the final day of exposure revealed evidence for significant pulmonary
inflammation following CAPs exposures, especially in chronic bronchitic rats, as indexed by
significant increases in neutrophils, lymphocytes, and total lavage protein. The authors
concluded that these results suggested two distinct mechanistic responses to inhaled particles:
(1) a stressor-type pulmonary function reaction (typified by increases in air flow and volume)
and (2) acute pulmonary inflammation characterized by cellular influx (especially neutrophils).
Zelikoff et al. (2003) reported effects on pulmonary or systemic immune defense
mechanisms in Fischer rats exposed by inhalation to filtered air or to New York City CAPs
(90 to 600 |ig/m3; mean = 345 |ig/m3) for 3 h prior to the IT instillation of Streptococcus
pneumoniae (2 - 4 x 107 organisms delivered dose). The number of lavageable cells (AM and
PMN) increased in both control and experimental groups, but were elevated faster and were
twice as high in the CAPs-exposed group, as well as staying elevated longer. Lymphocyte
values and white blood cell (WBC) counts were significantly increased 24 and 72 h postinfection
in both groups. CAPs exposure resulted in a decline in TNF-a and IL-6 levels three days
postinfection compared to bacteria-only exposed rats; but the differences were not significant.
The New York City CAPs exposure significantly increased bacterial burdens at 24 h after
infection. Thereafter, CAPs-exposed animals exhibited significantly lower bacterial burdens.
In another study, Zelikoff et al. (2003) also evaluated the effects of New York City CAPs
exposure (65 to 150 |ig/m3; mean =107 |ig/m3) in rats following a single 5 h exposure to IT
instilled Streptococcus pneumoniae. The CAPs exposure significantly reduced percentages of
lavageable PMN 24 h following CAPs exposure and remained well below control levels for up
to 3 days, but lavageable AM was significantly increased in the CAPs-exposed animals. CAPs
7-43
-------
exposure also reduced the levels of TNF-a, IL-1, and IL-6. The bacterial burden decreased in
both exposed groups over time; however, CAPs exposed animals had a significantly greater
burden after 24 h than did control rats. Lymphocyte and monocyte levels were unaffected by the
CAPs exposures.
Bouthillier et al. (1998) reported that a 4-h exposure of rats by nose-only inhalation to
40 mg/m3 of resuspended Ottawa ambient PM (MMAD = 4 to 5 jim) produced no evident acute
lung injury. However, lung NO production decreased and macrophage inflammatory protein-2
from lung lavage cells increased at 4 h after exposure, as did plasma levels of endothelin-1
(a powerful cardiac cytotoxic agent). The lack of acute lung injury demonstrated may be due, in
part, to only 60% of the PM being inhalable (see Figure 7A-4) and too much of the PM being
trapped in the nose.
7.3.1.2 Intratracheal and Intrabronchial Instillation of Ambient Particulate Matter
Other newly-available studies (Table 7-5) that evaluated acute effects of intratracheal or
intrabronchial instillation of ambient PM extracts from filters obtained from various locations
have found evidence indicating that exposures to such ambient PM materials can cause lung
inflammation and injury.
Costa and Dreher (1997) showed that instillation in rats of relatively high doses of PM
samples from four ambient airsheds (St. Louis, MO; Washington, DC; Dusseldorf, Germany;
and Ottawa, Canada) and from three combustion sources (two oil and one coal fly ash) resulted
in acute inflammatory responses, as indexed by increases in lung PMNs and eosinophils 24 h
after instillation. Biomarkers of permeability (total protein and albumin) and cellular injury,
lactic dehydrogenase (LDH), were also increased. Animals were dosed with (1) an equal dose
by mass (nominal 2.5 mg/rat) of each PM mixture or with (2) doses based on normalization of
each PM mass to a metal content of 46 jig/dose and 35.5 jig of total metals (Cu, Fe, V, Zn) for
the ambient PM and ROFA comparison. The relative potencies of the combustion-source
particles in producing the acute inflammatory effects ranked in the order: DOFA > ROFA »
CFA = saline vehicle, reflecting closely the much higher amounts of bioavailable metals in the
oil fly ash than in the coal-derived fly ash. Analogously, the ambient PM extracts exhibited, on
a per mass basis, much less potency in inducing inflammatory responses than the oil fly ash
extracts (e.g., ROFA), with the Ottawa extract exerting notably stronger effects than ambient
7-44
-------
TABLE 7-5. RESPIRATORY EFFECTS OF INSTILLED AMBIENT PARTICIPATE MATTER IN
LABORATORY ANIMALS AND HUMAN SUBJECTS3
Species, Gender,
Strain, Age, etc.
Humans, healthy
nonsmokers;
21 M, 3 F;
26.4 ±2. 2 years old
Rats, male S-D
60 days old
Rats, S-D
60 days old
n = 8/fraction
Rats, male,
S-D
60 days old
Rats, Wistar
(HAN strain)
Hamsters, Syrian
golden, male,
90-125 g
Particle
Provo, UT
PM10 filters (10 ys
old)
Provo, UT
TSP filters
(10 ys old)
Provo, UT
TSP filters (10 ys
old), soluble and
insoluble extracts
Provo, UT
TSP.
Collected 1982.
Edinburgh PM10
filters
Carbon black (CB)
Ultrafme CB
Kuwaiti oil fire
particles;
urban particles
from St. Louis, MO
Exposure
Technique
Intrabronchial
instillation
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
Instillation
Intratracheal
instillation
Intratracheal
instillation
Concentration Particle Size
SOOugofPM N/A
extract in lOmL
saline
0.25, 1.0, 2.5, N/A
S.OmgofPM
extract in 0.3 mL
saline
100, 150, 500, and N/A
1000 jig
of PM extract in
0. 5 mL saline
100, 500, 1500 N/A
jig PM in 0.5 mL
saline
Range of 50 to PM10
125 ug in 0.2 mL CB = (200-500 nm)
phosphate UCB = 20 nm
buffered saline
0. 15, 0.75, and Oil fire particles:
3.75 mg/100 g < 3.5 urn, 10 days
of 24-h samples
Exposure
Duration; Time
to PEb Analysis Particle Effects/Comments
24 h BAL Inflammation (PMN) and pulmonary injury
produced by particles collected during steel
mill operation was greater than during period
of mill closure.
24 h Dose-dependent increase in inflammation
(PMN) and pulmonary injury produced by
particles collected during steel mill operation
was greater than for during period of mill
closure for all exposed groups.
24 h Dose-dependent increase in inflammation
(PMN) and lavage fluid protein. Effect was
greater with the soluble fraction containing
more metal (Zn, Fe, Cu) except for the
100 ug exposed group.
24 h BAL Increased BAL protein and PMN at > 500 jig
dose. Also proliferation of bronchiolar
epithelium and intraalveolar hemorrhage at
500 ug dose.
Sacrificed at 6 h Increased PMN, protein, and LDH following
50-125 ug PM10; greater response with
ultrafine CB but not CB; decreased GSH
level in BAL; free radical activity (deplete
supercoil DNA); leukocytes from treated
animals produced greater NO and TNF.
Sacrificed 1 Dose-dependent increases in PMN, albumin,
and 7 days LDH, and p-N-acetylglucosaminidase and
postinstillation myeloperoxidase, decrease in AM.
Acute toxicity of the particles found in
smoke from Kuwaiti oil fires roughly similar
to that of urban particles.
Reference
Ohio and
Devlin
(2001)
Dye et al.
(2001)
Ohio et al.
(1999a)
Kennedy
etal. (1998)
Li et al.
(1996, 1997)
Brain et al.
(1998)
LDH = lactate dehydrogenase
PMN = polymorphonuclear leukocytes
PE = Post Exposure
-------
PM extracts from the other cities. However, when the exposures were normalized to match
metal content, there was little difference between the ambient PM and ROFA effects.
Interestingly, the most potent ambient PM (Ottawa) both was the freshest one collected (3 years
versus 10 years old) and had the highest bioavailable metal content of the ambient PM. Thus,
this study demonstrated, overall, that the lung dose of bioavailable transition metals, not just
instilled PM mass, was the primary determinant of the acute inflammatory response.
Kennedy et al. (1998) reported a similar dose-dependent inflammation (i.e., increase in
protein and PMN in lavage fluid, proliferation of bronchiolar epithelium, and intraalveolar
hemorrhage) in rats instilled with water-extracted particles in TSP samples collected from Provo,
UT in 1982. The particulate extract mixture was comprised of 1.0 mg/g Zn, 0.04 mg/g Ni,
2.2 mg/g Fe, 0.01 mg/g V, 1.4 mg/g Cu, 1.7 mg/g Pb, and 78 mg/g SO42+ in 500 mL saline
solution. Doses of 0, 150, 500, and 1500 jig were instilled; and effects were seen at > 500 jig.
This study also indicated that the metal constituent, in this case PM-associated Cu, was a
plausible cause of the outcome based on IL-8 secretion and enhanced activation of the
transcription factor NF-kB in cultured epithelium.
Toxicological studies of ambient PM collected from around Provo, UT (Utah Valley) in the
late 1980s are also particularly interesting (Ohio and Devlin, 2001; Dye et al., 2001). Earlier
epidemiologic studies by Pope (1989, 1991) showed that exposures to PM10 during closure of an
open-hearth steel mill over a 13-mo period beginning in 1987 were associated with reductions in
several health endpoints, e.g., hospital admissions for respiratory diseases, as discussed in the
1996 PM AQCD. Ambient PM was collected near the steel mill during the winter of 1986
(before closure), throughout 1987 (during closure), and again in 1988 (after plant reopening).
The fibrous glass hi-vol filters were stored, folded PM-side inward, in plastic sleeves at room
temperature and humidity (Dye et al., 2001). A description of the in vivo toxicological studies
follows; pertinent in vitro studies (e.g., Wu et al., 2001; Soukup et al., 2000; Frampton et al.,
1999) are discussed in Section 7.4.2.1.
Ohio and Devlin (2001) investigated biologic effects of PM from the Utah Valley to
determine if the biological responses mirrored the epidemiologic findings, with greater injury
occurring after exposure to an equal mass of particles from those years when the mill was in
operation. Aqueous extracts of the filters collected prior to temporary closure of the steel mill,
during the closure, and after its reopening were instilled (500 jig of extract in 10 mL of sterile
7-46
-------
saline) through a bronchoscope into the lungs of nonsmoking human volunteers. Twenty-four
hours later, the same subsegment was lavaged. Exposure to aqueous extracts of PM collected
before closure and after reopening of the steel mill provoked a greater inflammatory response
than PM extracts from filters taken during the plant shutdown. These results are suggestive of
pulmonary effects of experimental exposure of humans to Utah Valley PM that parallel health
outcomes observed in epidemiologic studies of the human population exposed under ambient
conditions.
Dye et al. (2001) also examined effects of Utah Valley ambient PM on respiratory health in
laboratory animals. Sprague-Dawley rats were intratracheally instilled with equivalent masses
of aqueous extracts (0, 0.83, 3.3, 8.3, or 16 mg extract/kg body weight in 0.3 mL saline) from
filters originally collected during the winter before, during, and after closure of the steel mill.
Twenty-four hours after instillation, rats exposed to extracts of particles collected when the plant
was open developed significant pulmonary injury and neutrophilic inflammation. Additionally,
50% of rats exposed to these extracts had increased airway responsiveness to acetylcholine,
compared to 17 and 25% of rats exposed to saline or the extracts of particles collected when the
plant was closed. By 96 h, these effects were largely resolved, except for increases in lung
lavage fluid neutrophils and lymphocytes in rats exposed to PM extracts from prior to the plant
closing. Analogous effects were observed with lung histologic assessment. Chemical analysis
of extract solutions demonstrated that nearly 70% of the mass in all three extracts appeared to be
sodium-based salts derived from the glass filter matrix (Ca2+). However, extracts of particles
collected when the plant was open contained more sulfate, cationic salts (e.g., Ca, K, Mg), and
certain metals (e.g., Cu, Zn, Fe, Pb, As, Mn, Ni). Although total metal content was < 1% of the
extracts by mass, the greater quantity detected in the extracts of particles collected when the
plant was open suggested that metals may be among the important determinants of the observed
pulmonary toxicity. The authors concluded that the pulmonary effects induced in rats by
exposure to aqueous extracts of local ambient PM filters were in good accord with the
epidemiologic reports of adverse respiratory health effects in Utah Valley residents and with
results from the Molinelli et al. (2002) in vitro study of Utah Valley PM filter extract effects on
human epithelial cells (discussed later in Section 7.4). As with use of other ambient PM, the use
in the Utah Valley studies of dust collected more than 10 years earlier introduces uncertainties
associated with the age and handling of the filters.
7-47
-------
Also of interest are some other new instillation study results. For example, Li et al. (1996,
1997) reported that instillation of ambient PM10 (50-125 jig in 0.2 mL buffered saline) collected
in Edinburgh, Scotland, also caused pulmonary injury and inflammation in rats. In addition,
Brain et al. (1998) examined the effects of instillation of particles (< 3.5 jim) that resulted from
the Kuwaiti oil fires in 1991 compared to effects of urban PM collected in St. Louis (NIST SRM
1648, collected in a bag house in the early 1980s). Brain et al. (1998) showed that, on an equal
mass basis, the acute toxicity of the Kuwaiti oil fire particles was similar to that of urban
particles collected in the United States. At all exposure levels (0.15, 0.75, and 3.75 mg/100 g
body weight), both the Kuwaiti oil fire and St. Louis urban particles significantly increased BAL
neutrophils, macrophages, and levels of albumin and other biomarkers (LDH, MPO) of
lung inflammation.
The fact that instillation of ambient PM collected from different geographical areas has
been shown to cause pulmonary inflammation and injury tends to support epidemiologic studies
that report increased PM-associated respiratory effects in populations living in some of the same
geographical areas (e.g., Utah Valley). On the other hand, the potential exists that lower, more
"realistic" doses associated with ambient PM exposures may activate cells and signaling
pathways not observed with much higher than ambient experimental doses, such that lower-dose
mechanisms may be overwhelmed. Thus, high-dose instillation studies may actually produce
different effects on the lung than inhalation exposures at lower concentrations or doses more
closely paralleling those seen with ambient PM exposures.
7.3.2 ROFA and Other Combustion Source-Related Particles
Because combustion emission sources contribute to the overall ambient air particulate
burden (Spengler and Thurston, 1983), a number of the studies investigating the response of
laboratory animals to particle exposures have used combustion source-related particles (see
Tables 7-6 and 7-7). For example, the residual oil fly ash (ROFA) samples used in many
toxicological studies have been collected from a variety of sources, e.g., boilers, bag houses used
to control emissions from power plants, and from particles emitted downstream of such
collection devices. ROFA has a high content of water soluble sulfate and metals, accounting for
82 to 92% of water-soluble mass, while the water-soluble mass fraction in ambient air varies
from low teens to more than 60% (Costa and Dreher, 1997; Prahalad et al., 1999). More than
7-48
-------
VO
TABLE 7-6. RESPIRATORY EFFECTS OF INTRATRACHEALLY INSTILLED ROFA AND OTHER COMBUSTION
SOURCE-RELATED PARTICULATE MATTER IN HEALTHY LABORATORY ANIMALS3
Species, Gender,
Strain, Age, etc.
Mice, female,
NMRI, 28-32 g
Rats, male, S-D,
60 days old
Rats, male, S-D,
65 days old
Rats, S-D,
65 days old
Particle
Coal fly ash (CFA)
Copper smelter dust
(CMP)
Tungsten carbide
(TC)
ROFA (Florida),
DOFA (Boston),
CFA (RTF, NC)
Ambient PM
(St. Louis; Wash,
DC; Ottawa;
Dusseldorf)
ROFA
(Florida)
ROFA
(Florida)
Dose Particle Size
CMP: 20 ug arsenic/kg, or N/A
CMP: 100 mg particles/kg,
TC alone (100 mg/kg), CFA
alone (100 mg/kg [i.e., 20 ug
arsenic/kg]), CMP mixed
with TC (CMP, 13. 6 mg/kg
[i.e., 20 ug arsenic/kg;
TC, 86.4 mg/kg]) and
Ca3(AsO4)2 mixed with TC
(20 ug arsenic/kg; TC
100 mg/kg)
Total mass: 2.5 mg/rat Combustion
or source PM:
Total transition metal: 1.78-4.17 um
46 ug/rat
Ambient PM:
3.27-4.09 um
2.5 mg (8.3 mg/kg) 1.95 um
500 ug/rat ROFA 1 .95 um
500 ug/rat ROFA plus
Time to PEb
Analysis
1,6, 30 days
post-treatment
lavage for total
protein content,
inflammatory
cell number and
type, and
TNF-a
production
Analysis at
24 and 96 h
following
instillation
Analysis at
24, 96 h PE
Analysis at 24 h
PE
Particle Effects/Comments
Mild inflammation for TC; Ca3(AsO4)2 caused
significant inflammation; CMP caused severe but
transient inflammation; CFA caused persistent
alveolitis. Cytokine production was upregulated in
TC-and Ca3(AsO4) treated animals after 6 and
30 days, respectively; a 90% inhibition of TNF-a
production was still observed at day 30 after CMP
administration 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 eosinophils, PMNs, albumin, and LDH
following exposure to ambient and combustion
source (ROFA, DOFA, CFA) particles; induction
of injury by test samples was determined primarily
by constituent metals and their bioavailability in
both ambient and combustion source PM.
Increased PMNs, protein, LDH at both time points;
bioavailable metals were causal constituents of
pulmonary injury.
ROFA-induced increased neutrophilic inflammation
was inhibited by DMTU treatment, indicating role
Reference
Broeckaert
etal. (1997)
Costa and
Dreher
(1997)
Dreher
etal. (1997)
Dye et al.
(1997)
DMTU
to reactive oxygen species.
-------
TABLE 7-6 (cont'd). RESPIRATORY EFFECTS OF INTRATRACHEALLY INSTILLED ROFA AND OTHER
COMBUSTION SOURCE-RELATED PARTICULATE MATTER IN HEALTHY LABORATORY ANIMALS3
Species, Gender,
Strain, Age, etc. Particle Dose Particle Size
Time to PE"
Analysis
Particle Effects/Comments
Reference
Rats, male,
S-D, 60 days old
Mice, female,
BALB/cJ
7-15 weeks
Rats, male, S-D
Mice, normal
and Hp,
105 day sold
Rats, male,
S-D, 60 days old
Rats, male,
S-D and F-344
(60 days old)
Two ROFA (Florida)
samples (Rl, R2)
also R2s
(supernatant)
2.5 mg (9.4 mg/kg)
ROFA, #6 lo-S
(Florida)
60 ug in saline
(dose 3 mg/kg)
ROFA
(Florida)
ROFA
(Florida)
ROFA
(Florida)
ROFA
(Florida)
500 ug
50 ug
1.0 mg in 0.5 mL saline 1.95 um
8.3 mg/kg
Rl: 1.88 um Analysis at Four of 24 animals treated with R2 or R2s died; none Gavett et al.
4 days PE of Rls animals; more AM, PMN, eosinophils (1997)
R2: 2.03 um 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 worse in R2 and R2s groups.
Rl had twice the saline-leachable sulfate, Ni, and V
and 40 times Fe as R2; R2 had 31 times higher Zn.
<2.5 um Analysis at 1, 3, ROFA caused increases in eosinophils, IL-4 and IL-5 Gavett et al.
8, 15 days and airway responsiveness in ovalbumin-sensitized (1999)
postexposure 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.
3.6 um Analysis at Ferritin and transferrin were elevated; greatest Ghio et al.
4 and 96 h increase in ferritin, lactoferrin, transferrin occurred (1998a)
postexposure 4 h postexposure.
1.95 um Analysis at Diminished lung injury (e.g., decreased lavage fluid Ghioetal.
24 h PE ascorbate, protein, lactate dehydrogenase, (2000b)
inflammatory cells, cytokines) in Hp mice lacking
transferrin; associated with increased metal storage
and transport proteins.
Analysis at Increased PMNs, protein. Kadiiska
24hPE etal. (1997)
1.95 um Sacrificed at Increase in neutrophils in both S-D and F-344 rats; Kodavanti
og = 2.14 24 h PE a time-dependent increase in eosinophils occurred in etal. (1996)
S-D rats but not in F-344 rats.
-------
TABLE 7-6 (cont'd). RESPIRATORY EFFECTS OF INTRATRACHEALLY INSTILLED ROFA AND OTHER
COMBUSTION SOURCE-RELATED PARTICULATE MATTER IN HEALTHY LABORATORY ANIMALS3
Species, Gender,
Strain, Age, etc. Particle Dose
Rats, male, S-D, ROFA 8.3 mg/kg
WISTAR, and (Florida)
F-344 (60 days
old)
Time to PE"
Particle Size Analysis
1 .95 um Sacrificed at 6,
og = 2.14 24, 48, and
72 h; 1,3,
and 12 weeks
Particle Effects/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 WISTAR 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.
(1997b)
Rats, male, S-D,
60 days old
Rats, male, S-D,
60 days old
ROFA
(Florida)
Fe2(S04)3,
VS04,
NiS04
10 compositionally
different ROFAs
from Boston area
power plant
8.3 mg/kg
ROFA-equivalent dose of
metals
0.83, 3.3, 8.3 mg/kg
Rats, male, S-D
60-day-old
treated with
MCT (60 mg/kg)
Rats, male,
WKY and SH,
11-13 weeks old
ROFA
(Florida)
ROFA
(Florida)
VS04,
NiSO4, or saline
0, 0.83, 3.3 mg/kg
3.3 mg/mL/kg
1.5 umol/kg
1.95 um Analysis at 3, ROFA-induced pathology lesions were as severe as Kodavanti
og = 2.14 24, and 96 h, those caused by Ni. Metal mixture caused less injury etal.
postinstillation than ROFA or Ni alone; Fe was less pathogenic. (1997a)
Cytokine and adhesion molecule gene expression
occurred as early as 3 h after exposure. V-induced
gene expression was transient, butNi caused
persistent expression and injury.
1.99-2.67 um Sacrificed at ROFA-induced increases in BAL protein and LDH, Kodavanti
24 h but not PMN, associated with water-leachable total et al.
metal, Ni, Fe, and S; BALF neutrophilic (1998a)
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 + V. Only data for the 8.3 mg/kg dosed
group were reported.
1.95 um 24-96 h Dose-dependent increase in BALF protein and LDH Kodavanti
og = 2.19 activity and neutrophilic inflammation. Effects were et al. (1999)
variable due to high mortality. 58% of rats exposed
to ROFA died within 96 h.
1.95 um 1 and 4 days; Increased BALF protein and LDH alveolitis with Kodavanti
postinstillation macrophage accumulation in alveoli; increased et al. (2001)
og = 2.14 analysis at 6 or neutrophils in BAL. Increased pulmonary protein
24 h 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.
-------
to
TABLE 7-6 (cont'd). RESPIRATORY EFFECTS OF INTRATRACHEALLY INSTILLED ROFA AND OTHER
COMBUSTION SOURCE-RELATED PARTICULATE MATTER IN HEALTHY LABORATORY ANIMALS3
Species, Gender,
Strain, Age, etc.
Rats, female,
Brown Norway
8-10 weeks old
Rats, male, S-D,
60 days old
Rats, male, S-D;
60 days old
Rats, male, S-D,
60 days old
Rats, male, S-D,
60 days old
Rats, S-D
Particle
ROFA (Florida)
and HDM
ROFA #6
(Florida)
DOFA (NC)
ROFA #6
(Florida)
NiSO4
VS04
ROFA
(Cayman Chemical,
Ann Arbor, MI)
ROFA, #6 LoS
(Florida)
DPM
Dose
200 ug or 1000 ug
1000 uginO.5 mL saline
1000 ug in 0.5 mL saline
3.3 mg/mL/kg; ROFA
equivalent dose of metals
400 and 1000 ug/mL
(200 and 500 ug ROFA in
0.5 mL saline)
500 ug in 0.5 mL saline
500 ug in 0.5 mL saline
Time to PE"
Particle Size Analysis Particle Effects/Comments
1.95 N/A ROFA enhanced the response to house dust mite
(HDM) antigen challenge. Eosinophil numbers and
LDH were increased in highest exposed groups.
BAL protein and IL-10 were increased in both
ROFA groups + HDM versus HDM alone.
1.95±0.18um 15minto24h Production of acetaldehyde increased at 2 h
postinstillation.
1 5 min to 24 h ROFA induced production of acetaldehyde with a
peak at about 2 h. No acetaldehyde was seen in
plasma at any time. DOFA increased acetaldehyde,
as did V, Fe.
1 .9 um 3 or 24 h Inflammatory and stress responses were upregulated;
og = 2.14 the numbers of genes upregulated were correlated
with metal type and ROFA
N/A 12 h post-IT ROFA increased PGE2 via cycloxygenase expression
in the 400 ug/mL group. PGE2 depressed in
1000 ug/mL group by COX2 inhibitor.
3.6 um 1,4, or 24 h Mild and variable inflammation at 4 h;
no pronounced inflammation until 24 h when there
were marked increases in P-Tyr and P-MAPKS.
N/A 3 times/week, Decreased concentration of lav age ascorbate. Urate
2 weeks and glutathione concentrations unchanged; elevated
MIP-2 and TNF; total cell count increased; lavage
protein and LDH increased; increased total lavage
iron concentration.
Reference
Lambert
etal. (1999)
Madden
etal. (1999)
Nadadur
etal.
(2000);
Nadadur
and
Kodavanti
(2002)
Samet et al.
(2000)
Silbajoris
et al. (2000)
Ohio etal.
(2000b)
aCFA = Coal fly ash
CMP = Copper smelter dust
DOFA = Domestic oil-burning furnace fly ash
ROFA = Residual oil fly ash
TC = Tungsten carbide
Fe2(SO4) = Iron sulfate
VSO4 = Vanadium sulfate
NiSO4 = Nickel sulfate
LoS = low sulfur
MCT = Monocrotaline
OVA = Ovalbumin
b PE = Post Exposure
-------
TABLE 7-7. RESPIRATORY EFFECTS OF INHALED AND INSTILLED ROFA AND OTHER COMBUSTION SOURCE-
RELATED PARTICULATE MATTER IN COMPROMISED LABORATORY ANIMAL MODELS3
Species, Gender,
Strain, Age, etc.
Inhalation
Rats, male,
WISTAR
BorWISW strain
n = 20
Mice, BALB/C,
2-day-old,
sensitized to
ovalbumin (OVA)
Rats, S-D, 250 g
MCT
Rats, male, S-D
60-day-old treated
with MCT
(60 mg/kg)
Exposure Concentration/
Particle Technique Dose Particle Size
Coal fly ash Inhalation 0,11, 32, and 1.9-2.6 |im
(CFA) (chamber) 103mg/m3 og=1.6-1.8
Aerosolized Nose-only 50 mg/mL N/A
ROFA inhalation
leachate
ROFA Inhalation 580 ± 110 |ig/m3 2.06 urn
(Boston) °g=1-57
ROFA Nose-only 15mg/m3 1.95 urn
(Florida) inhalation og = 2.14
Exposure
Duration; Time
to PEb Analysis
6 h/day,
5 days/week,
for 4 weeks
30 min
6 h/day for 3 days
6 h/day for 3 days
analysis at 0 or 1 8 h
Particle Effects/Comments
At 103 mg/m3, type II cell proliferation, mild
fibrosis and increased perivascular
lymphocytes seen. At lowest concentration,
main changes seen 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.
Mortality seen only in MCT rats exposed to
ROFA. Neutrophils in lavage fluid increased
significantly in MCT rats exposed to ROFA
versus filtered air. MIP-2 mRNA expression
induced in lavage cells 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).
Reference
Dormans et al.
(1999)
Hamada et al.
(1999)
Killing sworth
etal. (1997)
Kodavanti
etal. (1999)
-------
TABLE 7-7 (cont'd). RESPIRATORY EFFECTS OF INHALED AND INSTILLED ROFA AND OTHER COMBUSTION
SOURCE-RELATED PARTICULATE MATTER IN COMPROMISED LABORATORY ANIMAL MODELS3
Species, Gender,
Strain, Age, etc.
Inhalation (cont'd)
Rats, male, WKY and
SH, 1 1-1 3 weeks old
Particle
ROFA
(Florida)
Exposure
Technique
Nose-only
Inhalation
Concentration/
Dose Particle Size
15 mg/m3 1.95 |im
og = 2.14
Exposure
Duration; Time
to PEb Analysis
6 h/day x 3 days,
analysis at 0 or 18 h
Particle Effects/Comments
More pulmonary injury in SH rats.
Increased RBCs in BAL of SH
Reference
Kodavanti
et al. (2000b)
rats. ROFA increased airway
reactivity to acetylcholine in both
SH and WKY rats. Increased
protein, albumin, and LDH in
BAL 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.
Mice, male,
Swiss- Webster, 6-8
weeks old (A/J, 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
Rats, F-344
8 weeks, 20 months old
Mice, TSK
14- 17 months old
Rats, male, S-D,
MCT-treated
Carbon black Nose only
Regal 660 inhalation
Carbon-
associated
S04=
Carbon Inhalation
Fluorescent Inhalation
microspheres
10 mg/m3 (carbon) 0.29 urn ± 2.7 ^m
10 ppm SO2
285 ng/m3
(average
concentration of
particle-associated
sulfates)
100 ng/m3 and/or UF
1.0 ppm O3
following
exposure to
endotoxin (12 min
to 70 EU)
3. 85 ±0.81 1.38 ±0.10 urn
mg/m3 og= 1.8 ±0.28
4 h Differences in inflammatory Ohtsuka et al.
responses (PMN) across strains. (2000a,b)
Appears to be genetic component
to the observed differences in
susceptibility.
6 h Small effect on lung inflammation Elder et al.
and activation of inflammatory (2000a,b)
cells. Effects enhanced in
compromised lung and in older
animals. Greatest effect in
compromised lung exposed to UF
carbon and O3.
3 h/day x 3 days MCT-treated animals had fewer Madl et al.
microspheres in their (1998)
macrophages, probably because of
impaired chemo taxis.
-------
TABLE 7-7 (cont'd). RESPIRATORY EFFECTS OF INHALED AND INSTILLED ROFA AND OTHER COMBUSTION
SOURCE-RELATED PARTICULATE MATTER IN COMPROMISED LABORATORY ANIMAL MODELS3
Species, Gender,
Strain, Age, etc.
Instillation
Rats, male, S-D;
60-day-old; WKY and
SH; cold-stressed SH;
O3-exposed SH;
MCT-treated SH
Rats, male, S-D
(200 g)
Exposure Concentration/
Particle Technique Dose
ROFA Intratracheal 0, 0.25, 1.0, and
(source not instillation 2.5mg/rat
specified),
Ottawa dust,
MSH Vol.
Ash
Diesel, Intratracheal 1 mg in 0.4 mL.
SiO2, instillation
carbon black
Particle Size
1.95 |im
DEP collected
as TSP-
disaggregated in
solution by
sonication (20 nm);
SiO2 (7 nm);
carbon black
Exposure
Duration; Time
to PEb Analysis
96 h post-IT
Necropsy at 2, 7,
21,42, and 84 days
postinstillation
Particle Effects/Comments Reference
ROFA instillation caused acute, Watkinson
dose-related increase in pulmonary et al.
inflammation. Data on Ottawa (2000a,b)
dust and volcanic ash not reported.
Amorphous SiO2 increased Murphy et al.
permeability and neutrophilic (1998)
inflammation. Carbon black
and DEP translocated to
interstitum and lymph nodes
by 12 weeks.
aCFA = Coal fly ash
CMP = Copper smelter dust
DOFA = Domestic oil-burning furnace fly ash
ROFA = Residual oil fly ash
TC = Tungsten carbide
Fe2(SO4) = Iron sulfate
VSO4 = Vanadium sulfate
NiSO4 = Nickel sulfate
LoS = low sulfur
MCT = Monocrotaline
OVA = Ovalbumin
PE = Post Exposure
-------
90% of the metals in ROFA are transition metals; whereas these metals typically represent only a
very small subfraction of the total ambient PM mass of U.S. monitoring samples. Thus, the dose
of bioavailable metal that is delivered to the lung when ROFA is instilled into a laboratory
animal can be orders of magnitude greater than an ambient PM dose, even under a worst-case
scenario. Transition metals generate reactive oxygen species (ROS) pertinent to understanding
of one proposed mechanism of PM toxicity and of PM components possibly contributing to toxic
responses.
Intratracheal instillation of various doses of ROFA suspension has been shown to produce
severe inflammation, an indicator of pulmonary injury that includes recruitment of neutrophils,
eosinophils, and monocytes into the airway. The biological effects of ROFA in rats have been
shown to depend on aqueous teachable chemical constituents of the particles (Dreher et al.,
1997; Kodavanti et al., 1997a). A leachate prepared from ROFA, containing predominantly Fe,
Ni, V, Ca, Mg, and sulfate, produced lung injury similar to that induced by the complete ROFA
suspension (Dreher et al., 1997). Depletion of Fe, Ni, and V from the ROFA leachate eliminated
its pulmonary toxicity. Correspondingly, minimal lung injury was observed in animals exposed
to saline-washed ROFA particles. A surrogate transition metal sulfate solution containing Fe, V,
and Ni largely reproduced the lung injury induced by ROFA. Interestingly, ferric sulfate and
vanadium sulfate antagonized the pulmonary toxicity of nickel sulfate. Interactions between
different metals and the acidity of PM were found to influence the severity and kinetics of lung
injury induced by ROFA and its soluble transition metals.
To further investigate the response to ROFA with differing metal and sulfate composition,
Kodavanti et al., (1997a) instilled male Sprague-Dawley rats (60 days old) intratracheally with
ROFA (2.5 mg/rat) or metal sulfates (Fe -0.54 |imole [105 |ig]/rat, V -1.7 |imole [245 |ig]/rat,
and Ni -1.0 jimole [263 |ig]/rat, individually or in combination). Transition metal sulfate
mixtures caused less injury than ROFA or Ni alone, suggesting metal interactions. This study
also showed that V-induced effects were less severe than that of Ni and were transient. Ferric
sulfate was least pathogenic. Cytokine gene expression was induced prior to the pathology
changes in the lung, and the kinetics of gene expression suggested persistent injury by nickel
sulfate. Another study by the same investigators was performed using 10 different ROFA
samples collected at various sites within a power plant burning residual oil (Kodavanti et al.,
1998a). Animals received intratracheal instillations of either saline (control), or a saline
7-56
-------
suspension of whole ROFA (< 3.0 jim MMAD for all ground PM) at three doses (0.83, 3.33, or
8.33 mg/kg). This study showed that ROFA-induced PMN influx was associated with its water-
leachable V content; but protein leakage was associated with water-leachable Ni content.
ROFA-induced in vitro activation of AMs was highest with ROFA containing teachable V but
not with Ni plus V, suggesting that the potency and the mechanism of pulmonary injury may
differ between emissions containing bioavailable V and Ni.
Other studies have shown that soluble metal components play an important role in the
toxicity of emission source particles. Gavett et al. (1997) investigated the effects of two ROFA
samples of equivalent diameters, but having different metal and sulfate content, on pulmonary
responses in Sprague-Dawley rats. ROFA sample 1 (Rl; the same emission particles used by
Dreher et al. [1997]) had approximately twice as much saline-leachable sulfate, Ni, and V, and
40 times as much Fe as ROFA sample 2 (R2), whereas R2 had a 31-fold higher Zn content. Rats
were instilled with suspensions of 2.5 mg R2 in 0.3 mL saline, the supernatant of R2 (R2s), the
supernatant of 2.5 mg Rl (Rls), or saline only. By 4 days after instillation, 4 of 24 rats treated
with R2s or R2 had died. None treated with Rls or saline died. Pathological indices, such as
alveolitis, early fibrotic changes, and perivascular edema, were greater in both R2 groups. In
surviving rats, baseline pulmonary function parameters and airway hyperreactivity to
acetylcholine were significantly worse in the R2 and R2s groups than in the Rl s groups. Other
than BAL neutrophils, which were significantly higher in the R2 and R2s groups, no other
inflammatory cells (macrophages, eosinophils, or lymphocytes) or biochemical parameters of
lung injury were significantly different between the R2 and R2s groups and the Rls group.
Although (a) soluble forms of Zn had been found in guinea pigs to produce a greater pulmonary
response than other sulfated metals (Amdur et al., 1978) and (b) the level of Zn was 30-fold
greater in R2 than Rl, the precise mechanisms by which Zn may induce such responses are
unknown. Still, these results show that the composition of soluble metals and sulfate is critical
in the development of airway hyperractivity and lung injury produced by ROFA, albeit at very
high instilled doses.
Dye et al. (1997) pretreated rats with an intraperitoneal (IP) injection of 500 mg/kg
dimethylthiourea (DMTU) or saline, followed 30 min later by intratracheal instillation of either
acidic saline (pH = 3.3) or an acidified suspension of ROFA (500 |ig/rat). Dimethylthiourea
reduces the activity of the reactive oxygen species. The systemic administration of DMTU
7-57
-------
impeded development of the cellular inflammatory response to ROFA but did not ameliorate
biochemical alterations in BAL fluid. In a subsequent study, it was determined that oxidant
generation, possibly induced by soluble V compounds in ROFA, is responsible for the
subsequent rat tracheal epithelial cells gene expression, inflammatory cytokine production
(MIP-2 and IL-6), and cytotoxicity (Dye et al., 1999).
In parallel work on the potential importance of metals in mediating ambient PM effects,
Kodavanti et al. (2002b) studied the role of Zn in PM-induced health effects in several animal
models. Male Sprague-Dawley (SD) rats were instilled IT with ROFA in saline (0.0, 0.8, 3.3,
or 8.3 mg/kg) from a Boston area power plant. Also, in order to evaluate the potential role of
teachable Zn, additional rats were instilled with either saline, whole ROFA suspension, the
saline teachable fraction of ROFA, the parti culate fraction of ROFA (8.3 mg/kg, soluble Zn =
14.5 |ig/mg ROFA), or ZnSO4 (0.0, 33.0, or 66.0 |ig/kg Zn). Three rat strains that differ in PM
susceptibility, i.e., male SD, normotensive WKY, and spontaneously hypertensive (SH) rats,
were exposed at age 90 days nose-only to either filtered air or ROFA (2, 5, 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, or 16 weeks)
and assessed at 2 days postexposure. Intratracheal exposures to whole ROFA suspensions
produced a dose-dependent increase in protein/albumin permeability and neutrophilic
inflammation. Pulmonary protein/albumin leakage and neutrophilic inflammation caused by the
teachable fraction of ROFA and ZnSO4 were comparable to effects of the whole suspension.
However, protein/albumin leakage was not associated with the particulate fraction, although
significant neutrophilic inflammation did occur after instillation. With ROFA nose-only
inhalation, acute exposures (10 mg/m3 only) for 4 days resulted in small increases BAL protein
and N-acetyl glucosaminidase activities (-50% above control); but, unlike with IT exposures, no
neutrophilic influx was detectable in BAL from any of the inhalation groups. The only major
effect of acute and long-term ROFA inhalation was a dose- and time-dependent increase in
alveolar macrophages (AM), regardless of rat strain. Histological evidence also showed dose-
and time-dependent accumulations of particle-loaded AM. Particles were also evident in
interstitial spaces and in the lung-associated lymph nodes following the inhalation exposures
(SH > WKY = SD). There were strain-related differences in peripheral WBC counts and plasma
fibrinogen, but no major ROFA inhalation effect. The authors attributed the differences in
pulmonary responsiveness to ROFA between IT and inhalation exposures to the dose of
7-58
-------
bioavailable zinc; the IT ROFA exposures, but not acute and long-term inhalation of up to
10 mg/m3, caused neutrophilic inflammation.
In addition to transition metals, other components in fly ash also may cause lung injury.
The effects of arsenic compounds in coal fly ash or copper smelter dust on the lung integrity and
on the ex vivo release of TNF-a by alveolar phagocytes were studied by Broeckaert et al. (1997).
Female NMRI mice were instilled with different particles normalized for arsenic content
(20 ng/kg body weight [i.e., 600 ng arsenic/mouse]) and particle load (100 mg/kg body weight
[i.e., 3 mg/mouse]). Mice received tungsten carbide (TC) alone, coal fly ash (CFA) alone,
copper smelter dust (CMP) mixed with TC, and Ca3(AsO4)2 mixed with TC (see Table 7-6 for
concentration details). Copper smelter dust caused a severe but transient inflammatory reaction;
whereas a persisting alveolitis (30 days postexposure) was seen after treatment with coal fly ash.
Also, TNF-a production in response to lipopolysaccharide (LPS) by alveolar phagocytes was
significantly inhibited at day 1, but was still observed at 30 days after administration of CMP
and CFA. Although arsenic was cleared from the lung tissue 6 days after Ca3(AsO4)2
administration, a significant fraction persisted (10 to 15% of the As administered) in the lung of
CMP- and CFA-treated mice at day 30 postexposure. Hence, suppression of TNF-a production
may be dependent on the slow elimination of particles and their metal content from the lung.
Antonini et al. (2002) investigated effects of preexposure to ROFA on lung defenses and
injury after pulmonary challenge with Listeria monocytogenes, a bacterial pathogen. Male SD
rats were dosed IT at day 0 with saline (control) or ROFA (0.2 or 1 mg/100 g body weight).
Three days later, both groups of rats were instilled IT with a low (5 x 103) or high (5 x 10s) dose
of L. monocytogenes. Chemiluminescence (CL) and nitric oxide (NO) production, two indices
of AM function, were measured for BAL cells from the right lungs. The left lungs and spleens
were homogenized, cultured, and colony-forming units were counted after overnight incubation.
Exposure to ROFA and the high dose of L. monocytogenes led to marked lung injury and
inflammation as well as to an increase in mortality, compared with rats treated with saline and
the high dose of L. monocytogenes. Preexposure to ROFA significantly enhanced injury and
delayed pulmonary clearance of L. monocytogenes at both bacterial doses when compared to the
saline-treated control rats. ROFA had no effect on AM CL but caused a significant suppression
of AM NO production. The authors concluded that acute exposure to ROFA slowed pulmonary
7-59
-------
clearance of L. monocytogenes and altered AM function, changes that could lead to increased
susceptibility to lung infection in exposed populations.
In summary and as indicated in Table 7-6, intratracheally instilled high doses of ROFA
produced acute lung injury and inflammation. Water soluble metals in ROFA appear to play a
key role in the acute effects of instilled ROFA through the production of reactive oxygen
species. These ROFA studies clearly show that combustion-generated particles with a high
metal content can cause substantial lung injury; but how well such effects can be extrapolated to
help understand ambient PM exposure effects in humans remains to be more fully established.
Results listed in Table 7-7 indicate that in a few cases both normal and compromised
animals have similar responses to ROFA and other combustion source-related PM. However,
most studies show that responses are seen at lower concentrations/doses in compromised
animals. Additionally, ROFA has been shown to exacerbate lesions and inflammation created
by MCT-pretreatment and to induce more pulmonary injury in SH rats. Strain-specific
differences have also been noted in inflammatory responses to carbon and ROFA, indicating a
genetic component to the differences in susceptibility.
7.3.3 Metals
Results from occupational and laboratory animal studies reviewed in the 1996 PM AQCD
indicated that acute exposures to very high levels (hundreds of |ig/m3 or more) or chronic
exposures to lower levels (as low as 15 |ig/m3) of metallic particles could affect the respiratory
tract. It was concluded, on the basis of data available at that time, that the metals at typical
concentrations present in the ambient atmosphere (1 to 14 |ig/m3) were not likely to have a
significant acute effect in healthy individuals. This included metals such as As, Cd, Cu, Ni, V,
Fe, and Zn. Other metals found at concentrations less than 0.5 |ig/m3 were not reviewed in the
1996PM AQCD.
More recently published data from controlled experimental exposure studies, however, are
suggestive of particle-associated metals possibly being among PM components contributing to
health effects attributed to ambient PM. Included among such studies are a number of the
"ambient PM," ROFA, and other "combustion-source" studies assessed in the preceding two
sections which included analyses of potential contributions of metals to observed effects. Other
7-60
-------
new studies on effects of laboratory-generated metals/metal compounds are summarized in
Table 7-8.
Iron is the most abundant of the elements capable of catalyzing oxidant generation and is
also present in ambient urban particles. Lay et al. (1998) and Ohio et al. (1998b) tested the
hypothesis that the human respiratory tract will attempt to diminish the added, iron-generated
oxidative stress. They examined cellular and biochemical responses of human subjects instilled,
via the intrapulmonary route, with a combination of iron oxyhydroxides that introduced an
oxidative stress to the lungs. Saline alone and iron-containing particles suspended in saline were
instilled into separate lung segments of human subjects. Subjects underwent bronchoalveolar
lavage at 1 to 91 days after instillation of 2.6-|im diameter iron oxide (~5 mg or 2.1 x 108
particles) agglomerates. Lay and colleagues found iron oxide-induced inflammatory responses
in both the alveolar fraction and the bronchial fraction of the lavage fluid at 1 day after
instillation. Lung lavage 24 h after instillation revealed decreased transferrin concentrations and
increased ferritin and lactoferrin concentrations, consistent with a host-generated response to
decrease the availability of catalytically reactive iron (Ohio et al., 1998b). Normal iron
homeostasis returned within 4 days of the iron particle instillation. The same iron oxide
preparation, which contained a small amount of soluble iron, produced similar pulmonary
inflammation in rats. In contrast, instillation of rats with two iron oxide preparations that
contained no soluble iron failed to produce injury or inflammation, thus suggesting that soluble
iron was responsible for the observed intrapulmonary changes.
In a subsequent inhalation study, Lay et al. (2001) studied the effect of iron oxide particles
on lung epithelial cell permeability. Healthy, nonsmoking human subjects inhaled 12.7 mg/m3
low- and high-solubility iron oxide particles (MMAD =1.5 jim and og = 2.1) for 30 minutes.
Neither pulmonary function nor alveolar epithelial permeability, as assessed by pulmonary
clearance of technetium-labeled DPTA, was changed at 0.5 or 24 h after exposure to either type
of iron oxide particle. Ohio et al. (2001) reported a case study, however, in which acute
exposure to oil fly ash from a domestic oil-fired stove produced diffuse alveolar damage,
difficulty in breathing, and symptoms of angina. Elemental analyses revealed high metal content
(Fe, V, etc.) in fly ash samples; and other evaluations suggested that the high metal content of oil
fly ash altered the epithelial cell barrier in the alveolar region.
7-61
-------
TABLE 7-8. RESPIRATORY EFFECTS OF INHALED AND INSTILLED METAL PARTICLES IN HUMAN SUBJECTS
AND LABORATORY ANIMALS
to
Species, Gender,
Strain, Age, etc.
Inhalation
Humans, healthy
nonsmokers;
8 M, 8 F;
1 8-34 years old
Rats, SD;
60 days old
Instillation
Humans, healthy
nonsmokers;
12 M, 4 F;
18-3 5 years old
Humans, healthy
nonsmokers;
27 M, 7 F; 20-36
years old.
Mice, NMRI;
Mouse peritoneal
macrophage
Exposure
Exposure Duration/Time
Particle Technique Concentration Particle Size to Analysis
Fe2O3 Inhalation 12.7mg/m3 1.5 |im 30 min
og = 2.1
VSO4 Inhalation 0.3-1.7mg/m3 N/A 6h/dayx4days
NiS04 0.37-2.1mg/m3
Colloidal Bronchial SmginlOmL 2.6 |im 1,2, and 4 days
iron oxide instillation after instillation
Fe2O3 Intrapulmonary 3 x 10s microspheres 2.6 |im N/A
instillation in 10 mL saline.
MnO2 Intratracheal 0.037,0.12,0.75, surface area Sacrificed at 5 days
instillation; 2.5mg/animal of 0.16, 0.5,
in vitro 17, 62 m2/g
Particle Effects/Comments
No significant difference in 98mTc-DTPA
clearance half-times, DLCO, or spirometry
V did not induce any significant changes in
BAL or HR. Ni caused delayed
bradycardia, hypothermia, and
arrhythmogenesis at> 1.3 mg/m3. Possible
synergistic effects were found.
L-ferritin increased after iron oxide particle
exposure; transferrin was decreased. Both
lactoferrin and transferrin receptors were
increased.
Initially -induced transient inflammation
(neutrophils, protein, LDH, IL-8) resolved
by 4 days postinstillation.
LDH, protein and cellular recruitment
increased in a dose-related manner with
increasing surface area for particles with
surface areas of 17 and 62 nvYg; freshly
ground particles with surface areas of
0.5 m2/g had enhanced cytotoxicity.
Reference
Lay et al.
(2001)
Campen
etal. (2001)
Ohio et al.
(1998b)
Lay et al.
(1998)
Lison et al.
(1997)
-------
TABLE 7-8 (cont'd). RESPIRATORY EFFECTS OF INHALED AND INSTILLED METAL PARTICLES IN HUMAN
SUBJECTS AND LABORATORY ANIMALS
Species, Gender,
Strain, Age, etc. Particle
Instillation (cont'd)
Rats, Female, CD NaVO3
VOSO4
VA
Mice, Swiss EHC-93
soluble
metal
salts
Rats, M, F344, TiO2
175-225 g
Rats, M, F344, TiO2
175-225 g
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
inhalation and
Intratracheal
instillation
Intratracheal
inhalation and
Intratracheal
instillation
Concentration
21 or 210 ng V/kg
(NaVO3, VOSO4
soluble)
42 or 420 ng V/kg
(V2OS) less soluble
lmgin0.1mLH2O
Inhalation at
125 mg/m3 for 2 h;
Instillation at
500 |ig for fine,
750 |ig for ultrafine
Inhalation at
125 mg/m3 for 2 h;
Instillation at 500 |ig
for fine, 750 |ig for
ultrafine
Exposure
Duration/Time
Particle Size to Analysis
N/A 1 h or 10 days
following
instillation
0.8 ±0.4 urn 3 days
Fine: 250 nm Inhalation exposure,
Ultrafine: 2 h; sacrificed at 0,
21 nm 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
Particle Effects/Comments
PMN influx was greatest following VOSO4,
lowest for V2O5 (no effect at lowest
concentration); 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.
Solution containing all metal salts (Al, Cu,
Fe, Pb, Mg, Ni, Zn) or ZnCl alone
increased BAL inflammatory cells and
protein.
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.
Reference
Pierce et al.
(1996)
Adamson
et al. (2000)
Osier and
Oberdorster
(1997)
Osier et al.
(1997)
CdO = Cadmium oxide
Fe2O3 = Iron oxide
MgO = Magnesium oxide
MnO2 = Manganese oxide
NaVO3 = Sodium metavanadate
TiO2 = Titanium oxide
VOSO4 = Vanadyl sulfate
V2O5 = Vanadium pentoxide
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
-------
Several of the other studies summarized in Table 7-8 provide evidence that several
different metal salts (when instilled intratracheally in rats or mice at relatively high doses) can
produce inflammatory responses in the lung as indicated by various markers (e.g., increased
BAL PMNs or other inflammatory cells, induction of cytokines, etc). Two of the studies (by
Osier and Oberdorster, 1997; Osier et al., 1997) further indicate that (a) ultrafme metal particles
are more effective than fine particles in producing the inflammation and (b) intratracheal
inhalation is less effective than instillation in producing the inflammation. Analogously,
Campen et al. (2001) did not observe any significant changes in BAL markers with 6 h/day
inhalation exposure of rats for 4 days to VOSO4 at concentrations ranging up to 2.1 mg/m3. The
results of these metal studies and their potential significance are elaborated on further in later
sections (e.g., Section 7.4.4.1) of this chapter.
7.3.4 Acid Aerosols
Extensive earlier studies (conducted up to the early 1990s) on the effects of controlled
exposures to aqueous acid aerosols on various aspects of lung function in humans and laboratory
animals were reviewed in an EPA Acid Aerosol Issue Paper (U.S. Environmental Protection
Agency, 1989) and in the 1996 PM AQCD. Methodology and measurement methods for
controlled human exposure studies were also reviewed elsewhere (Folinsbee et al., 1997).
The studies summarized in the 1996 PM AQCD illustrate that aqueous acidic aerosols have
minimal effects on symptoms and mechanical lung function in young healthy adult volunteers at
concentrations as high as 1000 |ig/m3. Asthmatic subjects appear to be more sensitive to the
effects of acidic aerosols on mechanical lung function. Responses have been reported in
adolescent asthmatics at concentrations as low as 68 |ig/m3, and modest bronchoconstriction has
been seen in adult asthmatics exposed to concentrations > 400 |ig/m3, but the available data are
not consistent. However, at levels as low as 100 |ig/m3, acid aerosols can alter mucociliary
clearance. Brief exposures (< 1 h) to low concentrations (« 100 |ig/m3) may accelerate clearance
while longer (multihour) exposures to higher ones (> 100 |ig/m3) can depress clearance.
Some earlier acid aerosol studies not assessed in the 1996 PM AQCD or published more
recently are summarized in Table 7-9. For example, Frampton et al. (1992) found that acid
aerosol exposure in humans (1000 |ig/m3 H2SO4 for 2 h) did not result in airway inflammation
and there was no evidence of altered macrophage host defenses. Also, Leduc et al. (1995) found
7-64
-------
TABLE 7-9. RESPIRATORY EFFECTS OF ACID AEROSOLS IN HUMANS AND LABORATORY ANIMALS3
Species, Gender,
Strain, Age, etc.
Humans, healthy
nonsmokers;
10M,2F;
20-39 years old
Humans, asthmatic;
13 M, 11 F
Rabbits,
New Zealand white
Humans, healthy
nonsmokers; 10M,
2 1-37 years old
Rats, female,
F-344; Guinea Pigs,
female, Hartley
Dogs, beagle,
healthy; n = 16
Mice, BALB/c,
6-8 weeks old,
normal and
sensitized to OVA
(ovalbumin)
Particle
H2S04
aerosol
NaCl
(control)
H2S04
aerosol
NH+4/SO-4
aerosol
H2S04
H2S04
aerosol
Neutral
sulfite aerosol
Acidic sulfate
aerosol
Ammonium
Bisulfite
(NH4HS04)
Exposure
Technique
Inhalation
Inhalation
by face
mask
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation,
nose only
Concentration Particle Size
1000 jjg/m3 0.8-0.9 urn
MMAD
500 ng/m3 9 urn
MMAD
Turn
MMAD
1000 ng/m3 0.8 urn
og1.6
94 mg/m3 0.80
og1.89
43 mg/m3 0.93
og2.11
1.5 mg/m3 1.0 |im
MMAD
og = 2.2
5. 7 mg/m3 1.1 urn
MMAD
og = 2.0
78 ng/m3 0.53 |im
972|ig/m3 0.45 jim
235 |ig/m3 0.085 jim
(MMD)
Exposure
Duration
2h;
analysis
18 h
Ih
2h
4h
16.5 h/day
for 13
months
6 h/day for
13 months
4 h/day for
3 days,
analyzed at
1 or 4 days
PE
Particle Effects/Comments
No inflammatory responses; slight increase in BAL
protein and slight decrease in albumin in H2SO4
subjects compared to NaCl. No effect on bacterial
killing by macrophages was found.
Exposure to simulated natural acid fog did not
induce bronchoconstriction nor change bronchial
responsiveness in asthmatics.
No inflammatory response; LDH activity in BAL
elevated in both species; effect on bacterial killing
by humans was inconclusive.
Acid aerosol increased surfactant film
compressibility in guinea pigs.
Long-term exposure to particle-associated sulfur
and hydrogen ions caused only subtle respiratory
responses and no change in lung pathology.
No changes in BAL NAG, LDH, or protein with
any of the particles. Small, non-relevant changes
IL-4, IL-6 and TNFa. No treatment related effects
on lung histopathology or serum IgE levels.
Suggest that increases in asthma are not due to
NH4HS04)
Reference
Frampton
etal. (1992)
Leduc et al.
(1995)
Zelikoffetal.
(1997)
Lee et al.
(1999)
Heyder et al.
(1999)
Cassee et al.
(1998a)
-------
TABLE 7-9 (cont'd). RESPIRATORY EFFECTS OF ACID AEROSOLS IN HUMANS AND LABORATORY ANIMALSa
Oi
Species, Gender,
Strain, Age, etc.
Mice, BALB/c,
6-8 weeks old,
normal and
sensitized to OVA
Mice, BALB/c,
6-8 weeks old,
normal and
sensitized to OVA
Rats, SD,
6-8 weeks old,
normal and
MCT-treated
Particle
Ammonium
Ferrosulfate
(NH4)2Fe(S04)2
•6H2O
Ammonium
Nitrate
(NH4N03)
Ammonium
Bisulfite,
Ammonium
Ferrosulfate,
Ammonium
Nitrate, or
Fine CB
Exposure
Technique
Inhalation,
nose only
Inhalation,
nose only
Inhalation,
nose only
Concentration
250 [ig/m3
140 Lig/m3
250 [ig/m3
70-420
275-410
[ig/m3
2-9 mg/m3
Particle
Size
0.459 |im
(MMD)
0.3 |im
0.03 |^m
(CMD)
0.070-0.1
0.57-0.64
0.6 MMD
Exposure
Duration
4h/day
for 3 days,
analyzed at
1 day PE
4h/day
for 3 days,
analyzed at
1 day PE
4h/day
for 3 days,
analyzed at
1 day PE
Particle Effects/Comments
No changes in BAL NAG, LDH, or protein.
Marginal, nonsignificant changes in TNFoc and
cell differential. No indications of enhanced
allergic response with (NH4)2Fe(SO4)2«6H2O.
Increases in BAL NAG with 0.3 [im particle
only. Increased neutrophils and decreased
AMs with 0.03 [im particle only. No other
exposure-related effects. Authors concluded
that both mass concentration and specific size
of the particles determine adverse effects of
NH4NO3 exposure.
No significant exposure-related changes in
BAL NAG, LDH, protein, cytokines, lung
histopathology or phagocytic activity with
any of the particles.
Referen
ce
Cassee
etal.
(1998b)
Cassee
etal.
(1998c)
Cassee
etal.
(2002)
aH2SO4=Sulfuricacid
BAL = Bronchoalveolar lavage
LDH = Lactate dehydrogenase
MMAD = Mass median aerodynamic diameter
MMD = Mass median diameter
og = Geometric standard deviation
CMD = Count median diameter
-------
no increase in bronchoconstriction or bronchial responsiveness among asthmatic human adults
exposed for 1 h via facemask to 500 |ig/m3 of simulated acid fog containing ammonium sulfate
or H2SO4 aerosol.
Zelikoff et al. (1997) compared the responses of rabbits and humans exposed for 3 h to
similar concentrations (i.e., 1000 |ig/m3) of H2SO4 aerosol. For both rabbits and humans, there
was no evidence of PMA infiltration into the lung and no change in BAL fluid protein level,
although there was an increase in LDH in rabbits but not in humans. Macrophages showed
somewhat less antimicrobial activity in rabbits; but insufficient data were available for humans.
Superoxide production by macrophages was somewhat depressed in both species. Macrophage
phagocytic activity was also slightly reduced in rabbits but not in humans.
Ohtsuka et al. (2000a,b) also showed that a single 4 h exposure of mice to acid-coated
carbon particles at a high mass concentration of 10,000 |ig/m3 carbon black caused decreased
phagocytic activity of alveolar macrophages, even in the absence of lung injury. However,
in another study, Lee et al. (1999) found little effect on female rats or guinea pigs of an
inhalation exposure for 4 h to very high concentrations (43 or 94 mg/m3) of H2SO4 aerosol.
In another study, Heyder et al. (1999) exposed healthy beagle dogs by inhalation to
1.5 mg/m3 of acidic neutral sulfate aerosol for 16.5 h/day for 13 months or to acidic sulfate
aerosol at 5.7 mg/m3 for 6 h/day for 13 months. Interestingly, such chronic exposure to particle-
associated hydrogen and sulfur ions at very high concentrations resulted in only some subtle
respiratory responses, but no evident lung pathology.
Cassee and colleagues, in reports for the National Institute of Public Health and the
Environment, Bilthoven, the Netherlands (Cassee et al., 1998a,b,c), have examined the effects of
sulfate and nitrate aerosols in several compromised animal models. They used mice sensitized to
OA as a model of allergic asthma and compared them to normal mice. One study (Cassee et al.,
1998a) used exposures consisting of either fine ammonium bisulfate at 78 |ig/m3 (0.53 |im
MMD) or 972 |ig/m3 (0.45 |im MMD), or ultra fine particles at 235 |ig/m3 (0.085 |im MMD), for
exposure periods of 4 h/day for 3 consecutive days. Animals were analyzed at 1 or 4 days PE for
various cellular, biochemical, and immunological endpoints. No changes were seen in BAL
NAG, LDH, or protein with any of the particles. Only small changes were seen in the cytokines
IL-4, IL-6 and TNF-a, which were not considered relevant. Additionally there were no
treatment related effects on lung histopathology or serum IgE levels. The authors conclude that
7-67
-------
ammonium bisulfate exerts only marginal responses in this compromised model, and suggest
that the finding of increases in asthma in epidemiological studies is not due to this component of
PM.
A second study by this group (Cassee et al, 1998b) using the same exposure regimen, but
with ammonium ferrosulfate at a concentration of 250 |ig/m3 (0.459 jim MMD), found only
marginal changes in TNF-a and cell differential, which were not significant. A third study
(Cassee et al., 1998c), again using the asthmatic mouse model, assessed exposures of 140 |ig/m3
(0.3 |im CMD) or 250 |ig/m3 (0.03 jim CMD) ammonium nitrate aerosols. This particulate
exposure differed from the sulfates in that it caused increases in B AL NAG with exposure to the
smaller particle, and increased neutrophils and decreased AMs with exposure to the larger
particle. Other parameters showed no exposure-related effects. The authors stated that, as the
effects were mostly seen after exposure to the fine, rather than ultrafme, particle, both mass
concentration and specific size of the particles determine adverse effects.
Subsequent studies (Cassee et al., 2002) utilized an animal model of pulmonary
hypertension, MCT treatment, to study the effects of ammonium bisulfate, ammonium
ferrosulfate, and ammonium nitrate exposure. Exposures lasting 4 h/day for 3 consecutive days
used concentrations of ultrafmes at 70+ H2O/|ig/m3 (0.070 to 0.1 jim MMD), concentrations of
fine particles at 275 to 410 |ig/m3 (0.57 to 0.64 MMD) and fine CB aerosol at 2 to 9 mg/m3
(0.6 |im). Animals were examined at 1 day PE, using the same endpoints as in the three previous
studies in addition to phagocytic activity. As with the asthma model, no significant exposure-
related effects were seen with any of the particles.
Schlesinger and Cassee (2003) reviewed the literature on nitrate and sulfate secondary
inorganic particles. They concluded that, in healthy humans and animals and in the limited
number of compromised animal models studied, exposure to environmentally relevant levels of
these particles has little biological potency. They also discussed the chemical basis of toxicity of
these secondary inorganic particles and state that acidic particles, upon contact with epithelial
lining fluid (ELF), can be neutralized by the endogenous ammonia present. Additionally, the
mucus lining the airway buffers the acidic particles. This neutralization and buffering modulate
the effect of the particles, but the capacity of these systems may be reduced in compromised
individuals (Holma, 1989). Sarangapani and Wexler (1996) have modeled this defense system
7-68
-------
and predict greater neutralization for small particles (< 0.1 jim) than for larger particles
(> 1.0 jim).
Schlesinger and Cassee (2003) also stated that the available data indicate an acute exposure
of > 1000 |ig/m3 is necessary to affect pulmonary function in healthy humans. The dosage for
adverse effects in asthmatics is 68 to 100 |ig/m3, although the data are inconsistent. Transient
effects on mucociliary clearance are seen at sulfuric acid aerosol concentrations of 100 mg/m3,
with no differences observed between asthmatics and normal individuals. Their evaluation of
chronic exposure studies shows that concentrations of 100 to 250 |ig/m3 elicit changes in
secretory cell function, mucociliary clearance, and nonspecific airway hyperresponsiveness.
The review also stated that, with acute exposures, effects seen are a function of exposure
concentration (C) and duration or time (T) and that a threshold appears to exist for both C and T.
There is little evidence linking direct acute or chronic exposures to aqueous acid aerosols
to acute respiratory effects or chronic long pathology, except at much higher than current
ambient levels.
7.3.5 Diesel Particulate Matter
Studies of controlled exposures to diesel exhaust (DE) and/or diesel particles (DPM) were
previously evaluated in detail in two prior assessment documents, one by the Health Effects
Institute (1995) and the other by the U.S. Environmental Protection Agency (2002). As noted in
these documents, in addition to carcinogenic effects of exposure to diesel exhaust (DE), there are
significant noncancer health effects observed with high levels of exposure.
Acute (short-term exposure) effects in both humans and laboratory animals include eye,
throat, and bronchial irritation; neurophysiological symptoms include lightheadedness and
nausea; respiratory effects include cough and phlegm; and immunologic effects such as
exacerbation of allergenic responses to allergens. Chronic (long-term exposure) effects, as
determined mainly from animal studies, include a spectrum of dose-dependent inflammation and
histopathological changes in lung.
The most salient findings of the EPA 2002 Health Assessment Document for Diesel
Engine Exhaust noncancer health effects are first briefly recapitulated (at times verbatim) below.
Then some of these findings are elaborated upon further and the results of additional new studies
are discussed.
7-69
-------
7.3.5.1 Salient Findings from U.S. EPA 2002 Diesel Document
The EPA 2002 Diesel Document (U.S. Environmental Protection Agency, 2002) indicated
that acute human exposure to DE elicits subjective complaints of eye, throat, and bronchial
irritation and neurophysiological symptoms including headache, lightheadedness, nausea,
vomiting, and numbness and tingling of limbs. With increasing concentrations of DE, the odor
is detected more rapidly and the severity of symptoms increase. Studies of occupationally-
exposed workers demonstrated that there are minimal, generally not statistically significant,
increases in respiratory symptoms and decreases in lung function (FVC, FEVl3 PEFR, and
^EFJ2S.7S) during the course of the work shift. Smokers showed greater decrements in respiratory
functions and increased incidence of respiratory symptoms with DE exposure compared to
nonsmokers. Taken as a whole, both experimental and epidemiologic studies were not found to
show any consistent pattern of acute DE exposure effects on human pulmonary function or
respiratory symptoms. On the other hand, controlled human exposure studies were found to
have shown that acute exposures to DE induce airway inflammation (Rudell et al., 1990, 1994)
and to cause changes in peripheral blood (Salvi et al., 1999) in healthy humans, as further
elaborated on below.
As for chronic exposure effects, epidemiologic studies of chronic DE exposures which
occur in occupationally-exposed workers such as bus garage workers, miners, and railroad yard
workers were found to indicate an absence of excess risk of chronic respiratory disease
associated with exposure. Some respiratory symptoms, (primarily cough, phlegm, or chronic
bronchitis) were seen in a few studies; and two studies found statistically significant decrements
in baseline pulmonary functions, though most studies did not find changes in these parameters.
There was little evidence detected for adverse effects of DE on other organ systems, including
the cardiovascular system. The 2002 Diesel Document cautioned that interpretation of these
epidemiologic studies is difficult because of some methodological problems that include
incomplete information regarding effects of potentially confounding variables (smoking and
exposure to other toxicants concurrently) and the short durations and low intensity of the
exposures.
The 2002 EPA Diesel Document further noted that acute exposure of laboratory animals to
DE had been shown to cause mild functional effects, but only at high concentrations (> 6 mg/m3
DPM) and durations (20 h/day; Pepelko et al., 1980a). However, short-term exposures to even
7-70
-------
low levels of DE were found to elicit pathophysiological effects such as accumulation of DPM in
lung tissue, inflammation, AM aggregation and accumulation near the terminal bronchioles,
Type II cell proliferation, and thickening of the alveolar walls adjacent to AM.
Chronic DPM exposures were found to have little effects on survival in rodents. Some
evidence of reduced body weight in rats was seen with exposure concentrations of > 1.5 mg/m3
DPM and durations of 16 to 20 h/day, 5 days/week for 104 to 130 weeks (Heinrich et al., 1995;
Nikula et al, 1995). Species-specific changes in organ weights were reported with cats having
decreased lung and kidney weights with exposure and rodents having increased lung weights,
lung to body-weight ratios, and heart to body-weight ratios. The LOEL for these effects in rats
was 1 to 2 mg/m3 DPM for 7h/day, 5 days/week for 104 weeks (Brightwell et al., 1986; Heinrich
etal., 1986a,b).
Chronic exposures were also found to impair pulmonary function in rodents, cats, and
monkeys. Parameters affected by DE exposure included lung compliance, resistance, diffusing
capacity, volume and ventilatory performance. The exposure levels at which pulmonary
function was affected differed among species: 1.5 and 3.5 mg/m3DPM in rats (Gross, 1981;
Mauderly et al., 1988; McClellan et al 1986), 4.24 and 6 mg/m3 PDM in hamsters (Vinegar et al,
1980, 1981a,b), 11.7 mg/m3 in cats (Pepelko et al, 1980b, 1981), and 2 mg/m3 in cynomolgus
monkeys (only level tested in this species; Lewis et al, 1989). Exposures were typically 7 to
8 h/day, 5 days/week for 104 to 130 weeks and resulted in restrictive lung disease in all species
except monkeys. Gross (1981) estimated that observed changes in expiratory flow rates in rats
indicated a LOEL of 1.5 mg/m3 for chronic exposures. Obstructive airway disease was
evidenced in monkeys exposed chronically to 2 mg/m3 DE. This disparity with other species
tested is probably due to differences in anatomy and physiology, dose delivered, dose retained,
site of deposition, and effectiveness of clearance and repair mechanisms.
Histopathological effects have also been reported with chronic DE exposures. These
typically included alveolar histiocytosis, AM aggregation, tissue inflammation, increase in
PMNs, hyperplasia of bronchiolar and alveolar Type II cells, thickened alveolar septa, edema,
fibrosis, emphysema, and lesions of the trachea and bronchi. These were accompanied by
histochemical changes in lung including increases in lung DNA, total protein, alkaline and acid
phosphatase, and glucose-6-phosphate dehydrogenase. Additionally, increased synthesis of
collagen and release of inflammatory mediators have also been observed with chronic exposures.
7-71
-------
There appears to be a threshold of exposure to DPM below which these histopathologic effects
are not observed. Reported no observed effect levels (NOELs) include: 0.11 to 0.35 mg/m3 for
rats (Ishinishi et al., 1986, 1988); 0.25 mg/m3 for guinea pigs (Barnhart et al., 1981, 1982); and
2 mg/m3 for cynomolgus monkeys (only level tested in this species; Lewis et al., 1989) for
exposures of 7 to 20 h/day, 5 to 5.5 days/week for 104 to 130 weeks.
Chronic exposures to DPM were further found to have an effect on airway clearance,
which in large part determines the pathological effects. Alveolar macrophages phagocytose
DPM as part of a multiphasic process of clearance. Exposures of > 1 mg/m3 DPM were shown to
have a detrimental effect on clearance (Wolff et al, 1987; Wolff and Gray, 1980), the net effect
being focal aggregations of particle-laden AMs in the peribronchiolar and alveolar regions and
also the hilar and mediastinal lymph nodes. As mentioned above, species differences exist in
anatomy, physiology, rate of uptake, deposition, clearance, size of AM population, rate of influx
of AM and leukocytes, and the relative efficiencies for removal of particles by the mucociliary
escalator and lymphatic transport system. Any decrease in AM function tends to reduce
clearance. It is mostly particles that are persistently retained in the lungs that impair clearance
and this occurs in F344 rats at a PM burden of 0.1 to 1 mg/g lung tissue (Health Effects Institute,
1995). Morrow (1988) estimated that AM loading of >60 jim3 PM impairs clearance and
> 600 |im3 causes clearance to cease. This results in agglomerated particle-laden AMs
remaining in the alveolar region and particles translocating to the pulmonary interstitium.
Consistent with impairment of AM function and clearance, reduction of an animal's
resistance to respiratory infection was found with exposure to DPM. This effect was seen after
an acute exposure of 5 to 8 mg/m3 for as little as 2 or 6 h. The effect is thought not to be due to
direct impairment of the lymphoid or splenic immune systems. Both animal and human acute
exposure data also suggest that DPM is a factor in the increasing incidence of allergic
hypersensitivity. Both the nonextractable carbon core and the organic fraction of DPM were
implicated in the effect. It was noted that synergies with DPM may increase the potency of
known airborne allergens, and DPM was posited to act as an adjuvant in immune responses.
Chronic DE exposures in rats lasting from birth to 28 days of age were also shown to have
behavioral effects on spontaneous locomotor activity and decrements in learning in adulthood
(Laurie et al, 1980). These findings were corroborated by physiological evidence of delayed
neuronal maturation (Laurie and Boyes, 1980, 1981). These studies, published in the early
7-72
-------
1980s, used exposures of 6 mg/m3 DPM for 8 h/day, 7 days/week. No recent studies have added
to this literature. Also, based on the weight of evidence of a number of studies, essentially no
effects were noted for reproductive and teratogenic effects in mice, rats, rabbits, and monkeys;
for clinical chemistry and hematology in rat, cat, hamster, and monkeys; and for enzyme
induction in the rat and mouse.
Key conclusions arrived at, based on the studies assessed in the U.S. EPA 2002 Health
Assessment Document for Diesel Engine Exhaust, included: (1) short-term exposure to the
DPM component of DE can result in allergenic inflammatory disorders of the airway; (2) acute
occupational exposures to DE can cause respiratory symptoms of cough, phlegm, chest tightness
and wheezing (all suggestive of an irritant mechanism) but do not generally cause pulmonary
function decrements; and (3) pulmonary histopathology (principally fibrosis) and chronic
inflammation are noncancer effects seen in laboratory animals, but noncancer effects in humans
from long-term chronic exposures to DPM are not evident. Also, current knowledge indicates
that the carbonaceous core of DPM is probably the causative agent of lung effects. Further,
progressive impairment of AM is a factor in the extent of lung injury. Lung effects occur in
response to DE exposures in several species and occur in rats at doses lower than those inducing
particle overload and a tumorigenic response.
It is important to note that several DE toxicity studies cited in the EPA 2002 Diesel
Document compared the effects of whole, unfiltered exhaust to those produced by the gaseous
components of the exhaust. A comparison of the toxic responses in laboratory animals exposed
to whole exhaust or filtered exhaust containing no particles demonstrates across studies that,
when the exhaust is sufficiently diluted to limit the concentrations of gaseous irritants (NO2
and SO2), irritant vapors (aldehydes), CO, or other systemic toxicants, the diesel particles are
clearly contributors to noncancer health effects, although additivity or synergism with the gases
cannot be ruled out. These toxic responses are both functional and pathological and represent a
cascading sequelae of lung pathology based on concentration and species. The diesel particles
plus gas exposures produced biochemical and cytological changes in the lung that are much
more prominent than those evoked by the gas phase alone. Such marked differences between
whole and filtered DE are also evident from general toxicological indices, such as decreases in
body weight and increases in lung weights, pulmonary function measurements, and pulmonary
histopathology (e.g., proliferative changes in Type II cells and respiratory bronchiolar epithelium
7-73
-------
fibrosis). Hamsters, under equivalent exposure regimens, have lower levels of retained DPM in
their lungs than rats and mice and, consequently, less pulmonary function impairment and
pulmonary pathology. These differences may result from lower DPM inspiration and deposition
during exposure, greater DPM clearance, or lung tissue less susceptible to the cytotoxicity of
deposited DPM.
The above past assessment findings, on the whole, tend to suggest the potential importance
of DPM contributing to at least some ambient PM-related toxic effects, particularly in urban
micro-environments with heavy diesel traffic. The findings of some DE- or DPM-related
controlled human exposure studies are elaborated on below and then are further interrelated to
pertinent laboratory animal studies discussed later in Section 7.5.3 (Particulate Matter Effects on
Allergic Hosts).
Pulmonary function and inflammatory markers (as assayed in induced sputum samples or
BAL) have been studied in human subjects exposed to either resuspended or freshly generated
and diluted DPM. In one controlled human exposure study, Sandstrom and colleagues (Rudell
et al., 1994) exposed eight healthy subjects in an exposure chamber to diluted exhaust from a
diesel engine for 1 h with intermittent exercise. Dilution of the DE was controlled to provide a
median NO2 level of-1.6 ppm. Median particle number was 4.3 x 106 /cm3, and median levels
of NO and CO were 3.7 and 27 ppm, respectively (particle size and mass concentration were not
provided). There were no effects on spirometry or on airway closing volume. Five of eight
subjects experienced unpleasant smell, eye irritation, and nasal irritation during exposure. BAL
performed 18 h after exposure was compared with a control BAL performed 3 weeks prior to
exposure. There was no control air exposure. Small, yet statistically significant, reductions
were seen in BAL mast cells, AM phagocytic function, and lymphocyte CD4 to CD8+ cell
ratios, along with a small increase in neutrophils. These findings suggest that DE may induce
mild airway inflammation in the absence of spirometric changes. Although this study generated
some potentially important information on the effect of DE exposure in humans, only one
exposure level was used, the number of subjects was low, a limited range of endpoints was
reported, and no comparisons to clean control exposures were provided. Several follow-up
studies have been done by the same and other investigators.
Rudell et al. (1996) later exposed 12 healthy volunteers to DE for 1 h in an exposure
chamber. Light work on a bicycle ergometer was performed during exposure. Random, double-
7-74
-------
blinded exposures included exposures to clean air, DE, or DE with particle numbers reduced
46% by a particle filter. The engine used was a new Volvo model 1990, a six-cylinder direct-
injection turbocharged diesel with an intercooler, run at a steady speed of 900 rpm during the
exposures. It is hard to compare this study with others, because neither exhaust dilution ratios
nor particle concentrations were reported. Based on concentrations of 27 to 30 ppm CO and of
2.6 to 2.7 ppm NO, however, estimated DPM concentrations likely equaled several mg/m3. The
most prominent symptoms during exposure were irritation of the eyes and nose, accompanied by
an unpleasant smell. Both airway resistance and specific airway resistance increased
significantly during the exposures. Despite the 46% reduction in particle numbers by the filter,
effects on symptoms and lung function were not significantly reduced. A follow-up study on the
usefulness of a particle filter confirmed the lack of effect of the filter on DE-induced symptoms
(Rudell et al., 1999). In this study, 10 healthy volunteers also underwent BAL 24 h after
exposure. Exposure to DE produced inflammatory changes in BAL, as evidenced by increases
in neutrophils and decreases in macrophage phagocytic function in vitro. A 50% reduction in
the particle number concentration by the particle filter did not alter these BAL cellular changes.
As reported in a series of studies (Rudell et al., 1990, 1996, 1999; Blomberg et al., 1998;
Salvi et al., 1999), significant increases in neutrophils and B lymphocytes, as well as in
histamine and fibronectin in airway lavage fluid, were not accompanied by decrements in
pulmonary function. Salvi et al. (1999) exposed healthy human subjects to diluted DE
(DPM = 300 |ig/m3 ) for 1 h with intermittent exercise. Bronchial biopsies obtained 6 h after DE
exposure showed a significant increase in neutrophils, mast cells, and CD4+ and CD8+
T lymphocytes, along with upregulation of the endothelial adhesion molecules ICAM-1 and
vascular cell adhesion molecule-1 (VCAM-1) and increases in the number of leukocyte function-
associated antigen-1 (LFA-1+) in the bronchial tissue. Importantly, extra-pulmonary effects
were observed in these subjects. Significant increases in neutrophils and platelets were found in
peripheral blood following exposure to DE.
In a follow-up investigation of potential mechanisms underlying the DE-induced airway
leukocyte infiltration, Salvi et al. (2000) exposed healthy human volunteers to diluted DE of
300 |ig/m3 on two separate occasions for (1 h each) in an exposure chamber. Fiber-optic
bronchoscopy was performed 6 h after each exposure to obtain endobronchial biopsies and
bronchial wash (BW) cells. These workers observed that diesel exhaust (DE) exposure enhanced
7-75
-------
gene transcription of interleukin-8 (IL-8) in the bronchial tissue and BW cells and increased
growth-regulated ontogeny-a protein expression and IL-8 in the bronchial epithelium; there was
also a trend toward an increase in interleukin-5 (IL-5) mRNA gene transcripts in the bronchial
tissue. Whether these effects were due to DPM or associated DE gaseous components (or both)
could not be disentangled with the study design used.
Nightingale et al. (2000) reported inflammatory changes in healthy volunteers exposed to
200 |ig/m3 resuspended DPM for 2 h under resting conditions in a double-blinded study. Small
but statistically significant increases in neutrophils and myeloperoxidase (an index of neutrophil
activation) were observed in sputum samples induced 4 h after exposure to DPM in comparison
to air. Exhaled CO was measured as an index of oxidative stress and was found to increase
maximally at 1 h after exposure. These biochemical and cellular changes occurred in the
absence of any decrements in pulmonary function, thus confirming that markers of inflammation
are more sensitive than pulmonary function measurements.
Because of the concern about inhalation of ambient particles by sensitive subpopulations,
(Nordenhall et al., 2001) also studied the effect of a 1 h exposure to DE (containing 300 |ig/m3
DPM, 1.2 ppm NO2, 3.4 ppm NO, 2.6 ppm HC, and 9.1 ppm CO) on 14 atopic asthmatics with
stable disease and on inhaled corticosteroid treatment. At 6 h after exposure, there was a
significant increase in airway resistance (p < 0.004) and in IL-6 in induced sputum (p < 0.048)
following exposure to DE versus filtered air. At 24 h after exposure, there was a significant
increase in the nonspecific airway responsiveness to inhaled methacholine. Although the DPM
exposure level was high relative to ambient PM levels, these findings may be important, as noted
by the authors, in terms of supporting epidemiologic evidence for increased asthma morbidity
associated with episodic exposure to ambient PM.
The IL-6 increase seen here 6 h after DE exposure in asthmatic subjects parallels similar
significant IL-6 increases in sputum 6 h after DE exposure of healthy subjects, suggesting that
the IL-6 release represents an acute response of both healthy and asthmatic persons to DE
exposures. Other work by Steerenberg, et al. (1998) showed that DE particles are effective in
inducing release of IL-6 from human bronchial epithelial cells (see Section 7.4).
The role of antioxidant defenses in protecting against acute DE exposure has also been
studied. Blomberg et al. (1998) investigated changes in the antioxidant defense network within
the respiratory tract lining fluids of human subjects following diesel exhaust exposure. Fifteen
7-76
-------
healthy, nonsmoking, asymptomatic subjects were exposed to filtered air or DE (containing
300 mg/m3 DPM) for 1 h on two separate occasions at least 3 weeks apart. Nasal lavage fluid
and blood samples were collected prior to, immediately after, and 5.5 h postexposure.
Bronchoscopy was performed 6 h after the end of DE exposure. Nasal lavage ascorbic acid
concentration increased 10-fold during DE exposure, but returned to basal levels 5.5 h
postexposure. Diesel exhaust had no significant effects on nasal lavage uric acid or GSH
concentrations and did not affect plasma, bronchial wash, or BAL antioxidant concentrations or
malondialdehyde or protein carbonyl concentrations. The authors concluded that the acute
increase in ascorbic acid in the nasal cavity induced by DE may help prevent further oxidant
stress in the upper respiratory tract of healthy individuals.
Seagrave et al. (2002) evaluated the inflammation and cytotoxicity created by exposure to
exhaust from a number of vehicles including automobiles, SUVs, and pickup trucks from 1976
to 2000. Both PM and vapor-phase semivolatile organic compound (SVOC) fractions were
collected, both at room temperature and in a cold environment. The PM and SVOC fractions
were recombined and tested for toxicity in male F344/CrlBR rats at age ~11 weeks. The
emission samples were intratracheally instilled at doses of 0.1 to 3 mg/rat. BAL was collected at
4 h for cytokine endpoints and collected at 24 h for examination of histopathology and lavage
parameters. Three different assays, histopathology, LDH and protein, were used to determine
the cytotoxicity of the emission samples. Total protein in BAL and LDH similarly ranked
cytotoxicity of the samples, and histology results created similar rankings, except for gasoline,
which was ranked least toxic by LDH and protein and fourth by histopathology. The authors
uniformly scaled the potencies and ranked the cytotoxicity as: gasoline engine emitting white
smoke > gasoline engine emitting black smoke > high emitter diesel > normal diesel 72 °F >
current diesel at 30 °F > normal gasoline 30 °F > normal gasoline 72 °F. Inflammatory
endpoints examined were total leukocytes, macrophages, PMNs/mL BALF , MIP-2, TNF-a, and
histopathology. There was good agreement among data for total leukocytes, PMNs, and
macrophages for which the three highest emissions were ranked: gasoline engine emitting white
smoke > gasoline engine emitting black smoke = high emitter diesel. These three exhausts as
had equally high inflammatory effects (as indicated by increases in MIP-2, but were less
consistent for effects on TNF-a, it being suppressed in some samples and slightly increased in
others. Uniformly scaled potencies for inflammation using total leukocytes, PMA, macrophages,
7-77
-------
histopathology, and MIP-2 endpoints were: gasoline engine emitting white smoke > gasoline
engine emitting black smoke > high emitter diesel > current diesel at 30 °F > normal gasoline
72 °F > normal gasoline 30 °F > normal diesel 72 °F.
7.3.6 Ambient Bioaerosols
Bioaerosols are airborne particles consisting of large molecules or volatile compounds that
are living, contain living organisms, or have been released from living organisms. Major types
of bioaerosol particles encountered in ambient (outdoor) air, indoor air, and/or in contaminated
indoor or outdoor dusts that can be resuspended into air include: (1) intact pollen and pollen
fragments; (2) fungi, their spores, and other fungal byproducts; (3) humic-like substances
(HULIS), which include bioaerosols, biomass combustion generated and secondary organic
compounds and other plant debris; (4) certain animals or associated debris, e.g., dust mites or
their excreta, shed mammalian or avian skin cells, etc.; (5) bacteria or fragments thereof, e.g.,
endotoxins consisting of proteins and lipopolysaccharides (LPS) that comprise portions of cell
walls of Gram-negative bacteria; (6) (1^3)-p-D-glucan, a polyglucose compound in the cell
walls of Gram-positive bacteria, fungi, and plants; and (7) viruses.
Such particles are suspended and/or transported in air as distinct separate entities or
adhered to other organic and nonorganic particles or in water droplets. Biological particles can
range in size from 0.01 |im (viruses) to > 20 jim (some pollen), with the smaller ones < 10.0 jim
being inhalable and, upon inhalation, being capable of penetrating into tracheobronchial (TB)
and alveolar (A) regions of the lower respiratory tract — thus creating potentially serious health
problems for sensitive human populations.
The relationship between bioaerosol exposure and illness is complex. Numerous studies
published since the 1996 PM AQCD have produced extensive new information which has
greatly enhanced our knowledge regarding environmental occurrence of such biological
aerosols, their health effects, and possible combined influences of their being copresent along
with other biological and/or nonbiological particles in ambient air. In particular, there is
growing recognition that bioaerosols may contribute to health effects related to ambient PM
exposures partly through their own direct toxic effects and/or in combination with other PM that
carries biologically-derived materials which may elicit untoward effects.
7-78
-------
Appendix 7B recapitulates a number of key points regarding ambient bioaerosols derived
from the 1996 PM AQCD and goes on to update and integrate information derived from newer
studies, as well. This includes background information on types and sources of ambient
bioaerosols, factors affecting their dispersal and airborne concentrations, and both epidemiologic
and toxicologic evaluations of health effects associated with different classes of them. As such,
some of the materials discussed may have been touched on in other chapters, but are brought
together in Appendix 7B and summarized here to provide a coherent overall picture related to
bioaerosols as potentially important contributors to ambient PM-related health effects.
A large number of studies show relationships between exposure to bioaerosols and airways
inflammation and other signs/symptoms of allergic/asthmatic responses. Generally these
exposures are most often associated with: certain occupational settings (cotton milling, grain
workers, feed mill employees, farmers); humid and poorly ventilated indoor environments where
moisture/dampness can harbor these organisms; and households having domestic animals/pets
(Wheatley and Platts-Mills, 1996).
Bioaerosols mainly tend to be in the coarser fraction of ambient PM, but some (e.g., fungal
spores, pollen fragments) are in the fine fraction as well. Flowering plants, trees, and grasses
produce pollen, the species and quantity being determined by region, season, and meteorological
factors (especially humidity/moisture levels). For example, increased levels of grass pollen
allergens following thunderstorms have been linked to increased levels of asthma attacks i.e.,
"thunderstorm asthma" (Bellomo et al., 1992; Ong, 1994; Rosas et al., 1998; Schappi et al.,
1999). Wind-pollinated plants produce large grains >10-20 jim, which when intact, deposit in
upper airways, inducing allergic rhinitis. However, rupture of these grains following rain events
generates allergen-containing cytoplasmic pollen fragments that constitute respirable particles
(-0.1 to 5.0 |im) associated with exacerbation of asthma.
Very importantly, it is now known that interactions between aerosolized allergen-laden
pollen debris and other types of ambient airborne particles occur. Pollen, in addition to
containing cytoplasmic allergens, has also been shown to be a carrier of other allergenic
materials. Several different types of immunoactive, allergenic materials (e.g., Gram-negative
and Gram-positive bacteria; endotoxin, fungi) have been shown to be associated with grass and
tree pollens in Poland (Spiewak, 1996a,b). Also, Taylor et al. (2002) suggest that the polycyclic
hydrocarbon component of diesel exhaust may interact with allergen-laden pollen debris in a
7-79
-------
synergistic combination to explain, in part, the notable increase in the prevalence of pollen-
induced asthma during the past 50 years. Ormstad et al. (1998) and Knox et al. (1997)
demonstrated that DPM (especially < 2.5 jim) can act as a carrier for plant (and animal)
allergens and, further, may act as a mechanism whereby plant allergens can become concentrated
in air and trigger asthma attacks. Additionally, evidence from Behrendt et al. (1992, 1995, 1997,
2001) show that pollen grains may incorporate other atmospheric pollutants that alter the pollen
surface, leading to exocytosis of proteinaceous material and increased allergen release. As for
health-related studies of pollen effects (see Table 7B-2), Hastie and Peters (2001) evaluated
in vivo ragweed allergen exposure (via bronchoscopic segmented ragweed challenge) effects on
ciliary activity of bronchial epithelial cells harvested 24 h after challenge in nonallergic human
adults and in allergic subjects with severe inflammatory response. Allergic subjects with mild
inflammatory changes showed slight increases in albumin and doubling of bronchoalveolar cell
levels whereas allergic subjects with severe inflammatory changes showed a 12-fold increase in
albumin and a 9-fold increase in bronchoalveolar cell levels.
In another study of mice, a mixture of DPM and Japanese cedar pollen caused increased
IgE and IL-4 production compared to pollen alone (Fujimaki et al., 1994). Synergistic
relationships were also observed with DPM and ragweed allergen in the production of specific
cytokines (Diaz-Sanchez et al., 1997).
In an epidemiologic study in the Netherlands, Brunekreef et al. (2000) found a positive
correlation between mortality rates and pollen concentration, suggesting that pollen-associated
acute exacerbation of allergic inflammation may cause death among some compromised
individuals. Increases in hospitalizations for asthma have also been reported to be correlated
with pollen exposure in Mexico City (Rosas et al., 1998) and London (Celenza et al., 1996), as
have increased asthma incidence and medication use (Delfino et al., 1996, 1997).
In summary, newly available information indicates release of allergen-laden material from
pollen-spores in respirable-sized aerosols; suggests possible ways by which binding of such
material to other airborne particles (e.g., DPM) may concentrate such allergens in ambient air or,
once inhaled, jointly exacerbate allergic reactions in susceptible human populations; and
indicates that pollen itself may act as a carrier for other allergenic materials. Thus, new
information about the synergistic relationship between plant allergens and other forms of PM
7-80
-------
suggest a possible mechanism which may explain, in part, the increased morbidity (especially
asthma) and mortality associated with increased pollen levels.
In addition to pollen, other plant-related bioaerosols are generated by human activities such
as the storage, handling and transport of plant material; and they, too, can cause adverse health
effects. A growing database suggests that plant debris is a significant contributor to organic
aerosols at continental sites. This debris has a considerable component that is insoluble. Humic-
like substances (HULIS), originating from biomass fires and secondary atmospheric reactions,
comprise up to 24% of organic carbon in some aerosol samples. In many areas in the western
United States there are episodic or seasonal increases in plant-derived bioaerosol material from
biomass burning emissions. These controlled agricultural burns, forest fires and domestic wood
burning all contribute to ambient PM in these regions.
Fungi, growing on dead organic matter, are ubiquitous and produce huge quantities of
aerosols, including spores, body fragments, and fragments of decomposed substrate material.
Fungal spores, ranging in size from 1.5 jim to > 100 jim, form the largest and most consistently
present component of outdoor bioaerosols. These cause allergic rhinitis and asthma, while
allergic fungal sinusitis and allergic bronchopulmonary mycoses are caused by fungi colonizing
thick mucus in the sinuses or lungs of allergic individuals. Yang and Johanning (2002) have
shown that, once an individual is sensitized to the fungi, small concentrations can trigger an
asthma attack or other allergic response.
Several studies have found relationships between exposure to fungi and their byproducts in
respiratory illnesses and immune pathology (Hodgson et al., 1998; Tuomi et al., 2000; Yang and
Johanning, 2002). Larsen et al. (1996) showed non-immunological histamine release from
leukocytes exposed to a suspension of fungal spores and hyphal fragments. Some fungal
byproducts have also been shown to stop ciliary activity in vitro and may act to produce general
intoxication by macroorganisms through the lung tissue or to enhance bacterial or viral infection
(Pieckova and Kunova, 2002; Yang and Johanning, 2002).
Fungal concentrations in most parts of the world have a pattern of peak levels in the
summer and early fall, but low levels in the winter months. Outdoor air fungal composition
affects culturable fungal propagules indoors, but it appears that the levels of fungi inside do not
just reflect the outdoor levels. Analogous to pollen exposures, fungal spore exposures have been
7-81
-------
positively correlated with asthma hospital admissions of children in Mexico City (Rosas et al.,
1998) and with asthma deaths in Chicago (Targonski et al., 1995). Airborne fungal
concentrations of > 1000 spores/m3 were reported to be associated with asthma deaths among 5-
to 34-year-old persons in Chicago between 1985 and 1989 (Targonski et al., 1995). The odds of
death occurring on days with airborne fungal concentrations of > 1000 spores/m3 were 2.16 times
higher than mortality on other days.
All classes of animals including humans, house pets, wild and domesticated birds, and
insects produce bioaerosols capable of producing hypersensitivity diseases. Dust mites and
cockroaches are prolific insects from which fecal material and shed body parts create allergens
that are a major causes of sensitization in children (Burge, 1995).
Bacteria and viruses are infectious agents that are released from hosts in droplets exhaled
from the respiratory tract. The antigenic component of bacteria can be the whole living bacteria
or enzymes or cell wall components of the bacteria. Viruses, composed of either DNA or RNA
surrounded by a protein coat, utilize living cells for reproduction. Virus are extremely small
(«1 |im), but the infectious droplets are usually larger (1 to 10 jim).
Endotoxins are present in the outer cell membrane of gram-negative (Gram -) bacteria.
Heederik et al. (2000) noted that animal feces and plant materials contaminated with bacteria
contribute most to organic dust-related endotoxin exposure. Although there is strong evidence
that inhaled endotoxin plays a major role in the toxic effects of bioaerosols encountered in the
work place (Castellan et al., 1984, 1987; Rose et al., 1998; Vogelzang et al., 1998; Zock et al.,
1998), it is not clear as to what extent typical ambient concentrations of endotoxin are sufficient
to produce toxic pulmonary or systemic effects in healthy or compromised individuals.
Endotoxins act on cells in the respiratory system by binding to receptors and triggering
production of cytokines, which initiates a cascade of inflammation, smooth muscle constriction,
and vasodilation (Young et al., 1998). Table 7B-3 summarizes studies of the respiratory effects
of inhaled endotoxin-laden ambient bioaerosols. Some new occupational exposure studies
suggest declines in lung function due to exposure to endotoxins in pig farm waste (Vogelzang
et al., 1998) and potato processing (Zock et al., 1998). Also, increases in BAL lymphocytes
were observed in life guards exposed to endotoxins at a swimming pool (Rose et al., 1998).
However, these studies do not rule out the effects of other agents in the complex airborne
7-82
-------
organic aerosols that may contribute to the functional and cellular effects observed. Still, the
authors noted that their results support the selection of the lower of two proposed (Clark, 1986;
Palchak et al., 1988) occupational exposure threshold levels of 30 or 100 ng/m3 for airborne
endotoxin.
Two German cities 80 km apart with a differing prevalence of hay fever and allergic
sensitization in children were studied for the possible effects of endotoxin (Heinrich et al.,
2002a,b), but the researchers could not attribute observed differences between the towns in
respiratory disease prevalence to endotoxin levels. Later work by this group showed higher
concentration and absolute mass of endotoxin in coarse-mode particles versus fine particles.
Levels of endotoxin were also seasonal, demonstrating increased levels in late spring/summer
and lower levels in winter.
Dose-response studies in healthy human adults exposed to doses ranging up to 50 jig
endotoxin, by the inhalation route, suggested a threshold for pulmonary and systemic effects for
endotoxin between 0.5 and 5.0 jig (Michel et al., 1997). Inhalation in heterodisperse droplets
with a MMAD of 1 to 4 jim of 5 or 50 jig of LPS, but not 0.0 or 0.5 jig increased PMNs in blood
and sputum. Another controlled human exposure study of endotoxin, involving inhalation of
lipopolysaccharide (LPS: the purified derivative of endotoxin) by known smokers showed
increases in myeloperoxidase and eosinophilic cationic protein, decreases in FEVj and FVC, and
irritation, dry cough, breathlessness, and tiredness at a LPS dose of 40 jig (Thorn and Rylander,
1998a).
Also, Monn and Becker (1999) also examined effects of size fractionated outdoor PM on
human monocytes and found cytokine induction characteristic of endotoxin activity in the
coarse-size fraction but not in the fine fraction.
Certain gram-positive bacteria, fungi, and plants contain the polyglucose compound (1^3)-
p-D-glucan in their cells walls, which has been shown to induce stimulation of the
reticuloendothelial system, activation of PMNs, AMs, and complement. Heederik et al. (2000)
found T-lymphocyte activation and proliferation with glucan exposure of experimental animals.
In homes in Sweden where (1 -»3)-P-D-glucan levels ranged between 0 and 19 ng/m3, Thorn and
Rylander (1998b) found that there was a significantly larger number of atopic subjects in the
> 65 year old group exposed to > 3 ng/m3 (1 ^3)-P-D-glucan in their homes. Rylander (1996)
also found that an acute exposure to(l - 3)-p~D-glucan can produce symptoms of airway
7-83
-------
inflammation in normal human subjects without a history of airway reactivity after exposing
subjects to 210 ± 147 ng/m3 (1 -> 3)-p~D-glucan for 3 separate 4 h sessions 5 to 8 days apart.
Douwes et al. (1998) examined the relationship between exposure to (1 -» 3)-p~D-glucan
and endotoxins and peak expiratory flow (PEF) in children (ages 7 to 11 years) with and without
chronic respiratory symptoms. As indicated by linear regression analysis, peak expiratory flow
variability in the children with chronic respiratory symptoms was strongly associated with
(1^3)-p-D-glucan levels in dust from living room floors. The association was strongest for
atopic children with asthma.
Thus, new research is focusing on the bioaerosol component of ambient PM. Findings
include a greater-than-realized impact of pollen on asthma, and synergistic associations between
pollen and other PM, with the potential for increased risk of adverse health effects. Fungi, and
especially the spores, are the largest component of outdoor bioaerosols and have been linked to
allergic rhinitis, asthma, sinusitis, and allergic bronchopulmonary mycoses. Much research is
being directed at characterizing the mechanism by which bacterial endotoxins and (H3)-P-D-
glucan cause adverse health effects, especially in compromised individuals.
Of much importance are the seasonal variations in ambient air concentrations of all types
of airborne allergens (both plant- and animal-derived) typically observed in temperate climate
areas. Typically, (given that warmer, humid conditions tend to facilitate pollen, fungal and
bacterial growth) outdoor levels of pollen fragments, fungal materials, endotoxins, and glucans
all tend to increase in the spring/summer months and decrease to low ambient levels in late
fall/winter months in most U.S. and other temperate areas. Also of much importance are
increased levels of cellulose and other plant debris in respirable size fractions of ambient
aerosols during the spring and summer months — plant materials that can act as carriers for
allergenic materials (bacterial, fungal, etc.). The copresence in ambient air of other biological
particles capable of acting as carriers of such allergens would probably enhance the risk of
allergic/asthmatic reactions to them. Pertinent to this, it is of interest to note, that endotoxin
concentrations tend to be higher in coarse fraction ambient PM samples than in fine (< 2.5 jim)
ambient PM samples; but endotoxin concentrations are typically very low, rarely exceeding
0.5 EU/m3.
7-84
-------
7.3.7 Summary of Respiratory Effects
The respiratory effects of PM having varying physical and chemical characteristics have
been extensively studied for more than 30 years using a wide range of techniques and with
exposure durations ranging from brief periods to months. The most extensively studied
materials have been sulfates and acid aerosols formed as secondary pollutants in the atmosphere.
Fly ash from coal-fired power plants or other coal-combustion sources has been less extensively
studied. The toxicological data available today provide little basis for concluding that these
specific PM constituents have substantial respiratory effects at current ambient levels of
exposure. Recently, ROFA, a very specific kind of PM, has been studied extensively and found
to produce a range of respiratory effects, especially lung inflammation at higher concentrations
or doses.
Recent studies evaluating controlled human exposures to concentrated ambient particles
(CAPs) from diverse locations (e.g., Boston, New York City , Los Angeles, Toronto, and
Chapel Hill, NC) have found little or no effects on pulmonary function or respiratory symptoms
in healthy human adults acutely exposed (for 2 h) to CAPs concentrations that ranged from about
25 up to about 300 |ig/m3. Some indications of mild lung inflammation were reported with such
exposures in some of the studies, but not others. Analogous controlled exposures to CAPs of
rats, hamsters, and dogs at concentrations varying across a range of-100 to 1000 |ig/m3 for
1-6 h/day for 1 to 3 days yielded similar minimal effects on respiratory functions, but some
signs of mild inflammation in normal healthy animals and somewhat enhanced indications of
lung inflammation in at least one compromised animal model of chronic bronchitis. Another
study found some indications of mild impairment of lung immune defense functions and
exacerbation of bacterial infection with an acute (3 h) exposure of rats to New York City CAPs
(at 100-350 |ig/m3). There is also new evidence for the transition metal components of ROFA
and ambient PM from diverse locations having a mediating role in producing injury.
There still remains, however, a critical need for the systematic conduct of studies of the
potential respiratory effects of major components of PM from different regions of the U.S., in
recognition that PM of different composition and from different sources can vary markedly in its
potency for producing different respiratory effects.
7-85
-------
7.4 CARDIOVASCULAR AND RESPIRATORY PATHOPHYSIOLOGY
AND TOXICITY: IN VITRO PM EXPOSURES
7.4.1 Introduction
Toxicological studies play an important role in providing evidence by which to evaluate
the biological plausibility of health effects associations with ambient PM exposure observed in
epidemiologic studies. At the time of 1996 PM AQCD (U.S. Environmental Protection Agency,
1996a) little was known about potential mechanisms that could explain associations between
morbidity and mortality and ambient airborne PM observed in human populations studies. One
of the difficulties in trying to sort out possible mechanisms is the nature of ambient PM mixes.
Ambient PM has diverse physicochemical properties, ranging from physical characteristics of
the particles to chemical components in or on the surface of the particles. Any one of these
properties could change at any time in the ambient exposure atmosphere, making it difficult to
duplicate actual properties of ambient PM in a controlled experiment. As a result, controlled
exposure studies have as yet neither clearly identified those particle properties nor specific
mechanisms by which ambient PM may affect biological systems. However, new in vitro
toxicologic studies that have become available since the 1996 PM AQCD have provided
additional information useful to help explain how ambient particles may exert toxic effects on
the respiratory and cardiovascular systems. Such studies are summarized in Tables 7-10
(ambient PM) and 7-11 (ROFA and other combustion source PM) and are discussed in the next
several subsections.
In vitro exposure is a useful technique by which to obtain information on potentially
hazardous PM constituents and mechanisms of PM injury, especially when only limited amounts
of PM test material are available. For example, respiratory epithelial cells lining the airway
lumen have been featured in numerous studies involving airborne pollutants and show
inflammatory responses similar to that of human primary epithelial cultures. Also, alveolar
macrophage cells from humans, rats, or other species have been employed in vitro to evaluate
effects on phagocytosis and various other aspects of lung defense mechanisms. Limitations of
in vitro studies include possible alterations in physiochemical characteristics of PM because of
collection and resuspension processes, exposure conditions that do not fully simulate air-cell
interface conditions within the lungs, and difficulties in estimating comparable dosage delivered
to target cells in vivo. Also, doses delivered in vitro, like intratracheal administration, can be
7-86
-------
TABLE 7-10. IN VITRO EFFECTS OF AMBIENT PARTICIPATE MATTER AND
PARTICIPATE MATTER CONSTITUENTS
oo
Species, Cell Particle or
Type, etc. " Constituent'
BEAS-2B TSP
Primary culture (Provo, Utah)
human tracheal and
bronchial epithelial
cells
BEAS-2B PM10 extract
(Provo, UT)
Cell Count Concentration Particle Size Exposure Duration
2 x 105 cells/mL TSP filter samples N/A (TSP samples, Sacrificed at 24 h
(36.5 mg/mL) comprised 50 to
agitated in 60% PM10)
deionized H2O2 for
96 h, centrifuged at
1200gfor30min,
lyophylized and
resuspended in
deionized H2O2
or saline
100 - 500 ug/well
125,250, PM10 2 and 24 h
500 ug/mL
Effect of Particles Reference
Provo particles caused cytokine- Kennedy et al.
induced neutrophil-chemoattractant- (1998)
dependent inflammation of rat lungs;
Provo particles stimulated IL-6 at 500
ug/mL and IL-8 at > 200 ug/mL,
increased IL-8 mRNA at 500 ug/mL
and ICAM-1 at 100 ug/mL in
BEAS-2B cells, and stimulated IL-8
secretion at > 125 ug/mL in primary
cultures of human tracheal and
bronchial epithelial 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
Dose-dependent increase in IL-6 and Frampton et al.
IL-8 induced at all doses after 24 h for (1999)
BEAS-2B
NHBE
BEAS-2B
(Provo, UT)
TSP soluble and
insoluble extract
Utah Valley PM1(
extract
500 ng/mL
TSP
24 h
50, 100, 200 ug/mL PM10
24 h
cells by particles collected while
steel mill in operation (years 1 and 3).
Increase noted for year 2 for particles
taken during plant closure, but not
dose-dependent; and particles
collected during plant closure had the
lowest concentrations of soluble Fe,
Cu, and Zn. Cytotoxicity seen at
500 |ig/mL.
Water soluble fraction caused greater
release of IL-8 than insoluble fraction.
The effect was blocked by
deferoxamine and presumably
because of metals (Fe, Cu, Zn, Pb).
Dose-dependent increase in
expression of IL-8 produced at
> 50 ug/mL by particles collected
when the steel mill was in operation;
effects seen at lowest dose tested.
Ohio et al.
(1999a)
Wuetal. (2001)
-------
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF AMBIENT PARTICIPATE MATTER AND
PARTICIPATE MATTER CONSTITUENTS
oo
oo
Species, Cell
Type, etc. "
Human AM
Human and
rat AM
Human AM and
blood monocytes
M and F
20-35 year
Human lung
epithelial (A549)
cells
0X174 RFI DNA
Particle or
Constituent' Cell Count
PM10 extract 2 x 105 cells/mL
(Provo, UT)
Four urban air particles 2.5 x 105 cells/mL
(UAP): St. Louis;
Wash, DC; Ottawa
ERC-93; Dusseldorf).
ROFA (Florida)
MSH Vol. Ash
DPM
Silica
Urban air particles 2 x 105 cells/mL
(UAP): St. Louis SRM
1648; Washington, DC,
SRM 1649; Ottawa,
Canada, EHC-93
ROFA (Florida #6)
MSH Vol. Ash
Urban particles: 20,000 cells/cm2
SRM 1648, St. Louis
SRM 1649,
Washington, DC
Concentration
500 ug/mL
Urban and DPM:
12,27, 111, 333, or
1000 ug/mL
SiO2 and TiO2:
4, 12, 35, 167,
or 500 ug/mL
Fe203: 1:1,3:1;
10:1 particles/cell
ratio
33 or 100 ug/mL
100 ug/cm2for Fe
mobilization assay
Particle Size Exposure Duration
PM10 24 h
Urban particles: 2 h for cytotoxicity,
0.3-0.4 urn 16-18 h for cytokine
DPM: 0.3 urn assay;
ROFA: 0.5 um chemiluminescence
MSH Vol. Ash: at 30 minutes
1.8 urn
Silica: 05-10 urn
TiO2: < 5 urn
Latex: 3.8 um
0.2 to 0.7 urn 3, 6, or 18-20 h
SRM 1648: Up to 25 h
50% < 10 urn
SRM 1649:
30% < 10 urn
Effect of Particles
AM phagocytosis of (FITC)-labeled
Saccharomyces cerevisiae inhibited
30% by particles collected before
steel mill closure.
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). The effects were seen in
human AM at UAP concentrations of
> 56 ug/mL and in rat AM at all
exposures. ROFA induced strong
chemiluminescence (all cone, in
humans and > 35 ug/mL in rats) but
had no effects on TNF production.
AM and MO phagocytosis inhibited
by exposure to 100 ug/mL UAP for
18h. UAP caused decreased
expression of p2-integrins involved in
antigen presentation and phagocytosis
in the AMs exposed to 100 ug/mL.
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.
Reference
Soukup et al.
(2000)
Becker et al.
(1996)
Becker and
Soukup (1998)
Smith and Aust
(1997)
-------
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF AMBIENT PARTICIPATE MATTER AND
PARTICIPATE MATTER CONSTITUENTS
oo
VO
Species, Cell
Type, etc. "
Rat AM
Rat AM
Human
erythrocytes;
mouse monocyte-
macrophage cell
RAW 264.7
Peripheral blood
monocytes
0X174 RF1 DNA
Particle or
Constituent' Cell Count Concentration
UAP (St. Louis) 1 x 106 for 25 to 200 ug/mL
DPM TNF-ct
secretion;
3 x 106 cells/mL
for gene
egression
PM10 (from 10 ug/cm2
Mexico City
1993);
MSH Vol. Ash
PM10.25; PM25 1 x 106 cells/mL 5 doses across
(Rome, Italy) Raw cells range of 0 to
80 ug/mL saline
solution
Organic extract 1 x 10" cells/mL 5, 10, 21, 42, 85,
of TSP, Italy 340 ug
PM10 3.7 or 7.5
(Edinburgh, ug/assay
Scotland)
Exposure
Particle Size Duration
DPM: 1.1-1. 3 urn 2 h incubation;
supernatant
collected
following 1 8 h
of culture
< 10 urn 24 h
PM2 5 1 h for hemolysis
PM10_2 5 24 h for oxidative
stress
N/A, collected 2 h
from high-volume
sampler (60 m3/h)
PM10 8 h
Effect of Particles
Dose-dependent increase in TNF-ct, IL-6, CINC,
MIP-2 gene expression by UAP but not DPM (TNF-ct
increase at all doses with peak at 200 ug/mL).
Cytokine production not related to ROS but inhibited
by polymyxin B; LPS detected on UAP but not DPM.
Endotoxin responsible for cytokine gene expression
induced by UAP in AM. Increase in gene expression
determined semiquantitatively.
PM10 stimulated AMs to induce up-regulation of
PDGF °= receptor on myofibroblasts. Endotoxin and
metal components of PM10 stimulate release of IL-p.
This is a possible mechanism for PM10-induced airway
remodeling.
Increased hemolysis of erythrocytes linearly related
to PM2 5 doses across 0 to 80 ug/mL range, but not
to PM10_25 below 50 ug/mL. However, little difference
seen between PM2 5 and PM10_2 5 effects based on
surface per volume unit of suspension, suggesting that
oxidative stress on cell membranes is related to PM
surface area. PM2 5 also caused dose-dependent
decrease in viability and increased markers of
inflammation in RAW 264.7 cells, but significance
levels not reported. .
Superoxide anion generation was inhibited at
a particulate concentration of 0. 17 mg/mL (340 ug)
when stimulated with PMA; dose-dependent increase
in LDH; at 0. 17 mg/mL LDH increased 50%;
disintegration of plasma membrane.
Significant free radical activity on degrading
supercoiled DNA at both concentrations; mainly
because of hydro xyl radicals (inhibited by mannitol);
Reference
Dong et al.
(1996)
Bonner et al.
(1998)
Diociaiuti et al.
(2001)
Fabiani et al.
(1997)
Gilmour et al.
(1996)
Fe involvement (DEF-B conferred protection); more
Fe3+ was released compared to Fe2+, especially at pH
4.6 than at 7.2.
-------
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF AMBIENT PARTICIPATE MATTER AND
PARTICIPATE MATTER CONSTITUENTS
Species, Cell Type,
etc."
Supercoiled
DNA
Particle or
Constituent*
PM10 from
Edinburgh,
Scotland
Cell Count Concentration Particle Size
996.2 ±181. 8 PM10
ug/filter in
100 uL
Exposure
Duration Effect of Particles
8 h PM10 caused damage to DNA; mediated by hydroxyl
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.
Reference
Donaldson et al.
(1997)
Human AM from
smokers (mean age
68) and
nonsmokers (mean
age 72), male and
female
EHC-93
(Ottawa)
ROFA
latex beads
carbon particles
0.5 x 106
cells/mL
0.01-O.lmg/mL
: 10 urn
0.1,1, and 10 urn
2, 4, 8, 12, TNF-a increased at 0.01 to 0.1 mg/mL EHC-93 and at Van Eeden, et al.
and 24 h (only 0.1 mg/mL latex, carbon and ROFA. EHC93 at (2001)
24 h data 0.1 mg/mL increased levels of IL6, IL-1 p, MIP-1 a,
shown) and GM-CSF
Human AM
from age 62 ± 5
smokers
Rabbit AM,
6 weeks old
Rat AM and AM
primed with LPS
Rat AM and AM
primed with LPS
EHC-93
(Ottawa)
PM25 (Boston) 1 x 106
Indoor and outdoor cells/mL
Boston CAPS,
separated in
soluble/insoluble
fractions
SRM1649, iron
oxide, carbon
black, diesel dust
0.5x10' 0.01-0.1 mg/mL 4-5ummass 2,4,8,12, 0.1 mg/mL produced significant increase in TNF-a. Mukaeetal.
cells/mL median diameter and 24 h Instillation of supernatants from human and rabbit PM- (2000)
(only 24 h exposed AMs into the lungs of rabbits caused increases
data shown) in circulating PMNs and circulating band cells and
shortening the transit time of PNMs through mitotic
and postmitotic bone narrow pools.
100 ug/mL < 2.5 um 20 h Increased TNF production in both indoor and outdoor Long et al. (2001)
exposures. LPS-primed AMs had greater responses.
Indoor PM2 5 caused significantly more TNF
production than outdoor PM2 5.
2.4x10' 100 ug/mL Fe, CV and DD 20 h Priming enhanced AM release of TNF and MIP-2 in Imrich et al.
cells/mL all < 1 um, UAP response to UAP and some CAPs samples. Other (2000)
was 30% larger CAPs and CB, DD, Fe did not induce cytokines.
Toxicity associated with insoluble fractions. The
activation state of the AM determines which particle-
associated components are most bioactive.
Mouse AM
Boston CAPs
1 x 10"
cells/mL
-5-120 ug/mL
< 2.5 urn
5h
Soluble and insoluble CAPs caused MIP-2 and TNF-a
production. Cytokine induction and endotoxin content
was associated with the insoluble fraction. PB
neutralization of endotoxin abrogated > 80% of TNF-a
induction, but inhibited MIP-2 production by only
40%.
Ning et al. (2000)
-------
TABLE 7-10 (cont'd). IN VITRO EFFECTS OF AMBIENT PARTICIPATE MATTER AND
PARTICIPATE MATTER CONSTITUENTS
Species, Cell
Type, etc. "
Rat AM
Human PMN
RLE-6TN cells
(type II like
cell line)
Human AM
Human AM
from healthy
males and
females, age
20-35 years
CHO
expressing
CD 14 and
TLR2 or TLR4
Particle or
Constituent" Cell Count
Switzerland PM 4 x 105/mL
collected during the
four seasons.
Aqueous and organic 1 x 106
extracts of TSP cells/mL
(from Dusseldorf and
Duisburg, Germany)
PM25 1 x 106
(Burlington, VT); cells/mL
Fine/ultrafine TiO2
Chapel Hill PM 2 x 106
extract; both H20 cells/mL
soluble(s) and
insoluble(is)
EHC-93 (Ott) 2-3 x 105
MSH Vol. Ash cells/mL
ROFA (Niagra, NY)
silica
PM bacteria from
Chapel Hill, NC
ambient air
Concentration
0.5 land 0.78
mg/mL (aqueous
extracts)
0.42-0.65
mg/mL (organic
extracts)
a-quartz,
[0-200 ug/mL]
1,2.5,5,
10 ug/mL PM2 5,
or up to 5 ug/mL
Ti02
100 ug/mL
PM - 30 ug/mL;
bacteria -
103-2xl06/tube
Exposure
Particle Size Duration
< 10 urn 40 h
Collected by high Up to
volume sampler, 35 min
90% < 5 urn,
50% < lum, max
at 0.3-0.45 urn
Water and
dichloromethane
used to yield
aqueous and
organic extracts
PM25: 39 nm 24 and 48 h
Fine TiO2: exposure
159 nm
UFTiO2: 37 nm
PM25 24 h
PM10.,5
2.5-10 um overnight
Effect of Particles
All exposures produced significant toxicity in MTT
assay. Spring and summer samples induced the most
TNF-a. Oxidative response was greatest in non-winter
months.
PM aqueous extract significantly stimulated the
production and release of ROS at 0.42 mg/mL 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. Phagocytosis
not affected.
PM increases in c-Jun kinase activity at > 10 ug/mL,
levels of phosphorylated c-Jun immunoreactive protein
at > 5 ug/mL; and transcriptional activation of activator
protein- 1-dependent gene expression; elevation in
number of cells incorporating 5'-bromodeoxyuridine at
> 1 ug/mL. . UF TiO2 increased c-Jun kinase activity
compared to fine TiO2.
Increased cytokine production (IL-6, TNF-a, MCP-1);
isPM10 > sPM10 > isPM25; sPM25 was inactive;
endotoxin was partially responsible.
Three times more gram+ bacteria were required to elicit
the same level of cytokine induction as gram- bacteria.
This induction was inhibited by anti-CD 14 and required
serum. TLR4 was involved in PM10_2 5 and gram-
induced activation. TLR2 activation was induced by
both gram + and - bacteria and by PM.
Reference
Monn et al.
(2003)
Hitzfeld et al.
(1997)
Timblin et al.
(1998)
Soukup and
Becker (2001)
Becker et al.
(2002)
-------
-------
Rat AM
NHBE
BEAS-2B
VO
TABLE 7-11. IN VITRO EFFECTS OF ROFA AND OTHER COMBUSTION-SOURCE
PARTICULATE MATTER CONSTITUENTS
Species, Cell
Type, etc. "
Particle or Constituent'
Cell Count
Concentration
Particle Size
Exposure
Duration
Effect of Particles
Reference
ROFA (Florida), iron
sulfate, nickel sulfate,
vanadyl sulfate
Latex particles with
metal complexed on the
surface
ROFA (Florida)
0.5-l.OxlO6 0.01-1.Omg/mL 3.6umMMAD Up to At all concentrations, increased Ghioetal.
cells/mL (dust) 400 min chemiluminescence, inhibited by DEF and (1997a)
hydroxyl radical scavengers; solutions of metal
0.945 um sulfates and metal-complexed latex particles
(latex beads) similarly elevated chemiluminescence. Effects
were generally dose-dependent, with largest dose
creating effects over the shortest period of
exposure.
5, 50, 200 ug/mL 3.6 um 2 and 24 h mRN A for ferritin did not change; ferritin protein Ghioetal.
increase at > 50 ug/mL; mRNA for transferrin (1998c)
receptor decreased at >50 ug/mL; mRNA for
lactoferrin increased; transferrin decreased at
> 50 ug/mL, whereas lactoferrin increased at > 50;
deferoxamine alone increased lactoferrin mRNA;
effects significant for two highest exposure
following 24 h exposure.
BEAS-2B ROFA (Florida)
respiratory
epithelial cells
NHBE cells ROFA (Florida)
Primary cultures ROFA; (Florida)
of RTE metal solutions
Hamster AM ROFA or CAPs
(Boston)
100 ug/mL
0, 5, 50, or
200 ug/mL (actual
dose delivered
1.6-60 ug/cm2)
5, 10, or 20 ug/cm2
0.5 x 106 ROFA: 0, 25, 50,
cells/mL 100, or 200 ug/mL
CAPs: 1:5, 1:10,
1:20 (described as
4, 10, 20 ug/mL)
3.6 um
< 10 um
1.95 umMMAD
CAPs:
0.1-2.5 um
(from Harvard
concentrator)
TiO2: 1 um
5 min - 1 h
2 or 24 h
Analysis at
24 h
30 min
incubation,
analysis
immediately
following
Lactoferrin binding with PM metal occurred
within 5 min and Fe (m), but not Ni, increased the
concentration of lactoferrin receptor.
Increase in expression of the cytokines IL-6 and
IL-8 at all exposure concentrations; TNF-a
increased at > 50 ug/mL; inhibition by DMTU or
deferoxamine.
ROFA, V, or Ni + V (at > 10 ug/cm2), 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.
Dose-dependent increase in AM oxidant stress
with both ROFA and CAPs (at 4 ug/mL). 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.
Ohio et al.
(1999b)
Carter et al.
(1997)
Dye et al.
(1999)
Goldsmith
etal. (1997)
-------
TABLE 7-11 (cont'd). IN VITRO EFFECTS OF ROFA AND OTHER COMBUSTION-SOURCE
PARTICULATE MATTER CONSTITUENTS
Species, Cell type,
etc."
Hamster AM
Mouse AM
Human AM
BEAS-2B
BEAS-2B
Primary GPTE cells
Human blood
monocytes and
neutrophils (PMN)
Particle or
Constituent* Cell Count
ROFA, CAPs, and 0.5 x 106
their water-soluble cells/mL
and particulate
fractions
(Boston)
ROFA (Florida) 1 x 106
UAP (#1648, 1649) cells/mL
MSH Vol. Ash
ROFA
(Florida, #6 LoS)
ROFA 5 x 106
(Florida) cells/mL
ROFA (Florida) 2 - 5 x 105
DOFA (Durham) cells/cm2
STL (St. Louis)
WDC (Wash., DC)
OT (Ottawa)
MSH Vol. Ash
ROFA (Florida); 2 x 105 cells/
CFA (Linden, NJ; 0.2 mL
Niagra, NY;
Western U.S.).
SRM 1649
(Dusseldorf,
Eliz. City, NJ;
Charlottesville);
MSH Vol. Ash
Concentration Particle Size
ROFA: 25, 50, CAPs =
100, 200 ug/mL, 0. 1-2.5 urn
50 and 250
ug/mL, and 100,
200, 400 ug/mL
CAPs: 38-180 ROFA =1.0 urn
ug/mL
0, 25, 100, or Volume median
200 ug/mL diameter:
ROFA: 1.1 urn
#1648: 1.4 urn
#1649: 1.1 urn
MSH: 2.3 urn
0, 6, 12, 25, or 1.96 urn
50 ug/mL
0,0.5, or 2.0 mg 1.95 urn
in 10 mL
6.25, 12.5, 25, N/A
and 50 ug/cm2
100 ug N/A
25, 50, 100, 150,
200 ug
Exposure
Duration Effect of Particles
30 min ROFA (particles -50, 100, and 200 ug/mL and water
soluble components - 200 ug/mL only dose tested) and
CAPs (all doses for particulate fraction and 150-180
ug/mL for soluble fraction - only dose tested)
caused increases in DCFH oxidation; CAPs samples
and components showed substantial day-to-day
variability in their oxidant effects; ROFA increased
MIP-2 in hamster AMs exposed to 50 or 250 ug/mL
and TNF-a production in mouse AM exposed to 100,
200, 400 ug/mL. Effects inhibited by NAC.
24 h ROFA highly toxic; urban PM toxic at 200 ug/mL;
ROFA produced significant apoptosis as low as
25 ug/mL; UAP produced apoptosis at 100 ug/mL;
ROFA and UAP also affect AM phenotype:
increased immune stimulatory, whereas decreased
immune suppressor phenotype.
1 to 24 h Transient activation at 50 ug/mL of IL-6 gene by NF-
KB activation and binding to specific sequences in
promoter of IL-6 gene at all dose levels; inhibition of
NF-KB activation by DEF and NAC; activation
NF-B may be a critical first step in the inflammatory
cascade following exposure to ROFA particles.
1 h ROFA induced production of acetaldehyde in dose-
dependent fashion. No effects on cell viability.
4, 8, and 24 h ROFA the most toxic (effects seen at 12.5 ug/cm2),
enhancing mucin secretion at 50 ug/cm2 and causing
toxicity, assessed by LDH release at > 25 ug/cm2.
DOFA produced significant effect at 25 ug/cm2.
Several other particles toxic at 50 ug/cm2 for 24 h.
40 min ROS generation, measured by LCL increase in PMN
and monocytes; PMN effects were correlated with Si,
Fe, Mn, Ti, and Co content but not V, Cr, Ni, and Cu.
Deferoxamine, a metal ion-chelator, did not affect LCL
in PMN, suggesting that metal ions are not related to
induction of LCL. Effects were generally dose-
dependent, with effects seen at lowest dose.
Reference
Goldsmith et al.
(1998)
Holian et al.
(1998)
Quay et al.
(1998)
Madden et al.
(1999)
Jiang et al.
(2000)
Prahalad et al.
(1999)
-------
TABLE 7-11 (cont'd). IN VITRO EFFECTS OF ROFA AND OTHER COMBUSTION-SOURCE
PARTICULATE MATTER CONSTITUENTS
Species, Cell Type,
etc."
Human
BEAS-2B
Human
BEAS-2B
Human airway
epithelium-derived
cell lines BEAS-2B
Human lung
mucoepidermoid
carcinoma cell line,
NCI-H292
Rat, Long Evans
epithelial cells
BEAS-2B
Particle or
Constituent* Cell Count
ROFA (Florida)
ROFA (Florida)
Synthetic ROFA
(soluble Ni, Fe,
andV)
Particle
components As,
Cr, Cu, Fe, Ni, V,
andZn
ROFA 1 x 106
(Florida) cells/mL
CFA 1 x 10" cells/
PFA 100 uL
a-quartz.
ROFA
(Birmingham, AL)
188mg/gofVO
Concentration Particle Size
2, 20, or 1.96 urn
60 ug/cm2
ROFA: ROFA: 1.96 urn
0-200 ug/mL Synthetic ROFA:
Synthetic ROFA N/A (soluble)
(100 ug/mL):
Ni, 64 uM
Fe, 63 uM
V, 370 mM
500 uM of As, N/A (soluble)
F, Cr (III), Cu,
V, Zn
10, 30, 100 N/A
ug/mL
1.5-3.0 urn
17.7 urn
2.5 urn
100 ug/mL N/A
Exposure
Duration
2 or 24-h
exposure
5 min to 24 h
20 min;
analyses
conducted
6 and 24 h
following
exposure
6 and 24 h
3h
2-6 h
Effect of Particles
Epithelial cells exposed to ROFA at > 20 ug/cm2 for
24 h secreted substantially increased amounts of the
PHS products prostaglandins E2 and F2a; ROFA-
induced increase in prostaglandin synthesis was
correlated with a marked increase in PHS activity.
Tyrosine phosphatase activity, which was known to
be inhibited by vanadium ions, was markedly
diminished after ROFA treatment at > 50 ug/mL;
effects were dose- and time-dependent; ROFA
exposure induces vanadium ion-mediated inhibition of
tyrosine phosphatase activity, leading to accumulation
of protein phosphotyrosines in cells.
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. Most
effects seen by 6 h postexposure.
Epithelial cells secreted increased mucin at
> 10 ug/mL and lysozyme > 30 ug/mL; effect was
time- and concentration-dependent; effects significant
for mucin at the lowest exposure dose for both
exposure periods; effects on lysozyme only significant
at highest dose for 6 h exposure and two highest doses
for 24 h exposure; caused by V-rich fraction (18.8%).
CFA produced highest level of hydro xyl radicals; no
relationship between hydroxy/radical generation and
CFA particle size, surface area, quartz, or iron content,
but positive correlation noted with iron mobilization.
ROFA caused increased intracellular Ca2+, IL-6, IL-and
TNF-a through activation of capsicin-
and pH-sensitive receptors; effects seen at the lowest
Reference
Samet et al.
(1996)
Samet
et al. (1997)
Samet et al.
(1998)
Longphere et al.
(2000)
Van Maanen
et al. (1999)
Veronesi et al.
(1999a)
dose tested.
-------
TABLE 7-11 (cont'd). IN VITRO EFFECTS OF ROFA AND OTHER COMBUSTION-SOURCE
PARTICULATE MATTER CONSTITUENTS
Species, Cell Type,
etc."
Human lung
epithelial (A549)
cells
Male (Wistar) rat
lung macrophages
Human lung
epithelial (A549)
cells
Rat (Wistar) AM
RAM cells
(a rat AM cell line)
Human bronchial
epithelial cells,
asthmatic (ASTH)
nonasthmatic
(NONA)
Human bronchial
epithelial cells
(smokers)
Particle or
Constituent" Cell Count
ROFA (Boston), 2.5 x 105
a-quartz, TiO2 cells/mL
Urban dust SRM 2 x 105
1649, TiO2, quartz cells/mL
TiO2, Fe2O3, CAP 3 x 105
(Boston), and the cells/mL
fibrogenic particle
a-quartz
TiO2 1 x 106
cells/mL
DPM
DPM
Concentration Particle Size
25-200 ug/mL N/A
0-100 ng/mL 0.3 -0.6 urn
TiO3: 40 N/A
ug/mL; Fe2O3:
100 ng/mL,
a-quartz: 200
ug/mL;
CAP: 40 ug/mL
20, 50, or N/A
80 ug/mL
10-100ug/mL 0.4 urn
10-100 ug/mL 0.4 urn
in culture
medium
50 |ig/mL
filtered solution
Exposure
Duration Effect of Particles
60 min Exposure of A549 cells to ROFA at > 50 ug/mL,
a-quartz at > 25 ug/mL, but not TiO2, caused increased
IL-8 production in TNF-a primed cells.
18 h Cytotoxicity ranking was quartz > SRM 1649 > TiO2,
based on cellular ATP decrease and LDH, acid
phosphatase, and p-glucuronidase release. Effects
were noted at the lowest exposure dose.
24 h TiO2 > Fe2O3 > a-quartz > CAP in particle binding;
binding of particle was found to be calcium-dependent
for TiO2 and Fe2O3, while a-quartz binding was
calcium-independent; scavenger receptor, mediate
particulate binding; a-quartz caused a dose-dependent
production of IL-8 (> 26.6 ug/cm2). 11-8 was not
present in TiO2 and CAPs treated cells.
4 h Opsonization of TiO2 with surfactant components
resulted in a modest dose-dependent increase in AM
uptake compared with that of unopsonized TiO2 at
> 50 ug/mL; surfactant components increase AM
phagocytosis of particles.
2, 4, 6, 24 h DPM caused no gross cellular damage. Ciliary beat
frequency was attentuated at all doses. DPM caused
IL-8 release at 10 ug/mL in ASTH and at 50 ug/mL in
NONA. Higher concentrations (50 and 100 ug/mL)
DPM suppressed IL-8 and GM-CSF, in ASTH cells.
24 h DPM attenuated ciliary beating. Release of IL-8
protein increased by exposure to > 50 ug/mL DPM in
culture medium, but 10-fold higher increase by DPM
filtered solution. GM-CSF and CAM-1 increased after
50-100 ug/mL.
Reference
Stringer and
Kobzik(1998)
Nadeau et al.
(1996)
Stringer et al.
(1996)
Stringer and
Kobzik
(1996)
Bayram et al.
(1998a)
Bayram et al.
(1998b)
Cell types: RTE = Rat tracheal epithelial cells; GPTE = Guinea pig tracheal epithelial cells; NHBE = Normal human bronchial epithelial cells; A549 = Human lung epithelial cell line;
BEAS-2B = human airway epithelial cell line; AM = Alveolar macrophage.
CAP = Concentrated ambient particles
UAP = Urban ambient PM
CFA = Coal fly ash
ROFA = Residual oil fly ash
DOFA = Domestic oil fly ash
PFA = Pulverized fuel ash
DPM = Diesel particulate matter
VO = Vanadate oxide
TiO, = Titanium oxide
TSP = Total suspended particles
MSH = Mt. Saint Helen volcanic ash
DEF = Deferoxamine
-------
very high on a cellular basis, thus requiring much caution in attempting to extrapolate in vitro
findings to in vivo exposure conditions. It would be useful if in vitro studies included, in
addition to the high doses, lower doses more comparable to environmental doses predicted to
occur at the cellular level under in vivo conditions. Even with these limitations, however,
in vitro studies do provide a useful approach by which to explore potential cellular and
molecular mechanisms by which PM mediates health effects, allowing mechanisms identified
in vitro to be evaluated later in vivo or possibly helping to confirm mechanisms suggested by the
results of in vivo studies.
The following subsection discusses the more recently published studies that provide an
in vitro approach toward identifying potential mechanisms by which PM mediates
cardiovascular and respiratory health effects. Based on these available data the ensuing
subsection then discusses the potential mechanisms in relation to PM size or chemical
characteristics.
7.4.2 Ambient Particle Effects
Newly available in vitro studies since the 1996 PM AQCD include many that exposed
airway epithelial cells, alveolar macrophages, or blood monocytes and erythrocytes to aqueous
extracts of ambient PM to investigate cellular processes, e.g., oxidant generation and cytokine
production, that may contribute to respiratory and/or cardiovascular pathophysiological
responses seen in vivo. The ambient PM evaluated includes samples collected from: Boston,
MA (Goldsmith et al., 1998); North Provo, UT (Ohio et al., 1999a,b); St. Louis, MO (SRM
1648, Dong et al., 1996; Becker and Soukup, 1998); Washington, DC (SRM 1649, Becker and
Soukup, 1998); Ottawa, Canada (EHC-93, Becker and Soukup, 1998); Dusseldorf and Duisburg,
Germany (Hitzfeld et al., 1997), Mexico City (Bonner et al., 1998), Terni, Italy (Fabiani et al.,
1997); and Rome, Italy (Diociaiuti et al., 2001).
Because soluble metals from ROFA-like particles have been shown to be associated with
pathophysiologic/toxic effects, several new studies have investigated whether the soluble
components of ambient PM may exert similar effects. For example, extracts of ambient PM
samples collected from North Provo, UT, (during 1981 and 1982) were used to test whether the
soluble components or ionizable metals, which accounted for only -0.1% of the mass, are
responsible for the biological activity of the extracted PM components. These North Provo
7-97
-------
extracts stimulated IL-6 and IL-8 production, as well as increased IL-8 mRNA and enhanced
expression of intercellular adhesion molecule-1 (ICAM-1) in human airway epithelial
(BEAS-2B) cells (Kennedy et al., 1998). Cytokine secretion was preceded by activation of
nuclear factor kappa B (NF-KB) and was reduced by treatment with superoxide dismutase
(SOD), Deferoxamine (DBF), or N-acetylcysteine (NAC). The addition of similar quantities
of Cu2+ as found in the Provo extract replicated the biological effects observed with particles
alone. When normal constituents of airway lining fluid (mucin or ceruloplasmin) were added to
BEAS cells, particulate-induced secretion of IL-8 was modified. Mucin reduced IL-8 secretion;
whereas ceruloplasmin significantly increased IL-8 secretion and activation of NF-KB. The
authors suggest that copper ions may cause some of the biologic effects of inhaled PM in the
Provo region and may provide an explanation for the sensitivity of asthmatics to Provo PM seen
in epidemiologic studies. Also, release of IL-8 from BEAS-2B cells, oxidant generation
(thiobarbituric acid reactive products), and PMN influx in rats exposed to these samples
correlated with sulfate content and the ionizable concentrations of metals in such Provo-PM
extracts (Ohio et al., 1999a,b).
Frampton et al. (1999) examined the effects of the same ambient PM samples collected
from Utah Valley in the late 1980s (see Section 7.2.1). Aqueous extracts of the filters were
analyzed for metal and oxidant production and added to BEAS-2B cultures for 2 or 24 h.
Particles collected in 1987, when the steel mill was closed had the lowest concentrations of
soluble Fe, Cu, and Zn and showed the least oxidant generation. Ambient PM collected before
and after plant closing induced expression of IL-6 and IL-8 in a dose-response relationship (125,
250, and 500 |ig/mL) with effects seen at all doses. Ambient PM collected after reopening of the
steel mill also caused cytotoxicity, as demonstrated by microscopy and LDH release at the
highest concentration used (500 |ig/mL).
Molinelli et al. (2002) also exposed BEAS-2B cultures for 24 h to an aqueous extract of
PM collected in the Utah Valley. A portion of the extract was treated with Chelex, an agent that
removes transition metals from solution. Cells incubated with the untreated extract showed a
significant concentration-dependent increase at the lowest concentration of 62.5 |ig/mL in the
inflammatory mediator IL-8 when compared to the control cells. However, cells incubated with
Chelex-treated extract produced no change (relative to control) in IL-8. They exposed rats
in vivo for 24 h to the same treatments as the in vitro cells and found significant increases in
7-98
-------
lactate dehydrogenase (LDH) and total protein in the rats exposed to the untreated extract and to
the Chelex-treated extract with metals added back to achieve original concentrations. There was
an attenuation of the observed LDH and total protein increases in the rats instilled with the
Chelex-treated extract. The authors concluded that removal of metal cations attenuates cellular
responses to the aqueous extract and suggest a role for transition metal involvement in
PM-associated increases in morbidity and mortality.
Soukup et al. (2000), using similar ambient PM extracts as did Frampton et al. (1999),
examined effects on human alveolar macrophages (AM). The phagocytic activity and oxidative
response of AMs were measured after segmental instillation of aqueous extracts from the Utah
Valley or after overnight in vitro cell culture exposure to such ambient PM at 500 |ig/mL.
Ambient PM collected before closure of the steel mill inhibited AM phagocytosis of (FITC)-
labeled Saccharomyces cerevisiae by 30%; but no significant effect on phagocytosis was seen
with the other two extracts. Furthermore, although extracts of ambient PM collected before and
after plant closure inhibited oxidant activity of AMs when incubated overnight in cell culture,
only the former particles caused an immediate oxidative response in AMs. Host defense effects
were attributed to apoptosis, which was most evident for exposures to particles collected before
plant closure. Interpretation of loss of these effects by chelation removal of the metals was
complicated by the observed differences in apoptosis despite similar metal contents of ambient
PM collected during the steel mill operation.
Wu et al. (2001) investigated intracellular signaling mechanisms related to pulmonary
responses to Utah Valley PM extracts. Human primary airway epithelial cells were exposed to
aqueous extracts of PM at doses of 50, 100, or 200 |ig/mL collected from the year before, during,
and after the steel mill closure in Utah Valley. Transfection with kinase-deficient extracellular
signal-regulated kinase (ERK) constructs partially blocked the PM-induced IL-8 promoter
reporter activity that was present at all doses. The mitogen-activated protein kinase/ERK kinase
(MEK) activity inhibitor PD-98059 significantly abolished IL-8 released in response to the PM,
as did the epidermal growth factor (EGF) receptor kinase inhibitor AG-1478. Western blotting
showed that the PM-induced phosphorylation of EGF receptor tyrosine, MEK1/2, and ERK1/2
could be ablated with AG-1478 or PD-98059. The results indicate that the potency of Utah
Valley PM collected during plant closure was lower than that collected while the steel mill was
7-99
-------
was in operation and imply that Utah Valley PM can induce IL-8 expression at concentrations
> 50 |ig/mL partially through the activation of the EGF receptor signaling.
There are regional as well as daily variations in the composition of ambient PM and, hence,
its biological activities. For example, concentrated ambient PM (CAP), from Boston urban air
has substantial day-to-day variability in its composition and oxidant effects (Goldsmith et al.,
1998). Similar to Utah PM, the water-soluble component of Boston CAPs significantly
increased AM oxidant production and inflammatory cytokine (MIP2 and TNF-a) production
over negative control values. These effects could be blocked by metal chelators or antioxidants,
suggesting important roles for metals in contributing to the observed Boston particle effects on
AM function.
In another study, the effects of water soluble as well as organic components (extracted in
dichloromethane) of ambient PM were investigated by exposing human PMN to PM extracts
(Hitzfeld et al., 1997). PM was collected with high-volume samplers in two German cities,
Dusseldorf and Duisburg, sites having high traffic and high industrial emissions, respectively.
Organic, but not aqueous, extracts of PM at concentrations of 0.03 |ig/mL alone significantly
stimulated production and release of ROS in resting human PMN. The effects of the PM
extracts were inhibited by SOD, catalase, and sodium azide (NaN3). Similarly, the organic
fraction (extractable by acetone) of ambient PM from Terni, Italy, was shown to produce
cytotoxicity, superoxide release in response to PMA, and zymosan in peripheral monocytes
(Fabiani et al., 1997).
Becker and Soukup (1998) found interesting differences between biological activity of PM
materials from urban air particle (UAP) sources (baghouse collection in St. Louis and Ottawa),
ROFA samples from a power plant, and Mt. St. Helen volcanic ash stored since 1980. Exposure
of human AMs and blood-derived monocytes (MO) to 100 |ig/mL of UAP (0.2 to 0.7 jim
MMAD originally) both from Boston and St. Louis reduced expression of certain receptors
(important for recognition of microbial entities), the phagocytosis of bioparticles (yeast cell
walls), and oxidant generation (an important bactericidal mechanism) in both AM and MO.
All of these were little affected at 33 |ig/mL of UAP. Exposure to 100 |ig/mL of ROFA (0.5 jim
MMAD originally) also significantly decreased AM (but not MO) phagocytosis (likely due to
ROFA cytotoxic effects on AM), but the volcanic ash had little effect on phagocytosis. The
oxidative burst response was significantly decreased by ROFA in both AM and MO, but only in
7-100
-------
the AM by the volcanic ash. Administration of 10 mg/mL of lipopolysaccaride (LPS), the active
endotoxin component, reduced AM receptor expression similar to UAP, but did not reduce all
the same receptor expression as UAP in MO. The authors noted that their results indicated
(a) differences in biological activity between urban air-related particles (both baghouse collected
and ROFA) and the more inert Mt. St. Helens volcanic ash particles (that had little effect on any
of the receptors or phagocytosis functions studied); and (b) that UAP endotoxin content may be
an important effector in UAP-modulation of some, but certainly not all, macrophage functions.
The findings of Dong et al. (1996) also suggest that biological activity of some ambient
PM materials may result from the presence of endotoxin on the particles. Using the same urban
particles (SRM 1648), cytokine production (TNF-a, IL-1, IL-6, CINC, and MIP-2) was
increased in macrophages following treatment with 50 to 200 |ig/mL of urban PM (Dong et al.,
1996). The urban particle-induced TNF-a secretion was abrogated completely by treatment with
polymyxin B (an antibiotic that blocks LPS-associated activities), but not by antioxidants.
The potential involvement of endotoxin, at least partially, in some PM-induced biological
effects has been explored further by Bonner et al. (1998) and Soukup and Becker (2001).
Bonner et al. (1998) used urban PM10 at a concentration of 10 jig/cm2 collected from north,
south, and central regions of Mexico City with SD rat AM to examine PM effects on platelet-
derived growth factor (PDGF) receptors on lung myofibroblasts. Mexico City PM10 (but not
volcanic ash) stimulated secretion of upregulatory factors for the PDGF a receptor, possibly via
IL-1 p. In the presence of an endotoxin-neutralizing protein, the Mexico City PM10 effect on
PDGF was blocked partially, suggesting that LPS was partly responsible for the PM10 effect.
In addition, both LPS and V (both present in the PM10) acted directly on lung myofibroblasts,
even though the ambient V levels in Mexico City PM10 were probably not high enough to exert
an independent effect. The authors concluded that PM10 exposure could lead to airway
remodeling by enhancing myofibroblast replication and chemotaxis.
Soukup and Becker (2001) used a dichotomous sampler to collect fresh PM2 5 and PM10_2 5
from the ambient air of Chapel Hill, NC and compared the activity of these two particle size
fractions. Both water soluble and insoluble components at concentrations of 100 |ig/mL were
assessed for cytokine production, inhibition of phagocytosis, and induction of apoptosis. The
insoluble PM10_25 fraction was the most potent in terms of inducing cytokines and increasing
oxidant generation, thus suggesting the importance of the coarse fraction in contributing to
7-101
-------
ambient PM health effects. Endotoxin appeared to be responsible for much of the cytokine
production, whereas inhibition of phagocytosis was induced by other moieties in the coarse
material. None of the activities were inhibited by the metal chelator deferoxamine.
Diociaiuti et al. (2001) compared the in vitro toxicity of fine (PM25) and thoracic
coarse (PM10_2 5) fraction particulate materials extracted from sampling filters collected in an
urban area of Rome (average 24-h levels of 31 and 19 |ig/m3, respectively). Cell cultures were
exposed to the extracted PM materials suspended in saline at doses ranging from 0 to 80 |ig/mL.
The in vitro toxicity assays included evaluation of human red blood cell hemolysis, cell viability,
and nitric oxide (NO) release in the RAW 264.7 macrophage cell line. There was a dose-
dependent hemolysis in human erythrocytes when they were incubated with either fine or coarse
particles, but the hemolytic potential was greater for the fine than for the coarse particles in equal
mass concentration, the fine particles being linearly related from 0 to 80 |ig/mL, but the coarse
ones not showing any effect below 50|ig/mL. However, when data were expressed in terms
of PM surface area per volume of suspension, the hemolytic activity of the fine fraction was not
markedly different from that of the coarse fraction, thus suggesting that oxidative stress induced
by PM on the blood cell membranes could be due mainly to the interaction between the particle
surfaces and the cell membranes. Although RAW 264.7 cells challenged with fine and coarse
particles showed decreased viability and an increased release of NO (a key inflammatory
mediator) both effects were not dose-dependent in the tested concentration range. The fine
particles were the most effective in inducing these effects either when the data were expressed as
mass concentration or as surface area per unit volume. The authors concluded that these
differences in biological activity were due to differences in the physicochemical nature of the
particles, their noting that the possible causative agent in the fine fraction could be carbon-rich
particles with an associated sulfur-containing acidic component.
7.4.3 Comparison of Ambient and Combustion Source-Related Particles
In vitro toxicology studies using AMs as target cells (Imrich et al., 2000; Long et al., 2001;
Ning et al., 2000; Mukae et al., 2000, 2001; Van Eeden et al., 2001) have found that urban
ambient air particles are much more potent for inducing cellular responses than individual
combustion particles such as diesel or ROFA. Metals, on the other hand, do not seem to affect
cytokine production, consistent with studies showing that ROFA does not induce macrophage
7-102
-------
cytokine production. However, in some studies (Long et al., 2001), cytokine responses in AMs
appeared to be correlated with LPS content of the PM samples tested. These results may be
important because LPS is an important component associated with both ambient coarse and fine
particles (Menetrez et al., 2001).
Van Eeden et al. (2001) compared effects of EHC-93 atmospheric PM sample materials
from Ottawa, ROFA, and different size latex particles on cytokine induction by human AMs.
The EHC-93 particles produced greater than 8-fold induction of various cytokines (including
IL-1, TNF, GMCSF) at concentrations as low as 0.01 mg/mL. The other particles at
concentrations of 0.1 mg/mL only induced these cytokines by ~2-fold. Using the same EHC-93
particles, Mukae et al. (2000, 2001) found that inhalation exposure of 0.1 mg/mL also stimulated
bone marrow band cell-granulocyte precursor production and that the magnitude of the response
was correlated with the amount of phagocytosis of the particles by AMs. These results may
indicate that macrophages produce factors which stimulate bone marrow, including IL-6 and
GMCSF. In fact, AMs exposed in vitro to these particles released cytokines; and when the
supernatant of PM-stimulated macrophages was instilled into rabbits, the bone marrow was
stimulated.
In a series of studies using the same ROFA samples collected from an oil-burning power
plant in Florida, several in vitro experiments have investigated the biochemical and molecular
mechanisms involved in ROFA-induced cellular injury. Prostaglandin metabolism in cultured
human airway epithelial cells (BEAS-2B and NHBE) exposed to ROFA was investigated by
Samet et al. (1996). Epithelial cells exposed to 20 jig/cm2 ROFA for 24 h secreted substantially
increased amounts of prostaglandins E2 and F2a. The ROFA-induced increase in prostaglandin
synthesis was correlated with a marked increase in activity of the prostaglandin H synthase-2
(PHS-2) as well as mRNA coded for this enzyme. In contrast, expression of the PHS1 form of
the enzyme was not affected by ROFA treatment of airway epithelial cells. These investigators
further demonstrated that the ROFA induced a significant dose- and time-dependent increase in
protein tyrosine phosphate, an important index of signal transduction activation leading to a
broad spectrum of cellular responses. Concentrations used were 5 to 200 |ig/mL, with effects
seen at > 50 mg/mL ROFA. ROFA-induced increases in protein phosphotyrosines were
associated with its soluble fraction and were mimicked by V-containing solutions but not Fe or
Ni solutions (Samet et al., 1997).
7-103
-------
ROFA also stimulates respiratory cells to secrete inflammatory cytokines such as IL-6,
IL-8, and TNF. Normal human bronchial epithelial cells exposed to Florida ROFA produced
significant amounts of IL-8, IL-6, and TNF, as well as mRNAs coding for these cytokines
(Carter et al., 1997). Increases in cytokine production were dose-dependent. The cytokine
production was inhibited by the addition of metal chelator, DBF, or the free radical scavenger
dimethylthiourea (DMTU). Similar to the data of Samet et al. (1997), V but not Fe or Ni
compounds were responsible for these effects. Cytotoxicity and decreased cellular glutathione
levels in primary cultures of rat tracheal epithelial (RTE) cells exposed to suspensions of ROFA
indicated that respiratory cells exposed to ROFA were under oxidative stress. Treatment with
buthionine sulfoxamine (an inhibitor of y-glutamyl cysteine synthetase) augmented ROFA-
induced cytotoxicity; whereas treatment with DMTU that inhibited ROFA-induced cytoxicity
further suggested that ROFA-induced cell injury may be mediated by hydroxyl-radical-like
reactive oxygen species (ROS; Dye et al., 1997). Using BEAS-2B cells, a time- and dose-
dependent increase in IL-6 mRNA induced by ROFA was shown to be preceded by the
activation of nuclear proteins, for example, NF-KB (Quay et al., 1998). Taken together, these
studies indicate that exposure to ROFA in high doses increases oxidative stress, perturbs protein
tyrosine phosphate homeostasis, activates NF-KB, and up-regulates inflammatory cytokine and
prostaglandin synthesis and secretion to produce lung injury.
Stringer and Kobzik (1998) observed that "primed" lung epithelial cells exhibited
enhanced cytokine responses to certain particulate materials. Compared to normal cells,
exposure of tumor necrosis factor (TNF)-a-primed A549 cells to ROFA (Boston area) or
a-quartz caused increased IL-8 production in a concentration-dependent manner for particle
concentrations ranging from 0 to 200 jig/mL. Addition of the antioxidant NAC (1.0 mM)
decreased ROFA and a -quartz-mediated IL-8 production by about 50% in both normal and
TNF-a-primed A549 cells. Exposure of A549 cells to ROFA caused an increase in oxidant
levels that could be inhibited by NAC. These data suggest that (1) lung epithelial cells primed
by inflammatory mediators show increased cytokine production after exposure to PM and
(2) oxidant stress is an important mechanism for this response.
Imrich et al. (2000) found that, when mouse AMs were stimulated with CAPs (PM2 5) at a
concentration of 100 |ig/mL, the resulting TNF responses could be inhibited by an endotoxin
neutralizing agent [e.g., polymyxin-B (PB)]. Because the MIP-2 response (IL-8) was only partly
7-104
-------
inhibited by PB, however, the authors concluded that endotoxin primed AM cells to respond to
other particle components. In a related study (Ning et al., 2000), the use of PB showed that
particle-absorbed endotoxin in CAPs suspensions caused activation of normal (control) AMs,
while other (nonendotoxin) components were predominantly responsible for the enhanced
cytokine release observed for primed AMs incubated with CAPs. The non-LPS component was
not identified in this study; however, the AM biological response did not correlate with any of
several elements quantified within the insoluble CAPs samples (e.g., Al, Cd, Cr, Cu, Fe, Mg,
Mn, Ni, S, Ti, V).
Osornio-Vargas et al. (2003) compared exposures of mouse monocytes to PM2 5 or to PM10
collected in either southeastern or northern Mexico City, and characterized as to metal and
endotoxin content. Tumor necrosis factor-a and IL-6 were measured from exposures both with
and without recombinant endotoxin-neutralizing protein (rENP). The southeastern PM10 samples
had the highest endotoxin levels, which correlated with greater cytokine secretion. rENP
reduced cytokine secretion by 50-75%, suggesting to the authors that the endotoxin-independent,
transition metal-dependent mechanism for fine PM contributed to cytotoxicity effects, whereas
an endotoxin-dependent mechanism was responsible for the proinflammatory response to PM10.
Rat AM exposed to PM10 collected from both rural and urban sites in Switzerland during
all four seasons demonstrated increased cytotoxicity from all PM samples (Monn et al., 2003).
TNF-a and oxidative radical release were highest with PM collected during non-winter months.
ENP inhibited cytotoxic effects and oxidative radical release, suggesting that endotoxin in some
ambient PM10 samples during warm months may affect some types of macrophage activity.
In central Taiwan, Huang et al. (2002) collected PM25 and PM10 samples, to which RAW
264.7 cells, a mouse monocyte-macrophage cell line, were then exposed at 40 |ig/mL. After 6 h
exposure, either with or without polymyxin B, TNF-a levels were assayed. Higher TNF-a
secretion was stimulated by PM10-exposed cells; and PB-inhibited TNF-a by 42% and 32%
in PM10 and PM2 5 exposures, respectively, suggesting that endotoxin may contribute more to
TNF-a stimulation by coarse than fine fraction particles.
Becker et al. (2002) hypothesized that PM activates receptors involved in recognition of
microbial cell structures. They coated model pollution particles with either gram-negative
(Pseudomonas) or gram-positive (Staphloccocus or Streptococcus) bacteria. Three times more
gram-positive bacteria were required to elicit the same level of cytokine induction as gram-
7-105
-------
negative bacteria. This inhibition was inhibited by anti-CD14 and required serum. This study
further found a suggested role of Toll-like receptors (TLR) in PM recognition, thus implicating
likely bacterial components as a factor contributing to PM-induced AM inflammatory responses.
Becker et al. (2003) exposed human AM to ultrafine (< 0.1 |im), fine (PM01_25) or
coarse (PM2 5_10) particles collected in two urban sites in the Netherlands. IL-6 induction levels
and reductions in CD1 Ib phagocyte receptor expression were positively correlated with particle
size. Induction of IL-6 was inhibited by an antibody to CD 14. Yeast-induced oxidative burst
and inhibition of phagocytosis of opsonized yeast was also correlated with size, with the
ultrafine particles having no effect. The authors concluded that human AM recognize microbial
cell structures, which are more prevalent in larger particles, and that exposure to PM is
associated with inflammatory events and decreased pulmonary defenses.
In summary, exposure of lung epithelial cells to ambient PM or ROFA leads to increased
production of cytokines and the effects may be mediated, at least in part, through production of
ROS. Day-to-day variations in the components of PM (such as soluble transition metals) which
may be critical to eliciting the response are suggested. The involvement of organic components
(e.g., endotoxins) in ambient PM was also suggested by some studies as contributing to
ambient PM (especially coarse thoracic PM10_25) effects on some types of AM functions.
7.4.4 Potential Cellular and Molecular Mechanisms
The numerous studies assessed in the foregoing sections provide evidence for various types
of PM effects on cardiopulmonary system components and functions. Considerable interest and
research attention has been accorded to effects aimed at characterizing specific cellular and
molecular mechanisms underlying PM effects. The ensuing sections highlight information
derived in part from in vivo, but more so, from in vitro, studies that supports identification of
several general types of mechanisms as mediating various PM-induced pathophysiological
responses likely underlying PM effects on cardiopulmonary and other functions. This includes,
in particular, evidence for important involvement in mediating PM effects of (a) reactive oxygen
species; (b) intracellular signaling mechanisms; and (c) other types of mechanisms, e.g., impacts
on sensory nerve receptors.
7-106
-------
7.4.4.1 Reactive Oxygen Species
Transition metals are capable of catalyzing the production of reactive free radicals such as
the hydroxyl ( 'OH) radical through the following reaction:
+2
Fe+2 + H2O2 -»• Fe3+ *OH + HO
(7-1)
It should be noted that the actual reaction is more complex than given above and is commonly
referred to as the iron-catalyzed Haber-Weiss Reaction or Fenton's Reaction. Fe3+ produced in
the above reaction can be reduced to Fe2+ by reactions such as:
Fe3N
-Of -
> Fe2+4
-o2
(7-2)
where [O; | is the superoxide radical. Hydrogen peroxide is formed by
2 Of + 2 H+ ->• H2O2 +OJ
(7-3)
This reaction is catalyzed by superoxide dismutase (SOD). SODs are present as Cu-Zn SOD in
cytoplasm, Mn-SOD in mitochondria and extra-cellularly as Cu SOD. Iron will participate in
the above reactions and hydroxyl radicals will continue to be generated so long as there are
sufficient reductants and H2O2 present. In addition to hydroxyl radicals, electronically excited
O2 produced in the reactions given above may also be involved in promoting cellular damage.
Soluble metals from inhaled PM dissolved into the fluid lining of the airway lumen can
react directly with biological molecules (acting as a reductant in the above reactions) to produce
ROS. For example, ascorbic acid in the human lung epithelial lining fluid can react with Fe(III)
from inhaled PM to cause single strand breaks in supercoiled plasmid DNA, 4>X174 RFI (Smith
and Aust, 1997). The DNA damage caused by some PM10 samples can be inhibited by mannitol,
an hydroxyl radical scavenger, further confirming the involvement of free radicals in these
7-107
-------
reactions (Gilmour et al., 1996; Donaldson et al., 1997; Li et al., 1997). Because the clear
supernatant of the centrifuged PM10 suspension contained all of the suspension activity, the free
radical activity is derived either from a fraction that is not centrifugable (10 min at 13,000 rpm
on a bench centrifuge) or the radical generating system is released into solution (Gilmour et al.,
1996; Donaldson et al., 1997; Li et al., 1997).
In addition to measuring the interactions of ROS and biomolecules directly, the role of
ROS in PM-induced lung injury also can be assessed by measuring the electron spin resonance
(ESR) spectrum of radical adducts or fluorescent intensity of dichlorofluorescin (DCFH), an
intracellular dye that fluoresces on oxidation by ROS. Alternatively, ROS can be inhibited using
free radical scavengers, such as DMTU; antioxidants, such as glutathione or NAC; or antioxidant
enzymes, such as SOD. The diminished response to PM after treatment with these antioxidants
may indicate the involvement of ROS; but, some antioxidants (e.g., thiol-containing) can interact
with metal ions directly.
As described earlier, Kadiiska et al. (1997) used the ESR spectra of 4-POBN [a-(4-pyridyl-
l-oxide)-N-tert-butylnitrone] adducts to measure ROS in rats instilled with ROFA and
demonstrated the association between ROS production within the lung and soluble metals in
ROFA. Using DMTU to inhibit ROS production, Dye et al. (1997) had shown that systemic
administration of DMTU impeded development of the cellular inflammatory response to ROFA,
but did not ameliorate biochemical alterations in BAL fluid. Goldsmith et al. (1998), as
described earlier, showed that ROFA and CAPs caused increases in ROS production in AMs.
The water-soluble component of both CAPs and ROFA significantly increased AM oxidant
production over negative control values. In addition, increased PM-induced cytokine production
was inhibited by NAC. Li et al. (1996, 1997) instilled rats with PM10 particles (collected on
filters from an Edinburgh, Scotland monitoring station). Six hours after intratracheal instillation
of PM10, they observed a decrease in glutathione (GSH) levels in the BAL fluid. Although this
study does not describe the composition of the PM10, the authors suggest that changes in GSH,
an important lung antioxidant, support the contention that the free radical activity of PM10 is
responsible for its biological activity in vivo.
In addition to ROS generated directly by PM, resident or newly recruited AMs or PMNs
also are capable of producing these reactive species on stimulation. The ROS produced during
the oxidative burst can be measured using a chemiluminescence (CL) assay. With this assay,
7-108
-------
AM CL signals in vitro have been shown to be greatest with ROFA containing primarily soluble
V and were less with ROFA containing Ni plus V (Kodavanti et al., 1998a). As described
earlier, exposures to Dusseldorf and Duisburg PM increased the resting ROS production in
PMNs, which could be inhibited by SOD, catalase, and sodium azide (Hitzfeld et al., 1997).
Stringer and Kobzik (1998) showed that addition of NAC (1.0 mM) decreased ROFA-mediated
IL-8 production by approximately 50% in normal and TNF-a-primed A549 cells. In addition,
exposures of A549 cells to ROFA caused a substantial (and NAC inhibitable) increase in oxidant
levels as measured by DCFH oxidation. In human AMs, Becker et al. (1996) found a CL
response for ROFA, but not urban air particles (Ottawa and Dusseldorf) or volcanic ash.
Metal compounds of PM are the most probable species capable of catalyzing ROS
generation on exposure to airborne PM. To determine elemental content and solubility in
relation to their ability to generate ROS, PMN or monocytes were exposed to a wide range of air
particles from divergent sources (one natural dust, two types of oil fly ash, two types of coal fly
ash, five different ambient air samples, and one carbon black sample), and CL production was
measured over a 20-min period postexposure (Prahalad et al., 1999). Percent of sample mass
accounted for by X-ray fluorescence (XRF) detectable elements was 1.2% (carbon black); 22 to
29% (natural dust and ambient air particles); 13 to 22% (oil fly ash particles); and 28 to 49%
(coal fly ash particles). The major proportion of elements in most of these particles were
aluminosilicates and insoluble Fe, except oil derived fly ash particles in which soluble V, and Ni
were in highest concentration, consistent with particle acidity as measured in the supernatants.
All particles induced CL response in cells, except carbon black. The CL response of PMNs in
general increased with all washed particles, with oil fly ash and one urban air particle showing
statistical differences between deionized water washed and unwashed particles. These CL
activities were significantly correlated with the insoluble Si, Fe, Mn, Ti, and Co content of the
particles. No relationship was found between CL and soluble transition metals such as V, Cr, Ni,
and Cu. Pretreatment of the particles with a metal ion chelator, DEF, did not affect CL
activities. Particle sulfate content and acidity of the particle suspension did not correlate with
CL activity.
Soluble metals can be mobilized into the epithelial cells or AMs to produce ROS
intracellularly. Size-fractionated coal fly ash particles (2.5, 2.5 to 10, and < 10 jim) of
bituminous b (Utah coal), bituminous c (Illinois coal), and lignite (Dakota coal) were used to
7-109
-------
compare the amount of iron mobilization in A549 cells and by citrate (1 mM) in cell-free
suspensions (Smith et al., 1998). Iron was mobilized by citrate from all three size fractions of all
three coal types. More iron, in Fe(III) form, was mobilized by citrate from the < 2.5-|im fraction
than from the > 2.5-|im fractions. In addition, the amount of iron mobilized was dependent on
the type of coal used to generate the fly ash (Utah coal > Illinois coal = Dakota coal) but was not
related to the total amount of Fe present in the particles. Ferritin (an iron storage protein) levels
in A549 cells increased by as much as 12-fold in cells treated with coal fly ash (Utah
coal > Illinois coal > Dakota coal). More ferritin was induced in cells treated with the < 2.5-jim
fraction than with the > 2.5-|im fractions. Mossbauer spectroscopy of a fly ash sample showed
that the bioavailable Fe was assocated with the glassy aluminosilicate fraction of the particles
(Ball et al., 2000). As with the bioavailability of Fe, there was an inverse correlation between
the production of IL-8 and fly ash particle size, with the Utah coal fly ash being the most potent.
Using ROFA and colloidal iron oxide, Ohio et al. (1997b; 1998a,b,c; 1999c; 2000b) have
shown that exposures to these particles disrupted Fe homeostasis and induced the production of
ROS in vivo and in vitro. Treatment of animals or cells with metal-chelating agents such as
DEF with an associated decrease in response has been used to infer the involvement of metal
in PM-induced lung injury. Metal chelation by DEF (1 mM) caused significant inhibition of
particulate-induced AM oxidant production, as measured using DCFH (Goldsmith et al., 1998).
DEF treatment also reduced NF-KB activation and cytokine secretion in BEAS-2B cells exposed
to Provo PM (Kennedy et al., 1998). However, treatment of ROFA suspension with DEF was
not effective in blocking teachable metal induced acute lung injury (Dreher et al., 1997). Dreher
et al. (1997) indicated that DEF could chelate Fe(III) and V(II), but not Ni(II), suggesting that
metal interactions played a significant role in ROFA-induced lung injury.
Other than Fe, several V compounds have been shown to increase mRNA levels for
selected cytokines in BAL cells and to induce pulmonary inflammation (Pierce et al., 1996).
NaVO3 and VOSO4, highly soluble forms of V, tended to induce pulmonary inflammation and
inflammatory cytokine mRNA expression more rapidly and intensely than the less soluble form,
V2O5, in rats. Neutrophil influx was greatest following exposure to VOSO4 and lowest after
exposure to V2O5. However, metal components of fly ash have not been shown to consistently
increase ROS production from bovine AM treated with combustion particles (Schliiter et al.,
1995). As(III), Ni(II), and Ce(III), major components of fly ash, did inhibit the secretion of
7-110
-------
superoxide anions (O2") and hydrogen peroxide, but in the same study, O2" were lowered by
Mn(II) and Fe(II), whereas V(IV) increased O2" and H2O2. In contrast, Fe(III) increased O2"
production, showing that the oxidation state of a metal may influence its oxidant-generating
properties. Other fly ash components, e.g., Cd(II), Cr(III), and V(V), had no effects on ROS.
It is likely that a combination of several metals rather than a single metal in ambient PM is
responsible for PM-induced cellular responses. For example, V and Ni+V but not Fe or Ni alone
(in saline with the final pH at 3.0) resulted in increased epithelial permeability, decreased
cellular glutathione, cell detachment, and lytic cell injury in rat tracheal epithelial cells exposed
to soluble salts of these metals at equivalent concentrations found in ROFA (Dye et al., 1999).
Treatment of V-exposed cells with buthionine sulfoximine further increased cytotoxicity.
Conversely, treatment with radical scavenger DMTU inhibited the effects in a dose-dependent
manner. These results suggest that soluble metal or combinations of several metals in ROFA
may be responsible for these effects.
Similar to combustion particles such as ROFA, the biological response to exposure to
ambient PM also may be influenced by the metal content of the particles. Human subjects were
instilled with 500 jig (in 20 mL sterile saline) of Utah Valley dust (UVD1, 2, 3, collected during
3 successive years) in the left segmental bronchus and on the right side with sterile saline as
control, followed by phagocytic cells being obtained from the segmental bronchi at 24 h
postinstillation. Alveolar macrophage from subjects instilled with UVD, obtained by BAL 24 h
postinstillation, were incubated with fluoresceinated yeast (Saccharomyces cerevisiae) to assess
their phagocytic ability. Although the same proportion of AMs that were exposed to UVD
phagocytized yeast, AMs exposed to UVD1, (which was collected while a local steel mill was
open), took up significantly less particles than AMs exposed to other extracts (UVD2 when the
steel mill was closed and UVD3 when the plant reopened). AMs exposed to UVD1 also
exhibited a small decrease in oxidant activity (using dihydrorhodamine-123, DHR). AMs from
healthy volunteers were incubated in vitro with the various UVD extracts to assess whether
effects on human AMs could be observed similar to those seen following in vivo exposure. The
percentage of AMs that engulfed yeast particles was significantly decreased by exposure to
UVD1 at 100 |ig/mL, but not at 25 |ig/mL. However, the amount of particles engulfed was the
same following exposure to all three UVD extracts. AMs also demonstrated increased oxidant
stress (using CL) after in vitro exposure to UVD1, and this effect was not abolished with
7-111
-------
pretreatment of the extract with the metal chelator DEF. As with the AMs exposed to UVD
in vivo, AM exposed to UVD in vitro had a decreased oxidant activity (DHR assay). UVD1
contains 61 times and 2 times the amount of Zn compared to UVD 2 and UVD3, respectively;
whereas UVD3 contained 5 times more Fe than UVD1. Ni and V were present only in trace
amounts. Using similarly extracted samples, Frampton et al. (1999) exposed BEAS-2B cells for
2 and 24 h. Similar results were observed for oxidant generation in these cells (i.e., UVD 2,
which contains the lowest concentrations of soluble Fe, Cu, and Zn, produced the least
response). Only UVD 3 produced cytotoxicity at a dose of 500 |ig/mL. UVD 1 and 3, but not 2,
induced expression of IL-6 and IL-8 in a dose-dependent fashion. Taken together, the above
results showed that the biological response to ambient particle extracts is heavily dependent on
the source and the chemical composition of PM.
7.4.4.2 Intracellular Signaling Mechanisms
In has been shown that the intracellular redox state of the cell modulates the activity of
several transcription factors, including NF-KB, a critical step in the induction of a variety of
proinflammatory cytokine and adhesion-molecule genes. NF-KB is a heterodimeric protein
complex that in most cells resides in an inactive state in the cell cytoplasm by binding to
inhibitory kappa B alpha (IicBa). On appropriate stimulation by cytokines or ROS, IicBa is
phosphorylated and subsequently degraded by proteolysis. The dissociation of IicBa from
NF-KB allows the latter to translocate into the nucleus and bind to appropriate sites in the DNA
to initiate transcription of various genes. Two in vitro studies (Quay et al., 1998; Kennedy et al.,
1998) have shown the involvement of NF-KB in particulate-induced cytokine and ICAM-1
production in BEAS-2B cells. Cytokine secretion was preceded by activation of NF-KB and was
reduced by treatment with antioxidants or metal chelators. These results suggest that metal-
induced oxidative stress may contribute to the initiation phase of the inflammatory cascade
following some PM exposures.
A second well-characterized human transcription factor, AP-1, also responds to the
intracellular ROS concentration. AP-1 exists in two forms, either in a homodimer of c-Jun
protein or a heterodimer consisting of c-Jun and c-Fos. Small amounts of AP-1 already exist in
the cytoplasm in an inactive form, mainly as phosphorylated c-Jun homodimer. Many different
oxidative stress-inducing stimuli, such as UV light and IL-1, can activate AP-1. Exposure of rat
7-112
-------
lung epithelial cells to ambient PM in vitro resulted in increases in c-Jun kinase activity, levels
of phosphorylated c-Jun immunoreactive protein, and transcriptional activation of AP-1-
dependent gene expression (Timblin et al., 1998). This study showed that interaction of ambient
particles with lung epithelial cells initiates a cell signaling cascade related to aberrant cell
proliferation.
Early response gene transactivation has been linked to the development of apoptosis,
a potential mechanism to account for PM-induced changes in cellular response. Apoptosis of
human AMs exposed to urban PM or ROFA (25 |ig/mL) was observed by Holian et al. (1998).
In addition, both urban PM and ROFA upregulated the expression of the RFD1+ (immune
stimulating macrophages) AM phenotype; whereas only ROFA decreased the RFDl+7+
(suppressor macrophages) phenotype. It has been suggested that an increase in the AM
phenotype ratio of RFDl+/RFDl+7+ may be related to disease progression in patients with
inflammatory diseases. These data showed that urban PM and ROFA can induce apoptosis of
human AMs and increase the ratio of AM phenotypes toward a higher immune active state and
may contribute to or exacerbate lung inflammation.
Inhaled fine and coarse particles are deposited in the epithelial lining of the nasal and
tracheal airways. Somatosensory neurons located in the dorsal root ganglia (DRG) innervate the
upper thoracic region of the airways and extend their terminals over and between the epithelial
lining of the lumen. Given this anatomical proximity, the sensory fibers and the tracheal
epithelial cells that they innervate encounter inhaled pollutants, such as PM, early during
inhalation. The differential responses of these cell types to PM derived from various sources
(i.e., industrial, residential, volcanic) were examined with biophysical and immunological
endpoints (Veronesi et al., 2002a). Although the majority of PM tested stimulated IL-6 release
in both BEAS-2B epithelial cells and DRG neurons in a receptor-mediated fashion, the degree of
these responses was markedly higher in the sensory neurons. Epithelial cells are damaged or
denuded in many common health disorders (e.g., asthma, viral infections), allowing PM particles
to directly encounter the sensory terminals and their acid-sensitive receptors.
Another intracellular signaling pathway that could lead to diverse cellular responses such
as cell growth, differentiation, proliferation, apoptosis, and stress responses to environmental
stimuli, is the phosphorylation-dependent, mitogen-activated protein kinase (MAPK).
Significant dose- and time-dependent increases in protein tyrosine phosphate levels have been
7-113
-------
seen in BEAS cells exposed to 100 |ig/mL ROFA for periods ranging from 5 min to 24 h (Samet
et al., 1997). In a subsequent study, the effects of As, Cr, Cu, Fe, Ni, V, and Zn on the MAPK,
extracellular receptor kinase (ERK), c-Jun N-terminal kinase (INK), and p38 in BEAS cells were
investigated (Samet et al., 1998). Arsenic, V, and Zn induced a rapid phosphorylation of MAPK
in BEAS cells. Activity assays confirmed marked activation of ERK, INK, and P38 in BEAS
cells exposed to As, V, and Zn; Cr and Cu exposure resulted in a relatively small activation of
MAPK; whereas Fe and Ni did not activate MAPK. Similarly, the transcription factors c-Jun
and ATF-2, substrates of INK and p38, respectively, were markedly phosphorylated in BEAS
cells treated with As, Cr, Cu, V, and Zn. The same acute exposure to As, V, or Zn that activated
MAPK was sufficient to induce a subsequent increase in IL-8 protein expression in BEAS cells.
All exposures were noncytotoxic based on measurement of LDH release and microscopic
examination of trypan blue or propidium iodide exclusion (Samet et al., 1996). These data
suggest that MAPK may mediate metal-induced expression of inflammatory proteins in human
bronchial epithelial cells. The ability of ROFA to induce activation of MAPKs in vivo was
demonstrated by Silbajoris et al. (2000; see Table 7-6). In addition, Gercken et al. (1996)
showed that the ROS production induced by PM was markedly decreased by the inhibition of
protein kinase C as well as phospholipase A2. Comparisons of in vitro and in vivo exposures of
ROFA to airway epithelial cells requires consideration of in vivo dosimetry and ambient
concentrations. Therefore, such extrapolations must be made with caution.
The major cellular response downstream of ROS and the cell signaling pathways described
above is the production of inflammatory cytokines or other reactive mediators. In an effort to
determine the contribution of cyclooxygenase to the pulmonary responses to ROFA exposure
in vivo, Samet et al. (2000) intratracheally instilled Sprague-Dawley rats with ROFA (200 or
500 jig in 0.5 mL saline). These animals were pretreated ip with 1 mg/kg NS398, a specific
prostaglandin H synthase 2 (COX2) inhibitor, 30 min prior to intratracheal exposure. At 12 h
after intratracheal instillations, ip injections (1 mL of NS398 in 20% ethanol in saline) were
repeated. ROFA treatment induced a marked increase in the level of PGE2 recovered in the BAL
fluid, which was effectively decreased by pretreating the animals with the COX2 inhibitor.
Immunohistochemical analyses of rat airway showed concomitant expression of COX2 in the
proximal airway epithelium of rats treated with soluble fraction of ROFA. This study further
showed that, although COX2 products participated in ROFA-induced lung inflammation, the
7-114
-------
COX metabolites are not involved in IL-6 expression nor the influx of PMN into the airway.
However, the rationale for the use of intraperitoneal challenge was not elaborated.
The production of cytokines and mediators also has been shown to depend on the type
of PM used in the experiments. A549 cells (a human airway epithelial cell line) were exposed
in vitro to several particulate materials: carbon black (CB, Elftex-12, Cabot Corp.), diesel soot
from two sources (ND from NIST, LD produced from General Motors LH 6.2 V8 engine at light
duty cycle), ROFA (from the heat exchange section of the Boston Edison), OAA (Ottawa
ambient air PM, EHC-93), SiO2, andM3S2 at 0.01, 0.03, 0.1, 0.3, 1.0, 3.0, 100, 300, 1000 |ig/cm2
for 18 h (Seagrave and Nikula, 2000). Endpoints included loss of adherence to tissue culture
substratum as evaluated by crystal violet staining, cell death measured by LDH release, mitotic
fraction and apoptosis, release of IL-8 measured by enzyme-linked immunosorbent assay, and
release of alkaline phosphatase measured by enzymatic activity using paranitrophenol phosphate.
Results indicated that (a) SiO2 and Ni3S2 caused dose dependent acute toxicity and apototic
changes; (b) ROFA and ND were acutely toxic only at the highest concentrations; (c) SiO2 (30,
100, 300 |ig/cm2) and Ni3S2 (10, 30, 100, 300 |ig/cm2) increased IL-8 (three and eight times over
the control, respectively) but suppressed IL-8 release at the highest concentration; (d) OAA and
ROFA also induced IL-8 but to a lesser degree; and (e) both diesel soots suppressed IL-8
production. The authors speculated that the suppression of IL-8 release may contribute to
increased respiratory disease as a result of decreased response to infectious agents. Silicon
dioxide and Ni3S2 increased the release of alkaline phosphatase, a marker of toxic responses,
only slightly. The less acutely toxic compounds caused significant release of alkaline
phosphatase. The order of potency in alkaline phosphatase production was OAA > LD =
ND > ROFA » SiO2 = Ni3S2. These results indicate that the type of particle used has a strong
influence on the biological response.
Dye et al. (1999) carried out reverse transcriptase-polymerase chain reactions on RNA
from rat tracheal epithelial cells to evaluate changes in steady-state gene expression of IL-6,
MIP-2, and inducible NOS (iNOS) in cells exposed for 6 h to ROFA (5 |ig/cm2) and Ni, V, or Ni
and V (water-soluble equivalent metal solution [pH 3.0]). Expression of MIP-2 and IL-6 genes
was significantly upregulated as early as 6 h post-ROFA-exposure in rat tracheal epithelial cells;
whereas gene expression of iNOS was maximally increased 24 h postexposure. Vanadium but
not Ni appeared to be mediating the effects of ROFA on gene expression. Treatment with
7-115
-------
DMTU (4 and 40 mm) inhibited both ROFA and V induced gene expression in a dose-dependent
manner.
It appears that many biological responses are produced by PM whether it is composed of a
single component or a complex mixture. The newly developed gene DNA microarrays monitors
the expressions of many mediator genes that regulate complex and coordinated cellular events
involved in tissue injury and repair. Using an array consisting of 27 rat genes representing
inflammatory and anti-inflammatory cytokines, growth factors, adhesion molecules, stress
proteins, metalloproteinases, vascular tone regulatory molecules, transcription factors, surfactant
proteins and antioxidant enzymes, Nadadur et al. (2000) measured pulmonary effects in rats 3
and 24 h following intratracheal instillation of ROFA (3.3 mg/kg), NiSO4 (1.3 |imol/kg), and
VSO4 (2.2 jimol/kg). Their data revealed a 2- to 3-fold increase in the expression of IL-6 and
TEVIP-1 at 24 h post-Ni exposure. The expression of cellular fibronectin (cFn-EIIIA) and iNOS
increased 24 h following ROFA exposure. Cellular fibronectin, interferon, iNOS and ICAM-1
was increased 24 h following Ni exposure and IL-6 was increased 24 h postexposure in
V-exposed animals. There was a modest increase in the expression of surfactant protein A
(SP-A) and p-actin genes. There was a 2-fold increase in the expression of IL-6 24 h following
exposure to ROFA, Ni, and V using the Northern blot analysis. A densitometric scan of an
autoradiograph of blots stripped and reprobed with SP-A cDNA insert indicated a minimal
increase in the expression of SP-A, both 3 and 24 h postexposure in all test groups. The findings
in this study suggest that DNA microarray may provide a tool for screening the expression
profile of tissue specific markers following exposure to PM. However, care should be taken in
reviewing such findings because of the variations in dose, instillation versus inhalation, and the
time-course for gene expression.
To investigate the interaction between respiratory cells and PM, Kobzik (1995) showed
that scavenger receptors are responsible for AM binding of unopsonized PM and that different
mechanisms mediate binding of carbonaceous dusts such as DPM. In addition, surfactant
components can increase AM phagocytosis of environmental particles in vitro, but only slightly
relative to the already avid AM uptake of unopsonized particles (Stringer and Kobzik, 1996).
Respiratory tract epithelial cells are also capable of binding with PM to secrete cytokine IL-8.
Using a cell line A549, Stringer et al. (1996) found that binding of particles to epithelial cells
was Ca-dependent for TiO2 and Fe2O3, while a-quartz binding was not Ca-dependent.
7-116
-------
In addition, as observed in AMs, PM binding by A549 cells also was mediated by scavenger
receptors, albeit those distinct from the heparin-insensitive acetylated-LDL receptor.
Furthermore, a-quartz, but not TiO2 or CAPs, caused a dose-dependent production of IL-8
(range 1 to 6 ng/mL), demonstrating a particle-specific spectrum of epithelial cell cytokine
(IL-8) response.
7.4.4.3 Particle Charge and Stimulation of Sensory Nerve Receptors
Colloidal particles carry an inherently negative surface charge (i.e., zeta potential) that
attracts protons from their vaporous milieu. These protons form a neutralizing, positive ionic
cloud around the individual particle (Hunter, 1981). Several early studies in the 1980s
demonstrated the influence of surface charge on toxicity of particulates. Work by Heck and
Costa (1983), for example, found crystalline MS, Ni3S2, and NiO, all particulates with strong
zeta potentials (i.e., electronegativity) to be more rapidly phagocytosed than amorphous
positively-charged MS. Additionally, they found that freshly suspended amorphous MS
particles were phagocytosed to a greater degree than particles aged for 1 to 7 days in water or
culture medium. These data suggested to the authors that a loss of negative charge during aging
is responsible for decreased phagocytosis, and correspondingly, decreased carcinogenicity. The
significance of particulate aging was also shown by Vallyathan et al. (1988), who compared the
effects of freshly-ground and aged silica in isolated rat AM. Freshly ground silica produced
greater respiratory burst, hydrogen peroxide release, superoxide anion secretion, and cytotoxicity
in AM than the aged dust. The authors suggested that this freshly fractured silica dust is
responsible for the pathogenesis of acute silicosis. A similar relationship between fresh and aged
particles was also observed with coal dust (Dalai et al., 1989). Both bituminous (72% carbon)
and anthracite (95% carbon) coal were ground and assayed for free radical concentration by
electron spin resonance spectroscopy. The freshly-ground anthracite coal had both greater
concentrations of free radicals and greater free radical activity than the bituminous coal. This
free radical activity correlated with toxicity of the coal dust. After several hours of being
ground, the coal dust lost significant free radical activity. Further, Dalai et al. examined lung
tissue from coal miners at autopsy and found free radicals, which they suggested may be
available for biological effects years after deposition.
7-117
-------
New insights with regard to the potential importance of particle charge have begun to
emerge in connection with the delineation of the important role of sensory nerve receptors in the
initiation of PM inflammation, as demonstrated in a series of newly available studies.
Neuropeptide and acid-sensitive sensory irritant (i.e., capsaicin, VR1) receptors were first
identified on BEAS-2B cells. To address whether PM could initiate airway inflammation
through these acid sensitive sensory receptors, BEAS-2B cells were exposed to ROFA from
Birmingham, AL. The BEAS-2B cells responded with an immediate increase in [Ca2+]; at
100 |ig/mL ROFA, followed by a concentration-dependent release of inflammatory cytokines
(i.e., IL-6, IL-8, TNF-a) and their transcripts at doses of 12 to 200 |ig/mL (significance levels
not given; Veronesi et al., 1999a). To test the relevance of neuropeptide or capsaicin VR1
receptors to these changes, BEAS-2B cells were pretreated with neuropeptide receptor
antagonists or capsazepine (CPZ), the antagonist for the capsaicin (i.e., VR1) receptor. The
neuropeptide receptor antagonists reduced ROFA-stimulated cytokine release by 25 to 50%.
However, pretreatment of cells with CPZ inhibited the immediate increases in [Ca2+];,
diminished transcript (i.e., IL-6, IL-8, TNF-a) levels and reduced IL-6 cytokine release to
control levels (Veronesi et al., 1999a). The above studies suggested that ROFA inflammation
was mediated by acid sensitive VR1 receptors located on the sensory nerve fibers that innervate
the airway and on epithelial target cells.
Since VR1 irritant receptors were found to respond to acidity (i.e., protonic charge),
experiments have been conducted to see if the surface charge carried by ROFA or other PM
particles could biologically activate cells and stimulate inflammatory cytokine release.
Oortgiesen et al. (2000) measured the mobility of ROFA particles in an electrically charged field
(i.e., micro-electrophoresis) microscopically and calculated their zeta potential. Next, synthetic
polymer microspheres (SPM) i.e., polymethacrylic acid nitrophenyl aery late microspheres were
prepared with attached carboxyl groups to yield SPM particles with a geometric diameter of
2 ± 0.1 and 6 ± 0.3 jim and with negative zeta potentials (-29 mV) similar to ROFA particles.
These SPM acted as ROFA surrogates with respect to their size and surface charge, but lacked
all contaminants thought to be responsible for its toxicity (e.g., transition metals, sulfates,
volatile organics and biologicals). Concentrations of SPM at 18.8 |ig/mL and ROFA particles
from Birmingham, AL at 50 |ig/mL were used to test BEAS-2B cells and mouse DRG sensory
neurons, both targets of inhaled PM. Equivalent degrees of biological activation (i.e., increase in
7-118
-------
intracellular calcium, [Ca2+];, IL-6 release) occurred in both cell types in response to either
ROFA or SPM, and both responses could be reduced by antagonists to VR1 receptors or acid-
sensitive pathways. Neutrally charged SPM (i.e., zeta potential of 0 mV), however, failed to
stimulate increases in [Ca2+];or IL-6 release (Oortgiesen et al., 2000).
To expand on these findings (Veronesi et al., 2002b), a larger set of PM was obtained from
urban (St. Louis, Ottawa), residential (wood stove), volcanic (Mt. St. Helens), and industrial (oil
fly ash, coal fly ash) sources. Each PM sample was described physicochemically (i.e., size and
number of particles, acidity, zeta potential) and used to test BEAS-2B epithelial cells at a
concentration of 10 |ig/mL. The resulting biological effect (i.e., increases in [Ca2+];, IL-6
release) was related to their physicochemical characteristics. When examined by linear
regression analysis, the only measured physicochemical property that correlated with increases
in [Ca2+]; and IL-6 release was the zeta potential of the visible particles (r2 > 0.97).
A recent study (Agopyan et al., 2003) has evaluated the hypothesis that the mechanism by
which negatively-charged PM acts on bronchial epithelial cells and sensory neurons is by
activation of capsaicin-sensitive vanilloid or acid-sensing receptors. They used BEAS-2B cells,
HEK293 cells expressing vanilloid receptors, and rat trigeminal ganglion neurons to which they
exposed negatively charged 2 and 10 jam polystyrene carboxylate-modified particles. They
found that the negatively-charged particles can activate capsazepine- and amiloride-sensitive
proton-gated receptors. This activation leads to increases in intracellular Ca2+, cyclic AMP, and
IL-6 release. Corroborating this study, Veronesi et al. (2003) conducted similar experiments
using positively- or negatively-charged synthetic polystyrene micells exposures to BEAS-2B
cells. They reported increases in intracellular Ca2+ and IL-6, which they attributed to the
negative charge on the particles. This negative charge, they hypothesize, acquires a cloud of
protons which then activates the proton-sensitive vanilloid and acid-sensitive receptors.
Thus, both older and newer studies have provided evidence that both the charge and the
age of the PM are important factors in its toxicity. Also, several newer in vitro studies
demonstrate a plausible neurogenic basis for PM inflammation by which the positively-charged
proton cloud associated with negatively-charged colloidal PM particles can activate acid-
sensitive VR1 receptors found on human airway epithelial cells and sensory terminals. This
activation is thought to result in an immediate influx of calcium and release of inflammatory
7-119
-------
cytokines and neuropeptides, which then initiate and sustain inflammatory events in the airways
via neurogenic inflammation (Veronesi and Oortgiesen, 2001).
7.4.4.4 Other Potential Cellular and Molecular Mechanisms
A potential mechanism involved in the alteration of surface tension may be related to
changes in the expression of matrix metalloproteinases (MMPs), such as pulmonary matrilysin
and gelatinase A and B, and tissue inhibitor of metalloproteinase (TIMP; Su et al., 2000a,b).
Sprague-Dawley rats exposed to ROFA by intratracheal injection (2.5 mg/rat) had increased
mRNA levels of matrilysin, gelatinase A, and TIMP-1. Gelatinase B, not expressed in control
animals, was increased significantly from 6 to 24 h following ROFA exposure. Alveolar
macrophages, epithelial cells, and inflammatory cells were major cellular sources for the
pulmonary MMP expression. The expression of Gelatinase B in rats exposed to the same dose
of ambient PM (< 1.7 jim and 1.7 to 3.7 jim) collected from Washington, DC, was significantly
increased as compared to saline control; whereas the expression of TIMP-2 was suppressed.
Ambient PM between 3.7 and 20 jim also increased the Gelatinase B expression. Increases in
MMPs, which degrade most of the extracellular matrix, suggest that ROFA and ambient PM can
similarly increase the total pool of proteolytic activity to the lung and contribute in the
pathogenesis of PM-induced lung injury. Since no control particles were used in this study, the
results must be interpreted with caution because it is possible that any particle administered in
high doses could have a similar effect.
7.4.5 Specific Particle Size and Surface Area Effects
Most particles used in laboratory animal toxicology studies are greater than 0.1 jim in size.
However, the enormous number and huge surface area of ultrafme particles highlight the likely
importance of considering the size of the particle in assessing response. Ultrafme particles with
a diameter of 20 nm, when inhaled at the same mass concentration, have a number concentration
that is approximately 6 orders of magnitude higher than for a 2.5-|im diameter particle; particle
surface area is also greatly increased (Table 7-12).
Many studies assessed in 1996 PM AQCD, as well as here, suggest that the surface of
particles or substances released from the surface (e.g., transition metals, organics) interact with
biological substrates, and that surface-associated free radicals or free radical-generating systems
7-120
-------
TABLE 7-12. NUMBERS AND SURFACE AREAS OF MONODISPERSE
PARTICLES OF UNIT DENSITY OF DIFFERENT SIZES AT A MASS
CONCENTRATION OF 10 ug/m3
Particle Diameter
(um)
0.02
0.1
0.5
1
2.5
Particle Number
(per cm3 air)
2400000
19100
153
19
1.2
Particle Surface Area
(um2 per cm3 air)
3016
600
120
60
24
Source: Oberdorster (1996a).
may be responsible for toxicity. Thus, if ultrafine particles were to cause toxicity by a transition
metal-mediated mechanism, for example, then the relatively large surface area for a given mass
of ultrafine particles would imply high concentrations of transition metals being available to
cause oxidative stress to cells.
Two groups have examined toxicity differences between fine and ultrafine particles, with
the general finding that ultrafine particles show a significantly greater response at similar mass
doses (Oberdorster et al., 1992; Li et al., 1996, 1997, 1999). However, only a few studies have
investigated the ability of ultrafine particles to generate a greater oxidative stress when compared
to fine particles of the same material. Osier and Oberdorster (1997) compared the response of
rats (F344) exposed by intratracheal inhalation to "fine" (-250 nm) and "ultrafine" (-21 nm)
TiO2 particles and found the ultra fine particles to be more effective in producing lung
inflammatory responses as indexed by several markers.
Consistent with these in vivo studies, Finkelstein et al. (1997) has shown that exposing
primary cultures of rat Type II cells to 10 |ig/mL ultrafine TiO2 (20 nm) causes increased TNF
and IL-1 release throughout the entire 48-h incubation period. In contrast, fine TiO2 (200 nm)
had no effect. In addition, ultrafine polystyrene carboxylate-modified microspheres (UFP,
fluorospheres, Molecular Probes, 44 ± 5 nm) have been shown to induce a significant
enhancement of both substance P(SP) and histamine release after administration of capsaicin
7-121
-------
(1CT4 M), to stimulate C-fiber, and carbachol (1CT4 M), a cholinergic agonist in rabbit
intratracheally instilled with UFP (Nemmar et al., 1999). A significant increase in histamine
release also was recorded in the UFP-instilled group following the administration of both SP
(1(T6 M) plus thiorpan (10~5 M) and compound 48/80 (C48/80, 1(T3 M) to stimulate mast cells.
Bronchoalveolar lavage analysis showed an influx of PMN, an increase in total protein
concentration, and an increase in lung wet weight/dry weight ratio. Electron microscopy showed
that both epithelial and endothelial injuries were observed. The pretreatment of rabbits in vivo
with a mixture of either SR 140333 and SR 48368, a tachykinin NKj and NK2 receptor
antagonist, or a mixture of terfenadine and cimetidine, a histamine Hx and H2 receptor
antagonist, prevented UFP-induced PMN influx and increased protein and lung WW/DW ratio.
It is believed that ultrafme particles cause greater cellular injury because of the relatively
large surface area for a given mass. In addition, the fate of ultrafmes after deposition is also
different in that they interact more rapidly with epithelial target cells rather than to be
phagocytized by AMs. However, in a study that compared the response to CB particles of two
different sizes, Li et al. (1999) demonstrated that in the instillation model, a localized dose of
particle over a certain level causes the particle mass to dominate the response, rather than the
surface area. Ultrafme CB (ufCB, Printex 90), 14 nm in diameter, and fine CB (CB, Huber 990),
260 nm in diameter, were instilled intratracheally in rats, and BAL profile at 6 h was assessed.
At mass of 125 jig or below, ufCB generated a greater response (increase LDH, epithelial
permeability, decrease in GSH, TNF, and NO production) than fine CB at various times
postexposure. However, higher doses of CB caused more PMN influx than the ufCB. In
contrast to the effect of CB, which showed dose-related increasing inflammatory response, ufCB
at the highest dose caused less of a neutrophil influx than at the lower dose, confirming earlier
work by Oberdorster et al. (1992). Moreover, when the PMN influx was expressed as a function
of surface area, CB produced greater response than ufCB at all doses used in this study.
Although particle interstitialization with a consequent change in the chemotatic gradient for
PMN was offered as an explanation, these results need further scrutiny. Moreover, these
findings imply that mass is relatively less important than surface area and that the latter metric
may be more useful for assessing PM toxicity. However, it is unclear if this finding is restricted
to the particular endpoints addressed and/or CB, the PM compound studied.
7-122
-------
Oberdorster et al. (2000) reported on a series of studies in rats and mice using ultrafine
particles of various chemical composition. In rats sensitized with endotoxin (70 EU) and
exposed to ozone (1 ppm) plus ultrafine carbon particles (-100 |ig/m3), they found a 9-fold
greater release of ROS in old rats (20 months) than in similarly treated young rats (10 weeks).
Exposure to ultrafine PM alone in sensitized old rats also caused an inflammatory response.
Although the potential mechanisms of ultrafme-induced lung injury remain unclear, it is
likely that ultrafine particles, because of their small size, are not effectively phagocytized by
AMs and can easily penetrate the airway epithelium, gaining access to the interstitium. Using
electron microscopy, Churg et al. (1998) examined particle uptake in rat tracheal explants.
Explants were exposed to either fine (0.12 |im) or ultrafine (0.021 jim) TiO2 particles and
examined after 3 or 7 days. They found both size particles in the epithelium at both time points;
but, in the subepithelial tissues, only at day 7. The volume proportion (the volume of TiO2 over
the entire volume of epithelium or subepithelium area) of both fine and ultrafine particles in the
epithelium increased from 3 to 7 days. It was greater for ultrafine at 3 days but was greater for
fine at 7 days. The volume proportion of particles in the subepithelium at day 7 was equal for
both particles, but the ratio of epithelial to subepithelial volume proportion was 2:1 for fine and
1:1 for ultrafine. Ultrafine particles persisted in the tissue as relatively large aggregates; whereas
the size of fine particle aggregates became smaller over time. Ultrafine particles appeared to
enter the epithelium faster and, once in the epithelium, a greater proportion of them were
translocated to the subepithelial space compared to fine particles. However, the authors assumed
that the volume proportion is representative of particle number and the number of particles
reaching the interstitial space is directly proportional to the number applied (i.e., there is no
preferential transport from lumen to interstitium by size). These data are in contrast to the
results of instillation or inhalation of fine and ultrafine TiO2 particles reported earlier (Ferin
et al., 1990, 1992). However, the explant and intratracheal instillation test systems differ in
many aspects, making direct comparisons difficult. Limitations of the explant test system
include traumatizing the explanted tissue, introducing potential artifacts through the use of liquid
suspension for exposure, the absence of inflammatory cells, and possible overloading of the
explants with dust.
7-123
-------
Two studies examined the influence of specific surface area on biological activity (Lison
et al., 1997; Oettinger et al., 1999). The biological responses to various MnO2 dusts with
different specific surface area (0.16, 0.5, 17, and 62 m2/g) were compared in vitro and in vivo
(Lison et al., 1997). In both systems, the results show that the amplitude of the response is
dependent on the total surface area that is in contact with the biological system, indicating that
surface chemistry phenomena are involved in the biological reactivity. Freshly ground particles
with a specific surface area of 5 m2/g also were examined in vitro. These particles exhibited an
enhanced cytotoxic activity that was almost equivalent to that of particles with a specific surface
area of 62 m2/g, indicating that undefined reactive sites produced at the particle surface by
mechanical cleavage also may contribute to the toxicity of insoluble particles.
In another study (Oettinger et al., 1999), two types of CB particles were used: (1) Printex
90 or P90 (formed by controlled combustion and consisting of defined granules with specific
surface area of 300 m2/g and particle size of 14 nm) is predominantly loaded with metallic
components (< 100 ppm Fe; < 50 ppm Pb; < 10 ppm Se; < 10 ppm As; < 10 ppm Zn); and
(2) soot FR 101 (with specific surface area of 20 m2/g, particle size of < 95 nm) has the ability to
adsorb poly cyclic and other carbons. Exposure of AMs to 100 |ig/106 cells of FR 101 and P90
resulted in a 1.4- and 2.1-fold increase in ROS release, respectively. These exposures also
caused a 4-fold up-regulation of NF-KB gene expression. This suggests that carbon particles
with larger surface area produce greater biological response than carbon particles with smaller
surface area. Another study by Schliiter et al. (1995), showed that by exposing bovine AMs to
metal oxide coated 5 jim silica particles, most of the metal coatings (As, Ce, Fe, Mn, Ni, Pb, and
V) had no effect on ROS production by these cells. However, coating with CuO markedly
lowered the O2" and H2O2, whereas V(IV) increased both reactive oxygen intermediates (ROI).
This study suggests that, in addition to specific surface area, chemical composition of the
particle surface also influences its cellular response.
Thus, ultrafme particles apparently have the potential to significantly contribute to the
adverse effects of PM. These studies, however, have not considered the portion of ambient
ultrafme particles not solid in form. Droplets (e.g., sulfuric acid droplets) and organic-based
ultrafme particles exist in the ambient environment; and they can spread, disperse, or dissolve
after contact with liquid surface layers and may thereby contribute further to PM-related effects.
7-124
-------
7.5 FACTORS AFFECTING SUSCEPTIBILITY TO PARTICULATE
MATTER EXPOSURE EFFECTS
Susceptibility of an individual to adverse health effects of PM can vary depending on a
variety of host factors such as age, physiological activity profile, genetic predisposition, or
preexistent disease. The potential for preexistent disease to alter pathophysiological responses to
toxicant exposure is widely acknowledged but poorly understood due, in part, to the statistical
limitations of toxicological studies noted earlier. Epidemiologic studies have demonstrated that
the effects of PM exposure tend to be more evident in populations with preexisting disease; and
it is logical that important mechanistic differences may exist among these populations.
However, because of inherent variability (necessitating large numbers of subjects) and ethical
concerns associated with using diseased subjects in clinical research studies, a solid database on
human susceptibilities is lacking. For more control over both environmental and host variables,
animal models are often used. Many laboratory studies have shown alterations in a variety of
endpoints in experimental animals following exposure to laboratory-generated particles. These
findings (e.g., increased pulmonary inflammation, increased airway resistance, and decrements
in pulmonary host defenses) may be of limited value because of inherent differences between the
laboratory-generated particles and actual ambient air particle mixes. Thus, care must be taken in
extrapolation from animal models of human disease to humans. Rodent models of human
disease, their use in toxicology, and the criteria for judging their appropriateness as well as their
limitations must be considered (Kodavanti et al., 1998b; Kodavanti and Costa, 1999; Costa,
2000; Conn et al., 2000; Bice et al., 2000; Mauderly et al., 2000; Muggenburg et al., 2000b).
7.5.1 Pulmonary Effects of Particulate Matter in Compromised Hosts
Epidemiologic studies suggest that there may be subsegments of human populations that
are especially susceptible to effects from inhaled particles (see Chapter 8). The elderly with
chronic cardiopulmonary disease, those with pneumonia and possibly other lung infections, and
those with asthma (at any age) appear to be at higher risk than healthy people of similar age.
Apropos to this, although many of the newly available toxicology studies used healthy adult
animals, a growing number of other newer studies examined effects of ambient or surrogate
particles in compromised host models. For example, Costa and Dreher (1997) used a rat model
of cardiopulmonary disease to explore the question of susceptibility and possible mechanisms by
7-125
-------
which PM effects are potentiated. Rats with advanced monocrotaline (MCT)-induced
pulmonary vasculitis/hypertension were given intratracheal instillations of ROFA (0, 0.25,
1.0, and 2.5 mg/rat). The MCT animals had a marked neutrophilic inflammation. In the context
of this inflammation, ROFA induced a 4- to 5-fold increase in BAL PMNs. There was also a
ROFA dose-dependent increased mortality at 96 h postexposure.
As discussed previously, Kodavanti et al. (1999) also studied PM effects in the MCT rat
model of pulmonary disease. Rats treated with 60 mg/kg MCT were exposed to 0, 0.83 or
3.3 mg/kg ROFA by intratracheal instillation and to 15 mg/m3 ROFA by inhalation. Both
methods of exposure caused inflammatory lung responses; and ROFA exacerbated the lung
lesions, as shown by increased lung edema, inflammatory cells, and alveolar thickening.
The manner in which MCT can alter the response of rats to inhaled particles was examined
by Madl and colleagues (1998). Rats were exposed to fluorescent colored microspheres (1 |im)
2 weeks after treatment with MCT. In vivo phagocytosis of the microspheres was altered in the
MCT rats in comparison with control animals. Fewer microspheres were phagocytized in vivo
by AMs, and there was a concomitant increase in free microspheres overlaying the epithelium at
airway bifurcations. The decrease in in vivo phagocytosis was not accompanied by a similar
decrease in vitro. Macrophage chemotaxis, however, was impaired significantly in MCT rats
compared with control rats. Thus, MCT appeared to impair particle clearance from the lungs via
inhibition of macrophage chemotaxis.
Respiratory infections are common in all individuals. The infections are generally cleared
quickly, depending on the virulence of the organism; however, in individuals with immunologic
impairment or lung diseases such a COPD, the residence time in the lung is extended. A variety
of viral and bacterial agents have been used to develop infection models in animals. Viral
infection models primarily use mice and rats. The models focus on the proliferation and
clearance of the microorganisms and the associated pulmonary effect. The models range from
highly virulent and lethal (influenza A/Hong Kong/8/68, H3N2) to nonlethal (rat-adapted
influenza virus model [RAIV]). The lethal model terminates in extensive pneumonia and lung
consolidation. Less virulent models (A/Port Chalmers/1/73 and H3N2) exhibit airway epithelial
damage and immune responses. The nonlethal model exhibits airway reactivity that subsides,
with recovery being complete in about 2 weeks (Kodavanti et al., 1998b). Bacterial infection
models mimic the chronic bacterial infections experienced by humans with other underlying
7-126
-------
disease conditions. The animal models develop signs similar to those in humans but to a milder
degree. To mimic the chronic infections, the bacteria are encased in agar beads to prevent rapid
clearance. Generally, the models involve preexposure to the irritant followed by the bacterial
challenge. More recently, bacterial infection models have involved pre-exposure by the bacteria
followed by exposure to the irritant (Kodavanti et al., 1998b).
Elder et al. (2000a,b) exposed 8-week to 22-months old Fischer 344 rats and 14- to 17-
months old Tsk mice to 100 |ig/m3 of ultrafine carbon (UF) and/or 1.0 ppm O3 for 6 h after a
12-min exposure to a low dose (70 EU) of endotoxin (lipopolysaccharide, LPS). The ultrafine
carbon had a small effect on lung inflammation and inflammatory cell activation. The effects
were enhanced in the compromised lung and in older animals. The greatest effect was in the
compromised lung exposed to both ultrafine carbon and ozone.
Chronic bronchitis is the most prevalent of the COPD-related illnesses. In humans, chronic
bronchitis is characterized by pathologic airway inflammation and epithelial damage, mucus cell
hyperplasia and hypersecretion, airway obstruction and in advance cases, airway fibrosis. The
most widely used animal models of bronchitis (rat and dog) are those produced by subchronic
exposure to high sulfur dioxide (SO2) concentrations (150 to 600 ppm) for 4 to 6 weeks.
Exposure to SO2 produces changes in the airways similar to those of chronic bronchitis in
humans. There is an anatomical difference between the rat and the human in the absence of
submucosal glands in the rat. However, like humans, rats exhibit increased airway
responsiveness to inhaled bronchoconstricting agonists. Sulfur dioxide-induced lesions include
increased numbers of epithelial mucus-producing cell, loss of cilia, airway inflammation,
increased pro-inflammatory cytokine expression, and thickening of the airway epithelium.
When the cause of the chronic bronchitis is removed, the pathology slowly reverses. The time
course and the extent of reversal differs between the human and rodent. Consequently, care
should be exercised when applying this model (Kodavanti et al., 1998b).
The SO2-induced model of chronic bronchitis has been used to examine the potential
interaction of PM with preexisting lung injury. Clarke and colleagues pretreated SD rats for
6 weeks with air or 170 ppm SO2 for 5 h/day and 5 days/week (Clarke et al., 1999; Saldiva et al.,
2002). Exposure to CAPs for 5 h/day for 3 days to PM concentrations ranging from 73.5 to
733 |ig/m3 produced significant changes in both cellular and biochemical markers in lavage
fluid. In comparison to control animal values, protein was increased ~3-fold in SO2-pretreated
7-127
-------
animals exposed to concentrated ambient PM. Lavage fluid neutrophils and lymphocytes were
increased significantly in both groups of rats exposed to concentrated ambient PM, with greater
increases in both cell types in the SO2-pretreated rats. Thus, exposure to concentrated ambient
PM produced adverse changes in the respiratory system, but no deaths, in both normal rats and
in a rat model of chronic bronchitis.
Clarke et al. (2000b) next examined the effect of CAPs from Boston, MA, in normal rats of
different ages. Unlike the earlier study that used Sprague-Dawley rats, 4- and 20-mo-old Fischer
344 rats were examined after exposure to concentrated ambient PM for 5 h/day for 3 consecutive
days. They found that exposure to the daily mean concentrations of 80, 170, and 50 |ig/m3 PM,
respectively, produced statistically significant increases in total neutrophil counts (over 10-fold)
in lavage fluid of the young, but not the old, rats. Thus, repeated exposure to relatively low
concentrations of ambient PM produced an inflammatory response, although the actual percent
neutrophils in the concentrated ambient PM-exposed young adult rats was low (-3%). On the
other hand, Gordon et al. (2000) found no evidence of neutrophil influx in the lungs of normal
and MCT-treated Fischer 344 rats exposed in nine separate experiments to concentrated ambient
PM from New York, NY at concentrations as high as 400 |ig/m3 for a 6-h exposure or 192 |ig/m3
for three daily 6-h exposures. Similarly, normal and cardiomyopathic hamsters showed no
evidence of pulmonary inflammation or injury after a single exposure to the same levels of
concentrated ambient PM. Gordon and colleagues did report a statistically significant doubling
in protein concentration in lavage fluid in MCT-treated rats exposed for 6 h to 400 |ig/m3 New
York City CAPs.
Kodavanti and colleagues (1998b) also have examined the effect of CAPs in normal rats
and rats with SO2 chronic bronchitis. Among the four separate exposures to PM, there was a
significant increase in lavage fluid protein in bronchitic rats from only one exposure protocol in
which the rats were exposed to 444 and 843 |ig/m3 PM on 2 consecutive days (6 h/day).
Neutrophil counts were increased in bronchitic rats exposed to concentrated ambient PM in three
of the four exposure protocols, but was decreased in the fourth protocol. No other changes in
normal or bronchitic rats were observed, even in the exposure protocols with higher PM
concentrations. Thus, rodent studies have demonstrated that inflammatory changes can be
produced in normal and compromised animals exposed to CAPs. These findings are important
7-128
-------
because only a limited number of studies have used real-time inhalation exposures to actual
ambient urban PM.
Pulmonary function measurements are often less invasive than other means to assess the
effects of inhaled air pollutants on the mammalian lung. After publication of the 1996 PM
AQCD, a number of investigators examined the response of rodents and dogs to inhaled ambient
particles. In general, these investigators have demonstrated that ambient PM has minimal effects
on pulmonary function. Gordon et al. (2000) exposed normal and MCT-treated rats to filtered
air or 181 |ig/m3 concentrated ambient PM for 3 h. For both normal and MCT-treated rats,
no differences in lung volumes or CO diffusing capacities monoxide were observed between the
air- or PM-exposed animals at 3 or 24 h after exposure. Similarly, in cardiomyopathic hamsters,
concentrated ambient PM had no effect on these same pulmonary function measurements.
Other pulmonary function endpoints have been studied in animals exposed to CAPs.
Clarke et al. (1999) observed that tidal volume was increased slightly in both control rats and
rats with SO2-induced chronic bronchitis exposed to 206 to 733 |ig/m3 PM on 3 consecutive
days. No changes in peak expiratory flow, respiratory frequency, or minute volume were
observed after exposure to CAPs. In the series of dog studies by Godleski et al. (2000; also see
Section 7.3), no signficant changes in pulmonary function were observed in normal mongrel
dogs exposed to CAPs, although a 20% decrease in respiratory frequency was observed in dogs
that underwent coronary artery occlusion and were exposed to PM. Thus, studies using normal
and compromised animal models exposed to CAPs have found minimal biological effects of
ambient PM on pulmonary function.
Kodavanti et al. (2000b; 2001) used genetically predisposed spontaneously hypertensive
(SH) rats as a model of cardiovascular disease to study PM-related susceptibility. The SH rats
were more susceptible to acute pulmonary injury from intratracheal ROFA exposure than
normotensive control Wistar Kyoto (WKY) rats (Kodavanti et al., 2001). The primary metal
constituents of ROFA, V and Ni, caused differential species-specific effects. Vanadium, which
was less toxic than Ni in both strains, caused inflammatory responses only in WKY rats, but Ni
was injurious to both WKY and SH rats (SH > WKY). This differential responsiveness of V and
Ni was correlated with their specificity for airway and parenchymal injury, discussed in another
study (Kodavanti et al., 1998b). When exposed to the same ROFA by inhalation (15 mg/m3,
6 h/day, 3 days), protein levels in BAL of both the WKY and SH rats increased significantly, but
7-129
-------
the increase in SH rats was greater than that of the WKY rats (Kodavanti et al., 2000b). The SH
rats exhibited a hemorrhagic response to ROFA. Oxidative stress was much higher in ROFA-
exposed SH rats than matching WKY rats. Also, SH rats, unlike WKY rats, showed a
compromised ability to increase BAL glutathione in response to ROFA, suggesting a potential
link to increased susceptibility. However, LDH and NAG activities were higher in WKY rats.
Lactate dehydrogenase was slightly higher in SH rats instilled with ROFA (Kodavanti et al.,
2001). Cardiovascular effects were characterized by ST-segment area depression of the ECG in
ROFA-exposed SH but not WKY rats. When the same rats were exposed to ROFA by inhalation
to 15 mg/m3, 6 h/day, 3 days/week for 1,2, or 4 weeks compared to intratracheal exposure to 0,
1.0, 5.0 mg/kg in saline (Kodavanti et al., 2002a), differences in effects were dependent on the
length of exposure. After acute exposure, increased plasma fibrinogen was associated with lung
injury; longer-term, episodic ROFA exposure resulted in progressive protein leakage and
inflammation that was significantly worse in SH rats versus WKY rats. These studies
demonstrate the potential utility of cardiovascular disease models for the study of PM health
effects and show that genetic predisposition to oxidative stress and cardiovascular disease may
play a role in increased sensitivity to PM-related cardiopulmonary injury.
On the basis of in vitro studies, Sun et al. (2001) predicted that the antioxidant and lipid
levels in the lung lining fluid may determine susceptibility to inhaled PM. In a subsequent study
from the same laboratory, Norwood et al. (2001) conducted inhalation studies on guinea pigs to
test this hypothesis. On the basis of dietary supplementation or depletion of ascorbic acid and
glutathione (GSH) the guinea pigs were divided into four groups: (+C + GSH), (+C - GSH),
(-C + GSH), and (-C - GSH). All groups were exposed (nose-only) for 2 h to clean air or
ROFA (< 2.5 |im) at 19-25 mg/m3. Nasal lavage and BAL fluid and cells were examined at 0 h
and 24 h postexposure. Exposure to ROFA increased lung injury in the (-C-GSH) group only (as
shown by increased BAL fluid protein, LDH, and PMNs and decreased BAL macrophages) and
resulted in lower antioxidant concentrations in BAL fluid than were found with single
deficiencies.
In summary, although more of these studies are just beginning to emerge and are only now
being replicated or followed more thoroughly to investigate underlying mechanisms, they do
provide evidence suggestive of enhanced susceptibility to inhaled PM in "compromised" hosts.
7-130
-------
7.5.2 Genetic Susceptibility to Inhaled Particles and Their Constituents
A key issue in understanding adverse health effects of inhaled ambient PM is identification
of which classes of individuals are susceptible to PM. Although factors such as age and health
status have been studied in both epidemiology and toxicology studies, some investigators have
begun to examine the importance of genetic susceptibility in the response to inhaled particles
because of evidence that genetic factors play a role in the response to inhaled pollutant gases.
To accomplish this goal, investigators typically have studied the interstrain response to particles
in rodents. The response to ROFA instillation in different strains of rats has been investigated by
Kodavanti et al. (1996, 1997a). In the first study, male SD and F-344 rats were instilled
intratracheally with saline or ROFA particles (8.3 mg/kg). ROFA instillation produced an
increase in lavage fluid neutrophils in both SD and F-344 rats; whereas a time-dependent
increase in eosinophils occurred only in SD rats. In the subsequent study (Kodavanti et al.,
1997a), SD, Wistar (WIS), and F-344 rats (60 days old) were exposed to saline or ROFA
(8.3 mg/kg) by intratracheal instillation and examined for up to 12 weeks. Histology indicated
focal areas of lung damage showing inflammatory cell infiltration as well as alveolar, airway,
and interstitial thickening in all three rat strains during the week following exposure. Trichrome
staining for fibrotic changes indicated a sporadic incidence of focal alveolar fibrosis at 1, 3, and
12 weeks in SD rats; whereas WIS and F-344 rats showed only a modest increase in trichrome
staining in the septal areas. One of the isoforms of fibronectin mRNA was upregulated in
ROFA-exposed SD and WIS rats, but not in F-344 rats. Thus, in rats there appears to be a
genetic based difference in susceptibility to lung injury induced by instilled ROFA.
Differences in the degree of pulmonary inflammation have been described in rodent strains
exposed to airborne pollutants. To understand the underlying causes, signs of airway
inflammation (i.e., airway hyper-responsiveness, inflammatory cell influx) were established in
responsive (BABL/c) and non-responsive (C57BL/6) mouse strains exposed to ROFA (Veronesi
et al., 2000). Neurons taken from the ganglia (i.e., DRG) that innervate the nasal and upper
airways were cultured from each mouse strain and exposed to 25 or 50 |ig/mL ROFA for 4 h.
The difference in inflammatory response noted in these mouse strains in vivo was retained in
culture, with C57BL/6 neurons showing significantly lower signs of biological activation (i.e.,
increased intracellular calcium levels) and cytokine (i.e., IL-6, IL-8) release relative to BALB/c
mice. RT-PCR and immunocytochemistry indicated that the BALB/c mouse strain had a
7-131
-------
significantly higher number of neuropeptide and acid-sensitive (i.e., NK1, VR1) sensory
receptors on their sensory ganglia relative to the C57BL/6 mice. Such data indicate that
genetically-determined differences in sensory inflammatory receptors can influence the degree
of PM-induced airway inflammation.
Kleeberger and colleagues have examined the role that genetic susceptibility plays in the
effect of inhaled acid-coated particles on macrophage function. Nine inbred strains of mice were
exposed nose-only to very high doses of carbon particles coated with acid (10 mg/m3 carbon
with 285 |ig/m3 sulfate) for 4 h (Ohtsuka et al., 2000a). Significant inter-strain differences in
Fc-receptor-mediated macrophage phagocytosis were seen with C57BL/6J mice being the most
sensitive. Although neutrophil counts were increased more in C3H/HeOuJ and C3H/HeJ strains
of mice than in the other strains, the overall magnitude of change was small and not correlated
with the changes in macrophage phagocytosis. In follow-up studies using the same type particle,
Ohtsuka et al. (2000a,b) performed a genome-wide scan with an intercross cohort derived from
C57BL/6J and C3H/HeJ mice. Analyses of phenotypes of segregant and nonsegregant
populations derived from these two strains indicate that two unlinked genes control
susceptibility. They identified a 3-centiMorgan segment on mouse chromosome 17 which
contains an acid-coated particle susceptibility locus. Interestingly, this quantitative trait locus
(a) overlaps with those described for ozone-induced inflammation (Kleeberger et al., 1997) and
acute lung injury (Prows et al., 1997) and (b) contains several promising candidate genes that
may be responsible for the observed genetic susceptibility for macrophage dysfunction in mice
exposed to acid-coated particles.
Leikauf and colleagues (Leikauf et al., 2000; Wesselkamper et al., 2000; McDowell et al.,
2000; Prows and Leikauf, 2001; Leikauf et al., 2001) have identified a genetic susceptibility in
mice that is associated with mortality following exposures to high concentrations (from 15 to
150 |ig/m3) of a "NiSO4" aerosol (0.22 |im MMAD) for up to 96 h. These studies also have
preliminarily identified the chromosomal locations of a few genes that may be responsible for
this genetic susceptibility. Though death is a somewhat crude endpoint in genetic studies, this
finding is significant in light of the toxicology studies demonstrating that bioavailable, first-row
transition metals participate in acute lung injury following exposure to emission and ambient air
particles. Similar genes may be involved in human responses to particle-associated metals; but
7-132
-------
additional studies are needed to determine whether the identified metal susceptibility genes are
involved in human responses to ambient levels of particulate-associated metals.
One study has examined the interstrain susceptibility to ambient particles. C57BL/6J and
C3H/HeJ mice were exposed to 250 |ig/m3 concentrated ambient PM2 5 for 6 h and examined at
0 and 24 h after exposure for changes in lavage fluid parameters and cytokine mRNA expression
in lung tissue (Shukla et al., 2000). No interstrain differences in response were observed.
Surprisingly, although no indices of pulmonary inflammation or injury were increased over
control values in the lavage fluid, increases in cytokine mRNA expression were observed in both
murine strains exposed to PM2 5. Although the increase in cytokine mRNA expression was
generally small (approximately 2-fold), the effects on IL-6, TNF-a, TGF-P2, and y-interferon
were consistent.
Thus, a few studies have begun to demonstrate that genetic susceptibility can play a role in
the response to inhaled particles. However, the doses of PM administered in these studies,
whether by inhalation or instillation, were extremely high in comparison to ambient PM levels.
Similar strain differences in response to inhaled metal particles have been observed by other
investigators (McKenna et al., 1998; Wesselkamper et al., 2000), although the concentration of
metals used in these studies were also more relevant to occupational rather than environmental
exposure levels. The extent to which genetic susceptibility plays as significant a role in the
adverse effects of ambient PM as does age or health status remains to be determined.
7.5.3 Participate Matter Effects on Allergic Hosts
Relatively little is known about the effects of inhaled particles on humoral (antibody) or
cell-mediated immunity. Alterations in the response to a specific antigenic challenge have been
observed in animal models at high concentrations of acid sulfate aerosols (above 1000 |ig/m3;
Pinto et al., 1979; Kitabatake et al., 1979; Fujimaki et al., 1992). Several studies have reported
an enhanced response to nonspecific bronchoprovocation agents, such as acetylcholine and
histamine, after exposure to inhaled particles. This nonspecific airway hyperresponsiveness,
a central feature of asthma, occurs in animals and human subjects exposed to sulfuric acid under
controlled conditions (Utell et al., 1983; Gearhart and Schlesinger, 1986). Although, its
relevance to specific allergic responses in the airways of atopic individuals is unclear, it
demonstrates that the airways of asthmatics may become sensitized to either specific or
7-133
-------
nonspecific triggers that could result in increases in asthma severity and asthma-related hospital
admissions (Peters et al., 1997; Jacobs et al., 1997; Lipsett et al., 1997). Combustion particles
also may serve as carrier particles for allergens (Knox et al., 1997).
A number of in vivo and in vitro studies have demonstrated that particles (PM) can alter
the immune response to challenge with specific antigens and suggest that PM may act as an
adjuvant. For example, studies have shown that treatment with diesel particulate matter (DPM)
enhances the secretion of antigen-specific IgE in mice (Takano et al., 1997) and in the nasal
cavity of human subjects (Diaz-Sanchez et al., 1996, 1997; Ohtoshi et al., 1998; Nel et al., 2001).
Because IgE levels play a major role in allergic asthma (Wheatley and Platts-Mills, 1996),
upregulation of its production could lead to an increased response to inhaled antigen in particle-
exposed individuals.
Van Zijverden et al. (2000) and Van Zijverdan and Granum (2000) used mouse models to
assess the potency of particles (diesel, CB, silica) to adjuvate an immune response to a protein
antigen. All types of particles exerted an adjuvant effect on the immune response to
coadministered antigen, apparently stimulated by the particle core rather than the attached
chemical factors. Different particles, however, stimulated distinct types of immune responses.
In one model (Van Zijverden et al., 2001), BALB/c mice were intranasally treated with a mixture
of antigen (model antigen TNP-Ovalbumin, TNP-OVA) and particles on three consecutive days.
On day 10 after sensitization, mice were challenged with the antigen TNP-O VA alone, and five
days later the immune response was assessed. Diesel particulate matter, as well as CB, were
capable of adjuvating the immune response to TNP-OVA as evidenced by an increase of TNP-
specific antibody (IgGl and IgE) secreting B cells antibodies in the lung-draining lymph nodes.
Increased antigen-specific IgGl, IgG2a, and IgE isotypes were measured in the serum, indicating
that the response resulted in systemic sensitization. Importantly, an increase of eosinophils in
the BAL was observed with CB. Companion studies with the intranasal exposure model showed
that the adjuvant effect of CB particles was even more pronounced when the particles were given
during both the sensitization and challenge phases; whereas administration during the challenge
phase caused only marginal changes in the immune response. These data show that PM can
increase both the sensitization and challenge responses to a protein antigen, and the immune
stimulating activity of particles appears to be a time-dependent process, suggesting that an
7-134
-------
inflammatory microenvironment (such as may be created by the particles) is crucial for
enhancing sensitization by particles.
Only a small number of studies have examined mechanisms underlying the enhancement
of allergic asthma by ambient urban particles. Ohtoshi et al. (1998) reported that a coarse-size
fraction of resuspended ambient PM, collected in Tokyo, induced the production of granulocyte
macrophage colony stimulating factor (GMCSF), an upregulator of dendritic cell maturation and
lymphocyte function, in human airway epithelial cells in vitro. In addition to increased GMCSF,
epithelial cell supernatants contained increased IL-8 levels when incubated with DPM, a
principal component of ambient particles collected in Tokyo. Although the sizes of the two
types of particles used in this study were not comparable, the results suggest that ambient PM, or
at least the DPM component of ambient PM, may be able to upregulate the immune response to
inhaled antigen through GMCSF production. Similarly, Takano et al. (1998) has reported airway
inflammation, airway hyperresponsiveness, and increased GMGSF and IL-5 in mice exposed to
diesel exhaust.
In a study by Walters et al. (2001), PM10 was found to induce airway hyperresponsiveness,
suggesting that PM exposure may be an important factor contributing to increases in asthma
prevalence. Naive mice were exposed to a single dose (0.5 mg/mouse) of ambient PM, coal fly
ash, or diesel PM. Exposure to PM10 induced increases in airway responsiveness and BAL
cellularity; whereas diesel PM induced significant increases in BAL cellularity, but not airway
responsiveness. On the other hand, coal fly ash exposure did not elicit significant changes in
either of these parameters. Ambient PM-induced airway hyperresponsiveness was sustained
over 7 days. The increase in airway responsiveness was preceded by increases in BAL
eosinophils; whereas a decline in airway responsiveness was associated with increases in
macrophages. Thus, ambient PM can induce asthma-like parameters in naive mice.
Several other studies have examined in greater detail the contribution of the particle
component and the organic fraction of DPM to allergic asthma. Tsien et al. (1997) treated
transformed IgE-producing human B lymphocytes in vitro with the organic extract of DPM. The
organic phase extraction had no effect on cytokine production but did increase IgE production.
In these in vitro experiments, DPM appeared to be acting on cells already committed to IgE
production, thus suggesting a mechanism by which the organic fraction of combustion particles
can directly affect B cells and influence human allergic asthma.
7-135
-------
Cultured epithelial cells from atopic asthmatics show a greater response to DPM exposure
when compared with cells from nonatopic nonasthmatics. IL-8, GM-CSF, and soluble ICAM-1
increased in response to DPM at a concentration of 10 |ig/mL DPM (Bayram et al., 1998a,b).
This study suggests that particles could modulate airway disease through their actions on airway
epithelial cells. This study also suggests that bronchial epithelial cells from asthmatics are
different from those of nonasthmatics in regard to their mediator release in response to DPM.
Sagai and colleagues (1996) repeatedly instilled mice with DPM for up to 16 weeks and
found increased numbers of eosinophils, goblet cell hyperplasia, and nonspecific airway
hyperresponsiveness, changes which are central features of chronic asthma (National Institutes
of Health, 1997). Takano et al. (1997) extended this line of research and examined the effect of
repeated instillation of DPM on the antibody response to antigen OVA in mice. They observed
that antigen-specific IgE and IgG levels were significantly greater in mice repeatedly instilled
with both DPM and OVA. Because this upregulation in antigen-specific immunoglobulin
production was not accompanied by an increase in inflammatory cells or cytokines in lavage
fluid, it would suggest that, in vivo, DPM may act directly on immune system cells, as described
in the work by Tsien et al. (1997). Animal studies have confirmed that the adjuvant activity of
DPM also applies to the sensitization of Brown-Norway rats to timothy grass pollen
(Steerenberg et al., 1999).
Steerenberg et al. (2003) expanded on these findings using a range of PM collected for the
EU study, "Respiratory Allergy and Inflammation Due to Ambient Particles." The Brown
Norway rat (BN) was utilized as a pollen allergy model and the B ALB/c mouse was used as an
OVA allergy model. PM included were two DEP samples (DEPj from Lovelace Respiratory
Research Institute, DEPn from NIST, SRM 2975), ROFA (collected in Niagra, NY), Ottawa dust
(EHC-93), and road tunnel dust (RTD, collected in Noord tunnel near Hendrik-Ido Ambacht,
NL). Endotoxins in the PM were below detection levels except for EHC-93 (50 ng
endotoxin/mL) and RTD (1 ng endotoxin/mL). The animals were exposed to either just allergen
or allergen and PM combined during the sensitization and/or challenge phases. In the BN pollen
model only DEPj stimulated IgE and IgG response to pollen allergens. The pollen + PM rats had
fewer eosinophilic granulocytes than rats exposed to pollen alone. In the BALB/c OVA model
all of the PM samples when coexposed with OVA during the sensitization phase (but not the
challenge phase), created increases in IgE serum responses. Both histopathological examination
7-136
-------
of the lung and BAL analysis showed inflammatory response in the lung, predominantly due to
an influx of eosinophilic granulocytes. Increases were also seen in BAL levels of IL-4. The
authors ranked the adjuvant capacity of the particles tested based on the OVA model results as:
RTD > ROFA > EHC-93 > DEPj > DEPn.
Diaz-Sanchez and colleagues (1996) have continued to study the mechanism of DPM-
induced upregulation of allergic response in the nasal cavity of human subjects. In one study,
a 200-|iL aerosol bolus containing 0.15 mg of DPM was delivered into each nostril of subjects
with or without seasonal allergies. In addition to increases in IgE in nasal lavage fluid (NAL),
they found an enhanced production of IL-4, IL-6, and IL-13, cytokines known to be B cell
proliferation factors. The levels of several other cytokines also were increased, suggesting a
general inflammatory response to a nasal challenge with DPM. In a following study, these
investigators delivered ragweed antigen, alone or in combination with DPM, on two occasions,
to human subjects with both allergic rhinitis and positive skin tests to ragweed (Diaz-Sanchez
et al., 1997). They found that the combined challenge with ragweed antigen and DPM produced
significantly greater antigen-specific IgE and IgG4 in NAL. A peak response was seen at 96 h
postexposure. The combined treatment also induced expression of IL-4, IL-5, IL-10, and IL-13,
with a concomitant decrease in expression of Thl-type cytokines. Although the treatments were
not randomized (antigen alone was given first to each subject), the investigators reported that
pilot work showed no interactive effect of repeated antigen challenge on cellular and
biochemical markers in NAL. Diesel PM also resulted in the nasal influx of eosinophils,
granulocytes, monocytes, and lymphocytes, as well as the production of various inflammatory
mediators. The combined DPM plus ragweed exposure did not increase the rhinitis symptoms
beyond those of ragweed alone. Thus, DPM can produce an enhanced response to antigenic
material in the nasal cavity.
Extrapolation of these findings of enhanced allergic response in the nose to extremely high
concentrations of DPM to the human lung would suggest that ambient combustion particles
containing any ambient PM may have significant effects on allergic asthma. A study by
Nordenhall et al. (2001) has addressed the effects of diesel PM on airway hyperresponsiveness,
lung function and airway inflammation in a group of atopic asthmatics with stable disease. All
were hyperresponsive to methacholine. Each subject was exposed to DE (DPM = 300 |ig/m3)
and air for 1 h on two separate occasions. Lung function was measured before and immediately
7-137
-------
after the exposures. Sputum induction was performed 6 h, and methacholine inhalation test 24 h,
after each exposure. Exposure to DE was associated with a significant increase in the degree of
hyperresponsiveness, as compared to after air, a significant increase in airway resistance and in
sputum levels of interleukin (IL)-6 (p = 0.048). No changes were detected in sputum levels of
methyl-histamine, eosinophil cationic protein, myeloperoxidase, and IL-8.
These studies provide biological plausibility support for the exacerbation of allergic asthma
likely being associated with episodic exposure to PM. Although DPM may make up only a
fraction of the mass of urban PM, because of their small size, DPM may represent a significant
fraction of the ultrafine particle mode in urban air, especially in cities and countries that rely
heavily on diesel-powered vehicles.
In an examination of the effect of concentrated ambient PM on airway responsiveness in
mice, Goldsmith et al. (1999) exposed control and OVA-sensitized mice to an average
concentration of 787 |ig/m3 PM for 6 h/day for 3 days. Although OVA sensitization itself
produced an increase in the nonspecific airway responsiveness to inhaled methylcholine (MCh),
concentrated ambient PM did not change the response to MCh in OVA-sensitized or control
mice. For comparison, these investigators examined the effect of inhalation of an aerosol of the
active soluble fraction of ROFA on control and OVA-sensitized mice and found that ROFA
could produce nonspecific airway hyperresponsiveness to MCh in both control and OVA-
sensitized mice. Similar increases in airway responsiveness have been observed after exposure
to ROFA in normal and OVA-sensitized rodents (Gavett et al., 1997, 1999; Hamada et al., 1999,
2000).
Gavett et al. (1999) have investigated the effects of ROFA (intratracheal instillation) in
OVA sensitized and challenged mice. Instillation of 3 mg/kg (~ 60 jig) ROFA induced
inflammatory and physiological responses in the OVA mice that were related to increases in Th2
cytokines (IL-4, IL-5). Compared to OVA sensitization alone, ROFA induced greater than
additive increases in eosinophil numbers and in airway responsiveness to MCh.
Hamada et al. (1999, 2000) have examined the effect of a ROFA leachate aerosol in a
neonatal mouse model of allergic asthma. In the first study, neonatal mice sensitized by
intraperitoneal (ip) injection with OVA developed airway hyperresponsiveness, eosinophilia, and
elevated serum anti-OVA IgE after a challenge with inhaled OVA. Exposure to the ROFA
7-138
-------
leachate aerosol had no marked effect on the airway responsiveness to inhaled MCh in
nonsensitized mice, but did enhance the airway hyperresponsiveness to MCh produced in
OVA-sensitized mice. No other interactive effects of ROFA exposure with OVA were observed.
In a subsequent study, Hamada et al. clearly demonstrated that, whereas inhaled OVA alone was
not sufficient to sensitize mice to a subsequent inhaled OVA challenge, pretreatment with a
ROFA leachate aerosol prior to the initial exposure to aerosolized OVA resulted in an allergic
response to the inhaled OVA challenge. Thus, exposure to a ROFA leachate aerosol can alter
the immune response to inhaled OVA both at the sensitization stage at an early age and at the
challenge stage.
Lambert et al. (1999) and Gilmour et al. (2001) also examined the effect of ROFA on a
rodent model of pulmonary allergy. Rats were instilled intratracheally with 200 or 1000 jig
ROFA 3 days prior to sensitization with house dust mite (HDM) antigen. HDM sensitization
after 1000 jig ROFA produced increased eosinophils, LDH, BAL protein, and IL-10 relative to
HDM alone. Although ROFA treatment did not affect antibody levels, it did enhance pulmonary
eosinophil numbers. The immediate bronchoconstrictive and associated antigen-specific IgE
response to a subsequent antigen challenge was increased in the ROFA-treated group in
comparison with the control group. Together, these studies suggest that components of ROFA
can augment the immune response to antigen.
Evidence that metals are responsible for augmentation of an allergic sensitization by
ROFA was demonstrated by Lambert et al. (2000). In this follow-up study, BN rats were
instilled with 1 mg ROFA or the three main metal components of ROFA (Fe, V, or Ni) prior to
sensitization with instilled HDM. The three individual metals augmented different aspects of the
immune response to HDM: Ni and V produced an enhanced immune response to the antigen as
seen by higher HDM-specific IgE serum levels after an antigen challenge at 14 days after
sensitization, Ni and V also produced an increase in the lymphocyte proliferative response to
antigen in vitro; and the antigen-induced bronchoconstrictive response was greater only in
Ni-treated rats. Thus, instillation of metals at concentrations equivalent to those present in the
ROFA leachate mimicked the response to ROFA, suggesting that the metal components of
ROFA may be responsible for increased allergic sensitization observed in ROFA-treated
animals.
7-139
-------
Although these studies demonstrate that inhalation or instillation of ROFA augments the
immune response in allergic hosts, the applicability of these findings to ambient PM is an
important consideration. Goldsmith et al. (1999) compared the effects of inhalation of CAPs for
6 h/day for 3 days versus the effect of a single exposure to a ROFA leachate aerosol on the
airway responsiveness to MCh in OVA-sensitized mice. Exposure to ROFA leachate aerosols at
a concentration of 50 ng/mL significantly enhanced the airway hyperresponsiveness in OVA-
sensitized mice; whereas exposure to CAPs (average concentration of 787 |ig/m3) had no effect
on airway responsiveness in six separate experiments. Thus, the effect of the ROFA leachate
aerosols on the induction of airway hyperresponsiveness in allergic mice was significantly
different than that of high concentrations of ambient PM. Although airway responsiveness was
examined at only one postexposure time point, these findings do suggest that a great deal of
caution should be used in interpreting the results of studies using ROFA particles or leachates in
efforts to investigate the biologic plausibility of the adverse health effects of ambient PM.
7.5.4 Resistance to Infectious Disease
Development of an infectious disease requires both the presence of an appropriate
pathogen and host susceptibility to the pathogen. There are numerous specific and nonspecific
host defenses against microbes, and the ability of inhaled particles to modify resistance to
bacterial infection could result from a decreased ability either to clear or to kill microbes.
Rodent infectivity models have frequently been used to examine effects of inhaled particles on
host defense and infectivity. Mice or rats are challenged with a bacterial or viral load either
before or after exposure to the particles (or gas) of interest; mortality rate, survival time, or
bacterial clearance are then examined. Numerous studies that used the infectivity model to
assess inhaled PM effects were assessed previously (U.S. Environmental Protection Agency,
1982, 1989, 1996a). In general, acute exposure to sulfuric acid aerosols at concentrations up to
5,000 |ig/m3 were not very effective in enhancing mortality in a bacterially-mediated murine
model. In rabbits, however, sulfuric acid aerosols altered anti-microbial defenses after exposure
to 750 |ig/m3 for 2 h/day for 4 days (Zelikoff et al., 1994). Acute or short-term repeated
exposures to high concentrations of relatively inert particles have produced conflicting results.
7-140
-------
Carbon black (10,000 |ig/m3) was found to have no effect on susceptibility to bacterial infection
(Jakab, 1993); whereas TiO2 (20,000 |ig/m3) decreased the clearance of microbes and the
bacterial response of lymphocytes isolated from mediastinal lymph nodes (Gilmour et al.,
1989a,b). Also, exposure to DPM (2 mg/m3, 7 h/day, 5 days/week for 3 and 6 months) has been
shown to enhance the susceptibility of mice to the lethal effects of some, but not all, microbial
agents (Hahon et al., 1985). Pritchard et al. (1996) observed in CD-I mice exposed by
instillation to particles (0.05 mL of a 1.0 mg/mL suspension) with a high concentration of metals
(e.g., ROFA), that the increased mortality rate after streptococcus infection was associated with
the amount of metal in the PM. Thus, the pulmonary defense responses to microbial agents has
been altered at relatively high particle concentrations in animal models, with observed effects
being highly dependent on the microbial challenge and the test animal studied.
There are a few more recent studies that have examined mechanisms potentially
responsible for the effect of PM on infectivity. In one study, Cohen and colleagues (1997)
examined the effect of inhaled V on immunocompetence. Healthy rats were repeatedly exposed
first to 2 mg/m3 V, as ammonium metavanadate, and then instilled with polyinosinic-
polycytidilic acid (poly I:C), a double-stranded polyribonucleotide that acts as a potent
immunomodulator. Increases in lavage fluid protein and neutrophils were greater in animals
preexposed to V. Similarly, IL-6 and interferon-gamma were increased in V-exposed animals.
Alveolar macrophage function, as determined by zymosan-stimulated superoxide anion
production and by phagocytosis of latex particles, was also depressed more after poly I:C
instillation in V-exposed rats as compared to filtered air-exposed rats. These findings provide
evidence that inhaled V, a trace metal found in combustion particles and shown to be toxic
in vivo in studies using instilled or inhaled ROFA (Dreher et al., 1997; Kodavanti et al., 1997b,
1999), has the potential to inhibit the pulmonary response to microbial agents. However, it must
be remembered that these effects were found at very high exposure concentrations of V, and care
must be taken in extrapolating the results to ambient exposures of healthy individuals or those
with preexisting cardiopulmonary disease to trace concentrations (~3 orders of magnitude lower
concentration) of metals in ambient PM.
7-141
-------
7.6 RESPONSES TO PARTICULATE MATTER AND GASEOUS
POLLUTANT MIXTURES
Ambient PM itself is comprised of mixtures of particles of varying size and composition
which coexist in outdoor and indoor air with a number of ubiquitous co-pollutant gases (e.g., O3,
SO2, NO2, CO) and innumerable other non-PM components that are not routinely measured. The
following discussion examines effects of mixtures of ambient PM or PM constituents with
gaseous pollutants, as evaluated by studies summarized in Table 7-13. Toxicological
interactions between PM and gaseous co-pollutants may be antagonistic, additive, or synergistic
(Mauderly, 1993). The presence and nature of any interaction likely depends on chemical
composition, size, concentration and ratios of pollutants in the mixture, exposure duration, and
the endpoint being examined. It is difficult to predict a priori from the presence of certain
pollutants whether any interaction will occur and, if so, whether it will be synergistic, additive,
or antagonistic.
Mechanisms responsible for the various forms of interaction are speculative. In terms of
potential health effects, the greatest hazard from pollutant interaction is the possibility of
synergy between particles and gases, especially if effects occur at concentrations at which no
effects occur when individual constituents are inhaled. Various physical and chemical
mechanisms may underlie synergism. For example, physical adsorption or absorption of some
other material on a particle could result in transport to more sensitive sites or accumulation at
sites where this material would not normally be deposited in toxic amounts. This physical
process may explain, for example, interactions found in studies of mixtures of CB and
formaldehyde or of CB and acrolein (Jakab, 1992, 1993). However, an earlier study
(Rothenberg et al., 1989) has demonstrated that, based on the physical and chemical
characteristics of formaldehyde, a 1-h 1-ppm exposure to formaldehyde and dust would result in
deposition of-500 jig of formaldehyde vapor into the upper respiratory tract and deposition of
only 2 to 50 ng of formaldehyde adsorbed to dust into the pulmonary compartment. Thus, an
important factor in PM/gaseous mixture dose evaluation is the equilibrium that exists between
the vapor phase of the gas and the particle-associated gas.
Also, chemical interactions between PM and gases can occur on particle surfaces, thus
forming secondary products whose surface layers may be more active lexicologically than the
primary materials and that can then be carried to a sensitive site. The hypothesis of such
7-142
-------
TABLE 7-13. RESPIRATORY AND CARDIOVASCULAR EFFECTS OF PM AND
GASEOUS POLLUTANT MIXTURES
Species, Gender, Strain
Age, or Body Weight
Humans; healthy
15 M, 10 F,
34.9±10 years of age
Mice, BALB/c,
3 days old
Gases and PM
CAPs + O3
Filtered air
(control)
CAPs
(Boston)
03
CAPs + O3
Exposure
Exposure Technique Concentration
Inhalation 150 ug/m3 CAPs
120 ppm O3
1.6 ug/m3PM25
8.5 ppb O3
Inhalation 0-1500 ug/m3
0.3 ppm
100-500 ug/m3 +
0.3 ppm
Particle Exposure Cardiopulmonary Effects of Inhaled
Size Duration PM and Gases
PM2 5 2 h PM2 5 CAPs + O3 exposure increased acute
brachial artery vasoconstriction (as
PM2 5 determined by vascular ultrasonography
2h performed before and 10 min after
exposure), but not endothelial-dependent or
-independent nitroglycerine-mediated
dilation. Lack of comparison between
exposure to PM or O3 alone precludes
attribution of observed effects to PM or O3
alone or to joint effect.
PM2 5 5 h 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. For every 100
Hg/m3 CAPs, Penh increased 0.86%.
Reference
Brook et al.
(2002)
Kobzik et al.
(2001)
Rats
Resuspended Inhalation
Ottawa (whole-body)
urban PM and O3
5,000 or 50,000 ug/m3
PM and 0.8 ppm O3
Single 4-h PM alone caused no change in cell
exposure proliferation in bronchioles or parenchyma.
Coexposure at both dose levels 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.
Vincent et al.
(1997)
Rats
Rats, F344: male and
female 9 weeks old
Ottawa urban PM
and O3
Ambient particles
and gases
Inhalation
Natural 23 h/day
exposure to filtered and
unfiltered Mexico City
air.
40,000 ug/m3 and
0.8 ppm O3
0.0 18 ppm O3
3.3 ppb CH20
0.068 mg/m3 TSP
0.032 mg/m3 PM10
0.016 mg/m3 PM25
4.5 um Single 4-h
MMAD exposure
followed by
20 h clean air
23 h/day
for 7 weeks
Coexposure to particles potentiated
O3-induced septal cellurity. Enhanced
septal thickening associated with elevated
production of macrophage inflammatory
protein-2 and endothelin 1 by lung lavage
cells.
Histopathology examination revealed no
nasal lesions in exposed or control rats;
tracheal and lung tissue from both groups
showed similar levels of minor
abnormalities.
Bouthillier
etal. (1998)
Moss et al.
(2001)
-------
TABLE 7-13 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF PM AND
GASEOUS POLLUTANT MIXTURES
Species, Gender,
Strain Age, or
Body Weight
Humans,
children:
healthy (N = 15)
asthma (N = 26);
Age 9-12 years
Rats
Rats, S-D, male,
250-300 g
Rats, S-D
300 g
Mice, Swiss:
female, age
5 weeks
Rats, Fischer
NNia, male, 22 to
24 months old
Gases and PM Exposure Technique
H2SO4, Inhalation
SO2, and O3 (chamber)
H2SO4 and O3 Inhalation,
whole body
H2SO4 and O3 Inhalation,
nose-only
H2SO4-coated carbon and Inhalation,
O3. nose-only
Carbon and SO2 Inhalation,
flow-past,
nose-only
Carbon, ammonium Inhalation
bisulfate, and O3
Exposure
Concentration
100 ± 40 ug/m3
H2SO4, 0.1 ppmSO2,
and 0.1 ppmO3
20 to 150 ug/m3
H2SO4and0.12or0.2
ppm O3
500 ug/m3 H2SO4
aerosol (two different
particle sizes),
with or without
0.6 ppm O3
0.2 ppm O3
+ 50 ug/m3 C
+ 100 ug/m3 H2S04
0.4 ppm O3
+250 ug/m3 C
+500 ug/m3 H2S04
10,000 ug/m3 carbon
with or without 5 to
20 ppm SO2 at 10% or
85% RH
50 ug/m3 carbon +
70 ug/m3 ammonium
bisulfate + 0.2 ppm O3
or 100 ug/m3 carbon
+140 ug/m3
ammonium bisulfate
+ 0.2ppmO3
Particle
Size
0.6 um H2SO4
0.4 to 0.8 um
Fine (0.3 um
MMD, og =
1.7) and
ultrafine
(0.06 um,
og=1.4)
0.26 um
og = 2.2
0.3 um
MMAD
°8 = 2'7
0.4 um
MMAD
°g = 2-°
Exposure Duration
Single 4-h exposure
with intermittent
exercise
Intermittent
(12 h/day) or
continuous exposure
for up to 90 days
4 h/day for 2 days
4 h/day for 1 day or
5 days
Single 4-h exposure
4 h/day,
3 days/week for
4 weeks
Cardiopulmonary Effects of Inhaled
PM and Gases
Positive association seen between acid
dose and respiratory symptoms, but not
spirometry, in asthmatic children.
No significant changes in healthy
children.
No interactive effect of H2SO4 and O3 on
biochemical and morphometric endpoints.
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.
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.
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.
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
Linn et al.
(1997)
Last and
Pinkerton
(1997)
Kimmel
etal.
(1997)
Kleinman
et al. (1999)
Jakab et al.
(1996)
Clarke et al.
(2000c)
Bolarin
etal.
(1997)
-------
TABLE 7-13 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF PM AND
GASEOUS POLLUTANT MIXTURES
Species, Gender,
Strain Age, or
Body Weight Gases and PM
Rats Elemental carbon + O3 +
ammonium bisulfate
Rats, F344/N O3 + nitric acid
male NO2 + carbon particles +
ammonium bisulfate
Rats, F344/N O3
male HNO3
O3 + HNO3
Exposure Particle
Exposure Technique Concentration Size
Inhalation 0.2 ppm O3 + 0.46 um
elemental carbon 0.3 um
50 um/m3 +
ammonium bisulfate
70 ug/m3
Inhalation low: 0. 16 ppm + 0.3 um
0.11 ppm + 0.05
mg/m3 + 0.03 mg/m3
medium: 0.3 ppm +
0.21 ppm +
0.06 mg/m3 +
0. 1 mg/m3
high: 0.59 ppm +
0.39 ppm + 0.1 mg/m3
+ 0.22 mg/m3
Inhalation 0.151 ± 0.003 ppm
51.1 ± 5.4ug/m3
0.152 ±0.003 ppm +
49.9 ±7.0 ug/m3
Exposure Duration
4h/day
3 days/week
4 weeks
4h/day
3 days/week
4 weeks
4h/day
3 days/week
40 weeks
Cardiopulmonary Effects of Inhaled
PM and Gases
Increased macrophage phagocytosis and
increased respiratory burst; decreased
lung collagen.
Dose-dependent decrease in macrophage
Fc-receptor mediated-phagocytosis (only
significant in high dose group),
nonsignificant increase in epithelial
permeability and proliferation, altered
breathing pattern in high dose group.
Increased lung putrescine content in all
exposed rats. Synergistic effect.
Reference
Kleinman
etal.
(2000)
Mautz et al.
(2001)
Sindhu et al.
(1998)
-------
chemical interactions has been examined in gas and particle exposure studies by Amdur and
colleagues (Amdur and Chen, 1989; Chen et al., 1992) and Jakab and colleagues (Jakab and
Hemenway, 1993; Jakab et al., 1996). These investigators have suggested that synergism occurs
as secondary chemical species are produced, especially under conditions of increased
temperature and relative humidity.
Another potential mechanism of gas-particle interaction may involve a pollutant-induced
change in the local microenvironment of the lung, enhancing the effects of the co-pollutant.
For example, Last et al. (1984) suggested that the observed synergism between ozone (O3) and
acid sulfates in rats was due to a decrease in the local microenvironmental pH of the lung
following deposition of acid, enhancing the effects of O3 by producing a change in the reactivity
or residence time of reactants, such as radicals, involved in O3-induced tissue injury.
One newly available controlled exposure study evaluated the effects of a combined
inhalation exposure to PM25 CAPs and O3 in human subjects. In a randomized, double-blind
crossover study, Brook et al. (2002) exposed 25 healthy male and female subjects, 34.9 ±10
(SD) years of age, to filtered ambient air containing 1.6 |ig/m3 PM25 and 8.5 ppb O3 (control) or
to unfiltered air containing 150 |ig/m3 PM25 CAPs and 120 ppb O3 while at rest for 2 h. Blood
pressure was measured and high-resolution brachial artery ultrasonography (BAUS) was
performed prior to and 10 min after exposure. The BAUS technique was used to measure
brachial artery diameter (BAD), endothelium-dependent flow-mediated dilation (FMD), and
endothelial-independent nitroglycerine-mediated dilation (NMD). Although no changes in blood
pressure or endothelial-dependent or endothelial-independent dilatation were observed, a small
(2.6%) but statistically significant (p = 0.007) decrease in BAD was observed in CAPs plus O3
exposures (-0.09 mm) when compared to filtered air exposures (+0.01 mm). Preexposure BAD
showed no significant day-to-day variation (± 0.03 mm). This finding suggests that combined
exposure to a mixture of PM25 CAPs plus O3 produces vasoconstriction, potentially via
autonomic reflexes or as the result of an increase in circulating endothelin, as has been described
in rats exposed to urban PM (Vincent et al., 2001). It is not known, however, whether this effect
is caused by CAPs or O3 alone. The likelihood that analogous vasoactive responses could be
found at ambient PM25 and O3 levels typically found in some U.S. urban locations is enhanced
by the fact that such responses would likely have been seen at distinctly lower exposure levels
7-146
-------
had the PM and O3 exposures occurred during light, moderate, or heavy exercise (which
enhances delivery of both PM and O3 to lower regions of the respiratory tract).
The interaction of PM and O3 was further examined in a murine model of O VA-induced
asthma. Kobzik et al. (2001) investigated whether coexposure to inhaled, concentrated PM from
Boston, MA and to O3 could exacerbate asthma-like symptoms. On days 7 and 14 of life, half of
the BALB/c mice used in this study were sensitized by ip injection of OVA and then exposed to
OVA aerosol on three successive days to create the asthma phenotype. The other half received
the ip OVA, but were exposed to a phosphate-buffered saline aerosol (controls). The mice were
further subdivided (n >6I/group) and exposed for 5 h to CAPs, ranging from 63 to 1569 |ig/m3,
0.3 ppm O3, CAPs + O3, or to filtered air. Pulmonary resistance and airway responsiveness to an
aerosolized MCh challenge were measured after exposures. A small, statistically significant
increase in pulmonary resistance and airway responsiveness, respectively, was found in both
normal and asthmatic mice immediately after exposure to CAPs alone and to CAPs + O3, but not
to O3 alone or to filtered air. By 24 h after exposure, the responses returned to baseline levels.
No significant increases in airway inflammation were seen after any of the pollutant exposures.
In this well-designed study of a small-animal model of asthma, CAPs and O3 did not appear to
be synergistic. In further analysis of the data using specific elemental groupings of the CAPs,
the acutely increased pulmonary resistance was found to be associated with the AISi fraction of
PM. Thus, some components of concentrated PM2 5 may affect airway caliber in sensitized
animals.
Linn and colleagues (1997) examined the effect of a single exposure to 60 to 140 |ig/m3
H2SO4, 0.1 ppm SO2, and 0.1 ppm O3 in healthy (N = 15) and asthmatic children (N = 26).
The children performed intermittent exercise during the 4-h exposure to increase the inhaled
dose of the pollutants. An overall effect on the combined group of healthy and asthmatic
children was not observed. The combined pollutant exposure had no effect on spirometry in
asthmatic children, and no changes in symptoms or spirometry were observed in healthy
children. A positive association between acid concentration and symptoms was seen, however,
in the subgroup of asthmatic children. Thus, the effect of combined exposure to PM and gaseous
co-pollutants appeared to have less effect on asthmatic children exposed under controlled
laboratory conditions in comparison with field studies of children attending summer camp
(Thurston et al., 1997). However, prior exposure to H2SO4 aerosol may enhance the subsequent
7-147
-------
response to O3 exposure (Linn et al., 1994; Frampton et al., 1995); and the timing and sequence
of the exposures may be important.
Vincent et al. (1997) exposed rats to 5 or 50 mg/m3 of resuspended Ottawa urban ambient
particles for 4 h in combination with 0.8 ppm O3. Although PM alone caused no change in cell
proliferation (3H-thymidine labeling), coexposure to either concentration of resuspended PM
with O3 greatly potentiated the proliferative effects of exposure to O3 alone. These interactive
changes occurred in epithelial cells of the terminal bronchioles and the alveolar ducts.
These findings using resuspended ambient PM, although at high concentrations, are consistent
with studies showing interactions between sulfuric acid (H2SO4) aerosols and O3.
Kimmel and colleagues (1997) examined the effect of acute coexposure to O3 (0.6 ppm)
and fine (MMD = 0.3 jim) or ultrafme (MMD = 0.06 jim) H2SO4 aerosols (0.5 mg/m3) on rat
lung morphology. They determined morphometrically that alveolar septal volume was increased
in animals coexposed to O3 and ultrafme, but not fine, H2SO4. Interestingly, cell labeling, an
index of proliferative cell changes, was increased only in animals coexposed to fine H2SO4 and
O3, as compared to animals exposed to O3 alone. Importantly, Last and Pinkerton (1997), in
extending their previous work, found that subchronic exposure to acid aerosols (20 to 150 |ig/m3
H2SO4) had no interactive effect on the biochemical and morphometric changes produced by
either intermittent or continuous O3 exposure (0.12 to 0.2 ppm). Thus, the interactive effects of
O3 and acid aerosol coexposure in the lung disappeared during the long-term exposure.
Kleinman et al. (1999) examined the effects of exposure to O3 (0.2 and 0.4 ppm) plus fine
(MMAD = 0.26 |im) H2SO4-coated carbon particles (100, 250, and 500 |ig/m3) for 1 or 5 days.
They found that the inflammatory response with the O3-particle mixture was greater after 5 days
(4 h/day) than after Day 1. This contrasted with O3 exposure alone (0.4 ppm), which caused
marked inflammation on acute exposure, but no inflammation after 5 consecutive days of
exposure.
Kleinman et al. (2000) examined the effects of a mixture of elemental carbon particles
(50 |ig/m3), O3 (0.2 ppm), and ammonium bisulfate (70 |ig/m3) on rat lung collagen content and
macrophage activity in senescent rats. Exposures were nose-only, 4 h/day, 3 consecutive days
per week for a total of 4 weeks. Decreases in lung collagen, and increases in macrophage
respiratory burst and phagocytosis were observed. They found small changes in macrophage
function and in injury to cells of the lung parenchyma, with exposures to just carbon and
7-148
-------
ammonium bisulfate. With the addition of O3, changes in those biological responses became
significant. These results suggest that (a) O3 may enhance the toxicity of inhaled particles in
terms of the above types of pathophysiologic responses and/or (b) conversely, PM25 exposure
may enhance O3-induced toxicity in aged rats. Mautz et al. (2001) used a similar mixture (i.e.,
elemental carbon particles, O3, ammonium bisulfate, but with NO2 also) and exposure regimen as
Kleinman et al. (2000). There were decreases in pulmonary macrophage Fc-receptor binding
and phagocytosis and increases in acid phosphatase staining. Bronchioalveolar epithelial
permeability and cell proliferation were increased. Altered breathing patterns were also seen,
but with some adaptation evident over the course of repeated O3 exposure.
Other studies have also examined interactions between carbon particles and gaseous
co-pollutants. Jakab et al. (1996) and Clarke et al. (2000c) challenged mice with a single 4-h
exposure to a high concentration of carbon particles (10 mg/m3) in the presence of 10 ppm SO2
(-140 jig cpSO42") at low and high relative humidities. Macrophage phagocytosis was depressed
significantly only in mice exposed to the combined pollutants under high relative humidity
(85%) conditions. There was no evidence of an inflammatory response based on total cell
counts and differential cell counts from BAL; however, macrophage phagocytosis remained
depressed for 7 to 14 days. Intrapulmonary bactericidal activity also was suppressed and
remained suppressed for 7 days. This study suggests that fine carbon particles can serve as an
effective carrier for acidic sulfates where chemical conversion of adsorbed SO2 to acid sulfate
species occurred. Interestingly, the depression in macrophage function was present as late as
7 days postexposure.
Bolarin et al. (1997) exposed rats to 50 or 100 |ig/m3 carbon particles in combination with
ammonium bisulfate (70 or 140 |ig/m3) and O3 (0.2 ppm) for 4 h/day, 3 days/week for 4 weeks.
Despite 4 weeks of exposure, they observed no changes in protein concentration in lavage fluid
or blood prolyl 4-hydroxylase, an enzyme involved in collagen metabolism. Slight decreases in
plasma fibronectin were present in animals exposed to the combined pollutants versus O3 alone.
Thus, as previously noted, the potential for adverse effects in the lungs of animals challenged
with a combined exposure to particles and gaseous pollutants is dependent on numerous factors,
including the gaseous co-pollutant, concentration, and time.
The effects of O3 modifying the biological potency of PM (diesel PM and CB) were
examined by Madden et al. (2000). Reaction of NIST Standard Reference Material # 2975
7-149
-------
diesel PM with 0.1 ppm O3 for 48 h increased the potency (compared to unexposed or
air-exposed diesel PM) to induce neutrophil influx, total protein, and LDH in lung lavage fluid in
response to intratracheal instillation. Exposure of the diesel PM to high, nonambient O3
concentration (1.0 ppm) attenuated the increased potency, suggesting destruction of the bioactive
reaction products. Unlike the diesel particles, CB particles exposed to 0.1 ppm O3 did not
exhibit an increase in biological potency, which suggested that the reaction of organic
components of the diesel PM with O3 contributed to the increased potency. Reaction of particle
components with O3 was ascertained by chemical determination of specific classes of organic
compounds.
In a complex series of exposures, Oberdorster and colleagues examined the interaction of
ultrafine carbon particles (100 |ig/m3) and O3 (1 ppm) in young and old F-344 rats that were
pretreated with aerosolized endotoxin (Elder et al., 2000a,b). In old rats, exposure to singlet
ultrafine carbon and O3 produced an interaction that resulted in a greater influx in neutrophils
than that produced by either agent alone. This interaction was not seen in young rats. Oxidant
release from lavage fluid cells was also assessed and the combination of endotoxin, carbon
particles, and O3 produced an increase in oxidant release in old rats. This combination produced
the opposite response in the cells recovered from the lungs of the young rats, indicating that the
lungs of the aged animals underwent greater oxidative stress in response to this complex
pollutant mix of particles, O3, and a biogenic agent.
The effects of gaseous pollutants on PM-mediated responses also have been examined by
in vitro studies, though to a limited extent. Churg et al. (1996) demonstrated increased uptake of
asbestos or TiO2 into rat tracheal explant cultures in response to 10 min O3 (up to 1.0 ppm)
preexposure. These data suggest that O3 may increase the penetration of some types of PM into
epithelial cells. Additionally, Madden et al. (2000) demonstrated a greater potency for ozonized
diesel PM to induce prostaglandin E2 production from human epithelial cell cultures, suggesting
that O3 can modify the biological activity of PM derived from diesel exhaust.
In summary, the newly available combined (PM and gaseous co-pollutant) studies provide
only relatively limited evidence for additive or interactive joint PM/gaseous pollutant effects on
one or the other few health endpoints evaluated. For example, recent studies have demonstrated
that coexposures of CAPs and O3 cause potentiation of proliferative changes in the epithelium of
terminal bronchioles (Vincent et al., 1997) and enhanced septal cellularity (Bouthillier et al.,
7-150
-------
1998) seen with O3 exposure alone. Both combined CAPs/O3 and O3-alone exposure in a mouse
asthma model (Kobzik et al., 2001) showed increases in airway responsiveness and pulmonary
resistance, thus indicating a lack of synergism with the combined exposure. Mixtures of
elemental carbon particles, O3, and ammonium bisulfate showed changes in lung collagen,
AM respiratory burst, and phagocytosis (Kleinman et al, 2000), the results are ambiguous as to
whether PM was enhancing the effects of O3 or the converse. A short exposure of combined
carbon particle/SO2 caused depressed AM phagocytosis and suppressed intrapulmonary
bactericidal activity which lasted for a week (Jakab et al., 1996; Clarke et al., 2000c). On the
other hand, other studies using coexposures of PM and gases have demonstrated no changes in
histopathological (Moss et al., 2001) or biochemical and morphometric endpoints (Last and
Pinkerton, 1997).
The mechanisms by which interactions between PM and gases occur is thought to be by:
(1) formation of secondary products by chemical interactions between the gas and the particle,
(2) adherence of material to the particle and subsequent transport to sensitive sites, and/or
(3) pollutant-induced change in the local microenvironment of the lung (e.g., by decreasing
the pH). All of these interactions have the potential to create antagonistic, additive, or
synergistic interactions between PM and gases, which can modify their individual effects.
7.7 QUANTITATIVE COMPARISONS OF EXPERIMENTAL PM
EFFECTS ON CARDIOVASCULAR/RESPIRATORY ENDPOINTS
IN HUMANS AND LABORATORY ANIMALS
7.7.1 Introduction
The extensive literature assessed in the foregoing sections provides considerable new
information on experimentally-induced effects of various types of PM on cardiovascular and
respiratory endpoints. The ensuing subsections attempt to characterize salient exposure/dose-
effect relationships; including comparisons between lowest observed effect levels (LOELs)
reported thus far for normal and compromised subjects. In the sections that follow, for both the
cardiovascular/systemic and the respiratory effects, the LOELs derived from inhalation studies
are first discussed and then those from instillation studies, followed by discussion of in vitro
observed effect levels. In addition, efforts are made to delineate key factors important in
attempting to extrapolate observed effects across species (rat to human) and/or to human ambient
7-151
-------
exposure conditions and to provide illustrative examples of some extrapolation modeling
outcomes.
7.7.1.1 Cardiovascular and Systemic Effects of Inhaled Particulate Matter
Newly available studies examining the cardiovascular and systemic effects of inhaled PM
have for the most part not carried out dose-response evaluations using multiple exposure levels
in the same study. However, various types of effects reported across a wide range of
concentrations for various types of PM tested at single exposure levels do allow one to gain
some impressions about possible lowest-observed-effect-levels (LOELs).
Concentrated ambient particles were used in a number of studies examining cardiovascular
and systemic endpoints, but none were done in a manner so as to allow clear delineation of dose-
response relationships for the endpoints evaluated. In fact, as CAPs vary from day to day, any
comparisons at different times do not comprise a true dose-response study. Probably the lowest
concentrations of U.S. ambient air observed to experimentally induce any cardiovascular effect
in humans were those in the study by Ohio et al. (2000a). Healthy adult subjects were divided in
three groups plus a fourth group receiving filtered air as control. All three exposure groups
(which had PM25 levels averaging 47.2 ± 5.3, 107.4 ± 9.3, and 206 ± 14.1 |ig/m3 after
concentration) showed increases in plasma fibrinogen which were accompanied by increases in
BAL neutrophils. The study did not show a dose response, which would have necessitated
exposure of the subjects at differing levels of CAPs. Another very small human exposure study
was also reported by Petrovic et al. (2000) to show a trend toward increased blood fibrinogen in
two of four human adults exposed for 2 h to Toronto CAPs ranging up to -125 |ig/m3. Also,
in laboratory animal studies, Godleski et al. (2000) used CAPs from the Boston area and found
effects on heart rate and ECG at a CAPs dose of-100 to 1000 |ig/m3 in some dogs (though the
lowest exposure that produced these effects was not determined). Both normal dogs and dogs
compromised by coronary occlusion were reported to be affected. However, a study exposing
both F344 rats and hamsters to CAPs collected in Manhattan at concentrations ranging from
132 to 919 |ig/m3 had contrasting findings (Gordon et al., 2000). That is, in hamsters and in rats,
both normal and MCT-treated, there was an increase in HR and peripheral blood cell differential
counts, but no other cardiovascular effects were observed. In contrast to these studies, a number
of other studies have examined a wide range of cardiovascular endpoints and found no changes
7-152
-------
in cardiovascular parameters, e.g., following inhalation exposures of dogs to Boston CAPS at 3
to 360 |ig/m3 (Clarke et al., 2000a), of rats to NYC CAPs at 95 to 341 |ig/m3 (Nadziejko et al.,
2002), and of humans to ultrafine carbon particles at 10 |ig/m3 (Frampton, 2001).
A number of other studies have examined cardiovascular and systemic endpoints at much
higher concentrations, using UAP and ROFA. For example, two other new studies showed
effects of Ottawa UAP on levels of endothelin. Bouthillier et al. (1998) found that 40 mg/m3
Ottawa UAP in Fischer 344 rats caused an increase in plasma endothelin-1 levels without
causing acute lung injury. Also, Vincent et al. (2001) found that 48 mg/m3 Ottawa UAP caused
increases in both endothelin-1 and endothelin-3 in Wistar rats, the endothelins being likely
powerful cardiotoxic agents. However, the relevance of these effects of exposures to such
extremely high ambient PM concentrations to evaluation of current ambient PM exposures in the
U.S. is questionable.
Two studies reported arrhythmias in response to ROFA exposure. Wellenius et al. (2002)
exposed healthy SD rats and rats with a model of myocardial infarction to 3 mg/m3 Boston
ROFA and found arrhythmias, ECG abnormalities, and decreases in HRV in the compromised
animals. Watkinson et al. (2000b) exposed healthy SD rats and rats with cold stress, O3
preexposure, or MCT to 15 mg/m3 ROFA (source not reported). They observed increased
arrhythmias, decreased heart rates, and hypothermia in the compromised animals. The same
concentration in SH rats caused cardiomyopathy, monocytic cell infiltration, and increased
expression of cardiac cytokines IL-6 and TGF-p. The ROFA-exposed SH rats also showed
increased ECG abnormalities compared to air-exposed SH rats. In another study, Muggenburg
et al. (2000a) exposed beagles to 3 mg/m3 Boston ROFA and found no effects on ECG and a
trend toward decreased heart rate, these overall results not being consistent with the Godleski
et al. (2000) findings noted above.
Comparisons were made also between normotensive WKY rats and SH rats following an
exposure to 15 mg/m3 Boston ROFA (Kodavanti et al., 2002a). The SH animals demonstrated
increased plasma fibrinogen and small but consistent decreased total white blood cell numbers,
the decrease being due mostly to decreased numbers of lymphocytes. Boston ROFA was used in
another study by Kodavanti et al. (2003) at a concentration of 2, 5, or 10 mg/m3 (for 6 h/day for
4 consecutive days) to compare cardiovascular endpoints in normal SD, WKY, and SH rats.
In this exposure paradigm, no significant effects were seen for any of the rat strains. However,
7-153
-------
with a second exposure paradigm (10 mg/m3 for 6 h/day for 16 weeks) WKY rats showed
cardiac lesions in the form of randomly distributed foci of fibrosis and inflammation in the
ventricles and the interventricular septum. Also, normal SD and MCT-treated rats exposed to
0.58 mg/m3 Boston ROFA showed increased expression of MIP-2, predominantly in heart
macrophages, in the MCT-treated animals.
In general, in the studies noted above, lower inhalation doses of CAPs than ROFA have
been found to elicit at least some cardiovascular effects. Some PM2 5 studies have demonstrated
effects (increased blood fibrinogen) at concentrations as low as -50 to 330 |ig/m3, whereas other
studies of CAPs at similar or higher concentrations did not show effects at such levels. Some of
the limitations of CAPs studies were discussed earlier in Section 7.1.1 Methodological
Considerations, and these must be kept in mind when interpreting CAPs data. Studies of UAP or
ROFA at much higher concentrations have also reported effects in healthy animals, but the
relevance for evaluation of health effects associated with current ambient PM levels in the
United States is unclear. The lack of more data from studies completing dose-response
evaluations highlights a need for more rigorous future evaluation of dose-effect relationships.
7.7.1.2 Cardiovascular and Systemic Effects of Instilled Particulate Matter
Recent studies characterizing the cardiovascular and systemic effects of instilled PM show
that most effects have been seen in a dose range of 0.7 to 9 mg/kg body weight. To better
compare studies here, all reported instillation study doses were converted to mg/kg body weight.
Urban air particles (UAPs) were used in several studies that evaluated changes in heart
rate, temperature, and blood parameters. Ottawa UAP at a dose of 7 mg/kg was instilled in aged
(15 months old) SH rats (Watkinson et al., 2000a, 2000b). Effects seen at this dose were
hypothermia and bradycardia. Mukae et al. (2001) exposed female New Zealand White rabbits
to 2 mg/kg Ottawa UAP and found a number of altered cardiovascular endpoints; e.g., increases
in circulating PMN band cell numbers and in the size of the bone marrow mitotic pool of PMNs,
as well as a shortened transit time of PMN through the postmitotic pool in marrow. Suwa et al.
(2002) evaluated similar endpoints with the same UAP, using 1.6 mg/kg in female Watanable
heritable hyperlipidemic rabbits. The same increases in PMN parameters were observed in this
study, along with progression of atherosclerotic lesions, increases in plaque cell turnover,
extracellular lipid pools, and total lipids in aortic lesions.
7-154
-------
Dose-response evaluations were carried out in several ROFA studies. Boston or Florida
was the origin of ROFA used in many of the IT studies that investigated cardiovascular and
systemic effects, (though several research groups neglected to report the source of the ROFA).
Arrhythmias were observed with both normal and MCT-treated rats at a dose of 3 mg/kg of
Florida ROFA (Watkinson et al., 1998). In another study by the same investigators, a dose of
0.7 mg/kg of ROFA (origin not reported) in SH rats was shown to create significant arrhythmias,
whereas 7 mg/kg induced arrhythmias in normal rats (Watkinson et al., 2000a, 2000b). One
study demonstrated ECG abnormalities in SD rats compromised by either MCT-pretreatment or
cold stress (10 °C) at a dose of 3 mg/kg ROFA (origin not reported). Bradycardia was seen in a
number of studies at doses of 0.7 to 7 mg/kg ROFA. Campen et al. (2000) observed a decreased
heart rate at 3 mg/kg ROFA (origin not reported) in normal SD rats and at 0.7 mg/kg in rats
compromised by cold stress, O3-pre-exposure, or MCT. Florida ROFA, tested at 7 mg/kg only,
caused bradycardia in both normal SD and MCT-treated rats (Costa and Dreher, 1997). Aged
SH rats displayed decreased heart rates when exposed to 1.4 mg/kg ROFA (origin not reported;
Watkinson et al., 2000a).
The systemic response of hypothermia has been observed in normal SD rats and SD rats
compromised by cold stress, O3-preexposure, or MCT when exposed to 0.7 mg/kg ROFA (origin
not reported; Campen et al., 2000). Watkinson et al. (2000b) observed the hypothermic response
in compromised rats at 1.4 mg/kg and in normal SD rats at 7 mg/kg. Increases in plasma
fibrinogen have been observed in normal SD rats following exposure to 8.3 mg/kg Florida
ROFA (Gardner et al., 2000). However, other hemostatic parameters and cardiovascular risk
factors, such as activated partial thromboplastin time, prothrombin time, plasma viscosity, and
complete blood count, were unaltered by the exposure. Kodavanti et al. (2002a) compared the
response to Boston ROFA at doses of 1 and 5 mg/kg in normal WKY and SH rats. Both strains
demonstrated increased plasma fibrinogen at the 5 mg/kg dose, while only the WKY rats showed
increased hematocrit at that dose. The SH rats demonstrated decreases in blood lymphocytes
and increases in blood neutrophils at 5 mg/kg. Lethality was observed in MCT-treated SD rats
exposed to 3 or 7 mg/kg Florida ROFA (Costa and Dreher, 1997).
Overall, then, some cardiovascular and systemic effects of instilled ambient PM were
observed at instilled doses of-1.5 to 7 mg/kg body weight. Some effects were also apparent
with exposures to ROFA in a dose range of 0.7 to 9 mg/kg body weight. In many cases the
7-155
-------
compromised animals in these studies showed effects at lower doses than normal counterparts.
Again, instillation studies must be viewed in light of the caveats mentioned earlier regarding
possible alteration of the physiochemical characteristics due to collection, storage, and
resuspension.
7.7.1.3 Respiratory Effects of Inhaled Particulate Matter
The few available inhalation studies of ambient PM respiratory effects in humans have
yielded consistent results in finding little or no indications of pulmonary function decrements or
increased respiratory symptoms among healthy adults exposed for 2 h to CAPs from several
locations (Toronto, Los Angeles, Chapel Hill, NC) at concentrations across a range of-25 up to
-300 |ig/m3 (Ohio et al., 2000a; Petrovic et al., 2000; Gong et al., 2000; Gong et al., 2003).
On the other hand, some of these studies did find indications of mild lung inflammatory
responses, although some were of unclear health significance.
Relatively few laboratory animal studies have been done to examine the respiratory effects
of inhaled PM, versus instillation studies. Mongrel dogs were exposed to Boston CAPs for
6 h/day for 3 days at concentrations varying from -100 to 1000 |ig/m3 (Godleski et al., 2000).
The only small effects seen were decreased respiratory rate over time and some increases in
BAL neutrophils. Also, Clarke et al. (1999) exposed SD rats, both normal and SO2-pretreated
bronchitic rats to Boston CAPs for 5 h/day for 3 days at concentrations of 200, 600, and
730 |ig/m3 (mean CAPs level for each day). With such CAPs exposures, PEF and TV were
increased in the bronchitic rats; and increased levels of BAL protein and percent neutrophils
were seen in both normal and bronchitic rats. Comparing this same bronchitic model to normal
SD rats, Kodavanti et al. (2000b) observed similar responses to CAPs collected in Research
Triangle Park, NC. That is, at a CAPs concentration of 650 |ig/m3, bronchitic rats had increased
levels of BAL protein and neutrophils compared to CAPs-exposed normal SD rats. To test the
effects on the respiratory system of combined bacterial infections and CAPs exposures, F-344
rats were exposed for 3 h to NYC CAPs at a mean concentration of 225 |ig/m3 (Zelikoff et al.,
2003). The CAPs exposures had little effect on respiratory parameters when they preceded lung
infection with IT-administered Streptococcus pneumoniae; but CAPs exposure of previously-
infected rats caused reductions in basal superoxide, decreased percentages of neutrophils, and
increased bacterial burdens. These CAPs studies most clearly provide indications that exposure
7-156
-------
to ambient PM from several locations for 1 to 6 h/day for 1 to 3 days at concentrations across a
range of-200 to 700 |ig/m3 can cause (a) some lung inflammation in normal and compromised
animals and (b) exacerbation of preexisting respiratory infection.
In a study of combustion emission source materials, Killingsworth et al. (1997) exposed by
inhalation both normal SD and MCT-treated SD rats to Boston ROFA at a concentration of
580 |ig/m3. Consequent respiratory effects included increases in neurotrophils in MCT-treated
rats and increases in MIP-2 mRNA in normal SD rats. Kodavanti's group (1999, 2000b, 2002a)
also completed a number of studies that examined a range of endpoints using a concentration of
15 mg/m3 of Boston or Florida ROFA. One inhalation study (Kodavanti et al., 2002a) used
Boston ROFA inhalation at a concentration of 15 mg/m3 in both WKY and SH rats to compare
normal and cardiovascular compromised animals. Effects seen at this concentration were
increases in PMN, AM, BAL protein, LDH, and lung lesions in both rat strains. Only the WKY
rats showed increased glutathione in this study. This group also completed two inhalation
studies utilizing Florida ROFA, also at a concentration of 15 mg/m3. Comparisons made of
normal SD and MCT-treated rats (Kodavanti et al., 1999) showed that both the normal and
MCT-treated rats display lung lesions at this dose; and both groups displayed increases in BAL
protein, LDH levels, and IL-6 levels. The healthy SD rats also showed increased MIP-2
expression. A subsequent comparison of WKY and SH rats (Kodavanti et al., 2000a) showed
that Florida ROFA had very similar effects on most respiratory parameters in both strains.
Airway hyper-reactivity, lung lesions, AM counts, RBCs in BAL, BAL protein, BAL AM, BAL
oxidants, and IL-6 all increased with ROFA exposure. As in the previous study, only the normal
animals exhibited increased MIP-2 expression. Only one study of combustion source materials
(Dormans et al., 1999) reported completing a dose-response evaluation. This laboratory used
exposures of 0, 10, 30, and 100 mg/m3 CFA and only observed a fibrotic reaction at 100 mg/m3,
thus confirming the relatively inert nature of CFA in comparison to ROFA.
Thus, from among the growing number of animal studies in the literature describing the
respiratory effects of inhaled PM, dose-response characterizations were generally not reported.
So possibilities for reliably estimating LOELs from these data or for attempting extrapolations to
human exposures are limited. Probably of most pertinence, for present purposes, are
(a) indications from several studies that inhalation exposures to CAPs of several species (rats,
hamsters, dogs) for 1 to 6 h/day for 1 to 3 days had little or no effect on pulmonary function, but
7-157
-------
induced some signs of lung inflammation in healthy animals and enhanced inflammatory
responses in chronic bronchitic rats at CAPs concentrations varying across a range of-100 to
1000 |ig/m3 (with the inflammatory responses being most clearly shown in a range of-200 to
700 |ig/m3); and (b) some exacerbation of respiratory infection following acute (3 h) exposure to
New York City CAPs at -225 |ig/m3. On the other hand, analogous but more intensive
inflammatory responses were reported for ROFA responses at a concentration of 15 mg/m3, and
CFA was not found to produce any effects until indications of fibrotic changes were seen at
100 mg/m3.
7.7.1.4 Respiratory Effects of Instilled Particulate Matter
Recent studies characterizing respiratory effects of instilled PM indicate that most effects
are observed for endpoints such as PMN, AM, protein, and LDH accumulation in BAL at a dose
range of 0.7 to 10 mg/kg body weight in rats, mice and hamsters. Changes in cytokines and
oxidant formation have been seen in a similar concentration range in rats. Dose-response
evaluations were carried out in about one third of these studies. To better compare studies, all
doses from the instillation studies were converted to mg/kg body weight in the instances where
researchers reported a dose per animal and an average weight for the animals.
Three studies examined the respiratory effects of instilled ambient PM collected in the
Utah Valley near a steel mill that was closed during 1987. Filters were collected before, during,
and after the closing and the PM was water-extracted for use in the studies. Ohio and Devlin
(2001) intrabronchially instilled the Utah ambient PM in healthy humans at a dose of 0.007
mg/kg and found increases in the cytokines IL-8, TNF, and IL-lp following exposure to the
extracted PM collected while the plant was open. Other parameters increased by this PM
exposure were fibrinogen, fibronectin, PMN, and BAL protein, and tissue factor. Dye et al.
(2001) found analogous increases with this Utah ambient PM sample in male SD rats at much
higher exposure levels. Exposure doses of 3 mg/kg increased BAL LDH, PMN, and total cells
counts, whereas doses of 8 mg/kg increased lung lesions and airway reactivity. Sprague-Dawley
rats exposed to pre- and post-closure Utah PM demonstrated increased BAL PMN and protein at
a concentration of 1.8 mg/kg and increased oxidant formation at a dose of 3.6 mg/kg (Ohio et al.,
1999a).
7-158
-------
The respiratory effects of UAPs were compared to the effects of ambient PM collected
from the Kuwaiti oil fires of 1991 in male Syrian golden hamsters (Brain et al., 1998). Doses of
1.5, 7.5, and 37.5 mg/kg were used; and at the lowest doses, increases in PMNs were observed.
At a dose of 7.5 mg/kg, effects seen were increases in BAL AM, protein and LDH.
Instillation of ROFA has been used in many of the newer studies examining the respiratory
effects of PM. Residual oil fly ash collected at a temperature of 250-300°C, downstream from
the cyclone of a power plant in Florida burning low-sulfur residual oil was the ROFA most
commonly used in the following studies. Kodavanti et al. (1997a) found increases in lung
lesions, PMN, AM, MIP-2, and IL-6 following instillations of 8.3 mg/kg ROFA in male SD rats.
Evaluating possible strain differences, Kodavanti et al. (1997a) found similar inflammatory cell
infiltration and alveolar, airway, and interstitial thickening in SD, Wistar, and F-344 rats at the
same 8.3 mg/kg ROFA concentration. Another comparison of SD and F-344 rats showed
increases in neutrophils in both strains at 8.3 mg/kg (Kodavanti et al., 1996). Florida ROFA at a
slightly higher concentration, 9.4 mg/kg, caused increases in airway hyper-reactivity, AM, PMN,
LDH, and protein in male SD rats (Gavett et al. (1997). The same laboratory (Gavett et al.,
1999) found similar effects in BALB/cJ mice at a dose of 3 mg/kg. Kadiiska et al. (1997)
exposed male SD rats to 3.3 mg/kg Florida ROFA and found increases in both PNM and protein.
A dose-response study was completed by Kovavanti et al. (2001) that showed that much lower
exposures to Florida ROFA could elicit the same effects. At 0.83 mg/kg there were increases in
BAL protein, LDH, and PMN in both WKY and SH rats. In the same study, increases in AM
were seen at 0.83 mg/kg in SH rats and at 3.3 mg/kg in WKY rats. Another rat strain, BN, was
shown to have increased production of LDH at 5.8 mg/kg and increased BAL protein at
1.1 mg/kg Florida ROFA, (Lambert et al., 1999), thus demonstrating some similarity of observed
effects across rat strains. In line with these studies, Madden et al. (1999) found increased
production of acetaldehyde at a concentration of 3.6 mg/kg Florida ROFA. In another study
(Costa and Dreher, 1997), comparing Florida ROFA, domestic oil fly ash (DOFA), CFA, and
four UAPs (St. Louis, Washington DC, Dusseldorf, and Ottawa), instillations of 7 mg/kg caused
increases in PMNs albumin and LDH.
One study reported on respiratory effects of diesel particles (Ghio et al., 2000c) instilled
into SD rats at a concentration of 1.8 mg/kg. Effects seen included increases in BAL protein,
LDH, PMN, MIP-2, TNF, and total cells. A decrease in glutathione was also observed.
7-159
-------
Overall, the studies of respiratory effects of instilled PM materials have provided limited
but interesting information. Perhaps of most importance are the observations of increased levels
of inflammatory indicators in BAL samples from adult humans exposed via instillation to as
little as 0.007 mg/kg (i.e., 7 |ig/kg) of Utah Valley ambient PM10 extract obtained during the
time the nearby steel mill was operating. Analogously, increased oxidant formation and BAL
LDH, PNM and total cell counts were seen in rats instilled with 1.8 or 3.6 mg/kg of the pre-
closure Utah ambient PM. As for other laboratory animal studies, effects in animals were
observed for the commonly assayed respiratory endpoints, in a PM dose range of 0.7 to
10 mg/kg body weight. ROFA was the most commonly used PM for instillation studies,
reflecting a significant data gap for other types of PM and leaving open the question of how
relevant many of the results might be for assessing ambient air PM effects at concentrations
pertinent to current U.S. conditions.
7.7.1.5 In Vitro Effects of Particulate Matter on ng/Cell Dose Basis
A number of the in vitro studies described previously have reported cell numbers used in
the exposures. Based on this information, it is possible to determine the actual dose applied on a
per cell basis. This information is important if any comparisons are to be made across studies
and, further, so that extrapolations between in vitro studies and in vivo studies may be attempted.
In most of these experiments, cells are plated at 1 x 104 to 5 x 106 cells per mL of media, with an
average of about a half million cells per experiment. Studies where cell counts were reported,
but wherein cells were given additional time to proliferate or possibly grow until confluent, were
not used to make PM/cell determinations. Researchers in most cases have carried out dose-
response evaluations, so that LOELs can be determined for the endpoints studied. Overall, there
is some consistency among studies as to concentrations of PM required to elicit effects, most of
which fall within a range across an order of magnitude difference between 0.02 and 0.2 ng/cell.
Urban air particles, ROFA , and CAPS have been most commonly used in these studies.
In studies with exposures of both rat and human AM to UAP from St. Louis, Ottawa, Dusseldorf,
and Florida, effects on cytokine production were consistently seen at doses as low as 0.02 ng/cell
(Becker et al., 1996; Van Eden et al., 2001; Mukae et al., 2000). Washington DC and Boston
UAP were both shown to increase TNF-a production in rat AM at doses of 0.1 ng/cell (Imrich
et al., 2000). Interestingly, it appears that cytokine production is induced at slightly lower
7-160
-------
concentrations in human AM than in rat AM (Becker et al., 1996). Oxidant formation is also
induced by UAP exposure in vitro. St. Louis, Ottawa, and Dusseldorf UAP all were shown to
induce ROS at doses of 0.05 to 0.5 ng/cell in human AM and blood monocytes (Becker et al.,
1996; Becker and Soukup, 1998). Exposure of human AM and blood monocytes to St. Louis
and Ottawa UAP inhibited phagocytosis as doses of 0.5 and 0.02 ng/cell, respectively (Becker
and Soukup, 1998; Van Eeden et al., 2001). Washington DC UAP has been shown to decrease
viability in rat AM at doses of 0.05 ng/cell (Nadeau et al., 1996) and 0.1 ng/cell (Imrich et al.,
2000). Washington, DC UAP has also been shown to increase levels of apoptosis at 0.2 ng/cell
in human AM (Holian et al., 1998) and to deplete ATP at 0.5 ng/cell in rat AM (Nadeau et al.,
1996).
In vitro effects on a wide range of endpoints have been observed with ROFA from Florida
and Boston. Tumor necrosis factor-a has been induced with Florida ROFA in human AM at
0.02 ng/cell (Van Eeden et al., 2001) and with Boston ROFA in mouse RAW 264.7 cells at
0.2 ng/cell. Alabama ROFA has been shown to induce IL-6 production in BEAS-2B cells at
0.08 ng/cell (Oortgeisen et al., 2000). Oxidant formation has been induced by Florida ROFA in
human AM and human blood monocytes at 0.05 ng/cell (Becker et al., 1996; Becker and
Soukup, 1998). A slightly higher concentration of Florida ROFA (0.15 ng/cell)was shown to
induce oxidant formation in rat AM (Ohio et al., 1997a; Becker et al., 1996). Boston ROFA was
found to induce both oxidant formation and inhibition of phagocytosis at 0.1 ng/cell in hamster
AM (Goldsmith et at., 1997; Goldsmith et al., 1998). Florida ROFA in human lung
mucoepidermoid carcinoma cells induced mucin secretion at 0.01 ng/cell and lysozyme at
0.03 ng/cell (Longphere et al., 2000). Other observed effects of Florida ROFA include increased
apoptosis in human AM at 0.025 ng/cell (Holian et al., 1998), increased acetaldehyde production
in BEAS-2B cells at 0.04 ng/cell (Madden et al., 1999), and increased calcium release in BEAS-
2B cells at 0.08 ng/cell (Oortgiesen et al., 2000).
Fewer studies have used in vitro exposures to CAPs. Goldsmith et al. (1998) induced
oxidant formation in hamster AM with Boston CAPs at 0.08 ng/cell. This group (Goldsmith
et al., 1997) also demonstrated oxidant formation and inhibition of phagocytosis in hamster AM
with Boston CAPs at doses of 0.01 ng/cell. One PM10 study was found that reported cell
numbers (Soukup and Becker, 2001). In that study, Chapel Hill PM10 (both the soluble and
insoluble fractions) caused increased production of IL-6, TNF-a, and MCP-1 at
7-161
-------
0.01 ng PM10/cell and inhibition of phagocytosis at 0.05 ng/cell in human AMs. Long et al.
(2001) reported that Boston area PM25 caused increase production of TNF-a at 0.2 ng/cell.
Kennedy et al. (1998) exposed BEAS-2B cells to Provo total suspended particulate (TSP) and
found increased IL-6 release at 2.5 ng/cell and increased IL-8 release at 1.0 ng/cell. They also
exposed human primary tracheal epithelial cells to TSP and saw effects at 0.06 ng/cell, but did
not report dose-response information for these effects.
Thus, it appears that the most commonly studied in vitro endpoints have very similar
LOELs across many types of PM evaluated, which range from about 0.02 to 0.2 ng/cell.
As more in vitro studies are completed with information regarding specific PM exposure
parameters and cell numbers used, clearer patterns should begin to emerge with regard to relative
toxicities by PM class, cell type, and endpoints affected.
7.7.2 Interspecies Comparisons of Experimental Results
7.7.2.1 Introduction
Much of the new toxicologic data assessed in this chapter has been derived from either
(a) in vivo exposures of human subjects or laboratory animals via inhalation exposures or
instillation of PM materials or (b) in vitro exposures of various (mostly respiratory tract) cells or
tissues to diverse types of PM. The experimental exposure conditions used in these studies are
typically different from those experienced through inhalation of airborne PM by human
populations in ambient environments. Most notably, the exposure concentrations used in many
of the experimental studies are well above ambient PM levels. Therefore, consideration of the
relevance of effects demonstrated under experimental conditions compared to the effects
observed in humans following ambient PM exposures is useful, especially to the extent that
quantitative extrapolation of experimental results across species or to ambient conditions may be
feasible based on currently available data.
Appendix 7A provides an analysis of the relationship between rat and human lung doses
predicted for various exposure scenarios ranging from ambient PM exposures to PM instillations
into the lung. In many studies, both toxicologic and epidemiologic, health endpoints are
presented and analyzed as a function of exposure concentration. However, it is generally
accepted that the dose to target cells or tissues, rather than exposure concentration per se, is
responsible for adverse responses. As discussed in Appendix 7A, establishing a firm linkage
7-162
-------
between exposure and dose requires that consideration be given to particle characteristics and
biological normalizing factors. Optimally, the dose metrics and normalizing factors should be
based on the biological mechanisms mediating an effect. For some effects, the mass of
soluble PM depositing in a region of the lung may be an appropriate dose metric. For example,
an appropriate normalizing factor for soluble PM could be the surface area of the airways for
irritants, whereas body mass might be more suitable when considering systemic effects.
There are two principle applications for the dosimetric assessments presented in
Appendix 7A. First, experimental exposure concentrations can be estimated that should result in
the same tissue dose in a rat as received by a human exposed to various levels of ambient PM as
a function of dose metric, normalizing factor, and level of human exertion. As no single dose
metric nor normalizing factor appears to be appropriate for all situations, numerous scenarios
were considered in Appendix 7A. The parameters chosen can dramatically affect the rat
exposure concentration estimated to be required to provide a normalized dose equivalent to that
occurring in a human, as illustrated in Tables 7A-7a through 7A-9b in Appendix 7A. Second,
the dose to the lung can be estimated for both animal and human inhalation studies. These
analyses make it possible to compare biological responses as a function of dose rather than just
exposure. Equal lung doses should not be assumed in comparing studies, even if PM mass
concentrations, animal species, and exposure times are identical because of variations in
individual breathing patterns, lung anatomy, and particle deposition fractions. Differences in the
aerosol size distributions to which animals are exposed also affect dose delivered or retained.
For example, in a comparison of several CAPs studies, one study was estimated to have 1.7
times the alveolar dose of another study despite a 10% lower exposure concentration in the first
study. Thus, to make accurate estimates of dose, it is essential to have accurate and complete
information regarding exposure conditions, i.e., not only concentration and duration of exposure,
but also the aerosol size distribution and the level of exertion (and hence breathing rates) for
exposed subjects.
It is obviously not feasible, given the complexity involved, to attempt extrapolation
modeling for more than a few illustrative health endpoints that were evaluated in the multitude
of studies assessed in this Chapter. Nor would such an effort necessarily be particularly useful
for present purposes. However, providing some modeling results that estimate comparative
exposure concentrations/doses demonstrated experimentally in animal or human studies to be
7-163
-------
effective in producing a few important types of health endpoints should be of value in helping to
provide a context by which to gauge the potential relevance of experimental results for ambient
human exposure conditions.
7.7.2.2 Dosimetric Intercomparison for PMN Influx as a Marker for Lung Inflammation
Various types of particulate materials (both ambient PM and combustion source particles)
have been shown to cause inflammation of the lung by migration of PMNs (predominantly
neutrophils) into the airways. These cells are initially produced by bone marrow and, along with
AM, constitute an important defense mechanism triggered by invasion of PM, bacteria, or some
other foreign matter. The PMNs, once in the lung, ingest PM and then may degranulate, forming
hydrogen peroxide and superoxide anions. Excessive quantities of PM in the lung can cause the
lysosomal enzymes in PMNs to enter the extracellular fluid, creating further inflammatory
responses. Additionally, PMNs produce thromboxanes, prostaglandins, and leukotrienes.
Three recent studies provide data on PMN increases following CAPs exposure that allow
comparison of rat to human responses. Kodavanti et al. (2000b) exposed both healthy SD rats
and rats with SO2-induced bronchitis to CAPs collected in Research Triangle Park, NC. The
particle size distribution in this study averaged 0.98 jim MMAD (og = 1.41), the average
concentration was 740 |ig/m3, and exposures consisted of whole-body inhalation of 6 h/day for
2 or 3 days. Inflammation was assessed immediately after exposure or 18 h later. Increases in
BAL PMNs were seen only in the CAPs-exposed bronchitic rats compared to healthy
CAPs-exposed rats in 2 of 4 separate experiments when rats were lavaged immediately
postexposure. The healthy CAPs-exposed rats had no significant differences in PMN counts
compared to healthy air-exposed rats. In a similar study, Clarke et al. (1999) exposed healthy
and SO2-induced bronchitic rats to Boston CAPs at an average concentration of 515 |ig/m3.
Particle size averaged 0.18 jim MMAD (og = 2.9) and exposures consisted of whole-body
inhalation for 5 h/day for 3 consecutive days. PMN levels were assayed 24 h after exposure.
Increases in PMNs (both in terms of total PMN counts and as PMN as percent of total cell count)
in both the normal and bronchitic rats were seen with CAPs exposure. It is possible to compare
these two rat studies to a human study, wherein Ohio et al. (2000a) exposed human subjects to
Chapel Hill CAPs. In that study, healthy human volunteers were exposed to -120 |ig/m3 for 2 h,
with 15 minute periods of exercise alternating with 15 minutes of rest. Particle size was 0.65 jim
7-164
-------
MMAD (og = 2.35) and BAL analysis was at 18 h PE. Consistent with data from the rat studies,
the total numbers of PMNs increased with the human CAPs exposure.
Appendix Section 7A.7.2 compares tissue doses predicted to occur in human and rat CAPs
exposures using the Ohio et al. (2000a), Kodavanti et al. (2000b), and Clarke et al. (1999)
studies. The Kodavanti et al. (2000a) study consisted of five separate CAPS experiments and the
retained CAPs determinations were made from the experiment that utilized the 18 h PE time
point. Comparisons of these rat and human studies indicate that in order to obtain the noted
similarities in PMN responses observed, rats actually received a far greater alveolar region dose
than humans. That is, 60 to 500% increases in PMN numbers were observed in the rat studies
with estimated retained alveolar surface area doses of 28 to 47 |ig/m2 of CAPs PM; whereas a
300% increase in PMNs was seen in humans with estimated doses of only 0.7 |ig/m2 of alveolar
tissue. This suggests that even healthy humans may be more susceptible to the inflammatory
effects of CAPs than are rats. Table 7-14, allows more specific comparisons. Note that the
dosimetry model used for these calculations considers only the insoluble component of PM.
Consideration of the soluble fraction of the PM would, of course, create a more complete picture
of the differences between rats and humans. Interpretation of these data regarding increases in
PMN should also be tempered by the caveats that percent increase reported are influenced by
basal levels of PMN and that composition, concentrations, and size distributions of CAPs can
vary substantially from place to place and from day to day even at the same location.
TABLE 7-14. CAPs: RAT AND HUMAN INHALATION STUDY COMPARISONS
Study
Kodavanti
et al. (2000a)
Clarke et al.
(1999)
Ohio et al.
(2000a)
Species
SDrat
SO2-SD
SDrat
S02-SD
humans
Exposure
Cone.
Particle (ug/m3)
RTF 740
CAPs
Boston 515
CAPs
Chapel 47
Hill
CAPs
MADD
K)
0.98
(1.41)
0.18
(2.9)
0.65
(2.35)
Exposure
duration
6 h/day for
2-3 days
5 h/day
for 3 days
2h
Analysis
PE Change in PMN
<3h 255% t PMN
in 2 of 4 exp
1 8 h (bronchitic rats
only) no change
in PMN
24 h 500%! PMN
367% IPMN
18 h 267%! PMN
Estimated
alveolar dose
per surface
area
ND
28 ug/m2
retained
47 ug/m2
retained
0.7 ug/m2
retained
7-165
-------
7.7.2.3 Inhibition of Phagocytosis by PM Exposure
Phagocytosis is a form of endocytosis wherein bacteria, dead tissue, or other foreign
material (e.g., inhaled ambient particles) are engulfed by cells such as AM, MO, or PMN as part
of normal lung defense mechanisms. Hence, increases in numbers of AM, MO, or PMN cells in
lung tissue represent one indicator of mobilization of lung defenses in response to infection or
deposition of inhaled particles. Once ingested by AM, lysosomes act to digest engulfed
materials. Inhibition of the phagocytosis by AM would signal interference with lung defense
mechanisms by which inhaled bacteria and viruses are killed or other foreign particles are
detoxified and/or cleared from the lung. Also, if an AM is overwhelmed by the amount or
toxicity of ingested material, that material may be released along with the AM's digestive
enzymes onto the alveolar surface and numbers of AM or their phagocytic activities may
decrease.
A number of experimental (especially in vitro) studies have demonstrated, that in some
instances, one or another type of PM has caused an inhibition of phagocytosis. As with other
endpoints affected by PM, this inhibitory effect is determined by the size and composition of the
specific particulate materials tested.
For example, Becker and Soukup (1998) exposed human AM to UAP from St. Louis
(0.2 to 0.7 jim MMAD) and ROFA from Florida (0.5 jim MMAD). Exposures periods were
18 to 20 h at 100 jig/mL per 2 x 10s cells/mL for a dose per cell of 0.5 ng/cell. AM had a 50%
decrease in phagocytosis of Saccharomyces cerevisiae with St. Louis UAP and a 30% decrease
with ROFA, which the authors attributed to the toxicity of ROFA. The authors noted decreased
phagocytosis in cells with both high and low particle burden, and further, that inhibited
phagocytosis was more pronounced in the cells with a low burden. They attributed this effect to
soluble fine constituents of the UAP more so than to particle-bound insoluble constituents.
These researchers (Soukup and Becker, 2001) extended these findings with human AM
exposures to Chapel Hill CAPs. They separated the CAPs into PM2 5 (soluble and insoluble
components) and PM10, (soluble and insoluble) and exposed 2><105 cells/mL to 12.5, 25 or
100 |ig/mL for 18 h. Phagocytosis was then assayed with fluorescein-tagged, zymosan particles.
They observed a dose-dependent decrease in uptake with the insoluble PM10 (12% at
12.5 |ig/mL, 30% at 25 |ig/mL, and 50% at 100 |ig/mL). This correlates with doses per cell of
0.06, 0.12, and 0.5 ng/cell, respectively. There was a similar percentage of AM that appeared to
7-166
-------
be undergoing apoptosis. They postulated that the decrease in phagocytosis is due to the cells
undergoing programmed cell death. In another study by this group, Soukup et al. (2000) found
decreased phagocytosis of yeast particles in human AM exposed to Provo PM10 (Utah Valley
Dust) collected before a steel plant closed. Exposures of 100 |ig/mL to 2x 10s cells/mL (for a
dose of 0.5 ng/cell) caused a 30% decrease in phagocytosis. Particles collected during and after
the steel mill closure did not cause a similar change in phagocytosis, even though the amount of
particles engulfed was the same for all samples of the dust. They suggested that the metal
content may not be predictive of decreases in AM phagocytic responses. In another study,
Van Eeden et al. (2001) studied human AM exposed to UAP (Ottawa) and ROFA (Florida), both
reported to be < 10 jim in diameter, and found an inhibition of phagocytosis at 100 ug/mL,
which they attribute to toxicity to the cells. At 24 h PE, phagocytosis was determined by visual
inspection of the cells. Cells were plated at a concentration of 0.5 x 106 and phagocytosis was
decreased at a dose of 0.2 ng/cell.
These in vitro studies of human AM may be compared to three available studies that
investigated animal AM responses to vitro PM exposures. Any conclusions drawn from these
comparisons must be tempered with the understanding that the data obtained from differing cell
types, culture conditions, and PM species have inherent limitations. Renwick et al. (2001) used
a mouse macrophage cell line (J774.2 M<3>) to evaluate inhibition of phagocytosis by both fine
and ultrafme particles. They used fine carbon black CB (260.2 nm diameter), ultrafine CB
(UCB, 14.3 nm), fine TiO2 (250.0 nm) and ultrafine TiO2 (UTiO2, 29.0 nm). The cultured cells
at a concentration of 5* 106 cells/mL were exposed to particles at concentrations of 15.6, 31, 63,
or 125 |ig/mL for 8h, after which phagocytosis was assessed using 2 jim fluorescent latex beads.
Phagocytosis was inhibited by UCB at 63 |ig/mL and by all the particles at 125 |ig/mL, which
corresponds to doses of 0.013 and 0.025 ng/cell, respectively.
Goldsmith et al. (1997) exposed hamster AM to Boston CAPs (1 |im) or Boston ROFA
(0.1 to 2.5 |im) for 30 minutes. They measured right angle light scatter to determine cell
granularity, as an indicator of phagocytosis. At concentrations of up to 20 |ig/mL of CAPs and
200 |ig/mL of ROFA, they observed no inhibition of phagocytosis. They listed two limitations
of this type of assay to quantify phagocytosis; (1) it provides only a relative measure, not
absolute numbers or mass of particles and, (2) it requires cells to be in suspension; whereas in
the previously mentioned studies, the AM are adherent and thus capable of functioning in a more
7-167
-------
realistic manner. Further, as the exposure was only 30 minutes, it is difficult to compare their
results to those of studies using exposures over 18 h.
To make comparisons between rodent and human studies investigating the inhibition of
AM phagocytosis by PM, an understanding of the species-specific differences in AM should be
noted. Appendix 7A Section 7.3 discusses rat AM function versus volumetric loading. The
volume, and presumably the capacity, of AM in rodents are smaller than for human AM.
A human AM has an internal volume (not including the cell nucleus) of-1350 jim3, compared to
the SD rat (1010 |im3), F344 rat (760 |im3), hamster (420 |im3), or mouse (370 |im3; Miller,
2000). The phagocytic activity of an AM is thought to slowly decrease above a particle loading
of-6% of its interior volume. At about 60% loading, alveolar macrophages become immobile.
Considering the phagocytosis of a single particle, a human AM would likely become
immobilized following the ingestion of a 11.6 |im diameter particle (0.816 ng assuming unit
density), whereas a mouse AM could only ingest a 7.5 jim diameter particle (0.221 ng, unit
density). Typically, considerably smaller particles than these (7.5 to 11.6 jim particles) deposit
in the alveolar region of the lung and become phagocytosed by AM. In addition to the volume
occupied by engulfed particles, another 30% or more of an AM capacity is lost to void spaces
between particles packed within an AM . Table 7-15 compares several in vitro studies of human
and rodent AM function and makes estimations of the AM loads based on reported PM
characteristics.
Only one study was found that used both rat and human AM to compare the effects of PM
on AM phagocytosis. Seemayer et al. (1990) exposed AM isolated from BAL to UAP from
Duisburg (F.R.G). Both species demonstrated a reduction in phagocytic activity (% cells with
> 2 particles) and phagocytic capacity (particles per cell), with little effect on cell viability.
Of note, this study indicated a greater inhibition of phagocytosis in human AM compared to rat
AM by both criteria, suggesting that human AM are more sensitive to the effects of PM than rat.
Unfortunately, given that the study only reported the volume of air from which particles were
collected and not particle mass, size, or composition, it is not possible to compare the data with
more recent studies. In summary, the above comparisons provide interesting results suggesting
that human AM may be at least as sensitive to ambient PM as to ROFA.
Whereas two studies reported no change in AM phagocytosis with exposures of human
cells to doses of Ottawa ambient PM or ROFA (Florida) ranging up to 0.4 ng/cell, other studies
7-168
-------
TABLE 7-15. INTERSPECIES COMPARISONS OF PARTICLE EFFECTS ON ALVEOLAR
MACROPHAGE PHAGOCYTOSIS
Study Species
Becker and human
Soukup(1998)
Soukup et al. human
(2000)
Soukup and human
Becker (2001)
Goldsmith hamster
etal. (1997)
VanEeden et al. human
(2001)
Renwick et al. mouse
(2001) macrophage
cell line
J774.2 MO
PM
UAP (St. Louis)
ROFA (Florida)
CAPs (Chapel Hill
- separated into
soluble and
insoluble
components
PM10 (Utah
Valley)
CAPs (Boston)
ROFA (Boston)
UAP (Ottawa)
ROFA (Florida)
CB
UCB
TiO2,
UTiO2
Cone
100 ug
per 2 x 105 cells
12.5
25
100 ug
per 2 x 105 cells
100 ug
per 2 x 106 cells
4ug
10
20
25
50
100
200
per 0.5 xlO6 cells
10 ug
100 per
O.SxlO6 cells
15, 31,63, or
125 ug
per 5 xlO6 cells
Exposure Change in
duration Particle size phagocytosis
18-20h 0.2-0.7 urn 150%
0.5 130%
18 h 2.5 or 10 urn 112%
130%
150% with
insoluble PM10
only
overnight 10 urn 130%
30 min 1 um no change
0.1-2.5
2,4,8,12,24 h < W ^im no change
(only 24 h data 1
reported)
8h 0.260 urn 1@125 ug
0.014 1@63
0.250 1@125
0.029 1@ 125
Estimate
dose/cell
0.5 ng
0.06 ng
0.12
0.5
0.05 ng
0.008 ng
0.02
0.04
0.05
0.1
0.2
0.4
0.02 ng
0.2
0.025 ng
0.013
0.025
0.025
Estimated
% of cell filled
53
6.3
13
53
5.3
2.7
6.9
14
17
34
69
140
2.1
21
9.7
5.0
9.7
9.7
CB = carbon black
UCB = ultrafine carbon black
TiO2 = titanium dioxide
UTiO, = ultrafine titanium dioxide
-------
showed decreased phagocytosis in human AM's exposed to 0.05 to 0.5 ng/cell of Chapel Hill
CAPs, Utah Valley PM10 extract, St. Louis ambient PM, or Florida ROFA. However, additional
more systematic work is necessary to fully characterize the phagocytic dose-response to various
species of PM. The efficient removal of inhaled PM by viable, functioning AM cells is a critical
respiratory defense mechanism, one thought likely to be impaired by at least some types of
ambient PM constituents.
7.8 MUTAGENICITY/GENOTOXICITY EFFECTS
The majority of newly-published PM research since the 1996 PM AQCD have focused on
acute cardiovascular or respiratory effects associated with short-term exposure to ambient PM or
selected constituents. However, the new epidemiologic analyses by Pope et al. (2002) not only
further substantiate associations between long-term exposure to ambient PM and increases in
cardiopulmonary mortality but also provide the strongest evidence yet linking such PM
exposures to lung cancer. In view of these new ambient PM-carcinogenicity findings (and others
from earlier epidemiologic studies), salient results both from some older studies (pre-1996 PM
AQCD) and newly available ones are discussed below with regard to evaluations of mutagenic
or other genotoxic effects of ambient PM, its constituents, and/or combustion emission source
particles thought to be useful as indices of likely carcinogenic potential of such materials. The
pertinent studies discussed below are summarized in Tables 7-16, 7-17, and 7-18.
7.8.1 Ambient Particulate Matter Effects
A limited number of new in vitro studies have examined the mutagenic and/or other
genotoxic potential of ambient PM from various geographic locations in the U.S. or elsewhere;
and, in general, they show some evidence that appears to support the biologic plausibility of lung
cancer effects being causally related to long-term exposure to ambient PM, as implied by the
epidemiologic findings.
The World Health Organization (1993) has found that the induction of sister chromatid
exchanges (SCE) is a sensitive cytogenic endpoint for the demonstration of genotoxic activity of
environmental mutagens and carcinogens. In vitro SCE assays using various types of human or
laboratory animal cells have been used in new studies, along with other techniques, to evaluate
7-170
-------
TABLE 7-16. MUTAGENIC/GENOTOXIC EFFECTS OF AMBIENT PARTICIPATE MATTER
Species, gender,
strain age, or
body weight
Human hlAlv2
Human hlAlv2
Cultured tracheal
epithelial cells
from Hamster
(Syrian golden,
young) or rat
Human
bronchioepithelial
cell line
(BEAS-2B)
Particle or Exposure
Constituent Technique
Ambient PM from in vitro
Los Angeles, San
Nicolas Island,
Long Beach, Azusa,
and Rubidoux, CA
Composite of in vitro
ambient PM from
Los Angeles, San
Nicolas Island,
Long Beach, Azusa,
and Rubidoux, CA.
Fractionated into
nonpolar, polar and
semipolar
components
Ambient PM: in vitro
industrial or high
traffic areas
(Germany)
Urban PM2 5 in vitro
Urban PM10
Industrial PM25
Industrial PM10
Rural PM25
Rural PM10
(Germany)
Concentration Particle
Dose Characteristics
(ug/mL) Size (urn); ng
120 ug HOC per 12 < 2um
mL assay
300 - 1200 ug HOC <2\im
per 12 mL assay
Not given Dichloromethane
extraction of high
volume samples.
6.6 to 26.5 ug/mL Dichloromethane
1.7 to 6.9 ug/mL extraction of coarse
10.8 to 43.2 ug/mL (PM10) and fine (PM25)
5.8 to 23.1 ug/mL fractions.
7.8 to 34.1 ug/mL
3.7 to 14.4 ug/mL
Effects of Particles on
Mammalian Cells or
Exposure Duration Bacteria
72h No seasonal variation in
mutagenic potency, leading
authors to suggest that
mutagenicity is due to
ubiquitous emission sources
like vehicle traffic or
stationary combustion rather
than isolated point sources.
LA air had 10 times more
mutagenicity than
background levels (San
Nicolas)
72 h Most of the mutatenic
potency was in the
unsubstituted PAC fraction.
2-nitrofluoranthene and 6H-
benzo [cd] pyrene were
semipolar mutagens.
Dilutions of extracted Dose-related increases in
organic phase of particles sister chromatid exchanges
incubated with cells for seen in both species. PM
48 h. from industrial sample had
LOELofO.ll m3air/mL
medium. PM from high
traffic area had LOEL of
0.16 m3 air/mL medium.
Dilutions of extracted Significant increases in sister
organic phase of size- chromatid exchanges were
segregated particles greater from all sampling
incubated with cells for sites at all doses of PM10 and
72 h. PM2 5 from urban and
industrial regions.
Extraction phase of coarse
particles produced fewer
sister chromatid exchanges
than did the fine particles.
Reference
Hannigan,
etal. (1997)
Hannigan
etal. (1998)
Hornberg
etal. (1996)
Hornberg
etal. (1998)
-------
TABLE 7-16 (cont'd). MUTAGENIC/GENOTOXIC EFFECTS OF AMBIENT PARTICIPATE MATTER
to
Species, Gender,
Strain Age, or
Body Weight
Kidney cells from
hamster
(Syrian golden,
8-10 weeks old)
Cultured
hepatoma cells
Liver tumor cell
line (HEPAlclc7)
Particle or Exposure
Constituent Technique
Ambient PM from in vitro
urban and industrial
areas
(Germany)
Ambient PM in vitro
(Netherlands)
Urban air PM in vitro
DEP
Rubber industries
PM
Metal industries PM
Poultry/swine farm
Compost
Concentration
Dose
(ug/mL)
Not given
Not given
6-12 ug
17-37 ug
36-47 ug
32-175 ug
81-137 ug
42 ug
Particle
Characteristics
Size (urn); (ig
Dichloromethane
(DCM) extraction of
high volume samples.
Acetone/DCM
extraction of high
volume samples.
Aqueous and organic
extraction from filters
of particles collected
with high volume
samplers.
Exposure Duration
Dilutions of extracted
organic phase of particles
incubated with cells for
18 h folio wed by
infection with simian
virus SV-40.
Dilutions of extracted
organic phase of particles
incubated with cells for
6 or 48 h.
4h
Effects of Particles on
Mammalian Cells or
Bacteria
Significantly greater SV-40-
induced transformation of
hamster kidney cells
pre-treated with organic
extractions of urban
particles/extracted from
4 m3 air.
Extracts of ambient PM both
upwind and downwind of
highway had genotoxic
effects, although PAH
content was greater in the
downwind samples.
Inhibition of gap-junctional
intercellular communication
(GJIC) only significant in
cells treated with aqueous
extract of diesel, compost, or
rubber particles.
Reference
Seemayer and
Hornberg
(1998)
Hamers
(2000)
Alink(1998)
PAC = polyaromatic compounds.
-------
TABLE 7-17. MUTAGENIC/GENOTOXIC EFFECTS OF WOOD AND COAL COMBUSTION-SOURCE PM
Species, gender,
strain age, or
body weight
Salmonella
Salmonella:
TA98
TA100
Salmonella'.
TA98
Human WBC
Salmonella'.
TA98
TA100
Salmonella'.
TA100
Particle or
constituent
Emissions
from wood
(birch, pine,
and spruce)
combustion
Wood, diesel,
and coal
emissions
Emission
from open
fireplaces
Wood smoke
condensate
(Sigma)
Emission PM
- collected
throughout
year from
burning fields
Exposure
Technique
in vitro
Ames assay
in vitro
Ames assay
(comparing
standard
plate and
spiral
assays)
in vitro
Ames assay
32P-post-
labelling
analysis of
DNA
adducts
in vitro
Ames assay
in vitro
Ames assay
Particle
Mass Cone Characteristics Exposure
ug/mL or u,g/m3 Size (u,m) Duration
32 to 100 ug/m3 of PM and organic 48 h
PM and 2. 6 to fractions from
200 ug organics wood stoves
combustion
200 ug/plate 72 h
woodsmoke,
500 ug/plate DE
and 200 ug/plate
coal
PM extracted 1 week
with methanol
0,125,250,500, 48 h
750 and
100 ug/plate
130 units ug/m3 PM25 and 48 h
(winter) to
15 (summer),
170 (winter),
37 (summer) PM10
Effects of particles on mammalian
cells or bacteria
Organic fraction: mutagenic potency of
0.5-21 revertants/ug. The PM fraction
demonstrated only very low
mutagenicity
DE had greatest mutagenicity under all
conditions, creating both frameshift and
base-pair substitution mutations. Coal
just slightly less mutagenic than diesel
creating indirect-acting frameshift
mutations. Woodsmoke only weakly
mutagenic.
Control: 28 revertants/30 m3 (-S9) and
69 (+S9).
Combustion: 153 (-S9) and 369 (+S9).
No change in DNA adducts.
Not mutagenic at all doses
PM2 5: 30.6 revertants/m3 air volume
(winter) to 0.1 (summer) with increased
mutagenicity with S9 activation.
PM10: 28.1 (winter) to 0.7 (summer),
revertants/m3 air volume, S9 activation
increasing the mutagenicity.
Reference
Lofroth et al.
(1986)
Houk et al.
(1991)
Heussen et al.
(1994)
Putnam et al.
(1999)
Vinitketkumnuen
et al. (2002)
-------
TABLE 7-17 (cont'd). MUTAGENIC/GENOTOXIC EFFECTS OF WOOD AND COAL COMBUSTION-SOURCE PM
Species, gender,
strain age, or
body weight
Salmonella'.
TA98
TA100
Salmonella:
TA98
TA100
TA1535
TA1537
TA1538
Salmonella'.
TA98
TA100
TA104
Particle or
constituent
Wood
burning
emission PM
and gas phase
Coal fly-ash
from
fluidized-bed
(FBC) and
conventional
combustion
(CC) plants
Extracts from
smoky coal,
China
Exposure Mass Cone
Technique ug/mL or u,g/m3
in vitro
Ames assay
in vitro
Ames assay
in vitro
Ames assay
PCR and
DNA
sequencing
Particle
Characteristics Exposure
Size (u,m) Duration
Two smoke 48 h
samples
collected, PM
and gas phase
< S^m 72 h
FBC mean
diameter 0.54 |im
CC mean
diameter 1.05 |im
Effects of particles on mammalian
cells or bacteria
12 x 106 revertants/kg using TA100-S9
and 3.5 * 106 revertants/kg using
TA98-S9. Emissions can cause both
frameshift and base pair substitution
mutations. The gas phase of the wood
smoke emission contributed to more than
60% of the direct-acting mutagenicity
FBC mutagenic in TA98 (3.32
revertants/mg) and TA 1538 (3.31), both
without activation. S9 decreased the
mutagenicity of FBC in TA98 and TA
1538. FBC had no mutagenic response
inTA1537andTA1535.
TA98 + S9 mutagenic at > 10 ng/plate;
TA98-S9 not mutagenic.
Mutation spectrum hotspot.
TA100 + S9 mutagenic at > 10 ug/plate;
TA100-S9 at > 50 ng/plate. Mutation
spectrum: GC - TA or GC - AT
trans versions. TA104: no mutagenicity
Reference
Kim Oanh et al.
(2002)
Munford and
Lewtas(1982)
Granville et al.
(2003)
-------
TABLE 7-18. MUTAGENIC/GENOTOXIC EFFECTS OF MOBILE COMBUSTION-SOURCE PM
Species, gender,
strain age, or
body weight
Salmonella
Salmonella'.
TA98
TA100
Salmonella'.
TA98
TA98NR
Mutagenicity:
S. typhimurium
Micro some assay
Cytotoxicity:
Mouse fibroblast
cell line < 1.292
Particle or Exposure Mass Cone
constituent Technique ug/mL or u,g/m3
PM in vitro
diesel Ames assay
gasoline
PM from in vitro
Diesel, Ames assay
Gasoline,
Gasoline +
alcohol,
liquified
petroleum
Fractionate in vitro Particle emission
exhaust of Ames assay values for the
gasoline and vehicles were
diesel engines 0.021g/kmand
0.23 g/km
Diesel in vitro Not given
exhaust Ames assay
particles
(DEP):
petroleum
DEP vs.
rapeseed oil
methyl ester
(RME) DEP
Particle
Characteristics Exposure
Size (u,m) Duration
48 h
48 h
48 h
Dichloromethane 48 h
extraction of incubation
particles with TA98
collected from and TA 100
diesel engine run strains.
with diesel fuels
with low or high
sulfur and plant
oil fuel.
Effects of particles on mammalian
cells or bacteria
DE mutagenic response was 800 (PM
fraction) revertants/g fuel used and
210 (condensate fraction). Gasoline 24
(PM fraction) and 39 (condensate).
LP cars: 10 rev/L exhaust.
Gasoline and gasoline + alcohol:
10-50 rev/L exhaust).
Light-duty diesel: 50-250 rev/L7
Most polar subfraction was most
mutagenic
TA98-S9: 7.8 rev/m3(g); 6.1 (d).
TA98+S9: 3.3 (g);1.5 (d).
TA98NR-S9: 3.7 (g); 4.1 (d).
TA98NR response generally lower than
TA98 response. Both had similar
TA98NR-S9 response, but differed
significantly in TA98-S9 response.
Revertants were 2- to 10-fold higher
with high sulfur diesel fuel particles.
Cytotoxicity in fibroblast cells higher
for RME.
Reference
Lofroth(1981)
Rannug(1983)
Strandell et al.
(1994)
Hunger (2000)
-------
TABLE 7-18 (cont'd). MUTAGENIC/GENOTOXIC EFFECTS OF MOBILE COMBUSTION-SOURCE PM
Species, Gender,
Strain Age, or
body weight
Salmonella'.
TA98
TA100
Salmonella:
TA98
Calf thymus
DNA
Particle or Exposure
constituent Technique
PM and in vitro
SVOC of Ames assay
exhaust from
diesel and
gasoline
engines
PM and in vitro
SVOC of Ames assay
exhaust from
diesel and Adduct
gasoline formation
engines
Particle
Mass Cone Characteristics Exposure
ug/mL or u,g/m3 Size (u,m) Duration
25-500 ng/plate
Ames assay:
30-500 ng/PM
(-S9) 10-1000
Hg/PM (+S9)
Adduct: 150
Hg/PM gasoline
extracts.
High doses
(42-150 ng/PM)
and low doses
(7.5-18.5 ng/PM)
of gasoline and
diesel extracts
Effects of particles on mammalian
cells or bacteria
Mutagenicity rankings: TA98: current
diesel at 30 °F > high emitter diesel
> gas emitting white smoke > normal
gasoline 72 °F > normal diesel 72 °F
> gas emitting black smoke.
TA100: current diesel at 30 °F > gas
emitting white smoke > high emitter
diesel > normal diesel 72 °F > gas
emitting black smoke > normal gasoline
30 °F > normal gasoline 72 °F.
Gas SVOC fraction less mutagenic than
PM fraction, but formed more DNA
adducts. Diesel PM and gasoline
extracts formed more S9-mediated
adducts with increasing doses, but no
dose response. Diesel extracts formed
higher levels of adducts than gasoline
extracts, especially in the presence of
XO. Results suggest high nitro-PAH
levels in diesel extract.
Reference
Seagrave et al.
(2002)
Pohjola et al.
(2003)
-------
the genotoxic potential of ambient PM samples, ambient PM constituents and/or PM emission
source constituents. A caveat to interpretation of these data is that there is not a simple linear
relationship between mutagenic potential and carcinogenic potential in animals or humans.
Additionally, not all Ames assays are equivalent in terms of predicting mutagenicity. These
studies, listed in Table 7-18, have focused mainly on the ability of the organic fraction of
ambient PM to induce mutagenic effects in mammalian cell lines and bacteria.
Probably of most direct relevance and usefulness for assessing U.S. ambient air
carcinogenic potential, Hannigan et al. (1997) examined the mutagenicity of PM from five
monitoring sites in southern California. San Nicolas Island in upwind Los Angeles was
considered to be a background site with low levels of PM. Central Los Angeles was
characterized as a region of high PM resulting from heavy vehicle traffic. Long Beach was
another high PM site studied with PM originating from power plants and oil refineries. The two
other high PM sites chosen were Azusa and Rubidoux, which were considered receptor sites
located downwind from high density primary emission sources. Mutagenic activity of air
samples collected in 1993 were assayed using a cultured human cell assay in addition to the
standard Ames bacterial mutation assay, which has limitations in terms of relevancy to human
mutagenicity. The human cell assay utilizes hi Alv2 cells, which test mutagenic activity at the
thymidine kinase locus. The cells contain a plasmid pHSRAA with two copies of human
CYP1A1 cDNA, which confers resistance to 1-histidinol. CYP1A1 is a cytochrome P450
capable of activation of PAH. Air samples were collected throughout the year at all sites using a
dichotomous sampler. Both seasonal and spatial differences in component elemental and
organic carbon were observed. However, both the human cell mutagenicity assay and a
Salmonella TM677 forward mutation assay showed no systematic seasonal pattern of changes in
mutagenicity. These results suggested to the authors that the proportion of mutagenic
compounds in the fine organic aerosol mass does not change throughout the year and that
perhaps the emission sources that show seasonal variation do not contribute in a major way to
the mutagenicity of the PM. They thusly concluded that, in the Los Angeles area, primary
particulate emissions from sources that operate on a year-round basis are the important human
cell mutagens. Further, since they found very similar mutagenic potencies at all four widely-
separated high PM sites, they suggested that the mutagenicity is due most likely to ubiquitous
emissions sources rather than to isolated point sources.
7-177
-------
To ascertain which components of the Los Angeles area PM were responsible for the
observed mutagenicity, Hannigan et al. (1998) extended these findings by combining the four
high PM samples and a background site sample described above into a composite sample that
was then separated by liquid chromatography into fractions of organic chemicals of similar
polarity and functionality. A primary fractionation separated the composite sample into four
fractions, designated nonpolar 1, nonpolar 2, semipolar, and polar. To further isolate the
mutagens, additional fractionation steps were done by HPLC. The mutagenic potency of the
unfractionated sample was 150 induced mutant fraction (IMF) per mass of fine particulate
organic carbon or IMF (x 106)/mg of EOC. They found that six unsubstituted polyaromatic
compounds (PACs) were responsible for much of the mutagenicity. These included
benzo[k]fluoranthene, indeno[l,2,3-cd]pyrene, benzo[b]fluoranthene, benzo[g,h,i]perylene,
benzo[a]pyrene, and cyclopenta[cd]pyrene. Benzo[k]fluoranthene and benzo[b]fluoranthene
sources include vehicle exhaust and natural gas combustion. The source of the other four PACs
is mainly gasoline and diesel vehicles. Additionally, two semipolar mutagens were identified:
2-nitrofluoranthene and 6H-benzo[cd]pyren-6-one. The first compound is a product of
atmospheric chemical reactions and the other is emitted by noncatalyst gasoline-powered
vehicles. The authors estimated that greater than half of the mutagenicity may be attributed to
the semipolar and polar fractions of the sample.
Some additional evidence for mutagenetic properties of ambient PM derive from several
European studies. For example, Hornberg et al. (1996), evaluated genotoxic effects on cultured
rodent (rat; Syrian golden hamster) tracheal epithelium cells exposed in vitro to ambient PM
collected on hi-vol (TSP) sampler filters during winter 1991 in a heavily industrialized city
(Duisburg) or in another area (Dusseldorf) of Germany dominated by high density vehicular
traffic. Exposure to ambient PM extracted (by dichloromethane or DCM) from filters from both
types of locations induced highly significant dose-dependent increases in SCE in the tracheal
cells of both rodent species. The authors noted that it was remarkable that even quantities of
chemical substances equivalent to airborne PM from just 0.11 to 3.5 m3 air for the samples from
the heavy industry area and from 0.16 to 10.2 m3 for the heavy traffic area induced significant
genotoxic effects (i.e., ~2-fold increases in SCE).
Hornberg et al. (1998) also evaluated genotoxic effects on human tracheal epithelial cells
of fine (PM25) and coarse (PM10) fractions of ambient PM collected during winter, 1996 on
7-178
-------
dichotomous sampler filters in an urban area (Diisseldorf), an industrial area (Duisburg) and a
rural area (Borken) of Germany. Both ambient PM10 and especially PM25 extracted (by DCM)
from filters for all three areas significantly increased SCE in the human bronchioepithelial cell
line (BEAS-2B) cultured in vitro for a 72 h exposure. The authors noted that the fine
fraction (PM2 5) exerted stronger genotoxic activity than the PM10 from a given area and that,
whereas the Diisseldorf and Duisburg ambient PM materials had comparable genotoxic activity,
samples from the rural area (Borken) showed lower genotoxicity. The fine fraction PM2 5
(equivalent to airborne PM substances from < 0.5 m3 of air) exerted strong genotoxicity.
The PM2 5 and PM10 extracted PM from the filters were reported to have been drawn from
ambient air having concentrations of: 18.4 and 4.8 |ig/m3 for Diisseldorf; 45 and 24.1 |ig/m3 for
Duisburg, and 21.8 and 10 |ig/m3 for Borken, respectively (all of which were in 1 mL of medium
for exposures).
Based on the above results, Hornberg et al. (1996, 1998) concluded that the increases
observed in SCE of tracheal epithelium cells with in vitro exposures to ambient PM materials are
indicative of genotoxic activity of such materials and increased risks for humans due to such
genotoxicity activity. However, insufficient information was provided by which to estimate the
actual exposure doses to the cell cultures in the Hornberg studies. Nevertheless, their results still
appear to provide qualitative evidence for mutagenic effects of ambient PM (especially the fine
fraction drawn from heavily industrialized or trafficked areas). The authors also noted that the
tracheobronchial epithelium is the site of one of the most common cancer in humans, i.e.,
bronchogenic carcinoma (Tomatis, 1990).
Further evidence for the likely carcinogenic potential of ambient PM, in addition to the
above findings, is derived from a study by Seemayer and Hornberg (1998), which employed a
bioassay for enhancement of malignment cell transformation in vitro. Exponentially growing
cell cultures from the Syrian golden hamster were exposed for 18 h to varying concentrations
of PM materials extracted (by DCM) from hi-vol sampler filters that collected ambient PM from
Diisseldorf or Duisburg, Germany in winter, 1990. Control and PM-exposed cultures were then
infected with the papovivarus simian virus (SV-40). There was a strong dose-dependent
enhancement of cell transformation frequency in the kidney cell cultures as a function of varying
pretreatment concentrations of ambient PM extracts. Inoculation of transformed cells into
syngeneic animals produced a high percentage of malignant tumors, mostly sarcomas. Positive
7-179
-------
control cultures pretreated with benzo[a]pyrene (BaP) showed similar dose-dependent
enhancement of malignant cell transformations. The authors also noted that the human
papovaruses BK and JC are ubiquitous and infect a large proportion of human populations
worldwide (Monini et al., 1995); and that interactions of environmental carcinogens and viruses
should be considered in human carcinogenesis.
Using a different type of bioassay from Hornberg and colleagues, Hamers et al. (2000)
evaluated the genotoxicity of ambient PM collected by hi-vol sampler at several sites in the
Netherlands: (1) one site next to a highway traffic point (density of vehicle passages/day =
63 x 103); (2) another next to a higher density (93 x 103 vehicle passages/day) traffic point; and
(3) a third in a natural conservation area (with extensive non-manured grasslands and cattle
grazing) thought to have background levels of diffuse air pollution. Extracts of PM filter
materials, collected from each of these sites in 1997 and/or 1998, were tested for genotoxic
activity in the umu-assay (using S. typhimurium). Arylhydrocarbon-receptor activation was also
assessed by DR-CALUX-assay, using a stably transfected H4IIE hepatoma cell line. Extracts of
ambient PM collected downwind from the highway (west-wind) traffic points had increased
genotoxicity that appeared to be attributable at least in part to polycyclic aromatic hydrocarbons
(PAHs) from traffic exhaust. The extracts of ambient PM collected upwind of the highway
(eastern wind) had a different composition of compounds (probably including some transported
from nearby Germany), with higher genotoxicity less related to highway-emitted PAH-like
compounds. Of interest, even the rural site ambient PM extracts showed some genotoxic
activity. The authors concluded that their results showed that the presence of pollutants with
genotoxic or PAH-like characteristics pose an undesirable mutagenic risk.
In another study using a less conventional endpoint, Alink, et al. (1998) compared effects
on gap-junctional intercellular communications (GJIC) in liver tumor (HEPAlclc?) cells of
in vitro exposures to PM from urban air (geographic area not stated), rubber and metal
industries, diesel exhaust, and biological sources (i.e., poultry/pig farming, compost industry).
Only diesel and rubber sample filter extract suspensions significantly inhibited GJIC, with up to
83% of the inhibition attributed to the particles per se. More active organics were reported to
have been extracted from the rubber industry particles than from the diesel particles by organic
solvents. The authors interpreted their results as suggesting that cancer promoting potential (as
7-180
-------
indexed by GJIC inhibition) may vary widely depending on particle source and type, possibly
due to the particles per se or to surface-bound bio-active material.
Taken together, the results of the above studies provide new evidence indicative of
ambient PM (especially the fine fraction) having mutagenic properties, thus supporting the
plausibility of epidemiologic evidence linking ambient PM (especially fine PM) to lung cancer.
The results further suggest likely contributions to the observed mutagenicity of ambient PM of
industrial or motor vehicle combustion sources (which are important emission sources for
fine PM). The ensuing subsections discuss studies that evaluated the mutagenic/genotoxic
potential of several types of major combustion sources known to contribute to ambient PM
(especially fine PM) in many U.S. regions.
7.8.2 Wood and Coal Combustion-Source Effects
Emissions from the combustion of wood and coal, as well as combustion of oil fuels
(diesel and gasoline) by mobile source vehicles, all contribute to ambient PM. A number of
studies have been done to evaluate the mutagenicity and genotoxicity of these combustion
emissions and to compare their relative mutagenic/genotoxic potentials. Table 7-17 summarizes
wood/coal combustion studies, discussed first below. These include some earlier studies,
conducted prior to the 1996 PM AQCD, given the only very limited more recent evaluation of
wood/coal combustion mutagenic/genotoxic effects.
7.8.2.1 Biomass/Wood Burning
Early studies by Lofroth et al. (1986) used the Ames Salmonella assay to determine the
mutagenicity of emissions from wood (birch, pine, and spruce) burned in conventional wood
stoves. Both the PM fraction and the condensable organic fraction were applied in doses of
0.6 to 4.3 liter flue gas per plate (for a range of 32 to 100 mg/m3 of PM and 2.6 to 200 mg
organics). The wide range of doses was due to use of both updraft and downdraft stoves, the
latter generating far less combustion emissions. The organic fraction had a mutagenic potency
of 0.5 to 21 revertants/ug (rev/jig). The PM fraction demonstrated only very low mutagenicity.
The authors compared these results to earlier studies by Lofroth (1981) and Rannug (1983) of
gasoline and diesel cars. On a revertant per hour basis, wood stoves produced 6 x io6, gasoline
cars 0.5 to 3 x io6, and diesel cars 3 to 20 x io6 rev/jig.
7-181
-------
In testing the spiral Salmonella assay, Houk et al. (1991) compared the mutagenicity of
wood smoke, automotive diesel exhaust and coal combustion emission. This automated
bacterial mutagenicity assay dispenses Salmonella, the test agent, and the S9 mix in a spiral
pattern on an agar plate, creating a uniform density of bacteria and a gradient of test agent.
Doses of 200 jig/plate woodsmoke, 500 jig/plate DE and 200 jig/plate coal emission were used
on standard Ames assay plates to compare the two assays. Exposures to strains TA98 and
TA100, both with and without metabolic activation by S9, showed that DE had the highest
mutagenicity. Results from the TA98 and TA100 experiments suggested to the authors that the
mutagenic activity is due to both nitrated polynuclear aromatics creating frameshift mutations
and nonpolar compounds creating base-pair substitution mutations. Coal was just slightly less
mutagenic than diesel; and the data suggested that indirect-acting frameshift mutations were
occurring, which is in agreement with Mumford et al. (1987). Woodsmoke was found to be only
weakly mutagenic in both strains.
Heussen et al. (1994) collected respirable PM from homes in Wageningen, NL, a region
with no significant industrial pollution. For a 1-month control period, the fireplaces were not
used in five homes. Wood was then burned in open fireplaces for 4h/day during the evening for
1 week. PM was collected during these same periods in the homes and, also, from some outdoor
sampling sites to correct for possible infiltration of ambient mutagens into the homes.
Nonsmoking subjects from these homes gave blood samples during the control period, at the
beginning of the combustion period, at the end of the combustion period, and 1 week later.
Blood was assayed using 32P-postlabeling analysis of DNA adducts from white blood cells. PM
samples were assayed by the Ames test (using strain TA98), both with and without S9 activation.
The mutagenicity of samples from all 5 homes was increased after the week of fireplace usage.
Control values averaged 28 revertants/30 m3 (-S9) and 69 (+S9), whereas combustion samples
averaged 153 (-S9) and 369 (+S9), indicating stronger indirect mutagenicity. However, there
was no correlative combustion-related increase in formation of DNA adducts in white blood
cells. The authors suggest several reasons for a lack of correlation between the two endpoints:
(1) the exposure was too short and/or the dose was too low; (2) white blood cells may not be a
suitable cell type for detection of adducts with this type of exposure; and (3) the actual genotoxic
damage that occurred my not be detectable by this method of 32P-postlabeling analysis.
7-182
-------
The mutagenicity and toxicity of wood smoke condensate was assayed by Putnam et al.,
(1999) using the Ames assay and neutral red uptake, respectively. The wood smoke condensate
was prepared from hardwoods by distillation and filtration, removing 'insoluble tars' and low
boiling point substances. The wood smoke condensate was tested for cytoxicity at
concentrations of 0, 10, 25, 50, 75, 100, and 150 |ig/mL and found to be toxic beginning at
25 |ig/mL. Mutagenicity was tested using S. typhimurium strains TA98 and TA100, both with
and without S9 activation. Concentrations of 0, 125, 250, 500, 750, and 100 jig/plate were used
and the wood smoke condensate was found to be nonmutagenic at all the doses.
More recent studies have examined the mutagenicity of biomass combustion in Chaing
Mai, Thailand (Vinitketkumnuen et al., 2002). Large open fires created by farmers burning
fields and grass in the winter months correlate with increased in PM at that time of year.
Twenty-four hour PM2 5 and PM10 samples were collected on Teflon filters at four outdoor
sampling sites over a period of 1 month and then pooled, dissolved in 1 mL DMSO and used for
the Ames assay (TS100 strain/0.05 mL sample per assay). Monthly averages of PM25 ranged
from -130 |ig/m3 (winter) to 15 |ig/m3 (summer) and for PM10, 170 |ig/m3 to 37 |ig/m3,
respectively. Mutagencity, expressed as number of revertants/m3 air volume, for PM25 ranged
from 30.6 (winter) to 0.1 (summer), with the mutagenicity being increased with S9 activation.
For PM10 the mutagenicity ranged from 28.1 (winter) to 0.7 (summer) revertants/m3 air volume,
again with S9 activation increasing the mutagenicity. One of the collection sites showed a much
higher mutagenicity level than the other three, which might be explained by a local source of
diesel exhaust. Mobile source emissions contribute to ambient PM levels in Chaing Mai, but as
they remain constant throughout the year, the authors suggested that the increased mutagenicity
observed in the winter months is due to biomass combustion.
Kim Oanh et al. (2002) examined the mutagenicity and toxicity of emissions from various
cooking sources, including wood and kerosene. The wood (Pterocarpus indicus) was burned in a
single-stage ceramic cookstove and two samples were collected, a PM phase consisting of
the PM collected on the filter and the PM rinsate and a gas phase consisting of XAD-2, the
condensate knockout and the rinsate. Toxicity was assayed using a Microtox bioassay and
mutagenicity was assayed using the Ames test with TA98 and TA100 strains, both with and
without metabolic activation by S9. The highest mutagenicity factor was observed from wood
fuel which produced 12 x 106 revertants/kg using TA100-S9 and 3.5 x 106 revertants/kg using
7-183
-------
TA98-S9. These results indicate that the wood smoke emissions can cause both frameshift and
base pair substitution mutations. The gas phase of the wood smoke emission contributed more
than 60% of the direct-acting mutagenicity.
7.8.2.2 Coal Combustion
Coal fly ash samples (< 3 jim) from a southern U.S. conventional combustion (CC) plant
and from a Linden, NJ fluidized-bed combustion (FBC) plant (both burning Pennsylvania
eastern bituminous coal) were collected by Mumford and Lewtas (1982). They compared the
mutagenicity of PM from the two sources using the Ames assay with test trains TA98, TA100,
TA1535, TA1537, and TA1538. FBC was mutagenic in TA98 (3.32 revertants/mg) and TA
1538 (3.31), both without S9 metabolic activation. S9 decreased the mutagenicity of FBC in
TA98 and TA 1538. FBC had no mutagenic response in TA1537 and TA1535, with or without
S9. Thus, FBC fly ash appeared to create direct-acting frameshift mutations. In all of the five
strains utilized, the CC fly ash demonstrated no mutagenicity.
Studies characterizing the health effects of coal emissions have focused mainly on the
mutagenicity and carcinogen!city of coal smoke exposure in regions of China with a
predominance of indoor burning of "smoky" coal (Mumford et al., 1987; Chapman et al., 1988;
Mumford et al., 1999). These regions have very high rates of lung cancer mortality that have
been linked to exposure to unvented coal smoke.
Mumford et al., (1987) collected indoor air samples from homes burning smoky coal and
from homes burning wood and smokeless coal in open hearths in kitchens. The PM levels inside
the homes burning smoky coal averaged 23 mg/m3, compared to 1.8 mg/m3 for homes burning
smokeless coal. The distribution of PM size in the homes burning smoky coal was bimodal,
with half the particles < 1 |im and half 1 |im to 10 jim, whereas particle size in wood-burning
homes ranged from 1 to 30 jam. Fractionation of the filter extracts created aliphatic, aromatic,
moderately polar, and polar components. High concentrations of organic matter were present in
the smoky coal (72 to 82%) compared to 27% for the smokeless coal and 55% for the wood
sample. Further, the highest PAH levels were found in smoky coal samples. Both neat samples
and fractions were tested for mutagenicity by the Ames assay using strain T98, with and without
metabolic activation by S9. Most of the samples required S9 activation for mutagenicity,
suggesting to the authors the presence of PAH. Smoky coal samples had the highest
7-184
-------
mutagenicity (60-17 revertants/m3 air) compared to wood (11) and smokeless coal (1.3). The
fractions that displayed the most mutagenicity were the polar and aromatic, which were shown
by GC/MS to consist primarily of nitrogen- and oxygen-containing compounds (polar fraction)
and PAH, methylated PAH, and nitrogen heterocyclic compounds (aromatic fraction).
A retrospective epidemiologic study (Lan et al., 2002) evaluated the incidence of lung
cancer in a cohort of farmers born in Yunnan Province. The farmers were raised in homes
burning smoky coal in unvented firepits and 81% changed to homes utilizing stoves with
chimneys that reduced indoor levels of PM10 (2.08 mg/m3 to 0.71 mg/m3, respectively). After
stove improvement, Lan et al. observed a long-term reduction in lung cancer incidence,
calculating risk ratios of 0.59 in men and 0.54 in women.
Very recent work (Granville et al., 2003) has focused on the mutation spectra of coal
smoke emissions from the Yunnan Province. Smoky coal extracts from the same source as
above (Mumford et al., 1987) at doses of 0, 10, 25, 50, and 100 jig/plate were used in the Ames
assay with strains TA98, TA100, and TA104, both with and without S9 activation. Molecular
analysis of the revertants was then done to identify the mutations. Coal smoke extract was
mutagenic in TA98 in the presence of S9 at doses > 10 jig/plate and not mutagenic without S9
activation. The extract was mutagenic in TA100 with S9 at doses > 10 jig/plate and without S9
at doses > 50 jig/plate, but it was not mutagenic in the TA104 strain. The authors interpreted
these results to suggest that the coal extract induced mutations primarily at GC sites and that
PAHs were probably involved in the mutations because of the greater mutagenicity in
TA100+S9 compared to TA98+S9. The mutation spectrum in TA98 showed that the extract
induced only the hotspot mutation, which is a 2-base deletion in an 8-base GC repeat. This
suggested to the authors that about 70% of the mutations in TA98 were due to standard PAH
compounds in the coal smoke. The mutation spectrum in TA100 showed that most of the
mutations were GC -» TA or GC^ AT transversions. The authors then compared these mutation
spectra with KRAS and TP53 mutation spectra observed in lung tumors from nonsmoking
women exposed to coal smoke emissions. They found similarities in the GC -» TA transversions
in TA100 with human mutations due to PAH exposures.
Thus, overall, the above findings link exposures to smoky coal in China to human cancer.
However, notable differences exist between the high concentrations of combustion products of
7-185
-------
smoky coal to which the Chinese populations are exposed versus the much lower exposures to
coal combustion emissions products to which U.S. populations are currently exposed.
7.8.3 Mobile Combustion-Source Effects
Numerous studies have linked mutagenic/carcinogenic effects to diesel and gasoline
exhaust and/or to particles contained therein, as summarized in Table 7-18 and discussed below.
7.8.3.1 Diesel
Results such as those noted in Table 7-18 add further to an extensive database on diesel-
related mutagenicity that was thoroughly reviewed in the 2002 U.S. EPA Diesel Document (U.S.
EPA, 2002) alluded to earlier. Important information drawn from that document's evaluation of
diesel-related mutagenic properties is recapitulated below (at times verbatim) with particular
emphasis on findings bearing on the role of PM components of diesel exhaust.
As noted in the 2002 Diesel Document, use of mutagenicity data as an approach to
evaluating potential carcinogenicity of diesel emissions is based on the premise that genetic
alterations are found in all cancers and that several of the chemicals found in diesel emissions
possess mutagenic activity in a variety of genetic assays. These genetic alterations can be
produced by gene mutations, deletions, translocations, aneuploidy, or amplification of genes.
hence no single genotoxicity assay should be expected to predict carcinogenicity. Also, because
of the inherent biological differences of measured endpoints, both within genotoxicity assays and
between genotoxicity assays and cancer bioassays, a direct extrapolation should not be expected.
Indeed, most genotoxicity data are generated with in vitro assays that frequently employ test
agent concentrations orders of magnitude greater than encountered in environmental situations.
With diesel emissions or other mixtures, other complications also arise due to the complexity of
the materials tested.
Since 1978, more than 100 publications have been reported for genotoxicity assays used
with whole diesel emissions (DE), the volatile and paniculate (DPM) fractions (including
extracts), or individual chemicals found in diesel emissions. Interest in the contribution of
mutagens to carcinogenicity was high in the early 1980s and the lack of long- term rodent
carcinogenicity data for DE led to use of semiquantitative mutagenicity (and in vitro cell
transformation) data for DE to augment epidemiology studies of diesel-related carcinogenic
7-186
-------
effects. The number of chemicals in diesel emissions is very large; and many have been shown
to exhibit mutagenic activity in a variety of assay systems (see Claxton, 1983). Among some of
the mutagenically active compounds found in the gas phase of diesel exhaust are ethylene,
benzene, 1,3 -butadiene, acrolein and several PAHs, all of which are also present in comparable
or greater amounts in gasoline exhaust. Of the diesel particle-associated chemicals, several
PAHs and nitro-PAHs have been the focus of mutagenic investigations both in bacteria and in
mammalian cell systems.
Numerous studies have evaluated mutagenic effects of DE and/or DPM. In one early
study, Huisingh et al. (1978) showed that dichloromethane extracts from DPM were mutagenic
in strains TA1537, TA1538, TA98, and TA100 of S. typhimurium, both with and without rat
liver S9 activation, based on data from several fractions as well as DPM from different vehicles
and fuels. Similar results with diesel extracts from various engines and fuels were reported by
several others using the salmonella frameshift-sensitive strains TA1537, TA1538, and TA98
(Siak et al., 1981; Claxton, 1981; Dukovich et al., 1981; Brooks et al., 1984). Mutagenic activity
was also seen in Salmonella forward mutation assays measuring 8-azaguanine resistance
(Claxton and Kohan, 1981) and inE1. coli mutation assays (Lewtas, 1983).
One approach to identifying significant mutagens in chemically complex environmental
samples (e.g., DE or ambient PM extracts) is the combination of short-term bioassays with
chemical fractionation (Scheutzle and Lewtas, 1986). The analysis is most frequently carried out
by sequential extraction with increasingly polar or binary solvents. Fractionation by silica-
column chromatography separates compounds by polarity or into acidic, basic, and neutral
fractions. The resulting fractions are difficult to characterize by chemical methods, but the
bioassay analysis can be used to determine fractions for further analysis. In most applications,
salmonella strain TA98 without the addition of S9 has been used as the indicator for mutagenic
activity.
Generally, a variety of nitrated polynuclear aromatic compounds have been found that
account for a substantial portion of the mutagenicity (Liberti et al., 1984; Schuetzle and Frazier,
1986; Schuetzle and Perez, 1983). However, not all bacterial mutagenicity has been identified in
this way, and the identity of the remaining mutagenic compounds remains unknown. The
nitrated aromatics thus far identified in diesel engine exhaust (DE) were the subject of review in
an IARC monograph on DE (International Agency for Research on Cancer, 1989). In addition to
7-187
-------
qualitative identification of mutagenic chemicals, several investigators have used numerical data
to express mutagenic activity as activity per distance driven or mass of fuel consumed. These
types of calculations have been the basis for estimates that the nitroarenes (both mono- and
dinitropyrenes) contribute a significant amount of the total mutagenic activity of the whole
extract (Nishioka et al., 1982; Salmeen et al., 1982; Nakagawa et al., 1983). More recently,
Crebelli et al. (1995) used salmonella to examine the effects of different fuel components. They
reported that although mutagenicity was highly dependent on aromatic content, especially di- or
triaromatics, there was no clear effect of sulfur content of the fuel. Later, however, Sjogren et al.
(1996), using multivariate statistical methods with ten diesel fuels, concluded that the most
influential chemical factors in salmonella mutagenicity were sulfur content, certain PAHs
(1-nitropyrene), and naphthenes.
Matsushita et al. (1986) tested particle-free DE gas and benzene nitroderivatives and
PAHs, identified as components of DE gas. The particle-free exhaust gas was positive in both
TA100 and TA98, but only without S9 activation. Of the 94 nitrobenzene derivatives tested,
61 were mutagenic and most showed greatest activity in TA100 without S9; whereas 28 of
50 PAHs tested were mutagenic, all required the addition of S9 for detection, and most appeared
to show a stronger response in TA100. When 1,6-dinitropyrene was mixed with various PAHs
or an extract of heavy-duty (HD) DE, the mutagenic activity in TA98 was greatly reduced when
S9 was absent but increased significantly with S9 present. These latter results suggest that
caution should be used in estimating mutagenicity (or other toxic effects) of complex mixtures
from the specific activity of individual components.
Mitchell et al. (1981) reported mutagenic activity of DPM extracts of diesel emissions in
the mouse lymphoma L5178Y mutation assay. Positive results were seen both with and without
S9 activation in extracts from several different vehicles, with mutagenic activity only slightly
lower in the presence of S9. These findings were confirmed in a numerous other mammalian
cell systems using several different genetic markers. Casto et al. (1981), Chescheir et al. (1981),
Li and Royer (1982), and Brooks et al. (1984) all reported positive responses at the HPRT locus
in Chinese hamster ovary (CHO) cells. Morimoto et al. (1986) used the APRT and Ouar loci in
CHO cells; Curren et al. (1981) used Ouar in BALB/c 3T3 cells. In all of these studies,
mutagenic activity was observed without S9 activation. Liber et al. (1981) used the thymidine
kinase (TK) locus in the TK6 human lymphoblast cell line and observed induced mutagenesis
7-188
-------
only in the presence of rat liver S9 when testing a methylene chloride extract of DE. Barfknecht
et al. (1982) also used the TK6 assay to identify some of the chemicals responsible for this
activation-dependent mutagenicity; and they suggested that 1-methylphenanthrene,
9-methylphenanthrene, and fluoranthene could account for over 40% of the observed activity.
Specific-locus mutations were not induced in (C3H x 101)F1 male mice exposed to
DE 8 h/day, 7 days/week for either 5 or 10 weeks (Russell et al., 1980). The exhaust was a
1:18 dilution and the average particle concentration was 6 mg/m3. After exposure, males were
mated to T-stock females and matings continued for the reproductive life of the males. The
results were unequivocally negative; no mutants were detected in 10,635 progeny derived from
postspermatogonial cells or in 27,917 progeny derived from spermatogonial cells.
Additional evidence for cytotoxic and mutagenic effects of particles emitted from diesels
comes from a study by Bunger et al. (2000). Filter sample particles, collected from diesel
emissions generated by a tractor engine during combustion of conventional petroleum diesel fuel
or diesel fuel containing rapeseed oil methyl ester (RME), were extracted by DCM and their
cytotoxicity was then evaluated by the neutral red assay and their mutagenicity by the
S. typhimurium assay. The diesel petroleum fuel emissions had much higher numbers of smaller
particles than the RME emissions. However, 4-fold stronger toxic effects on mouse fibroblast
cells were exerted by RME extracts from filters taken at "idling" but not at "rated" power load
modes. Both types of extracts were significantly mutagenic at both load modes in both the TA98
and TA100 strain bioassays, but the petroleum fuel extracts had 4-fold more mutagenic effect in
the TA98 and 2-fold more in the TA100 strain assays than did RME extracts. The authors
attributed the lower mutagenic potency of the RME diesel emissions to lower sulfur and PAH
content in the RME emissions.
Hou et al. (1995) measured DNA adducts and hprt mutations in peripheral lymphocytes of
47 bus maintenance workers and 22 control individuals. All were nonsmoking men from
garages in the Stockholm area; the exposed group consisted of 16 garage workers, 25 mechanics,
and 6 other garage workers. There were no exposure data, but the three groups were considered
to be of higher to lower exposure to diesel engine exhaust, respectively. Levels of DNA adducts
determined by 32P-postlabeling were significantly higher in workers than controls (3.2 versus
2.3 x 10"8), but not hprt mutant frequencies (8.6 versus 8.4 x 10"6). Although group mean mutant
frequencies were not different, both adduct level and mutagenicity were highest among the
7-189
-------
16 most exposed, and mutant frequency was significantly correlated with adduct level.
All individuals were genotyped for glutathione transferase GSTM1 and aromatic amino
transferase NAT2 polymorphism. Neither GSTM1 nulls norNAT2 slow acetylators exhibited
effects on either DNA adducts or hprt mutant frequencies.
Driscoll et al. (1996) exposed Fischer 344 male rats to aerosols of CB (1.1, 7.1, and
52.8 mg/m3) or air for 13 weeks (6 h/day, 5 days/week) and measured hprt mutations in alveolar
type II cells in animals immediately after exposure and at 12 and 32 weeks after the end of
exposure. The two higher exposures caused significant increases in mutant frequency. Whereas
the mutant frequency from the 7.1 mg/m3 group returned to control levels by 12 weeks, that of
the high-exposure group was still higher than controls even after 32 weeks. Carbon black
particles have very little adsorbed PAHs; hence a direct chemically induced mechanism is highly
unlikely. Induction of hprt mutations were also seen for rat alveolar epithelial cells after
intratracheal instillation with CB, quartz, and TiO2 (Driscoll et al., 1997). All three types of
particles elicited an inflammatory response as shown by significant increases of neutrophils in
BAL fluid. The neutrophils in BAL are the source of ROS. DNA damage resulting from ROS is
a secondary genotoxicity, and this effect is seen only at high doses. Culturing the BAL from
exposed rats with a rat lung epithelial cell line also resulted in elevation of hprt mutational
response. This response was effectively eliminated when catalase was included in the incubation
mixture, providing evidence for cell-derived oxidative damage. The oxidative damage of CB
appeared to be a threshold exposure dose-response phenomenon.
Recently, Sato et al. (2000) exposed male Big Blue transgenic F344 rats to diluted DE
(1 and 6 mg/m3 suspended particle concentration) for 4 weeks. Mutant frequency in lung DNA
was significantly elevated (4.8* control) at 6 mg/m3 but not at 1 mg/m3. Lung DNA adduct
levels measured by 32P-postlabeling and 8-hydroxydeoxyguanosine measured by HPLC were
elevated at both particle concentrations, but to a lesser extent than mutant frequencies. Sequence
analysis of mutants indicated that some, but not all, of the mutations could be explained by an
oxidative damage mechanism.
Other diesel studies have evaluated chromosome effects. Mitchell et al. (1981) and Brooks
et al. (1984), for example, reported increased SCE in CHO cells exposed to DPM extracts of
emissions from both LD and HD diesel engines. Morimoto et al. (1986) observed increased SCE
from both LD and HD DPM extracts in PAH-stimulated human lymphocyte cultures. Tucker
7-190
-------
et al. (1986) exposed human peripheral lymphocyte cultures from four donors to direct DE for
up to 3 h. Samples were taken at 16, 48, and 160 min of exposure. Cell cycle delay was
observed in all cultures; and significantly increased SCE levels were reported for two of the four
cultures. Structural chromosome aberrations were induced in CHO cells by DPM extracts from
a Nissan diesel engine (Lewtas, 1983) but not by similar extracts from an Oldsmobile diesel
engine (Brooks et al., 1984).
DPM dispersed in an aqueous mixture containing dipalmitoyl lecithin (DPL), a component
of pulmonary surfactant or extracted with DCM induced similar responses in SCE assays in
Chinese hamster V79 cells (Keane et al., 1991), micronucleus tests in V79 and CHO cells
(Gu et al., 1992), and unscheduled DNA synthesis (UDS) in V79 cells (Gu et al., 1994). After
separating the samples into supernatant and sediment fractions, mutagenic activity was confined
to the sediment fraction of the DPL sample and the supernatant of the DCM sample. These
findings suggest that the mutagenic activity of DPM inhaled into the lungs could be made
bioavailable through solubilization and dispersion of pulmonary surfactants. In a later study in
the same laboratory, Liu et al. (1996) found increased micronuclei in V79 cells treated with
crystalline quartz and a noncrystalline silica, but response was reduced after pretreatment of the
particles with the simulated pulmonary surfactant.
Guerrero et al. (1981) observed a linear concentration-related increase in SCE in lung cells
cultured after intratracheal instillation of DPM at doses up to 20 mg/hamster. However, they did
not observe any increase in SCE after 3 mos of inhalation exposure to DE particles at 6 mg/m3.
Also, Pereira et al. (198la) exposed female Swiss mice to by inhalation DE 8 h/day, 5 days/week
for 1, 3, and 7 weeks. The incidence of micronuclei and structural aberrations was similar in
bone marrow cells of both control and exposed mice.
Pereira et al. (1982) measured SCE in embryonic liver cells of Syrian hamsters. Pregnant
females were exposed to DE diluted with air 1:9 to contain about 12 mg/m3 particles from days
5 to 13 of gestation or injected intraperitoneally with diesel particles or particle extracts on
gestational day 13 (18 h before sacrifice). Neither the incidence of SCE nor mitotic index was
affected by exposure to DE. The injection of DPM extracts but not DPM resulted in a dose-
related increase in SCE; however, the toxicity of the DPM was about 2-fold greater than the
DPM extract.
7-191
-------
In a study using mammalian germ cells, Russell et al. (1980) reported no increase in either
dominant lethals or heritable translocations in males of T-stock mice exposed by inhalation to
DE. In the dominant lethal test, T-stock males were exposed for 7.5 weeks and immediately
mated to females of different genetic backgrounds. There were no differences from controls in
any of the parameters measured. For heritable translocation analysis, T-stock males were
exposed for 4.5 weeks and mated to (SEC * C57BL/6) females, and the Fl males were tested for
the presence of heritable translocations. Although no translocations were detected among
358 progeny tested, the historical control incidence is < 1/1,000.
A number of studies have measured other types of genotoxic effects (e.g., increased DNA
adducts) in animals exposed to DPM, CB or other particles, as reviewed by Shirname-More
(1995). Although modest increases in DNA adducts have been observed in lung tissue of rats
after inhalation of DPM (Wong et al., 1986; Bond et al., 1990), the increases are small in
comparison with those induced by chemical carcinogens present in DE (Smith et al., 1993).
While Gallagher et al. (1994) found no increases in total DNA adducts in lung tissue of rats
exposed to DE, CB, or titanium dioxide, they did observe an increase in an adduct with
migration properties similar to nitrochrysene and nitro-benzo(a)pyrene adducts from diesel but
not CB or TiO2 exposures. The majority of the studies used the 32P postlabeling assay to detect
adducts. Although this method is sensitive, chemical identity of adducts can only be inferred if
an adduct spot migrates to the same location as a known prepared adduct.
DNA adducts have also been measured in humans occupationally exposed to DE. Distinct
adduct patterns were found among garage workers occupationally exposed to DE compared to
nonexposed controls (Nielsen and Autrup, 1994). Furthermore, the findings were concordant
with adduct patterns observed in groups exposed to low concentrations of PAHs from
combustion processes. Hemminki et al. (1994) also reported significantly elevated levels of
DNA adducts in lymphocytes from garage workers with known DE exposure compared with
unexposed mechanics. Hou et al. (1995) found elevated adduct levels in bus maintenance
workers exposed to DE. Although no difference in mutant frequency was observed between the
groups, the adduct levels were significantly different (3.2 versus 2.3 x 10"8). Nielsen et al.
(1996) reported significantly increased levels of three biomarkers (lymphocyte DNA adducts,
hydroxyethylvaline adducts in hemoglobin, and 1-hydroxypyrene in urine) in DE-exposed bus
garage workers.
7-192
-------
The role of oxidative damage in causing mutations has received increasing attention. More
than 50 different chemicals have been studied in rodents usually measuring the formation of
8-hydroxydeoxyguanosine (8-OH-dG), a highly mutagenic adduct (Loft et al., 1998). Dose-
dependent increases in that mutagenic DNA adduct were found in mouse lung DNA after
intratracheal instillation of diesel particles (Nagashima et al., 1995). Mice fed on a high-fat diet
showed an increased response, whereas the responses were partially reduced when the
antioxidant, p-carotene, was included in the diet (Ichinose et al., 1997). Oxidative damage also
has been measured in rat lung tissue after intratracheal instillation of quartz (Nehls et al., 1997)
and in rat AMs after in vitro treatment with silica dust (Zhang et al., 2000). Arimoto et al.
(1999) found that redissolved methanol extracts of DPM also induced the formation of 8-OH-dG
adducts in L120 mouse cells. The response was dependent on both DPM concentration and
P450 reductase. The potential role of oxidative damage in DE carcinogenesis is discussed in
more detail in the U.S. EPA Diesel Document (U.S. Environmental Protection Agency, 2002).
7.8.3.2 Gasoline
In addition to the above studies of DE and DPM effects, other studies have also evaluated
mutagenic/genotoxic effects of gasoline combustion emissions and/or compared the potencies of
such emissions to DE or DPM potencies.
In an early study, Lofroth (1981) compared the mutagenic activity of PM from diesel and
gasoline engine exhaust and found both to be mutagenic in the Ames assay, in the absence of
mammalian metabolic activation. Both paniculate and condensate fractions were tested.
Expressed in units of revertants/g fuel used, diesel exhaust mutagenic response was 800 (PM
fraction) and 210 (condensate fraction), which was far greater than the mutagenic response of
gasoline (24 and 39, respectively). In another older study, Rannug (1983) collected both
particulate and gas phase components from motor vehicle exhaust from medium- and heavy-duty
diesel vehicles and light-duty cars burning gasoline and other fuels. The Ames assay was used
with strains TA98 and TA100, both with and without S9 activation. The particulate phase of the
exhaust created < 20,000 revertants/km (corresponding to 10 rev/L exhaust) in cars burning
liquified petroleum and cars with catalysts, classified by the authors as the low mutagenicity
group. Light-duty diesels produced > 100,000 rev/km with the highest effect of up to
700,000 rev/km seen with TA100-S9 (corresponding to 50-250 revertants/L exhaust) and were
7-193
-------
classified by the authors as the high mutagenicity group. Engines burning gasoline or gasoline-
alcohol fuels created exhaust from which the particulate phase gave 10,000 to 20,000 rev/km (or
10-50 rev/L exhaust). In general, the newer vehicles tested produced exhaust with slightly less
mutagencity than older models. Also, more mutagenesis was seen in exhaust from cold starts
(0 °C) than in starts at 23 °C.
Strandell et al. (1994) fractionated the extracts of gasoline and diesel exhaust from Volvos
to find the most potent mutagens among the subtractions. Particle emission values for the
vehicles were 0.021 g/km and 0.23 g/km, respectively. Mutagenicity testing was done with the
Ames assay with strain TA98, both with and without S9 metabolic activation, and with strain
TA98NR, the nitro reductase-deficient strain used to determine the presence of nitro aromatic
mutagens. The subtraction that was most polar also demonstrated the most mutagenicity (51%
of the total mutagenicity for gasoline and 39% of the total for diesel). This fraction contained
low-boiling point components and some phenol derivatives. Quantitatively, this subfraction's
mutagenicity using TA98-S9, TA98+S9 and TA98NR-S9 was 7.8, 3.3 and 3.7 rev/m3,
respectively, for gasoline and 6.1, 1.5, and 4.1 for diesel. The TA98NR response was generally
lower than the TA98 response, which the authors suggested was due to the presence of nitro-
PAH in the fractions. Both fuels had a similar TA98NR-S9 response, but differed significantly
in their TA98-S9 response, which suggests difference in some nitro-reductase-dependent
mutagens. A reduction in mutagenicity was observed with the addition of S9 activation, which
the authors attribute to enzymatic deactivation of direct-acting mutagens or possible activation or
deactivation of unknown compounds.
A more recent study (Seagrave, et al., 2002) using vehicles including automobiles, SUVs
and pickup trucks from 1976 to 2000 evaluated the genotoxicity of gasoline and diesel emissions
from normal vehicles, high emitters, and gasoline vehicles emitting smoke. Both PM and
semivolatile organic compound (SVOC) fractions were collected, both at room temperature and
in a cold environment. The PM and SVOC fractions were recombined and tested for
mutagencity using the Ames assay with strains TA98 and TA100, both with and without S9
activation. All of the samples caused mutations in both strains with doses from 25 to
5000 jig/plate (LOEL not given). With most samples evidence pointed to a direct-acting
mutagenesis effect due to the results of TA98 both with and without S9 activation. The assay
using TA100 showed greater mutagenicity in the exhausts from the high emitter diesel, the white
7-194
-------
smoker gasoline, and the black smoker gasoline. As seen in the Rannug (1983) study, emissions
samples collected from cold engines were more mutagenic than those collected at room
temperature. The authors ranked the mutagenic potency based on the TA98 results as: current
diesel at 30 °F > high emitter diesel > gasoline engine emitting white smoke > normal gasoline >
normal diesel 72 °F > gasoline engine emitting black smoke. The mutagenic potency based on
the TA100 results were: current diesel at 30 °F > gasoline engine emitting white smoke > high
emitter diesel > normal diesel 72 °F > gasoline engine emitting black smoke > normal gasoline
30 °F > normal gasoline 72 °F. The authors' goal with this work was to examine the various
bioassays available to ascertain which are most useful in determining differences in
mutagenicity, toxicity, and inflammation. Significant findings indicate that both diesel and
gasoline exhaust emissions are mutagenic, with diesel being more mutagenic in general. The
increase in mutagenicity of gasoline samples with S9 activation indicates the role of PAH in this
effect. Decreased mutagenicity by the addition of S9 in the diesel sample collected at 30 °F
suggested to the authors that the mutagenicity of the exhaust may be due to nitroarenes.
Pohjola et al., (2003) used the extractable organic material from the PM and SVOC
gasoline and diesel exhaust fractions to examine their ability to induce mutations in Salmonella
strain TA98 and to form adducts in calf thymus DNA. Doses used in the Ames assay were
30 to 500 |ig/PM for the -S9 experiments and 10 to 1000 |ig/PM for the +S9 experiments.
Doses used for oxidative and reductive activation of PAHs were 18-300 jig PM for gasoline and
75-1,500 jig PM for diesel. Using the 32P-postlabeling method, 4 jig of DNA was analyzed for
bulky aromatic DNA adducts. Only the gasoline was tested in the Ames assay. The PM fraction
had higher mutagenicity, which averaged 431 rev/mg PM (- S9) and 487 rev/mg PM (+S9). The
SVOC fraction had only 106 rev/mg PM (- S9) and 98 rev/mg PM (+S9). However, the SVOC
fraction formed more DNA adducts. PAH-DNA adduct formation with 150 jig PM gasoline
extracts in calf thymus DNA ranged from 3.7 to 8.3 (-S9), 7.7 to 56 (+S9), 5.2 to 18 (-XO), and
19-60 (+XO) adducts/108 nucleotides/mg PM. Comparisons were made of PAH-DNA adduct
levels using high doses (42 to 150 jig PM) and low doses (7.5 to 18.5 jig PM) of gasoline and
diesel extracts. Diesel PM and gasoline extracts formed more S9-mediated adducts with
increasing doses, but gasoline did not have a linear dose-response. The authors suggested that
complex interactions and/or inhibition of S9 caused lower concentrations of both gasoline and
diesel extracts to bind DNA with greater efficiency than 8-fold higher doses. Diesel extracts
7-195
-------
formed higher levels of adducts than gasoline extracts, especially in the presence of XO
(reductive activation), indicating possible high levels of nitro PAHs. Diesel extracts were also
more mutagenic than gasoline extracts in the - S9 (direct-acting) assays, which further suggests
higher concentrations of nitro-PAHs in diesel exhaust. This study corroborates earlier research
suggesting that diesel exhausts extracts are more mutagenic than gasoline extracts, and that
diesel's mutagencicity can be attributed, in part, to DNA adduct formation.
7.8.4 Summary of Mutagenic/Genotoxic Effects
A number of recent in vivo and in vitro studies have suggested that ambient urban PM is
mutagenic. Research evaluating the mutagenicity of ambient PM from the Los Angeles area has
pointed to ubiquitous emission sources as being responsible for mutagenic activity observed in
vitro (Hannigan et al., 1997, 1998). Fractionation of those ambient samples and subsequent
mutagenicity assessments have indicated that six unsubstituted polyaromatic compounds and
two semi-polar compounds are the likely mutagens. Mutagenicity of urban air from Germany
has also been demonstrated (Hornberg et al., 1996, 1998; Seemayer and Hornberg, 1998), with
evidence showing that the fine fraction of PM exerted greater toxicity. Additionally,
ambient PM from high traffic areas in the Netherlands also induced genotoxic activity.
Emissions from wood/biomass burning have been shown to be mutagenic. Studies of
human exposures in China (Vinitketkumnuen et al., 2002) and the Netherlands (Heussen et al.,
1994), examining both chronic seasonal and acute exposures, have demonstrated increased
mutagenicity with environmental exposures. Characterization of wood smoke fractions to assign
mutagenicity have shown that the organic fraction is mutagenic and that the condensate is not.
Wood smoke emissions can cause both frameshift and base pair mutations but have not yet
demonstrated the production of DNA adducts.
Emissions from coal combustion have been shown to be mutagenic, especially the polar
and aromatic fractions. Research in China examining populations with high lung cancer rates
have shown that emission samples from homes burning smoky coal are mutagenic in the Ames
assay, and implicate PAHs as contributors to the mutagenicity (Mumford et al., 1987, 1999; Lan
et al, 2002). Recent work (Granville et al., 2003) characterizing the mechanism of genotoxicity
has examined the mutation spectra of coal smoke emissions from these Chinese homes.
Sequencing the revertants has shown that the mutations in Salmonella exposed to coal smoke
7-196
-------
extract are similar to mutations seen in lung tumors of women exposed environmentally to the
coal smoke, which differs in notable ways from coal combustion emissions in the United States.
Extensive studies have demonstrated mutagenic activity in both particulate and gaseous
fractions of DE. By sequential fractionation of DE, apportionment of the mutagenicity is
possible, which has implicated nitrated polynuclear aromatic compounds as being responsible
for a substantial portion of the mutagenicity. Other mutagenically active compounds include
ethylene, benzene, 1,3-butadiene, acrolein, and several PAHs in the gas phase. In addition to
Ames assay studies, the induction of gene mutations has been reported in several in vitro
mammalian cell lines after exposure to extracts of DPM. Structural chromosome aberrations and
SCE in mammalian cells have been induced by DE particles and extracts.
Early studies comparing the mutagenicity of gasoline and diesel exhaust showed that
the PM component of the exhaust is more mutagenic than the condensate fraction, and that
overall, diesel exhaust is more mutagenic than gasoline exhaust. More mutagenicity is also
observed in exhaust from cold starts than from exhausts at room temperature. Examining the
fractional mutagenicity of gasoline and diesel exhausts, it was shown that, as with coal smoke,
the polar component has the most mutagenicity, and further, that nitro-PAH is present in the
fraction. A comprehensive study comparing gasoline and diesel exhaust genotoxicity, using
both the PM and SVOC fractions, demonstrated that both exhausts are mutagenic, but, in
general, diesel exhaust is more mutagenic. Further, the study implicates PAH and nitroarenes in
the genotoxicity. Another current study corroborates these finding, and includes data suggesting
that DNA adduct formation is a component of the mutagenicity.
Exact comparisons of the mutagenicity of combustion emissions of these fuels are not
possible because data provided in the studies vary so greatly in units in which mutagenicity is
expressed. Thus, there is qualitative evidence for the mutagenic/genotoxic potential of both
ambient PM and some fuel combustion products. Many of the published in vitro studies failed to
provide details regarding the dose of PM extract delivered to the cells in vitro. In general, equal
volumes of air or amounts of time were sampled and reported, but only limited, if any,
characterization of the amount of PM mass or size was done or reported in many studies. Thus,
any quantitative extrapolation of the reported findings would be quite difficult. Nevertheless,
they collectively do appear to provide some evidence tending to substantiate the biologic
7-197
-------
plausibility of, and/or elucidating potential mechanisms underlying, reported epidemiologic
associations between long-term human exposure to ambient PM and lung cancer.
7.9 INHALED PARTICLES AS POTENTIAL CARRIERS OF
TOXIC AGENTS
Particle-Bound Water
In Chapter 2, it was noted that, although water vapor is not considered a pollutant per se
and particle-bound water is not measured as part of the ambient PM mass typically monitored for
regulatory purposes, particle bound water may serve as a carrier for other pollutants. Wilson
(1995) proposed that water-soluble gases that are usually largely removed by deposition to wet
surfaces in the upper (ET) portion of the respiratory tract could be dissolved in particle bound
water and, thereby, be carried into the lower regions of the respiratory tract. Such water-soluble
gases commonly found in polluted air masses include: oxidant species (e.g., O3, H2O2, and
organic peroxides); acid gases (e.g., SO2, HC1, HNO3, HONO, and formic acid); and polar
organic species (e.g., formaldehyde). Thus, water may be a vector by which these gases may be
delivered in enhanced proportions to the TB and A regions of the deep lung.
Kao and Friedlander (1995) also noted that, in evaluating health effects of ambient aerosol
components, it is "important to realize that the chemical analyses of routinely collected
particulate samples are not necessarily an accurate representation of the atmosphere." They
further noted that many short-lived chemical species in the gas or particle phase, such as free
radicals, may not be present in the sampled materials when analyzed hours to weeks (or longer)
after being collected on filters and stored. Also, the unmeasured metastable species may be
much more biochemically active than "dead" components collected or remaining on analyzed
filters. They concluded that, "since inhalation toxicology studies using both human and animal
subjects often do not include the potential for metastable species and reactive intermediaries to
be present, they could greatly underestimate the effects seen in field or epidemiologic studies."
Friedlander and Yeh (1998) elaborated further on the fact that the aqueous component of the
atmospheric submicron aerosol contains short-lived reactive chemical species. That is, they
explained that submicron atmospheric aerosols contain several types of components, which
importantly include very short-lived reaction intermediates, such as hydrogen peroxides,
7-198
-------
aldehydes, and organic acids found in cloud and rain water. Friedlander and Yeh (1998) further
noted (1) that particle phase concentrations of hydrogen peroxide fall in a range for which
significant biochemical effects were elicited with treatment of respiratory tract epithelial calls;
(2) that this may help to explain epidemiologic study results showing significant health effects to
be associated with fine-mode aerosols or sulfate (the submicron sulfate-containing aerosol often
being the product of atmospheric reactions involving hydrogen peroxide), and (3) that such
aerosols may be serving as a surrogate or indicator for the hydrogen peroxide or other reactive
species.
Wexler and Sarangapani (1998) used a physical model of "gas-particle-mucus heat and
mass transport in the human airways" to investigate the transport by particles of soluble vapors
to the tracheobronchial and air exchange regions of the lung. When the atmospheric aerosol is
inhaled, water soluble gases will begin to dissolve in the mucus on the surface of the airways.
However, hygroscopic particles will add particle-bound water in the high relative humidity of
the respiratory tract and more soluble gas can dissolve in the particle. The amount of soluble gas
in the particle will depend on the solubility of the gas (expressed as the Henry's Law
coefficient), the size of the particle, and the position of the particle in the respiratory tract. In the
presence of particles, the pattern of deposition of soluble gases may be moved deeper into the
respiratory tract. Very soluble gases, such as H2O2 and formaldehyde will still be almost
completely removed from the gas phase to the mucus on the airways. However, soluble gases
dissolved in particles may be carried into the air exchange region. If equilibrium is reached
rapidly, such highly soluble gases will evaporate from particles smaller than 0.1 jim diameter
before the particles reach the air exchange region. However, particles larger than -0.3 jim
diameter can efficiently carry such gases into the air exchange region.
Wexler and Sarangapani (1998) point out that due to the small volume of particle-bound
water, even in the case of highly soluble gases, only on the order of 1% of the soluble gas will be
found in the particles. However, particles will change the pattern of vapor deposition and
particles will carry dissolved gases deeper into the respiratory tract where the particles can
deposit on air exchange surfaces not protected by mucus. Furthermore, the Wexler and
Sarangapani (1998) analysis was based on considerations of physical solubility only. If adducts
or complexes form, such peroxohydrates from H2O2 (Friedlander and Yeh, 1998; Elvers et al.,
1991), or if the gas reacts chemically with water, as SO2 does to form SO2 (aq), H2SO3 (aq),
7-199
-------
and HSO 3 (aq) (Schwartz, 1984), the solubility of the gas may be increased greatly and the time
to reach equilibrium may be increased. Both factors would enable particles to transfer greater
quantities of dissolved gases to the air exchange region.
Morio et al. (2001) evaluated whether hygroscopic components of PM may transport H2O2
into the lower respiratory tract and induce tissue injury. Rats were exposed via inhalation
to (NH4)2 SO4 (0.3 to 0.4 |im MMD) at 215 or 249 |ig/m3 or H2O2 at 10, 20, or 100 ppb alone or
in combination for 2 h. No major effect was observed on BAL cell number or viability or on
protein content or LDH levels immediately or 24 h post exposure. However, rats treated with the
combination of sulfate and peroxide showed increased TNF°= produced by AMs and increased
numbers of neutrophils in pulmonary capillaries (as seen via EM). These results and other
effects on NO levels were interpreted by the authors as showing that biological effects of inhaled
PM are augmented by coexposure to sulfate and peroxide, including altered production of
cytokine mediators by AM.
The information summarized above has substantial implications for interpreting and
understanding the vast array of epidemiological and toxicologic results discussed in preceding
sections of this chapter and earlier chapters of this document. Their full significance becomes
more evident when considered in light of dosimetric information discussed in Chapter 6. It is
worth restating a few basic points here from Chapter 6 and expanding on them further with
regard to the importance of dosimetric considerations in relation to particles as carriers of other
toxic agents.
First, particle size is one of the most basic parameters governing particle behavior and
deposition in the respiratory tract. Particles between 0.3 and 0.7 jim in diameter have minimal
deposition in the respiratory tract. Above and below this range of minimum deposition, the
efficiency of deposition increases. The pattern of deposition within the respiratory tract also
slowly shifts from the alveolar region to the TB and ET regions with increasing particle size over
1 to 2 jim MMAD and with decreasing particle size below 0.1 jim.
Hygroscopicity, the propensity of a material for taking up and retaining moisture, is an
important property of some ambient particle species and affects respiratory tract deposition.
Such particles can increase in size in humid air in the atmosphere or in the respiratory tract and,
when inhaled, deposit according to their hydrated size rather than their initial size. Compared to
nonhygroscopic particles of the same initial size, deposition of hygroscopic aerosols in different
7-200
-------
regions varies, depending on initial size: hygroscopicity generally increases total deposition for
particles with initial sizes larger than -0.5 jam, but decreases deposition for particles between
-0.01 and 0.5 and again increases deposition for particles < 0.01 jim. Thus, under high humidity
conditions, there is increased deposition of smaller (nucleation-mode; < 0.01 jim) ultrafine
particles and of larger accumulation-mode (> 0.5 jim) particles, the latter of which can grow to
sizes exceeding 1.0 jim and both of which would contain enhanced amounts of particle bound
water and other toxic agents (e.g., SO2, peroxide, formaldehyde) dissolved therein.
Enhanced particle retention occurs on carinal ridges in the trachea and segmental bronchi;
and deposition "hot spots" occur at airway bifurcations or branching points. Peak deposition
sites shift from distal to proximal sites as a function of particle size, with greater surface dose in
conducting airways than in the A region for all particle sizes. To some extent then, the growth
of ultrafine and accumulation mode particles under humid conditions would also likely increase
"hot spot" deposition at airway branching points and thereby increase PM doses to tissues at
those points.
Ventilation rate, gender, age, and respiratory disease status all affect total and regional
respiratory tract particle deposition. Of likely most concern from among all these factors
affecting respiratory particle deposit patterns are altered PM deposition patterns due to
respiratory disease status that may put certain groups of adults (including some elderly) and
children at greater risk for PM effects. Importantly, COPD contributes to more heterogenous
deposition patterns and differences in regional deposition. One study indicates that people with
COPD tend to breathe faster and deeper than those with normal lungs (i.e., about 50% higher
resting ventilation) and have -50% greater deposition than age-matched healthy adults under
typical breathing conditions, with average deposition rates 2.5 times higher under elevated
ventilation rates. Enhanced deposition appears to be associated more with the chronic bronchitic
than the emphysematous component of COPD. In this and other new studies, fine-particle
deposition increased markedly with increased degree of airway obstruction. With increasing
airway obstruction and uneven airflow because of irregular obstruction patterns, particles tend to
penetrate more into remaining better ventilated lung areas, leading to enhanced focal deposition
at airway bifurcations and alveoli in those A region areas. In contrast, TB deposition increases
with increasingly more severe bronchoconstrictive states, as occur with asthmatic conditions.
7-201
-------
Disease states can also alter clearance rates for removal of deposited particles from the
lung. Bronchial mucus transport is slowed by asthma, chronic bronchitis, bronchial carcinoma,
and various acute respiratory infections - all being disease conditions expected to increase
retention of deposited particle material and, thereby, increase the probability of toxic effects
from inhaled ambient PM components reaching the TB region. Also, spontaneous coughing, an
important TB region clearance mechanism, does not appear to fully compensate for impaired
mucociliary clearance in small airways and may become depressed with worsening airway
disease, as seen in COPD patients. Clearance of particles from the A region by AMs and their
mucociliary transport is usually rapid (< 24 h), but alveolar region clearance rates are decreased
in human COPD sufferers and slowed by acute respiratory infections; and the viability and
functioning of AMs are reduced in human asthmatics and in animals with viral lung infections.
All this suggests that persons with asthma, chronic bronchitis, or acute lung infections are
likely to experience increased deposition and retention of inhaled particles and to be at increased
risk for ambient PM exposure effects. Such individuals can reasonably be expected to be put at
even greater risk when inhaling ambient PM under high humidity conditions (with increased
delivery of peroxides, SO2, and other noxious agents into the deep lung in particle-bound water
and enhanced "hot spot" deposition of hygroscopic aerosols at branching points in bronchial
airways).
Bioaerosols as Contributors to Ambient PM Effects
Bioaerosols, from sources such as plants, fungi, and microorganisms, range in size from
0.01 |im to > 20 |im. They comprise a small fraction of ambient PM, but likely contribute to
some types of ambient PM-related health effects exposure.
Intact pollen grains from flowering plants, trees and grasses are by far most abundant in
warm/humid spring and summer months and can deposit in upper airways to induce allergic
rhinitis. Allergen-laden cytoplasmic fragments (-0.1 to 0.4 |im in size) of pollen grains (which
rupture under high moisture conditions) can enter the deep lung, where they can exacerbate
asthma. Binding of allergen-laden pollen cytoplasmic fragments to ambient fine particles (e.g.,
DPM) has also been observed; and synergistic interactions between pollen debris and other
ambient PM (e.g., the poly cyclic hydrocarbon component of DE) are thought to be a mechanism
that may explain the increased incidence of asthma morbidity and mortality. Pollen granules can
7-202
-------
also act as vectors for binding of other bioaerosols (e.g., endotoxins, fungi or fragments, glucans)
and thereby enhance their inhalation and deposition in the respiratory tract.
Fungal spores and fungi fragments are among the largest and most consistently present
bioaerosols found outdoors (levels being higher during warm/humid months). They cause
allergic rhinitis and asthma, which is highly dependent on seasonal variations in concentration.
Exposures have been linked to asthma hospitalization and death.
Bacteria and viruses are significant bioaerosols. Much of the toxicity of bacteria is due to
the endotoxins present in the outer cell membrane, which trigger production of cytokines and a
cascade of inflammation. Ambient airborne concentrations of endotoxins vary with seasons
(being higher in warm/humid periods and low in colder months) and tend to be higher in samples
of coarse-mode than in fine-mode ambient PM. Another cell wall component of bacteria and
fungi, (1^3)-p~D-glucan, has also been shown to cause respiratory inflammation.
Animals and insects produce bioaerosols capable of producing hypersensitivity diseases.
Most notably, exposure to dust mite and cockroach material has been linked to sensitization in
children. However, indoor exposures to such materials probably are of most importance with
regard to human exposures to such materials.
It thus appears that certain ambient bioaerosols (e.g., pollen, fungi, endotoxins, glucans)
that become abundant during warm/humid weather may contribute to seasonal increases in PM-
associated risk during spring/summer months, but not during colder winter months. The
copresence of nonbiological particles, serving as vectors concentrating such bioaerosols and
enhancing their delivery into the deep lung, appears to likely be important.
Summary and Conclusions
It has been proposed that particles also may act as carriers to transport toxic gases into the
deep lung. Water-soluble gases, which would be removed by deposition to wet surfaces in the
upper respiratory system during inhalation, could dissolve in particle-bound water and be carried
with the particles into the deep lung. Equilibrium calculations indicate that particles do not
increase vapor deposition in human airways. However, these calculations do show that soluble
gases are carried to higher generation airways (i.e., deeper into the lung) in the presence of
particles than in the absence of particles. In addition, species such as SO2 and formaldehyde
react in water, reducing the concentration of the dissolved gas-phase species and providing a
7-203
-------
kinetic resistence to evaporation of the dissolved gas. Thus, the concentration of the dissolved
species may be greater than that predicted by the equilibrium calculations. Of much concern,
particle-bound water appears to be a means by which dissolved hydrogen peroxide and other
short-lived reactive oxygen species can be carried into lower respiratory tract regions and
contribute to the induction of inflammatory responses. Also, certain other toxic species (e.g.,
NO, NO2, benzene, PAHs, nitro-PAH, a variety of allergens) may be absorbed onto solid
particles and carried into the lungs. Thus, ambient particles may play important roles not only in
inducing direct health impacts of their constituent components but also in facilitating delivery of
toxic gaseous pollutants or bioagents into the lung and may, thereby, serve as key mediators of
health effects caused by the overall air pollutant mix.
7.10 INTERPRETIVE SUMMARY OF PM TOXICOLOGY FINDINGS
Toxicological studies can play an integral role in addressing several key important
questions regarding ambient PM health effects:
(1) What types of pathophysiological effects are exerted by ambient PM or constituent
substances and what are potential mechanisms that likely mediate various PM
health effects?
(2) What PM characteristics (size, chemical composition, etc.) cause or contribute to
health effects?
(3) What types of interactive effects of particles and gaseous co-pollutants have been
demonstrated?
(4) What susceptible subgroups are at increased risk for ambient PM health effects and
what factors contribute to increased susceptibility?
(5) What are the exposure-response relationships of ambient PM and how can that
information be extrapolated to human exposures?
This summary focuses on highlighting salient findings that reflect the progress made by
toxicological studies towards addressing these questions. All these questions have important
implications bearing on the matter of biological plausibility of epidemiologically-observed
ambient PM effects.
7-204
-------
One overarching issue in the interpretation of toxicology study results is the relevance of
findings from experimental human or animal studies using controlled exposure/dose
concentrations that are high relative to the much lower ambient pollutant exposure levels that
apply within the context of pertinent epidemiology studies. To provide insight on this issue,
EPA conducted a series of illustrative analyses using dosimetric modeling of the type discussed
in Chapter 6; these analyses are described in detail in Appendix 7A. First, taking into account
certain key points regarding dose metrics, one of the publically available dosimetry models (the
MMPD model) discussed in Section 6.6.4 of Chapter 6 and in Appendix 7A Section 3 was
employed to compare estimates of deposited and/or retained respiratory tract PM doses in the
human and rat lung using different dose metrics as described in Table 7A-4. The second
approach involved application of the same publically-available model (a) to estimate likely
respiratory tract doses (again using various dose metrics) resulting from experimental exposures
(via PM inhalation or instillation) of human or laboratory animals (rats) actually employed in
representative published PM toxicology studies assessed in this chapter and (b) to estimate likely
ambient PM exposure concentrations that would be needed in order to obtain comparable human
and rat PM respiratory tract doses. Rats clear PM from the respiratory tract much faster than
humans, which is more important for long-term exposure comparisons. Rats also have a lower
deposition fraction, which is more important for comparing acute effects. MPPD modeling
indicates that higher exposure concentrations in the rat may be needed, in certain cases, in order
to evaluate toxicological endpoints predictive of health outcomes in humans and to investigate
biological mechanisms. The higher doses needed depend on the health endpoint measured.
For example, the modeling results suggest that higher PM concentration exposures in rats may
be needed to achieve nominally similar inflammatory responses relative to the human.
7.10.1 Particulate Matter Health Effects and Potential Mechanisms of Action
Numerous epidemiologic analyses discussed in Chapter 8 have shown associations
between ambient PM levels and increased risk for cardiorespiratory effects, as well as for lung
cancer. Findings since 1996 have provided evidence supporting many hypotheses regarding
induction of PM effects; and this body of evidence has grown substantially. Various toxicologic
studies using PM having diverse physicochemical characteristics have shown that such
7-205
-------
characteristics have a great impact on the specific response that is observed. Thus, there appear
to be multiple biological mechanisms that may be responsible for observed morbidity/mortality
due to exposure to ambient PM, and these mechanisms appear to be highly dependent on the
type and dose of particle in the exposure atmosphere. It also appears that many biological
responses are produced by PM whether it is composed of a single component or a complex
mixture.
The following discussion focuses on summarizing key lines of toxicological evidence
useful in (a) delineating various types of health effects attributable to PM exposures, and
(b) identifying potential pathophysiological mechanisms by which the effects of particle
exposure are mediated. Major emphasis is placed on discussions of PM effects on the
cardiopulmonary system, and some attention is accorded to PM-related mutagenic/genotoxic
effects of relevance to evaluating the carcinogenic potential of ambient PM or constituent
substances.
7.10.1.1 Direct Pulmonary Effects
When the 1996 PM AQCD was written, the lung was thought to be the primary organ
affected by particulate air pollution. Although the lung still is a primary organ affected by PM
inhalation, there is growing toxicological and epidemiologic evidence that the cardiovascular
system is also affected and may be a co-primary organ system related to certain health endpoints
such as mortality. Nonetheless, understanding how particulate air pollution affects respiratory
system functions or exacerbates respiratory disease remains an important goal. The
toxicological evidence from controlled exposures to ambient PM or constituents appear to
support three hypothesized mechanisms for PM inducing direct pulmonary effects: (1) lung
injury and inflammation; (2) increased airway reactivity and exacerbation of asthma; and
(3) impaired lung defense mechanisms and increased susceptibility to respiratory infections.
An important caveat in interpretation of the toxicological data is that the high doses used in
many of the studies may produce different effects on the lung than inhalation exposures at lower
ambient concentrations. That is, "realistic" doses associated with ambient PM exposures may
activate cells and pathways entirely disparate from those activated at high experimental doses.
7-206
-------
Lung Injury and Inflammation
Particularly compelling evidence pointing towards ambient PM causing lung injury and
inflammation derives from the study of extracts of ambient PM materials on filters collected
from community air monitors before, during and after the temporary closing of a steel mill in
Utah Valley. Ohio and Devlin (2001) found that intratracheal instillation of filter extract
materials in human volunteers provoked greater lung inflammatory responses for materials
obtained before and after the temporary closing versus that collected during the plant closing.
The instilled dose of 500 jig of extract material was calculated by Ohio and Devlin to result in
focal lung deposition in the lingula roughly equivalent to 5 times more than would be deposited
if an active person experienced 24-h inhalation exposure to 100 |ig/m3 PM10 (during wintertime
temperature inversions in Utah Valley 24-h PM10 levels can exceed 100 |ig/m3). Moreover,
100 jig of filter extract collected during the winter before the temporary plant closure similarly
instilled into the lungs of human volunteers also increased levels of neutrophils, protein, and
inflammatory cytokines. Dosimetric calculations presented in Appendix 7A suggest that the
instilled dose used in the Ohio and Devlin (2001) study could be experienced by persons
exposed for about 6 to 9 weeks to ambient PM in the Utah Valley. Further, the instillation in
rats (Dye et al., 2001) of extract materials from before and after the plant closing resulted in a
50% increase in airway hyperresponsiveness to acetylcholine compared to 17 or 25% increases
with saline or extract materials for the period when the plant was closed, respectively. Analysis
of the extract materials revealed notably greater quantities of metals for when the plant was
opened, thus suggesting that such metals (e.g., Cu, Zn, Fe, Pb, As, Mn, Ni) may importantly
contribute to the pulmonary toxicity observed in the controlled exposure studies, as well as to
health effects shown epidemiologically to vary with PM exposures of Utah Valley residents
before, during, and after the steel mill closing.
Still other toxicological studies point towards lung injury and inflammation being
associated with exposure of lung tissue to complex combustion-related PM materials, with
metals again being among some ambient PM constituents identified as contributors. For
example, in the last few years, numerous studies have shown that high doses/concentrations of
instilled and inhaled ROFA, a product of fossil fuel combustion, can cause substantial lung
injury and inflammation. The toxic effects of ROFA are largely caused by its high content of
soluble metals including Fe, Ni, and V. Some of the pulmonary effects of ROFA can be
7-207
-------
reproduced by equivalent exposures to soluble metal salts. In contrast, controlled exposures of
animals to sulfuric acid aerosols, acid-coated carbon, and sulfate salts cause little lung injury or
inflammation, even at high concentrations. Inhalation of concentrated ambient PM (which
contains only small amounts of metals) by laboratory animals at concentrations in the range of
100 to 1000 |ig/m3 have been shown in some (but not all) studies to cause mild pulmonary injury
and inflammation. Rats with SO2-induced bronchitis and MCT-treated rats have been reported
to have a greater inflammatory response to concentrated ambient PM than normal rats. These
studies suggest that exacerbation of respiratory disease by ambient PM may be caused in part by
lung injury and inflammation.
There are also new in vitro data indicating a potential neurogenic basis for the effects of
particulate matter (Veronesi et al., 1999a,b; Oortgiesen et al., 2000; Veronesi et al., 2002b).
More specifically, these researchers hypothesize that the proton cloud associated with negatively
charged colloidal PM particles could activate acid sensitive VR1 receptors found on human
airway epithelial cells and sensory terminals; this activation, in turn, results in an immediate
influx of calcium and the release of inflammatory neuropeptides and cytokines, which initiate
and sustain inflammatory events in the pathophysiology of neurogenic inflammation. This
implies that a wide variety of particulate substances, from many different types of sources (both
natural and anthropogenic), falling across wide size ranges (from ultrafine through accumulation
mode and including small, < 10 |im, coarse fraction particles), and of highly diverse chemical
composition could possibly exert neurogenically-mediated pathophysiological effects depending
on shared physical properties of their surface molecules (i.e., negative charges surrounded by a
proton cloud).
Increased Airway Reactivity and Exacerbation of Asthma
The strongest evidence supporting this hypothesis is from studies on DPM. Diesel
particulate matter has been shown to increase production of antigen-specific IgE in mice and
humans (summarized in Section 7.2.1.2). In vitro studies have suggested that both organic
fraction and the CB fraction of DPM are involved in the increased IgE production. ROFA
leachate also has been shown to enhance antigen-specific airway reactivity in mice (Goldsmith
et al., 1999), indicating that soluble metals can also enhance an allergic response. However, in
this same study, exposure of mice to concentrated ambient PM did not affect antigen-specific
7-208
-------
airway reactivity. Thus the available evidence is inconclusive with regard to increased airway
activity as a possible PM mechanism.
Increased Susceptibility to Respiratory Infections
A few newly published studies have provided some evidence for ambient PM potentially
affecting lung defense mechanisms and increasing susceptibility to infection. The studies of
Zelikoff et al. (2003) showed that brief exposures (3 to 5 h) of Fischer rats to New York City
CAPs (-225 |ig/m3) either before or after IT-instillation of Streptococcus pneumoniae increased
numbers of lavageable AMs and increased bacterial burden over control levels at 24 h
postinfection. Similarly, Antonini et al. (2002) found that preexposure to ROFA (0.2 or
1.0 mg/100 g body weight) of SD rats 3 days before IT instillation of Listeria monocytogenes
(a bacterial pathogen) led to notable lung injury, slowed clearance of the bacteria, and reduced
AM NO production, although AM numbers were not reduced. Lastly, new studies by Ohtsuka
et al. (2000a,b), showing decreased phagocytic activity of AMs in mice after a 4 h inhalation
exposure to acid-coated carbon particles (albeit at a high mass concentration of 10 mg/m3), are
suggestive of possible impairment of an important lung defense mechanism even in the absence
of lung injury.
7.10.1.2 Cardiovascular and Other Systemic Effects Secondary to Lung Injury
When the 1996 PM AQCD was written, it was thought that cardiovascular-related
morbidity and mortality most likely would be sequelae occurring secondary to impairment of
oxygenation or some other consequence of lung injury and inflammation. Newly available
toxicologic studies provide evidence regarding such possibilities, as discussed below.
Impairment of Oxygenation and Increased Work of Breathing That Adversely Affect
the Heart Secondary to Lung Injury
Results from most of the new toxicology studies in which animals (normal and
compromised) were exposed to concentrated ambient PM (even at concentrations many times
higher than would be encountered in the United States) indicate that ambient PM is unlikely to
cause severe disturbances in oxygenation or pulmonary function. One study, of PM effects in a
severely compromised animal model, may hint at possible PM pathophysiologic effects
mediated via hypoxemia. Specifically, the instillation of ROFA (0, 0.25, 1.0, 2.5 mg) was
7-209
-------
shown (Watkinson et al., 2000a,b) to increase (to 50%) the mortality rate observed in MCT-
treated rats with pulmonary hypertension. Although blood oxygen levels were not measured in
this study, there were ECG abnormalities consistent with severe hypoxemia in about half of the
rats that subsequently died. Given the severe inflammatory effects of instilled ROFA at high
doses and the fact that MCT-treated rats have increased lung permeability as well as pulmonary
hypertension, it is plausible that instilled ROFA may cause severe hypoxemia leading to death in
this rat model; but the relevance of such effects of high ROFA exposures in a severely
compromised rat model to human ambient PM exposure effects is unclear. More research is
needed on the effects of PM on arterial blood gases and pulmonary function to fully address the
above hypothesis.
Systemic Hemodynamic Effects Secondary to Lung Inflammation and Increased
Cytokine Production
It has been suggested that systemic effects of particulate air pollution may result from
activation of cytokine production in the lung (Li et al., 1997). Results from some studies of
compromised animal models provide some support for this idea. For example, there was a
significant decrease in the time of onset of ischemic ECG changes following coronary artery
occlusion in PM-exposed dogs compared to controls (Godleski et al., 2000). Analogously,
Wellenius et al. (2002) found, in another animal model (i.e., left ventricular MI induced by
thermocoagulation), that 41% of the MI rats exhibited one or more premature ventricular
complexes (PVCs) during baseline periods 12 to 18 h after surgery; and exposure to ROFA (but
not to CB or room air) increased arrhythmia frequency in animals with prior PVCs and
decreased their HRV. Also, severely compromised MCT-treated rats exposed to high
concentrations of inhaled ROFA (15,000 |ig/m3, 6 h/day for 3 days) showed increased
pulmonary cytokine gene expression, bradycardia, hypothermia, and increased arrhythmias
(Watkinson et al., 2000a,b). On the other hand, SH rats manifested similar cardiovascular
responses to inhaled ROFA (except that they also developed ST segment depression), but
without any increase in pulmonary cytokine gene expression.
Other studies of normal dogs exposed to concentrated ambient PM (322 |ig/m3,
MMAD = 0.23 to 0.34 jim) showed minimal pulmonary inflammation and no positive staining
for IL-8, IL-1, or TNF in airway biopsies (Godleski et al., 2000). In addition, several other
studies (e.g., Muggenburg et al., 2000a,b) of normal dogs and/or rats failed to show changes in
7-210
-------
ECG consistent with the types observed in the above studies of compromised models. Thus, the
link between PM-induced changes in the production of cytokines in the lung and effects on
cardiovascular function is not clear, and more basic information on the effects of mild
pulmonary injury on cardiovascular function is needed to more fully evaluate this hypothesis.
Increased Blood Coagulability Secondary to Lung Inflammation
There is abundant evidence linking small prothrombotic changes in the blood coagulation
system to increased long-term risk of heart attacks and strokes. However, the published
toxicological evidence bearing on whether moderate lung inflammation causes increased blood
coagulability is very mixed and inconsistent.
Several new studies have investigated possible effects of ambient PM or surrogate particles
on blood chemistry constituents that would be indicative of increased blood coagulability.
For example, Ohio et al. (2000a) have shown that inhalation of concentrated ambient PM (at
> 47 |ig/m3) in healthy nonsmokers causes increased levels of blood fibrinogen. Gardner et al.
(2000) have also shown that a high dose (8300 |ig/kg) of instilled ROFA in rats causes increased
blood levels of fibrinogen, but no effect was seen at lower doses. Gordon et al. (1998) also
reported increased blood platelets and neutrophils in control and MCT-treated rats on some, but
not all, days when exposed to concentrated NYC ambient PM (150 to 400 |ig/m3) for 3 h. These
differences in blood parameters were present at 3 h PE but not 24 h PE.
On the other hand, exposure of normal dogs to concentrated ambient PM from Boston
(-100 to 1000 |ig/m3) had no effect on fibrinogen levels (Godleski et al., 2000). Nor were any
significant effects on blood fibrinogen or other factors (e.g., blood platelets, tissue plasminogen
activator, Factor VII, etc.) involved in the coagulation cascade seen with exposure of normal rats
to concentrated NYC ambient PM (-130 to 900 |ig/m3), as reported by Nadziejko et al. (2002).
Frampton (2001) also reported finding no effects on fibrinogen or clotting Factor VII in healthy,
nonsmoking human adults exposed to 10 |ig/m3 ultrafine carbon for 2 h via mouthpiece
inhalation while at rest. All these latter results, indicative of little effect of PM exposure on
blood coagulation factors in healthy humans or laboratory animals, stand in contrast to the
above-noted highly suggestive ambient PM-induced increases in fibrinogen seen by Ghio et al.
(2000a) in healthy human adult volunteers.
7-211
-------
The coagulation system is as multifaceted and complex as the immune system; and there
are many other sensitive and clinically significant parameters that should, in addition to
fibrinogen, show more extensive and consistent patterns of change reflective of PM effects on
blood coagulation. Thus, it is premature to draw any strong conclusions about the relationship
between PM and blood coagulation.
Hematopoiesis Effects Secondary to PM Interactions With the Lung
Terashima et al. (1997) found that instillation of fine carbon particles (20,000 jig/rabbit)
stimulated release of PMNs from bone marrow. In further support of this hypothesis, Gordon
and colleagues reported that the percentage of PMNs in the peripheral blood increased in rats
exposed to ambient PM in some but not all exposures. On the other hand, Godleski et al. (2000)
found no changes in peripheral blood counts of dogs exposed to concentrated ambient PM.
Thus, consistent evidence that PM ambient concentrations can affect hematopoiesis remains to
be demonstrated.
7.10.1.3 Direct Effects on the Heart
Although the data are still limited, two types of hypothesized direct effects of PM on the
heart are noted below.
Effects on the Heart Secondary to Uptake of Particles into the Circulation and/or Release of
Soluble Substances into the Circulation
Drugs can be rapidly and efficiently delivered to the systemic circulation by inhalation.
This implies that the pulmonary vasculature absorbs inhaled materials, including charged
substances such as small proteins and peptides. Such PM materials could conceivably be rapidly
transported to the heart, where they might exert effects directly on cardiac vasculature or heart
muscle itself. Alternatively, they could also exert very rapid effects on cardiac function through
stimulation of nerve ending receptors in lung tissue, resulting in secretion of inflammatory
messenger substances and/or activation of neurally-mediated autonomic reflexes. This raises the
question of how inhaled particles could affect the autonomic nervous system. Activation of
neural receptors in the lung is a logical area to investigate.
7-212
-------
Epithelial cells lining lower respiratory tract airways are damaged or denuded in many
common health disorders (e.g., asthma, viral infections, etc.), which may allow inhaled PM to
directly encounter sensory nerve terminals and their acid-sensitive receptors. In vitro studies by
Veronesi and colleagues provide interesting evidence indicating that (a) ROFA-induced
inflammation is mediated by acid-sensitive VR1 receptors on sensory nerve fibers that innervate
the airways and on surrounding bronchial epithelial cells (Veronesi et al., 1999a,b);
(b) negatively-charged, but not neutrally-charged (i.e., zeta potential = 0 mV), particles in
ROFA, synthetic polymer aerosols, or extracts from urban St. Louis, residential (woodstove),
volcanic (Mt. St. Helens), and industrial (coal and oil fly ash) sources activate the VR1 receptors
(Oortgiesen et al., 2000), with their zeta potential being the key physiochemical property
correlated with consequent increases in Ca2+ and IL-6 release (Veronesi et al., 2002b); and
(c) the receptor activation causing release of inflammatory cytokines and neuropeptides initiates
and sustains inflammatory effects in the airways (Veronesi and Oortgeisen, 2001).
Inhaled Particulate Matter Effects on Autonomic Control of the Heart and
Cardiovascular System
Besides the above studies, it is worth noting that earlier studies in conscious rats have
shown that inhalation of wood smoke causes marked changes in sympathetic and
parasympathetic input to the cardiovascular system that are mediated by neural reflexes
(Nakamura and Hayashida, 1992).
While changes in heart rate variability and conduction system function with ambient PM
exposure have been reported in some animal studies, not all studies have shown consistent
alterations (Godleski et al., 2000; Gordon et al., 2000; Watkinson et al., 2000a,b; Campen et al.,
2000; Muggenburg et al., 2000b; Frampton, 2001). In several human panel studies described in
Chapter 8, similar discrepancies were noted. Some of these studies included endpoints related to
respiratory effects but few significant adverse respiratory changes were detected. This raises the
possibility that ambient PM may have effects on the heart that are independent of adverse
changes in the lung. There is certainly precedent for this idea. For example, tobacco smoke
(which is a mixture of combustion-generated gases and PM) causes cardiovascular disease by
mechanisms that are independent of its effect on the lung.
7-213
-------
7.10.1.4 Mutagenic/Genotoxic Effects of PM
As discussed in Chapter 8, the Pope et al. (2002) analysis of the American Cancer Society
longer-term database provides evidence for chronic ambient PM exposure being associated with
increased risk of lung cancer.
Ambient urban PM from sources such as the Los Angeles area, Germany, and the
Netherlands has been shown to possess mutagenic activity in both in vivo and in vitro assays.
The mutagenicity is dependent upon the chemical composition of the PM and also the size of the
particles. Both unsubstituted polyaromatic compounds and semi-polar compounds are thought
to be highly mutagenic components of urban ambient PM. Additionally, the fine fraction
appears to be more mutagenic than the coarse fraction.
Emissions from wood/biomass burning have been shown to be mutagenic in studies of
human exposures in China and the Netherlands. Mutagens from wood smoke emissions can
cause both frameshift and base pair mutations but have not yet demonstrated the production of
DNA adducts.
A large body of work examining emissions from coal combustion in China has
demonstrated the mutagenicity of both the polar and aromatic fractions. Populations with high
lung cancer rates have been linked to exposures of the PAH component of coal smoke.
By sequencing of mutations, a direct link has been established between the mutagenesis assay
results and human lung cancer.
The U.S. EPA Diesel Document (U.S. Environmental Protection Agency, 2002) was cited
earlier in this chapter as discussing a number of studies utilizing mutagenicity/genotoxicity
assays with diesel emissions; and key information from that document on a number of studies
indicative of diesel emission particle-induced gene mutations, chromosome effects, or other
genotoxic effects (e.g., altered DNA adduct patterns, increases in mutagenic DNA, adduct-
related vulnerability to oxidative damage) was recounted. Additional findings were also noted
which show that, although 50 to 90% of the total mutagenicity of diesel exhaust is likely
attributable to its gaseous components, nitrated polynuclear aromatic compounds (PAHs) also
appear to account for a notable portion of the mutagenicity. Some results (but not others) further
appear to implicate sulfur in diesel emissions as contributing to mutagenic effects. Lastly, of
much interest are findings by Driscoll et al. (1996, 1997) showing increased hprt mutations in rat
alveolar type II cells with inhalation exposure to CB particles or with intratracheal instillation of
7-214
-------
CB or two other (quartz, TiO2) particles. All three types of particles elicited increased
inflammatory responses, which the authors suggest leads to increased epithelial cell proliferation
and consequently, mutations. Overall, the new studies are highly indicative of mutagenic and
other secondary genotoxic effects of ambient PM and/or various specific constituents (e.g.,
comparisons of gasoline and diesel exhaust show that the PM component is more mutagenic than
the condensate fraction in both exhausts). Additionally, the polar component has the most
mutagenicity. Other components of both gasoline and diesel exhaust thought to contribute to the
mutagenicity are PAHs, nitro-PAH, and nitroarenes. DNA adduct formation is one mechanism
whereby these mobile combustion products are thought to induce carcinogenesis.
In summary, both ambient PM and combustion products of coal, wood, diesel, and gasoline
are mutagenic/genotoxic, though exact comparisons of the mutagenicity of combustion
emissions of these fuels are not possible. Additionally, the data currently available allow some
clues as the to potential mechanisms underlying these health effects.
7.10.2 Links Between Specific Particulate Matter Components and
Health Effects
The plausibility of epidemiologically-demonstrated associations between ambient PM and
increases in morbidity and mortality has been questioned because adverse cardiopulmonary
effects have been observed among human populations at very low ambient PM concentrations.
To date, experimental toxicology studies have provided some intriguing, but limited, evidence
for ambient PM mixes or specific PM components potentially being responsible for reported
health effects of ambient PM. Overall, the new studies suggest that some of particles are more
toxic than others. New findings substantiating the occurrence of health effects in response to
controlled exposures to ambient PM mixes and/or their constituent substances are useful in
demonstrating or clarifying potential contributions of physical/chemical factors of constituent
particles are discussed below.
7.10.2.1 Ambient Particle Studies
Studies using concentrated ambient particles CAPs studies are probably most useful in
helping to substantiate that particles present in "real-world" ambient air mixes are indeed
capable of inducing notable pathophysiological effects under controlled exposure conditions and
7-215
-------
to clarify further factors affecting increased susceptibility of "at risk" groups for PM effects.
CAPs studies, on the other hand, tend to be somewhat less helpful than other toxicologic
approaches in helping to delineate the specific characteristics of PM producing toxicity and
potential underlying mechanisms. Some, but not all, studies with inhaled (CAPs) have found
cardiopulmonary changes in rodents and dogs at high concentrations of fine PM. However, no
comparative studies to examine the effects of ultrafine and coarse ambient PM have been done.
Studies using collected urban PM for intratracheal administration to healthy and
compromised animals have also produced interesting new information. Despite the difficulties
associated with extrapolating from the bolus delivery used in such studies, they have provided
evidence indicating that the chemical composition of ambient particles can have a major
influence on toxicity. Instillation of rats with filter extracts of ambient air particles collected
from Ottawa CN air (Watkinson, et al., 2002a,b) at 2.5 mg, for example, induced pronounced
biphasic hypothermia, severe drop in heart rate, and increased arrhythmias; this was in contrast
to no cardiac effects seen with comparable instilled dose of Mt. St. Helens volcanic ash (shown
by many studies to be relatively inert lexicologically). Similarly, dose-dependent increases in
PMNs, other markers of lung inflammation, and decreases in AMs were seen with intratracheal
exposures of hamsters to urban ambient particles from St. Louis or Kuwaiti oil fire particles
(Brain etal., 1998).
Importantly, it has become evident that, although the concentrated ambient PM (CAPs)
studies have provided important exposure-response information for some PM size fractions
(especially PM2 5), they have not, to date, been very helpful in identifying specific toxic
components in urban PM. Insufficient attention has been accorded to characterization of day-to-
day variations in specific PM constituents in order to relate such variations to observed variable
health responses to CAPs exposures. Also, because only a limited number of exposures using
CAPs can be reasonably conducted by a given laboratory in a particular urban environment,
there may be insufficient information to conduct factor analyses on exposure/response matrices.
This may also hinder principal component analysis techniques that are useful in identifying
particle components responsible for health outcomes. New particle concentrator systems now
coming on-line at the U.S. EPA and elsewhere that permit selective concentration of ultrafine,
fine, and thoracic coarse PM hold promise for enhancing our understanding of PM
characteristics producing toxicity. CAPs studies also hold promise for helping to identify
7-216
-------
susceptibility factors in animal models, and permit examination of mechanisms related to PM
toxicity.
7.10.2.2 Acid Aerosols
There is relatively little new information on the effects of acid aerosols. The 1996 PM
AQCD previously assessed acid aerosol health effects and concluded that acid aerosols cause
little or no change in pulmonary function in healthy subjects, but asthmatics may develop small
changes in pulmonary function. This conclusion is further supported by the new study of Linn
and colleagues (1997) in which children (26 children with allergy or asthma and 15 healthy
children) were exposed to sulfuric acid aerosol (100 |ig/m3) for 4 h. There were no significant
effects on symptoms or pulmonary function when data for the entire group were analyzed, but
the allergy group had a significant increase in symptoms after the acid aerosol exposure. Thus,
acid aerosol health effects may be a possible causal physical property for some types of
PM-related respiratory symptoms. However, it is unlikely that particle acidity alone could
account for the pulmonary function effects (Dreher, 2000).
7.10.2.3 Metals
The 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) mainly relied on
data related to occupational exposures to evaluate the potential toxicity of metals in contributing
to health effects associated with ambient PM exposures. Since that time, numerous newly
published in vivo and in vitro studies using exposures to ambient PM extracts, ROFA, other
combustion source emission materials (e.g., CFA, etc.), or specific soluble transition metals have
contributed substantial further information on the health effects of particle-associated soluble
metals. Although there are some uncertainties about differential effects of one transition metal
versus another, some water soluble metals (e.g., Ni, V, Zn, Fe) leached from ambient filter
extracts or ROFA have been shown consistently (albeit at high concentrations) to cause cell
injury and inflammatory changes in vitro and in vivo.
Perhaps most notable in this argument are the Utah Valley studies that have linked the
toxicology (in vitro cell culture as well as human and rodent instillation) with published
epidemiological findings. In these studies, filter extracts of Utah Valley PM collected from the
state/federal sampling sites yielding aerometric data used to ascribe the impact of PM on hospital
7-217
-------
admissions and population mortality rates showed remarkable qualitative coherence with
toxicological and clinical endpoints (BAL fluid markers, lung dysfunction) among the human
and rodent test subjects. Moreover, the data were themselves consistent with the hypothesized
underlying mode of action (oxidant generation, inflammation) for metal-associated PM
cardiorespiratory effects (Frampton et al., 1999; Dye et al., 2001; Ohio and Devlin, 2001;
Soukup et al., 2000; Wu et al., 2001; Pagan et al., 2003). Studies comparing human (Ohio and
Devlin, 2001) and rat (Dye et al., 2001) exposures to both high and low metal content PM
collected near a steel plant, showed convincingly that the metal content of the PM, and not the
mass, was a major determinant of the toxicity of the PM. Both species showed similar
inflammatory responses to exposures from PM with high metal content (collected while the steel
mill was operating). Hence, this rich data set provides an important linkage across study
disciplines used in the human and animal toxicology as well as in the in vitro studies.
Even though it is clear that combustion particles that have a high content of soluble metals
can cause lung injury in compromised animals and correlate well with epidemiological findings
in some cases (e.g., Utah Valley Studies), it has not been fully established that the small
quantities of metals (typically < 0.5 to 1.0 |ig/m3) associated with current U.S. ambient PM mass
concentrations exhibit greater toxicity than other PM components typically present in ambient
air. In studies in which various ambient and emission source particulates were instilled into rats,
the soluble metal content did appear to be one important determinant of lung injury (Costa and
Dreher, 1997). However, one published study (Kodavanti et al., 2000b) has compared the
effects of inhaled ROFA (at 1 mg/m3) to concentrated ambient PM (four experiments, at mean
concentrations of 475 to 900 |ig/m3) in normal and SO2-induced bronchitic rats. A statistically
significant increase in at least one lung injury marker was seen in bronchitic rats with one out of
four of the concentrated ambient exposures; whereas inhaled ROFA had no effect, even though
the content of soluble Fe, V, and Ni was much higher in the ROFA sample than in the
concentrated ambient PM.
Nevertheless, other particularly interesting new findings do point toward ambient PM
exacerbation of allergic airway hyperresponsiveness and/or antigen-induced immune responses.
Both metal and diesel particles have been implicated, with an expanding array of new studies
showing DPM in particular as being effective in exacerbating allergic asthmatic responses.
7-218
-------
7.10.2.4 Diesel Exhaust Particles
As described in Section 7.5.3, there is growing toxicological evidence that, analogously to
several other types of PM (silica, CB, road dust, etc.), diesel PM may exacerbate allergic
responses to inhaled antigens. The organic fraction of diesel exhaust PM has been linked to
eosinophil degranulation and induction of cytokine production, suggesting that the organic
constituents of diesel PM are the responsible part for the immune effects. It is important to
compare the immune effects of diesel PM to other combustion source-specific emissions, as well
as CAPs, to determine the relative potency of diesel exhaust PM in contributing to the incidence
and severity of allergic rhinitis and asthma. It is also notable that rather direct evidence has been
obtained which demonstrates adherence of allergen-laden pollen cytoplasm fragments to diesel
particles, providing a likely mechanism by which diesel PM acts to concentrate bioaerosol
materials and to increase their focal accumulation in lower regions of the respiratory tract. The
adherence of allergen-laden pollen to other types of PM has not been investigated. Other
evidence substantiates mutagenic/genotoxic effects of diesel emission particles (e.g., PAHs),
consistent with qualitative findings in several studies of increased lung cancer effects being
associated with long-term, occupational exposure to diesel emissions.
7.10.2.5 Organic Compounds
Published research on the acute effects of particle-associated organic carbon constituents is
conspicuous by its relative absence, except for diesel exhaust particles. Like metals, organics are
common constituents of combustion-generated particles and have been found in ambient PM
samples over a wide geographical range. Organic carbon constituents comprise a substantial
portion of the mass of ambient PM (10 to 70% of the total dry mass [Turpin, 1999]). The
organic fraction of ambient PM has been evaluated for its mutagenic effects. Although the
organic fraction of ambient PM is a poorly characterized heterogeneous mixture of an unknown
number of different compounds, organic compounds remain a potential causal property for PM
health effects due to the contribution of exhaust particles from various sources to the fine PM
fraction (Dreher, 2000). Strategies have been proposed for examining the health effects of this
potentially important constituent (Turpin, 1999).
7-219
-------
7.10.2.6 Ultrafine Particles
Studies of various types of ultrafme particles have demonstrated a significantly greater
inflammatory response than that seen with fine particles of the same chemical composition at
similar mass doses (Oberdorster et al., 1992; Li et al., 1996, 1997, 1999).
In other more limited studies, ultrafmes also have generated greater oxidative stress in
experimental animals. Inhalation exposure of normal rats to ultrafme carbon particles generated
by electric arc discharge (100 |ig/m3 for 6 h) caused minimal lung inflammation per unit mass
(Elder et al., 2000a,b), compared to ultrafme PTFE or metal particles. On the other hand,
instillation of 125 jig of ultrafme CB (20 nm) caused substantially more inflammation per unit
mass than did the same dose of fine particles of CB (200 to 250 nm), suggesting that ultrafme
particles may cause more inflammation per unit mass than larger particles (Li et al., 1997).
However, the chemical constituents of the two sizes of CB used in this study were not analyzed,
and it cannot be assumed that the chemical composition was the same. Further, when the
particle surface area is used as a dose metric, the inflammatory response to both fine and
ultrafme particles may be basically the same (Oberdorster, 1996b; Oberdorster et al., 2000; Li
etal., 1996).
With regard to acid aerosols, studies of low concentrations of ultrafme sulfuric acid and
metal oxide particles have demonstrated effects in the lung. However, it is possible that inhaled
ultrafme particles may have systemic effects that are independent of effects on the lung. Thus,
there is still insufficient toxicological evidence to elucidate clearly the extent to which ambient
concentrations or high number counts of ultrafme particles may differentially contribute to
increased health effects risks associated with ambient PM air pollution.
7.10.2.7 Bioaerosols
Bioaerosols, from sources such as plants, fungi, and microorganisms, range in size from
0.01 to |im to > 20 |im. They comprise a small fraction of ambient PM, but have been shown to
contribute to the adverse health affects from PM exposure. Pollen from flowering plants, trees
and grasses, deposits in upper airways to induce allergic rhinitis. Allergen-ladened cytoplasmic
fragments of ruptured pollen grains can enter the deep lung, where they can exacerbate asthma.
Synergistic interactions between pollen debris and other ambient PM (e.g., the polycyclic
hydrocarbon component of DE) are thought to be a mechanism that may explain the increased
7-220
-------
incidence of asthma morbidity and mortality. Human handling and burning of plant material
contributes to increased bioaerosol levels, which have been shown to have adverse health effects.
Fungi and fungal spores are the largest and the most consistently present outdoor bioaerosol.
They cause allergic rhinitis and asthma, which is highly dependent on seasonal variations in
concentration. Exposures have been linked to asthma hospitalization and death. Bacterial and
viruses are significant bioaerosols. Much of the toxicity of bacteria is due to the endotoxins
present in the outer cell membrane, which trigger production of cytokines and a cascade of
inflammation. Concentrations of endotoxins are seasonal (higher under warm, humid
conditions) and tend to be higher in samples of coarse-mode than in fine-mode ambient PM.
Another cell wall component of bacteria and fungi, (l^S)-p-D-glucan, has also been shown to
cause respiratory inflammation.
7.10.3 PM Interactions with Gaseous Co-Pollutants
Particulate matter exists in an atmospheric milieu of ubiquitous co-pollutant gases, all of
which have the potential for antagonistic, additive, or synergistic interactions with PM and
which can modify the toxicologic health effects. The mechanisms by which interactions
between PM and gases are thought likely to occur are by: (1) formation of secondary products
by chemical interactions between the gas and the particle, (2) adherence of material to the
particle and subsequent transport to sensitive sites, and/or 3) pollutant-induced change in the
local microenviroment of the lung (e.g., by decreasing the pH). All of these interactions have
the potential to create antagonistic, additive, or synergistic interactions between PM and gases,
which could potentially greatly modify their individual effects.
New controlled human exposure toxicology studies provide interesting findings on effects
of combined exposures to PM and other pollutants. One study, by Linn et al. (1997), found a
positive association between acid concentration and respiratory symptoms (but not spirometry)
among allergic/asthmatic children (in comparison to clean-air exposure results) following a
single 4-h exposure to 60 to 140 |ig/m3 H2SO4, 0.1 ppm SO2, and 0.1 ppm O3 while undergoing
intermittent exercise. No changes were seen among healthy children. However, the
experimental design did not include comparison of effects of the overall mixture versus those of
individual components (e.g., H2SO4 alone, O3 alone), thus precluding discernment of possible
interactive effects.
7-221
-------
Recent animal studies have also evaluated effects of coexposures to particles and gases.
In one study, both combined CAPs/O3 and O3-alone exposure in a mouse asthma model (Kobzik
et al., 2001) showed similar increases in airway responsiveness and pulmonary resistance, thus
indicating a lack of synergism with the combined exposure. Mixtures of elemental carbon
particles, O3, and ammonium bisulfate showed enhanced changes in lung collagen, AM
respiratory burst, and phagocytosis (Kleinman et al, 2000), although the results were ambiguous
as to whether PM was enhancing the effects of O3 or the converse. A short exposure of
combined carbon particle/SO2 caused depressed AM phagocytosis and suppressed
intrapulmonary bactericidal activity which lasted for a week (Jakab et al., 1996; Clarke et al.,
2000c). Other studies using co-exposures of PM and gases have demonstrated no changes in
histopathological (Moss et al., 2001) or biochemical and morphometric endpoints (Last and
Pinkerton, 1997). Additionally, 4-week exposures of rats to a mixture of carbon particles,
ammonium bisulfate, and O3 caused no changes in BAL parameters and only a small decrease in
plasma fibronectin compared to O3 alone (Bolarin et al., 1997).
7.10.4 Susceptibility
Progress has been made in understanding the role of individual susceptibility to
ambient PM effects. Studies have consistently shown that older animals or animals with certain
types of compromised health, either genetic or induced, are more susceptible to instilled or
inhaled particles, although the increased animal-to-animal variability in these models has created
greater uncertainty for the interpretation of the findings (Clarke et al., 1999, 2000a,b; Kodavanti
et al, 1998b, 2000a, 2001; Gordon et al., 2000; Ohtsuka et al., 2000b; Wesselkamper et al., 2000;
Leikauf et al., 2000; Saldiva et al., 2002). Moreover, because PM seems to affect broad
categories of disease states, ranging from cardiac arrhythmias to pulmonary infection, it is
difficult to know what disease models to use in evaluating the biological plausibility of adverse
health effects of PM.
Compromised hosts are a significant susceptible population which includes individuals
with asthma, individuals with pneumonia or other lung infections, and the elderly with chronic
cardiopulmonary disease. To better understand the effects of PM in this population, researchers
are increasingly using compromised host models such as a rat model of cardiopulmonary disease
that utilize MCT-treatment to induce pulmonary vasculitis/hypertension. This model has
7-222
-------
demonstrated ROFA-induced increased neutrophilic inflammation, exacerbated lung lesions,
increased lung edema, alveolar thickening, and decreased phagocytosis of particles in some
studies (Costa and Dreher, 1997; Kodavanti et al., 1999; Madl et al., 1998). Another report
using MCT-treated rats did not find similar inflammatory responses or changes in pulmonary
function following CAPs exposures (Gordon et al., 2000). Though the MCT-treated rat has been
used extensively for modeling human cardiopulmonary disease, it does have limitations. There
is clearly a need for new and better animal models to examine human pathophysiology
associated with PM exposure.
Animals infected with bacteria or viruses are used to model humans with respiratory
infections; and they have been shown to have increased inflammatory response with PM
exposure (Kodavanti et al., 1998b; Elder et al., 2000a,b). Rats pre-treated with high SO2
exposures have been used to model chronic bronchitis and have shown interaction of CAPs with
preexisting lung injury which produce changes in cellular and biochemical markers in lavage
fluid and increases in tidal volume (Clarke et al., 1999; Saldiva et al., 2002; Kodavanti et al.,
1998b).
Genetic susceptibility to the effects of PM are becoming increasingly apparent as various
strains of rodents are characterized for strain-specific responses. Rat strains such as SD, Fischer-
344, and Wistar demonstrate unique inflammatory and histological responses to ROFA
(Kodavanti et al., 1996, 1997b). Also, genetically predisposed SH rats have been used to model
cardiovascular disease to evaluate potential increased susceptibility to PM-associated effects.
SH rats demonstrate greater oxidative stress and cardiovascular responses than their normal
counterparts in response to ROFA exposure. Inter-strain differences in airway
hyperresponsiveness, inflammation, Fc-receptor-mediated AM phagocytosis, and mortality, too,
have been demonstrated in mouse strains such as BALB/c, C57BL/6, C3FI/HeJ, and others
(Veronesi et al., 2000; Ohtsuka et al., 2000a,b; Prows et al., 1997; Leikauf et al., 2000,
Wesselkamper, et al., 2000) in response to PM exposure. The extent to which genetic
susceptibility plays a role in humans remains to be determined. Use of genomics, proteomics,
and bioinformatics technologies will allow further characterization of the differences in
susceptibility to PM.
7-223
-------
Newly available studies also suggest that individuals with allergic disorders are likely more
susceptible to PM effects than are nonallergic persons. Relatively little is known about the
effects of inhaled PM on humoral (antibody) or cell-mediated immunity. However, both in vivo
and in vitro studies have shown that various types of PM can alter immune responses to
challenge to antigens and may act as an adjuvant (Van Zijverden et al., 2000; Van Zijverdan and
Granum, 2000). Steerenberg et al. (2003) found adjuvant activity to be associated with several
types of PM tested (e.g., road tunnel dust, ROFA, DPM).
ROFA has been shown to enhance allergic sensitization in a number of studies (e.g.,
Hamada et al., 1999; Lambert., 1999; Gilmour et al., 2001), but the applicability of these
findings to ambient PM is not certain. A study has shown that a single exposure to ROFA elicits
a greater effect on airway hyperresponsiveness than a 3-day exposure to CAPs (Goldsmith et al.,
1999).
Particularly interesting new findings point toward ambient PM exacerbation of allergic
airway hyperresponsiveness and/or antigen-induced immune responses. Both metals and diesel
particles have been implicated, with an expanding array of new studies showing DPM as one
particle that is effective in exacerbating allergic asthma responses (Takano et al., 1997; Nel
et al., 2001; Van Zijerden et al., 2000, 2001; Walters et al., 2001; Nordenhall et al., 2001;
Hamada et al., 1999, 2000; Lambert et al., 1999; Gilmour et al., 2001).
7-224
-------
REFERENCES
Adamson, I. Y. R.; Prieditis, H.; Hedgecock, C.; Vincent. R. (2000) Zinc is the toxic factor in the lung response to
an atmospheric paniculate sample. Toxicol. Appl. Pharmacol. 166: 111-119.
Agopyan, N.; Li, L.; Yu, S.; Simon, S. A. (2003) Negatively charged 2- and 10-|_im particles activate vanilloid
receptors, increase cAMP, and induce cytokine release. Toxicol. Appl. Pharmacol. 186: 63-76.
Alink, G. M; Sjogren, M; Bos, R. P.; Doekes, G.; Kromhout, H.; Scheepers, P. T. J. (1998) Effect of airborne
particles from selected indoor and outdoor environments on gap-junctional intercellular communication.
Toxicol. Lett. 96/97: 209-213.
Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects.
In: Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC. Environ.
Health Perspect. 79: 147-150.
Amdur, M. O.; Dubriel, M.; Creasia, D. A. (1978) Respiratory response of guinea pigs to low levels of sulfuric acid.
Environ. Res. 15: 418-423.
Andrews, N. P.; Gralnick, H. R.; Merryman, P.; Vail, M.; Quyyumi, A. A. (1996) Mechanisms underlying the
morning increase in platelet aggregation: a flow cytometry study. J. Am. Coll. Cardiol. 28: 1789-1795.
Antonini, J. M.; Roberts, J. R.; Jernigan, M. R.; Yang, H.-M; Ma, J. Y. C.; Clarke, R. W. (2002) Residual oil fly ash
increases the susceptibility to infection and severely damages the lungs after pulmonary challenge with a
bacterial pathogen. Toxicol. Sci. 70: 110-119.
Arimoto, T.; Yoshikawa, T.; Takano, H.; Kohno, M. (1999) Generation of reactive oxygen species and
8-hydroxy-2'-deoxyguanosine formation from diesel exhaust particle components in L1210 cells. Jpn. J.
Pharmacol. 80: 49-54.
Ball, B. R.; Smith, K. R.; Veranth, J. M.; Aust, A. E. (2000) Bioavailability of iron from coal fly ash: mechanisms of
mobilization and of biological effects. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
Toxicol. 12(suppl. 4): 209-225.
Baluk, P.; Nadel, J. A.; McDonald, D. M. (1992) Substance P-immunoreactive sensory axons in the rat respiratory
tract: a quantitative study of their distribution and role in neurogenic inflammation. J. Compar. Neurol.
319:586-598.
Barfknecht, T. R.; Hites, R. A.; Cavaliers, E. L.; Thilly, W. G. (1982) Human cell mutagenicity of polycyclic
aromatic hydrocarbon components of diesel emissions. In: Lewtas, J., ed. Toxicological effects of emissions
from diesel engines: proceedings of the Environmental Protection Agency 1981 diesel emissions symposium;
October 1981; Raleigh, NC. New York, NY: Elsevier Biomedical; pp. 277-294. (Developments on toxicology
and environmental science: v. 10).
Barnhart, M. I.; Chen, S.-T.; Salley, S. O.; Puro, H. (1981) Ultrastructure and morphometry of the alveolar lung of
guinea pigs chronically exposed to diesel engine exhaust: six month's experience. J. Appl. Toxicol. 1: 88-103.
Barnhart, M. I.; Salley, S. O.; Chen, S.-T.; Puro, H. (1982) Morphometric ultrastructural analysis of alveolar lungs of
guinea pigs chronically exposed by inhalation to diesel exhaust (DE). In: Lewtas, J., ed. Toxicological effects
of emissions from diesel engines: proceedings of the Environmental Protection Agency diesel emissions
symposium; October, 1981; Raleigh, NC. New York, NY: Elsevier Biomedical; pp. 183-200. (Developments
in toxicology and environmental science: v. 10).
Bayram, H.; Devalia, J. L.; Khair, O. A.; Abdelaziz, M. M.; Sapsford, R. J.; Sagai, M.; Davies, R. J. (1998a)
Comparison of ciliary activity and inflammatory mediator release from bronchial epithelial cells of nonatopic
nonasthmatic subjects and atopic asthmatic patients and the effect of diesel exhaust particles in vitro.
J. Allergy Clin. Immunol. 102: 771-782.
Bayram, H.; Devalia, J. L.; Sapsford, R. J.; Ohtoshi, T.; Miyabara, Y.; Sagai, M.; Davies, R. J. (1998b) The effect of
diesel exhaust particles on cell function and release of inflammatory mediators from human bronchial
epithelial cells in vitro. Am. J. Respir. Cell Mol. Biol. 18: 441-448.
Becker, S.; Soukup, J. M. (1998) Decreased CD11B expression, phagocytosis, and oxidative burst in urban
paniculate pollution-exposed human monocytes and alveolar macrophages. J. Toxicol. Environ. Health Part A
55: 455-477.
Becker, S.; Soukup, J. M.; Gilmour, M. I.; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
Pharmacol. 141:637-648.
Becker, S.; Fenton, M. J.; Soukup, J. M. (2002) Involvement of microbial components and toll-like receptors 2 and
4 in cytokine responses to air pollution particles. Am. J. Respir. Cell Mol. Biol. 27: 611-618.
7-225
-------
Becker, S.; Soukup, J. M.; Sioutas, C.; Cassee, F. R. (2003) Response of human alveolar macrophages to ultrafme,
fine and coarse urban air pollution particles. Exp. Lung Res. 29: 29-44.
Behrendt, H.; Becker, W. M.; Friedrichs, K. H.; Darsow, U.; Tomingas, R. (1992) Interaction between aeroallergens
and airborne particulate matter. Int. Arch. Allergy Immunol. 99: 425-428.
Behrendt, H.; Friedrichs, K.-H.; Kramer, U.; Hitzfeld, B.; Becker, W.-M.; Ring, J. (1995) The role of indoor and
outdoor air pollution in allergic diseases. In: Johansson, S. G. O., ed. Prog. Allergy Clin. Immunol.,
Proceedings of the 15th international congress, Allergol. Clin. Immunol.; 1994. Seattle, WA: Hogrefe &
Huber; pp. 83-89.
Behrendt, H.; Becker, W. M.; Fritzsche, C.; Sliwa-Tomczok, W.; Tomczok, J.; Friedrichs, K. H.; Ring, J. (1997) Air
pollution and allergy: experimental studies on modulation of allergen release from pollen by air pollutants.
Int. Arch. Allergy Immunol. 113: 69-74.
Behrendt, H.; Kramer, U.; Schafer, T.; Kasche, A.; Eberlein-Konig, B.; Darsow, U.; Ring, J. (2001)
Allergotoxicology-a research concept to study the role of environmental pollutants in allergy. Allergy Clin.
Immunol. Int. 13: 122-128.
Belisario, M. A.; Buonocore, V.; De Marinis, E.; De Lorenzo, F. (1984) Biological availability of mutagenic
compounds adsorbed onto diesel exhaust particulate. Mutat. Res. 135: 1-9.
Bellomo, R.; Gigliotti, P.; Treloar, A.; Holmes, P.; Suphioglu, C.; Singh, M. B.; Knox, Bruce. (1992) Two
consecutive thunderstorm associated epidemics of asthma in the city of Melbourne: the possible role of rye
grass pollen. Med. J. Aust. 156: 834-837.
Bice, D. E.; Seagrave, JC.; Green, F. H. Y. (2000) Animal models of asthma: potential usefulness for studying health
effects of inhaled particles. Inhalation Toxicol. 12: 829-862.
Blomberg, A.; Sainsbury, C.; Rudell, B.; Frew, A. J.; Holgate, S. T.; Sandstrom, T.; Kelly, F. J. (1998) Nasal cavity
lining fluid ascorbic acid concentration increases in healthy human volunteers following short term exposure
to diesel exhaust. Free Radical Res. 28: 59-67.
Bolarin, D. M.; Bhalla, D. K.; Kleinman, M. T. (1997) Effects of repeated exposures of geriatric rats to ozone and
particle-containing atmospheres: an analysis of bronchoalveolar lavage and plasma proteins. Inhalation
Toxicol. 9: 423-434.
Bond, J. A.; Mauderly, J. L.; Wolff, R. K. (1990) Concentration- and time-dependent formation of DNA adducts in
lungs of rats exposed to diesel exhaust. Toxicology 60: 127-135.
Bonner, J. C.; Rice, A. B.; Lindroos, P. M.; O'Brien, P. O.; Dreher, K. L.; Rosas, I.; Alfaro-Moreno, E.;
Osornio-Vargas, A. R. (1998) Induction of the lung myofibroblast PDGF receptor system by urban ambient
particles from Mexico City. Am. J. Respir. Cell Mol. Biol. 19: 672-680.
Bouthillier, L.; Vincent, R.; Goegan, P.; Adamson, I. Y. R.; Bjarnason, S.; Stewart, M.; Guenette, J.; Potvin, M.;
Kumarathasan, P. (1998) Acute effects of inhaled urban particles and ozone: lung morphology, macrophage
activity, and plasma endothelin-1. Am. J. Pathol. 153: 1873-1884.
Brain, J. D.; Long, N. C.; Wolfthal, S. F.; Dumyahn, T.; Dockery, D. W. (1998) Pulmonary toxicity in hamsters of
smoke particles from Kuwaiti oil fires. Environ. Health Perspect. 106: 141-146.
Braunwald, E. (1997) Cardiovascular medicine at the turn of the millennium: triumphs, concerns, and opportunities.
N. Engl. J. Med. 337: 1360-1369.
Brightwell, J.; Fouillet, X.; Cassano-Zoppi, A.-L.; Gatz, R.; Duchosal, F. (1986) Neoplastic and functional changes
in rodents after chronic inhalation of engine exhaust emissions. In: Ishinishi, N.; Koizumi, A.; McClellan, R.
O.; Stober, W., eds. Carcinogenic and mutagenic effects of diesel engine exhaust: proceedings of the
international satellite symposium on lexicological effects of emissions from diesel engines; July; Tsukuba
Science City, Japan. Amsterdam, Holland: Elsevier Science Publishers B. V.; pp. 471-485. (Developments in
toxicology and environmental science: v. 13).
Broeckaert, F.; Buchet, J. P.; Huaux, F.; Lardot, C.; Lison, D.; Yager, J. W. (1997) Reduction of the ex vivo
production of tumor necrosis factor alpha by alveolar phagocytes after administration of coal fly ash and
copper smelter dust. J. Toxicol. Environ. Health 51: 189-202.
Brook, R. D.; Brook, J. R.; Urch, B.; Vincent, R.; Rajagopalan, S.; Silverman, F. (2002) Inhalation of fine particulate
air pollution and ozone causes acute arterial vasoconstriction in health adults. Circulation 105: 1534-1536.
Brooks, A. L.; Li, A. P.; Dutcher, J. S.; Clark, C. R.; Rothenberg, S. J.; Kiyoura, R.; Bechtold, W. E.; McClellan,
R. O. (1984) A comparison of genotoxicity of automobile exhaust particles from laboratory and
environmental sources. Environ. Mutagen. 6: 651-668.
Brunekreef, B.; Hoek, G.; Fischer, P.; Spieksma, F. T. M. (2000) Relation between airborne pollen concentrations
and daily cardiovascular and respiratory-disease mortality. Lancet 355: 1517-1518.
7-226
-------
Hunger, I; Krahl, I; Baum, K.; Schroder, O.; Mtiller, M.; Westphal, G.; Ruhnau, P.; Schulz, T. G.; Hallier, E.
(2000) Cytotoxic and mutagenic effects, particle size and concentration analysis of diesel engine emissions
using biodiesel and petrol diesel as fuel. Arch. Toxicol. 74: 490-498.
Burge, H. A. (1995) Bioaerosols in the residential environment. In: Cox, C. S.; Wathes, C. M., eds. Bioaerosols
handbook. Boca Raton, FL: CRC Press, Inc.; pp. 579-597.
Campen, M. I; Costa, D. L.; Watkinson, W. P. (2000) Cardiac and thermoregulatory toxicity of residual oil fly ash
in cardiopulmonary-compromised rats. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the third
colloquium on paniculate air pollution and human health (second special issue); June, 1999; Durham, NC.
Inhalation Toxicol. 12(suppl. 2): 7-22.
Campen, M. I; Nolan, J. P.; Schladweiler, M. C. I; Kodavanti, U. P.; Evansky, P. A.; Costa, D. L.; Watkinson,
W. P. (2001) Cardiovascular and thermoregulatory effects of inhaled PM-associated transition metals:
a potential interaction between nickel and vanadium sulfate. Toxicol. Sci. 64: 243-252.
Campen, M. J.; Nolan, J. P.; Schladweiler, M. C. J.; Kodavanti, U. P.; Costa, D. L.; Watkinson, W. P. (2002)
Cardiac and thermoregulatory effects of instilled paniculate matter-associated transition metals in healthy and
cardiopulmonary-compromised rats. J. Toxicol. Environ. Health Part A 65:1615-1631.
Campen, M. J.; McDonald, J. D.; Gigliotti, A. P.; Seilkop, S. K.; Reed, M. D.; Benson, J. M. (2003) Cardiovascular
effects of inhaled diesel exhaust in spontaneously hypertensive rats. Cardiovasc. Toxicol. 3: 353-361.
Carter, J. D.; Ohio, A. J.; Samet, J. M.; Devlin, R. B. (1997) Cytokine production by human airway epithelial cells
after exposure to an air pollution particle is metal-dependent. Toxicol. Appl. Pharmacol. 146: 180-188.
Cassee, F. R.; Dormans, J. A. M. A.; VanLoveren, H.; VanBree, L.; Rombout, P. J. A. (1998a) Toxicity of ambient
paniculate matter (PM10). I. Acute toxicity study in asthmatic mice following 3-day exposure to ultrafine and
fine ammonium bisulfate, a model compounds for secondary aerosol fraction of PM10. Bilthoven, The
Netherlands: National Institute of Public Health and the Environment (RIVM); report no. 650010 010.
Cassee, F. R.; Dormans, J. A. M. A.; VanLoveren, H.; VanBree, L.; Rombout, P. J. A. (1998b) Toxicity of ambient
paniculate matter. II. Acute toxicity study in asthmatic mice following 3-day exposure to fine ammonium
ferrosulfate, a model compounds for secondary aerosol fraction of PM10. Bilthoven, The Netherlands:
National Institute of Public Health and the Environment (RIVM); report no. 650010 Oil.
Cassee, F. R.; Dormans, J. A. M. A.; VanLoveren, H.; VanBree, L.; Rombout, P. J. A. (1998c) Toxicity of ambient
paniculate matter. III. Acute toxicity study in (asthmatic) mice following 3-day exposure to ultrafine and fine
ammonium nitrate, a model compound for secondary aerosol fraction of PM10. Bilthoven, The Netherlands:
National Institute of Public Health and the Environment (RIVM); report no.651101013.
Cassee, F. R.; Arts, J. H. E.; Fokkens, P. H. B.; Spoor, S. M.; Boere, A. J. F.; VanBree, L.; Dormans, J. A. M. A.
(2002) Pulmonary effects of ultrafine and fine ammonium salts aerosols in healthy and monocrotaline-treated
rats following short-term exposure. Inhalation Toxicol. 14: 1215-1229.
Castellan, R. M.; Olenchock, S. A.; Hankinson, J. L.; Millner, P. D.; Cocke, J. B.; Bragg, C. K.; Perkins, H. H., Jr.;
Jacobs, R. R. (1984) Acute bronchoconstriction induced by cotton dust: dose-related responses to endotoxin
and other dust factors. Ann. Intern. Med. 102: 157-163.
Castellan, R. M.; Olenchock, S. A.; Kinsley, K. B.; Hankinson, J. L. (1987) Inhaled endotoxin and decreased
spirometric values: an exposure-response relation for cotton dust. N. Engl. J. Med. 317: 605-610.
Casto, B. C.; Hatch, G. G.; Huang, S. L.; Lewtas, J.; Nesnow, S.; Waters, M. D. (1981) Mutagenic and carcinogenic
potency of extracts of diesel and related environmental emissions: in vitro mutagenesis and oncogenic
transformation. Environ. Int. 5: 403-409.
Celenza, A.; Fothergill, J.; Kupek, E.; Shaw R. J. (1996) Thunderstorm associated asthma: a detailed analysis of
environmental factors. Br. Med. J. 312: 604-607.
Chang, M. C.; Sioutas, C.; Kim, S.; Gong, H., Jr.; Linn, W. S. (2000) Reduction of nitrate losses from filter and
impactor samplers by means of concentration enrichment. Atmos. Environ. 34: 85-98.
Chapman, R. S.; Mumford, J. L.; Harris, D. B.; He, X.; Jiang, W.; Yang, R. (1988) The epidemiology of lung cancer
in Xuan Wei, China: current progress, issues, and research strategies. Arch. Environ. Health 43: 180-185.
Chen, L. C.; Fine, J. M.; Qu, Q.-S.; Amdur, M. O.; Gordon, T. (1992) Effects of fine and ultrafine sulfuric acid
aerosols in guinea pigs: alterations in alveolar macrophage function and intracellular pH. Toxicol. Appl.
Pharmacol. 113: 109-117.
Chescheir, G. M., Ill; Garrett, N. E.; Shelburne, J. D.; Huisingh, J. L.; Waters, M. D. (1981) Mutagenic effects of
environmental particulates in the CHO/HGPRT system. In: Waters, M. D.; Sandhu, S. S.; Huisingh, J. L.;
Claxton, L.; Newnow, S., eds. Short-term bioassays in the analysis of complex environmental mixtures II.
New York, NY: Plenum Press; pp. 337-350. (Environmental science research: v. 22).
Churg, A.; Brauer, M.; Keeling, B. (1996) Ozone enhances the uptake of mineral particles by tracheobronchial
epithelial cells in organ culture. Am. J. Respir. Crit. Care Med. 153: 1230-1233.
7-227
-------
Churg, A.; Stevens, B.; Wright, J. L. (1998) Comparison of the uptake of fine and ultrafine TiO2 in a tracheal explant
system. Am. J. Physiol. 274: L81-L86.
Clark, S. (1986) Report on prevention and control. Am. J. Ind. Med. 10: 267-273.
Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.; Sioutas, C.; Paulauskis, J.; Coull, B.; Ferguson,
S.; Godleski, J. J. (1999) Urban air paniculate inhalation alters pulmonary function and induces pulmonary
inflammation in a rodent model of chronic bronchitis. Inhalation Toxicol. 11: 637-656.
Clarke, R. W.; Coull, B.; Reinisch, U.; Catalano, P.; Killingsworth, C. R.; Koutrakis, P.; Kavouras, L; Murthy,
G. G. K.; Lawrence, J.; Lovett, E.; Wolfson, J. M; Verrier, R. L.; Godleski, J. J. (2000a) Inhaled concentrated
ambient particles are associated with hematologic and bronchoalveolar lavage changes in canines. Environ.
Health Perspect. 108: 1179-1187.
Clarke, R. W.; Catalano, P.; Coull, B.; Koutrakis, P.; Krishna Murthy, G. G.; Rice, T.; Godleski, J. J. (2000b)
Age-related responses in rats to concentrated urban air particles (CAPs). Inhalation Toxicol.
12(suppl. 1): 73-84.
Clarke, R. W.; Antonini, J. M.; Hemenway, D. R.; Frank, R.; Kleeberger, S. R.; Jakab, G. J. (2000c) Inhaled
particle-bound sulfate: effects on pulmonary inflammatory responses and alveolar macrophage function.
Inhalation Toxicol. 12: 169-186.
Claxton, L. D. (1981) Mutagenic and carcinogenic potency of diesel and related environmental emissions:
Salmonella bioassay. Environ. Int. 5: 389-391.
Claxton, L. D. (1983) Characterization of automotive emissions by bacterial mutagenesis bioassay: a review.
Environ. Mutagen. 5: 609-631.
Claxton, L.; Kohan, M. (1981) Bacterial mutagenesis and the evaluation of mobile-source emissions. In: Waters,
M. D.; Sandhu, S. S.; Huisingh, J. L.; Claxton, L.; Nesnow, S., eds. Short-termbioassays in the analysis of
complex environmental mixtures II: proceedings of the second symposium on the application of short-term
bioassays in the fractionation and analysis of complex environmental mixtures; March 1980; Williamsburg,
VA. New York, NY: Plenum Press; pp. 299-317. (Hollaender, A.; Welch, B. L.; Probstein, R. F., eds.
Environmental science research series: v. 22).
Cohen, M. D.; Becker, S.; Devlin, R.; Schlesinger, R. B.; Zelikoff, J. T. (1997) Effects of vanadium upon
polyl:C-induced responses in rat lung and alveolar macrophages. J. Toxicol. Environ. Health 51: 591-608.
Conn, C. A.; Green, F. H. Y.; Nikula, K. J. (2000) Animal models of pulmonary infection in the compromised host:
potential usefulness for studying health effects of inhaled particles. Inhalation Toxicol. 12: 783-827.
Costa, D. L. (2000) Paniculate matter and cardiopulmonary health: a perspective. Inhalation Toxicol.
12(suppl. 3): 35-44.
Costa, D. L.; Dreher, K. L. (1997) Bioavailable transition metals in paniculate matter mediate cardiopulmonary
injury in healthy and compromised animal models. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of
the sixth international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1053-1060.
Crebelli, R.; Conti, L.; Crochi, B.; Carera, A.; Bertoli, C.; Del Giacomo, N. (1995) The effect of fuel composition on
the mutagenicity of diesel engine exhaust. Mutat. Res. 346: 167-172.
Curren, R. D.; Kouri, R. E.; Kim, C. M.; Schechtman, L. M. (1981) Mutagenic and carcinogenic potency of extracts
from diesel related environmental emissions: simultaneous morphological transformation and mutagenesis in
BALB/c 3T3 cells. Environ. Int. 5: 411-415.
Dalai, N. S.; Suryan, M. M.; Vallyathan, V.; Green, F. H.; Jafari, B.; Wheeler, R. (1989) Detection of reactive free
radicals in fresh coal mine dust and their implication for pulmonary injury. Ann. Occup. Hyg. 33: 79-84.
Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154: 633-641.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med. 155:
568-576.
Diaz-Sanchez, D.; Tsien, A.; Casillas, A.; Dotson, A. R.; Saxon, A. (1996) Enhanced nasal cytokine production in
human beings after in vivo challenge with diesel exhaust particles. J. Allergy Clin. Immunol. 98: 114-123.
Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997) Combined diesel exhaust paniculate and ragweed
allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
Di Minno, G.; Mancini, M. (1990) Measuring plasma fibrinogen to predict stroke and mycardial infarction.
Arteriosclerosis 10: 1-7.
7-228
-------
Diociaiuti, M; Balduzzi, M; De Berardis, B.; Cattani, G.; Stacchini, G.; Ziemacki, G.; Marconi, A.; Paoletti, L.
(2001) The two PM25 (fine) andPM25-10 (coarse) fractions: evidence of different biological activity. Environ.
Res. A 86: 254-262.
Donaldson, K.; Brown, D. M.; Mitchell, C.; Dineva, M.; Beswick, P. H.; Gilmour, P.; MacNee, W. (1997) Free
radical activity of PM10: iron-mediated generation of hydroxyl radicals. In: Driscoll, K. E.; Oberdorster, G.,
eds. Proceedings of the sixth international meeting on the toxicology of natural and man-made fibrous and
non-fibrous particles; September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1285-1289
Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor a expression from alveolar
macrophages treated with urban air particles. Exp. Lung Res. 22: 577-592.
Dormans, J. A. M. A.; Steerenberg, P. A.; Arts, J. H. E.; Van Bree, L.; De Klerk, A.; Verlaan, A. P. J.; Bruijntjes,
J. P.; Beekhof, P.; Van Soolingen, D.; VanLoveren, H. (1999) Pathological and immunological effects of
respirable coal fly ash in male wistar rats. Inhalation Toxicol. 11:51-69.
Douwes, J.; Zuidhof, A.; Doekes, G.; Van der Zee, S.; Wouters, I.; Boezen, H. M.; Brunekreef, B. (2000)
(l->3)-p-D-glucan and endotoxin in house dust and peak flow variability in children. Am. J. Respir. Crit.
Care Med. 162: 1348-1354.
Dreher, K. L. (2000) Paniculate matter physicochemistry and toxicology: in search of causality—a critical
perspective. Inhalation Toxicol. 12(suppl. 3): 45-57.
Dreher, K. L.; Jaskot, R. H.; Lehmann, J. R.; Richards, J. H.; McGee, J. K.; Ohio, A. J.; Costa, D. L. (1997)
Soluable transition metals mediate residual oil fly ash induced acute lung injury. J. Toxicol. Environ. Health
50: 285-305.
Driscoll, K. E.; Carter, J. M.; Howard, B. W.; Hassenbein, D. G.; Pepelko, W. I.; Baggs, R. B.; Oberdorster, G.
(1996) Pulmonary inflammatory, chemokine, and mutagenic responses in rats after subchronic inhalation of
carbon black. Toxicol. Appl. Pharmacol. 136: 372-380.
Driscoll, K. E.; Deyo, K. C.; Carter, J. M.; Howard, B. W.; Hassenbein, D. G.; Bertram, T. A. (1997) Effects of
particle exposure and particle-elicited inflammatory cells on mutation in rat alveolar epithelial cells.
Carcinogenesis 18: 423-430.
Driscoll, K. E.; Costa, D. L.; Hatch, G.; Henderson, R.; Oberdorster, G.; Salem, H.; Schlesinger, R. B. (2000)
Intratracheal instillation as an exposure technique for the evaluation of respiratory tract toxicity: uses and
limitations. Toxicol. Sci. 55: 24-35.
Dukovich, M.; Yasbin, R. E.; Lestz, S. S.; Risby, T. H.; Zweidinger, R. B. (1981) The mutagenic and SOS-inducing
potential of the soluble organic fraction collected from diesel paniculate emissions. Environ. Mutagen.
3:253-264.
Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1997) Epithelial injury induced by exposure to residual oil
fly-ash particles: role of reactive oxygen species? Am. J. Respir. Cell Mol. Biol. 17: 625-633.
Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1999) Role of soluble metals in oil fly ash-induced airway
epithelial injury and cytokine gene expression. Am. J. Physiol. 277: L498-L510.
Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. J.; Costa, D. L.
(2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
studies in Utah Valley residents. Environ. Health Perspect. Suppl. 109(3): 395-403.
Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000a) Endotoxin priming affects the lung
response to ultrafine particles and ozone in young and old rats. In: Phalen, R. F., ed. Inhalation Toxicology:
proceedings of the third colloquium on paniculate air pollution and human health (first special issue); June,
1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 85-98.
Elder, A. C. P.; Gelein, R.; Finkelstein, J. N.; Cox, C.; Oberdorster, G. (2000b) Pulmonary inflammatory response to
inhaled ultrafine particles is modified by age, ozone exposure, and bacterial toxin. In: Grant, L. D., ed.
PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 227-246.
Elvers, B.; Hawkins, S.; Schulz, G., eds. (1991) Peroxo compounds, inorganic. In: Ullmann's encyclopedia of
industrial chemistry. V. A19. 5th ed. New York, NY: VCH Publishers; pp. 177-197.
Fabiani, R.; Pampanella, L.; Minelli, A.; Mezzasoma, I.; Vecchiarelli, A.; Morozzi, G. (1997) Effect of airborne
paniculate extracts on monocyte oxidative metabolism. J. Environ. Pathol. Toxicol. Oncol. 16: 195-199.
Federal Register. (1997) National ambient air quality standards for paniculate matter; final rule. F. R. (July 18)
62: 38,652-38,752.
Ferin, J.; Oberdorster, G.; Penney, D. P.; Soderholm, S. C.; Gelein, R.; Piper, H. C. (1990) Increased pulmonary
toxicity of ultrafine particles? I. Particle clearance, translocation, morphology. J. Aerosol Sci. 21: 381-384.
Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
Respir. Cell Mol. Biol. 6: 535-542.
7-229
-------
Finkelstein, J. N.; Johnston, C.; Barrett, T.; Oberdorster, G. (1997) Particulate-cell interactions and pulmonary
cytokine expression. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
NY. Environ. Health Perspect. Suppl. 105(5): 1179-1182.
Folinsbee, L. J.; Kim, C. S.; Kehrl, H. R.; Prah, J. D.; Devlin, R. B. (1997) Methods in human inhalation toxicology.
In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press LLC; pp. 607-670.
Frampton, M. W. (2001) Systemic and cardiovascular effects of airway injury and inflammation: ultrafine particle
exposure in humans. Environ. Health Perspect. Suppl. 109(4): 529-532.
Frampton, M. W.; Voter, K. Z.; Morrow, P. E.; Roberts, N. J., Jr.; Gulp, D. J.; Cox, C.; Utell, M. J. (1992) Sulfuric
acid aerosol exposure in humans assessed by bronchoalveolar lavage. Am. Rev. Respir. Dis. 146: 626-632.
Frampton, M. W.; Morrow, P. E.; Cox, C.; Levy, P. C.; Condemi, J. J.; Speers, D.; Gibb, F. R.; Utell, M. J. (1995)
Sulfuric acid aerosol followed by ozone exposure in healthy and asthmatic subjects. Environ. Res. 69: 1-14.
Frampton, M. W.; Ohio, A. J.; Samet, J. M.; Carson, J. L.; Carter, J. D.; Devlin, R. B. (1999) Effects of aqueous
extracts of PM10 filters from the Utah Valley on human airway epithelial cells. Am. J. Physiol.
277: L960-L967.
Friedlander, S. K.; Yeh, E. K. (1998) The submicron atmospheric aerosol as a carrier of reactive chemical species:
case of peroxides. Appl. Occup. Environ. Hyg. 13: 416-420.
Fujimaki, H.; Katayama, N.; Wakamori, K. (1992) Enhanced histamine release from lung mast cells of guinea pigs
exposed to sulfuric acid aerosols. Environ. Res. 58: 117-123.
Fujimaki, H.; Nohara, O.; Ichinose, T.; Watanabe, N.; Saito, S. (1994) IL-4 production in mediastinal lymph node
cells in mice intratracheally instilled with diesel exhaust particulates and antigen. Toxicology 92: 261-268.
Galatius-Jensen, S.; Wroblewski, H.; Emmeluth, C.; Bie, P.; Haunse, S.; Kastrup, J. (1996) Plasma endothelin in
congestive heart failure: a predictor of cardiac death? J. Card. Failure 2: 71-76.
Gallagher, J.; Heinrich, U.; George, M.; Hendee, L.; Phillips, D. H.; Lewtas, J. (1994) Formation of DNA adducts in
rat lung following chronic inhalation of diesel emissions, carbon black and titanium dioxide particles.
Carcinogenesis (London) 15: 1291-1299.
Gardner, S. Y.; Lehmann, J. R.; Costa, D. L. (2000) Oil fly ash-induced elevation of plasma fibrinogen levels in rats.
Toxicol. Sci. 56: 175-180.
Gavett, S. H.; Madison, S. L.; Dreher, K. L.; Winsett, D. W.; McGee, J. K.; Costa, D. L. (1997) Metal and sulfate
composition of residual oil fly ash determines airway hyperreactivity and lung injury in rats. Environ. Res.
72: 162-172.
Gavett, S. H.; Madison, S. L.; Stevens, M. A.; Costa, D. L. (1999) Residual oil fly ash amplifies allergic cytokines,
airway responsiveness, and inflammation in mice. Am. J. Respir. Crit. Care Med. 160: 1897-1904.
Gearhart, J. M.; Schlesinger, R. B. (1986) Sulfuric acid-induced airway hyperresponsiveness. Fundam. Appl.
Toxicol. 7: 681-689.
Gercken, G.; Berg, I.; Dorger, M.; Schluter, T. (1996) Mechanisms of particle-induced activation of alveolar
macrophages. Toxicol. Lett. 88: 121-129.
Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
Am. J. Respir. Crit. Care Med. 164: 704-708.
Ohio, A. J.; Meng, Z. H.; Hatch, G. E.; Costa, D. L. (1997a) Luminol-enhanced chemiluminescence after in vitro
exposures of rat alveolar macrophages to oil fly ash is metal dependent. Inhalation Toxicol. 9: 255-271.
Ohio, A. J.; Piantadosi, C. A.; Crumbliss, A. L. (1997b) Hypothesis: iron chelationplays avital role in neutrophilic
inflammation. BioMetals 10: 135-142.
Ohio, A. J.; Richards, J. H.; Dittrich, K. L.; Samet, J. M. (1998a) Metal storage and transport proteins increase after
exposure of the rat lung to an air pollution particle. Toxicol. Pathol. 26: 388-394.
Ohio, A. J.; Carter, J. D.; Richards, J. H.; Brighton, L. E.; Lay, J. C.; Devlin, R. B. (1998b) Disruption of normal
iron homeostasis after bronchial instillation of an iron-containing particle. Am. J. Physiol. 274: L396-L403.
Ohio, A. J.; Carter, J. D.; Samet, J. M.; Reed, W.; Quay, J.; Dailey, L. A.; Richards, J. H.; Devlin, R. B. (1998c)
Metal-dependent expression of ferritin and lactoferrin by respiratory epithelial cells. Am. J. Physiol.
274: L728-L736.
Ohio, A. J.; Stonehuerner, J.; Dailey, L. A.; Carter, J. D. (1999a) Metals associated with both the water-soluble and
insoluble fractions of an ambient air pollution particle catalyze an oxidative stress. Inhalation Toxicol.
11:37-49.
Ohio, A. J.; Carter, J. D.; Dailey, L. A.; Devlin, R. B.; Samet, J. M. (1999b) Respiratory epithelial cells demonstrate
lactoferrin receptors that increase after metal exposure. Am. J. Physiol. 276: L933-L940.
Ohio, A. J.; Stoneheurner, J.; McGee, J. K.; Kinsey, J. S. (1999c) Sulfate content correlates with iron concentrations
in ambient air pollution particles. Inhalation Toxicol. 11: 293-307.
7-230
-------
Ohio, A. I; Kim, C.; Devlin, R. B. (2000a) Concentrated ambient air particles induce mild pulmonary inflammation
in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
Ohio, A. J.; Carter, J. D.; Richards, J. H.; Crissman, K. M; Bobb, H. H.; Yang, F. (2000b) Diminished injury in
hypotransferrinemic mice after exposure to a metal-rich particle. Am. J. Physiol. 278: L1051-L1061.
Ohio, A. J.; Richards, J. H.; Carter, J. D.; Madden, M. C. (2000c) Accumulation of iron in the rat lung after trachea!
instillation of diesel particles. Toxicol. Pathol. 28: 619-627.
Ohio, A. J.; Gilbey, J. G.; Roggli, V. L.; Richards, J. H.; McGee, J. K.; Carson, J. L.; Devlin, R. B.; Cascio, W. E.
(2001) Diffuse alveolar damage after exposure to an oil fly ash. Am. J. Respir. Crit. Care Med. 164:
1514-1518.
Gift, J. S.; Faust, R. A. (1997) Noncancer inhalation toxicology of crystalline silica: exposure-response assessment.
J. Exposure Anal. Environ. Epidemiol. 7: 345-358.
Gilmour, M. L; Taylor, F. G. R.; Baskerville, A.; Wathes, C. M. (1989a) The effect of titanium dioxide inhalation on
the pulmonary clearance of Pasteurella haemolytica in the mouse. Environ. Res. 50: 157-172.
Gilmour, M. L; Taylor, F. G. R.; Wathes, C. M. (1989b) Pulmonary clearance of Pasteurella haemolytica and
immune responses in mice following exposure to titanium dioxide. Environ. Res. 50: 184-194.
Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
53: 817-822.
Gilmour, M. L; Daniels, M.; McCrillis, R. C.; Winsett, D.; Selgrade, M. K. (2001) Air pollutant-enhanced
respiratory disease in experimental animals. Environ. Health Perspect. Suppl. 109(4): 619-622.
Godleski, J. J.; Verrier, R. L.; Koutrakis, P.; Catalano, P. (2000) Mechanisms of morbidity and mortality from
exposure to ambient air particles. Cambridge, MA: Health Effects Institute; research report no. 91.
Goldsmith, C.-A.; Frevert, C.; Imrich, A.; Sioutas, C.; Kobzik, L. (1997) Alveolar macrophage interaction with air
pollution particulates. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
NY. Environ. Health Perspect. Suppl. 105(5): 1191-1195.
Goldsmith, C.-A. W.; Imrich, A.; Danaee, H.; Ning, Y. Y.; Kobzik, L. (1998) Analysis of air pollution
particulate-mediated oxidant stress in alveolar macrophages. J. Toxicol. Environ. Health Part A 54: 529-545.
Goldsmith, C.-A. W.; Hamada, K.; Ning, Y. Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
(1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
Inhalation Toxicol. 11: 981-998.
Gong, H., Jr.; Sioutas, C.; Linn, W. S.; Clark, K. W.; Terrell, S. L.; Terrell, L. L.; Anderson, K. R.; Kim, S.; Chang,
M.-C. (2000) Controlled human exposures to concentrated ambient fine particles in metropolitan
Los Angeles: methodology and preliminary health-effect findings. In: Phalen, R. F., ed. Inhalation
Toxicology: proceedings of the third colloquium on paniculate air pollution and human health (first special
issue); June, 1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 107-119.
Gong, H., Jr.; Linn, W. S.; Sioutas, C.; Terrell, S. L.; Clark, K. W.; Anderson, K. R.; Terrell, L. L. (2003) Controlled
exposures of healthy and asthmatic volunteers to concentrated ambient fine particles in Los Angeles.
Inhalation Toxicol. 15: 305-325.
Gordon, T.; Nadziejko, C.; Schlesinger, R.; Chen, L. C. (1998) Pulmonary and cardiovascular effects of acute
exposure to concentrated ambient paniculate matter in rats. Toxicol. Lett. 96-97: 285-288.
Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
Granville, C. A.; Hanley, N. M.; Mumford, J. L.; DeMarini, D. M. (2003) Mutation spectra of smoky coal
combustion emissions in Salmonella reflect the TP53 and KRAS mutations in lung tumor from smoky
coal-exposed individuals. Mutat. Res. 525: 77-83.
Gross, K. B. (1981) Pulmonary function testing of animals chronically exposed to diluted diesel exhaust. J. Appl.
Toxicol. 1: 116-123.
Gu, Z. W.; Zhong, B. Z.; Nath, B.; Whong, W.-Z.; Wallace, W. E.; Ong, T.-M. (1992) Micronucleus induction and
phagocytosis in mammalian cells treated with diesel emission particles. Mutat. Res. 279: 55-60.
Gu, Z. W.; Zhong, B. Z.; Keane, M. J.; Whong, W. Z.; Wallace, W. E.; Ong, T. (1994) Induction of unscheduled
DNA synthesis in V79 cells by diesel emission particles dispersed in simulated pulmonary surfactant. In:
Dodgson, J.; McCallum, R. I., eds. Inhaled Particles VII: proceedings of an international symposium on
inhaled particles organised by the British Occupational Hygiene Society; September 1991; Cambridge, UK.
Ann. Occup. Hyg. 38(suppl. 1): 345-349.
Guerrero, R. R.; Rounds, D. E.; Orthoefer, J. (1981) Genotoxicity of Syrian hamster lung cells treated in vivo with
diesel exhaust particulates. Environ. Int. 5: 445-454.
7-231
-------
Hahon, N.; Booth, J. A.; Green, F.; Lewis, T. R. (1985) Influenza vims infection in mice after exposure to coal dust
and diesel engine emissions. Environ. Res. 37: 44-60.
Hamada, K.; Goldsmith, C.-A.; Kobzik, L. (1999) Increased airway hyperresponsiveness and inflammation in a
juvenile mouse model of asthma exposed to air-pollutant aerosol. J. Toxicol. Environ. Health 58: 129-143.
Hamada, K.; Goldsmith, C.-A.; Goldman, A.; Kobzik, L. (2000) Resistance of very young mice to inhaled allergen
sensitization Is overcome by coexposure to an air-pollutant aerosol. Am. J. Respir. Crit. Care Med.
161: 1285-1293.
Hamers, T.; Van Schaardenburg, M. D.; Felzel, E. C.; Murk, A. J.; Koeman, J. H. (2000) The application of reporter
gene assays for the determination of the toxic potency of diffuse air pollution. Sci. Total Environ.
262: 159-174.
Hannigan, M. P.; Cass, G. R.; Penman, B. W.; Crespi, C. L.; Lafleur, A. L.; Busby, W. F., Jr.; Thilly, W. G. (1997)
Human cell mutagens in Los Angeles air. Environ. Sci. Technol. 31: 438-447.
Hannigan, M. P.; Cass, G. R.; Penman, B. W.; Crespi, C. L.; Lafleur, A. L.; Busby, W. F., Jr.; Thilly, W. G.;
Simoneit, B. R. T. (1998) Bioassay-directed chemical analysis of Los Angeles airborne paniculate matter
using a human cell mutagenicity assay. Environ. Sci. Technol. 32: 3502-3514.
Harder, S. D.; Soukup, J. M.; Ohio, A. J.; Devlin, R. B.; Becker, S. (2001) Inhalation of PM25 does not modulate
host defense or immune parameters in blood or lung of normal human subjects. Environ. Health Perspect.
Suppl. 109(4): 599-604.
Hastie, A. T.; Peters, S. P. (2001) Interactions of allergens and irritants in susceptible populations in producing lung
dysfunction: implications for future research. Environ. Health Perspect. 109(suppl. 4): 605-607.
Health Effects Institute. (1995) Diesel exhaust: a critical analysis of emissions, exposure, and health effects:
a special report of the Institute's Diesel Working Group. Cambridge, MA: Health Effects Institute.
Heck, J. D.; Costa, M. (1983) Influence of surface charge and dissolution on the selective phagocytosis of potentially
carcinogenic paniculate metal compounds. Cancer Res. 43: 5652-5656.
Heederik, D.; Douwes, J.; Wouters, I.; Doekes, G. (2000) Organic dusts: beyond endotoxin. Inhalation Toxicol.
12(suppl. 3): 27-33.
Heinrich, U.; Muhle, H.; Takenaka, S.; Ernst, H.; Fuhst, R.; Mohr, U.; Pott, F.; Stober, W. (1986a) Chronic effects
on the respiratory tract of hamsters, mice, and rats after long-term inhalation of high concentrations of filtered
and unfiltered diesel engine emissions. J. Appl. Toxicol. 6: 383-395.
Heinrich, U.; Pott, F.; Rittinghausen, S. (1986b) Comparison of chronic inhalation effects in rodents after long-term
exposure to either coal oven flue gas mixed with pyrolized pitch or diesel engine exhaust. In: Ishinishi, N.;
Koizumi, A.; McClellan, R. O.; Stober, W., eds. Carcinogenic and mutagenic effects of diesel engine exhaust:
proceedings of the international satellite syposium on lexicological effects of emissions from diesel engines;
July; Tsukuba Science City, Japan. Amsterdam, Holland: Elsevier Science Publishers B. V.; pp. 441-457.
(Developments in toxicology and environmental science: v. 13).
Heinrich, U.; Fuhst, R.; Rittinghausen, S.; Creutzenberg, O.; Bellmann, B.; Koch, W.; Levsen, K. (1995) Chronic
inhalation exposure of Wistar rats and two different strains of mice to diesel engine exhaust, carbon black,
and titantium dioxide. Inhalation Toxicol. 7: 533-556.
Heinrich, J.; Hoelscher, B.; Frye, C.; Meyer, L; Pitz, M.; Cyrys, J.; Wjst, M.; Neas, L.; Wichmann, H.-E. (2002a)
Improved air quality in reunified Germany and decreases in respiratory symptoms. Epidemiology 13:
394-401.
Heinrich, J.; Hoelscher, B.; Frye, C.; Meyer, I.; Wjst, M.; Wichmann, H. E. (2002b) Trends in prevalence of atopic
diseases and allergic sensitization in children in eastern Germany. Eur. Respir. J. 19: 1040-1046.
Heinrich, J.; Pitz, M.; Bischof, W.; Krug, N.; Borm, P. J. A. (2003) Endotoxin in fine (PM2 5) and coarse (PM25.10)
particle mass of ambient aerosols. A tempero-spatial analysis. Atmos. Environ. 37: 3659-3667.
Hemminki, K.; Soderling, J.; Ericson, P.; Norbeck, H. E.; Segerback, D. (1994) DNA adducts among personnel
servicing and loading diesel vehicles. Carcinogenesis 15: 767-769.
Heussen, G. A. H.; Bouman, H. G. M.; Roggeband, R.; Baan, R. A.; Alink, G. M. (1994) 32P-postlabelling analysis
of DNA adducts in white blood cells of humans exposed to residential wood combustion paniculate matter.
Environ. Mol. Mutagen. 23: 121-127.
Heyder, J.; Beck-Speier, I.; Busch, B.; Dirscherl, P.; Heilmann, P.; Perron, G. A.; Josten, M.; Karg, E.; Kreyling,
W. G.; Lenz, A.-G.; Maier, K. L.; Miaskowski, U.; Platz, S.; Reitmeir, P.; Schulz, H.; Takenaka, S.;
Ziesenis, A. (1999) Health effects of sulfur-related environmental air pollution. I. Executive summary.
Inhalation Toxicol. 11: 343-359.
Hitzfeld, B.; Friedrichs, K. H.; Ring, J.; Behrendt, H. (1997) Airborne paniculate matter modulates the production of
reactive oxygen species in human polymorphonuclear granulocytes. Toxicology 120: 185-195.
7-232
-------
Hodgson, M. I; Morey, P.; Leung, W. Y.; Morrow, L.; Miller, D.; Jarvis, B. B.; Robbins, H.; Halsey, J. F.;
Storey, E. (1998) Building-associated pulmonary disease from exposure to Stachybotrys chartamm and
Aspergillus versicolor. J. Occup. Environ. Med. 40:241-249.
Holian, A.; Hamilton, R. F., Jr.; Morandi, M. T.; Brown, S. D.; Li, L. (1998) Urban particle-induced apoptosis and
phenotype shifts in human alveolar macrophages. Environ. Health Perspect. 106: 127-132.
Holma, B. (1989) Effects of inhaled acids on airway mucus and its consequences for health. In: Symposium on the
health effects of acid aerosols; October 1987; Research Triangle Park, NC. Environ. Health Perspect.
79: 109-113.
Hornberg, C.; Maciuleviciute, L.; Seemayer, N. H. (1996) Sister chromatid exchanges in rodent trachea! epithelium
exposed in vitro to environmental pollutants. Toxicol. Lett. 88: 45-53.
Hornberg, C.; Maciuleviciute, L.; Seemayer, N. H.; Kainka, E. (1998) Induction of sister chromatid exchanges
(SCE) in human trachea! epithelial cells by the fractions PM-10 and PM-2.5 of airborne particulates. Toxicol.
Lett. 96/97: 215-220.
Hou, S.; Lambert, B.; Hemminki, K. (1995) Relationship between hprt mutant frequency, aromatic DNA adducts
and genotypes for GSTM1 andNAT2 in bus maintenance workers. Carcinogenesis 16: 1913-1917.
Houk, V. S.; Early, G.; Claxton, L. D. (1991) Use of the spiral Salmonella assay to detect the mutagenicity of
complex environmental mixtures. Environ. Mol. Mutagen. 17: 112-121.
Huang, S. L.; Cheng, W. L.; Lee, C. T.; Huang, H. C.; Chan, C. C. (2002) Contribution of endotoxin in macrophage
cytokine response to ambient particles in vitro. J. Toxicol. Environ. Health A 65: 1261-1272.
Huisingh, J.; Bradow, R.; lungers, R.; Claxton, L.; Zweidinger, R.; Tejada, S.; Bumgarner, J.; Duffield, F.;
Waters, M.; Simmon, V. F.; Hare, C.; Rodriguez, C.; Snow, L. (1978) Application of bioassay to the
characterization of diesel particle emissions. In: Waters, M. D.; Nesnow, S.; Huisingh, J. L.; Sandhu, S. S.;
Claxton, L., eds. Application of short-term bioassays in the fractionation and analysis of complex
environmental mixtures: [proceedings of a symposium; February; Williamsburg, VA]. New York, NY:
Plenum Press; pp. 383-418. (Hollaender, A.; Probstein, F.; Welch, B. L., eds. Environmental science research:
v. 15).
Hunter, R. J. (1981) Zeta potential in colloid science: principles and applications. Orlando, FL: Academic Press,
pp. 1-124. (Ottewill, R. H.; Rowell, R. L., eds. Colloid Science).
Ichinose, T.; Yajima, Y.; Nagashima, M.; Takenoshita, S.; Nagamachi, Y.; Sagai, M. (1997) Lung carcinogenesis
and formation of 8-hydroxydeoxyguanosine in mice by diesel exhaust particles. Carcinogenesis 18: 185-192.
Imrich, A.; Ning, Y. Y.; Kobzik, L. (2000) Insoluble components of concentrated air particles mediate alveolar
macrophage responses in vitro. Toxicol. Appl. Pharmacol. 167: 140-150.
International Agency for Research on Cancer. (1989) Diesel and gasoline engine exhausts and some nitroarenes.
Lyon, France: International Agency for Research on Cancer; pp. 41-185. (IARC monographs on the
evaluation of carcinogenic risks to humans: v. 46).
Ishinishi, N.; Kuwabara, N.; Nagase, S.; Suzuki, T.; Ishiwata, S.; Kohno, T. (1986) Long-term inhalation studies on
effects of exhaust from heavy and light duty diesel engines on F344 rats. In: Ishinishi, N.; Koizumi, A.;
McClellan, R. O.; Stober, W., eds. Carcinogenic and mutagenic effects of diesel engine exhaust: proceedings
of the international satellite symposium on lexicological effects of emissions from diesel engines; July;
Tsukuba Science City, Japan. Amsterdam, Holland: Elsevier Science Publishers B. V.; pp. 329-348.
(Developments in toxicology and environmental science: v. 13).
Ishinishi, N.; Kuwabara, N.; Takaki, Y.; Nagase, S.; Suzuki, T.; Nakajima, T.; Maejima, K.; Kato, A.; Nakamura, M.
(1988) Long-term inhalation experiments on diesel exhaust. In: Diesel exhaust and health risks: results of the
HERP studies. Tsukuba, Ibaraki, Japan: Japan Automobile Research Institute, Inc., Research Committee for
HERP Studies; pp. 11-84.
Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
1983-1992. Environ. Health Perspect. 105: 980-985.
Jakab, G. J. (1992) Relationship between carbon black particulate-bound formaldehyde, pulmonary antibacterial
defenses, and alveolar macrophage phagocytosis. Inhalation Toxicol. 4: 325-342.
Jakab, G. J. (1993) The toxicologic interactions resulting from inhalation of carbon black and acrolein on pulmonary
antibacterial and antiviral defenses. Toxicol. Appl. Pharmacol. 121: 167-175.
Jakab, G. J.; Hemenway, D. R. (1993) Inhalation coexposure to carbon black and acrolein suppresses alveolar
macrophage phagocytosis and TNF-a release and modulates peritoneal macrophage phagocytosis. Inhalation
Toxicol. 5: 275-289.
Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
7-233
-------
Jiang, N.; Dreher, K. L.; Dye, J. A.; Richards, J. H.; Martin, L. D.; Adler, K. B. (2000) Residual oil fly ash induces
cytotoxicity and mucin secretion by guinea pig tracheal epithelial cells via an oxidant-mediated mechanism.
Toxicol. Appl. Pharmacol. 163: 221-230.
Kadiiska, M. B.; Mason, R. P.; Dreher, K. L.; Costa, D. L.; Ohio, A. J. (1997) In vivo evidence of free radical
formation in the rat lung after exposure to an emission source air pollution particle. Chem. Res. Toxicol.
10: 1104-1108.
Kao, A. S.; Friedlander, S. K. (1995) Temporal variations of paniculate air pollution: a marker for free radical
dosage and adverse health effects? Inhalation Toxicol. 7: 149-156.
Keane, M. J.; Xing, S.-G.; Harrison, J. C.; Ong, T.; Wallace, W. E. (1991) Genotoxicity of diesel-exhaust particles
dispersed in simulated pulmonary surfacta t. Mutat. Res. 260: 233-238.
Kennedy, T.; Ohio, A. J.; Reed, W.; Samet, J.; Zagorski, J.; Quay, J.; Carter, J.; Dailey, L.; Hoidal, J. R.; Devlin,
R. B. (1998) Copper-dependent inflammation and nuclear factor-KB activation by paniculate air pollution.
Am. J. Respir. CellMol. Biol. 19: 366-378.
Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
(1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly ash
inhalation. Inhalation Toxicol. 9: 541-565.
Kim, S.; Chang, M.-C.; Kim, D.; Sioutas, C. (2000a) A new generation of portable coarse, fine, and ultrafine particle
concentrators for use in inhalation toxicology. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings of the
third colloquium on paniculate air pollution and human health (first special issue); June, 1999; Durham, NC.
Inhalation Toxicol. 12(suppl. 1): 121-137.
Kim, S.; Sioutas, C.; Chang, M.-C.; Gong, H., Jr. (2000b) Factors affecting the stability of the performance of
ambient fine-particle concentrators. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
Toxicol. 12(suppl. 4): 281-298.
Kimmel, T. A.; Chen, L. C.; Bosland, M. C.; Nadziejko, C. (1997) Influence of acid aerosol droplet size on
structural changes in the rat lung caused by acute exposure to sulfuric acid and ozone. Toxicol. Appl.
Pharmacol. 144: 348-355.
Kim Oanh, N. T.; Nghiem, L. H.; Phyu, Y. L. (2002) Emission of poly cyclic aromatic hydrocarbons, toxicity, and
mutagenicity from domestic cooking using sawdust briquettes, wood, and kerosene. Environ. Sci. Technol.
36: 833-839.
Kitabatake, M.; Imai, M.; Kasama, K.; Kobayashi, L; Tomita, Y.; Yoshida, K. (1979) Effects of air pollutants on the
experimental induction of asthma attacks in guinea pigs: sulfuric acid mist and mixture of the mist and sulfur
dioxide. Mie Med. J. 29: 29-36.
Kleeberger, S. R.; Levitt, R. C.; Zhang, L.-Y.; Longphre, M.; Harkema, J.; Jedlicka, A.; Eleff, S. M.; DiSilvestre, D.;
Holroyd, K. J. (1997) Linkage analysis of susceptibility to ozone-induced lung inflammation in inbred mice.
Nat. Genet. 17: 475-478.
Kleiger, R. E.; Bigger, J. T.; Bosner, M. S.; Chung, M. K.; Cook, J. R.; Rolnitzky, L. M.; Steinman, R.; Fleiss, J. L.
(1991) Stability over time of variables measurtng heart rate variability in normal subjects. Am. J. Cardiol. 68:
626-630.
Kleinman, M. T.; Leaf, D. A.; Kelly, E.; Caiozzo, V.; Osann, K; O'Niell, T. (1998) Urban angina in the mountains:
effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch. Environ.
Health 53: 388-397.
Kleinman, M. T.; Mautz, W. J.; Bjarnason, S. (1999) Adaptive and non-adaptive responses in rats exposed to ozone,
alone and in mixtures, with acidic aerosols. Inhalation Toxicol. 11: 249-264.
Kleinman, M. T.; Bufalino, C.; Rasmussen, R.; Hyde, D.; Bhalla, D. K.; Mautz, W. J. (2000) Toxicity of chemical
components of ambient fine paniculate matter (PM25) inhaled by aged rats. J. Appl. Toxicol. 20: 357-364.
Knox, R. B.; Suphioglu, C.; Taylor, P.; Desai, R.; Watson, H. C.; Peng, J. L.; Bursill, L. A. (1997) Major grass
pollen allergen Lol p 1 binds to diesel exhaust particles: implications for asthma and air pollution. Clin. Exp.
Allergy 27: 246-251.
Kobzik, L. (1995) Lung macrophage uptake of unopsonized environmental particulates: role of scavenger-type
receptors. J. Immunol. 155: 367-376.
Kobzik, L.; Goldsmith, C.-A. W.; Ning, Y. Y.; Qin, G.; Morgan, B.; Imrich, A.; Lawrence J.; Murthy, G. G. K.;
Catalano, P. J. (2001) Effects of combined ozone and air pollution particle exposure in mice. Boston, MA:
Health Effects Institute; research report no. 106. Available: http://www.healtheffects.org/Pubs/Kobzik.pdf [27
January 2003].
Kodavanti, U. P.; Costa, D. L. (1999) Animal models to study for pollutant effects. In: Holgate, S. T.; Samet, J. M.;
Koren, H. S.; Maynard, R. L., eds. Air pollution and health. New York, NY: Academic Press; pp. 165-197.
7-234
-------
Kodavanti, U. P.; Jaskot, R. H.; Bonner, J.; Badgett, A.; Dreher, K. L. (1996) Eosinophilic lung inflammation in
particulate-induced lung injury: technical consideration in isolating RNA for gene expression studies.
Exp. Lung Res. 22: 541-554.
Kodavanti, U. P.; Jaskot, R. H.; Costa, D. L.; Dreher, K. L. (1997a) Pulmonary proinflammatory gene induction
following acute exposure to residual oil fly ash: roles of particle-associated metals. Inhalation Toxicol.
9: 679-701.
Kodavanti, U. P.; Jaskot, R. H.; Su, W. Y.; Costa, D. L.; Ohio, A. J.; Dreher, K. L. (1997b) Genetic variability in
combustion particle-induced chronic lung injury. Am. J. Physiol. 272: L521-L532.
Kodavanti, U. P.; Hauser, R.; Christian!, D. C.; Meng, Z. H.; McGee, J.; Ledbetter, A.; Richards, J.; Costa, D. L.
(1998a) Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals.
Toxicol. Sci. 43:204-212.
Kodavanti, U. P.; Costa, D. L.; Bromberg, P. A. (1998b) Rodent models of cardiopulmonary disease: their potential
applicability in studies of air pollutant susceptibility. Environ. Health Perspect. Suppl. 106(1): 111-130.
Kodavanti, U. P.; Jackson, M. C.; Ledbetter, A. D.; Richards, J. R.; Gardner, S. Y.; Watkinson, W. P.; Campen,
M. J.; Costa, D. L. (1999) Lung injury from intratracheal and inhalation exposures to residual oil fly ash in a
rat model of monocrotaline-induced pulmonary hypertension. J. Toxicol. Environ. Health Part A 57: 101-121.
Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Watkinson, W. P.; Campen, M. J.; Winsett, D. W.;
Richards, J. R.; Crissman, K. M.; Hatch, G. E.; Costa, D. L. (2000a) The spontaneously hypertensive rat as a
model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress
from inhaled emission particulate matter. Toxicol. Appl. Pharmacol. 164: 250-263.
Kodavanti, U. P.; Mebane, R.; Ledbetter, A.; Krantz, T.; McGee, J.; Jackson, M. C.; Walsh, L.; Hilliard, H.; Chen,
B. Y.; Richards, J.; Costa, D. L. (2000b) Variable pulmonary responses from exposure to concentrated
ambient air particles in a rat model of bronchitis. Toxicol. Sci. 54: 441-451.
Kodavanti, U. P.; Schladweiler, M. C. J.; Richards, J. R.; Costa, D. L. (2001) Acute lung injury from intratracheal
exposure to fugitive residual oil fly ash and its constituent metals in normo- and spontaneously hypertensive
rats. Inhalation Toxicol. 13: 37-54.
Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Hauser, R.; Christiani, D. C.; McGee, J.; Richards, J. R.;
Costa, D. L. (2002a) Temporal association between pulmonary and systemic effects of particulate matter in
healthy and cardiovascular compromised rats. J. Toxicol. Environ. Health Part A 65: 1545-1569.
Kodavanti, U. P.; Schladweiler, M. C. J.; Ledbetter, A. D.; Hauser, R.; Christiani, D. C.; Samet, J. M.; McGee, J.;
Richards, J. H.; Costa, D. L. (2002b) Pulmonary and systemic effects of zinc-containing emission particles in
three rat strains: multiple exposure scenarios. Toxicol. Sci. 70: 73-85.
Kodavanti, U. P.; Moyer, C. F.; Ledbetter, A. D.; Schladweiler, M. C.; Costa, D. L.; Hauser, R.; Christiani, D. C.;
Nyska, A. (2003) Inhaled environmental combustion particles cause myocardial injury in the Wistar Kyoto
rat. Toxicol. Sci. 71: 237-245.
Lambert, A. L.; Dong, W.; Winsett, D. W.; Selgrade, M. K.; Gilmour, M. I. (1999) Residual oil fly ash exposure
enhances allergic sensitization to house dust mite. Toxicol. Appl. Pharmacol. 158: 269-277'.
Lambert, A. L.; Dong, W.; Selgrade, M. K.; Gilmour, M. I. (2000) Enhanced allergic sensitization by residual oil fly
ash particles is mediated by soluble metal constituents. Toxicol. Appl. Pharmacol. 165: 84-93.
Lan, Q.; Chapman, R. S.; Schreinemachers, D. M.; Tian, L. W.; He, X. Z. (2002) Household stove improvement and
risk of lung cancer in Xuanwei, China. J. Nat. Cancer Inst. 94: 826-835.
La Rovere, M. T.; Pinna, G. D.; Maestri, R.; Mortara, A.; Capomolla, S.; Febo, O.; Ferrari, R.; Franchini, M.;
Gnemmi, M.; Opasich, C.; Riccardi, P. G.; Traversi, E.; Cobelli, F. (2003) Short-term heart rate variability
strongly predicts sudden cardiac death in chronic heart failure patients. Circulation 107: 565-570.
Larsen, F. O.; Christensen, L. H. R.; Clementsen, P.; Gravesen, S.; Stahl Skov, P.; Norn, S. (1996) Microfungi in
indoor air are able to trigger histamine release by non-IgE-mediated mechanisms. Inflammation Res.
45(suppl. 1): S23-S24.
Last, J. A.; Pinkerton, K. E. (1997) Chronic exposure of rats to ozone and sulfuric acid aerosol: biochemical and
structural responses. Toxicology 116: 133-146.
Last, J. A.; Hyde, D. M.; Chang, D. P. Y. (1984) A mechanism of synergistic lung damage by ozone and a respirable
aerosol. Exp. Lung Res. 7: 223-235.
Laurie, R. D.; Boyes, W. K. (1980) Neurophysiological alterations due to diesel exhaust exposure during the
neonatal life of the rat. In: Pepelko, W. E.; Danner, R. M.; Clarke, N. A., eds. Health effects of diesel engine
emissions: proceedings of an international symposium, v. 2; December 1979; Cincinnati, OH. Cincinnati, OH:
U.S. Environmental Protection Agency, Health Effects Research Laboratory; pp. 713-727; EPA report no.
EPA-600/9-80-057b. Available from: NTIS, Springfield, VA; PB81-173817.
7-235
-------
Laurie, R. D.; Boyes, W. K. (1981) Neurophysiological alterations due to diesel exhaust exposure during the
neonatal life of the rat. Environ. Int. 5: 363-368.
Laurie, R. D.; Boyes, W. K.; Wessendarp, T. (1980) Behavioral alterations due to diesel exhaust exposure. In:
Pepelko, W. E.; Banner, R. M; Clarke, N. A., eds. Health effects of diesel engine emissions: proceedings of
an international symposium, v. 2; December 1979; Cincinnati, OH. Cincinnati, OH: U.S. Environmental
Protection Agency, Health Effects Research Laboratory; pp. 698-712; EPA report no. EPA-600/9-80-057b.
Available from: NTIS, Springfield, VA; PB81-173817.
Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular distribution
of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell Mol. Biol. 18: 687-695.
Lay, J. C.; Zeman, K. L.; Ohio, A. J.; Bennett, W. D. (2001) Effects of inhaled iron oxide particles on alveolar
epithelial permeability in normal subjects. Inhalation Toxicol. 13: 1065-1078.
Leduc, D.; Fally, S.; De Vuyst, P.; Wollast, R.; Yernault, J.-C. (1995) Acute exposure to realistic acid fog: effects on
respiratory function and airway responsiveness in asthmatics. Environ. Res. 71: 89-98.
Lee, M. M.; Green, F. H. Y.; Roth, S. H.; Karkhanis, A.; Bjarnason, S. G.; Schtirch, S. (1999) Sulfuric acid aerosol
induces changes in alveolar surface tension in the guinea pig but not in the rat. Exp. Lung Res. 25: 229-244.
Leikauf, G. D.; McDowell, S. A.; Gammon, K.; Wesselkamper, S. C.; Bachurski, C. J.; Puga, A.; Wiest, J. S.;
Leikauf, J. E.; Prows, D. R. (2000) Functional genomics of particle-induced lung injury. Inhalation Toxicol.
12: 59-73.
Leikauf, G.; McDowell, S. A.; Wesselkamper, S. C.; Miller, C. R.; Hardie, W. D.; Gammon, K.; Biswas, P. P.;
Korfhagen, T. R.; Bachurski, C. J.; Wiest, J. S.; Willeke, K.; Bingham, E.; Leikauf, J. E.; Aronow, B. J.;
Prows, D. R. (2001) Pathogenomic mechanisms for paniculate matter induction of acute lung injury and
inflammation in mice. Boston, MA: Health Effects Institute; research report no. 105. Available:
http://www.healtheffects.org/pubs-research.htm [15 May, 2003].
Lewis, T. R.; Green, F. H. Y.; Moorman, W. J.; Burg, J. R.; Lynch, D. W. (1989) A chronic inhalation toxicity study
of diesel engine emissions and coal dust, alone and combined. J. Am. Coll. Toxicol. 8: 345-375.
Lewtas, J. (1983) Evaluation of the mutagenicity and carcinogenicity of motor vehicle emissions in short-term
bioassays. Environ. Health Perspect. 47: 141-152.
Li, A. P.; Royer, R. E. (1982) Diesel-exhaust-particle extract enhancement of chemical-induced mutagenesis in
cultured Chinese hamster ovary cells: possible interaction of diesel exhaust with environmental chemicals.
Mutat. Res. 103:349-355.
Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1996) Free radical activity and pro-inflammatory effects of
paniculate air pollution (PM10) in vivo and in vitro. Thorax 51: 1216-1222.
Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
paniculate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international
meeting on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake
Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
Li, X. Y.; Brown, D.; Smith, S.; MacNee, W.; Donaldson, K. (1999) Short-term inflammatory responses following
intratracheal instillation of fine and ultrafine carbon black in rats. Inhalation Toxicol. 11: 709-731.
Liber, H. L.; Andon, B. M.; Kites, R. A.; Thilly, W. G. (1981) Diesel soot: mutation measurements in bacterial and
human cells. Environ. Int. 5: 281-284.
Liberti, A.; Ciccioli, P.; Cecinato, A.; Brancaleoni, E.; Di Palo, C. (1984) Determination of nitrated-polyaromatic
hydrocarbons (nitro-PAHs) in environmental samples by high resolution chromatographic techniques. J. High
Resolut. Chromatogr. Chromatogr. Commun. 7: 389-397.
Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D. (1994) Effects of prolonged,
repeated exposure to ozone, sulfuric acid, and their combination in healthy and asthmatic volunteers. Am. J.
Respir. Crit. Care Med. 150: 431-440.
Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
California. Environ. Health Perspect. 105: 216-222.
Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
Liu, X.; Keane, M. J.; Zong, B.-Z.; Ong, T.-M.; Wallace, W. E. (1996) Micronucleus formation in V79 cells treated
with respirable silica dispersed in medium and simulated pulmonary surfactant. Mutat. Res. 361: 89-94.
7-236
-------
Lofroth, G. (1981) Comparison of the mutagenic activity in carbon paniculate matter and in diesel and gasoline
engine exhaust. In: Waters, M. D.; Sandhu, S. S.; Huisingh, J. L.; Claxton, L.; Nesnow, S., eds. Short-term
bioassays in the analysis of complex environmental mixtures II: proceedings of the second symposium on the
application of short-term bioassays in the fractionation and analysis of complex environmental mixtures;
March 1980; Williamsburg, VA. New York, NY: Plenum Press; pp. 319-336. (Hollaender, A.; Welch, B. L.;
Probstein, R. F., eds. Environmental science research series: v. 22).
Lofroth, G.; Lazaridis, G.; Rudling, L. (1986) Mutagenicity assay of emission extracts from wood stoves:
comparison with other emission parameters. Sci. Total Environ. 58: 199-208.
Loft, S.; Deng, X. S.; Tuo, J.; Wellejus, A.; Serensen, M.; Poulsen, H. E. (1998) Experimental study of oxidative
DNA damage. Free Rad. Res. 29: 525-539.
Long, C. M.; Suh, H. H.; Kobzik, L.; Catalano, P. J.; Ning, Y. Y.; Koutrakis, P. (2001) A pilot investigation of the
relative toxicity of indoor and outdoor fine particles: in vitro effects of endotoxin and other paniculate
properties. Environ. Health Perspect. 109: 1019-1026.
Longphre, M.; Li, D.; Matovinovic, E.; Gallup, M.; Samet, J. M.; Basbaum, C. B. (2000) Lung mucin production is
stimulated by the air pollutant residual oil fly ash. Toxicol. Appl. Pharmacol. 162: 86-92.
Lowe, G. D. O.; Lee, A. J.; Rumley, A.; Price, J. F.; Fowkes, F. G. R. (1997) Blood viscosity and risk of
cardiovascular events: The Edinburgh Study. Br. J. Haematol. 96: 168-173.
Madden, M. C.; Thomas, M. J.; Ohio, A. J. (1999) Acetaldehyde (CH3CHO) production in rodent lung after
exposure to metal-rich particles. Free Radical Biol. Med. 26: 1569-1577.
Madden, M. C.; Richards, J. H.; Dailey, L. A.; Hatch, G. E.; Ohio, A. J. (2000) Effect of ozone on diesel exhaust
particle toxicity in rat lung. Toxicol. Appl. Pharmacol. 168: 140-148.
Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
macrophage function, and microfilament arrangement in monocrotaline-treated rats. Toxicol. Appl.
Pharmacol. 153: 28-38.
Matsushita, H.; Goto, S.; Endo, O.; Lee, J.-H.; Kawai, A. (1986) Mutagenicity of diesel exhaust and related
chemicals. In: Ishinishi, N.; Koizumi, A.; McClellan, R. O.; Stober, W., eds. Carcinogenic and mutagenic
effects of diesel engine exhaust: proceedings of the international satellite symposium on lexicological effects
of emissions from diesel engines; July; Tsukuba Science City, Japan. Amsterdam, The Netherlands: Elsevier
Science Publishers B. V.; pp. 103-118. (Developments on toxicology and environmental science: v. 13).
Mauderly, J. L. (1993) Toxicological approaches to complex mixtures. Environ. Health Perspect. Suppl.
101(4): 155-165.
Mauderly, J. L. (2000) Animal models for the effect of age on susceptibility to inhaled paniculate matter. Inhalation
Toxicol. 12: 863-900.
Mauderly, J. L.; Gillett, N. A.; Henderson, R. F.; Jones, R. K.; McClellan, R. O. (1988) Relationships of lung
structural and functional changes to accumulation of diesel exhaust particles. In: Dodgson, J.; McCallum, R.
I.; Bailey, M. R.; Fisher, D. R., eds. Inhaled particles VI: proceedings of an international symposium and
workshop on lung dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1):
659-669.
Mautz, W. J.; Kleinman, M. T.; Bhalla, D. K.; Phalen, R. F. (2001) Respiratory tract responses to repeated inhalation
of an oxidant and acid gas-particle air pollutant mixture. Toxicol. Sci. 61: 331-341.
McClellan, R. O.; Bice, D. E.; Cuddihy, R. G.; Gillett, N. A.; Henderson, R. F.; Jones, R. K.; Mauderly, J. L.;
Pickrell, J. A.; Shami, S. G.; Wolff, R. K. (1986) Health effects of diesel exhaust. In: Lee, S. D.; Schneider,
T.; Grant, L. D.; Verkerk, P. J., eds. Aerosols: research, risk assessment and control strategies: proceedings of
the second U.S.-Dutch international symposium; May 1985; Williamsburg, VA. Chelsea, MI: Lewis
Publishers, Inc.; pp. 597-615.
McDowell, S. A.; Gammon, K.; Bachurski, C. J.; Weist, J. S.; Leikauf, J. E.; Prows, D. R. Leikauf, G. D. (2000)
Differential gene expression in the initiation and progression of nickel-induced acute lung injury. Am. J.
Respir. Cell Mol. Biol. 23: 466-474.
McKenna, I. M.; Gordon, T.; Chen, L. C.; Anver, M. R.; Waalkes, M. P. (1998) Expression of metallothionein
protein in the lungs of Wistar rats and C57 and DBA mice exposed to cadmium oxide fumes. Toxicol. Appl.
Pharmacol. 153: 169-178.
Meade, T. W.; Ruddock, V.; Stirling, Y.; Chakrabarti, R.; Miller, G. J. (1993) Fibrinolytic activity, clotting factors,
and long-term incidence of ischaemic heart disease in The Northwick Park Heart Study. Lancet
342: 1076-1079.
Menetrez, M. Y.; Foarde, K. K.; Ensor, D. S. (2001) An analytical method for the measurement of nonviable
bioaerosols. J. Air Waste Manage. Assoc. 51: 1436-1442.
7-237
-------
Michel, O.; Nagy, A.-M.; Schroeven, M; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
Miller, F. J. (2000) Dosimetry of particles in laboratory animals and humans in relationship to issues surrounding
lung overload and human health risk assessment: a critical review. Inhalation Toxicol. 12: 19-57.
Mitchell, A. D.; Evans, E. L.; Jotz, M. M.; Riccio, E. S.; Mortelmans, K. E.; Simmon, V. F. (1981) Mutagenic and
carcinogenic potency of extracts of diesel and related environmental emissions: in vitro mutagenesis and
DNA damage. Environ. Int. 5: 393-401.
Molinelli, A. R.; Madden, M. C.; McGee, J. K.; Stonebuerner, J. G.; Ohio, A. J. (2002) Effect of metal removal on
the toxicity of airborne paniculate matter from the Utah Valley. Inhalation Toxicol. 14: 1069-1086.
Monini, P.; de Lellis, L.; Barbanti-Brodano, G. (1995) Association of BK and JC human Polyomaviruses and SV40
with human tumors. In: Barbanti-Brodano, G.; Bendinelli, M.; Friedman, H., eds. DNA tumor
viruses-oncogenic mechanisms. New York, NY: Plenum; pp. 51-73.
Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
155: 245-252.
Monn, C.; Naef, R.; Koller, T. (2003) Reactions of macrophages exposed to particles <10 |_im Environ. Res.
91: 35-44.
Morimoto, K.; Kitamura, M.; Kondo, H.; Koizumi, A. (1986) Genotoxicity of diesel exhaust emissions in a battery
of in-vitro short-termbioassays. In: Ishinishi, N.; Koizumi, A.; McClellan, R. O.; Stober, W., eds.
Carcinogenic and mutagenic effects of diesel engine exhaust: proceedings of the international satellite
symposium on lexicological effects of emissions from diesel engines; July; Tsukuba Science City, Japan.
Amsterdam, The Netherlands: Elsevier Science Publishers B. V.; pp. 85-102. (Developments in toxicology
and environmental science: v. 13).
Morio, L. A.; Hooper, K. A.; Brittingham, J.; Li, T.-H.; Gordon, R. E.; Turpin, B. J.; Laskin, D. L. (2001) Tissue
injury following inhalation of fine paniculate matter and hydrogen perioxide is associated with altered
production of inflammatory mediators and antioxidants by alveolar macrophages. Toxicol. Appl. Pharmacol.
177: 188-199.
Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
10: 369-384.
Moss, O. R.; Gross, E. A.; James, R. A.; Janszen, D. B.; Ross, P. W.; Roberts, K. C.; Howard, A. M.; Harkema,
J. R.; Calderon-Garciduenas, L.; Morgan, K. T. (2001) Respiratory tract toxicity in rats exposed to Mexico
City air. Cambridge, MA: Health Effects Institute; research report no. 100. Available:
http://www.healtheffects.org/pubs-research.htm [15 May, 2003].
Muggenburg, B. A.; Tilley, L.; Green, F. H. Y. (2000a) Animal models of cardiac disease: potential usefulness for
studying health effects of inhaled particles. Inhalation Toxicol. 12: 901-925.
Muggenburg, B. A.; Barr, E. B.; Cheng, Y. S.; Seagrave, JC.; Tilley, L. P.; Mauderly, J. L. (2000b) Effect of inhaled
residual oil fly ash on the electrocardiogram of dogs. In: Grant, L. D., ed. PM2000: paniculate matter and
health. Inhalation Toxicol. 12(suppl. 4): 189-208.
Muggenburg, B. A.; Benson, J. M.; Barr, E. B.; Kubatko, J.; Tilley, L. P. (2003) Short-term inhalation of paniculate
transition metals has little effect on the electrocardiograms of dogs having preexisting cardiac abnormalities.
Inhalation Toxicol. 15: 357-371.
Mukae, H.; Hogg, J. C.; English, D.; Vincent, R.; VanEeden, S. F. (2000) Phagocytosis of paniculate air pollutants
by human alveolar macrophages stimulates the bone marrow. Am. J. Physiol. 279: L924-L931.
Mukae, H.; Vincent, R.; Quinlan, K.; English, D.; Hards, J.; Hogg, J. C.; Van Eeden, S. F. (2001) The effect of
repeated exposure to paniculate air pollution (PM10) on the bone marrow. Am. J. Respir. Crit. Care Med. 163:
201-209.
Mumford, J. L.; Lewtas, J. (1982) Mutagenicity and cytotoxicity of coal fly ash from fluidized bed and conventional
combustion. J. Toxicol. Environ. Health 10: 565-586.
Mumford, J. L.; He, X. Z.; Chapman, R. S.; Cao, S. R.; Harris, D. B.; Li, X. M.; Xian, Y. L.; Jiang, W. Z.; Xu,
C. W.; Chuang, J. C.; Wilson, W. E.; Cooke, M. (1987) Lung cancer and indoor air pollution in Xuan Wei,
China. Science (Washington, DC) 235: 217-220.
Mumford, J. L.; Tian, D.; Younes, M.; Hu, F.; Lan, Q.; Ostrowski, M. L.; He, X.; Feng, Z. (1999) Detection of p53
protein accumulation in sputum and lung adenocarcinoma associated with indoor exposure to unvented coal
smoke in China. Anticancer Res. 19: 951-958.
Murphy, S. A.; BeruBe, K. A.; Pooley, F. D.; Richards, R. J. (1998) The response of lung epithelium to well
characterised fine particles. Life Sci. 62: 1789-1799.
7-238
-------
Nadadur, S. S.; Kodavanti, U. P. (2002) Altered gene expression profiles of rat lung in response to an emission
paniculate and its metal constituents. J. Toxicol. Environ. Health Part A 65:1333-1350.
Nadadur, S. S.; Schladweiler, M. C. J.; Kodavanti, U. P. (2000) A pulmonary rat gene array for screening altered
expression profiles in air pollutant-induced lung injury. Inhalation Toxicol. 12: 1239-1254.
Nadeau, D.; Vincent, R.; Kumarathasan, P.; Brook, J.; Dufresne, A. (1996) Cytotoxicity of ambient air particles to
rat lung macrophages: comparison of cellular and functional assays. Toxicol. in Vitro 10: 161-172.
Nadziejko, C.; Fang, K.; Chen, L. C.; Cohen, B.; Karpatkin, M.; Nadas, A. (2002) Effect of concentrated ambient
paniculate matter on blood coagulation parameters in rats. Boston, MA: Health Effects Institute. Available:
http://www.healtheffects.org/Pubs/Nadziejko.pdf.
Nagashima, M.; Kasai, H.; Yokota, J.; Nagamachi, Y.; Ichinose, T.; Sagai, M. (1995) Formation of an oxidative
DNA damage, 8-hydroxydeoxyguanosine, in mouse lung DNA after intratracheal instillation of diesel exhaust
particles and effects of high dietary fat and beta-carotene on this process. Carcinogenesis 16: 1441-1445.
Nakagawa, R.; Kitamori, S.; Horikawa, K.; Nakashima, K.; Tokiwa, H. (1983) Identification of dinitropyrenes in
diesel-exhaust particles: their probable presence as the major mutagens. Mutat. Res. 124: 201-211.
Nakamura, T.; Hayashida, Y. (1992) Autonomic cardiovascular responses to smoke exposure in conscious rats.
Am. J. Physiol. 262: R73 8-745.
National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
publication no. 97-4051.
Nehls, P.; Seiler, F.; Rehn, B.; Greferath, R.; Bruch, J. (1997) Formation and persistence of 8-oxoguanine in rat lung
cells as an important determinant for tumor formation following particle exposure. In: Driscoll, K. E.;
Oberdorster, G., eds. Proceedings of the sixth international meeting on the toxicology of natural and
man-made fibrous and non-fibrous particles; September 1996; Lake Placid, NY. Environ. Health Perspect.
Suppl. 105(5): 1291-1296.
Nel, A. E.; Diaz-Sanchez, D.; Li, N. (2001) The role of paniculate pollutants in pulmonary inflammation and
asthma: evidence for the involvement of organic chemicals and oxidative stress. Curr. Opin. Pulm. Med.
7: 20-26.
Nemmar, A.; Delaunois, A.; Nemery, B.; Dessy-Doize, C.; Beckers, J.-F.; Sulon, J.; Gustin, P. (1999) Inflammatory
effect of intratracheal instillation of ultrafine particles in the rabbit: role of C-fiber and mast cells.
Toxicol. Appl. Pharmacol. 160: 250-261.
Nemmar, A.; Hoylaerts, M. F.; Hoet, P. H. M.; Dinsdale, D.; Smith, T.; Xu, H.; Vermylen, J.; Nemery, B. (2002)
Ultrafine particles affect experimental thrombosis in an in vivo hamster model. Am. J. Respir. Crit. Care Med.
166: 998-1004.
Nielsen, P. S.; Autrup, H. (1994) Diesel exhaust-related DNA adducts in garage workers. Clin. Chem.
40: 1456-1458.
Nielsen, P. S.; Andreassen, A.; Farmer, P. B.; 0vrebe, S.; Autrup, H. (1996) Biomonitoring of diesel
exhaust-exposed workers. DNA and hemoglobin adducts and urinary 1-hydroxypyrene as markers of
exposure. Toxicol. Lett. 86: 27-37.
Nightingale, J. A.; Maggs, R.; Cullinan, P.; Donnelly, L. E.; Rogers, D. F.; Kinnersley, R.; Chung, K. F.; Barnes,
P. J.; Ashmore, M.; Newman-Taylor, A. (2000) Airway inflammation after controlled exposure to diesel
exhaust particulates. Am. J. Respir. Crit. Care Med. 162: 161-166.
Nikula, K. J.; Snipes, M. B.; Barr, E. B.; Griffith, W. C.; Henderson, R. F.; Mauderly, J. L. (1995) Comparative
pulmonary toxicities and carcinogenicities of chronically inhaled diesel exhaust and carbon black in F344
rats. Fundam. Appl. Toxicol. 25: 80-94.
Ning, Y.; Imrich, A.; Goldsmith, C. A.; Qin, G.; Kobzik, L. (2000) Alveolar macrophage cytokine production in
response to air particles in vitro: role of endotoxin. J. Toxicol. Environ. Health A 59: 165-180.
Nishioka, M. G.; Petersen, B. A.; Lewtas, J. (1982) Comparison of nitro-aromatic content and direct-acting
mutagenicity of diesel emissions. In: Cooke, M.; Dennis, A. J.; Fisher, G. L., eds. Polynuclear aromatic
hydrocarbons: physical and biological chemistry, proceedings of the sixth international symposium on
polynuclear aromatic hydrocarbons; October 1981; Columbus, OH. Columbus, OH: Battelle Press;
pp. 603-613.
Nordenhall, C.; Pourazar, J.; Ledin, M. C.; Levin, J. O.; Sandstrom, T.; Adelroth, E. (2001) Diesel exhaust enhances
airway responsiveness in asthmatic subjects. Eur. Respir. J. 17: 909-915.
Norwood, J., Jr.; Ledbetter, A. D.; Doerfler, D. L.; Hatch, G. E. (2001) Residual oil fly ash inhalation in guinea pigs:
influence of absorbate and glutathione depletion. Toxicol. Sci. 61: 144-153.
7-239
-------
Oberdorster, G. (1996a) Effects of ultrafine particles in the lung and potential relevance to environmental particles.
In: Marijnissen, J. C. M; Gradon, L., eds. Proceedings of Aerosol inhalation, lung transport, deposition and
the relation to the environment: recent research frontiers, an international workshop; September, 1995;
Warsaw, Poland. London, United Kingdom: Kluwer Academic; pp. 165-174.
Oberdorster, G. (1996b) Significance of particle parameters in the evaluation of exposure-dose-response
relationships of inhaled particles. Inhalation Toxicol. 8(suppl.): 73-89.
Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer particles
for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
Oberdorster, G.; Finkelstein, J. N.; Johnston, C.; Gelein, R.; Cox, C.; Baggs, R.; Elder, A. C. P. (2000) Acute
pulmonary effects of ultrafine particles in rats and mice. Cambridge, MA: Health Effects Institute; research
report no. 96.
Oettinger, R.; Drumm, K.; Knorst, M.; Krinyak, P.; Smolarski, R.; Kienast, K. (1999) Production of reactive oxygen
intermediates by human macrophages exposed to soot particles and asbestos fibers and increase in NF-kappa
B p50/p!05 mRNA. Lung 177: 343-354.
Ohtoshi, T.; Takizawa, H.; Okazaki, H.; Kawasaki, S.; Takeuchi, N.; Ohta, K.; Ito, K. (1998) Diesel exhaust
particles stimulate human airway epithelial cells to produce cytokines relevant to airway inflammation
in vitro. J. Allergy Clin. Immunol. 101: 778-785.
Ohtsuka, Y.; Clarke, R. W.; Mitzner, W.; Brunson, K.; Jakab, G. J.; Kleeberger, S. R. (2000a) Interstrain variation in
murine susceptibility to inhaled acid-coated particles. Am. J. Physiol. L469-L476.
Ohtsuka, Y.; Brunson, K. J.; Jedlicka, A. E.; Mitzner, W.; Clarke, R. W.; Zhang, L.-Y.; Eleff, S. M.; Kleeberger,
S. R. (2000b) Genetic linkage analysis of susceptibility to particle exposure in mice. Am. J. Respir. Cell Mol.
Biol. 22: 574-581.
Omland, T.; Lie, R. T.; Aakvaag, A.; Aarsland, T.; Dickstein, K. (1994) Plasma endothelin determination as a
prognostic indicator of 1-year mortality after acute myocardial infarction. Circulation 89: 1573-1579.
Ong, E. K. (1994) Grass pollen allergens: molecular characterization and environmental monitoring [dissertation].
Melbourne, Australia: The University of Melbourne.
Ong, T.; Whong, W. Z.; Xu, J.; Burchell, B.; Green, F. H.; Lewis, T. (1985) Genotoxicity studies of rodents exposed
to coal dust and diesel emission particulates. Environ. Res. 37: 399-409.
Oortgiesen, M.; Veronesi, B.; Eichenbaum, G.; Kiser, P. F.; Simon, S. A. (2000) Residual oil fly ash and charged
polymers activate epithelial cells and nociceptive sensory neurons. Am. J. Physiol. 278: L683-L695.
Ormstad, H.; Johansen, B. V.; Gaarder, P. I. (1998) Airborne house dust particles and diesel exhaust particles as
allergen carriers. Clin. Exp. Allergy 28: 702-708.
Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
Fundam. Appl. Toxicol. 40: 220-227.
Osier, M.; Baggs, R. B.; Oberdorster, G. (1997) Intratracheal instillation versus intratracheal inhalation: influence of
cytokines on inflammatory response. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles; September
1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1265-1271.
Osornio-Vargas, A. R.; Bonner, J. C.; Alfaro-Moreno, E.; Martinez, L.; Garcia-Cuellar, C.; Resales, S. P.;
Miranda, J.; Rosas, I. (2003) Proinflammatory and cytotoxic effects of Mexico City air pollution paniculate
matter in vitro are dependent on particle size and composition. Environ. Health Perspect. Ill: 1289-1293.
Pagan, L; Costa, D. L.; McGee, J. K.; Richards, J. H.; Dye, J. A. (2003) Metals mimic airway epithelial injury
induced by in vitro exposure to Utah Valley ambient paniculate matter extracts. J. Toxicol. Environ. Health A
66: 1087-1112.
Palchak, R. B.; Cohen, R.; Ainslie, M.; Hoerner, C. L. (1988) Airborne endotoxin associated with industrial-scale
production of protein products in gram-negative bacteria. Am. Ind. Hyg. Assoc. J. 49: 420-421.
Pepelko, W. E.; Mattox, J. K.; Yang, Y. Y.; Moore, W., Jr. (1980a) Pulmonary function and pathology in cats
exposed 28 days to diesel exhaust. J. Environ. Pathol. Toxicol. 4: 449-458.
Pepelko, W. E.; Mattox, J.; Moorman, W. J.; Clark, J. C. (1980b) Pulmonary function evaluation of cats after one
year of exposure to diesel exhaust. In: Pepelko, W. E.; Danner, R. M.; Clarke, N. A., eds. Health effects of
diesel engine emissions: proceedings of an international symposium, v. 2; December 1979; Cincinnati, OH.
Cincinnati, OH: U.S. Environmental Protection Agency, Health Effects Research Laboratory; pp. 757-765;
EPA report no. EPA-600/9-80-057b. Available from: NTIS, Springfield, VA; PB81-173817.
Pepelko, W. E.; Mattox, J.; Moorman, W. J.; Clark, J. C. (1981) Pulmonary function evaluation of cats after one year
of exposure to diesel exhaust. Environ. Int. 5: 373-376.
7-240
-------
Pereira, M. A.; Connor, T. H.; Meyne, I; Legator, M. S. (1981a) Metaphase analysis, micronuclei assay, and urinary
mutagenicity assay of mice exposed to diesel emissions. Environ. Int. 5: 435-438.
Pereira, M. A.; Sabharwal, P. S.; Gordon, L.; Wyrobek, A. J. (1981b) The effect of diesel exhaust on sperm-shape
abnormalities in mice. Environ. Int. 5: 459-460.
Pereira, M. A.; Sabharwal, P. S.; Kaur, P.; Ross, C. B.; Choi, A.; Dixon, T. (1981c) In vivo detection of mutagenic
effects of diesel exhaust by short-term mammalian bioassays. Environ. Int. 5: 439-443.
Pereira, M. A.; McMillan, L.; Kaur, P.; Gulati, D. K.; Sabharwal, P. S. (1982) Effect of diesel exhaust emissions,
particulates, and extract on sister chromatid exchange in transplacentally exposed fetal hamster liver. Environ.
Mutagen. 4: 215-220.
Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
Petrovic, S.; Urch, B.; Brook, J.; Datema, J.; Purdham, J.; Liu, L.; Lukic, Z.; Zimmerman, B.; Toiler, G.;
Downar, E.; Corey, P.; Tarlo, S.; Broder, L; Dales, R.; Silverman, F. (2000) Cardiorespiratory effects of
concentrated ambient PM2.5: a pilot study using controlled human exposures. Inhalation Toxicol.
12(suppl. 1): 173-188.
Pieckova, E.; Kunova, Z. (2002) Indoor fungi and their ciliostatic metabolites. Ann. Agric. Environ. Med. 9: 59-63.
Piedimonte, G.; Hoffman, J. I. E.; Husseini, W. K.; Hiser, W. L.; Nadel, J. A. (1992) Effect of neuropeptides
released from sensory nerves on blood flow in the rat airway microcirculation. J. Appl. Physiol.
72: 1563-1570.
Pierce, L. M.; Alessandrini, F.; Godleski, J. J.; Paulauskis, J. D. (1996) Vanadium-induced chemokine mRNA
expression and pulmonary inflammation. Toxicol. Appl. Pharmacol. 138: 1-11.
Pinto, M.; Birnbaum, S. C.; Kadar, T.; Goldberg, G. M. (1979) Lung injury in mice induced by factors acting
synergistically with inhaled paniculate antigen. Clin. Immunol. Immunopathol. 13: 361-368.
Pohjola, S. K.; Lappi, M.; Honkanen, M.; Savela, K. (2003) Comparison of mutagenicity and calf thymus DNA
adducts formed by the paniculate and semivolatile fractions of vehicle exhausts. Environ. Mol. Mutagen. 42:
26-36.
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and
Cache Valleys. Arch. Environ. Health 46: 90-97.
Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
Assoc. 287: 1132-1141.
Prahalad, A. K.; Soukup, J. M.; Inmon, J.; Willis, R.; Ohio, A. J.; Becker, S.; Gallagher, J. E. (1999) Ambient air
particles: effects on cellular oxidant radical generation in relation to paniculate elemental chemistry. Toxicol.
Appl. Pharmacol. 158: 81-91.
Pritchard, R. J.; Ohio, A. J.; Lehmann, J. R.; Winsett, D. W.; Tepper, J. S.; Park, P.; Gilmour, M. I.; Dreher, K. L.;
Costa, D. L. (1996) Oxidant generation and lung injury after paniculate air pollutant exposure increase with
the concentrations of associated metals. Inhalation Toxicol. 8: 457-477.
Prows, D. R.; Leikauf, G. D. (2001) Quantitative trait analysis of nickel-induced acute lung injury in mice. Am. J.
Respir. Cell Mol. Biol. 24: 740-746.
Prows, D. R.; Shertzer, H. G.; Daly, M. J.; Sidman, C. L.; Leikauf, G. D. (1997) Genetic analysis of ozone-induced
acute lung injury in sensitive and resistant strains of mice. Nat. Genet. 17: 471-474.
Putnam, K. P.; Bombick, D. W.; Avalos, J. T.; Doolittle, D. J. (1999) Comparison of the cytotoxic and mutagenic
potential of liquid smoke food flavourings, cigarette smoke condensate and wood smoke condensate.
FoodChem. Toxicol. 37: 1113-1118.
Quay, J. L.; Reed, W.; Samet, J.; Devlin, R. B. (1998) Air pollution particles induce IL-6 gene expression in human
airway epithelial cells via NF-KB activation. Am. J. Respir. Cell Mol. Biol. 19: 98-106.
Rannug, U.; Sundvall, A.; Westerholm, R.; Alsberg, T.; Stenberg, U. (1983) Some aspects of mutagenicity testing of
the paniculate phase and the gas phase of diluted and undiluted automobile exhaust. In: Waters, M. D.;
Sandhu, S. S.; Lewtas, J.; Claxton, L.; Chernoff, N.; Newnow, S., eds. Short-term bioassays in the analysis of
complex environmental mixtures III. New York, NY: Plenum Press; pp. 3-16. (Environmental science
research: v. 27).
Renwick, L. C.; Donaldson, K.; Clouter, A. (2001) Impairment of alveolar macrophage phagocytosis by ultrafine
particles. Toxicol. Appl. Pharmacol. 172: 119-127.
7-241
-------
Rosas, I.; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, I; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis
of the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
emergency admissions to a hospital in Mexico City. Allergy 53: 394-401.
Rose, C. S.; Martyny, J. W.; Newman, L. S.; Milton, D. K.; King, T. E., Jr.; Beebe, J. L.; McCammon, J. B.;
Hoffman, R. E.; Kreiss, K. (1998) "Lifeguard lung": endemic granulomatous pneumonitis in an indoor
swimming pool. Am. J. Public Health 88: 1795-1800.
Rothenberg, S. J.; Nagy, P. A.; Pickrell, J. A.; Hobbs, C. H. (1989) Surface area, adsorption, and desorption studies
on indoor dust samples. Am. Ind. Hyg. Assoc. J. 50: 15-23.
Rudell, B.; Sandstrom, T.; Stjernberg, N.; Kolmodin-Hedman, B. (1990) Controlled diesel exhaust exposure in an
exposure chamber: pulmonary effects investigated withbronchoalveolar lavage. J. Aerosol Sci. 21(suppl. 1):
S411-S414.
Rudell, B.; Sandstrom, T.; Hammarstrom, U.; Ledin, M.-L.; Horstedt, P.; Stjernberg, N. (1994) Evaluation of an
exposure setup for studying effects of diesel exhaust in humans. Int. Arch. Occup. Environ. Health 66: 77-83.
Rudell, B.; Ledin, M.-C.; Hammarstrom, U.; Stjernberg, N.; Lundback, B.; Sandstrom, T. (1996) Effects on
symptoms and lung function in humans experimentally exposed to diesel exhaust. Occup. Environ. Med. 53:
658-662.
Rudell, B.; Blomberg, A.; Helleday, R.; Ledin, M.-C.; Lundback, B.; Stjernberg, N.; Horstedt, P.; Sandstrom, T.
(1999) Bronchoalveolar inflammation after exposure to diesel exhaust: comparison between unfiltered and
particle trap filtered exhaust. Occup. Environ. Med. 56: 527-534.
Russell, L. B.; Generoso, W. M.; Oakberg, E. F.; Russell, W. L.; Bangham, J. W.; Stelzner, K. F. (1980) Tests for
heritable effects induced by diesel exhaust in the mouse. In: Biology division progress report: for period of
October 1, 1978 - May 31, 1980. Oak Ridge, TN: Oak Ridge National Laboratory; report no. ORNL-5685.
Rylander, R. (1996) Airway responsiveness and chest symptoms after inhalation of endotoxin or (1 -> 3)-p-D-glucan.
Indoor Built Environ. 5:106-111.
Sagai, M.; Furuyama, A.; Ichinose, T. (1996) Biological effects of diesel exhaust particles (DEP). III. Pathogenesis
of asthma like symptoms in mice. Free Radical Biol. Med. 21: 199-209.
Saldiva, P. H. N.; Clarke, R. W; Coull, B. A.; Stearns, R. C.; Lawrence, J.; Krishna-Murthy, G. G.; Diaz, E.;
Koutrakis, P.; Sun, H.; Tsuda, A.; Godleski, J. J. (2002) Lung inflammation induced by concentrated ambient
air particles is related to particle composition. Am. J. Respir. Crit. Care Med. 165: 1610-1617.
Salmeen, L; Durisin, A. M.; Prater, T. J.; Riley, T.; Schuetzle, D. (1982) Contribution of 1-nitropyrene to
direct-acting Ames assay mutagenicities of diesel paniculate extracts. Mutat. Res. 104: 17-23.
Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (1999) Acute inflammatory
responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.
Salvi, S. S.; Nordenhall, C.; Blomberg, A.; Rudell, B.; Pourazar, J.; Kelly, F. J.; Wilson, S.; Sandstrom, T.; Holgate,
S. T.; Frew, A. J. (2000) Acute exposure to diesel exhaust increases IL-8 and GRO-aproduction in healthy
human airways. Am. J. Respir. Crit. Care Med. 161: 550-557.
Samet, J. M.; Reed, W.; Ohio, A. J.; Devlin, R. B.; Carter, J. D.; Dailey, L. A.; Bromberg, P. A.; Madden, M. C.
(1996) Induction of prostaglandin H synthase 2 in human airway epithelial cells exposed to residual oil fly
ash. Toxicol. Appl. Pharmacol. 141: 159-168.
Samet, J. M.; Stonehuerner, J.; Reed, W.; Devlin, R. B.; Dailey, L. A.; Kennedy, T. P.; Bromberg, P. A.; Ohio, A. J.
(1997) Disruption of protein tyrosine phosphate homeostasis in bronchial epithelial cells exposed to oil fly
ash. Am. J. Physiol. 272: L426-L432.
Samet, J. M.; Graves, L. M.; Quay, J.; Dailey, L. A.; Devlin, R. B.; Ohio, A. J.; Wu, W.; Bromberg, P. A.; Reed, W.
(1998) Activation of MAPKs in human bronchial epithelial cells exposed to metals. Am. J. Physiol.
275:L551-L558.
Samet, J. M.; Ohio, A. J.; Costa, D. L.; Madden, M. C. (2000) Increased expression of cyclooxygenase 2 mediates
oil fly ash-induced lung injury. Exp. Lung Res. 26: 57-69.
Sarangapani, R.; Wexler, A. S. (1996) Growth and neutralization of sulfate aerosols in human airways. J. Appl.
Physiol. 81:480-490.
Sato, H.; Sone, H.; Sagai, M.; Suzuki, K. T.; Aoki, Y. (2000) Increase in mutation frequency in lung of Big Blue® rat
by exposure to diesel exhaust. Carcinogenesis 21: 653-661.
Schappi, G. F.; Taylor, P. E.; Pain, M. C. F.; Cameron, P. A.; Dent, A. W.; Staff, I. A.; Suphioglu, C. (1999)
Concentrations of major grass group 5 allergens in pollen grains and atmospheric particles: implications for
hay fever and allergic asthma sufferers sensitized to grass pollen allergens. Clin. Exp. Allergy 29: 633-641.
Schlesinger, R. B.; Cassee, F. (2003) Atmospheric secondary in organic paniculate matter: the lexicological
perspective as a basis for health effects risk assessment. Inhalation Toxicol. 15: 197-235.
7-242
-------
Schliiter, T.; Berg, I.; Dorger, M; Gercken, G. (1995) Effect of heavy metal ions on the release of reactive oxygen
intermediates by bovine alveolar macrophages. Toxicology 98: 47-55.
Schuetzle, D.; Frazier, J. A. (1986) Factors influencing the emission of vapor and paniculate phase components from
diesel engines. In: Ishinishi, N.; Koizumi, A.; McClellan, R. O.; Stober, W., eds. Carcinogenic and mutagenic
effects of diesel engine exhaust: proceedings of the international satellite symposium on lexicological effects
of emissions from diesel engines; July; Tsukuba Science City, Japan. Amsterdam, The Netherlands: Elsevier
Science Publishers B. V.; pp. 41-63. (Developments in toxicology and environmental science: v. 13).
Schuetzle, D.; Lewtas, J. (1986) Bioassay-directed chemical analysis in environmental research. Anal. Chem.
58: 1060A-1076A.
Schuetzle, D.; Perez, J. M. (1983) Factors influencing the emissions of nitrated-polynuclear aromatic hydrocarbons
(nitro-PAH) from diesel engines. J. AirPollut. Control Assoc. 33: 751-755.
Schwartz, S. E. (1984) Gas- and aqueous-phase chemistry of HO2 in liquid water clouds. J. Geophys. Res. [Atmos.]
89: 11,589-11,598.
Seagrave, J. C.; Nikula, K. J. (2000) Multiple modes of responses to air pollution paniculate materials in
A549 alveolar type II cells. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12(suppl. 4): 247-260.
Seagrave, J.; McDonald, J. D.; Gigliotti, A. P.; Nikula, K. J.; Seilkop, S. K.; Gurevic, M.; Mauderly, J. L. (2002)
Mutagenicity and in vivo toxicity of combined paniculate and semivolatile organic fractions of gasoline and
diesel engine emissions. Toxicol. Sci. 70: 212-226.
Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Paniculate air pollution and acute health effects. Lancet
(8943): 176-178.
Seemayer, N. H.; Happel, A.; Behrendt, H.; Hadnagy, W.; Tomingas, R. (1990) Comparison of cytotoxicity of
airborne particulates to rat and human macrophages. J. Aerosol Sci. 21: 387-391.
Seemayer, N. H.; Hornberg, C. (1998) Malignant transformation of Syrian hamster kidney cells in vitro by
interaction of airborne particulates and simian virus (SV-) 40. Toxicol. Lett. 96/97: 231-238.
Shirname-More, L. (1995) Genotoxicity of diesel emissions. Part 1: mutagenicity and other genetic effects. Diesel
exhaust: a critical analysis of emissions, exposure, and health effects. A special report of the Institute's
Working Group. Cambridge, MA: Health Effects Institute, pp. 222-242.
Shukla, A.; Timblin, C.; BeruBe, K.; Gordon, T.; McKinney, W.; Driscoll, K.; Vacek, P.; Mossman, B. T. (2000)
Inhaled paniculate matter causes expression of nuclear factor (NF)-KB-related genes and oxidant-dependent
NF-KB activation in vitro. Am. J. Respir. Cell Mol. Biol. 23: 182-187.
Siak, J. S.; Chan, T. L.; Lees, P. S. (1981) Diesel paniculate extracts in bacterial test systems. Environ. Int.
5: 243-248.
Silbajoris, R.; Ohio, A. J.; Samet, J. M.; Jaskot, R.; Dreher, K. L.; Brighton, L. E. (2000) In vivo and in vitro
correlation of pulmonary map kinase activation following metallic exposure. Inhalation Toxicol. 12: 453-468.
Sindhu, R. K.; Mautz, W. J.; Kikkawa, Y. (1998) Chronic exposure to ozone and nitric acid vapor results in
increased levels of rat pulmonary putrescine. Arch. Toxicol. 72: 445-449.
Sioutas, C.; Koutrakis, P.; Burton, R. M. (1995a) A technique to expose animals to concentrated fine ambient
aerosols. Environ. Health Perspect. 103: 172-177.
Sioutas, D.; Koutrakis, P.; Ferguson, S. T.; Burton, R. M. (1995b) Development and evaluation of a prototype
ambient particle concentrator for inhalation exposure studies. Inhalation Toxicol. 7: 633-644.
Sioutas, C.; Kim, S.; Chang, M.; Terrell, L. L.; Gong, H., Jr. (2000) Field evaluation of a modified DataRAM MIE
scattering monitor for real-time PM2 5 mass concentration measurements. Atmos. Environ. 34: 4829-4838.
Sjogren, B. (1997) Occupational exposure to dust: inflammation and ischaemic heart disease. Occup. Environ. Med.
54: 466-469.
Sjogren, M.; Li, H.; Banner, C.; Rafter, J.; Westerholm, R.; Rannug, U. (1996) Influence of physical and chemical
characteristics of diesel fuels and exhaust emissions on biological effects of particle extracts: a multivariate
statistical analysis often diesel fuels. Chem. Res. Toxicol. 9: 197-207.
Smith, K. R.; Aust, A. E. (1997) Mobilization of iron from urban particulates leads to generation of reactive oxygen
species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chem. Res. Toxicol.
10: 828-834.
Smith, B. A.; Fullerton, N. F.; Aidoo, A.; Heflich, R. H.; Beland, F. A. (1993) DNA adduct formation in relation to
lymphocyte mutations and lung tumor induction in F344 rats treated with the environmental pollutant
1,6-dinitropyrene. Environ. Health Perspect. 99: 277-280.
Smith, K. R.; Veranth, J. M.; Lighty, J. S.; Aust, A. E. (1998) Mobilization of iron from coal fly ash was dependent
upon the particle size and the source of coal. Chem. Res. Toxicol. 11: 1494-1500.
7-243
-------
Soukup, J. M; Becker, S. (2001) Human alveolar macrophage responses to air pollution particulates are associated
with insoluble components of coarse material, including paniculate endotoxin. Toxicol. Appl. Pharmacol.
171: 20-26.
Soukup, J. M.; Ohio, A. J.; Becker, S. (2000) Soluble components of Utah Valley paniculate pollution alter alveolar
macrophage function in vivo and in vitro. Inhalation Toxicol. 12: 401-414.
Spengler, J. D.; Thurston, G. D. (1983) Mass and elemental composition of fine and coarse particles in six U.S.
cities. J. AirPollut. Control Assoc. 33: 1162-1171.
Spiewak, R.; Krysiriska-Traczyk, E.; Sitkowska, J.; Dutkiewicz, J. (1996a) Microflora of allergenic pollens-a
preliminary study. Ann. Agric. Environ. Med. 3: 127-130.
Spiewak, R.; Skorska, C.; Prazmo, Z.; Dutkiewicz, J. (1996b) Bacterial endotoxin associated with pollen as a
potential factor aggravating pollinosis. Ann. Agric. Environ. Med. 3: 57-59.
Steerenberg, P. A.; Zonnenberg, J. A. J.; Dormans, J. A. M. A.; Joon, P. N. T.; Wouters, I. M.; VanBree, L.;
Scheepers, P. T. J.; Van Loveren, H. (1998) Diesel exhaust particles induced release of interleukin 6 and 8 by
(primed) human bronchial epithelial cells (BEAS 2B) in vitro. Exp. Lung Res. 24: 85-100.
Steerenberg, P. A.; Dormans, J. A. M. A.; Van Doom, C. C. M.; Middendorp, S.; Vos, J. G.; Van Loveren, H.
(1999) A pollen model in the rat for testing adjuvant activity of air pollution components. Inhalation Toxicol.
11: 1109-1122.
Steerenberg, P. A.; Withagen, C. E.; Dormans, J. A. M. A.; VanDalen, W. J.; Van Loveren, H.; Casee, F. R. (2003)
Adjuvant activity of various diesel exhaust and ambient particle in two allergic models. J. Toxicol. Environ.
Health A 66: 1421-1439.
Strandell, M.; Zakrisson, S.; Alsberg, T.; Westerholm, R.; Winquist, L.; Rannug, U. (1994) Chemical analysis and
biological testing of a polar fraction of ambient air, diesel engine, and gasoline engine paniculate extracts. In:
Symposium of risk assessment of urban air: emissions, exposure, risk identification, and risk quantitation;
May-June 1992; Stockholm, Sweden. Environ. Health Perspect. 102(suppl. 4): 85-92.
Stringer, B.; Kobzik, L. (1996) Alveolar macrophage uptake of the environmental paniculate titanium dioxide:
role of surfactant components. Am. J. Respir. Cell Mol. Biol. 14: 155-160.
Stringer, B.; Kobzik, L. (1998) Environmental particulate-mediated cytokine production in lung epithelial cells
(A549): role of preexisting inflammation and oxidant stress. J. Toxicol. Environ. Health Part A 55: 31-44.
Stringer, B.; Imrich, A.; Kobzik, L. (1996) Lung epithelial cell (A549) interaction with unopsonized environmental
particulates: quantitation of particle-specific binding and IL-8 production. Exp. Lung Res. 22: 495-508.
Su, W.-Y.; Jaskot, R. H.; Richards, J.; Abramson, S. R.; Woessner, J. F., Jr.; Yu, W.-H.; Dreher, K. L. (2000a)
Induction of pulmonary matrilysin expression by combustion and ambient air particles. Am. J. Physiol.
279: L152-L160.
Su, W.-Y.; Jaskot, R. H.; Dreher, K. L. (2000b) Paniculate matter induction of pulmonary gelatinase A, gelatinase
B, tissue inhibitor of metalloproteinase expression. In: Phalen, R. F., ed. Inhalation Toxicology: proceedings
of the third colloquium on paniculate air pollution and human health (second special issue); June, 1999;
Durham, NC. Inhalation Toxicol. 12(suppl. 2): 105-119.
Sun, G.; Crissman, K.; Norwood, J.; Richards, J.; Slade, R.; Hatch, G. E. (2001) Oxidative interactions of synthetic
lung epithelial lining fluid with metal-containing particulate matter. Am. J. Physiol. 281: L807-L815.
Suwa, T.; Hogg, J. C.; Quinlan, K. B.; Ohgami, A.; Vincent, R.; VanEeden, S. F. (2002) Particulate air pollution
induces progression of atherosclerosis. J. Am. Coll. Cardiol. 39: 935-942.
Takano, H.; Yoshikawa, T.; Ichinose, T.; Miyabara, Y.; Imaoka, K.; Sagai, M. (1997) Diesel exhaust particles
enhance antigen-induced airway inflammation and local cytokine expression in mice. Am. J. Respir. Crit.
Care Med. 156: 36-42.
Takano, H.; Ichinose, T.; Miyabara, Y.; Shibuya, T.; Lim, H.-B.; Yoshikawa, T. Sagai, M. (1998) Inhalation of
diesel exhaust enhances allergen-related eosinophil recruitment and airway hyperresponsiveness in mice.
Toxicol. Appl. Pharmacol. 150: 328-337.
Tan, W. C.; Qiu, D.; Liam, B. L.; Ng, T. P.; Lee, S. H.; Van Eeden, S. F.; D'Yachkova, Y.; Hogg, J. C. (2000) The
human bone marrow response to acute air pollution caused by forest fires. Am. J. Respir. Crit. Care Med.
161: 1213-1217.
Targonski, P. V.; Persky, V. W.; Ramekrishnan, V. (1995) Effect of environmental molds on risk of death from
asthma during the pollen season. J. Allergy Clin. Immunol. 95: 955-961.
Taylor, P. E.; Flagan, R. C.; Valenta, R.; Glovsky, M. M. (2002) Release of allergens as respirable aerosols: a link
between grass pollen and asthma. J. Allergy Clin. Immunol. 109: 51-56.
Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; VanEeden, S. F. (1997) Phagocytosis of small carbon particles
(PM10) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
Am. J. Respir. Crit. Care Med. 155: 1441-1447.
7-244
-------
Terashima, T.; Amakawa, K.; Matsumaru, A.; VanEeden, S.; Hogg, J. C.; Yamaguchi, K. (2001) BAL induces an
increase in peripheral blood neutrophils and cytokine levels in healthy volunteers and patients with
pneumonia. Chest 119: 1724-1729.
Thorn, J.; Rylander, R. (1998a) Inflammatory response after inhalation of bacterial endotoxin assessed by the
induced sputum technique. Thorax 53: 1047-1052.
Thorn, J.; Rylander, R. (1998b) Airways inflammation and glucan in a rowhouse area. Am. J. Respir. Crit. Care
Med. 157: 1798-1803.
Thurston, G. D.; Lippmann, M; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
Timblin, C.; BeruBe, K.; Chrug, A.; Driscoll, K.; Gordon, T.; Hemenway, D.; Walsh, E.; Cummins, A. B.;
Vacek, P.; Mossman, B. (1998) Ambient paniculate matter causes activation of the c-jun
kinase/stress-activated protein kinase cascade and DNA synthesis in lung epithelial cells. Cancer Res.
58: 4543.4547.
Tofler, G. H.; Brezinski, D.; Schafer, A. I.; Czeisler, C. A.; Rutherford, J. D.; Willich, S. N.; Gleason, R. E.;
Williams, G. H.; Muller, J. E. (1987) Concurrent morning increase in platelet aggregability and the risk of
myocardial infarction and sudden cardiac death. N. Engl. J. Med. 316: 1514-1518.
Tomatis, L. (1990) Air pollution and cancer: an old and new problem. In: Tomatis, L., ed. Air pollution and human
cancer. Berlin, Germany: Springer-Verlag; pp. 1-7.
Tsien, A.; Diaz-Sanchez, D.; Ma, J.; Saxon, A. (1997) The organic component of diesel exhaust particles and
phenanthrene, a major polyaromatic hydrocarbon constituent, enhances IgE production by IgE-secreting
EBV-transformed human B cells in vitro. Toxicol. Appl. Pharmacol. 142: 256-263.
Tucker, J. D.; Xu, J.; Stewart, J.; Baciu, P. C.; Ong, T.-M. (1986) Detection of sister chromatid exchanges induced
by volatile genotoxicants. Teratog. Carcinog. Mutagen. 6: 15-21.
Tuomi, T.; Engstrom, B.; Niemela, R.; Svinhufvud, J.; Reijula, K. (2000) Emission of ozone and organic volatiles
from a selection of laser printers and photocopiers. Appl. Occup. Environ. Hyg. 15: 629-634.
Turpin, B. J. (1999) Options for characterizing organic paniculate matter. Environ. Sci. Technol. 33: 76A-79A.
Ulrich, M. M. W.; Alink, G. M.; Kumarathasan, P.; Vincent, R.; Boere, A. J.; Cassee, F. R. (2002) Health effects
and time course of paniculate matter on the cardiopulmonary system in rats with lung inflammation.
J. Toxicol. Environ. Health Part A 65: 1571-1595.
U.S. Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
U.S. Environmental Protection Agency. (1989) An acid aerosols issue paper: health effects and aerometrics.
Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
Assessment Office; EPA report no. EPA-600/8-88-005F. Available from: NTIS, Springfield, VA;
PB91-125864.
U.S. Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
U.S. Environmental Protection Agency. (1996b) Ambient levels and noncancer health effects of inhaled crystalline
and amorphous silica: health issue assessment. Research Triangle Park, NC: National Center for
Environmental Assessment; EPA report no. EPA/600/R-95/115.
U.S. Environmental Protection Agency. (2002) Health assessment document for diesel engine exhaust. Washington,
DC: Office of Research and Development, National Center for Environmental Assessment; report no.
EPA/600/8-90/057F. Available: http://cfpub.epa.gov/ncea/ [22 May, 2003].
Utell, M. J.; Morrow, P. E.; Hyde, R. W. (1983) Latent development of airway hyperreactivity in human subjects
after sulfuric acid aerosol exposure. J. Aerosol Sci. 14: 202-205.
Vallyathan, V; Shi, X.; Dalai, N. S.; Irr, W.; Castranova, V (1988) Generation of free radicals from freshly
fractured silica dust: potential role in acute silica-induced lung injury. Am. Rev. Respir. Dis. 138: 1213-1219.
Van Eeden, S. F.; Tan, W. C.; Suwa, T.; Mukae, H.; Terashima, T.; Fujii, T.; Qui, D.; Vincent, R.; Hogg, J. C.
(2001) Cytokines involved in the systemic inflammatory response induced by exposure to paniculate matter
air pollutants (PM10). Am. J. Respir. Crit. Care Med. 164: 826-830.
VanMaanen, J. M.; Bonn, P. J.; Knaapen, A.; VanHerwijnen, M.; Schilderman, P. A.; Smith, K. R.; Aust, A. E.;
Tomatis, M.; Fubini, B. (1999) in vitro effects of coal fly ashes: hydroxyl radical generation, iron release, and
DNA damage and toxicity in rat lung epithelial cells. Inhalation Toxicol. 11: 1123-1141.
Van Zijverden, M.; Granum, B. (2000) Adjuvant activity of paniculate pollutants in different mouse models.
Toxicology 152: 69-77.
7-245
-------
Van Zijverden, M; Van der Fiji, A.; Bol, M; VanPinxteren, F. A.; De Haar, C.; Penninks, A. H.; VanLoveren, H.;
Pieters, R. (2000) Diesel exhaust, carbon black, and silica particles display distinct Thl/Th2 modulating
activity. Toxicol. Appl. Pharmacol. 168: 131-139.
Van Zijverden, M.; De Haar, C.; Van Beelen, A.; Van Loveren, H.; Penninks, A.; Pieters, R. (2001)
Co-administration of antigen and particles optimally stimulates the immune response in an intranasal
administration model in mice. Toxicol. Appl. Pharmacol. 177: 174-178.
Veronesi, B.; Oortgiesen, M. (2001) Neurogenic inflammation and paniculate matter (PM) air pollutants.
Neurotoxicology 22: 795-810.
Veronesi, B.; Oortgiesen, M.; Carter, J. D.; Devlin, R. B. (1999a) Paniculate matter initiates inflammatory cytokine
release by activation of capsaicin and acid receptors in a human bronchial epithelial cell line. Toxicol. Appl.
Pharmacol. 154: 106-115.
Veronesi, B.; Carter, J. D.; Devlin, R. B.; Simon, S. A.; Oortgiesen, M. (1999b) Neuropeptides and capsaicin
stimulate the release of inflammatory cytokines in a human bronchial epithelial cell line. Neuropeptides
33:447-456.
Veronesi, B.; Oortgiesen, M.; Roy, J.; Carter, J. D.; Simon, S. A.; Gavett, S. H. (2000) Vanilloid (capsaicin)
receptors influence inflammatory sensitivity in response to paniculate matter. Toxicol. Appl. Pharmacol. 169:
66-76.
Veronesi, B.; De Haar, C.; Roy, J.; Oortgiesen, M. (2002a) Paniculate matter inflammation and receptor sensitivity
are target cell specific. Inhalation Toxicol. 14: 159-183.
Veronesi, B.; De Haar, C.; Lee, L.; Oortgiesen, M. (2002b) The surface charge of paniculate matter related to
cytokine release in human bronchial epithelial cells (BEAS-2B). Toxicol. Appl. Pharmacol. 178: 144-154.
Veronesi, B.; Wei, G.; Zeng, J. Q.; Oortgiesen, M. (2003) Electrostatic charge activates inflammatory vanilloid
(VR1) receptors. Neurotoxicology 24: 463-473.
Vincent, R.; Bjarnason, S. G.; Adamson, I. Y. R.; Hedgecock, C.; Kumarathasan, P.; Guenette, J.; Potvin, M.;
Goegan, P.; Bouthillier, L. (1997) Acute pulmonary toxicity of urban paniculate matter and ozone. Am. J.
Pathol. 151: 1563-1570.
Vincent, R.; Kumarathasan, P.; Goegan, P.; Bjarnason, S. G.; Guenette, J.; Berube, D.; Adamson, I. Y.; Desjardins,
S.; Burnett, R. T.; Miller, F. J.; Battistini, B. (2001) Inhalation toxicology of urban ambient paniculate matter:
acute cardiovascular effect in rats. Boston, MA: Health Effects Institute; research report no. 104. Available:
http://www.healtheffects.org/Pubs/Vincent.pdf [29 January 02].
Vinegar, A.; Carson, A. I.; Pepelko, W. E. (1980) Pulmonary function changes in Chinese hamsters exposed six
months to diesel exhaust. In: Pepelko, W. E.; Danner, R. M.; Clarke, N. A., eds. Health effects of diesel
engine emissions: proceedings of an international symposium, v. 2; December, 1979; Cincinnati, OH.
Cincinnati, OH: U.S. Environmental Protection Agency, Health Effects Research Laboratory; pp. 749-756;
EPA report no. EPA-600/9-80-057b. Available from: NTIS, Springfield, VA; PB81-173817.
Vinegar, A.; Carson, A.; Pepelko, W. E.; Orthoefer, J. O. (1981a) Effect of six months of exposure to two levels of
diesel exhaust on pulmonary function of Chinese hamsters. Fed. Proc. 40: 593.
Vinegar, A.; Carson, A.; Pepelko, W. E. (1981b) Pulmonary function changes in Chinese hamsters exposed six
months to diesel exhaust. Environ. Int. 5: 369-371.
Vinitketkumnuen, U.; Kalayanamitra, K.; Chewonarin, T.; Kamens, R. (2002) Paniculate matter, PM10 & PM2.5
levels, and airborne mutagenicity in Chiang Mai, Thailand. Mutat. Res. 519: 121-131.
Vogelzang, P. F. J.; Van Der Gulden, J. W. J.; Folgering, H.; Kolk, J. J.; Heederik, D.; Preller, L.; Tielen, M. J. M.;
Van Schayck, C. P. (1998) Endotoxin exposure as a major determinant of lung function decline in pig
farmers. Am. J. Respir. Crit. Care Med. 157: 15-18.
Wagner, J. G.; Hotchkiss, J. A.; Harkema, J. R. (2001) Effects of ozone and endotoxin coexposure on rat airway
epithelium: potentiation of toxicant-induced alterations. Environ. Health Perspect. 109(suppl. 4): 591-598.
Walters, D. M.; Breysse, P. N.; Wills-Karp, M. (2001) Ambient urban Baltimore particulate-induced airway
hyperresponsiveness and inflammation in mice. Am. J. Respir. Crit. Care Med. 164: 1438-1443.
Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil fly
ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Kodavanti, U. P.; Dreher, K. L.; Su, W.-Y.; Highfill, J. W.; Costa,
D. L. (2000a) Cardiovascular effects following exposure to paniculate matter in healthy and
cardiopulmonary-compromised rats. In: Heinrich, U.; Mohr, U., eds. Relationships between acute and chronic
effects of air pollution. Washington, DC: ISLI Press, pp. 447-463.
Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U.P.; Highfill, J. W.; Costa, D. L. (2000b)
Thermoregulatory effects following exposure to paniculate matter in healthy and
cardiopulmonary-compromised rats. J. Therm. Biol. 25: 131-137.
7-246
-------
Watkinson, W. P.; Campen, M. J.; Nolan, J. P.; Costa, D. L. (2001) Cardiovascular and systemic responses to
inhaled pollutants in rodents: effects of ozone and paniculate matter. Environ. Health Perspect. 109(suppl. 4):
539-546.
Waxman, M. B.; Cameron, D.; Wald, R. W. (1994) Vagal activity and ventricular tachyarrhythmias. In: Levy, M.;
Schwartz, P. Vagal control of the heart: experimental basis and clinical implications. Armonk, NY: Futura
Publishing Co.; pp. 579-612.
Wellenius, G. A.; Saldiva, P. H. N.; Batalha, J. R. F.; Murthy, G. G. K.; Coull, B. A.; Verrier, R. L.; Godleski, J. J.
(2002) Electrocardiographic changes during exposure to residual oil fly ash (ROFA) particles in a rat model
of myocardial infarction. Toxicol. Sci. 66: 327-335.
Wesselkamper, S. C.; Prows, D. R.; Biswas, P.; Willeke, K.; Bingham, E.; Leikauf, G. D. (2000) Genetic
susceptibility to irritant-induced acute lung injury in mice. Am. J. Physiol. 279: L575-L582.
Wexler, A. S.; Sarangapani, R. (1998) Particles do not increase vapor deposition in human airways. J. Aerosol Sci.
29: 197-204.
Wheatley, L. M.; Platts-Mills, T. A. E. (1996) Perennial allergens and the asthma epidemic. Sci. Med. 3: 6-13.
Willich, S. N.; Levy, D.; Rocco, M. B.; Toiler, G. H.; Stone, P. H.; Muller, J. E. (1987) Circadian variation in the
incidence of sudden cardiac death in the Framingham Heart Study population. Am. J. Cardiol. 60: 801-806.
Wilson, W. E. (1995) Aerosol exposure, physics, and chemistry. Inhalation Toxicol. 7: 769-772.
Wilson, A. F.; Novey, H. S.; Berke, R. A.; Surprenant, E. L. (1973) Deposition of inhaled pollen and pollen extract
in human airways. N. Engl. J. Med. 288: 1056-1058.
Wolff, R. K.; Gray, R. L. (1980) Tracheal clearance of particles. In: Diel, J. H.; Bice, D. E.; Martinez, B. S., eds.
Inhalation Toxicology Research Institute annual report: 1979-1980. Albuquerque, NM: Lovelace Biomedical
and Environmental Research Institute; p. 252; report no. LMF-84.
Wolff, R. K.; Henderson, R. F.; Snipes, M. B.; Griffith, W. C.; Mauderly, J. L.; Cuddihy, R. G.; McClellan, R. O.
(1987) Alterations in particle accumulation and clearance in lungs of rats chronically exposed to diesel
exhaust. Fundam. Appl. Toxicol. 9: 154-166.
Wong, D.; Mitchell, C. E.; Wolff, R. K.; Mauderly, J. L.; Jeffrey, A. M. (1986) Identification of DNA damage as a
result of exposure of rats to diesel engine exhaust. Carcinogenesis 7: 1595-1597.
World Health Organization. (1993) Biomarkers and risk assessment: concepts and principles. Geneva, Switzerland:
World Health Organization. (Environmental health criteria: v. 155).
Wu, W.; Samet, J. M.; Ohio, A. J.; Devlin, R. B. (2001) Activation of the EOF receptor signaling pathway in human
airway epithelial cells exposed to Utah Valley PM. Am. J. Physiol. 281: L483-L489.
Yang, C. S.; Johanning, E. (2002) Airborne fungi and mycotoxins. In: Hurst, C. J., Crawford, R. L.; Knudsen, G. R.;
Mclnerney, M. J.; Stetzenbach, L. D. eds. Manual of environmental microbiology. 2nd ed. Washington, DC:
American Society for Microbiology; pp. 839-852.
Young, R. S.; Jones, A. M.; Nicholls, P. J. (1998) Something in the air: endotoxins and glucans as environmental
troublemakers. J. Pharm. Pharmacol. 50: 11-17.
Zelikoff, J. T.; Sisco, M.; Cohen, M. D.; Frampton, M. W.; Utell, M. J; Schlesinger, R. B. (1994) Interspecies
comparison of immunotoxicity of inhaled sulfuric acid. II. New Zealand white rabbits. In: 1994 international
conference sponsored by the American Lung Association and the American Thoracic Society; May; Boston,
MA. Am. J. Respir. Crit. Care Med. 149: A621.
Zelikoff, J. T.; Frampton, M. W.; Cohen, M. D.; Morrow, P. E.; Sisco, M.; Tsai, Y.; Utell, M. J.; Schlesinger, R. B.
(1997) Effects of inhaled sulfuric acid aerosols on pulmonary immunocompetence: a comparative study in
humans and animals. Inhalation Toxicol. 9: 731-752.
Zelikoff, J. T.; Chen, L. C.; Cohen, M. D.; Fang, K.; Gordon, T.; Li, Y.; Nadziejko, C.; Schlesinger, R. B. (2003)
Effects of inhaled ambient paniculate matter on pulmonary antimicrobial immune defense. Inhalation
Toxicol. 15: 131-150.
Zhang, Z.; Shen, H. M.; Zhang, Q. F.; Ong, C.-N. (2000) Involvement of oxidative stress in crystalline
silica-induced cytotoxicity and genotoxicity in rat alveolar macrophages. Environ. Res. 82: 245-252.
Zock, J.-P.; Hollander, A.; Heederik, D.; Douwes, J. (1998) Acute lung function changes and low endotoxin
exposures in the potato processing industry. Am. J. Ind. Med. 33: 384-391.
7-247
-------
APPENDIX 7A. RAT-TO-HUMAN
DOSE EXTRAPOLATION
7A.1 INTRODUCTION
As noted at the outset of Chapter 7, the 1997 revisions to the PM NAAQS (Federal
Register, 1997) were largely based on newly emerging epidemiologic evidence that showed
associations between (a) ambient PM measured at community monitoring stations and
(b) increased risks for mortality and morbidity (especially cardiorespiratory-related) among
human populations exposed to contemporary U.S. ambient PM concentrations. However, little
experimental toxicology data from controlled laboratory animal or human exposure studies were
then available that provided more direct evidence supporting the plausibility of the PM-
mortality/morbidity relationships observed at relatively low ambient PM concentrations.
Since completion of the 1996 PM AQCD supporting the 1997 PM NAAQS decisions,
numerous hypotheses have been advanced and extensive new toxicologic evidence generated
with regard to possible pathophysiologic mechanisms by which PM exposures (even at low
ambient concentrations) might induce increased morbidity and/or mortality. Much of the new
toxicologic data (as addressed in Chapter 7) has involved either (a) experimental in vivo
exposures of human subjects and/or laboratory animals via inhalation exposures and/or
instillation of PM materials into the lung or trachea or (b) in vitro exposures of various (mostly
respiratory tract) cells or tissues to diverse types of PM. The exposure conditions used in these
studies were typically different from those experienced through inhalation of ambient PM.
Therefore, the relevance of the effects observed under experimental conditions compared to the
effects observed in humans following ambient PM exposures needed evaluation.
To address this issue, the EPA has conducted analyses of relationships between rat and
human lung doses predicted for various exposure scenarios ranging from ambient PM exposures
to PM instillations into the lung. This appendix begins in Section 7A.2 by presenting basic
principles such as the relationship between PM exposure and PM dose in the lung. The section
then introduces the concept of determining PM exposures for rats which lead to PM doses in the
rat lung equivalent to those received by humans. The mathematical model used herein for
7A-1
-------
interspecies comparisons is discussed in Section 7A.3. Particle dosimetry in the lung was
described in Chapter 6; however, additional details regarding differences in particle dosimetry
between rats and humans are discussed in Section 7A.4. Section 7A.5 expands on the equivalent
dose concept and illustrates the variability in PM exposure concentrations that could be required
for rats to have the same dose as a human as a function of dose metric, normalizing factor, and
level of human exertion. Section 7A.5 provides information that can be used to estimate the
exposure concentrations required to give a rat a dose equivalent to the dose that would be
received by a human exposed to various levels of ambient PM. In Section 7A.7, the dosimetric
modeling techniques discussed earlier are used to compare doses received by rats and humans
from experimental exposures. That dosimetry alone cannot explain all differences in response
between rats and humans is discussed in Section 7A.6 and again in Section 7A.8. Readers not
interested in the comprehensive analyses of dosimetric issues presented in Sections 7A.2 through
7A.6 may wish to skip to Section 7A.7, where several specific studies are compared and
contrasted and then further discussed in Section 7A.8. Finally, conclusions based on the
analyses are presented in Section 7A.9.
7A.2 QUANTITATIVE INTERSPECIES EXTRAPOLATION
Much of the information on the toxicity of PM comes from studies in which laboratory rats
were exposed to PM by inhalation or instillation. For optimal use of these toxicologic data,
estimates of PM exposures that would result in similar human doses are needed. The premise of
such comparisons is that comparable doses should cause comparable effects. It is the tissue
dose, rather than exposure per se, that is responsible for observed responses, making it essential
to first consider the dose to the lung that might occur during an exposure to PM.
The rate of deposition in a specific region of the respiratory tract resulting from the
inhalation of PM may be given as
Dr (t) = C(t) x f(t) x VT(t) x DFr(t) (7A-1)
7A-2
-------
where: Dr is the rate of deposition per unit time in region r; C is the PM exposure concentration
and may be expressed as particle mass, surface area, or number per unit volume; f is breathing
frequency in breaths per unit time; VT is tidal volume, i.e., the volume of air inhaled per breath;
and DFr is the fraction of inhaled particles deposited in region r.
It should be noted that all of the variables in Equation 7A-1 can potentially vary over time.
The effect of activity or exertion level on VT and f was presented in Tables 6-3 and 6-6. Within
an individual, the variability in DFr over time is largely attributable to variations in inhaled
particle size, f, VT, and route of breathing, i.e., mouth versus nose (ICRP, 1994). Intersubject
and interspecies variability in DFr is additionally affected by morphologic differences in the size
and structure of the respiratory tract.
Health effects may be due to deposited dose, retained dose, or a combination of both.
Some effects associated with acute exposures may simply be a function of the deposited dose in
a region (Dr) of the respiratory tract, given by
Dr = J D, (t) dt (7A-2)
*x\t
where At is the exposure time interval.
For chronic exposures, it may be useful to consider both the retained dose due to long term
exposure and the acute deposited dose. The PM dose retained in a region of the lung is
determined by the balance between rate of input and the rate of removal. The PM burden (Br) in
a lung region may be expressed as
= Ł)r(t)-ArBr(t) (7A-3)
where A r is the clearance rate constant for region r. It should be noted that transfer into region r
from another region may also occur. Such situations in which a region receives a portion of its
burden from another region are common in the lung, e.g., the mucus clearance of the segmental
bronchi into the lobar bronchi, which clear into the main bronchi, which in turn clear into the
7A-2
-------
trachea. In addition, the clearance from one region can transfer burden into more than one other
compartment, e.g., soluble particles in the airways may be cleared into the blood as well as via
the mucus. The discussion herein of retention is mainly limited to poorly soluble particles.
However, multiple pathways for clearance of insoluble particles exist such as from the alveoli
into the lymph and into the terminal bronchioles via macrophages.
For instillations into the lung, the total instilled dose can be characterized fairly well.
For inhalation studies, however, the dose is not always known and must instead be calculated
using a dosimetric model that may be based on empirical relationships, theoretical calculations,
or a combination. The following discussion is based on the application of dosimetry to
interspecies extrapolation as given in the scientific literature (U.S. Environmental Protection
Agency 1994, 1996; Jarabek, 1994, 1995).
For dosimetric calculations and comparisons, it is useful to assume that PM concentrations
and activity levels are constant over time. Further, it is convenient to separate the deposited dose
into one factor that depends on the exposure-related variables and a second factor that depends
on species, particle size, and activity level. Exposure, E, can be defined as
E = C x At (7A-4)
where: C is PM exposure concentration and At is exposure duration. A dose adjustment factor,
DAF, can also be defined as
DAF = f x Vt x DF (7A-5)
where it is understood that DF refers to specific regions of the lung. Retained dose can be
expressed similarly except that the DAF would include a retention fraction.
In order to compare a rat dose with a human dose that might have comparable biological
effects, it is useful to introduce the concept of dose normalization. Examples of normalized
doses are the dose per body mass, per lung mass, per lung area, per macrophage, or per other
biological or physiological parameters. A normalized dose (ND) is the dose (D) to the lung or
lung region divided by an appropriate normalizing factor (NF):
7A-4
-------
D Ex DAF
ND = = (7A-6)
NF NF V '
In Equation 7A-6, ND and DAF refer to specific regions of the lung and could apply to either a
rat or human. In the extrapolation modeling presented here, normalized doses are calculated for
rats and humans. The concept of dose normalization is not new to interspecies extrapolation of
toxicologic data. The ingested dose that produces no adverse effect in animals is normalized and
used to estimate a comparable human dose. Typically, an uncertainty factor of 10 is applied to
the estimated human dose unless a dosimetric adjustment is made, in which case, the uncertainty
factor is reduced to 3 (U.S. Environmental Protection Agency, 1994; Jarabek, 1995).
The objective of the analysis set forth here is to specify an exposure for one species and
determine an exposure for the second species, such that both species will receive equivalent
normalized doses,
NDR=NDH (7A-7)
where: subscripts refer to rats (R) and humans (H). Substituting in Equation 7A-6 gives
ER x (DAFR / NFR) = EH x (DAFH / NFH) (7A-8)
For a given human exposure, Equation 7A-8 may be solved to obtain the rat exposure such that
both species receive equivalent normalized doses. An equivalent exposure ratio (EqER)
represents the ratio of species' exposures that give equivalent doses.
ER (DAFn/NFn)
Thus, EqER is the factor by which a specified human exposure concentration must be multiplied
to obtain a rat exposure concentration yielding an equivalent dose. EqER can be calculated
7A-5
-------
directly from the DAF and NF for the two species provided that the dose is a linear function of
time and concentration. If the exposure time is the same for both species, Equation 7A-9 can be
reduced to
CR = EqERxCH (7 A-10)
If EqER is greater than 1, then the rat must receive a greater concentration than the human in
order to receive an equivalent dose.
7A.3 THE MULTIPLE PATH PARTICLE DOSIMETRY MODEL (MPPD)
The deposition and clearance of particles in the human and rat respiratory tract was
estimated using the publicly available Multiple Path Particle Dosimetry (MPPD) model1.
The MPPD model was developed by the CUT Centers for Health Research (CUT), USA,
in collaboration with the National Institute of Public Health and the Environment (RIVM),
the Netherlands, and the Ministry of Housing, Spatial Planning and the Environment, the
Netherlands. Other models of deposition and clearance, which are not necessarily publicly
available nor in a form easily suited for comparisons between particle disposition in rats and
humans, were discussed in Chapter 6 (Sections 6.6.1 to 6.6.3). General information about the
MPPD model was discussed in Chapter 6, Section 6.6.4.2; additional details relevant to this
appendix are provided here. Comparisons between MPPD-predicted deposition fractions of
monodisperse particles (0.01 to 10 jim) in humans during light exercise and in rats at rest were
provided in Chapter 6, Section 6.6.43. Differences between rats and humans in deposition
normalized to lung mass and lung surface were also provided. In this appendix, other
normalizing parameters are considered as is the clearance of particles from the lung.
The MPPD model may be used to predict the deposition of particles between 0.01 to 20 jim
in diameter in humans and rats. In the lung, the model considers deposition by the mechanisms
1 Some software problems encountered during the dosimetric modeling were fixed by the developers; and a revised
MPPD upgrade version is available on request from the CUT Centers for Health Research ().
7A-6
-------
of impact!on, sedimentation, and diffusion. The model does not consider particle interception,
charge, or hygroscopic growth. Although the lung geometries differ between species, the same
mathematical formulation may be used to calculate particle deposition in the rat as well as in the
human lung (Anjilvel and Asgharian, 1995). The extrathoracic particle deposition efficiencies
used in the MPPD model were adopted from the ICRP (1994) for humans and from Zhang and
Yu (1993) for rats. Model input parameters include airways morphology, particle properties
(size distribution, density, concentration), and breathing conditions (tidal volume, breathing
frequency, and mode of breathing). The effects of these parameters on deposition in rats and
humans were reported by Winter-Sorkina and Cassee (2002). The MPPD model also contains an
optional correction for the inhalability of particles during nasal breathing which may be applied
to both humans and rats (Menache et al., 1995). This correction becomes increasingly important
when particle size exceeds 1 |im (MMAD) for rats and 10 jim (MMAD) for humans. With
reference to Equation 7A-1, it should be noted that average exposure concentrations and average
breathing patterns are used to estimate particle deposition fractions and lung doses over discrete
time periods, i.e., the simulations presented herein do not consider temporal variations on a
breath-by-breath basis as suggested by Equation 7A-1.
Several types of normalized deposition predictions are available using the MPPD model.
Particle deposition fractions normalized to airway surface area provide an index of the average
dose of particles to epithelial cells. These data are useful in assessing generation-to-generation
variability but do not consider dose variability within a generation, e.g., between the carina and
airway wall. For this normalization, the MPPD model calculates the surface area of the airways
based on the diameter, length, and number of airways in a generation. These data are most
useful for the tracheobronchial airways since alveolar surface area is not included in the model's
calculations. For the alveolar region, the MPPD model calculates particle mass and number
deposited per alveolus and per macrophage. From Mercer et al. (1994), the model assumes
4.86 x 108 alveoli in humans and 1.97 x 107 alveoli in rats. From Miller (2000), the number of
alveolar macrophage (AM) per alveolus assumed in the model is 12.3 in humans and 1.5 in rats.
However, an influx of monocytes and macrophages into the alveoli occurs following acute
exposures to numerous pollutants, e.g., PM, O3, and NO (Oberdorster, 1988; Mercer, 1999;
Driscoll, 1988). Furthermore, the volume (and capacity) of a human AM is about 1.5 times that
7A-7
-------
of a rat macrophage (Miller, 2000). Hence, it is difficult to interpret a dose metric like the
predicted number of particles deposited per macrophage.
The balance between deposition and clearance affects tissue dose and lung burden. The
MPPD model considers the lung clearance of insoluble particles as a two-phase process. The
rapid first phase, tracheobronchial clearance, occurs via the action of the mucociliary escalator.
The second clearance phase is the slow removal of particles that have deposited in the alveolar
region of the lung. The model considers particle clearance from the alveolar surface via
macrophages to the distal TB airways and via the lymphatic system. The clearance of soluble
particles is not considered by the current MPPD model.
The MPPD model estimates mucus clearance of insoluble particles in the human and rat
lung by assuming a mass balance between the volume of mucus produced in the terminal
bronchioles and the volume exiting the trachea. By further assuming that the production of
mucus is the same in all terminal bronchioles, the mucus velocity in terminal bronchioles may be
determined given tracheal mucus velocity, tracheal diameter, and the number and diameter of
terminal bronchioles. Moving proximally from the terminal bronchioles, the mucus velocity in
each parent airway is based on its diameter and daughter airways' diameters and mucus
velocities. The mucus within an airway is assumed to travel at a constant velocity which
incorporates any discontinuity in mucus blanket. An implicit assumption in this mucus
clearance model is that particles are transported with the mucus blanket, i.e., there is no particle
size-dependent slow-cleared fraction from the airways as in the ICRP (1994) model. A more
detailed description of the MPPD mucus clearance model appears elsewhere (Asgharian et al.,
2001; Hofmann and Asgharian, 2003). Model simulations of tracheobronchial clearance,
presented herein, assumed tracheal mucus velocities of 1.9 mm/min in rats (Felicetti et al., 1981)
and 5.5 mm/min in humans (ICRP, 1994).
Clearance from the alveolar region of the lung is treated somewhat differently between
humans and rats by the MPPD model. For humans, the alveolar clearance model was adopted
from the ICRP (1994). In that model, the alveolar region consists of three compartments which
clear particles into the bronchioles at the rates of 0.02, 0.001, and 0.0001 day"1. Of the particles
deposited in the alveolar region, 30% was assumed in the fast compartment, 60% in the medium
rate, and 10% in the slow compartment. The slow compartment also clears via lymphatic
7A-8
-------
channels at a rate of 0.00002 day l. In rats, the MPPD model considers the overall alveolar
clearance rate as the sum of the transport rates to the terminal bronchioles and to the lymph. The
alveolar clearance rate constants are based on the pulmonary retention and lymphatic uptake of
titanium dioxide particles (MMAD = 1.44 |im, og = 1.71) following a 13-week exposure (6 h per
day, 5 day per week) to 10, 50, or 250 mg/m3 (Bermudez et al., 2002). Average postexposure
alveolar rate constants of 0.00693, 0.00214, and 0.00083 day"1 and postexposure pulmonary
burdens of approximately 1, 8, and 41 mg were observed for the 10, 50, or 250 mg/m3 exposures,
respectively. The rates of clearance observed by Bermudez et al. (2002), with the fastest
reported retention half-time being 100 days for the 10 mg/m3 exposure, are slow relative to
healthy rats which typically have a retention half-time of-70 days (Oberdorster, 1995). This
reduction in alveolar clearance rates suggests that rat's macrophage-mediated clearance was
somewhat impaired even at the lowest exposure concentration of 10 mg/m3 in the Bermudez
et al. (2002) study. Translocation into the lymph nodes increased in a concentration dependent
manner. Based on these data, the MPPD model assumes that the overall alveolar clearance
rate (AA) decreases with pulmonary burden (mA) in rats. Specifically, AA for rats equals
[0.03341 x exp (- 1.7759mAa3123) + 0.00072] day"1. Based on this equation, a clearance rate
typical of healthy rats (t1/2 ~ 70 days, AA = 0.01 day"1) occurs at a lung burden of 0.4 mg. The
assumed clearance rate from the alveoli to the lymphatic system is 0.00106 day"1. The MPPD
model, in effect, treats the clearance of particles from the alveolar surface (via macrophages) to
the distal airways as a pathway subject to saturation or overload in rats but not in humans.
The current version of the MPPD model does not offer the option of calculating clearance
for exposures to multiple polydisperse aerosol modes or for multiple activity levels. Also,
MPPD clearance calculations for rats during chronic exposures are quite computationally
intensive, taking approximately 10 minutes on a Pentium computer (2.8 GHz with 512 MB of
RAM) to determine retention at 1 year of exposure. For such cases, alveolar clearance was
calculated in a spreadsheet, instead of the MPPD model, using the deposition fraction (calculated
using the MPPD model) and the same clearance rate constants as used by the model. Based on
Equation 7A-3, the alveolar burden in rats was calculated as
7A-9
-------
BR(t) = DR(t- At) At + BR (t- At) exp(^/lAAt) (7A-1 1)
where: BR is the alveolar burden in a rat; t is time; DR is the dose rate to the alveolar region of
the rat; At is the time increment for the calculations and was selected to be -1% (or less) of the
retention half-time (i.e., 0.693 / AA); and AA is the overall alveolar clearance rate in the rat.
Alveolar burden in humans was computed similarly for the three alveolar compartments (see
above discussion) in humans as
BH(t) = {FH.DH(t-At)At + BH.(t-At) exp(-AH.At)} (7A-12)
i=l
BH is the alveolar burden in a human; FH is the fraction of alveolar deposition distributed to the
i
rth alveolar compartment; DH is the dose rate to the human alveolar region; At is the time
increment for the calculations and was selected to be -1% (or less) of the fastest compartment's
retention half-time (35 days); BH is the burden in the rth alveolar compartment; and AH is the
i i
clearance rate constant for the rth alveolar compartment.
7A.4 RAT AND HUMAN DOSIMETRY: INTERSPECIES DIFFERENCES
Before providing illustrative examples of how a dosimetric model may be used in rat-to-
human extrapolation, it is useful to discuss some of the many differences between rat and human
exposure and dosimetry.
7A.4.1 Anatomy
The structure and function of the respiratory tract differs in rats and humans in ways that
affect the deposition of particles in the lung. Rats are obligate nose breathers whereas most
humans are oronasal breathers who breathe through the nose when at rest but who breathe
increasingly through the mouth with increasing activity. It has been estimated that 13% of the
7 A-10
-------
human population are "mouth-breathers" (Niinimaa, 1981). This distinction is important
because the nose is a more efficient filter than the mouth for preventing the penetration of
particles into the lung. Thus, by breathing through the mouth, humans effectively increase the
amount of inhaled particles reaching the lung. Even when breathing through the nose, humans
have greater TB and A region deposition fractions for coarse particles compared to rats due to
the lower inhalability of particles larger than 3 jim in the rat. The structure of the human and rat
intrathoracic airways also differs in ways that affect the regional deposition pattern in the lung.
The branching structure of the lung is monopodial in rats and symmetrically dichotomous in
humans. A monopodial structure has the potential to allow increased penetration of large
particles into the A region. Rats also lack respiratory bronchioles, a site of early airway disease
in humans.
7A.4.2 Exposure Scenarios
7A.4.2.1 Exertion Level
The amount of PM inhaled is influenced by exertion level and the related lung ventilation
rate. Chapter 6 discussed how increasing exertion leads to greater deposition of PM in the
human lung due to changes in the mode of breathing (nasal to oronasal to oral) as well as the
inhalation of greater quantities of PM per unit time due to an increase in minute ventilation
(breaths per minute times the tidal volume in liters) (Figure 6-18). Humans typically experience
a range of breathing patterns during exposure to ambient PM, including those experienced
during light and heavy exertion as well as at rest and during sleep. In contrast, laboratory rats
are commonly at rest when exposed to PM by inhalation. It is not clear which human breathing
pattern is most appropriate for use in an extrapolation. However, just because the rat received its
dose while resting does not mean that only the dose received by a resting human should be of
interest. The quantity of PM inhaled during a specified time period is given by
PM (Inhaled) = CxfxV,xt = Cx minute ventilation x t (7A-13)
where C may be given in jig, jim2, or particle number per m3.
7 A-11
-------
Breathing patterns used in subsequent dosimetric calculations are given in Table 7A-1.
The minute ventilation, and therefore the mass of PM inhaled per unit time, will increase with
exertion level.
TABLE 7A-1. HUMAN AND RAT BREATHING PATTERNS USED IN DOSIMETRIC
CALCULATIONS
Activity
Breaths/min
Tidal volume, mL
Minute ventilation, L/min
Awake
Rest3
12
625
7.5
Slow
Walk3
16
813
13
Human
Light
Exertion a
19
1000
19
Moderate
Exertion a
28
1429
40
Heavy
Exertion b
26
1923
50
Rat
Awake
Rest3
102
2.1
0.214
1 Winter-Sorkinaand Cassee (2002), bICRP (1994).
7A.4.2.2 Size Distribution
The atmospheric aerosol to which people are exposed may be thought of in terms of three
particle classes: coarse mode particles (greater than about 1 jim in diameter), accumulation
mode particles (about 0.1 to 1.0 jim in diameter, although accumulation mode PM may grow
into 1 to 2.5 |im diameter size range at very high relative humidities), and ultrafine particles
(< 0.1 |im in diameter, including the nucleation and Aitken modes [see Chapter 2]). However,
in toxicologic studies, laboratory rats are rarely exposed to all three size classes at the same time.
Some experimental studies reported in the literature use diesel exhaust (ultrafine particles but
with some coagulation into the accumulation mode size range), concentrated accumulation mode
particles (concentrated air particles [CAPs]), or particles with a narrow size range within the
accumulation mode size range (e.g., studies of acid aerosol). A more recent development is the
ultrafine concentrator in which ultrafine particles are separated from larger particles, grown by
humidification, concentrated, and dehydrated to reconstitute ultrafine particles. In other studies,
rats have been exposed to particles produced by resuspension of bulk material or resuspension of
7 A-12
-------
particles previously collected from specific sources (e.g., resuspended oil fly ash, ROFA, or
from ambient air). Particles produced by resuspension are frequently passed through an inertial
separator (cyclone or impactor) to remove particles > 2.5 jim diameter, thus leaving particles
with a nominal MMAD between 1 and 2 jim. The particle size distribution is important because
the deposition fraction and the pattern of deposition in the lung varies with particle size.
Some studies suggest that particle surface area (Oberdorster et al., 1994, 2000) or possibly
particle number (Wichmann and Peters, 2000; Wichmann et al., 2000) may be as (or more)
important than mass in determining the extent of health effects. Figure 7A-la shows the mass
size distribution of a representative resuspended dust (MMAD = 2 jim, og = 2) overlaid on an
atmospheric mass size distribution. Figures 7A-lb and 7A-lc show the distribution of particle
surface area and number, respectively. The coarse mode and the resuspended PM mode
contribute little to the total particle surface area and contribute minimally to the particle number
concentration (note the logarithmic scale for number concentration). Particle characteristics
used in subsequent dosimetric calculations and some examples of deposition fractions calculated
with the MPPD model are given in Table 7A-2.
In many cases, it is difficult to find good quality and precise information on the size
distribution of particles used in laboratory exposure studies. Accumulation mode CAPs might
be expected to have a size distribution similar to the accumulation mode in the atmosphere.
However, most concentrators have an upper cut of 2.5 jim and do not concentrate particles below
about 0.1 to 0.15 jim. Hence, the lower tail of the accumulation mode will not be concentrated
while the lower tail of the coarse mode will be. Thus, in atmospheres not influenced by fog or
clouds, the size distribution of the CAPs might be bimodal or otherwise non-1 ognormal. In any
event, a single MMD and og probably will not be adequate to characterize the size distribution.
Values of og in excess of 2.5, for example as reported for some CAPs (Gordon et al., 2000,
2004), suggest a multimodal, rather than a monomodal, distribution. If a combined
accumulation/ultrafine concentration technique is used, the resulting CAPs would be expected to
contain several modes. Diesel exhaust, as generated for laboratory exposures, probably has a
nucleation mode and an Aitken mode, with some particles possibly having grown by coagulation
into the lower end of the accumulation mode. Thus, the size distribution of diesel particles
7 A-13
-------
140 •,
120 •
E
1> 100
g" 80
O)
2. 60
<
S 40-
<
20
0 •
0
70 -I
60
n
1 60
° 40
Cf
O) ,n
O 30
<
~- 20
(O
<
10
0
0
•ins .,
•in7 -
E 1°
0 ins.
Q
O
< 103 -
•inl •
0
Figure 7A-la,b,c. Size dist
the aver
resuspen
distribuf
| are show
Resuspended
-"- ,, /VST
/\ / \ A
/ \ / A \
/Accumulation\ / \
Aitken / (65) \ / / \
(10) / y / \
.01 0.1 1 10
/\ b. Surface Area
jkitken\
' \ A
\ / \
\ / \
\ / \ Resuspended
\ / Accumulation \ i
\ / \ T Coarse
\ / \ ^^ ^^ i
01 0.1 1 10
c. Number
Aitken\
\
\"'""" ""^,, Accumulation
A \
/ \ \
/ \ ^ \~~-»^ Resuspended
/ X \ \
1 / \ \ .,„- — — — .-\
/ \ /'t '\xCoarse
01 0.1 1 10
Particle Diameter, (jrn
ributions of the Aitken, accumulation, and coarse modes of
age urban aerosol (as reported by Whitby [1978]) and a
ided PM mode: (a) mass distribution, (b) surface area
tion, and (c) number distribution. Concentrations, in ug/cm3,
m under the name of each mode in (a).
7 A-14
-------
TABLE 7A-2. PARTICLE CHARACTERISTICS USED BY EPA IN MPPD MODEL
CALCULATIONS AND SOME EXAMPLES OF REGIONAL
DEPOSITION FRACTIONS
Size Distributions
Mass Mean Diameter, (im
Surface Mean Diameter, (im
Number Mean Diameter, (im
Geometric Standard Deviation, og
Density, g/ml
% Mass in Size Range
Fraction Deposited*
TB Region
A Region
Thoracic Region
Aitken a
0.031
0.023
0.013
1.7
1
6.7
0.19
0.32
0.51
Human
Accumulation a
0.31
0.19
0.069
2
1
43.3
0.062
0.1
0.16
Coarse"
5.7
3.3
1.1
2.1
1
50
0.024
0.055
0.079
Rat
Resuspended b
2
1.2
0.47
2
1
100
0.04
0.058
0.098
a Size distribution for human calculations from Whitby (1978).
b The size distribution for resuspended PM is based on several reported size distributions. To remove larger
particles, some studies have used a cyclone (Kodavanti et al., 2002; Dormans et al, 1999) or an impactor
(Killingsworth et al., 1997) often with a 50% cut point at 2.5 ^m diameter.
0 Calculated with the MPPD model for activity levels of light exertion for humans and rest for rats.
cannot be adequately modeled as a monomodal distribution. In addition, diesel exhaust contains
particles below 0.01 jim in diameter. Since the lower limit of the MPPD model is 0.01 jim,
it may underestimate the number of diesel exhaust particles depositing in the lung. The analysis
presented here is limited to particles between 0.01 and 20 jam in diameter.
7A.4.3 Quantities Calculated by Dosimetric Models
7A.4.3.1 Deposition Fraction (DF)
The fraction of inhaled particles deposited in various regions of the respiratory tract
depends on the particle size and the breathing pattern (breaths per minute, tidal volume, and
7 A-15
-------
whether breathing by nose or mouth). Examples of the ratio, DFH/DFR, for a resting rat and a
human at various activity levels for nasal and oral breathing are given in Figures 7A-2 and 7A-3.
The ratio increases rapidly for particle diameters above about 5 jim diameter due to
differences in inhalability as shown in Figure 7A-4. The DFH/DFR for the TB and A region
differs only by a small factor in the accumulation size range. Due to the lower inhalability of
coarse particles by the rat and differences in the nasal passages of the rat and human, the ratio is
quite variable for coarse particles. The ratio is also variable for ultrafine particles due partially
to differences in the removal of ultrafine particles in the extrathoracic region.
7A.4.3.2 Clearance
Poorly soluble fine and coarse particles deposited in the lung are cleared by a variety of
mechanisms as discussed in Chapter 6. However, the clearance rates from both the TB and
A regions are much higher for rats than for humans. Figures 7A-5a and 7A-5b show examples
of accumulation and clearance for the TB region for humans and rats. Note the different time
scales for the two figures. Because of these species differences in clearance rates, retention half-
times also vary by species. Retention half-times in the TB region are highly dependent on the
site of deposition, but generally range from 1 to 2 h in rats and 4 to 10 h in healthy humans
(Hoffmann and Asgharian, 2003).
Figure 7A-6 compares the longer term clearance of particles initially deposited in the
A region for several species (Oberdorster, 1988). A more recent review is given by Kreyling and
Scheuch (2000). Clearance from the A region is much slower than clearance from the TB region
for both humans and rats, while particles deposited in the A region are cleared more rapidly from
the rat than the human. For the A region, retention half-times are 60 to 80 days in rats but up to
2 years in humans.
7A.4.3.3 Retention
Figures 7A-5 and 7A-6 show the clearance of particles after exposure had ceased as a
fraction of the particles present in the lung at the time exposure ceased. For chronic exposures,
however, it is necessary to consider the retained dose. In comparing retention for rats and
humans, how much of the deposited PM remains in the lung after exposures of various
7 A-16
-------
0.01
0.01
0.1 1
Particle Diameter, |jm
Figure 7A-2a,b,c. The ratio of the predicted deposition fractions for human relative to rat
at rest, DFH/DFR, (a) the head region, (b) the TB region, and (c) the A
region for nasal breathing corrected for particle inhalability.
7 A-17
-------
n
E
3
X
(C
Qi
c
CO
X
LL
Q
ra
a:
c
(0
E
0.01
10
0.01
0.1 1
Particle Diameter, |jm
Figure 7A-3a,b,c. The ratio of the predicted deposition fractions for human relative to rat
at rest, DFH/DFR, (a) the head region, (b) the TB region, and (c) the A
region for oral breathing corrected for particle inhalability.
7 A-18
-------
0.0
1.0
Particle Diameter
10.0
20.0
Figure 7A-4. Inhalability curves for human and rat showing the fraction of PM that
enters the nose (based on empirical fit to experimental data given in Table 1
from Menache et al., 1995).
magnitudes and durations is of interest. The PM dose retained in a region of the lung is
determined by the balance between the rate of input (deposition) and the rate of removal
(clearance) as described by Equation 7A-3. Figure 7A-7a and 7A-7b show how the balance
between deposition (for 6 h) followed by clearance (for 18 h) leads to differences in retained
burdens between rats and humans and between the TB and A regions. The scenario depicted is
for a 6-h-per-day, 3-day exposure to 100 |ig/m3 of 2-|im diameter particles with a og of 2.0. The
y-axis on these figures is the fraction of total PM mass (i.e., the total mass that would be
deposited in the TB or A region over the 3-day exposure period) that is retained in the TB or
A region. As shown in Figure 7A-7a, because of the more rapid clearance of the rat, the fraction
of deposited mass retained in the TB region is much smaller for the rat than the human. The
maximum retained dose in the rat TB region is never greater than 0.07 of the total deposited
7 A-19
-------
a. Human
8 16 24 32 40
Post-exposure Time, Hours
48
b. Rat
— —• 0.1 urn
1 um
2 um
7 um
2 4 6 8 10
Post-exposure Time, Hours
12
Figure 7A-5. Predicted clearance curves for the TB region for poorly soluble particles for
(a) human and (b) rat. Note different time scales. The rat clears PM from
the TB region much faster than a human. Fraction of mass retained in the
TB region after 1 h of exposure to unit density particles of diameter shown.
Adapted from Hofmann and Asgharian (2003).
7A-20
-------
o
s
LJ_
-o 0.1 +
0)
Dog
— — — Guinea Pig
Man
— Hamster
Rat
0.01-1 1 1 1 1 1
0 100 200 300 400 500 600
Days After Inhalation
Figure 7A-6. Alveolar region clearance curves for measured poorly soluble particles in
several species. Note much higher clearance rate for rat compared to
human. From Oberdorster (1988).
dose; whereas, in the case of the human, the maximum retained TB dose reaches as high as
0.30 of the total deposited dose. Figure 7A-7b shows a similar plot for the A region. As shown
in Figure 7A-7b, clearance is slower, and retention is greater, in the A region than the TB region
for both rats and humans. However, retention in the rat is less than in the human due to the
faster clearance in the rat.
7A.4.3.4 Long-Term Burden from Chronic Exposure
PM contains components with various degrees of solubility. Some components of PM
deposited in the lung dissolve in seconds to minutes, and others within hours to days. However,
there are some PM components that are sufficiently insoluble that they remain in the lung for
months to years. If the exposure concentration, breathing rate, tidal volume, and any other
dosimetric variables remained constant, then the processes of clearance and removal would
eventually approach an equilibrium; and the amount of insoluble PM in the lung would approach
a steady-state value. In reality, the exposure concentration and dosimetric parameters will vary
7A-21
-------
-------
Rats are usually kept in a laboratory setting and breathe air that has been filtered and
conditioned and are, therefore, exposed to relatively clean air for the months prior to their
experimental exposure. In addition, rat exposures usually have a daily schedule of 6 h exposure
to an experimental atmosphere followed by 18 h exposure to relatively clean air for 5 days a
week. On the other hand, people are exposed to ambient and nonambient PM all their lives.
Because of its more rapid clearance rate, a rat will reach a near steady state retained dose of
poorly soluble particles in the A region in a few months; it will take more than 10 years for a
human to do so. Figure 7A-8 shows the accumulation of PM in the lung for chronic exposures
for a rat and a human. Exposure parameters and particle sizes used in the MPPD model
calculations and the calculated alveolar deposition fractions are given in Table 7A-3.
7A.4.4 Dose Metrics
For inhalation toxicology, several parameters are required to define a dose metric: a PM
indicator, a respiratory region, the time over which the dose is integrated, whether the dose is
deposited or retained, and whether the dose is incremental or accumulated. Thus, there are many
possible dose metrics. It is not clear which dose metric is most appropriate and it may be that
different health effects will be associated with different dose metrics. For example, for health
effects associated with soluble PM components, mass may be the most appropriate PM indicator
and deposited mass more appropriate than retained mass. For health effects associated with
poorly soluble ultrafme PM, the particle number or particle surface area might be the more
appropriate PM indicator and the retained dose more appropriate than the deposited dose.
For acute effects, the maximum deposited incremental dose may be the appropriate type of dose
metric. For chronic effects, the total, retained, long-term burden may be more appropriate.
For health effects associated with the rupture or inactivation of macrophages, the volume of
particles might be an appropriate PM indicator and either total retained incremental dose or long-
term burden the appropriate type of dose. Some possible parameters are listed in Table 7A-4.
7A-23
-------
-~ 10
BJ
E
I! 6
_re
o
o
> 4
CQ
a. Human
m
4 6
Time, years
10
0.1
0.2 0.3
Time, years
0.4
0.5
Figure 7A-8. Mass burden of poorly soluble PM predicted for the A region of (a) human
and (b) rat. Burdens were calculated by the MPPD model for the exposure
scenarios provided in Table 3. The rat reaches a near steady-state burden
in 6 months. After 10 years the human is approaching a steady-state
burden that is a 1,000 times larger (or approximately 5 times larger when
normalized to lung surface area or body mass). Note different time of scales
(x-axis) and \ig versus mg units (y-axis).
7A.4.5 Normalizing Factors and Other Differences Between Humans
and Rats
The human and rat doses may be scaled by a normalizing parameter to better quantify dose
to specific target sites of the respiratory tract. If epithelial cells are the target, the
tracheobronchial or alveolar surface area would be the most likely normalizing parameter. If the
interstitium is the target, then the lung mass or weight may be better parameters. If activation of
7A-24
-------
TABLE 7A-3. EXPOSURE SCENARIOS FOR ACCUMULATION OF
LONG-TERM BURDEN USED BY EPA IN MPPD MODEL CALCULATIONS
Exposure
Hours a day
Days a week
Total time
Concentration of insoluble PM
Particle Size (MMAD)
Geometric Standard Deviation (og)
Density, g/mL
Breathing pattern
Breaths per min
Tidal volume, mL
Alveolar Deposition Fraction a
Alveolar Surface Dose b, mg/m2
Human
24
7
10 years
10 (ig/m3
1 (im
1
1
Resting
12
625
0.0993
0.025 c
0.17d
Rat
6
5
6 months
10 (ig/m3
1 (im
1
1
Resting
102
2.1
0.0593
0.0039 c
a Calculated with MPPD model.
b Alveolar burdens from Figure 7A-8 and alveolar surface areas from Table 7A-5.
0 After 6 months of exposure.
d After 10 years of exposure.
TABLE 7A-4. PARAMETERS USED TO DEFINE A DOSE METRICa
PM Indicator
Respiratory Region
Type of Dose
1 Number, surface area, mass, or volume; total PM or of a specific
PM component
2 Nasal, tracheobronchial (TB), alveolar (A), thoracic (total lower
respiratory tract, TB + A), specific TB generation, alveolus, macrophage
or other target cells
3 Total, average, or maximum
4 Deposited or retained
5 Incremental dose (over and above long-term burden) or incremental dose
plus accumulated, long-term burden
a One parameter is chosen from each of the five rows to form a dose metric.
7A-25
-------
macrophages is a causal process, then the number of macrophages would be an appropriate
normalizing parameter. Respiratory parameters for the human and rat that may be used as
normalizing factors are shown in Table 7A-5.
TABLE 7A-5. CHARACTERISTICS OF HUMAN AND RAT LUNGS
Functional Residual Capacity, FRC, ml
Body Mass, g
Lung Mass, g
TB Area, m2
A Area, m2
Human
3300a
73000
1100b
0.4419d
57.22d
Rat
4.0a
330
1.65C
0.002346 e
0.2972 e
Human/Rat
825
221
667
188
193
1 Winter-Sorkina and Cassee (2002), bU.S. EPA (1996),c Takezawa (1980) for a 330g rat, dYeh and Schum
(1980) scaled to FRC,e Yeh et al. (1979) scaled to FRC.
The volume and surface area of the lung are variable. Across various species, lung volume
varies with body weight (Weibel, 1972). For humans, lung volume is also a function of age and
height (Morris et al., 1984). In addition, the volume and the surface area change as the lung
expands and contracts during breathing. The functional residual capacity (FRC), defined as the
volume of the lung at the end of a normal expiration during rest, has been chosen as a
normalizing parameter. The values of FRC chosen are the default values used in the MPPD
model (Winter-Sorkina and Cassee, 2002). The value of 3300 ml for an adult human male is
also the default value for the ICRP model (ICRP, 1994). The value of 4 ml for a rat is also the
predicted value for a 330 g rat based on data from Takezawa (1980).
Lung surface areas at the FRC were calculated from whole lung anatomic models for a
human (Yeh and Schum, 1980) and a rat (Yeh et al., 1979). These models are based on
morphometric measurements of silicone rubber casts of the tracheobronchial airways of one
human and one rat lung. The models give the number, diameter and length of the airways for
7A-26
-------
generations 1-24 and the diameter, length (height of a spherical segment) and total number of
alveoli. Values for generations distal to the terminal bronchioles were derived by the authors of
the references from assumptions and formulae and did not represent direct estimates. The
dimensions given correspond to the total lung capacity since the casts were made under a slight
positive pressure. Both papers give volume of the ducts and the total number and volume of the
alveoli but do not distribute the alveoli to the respiratory bronchioles. For the human, the
number of alveoli on each respiratory bronchiole was taken from Weibel (1963). For the rat, the
alveoli were distributed to the respiratory bronchiole in proportion to the surface of the
respiratory bronchioles. The airway dimensions were then scaled to give the specified FRC and
the corresponding surface areas were calculated from the linear dimensions treating the alveoli
as segments of a sphere. (The surface areas given in the references are for the cross-sectional
areas of the tubes.) A summary of the dosimetric differences between humans and rats is given
in Table 7A-6.
7A.5 DOSIMETRIC CALCULATION FOR EXTRAPOLATION
MODELING: COMPARING RATS TO HUMANS
7A.5.1 General Exposure Scenarios
7A.5.1.1 Acute Exposures
For the first series of extrapolation modeling, an acute exposure of 6-h in duration for
humans and rats is examined. Only an incremental dose is considered, ignoring the burden of
PM preexisting in the lung at the time of exposure. Typical of experimental exposure
conditions, a resting activity level is used for rats. For humans, three levels of activity are used:
resting, light exertion, and moderate exertion. For humans, oronasal (normal augmentor) and
oral breathing are considered. Breathing parameters are given in Table 7A-1. For the human
exposure, a near-roadway situation with exposure to all three atmospheric modes is used. Dose
is calculated for each mode separately and for all three modes together. For the rat, exposure is
considered to each of the three atmospheric modes separately. Exposures to resuspended
collected particles, e.g., residual oil fly ash (ROFA) or ambient particles collected on a filter,
7A-27
-------
TABLE 7A-6. DOSIMETRIC DIFFERENCES BETWEEN RATS AND HUMANS
Differences In: Rats (Experimental Exposures)
Humans (Mainly Ambient Exposure)
Anatomy Nasal breathers
Monopodial branching lung structure
Exertion Level Usually resting during exposure
Clearance Fasta
Prior Exposure Usually kept in clean or relatively clean air
in laboratory setting; only a few months of
low exposure prior to test exposure
PM Burden Retained dose approaches steady state after
several months, and at a lower fraction of
deposited dose than for a human
PM Size Experimental challenge exposures mostly
Distribution to particles of limited size distribution.
Representative size distributions:
Resuspended PM: MMD = 1.2 - 2.5 urn,
og=1.5-2.5
Diesel exhaust: < 0.2 um
CAPs: usually only the 0.1 to 2.5 um
size range is concentrated
Oronasal breathers
Dichotomous branching lung structure
Exposure occurs over a range from sleep
to heavy exercise or work
Slow
Mature or elderly humans likely will have
accumulated larger burdens of PM from prior
exposures than will have laboratory rats, on a
normalized basis
On the order of 10 years required for the
retained dose to approach steady state
Ambient exposure to all three atmospheric
modes:
Aitken (0.01-0.1 um), ag= 1.6-1.7
Accumulation (0.1-1 um), ag = 1.6-2.2
Coarse (1-100 um), ag = 1.8-2.4
Experimental CAPs exposures usually to
one mode.
1 Alveolar clearance rates may be a function of retained dose.
impactor, or electronic-air-cleaner plate are also considered for rats. The size distribution and
the fraction of particles in each mode used in the model simulations are given in Table 7A-2.
Doses were calculated with the MPPD model (described in Section 7A.3). Normalized doses
were calculated using several normalizing factors with the values given in Table 7A-5.
Three sets of comparisons are given:
(1) Rat and human each exposed to the same single mode (Aitken, accumulation, or
coarse);
(2) Rat exposed to resuspended PM, human exposed to all three modes; and
(3) Rat exposed to each of the three modes separately, human exposed to all three
modes together.
7A-28
-------
The concept of an equivalent exposure ratio (EqER) was discussed in 7A.2. If exposure
times are the same, the rat exposure concentration that will give a dose equivalent to that
received by a human at a specified concentration can be determined by multiplying the specified
human concentration by EqER, i.e.,
r1 SlFaFR ipC f7A-i(Vi
\_^ T? fr^sjsy—'VI i—'-L V y>v\\A-^ TT I ' -^*- A \J I
9^ mf
In Tables 7A-7a to 7A-9b, values of EqER are reported for some of the dose metrics
listed in Table 7A-4. For example, EqER x 100 will yield the rat exposure concentration
necessary to produce a dose equivalent to that received by a human at an exposure concentration
of 100 |ig/m3. For clarity, if EqER is greater than 1, the rat must be exposed to a higher
concentration than the human for effectively equivalent doses.
It may be worthwhile here to emphasize that dosimetry refers to the deposition of
particles in the lung and the removal of deposited particles by a variety of mechanisms. The
relationship between exposure and dose for different species, lung sizes, and breathing patterns
as well as for a variety of dose metrics and normalizing factors can be characterized. However,
this dosimetric analysis provides limited insight into the relative toxicity of particles on a
composition or size basis. Toxicologic studies suggest that particle toxicity varies with
composition and with size (for particles of the same composition). Particle composition also
varies between fine and coarse modes and to a lesser extent within modes. Thus, it is difficult to
interpret relative doses in terms of relative toxicity unless the exposures are to particles with
identical composition and size. This is especially a problem for coarse mode particles, because
differences in inhalability between humans and rats result in greater deposition of the large
coarse mode particles in the human than the rat. The subsequent tables report equivalent doses,
but equivalent dose does not necessarily mean equivalent toxicity. Dosimetric calculations can
be used to predict exposures to give equivalent doses of the same material for the purpose of
comparing responses across species. If adequate information were available on size and
composition of particles in the exposure mix, and toxicity as a function of size and composition
were known, then dosimetry could be used to calculate doses of equivalent toxicity.
7A-29
-------
7A.5.1.2 Rat and Human Each Exposed to One Mode of the Atmospheric Particle
Size Distribution
Tables 7A-7a and 7A-7b give results, in terms of EqER, for a series of simulations in
which the rat normalized dose due to exposure to a mode of the atmospheric particle distribution
was compared to a human normalized dose due to exposure to the same single mode. The
specific particle size and breathing parameters are given in Tables 7A-1 and 7A-2. The rat was
assumed to be resting, the usual condition for experimental exposures. Simulations were run for
four human breathing patterns: resting, light exertion, moderate exertion (normal augmentor),
and moderate exertion (oral breathing). Normalized doses to a specific mode (Aitken [At],
accumulation [Ac], and coarse [C] mode particles) were compared over one 6-h exposure period
for a variety of dose metrics based on particle mass, surface area, and number for several
normalizing parameters. Values of EqER for deposited mass per lung mass, body mass, or lung
area range from 0.09 to 5.5 (Table 7A-7a Section I). This means that to provide a normalized
dose to a rat equivalent to that a human would receive at an exposure of 100 |ig/m3, depending
on the dose metric chosen, the predicted EqCR would range from 9 |ig/m3 (TH deposition per
lung mass for a resting human for Aitken particles) to 550 |ig/m3 (TB deposition per unit TB
area for a human undergoing moderate exertion for coarse particles). For short-term retention in
the TB region, EqER values are higher, 0.67 to 33, because of the more rapid clearance of PM
from the rat TB region. For short-term retention in the A region, EqER values are lower,
0.06 to 4.05.
Dose metrics based on surface area or number are somewhat different from those based
on mass due to changes in DF since the median diameter decreases in going from mass to
surface area to number. For surface area dose metrics (Table 7A-7b, Sections IV to VI), EqER
values range from 0.09 to 4.6 for deposited dose metrics; 1.7 to 47 for short-term (6- or 24-h)
retention in TB regions; and 0.16 to 3.7 for short-term retention in the A region. For particle
number dose metrics (Table 7A-7b, Sections VII to IX), the EqER range is 0.09 to 2.1 for
deposited dose metrics, 1.4 to 15 for short-term retention in the TB region, and 0.14 to 1.8 for
short-term retention in the A region. The MPPD model has a lower particle size limit of
0.01 |im. Hence, it could not calculate the DF for the count distribution of the Aitken mode with
a a of 1.7, because approximately 30% of the particles are below 0.01. Therefore, the EqER for
7A-30
-------
TABLE 7A-7a. PREDICTED PARTICLE MASS DOSE METRICS FOR HUMAN AND RAT EACH EXPOSED TO ONE
ATMOSPHERIC MODE: EQUIVALENT EXPOSURE RATIO (EqER) FOR PM DOSE AFTER 6-HOUR EXPOSURE
FOR SEVERAL BREATHING PATTERNS.*
I. Deposited Mass
TH per Lung Mass, ng/g
TH per Body Mass, |ig/g
TH per Lung Area, ng/m2
TB per TB Area, |ig/m2
A per A Area, |ig/m2
Hg per Macrophage
II. Retained Mass in TB
j> 6-h Avg per lung mass, ng/g
oo
^ 24-h Avg per lung mass, ng/g
6-h Avg per body mass, |ig/g
24-h Avg per body mass, |ig/g
6-h Avg per TB area, |ig/m2
24-h Avg per TB area, |ig/m2
///. Retained Mass in A
Maximum per A Area
6-h Avg per lung mass, ng/g
24-h Avg per lung mass, ng/g
6-h Avg per body mass, jig/g
24-h Avg per body mass, ng/g
6-h Avg per A area, |ig/m2
24-h Avg per A area, |ig/m2
At
0.088
0.21
0.24
0.45
0.17
0.16
0.7
12
1.6
28
1.8
33
0.16
0.060
0.061
0.14
0.14
0.16
0.16
Resting
Ac
0.1
0.23
0.26
0.53
0.20
0.19
0.7
1.8
1.6
4.2
1.9
4.9
0.19
0.071
0.072
0.17
0.17
0.19
0.19
C
0.18
0.42
0.48
0.42
0.54
0.52
0.6
1.9
1.5
4.5
1.8
5.3
0.54
0.20
0.20
0.47
0.48
0.54
0.55
At
0.23
0.55
0.63
0.90
0.54
0.52
1.3
3.8
3.1
9.0
3.7
11
0.54
0.20
0.20
0.47
0.47
0.54
0.54
Light
Ac
0.23
0.53
0.61
1.2
0.47
0.46
1.4
3.5
3.4
8.1
4.0
9.5
0.46
0.17
0.17
0.40
0.41
0.46
0.47
C
0.24
0.56
0.64
0.42
0.83
0.81
0.5
1.6
1.3
3.8
1.5
4.4
0.83
0.31
0.31
0.72
0.74
0.83
0.84
Moderate,
At
0.49
1.1
1.3
1.6
1.2
1.2
2.4
6.7
5.6
16
6.5
19
1.2
0.45
0.45
1.0
1.1
1.2
1.2
Normal Augmentor
Ac
0.44
1.0
1.2
2.6
0.86
0.83
2.7
6.3
6.4
15
7.5
17
0.84
0.31
0.32
0.73
0.74
0.84
0.85
C
1.2
2.9
3.3
4.1
2.8
2.7
2.3
5.3
5.5
13
6.5
15
2.7
1.0
1.0
2.4
2.4
2.7
2.8
Moderate, Oral Breathing
At
0.47
1.1
1.3
1.6
1.2
1.2
2.3
6.5
5.4
15
6.4
18
1.2
0.44
0.44
1.0
1.0
1.2
1.2
Ac
0.42
0.98
1.1
2.4
0.82
0.79
2.6
6.0
6.1
14
7.1
16
0.80
0.30
0.30
0.69
0.70
0.80
0.81
C
1.7
4.1
4.7
5.5
4.0
3.9
3.4
7.7
7.9
18
9.3
21
4.0
1.5
1.5
3.5
3.5
4.0
4
*At, Aitken mode; Ac, accumulation mode; C, coarse mode; TH, thoracic region; TB, tracheobronchial region; A, alveolar region; SA, particle surface area; #, particle number.
-------
TABLE 7A-7b. PREDICTED PARTICLE SURFACE AREA AND NUMBER DOSE METRICS FOR HUMAN AND RAT
EACH EXPOSED TO ONE ATMOSPHERIC MODE: EQUIVALENT EXPOSURE RATIO (EqER) FOR PM DOSE
AFTER 6-HOUR EXPOSURE FOR SEVERAL BREATHING PATTERNS.*
IV. Surface Area of Particles Deposited
TH per Lung Mass, SA /g
TH per Body Mass, SA /g
TH per Lung Area, SA /m2
TB per TB Area, SA /m2
A per A Area, SA /m2
SA per Macrophage
V. Surface Area of Particles Retained in TB
6-h Avg per TB area, SA /m2
24-h Avg per TB area, SA /m2
K^ VI. Surface Area of Particles Retained in A
00
to 6-h Avg per A area, SA /m
24-h Avg A per A area, SA /m2
VII. Number of Particles Deposited
TH per Lung Mass, # /g
TH per Body Mass, # /g
TH per Lung Area, # /m2
TB per TB Area, # /m2
A per A Area, # /m2
# per Macrophage
VIII. Number of Particles Retained in TB
6-h Avg per TB area, # /m2
24-h Avg per TB area, # /m2
IX. Number of Particles Retained in A
6-h Avg per A area, # /m2
24-h Avg per A area, # /m2
M
0.09
0.21
0.24
0.43
0.16
0.15
1.7
4.9
0.16
0.16
0.09
0.21
0.24
0.39
0.14
0.13
1.6
4.7
0.14
0.14
Resting
Ac
0.09
0.21
0.25
0.55
0.18
0.18
2.3
8.0
0.18
0.18
0.09
0.20
0.23
0.49
0.17
0.17
2.0
5.2
0.17
0.17
C
0.17
0.41
0.47
0.45
0.49
0.47
2.2
8.8
0.48
0.49
0.12
0.29
0.33
0.36
0.32
0.31
1.4
3.8
0.31
0.31
At
0.24
0.56
0.65
0.86
0.56
0.54
3.6
11
0.55
0.56
0.25
0.60
0.69
0.80
0.61
0.59
3.6
10.7
0.62
0.62
Light
Ac
0.22
0.52
0.59
1.20
0.46
0.45
5.1
16
0.46
0.46
0.22
0.51
0.59
1.00
0.49
0.48
3.8
10.3
0.49
0.49
Moderate, Normal Augmentor
C
0.26
0.60
0.69
0.53
0.82
0.79
2.4
8.7
0.82
0.82
0.23
0.53
0.61
0.65
0.60
0.58
2.1
5.6
0.59
0.60
At
0.52
1.2
1.4
1.5
1.3
0.87
4.3
13
0.88
0.89
0.63
1.5
1.7
0.9
2.3
1.1
4.3
13
1.1
1.1
Ac
0.42
0.99
1.1
2.4
0.87
0.56
4.9
12
0.57
0.58
0.43
1.0
1.2
1.8
1.0
0.65
4.3
12
0.66
0.67
C
1.1
2.5
2.9
3.3
2.6
1.7
5.2
13
1.7
1.8
0.53
1.2
1.4
1.8
1.3
0.82
3.3
8.1
0.84
0.84
Moderate, Oral Breathing
At
0.55
1.3
1.5
1.6
1.4
1.4
7.0
20
1.4
1.4
0.63
1.5
1.7
1.6
1.8
1.8
6.9
21
1.8
1.8
Ac
0.43
.1.0
1.2
2.5
0.88
0.85
9.8
29
0.87
0.88
0.44
1.0
1.2
1.9
1.0
1.0
6.7
18
1.0
1.0
C
1.5
3.5
4.0
4.6
3.7
3.6
16
47
3.6
3.7
0.64
1.5
1.7
2.1
1.6
1.5
5.8
15
1.5
1.5
*At, Aitken mode; Ac, accumulation mode; C, coarse mode; TH, thoracic region; TB, tracheobronchial region; A, alveolar region; SA, particle surface area; #, particle number.
-------
number distribution of the Aitken mode is based on monodisperse particles of 0.013 jim
diameter, the number mean diameter of the Aitken mode.
7A.5.1.3 Exposure to Resuspended Combustion Particles
Experimental studies with rats have typically used only one particle size range, either Aitken
mode particles (exposure to diesel or auto exhaust), accumulation mode particles (CAPs or some
acid aerosol exposure studies), or resuspended PM. Resuspended PM, regardless of its initial
size distribution, if passed through a 2.5 jim cyclone or impactor, will have a MMAD between
1 and 2 jim and a og between 1.5 and 2.5. One can ask if it is appropriate to compare the rat
dose, from only one of the PM size ranges, to the human dose from only that size range when the
human is exposed to the entire atmospheric aerosol. The answer to this question may be brought
into focus by asking what size particle should be used to calculate the human dose to compare
with rat exposures to resuspended combustion particles, such as the stationary source
combustion particles (e.g., ROFA) used in many EPA studies. It would not be appropriate to use
as a basis for the human dose, or for the equivalent human exposure, an exposure to resuspended
particles. People do not typically breath resuspended particles with a MMAD of 2 jim and a
og of 2. As shown in Figure 7A-1, resuspended particles have minimal surface area or particle
number compared to the PM that a human would be exposed to in an urban atmosphere. Thus,
if the health effect of interest were related to particle surface area or particle number, it would
require very high doses of a typical resuspended PM to achieve surface area or number doses
equivalent to those received by a human. Tables 7A-8a and 7A-8b report calculated values of
EqER for the comparison of a rat exposed to resuspended PM (MMAD = 2 jim, og = 2) relative
to a human exposed to all three modes of the atmospheric size distribution for four human
exposure scenarios.
For a comparison of a rat exposed to resuspended PM for 6 h to a human exposed to ambient
PM near a roadway for 6 h, the EqER for mass-based metrics (Table 7A-8a) have a smaller
range than for the comparison of individual modes: 0.13 to 2.7 for deposited mass, 0.54 to 16
for mass retained in the TB region, and 0.12 to 2.0 for mass retained in the A region. However,
for dose metrics based on surface area or number, EqER values are very high because
7A-33
-------
TABLE 7A-8a. PREDICTED PARTICLE MASS DOSE METRICS FOR RAT
EXPOSED TO RESUSPENDED PM (e.g., ROFA), HUMAN EXPOSED TO ALL
THREE ATMOSPHERIC MODES: EQUIVALENT EXPOSURE RATIO (EqER)
FOR PM DOSE AFTER A 6-HOUR EXPOSURE FOR SEVERAL BREATHING
PATTERNS.*
/. Deposited Mass
TH per Lung Mass, ug/g
TH per Body Mass, ug/g
TH per Lung Area, ug/m2
TB per TB Area, ug/m2
A per A Area, ug/m2
ug per Macrophage
II. Retained Mass in TB
6-h Avg per lung mass, ug/g
24-h Avg per lung mass, ug/g
6-h Avg per body mass, ug/g
24-h Avg per body mass, ug/g
6-h Avg per TB area, ug/m2
24-h Avg per TB area, ug/m2
///. Retained Mass in A
Maximum per A Area
6-h Avg per lung mass, ug/g
24-h Avg per lung mass, ug/g
6-h Avg per body mass, ug/g
24-h Avg per body mass, ug/g
6-h Avg per A area, ug/m2
24-h Avg per A area, ug/m2
Resting
0.13
0.29
0.34
0.35
0.33
0.32
0.54
3.6
1.3
8.4
1.5
9.9
0.33
0.12
0.12
0.29
0.29
0.33
0.33
Light
0.25
0.59
0.67
0.61
0.73
0.70
0.84
2.4
2.0
5.6
2.3
6.6
0.72
0.27
0.27
0.62
0.63
0.72
0.73
Moderate,
Normal
Augmentor
0.72
1.7
1.9
2.3
1.7
1.7
2.0
5.1
4.7
12
5.5
14
1.7
0.63
0.64
1.5
1.5
1.7
1.7
Moderate Oral
Breathing
0.84
2.0
2.3
2.7
2.0
1.9
2.3
5.7
5.4
13
6.3
16
2.0
0.73
0.74
1.7
1.7
2.0
2.0
* At, Aitken mode; Ac, accumulation mode; C, coarse mode; TH, thoracic region; TB, tracheobronchial region;
A, alveolar region; SA, particle surface area; #, particle number.
7A-34
-------
TABLE 7A-8b. PREDICTED PARTICLE SURFACE AREA AND NUMBER DOSE
METRICS FOR RAT EXPOSED TO RESUSPENDED PM (e.g., ROFA), HUMAN
EXPOSED TO ALL THREE ATMOSPHERIC MODES: EQUIVALENT EXPOSURE
RATIO (EqER) FOR PM DOSE AFTER A 6-HOUR EXPOSURE FOR SEVERAL
BREATHING PATTERNS.*
IV. Surface Area of Particles Deposited
TH per Lung Mass, SA/g
TH per Body Mass, SA/g
TH per Lung Area, SA/m2
TB per TB Area, SA/m2
A per A, SA/m2
SA per Macrophage
V. Surface Area of Particles Retained In TB
6-h Avg per TB area, SA/m2
24-h Avg per TB area, SA/m2
VI. Surface Area of Particles Retained in A
6-h Avg per A area, SA/m2
24-h Avg A per A area, SA/m2
VII. Number of Particles Deposited
TH per Lung Mass, # /g
TH per Body Mass, # /g
TH per Lung Area, # /m2
TB per TB Area, # /m2
A per A Area, # /m2
# per Macrophage
VIII. Number of Particles Retained in TB
6-h Avg per TB area, # /m2
24-h Avg per TB area, # /m2
IX. Number of Particles Retained in A
6-h Avg per A area, # /m2
24-h Avg per A area, # /m2
Resting
1.7
4.0
4.6
6.7
3.5
27
31
94
3.4
3.4
2.6E + 03
6.1E + 03
7.0E + 03
2.0E + 04
3.0E + 03
2.9E + 03
9.0E + 04
2.3E + 05
2.9E + 03
2.9E + 03
Light
4.4
10
12
14
11
87
65
198
11
11
7.3E + 03
1.7E + 04
2.0E + 04
4. IE + 04
1.3E + 04
1.3E + 04
2.0E + 05
5.3E + 05
1.3E + 04
1.3E + 04
Moderate,
Normal
Augmentor
9.4
22
25
26
25
132
74
218
17
17
1.8E + 04
4.2E + 04
4.8E + 04
4.5E + 04
5.0E + 04
2.4E + 04
2.3E + 05
6.4E + 05
2.4E + 04
2.3E + 04
Moderate, Oral
Breathing
9.9
23
27
27
27
208
125
381
26
27
1.8E + 04
4.2E + 04
4.9E + 04
8.0E + 04
3.9E + 04
3.8E + 04
3.8E + 05
LIE + 06
3.9E + 04
3.9E + 04
* At, Aitken mode; Ac, accumulation mode; C, coarse mode; TH, thoracic region; TB, tracheobronchial region; A,
alveolar region; SA, particle surface area; #, particle number.
7A-35
-------
resuspended PM is lacking in smaller particles. Thus, for particle surface area-based dose
metrics (Table 7A-8b, Sections IV to VI), EqER values range from 1.3 to 380. For particle
number-based dose metrics (Table 7A-8b, Sections VII to IX), EqER range from 1,100 to
1,100,000.
7A.5.1.4 Rat Exposed to One Fraction, Human Exposed to All Three Modes of the
Atmospheric Particle Size Distribution
As suggested in 7A.5.1.2, it may not be appropriate to compare a rat dose from one particle
size fraction to a human dose from the same size fraction (as was reported in Tables 7A-7a and
7A-7b) because humans are exposed to the full range of particle sizes. Tables 7A-9a and 7A-9b
show EqER values derived from normalized doses calculated from the combined exposure to all
three particles size fractions for humans whereas rats were considered to be exposed to only one
of the three size fractions in a given individual study. Again, a wide range of EqER values is
found: from 0.03 to 4.1 for deposited mass, from 0.19 to 24 for retained mass in the TB region,
and from 0.03 to 3.9 for retained mass in the A region. For particle surface area- and particle
number-based dose metrics the ranges for EqER values are very high, 0.008 to 1,300 for surface
area and 0.01 to 1.3 x 107 for number.
7A.5.1.5 Discussion
Several conclusions can be drawn from the above comparisons. For a specific dose metric,
it is possible to calculate a rat exposure that will give a dose equivalent to that received by a
human or to calculate a human exposure that would give a dose equivalent to that received by a
rat. However, the doses will be equivalent only for the specific dose metric chosen; for other
dose metrics, the doses may not be equivalent. If the rat and human are exposed to particles with
the same size distributions, or if the comparisons are based only on deposited mass, the
variations among dose metrics would be moderate. However, if dose metrics include particle
surface area or number, the variations among doses between rats and humans, other than for the
specified dose metric, may be very large.
It is especially difficult to determine a human exposure to ambient PM that would yield a
human dose equivalent to a rat dose due to exposure to resuspended PM. Toxicity may depend
7A-36
-------
TABLE 7A-9a. PREDICTED PARTICLE MASS DOSE METRICS FOR RAT EXPOSED TO ONE MODE AT A TIME,
HUMAN EXPOSED TO ALL THREE ATMOSPHERIC MODES: EQUIVALENT EXPOSURE RATIO (EqER) FOR PM
DOSE AFTER A 6-HOUR EXPOSURE FOR SEVERAL BREATHING PATTERNS.*
/. Deposited Mass
TH per Lung Mass, ug/g
TH per Body Mass, ug/g
TH per Lung Area, ug/m2
TB per TB Area, ug/m2
A per A Area, ug/m2
ug Mass per Macrophage
//. Retained Mass in TB
6-h Avg per lung mass, ug/g
24-h Avg per lung mass, ug/g
6-h Avg per body mass, ug/g
24-h Avg per body mass, ug/g
6-h Avg per TB area, ug/m2
24-h Avg per TB area, ug/m2
///. Retained Mass in A
Maximum per A Area
6-h Avg per lung mass, ug/g
24-h Avg per lung mass, ug/g
6-h Avg per body mass, ug/g
24-h Avg per body mass, ug/g
6-h Avg per A area, ug/m2
24-h Avg per A area, ug/m2
At
0.033
0.078
0.089
0.14
0.071
0.069
0.19
1.2
0.45
2.8
0.52
3.3
0.071
0.026
0.026
0.062
0.062
0.071
0.071
Resting
Ac
0.10
0.24
0.27
0.57
0.20
0.19
0.82
4.8
1.9
11
2.2
13
0.20
0.073
0.074
0.17
0.17
0.20
0.20
C
0.22
0.51
0.59
0.54
0.64
0.62
0.81
5.6
1.9
13
2.2
15
0.64
0.24
0.24
0.56
0.56
0.64
0.65
At
0.07
0.15
0.18
0.25
0.16
0.15
0.30
0.80
0.70
1.9
0.82
2.2
0.15
0.057
0.057
0.13
0.13
0.15
0.16
Light
Ac
0.20
0.47
0.54
0.98
0.44
0.42
1.3
3.2
3.0
7.6
3.5
8.9
0.43
0.16
0.16
0.37
0.38
0.43
0.43
Moderate, Normal Augmentor
C
0.43
1.0
1.2
0.92
1.4
1.4
1.3
3.7
3.0
8.7
3.5
10
1.4
0.51
0.52
1.2
1.2
1.4
1.4
At
0.19
0.44
0.51
0.93
0.37
0.36
0.71
1.70
1.66
4.00
1.95
4.69
0.36
0.13
0.14
0.32
0.32
0.36
0.37
Ac
0.58
1.4
1.6
3.7
1.0
1.0
3.0
6.8
7.1
16
8.4
19
1.0
0.37
0.38
0.88
0.89
1.01
1.02
C
1.2
2.9
3.4
3.5
3.3
3.2
3.0
7.9
7.1
19
8.4
22
3.3
1.2
1.2
2.9
2.9
3.3
3.3
Moderate, Oral Breathing
At
0.22
0.52
0.60
1.1
0.43
0.42
0.8
1.9
1.9
4.5
2.2
5.3
0.42
0.16
0.16
0.37
0.37
0.42
0.43
Ac
0.68
1.6
1.8
4.4
1.2
1.2
3.5
7.7
8.2
18
9.6
21
1.2
0.44
0.44
1.0
1.0
1.2
1.2
C
1.5
3.4
3.9
4.1
3.9
3.7
3.5
8.9
8.1
21
9.5
24
3.8
1.4
1.4
3.3
3.4
3.8
3.9
* At, Aitken mode; Ac, accumulation mode; C, coarse mode; TH, thoracic region; TB, tracheobronchial region; A, alveolar region; SA, particle surface area; #, particle number.
-------
TABLE 7A-9b. PREDICTED PARTICLE SURFACE AREA AND NUMBER DOSE METRICS FOR RAT EXPOSED TO
ONE MODE AT A TIME, HUMAN EXPOSED TO ALL THREE ATMOSPHERIC MODES: EQUIVALENT EXPOSURE
RATIO, EqER, FOR PM DOSE AFTER A 6-HOUR EXPOSURE FOR SEVERAL BREATHING PATTERNS.*
IV. Surface Area of Particles Deposited
TH per Lung Mass, SA/g
TH per Body Mass, SA/g
TH per Lung Area, SA/m2
TB per TB Area, SA/m2
A per A Area, SA/m2
SA per Macrophage
V. Surface Area of Particles Retained in TB
6-h Avg per TB area, SA/m2
24-h Avg per TB area, SA/m2
^
*7 VI- Surface Area of Particles Retained in A
oo
6-h Avg per A area, SA/m2
24-h Avg A per A area, SA/m2
VII. Number of Particles Deposited
TH per Lung Mass, # /g
(TH per Body Mass, # /g
TH per Lung Area, # /m2
TB per TB Area, # /m2
A per A Area, # /m2
per Macrophage
VIII. Number of Particles Retained in TB
6-h Avg per TB area, # /m2
24-h Avg per TB area, # /m2
IX. Number of Particles Retained in A
6-h Avg per A area, # /m2
24-h Avg per A area, # /m2
At
0.008
0.019
0.021
0.035
0.015
0.12
0.052
0.16
0.015
0.015
0.006
0.014
0.016
0.026
0.009
0.009
0.040
0.12
0.009
0.009
Resting
Ac
0.17
0.40
0.46
1.30
0.28
2.2
2
5.3
0.27
0.27
3.2
7.5
8.6
29
3.5
3
42
107
3.4
3.5
C
7
16
18
23
15
120
38
120
15
15
4.4E + 04
l.OE + 05
1.2E + 05
2.5E + 05
5.7E + 04
5.5E + 04
4.0E + 05
1.1E + 06
5.6E + 04
5.7E + 04
At
0.021
0.048
0.056
0.072
0.049
0.38
0.11
0.33
0.048
0.049
0.017
0.040
0.046
0.054
0.041
0.040
0.086
0.26
0.042
0.042
Light
Ac
0.44
1.04
1.2
2.7
0.88
6.9
4.1
11
0.87
0.88
9.0
21
24
60
15
15
91
240
16
16
Moderate, Normal Augmentor
C
18
42
48
47
49
390
81
250
49
50
1.2E + 05
2.9E + 05
3.3E + 05
5.0E + 05
2.5E + 05
2.5E + 05
8.7E + 05
2.5E + 06
2.5E + 05
2.6E + 05
At
0.04
0.10
0.12
0.13
0.11
0.58
0.13
0.37
0.07
0.07
0.042
0.099
0.11
0.06
0.16
0.07
0.10
0.32
0.076
0.073
Ac
0.93
2.2
2.5
4.9
2.0
11
4.7
12
1.3
1.3
22
52
60
66
59
28
110
290
28
27
C
37
88
100
87
110
590
92
280
74
75
3.0E + 05
7. IE + 05
8.2E + 05
5.5E + 05
9.6E + 05
4.6E + 05
l.OE + 06
3. IE + 06
4.6E + 05
4.4E + 05
Moderate, Oral Breathing
At
0.05
0.11
0.12
0.14
0.12
0.91
0.21
0.64
0.12
0.12
0.043
0.10
0.11
0.10
0.12
0.12
0.17
0.52
0.12
0.12
Ac
1.0
2.3
2.7
5.2
2.1
17
8.0
21.4
2.1
2.1
22
52
60
115
46
45
180
480
46
46
C
40
93
110
92
120
920
160
490
120
120
3. IE + 05
7.2E + 05
8.2E + 05
9.7E + 05
7.6E + 05
7.4E + 05
1.7E + 06
5.0E + 06
7.5E + 05
7.6E + 05
* At, Aitken mode; Ac, accumulation mode; C, coarse mode; TH, thoracic region; TB, tracheobronchial region; A, alveolar region; SA, particle surface area; #, particle number.
-------
on both particle size and composition. In addition, the rat has a much smaller deposition fraction
for coarse particles than a human, and particle composition (and probably toxicity) varies with
mode (and possibly with size within the coarse mode). Therefore, it is difficult to estimate what
tissue doses would yield comparable toxicity in rats and humans for exposures in which rats and
humans are exposed to particles with different size distributions or for exposures involving
coarse PM.
It is possible to use dosimetric models to predict the tissue doses for specific dose metrics.
However, it is difficult to determine what exposures would yield doses of comparable toxicity
since toxicity may vary with particle size and composition as well as with the distribution of
deposition within the lung.
7A.5.1.6 Rat-to-Human Extrapolation of Long-Term PM Burden in the Alveolar Region
As discussed in 7A.4.3.4, differences in clearance, and resulting differences in the long-
term burden of poorly soluble PM retained in the lungs of humans and rats, must be considered
in extrapolation of chronic exposures. In the MPPD model, the first order alveolar clearance rate
constants for the human (for removal of poorly soluble particles from the A region to the TB
region) are independent of PM burden (see Appendix Equations 7A-3 and 7A-12). This is
probably a reasonable assumption for humans exposed to ambient PM. However, the first order
alveolar clearance rate (for removal of poorly soluble particles from the A region to the TB
region) for a rat, exposed to PM levels well in excess of ambient PM, depends on the particle
burden in the alveolar region (see Appendix Equations 7A-3 and 7A-11). Thus, the fraction of
deposited PM mass retained in the alveolar region, as estimated by the MPPD model, does not
depend on the PM burden in humans whereas it increases with increasing exposure concentration
and PM burden for rats. This phenomenon of the first order rate constant for alveolar clearance
decreasing with increased loading is illustrated in Figure 7A-9 for the exposure parameters and
particle size given in Table 7A-3.
For rat to human extrapolation of chronic exposure, we have chosen the dose metric of
retained mass of poorly soluble PM per unit lung surface area. Dosimetric modeling enables us
to estimate the exposure scenario to yield a retained dose in the rat equivalent to a retained dose
7A-39
-------
1.00
Concentration (|jg/m3)
105
0.1
0.2 0.3 0.4
Time, years
0.5
Figure 7A-9. Mass of poorly soluble PM predicted to be retained in the alveolar (A)
region of the rat lung as a fraction of the total mass deposited in the A
region during a 6-month exposure. The MPPD model estimates illustrated
in this figure are for the exposure scenario in Table 3 with the exception of
exposure concentrations, which are provided in the figure.
in a human. As illustrated in Figure 7A-10, a rat would require an exposure of 60 |ig/m3 versus
a human exposure of only 10 |ig/m3 to have the same retained PM mass per unit alveolar area
after 6-months exposure. For shorter exposure times, the rat equivalent dose would be less than
60 |ig/m3.
Suppose that it is necessary to give a rat an exposure such that after 6 months the rat dose
(in mass of PM retained per unit alveolar surface area) is the same as a human's steady state
dose (0.15 mg/m2) reached only after about 10-years of exposure. Figure 7A-11 shows the
accumulation of PM burden per unit area in a rat for various exposure concentrations. In order
to better interpolate the rat 6-month exposure concentration needed to yield the burden in a
human at steady state, Figure 7 A-12 shows a log log plot of burden versus exposure
7A-40
-------
OJ C
Q- •Ł:
c o)
.2 ^
"5) to
0) 0)
(0 0)
11
< CO
c »-
.2 10
Rat Exposure
Concentration
(Mg/m3
0
0.1
0.2 0.3 0.4
Time, years
0.5
Figure 7A-10. Mass of poorly soluble PM predicted to be retained in the alveolar (A) lung
region normalized to A surface area. The MPPD model estimates
illustrated in this figure are for the exposure scenarios in Table 3 with the
exception of rat exposure concentrations, which are provided in the figure.
For the exposure concentration illustrated, the burden in a rat is predicted
to reach a plateau or equilibrium during a 6-month exposure, whereas the
burden in a human continues to increase montonically with exposure.
concentration and the equation for the regression line. This equation can be used to calculate the
rat-equivalent exposure concentration. The equivalent rat exposure concentration is 300 |ig/m3.
If one assumes a human exposure to 50 |ig/m3 total PM of which 20% is insoluble, the rat would
have to be exposed to the same PM at a concentration of 1,500 |ig/m3 for 6 months (6 h a day,
5 days a week) in order to receive a dose or burden equivalent to the near steady-state dose or
burden of a human after exposure to 50 |ig/m3 24 ha day, 7 days a week, for 10 years. The rat
would, of course, receive a much greater dose of soluble PM than the human.
The previous discussion applies to obtaining an equivalent lung burden at some instant in
time. However, if a health outcome is related to the lung burden that existed over some period
7A-41
-------
800
0
1000
Rat Exposure
Concentration
(|jg/m3)
750
Human
(10 years at 10 ug/m3)
0.1
0.2 0.3 0.4
Time, years
10
Figure 7A-11. Mass of poorly soluble PM predicted to be retained in the alveolar (A) lung
region normalized to A surface area. The MPPD model estimates
illustrated in this figure are for the exposure scenarios in Table 3 with the
exception of rat exposure concentrations, which are provided in the figure.
Due to decreasing clearance rates with increasing burden, it takes longer
for the rat to reach equilibrium at the higher exposure concentrations.
Illustrated for comparative purposes, humans are predicted to reach a
burden of 0.17 mg/m2 after 10 years of exposure to 10 ug/m3.
of time, rather than the recent dose, then instantaneous lung burden is less important than the
integral of burden over time. Comparing the areas under the rat and human curves in Figure
7A-10 clearly illustrates that the human and rat lung burdens integrated over time (similar to the
concept of C x T) are different. In order to mimic the time course of lung burden in a human, it
would be necessary to use a constantly, or frequently, changing rat exposure concentration.
While it would be possible to obtain an equivalent rat lung burden of very poorly soluble PM,
either at a point in time or integrated over time, in either case the rat would receive a much
higher dose of soluble PM.
7A-42
-------
~ 1000
o> E
10 100 1000 10000
Exposure Concentration (ug/m3)
100000
Figure 7A-12. Mass of poorly soluble PM predicted to be retained in the alveolar (A)
lung region normalized to A surface area. The MPPD model estimates
illustrated in this figure are for the exposure scenarios in Table 3 with
the exception of exposure concentrations, which are illustrated. On a
log-log plot, particle retention per A surface area in rats after a 6-month
exposure is a linear function concentration, i.e, log(Retention [mg/m2])
= 1.17 x log(Concentration [ug/m3]) - 3.56, R2 = 0.999.
7A.5.1.7 Caveats
The simulations are based on a model, and while the model uses similar deposition
calculations for humans and rats, the results of the simulations are only considered to be
estimates. The particles were assumed to have a density of 1 g/cm3, making the physical and
aerodynamic diameters the same. The calculations for the number dose of At mode particles
used a single size, 0.013 jam, rather than a distribution since the MPPD model does not go below
0.01 |im diameter in particle size. No consideration was given to the difference between human
PM exposures and ambient PM concentrations nor to exposures to indoor-generated or
occupational PM. Thus, while the results may not be quantitatively accurate, the general
relationships between human and rat exposure may provide useful information in the attempt to
understand rat to human PM dose extrapolation.
7A-43
-------
7A.6 HEALTH STATUS: A NON-DOSIMETRIC CONSIDERATION
Clearly, many host factors may come into play when considering response to PM. While
the mechanistic reasons for enhanced responsiveness are poorly understood, some specific host
attributes or health conditions seem to be contributory. Chronic conditions such as diabetes,
chronic heart or vascular disease, or chronic lung disease generally have been shown to lead to
increased susceptibility. It appears that existent lung conditions which may increase or alter the
deposition or retention of PM provide one means (i.e., dose) by which risk is augmented. The
very old and the very young may also be more susceptible due to underlying disease, impaired or
immature defenses, perhaps exacerbated or associated with other factors such as poor nutrition.
Rats normally have higher concentrations of some of the major endogenous antioxidants than
people (e.g., ascorbate), and, thereby, may be better able to resist the effects of reactive oxygen
species thought to be generated by or in response to PM. However, rats also are subject to
"overload," a condition in which sufficiently high doses of PM overwhelm both their clearance
and antioxidant defenses. Under these conditions the rat lung is highly sensitive to PM, and
fibrosis and tumor formation can occur.
7A.7 COMPARATIVE DOSIMETRY FOR SPECIFIC PUBLISHED
STUDY EXAMPLES
This section describes specific human and rat PM exposure studies. The section is divided
into three main parts: one examining exposures by intratracheal instillation, a second exposure
by inhalation, and a third discussing overload in rats. The MPPD model served as the primary
means of estimating regional deposition fractions and retained doses for comparisons. The first
part of this section considers Utah Valley Dust (UVD) instillation studies conducted in humans
by Ohio and Devlin (2001) and in rats by Dye et al. (2001). Under the premise that equal tissue
doses might produce similar across-species responses, instilled doses are compared across
species, and inhalation exposure scenarios leading to comparable tissue doses are presented.
The second part examines Concentrated Ambient Particles (CAPs) inhalation studies conducted
in humans by Ohio et al. (2000) and in rats by Kodavanti et al. (2000) and Clarke et al. (1999).
7A-44
-------
Across-species dose comparisons are made for the same exposure durations and concentrations
used in each of the studies. The final part of this section discusses Clearance Overload in Rats
and derives exposure concentrations predicted to achieve varied levels of alveolar loading in
subchronically and chronically exposed rats.
7A.7.1 Utah Valley Dust
Table 7A-10a provides assumed exposure scenarios and alveolar doses based on the Utah
Valley epidemiology study by Pope (1989) in the context of instillation studies conducted in
humans by Ohio and Devlin (2001) and in rats by Dye et al. (2001). The hypothetical exposure
scenarios are for humans and rats in the Utah Valley during an "Open-Plant" period (December
1985 to January 1986). On 13 occasions during those 2 months, the 24-h average PM10 values
exceeded 300 |ig/m3. The 2-month average PM10 was 120 |ig/m3 (Pope, 1989). In order to
compare instilled doses with a dose received by inhalation, it is necessary to assume a size
distribution of the UVD. For this region of the U.S., PM10 might typically be expected to be
about 50% PM2 5 by mass (Chapter 3). However, because the steel mill accounted for the
majority of PM10, it was assumed that PM25 was likely closer to 80% of the mass, such as in a
highly polluted industrial area (Pinto et al., 1998).
The activity patterns of the exposed humans and rats are also provided in Table 7A-10a.
People were presumed generally sedentary, spending 50% of their time at rest and 50% of their
time in an activity similar to a slow walk. Rats were assumed always at rest. Tidal volumes and
breathing frequencies associated with these activity levels were provided earlier in Table 7A-1.
Based on these exposure conditions, people are predicted to deposit between 176 jig (nasal
breather) and 222 jig (oral breather) in the A region of the lung on a daily basis, whereas rats are
predicted to deposit 2 jig. Only the alveolar region of the lung was considered for comparison to
the instilled doses, because most material depositing in the tracheobronchial airways is rapidly
cleared.
Ohio and Devlin (2001) tested the hypothesis that the soluble components (-20% of
particle mass on average for 1986 - 1987 UVD) of UVD might differ between years when the
Geneva Steel Mill was open (1986 and 1988) versus when it was closed (1987) and that these
7A-45
-------
TABLE 7A-10a. UTAH VALLEY DUST: EXPOSURE SCENARIO
Utah Valley Dust, ambient exposures (December 1985-January 1986)
- 120 (ig/m3 PM10 (2-month average)
Assumed characteristics of Utah Valley Dust
- 80% Fine mode (MMAD = 0.31 pm; og = 2.03)
- 20% Coarse mode (MMAD = 5.7 (im; og = 2.1)
Activity level and route of breathing
Human Rat
- 12 h rest, 12 h slow walk3 - 24 h rest
- nasal and oral breathing - nasal breathing
Predicted Daily Mass Depositing in A region
Human Rat
- 176 (ig (nasal breathing) - 2.0 (ig
- 222 (ig (oral breathing)
a These values represent the presumed average amount of time over the course a day that a person might spend
either at rest (sitting or sleeping) or engaged in an activity similar in exertion to a slow walk.
differences might affect biological response. In their study of 24 healthy adults, UVD extracts
(500 |ig) from either 1986 (n = 8), 1987 (n = 8), or 1988 (n = 8) were instilled into the lingula of
the lung. As a control, saline was instilled into a subsegment of the right middle lobe of each
study participant. Extracts of UVD were prepared by agitating filter samples in deionized water
for 24 h. Following centrifugation, supernatants were removed and lyophilized. The desired
amounts of the resulting dry but soluble extracts for each year were then placed in sterile saline
for instillations. The estimated surface dose of the instilled material is -170 jig per m2 of
alveolar surface area (see Table 7A-10b). At 24-h postinstillation and relative to a saline control,
lavage fluid from subjects instilled with the 1986 and 1988 extracts contained significantly
increased total cells, neutrophils, protein, fibronectin, albumin, and cytokines. The extracts
of UVD from 1987 (the year the steel mill was closed) did not elicit a response different from
the saline control. Considering the inflammatory response, neutrophil levels were increased
7A-46
-------
TABLE 7A-10b. UTAH VALLEY DUST: HUMAN INSTILLATION STUDY
Instilled Mass and Surface Dose
Human Rat (Similarly exposed}
- 500 (ig to lingula3 (Ohio and Devlin, 2001) - 50 (ig to entire lung
- 170 (ig/m2 (lingular surface dose) - 170 (ig/m2 (whole lung surface dose)
Predicted Time to Achieve Instilled Surface Dose by Inhalation (assuming no A clearance)^
Human Rat (Similarly exposed}
- 55 days (nasal breathing) - 25 days
- 44 days (oral breathing)
Predicted Time to Achieve Instilled Surface Dose by Inhalation (adjusted for A clearance)11
Human Rat (Similarly exposed}
- 65 days (nasal breathing) - 32 days
- 50 days (oral breathing)
a The lingula is the lower anterior portion of the left upper lobe and is the left lung's homologue of the right middle
lobe. The volume of lobes relative to total lung capacity is 15.4% for the left upper lobe, 15.4% for the right
upper lobe, and 7.7% for the right middle lobe (Yeh and Schum, 1980). Based on the ratio of right middle lobe
to right upper lobe volume, the lingula was assumed one-third the volume of the left upper lobe or 5.1% of total
lung volume and lung surface area.
b Exposure scenario provided in Table 7A-10a.
3.5- and 2.9-fold by the 1986 and 1988 UVD extracts, respectively, but only 1.2-fold by the
1987 UVD extract.
Ohio and Devlin (2001) provided an estimate of the time it might take for their instilled
dose to occur by inhalation. They assumed a hypothetical ambient UVD exposure level of PM10
(100 jig/m3). The computations described in their discussion were based on a total lung DF of
0.42. They concluded that the dose instilled (500 jig) into the lingula of human volunteers was
roughly comparable to the PM deposited as the result of living about 5 days in the Utah Valley.
Strictly speaking, the Ohio and Devlin (2001) analysis is flawed in that they only instilled the
soluble fraction of UVD (-20% of particle mass on average for 1986-1988 UVD), whereas their
estimates of dose by inhalation are based on total PM10, which contains both soluble and
insoluble components. For simplicity, the analysis presented here also considered PM10 as
7A-47
-------
insoluble. The results of this analysis are provided in Table 7A-10b. It was estimated that
between 44 and 65 days would be required for a person to deposit the instilled dose on the basis
of mass per surface area. A comparable surface dose would occur in a rat after a month of
exposure.
The human dose estimate provided in Table 7A-10b differs from that of Ohio and Devlin
(2001) for a number of reasons. First, their DF included the nasal, TB, and A regions of the
lung. In contrast, the estimate provided here considered only the A region and had an average
DF of only 0.1. Second, the lingula is only about 5% of total lung volume, whereas the authors
assumed the lingula represented 10% of lung volume. This difference effectively doubled the
estimated surface dose from the instillation. Based on the present analysis, it appears possible to
achieve the instilled surface dose at the relatively high ambient PM10 concentrations. However,
this instilled dose would be achieved only from a subchronic exposure and not in the acute
manner in which it was delivered by instillation. Considered from the perspective of a single
exposure day, the corresponding estimated 24-h average PM exposure would need to be between
5.2 mg/m3 (oral breather) and 6.6 mg/m3 (nasal breather) for humans and 3.0 mg/m3 for rats.
In the study by Dye et al. (2001), rats received intratracheal instillations of soluble extracts
from UVD collected in 1986, 1987, and 1988. UVD extracts were prepared by agitating filter
samples in deionized water for 96 h. Following centrifugation, supernatants were removed and
lyophilized. The desired amounts of the resulting dry but soluble extracts for each year were
then placed in sterile saline for instillations. The 1986 UVD extracts were instilled at the doses
of 250, 1000, and 2500 jig. Largely driven by an influx of neutrophils, the BAL fluid collected
at 24 h postinstillation showed a dose dependent increase in total cell counts (see Figure 4 in Dye
et al., 2001). Neutrophil cell counts (BAL fluid cell counts x 103/mL) were 105, 245, and 370
for the 250-, 1000-, and 2500-jig doses, respectively. These increases in neutrophils are 10-, 22-,
and 34-fold [for the doses of 250, 1000, and 2500 jig, respectively] relative to an average
neutrophil level of 11 (BAL fluid cell counts x 103/mL) in the saline controls (n = 22). The
1987 UVD (collected the year the Geneva Steel Mill was closed) extract instilled at the dose of
5000 jig only increased neutrophil levels to 61 (BAL fluid cell counts x 103/mL). These findings
are generally consistent with Ohio and Devlin (2001) in that the 1987 dust extracts were far less
potent producers of an inflammatory response relative to 1986 and 1988 extracts.
7A-48
-------
Considering the 250-jig dose instilled by Dye et al. (2001), the surface dose to the entire rat
lung was computed to be 840 jig per m2 alveolar surface area and used as the dose-equivalent
parameter for comparison to humans. These data appear in Table 7A-10c (1). By inhalation and
ignoring particle clearance, an 840 jig per m2 alveolar surface area dose of PM could occur in
124 days for rats and between 215 and 272 days for humans at an ambient PM concentration of
120 |ig/m3 (see Table 7A-10a for exposure scenarios). When clearance is considered, however,
a lung burden equal to the instilled dose is not achievable in rats by inhalation, given the
exposure conditions provided in Table 7A-10a. Other exposure conditions in which the rat
would receive the instilled dose are provided in Table 7A-10c (2). One finds that the rat
instillation of 250 jig corresponded to a single 24-h exposure by inhalation to a concentration of
15 mg/m3 in the rats or roughly double this concentration for humans. A 30-day (24 h per day)
exposure would still require PM concentrations of 0.6 mg/m3 in the rats or about 1 mg/m3
in humans.
The simulations presented here show that ambient PM10 exposure periods on the order of
months in the Utah Valley (winter 1986) would be required for poorly soluble PM to reach
alveolar surface doses equivalent to those instilled by Ohio and Devlin (2001) and by Dye et al.
(2001). For a sedentary person, 24 h exposures of at least 5.2 mg/m3 would be required to reach
the doses instilled by Ohio and Devlin (2001). For resting rats, a 24 h exposure to 15 mg/m3 was
predicted to achieve the lowest dose instilled by Dye et al. (2001). However, only the soluble
extracts of UVD (-20% of total PM) were instilled in these studies. Approximately 5-fold
greater exposures, i.e. 26 mg/m3 instead of 5.2 mg/m3 for humans, would be required to reach the
instilled surface doses of soluble extracts. It appears very unlikely that a person in the Utah
Valley could have received the instilled doses via inhalation in an acute ambient exposure. The
dose rates for instillation versus inhalation are also quite obviously different. Thus, the health
effects from acute ambient exposures to soluble PM are difficult to predict based on these
instillation studies. This, however, does not imply that these studies are without merit. In both
the human and rat study, the 1987 dust extracts were far less potent producers of an
inflammatory response relative to 1986 and 1988 extracts. Although the pulmonary doses used
in these instillation studies may not be achieved via inhalation, these studies suggest that PM
composition can affect response.
7A-49
-------
TABLE 7A-10c(l). UTAH VALLEY DUST: RAT INSTILLATION STUDY
Instilled Mass and Surface Dose
Rat Human (Similarly exposed}
- 250 (ig to whole lung (Dye et al., 2001) - 48,000 p.g to whole lung
- 840 p.g/m2 (whole lung surface dose) - 840 p.g/m2 (whole lung surface dose)
Predicted Time to Achieve Instilled Surface Dose by Inhalation (assuming no A clearance)"0
Rat Human (Similarly exposed}
- 124 days - 272 days (nasal breathing)
- 215 days (oral breathing)
Predicted Time to Achieve Instilled Surface Dose by Inhalation (adjusted for A clearance)"0
Rat Human (Similarly exposed}
- indefinite timea - 3.0 years (nasal breathing)
- 2.0 years (oral breathing)
a The equilibrium lung burden for the exposure conditions is only 160 \ig. After one year of exposure, the
burden is within 2.5% of this equilibrium.
b Exposure scenario provided in Table 7A-10a.
TABLE 7A-10c(2). UTAH VALLEY DUST: RAT INSTILLATION STUDY
EXPOSURE SCENARIOS ACHIEVING INSTILLED DOSE
Predicted 24-h Exposure Concentration to Achieve Instilled Surface Dose by Inhalation"
Rat Human (Similarly exposed}
- 15,000 p.g/m3 - 32,500 p.g/m3 (nasal breathing)
- 26,000 p.g/m3 (oral breathing)
Predicted 30-day Exposure Concentration to Achieve Instilled Surface Dose by Inhalationa
Rat Human (Similarly exposed}
- 590 p.g/m3 - 1,200 p.g/m3 (nasal breathing)
- 950 p.g/m3 (oral breathing)
1 With the exception of exposure concentrations, the exposure scenario is provided in Table 7A-10a.
7A-50
-------
7A.7.2 Concentrated Ambient Particles (CAPs)
In this section, tissue doses predicted to occur in a human and two rat CAPs exposure
studies are determined. Ohio et al. (2000) exposed healthy young adult human subjects (n = 38)
in Chapel Hill, NC to an average 120 |ig/m3 CAPs for 2 h. Table 7A-1 la provides the predicted
tissue doses to the subjects that participated in this study as well as the doses that would be
predicted to occur in rats for similar exposure conditions (time and concentration). For this
particle size and exposure conditions, the dose to the A region of the lung is quite similar
between species. This dose elicited a mild inflammatory response but did not affect the
pulmonary function of the exposed subjects.
TABLE 7A-lla. CAPs: HUMAN INHALATION STUDY (Ghio et al., 2000)
Human CAPs Rat
Ghio et al. (2000)a {Similarly exposed}
b
MMAD (|im)
og
Concentration ((ig/m3)
Deposited TB Dose per SA ° ((ig/m2)
Deposited A Dose per SA (|ig/m2)
0.65
2.35
120
64
0.7
0.65
2.35
120
37
0.78
a Two-hour protocol with 15-min periods of heavy exercise (VE = 50 L/min) followed by 15-min of recovery
(VE = 13 L/min) repeated four times. Subjects were presumed to breathe as normal oronasal augmenters.
b Rats were presumed exposed at rest.
c Surface area of lung region.
Kodavanti et al. (2000) exposed healthy (n = 5) and bronchitic (n = 4) rats in Research
Triangle Park, NC to 590 |ig/m3 CAPs, 6 h per day, for 3 days. Table 7A-1 Ib provides the
predicted tissue doses in the rats and predicted doses for similarly exposed humans. As a
control, healthy (n = 4) rats were exposed 6 h per day for 3 days to filtered air. At 18 h after the
third exposure, the CAPs-exposed rats showed no significant inflammatory response despite the
high delivered and retained doses relative to controls. For clarification, in two of four additional
CAPs exposure protocols, Kodavanti et al. (2000) observed a significant neutrophil influx in
7A-51
-------
TABLE 7A-llb. CAPs: RAT INHALATION STUDY (Kodavanti et al., 2000)
Rat CAPs Human
Kodavanti et al. (2000)a {Similarly exposed}a
MMAD (urn)
°g
Concentration (p.g/m3)
Deposited TB Dose per SA ° (p-g/m2)
Deposited A Dose per SA (p.g/m2)
Retained TB Dose per SA (p-g/m2)
Retained A Dose per SA (p.g/m2)
0.98b
1.41b
590
1740
29
lld
28 d
0.98
1.41
590
642
8.8
43d
8.6d
a Exposure was for 6 h/day for 3 days, both rats and humans were presumed exposed at rest.
b Personal communication by study authors.
c Surface area of lung region.
d Retained dose at 18 h following the 3rd exposure.
bronchitic rats when lavaged within 3 h postexposure. However, data from the rats lavaged at
18 h postexposure are used here for comparison to the Ohio et al. (2000) and Clarke et al. (1999)
studies where lavages were performed at 18 and 24 h postexposure, respectively.
Clarke et al. (1999) exposed healthy (n = 12) and bronchitic (n = 12) rats in Boston, MA
to 515 |ig/m3 of CAPs, 5 h per day, for 3 days. Table 7A-1 Ic provides the predicted tissue doses
for rats in the Clarke et al. (1999) study and the predicted doses for similarly exposed humans.
Note that due to differences in the inhaled particle size, the rats in the Clarke et al. (1999) study
were predicted to receive a greater dose than the rats in the Kodavanti et al. (2000) study despite
a shorter exposure time and lower CAPs concentration. The dose of CAPs per alveolar surface
area was about 67 times greater in the rats (Clark et al., 1999) relative to the humans (Ohio et al.,
2001). The inflammatory response observed in healthy rats by Clarke et al. (1999), however,
was quantitatively similar to that observed by Ohio et al. (2000) in healthy humans.
7A-52
-------
TABLE 7A-llc. CAPs: RAT INHALATION STUDY (Clarke et al., 1999)
Rat CAPs Human
Clarke et al. (1999)a {Similarly exposed}'
MMAD (urn)
o\
g
Concentration (p.g/m3)
Deposited TB Dose per SA ° (p-g/m2)
Deposited A Dose per SA (p.g/m2)
Retained TB Dose per SA (p.g/m2)b
Retained A Dose per SA (p.g/m2) b
0.18b
2.9b
515
1580
48
16d
47 d
0.18
2.9
515
802
8.9
36d
8.8d
a Exposure was for 5 h/day for 3 days, both rats and humans were presumed exposed at rest.
b This is the size distribution of the ambient particles and may differ from the concentrated aerosol to which the
rats were exposed.
c Surface area of lung region.
d Retained dose at 24 h following the 3rd exposure.
7A.7.3 Clearance Overload in Rats
Unlike other laboratory animals and humans, rats appear susceptible to "overload"-related
effects due to impaired macrophage-mediated alveolar clearance. Numerous reviews have
discussed this phenomenon and the difficulties it poses for the extrapolation of chronic effects in
rats to humans (ILSI, 2000; Miller, 2000; Oberdorster, 1995, 2002; Morrow, 1994). In brief, rats
chronically exposed to high concentrations of insoluble particles, even those which may
generally be considered as nuisance or low toxicity dusts, experience a reduction in their alveolar
clearance rates. With continued exposure, some rats eventually develop pulmonary fibrosis and
both benign and malignant tumors. These high-dose effects are not observed at lower doses in
rats. Oberdorster (2002) proposed that high-dose effects observed in rats may be associated with
two thresholds. The first threshold is the pulmonary dose that results in a reduction in
macrophage-mediated clearance. The second threshold, occurring at a higher dose than the first,
is the dose at which antioxidant defenses are overwhelmed and pulmonary tumors develop.
In chronic exposure studies, maintaining pulmonary doses below these thresholds should lessen
7A-53
-------
the uncertainty in the extrapolation of effects observed in rats to those expected in humans. Here
the focus will be on the lower threshold, i.e., the dose capable of overwhelming macrophage-
mediated alveolar clearance in rats, and derive concentrations for chronic exposures below
which overload might be avoided.
Overload has been loosely defined as the alveolar burden causing a 2- to 4-fold reduction
in alveolar clearance rates relative to normal clearance rates (ILSI, 2000; Oberdorster, 1995).
There is some discrepancy between whether overload is related to deposited particle volume or
surface area (Miller, 2000; Oberdorster, 2002). Here, only the relationship between volume
loading and overload is considered. To be consistent with Morrow's (1988, 1994) analyses in
this discussion of overload, the following values are assumed for rats: lung weight, 1.5 grams;
displaced volume of an AM, 1000 |im3; number of AM, 2.5 x 107. Morrow (1988) suggested a
rat's macrophage-mediated clearance was impaired at a volumetric loading of 60 jim3 per AM
and that macrophage stasis occurred at a loading of 600 jam3. These volumes represent 6 and
60% of the AM's displaced volume and correspond to the volumetric loadings of 1,000 and
10,000 nL/g-lung, respectively. Clearance rates do not differ from control at the volume loading
of 100 nL/g-lung or 6 jim3 per AM (Morrow, 1994). Morrow (1994) described the relationship
between alveolar clearance rates (k, day"1) in rats and the particle volume loading (Va,
nL/g-lung) as k = 0.021 - 0.0052 x log(Va) for 100 < Va < 10,000 nL/g-lung. Based on this
equation and consistent with Morrow (1988), the loading that would cause a doubling of the
retention half-time (a loose definition of overload) can be determined to occur at 1,000 nL/g-
lung or 60 jim3 per AM. For comparison, from Table 2 in Oberdorster (1995), a loading of
1,400 nL/g-lung can be inferred as doubling retention half-times, fairly consistent with Morrow
(1994).
Based on the work of Morrow (1988, 1994), estimates of the volumetric loadings
associated with no effect on clearance (100 nL/g-lung), the onset of overload (1,000 nL/g-lung),
and AM stasis (10,000 nL/g-lung) can be determined. The goal here was to derive
concentrations for chronic exposures below which overload might be avoided. Miller (2000)
estimated the amount of time that it would take for a rat (F344) exposed to 10 mg/m3 for 24 h
per day to reach clearance stasis on the basis of volumetric loading. For monodisperse 1 jim
particles (DF = 0.04, VT = 2.1ml, f = 102 min"1), Miller estimated it would take about 80 days
7A-54
-------
(ignoring clearance) for the AM to become filled and reach stasis. Within the macrophage,
particles were assumed to be tightly packed spheres occupying a volume of 1.43 times greater
than the volume of the particles themselves, i.e., the porosity or void space between spheres is
0.3. Using the clearance kinetics from the MPPD model, an additional 10 days (90 days total)
would be required to reach stasis. This approach can also be used to determine the amount of
time required to reach lower levels of AM loading, or conversely, the exposure concentration
achieving a level of loading in a given period of time.
In Table 7A-12, particle concentrations for rat exposures predicted to cause various levels
of alveolar loading are shown. Alveolar loadings in this table refer to the volumes occupied by
unit density spheres. However, particle density cannot be ignored, because for a constant
MMAD, the physical size and volume of particles decreases with increasing density. Hence,
despite having the same MMAD, dense particles would achieve a lower volumetric loading than
unit density spheres for the same exposure concentration. The loading achieving stasis has been
reduced from 10 jiL/g-lung to 7 jiL/g-lung as an adjustment for the void space between packed
particles within macrophages. The onset of overload may also be considered as adjusted for
void space based on a reduction from 1.4 |iL/g-lung (Oberdorster, 1995) tol |iL/g-lung.
Although, this difference (1 versus 1.4 jiL/g-lung) may be due to variability between
experiments.
The volumetric loadings in the Table 7A-13 were estimated for an exposure scenario of 6 h
per days, 5 days per week. However, other exposure scenarios can easily be considered by
maintaining a constant weekly exposure. For instance, in rats exposed 6 h per day, 1 day per
week for 1 year to an aerosol (MMAD = 2|im, og = 1.5), a loading of 1 |iL/g-lung is predicted
for an exposure concentration of 12.5 mg/m3. This exposure concentration of 12.5 mg/m3 is
calculated as 2.5 mg/m3 (from table) x 30 h (used for table estimates) + 6 h (the desired weekly
exposure time).
The analysis of particle overload in rats presented here is somewhat simplistic in that it
only considered the accumulated volumetric burden of particles in the lung. More sophisticated
multicompartment models of AM-mediated clearance, based on particle volume (Stober, 1994)
and particle surface area (Tran, 2000), exist. An important consideration addressed by Stober
et al. (1994) is that not all AM carry the same burden. Another important AM-related
7A-55
-------
TABLE 7A-12. ESTIMATED EXPOSURE CONCENTRATIONS (mg/m3)
LEADING TO VARIED LEVELS OF ALVEOLAR LOADING AS A FUNCTION
OF PARTICLE SIZE AND EXPOSURE DURATION
Exposure MMADb
Time a (urn)
2 months 1
2
3
4
3 months 1
2
3
4
6 months 1
2
3
4
1 year 1
2
3
4
2 years 1
2
3
4
Alveolar
0.1
TVo effect 0.3
Exposure
1.1
1
1.3
1.8
0.8
0.8
1
1.3
0.6
0.6
0.7
1
0.6
0.5
0.7
0.9
0.6
0.5
0.6
0.9
Volume Loading0
1
Overload
(uL/g-lung)
3
7
Stasis
Concentration (mg/m3) to Achieve Above
Alveolar Loadings
3 9.1 25 57
2.7
3.4
4.8
2.2
1.9
2.5
3.5
1.5
1.4
1.7
2.4
1.3
1.1
1.4
2
1.2
1
1.3
1.9
8.1
10
15
6.2
5.5
7
10
3.9
3.5
4.4
6.2
2.8
2.5
3.2
4.5
2.4
2.1
2.7
3.8
22
29
40
16
15
19
26
9.5
8.4
11
15
6.1
5.4
6.9
9.7
4.4
3.9
5
7.1
50
64
90
36
32
41
58
20
18
22
32
12
10
13
19
7.6
6.8
8.6
12
aRats presumed exposed at rest for 6 h per day, 5 days per week.
b Geometric standard deviation of 1.5.
0 Effects of alveolar loading on macrophage-mediated clearance range from no effect at 0.1 uL/g-lung,
to overload at 1 uL/g-lung, to stasis at 7 uL/g-lung.
consideration is that particle uptake by AM depends on particle size. The efficiency of
phagocytosis by AM appears to be greatest for particles between 1.5 and 3 jim in diameter
(Oberdorster, 1988). Adamson and Bowden (1981) reported less phagocytic activity in rats
following instillation of 0.1 jim versus 1.0 jim latex spheres. In addition, Adamson and Bowden
(1981) identified 0.1 |im spheres in Type 1 epithelial cells, free in the interstitium, and in
7A-56
-------
interstitial macrophages — all of which were rarely seen for the larger 1.0 jim spheres.
Recognizing the importance of particle size on AM-mediated clearance, only values of MMAD
between 1 and 4 jim were included in the analysis of overload discussed here.
Particles formed by the aggregation of smaller particles also warrant special consideration
with regard to alveolar clearance kinetics. Ferin et al. (1992) conducted an inhalation study
using particle aggregates having MMAD of 0.78 and 0.71 |im, which were composed of
0.021 and 0.25 jim TiO2 primary particles, respectively. Rats were exposed to an aerosol
concentration of 23 to 23.5 mg/m3 (6 h/day, 5 days/week, 12 weeks). Postexposure retention
half-times were 174 days for fine primary particles (0.25 jim diameter) versus 501 days for the
ultrafine primary particles (0.021 jim diameter). With primary particle size affecting clearance
kinetics, this study supports the supposition that TiO2 particles may disaggregate following
deposition in the lung. Particle disaggregation following deposition in the lung appears to be
specific to TiO2 aerosols (Takenaka et al., 1986). Contrary to the analysis of overload as a
function of volume loading as presented in Table 7A-12, ultrafine TiO2 particles have a longer
retention than fine TiO2 particles, despite having the same alveolar loading on the basis of
particle volume. The overload of ultrafine particles might be function of particle surface area
rather than volume. Hence, depending on material from which an aggregate is composed, it may
be inappropriate to assess alveolar loading on the basis of particle volume.
7A.8 SUMMARY
The MPPD model was used to calculate concentrations of atmospheric and resuspended
PM that would be necessary to achieve doses in the rat comparable to those in humans breathing
ambient PM, as measured by a variety of dose metrics. The same model was then used to
estimate the differences in doses in rats and humans exposed to comparable types of ambient or
emission PM in salient published studies. Complementary approaches were used to analyze the
relationship between PM doses resulting from inhalation exposures or intratracheal instillation in
rats and PM doses in humans resulting from exposures during a variety of activities.
7A-57
-------
The MPPD model estimates in Table 7A-8a suggest that, depending on the activity level
and breathing pattern of the human, a rat may need to be exposed to between 33 and 200 |ig/m3
of resuspended PM over 6 h to receive an incremental dose in the A region per surface area
(measured as deposited mass) comparable to that of a healthy human working for 6 h near a busy
road and exposed to 100 |ig/m3 ambient PM10. To achieve an incremental dose retained in the rat
TB region per TB surface area (averaged over 6 h) comparable to that in the human, the rat
would need to be exposed to between 150 and 630 |ig/m3 (depending on the activity level and
breathing pattern of the human) of resuspended PM for 6 h. However, because of the more rapid
clearance in the rat, a higher exposure concentration of between 0.7 and 1.6 mg/m3 would be
required for the rat to achieve a retained TB dose per TB surface area (averaged over 24 h)
comparable to that in the human.
The chronic retention of PM in the A region of the human cannot be simulated in the rat
except under conditions in which the normal clearance process of the rat is inhibited. However,
giving high doses of PM to healthy mature rats will likely not simulate the response of humans
who are vulnerable because of heart or vascular disease, infectious diseases of the lung,
conditions such as diabetes, or acute or chronic stress. Therefore, development of rat models of
human vulnerabilities would enhance the value of using the rat in inhalation toxicology studies.
Understanding the interplay of dose and responsiveness in animal models as well as in the
human will substantially advance the ability to predict adverse health outcomes in the human
population.
In daily life, humans are exposed to PM in the atmosphere and inhale a complex profile of
Aitken, accumulation, and coarse mode particles covering a size range from below 0.1 to over
10 jim diameter. On the other hand, laboratory inhalation studies do not simulate the full size
distribution to which humans are exposed and in some cases do not simulate the chemical
composition or physical structure of atmospheric particles. Resuspended PM (e.g., ROFA-like
material or other bulk material) has a particle size intermediate between coarse and accumulation
modes but does not have the smaller sizes of the accumulation or Aitken modes. CAPs give a
better simulation of the chemical composition of atmospheric particles but typically concentrate
only one mode. For ultrafme particles, the physical structure and possibly the chemical
composition may be changed by going through growth and shrinkage during the concentration
7A-58
-------
process. Fresh diesel exhaust particles, especially if more concentrated than in a roadway, will
have a larger particle size than when diluted by vehicle turbulence. They will also differ in
physical structure and chemical composition from aged diesel particles. Acid aerosol studies
may also use particle sizes in the accumulation mode size range but usually do not contain the
metals and organic components found in atmospheric aerosols. Laboratory exposures of rats to
resuspended dust can simulate the dose of particle mass to the alveolar region but not dose
metrics based on particle surface area or number unless very high concentrations are used.
While the calculation of EqER for various dose metrics and normalizing factors is simple,
the interpretation of the resulting EqERs can be somewhat more ambiguous. Optimally, the
choice of dose metrics and normalizing factors should be based on the biological mechanisms
mediating an effect. For soluble compounds, the mass of PM depositing in a region of the lung
may be the most appropriate dose metric. For poorly soluble particles depositing in the
A region, particle surface area (for ultrafine PM) or particle volume (for coarse PM) may be
more appropriate dose metrics. The appropriateness of a normalizing factor is, in part,
determined by the site most affected by PM. For soluble compounds, an appropriate
normalizing factor could be the surface area of the airways for irritants whereas body mass
would be more logical when considering systemic effects. For insoluble compounds retained in
the lung, normalizing factors can range from the number of macrophages in an alveolus to the
mass of the lung. Due to the more rapid clearance in rats, larger rat exposure doses will be
required to simulate retained doses in humans than would be the case for deposited doses.
If dose metrics based on surface area or particle number are appropriate, rat exposure
concentrations using resuspended PM must be very high because resuspended PM contains
few accumulation mode or ultrafine particles.
It appears that no single dose metric nor normalizing factor is appropriate for all situations.
As illustrated in Tables 7A-7a through 7A-9b, the parameters chosen can drastically affect the
rat exposure concentration required to provide a normalized dose equivalent to that occurring in
a human. A rat exposure which simulates a human dose for one specific dose metric or
normalizing factor may provide a higher or lower dose as measured by a different dose metric or
normalizing factor. In addition, regardless of the dose metric and normalizing factor chosen, the
exposure concentration required for a rat to achieve an equivalent human dose increases with the
7A-59
-------
level of activity of the human being considered. From a purely dosimetric standpoint, the
complexity of interspecies extrapolation is obvious but not necessarily insurmountable.
Conclusions regarding rat to human comparisons may require the use of a variety of dose metrics
and normalizing factors depending on the degree to which biological mechanisms mediating an
effect are understood.
Instillation studies in both animals and humans have been criticized for lack of relevance
related to dose and means of administration. Ohio and Devlin (2001) instilled 500 jig of Utah
Valley Dust (UVD) extracts into the lingula of human volunteers (healthy young adults). This
instilled dose (about 170 jig per m2 alveolar surface area) elicited a robust inflammatory
response for the 1986 and 1988 extracts, but not the 1987 extract, suggesting that extract
composition is important. In a complementary animal study, the intratracheal instillation of rats
with 250 jig (840 jig per m2 alveolar surface area) of 1986 UVD extracts also caused an
inflammatory response (Dye et al., 2001). The neutrophilic response elicited by the 1986 UVD
extract instillations was about 3 times greater in the rats (10-fold PMN increase) than in humans
(3.5-fold PMN increase). On the basis of mass per alveolar surface area, however, the dose
delivered to the rats was about 5 times greater than delivered to the humans. This disparity
(3 times the response at 5 times the dose) is suggestive of a decreased susceptibility for an
inflammatory response in the rats relative to humans. However, baseline levels of PMN may
differ between studies and might account for some of the differences described here.
For comparison to delivery by inhalation, it was estimated that 44-65 days of exposure in
the Utah Valley during the winter 1985-1986 would have been required for a person to have
received a PM dose per alveolar surface area equivalent to that of instillations in the study by
Ohio and Devlin (2001) (see Table 7A-10b). However, it was estimated that a rat lung burden of
250 jig, the mass instilled by Dye et al. (2001), could not be achieved by inhalation at the
assumed ambient exposure scenario due to the rapid clearance in the rat (Table 7A-10c [1]).
Toxicologically, it is obvious that a different response might be expected between an instilled
dose (delivered as a bolus) versus the a subchronic delivery by inhalation. For a more acute
(24-h period) delivery by inhalation, humans would need to be exposed to ~6 mg/m3 and rats to
15 mg/m3 in order to reach the instilled doses used in the Ohio and Devlin (2001) and Dye et al.
(2001) studies, respectively. Dosimetrically, the relevance of both the human and the rat
7A-60
-------
instillation studies to exposure by inhalation are difficult to judge and it should again be noted
that the extracts contained only the soluble fraction of the UVD. However, both rat and human
instillation studies showed that the 1987 UVD (collected while the Gevena Steel Mill was
closed) extract was relatively less potent compared to the 1986 and 1988 extracts.
Several studies (one human and two rat) involving exposure by inhalation to CAPs provide
a seemingly more useful basis for comparing dose and response. Tables 7A-1 la, -1 Ib, and -lie
provided exposure conditions and estimated doses for the human study by Ohio et al. (2000), the
rat study by Kodavanti et al. (2000), and the rat study by Clarke et al. (1999), respectively.
Bronchial lavages were performed at 18 h postexposure in both the Ohio et al. (2000) and
Kodavanti et al. (2000) studies and at 24 h postexposure in the Clarke et al. (1999) study. At the
time of bronchial lavage, the estimated alveolar dose in the human study was 0.7 |ig/m2. This
dose produced a mild inflammatory response in young healthy human subjects. In the Clarke
et al. (1999) study, an increase in neutrophils in response to CAPs exposure was found in healthy
rats (air, -1%; CAPs, -7%) that was very similar to that observed in healthy humans (air, 2.7%;
CAPs, 8.1%) by Ohio et al. (2000). However, the alveolar tissue doses (mass per surface area)
are estimated to be 67 times greater in the rats than in the humans. The similarity in the
response, but disparity in dose, suggests that healthy rats may be less susceptible to
inflammatory effects of CAPs than healthy humans. In the Kodavanti et al. (2000) study, rats
were predicted to have 40 times the human dose in the Ohio et al. (2000) study but only 60% of
the dose delivered to the rats in the Clarke et al. (1999) study. Interestingly, neither the healthy
nor bronchitic rats in the Kodavanti et al. (2000) study showed a consistent inflammatory
response, again suggesting that rats are less susceptible to CAPs effects than healthy humans.
The composition, size distribution, and concentration of CAPs varies temporally and spatially.
Thus, some of the above-described differences between studies may be attributable to the
toxicity of the CAPs itself.
A key premise for the dosimetric analysis presented here is that comparable tissue doses
should cause comparable effects. From the preceding discussion of CAPs studies, however,
it appears that rats (whether healthy or compromised) may have a decreased inflammatory
response (indexed by PMN levels) relative to healthy humans at comparable tissue doses.
Decreased sensitivity of rats relative to humans may only occur in studies of several days
7A-61
-------
duration. For longer subchronic and chronic studies, rats appear susceptible to an overload of
their macrophage-mediated alveolar clearance. Under conditions of overload, rats may indeed
be more susceptible than humans, having decreased rates of alveolar clearance and antioxidant
defenses. Table 7A-12 provided exposure concentrations for chronic exposures below which
overload might be avoided.
7A.9 CONCLUSIONS
• Exposure concentrations can be estimated that give a rat the same dose as received by a
human exposed to various levels of ambient PM as a function of dose metric, normalizing
factor, and level of human exertion. The estimated concentrations will vary widely
depending on the selection of these parameters. While human and rat doses may be
matched for a specific dose metric, normalizing factor, and level of human exertion, the
dose estimated for other dose metrics and normalizing factors may be quite different.
Thus, it may not be possible to match all relevant dose metrics.
• The dosimetric calculations indicate that PM concentration exposures in rats, somewhat
higher than in humans, would be justified in certain conditions to achieve nominally
similar acute doses per surface area relative to the humans undergoing moderate to high
exertion. However, for resting or light exertions, lower rat exposure concentrations are
adequate to produce equivalent doses.
• Given that rats clear PM much faster than humans, the MPPD model results show that
much higher exposure concentrations in the rat are required to simulate the retained
burden of poorly soluble particles which builds up over years of human exposure.
• In resuspended PM, used in some inhalation studies, the smaller particles found in the
accumulation and Aitken modes of the atmospheric aerosol are aggregated onto (or into)
larger particles. Thus, for dose metrics based on particle surface area or number very high
exposure concentrations of resuspended PM for rats would be required to provide a dose
equivalent to that received by humans exposed to atmospheric aerosol.
• The biological mechanisms of PM toxicity are uncertain, as are the dose metrics most
appropriate for establishing human-rat equivalent doses. The concept of using dosimetric
calculations to provide a quantitative rat to human extrapolation depends on the
assumption that an equal dose to target cells or tissues will produce a similar response in
each species. At sufficiently high doses, however, the rat is subject to an overload
phenomenon. When this occurs in the rat, clearance slows and anti-inflammatory
defenses become depleted. Under these conditions, rats are more sensitive to PM than
humans and tumor formation and fibrosis may occur. At lower doses, healthy rats clear
PM faster than healthy humans and appear less sensitive to PM-induced inflammatory
responses than humans. Thus, it is essential for toxicological studies to characterize dose
7A-62
-------
to the fullest extent possible and to carefully consider dose-response relationships in both
rats and humans.
• Particle characteristics and biological normalizing factors which mediate effects should
be carefully considered in study design. It is important that investigators provide accurate
and complete information regarding exposure conditions (PM concentration, exposure
duration, and particle size distribution) so that respiratory doses can be calculated for
comparisons between studies.
• It is difficult to judge the relevance of human and rat instillation studies to ambient
exposures via inhalation dosimetrically. The UVD instillation studies discussed herein
used extracts containing only the soluble fraction of the UVD. The estimated alveolar
surface doses from UVD extract instillations are likely far greater than would have
occurred in residents of the Utah Valley during the winters of 1986-1988. However,
both rat and human instillation studies showed that the 1987 UVD (collected while the
Geneva Steel Mill was closed) extract was relatively less potent compared to the 1986
and 1988 extracts.
• Inflammatory responses were compared between several CAPs inhalation studies (one
human and two rat). Exposure concentration data did not provide a useful means for
comparing studies even when considering only the two rat studies. An analysis of
inflammatory response as a function of PM dose showed rats to be less sensitive than
humans in short duration studies of 3 days or less, although some of the variability
between studies may be attributable to differences in the CAPs itself.
• Calculation of PM dose to the lung requires data on exposure concentration, exposure
duration, and particle characteristics (solubility, hygroscopicity, size distribution, etc.)
as well as information about the exposed individual or animal (age, gender, respiratory
health, lung size, breathing conditions, etc.). Erroneous estimates of dose can occur from
missing or faulty data, e.g., a multimodal particle size distribution being characterized as
if unimodal. In order to examine the PM dose-response relationship across studies,
complete and accurate information about the exposure and the exposed subjects need to
be provided in the published literature.
7A-63
-------
REFERENCES
Adamson, I. Y. R.; Bowden, D. H. (1981) Dose response of the pulmonary macrophagic system to various
particulates and its relationship to transepithelial passage of free particles. Exp. Lung Res. 2: 165-175.
Anjilvel, S.; Asgharian, B. (1995) A multiple-path model of particle deposition in the rat lung. Fundam. Appl.
Toxicol. 28: 41-50.
Asgharian, B.; Hofmann, W.; Miller, F. J. (2001) Mucociliary clearance of insoluble particles from the
tracheobronchial airways of the human lung. J. Aerosol Sci. 32: 817-832.
Bermudez, E.; Mangum, J. B.; Asgharian, B.; Wong, B. A.; Reverdy, E. E.; Janszen, D. B.; Hext, P. M; Warheit,
D. B.; Everitt, J. I. (2002) Long-term pulmonary responses of three laboratory rodent species to subchronic
inhalation of pigmentary titanium dioxide particles. Toxicol. Sci. 70: 86-97.
Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.; Sioutas, C.; Paulauskis, J.; Coull, B.;
Ferguson, S.; Godleski, J. J. (1999) Urban air paniculate inhalation alters pulmonary function and induces
pulmonary inflammation in a rodent model of chronic bronchitis. Inhalation Toxicol. 11: 637-656.
Dormans, J. A. M. A.; Steerenberg, P. A.; Arts, J. H. E.; Van Bree, L.; De Klerk, A.; Verlaan, A. P. J.; Bruijntjes,
J. P.; Beekhof, P.; Van Soolingen, D.; VanLoveren, H. (1999) Pathological and immunological effects of
respirable coal fly ash in male wistar rats. Inhalation Toxicol. 11: 51-69.
Driscoll, K. E.; Schlesinger, R. B. (1988) Alveolar macrophage-stimulated neutrophil and monocyte migration:
effects of in vitro ozone exposure. Toxicol. Appl. Pharmacol. 93: 312-318.
Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. J.; Costa, D. L.
(2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
studies in Utah Valley residents. Environ. Health Perspect. Suppl. 109(3): 395-403.
Federal Register. (1997) National ambient air quality standards for paniculate matter; final rule. F. R. (July 18)
62: 38,652-38,752.
Felicetti, S. A.; Wolff, R. K.; Muggenburg, B. A. (1981) Comparison of trachea! mucous transport in rats, guinea
pigs, rabbits, and dogs. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 51: 1612-1617.
Ferin, J.; Oberdorster, G.; Penney, D. P. (1992) Pulmonary retention of ultrafine and fine particles in rats. Am. J.
Respir. CellMol. Biol. 6: 535-542.
Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles.
Am. J. Respir. Crit. Care Med. 164: 704-708.
Ohio, A. J.; Kim, C.; Devlin, R. B. (2000) Concentrated ambient air particles induce mild pulmonary inflammation
in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162: 981-988.
Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93.
Gordon, T.; Nadziejko, C.; Lung, C. C.; Schlesinger, R. (2004) Effects of concentrated ambient particles in rats and
hamsters: an exploratory study [erratum]. Boston, MA: Health Effects Institute; research report no. 93; p. 15.
Available: http://www.healtheffects.org/Pubs/Gordon-Table2.pdf (15 July, 2004).
Hofmann, W.; Asgharian, B. (2003) The effect of lung structure on mucociliary clearance and particle retention in
human and rat lungs. Toxicol. Sci. 73: 448-456.
International Commission on Radiological Protection (ICRP). (1994) Human respiratory tract model for radiological
protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP: v. 24, pp. 1-482).
International Life Sciences Institute (ILSI) Risk Science Institute Workshop Participants. (2000) The relevance of
the rat lung response to particle overload for human risk assessment: a workshop consensus report. In:
Gardner, D. E., ed. ILSI Risk Science Institute Workshop: The Relevance of the Rat Lung Response to
Particle Overload for Human Risk Assessment; March, 1998. Inhalation Toxicol. 12: 1-17.
Jarabek, A. M. (1994) Inhalation RfC methodology: dosimetric adjustments and dose-response estimation of
noncancer toxicity in the upper respiratory tract. Inhalation Toxicol. 6(suppl.): 301-325.
Jarabek, A. M. (1995) Interspecies extrapolation based on mechanistic determinants of chemical disposition. Hum.
Ecol. Risk Assess. 1: 641-662.
Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
(1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly ash
inhalation. Inhalation Toxicol. 9: 541-565.
Kodavanti, U. P.; Mebane, R.; Ledbetter, A.; Krantz, T.; McGee, J.; Jackson, M. C.; Walsh, L.; Hilliard, H.; Chen,
B. Y.; Richards, J.; Costa, D. L. (2000) Variable pulmonary responses from exposure to concentrated ambient
air particles in a rat model of bronchitis. Toxicol. Sci. 54: 441-451.
7A-64
-------
Kodavanti, U. P.; Schladweiler, M. C.; Ledbetter, A. D.; Hauser, R.; Christian!, D. C.; McGee, I; Richards, J. R.;
Costa, D. L. (2002) Temporal association between pulmonary and systemic effects of paniculate matter in
healthy and cardiovascular compromised rats. J. Toxicol. Environ. Health Part A 65: 1545-1569.
Kreyling, W. G.; Scheuch, G. (2000) Clearance of particles deposited in the lungs. In: Gehr, P.; Heyder, J.; eds.
Particle-lung interactions. New York, NY: Marcel Dekker, Inc.; pp. 323-376. (Lung Biology in Health and
Disease, v. 143.)
Menache, M. G.; Miller, F. J.; Raabe, O. G. (1995) Particle inhalability curves for humans and small laboratory
animals. Ann. Occup. Hyg. 39: 317-328.
Mercer, R. R. (1999) Morphometric analysis of alveolar responses of F344 rats to subchronic inhalation of nitric
oxide. Cambridge, MA: Health Effects Institute; research report no. 88.
Mercer, R. R.; Russell, M. L.; Crapo, J. D. (1994) Alveolar septal structure in different species. J. Appl. Physiol.
77: 1060-1066.
Miller, F. J. (2000) Dosimetry of particles in laboratory animals and humans in relationship to issues surrounding
lung overload and human health risk assessment: a critical review. Inhalation Toxicol. 12: 19-57.
Morris, A. H.; Kanner, R. E.; Crapo, R. O.; Gardner, R. M. (1984) Clinical pulmonary function testing: a manual of
uniform laboratory procedures. 2nd ed. Salt Lake City, UT: Intermountain Thoracic Society.
Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
10: 369-384.
Morrow, P. E. (1994) Mechanisms and significance of "particle overload." In: Mohr, U.; Dungworth, D. L.;
Mauderly, J. L.; Oberdorster, G., eds. Toxic and carcinogenic effects of solid particles in the respiratory tract:
[proceedings of the 4th international inhalation symposium]; March 1993; Hannover, Germany. Washington,
DC: International Life Sciences Institute Press; pp. 17-25.
Niinimaa, V.; Cole, P.; Mintz, S.; Shephard, R. J. (1981) Oronasal distribution of respiratory airflow. Respir.
Physiol. 43:69-75.
Oberdorster, G. (1988) Lung clearance of inhaled insoluble and soluble particles. J. Aerosol Med. 1: 289-330.
Oberdorster, G.; Ferin, J.; Lehnert, B. E. (1994) Correlation between particle size, in vivo particle persistence, and
lung injury. Environ. Health Perspect. 102(suppl. 5): 173-179.
Oberdorster, G. (1995) Lung particle overload: implications for occupational exposures to particles. Regul. Toxicol.
Pharmacol. 27: 123-135.
Oberdorster, G.; Finkelstein, J. N.; Johnston, C.; Gelein, R.; Cox, C.; Baggs, R.; Elder, A. C. P. (2000) Acute
pulmonary effects of ultrafine particles in rats and mice. Cambridge, MA: Health Effects Institute; research
report no. 96.
Oberdorster, G. (2002) Toxicokinetics and effects of fibrous and nonfibrous particles. Inhalation Toxicol. 14: 29-56.
Pinto, J. P.; Stevens, R. K.; Willis, R. D.; Kellogg, R.; Mamane, Y.; Novak, J.; Santroch, J.; Benes, I.; Lenicek, J.;
Bures, V. (1998) Czech air quality monitoring and receptor modeling study. Environ. Sci. Technol.
32: 843-854.
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Stober, W.; Morrow, P. E.; Koch, W.; Morawietz, G. (1994) Alveolar clearance and retention of inhaled insoluble
particles in rats simulated by a model inferring macrophage particle load distributions. J. Aerosol Sci.
25: 975-1002.
Takenaka, S.; Dornhofer-Takenaka, H.; Muhle, H. (1986) Alveolar distribution of fly ash and of titanium dioxide
after long-term inhalation by Wistar rats. J. Aerosol Sci. 17: 361-364.
Takezawa, J.; Miller, F. J.; O'Neil, J. J. (1980) Single-breath diffusing capacity and lung volumes in small laboratory
mammals. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 48: 1052-1059.
Tran, C. L.; Buchanan, D.; Miller, B. G.; Jones, A. D. (2000) Mathematical modeling to predict the responses to
poorly soluble particles in rat lungs. Inhalation Toxicol. 12(suppl. 3): 403-409.
U.S. Environmental Protection Agency. (1994) Methods for derivation of inhalation reference concentrations and
application of inhalation dosimetry. Research Triangle Park, NC: Office of Health and Environmental
Assessment, Environmental Criteria and Assessment Office; report no. EPA/600/8-88/066F. Available:
http://cfpubl.epa.gov/ncea/cfm/recordisplay.cfm?deid=71993 (15 June, 2004).
U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
Weibel, E. R. (1963) Morphometry of the human lung. New York, NY: Academic Press Inc.
Weibel, E. R. (1972) Morphometric estimation of pulmonary diffusion capacity: V. comparative morphometry of
alveolar lungs. Respir. Physiol. 14: 26-43.
7A-65
-------
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
Wichmann, H.-E.; Peters, A. (2000) Epidemiological evidence of the effects of ultrafine particle exposure. Philos.
Trans. R. Soc. London, Ser. A 358: 2751-2768.
Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
Cambridge, MA: Health Effects Institute; research report no. 98.
Winter-Sorkina, R. de; Cassee, F. R. (2002) From concentration to dose: factors influencing airborne paniculate
matter deposition in humans and rats. Bilthoven, The Netherlands: National Institute of Public Health and the
Environment (RIVM); report no. 650010031/2002. Available:
http://www.rivm.nl/bibliotheek/rapporten/650010031.html (13 June 2003).
Yeh, H. C.; Schum, G. M. (1980) Models of human lung airways and their application to inhaled particle deposition.
Bull. Math. Biol. 42: 461-480.
Yeh, H. C.; Schum, G. M.; Duggan, M. T. (1979) Anatomic models of the tracheobronchial and pulmonary regions
of the rat. Anat. Rec. 195: 483-492.
Zhang, L.; Yu, C. P. (1993) Empirical equations for nasal deposition of inhaled particles in small laboratory animals
and humans. Aerosol Sci. Technol. 19: 51-56.
7A-66
-------
APPENDIX 7B. AMBIENT BIO AEROSOLS
7B. 1 INTRODUCTION AND BACKGROUND INFORMATION ON
AMBIENT BIOAEROSOLS
The American Conference of Industrial Hygienists defines bioaerosols as airborne
particles, large molecules or volatile compounds that are living, contain living organisms, or
originate from living organisms. Such particles may be suspended in the air adhered to dust
particles or tiny droplets of water. Bioaerosols include fungal materials, pollen, bacteria,
viruses, endotoxins, and plant and animal debris, and range in size from 0.01 jim (viruses) to
well over 20 jim (pollen). They are naturally present in the environment and can pose a threat to
human health, especially for sensitive individuals for whom some bioaerosols, when inhaled,
may cause diseases such as asthma, allergic rhinitis, and respiratory infections. The 1996 PM
AQCD (U.S. Environmental Protection Agency, 1996), highlighted several examples of common
bioaerosol sources, particles, and agents, as listed in Table 7B-1 and discussed in several earlier
bioaerosols reviews, e.g., Cox (1987), Pope et al. (1993), Lighthart and Mohr (1994), and Cox
and Watties (1995).
TABLE 7B-1. EXAMPLES OF MAJOR SOURCES, TYPES OF PARTICLES,
AND DISEASE AGENTS ASSOCIATED WITH BIOAEROSOLS
Sources
Aerosol Particles
Disease Agents
Plants Pollen and pollen fragments, fragments of other
plant parts, spores (ferns, mosses), algal cells
Animals Skin scales, secretions (saliva, skin secretions),
excreta, body parts (arthropods)
Fungi Spores, hyphae, yeast cells, metabolites
(toxins, digested substrate material)
Bacteria Cells, fragments, metabolites
(toxins, digested substrate material)
Viruses Viral particles
Glycoprotein allergens
Glycoprotein allergens
Glycoprotein allergens, infectious
units, glucans, mycotoxins
Infectious units, allergens,
endotoxin, exotoxins
Infectious units
Source: Modified from 1996 PM AQCD (U.S. EPA, 1996).
7B-1
-------
7B.1.1 Plant Aerosols
Pollen. Among the best known plant aerosols are pollens produced by flowering plants,
including trees (e.g., pines, cedars, birch, elm, maple, oak, etc.), weeds (e.g., ragweed, sage,
etc.), and grasses (e.g., rye grass, Bermuda grass, etc.). Within these groupings, specific types
are regionally more common, e.g., ragweed more so in the eastern United States, birch during
the spring pollen season in New England, mountain cedar early in the year in the southwest, etc.
(Lewis et al., 1983). Outdoor pollen levels are determined by numbers of plants available for
pollen release, the amount of pollen produced per plant, factors controlling pollen release and
dispersion from the plant, and factors directly affecting the aerosols (Edmonds, 1979). Plant
numbers depend on many environmental factors (some human) that control plant prevalence,
e.g., numbers of plants that produced seed in the past year, disturbed ground available for seed
germination and growth, growing season, and other meteorological factors (temperature, rainfall,
day length, etc.). Pollen shed is controlled by temperature, humidity, wind, and rain. Air pollen
levels depend on all of these factors, as well as wind and rain conditions after release and on
surfaces available for impaction. Pollen grains are large complex particles that consist of
cellular material surrounded by a cell membrane and a complex wall. Pollen allergens are water-
soluble glycoproteins that rapidly diffuse from the grain when it contacts a wet surface and are
generally specific to the type of pollen, although large groups include a single allergen (e.g.,
many different kinds of grasses have similar allergens in their pollen grains). Several pollen
allergens have been characterized: Amb a I (ragweed), Bet v I (birch), Parj I (parietaria).
Other Natural Plant Aerosols. Other plant-derived particles naturally occurring in outdoor
air include algal cells; spores of mosses, liverworts, club mosses, and ferns; and fragments of all
kinds of plants. Very little has been reported about the prevalence or human impact of any of
these aerosol particles, but they are presumed to carry allergens.
Plant-Related Bioaerosols Generated by Human Activities (Grain Dust Latex, etc.).
Human activities that accumulate plant materials, e.g., storage, handling, and transport of farm
products (hay, straw, grain), composting, produce bioaerosols. Grain dusts that include
respirable-size particles (< 10 jim) are of particular interest. Soybean dust aerosols released
from freighters unloading the beans in port have been blamed for epidemics of asthma. Also,
human uses of some plant products can result in disease-causing aerosols (Alberts and Brooks,
1992), e.g., wood trimmer's disease (from inhalation of wood dust particles released during
7B-2
-------
high-speed wood cutting); sewage composting involving use of wood chips which can release
allergenic aerosols, and latex particles from automobile tires that can contaminate reentrained
roadway dust.
7B.1.2 Animal Aerosols
Mammalian Aerosols. All mammals produce aerosols. Human aerosols (skin scales,
respiratory secretions) generally do not cause disease except for agents of infection (see below).
Other mammals release aerosols that cause hypersensitivity diseases, the most common sources
being cats, dogs, farm animals, laboratory animals, and house mice — although all animals
release aerosols that could be sensitizing under appropriate conditions (Burge, 1995). Mammals
only cause human disease under appropriate exposure conditions, e.g., having a cat in a house or
handling of any animal. Cat allergens apparently become aerosolized on very small particles
(< 1 |im) shed from skin and saliva. Dog, mouse, and other rodent allergens may be borne on
dried urine particles, having sizes similar to those of cat allergen. Little is known about other
mammalian aerosols. Cat and dog allergens (Peld\ CanfT) have been characterized.
Avian Aerosols. Examples of wild and domesticated birds associated with disease-causing
aerosols include: starlings (histoplasmosis); pigeons (histoplasmosis, pigeon-breeders disease);
parrots (psittacosis); poultry (poultry-handlers disease); etc. Only the hypersensitivity diseases
(e.g., pigeon breeders and poultry handlers disease) are caused by "bird" aerosols per se. The
others are infectious diseases caused by agents inhabiting the birds (see below). The avian
aerosol-hypersensitivity diseases are almost exclusively confined to sites where birds are bred
and handled extensively, especially in indoor environments; and birds that release antigens
observed to cause human disease are those that congregate or are typically confined close to
people. Relatively little is known about avian aerosols. Probably skin scales, feather particles,
and fecal material are all released as antigen-containing aerosols. The antigens (allergens)
responsible for avian-related hypersensitivity diseases have yet to be well characterized.
Insect Aerosols. Dust mites are arthropods (family Pyrogliphidae) that include two
common species in temperate climates: Dermatophagoidesfarinae, which proliferates under
relatively dry conditions; and D. pteronyssinus, which dominates in more humid environments
(Arlian, 1989). Dust mites thrive where relative humidity consistently exceeds 60% and where
skin scales and fungal spores are available as food. Bedding and carpet dust are prime exposure
7B-2
-------
reservoirs. The mite itself is about 100 jim long, but excretes 20 jim membrane-bound fecal
particles that contain allergens that are a major cause of sensitization in children. The allergens
are digestive enzymes that gradually diffuse from fecal particles after deposition on mucous
membranes. Several dust mite allergens have been characterized, including: Derfl and II; and
Derp I and II (Platts-Mills and Chapman, 1987). Cockroaches are nocturnal insects belonging
to the Orthoptera (Mathews, 1989) that inhabit dark environments where food and water are
available. Cockroaches are very prolific, given favorable environmental conditions, and
population pressure eventually drives them into daylight in search of food. They shed body
parts, egg cases, and fecal particles (all of which probably carry allergens), but little is known
about the particles that actually carry the allergens. Two cockroach allergens have been
characterized: Bla g I, and Bla g II; and they are likely a major cause of asthma for some
populations of children. Fragments of gypsy moths and other insects that undergo massive
migrations can also become abundant in ambient air. Sizes, nature, and allergen content of such
particles have not been well studied; but cases of occupational asthma from exposure to insects
(e.g., sewer flies) have been reported.
Other Animal Allergens. It is likely that proteinaceous particles shed from any animal
could cause sensitization if exposure conditions are appropriate. For example, exposure to
proteins aerosolized during seafood processing have caused epidemics of asthma.
7B.1.3 Fungal Aerosols
Fungi are primarily filamentous microorganisms that reproduce and colonize new areas by
airborne spores. Most use complex nonliving organic material for food, require oxygen, and
have temperature optima within the human comfort range. The major structural component of
the cell wall is acetyl-glucosamine polymers (chitin). Cell walls also may contain p-glucans,
waxes, mucopolysaccharides, and many other substances. In the process of degrading organic
material, fungi produce CO2, ethanol, other volatile organic compounds, water, organic acids,
ergosterol, and other metabolites that include many antibiotics and mycotoxins.
Fungi colonize dead organic materials in both indoor and outdoor environments. Some
invade living plant tissue and cause many important plant diseases; some invade living animal
hosts, including people. Fungi are universally present in indoor environments unless specific
efforts are made for their exclusion (i.e., as in clean rooms). Kinds of fungi able to colonize
7B-4
-------
indoor materials are generally those with broad nutritional requirements (e.g., Cladosporium
sphaerospermum), those that can colonize dry environments (e.g., members of the Aspergillus
glaucus group), or organisms that readily degrade cellulose and lignin present in many indoor
materials (e.g., Chaetomium globosum, Stachybotrys atra, Merulius lacrymans). Yeasts (which
are unicellular fungi) and other hydrophilic taxa (e.g., Fusarium, Phialophora) are able to
colonize air/water interfaces. Moisture, in fact, is the most important factor determining indoor
fungal growth, since food sources are ubiquitous (Kendrick, 1992).
Particles that become airborne from fungal growth include spores (the unit of most fungal
exposure); fragments of the filamentous body of the fungus; and fragments of decomposed
substrate material. Fungal spores range from about 1.5 jim to > 100 jim in size and come in
many different shapes, the simplest being smooth spheres and the most complex large
multicellular branching structures. Most fungal spores are near unit density or less. Some
include large air-filled vacuoles. Fungal spores form the largest and most consistently present
component of outdoor bioaerosols. Levels vary seasonally, with lowest levels occurring during
periods of snow. While rain may initially wash large dry spores from the air, these are
immediately replaced by wet (hydrophilic) spores that are released in response to the rain.
Some kinds of spores are widespread in outdoor air (e.g., Cladosporium herbarum,
Alternaria tenuissimd). Others produced by fungi with more fastidious nutritional requirements
are only locally abundant. Typical indoor fungal aerosols are composed of particles penetrating
from outdoors, particles released from active growth on indoor substrates, and reaerosolized
particles that had settled into dust reservoirs. Indoor fungal aerosols are produced by active
forcible discharge of spores; by mechanisms intrinsic to the fungus that "shake" spores from the
growth surface; and, most commonly, by mechanical disturbance (e.g., air movement, vibration).
Allergic rhinitis and asthma are the only commonly reported diseases resulting from fungal
exposures outdoors, and which also commonly occur indoors. The allergens of fungi are
probably digestive enzymes that are released as the spore germinates. Other spore components
(of unknown function) may also be allergenic. Only very few fungal allergens (e.g., Alt a I,
Cla h I, and Aspfl), out of possibly hundreds of thousands, have been characterized.
Allergic fungal sinusitis and allergic bronchopulmonary mycoses occur when fungi
colonize thick mucous in the sinuses or lungs of allergic people. The patterns of incidence of
allergic fungal sinusitis may be explained in part by geographic variability in ambient fungal
7B-5
-------
exposures. This disease is most commonly caused by Bispora, Curvularia, and other dark-
spored fungi. Exposure patterns required for allergic bronchopulmonary mycoses are unknown.
This disease is usually caused by Aspergillus fumigatus. Coccidioidomycoses and
Histoplasmosis are infectious fungal diseases that result from outdoor exposures to Histoplasma
capsulatum (a fungus that contaminates damp soil enriched with bird droppings) and
Coccidioides immitis (a fungus that grows in desert soils). Indoor aerosol-acquired fungal
infections are rare and mostly restricted to immunocompromised people (Rippon, 1988).
Toxic agents produced by fungi include antibiotics, mycotoxins, and some cell-wall
components that have irritant or toxic properties. The antibiotics and mycotoxins are secondary
metabolites produced during fungal digestion of substrate materials, and their presence depends,
in part, on the nature of the substrate. The locations of the toxins in spores or other mycelial
fragments are unknown, as are the dynamics of release in the respiratory tract. Aerosol exposure
to fungal antibiotics in levels sufficient to cause disease is unlikely. Mycotoxicoses have been
reported as case studies from exposure to spores of Stachybotrys atra (Croft et al., 1986), and
epidemiologically for Aspergillus flavus (Baxter et al., 1981).
7B.1.4 Bacterial Aerosols
Most bacteria are unicellular, although some form "pseudo" filaments when cells remain
attached following cell division. The actinomycetes are bacteria that do form filaments and,
in some cases, dry spores designed for aerosol dispersal. Bacteria can be broadly categorized
into two groups based on a response to the Gram stain procedure. The cell walls of Gram
positive (Gram+) bacteria are able to absorb a purple stain; the cell walls of Gram negative
(Gram-) bacteria resist staining and contain endotoxin consisting of proteins, lipids, and
polysaccharides.
Most infectious agents are maintained in diseased hosts. A few, including Legionella
pneumophila, reside in water-filled environmental reservoirs such as water delivery systems,
cooling towers, air conditioners, and lakes, streams, oceans, etc. Infectious agents are often
released from hosts in droplets exhaled from the respiratory tract. Each droplet contains one or
more of the infectious agent, possibly other organisms, and respiratory secretions. Most droplets
are very large and fall quickly. Smaller droplets dry quickly to droplet nuclei, which range from
7B-6
-------
the size of the individual organism (< 1 jim for the smallest bacteria) to clumps of larger
organisms (> 10 jim for larger bacteria).
Environmental-source aerosols are produced by mechanical disturbances that include wind,
rain splash, wave action, and as occurs in air recirculation, in sprays of washes and coolants, and
in humidifiers. Particle sizes from all of these activity cover a wide range from well below 1 |im
to > 50 |im. The thermophilic actinomycetes produce dry aerial spores that require only slight
air movements to stimulate release. Each spore is about 1 jim in diameter.
Whole living bacteria are agents of infectious disease (e.g., tuberculosis, Legionnaires'
disease, etc.). For tuberculosis, a single virulent bacterial cell deposited in the appropriate part
of the lung can cause disease in a host without specific immunity. For Legionnaires' disease, the
number of organisms needed for disease development likely depends on how well the host's
general protective immune system is operating. Some bacteria release antigens that cause
hypersensitivity pneumonitis. The antigens may be enzymes (e.g., Bacillus subtilis enzymes
used in the detergent industry) or may be cell wall components (e.g., endotoxin or glucans).
7B.1.5 Viral Aerosols
Viruses are either RNA or DNA units surrounded by a protein coat that have no intrinsic
mechanism for reproduction, but rather require living cells (whose enzyme systems they utilize
to make new viral particles). Viruses can be crystallized and yet remain able to reproduce and
are often considered intermediates between nonlife and life. Because viruses require living cells
to reproduce, reservoirs for them are almost exclusively living organisms. Viruses, in rare cases,
even survive (but do not reproduce) in environmental reservoirs from which they are
re-aerosolized to cause disease. Hanta virus that causes severe respiratory disease in people
exposed to intense aerosols of infected mouse urine is an example of this. Viral aerosols are
produced when the infected organism coughs, sneezes, or otherwise forces respiratory or other
secretions into the air. The viral particles are coated with secretions from the host and, as in the
case for bacteria, there may be one to many in a single droplet. The size of a single viral particle
is very small (a small fraction of a |im), but infectious droplets more usually occur within a
larger size range (1 to 10 jim). Each kind of virus produces a specific disease, although some of
the diseases present with similar symptoms. Thus, the measles virus produces measles, the
chicken pox viruses produces chicken pox and shingles, etc. Influenza and common colds are
7B-7
-------
produced by a variety of viruses, all of which produce similar (but not necessarily identical)
symptoms.
7B.2 NEWLY AVAILABLE BIOAEROSOLS RESEARCH
Since the 1996 PM AQCD, a number of newly available studies provide interesting new
information pertinent to evaluating potential involvement of bioaerosols in contributing to health
effects associated with exposures to ambient PM. Of much interest are newly published findings
which (a) indicate greater contributions (than previously thought) of bioaerosols to airborne
ambient PM concentrations; (b) improve our understanding of factors and mechanisms affecting
release of some bioaerosol materials into ambient air; and (c) provide evidence indicative of
bioaerosols contributing to ambient PM-related health effects, including contributions made in
combination with other, nonbiological, PM components.
The fate of bioaerosols is dependent on a number of variables: geography, time of day,
moisture levels, air temperature/humidity, wind speed and direction, and seasonal variations in
the latter variables. Once airborne, depending on the particle size, bioaerosols may travel great
distances. As discussed in more detail below, bioaerosols generally represent a rather small
fraction of the measured urban ambient PM mass and are typically present at even lower
concentrations outdoors during cold seasons, when notable ambient PM effects have been
demonstrated (Ren et al., 1999; Kuhn and Ghannoum, 2003). Bioaerosols tend to be in the
coarser fraction of PM; but some bioaerosols (e.g., fungal spores, fragmented pollens,
nonagglomerated bacteria) are found in the fine fraction as well (Meklin et al., 2002a; Schappi,
1999), possibly due to reactions of the biological agents with ambient particles (Schappi et al.,
1999; Oikonen et al., 2003; Behrendt et al., 2001; Ormstad et al., 1998).
For the sake of bringing together information regarding bioaerosols, the following
discussions include new information on bioaerosol sources and factors affecting their dispersal
in ambient air as well as new studies on their health effects. The latter include not only
toxicology studies, but also some studies conducted in occupational settings or results from
epidemiology studies assessing health responses to airborne allergens or biological material.
To the extent that other aspects of air pollution evaluated in these epidemiology studies are
deemed pertinent and important, the results are discussed in Chapter 8. Tables 7B-2 and 7B-3
7B-8
-------
TABLE 7B-2. RESPIRATORY EFFECTS OF POLLEN/FUNGI AND PM EXPOSURES
td
Species, Gender,
Strain, Age, etc.
Humans
Human,
Netherlands
Particle
Pollen
Ambient PM
Pollens:
grass; sorrel;
birch; dock
Exposure
Technique
Ambient,
London
Ambient
(Leiden and
Helmond)
Concentration
Pollen count
increased from
3 7 to 130grains/L
3 h after
thunderstorm
Poaceae (grass)
78 pol. grains/m3
Betula (birch)
69 pol. grains/m3
Quercus (oak)
13 pol. grains/m3
Exposure
Particle Size Duration
Not 2 month study
characterized period
Deaths from
1986 to 1994
evaluated
Particle Effects/Comments
ER visits increases from 2.25
patients/day to 40 patients/day
following thunderstorm. Peak in
pollen levels about 9 h before peak
in ER visits.
Pollen concentrations only weakly
associated with air pollution. Grass
pollen levels were associated with
daily deaths from pneumonia and
COPD. Other pollens (birch, sorrel,
dock) were also positively correlated
with mortality.
Reference
Celenza et al.
(1996)
Brunekreef
et al. (2000)
Human, male and
female, normal
and allergic,
ages 26-29
Ragweed
Bronchoscopic
challenge
Not characterized
Not
characterized
3h
Human, male and
female,
21-49, with
allergic rhinitis,
nonsmoking
DPM
ragweed
Intranasal
spray
0.3mgin200 ul
saline
1,4,8 days
Nonallergic subjects: ragweed had Hastie and
little effect on ciliary activity; acid Peters (2001)
reduced activity
Allergic subjects: slight increase in
albumen and 2-fold increase in
BAL cell number.
Allergic subjects with severe
inflammatory changes had a 12-fold
increase in albumin and 9-fold
increase in BAL cell number
Combined DEP/ragweed challenge Diaz-Sanchez
induced ragweed-specific IgE, but not et al. (1997)
total IgE or IgE-secreting cell
numbers. Also caused a decrease in
IFNy and IL-2 and an increase in
IL-4, IL-5, IL-6, IL-10, and IL-13.
-------
TABLE 7B-2 (cont'd). RESPIRATORY EFFECTS OF POLLEN/FUNGI AND PM EXPOSURES
td
o
Species, Gender,
Strain, Age, etc.
Mice, Male,
BALB/c,
7 weeks old
Mice, Female,
BALB/c,
6-8 weeks old
Human,
asthmatic,
9- 16 years old
Human,
children < 15,
adults,
adults > 59
Human
Chicken tracheal
rings
Particle
DPM
Japanese
cedar pollen
OVA
SPM
DPM
(SRM1650)
PM fungal
spores
Ascomycetes
basidiospore
content in
ambient TSP
and PM10
Mold spores;
tree, grass,
and ragweed
pollen
Filamentous
fungi
Exposure
Technique
Intratracheal
Instillation
Injection into
mouse footpad
Ambient,
(San Diego, CA
area)
Ambient
(Mexico City)
Ambient
(Chicago)
in vitro
Concentration
0.3 mg/mouse
Img
10 LIB
1U f^g
24.8 ±11.1 iig/m3
2461 ± 1307/m3
56-98
78-156
100-207
100-1000
spores/m3
Variable
N/A
Exposure
Particle Size Duration
0.4 urn 3 times with
an interval of
3 weeks
< 2.5 urn 20-26 days
2.5 urn 12 h
10 urn
10 urn 1 year
analysis
7 -month
periods from
1985 through
1989
N/A 24, 48, 96 h
Particle Effects/Comments
IgE antibody and IL-4 production
increased. Slight increase in
IL-3 output.
DEP affects antigen-specific IgE
antibody response by enhanced
IL-4 production.
Adjuvant activity noted on the
production of IgE antibodies to OA.
Inhaler puffs increased by 1 .2 per
1000 fungal spores/m3. Positive
association between asthma severity
and PEFR and total fungal spores.
No significant relationship between
asthma severity and PM10 or pollen
exposure.
Statistically significant increase in
fungal spore exposure-related asthma
hospital admissions in children
(but not adults and seniors) in
Mexico City.
On days that mold spores were
> 1000 spores/m3, death caused by
asthma were 2.16 times greater than
days with spores < 1000; but no
increase in death seen with tree,
grass, or ragweed pollen.
Chloroform-extractable endo- and
exometabolites stopped tracheal
ciliary movement.
Reference
Fujimaki et al.
(1994)
Ormstad et al.
(1998);
Ormstad
(2000)
Delfino et al.
(1996, 1997)
Rosas et al.
(1998)
Targonski
etal. (1995)
Pieckova and
Kunova (2002)
SPM = suspended particulate matter
DPM = diesel particulate matter
OVA = ovalbumin
-------
TABLE 7B-3. RESPIRATORY EFFECTS OF INHALED ENDOTOXIN-LADEN AMBIENT BIOAEROSOLS
Species, Gender,
Strain, Age, etc.
Humans
(pig farmers),
82 symptomatic and
89 asymptomatic
n= 171
Humans (healthy);
32 male, 32 female,
16 to 50 years old
Humans (potato
plant workers),
low (37 male) and
high (20 male)
exposures
Humans (healthy);
5 male, 4 female,
24 to 50 years old
Particle
Dust
Endotoxin
Indoor pool
water spray
Endotoxin
Endotoxin
LPS1
(endotoxin)
Exposure Particle
Technique Concentration Size
Inhalation 2.63 mg/m3 N/A
og=1.3
105 ng/m3
og=1.5
Inhalation N/A 0.1-7.5 um
Inhalation low: 2 1.2 EU/m3 N/A
og= 1.6
high: 55.7 EU/m3
og = 2.1
Inhalation 0.5 ug 1-4 um
5.0 ug MMAD
50 ug
Exposure
Duration
5h/day
average
lifetime
exposure
N/A
8h
30 min
Particle Effects/Comments
Large decline in FEV[ (73 mL/year) and FVC
(55 mL/year) was significantly associated with
estimated long-term average exposure
to endotoxin at 105 ng/m3.
Recurring outbreaks of pool-associated
granulomatous pneumonitis (n = 33); case
patients had higher cumulative work hours.
Analysis indicated increased levels of
endotoxin in pool air and water.
Concentration-related decreased FEVl5 FVC,
and MMEF over the work shift; endotoxin
effects on lung function can be expected above
53 EU/m3 (« 4.5 ng/m3) over 8 h.
Significant decrease in PMN luminal-enhanced
chemiluminescence with 0.5 ug LPS; increase
in blood CRP and PMNs, and increase in
Reference
Vogelzang
etal. (1998)
Rose et al.
(1998)
Zock et al.
(1998)
Michel et al.
(1997)
sputum PMNs, monocytes, and MPO with
5.0 ug LPS; increase in body blood PMNs,
temperature, blood and urine CRP, sputum
PMNs, lymphocytes, monocytes, TNFa,
andECPwithSOugLPS.
Rats (Fischer 344),
8 wks to 22 mo old,
n = 3/group
LPS ' Inhalation 70 EU/m3
(endotoxin)
0.72 urn
og = 1.6
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.
Elder et al.
(2000a,b)
1 LPS = lipopoly saccharide.
-------
summarize salient features of newly available studies of respiratory effects of pollens/fungi and
endotoxins, respectively.
7B.2.1 Atmospheric Levels of Cellulose/Other Plant Debris Markers
Puxbaum and Tenze-Kunit (2003) investigated seasonal variations in atmospheric cellulose
levels (as a "microtracer" for airborne plant debris) in and around Vienna, Austria. The 9-month
average of "free" cellulose concentrations at the downtown site was 0.374 |ig/m3 (reflective of
0.75 |ig/m3 plant debris). Given an annual average for organic carbon (OC) at the downtown site
of 5.7 |ig/m3, plant debris appears to be more than a minor contributor to ambient organic aerosol
at that site. Unexpectedly, size distribution determinations via impactor measurements indicated
that the "free cellulose" (on a mass basis) comprised ~ 0.7% of ambient fine PM (0.1 - 1.5 jim),
forming a "wetable but insoluble part of the accumulation mode aerosol," as noted by Puxbaum
and Tenze-Kunit (2003). They further noted that the cellulose levels at the downtown site
showed maximum concentrations during the fall (probably due to increased biological activity
involving seed production and entrainment of other plant cellulose materials into the air).
Comparison of simultaneous measurements of cellulose at the downtown site to those from a
suburban site indicated that the ambient PM cellulose did not originate in notable amounts from
within the city.
The Puxbaum and Tenze-Kunit (2003) study adds further to a growing database which
points toward plant debris being a significant contributor to organic aerosols present at
continental sites. As discussed by Puxbaum and Tenze-Kunit, Rogge et al. (1993) and Zappoli
et al. (1999) have shown a considerable portion of the organic aerosols not to be soluble in water
or organic solvents, suggesting larger molecular sizes of the insoluble compounds. Also,
Matthias-Maser and Jaenicke (1995) found up to 40% of the number of particles > 0.2 jim (AD)
at a continental site to be of primary "biological origin". Puxbaum and Tenze-Kunit further
noted that Bauer et al. (2002) found fungal spores in the 2.15 to 10 |im fraction of organic
background aerosol at a mountain site to comprise on average, about 6% of the OC mass in the
coarse PM fraction. Also, they noted that the main constituents of the organic aerosol appear
to be humic-like substances (HULIS) that are present in continental aerosol samples at
concentrations (HULIS-carbon) that range from 7 to 24% of the OC mass (Havers et al., 1998;
Zappoli, et al., 1999; Facchini et al., 1999). The macromolecular HULIS materials likely have
7B-12
-------
many origins, e.g., from biomass fires (Facchini et al., 1999) or secondary atmospheric reactions
(Gelencser et al., 2003). It was further noted by Puxbaum and Tenze-Kunit that cellulose is also
contained in pollen at 3 to 7% dry mass (Stanley and Linskens, 1985).
Other new studies evaluated atmospheric levels of levoglucosan (LVG) and other markers
(e.g., palmitic acid, stearic acid) of biomass burning so as to investigate potential inputs of
materials from that source category to ambient PM. One study (Fraser and Lakshaman, 2000),
measuring effects in Texas of biomass fires in Mexico/Central America, found 0.2 to 1.2 |ig/m3
of LVG during episodes resulting from long-range transport of smoke haze. In another study,
Poore (2002) reported on LVG concentrations in PM2 5 samples taken at the Fresno, California
supersite during the year 2000. Highest levels of LVG (up to 4.05 |ig/m3) were found during late
fall/winter months (November-January), whereas LVG concentrations during spring/summer
months were near or below the detection limit of 0.01 |ig/m3. Analogous seasonal patterns
of variations in concentrations of palmitic and stearic acid were also seen for the Fresno
supersite PM2 5 samples. Given that agriculturally-related biomass burning in the Fresno area
is typically completed by the end of October, the elevated LVG levels during fall/winter months
were most likely derived from residential woodsmoke emissions. The same may also be true for
fall/winter increases in palmitic and stearic acid levels, although as noted by Poore (2002), both
of these acids are emitted from a variety of sources, including food production. In any case,
these results appear to be indicative of episodic or more prolonged seasonal increases in plant-
derived bioaerosol materials contributing to ambient PM levels in Texas and California, and by
analogy, other areas of the western United States where air quality is affected by biomass
burning emissions (e.g., from controlled burns on agricultural land, forest fires, or residential
fireplaces/woodstoves).
However, a comprehensive review (Anderson et al., 1992) of a number of studies of
rodents exposed to dry cellulose has shown that exposures in the range of mg/m3 do not cause
adverse effects on reproduction or development, nor do they increase in the incidence of cancer.
Also, one more recent study (Cullen et al., 2000) has demonstrated that inhalation of 1,000 ml
(-1) cellulose fibers in rats for 5 d/week for 3 weeks created only an early inflammatory
response which peaked at day 1 following the start of inhalation. The inflammation, including
levels of BAL TNFa, declined over the next 18 days. Thus, ambient exposures to dry cellulose
materials, per se, would appear to pose relatively low risk for health effects. It remains to be
7B-13
-------
more clearly delineated as the extent to which ambient airborne cellulose materials (especially
under wet conditions) may act as effective carriers of other bioaerosol (e.g., fungi, bacteria,
viruses) or nonbiologic materials and thereby serve to increase health risks associated with them.
7B.2.2 Pollen
With regard to pollen, important new insights are emerging with regard to: (a) factors
influencing the occurrence of asthmatic or other allergic responses to certain types of common,
widespread pollens; and (b) the likelihood that such bioaerosol-related asthma events are
enhanced by the presence in ambient air of other types of non-bioaerosol airborne particles.
More specifically, researchers in several countries have demonstrated links between epidemics
of "thunderstorm asthma" (characterized by notable increases in asthma attacks and upsurges in
hospital visits/admissions for asthma within hours after such storms) and increased levels of
grass pollen allergens among respirable airborne bioaerosol components (Bellomo et al., 1992;
Ong, 1994; Venables et al., 1997; Rosas et al., 1998; Newson et al., 1997; Schappi et al., 1999;
Girgis, et al., 2000).
Anemophilous plants (wind-pollinated plants) produce copious amounts of pollen, making
pollen from these plants the most abundant in the atmosphere and the most important in terms of
human exposure. Typically, exposure to pollen has been thought to only play a role in allergic
rhinitis because they are too large to penetrate into the lower airways. However, in more recent
years, there is evidence which indicates that pollen may in fact be associated with exacerbation
of asthma through the release of pollen allergens small enough in size to penetrate into lower
respiratory airways and/or via the binding of these allergens to other respirable size particles
(Suphioglu et al., 1992; Burge and Rogers, 2000; Knox et al., 1997; Schappi et al., 1999). More
specifically, although intact (unruptured) pollen grains are typically so large (often
> 10 to 20 |im) that, when inhaled, they mainly deposit in upper airways (nasopharyngeal areas),
grass pollen allergens are found in the cytoplasm of the pollen grains (Taylor et al., 1994); and,
upon the rupture of mature pollen grains, they are released as cytoplasmic fragments that
comprise respirable (-0.1 to 5.0 jim) particles (Schappi et al., 1999; Grote et al., 2000; Taylor
et al., 2002).
7B-14
-------
The release of allergens from the pollen grains is moisture dependent (Suphioglu et al.,
1992; Schappi et al., 1997, 1999). Suphioglu et al. (1992) reported the release of a major
allergen (LolplX) from the intracellular starch granules of rye grass when pollen grains were
ruptured during a rain storm. The allergen was small enough (< 3 jim) to penetrate to the lower
airways. The atmospheric concentration of the allergen showed a 50% increase on days
following a rain event. Asthmatic volunteers were exposed by aerosol mask to the starch
granules (volume of 1 mL nebulized, of a 0.34 |ig/m3 solution) or the pollen grain extracts.
Asthmatic volunteers (n = 4) that underwent inhalation challenge showed a typical early
response, described by the authors as a striking bronchial constriction following exposure to the
starch granules. The effect was not noted in volunteers exposed to pollen grain extracts.
Taylor and colleagues (2002) confirmed that the key trigger for rupture of rye grass and
Bermuda grass pollen is pollen grain contact with water, e.g., with the moistening of such pollen
by dew, fog, rainfall, or lawn watering. They also further provided evidence on the specific
sequence of events (and time periods) leading to appearance of the allergen-containing
cytoplasmic material in airborne respirable aerosols. Taylor et al. (2002) reported that, upon
drying within 1 to 6 h after rye grass or Bermuda grass pollen were moistened with water and
grain rupture occurred, allergen-containing cytoplasmic fragment particles were entrained into
the air by blowing air across the grass flowers or shaking them, with many thousands of such
fragments in the 0.1 to 4.7 jim size range (most below 0.4 jim) being collected by a Cascade
impactor. The dispersal of such allergen-laden particles following cycles of wetting and drying
of grass pollen, it was noted by Taylor et al. (a) may occur in response to such disturbances as
wind, lawn mowing, and recreational activities; (b) likely account for marked increases in
asthma attacks after thunderstorms; and (c) may also account for increased asthmatic symptoms
during grass flowering season after any moist weather conditions. Also, more recently, Taylor
et al. (2003) employed analogous experimental wetting/drying procedures, collection and
measurement of wind-released cytoplasmic fragments of birch tree pollen in the 0.03 to 4 |im
size range, and found them to contain Bet v 1 allergens.
Taylor et al. (2002) also highlighted possible bases for interactions between aerosolized
allergen-laden pollen debris and other types of ambient airborne particles. They noted, for
example, that diesel emission particles are a major contributor to urban respirable aerosols mass,
e.g., 18% in Pasadena, CA (Schauer et al., 1996), and have been implicated as a cause of allergic
7B-15
-------
rhinitis and asthma in mice and humans (Nel et al., 1998; Bayram et al., 1998; and Diaz-
Sanchez, et al., 2000). Taylor et al. further noted (a) that fine combustion particles and aerosols
of pollen allergens, because of their small size, may deposit in similar respiratory tract regions;
and (b) that synergistic combinations of allergen-laden pollen debris and PAHs found in fine
combustion aerosols may explain the notable increased prevalence of pollen-induced asthma
during the past 50 years.
Further possibilities exist with regard to possible ways that the copresence of grass pollens
and diesel paniculate matter (or perhaps other airborne particles) may contribute jointly to
enhanced probability of asthma symptoms occurring in susceptible human population groups.
More specifically, the EPA Health Assessment Document for Diesel Engine Exhaust (U.S.
Environmental Protection Agency, 2002) noted that Ormstad et al. (1998) investigated the
potential for DPM (as well as other suspended PM) to act as a carrier for allergens into the
airways and found both Canf 1 (dog) and Bet v 1 (birch pollen) on the surface of airborne PM
collected inside homes. They also found that DPM adhered to polycarbonate filters could bind
both of these allergens as well asFeldl (cat) andDerp 1 (house mite) allergens. The authors
concluded that soot particles in indoor air house dust may act as a carrier for several allergens in
indoor air. The EPA Diesel Document (2002) also noted that Knox et al. (1997) investigated
whether free grass pollen allergen molecules, released from pollen grains by osmotic shock
(Suphioglu et al., 1992) and dispersed in microdroplets of water in aerosols, can bind to DPM
mounted on copper grids in air. Using natural highly purified Lolp 1 (the major grass pollen
allergen), immunogold labeling with specific monoclonal antibodies, and a high-voltage
transmission electron-microscopic imaging technique, Knox et al. found binding of Lolp 1 to
DPM in vitro. They concluded that binding of Lolp 1 with DPM might be a mechanism by
which allergens can become concentrated in air and trigger asthma attacks.
In addition to suggesting that airborne diesel exhaust particles can act as carriers of
biological aerosols producing an enhanced allergic response (Knox et al., 1997; Diaz-Sanchez
et al., 1997; Fujimaki et al., 1994), some studies suggest that allergen carriers (e.g., pollen
grains) may incorporate other atmospheric pollutants that alter the pollen surface, leading to
altered protein and allergen release (Behrendt et al., 1992, 1995, 1997, 2001). Pollen grains
from an industrial region with high PAHs were shown to be agglomerated with airborne
particles. In vitro exposure of grass pollen to particles demonstrated ultrastructural changes at
7B-16
-------
the surface of the pollen and within the protoplasm, such as exocytosis of granular proteinaceous
material and increased allergen release (Behrendt et al., 1997).
Fujimaki et al. (1994) examined the effect of intratracheal instillation of a mixture of diesel
exhaust particles and Japanese cedar pollen on IgE antibody production and lymphokine
production in mice. IgE antibody production and IL-4 production in mediastinal lymph nodes
were significantly increased in mice instilled with the diesel exhaust particles and the cedar
pollen compared with the cedar pollen alone. There was a slight increase seen in IL-2 output.
Measurable levels of birch pollen-specific human IgE were noted in hu-PBL-SCID mice
previously stimulated with birch pollen. When the mice were exposed i.p. to the 25 jig birch
pollen plus 500 jig of diesel exhaust particles, IgE levels were twice as high as those for birch
pollen exposure only. Ormstad et al. (1998) found that Feld 1 (cat), Can/I (dog), Derp 1
(house dust mite) and Bet v 1 (birch pollen) allergens bind with soot particles from diesel
exhaust in the < 2.5 jim size range. When the particle mixture was injected in the footpad of
mice, adjuvant activity was noted on the production of IgE antibodies to ovalbumin (Ormstad,
2000). The authors suggested that it is likely that the soot particles alone were responsible for
some of the adjuvant activity. However, the particles may increase the IgE production to
allergens by modulating the immune response.
Diaz-Sanchez et al. (1997) studied possible synergistic relationships between diesel
exhaust particles (DPM) and ragweed allergen. Inconsistent and low levels of mucosal cytokine
mRNAs were found in ragweed sensitized subjects following intranasal challenge with ragweed
allergen alone. When the subjects were challenged with ragweed allergen and DPM there was a
decrease in Thl-type cytokines (IFN-y and IL-2) expression but an elevated expression of
mRNA for other cytokines (IL-4, IL-5, IL-6, IL-10, IL-13). Ragweed allergen and DPM also
produced a 16-fold increase in ragweed-specific IgE but not total IgE levels or IgE-secreting cell
numbers. Total and specific IgG-4 levels were enhanced , while total IgG levels were not.
Subject were given short ragweed Amb a I allergen, starting at 10 allergen units and increasing in
10-fold units until symptoms were noted. The diesel particles were administered for a total of
0.3 mg in 200 jiL of saline. Clones of deleted switch circular DNA (Se/Sji), representing
switching from ji to Ł from the nasal lavage cells, also were detected (Fujieda et al., 1998).
Brunekreef et al. (2000) suggested that airborne pollen associated with allergic responses
may pose more serious effects than previously thought. They evaluated the relationship between
7B-17
-------
the daily number of deaths in the Netherlands for the period of 1986 to 1994 and air pollution,
meteorological factors, and airborne pollen concentrations (analyzed as categorical variables).
The relationship between mortality and airborne pollen concentration was modeled using
Poisson regression with generalized additive models. The pollen mortality associations were
adjusted for long-term and seasonal trend, influenza morbidity, ambient temperature, humidity,
and indicators for the day of the week and holidays. The average number of daily deaths for the
study period was 332.5 (total), including 141.8 cardiovascular related deaths, 15.8 COPD related
deaths, and 9.8 pneumonia related deaths. Pollen concentrations were only weakly associated
with air pollution and there was no confounding by PM10, black smoke, sulphate and nitrate
aerosols, NO2, SO2, or O3. Poaceae (grass) pollens were associated with daily deaths due to
COPD and pneumonia. Other pollens, especially Betula (birch) and Rumex (sorrel, dock) were
also positively correlated with mortality. Information was not included on whether this
association was with daily deaths due to cardiovascular disease, COPD, and/or pneumonia. The
authors suggested that acute exacerbations of allergic inflammation associated with high pollen
exposures may also precipitate death due to cardiovascular disease, COPD, or pneumonia in
individuals suffering from these disorders.
Rosas et al. (1998) reported an association between asthma hospital admissions and grass
pollen exposure for children, adults, and seniors in Mexico City. The effects were noted for both
the wet (May through October) and dry (November through April) seasons. The number of
hospital admissions increased by a factor of 2 to 3 for children and adults on day when the grass
pollen concentrations were above 20 grains/m3. There was no association between asthma
exacerbation and tree pollen.
An association between asthma and emergency room (ER) visits was reported by Celenza
et al. (1996). During a 2-mo study period, the daily average number of emergency room visits
was 2.25 patients; but, following a thunderstorm, such visits increased to 40. There was a peak
in pollen levels about 9 h before the peak in asthma ER visits. Three hours after the storm, the
pollen count increased from 37 to 130 grains/L. There was no evidence that vehicle exhaust
pollutants were related to the increase in asthma ER visits.
Hastie and Peters (2001) studied the effect of in vivo ragweed allergen exposure (via
bronchoscopic segmented ragweed challenge) on ciliary activity of bronchial epithelial cells
harvested 24 h after challenge in human volunteers and allergic subjects with severe
7B-18
-------
inflammatory response. Nonallergic subjects with mild inflammatory response showed a
minimal ragweed allergen effect on ciliary activity, a slight increase in bronchoalveolar cells,
and a nonsignificant increase in albumin concentration. Allergic subjects with mild
inflammatory changes showed slight but significant increase in albumin concentration and a
2-fold increase in bronchoalveolar cell concentration. The allergic subjects with severe
inflammatory changes had a 12-fold increase in albumin concentration and a 9-fold increase in
bronchoalveolar cell concentration.
Delfmo et al. (1997,1996) conducted several studies evaluating the association between
asthma incidence and exposure to various air pollutants and fungal spores and pollen. There was
an association between exposure to air pollutants and fungal spores and symptom severity as
measured by inhaler usage. Inhaler puffs increased by 1.1/100 ppb O3 (14 to 87 ppb; 12-h
daytime average) and by up to 1.2/1,000 fungal spores/m3 (648 to 7,512 spores/m3) depending on
the species. Delfmo et al. (1997) found an association between asthma severity (asthma
symptom scores and inhaler use) and peak expiratory flow rate (PEFR) and total fungal spores.
Symptom severity was more strongly associated with basidiospore concentrations, especially
during the period of sporulation. There was no detected association between O3 exposure and
asthma severity in the Delfmo et al. (1996) study, possibly due to O3 measurement problems
(as suggested by the authors). There was also no significant relationship between asthma
severity and PM10 or pollen exposure, but, their concentrations during the study period were low,
26 |ig/m3 and 216 grains/m3, respectively.
In summary, newly available information indicates release of allergen-laden material from
pollen-spores in respirable-sized aerosols and suggests possible ways by which binding of such
material to other airborne particles (e.g., DPM) may concentrate such allergens in ambient air or,
once inhaled, jointly exacerbate allergic reactions in susceptible human populations. It should
also be noted that pollen itself may act as a carrier for other allergenic materials. Spiewak et al.
(1996a) found Gram-bacteria and endotoxin on the surface of pollens; and Spiewak et al.
(1996b) found concentrations of several immunotoxicant allergens (Gram+ and Gram- bacteria,
thermophilic actinomycetes, fungi) to range from 0 to 10,000 cfu/g of pollen from several
grasses or trees in Poland.
7B-19
-------
7B.2.3 Fungi and Their Byproducts
Fungal spores are known to cause allergic diseases. All fungi may be allergenic depending
on the dose. Once an individual is sensitized to a given fungus, small concentrations can trigger
an asthma attack or some other allergic response (Yang and Johanning, 2002). Unlike fungal-
induced allergic responses, fungal toxic inflammatory responses depend on concentrations of
airborne fungi or fungal fragments and are similar for most individuals. Fungi concentrations
are usually higher in the indoor environment; but, as noted earlier, outdoor airborne spores are
often the source of indoor fungal contamination (Koch et al., 2000).
Fungi produce a variety of byproducts, including mycotoxins and volatile organic
compounds. Mycotoxins have low volatility, making inhalation of volatile mycotoxins unlikely.
However, mycotoxins are an integral part of the fungus. Volatile organic compounds or VOCs
(derivatives of alcohols, ketones, hydrocarbons, and aromatics) are produced when the fungi are
actively growing. Concentrations of these VOCs are generally quite low and relationships
between exposure and health effects are unclear (Yang and Johanning, 2002).
A number of studies have suggested a relationship between exposure to fungi and their
byproducts in respiratory illnesses and immune pathology (Hodgson et al., 1998; Tuomi et al.,
2000; Yang and Johanning, 2002). Some fungal byproducts have been shown to stop ciliary
activity in vitro and may act to produce general intoxication of macroorganisms through the lung
tissue or to enhance bacterial or viral infection (Pieckova and Kunova, 2002; Yang and
Johanning, 2002). Larsen et al. (1996) showed nonimmunological histamine release from
leukocytes exposed to a suspension of fungal spores and hyphal fragments and suggested that the
fungal suspension possessed at least two histamine releasing components; an energy-dependent
release process and a cytotoxic release process.
Exposures to airborne fungal spores have been shown to be associated with increased
asthma attacks and asthma-related deaths. For example, airborne fungal concentrations
of > 1000 spores/m3 were reportedly associated with asthma deaths among 5 to 34 year olds
in Chicago between 1985 and 1989 (Targonski et al., 1995). The odds of death occurring on
days with airborne fungal concentrations of > 1000 spores/m3 were 2.16 times higher than other
days. Logistic regression analysis was used to compare the probability of deaths caused by
asthma as the result of tree, grass, and ragweed pollen and fungal spores. Fungal spores were
7B-20
-------
counted as a single group. Asthma deaths were obtained from death certificates. The deaths
were also related to personal, social, and medical access factors.
In a study by Rosas et al. (1998), there was a statistically significant increase in fungal
spore exposure-related asthma hospital admissions in children in Mexico City that was not seen
in adults and seniors. The highest spore (ascomycetes and basidiospore) levels were associated
with a 2- to 3-fold increase in hospital admissions per day. Ascomycetes and basidiospore
concentrations ranged from < 100 to 207 spores/m3 and from < 100 to > 1000 spores/m3,
respectively. There was an association with hospital admissions during both the wet and dry
season. There was no strong statistical association between asthma admissions and NO2 (mean:
0.102 and 0.164 ppm),O3 (mean: 0.204 and 0.187 ppm), SO2 (mean: 0.074 and 0.081 ppm),
TSP (mean: 78 and 156 |ig/m3) and PM10 (mean: 56 and 98 |ig/m3) concentrations during either
the wet or dry seasons, respectively.
Another example of serious health effects associated with exposures to ambient airborne
fungal material is the occurrence of coccidioidomycosis (or "Valley Fever") caused specifically
by the fungus Coccidioides immitis — a fungus which grows in soils in areas of low rainfall,
high summer temperatures, moderate winter temperatures, and which is emitted with the
airborne suspension or resuspension of the soil that supports it (i.e., in the southwestern USA,
parts of California, and parts of Mexico, Central and South America). At least 50,000 new
coccidioidal infections are estimated to occur each year in endemic areas of the United States,
many producing subclinical infections typified by lower respiratory tract signs and symptoms
largely indistinguishable from flu-like illnesses or other types of infections due to other
etiologies (Galgiani and Appel, 1990). However, in some patients, especially those > 65 years
old or immune-suppressed (e.g., due to HIV), more protracted serious outcomes can occur, e.g.,
chronic pneumonia or, rarely, more widespread systemic infection in lymph nodes, skin, bones,
meninges, etc., and can be life-threatening (Galgiani and Appel, 1990). Outbreaks of "Valley
Fever" due to exposure to airborne dust containing large amounts of Coccidioides-contaminated
soil materials have been documented both in Arizona during 1990 to 1995 (Morbidity and
Mortality Weekly Report, 1996) and among individuals exposed to dust clouds raised by
landslides in the San Joaquin Valley set off by the January 1994 Northridge, California
earthquake (Schneider et al., 1997). During dry conditions encountered in desert or other
endemic areas during drought periods, both natural dust storms and dust-generating human
7B-21
-------
agricultural activities and off-road vehicle use that disturbs the soil can reasonably be projected
as being likely to increase Coccidioides immitis infection risk.
Several newly-published studies have evaluated levels of fungi or their viable propagules
in ambient (outdoor) and/or indoor air in various areas of the U.S. or other countries in Europe or
East Asia. In an extensive 22-month study, Cooley et al. (1998) investigated the types of fungi
found in indoor and outdoor air at 48 schools in U.S. areas located along the Atlantic seaboard
and Gulf of Mexico. Five fungal genera consistently found in outdoor air comprised > 95% of
the outdoor air fungi detected: Cladosporium (81.5%); Penicillium (5.2%); Chrysosporium
(4.9%); Alternaria (2.8%); and Aspergillus (1.1%). An average of-700 colony-forming units
(CFU)/m3 of Cladosporium fungi were found in outdoor air (about 3 times that found indoors);
whereas relatively low concentrations of Penicillium (-30 CFU/m3) and the other species
(ranging from < 5 to -40 CFU/m3) were found in ambient air (compared to indoors). Penicillium
was most consistently found to be elevated in indoor "complaint areas", the growth of this rather
ubiquitous species being optimized between 10 to 25 °C and it predominating in complaint areas
with a wide range (23 to 67%) of relative humidity. Cooley et al. noted: (a) that Penicillium
spores are small (1 to 5 jim) and capable of entering the lower respiratory tract; and (b) that
bronchial challenges with Penicillium species spores cause immediate and delayed-type asthma
in sensitized subjects (Licorish et al., 1985).
In a detailed study of the nature and variation of fungi inside and outside homes in the
greater New Haven, CT area, Ren et al. (1999) found that fungi in living room, bedroom, and
outdoor air varied across seasons but did not differ seasonally in basement air. They reported
that Claudosporium spp. dominated both indoor and outdoor air during summer months, whereas
Penicillium and Aspergillus were dominant in indoor air in winter, but neither were dominant in
outdoor air during any season. Ren et al. further noted: (a) the fungi isolated in their study are
broadly the same as those found in European studies (Beaumont et al., 1984, 1985; Verhoeff
et al., 1988; Hunter and Lea, 1994); (b) the seasonal trend found by them for fungal propagules
indoors and outdoors were generally comparable with those reported by Hunter and Lea (1994)
for British homes, i.e., lowest in winter, increasing in spring, reaching the maximum in summer,
and decreasing in fall; (c) their results support current concepts that outdoor air may affect
cultural fungal propagules indoors, but the presence of cultural molds in indoor air may not
always reflect the presence of such molds in outdoor air, especially in problem indoor
7B-22
-------
environments; and (d) no associations were found between fungal types and their concentrations
in dust and in air, suggesting that types of fungi and concentrations measured in housedust do
not necessarily reflect those in indoor air, with air samples likely providing a more direct and
better measure of inhalation exposure to fungi. Lastly, Ren et al. (1999) noted that: 50% of the
342 air samples taken during the 1996-1997 study period had < 575 CFU/m3 total cultural fungal
propagules; 97% < 100 CFU/m3 of Alternaria; < 28% > 50 CFU/m3 of Aspergillus; and -90%
< 250 CFU/m3 of Penicillium; and none had Cladosporium spp. over the 3000 CFU/m3 level set
as an allergic threshold by Gravesen (1979).
Koch et al. (2000) obtained data on fungi concentrations in a study that evaluated if
differences in types of seasonal variations in concentrations of fungi in indoor and/or outdoor air
occur and could perhaps account for lower prevalence of allergies and asthma in western than in
eastern Germany. During 1995 to 1997, 405 homes in Hamburg (West) and Erfurt (East)
Germany were visited twice and samples of settled dust taken by vacuuming from carpets in the
living room. No significant differences were found between the two cities for total genera or
single fungi species (Alternaria, Aspergillus, Cladosporium, and Penicillium) with regard to
concentrations of viable fungi detected in settled housedust. Similar seasonal variations were
observed for outdoor air and indoor dust, i.e., with a late summer peak detected in outdoor air
(-2400 CFU/m3 viable fungi in August) and a parallel peak in such concentrations in housedust.
Koch et al. also noted: (a) that recent studies indicate that outdoor air spora influence the
presence of fungi in indoor environments, but indoor air levels of fungi in indoor environments
do not simply reflect the presence of fungi or spora in outdoor air; and (b) that the genera
commonly isolated in housedust (e.g., Cladosporium., Penicillium., Alternaria, Aspergillus)
reflect their relative occurrence in outdoor spore counts.
Takahashi (1997) evaluated fungal types and concentrations in indoor and outdoor air in
Yokahama, Japan and found the number of outdoor total fungal colony-forming units to vary
from < 13 to 2750 CFU/m3. Cladosporium spp. again was found to predominate in outdoor air,
followed by Alternaria spp. and Penicillium spp., with fungal concentrations peaking in
September. Outdoor fungal concentrations were significantly correlated with maximum,
minimum, and average temperature of the day, as well as average wind velocity of the day,
relative humidity, and precipitation for the month. The ranges of concentrations of fungi in
outdoor air were reported by Takahashi to be the same as reported for many European,
7B-23
-------
North American countries, and Israel — with most showing peak levels during the summer and
early fall (July to October) and lowest means during winter months (January to February).
In another East Asia study, Su et al. (2001) compared concentrations of airborne fungi,
endotoxin, and housedust mite allergens in the homes of asthmatic and nonasthmatic children in
southern Taiwan, where temperature and relative humidity are high throughout the year.
With regard to fungi, their results paralleled those of other studies noted above in many respects,
except for some differences in seasonal variations — not too surprisingly given the more
constant warm temperature/high humidity conditions in this study area. The most predominant
indoor genera were Cladosporium, Aspergillus, Penicillium, Alternaria, and yeast.
Cladosporium ranked highest, it being in -85% of the colonies from indoor samples and its
highest CFU/m3 concentration in winter and other seasonal variation patterns also applying for
the other types of fungi. Outdoor air Cladosporium levels were significantly correlated with
indoor air values during all seasons; and the indoor/outdoor concentrations for the other fungi
were also positively correlated during the spring. This suggests that outdoor levels of fungi
and/or their spores are important determinants of indoor air levels of fungi in southern Taiwan.
7B.2.4 Endotoxins
Endotoxins and lipopolysaccharides (LPS; chemically purified version of endotoxin) are
present in the outer cell membrane of all Gram-negative (Gram-) bacteria. Endotoxins are toxic
to most mammals. When released into the blood stream, it is thought that endotoxins/LPS
interact with receptors on monocytes and macrophages and other types of receptors on
endothelial cells, triggering the production of cytokines, which in turn stimulate production of
prostaglandins and leukotrienes, arachidonic acid metabolites (e.g., prostacyclin and
thromboxane A2, and nitric oxide). These mediators can induce physiological changes, e.g.,
inflammation, smooth muscle constriction, and vasodilatation (Young et al., 1998).
Some of the more recent inhalation studies on endotoxin exposure are summarized in
Table 7B-3. In vitro studies on particle-associated endotoxin are discussed in Section 7.5.2.2.
Heederik et al. (2000) noted that animal feces and plant materials contaminated with bacteria
contribute most to organic dust-related endotoxin exposure. Although there is strong evidence
that inhaled endotoxin plays a major role in the toxic effects of bioaerosols encountered in the
work place (Castellan et al., 1984, 1987; Rose et al., 1998; Vogelzang et al., 1998; Zock et al.,
7B-24
-------
1998), it is not clear as to what extent typical ambient concentrations of endotoxin are sufficient
to produce toxic pulmonary or systemic effects in healthy or compromised individuals.
Several new occupational exposure studies have yielded potentially useful information for
estimating exposure-response relationships for health effects associated with exposure to
airborne endotoxin. For example, Vogelzang et al. (1998) evaluated exposure-response
relationships for lung function decline in relation to endotoxin exposure of pig farmers in
the Netherlands. Long-term average exposure to endotoxin and dust was evaluated via personal
monitoring during summer and winter for a cohort of 171 pig farmers over a three-year period.
Mean age at start was 39.6 years and mean number of years worked in pig farming was
16.7 years. Linear regression analyses were used to analyze relationships between declines in
FEVj or FVC (based on measures taken in the first or third years of the studies) and dust
concentrations or endotoxin levels in the inhalable dust. Statistically significant (p < .05)
associations (correcting for age, baseline values, and smoking) were found by regression
analysis between estimated long-term average exposure (typically > 5 h/day) to endotoxin
(105 ng/m3) and annual decline in FEVLO (73 mL/year) and FVC (55 mL/year). The FVC, but
not the FEVj 0, declines were also significantly correlated with inhalable dust concentrations
(long-term average = 2.6 mg/m3). The FEVLO annual average decline is large in relation to the
expected age-related decline of 29 mL/year but equal to that of 73 mL/year reported by Iversen
et al. (1994) based on a 5-year study of farmers. The least exposed pig farmers in the Vogelzang
et al. study showed an average FEVLO decline similar to the expected age-related decline,
whereas the predicted decline for the most exposed pig farmers ranged up to 100 mL/year. The
authors noted that their results support the selection of the lower of two proposed (Clark, 1986;
Palchak et al., 1988) occupational exposure threshold levels of 30 or 100 ng/m3 for airborne
endotoxin.
Some health effects have been reported for occupational exposure to complex aerosols
containing endotoxin at concentrations likely more relevant to ambient levels. Zock et al. (1998)
reported a decline in FEVj (-3%) across a shift in a potato processing plant with up to
56 endotoxin units (EU)/m3 in the air. Rose et al. (1998) reported a high incidence (65%) of
BAL lymphocytes in lifeguards working at a swimming pool where endotoxin levels in the air
were on the order of 28 EU/m3. Although these latter two studies may point towards possible
pulmonary changes at low concentrations (25 to 50 EU/m3) of airborne endotoxin, it is not
7B-25
-------
possible to rule out the contribution to observed effects by other agents present in the complex
airborne organic aerosols in the occupational settings studied.
In another European study, Heinrich et al. (2003) recently carried out temporal-spatial
analyses of endotoxin in fine (PM25) and coarse (PM10_25) particle mass of ambient aerosols from
two East German towns about 80 km apart. The authors noted that one town, Hettstedt, showed
consistently higher prevalence of hay fever and strong allergic sensitization for children than the
prevalence rates seen in the other town, Zerbst, even into the late 1990's when levels of ambient
air pollutants (TSP, SO2) had converged in areas earlier differing in such air pollution levels
(Heinrich et al., 2002a,b). From January to June 2002, weekly PM25 and PM10_25 samples were
taken by dichotomous samplers in each of the two towns and analyzed for endotoxin in the
collected ambient PM. The arithmetic mean for the PM25 sample mass average 10.2 and
12.4 |ig/m3 for Hettstedt and Zerbst, respectively; and PM10_25 sample mass 6.1 and 6.8 |ig/m3,
respectively. Comparable ranges for Hettstedt and Zerbst were 0.3 to 25.8 and 4.2 to 26.3 |ig/m3
for PM25 and 1.2 to 10.6 and 3.0 to 10.7 |ig/m3 for PM10_25. Mass levels for both particle size
fractions showed notable week-to-week fluctuations (mostly closely parallel for both towns),
with weekly means in each town being highest in late March/early April. Airborne endotoxin
levels for both towns showed strong seasonality in parallel patterns for both the fine and the
coarse particle fractions, with endotoxin mass concentrations generally being low during late
winter/early spring in comparison to their generally increasing from late April to highest points
seen in early June (except for a brief episode of elevated endotoxin in Hettsted fine PM seen in
late January/early February). Fine PM endotoxin mass concentrations for Hettstedt (1.2 EU/mg3
arith. mean) were not significantly different from such concentrations for Zerbst (1.1 EU/mg3
arith. mean), but endotoxin levels expressed per mg dust were significantly higher in Zerbst,
suggesting that there may be a higher biogenic content or more bioactive particles in the Zerbst
fine PM fraction. The endotoxin levels in the coarse fraction were about 10 times those in the
fine fraction, whether expressed in EU/mg dust or EU/m3 air and did not significantly differ
between the two towns. The range of endotoxin levels for Hettsted were 0.2 to 3.6 EU/mg dust
and 0.002 to 0.21 EU/m3 air for PM2 5 versus 4.0 to 25.2 EU/mg dust and 0.01 to 0.24 EU/m3 air
for PM10_2 5. The comparable concentrations for Zerbst were 0.2 to 4.3 EU/mg dust and 0.004 to
0.031 EU/m3 for PM2 5 versus 3.1 to 24.2 EU/mg dust and 0.02-0.17 EU/m3 air for PM10.2 5. The
authors concluded that, given the generally similar levels and patterns in seasonal variations of
7B-26
-------
endotoxin concentrations in Hettstedt and Zerbst, it was unlikely that differential exposures to
endotoxin could explain differences in hay fever or allergic reaction prevalence between the two
towns.
The levels of endotoxin concentrations found in Hettsted and Zerbst are similar to those
reported for other ambient or rural aerosols and dusts, with those in coarse PM fractions
typically exceeding those in fine fractions, as noted by Heinrich et al. (2003). They also noted
that measurements in livestock buildings (poultry, pig, cattle) often show endotoxin levels up to
several thousand EU/mg dust, with levels in the inhalable PM10 fraction being higher by
-10-fold than in the fine PM. The finding of notably higher concentrations and absolute mass
amounts of endotoxin in coarse-mode particle samples versus fine particle samples thus appears
to hold, in general, across a number of geographic areas and for both occupational and
environmental situations. The authors also noted the seasonal variation observed in their study
with increased airborne levels of endotoxin in May and June apparently following increased
growth of fungi, other plants, and presumably of microbes due to increasing outdoor spring
temperatures under moderate climatic conditions in Germany. They also noted that increased
levels of plant-related materials and leaf surfaces (Rylander, 2002), as well as pollen surfaces
(Spiewak et al., 1996a), may provide additional sources of growth of Gram- bacteria (from
which endotoxin is derived). The seasonal variation in endotoxin concentrations observed by
Heinrich et al. appear to parallel those seen in other studies for ambient airborne endotoxin
levels (i.e., lower in winter and high during warmer weather in late spring/summer).
Park et al. (2000) evaluated endotoxin levels in indoor dust of 20 homes, indoor air of
15 homes, and outdoor air at two locations in the Boston, MA, area. Endotoxin levels in indoor
dust (from the bed and bedroom/kitchen floors) were not significantly associated with indoor
airborne endotoxin concentrations. The airborne endotoxin levels were, however, significantly
associated with absolute humidity; and a significant seasonal effect for kitchen dust
(spring > fall) and indoor airborne endotoxin (spring > winter) was seen, as was a significant
seasonal pattern for outdoor airborne endotoxin (summer > winter). The authors indicated that,
overall, the indoor airborne endotoxin levels (geometric mean = 0.64 EU/m3) were higher than
outdoor concentrations (geometric mean = 0.46 EU/m3); but seasonal variations were evident,
in that indoor airborne endotoxin levels were generally higher than outdoor airborne endotoxin
levels during September-April and lower than outdoor levels during late spring/summer
7B-27
-------
(May-August). Outdoor airborne endotoxin levels showed significant seasonality, varying by
more than 4-fold across seasons, with decreases in outdoor levels beginning at the end of
summer/early fall and remaining at lowest levels during winter before starting to increase again
with onset of the growing season in late spring. The authors noted that this pattern is consistent
with data suggesting that outdoor Gram- bacteria (and thus airborne) endotoxins are shed from
leaves of growing plants (Edmonds, 1979; Andrews, 1992). Further, the overall mean outdoor
airborne endotoxin levels at an urban sampling location (geometric mean = 0.51 EU/m3)
were somewhat (but not significantly) higher than at a suburban location (geometric mean =
0.39 EU/m3).
Thorn and Rylander (1998a) examined the effect of endotoxin inhalation on inflammatory
responses in 21 healthy subjects from 20 to 30 years old. All subjects were known smokers,
currently did not have a respiratory infection, no self-reported allergies or chronic bronchitis, and
no physician diagnosed asthma. Subjects were examined before exposure to up to 40 jig LPS.
The LPS was suspended in saline, aerosolized, and then delivered to the subject by a nebulizer
adjusted to give 4 jil per nebulizer dose. The subjects inhaled 20 puffs of LPS at a concentration
of 500 |ig/mL, for a total of 40 jig of inhaled LPS. Cell counts, eosinophilic cationic protein
(ECP), and myeloperoxidase (MPO) were monitored in the blood and sputum before and 24 h
following exposure. Following LPS inhalation, MPO was significantly increased in both the
blood and sputum and ECP was increased, but only significantly so in sputum. The ratio of
MPO and neutrophils was significantly decreased in blood and sputum. Spirometric testing
demonstrated a significant decrease in FEVj and FVC values following LPS inhalation. Some
subjects experienced symptoms (throat irritation, dry cough, breathlessness, unusual tiredness,
headache, and heaviness in the head) that developed 4 to 6 h after exposure and persisted for
6 to 8 h.
Michel et al. (1997) examined the dose-response relationships for effects of inhaled LPS
(the purified derivative of endotoxin) in normal healthy volunteers exposed to 0, 0.5, 5, and
50 jig of LPS solubilized in sterile saline administered via a jet nebulizer. The nebulizer
produces a calibrated aerosol consisting of heterodispersed droplets in the range of 1 to 4 jim
(MMAD). Inhalation of 5 jig of LPS resulted in significantly increased blood c-reactive protein
(CRP) and PMNs, as well as PMNs and monocytes in sputum. Inhalation of 50 jig LPS
significantly increased body temperature, blood PMNs, blood and urine CRP and sputum PMNs,
7B-28
-------
monocytes, and lymphocytes. At the higher concentration, a slight (3%) but nonsignificant
decrease in FEVj was also seen.
Other controlled exposure studies of laboratory animals (rat) by Elder et al. (2000a,b)
indicate that priming of the respiratory tract by inhaled endotoxin increases the effect of inhaled
ultrafme surrogate particles and ozone (as discussed in more detail in Section 7.6).
In vitro studies of potential endotoxin contributions to toxic effect of ambient PM are
discussed in Section 7.4.2.
7B.2.5 (1 - 3)-p-D-Glucan
Studies from different countries have reported relationships between damp/humid indoor
environments and various symptoms in both adults and children (Meklin et al., 2002b). Such
symptoms consist of eye, nose, and throat irritation, dry cough, headache, tiredness, and
sometime skin problems. Fungi and their byproducts (discussed above) and bacteria commonly
present in damp/humid indoor environments contain several substances that have known
inflammatory properties. Of the substances associated with these symptoms, (1^3)-p-D-glucan,
a polyglucose compound in the cell walls of fungi, certain Gram+ bacteria, and plants, has begun
to be accorded increasing attention.
The (1 -» 3)-p~D-glucan can induce several biological responses in vertebrates, including
stimulation of the reticulo-endothelial system, activation of neutrophils, macrophages, and
complement, and possibly activation of eosinophils. T-lymphocyte activation and proliferation
have been reported in experimental animals (Heederik et al., 2000). Rylander (1996) suggested
that an acute exposure to(l - 3)-p~D-glucan can produce symptoms of airway inflammation in
normal subjects without a history of airway reactivity after exposing subjects to 210 ± 147 ng/m3
(1 -» 3)-p~D-glucan for 3 separate 4 h sessions 5 to 8 days apart. Exposure to (1 -» 3)-p~D-glucan
alone did not significantly impact FEVj values; but there was a slight decrease in FEVj values
following administration of the two highest doses of methacholine (MCh). Methacholine was
administered in increasing doses in 3 min intervals for a total of 1.25 mg. Forced vital capacity
(FVC) and FEVj/FVC were also unchanged following (1 - 3)-p~D-glucan exposure and MCh
challenge. There was a significant, negative correlation between MCh-induced decrease in FEVj
values and the intensity of throat irritation after 1 h exposure. The intensity of nasal irritation
and stuffy nose and throat irritation was increased at 1 and 4 h. Dry cough, cough with phlegm,
7B-29
-------
chest tightness and wheezy chest was not affected. No effects on airway responsiveness or
inflammatory symptoms were noted in subjects exposed to endotoxins (9.9 ng/m3) under the
same exposure conditions.
Thorn and Rylander (1998b) examined the relationship between exposure to airborne
(1 - 3)-p-D-glucan and airways inflammation. The study was conducted on a group of
75 houses in Gothenburg, Sweden where there had been numerous complaints about dampness
and respiratory symptoms, fatigue, and mold odors. Measurements of (1 - 3)-p~D-glucan and
endotoxins in airborne dust were made with Limulus lysates. Study participants included
67 females and 62 males 18 to 83 years old and included 34 smokers and 9 physician-diagnosed
asthmatics. The average number of years the subjects lived in their house was 18 years (range
2 to 36 years). Study participants provided questionnaire information for assessment of organic
dust-induced effects. The questionnaire inquired about existing diseases states; occupation;
length of time the subject had lived in the house; the presence of pets; and the occurrence of
cough (dry or with phlegm); shortness of breath; nose, throat, and eye irritation; nasal and chest
congestion; and joint and muscle pains, headache, fatigue, and dermal disorders. Other
questions addressed subjective airway reactivity, chronic bronchitis, asthma, and episodes of
fever and influenza-like symptoms gone the next day. Chronic bronchitis was defined as a
cough with sputum for at least 3 months a year for a period of at least 2 years. Asthma was
defined as physician-diagnosed asthma. Spirometry was performed on test subjects to exclude
subjects with less than 70% of predicted values in FEVj and/or FEVj/FVC. Airway
responsiveness was assessed using MCh for a total of 1.2 mg MCh, administered in increasing
doses at 3-min intervals. Serum eosinophilic cationic protein (ECP), myeloperoxidase (MPO),
and C-reactive protein (CRP) were measured. Atopy was determined using the Phadiatop test to
measure the concentration of specific IgE antibodies against airborne allergens.
No detectable levels of endotoxin were found in the homes, but (1 -» 3)-p~D-glucan levels
ranged from 0 to 19 ng/m3. Of 75 homes studied, 20 had (1 -» 3)-p~D-glucan concentrations
below 1 ng/m3 and 13 homes had levels above 6 ng/m3. Twenty-four subjects had positive
Phadioatop test; but there was no significant correlation between exposure and atopy. However,
when evaluated by age, there was a significantly larger number of atopic subjects in the
> 65-year-old group exposed to > 3 ng/m3 (1 - 3)-p~D-glucan. There was a significant inverse
correlation between baseline FEVj and number of years the subjects lived in the house when
7B-30
-------
controlled for age, gender, cigarette smoking status, asthma, atopy, and pets among male
subjects < 65 years old that was not seen in the female subjects < 65 years old and in > 65 years
old subjects. The relationship was present only for those male subjects exposed to > 1 ng/m3
(1 -» 3)-p-D-glucan. Atopic subjects exposed to > 1 ng/m3(l -> 3)-p~D-glucan had significantly
higher serum MPO. Serum ECP and CRP were also higher in these subjects but not
significantly so.
Douwes et al. (2000) examined the relationship between exposure to (1 - 3)-p~D-glucan
and endotoxins and peak expiratory flow (PEF) in children (ages 7 to 11 years) with and without
chronic respiratory symptoms. The children were monitored twice a day for PEF variability.
House dust samples from living room and bedroom floors and the children's mattresses were
taken during the PEF monitoring period. As indicated by linear regression analysis (adjusting
for dust mite allergen levels, the presence of pets, and the type of flooring in the home), peak
expiratory flow variability in the children with chronic respiratory symptoms was strongly
associated with (1^3)-p-D-glucan levels in dust from living room floors when expressed in
micrograms per square meter. The association was strongest for atopic children with asthma.
7B.3 SUMMARY
Bioaerosols, from sources such as plants, fungi, and microorganisms, range in size from
0.01 to |im to > 20 |im. They comprise a small fraction of ambient PM, but have been shown to
contribute to health effects associated with PM exposure.
Pollen from flowering plants, trees and grasses, deposits in upper airways to induce allergic
rhinitis. Allergen-containing cytoplasmic fragments from ruptured pollen grains can enter the
deep lung, where they can exacerbate asthma. Synergistic interactions between pollen debris
and other ambient PM (e.g., the poly cyclic hydrocarbon component of DE) are thought to be a
mechanism that may explain the increased incidence of asthma morbidity and mortality.
Fungal spores and other fungal materials are the largest and most consistently present
outdoor bioaerosol. Some airborne fungal materials cause allergic rhinitis and asthma, which are
highly dependent on seasonal variations in airborne fungi concentrations (being highest in
spring/summer and lowest in winter). Exposures have been linked to asthma hospitalization and
death. Exposures to airborne dust containing elevated concentrations of a soil-dwelling fungus
7B-31
-------
common to dry areas of central California and certain desert areas of the southwestern United
States have been linked to outbreaks of "Valley Fever", a respiratory infection that can be
potentially deadly (especially for those > 65 years old and immune suppressed persons).
Human handling and burning of plant material contributes to increased airborne
bioaerosols, some of which have been shown to contribute to human health effects.
Animals and insects produce bioaerosols capable of producing hypersensitivity diseases.
Most notably, exposure to dust mite and cockroach material has been linked to sensitization in
children.
Also, bacteria and viruses are significant bioaerosols. Much of the toxicity of bacteria is
due to the endotoxins present in the outer cell membrane, which trigger production of cytokines
and a cascade of inflammation. Concentrations of endotoxins are seasonal (highest in warm
months — lowest in cold months), and are typically associated more with coarse-mode than with
fine-mode particles. Another component, (1^3)-p-D-glucan, of cell walls of fungi, certain
bacteria, and plants, has also been shown to cause respiratory inflammation.
7B-32
-------
REFERENCES
Alberts, W. M; Brooks, S. M. (1992) Advances in occupational asthma. Clin. Chest Med. 13: 281-302.
Anderson, R. L.; Owens, J. W.; Tinims, C. W. (1992) The toxicity of purified cellulose in studies with laboratory
animals. Cancer Lett. 63: 83-92.
Andrews, J. H.; Hirano, S. S., eds. (1992) Microbial ecology of leaves. New York, NY: Springer-Verlag.
Arlian, L. G. (1989) Biology and ecology of house dust mites, Dermatophagoides spp. and Euroglyphus spp.
Immunol. Allergy Clin. North Am. 9: 339-356.
Bauer, H.; Kasper-Giebl, A.; Loflund, M.; Giebl, H.; Hitzenberger, R.; Zibuschka, F.; Puxbaum, H. (2002) The
contribution of bacteria and fungal spores to the organic carbon content of cloud water, precipitation and
aerosols. Atmos. Res. 64: 109-119.
Baxter, C. S.; Wey, H. E.; Burg, W. R. (1981) A prospective analysis of the potential risk associated with inhalation
of aflatoxin-contaminated grain dusts. FoodCosmet. Toxicol. 19: 765-769.
Bayram, H.; Devalia, J. L.; Khair, O. A.; Abdelaziz, M. M.; Sapsford, R. J.; Sagai, M.; Davies, R. J. (1998)
Comparison of ciliary activity and inflammatory mediator release from bronchial epithelial cells of nonatopic
nonasthmatic subjects and atopic asthmatic patients and the effect of diesel exhaust particles in vitro.
J. Allergy Clin. Immunol. 102: 771-782.
Beaumont, F.; Kauffman, H. F.; Sluiter, H. J.; De Vries, K. (1984) A volumetric-aerobiologic study of seasonal
fungus prevalence inside and outside dwellings of asthmatic patients living in northeast Netherlands. Ann.
Allergy 53: 486-492.
Beaumont, F.; Kauffman, H. F.; Sluiter, H. J.; De Vries, K. (1985) Sequential sampling of fungal air spores inside
and outside the homes of mould-sensitive, asthmatic patients: a search for a relationship to obstructive
reactions. Ann. Allergy 55: 740-746.
Behrendt, H.; Becker, W. M.; Friedrichs, K. H.; Darsow, U.; Tomingas, R. (1992) Interaction between aeroallergens
and airborne particulate matter. Int. Arch. Allergy Immunol. 99: 425-428.
Behrendt, H.; Friedrichs, K.-H.; Kramer, U.; Hitzfeld, B.; Becker, W.-M; Ring, J. (1995) The role of indoor and
outdoor air pollution in allergic diseases. In: Johansson, S. G. O., ed. Prog. Allergy Clin. Immunol.,
Proceedings of the 15th international congress, Allergol. Clin. Immunol.; 1994. Seattle, WA: Hogrefe &
Huber; pp. 83-89.
Behrendt, H.; Becker, W. M.; Fritzsche, C.; Sliwa-Tomczok, W.; Tomczok, J.; Friedrichs, K. H.; Ring, J. (1997)
Air pollution and allergy: experimental studies on modulation of allergen release from pollen by air
pollutants. Int. Arch. Allergy Immunol. 113: 69-74.
Behrendt, H.; Kramer, U.; Schafer, T.; Kasche, A.; Eberlein-Konig, B.; Darsow, U.; Ring, J. (2001)
Allergotoxicology-a research concept to study the role of environmental pollutants in allergy. Allergy Clin.
Immunol. Int. 13: 122-128.
Bellomo, R.; Gigliotti, P.; Treloar, A.; Holmes, P.; Suphioglu, C.; Singh, M. B.; Knox, Bruce. (1992) Two
consecutive thunderstorm associated epidemics of asthma in the city of Melbourne: the possible role of rye
grass pollen. Med. J. Aust. 156: 834-837.
Brunekreef, B.; Hoek, G.; Fischer, P.; Spieksma, F. T. M. (2000) Relation between airborne pollen concentrations
and daily cardiovascular and respiratory-disease mortality. Lancet 355: 1517-1518.
Burge, H. A. (1995) Bioaerosols in the residential environment. In: Cox, C. S.; Wathes, C. M., eds. Bioaerosols
handbook. Boca Raton, FL: CRC Press, Inc.; pp. 579-597.
Burge, H. A.; Rogers, C. A. (2000) Outdoor allergens. Environ. Health Perspect. 108(suppl. 4): 653-659.
Castellan, R. M.; Olenchock, S. A.; Hankinson, J. L.; Millner, P. D.; Cocke, J. B.; Bragg, C. K.; Perkins, H. H., Jr.;
Jacobs, R. R. (1984) Acute bronchoconstriction induced by cotton dust: dose-related responses to endotoxin
and other dust factors. Ann. Intern. Med. 102: 157-163.
Castellan, R. M.; Olenchock, S. A.; Kinsley, K. B.; Hankinson, J. L. (1987) Inhaled endotoxin and decreased
spirometric values: an exposure-response relation for cotton dust. N. Engl. J. Med. 317: 605-610.
Celenza, A.; Fothergill, J.; Kupek, E.; Shaw R. J. (1996) Thunderstorm associated asthma: a detailed analysis of
environmental factors. Br. Med. J. 312: 604-607.
Centers for Disease Control and Prevention. (1996) Coccidioidomycosis-Arizona, 1990-1995. Morb. Mortal. Wkly.
Rep. 45: 1069-1073.
Clark, S. (1986) Report on prevention and control. Am. J. Ind. Med. 10: 267-273.
Cooley, J. D.; Wong, W. C.; Jumper, C. A.; Straus, D. C. (1998) Correlation between the prevalence of certain fungi
and sick building syndrome. Occup. Environ. Med. 55: 579-584.
Cox, C. (1987) Threshold dose-response models in toxicology. Biometrics 43: 511-523.
7B-33
-------
Cox, C. S.; Wathes, C. M., eds. (1995) Bioaerosols handbook. Boca Raton, FL: CRC Press, Inc.
Croft, W. A.; Jarvis, B. B.; Yatawara, C. S. (1986) Airborne outbreak of trichothecene toxicosis. Atmos. Environ.
20: 549-552.
Cullen, R. T.; Searl, A.; Miller, B. G.; Davis, J. M.; Jones, A. D. (2000) Pulmonary and intraperitoneal inflammation
induced by cellulose fibres. J. Appl. Toxicol. 20: 49-60.
Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154: 633-641.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997) Combined diesel exhaust paniculate and ragweed
allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
Diaz-Sanchez, D.; Jyrala, M.; Ng, D.; Nel, A.; Saxon, A. (2000) In vivo nasal challenge with diesel exhaust particles
enhances expression of the CC chemokines rantes, MlP-la, and MCP-3 in humans. Clin. Immunol.
97: 140-145.
Douwes, J.; Zuidhof, A.; Doekes, G.; Van der Zee, S.; Wouters, I.; Boezen, H. M.; Brunekreef, B. (2000)
(1 -> 3)-p-D-glucan and endotoxin in house dust and peak flow variability in children. Am. J. Respir. Crit.
Care Med. 162: 1348-1354.
Edmonds, R. L., ed. (1979) Aerobiology: the ecological systems approach. Stroudsburg, PA: Dowden, Hutchinson&
Ross, Inc. (US/IBP synthesis series 10).
Elder, A. C. P.; Gelein, R.; Finkelstein J. N.; Cox, C.; Oberdorster, G. (2000a) Endotoxin priming affects the lung
response to ultrafine particles and ozone in young and old rats. In: Phalen, R. F., ed. Inhalation toxicology:
proceedings of the third colloquium on paniculate air pollution and human health (first special issue); June,
1999; Durham, NC. Inhalation Toxicol. 12(suppl. 1): 85-98.
Elder, A. C. P.; Gelein, R.; Finkelstein, J. N.; Cox, C.; Oberdorster, G. (2000b) Pulmonary inflammatory response to
inhaled ultrafine particles is modified by age, ozone exposure, and bacterial toxin. In: Grant, L. D., ed.
PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 227-246.
Facchini, M. C.; Fuzzi, S.; Zappoli, S.; Andracchio, A.; Gelencser, A.; Kiss, G.; Krivacsy, Z.; Meszaros, E.;
Hansson, H.-C.; Alsberg, T.; Zebtihr, Y. (1999) Partitioning of the organic aerosol component between fog
droplets and interstitial air. J. Geophys. Res. [Atmos.] 104: 26,821-26,832.
Fraser, M. P.; Lakshmanan, K. (2000) Using levoglucosan as a molecular marker for the long-range transport of
biomass combustion aerosols. Environ. Sci. Technol. 34: 4560-4564.
Fujieda, S.; Diaz-Sanchez, D.; Saxon, A. (1998) Combined nasal challenge with diesel exhaust particles and allergen
induces in vivo IgE isotype switching. Am. J. Respir. Cell Mol. Biol. 19: 507-512.
Fujimaki, H.; Nohara, O.; Ichinose, T.; Watanabe, N.; Saito, S. (1994) IL-4 production in mediastinal lymph node
cells in mice intratracheally instilled with diesel exhaust particulates and antigen. Toxicology 92: 261-268.
Galgiani, J. N.; Ampel, N. M. (1990) Coccidioidomycosis in human immunodeficiency virus-infected patients.
J. Infect. Dis. 162: 1165-1169.
Gelencser, A.; Hoffer, A.; Kiss, G.; Tombacz, E.; Kurdi, R.; Bencze, L. (2003) In-situ formation of light-absorbing
organic matter in cloud water. J. Atmos. Chem. 45: 25-33.
Girgis, S. T.; Marks, G. B.; Downs, S. H.; Kolbe, A.; Car, G. N.; Paton, R. (2000) Thunderstorm-associated asthma
in an inland town in south-eastern Australia. Who is at risk? Eur. Respir. J. 16: 3-8.
Gravesen, S. (1979) Fungi as a cause of allergic disease. Allergy 34:135-154.
Grote, M.; Vrtala, S.; Niederberger, V.; Valenta, R.; Reichelt, R. (2000) Expulsion of allergen-containing materials
from hydrated rye grass (Lolium perenne) pollen revealed by using immunogold field emission scanning and
transmission electron microscopy. J. Allergy Clin. Immunol. 105: 1140-1145.
Hastie, A. T.; Peters, S. P. (2001) Interactions of allergens and irritants in susceptible populations in producing lung
dysfunction: implications for future research. Environ. Health Perspect. 109(suppl. 4): 605-607.
Havers, N.; Burba, P.; Lambert, J.; Klockow, D. (1998) Spectroscopic characterisation of humic-like substances in
airborne paniculate matter. J. Atmos. Chem. 29: 45-54.
Heederik, D.; Douwes, J.; Wouters, L; Doekes, G. (2000) Organic dusts: beyond endotoxin. Inhalation Toxicol.
12(suppl. 3): 27-33.
Heinrich, J.; Hoelscher, B.; Frye, C.; Meyer, L; Pitz, M.; Cyrys, J.; Wjst, M.; Neas, L.; Wichmann, H.-E. (2002a)
Improved air quality in reunified Germany and decreases in respiratory symptoms. Epidemiology
13: 394-401.
7B-34
-------
Heinrich, J.; Hoelscher, B.; Frye, C.; Meyer, I.; Wjst, M.; Wichmann, H. E. (2002b) Trends in prevalence of atopic
diseases and allergic sensitization in children in eastern Germany. Eur. Respir. J. 19: 1040-1046.
Heinrich, J.; Pitz, M.; Bischof, W.; Krug, N.; Borm, P. J. A. (2003) Endotoxin in fine (PM2 5) and coarse (PM25.10)
particle mass of ambient aerosols. A tempero-spatial analysis. Atmos. Environ. 37: 3659-3667.
Hodgson, M. J.; Morey, P.; Leung, W.-Y.; Morrow, L.; Miller, D.; Jarvis, B. B.; Robbins, H.; Halsey, J. F.; Storey,
E. (1998) Building-associated pulmonary disease from exposure to Stachybotrys chartarum andAspergillus
versicolor. J. Occup. Environ. Med. 40: 241-249.
Hunter, C. A.; Lea, R. G. (1994) The airborne fungal population of representative British homes. In: Samson, R. A.;
Flannigan, B.; Flannigan, M. E.; Verhoeff, A. P.; Adan, O. C. G.; Hoekstra, E. S., eds. Health implications of
fungi in indoor environments. New York, NY: Elsevier, pp. 141-153. (Air Quality Monographs, v. 2).
Iversen, M.; Brink, O.; Dahl, R. (1994) Lung function in a five year follow-up study of farmers. Ann. Agric.
Environ. Med. 1: 39-43.
Kendrick, B. (1992) The fifth kingdom. 2nd ed. Newburyport, MA: Focus Information Group.
Knox, R. B.; Suphioglu, C.; Taylor, P.; Desai, R.; Watson, H. C.; Peng, J. L.; Bursill, L. A. (1997) Major grass
pollen allergen Lol p 1 binds to diesel exhaust particles: implications for asthma and air pollution. Clin. Exp.
Allergy 27: 246-251.
Koch, A.; Heilemann, K.-J.; Bischof, W.; Heinrich, J.; Wichmann, H. E. (2000) Indoor viable mold spores - a
comparison between two cities, Erfurt (eastern Germany) and Hamburg (western Germany) Allergy
55: 176-180.
Kuhn, D. M.; Ghannoum, M. A. (2003) Indoor mold, toxigenic fungi, and Stachybotrys chartarum'. infectious
disease perspective. Clin. Microbiol. Rev. 16: 144-172.
Larsen, F. O.; Christensen, L. H. R.; Clementsen, P.; Gravesen, S.; Stahl Skov, P.; Norn, S. (1996) Microfungi in
indoor air are able to trigger histamine release by non-IgE-mediated mechanisms. Inflammation Res.
45(suppl. 1): S23-S24.
Lewis, W. H.; Vinay, P.; Zenger, V. E. (1983) Airborne and allergenic pollen of North America. Baltimore, MD:
The Johns Hopkins University Press.
Licorish, K.; Novey, H. S.; Kozak, P.; Fairshter, R. D.; Wilson, A. F. (1985) Role ofAlternaria andPenicillium
spores in the pathogenesis of asthma. J. Allergy Clin. Immunol. 76: 819-825.
Lighthart, B.; Mohr, A. J. (1994) Atmospheric microbial aerosols: theory and applications. New York, NY:
Chapman & Hall.
Mathews, K. P. (1989) Inhalant insect-derived allergens. Immunol. Allergy Clin. North Am. 9: 321-338.
Matthias-Maser, S.; Jaenicke, R. (1995) The size distribution of primary biological aerosol particles with radii
> 0.2 |_im in an urban/rural influenced region. Atmos. Res. 39: 279-286.
Meklin, T.; Reponen, T.; Toivola, M.; Koponen, V.; Husman, T.; Hyvarinen, A.; Nevalainen, A. (2002a) Size
distributions of airborne microbes in moisture-damaged and reference school buildings of two construction
types. Atmos. Environ. 36: 39-40.
Meklin, T.; Husman, T.; Vepsalainen, A.; Vahteristo, M.; Koivisto, J.; Halla-Aho, J.; Hyvarinen, A.;
Moschandreas, D.; Nevalainen, A. (2002b) Indoor air microbes and respiratory symptoms of children in
moisture damaged and reference schools. Indoor Air 12: 175-183.
Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
Nel, A. E.; Diaz-Sanchez, D.; Ng, D.; Hiura, T.; Saxon, A. (1998) Enhancement of allergic inflammation by the
interaction between diesel exhaust particles and the immune system. J. Allergy Clin. Immunol. 102: 539-554.
Newson, R.; Strachan, D.; Archibald, E.; Emberlin, J.; Hardaker, P.; Collier, C. (1997) Effect of thunderstorms and
airborne grass pollen on the incidence of acute asthma in England, 1990-94. Thorax 52: 680-685.
Oikonen, M.; Laaksonen, M.; Laippala, P.; Oksaranta, O.; Lilius, E.-M.; Lindgren, S.; Rantio-Lehtimaki, A.;
Anttinen, A.; Koski, K.; Eralinna, J.-P. (2003) Ambient air quality and occurrence of multiple sclerosis
relapse. Neuroepidemiology 22: 95-99.
Ong, E. K. (1994) Grass pollen allergens: molecular characterization and environmental monitoring [dissertation].
Melbourne, Australia: The University of Melbourne.
Ormstad, H. (2000) Suspended paniculate matter in indoor air: adjuvants and allergen carriers. Toxicology
152: 53-68.
Ormstad, H.; Johansen, B. V.; Gaarder, P. I. (1998) Airborne house dust particles and diesel exhaust particles as
allergen carriers. Clin. Exp. Allergy 28: 702-708.
Palchak, R. B.; Cohen, R.; Ainslie, M.; Hoerner, C. L. (1988) Airborne endotoxin associated with industrial-scale
production of protein products in gram-negative bacteria. Am. Ind. Hyg. Assoc. J. 49: 420-421.
7B-35
-------
Park, J.-H.; Spiegelman, D. L.; Burge, H. A.; Gold, D. R.; Chew, G. L.; Milton, D. K. (2000) Longitudinal study of
dust and airborne endotoxin in the home. Environ. Health Perspect. 108: 1023-1028.
Pieckova, E.; Kunova, Z. (2002) Indoor fungi and their ciliostatic metabolites. Ann. Agric. Environ. Med. 9: 59-63.
Platts-Mills, T. A. E.; Chapman, M. D. (1987) Dust mites: immunology, allergic disease, and environmental control.
J. Allergy Clin. Immunol. 80: 755-775.
Poore, M. W. (2002) LevoglucosaninPM2.5 at the Fresno supersite. J. Air Waste Manage. Assoc. 52: 3-4.
Pope, A. M.; Patterson, R.; Burge, H. (1993) Indoor allergens: assessing and controlling adverse health effects.
Washington, DC: National Academy Press.
Puxbaum, H.; Tenze-Kunit, M. (2003) Size distribution and seasonal variation of atmospheric cellulose. Atmos.
Environ. 37: 3693-3699.
Ren, P.; Jankun, T. M.; Leaderer, B. P. (1999) Comparisons of seasonal fungal prevalence in indoor and outdoor air
and in house dusts of dwellings in one northeast American county. J. Exposure Anal. Environ. Epidemiol.
9: 560-568.
Rippon, J. W. (1988) Medical mycology: the pathogenic fungi and the pathogenic actinomycetes. 3rd ed.
Philadelphia, PA: W. B. Saunders Company.
Rogge, W. F.; Mazurek, M. A.; Hildemann, L. M.; Cass, G. R.; Simoneit, B. R. T. (1993) Quantification of urban
organic aerosols at a molecular level: identification, abundance and seasonal variation. Atmos. Environ.
Part A 27: 1309-1330.
Rosas, L; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, J.; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis
of the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
emergency admissions to a hospital in Mexico City. Allergy 53: 394-401.
Rose, C. S.; Martyny, J. W.; Newman, L. S.; Milton, D. K.; King, T. E., Jr.; Beebe, J. L.; McCammon, J. B.;
Hoffman, R. E.; Kreiss, K. (1998) "Lifeguard lung": endemic granulomatous pneumonitis in an indoor
swimming pool. Am. J. Public Health 88: 1795-1800.
Rylander, R. (1996) Airway responsiveness and chest symptoms after inhalation of endotoxin or (1 -> 3)-p-D-glucan.
Indoor Built Environ. 5: 106-111.
Rylander, R. (2002) Endotoxin in the environment - exposure and effects. J. Endotoxin Res. 8: 241-252.
Schappi, G. F.; Taylor, P. E.; Staff, I. A.; Suphioglu, C.; Knox, R. B. (1997) Source of Bet v 1 loaded inhalable
particles from birch revealed Sex. Plant Reprod. 10: 315-323.
Schappi, G. F.; Taylor, P. E.; Pain, M. C. F.; Cameron, P. A.; Dent, A. W.; Staff, I. A.; Suphioglu, C. (1999)
Concentrations of major grass group 5 allergens in pollen grains and atmospheric particles: implications for
hay fever and allergic asthma sufferers sensitized to grass pollen allergens. Clin. Exp. Allergy 29: 633-641.
Schauer, J. J.; Rogge, W. F.; Hildemann, L. M.; Mazurek, M. A.; Cass, G. R. (1996) Source apportionment of
airborne paniculate matter using organic compounds as tracers. Atmos. Environ. 30: 3837-3855.
Schneider, E.; Hajjeh, R. A.; Spiegel, R. A.; Jibson, R. W.; Harp, E. L.; Marshall, G. A.; Gunn, R. A.; McNeil,
M. M.; Pinner, R. W.; Baron, R. C.; Burger, R. C.; Hutwagner, L. C.; Crump, C.; Kaufman, L.; Reef, S. E.;
Feldman, G. M.; Pappagianis, D.; Werner, S. B. (1997) A coccidioidomycosis outbreak following the
Northridge, California earthquake. JAMA J. Am. Med. Assoc. 277: 904-908.
Spiewak, R.; Krysiriska-Traczyk, E.; Sitkowska, J.; Dutkiewicz, J. (1996a) Microflora of allergenic pollens - a
preliminary study. Ann. Agric. Environ. Med. 3: 127-130.
Spiewak, R.; Skorska, C.; Prazmo, Z.; Dutkiewicz, J. (1996b) Bacterial endotoxin associated with pollen as a
potential factor aggravating pollinosis. Ann. Agric. Environ. Med. 3: 57-59.
Stanley, R. G.; Linskens, H. F. (1985) Pollen: biologie biochemie gewinnung und verwendung (Pollen: biology,
biochemistry, production and uses). Greifenberg, German Federal Republic: Urs Freund Verlag.
Su, H.-J.; Wu, P.-C.; Chen, H.-L.; Lee, F.-C.; Lin, L.-L. (2001) Exposure assessment of indoor allergens, endotoxin,
and airborne fungi for homes in Southern Taiwan. Environ. Res. 85: 135-144.
Suphioglu, C.; Singh, M. B.; Taylor, P.; Bellomo, R.; Holmes, P.; Puy, R.; Knox, R. B. (1992) Mechanism of
grass-pollen-induced asthma. Lancet 339: 569-572.
Takahashi,T. (1997) Airborne fungal colony-forming units in outdoor and indoor environments in Yokohama, Japan.
Mycopathologia 139: 23-33.
Targonski, P. V.; Persky, V. W.; Ramekrishnan, V. (1995) Effect of environmental molds on risk of death from
asthma during the pollen season. J. Allergy Clin. Immunol. 95: 955-961.
Taylor, P. E.; Staff, I. A.; Singh, M. B.; Knox, R. B. (1994) Localization of the two major allergens in rye-grass
pollen using specific monoclonal antibodies and quantitative analysis of immunogold labelling. Histochem. J.
26: 392-401.
Taylor, P. E.; Flagan, R. C.; Valenta, R.; Glovsky, M. M. (2002) Release of allergens as respirable aerosols: a link
between grass pollen and asthma. J. Allergy Clin. Immunol. 109: 51-56.
7B-36
-------
Taylor, P. E.; Flagan, R.; Miguel, A. G.; Valenta, R.; Glovsky, M. M. (2003) Identification of birch pollen respirable
particles [abstract]. Chest 123(suppl. 3): 433S.
Thorn, J.; Rylander, R. (1998a) Inflammatory response after inhalation of bacterial endotoxin assessed by the
induced sputum technique. Thorax 53: 1047-1052.
Thorn, J.; Rylander, R. (1998b) Airways inflammation and glucan in a rowhouse area. Am. J. Respir. Crit. Care
Med. 157: 1798-1803.
Tuomi, T.; Engstrom, B.; Niemela, R.; SvinhufVud, J.; Reijula, K. (2000) Emission of ozone and organic volatiles
from a selection of laser printers and photocopiers. Appl. Occup. Environ. Hyg. 15: 629-634.
U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
U.S. Environmental Protection Agency. (2002) Health assessment document for diesel engine exhaust. Washington,
DC: Office of Research and Development, National Center for Environmental Assessment; report no.
EPA/600/8-90/057F. Available: http://cfpub.epa.gov/ncea/ [22 May, 2003].
Venables, K. M.; Allitt, U.; Collier, C. G.; Emberlin, J.; Greig, J. B.; Hardaker, P. J.; Highham, J. H.; Laing-Morton,
T.; Maynard, R. L.; Murray, V.; Strachan, D.; Tee, R. D. (1997) Thunderstorm-related asthma-the epidemic
of 24/25 June 1994. Clin. Exp. Allergy 27: 725-736.
Verhoeff, A. P.; Van Wijnen, H. J.; Attwood, P.; Versloot, P.; Boleij, J. S. M.; VanReenen-Hoekstra, E. S.; Samson,
R. A. (1988) Quantification and qualification of airborne fungi in houses. A comparison of measurement
techniques. Amsterdam, The Netherlands: Municipal Health Service.
Vogelzang, P. F. J.; VanDer Gulden, J. W. J.; Folgering, H.; Kolk, J. J.; Heederik, D.; Preller, L.; Tielen, M. J. M.;
Van Schayck, C. P. (1998) Endotoxin exposure as a major determinant of lung function decline in pig
farmers. Am. J. Respir. Crit. Care Med. 157: 15-18.
Yang, C. S.; Johanning, E. (2002) Airborne fungi and mycotoxins. In: Hurst, C. J., Crawford, R. L.; Knudsen, G. R.;
Mclnerney, M. J.; Stetzenbach, L. D. eds. Manual of environmental microbiology. 2nd ed. Washington, DC:
American Society for Microbiology; pp. 839-852.
Yozwiak, M. L.; Lundergan, L. L.; Kerrick, S. S.; Galgiani, J. N. (1988) Symptoms and routine laboratory
abnormalities associated with coccidioidomycosis. West. J. Med. 149: 419-421.
Young, R. S.; Jones, A. M.; Nicholls, P. J. (1998) Something in the air: endotoxins and glucans as environmental
troublemakers. J. Pharm. Pharmacol. 50: 11-17.
Zappoli, S.; Andracchio, A.; Fuzzi, S.; Facchini, M. C.; Gelencser, A.; Kiss, G.; Krivacsy, Z.; Molnar, A.; Meszaros,
E.; Hansson, H.-C.; Rosman, K.; Zebtihr, Y. (1999) Inorganic, organic and macromolecular components of
fine aerosol in different areas of Europe in relation to their water solubility. Atmos. Environ. 33: 2733-2743.
Zock, J.-P.; Hollander, A.; Heederik, D.; Douwes, J. (1998) Acute lung function changes and low endotoxin
exposures in the potato processing industry. Am. J. Ind. Med. 33: 384-391.
7B-37
-------
8. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS
ASSOCIATED WITH AMBIENT
PARTICULATE MATTER
8.1 INTRODUCTION
Epidemiologic studies linking community ambient PM concentrations to health effects
played an important role in the 1996 PM Air Quality Criteria Document (PM AQCD; U.S.
Environmental Protection Agency, 1996a). Many of those studies reported that measurable
excesses in pulmonary function decrements, respiratory symptoms, hospital and emergency
department admissions, and mortality in human populations are associated with ambient levels
of various indicators of PM exposure, including most notably PM10 as well as other indicators of
fine-fraction particles (e.g., PM25). Numerous more recent epidemiologic studies discussed in
this chapter have also evaluated ambient PM relationships to morbidity and mortality, using
various PM indicators, with greater emphasis on PM25 and other indicators of fine-fraction
particles and, to much lesser extent, PM10_2 5 as an indicator of coarse thoracic particles.
The epidemiology studies assessed here are best considered in combination with
information on ambient PM concentrations presented in Chapter 3, studies of human PM
exposure (Chapter 5), and PM dosimetry and toxicology (Chapters 6 and 7). Such epidemiologic
studies contribute important information on associations between health effects and exposures of
human populations to "real-world" ambient PM; and they also help to identify susceptible
subgroups and associated risk factors. Chapter 9 provides an interpretive synthesis of
information drawn from this and other chapters.
This chapter opens with brief discussion of approaches used for identifying, presenting,
and assessing studies; general features of the different types of epidemiologic studies assessed
and key methodological issues that arise in analyzing and interpreting study results; and salient
aspects of epidemiologic evidence that are considered in their critical assessment. Sections 8.2
and 8.3 assess epidemiologic studies of PM effects on mortality and morbidity, respectively.
Section 8.4 then provides an interpretive assessment of the overall PM epidemiologic data base
3-1
-------
reviewed in Sections 8.2 and 8.3 in relation to various key issues and aspects of the evidence.
The overall key findings and conclusions for this chapter are summarized in Section 8.5.
8.1.1 Approaches for Identifying, Presenting, and Assessing Studies
Numerous PM epidemiologic papers have been published since completion of the 1996 PM
AQCD, and U.S. EPA (NCEA-RTP) has used a systematic approach to identifying pertinent
epidemiologic studies for consideration in this chapter. In general, an ongoing continuous
MEDLINE search has been employed in conjunction with other strategies to identify PM
literature pertinent to developing criteria for PM NAAQS. The literature search method is
similar to those used by others (e.g., Basu and Samet, 1999). A publication base was first
established by using MEDLINE and other data bases and a set of key words (particles, air
pollution, mortality, morbidity, cause of death, PM, etc.) in a search strategy which was later
reexamined and modified to enhance identification of pertinent published papers. Since
literature searches encounter not a static but a changing, growing stream of information, searches
were not run just for the most recent calendar quarter but were also backdated in an attempt to
capture references added to that time period since the previous search was conducted. Papers
were also added to the publication base by EPA staff (a) through review of advance tables of
contents of thirty journals in which relevant papers are published and (b) by requesting scientists
known to be active in the field to identify papers recently accepted for publication.
While the above search regime builds a certain degree of redundancy into the system,
which ensures good coverage of the relevant literature and lessens the possibility of important
papers being missed, additional approaches have augmented traditional search methods. First, at
the beginning of the process, a Federal Register Notice was issued, requesting information and
published papers from the public at large. Next, experts in this field serving as non-EPA chapter
authors were not only provided with the outcomes of EPA literature searches, but also were
charged with identifying pertinent literature on their own. Finally, a keystone in the literature
identification process was that, at several review stages in the process, both the public and
CASAC offered comments which identified additional potentially relevant publications. The
combination of these approaches produced a rather comprehensive collection of pertinent studies
8-2
-------
appropriate for review and assessment here. This collection of studies includes pertinent new
studies published or accepted for publication through April, 2002, as well as some published
since then that provide important new information bearing on key scientific issues.
Those epidemiologic studies that relate measures of ambient air PM to human health
outcomes are assessed in this chapter, whereas studies of (typically much higher) occupational
exposures are generally not considered here. Criteria used for selecting literature for the present
assessment include mainly whether a given study includes information on: (1) ambient PM
indices (e.g., PM10, PM2 5, PM10_2 5, etc.) of short- and long-term exposures as a key element;
(2) analyses of health effects of specific PM chemical or physical constituents (e.g., metals,
sulfates, nitrates or ultrafine particles, etc.) or indicators related to PM sources (e.g., motor
vehicle emissions, combustion-related particles, earth crustal particles); (3) evaluation of health
endpoints and populations not previously extensively researched; (4) multiple pollutant analyses
and other approaches to addressing issues related to potential confounding of effects and effects
modification; and/or (S)studies addressing important methodological issues (e.g., lag structure,
model specification, thresholds, mortality displacement) related to PM exposure effects.
In assessing the evidence, key points derived from the 1996 PM AQCD assessment of the
available information are first concisely highlighted. Then, key new information is presented in
succinct text summary tables for important new studies that have become available since the
1996 PM AQCD. More detailed information on various methods and results for these and other
newly available studies are summarized in tabular form in Appendices 8A and 8B. These
appendix tables are generally organized to include: information about (1) study location and
ambient PM levels; (2) description of study methods employed; (3) results and comments; and
(4) quantitative outcomes for PM measures. In the main body of the chapter, greater emphasis is
placed on integrating and interpreting findings from the array of evidence provided by the more
important newer studies than on detailed evaluation of each of the numerous newly available
studies. In presenting quantitative effects estimates in tables in the chapter and appendices,
study results were normalized to standard PM increments, as was done in the 1996 PM AQCD.
In selecting PM increments for use in this review, more recent air quality data were considered,
resulting in no changes to the increments previously used for short-term exposure studies, but
smaller increments than those used in the 1996 PM AQCD for long-term exposure studies. More
-------
specifically, the pollutant concentration increments used here to report relative risks (RR's) or
odds ratios for various health effects are as follow for short term (< 24 h) exposure studies:
50 |ig/m3 for PM10; 25 |ig/m3 for PM25 and PM10.25; 155 nmoles/m3 (15 |ig/m3 for SO42 ; and
75 nmoles/m3 (3.6 |ig/m3, if as H2SO4) for H+. For long-term exposure studies, the increments
used here are 20 |ig/m3 for PM10 and 10 |ig/m3 for PM25 and PM10_25.
Particular emphasis is focused in the text on those studies and analyses thought to provide
information most directly applicable for United States standard setting purposes. Specifically,
North American studies conducted in the U.S. or Canada are generally accorded more text
discussion than those from other geographic regions; and analyses using gravimetric (mass)
measurements are generally accorded more text attention than those using nongravimetric
ambient PM measures, e.g., black smoke (BS) or coefficient of haze (CoH). In addition,
emphasis is placed on text discussion of (a) new multicity studies that employ standardized
methodological analyses for evaluating PM effects across several or numerous cities and often
provide overall effects estimates based on combined analyses of information pooled across
multiple cities; (b) other studies providing quantitative PM effect-size estimates for populations
of interest; and (c) studies that consider PM as a component of a complex mixture of air
pollutants, including in particular the gaseous criteria pollutants (O3, CO, NO2, SO2).
In assessing the relative scientific quality of epidemiologic studies reviewed here and to
assist in interpreting their findings, the following considerations were taken into account, as was
done in the 1996 PM AQCD:
• To what extent are the aerometric data/exposure metrics used of adequate quality and
sufficiently representative to serve as credible exposure indicators, well reflecting
geographic or temporal differences in study population pollutant exposures in the range(s)
of pollutant concentrations evaluated?
• 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?
• Were the health endpoint measurements meaningful and reliable, including clear
definition of diagnostic criteria utilized and consistency in obtaining dependent
variable measurement
• Were the statistical analyses used appropriate and properly performed and interpreted,
including accurate data handling and transfer during analyses?
8-4
-------
• Were likely important covariates (e.g., potential confounders or effect modifiers)
adequately controlled for or taken into account in the study design and statistical analyses?
• Were the reported findings internally consistent, biologically plausible, and coherent in
terms of consistency with other known facts?
These guidelines provide benchmarks for judging the relative quality of various studies and
for focusing on the highest quality studies in assessing the body of epidemiologic evidence.
Detailed critical analysis of all epidemiologic studies on PM health effects, especially in relation
to all of the above questions, is beyond the scope of this document. Of most importance for
present purposes are those studies which provide useful qualitative or quantitative information
on exposure-effect or exposure-response relationships for health effects associated with ambient
air levels of PM currently likely to be encountered in the United States.
8.1.2 Types of Epidemiologic Studies Reviewed
Definitions of various types of epidemiologic studies assessed here were provided in the
1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) and are briefly summarized
here. Briefly, the epidemiologic studies are divided into mortality studies and morbidity studies.
Mortality studies evaluating PM effects on total (nonaccidental) mortality and cause-specific
mortality provide the most unambiguous evidence related to a clearly adverse endpoint. The
morbidity studies further evaluate PM effects on a wide range of health endpoints, such as:
cardiovascular and respiratory-related hospital admissions, medical visits, reports of respiratory
symptoms, self-medication in asthmatics, changes in pulmonary function; changes in
cardiovascular physiology/functions, and blood coagulation; low birthweight infants, etc.
The epidemiologic strategies most commonly used in PM health studies are of four types:
(1) ecologic studies; (2) time-series semi-ecologic studies; (3) prospective cohort studies; and
(4) case-control and crossover studies. In addition, time-series analyses or other analytic
approaches have been used in so-called intervention studies or "natural experiments." All of
these are observational studies rather than experimental studies. In general, the exposure of the
participant is not directly observed; and the concentration of airborne particles and other air
3-5
-------
pollutants at one or more stationary air monitors is used as a proxy for individual exposure to
ambient air pollution.
In ecologic studies, the responses are at a community level (for example, annual mortality
rates), as are the exposure indices (for example, annual average PM concentrations) and
covariates (for example, the percentage of the population greater than 65 years of age).
No individual data are used in the analysis; therefore, the relationship between health effect and
exposure calculated across different communities may not reflect individual-level associations
between health outcome and exposure. The use of proxy measures for individual exposure and
covariates or effect modifiers may also bias the results, and within-city or within-unit
confounding may be overlooked.
Time-series studies are more informative because they allow the study of associations
between changes in a health outcome and changes in exposure indicators preceding or
simultaneous with the outcome. The temporal relationship supports a conclusion of a causal
relation, even when both the outcome (for example, the number of nonaccidental deaths in a city
during a day) and the exposure (for example, daily air pollution concentration) are community
indices.
Prospective cohort studies use data from individuals, including health status (where
available), individual exposure (not usually available), and individual covariates or risk factors,
observed over time. The participants in a prospective cohort study are ideally recruited (using a
simple or stratified random sample) so as to represent a target population for which individual or
community exposure of the participants is known before and during the interval up to the time
the health endpoint occurs. The use of individual-level data is believed to give prospective
cohort studies greater inferential strength than other epidemiologic strategies. The use of
community-level or estimated exposure data, if necessary, may weaken this advantage, as it does
in time-series studies.
Case-control studies are retrospective studies in that exposure is determined after the
health endpoint occurs (as is common in occupational health studies). As Rothman and
Greenland (1998) describe it, "Case-control studies are best understood by defining a source
population, which represents a hypothetical study population in which a cohort study might have
8-6
-------
been conducted ... In a case-control study, the cases are identified and their exposure status is
determined just as in a cohort study . . . [and] a control group of study subjects is sampled from
the entire source population that gives rise to the cases . . . the cardinal requirement of control
selection is that the controls must be sampled independently of their exposure status."
The case-crossover design is suited to the study of a transient effect of an intermittent
exposure on the subsequent risk of an acute-onset health effect thought to occur shortly after
exposure. In the original development of the method, effect estimates were based on within-
subject comparisons of exposures associated with incident disease events with exposures at
times before the occurrence of disease, using matched case-control methods or methods for
stratified follow-up studies with spare data within each stratum. The principle of the analysis is
that the exposures of cases just before the event are compared with the distribution of exposure
estimated from some separate time period, the former being assumed to be representative of the
distribution of exposures for those individuals while they were at risk for the outcome of interest.
When measurements of exposure or potential effect modifiers are available on an
individual level, it is possible to incorporate this information into a case-crossover study (unlike
a time-series analysis). A disadvantage of the case-crossover design, however, is the potential
for bias due to time trends in the exposure time-series. Because case-crossover comparisons are
made between different points in time, the case-crossover analysis implicitly depends on an
assumption that the exposure distribution is stable over time (stationary). If the exposure time-
series is nonstationary and case exposures are compared with referent exposures systematically
selected from a different period in time, a bias may be introduced into estimates of the measure
of association for the exposure and disease. These biases are particularly important when
examining the small relative risks that appear to exist for PM health outcomes.
Intervention studies (often involving features of time-series or other types of analyses
noted above) provide another useful approach for evaluating possible causal relationships
between ambient air pollution variables (e.g., PM) and health effects in human populations.
In such studies, the effects of active interventions that result in reductions of one or another or
several air pollutants (constituting essentially a "found experiment") are evaluated in relation to
changes in mortality or morbidity outcomes among population groups affected by the reduction
3-7
-------
in air pollution exposure. To date, only a few epidemiologic studies have evaluated the
consequences of interventions that allow for comparison of PM-health outcome associations
before and after certain relatively discrete events resulting in notable changes in concentrations
of ambient PM and/or one or more other co-pollutants. Given that the etiology of health
outcomes related to PM or other air pollutants are typically also affected by other risk factors,
it is important for intervention studies not only to measure air pollution exposure and health
status before and after air pollution reductions but also to identify and evaluate potential effects
of other risk factors before and after the air pollution reductions. The proposition that
intervention studies can provide strong support for causal inferences was emphasized by Hill
(1965), as discussed further in Section 8.1.4. In his classic monograph (The Environment and
Disease: Association or Causation?), Hill (1965) addressed the topic of preventive action and its
consequences under Aspect 8, stating:
"Experiment: Occasionally it is possible to appeal to experimental, or semi-
experimental, evidence. For example, because of an observed association some
preventive action is taken. Does it in fact prevent? The dust in the workshop is
reduced, lubricating oils are changed, persons stop smoking cigarettes. Is the
frequency of the associated events affected? Here the strongest support for the
causation hypothesis may be revealed."
8.1.3 Overview of Key Methodological Issues
There are a number of methodological issues that arise in analyzing and interpreting
epidemiologic studies that are more fully discussed in Section 8.4 below. The following brief
overview of two such key issues is intended to orient the reader to these issues so as to provide
context for the presentation and assessment of the epidemiologic studies on mortality and
morbidity effects in Sections 8.2 and 8.3.
8.1.3.1 Issues Related to Use of Generalized Additive Models (GAM) in PM Epidemiology
In the spring of 2002, the original investigators of an important newly available multicity
study (the National Mortality and Morbidity Air Pollution Study; NMMAPS) cosponsored by the
Health Effects Institute (HEI) reported that use of the default convergence criteria setting used in
the GAM routine of certain widely-used statistical software (Splus) could result in biased
-------
estimates of air pollution effects when at least two nonparametric smoothers are included in the
model (Health Effects Institute letter, May 2002). The NMMAPS investigators also reported
(Dominici et al., 2002), as determined through simulation, that such bias was larger when the
size of risk estimate was smaller and when the correlation between the PM and the covariates
(i.e., smooth terms for temporal trend and weather) was higher. While the NMMAPS
investigators reported that reanalysis of the 90 cities air pollution-mortality data (using stringent
convergence criteria) did not qualitatively change their original findings (i.e., the positive
association between PM10 and mortality; lack of confounding by gaseous pollutants; regional
heterogeneity of PM, etc.), the reduction in the PM10 risk estimate was apparently not negligible
(dropping, upon reanalysis, from 2.1% to 1.4% excess deaths per 50 |ig/m3 increase in PM10)
with GAM using strict convergence criteria and a further reduction to 1.1% using a generalized
linear model (GLM).
Issues surrounding potential bias in PM risk estimates from time-series studies using GAM
analyses and default convergence criteria were raised by EPA and discussed in July 2002 at the
CAS AC review of the Third External Review Draft of this PM AQCD. In keeping with a follow
up consultation with CASAC in August 2002, EPA encouraged investigators for a number of
important published studies to reanalyze their data by using GAM with more stringent
convergence criteria, as well as by using GLM analyses with parametric smoothers that
approximated the original GAM model. EPA, working closely with HEI, also arranged for (a)
the resulting reanalyses first to be discussed at an EPA-sponsored Workshop on GAM-Related
Statistical Issues in PM Epidemiology held in November 2002; (b) then for any revamping of the
preliminary analyses in light of the workshop discussions; before (c) submittal by the
investigators of short communications describing the reanalyses approaches and results to EPA
and HEI for peer-review by a special panel assembled by HEI; and (d) the publication of the
short communications on the reanalyses, along with commentary by the HEI peer-review panel,
in an HEI Special Report (2003a). Some of the short communications included in the HEI
Special Report (2003a) included discussion of reanalyses of data from more than one original
publication because the same data were used to examine different issues of PM-mortality
8-9
-------
associations (e.g., concentration/response function, harvesting, etc.). In total, reanalyses were
reported for more than 35 originally published studies.
8.1.3.2 Confounding and Effect Modification
A pervasive problem in the analysis of epidemiologic data, no matter what design or
strategy, is the unique attribution of a given health outcome to a nominal causal agent (e.g., to
airborne particles in this document). The health outcomes attributed to particles are not specific;
and, as such, they may also be attributable to high or low temperatures, influenza and other
diseases, and/or exposure to other air pollutants. Some of these co-variables may be
confounders and others effect modifiers. The distinctions are important.
Confounding is "... a confusion of effects. Specifically, the apparent effect of the
exposure of interest is distorted because the effect of an extraneous factor is mistaken for or
mixed with the actual exposure effect (which may be null)" (Rothman and Greenland, 1998,
p. 120).
Causal events occur prior to some initial bodily response. A causal association may
usually be defined as an association in which alteration in the frequency or quality of one
category (e.g., level of PM in ambient air) is followed by a change in the other (e.g, increased
mortality). The concept of the chain mechanism is that many variables may be related to a
single effect through a direct-indirect mechanism. In fact, events are not dependent on single
causes. A given chain of causation may represent only a fraction of a web (MacMahon and
Pugh, 1970). A causal pathway refers to the network of relationships among factors in one or
more causal chains in which the members of the population are exposed to causal agents that
produce the observed health effect. The primary cause may be mediated by secondary causes
(possibly proximal to exposure) and may have either a direct effect on exposure or an indirect
effect through the secondary causes, or both, as illustrated below. A noncausal pathway may
involve factors not actually associated or correlated with population exposure to the pollutant of
interest, but are coincidentally (spuriously) also associated with health outcome.
The determination of whether a potential confounder is an actual confounder may be
elucidated from biological or physical knowledge about its exposure and health effects. Patterns
8-10
-------
of association in epidemiology may be helpful in suggesting where to look for this knowledge,
but do not replace it. In evaluating effects of ambient PM exposures, gaseous criteria pollutants
(CO, NO2, SO2, O3) are candidates for confounders because all of these are known to cause at
least some types of adverse health effects that are also associated with particles (CO more often
being associated with cardiovascular effects and the other gases with respiratory effects,
including symptoms and hospital admissions). In addition, the gaseous criteria pollutants may
be associated with particles for several reasons, including common sources and correlated
changes in response to wind and weather. Lastly, SO2 and NO2 may be precursors to sulfate and
nitrate components of ambient particle mixes, while NO2 contributes also to the formation of
organic aerosols during photochemical transformations.
The problem of disentangling the effects of other pollutants is especially difficult when
high correlation exists between ambient PM measurements and one or more of them.
For example, both CO and particles are emitted from motor vehicles. These and other fossil fuel
combustion sources also often emit SO2 and/or NO, which converts to NO2 upon emission.
SO2 and NO2, in turn, are precursors to sulfates and nitrates as two widely common contributors
to secondary ambient PM aerosol components. Ozone (O3) also contributes to ambient PM via
(a) hydroxyl radicals which oxidize SO2 to H2SO4 and NO2 to HNO3 and (b) participation in
chemical reactions underlying the formation of ultrafine particles from naturally occurring
terpenes, isoprene, and other hydrocarbons. A common source, such as combustion of gasoline
in motor vehicles emitting CO, NO2, and primary particles (and often resulting in high
correlations), may play an important role in confounding among these pollutants, as do weather
and seasonal effects. Even though O3 is a secondary pollutant also associated with emission
of NO2, it is often more variably correlated with ambient PM concentrations, depending on
location, season, etc. Levels of SO2 in the western U.S. are often quite low, so that secondary
formation of particle sulfates plays a much smaller role there, resulting in usually relatively little
confounding of SO2 with PM mass concentration in the West. On the other hand, in the
industrial Midwest and northeastern states, SO2 and sulfate levels during many of the
epidemiology studies were relatively high and highly correlated with fine particle mass
concentrations. If the correlation between PM and SO2 is not too high, it may be possible to
8-11
-------
estimate some part of their independent effects, which depend on the assumption of
independence under the particular model analyzed. If there is a causal pathway, then it may be
difficult to determine whether the observed relationship of exposure to health effect is a direct
effect of the exposure (to sulfate or fine PM as an example), an indirect effect mediated by the
potential confounder (e.g., exposure to SO2), or a mixture of these. Consideration of additional
(e.g., exposure, dosimetric, toxicologic) information beyond narrow reliance on observed
correlations among the PM measure(s), other pollutants, and health outcome indicators is often
useful in helping to elucidate the plausibility of PM or other pollutants being causally related to
statistically-associated health effects.
Some variables fall into the category of effect modifiers. "Effect-measure modification
differs from confounding in several ways. The main difference is that, whereas confounding is a
bias that the investigator hopes to prevent or remove from the effect estimate, effect-measure
modification is a property of the effect under study ... In epidemiologic analysis one tries to
eliminate confounding but one tries to detect and estimate effect-measure modification"
(Rothman and Greenland, 1998, p. 254). Examples of effect modifiers in some of the studies
evaluated in this chapter include environmental variables (such as temperature or humidity),
individual risk factors (such as education, cigarette smoking status, age in a prospective cohort
study), and community factors (such as percent of population > 65 years old). It is often possible
to stratify the relationship between health outcome and exposure by one or more of these risk
factor variables. Effect modifiers may be encountered (a) within single-city time-series studies
or (b) across cities in a two-stage hierarchical model or meta-analysis.
Potential confounding is usually much more difficult to identify; and several statistical
methods are available, none of them being completely satisfactory. The usual methods include
the following:
Within a city:
(A) Fit both a single-pollutant model and then several multipollutant models, and
determine if including the co-pollutants greatly changes the estimated effect;
(B) If the PM index and its co-pollutants are nearly multi-colinear, carry out a factor
analysis, and determine which gaseous pollutants are most closely associated with
PM in one or more common factors;
8-12
-------
Using data from several cities:
(C) Proceed as in Method A and pool the effect size estimates across cities for single-
and multipollutant models;
(D) Carry out a hierarchical regression of the PM effects versus the mean co-pollutant
concentration and determine if there is a relationship; and
(E) First carry out a regression of PM versus the co-pollutant concentration within each
city and the regression coefficient of PM versus health effect for each city. Then
fit a second-stage model regressing the PM-health effect coefficient versus the
PM-co-pollutant coefficient, concluding that the co-pollutant is a confounder if
there is an association at the second stage.
Each of the above methods (A through E) are subject to one or more disadvantages. The
multipollutant regression coefficients in method A, for example, may be unstable and have
greatly inflated standard errors, weakening their interpretation. In method B, the factors may be
sensitive to the choice of co-pollutants and the analysis method, and may be difficult to relate to
real-world entities. In method C, as with any meta-analysis, it is necessary to consider the
heterogeneity of the within-city effects before pooling them. Some large multicity studies have
revealed unexpected heterogeneity, not fully explained at present. While method D is sometimes
interpreted as showing confounding if the regression coefficient is nonzero, this is an argument
for effect modification, not confounding. Method E is sensitive to the assumptions being made;
for instance, if PM is the primary cause and the co-pollutant the secondary cause, then the two-
stage approach may be valid. However, if the model is mis-specified and there are two or more
secondary causes, some of which may not be identified, then the method may give misleading
results.
Given the wide array of considerations and possibilities discussed above, it is extremely
important to recognize that there is no single "correct" approach to modeling ambient PM-health
effects associations that will thereby provide the "right" answer with regard to precise
quantification of PM effect sizes for different health outcomes. Rather, it is clear that emphasis
needs to be placed here on (a) looking for convergence of evidence derived from various
acceptable analyses of PM effects on a particular type of health endpoint (e.g., total mortality,
respiratory hospital admissions, etc.); (b) according more weight to those well-conducted
analyses having greater power to detect effects and yielding narrower confidence intervals; and
8-13
-------
(c) evaluating the coherence of findings across pertinent health endpoints and effect sizes for
different health outcomes.
The issue of what PM effect sizes should be the main focus of presentation and discussion
in ensuing text - i.e., those derived from single-pollutant models including only PM or effect
sizes derived from multipollutant models that include one or more other co-pollutants along with
the PM indicator(s) - is an important one. Again, there is not necessarily any single "correct"
answer on this point. Implicit in arguments asserting that multipollutant model results must be
reported and accorded equal or more weight than single-pollutant model PM results is
a functional construct that has been widely used in epidemiologic modeling of health effects of
air pollution, a functional construct that considers the various air pollutants to be acting mainly
independently of one another in terms of their health effects, which may not necessarily be the
case. This may be causing either over- or under-estimation of PM health effects, depending on
the modeling choices made by the investigator and the study situation. For example, O3
and PM2 5 can share some similar oxidative formation and effect pathways in exerting adverse
health effects on the lung, yet are often modeled as independent pollutants or are placed in
models simultaneously, even though they may sometimes have high correlations over space and
time and in their human health effects. Another complication is that other pollutants can be
derived from like sources and may serve less as a measure of direct effects than as a marker of
pollution from a specific source. As an example noted earlier, SO2 and PM2 5 are often
predominantly derived from the same sources in a locale (e.g., coal-fired power plants in the
midwestern United States), so that putting these two pollutants in a model simultaneously may
cause a diminution of the PM25 coefficient that can be misleading.
One approach that has been taken is to look at pollutant interactions (either multiplicative
or additive, depending on the model assumed), but until we understand (and appropriately
model) underlying biological mechanisms, such models are assumptions on the part of the
researcher. Present modeling practices represent the best methods now available and provide
useful assessments of PM health effects. However, ultimately, more biological-plausibility
based models are needed that more accurately model pollutant interactions and allow more
biologically-based interpretations of modeling results.
8-14
-------
Until more is known about multiple pollutant interactions, it is important to avoid over-
interpreting model results regarding the relative sizes and significance of specific pollutant
effects, but instead to use biological plausibility in interpreting model results. For example, as
discussed later, Krewski et al (2000) found significant associations for both PM and SO2 in their
reanalysis for the Health Effects Institute of the ACS data set published by Pope et al. (1995).
Regarding these pollutant associations, they concluded that: "The absence of a plausible
toxicological mechanism by which sulfur dioxide could lead to increased mortality further
suggests that it might be acting as a marker for other mortality-associated pollutants." (Note:
Annual mean SO2 averaged < 10 ppb across -125 cities in the ACS data set.) Rather than letting
statistical significance be the sole determinant of the "most important" pollutant, the authors
utilized biological plausibility to dray conclusions about which association was most likely
driving the pollution-health effects association in question. Such biological
plausibility/mechanistic considerations need to be taken into account more broadly in the future
in modeling and assessing possible pollutant interactions in contributing to health effects
attributed to PM. In the meantime, the results from single-pollutant models of PM effects are
emphasized here, as being those most likely reflecting overall effects exerted by ambient PM
either acting alone and/or in combination with other ambient air pollutants.
8.1.4 Approach to Assessing Epidemiologic Evidence
The critical assessment of epidemiologic evidence presented in this chapter is conceptually
based upon consideration of salient aspects of the evidence of associations so as to reach
fundamental judgments as to the likely causal significance of the observed associations. In so
doing, it is appropriate to draw from those aspects initially presented in Hill's classic monograph
(Hill, 1965) and widely used by the scientific community in conducting such evidence-based
reviews. A number of these aspects are judged to be particularly salient in evaluating the body
of evidence available in this review, including the aspects described by Hill as strength,
experiment, consistency, plausibility, and coherence. Other aspects identified by Hill, including
temporality and biological gradient, are also relevant and considered here (e.g., in characterizing
lag structures and concentration-response relationships), but are more directly addressed in the
8-15
-------
design and analyses of the individual epidemiologic studies included in this assessment.
(As noted below, Hill's remaining aspects of specificity and analogy are not considered to be
particularly salient in this assessment.) As discussed below, these salient aspects are interrelated
and considered throughout the evaluation of the epidemiologic evidence presented in this
chapter, and are more generally reflected in the integrative synthesis presented in Chapter 9.
In the following sections, the general evaluation of the strength of the epidemiological
evidence reflects consideration not only of the magnitude of reported PM effects estimates and
their statistical significance, but also of the precision of the effects estimates and the robustness
of the effects associations. Consideration of the robustness of the associations takes into account
a number of factors, including in particular the impact of alternative models and model
specifications and potential confounding by co-pollutants, as well issues related to the
consequences of measurement error. Another aspect that is related to the strength of the
evidence in this assessment is the availability of evidence from "found experiments", or
so-called intervention studies, which have the potential to provide particularly strong support for
making causal inferences.
Consideration of the consistency of the effects associations, as discussed in the following
sections, involves looking across the results of multi- and single-city studies conducted by
different investigators in different places and times. In this assessment of ambient PM-health
effects associations, it is important to consider the aspect of consistency in the context of
understanding that ambient PM in different locations and at different times originates from
different sources, such that its composition and physical characteristics can vary greatly across
studies using the same indicator for size-differentiated PM mass. Other relevant factors are also
known to exhibit much variation across studies. These include, for example, the presence and
levels of co-pollutants, the relationships between central measures of PM and exposure-related
factors, relevant demographic factors related to sensitive subpopulations, as well as climatic and
meteorological conditions. Thus, in this case, consideration of consistency, and the related
heterogeneity of effects issue, is appropriately understood as an evaluation of the similarity or
general concordance of results, rather than an expectation of finding quantitative results within a
very narrow range. Particular weight is given in this assessment, consistent with Hill's views,
8-16
-------
to the presence of "similar results reached in quite different ways, e.g., prospectively and
retrospectively" (Hill, 1965). On the other hand, in light of complexities in the chemical and
physical properties of the mix of ambient PM and its spatial and temporal variations, Hill's
specificity of effects and analogy aspects are not viewed as being particularly salient here.
Looking beyond the epidemiological evidence, evaluation of the biological plausibility of
the PM-health effect associations observed in epidemiologic studies reflects consideration of
both exposure-related factors and dosimetric/toxicologic evidence relevant to identification of
potential biological mechanisms. Similarly, consideration of the coherence of health effects
associations reported in the epidemiologic literature reflects broad consideration of information
pertaining to the nature of the various respiratory- and cardiac-related mortality and morbidity
effects and biological markers evaluated in toxicologic and epidemiologic studies. These
broader aspects of the assessment are only touched upon in this chapter but are more fully
integrated in the discussion presented in Chapter 9.
In identifying these aspects as being particularly salient in this assessment, it is also
important to recognize that no one aspect is either necessary or sufficient for drawing inferences
of causality. As Hill (1965) emphasized:
"None of my nine viewpoints can bring indisputable evidence for or against the cause-
and-effect hypothesis and none can be required as a sine qua non. What they can do,
with greater or less strength, is to help us to make up our minds on the fundamental
question — is there any other way of explaining the set of facts before us, is there
any other answer equally, or more, likely than cause and effect?"
Thus, while these aspects frame considerations weighed in assessing the epidemiologic evidence,
they do not lend themselves to being considered in terms of simple formulas or hard-and-fast
rules of evidence leading to answers about causality (Hill, 1965). One, for example, cannot
simply count up the numbers of studies reporting statistically significant results for the various
PM indicator and health endpoints evaluated in this assessment and reach credible conclusions
about the relative strength of the evidence and the likelihood of causality. Rather, these
important considerations are taken into account throughout this assessment with a goal of
8-17
-------
producing an objective appraisal of the evidence (informed by peer and public comment and
advice), which includes the weighing of alternative views on controversial issues.
8.2 MORTALITY EFFECTS ASSOCIATED WITH AIRBORNE
PARTICULATE MATTER EXPOSURE
8.2.1 Introduction
The relationship of PM and other air pollutants to excess mortality has been studied
extensively and represents an important issue addressed in previous PM criteria assessments
(U.S. Environmental Protection Agency, 1986, 1996a). Recent findings are evaluated here
mainly for the two most important epidemiology designs by which mortality is studied: time-
series mortality studies (Section 8.2.2) and prospective cohort studies (Section 8.2.3). The time-
series studies mostly assess acute responses to short-term PM exposure, although some recent
work suggests that time-series data sets can also be useful in evaluating responses to exposures
over a longer time scale. Time-series studies use community-level air pollution measurements to
index exposure and community-level response (i.e., the total number of deaths each day by age
and/or by cause of death). Prospective cohort studies usefully complement time-series studies;
they typically evaluate human health effects of long-term PM exposures indexed by community-
level measurements, using individual health records with survival lifetimes or hazard rates
adjusted for individual risk factors.
8.2.2 Mortality Effects of Short-Term Particulate Matter Exposure
8.2.2.1 Summary of 1996 Particulate Matter Criteria Document Findings and Key Issues
The time-series mortality studies reviewed in the 1996 and other past PM AQCD's
provided much evidence that ambient PM air pollution is associated with increases in daily
mortality. The 1996 PM AQCD assessed about 35 PM-mortality time-series studies published
between 1988 and 1996. Of these studies, only five studies used GAM with default convergence
criteria. Recent reanalyses (Schwartz, 2003a; Klemm and Mason, 2003) using GAM with
stringent convergence criteria and other non-GAM approaches for one of these five studies, i.e.,
8-18
-------
the Harvard Six Cities time-series analysis (the only multeity study among the five studies),
essentially confirmed the original findings. Thus, information provided in the 1996 PM AQCD
can be summarized without major concern with regard to the GAM convergence issue. The
evidence derived from those studies was generally consistent with the hypothesis that PM is a
causal agent in contributing to short-term air pollution exposure effects on mortality.
The PM10 relative risk estimates derived from short-term PM10 exposure studies reviewed
in the 1996 PM AQCD suggested that an increase of 50 |ig/m3 in the 24-h average of PM10 is
most clearly associated with an increased risk of premature total nonaccidental mortality (total
deaths minus those from accident/injury), on the order of relative risk (RR) = 1.025 to 1.05 in the
general population or, in other words, 2.5 to 5.0% excess deaths per 50 |ig/m3 PM10 increase.
Higher relative risks were observed for the elderly and for those with preexisting
cardiopulmonary conditions. Also, based on the Schwartz et al. (1996a) analysis of Harvard
Six City data (as later confirmed in the reanalysis by Schwartz [2003a] and Klemm and Mason
[2003]), the 1996 PM AQCD found the RR (combined across the six cities) for excess total
mortality in relation to 24-h fine particle concentrations to be about 3% excess risk per
25 |ig/m3 PM2 5 increment.
While numerous studies reported PM-mortality associations, important issues needed to be
addressed in interpreting their findings. The 1996 PM AQCD evaluated in considerable detail
several critical issues, including: (1) seasonal confounding and effect modification;
(2) confounding by weather; (3) confounding by co-pollutants; (4) measurement error;
(5) functional form and threshold; (6) harvesting and life shortening; and (7) the role of PM
components. As important issues related to model specification became further clarified, more
studies began to address the most critical issues, some of which were at least partially resolved,
whereas others required still further investigation. The next several paragraphs summarize the
status of these issues at the time of the 1996 PM AQCD publication.
One of the most important components in time-series model specification is adjustment for
seasonal cycles and other longer-term temporal trends; not adequately adjusting for them could
result in biased RRs. Modern smoothing methods allow efficient fits of temporal trends and
reduce such statistical problems (but may also introduce some additional issues as discussed in
8-19
-------
later sections). Most recent studies controlled for seasonal and other temporal trends, and it was
considered unlikely that inadequate control for such trends seriously biased estimated PM
coefficients. Effect modification by season was examined in several studies. Season-specific
analyses are often not feasible in small-sized studies (due to marginally significant PM effect
size), but some studies (e.g., Samet et al., 1996; Moolgavkar and Luebeck, 1996) suggested that
estimated PM coefficients varied from season to season. It was not fully resolved, however, as
to whether these results represent real seasonal effect modifications or are due to varying extent
of correlation between PM and co-pollutants or weather variables by season.
While most available studies included control for weather variables, some reported
sensitivity of PM coefficients to weather model specification, leading some investigators to
speculate that inadequate weather model specifications may still have erroneously ascribed
residual weather effects to PM. Two PM studies (Samet et al., 1996; Pope and Kalkstein, 1996)
involved collaboration with a meteorologist and utilized more elaborate weather modeling, e.g.,
use of synoptic weather categories. These studies found that estimated PM effects were
essentially unaffected by the synoptic weather variables and also indicated that the synoptic
weather model did not provide better model fits in predicting mortality when compared to other
weather model specifications used in previous PM-mortality studies. Thus, these results
suggested at the time that the reported PM effects were not explained by more sophisticated
synoptic weather models. However, some analyses in both of these studies used GAM,
presumably with default convergence criteria, and therefore need to be interpreted with caution,
especially in light of their not having been reanalyzed with more stringent GAM convergence
criteria and/or by GLM or other types of modeling specifications. Also, reanalyses of other
studies originally using GAM with default convergence criteria have contributed to reopening of
renewed debate on weather model specification issues, as discussed later (in Section 8.4).
Many earlier PM studies considered at least one co-pollutant in the mortality regression,
and some also examined several co-pollutants. In most cases, when PM indices were significant
in single pollutant models, addition of a co-pollutant diminished the PM effect size somewhat,
but did not eliminate the PM associations. When multiple pollutant models were performed by
season, the PM coefficients became less stable, again, possibly due to PM's varying correlation
8-20
-------
with co-pollutants among season and/or smaller sample sizes. However, in many studies, PM
indices showed the highest significance (versus gaseous co-pollutants) in single and multiple
pollutant models. Thus, it was concluded that PM-mortality associations were not seriously
distorted by co-pollutants, but interpretation of the relative significance of each pollutant in
mortality regression as relative causal strength was difficult because of limited quantitative
information on relative exposure measurement/characterization errors among air pollutants.
Measurement error can influence the size and significance of air pollution coefficients
in time-series regression analyses and is also important in assessing confounding among
multiple pollutants, as varying the extent of such error among the pollutants could also influence
the corresponding relative statistical significance. The 1996 PM AQCD discussed several types
of such exposure measurement errors, including site-to-site variability and site-to-person
variability — errors thought to bias the estimated PM coefficients downward in most cases.
However, there was not sufficient quantitative information available to estimate such bias.
The 1996 PM AQCD also reviewed evidence for threshold and various other functional
forms of short-term PM mortality associations. Several studies appeared to suggest that
associations were seen monotonically below the existing PM standards. It was considered
difficult, however, to statistically test for a threshold from available data because of low data
density at lower ambient PM concentrations, potential influence of measurement error, and
adjustments for other covariates. Thus, the use of relative risk (rate ratio) derived from the
log-linear Poisson models was considered adequate and appropriate, although threshold-related
issues remained to be more fully resolved.
The extent of prematurity of death (i.e., mortality displacement or "harvesting") in
observed PM-mortality associations has important public-health-policy implications. At the
time of the 1996 PM AQCD review, only a few studies had investigated this issue. While one of
the studies suggested that the extent of such prematurity might be only a few days, this may not
be generalizable because this estimate was obtained for identifiable PM episodes. There was not
sufficient evidence to suggest the extent of prematurity for non-episodic periods from which
most of the recent PM relative risks were derived. The 1996 PM AQCD concluded:
8-21
-------
"In summary, most available epidemiologic evidence suggests that increased mortality
results from both short-term and long-term ambient PM exposure. Limitations of
available evidence prevent quantification of years of life lost to such mortality in the
population. Life shortening, lag time, and latent period of PM-mediated mortality are
almost certainly distributed over long time periods, although these temporal
distributions have not been characterized." (p. 13-45)
Only a limited number of PM-mortality studies analyzed fine particles and chemically
specific components of PM. The Harvard Six Cities Study (Schwartz et al., 1996a) analyzed
size-fractionated PM (PM2 5, PM10/15, and PM10/15.2 5) and PM chemical components (sulfates
and H+). The results suggested that, among the components of ambient PM, PM2 5 was most
significantly associated with mortality. Because the original study was conducted using GAM
with default convergence criteria, the data were recently reanalyzed (a) by Schwartz (2003a),
who provided reanalyzed PM25 results for each of the six cities and a combined risk estimate
across the six, but only excess risk estimates for individual cities for PM10/15.2.5, and
(b) by Klemm and Mason (2003), who analyzed PM2 5, PM10.15, PM10/15.2 5, and sulfate. Although
the excess risk estimates were somewhat lower than those in the original study, both the
Schwartz (2003a) and Klemm and the Mason (2003) reanalyses confirmed the original findings
with regard to the relative importance of fine versus coarse particles. While H+ was not
significantly associated with mortality in the original and an earlier analysis (Dockery et al.,
1992), the smaller sample size for H+ than for other PM components made a direct comparison
difficult. The 1996 PM AQCD also noted that mortality associations with BS or CoH reported
in earlier studies in Europe and the United States during the 1950s to 1970s most likely reflected
contributions from fine particles, as those PM indices had low 50% cut-points (< 4.5 jim).
Furthermore, certain respiratory morbidity studies showed associations between hospital
admissions/visits with components of PM in the fine particle range. Thus, the U.S. EPA 1996
PM AQCD concluded that there was adequate evidence to suggest that fine particles play
especially important roles in observed PM mortality effects.
Overall, then, the status of key issues as addressed in the 1996 PM AQCD can be
summarized as follows: (1) it was thought that the observed PM effects were unlikely to be
seriously biased by inadequate statistical modeling (e.g., control for seasonality); (2) it also
appeared unlikely that the observed PM effects were seriously confounded by weather (at least
8-22
-------
by synoptic weather models); (3) the observed PM effects appeared to some extent to be
confounded or modified by co-pollutants, and such extent may vary from season to season;
(4) determining the extent of confounding and effect modification by co-pollutants would likely
require knowledge of relative exposure measurement characterization error among pollutants
(there was not sufficient information on this); (5) no compelling evidence substantiating a
threshold for PM-mortality associations was available (statistically identifying a threshold from
existing data was also considered difficult, if not impossible); (6) some limited evidence for
harvesting, a few days of life-shortening, was reported for episodic periods (no study was yet
conducted to evaluate possible harvesting in non-episodic U.S. data); (7) only a relatively limited
number of studies suggested a causal role of fine particles in PM-mortality associations, but in
the light of historical data, biological plausibility, and the results from morbidity studies, a
greater role for fine particles than coarse particles was suggested in the 1996 PM AQCD as being
likely. The 1996 PM AQCD concluded:
The evidence for PM-related effects from epidemiologic studies is fairly strong, with
most studies showing increases in mortality, hospital admissions, respiratory symptoms,
and pulmonary function decrements associated with several PM indices. These
epidemiologic findings cannot be wholly attributed to inappropriate or incorrect
statistical methods, mis-specification of concentration-effect models, biases in study
design or implementation, measurement of errors in health endpoint, pollution exposure,
weather, or other variables, nor confounding of PM effects with effects of other factors.
While the results of the epidemiologic studies should be interpreted cautiously, they
nonetheless provide ample reason to be concerned that there are detectable human
health effects attributable to PM at levels below the current NAAQS. (p. 13-92)
8.2.2.2 Newly Available Information on Short-Term Mortality Effects
Since the 1996 PM AQCD, numerous new studies have examined short-term associations
between PM indices and mortality. Of these studies (more than 80), nearly 70% used GAM
(presumably with default convergence criteria). In the summer of 2002, U.S. EPA asked the
original investigators of some of these studies to reanalyze the data using GAM with more
stringent convergence criteria and GLM with parametric smoothers such as natural splines.
Because the extent of possible bias caused by the default criteria setting in the GAM models is
difficult to estimate for individual studies, the discussion here will focus only on those studies
that did not use GAM Poisson models and those studies that have reanalyzed data using more
-------
stringent convergence criteria and/or alternative approaches. Newly available U.S. and Canadian
studies on relationships between short-term PM exposure and daily mortality that meet these
criteria are summarized in Table 8-1. More detailed summaries of all the short-term exposure
PM-mortality studies, including other geographic areas (e.g., Europe, Asia, etc) are described in
Appendix Table 8A-1. These include the studies that apparently used GAM with default
convergence criteria, and those studies are noted as such. Information on study location and
period, levels of PM, health outcomes, methods, results, and reported risk estimates and lags is
provided in Table 8A-1. In addition to these summary tables, discussion in the text below
highlights findings from several multicity studies (Section 8.2.2.3) and single-city studies
(Section 8.2.2.4). Discussion of implications of new study results for types of issues identified
in foregoing text is mainly deferred to Section 8.4.
The summary of studies in Table 8-1 and 8A-1 (and in other tables) is not meant to imply
that all listed studies should be accorded equal weight in the overall interpretive assessment of
evidence regarding PM-associated health effects. In general, for those studies not clearly flawed
and having adequate control for confounding, increasing scientific weight should be accorded to
in proportion to the precision of their estimate of a health effect. Small studies and studies with
an inadequate exposure gradient generally produce less precise estimates than large studies with
an adequate exposure gradient. Therefore, the range of exposures (e.g., as indicated by the IQR),
the size of the study as indexed by the total number of observations (e.g., days) and total number
of events (i.e., total deaths), and the inverse variance for the principal effect estimate are all
important indices useful in determining the likely precision of health effects estimates and in
according relative scientific weight to the findings of a given study.
As can be seen in Tables 8-1 and 8A-1, many of the newly reported analyses continue to
show statistically significant associations between short-term (24 h) PM exposures indexed by a
variety of ambient PM measurements and increases in daily mortality in numerous U.S. and
Canadian cities, as well as elsewhere around the world. Several newly available PM
epidemiologic studies that conducted time-series analyses in multiple cities, as discussed first
below, are of particular interest.
8-24
-------
TABLE 8-1. RECENT U.S. AND CANADIAN TIME-SERIES STUDIES OF
PM-RELATED DAILY MORTALITY*
Reference
Type** Location(s)/Period
Pollutants
Comments
Multicity Mortality Studies in the U.S. and Canada
oo
I
to
PM10 studies using NMMAPS data
Samet et al. (2000a, b, c);
Dominici et al. (2000a, b);
Dominici et al. (2003a)
Daniels et al. (2000);
Dominici et al. (2003a)
Dominici et al. (2002)
Dominici et al. (2003a)
Studies using every day PM10 data
Schwartz (2000a);
Schwartz (2003b)
Schwartz (2000b);
Schwartz (2003b).
Bragaetal. (2001a);
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.
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)
Same ten U.S. cities as in
(Schwartz, 2000a)
PM10, O3, CO, NO2, SO2
PM,n only
PM10 only
PM10, O3, CO, NO2, SO2
PM10 only.
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 nonparametric spline model for
concentration- response functions. Average
response curve nearly linear.
Smooth nonparametric spline models for PM10
concentration-response functions. Average
response curves are nearly linear in the industrial
Midwest, Northeast regions, and overall, but
nonlinear (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 10 cities for
single day, moving average, and distributed lags.
Pooled estimates across cities evaluated for
deaths due to pneumonia, COPD, CVD, and
MI using distributed lags models.
-------
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
Multicity Mortality Studies in the U.S. and Canada (cont'd)
oo
I
to
Other multicity studies
Schwartz (1996a)
Schwartz (2003a)
Klemm et al., (2000);
Klemm and Mason (2003)
Laden etal.. (2000);
Schwartz (2003a)
Tsaietal. (1999, 2000)
Burnett et al. (2000);
Burnett and Goldberg (2003)
B
Six cities 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.
Same six cities as Harvard Six
City Study (Schwartz et al.,
1996a), 1979-1988.
Same six cities as Harvard Six
City Study (Schwartz et al.,
1996a), 1979-1988.
Camden, Elizabeth, and
Newark, NJ, 1981-1983.
Eight Canadian cities:
Montreal, Ottawa, Toronto,
Windsor, Calgary, Edmonton,
Winnipeg, Vancouver, 1986-
1996.
PM10, PM2 5, PM
sulfates
PM,,
PM2.5,PM10.2.5,
-MO
sulfates
Chemically speciated PM2 5
and factors aligned with
putative sources for each
city identified by specific
chemical elements as
tracers.
PM2 5, PM15, sulfates, trace
elements.
PM,,
PM2.5,PM10.2.5,
sulfates, O3, CO, NO2, SO2.
City-specific associations and combined effect
estimates recalculated for mortality due to all
causes (total), ischemic heart disease, COPD, and
pneumonia. Associations with PM2 5 recalculated
by several techniques, including natural splines,
penalized splines, etc. Associations with PM10_2 5
only recalculated by use of penalized splines for
individual cities.
Replicated Schwartz et al. (1996a) and did
additional sensitivity analyses.
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.
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.
The results of reanalysis indicate no clear
difference between PM2 5 and PM10_2 5 in
associations with mortality.
-------
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
oo
i
to
Moolgavkar (2000a);
Moolgavkar (2003).
Ostroetal. (1999a, 2000);
Ostro et al. (2003)
Fairley (1999);
Fairley (2003)
Schwartz etal. (1999)
Clyde (1999)
Lippmann et al. (2000);
Ito (2003)
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.
Coachella Valley (Palm
Springs), CA, 1989-1998.
A Santa Clara County (San Jose),
CA, 1989-1996.
B Spokane, WA, 1989-1995.
B Chicago/Cook County
1985-1990
Detroit, MI, 1985-1990; 1992-
1994 (separate analysis for two
periods).
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.
PM10 in earlier study, PM2 s
and PM10.2 5 in later study;
O3, CO, NO2. Reanalysis
reported PM risk estimates
only.
PM,,
PM2.5,PM10.2.5,
sulfates, nitrates, O3, CO,
NO2.
PM,n
PM,,
PM10, PM2 5, PM10.2.5,
sulfates, acidity, TSP, O3,
CO, NO2, SO2
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.
PM10 (-65% of which was coarse particles) and
PM10_2 5 (missing values predicted from PM10)
were associated with cardiovascular mortality.
PM2 5 was available for shorter period.
All significant in one-pollutant models, nitrates
significant in all multipollutant 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.
Various lags and fourth degree orthogonal
polynomials of PM10 evaluated using Bayesian
model averaging (BMA). Detected PM10
association with mortality in the elderly
(> 65 years old).
PM mass indices were more strongly associated
with mortality than sulfate or acidity. The extent
of association with health outcomes was similar
forPM25andPM10.25.
-------
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)
Chock et al. (2000)
B Pittsburgh, PA, 1989-1991.
PM10,PM25,PM10.25,03,
CO, NO,, SO,
Fine and coarse particle data on about !/2 of days
with PM10. Data split into ages < 15 and 75+, and
seasons. Significant effects for PM10 but not for
other size fractions.
Klemm and Mason (2000)
Clyde (2000a)
oo
i
to
oo
Schwartz (2000c);
Schwartz (2003a)
Lipfert et al. (2000a)
Mar et al. (2000);
Mar et a. (2003)
Clyde et al. (2000b)
B Atlanta, GA, 1998-1999
(one year).
B Birmingham, AL
August 1985 through
December 1988
Boston, MA, 1979-1986.
B Philadelphia, PA- Camden, NJ
seven- county area, 1995-1997.
Phoenix, AZ, near the EPA
platform monitor, 1995-1997.
B Phoenix, AZ, 1995-1997.
PM,,
PM2.5,PM10.2.5,
oxygenated hydrocarbons
(HC), elemental carbon
(EC), organic carbon (OC),
sulfates, acidity
PM,n
PM,
PM,,
PM2.5,PM10.2.5,
sulfates, acidity, metals,
O3, CO, NO2, SO2
PM,,
PM25,PM10.25,PM25
metals, EC, OC, O3, CO,
NO2, SO2, and source-
apportioned factor scores.
PM25andPM
No significant effects likely due to short time-
series (ca. one year).
PM10 lags up to 3 days and area-wide average of
PM10 for these lags evaluated using Bayesian
model averaging (BMA). Probability intervals
for PM10 effect included both one (suggesting no
effect) and higher estimates reported earlier by
other investigations.
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.
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_25 among 25 "best" models. Estimates
combined using Bayesian model averaging.
-------
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
VO
Smith et al. (2000)
Ostro (1995)
Murray and Nelson (2000)
B Phoenix, AZ (within city and
within county), 1995-1997.
B San Bernardino and Riverside
Counties, CA, 1980- 1986.
Goldberg et al. (2000,
2001a,b,c,d; 2003);
Goldberg and Burnett (2003)
Montreal, PQ, Canada, 1984-
1995
PM,, and PM
•2.5 mlu -1 lvJ-10-2.5
PM2 5 estimated from visual
range, O3
B Philadelphia, PA, 1973- 1990 TSP only
CoH and extinction were
available daily. PM2 5 and
PM10 every sixth day until
1992, daily through 1993.
Significant linear relationship with PM10_2 5,
not PM2 5 Piecewise linear models with possible
PM2 5 threshold for elderly mortality at
20-25 ug/m3.
Positive, significant PM2 5 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.
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.
*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 nonparametric 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 nonparametric smoother.
-------
8.2.2.3 New Multicity Studies
The new multicity studies are of interest here due to their evaluation of a wide range of PM
exposures and large numbers of observations, thus holding promise of possibly providing more
precise effects estimates than most smaller scale independent studies of single cities. Another
potential advantage of the multicity studies, over meta-analyses for multiple "independent"
studies, is consistency in data handling and model specifications that eliminates variation due to
study design. Also, unlike regular meta-analysis, they clearly do not suffer from potential
omission of negative analyses due to "publication bias." Furthermore, geographic patterns of air
pollution effects can be systematically evaluated in multiple-city analyses. Thus, results from
multicity studies have the potential to provide especially valuable evidence regarding relative
homogeneity and/or heterogeneity of PM-health effects relationships across geographic
locations. Also, many of the cities included in these multicity studies were ones for which no
time-series analyses had been previously reported. Most of these new multicity studies used
GAM Poisson models, but the data sets have recently been reanalyzed using GAM models with
more stringent convergence criteria, as well as by using GLM with parametric smoothers.
8.2.2.3.1 U.S. Multicity Studies
U.S. PM10 90-Cities NMMAPS Analyses
The National Morbidity, Mortality, and Air Pollution Study (NMMAPS) focused on time-
series analyses of PM10 effects on mortality during 1987-1994 in the 90 largest U.S. cities
(Samet et al., 2000a,b,c), in the 20 largest U.S. cities in more detail (Dominici et al., 2000a,b),
and PM10 effects on emergency hospital admissions in 14 U.S. cities (Samet et al., 2000a,b).
These NMMAPS analyses employed sophisticated statistical approaches addressing issues of
measurement error biases, co-pollutant evaluations, regional spatial correlation, and synthesis of
results from multiple cities by hierarchical Bayesian meta-regressions and metaanalyses. These
analyses provide extensive new information of much relevance to the setting of U.S. PM
standards, because no other study has examined so many U.S. cities in such a consistent manner.
That is, NMMAPS used only one consistent PM index (PM10) across all cities (based on PM10
samples collected only every 6 days in most of the 90 cities); death records were collected in a
8-30
-------
uniform manner; and demographic variables were uniformly addressed. The 90-cities analyses
studies employ multistage models (see Table 8-1) in which heterogeneity in individual city's
coefficients in the first stage Poisson models were evaluated in the second stage models with
city- or region-specific explanatory variables.
As noted earlier, the original investigators of the NMMAPS study reported in 2002 a
potential problem with using the GAM Poisson models with default convergence criteria
available in widely-used statistical software in estimating air pollution risks (Dominici et al.,
2002). The default convergence criteria were too lax to attain convergence in the setting of air
pollution, weather, and mortality/morbidity parameters where "small" PM regression
coefficients were estimated and at least two covariates were modeled with nonparametric
smoothers. The NMMAS investigators simulation analysis also suggested that the extent of bias
could be more serious when the magnitude of risk coefficient was smaller and when PM
correlations with covariates were stronger. The investigators have since reanalyzed the 90 cities
data, using more stringent convergence criteria as well as using fully parametric smoothers, and
reported revised results. The following description of the NMMAPS mortality study therefore
focuses on the results of the reanalysis of the 90 cities study.
In both the original and reanalyzed NMMAPS 90 cities studies, the combined estimates
of PM10 coefficients were positively associated with mortality at all the lags examined (0, 1, and
2 day lags), although the 1-day lag PM10 gave the largest overall combined estimate. Figure 8-1
shows the reanalyzed results for the estimated percent excess total deaths per 10 |ig/m3 PM10 at
lag 1 day in the 88 (90 minus Honolulu and Anchorage) largest cities, as well as (weighted
average) combined estimates for U.S. geographic regions depicted in Figure 8-2. The majority
of the coefficients were positive for the various cities listed along the left axis of Figure 8-1. The
estimates for the individual cities were first made separately. The cities were then grouped into
the 7 regions seen in Figure 8-2 (based on characteristics of the ambient PM mix typical of each
region, as delineated in the 1996 PM AQCD). The bolded segments represent the posterior
means and 95% posterior intervals of the pooled regional effects without borrowing information
from other regions. The triangle and bolded segment at the bottom of Figure 8-1 display the
combined estimate of overall nationwide effects of PM10 for all the cities.
8-31
-------
-6
% Increase in Mortality per 10 Unit Increase in PM10
-4-2024 6
1
San Jose— —
Salt Lake City-— o—
Stockton o
Modesto o
Overall Olympia
San Antonio
El Paso-
Albuquerque
Overall Lubbock ° _
_
San Bernardino— ——— o—
Fresno~—
Overall """""""
Minneapolis / St. Paul— o
Wichita o
Kansas City (Wyand otte)
Overall _
Chicago
Cleveland— o-
Columbus-
o
Indianapolis
Dayton- o
o
Grand Rapids o
Madison ZHŁ
Lex i n gto n — o
Overall
Philadelphia-
Baltimore —
Worchester
Washington——— o
Jersey City-
Arlington -^^^^^^^^^-^^-
Ove ra 1 1 ^^^^^^^~^^^^^^^~
Miami
St . Pete rsbu rg
Memphis
O
Jacksonville
Charlotte —
-o
Tulsa-
JJttkBcsk 0
o
Knoxville o
Jackson ^IIIIIIIIIIl
Overall ;
Overall
i
-o Seattle
o Oakland
Sacramento
o Tacoma
o Spokane
*-
Northwest
o Phoenix
o
o Oklahoma City
o Austin Southwest
o
Corpus Christ!
•-
n — Los Angeles
o Santa Ana / Anahei
O
-g-o Bakersfield
n Southern
California
-o Kansas Cityiciay, Jackson, piatiei Upper
~Des Moines Midwest
° .Q. lopeka
l-
-o Detroit
D — Pittsburgh
— Buffalo
^Cincinnati Industrial
° -o Louisville Midwest
-Akron
o Toledo
P 1 1 B i
I Eyansville
«-
O New York
O
~o Newark
o
O Rochester
o Boston
^=<^= Providence Northeast
-o-
O Syracuse
o ^^^^J^JDgsloiL^^^.
o— HHouston
----O--------------
-o- -
O ^Tampa
O
Orlando
O B irmingham _ .. .
o -- Southeast
-Nashville
o
-New Oreans
5 Baton Rouge
G re en s bo ro
••-
->-
Figure 8-1. 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, 2003b).
8-32
-------
Northwest
Southern
California
Upper
Midwest
Industrial
Midwest
Northeast
Figure 8-2. 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.
Note that there appears to be some regional-specific variation in the overall combined
estimates for all the cities in a given region. This can be discerned most readily in Figure 8-3,
which depicts overall region-specific excess risk estimates for 0-, 1-, and 2-day lags. For
example, the coefficients for the Northeast for any given lag are generally higher than for other
regions. The NMMAPS investigators noted that the extent of the regional heterogeneity seen
with the reanalysis was reduced slightly compared to the original finding (between-city standard
deviation changed from 0.112 to 0.088 in the unit of percent excess deaths per 10 |ig/m3 PM10),
but the pattern of heterogeneity remained the same. The overall national combined estimate
(i.e., at lag 1 day, 1.4% excess total deaths per 50 |ig/m3 increase in PM10 using GAM with
stringent convergence criteria) for the 90 cities is somewhat lower than the range of available
estimates reported in the 1996 PM AQCD (i.e., 2.5 to 5.0%) for the general U.S. population.
In the original 90 cities study, the weighted second-stage regression included five types of
county-specific variables: (1) mean weather and pollution variables; (2) mortality rate (crude
mortality rate); (3) sociodemographic variables (% not graduating from high school and median
8-33
-------
% Increase in Mortality per 10 Unit Increase in PM;c
Nc
Sc
Sc
Cs
Uf
Mi
In
Mi
Nc
Sc
O\
-1 -0.5 (
I I
NW.IagO mm
irthwest
NW.IagZ
1
uthwest SW|ag1
SC.IagO
uthern
ilifornia
SC.Iag2 —
'Per UMW.iagl
dwest
IMW.IagO —
Austria!
dwest
IMW.Iag2 —
jrtheast NE.Iagl
SE.IagO mmmi
utheast ~
SE.Iag2
rerall Overall. Iag1
) 0.5 1
I I
•
— **** NW.lagl
-mmmt —
mmm SW.IagO
— ^^m
-mmm SVtf.lag2
-mmm*
— mmmt
-•••i UMW.IagO
-mmmi
•mmmm — UMW.Iag2
mmm —
^mmm — iMW.Iagl
••• —
— mmm NE.IagO
mmm NE.Iag2
mmmi SE.Iagl
mmm
mmm — Overall. lagO
mmm — Overall. Iag2
Figure 8-3. 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 8-2.
Source: Dominici et al. (2002, 2003b).
8-34
-------
household income); (4) urbanization (public transportation); and (5) variables related to
measurement error (median of all pair-wise correlations between monitors). Some of these
variables were apparently correlated (e.g., mean PM10 and NO2, household income and
education) so that the sign of coefficients in the regression changed when correlated variables
were included in the model. Thus, while some of the county-specific variables were statistically
significant (e.g., mean NO2 levels), interpreting the role of these county-specific variables may
require caution. Regarding the heterogeneity of PM10 coefficients, the investigators concluded
that they "did not identify any factor or factors that might explain these differences."
Another important finding from Samet and coworkers' analyses was the weak influence of
gaseous co-pollutants on the PM10 effect size estimates (see Figure 8-4). In the reanalysis of
90 cities data, PM10 coefficients slightly increased when O3 was added to regression models.
Additions of a third pollutant (i.e., PM10 + O3 + another gaseous pollutant) hardly changed the
posterior means of PM10 effect size estimates, but widened the distribution. However, the
posterior probabilities that the overall PM10 effects are greater than zero remained at or above
0.96. The gaseous pollutants themselves in single-, two-, and three-pollutant models were less
consistently associated with mortality than PM10. Ozone was not associated with mortality using
year-round data; but, in season-specific analyses, it was associated with mortality negatively in
winter and positively in summer. SO2, NO2, and CO were weakly associated with mortality, but
additions of PM10 and other gaseous pollutants did not always reduce their coefficients, possibly
indicative of their independent effects. As noted in Section 8.1, CO and NO2 from motor
vehicles are likely confounders of PM2 5 and, thus, of PM10 when it is not dominated by the
coarse particle fraction. The investigators stated that the PM10 effect on mortality "was
essentially unchanged with the inclusion of either O3 alone or O3 with additional pollutants."
The reanalyses of the 90 cities data by the original NMMAPS investigators also included a
sensitivity analysis of lag Iday PM10 GLM results to the alternative degrees of freedom for
adjustment of the confounding factors: season, temperature, and dewpoint. The degrees of
freedom for each of these three smoothing terms were either doubled or halved, resulting in nine
scenarios in addition to the degrees of freedom in the original GLM model. The PM10 effect
posterior means were generally higher when the degrees of freedom were halved for season, and
8-35
-------
PM10 I
PM10+O3
PM10+03+N02
PM10+O3+SO2
PM10+O3+CO
-0.25 0.0 0.25 0.5 0.75
% Change in Mortality per 10 pg/m3 increase in PM10
Figure 8-4. 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. (2003b).
lower when they were doubled, ranging between 1.6% to 0.9% (the main GLM result was 1.1%)
excess total mortality per 50 |ig/m3 PM10 increase. These results underscore the fact that the
magnitude of sensitivity of the results due to model specification (in this case, degrees of
freedom alone) can be as great as the potential bias caused by the GAM convergence problem.
HEI (2003a) states that the revised NMMAPS 90 individual-city mortality results show
that, in general, the estimates of PM effect are shifted downward and the confidence intervals are
widened. In the revised analyses, a second stage meta-analysis was used to combine results on
effects of PM and other pollutants on health outcomes across cities. Tightening the convergence
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
8-36
-------
estimated 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
co-pollutants were introduced into the model (Samet et al., 2000a); and this conclusion has not
changed.
The extent of reduction in PM10 excess risk estimate due to the change in the convergence
criteria (2.3% per 50 |ig/m3 PM10 using default versus 1.4% using stringent) using GAM models
in the 90 cities study appears to be greater than those reported for most of the other reanalysis
studies. This may be partly due to the smaller risk estimate (2.3%) in the original study
compared to other studies (> 3%), as the smaller coefficient is likely more strongly affected as a
relative reduction. This may also be in part due to the more "aggressive" adjustment for possible
weather effects (discussed later) used in this study, which may have increased the concurvity
between PM and the covariates (which included four smoothing terms for weather adjustment).
Dominici et al. (2002) reported that the higher the concurvity, the larger the potential bias that a
GAM model with default convergence criteria could produce.
In summary, the 90-cities NMMAPS study provides useful information regarding:
(1) the magnitude of the combined PM10 risk estimate; (2) lack of sensitivity of PM10 risk
estimates to gaseous co-pollutants; (3) indications of regional heterogeneity in PM10 risk
estimates across the United States; (4) the shape of concentration-response relationship
(discussed later in Section 8.4); and (5) the range of sensitivity of PM10 risk estimates to the
extent of smoothing of covariates in the original weather model specification. A major
uncertainty not extensively examined in this study is the sensitivity of the PM10 risk estimates to
different weather model specifications.
U.S. 10-Cities Studies
In another set of multicity analyses, Schwartz (2000a,b), Schwartz and Zanobetti (2000),
Zanobetti and Schwartz (2000), Braga et al. (2000), and Braga et al. (2001a) analyzed 1987-1995
air pollution and mortality data from ten U.S. cities (New Haven, CT; Birmingham, AL;
8-37
-------
Pittsburgh, PA; Detroit, MI; Canton, OH; Chicago, IL; Minneapolis-St. Paul, MN; Colorado
Springs, CO; Spokane, WA; and Seattle, WA.) or subsets (4 or 5 cities) thereof. The selection of
these cities was based on the availability of daily (or near daily) PM10 data. All of these original
studies utilized GAM Poisson models with default convergence criteria. Of these studies,
Schwartz (2003b) reanalyzed the data from Schwartz (2000a), Schwartz (2000b), and Braga
et al. (200la) using GAM with stringent convergence criteria as well as alternative models such
as GLM with natural cubic splines or penalized splines, both of which are expected to give
correct standard errors. The main original results of the study were presented in the Schwartz
(2000a) paper; and the other studies noted above focused on each of several specific issues,
including potential confounding, effect modification, distributed lag, and threshold. In this
section, the results for the three reanalysis studies noted above are discussed.
In the reanalysis (Schwartz, 2003b) of the main results (Schwartz, 2000a), daily total
(nonaccidental) mortality in each of the 10 cities was fitted using a GAM Poisson model (with
stringent convergence criteria) or a GLM Poisson model with natural splines, adjusting for
temperature, dewpoint, barometric pressure, day-of-week, season, and time. The data were also
analyzed by season (November through April as heating season). The inverse-variance weighted
averages of the ten cities' estimates were used to combine results. PM10 (average of lag 0 and
1 days) was significantly associated with total deaths, and the effect size estimates were
comparable in summer and winter. Adjusting for other pollutants did not substantially change
the PM10 effect size estimates. The combined percent-excess-death estimate for total mortality
was 3.4% (CI: 2.6, 4.1)1 per 50 |ig/m3 increase in the average of lag 0 and 1 days PM10
(essentially unchanged from the original study) using GAM with stringent convergence criteria.
The PM10 risk estimate using GLM with natural splines was 2.8% (CI: 2.0, 3.6).
In the reanalysis (Schwartz, 2003b) of the study of multiday effects of air pollution
(Schwartz, 2000b), constrained (quadratic model over 0 through 5 day lags) and unconstrained
(0 through 5 day lags) distributed lag models were fitted in each city. The overall estimate was
computed using the inverse-variance weighted average of individual city estimates. Among the
1 95% Confidence Intervals for a given percent risk estimate are standardly provided in parenthesis following the
risk estimate in this chapter. For example, 95% CI = 2.6 to 4.1% is expressed as (CI: 2.6,4.1).
8-38
-------
results obtained using GAM with stringent convergence criteria, the PM10 effect size estimate
was 6.3% (CI: 4.9, 7.8) per 50 |ig/m3 increase for the quadratic distributed lag model, and
5.8% (CI: 4.4, 7.3) for the unconstrained distributed lag model. Corresponding values using the
penalized splines were somewhat smaller (-5.3%). These values are about twice the effect-size
estimate for single-day PM10 in the original report or the two-day mean PM10 reported in the
reanalysis above (this reanalysis did not report results for single-day or 2-day mean PM10).
Schwartz (2003b) also reanalyzed the data from Braga et al.'s (2001a) study to examine the
lag structure of PM10 association with specific cause of mortality in the 10 cities. Unconstrained
distributed lags for 0 through 5 days as well as two-day mean were fitted in each city for COPD,
pneumonia, all cardiovascular, and myocardial infarction deaths using GAM with stringent
convergence criteria and penalized spline models. Combined estimates by lag were obtained
across the 10 cities. The distributed lag estimates were generally larger than the two-day mean
estimates for COPD and pneumonia mortality, but they were comparable for all cardiovascular
and myocardial infarction mortality. For example, in the results using GAM with stringent
convergence criteria, the PM10 effect size estimate for COPD mortality was 11.0% (CI: 7.2, 14.8)
per 50 |ig/m3 increase for two-day mean model and 16.8% (CI: 8.3, 25.9) for the unconstrained
distributed lag model. Note that these values are substantially larger than those reported for total
nonaccidental deaths.
The PM10 risk estimates from these 10 cities analyses are larger than those from the
NMMAPS 90 cities study and these results suggest a possibility that PM effects may be
underestimated when only single-day PM indices are used. That is, if ambient PM effects on
mortality occur only very immediately, e.g., the same day, then the full risk would be reflected
by single day lag analyses. However, if PM-associated deaths occurred over a more extended
time, e.g., the next several days, then the fuller PM-related mortality risk would presumably be
more closely reflected by distributed lag models and, overall, would logically be higher than for
any single lag day. Differences in the number of cities analyzed, in weather model specification,
and/or in the extent of smoothing for temporal trends may also have contributed to the
differences in the size of PM10 risk estimates found by the NMMAPS 90 cities versus the
Schwartz 10 cities studies. These issues are further discussed in Section 8.2.2.3.5.
8-39
-------
Reanalyses of Harvard Six Cities Study
Both the original Harvard Six Cities Study time-series analysis (Schwartz et al., 1996a) and
the replication analysis by Klemm et al. (2000), which essentially replicated Schwartz et al.'s
original findings, used GAM Poisson models with default convergence criteria. Schwartz
(2003a) and Klemm and Mason (2003) conducted reanalyses of the Harvard Six Cities data to
address the GAM statistical issues.
Schwartz (2003a) not only reported risk estimates for PM2 5 and PM10_2 5, but also provided
results using several other spline smoothing methods (natural splines, B-splines, penalized
splines, and thin plate splines) in addition to GAM with stringent convergence criteria. The risk
estimate combined across the six cities per 25 |ig/m3 in PM25 (average of lag 0 and 1 day) using
GAM with stringent convergence criteria was 3.5% (CI: 2.5, 4.5), as compared to the original
value of 3.7% (CI: 2.7, 4.7). The corresponding value from a GLM model with natural splines
was 3.3% (CI: 2.2, 4.3); and the values using B-splines, penalized splines, and thin plate splines
were somewhat lower (3.0%, 2.9%, and 2.6%, respectively). However, when the Harvard Six
Cities were examined individually in the reanalysis of Schwartz using GLM and penalized
splines, Boston and St. Louis gave significant associations with PM2 5 and Steubenville gave a
significant association with "thoracic" coarse PM (i.e. PM10_2 5).
The Klemm and Mason (2003) reanalysis calculated risk estimates for PM2 5, PM10_2 5, PM10
(PM15 or PM10), and SO42 . They also conducted sensitivity analyses using GLM with natural
splines that approximated the degrees of freedom used in the LOESS smoothers in the GAM
models, as well as 12 knots per year and 4 knots per year for smoothing of temporal trends.
The PM2 5 and PM10_2 5 total nonaccidental mortality risk estimates combined across the six cities
per 25 |ig/m3 (average of lag 0 and 1 day) using GAM with stringent convergence criteria were
3.0% (CI: 2.1, 4.0) and 0.8% (CI: -0.5, 2.0), respectively. The corresponding PM10 mortality
excess risk estimate per 50 |ig/m3 (average of lag 0 and 1 day) was 3.6% (CI: 2.1, 5.0). In their
sensitivity analysis, increasing the degrees of freedom for temporal trends for natural splines in
GLM models from 4 knots/year to 12 knots/year markedly reduced PM risk estimates. For
example, the PM2 5 risk estimate per 25 |ig/m3 was reduced from 2% in the 4 knots/year model to
1% in the 12 knots/year model. The results showing the smaller PM risk estimates for larger
8-40
-------
degrees of freedom for smoothing of temporal trends are consistent with similar findings
reported for the reanalysis of the NMMAPS 90 cities study. It also should be noted that Klemm
and Mason (2003) reported positive (but not significant at p < .05) associations between total
mortality and PM10_25 in Steubenville, which parallels the Schwartz (2003a) finding of a positive
(and statistically significant at p < .05) total mortality association with PM10_2 5 for Steubenville.
Although PM effect estimates from the Klemm and Mason (2003) reanalysis are somewhat
smaller than those from Schwartz (2003a); e.g., 3.5% by Schwartz versus 3.0% by Klemm and
Mason for PM2 5 using strict convergence criteria, the results are very similar. Both studies also
showed that the comparable GLM models produced smaller risk estimates than GAM models.
8.2.2.3.2 Canadian Multicity Studies
Burnett et al. (2000) analyzed various PM indices (PM10, PM25, PM10_25, sulfate, CoH,
and 47 elemental component concentrations for fine and coarse fractions) and gaseous air
pollutants (NO2, O3, SO2, and CO) for association with total mortality in the 8 largest Canadian
cities: Montreal, Ottawa-Hull, Toronto, Windsor, Winnipeg, Calgary, Edmonton, and
Vancouver. This study differs from Burnett et al. (1998a) in that it included fewer cities but
more recent years of data (1986 to 1996 versus 1980 to 1991) and detailed analyses of particle
mass components by size and elemental composition. Each city's mortality, pollution, and
weather variables were separately filtered for seasonal trends and day-of-week patterns. The
residual series from all cities were then combined and analyzed in a GAM Poisson model.
In Burnett and Goldberg's reanalysis (2003) of the eight Canadian cities data, they only
evaluated the PM indices (PM2 5, PM10_2 5, and PM10) using GAM models with more stringent
convergence criteria. The reanalysis used co-adjustment regression (i.e., simultaneous
regression), rather than the regression with prefiltered data that was the main approach of the
original analysis. The reanalysis also considered several sensitivity analyses, including models
with and without day-of-week adjustment and several alternative approaches (fitting criteria and
extent of smoothing) to adjust for temporal trends using natural splines. Adjusting for temporal
trends, smoothing of same-day temperature, pressure, and day-of-week effects, the pooled PM
effect estimates across the eight Canadian cities were: 2.2% (CI: 0.1, 4.2) per 25 |ig/m3 increase
8-41
-------
in PM25; 1.8% (CI: -0.6, 4.4) per 25 |ig/m3 increase PM10.25; and 2.7% (CI: -0.1, 5.5) per
50 |ig/m3 increase PM10. These effect size estimates are fairly close to the estimates reported in
the original study, despite the differences in the regression approach (prefiltering and GAM with
default convergence criteria in the original study versus use of co-adjustment and GAM with
stringent convergence criteria in the reanalyses).
The temporal adjustment of the above model used LOESS smoothing with a span of
-0.022 (= 90 days/4,012 study days). Sensitivity analysis included several choices of degrees of
freedom for natural splines of temporal trend, with two fitting criteria (i.e., Bartlett's test for
white noise and Akaike Information Criterion [AIC]) and either using the same degrees of
freedom for all eight cities or varying degrees of freedom for each city. The PM risk estimates
based on natural splines were generally smaller than those based on LOESS smoothers. The PM
risk estimates also varied inversely with the number of knots for temporal trend. That is, the
more details of the temporal trend were described by natural splines, the smaller the PM risk
estimates became. The reported PM25 risk estimates per 25 |ig/m3 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 knot/year, knot/6
months, knot/3 months, knot/2 months, and knot/1 month, respectively. The corresponding
values for 25 |ig/m3 increase in PM10.25 were 3.9% (t = 3.42), 2.9% (t = 2.52), 2.1% (t = 1.69),
1.8% (t = 1.46), and 1.2% (t = 0.91), suggesting greater sensitivity of PM10_25 risk estimates to
the extent of temporal smoothing. The authors suggested that this was likely due to the stronger
correlation between (and temporal trends in) mortality and mass concentrations for PM10_2 5
(average correlation among cities of-0.45) than for PM25 (-0.36). Because the relative size and
significance of PM25 and PM10_25 risk estimates varied depending on the model and extent of
smoothing for temporal trend, it is difficult to compare the relative importance of the two size-
fractionated PM indices in this study; but, overall, they do not appear to be markedly different.
8.2.2.3.3 European Multicity APHEA Study Analyses
The Air Pollution and Health: A European Approach (APHEA) project is a multicenter
study of short-term effects of air pollution on mortality and hospital admissions within and
across a number of European cities having a wide range of geographic, climatic, air quality, and
8-42
-------
sociodemographic patterns. The obvious strength of this approach is its ability to evaluate
potential confounders or effect modifiers in a consistent manner. However, it should be noted
that PM indices measured in those cities varied. In APHEA1, the PM indices measured were
mostly black smoke (BS), except for: Paris and Lyon (PM13); Bratislava, Cologne, and Milan
(TSP); and Barcelnoa (BS and TSP). In APHEA2, 10 out of the 29 cities used direct PM10
measurements; and, in 11 additional cities, PM10 levels were estimated based on regression
models relating collocated PM10 measurements to BS or TSP. In the remaining 8 cities, only BS
measurements were available (14 cities had BS measurements). As discussed below, there have
been several papers published that present either a meta-analysis or pooled summary estimates of
these multicity mortality results: (1) Katsouyanni et al. (1997) — SO2 and PM results from
12 cities; (2) Touloumi et al. (1997) — ambient oxidants (O3 and NO2) results from six cities;
(3) Zmirou et al. (1998) — cause-specific mortality results from 10 cities (see Section 8.2.2.5);
(4) Samoli et al. (2001) — a reanalysis of APHEA1 using a different model specification (GAM)
to control for long-term trends and seasonality; and (5) Katsouyanni et al. (2001) — APHEA2,
with emphasis on the examination of confounding and effect modification. The original APHEA
protocol used sinusoidal terms for seasonal adjustment and polynomial terms for weather
variables in Poisson regression models. Therefore, publications 1 through 3 above are not
subject to the GAM default convergence issue. Publications 4 and 5 did use GAM Poisson
model with default convergence criteria, but the investigators have reanalyzed the data using
GAM with more stringent convergence criteria, as well as GLM with natural splines (Katsouyani
et al., 2003; Samoli et al., 2003). The discussions presented below on publications 4 and 5 are
focused on the results from the reanalyses.
APHEA1 Sulfur Dioxide and Particulate Matter Results for 12 Cities
The Katsouyanni et al. (1997) analyses evaluated data from the following cities: Athens,
Barcelona, Bratislava, Cracow, Cologne, Lodz, London, Lyon, Milan, Paris, Poznan, and
Wroclaw. In the western European cities, an increase of 50 |ig/m3 in BS or SO2 was associated
with a 3% (CI: 2.0, 4.0) increase in daily mortality; and 2% (CI: 1.0, 3.0) per 50 |ig/m3 increase
in estimated PM10 (based on PM10 = TSP*0.55 conversion). In the 31 central/eastern European
8-43
-------
cities, the increase in mortality was 0.6% (CI: 0.1, 1.1) per 50 |ig/m3 change in BS and 0.8%
(CI: 0.1, 2.4) per 50 |ig/m3 change for SO2. Estimates of cumulative effects of prolonged (two to
four days) exposure to air pollutants were comparable to those for one day effects. The effects
of both pollutants (BS, SO2) were stronger during the summer and were mutually independent.
Regarding the contrast between the western and central/eastern Europe results, the authors
speculated that this could be due to differences in exposure representativeness; differences in
pollution toxicity or mix; differences in proportion of sensitive subpopulation; and differences in
model fit for seasonal control. Bobak and Roberts (1997) commented that the heterogeneity
between central/eastern and western Europe could be due to the difference in mean temperature.
However, Katsouyanni and Touloumi (1998) noted that, having examined the source of
heterogeneity, other factors could apparently explain the difference in estimates as well as or
better than temperature.
APHEA1 Ambient Oxidants (Ozone and Nitrogen Dioxide) Results for Six Cities
Touloumi et al. (1997) reported on additional APHEA data analyses, which evaluated
(a) short-term effects of ambient oxidants on daily deaths from all causes (excluding accidents),
and (b) impacts on effect estimates for NO2 and O3 of including a PM measure (BS) in
multipollutant models. Six cities in central and western Europe provided data on daily deaths
and NO2 and/or O3 levels. Poisson autoregressive models allowing for overdispersion were
fitted. Significant positive associations were found between daily deaths and both NO2 and O3.
Increases of 50 |ig/m3 in NO2 (1-h maximum) or O3 (1-h maximum) were associated with a 1.3%
(CI: 0.9, 1.8) and 2.9% (CI: 1.0, 4.9) increase in the daily mortality, respectively. There was a
tendency for larger effects of NO2 in cities with higher levels of BS; that is, when BS was
included in the model, the coefficient for NO2 was reduced by half (but remained significant)
whereas the pooled estimate for the O3 effect was only slightly reduced. The authors speculated
that the short-term effects of NO2 on mortality might be confounded by other vehicle-derived
pollutants (e.g., airborne ambient PM indexed by BS measurements). Thus, while this study
reports only relative risk levels for NO2 and O3 (but not for BS), it illustrates the potential
8-44
-------
importance of confounding between NO2 and PM effects and relative limited confounding
between O3 and PM effects.
APHEA1: A Sensitivity Analysis for Controlling Long-Term Trends and Seasonality
The original study (Samoli et al., 2001) examined the sensitivity of APHEA1 results to
how the temporal trends were modeled (i.e., sine/cosine in the APHEA1 versus LOESS
smoother using GAM with default convergence criteria). Samoli et al. (2003) reanalyzed the
data using GAM with more stringent convergence criteria, as well as GLM with natural splines.
The reanalysis allowed a comparison of results across a fixed functional model (sine/cosine),
a nonparametric smoother (GAM with LOESS), and a parametric smoother (GLM with natural
splines). The combined estimate across cities for percent excess in total nonaccidental mortality
per 50 |ig/m3 increase in BS using GAM with stringent convergence criteria (2.3%; CI: 1.9, 2.7)
was bigger than that using sine/cosine (1.3%; CI: 0.9, 1.7). The GAM with stringent
convergence criteria reduced the combined estimate by less than 10% versus that from GAM
with default convergence criteria. The corresponding estimate using GLM with natural splines
(1.2%; CI: 0.7, 1.7) was comparable to that from the sine/cosine model but smaller than that
using GAM. The contrast between western and eastern Europe in the original APHEA1 study
(2.9% for west versus 0.6% for east) was less clear in results using GAM with stringent
convergence criteria (2.7% versus 2.1%) or GLM with natural splines (1.6% versus 1.0%). This
suggests that the apparent regional heterogeneity found in the original APHEA1 study could be
sensitive to model specification. Because the number of cities used in the APHEA1 study was
relatively small (eight western and five central-eastern cities), apparent regional heterogeneity
found in the earlier publications could be due to chance. These reanalysis findings also suggest
that the results are somewhat sensitive to model specification for temporal trends.
APHEA2: Confounding and Effect Modification Using Extended Data
The APHEA2 original study (Katsouyanni et al. 2001) included more cities (29 cities) and
a more recent study period (variable years in 1990-1997, compared to 1975-1992 in APHEA1).
Also, the APHEA2 original study used a GAM (with default convergence criteria) Poisson
8-45
-------
model with LOESS smoothers to control for season and trends. Katsouyanni et al. (2003)
reanalyzed the data using GAM with more stringent convergence criteria and two parametric
approaches: natural splines and penalized splines. Because the reanalysis GAM results changed
the PM10 risk estimates only slightly from the original estimates and the investigators mention
that the patterns of effect modification were preserved in their reanalyses regardless of model
specification, the qualitative description below of the effect modification relies on the original
study, but PM10 estimates for various models are from the reanalysis.
The analyses put emphasis on effect modification by city-specific factors. Thus, the city-
specific coefficients from the first stage of Poisson regressions were modeled in the second stage
regression using city-specific characteristics as explanatory variables. Inverse-variance
weighted pooled estimates (fixed-effects model) were obtained as part of this model. When
substantial heterogeneity was observed, the pooled estimates were obtained using random-effects
models. These city-specific variables included (1) air pollution level and mix, such as average
air pollution levels and PM/NO2 ratio (as an indicator of traffic-generated PM); (2) climatic
variables, such as mean temperature and relative humidity; (3) health status of the population,
such as the age-adjusted mortality rates, the percentage of persons over 65 years of age, and
smoking prevalence; and (4) geographic area (three regions: central-eastern, southern, and
north-western Europe). The study also addressed the issue of confounding by simultaneous
inclusion of gaseous co-pollutants in city-specific regressions and obtained pooled PM estimates
for each co-pollutant included. Unlike APHEA1, in which the region (larger PM estimates in
western Europe than in central-eastern Europe) was highlighted as the important factor,
APHEA2 found several effect modifiers. NO2 (i.e., index of high pollution from traffic) was an
important one. Cities with higher NO2 levels showed larger PM effects, as did cities with a
warmer climate. The investigators noted that this might be due to better indexing of population
exposures by outdoor community monitors (because of more open windows). Also, cities with
low standardized mortality rate showed larger PM effects. The investigators speculated that this
may be because a smaller proportion of susceptible people (to air pollution) are available in a
population with a large age-standardized mortality rate. Interestingly, in the pooled PM risk
estimates from models with gaseous pollutants, it was also NO2 that affected (reduced) PM risk
8-46
-------
estimates most. For example, in the fixed-effects models, -50% reductions in both PM10 and BS
coefficients were observed when NO2 was included in the model. SO2 only minimally reduced
PM coefficients; whereas O3 actually increased PM coefficients. Thus, in this analysis, NO2 was
implicated as a confounder, an effect modifier, and/or as an indicator of PM source. The overall
random-effects model combined estimates for total mortality for 50 |ig/m3 increase in PM10 were
3.0% (CI: 2.0, 4.1), 2.1% (CI: 1.2, 3.0), and 2.8% (CI: 1.8, 3.8), for GAM (stringent convergence
criteria), natural splines, and penalized splines models, respectively. The original excess
mortality risk estimate (3.1%) using GAM with default convergence criteria was thusly reduced
by 3 to 33% in the different reanalyses models. While the effect estimates varied somewhat
depending on the choice of GAM with LOESS, natural splines, or penalized splines, the patterns
of effect modification (by NO2, etc.) were preserved.
8.2.2.3.4 Comparison of Effect Size Estimates from Multicity Studies
Based on different pooled analyses of data combined across multiple cities, the percent
excess total (nonaccidental) deaths estimated per 50 |ig/m3 increase in PM10 in the above
multicity studies were (a) 1.4% using GAM (1.1% using GLM) at lag 1-day in the 90 largest
U.S. cities (the Northeast region results being about twice as high); (b) 3.4% using GAM (2.8%
using GLM) for average of 0 and 1 day lags in 10 U.S. cities; (c) 3.6% using GAM (2.7% using
GLM) for 1 day lag PM10 in the 8 largest Canadian cities; and (d) 3.0% using GAM (2.1% using
GLM) in APHEA2 for average of 0 and 1 day lags for 29 European cities during 1990-1997.
The estimate for NMMAPS 90 cities study is somewhat smaller than those for the other
multicity studies and the range reported in the previous PM AQCD (2.5 to 5%). There are
several possible explanations for this, including differences in (a) model specifications for
weather, (b) extent of smoothing to adjust for temporal trends, (c) use of different specific
smoothing approaches, and (d) consequent effects of each of these differences on ranges of
degrees of freedom assigned for different aspects of the analyses.
Model specification for weather appears to be one of the more crucial factors. The
NMMAPS 90 cities study used much more "aggressive" adjustment for possible weather effects
than did most other studies. The 90 cities analysis included four separate weather terms:
8-47
-------
(1) smoothing splines (natural splines when GLM was used) of same-day temperature with
6 degrees of freedom; (2) smoothing splines of the average of lag 1 through 3 day temperature
with 6 degrees of freedom; (3) smoothing splines of same-day dewpoint with 3 degrees of
freedom; and (4) smoothing splines of the average of lag 1 through 3 day dewpoint with
3 degrees of freedom. In contrast, most of the other studies used only one or two terms for
weather variables. For example, the Harvard Six Cites Study used a LOESS smoother (or
natural splines or other smoothers in reanalysis) of same-day temperature with a span of 0.5 and
a LOESS smoother of same-day dewpoint with a span of 0.5. Note, too, that the NMMAPS
90 cities study not only used more terms for weather effects, but it also used more degrees of
freedom for temperature than the Schwartz (2003b) analysis (according to the Klemm and
Mason [2003] reanalysis, the span of 0.5 in LOESS corresponds to -3.5 degrees of freedom).
It should be noted that the purpose for inclusion of dewpoint in these models is often explained
as to adjust for possible effects of humidity; but there are differing perspectives related to the
need for inclusion of dewpoint along with temperature, given that dewpoint and temperature
tend to be highly correlated (r > 0.9) in most cities. Thus, although the inclusion of these terms
may statistically (i.e., by AIC, etc.) provide a better fit, the epidemiologic implications of the use
of these terms are not yet fully clear. On the one hand, extreme temperatures, hot or cold, are
known to cause excess mortality and the combined effects of high temperature and high
humidity occurring together can cause especially high excess mortality. Thus, it is clear that
these models need somehow to control for high "heat index" effects when notable increases in
weather-related mortality occur (hence the need for heat index forecasts) and/or for cold-induced
deaths when sufficiently low temperatures occur in a given locale. On the other hand, it is also
not clear at this time as to whether these models may be overcorrecting for weather effects in the
more moderate range that is typical of much of the data. These issues are further discussed later
in Section 8.4.3.
Another factor that may contribute to the difference in PM risk estimates is the extent of
smoothing to adjust for temporal trends. Several of the reanalysis studies (Dominici et al., 2002;
Burnett and Goldberg, 2003; Ito, 2003; Klemm and Mason, 2003) consistently reported, though
to varying extents, that using more degrees of freedom for temporal trends tended to reduce PM
8-48
-------
coefficients. That is, when more details in the short-term fluctuations of mortality were ascribed
to temporal trends, PM risk estimates were reduced. For example, in Dominici et al.'s (2002)
sensitivity analysis, the PM10 risk estimate was larger (1.6% per 50 |ig/m3 increase in PM10) for
the GLM model with 3 degrees of freedom per year that the estimate using 7 degrees of freedom
(1.1%). Note that, in general, the presumed objective of including temporal trends in the
mortality regression is to adjust for potential confounding (measured or unmeasured) by time-
varying factors that change seasonally or in shorter time spans (e.g., influenza epidemics).
However, ascribing "too short" temporal fluctuations to these "confounding temporal trends"
may inadvertently take away PM effects. Because the "right" extent of smoothing is not known,
these sensitivity analyses are useful. In the reanalyses mentioned above, the PM risk estimates
changed, at times, by a factor of two when a range of degrees of freedom was applied even for a
model specification in which all the other terms were kept unchanged.
Based on the results from the reanalysis studies, it has become apparent that different
smoothing approaches can also affect PM risk estimates. For example, the models with natural
splines (parametric smoothing) appear, in general but not always, to result in smaller PM risk
estimates than GAM models with LOESS or smoothing splines. GAM models may possibly
suffer from biased standard error of risk estimates, but they also seem to fit the data better (i.e.,
based on AIC) than GLM models with natural splines. In any case, the choice of these
smoothers does not seem to affect PM risk estimates (-10 to 30%) as much as the range of
weather model specifications or the range of the degrees of freedom for temporal trends
adjustment do (as large as a factor of two).
A less explored issue is the effect of multiday effects of PM. The PM10 risk estimates
summarized above are either for a single-day lag (U.S. NMMAPS 90 cities study, Canadian 8
cities study, and APHEA1), or an average of two days (U.S. 10 cities study and APHEA2).
However, the reanalysis of U.S. 10 cities study data suggests that the multiday PM effect,
accounting for 0 through 5 day lag, could be twice as large as the effect sizes estimated from
single or two-day average models and even bigger (~3 to 4 fold) when more specific cause of
death categories are examined. This issue warrants further investigation.
8-49
-------
In summary, considering the wide variability in possible reasonable model specification
choices that can affect the PM risk estimates, the reported combined PM10 total nonaccidental
mortality risk estimates from multicity studies are in reasonably good agreement. That is,
they fall mainly in the range of-1.0 to 3.5% per 50 |ig/m3 increase in single or two-day
average PM10. Combinations of choices in model specifications (the number of weather terms
and degrees of freedom for smoothing of mortality temporal trends) alone may explain the extent
of the difference in PM10 risk estimates across studies. The range for these newly available
combined estimates from multicities studies overlap with the range of PM10 estimates (2.5 to 5%,
obtained from single cities studies) previously reported in the 1996 PM AQCD, but extends to
somewhat lower values.
8.2.2.4 U.S. Single-City Studies
In addition to the new multicity studies assessed above, many studies newly available since
the 1996 PM AQCD evaluated relationships between mortality and short-term exposure to PM
using data from individual cities. The results of such studies are summarized in tabular form
in Appendix 8A-1. The ensuing discussion focuses on the results of recent U.S. single-city
studies, especially those including PM10, PM2 5 and PM10_2 5 data. Results of analyses using PM2 5
and PM10_2 5 measurements are also discussed further in Section 8.2.2.5.
Lippmann et al. (2000; reanalyzed Ito, 2003) used aerometric data from Detroit which
included measurements of PM10, PM25, PM10_25, sulfate, H+, O3, SO2, NO2, and CO for a
1992-1994 study period. Associations with total (nonaccidental), cardiovascular, respiratory,
and other deaths were analyzed using GAM Poisson models, adjusting for season, temperature,
and relative humidity. Analyses were also done for an earlier 1985-1990 study period that
included measurements of PM10 and TSP along with the gaseous co-pollutants. Reanalyses were
done using stringent convergence criteria as well as natural splines, as well as additional
sensitivity analyses to examine the influence of alternative weather models and selection of
degrees of freedom on model results. In the reanalyses, 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 different across PM
8-50
-------
components; the original conclusion regarding the relative importance of PM components
remained the same. PM10, PM2 5, and PM10_2 5 were more significantly associated with mortality
outcomes than sulfate or H+. PM coefficients were generally not sensitive to inclusion of
gaseous pollutants. PM10, PM2 5, and PM10_2 5 effect size estimates were comparable in terms of
the 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- SO42 , were significantly associated with respiratory
mortality for the 1985-1990 period. The simultaneous inclusions of gaseous pollutants
with PM10 or TSP reduced the PM effect size by 0 to 34%. Effect size estimates for total,
circulatory, and "other" categories were smaller than for respiratory mortality.
Chock et al. (2000) evaluated associations between daily mortality in two age groups
(< 75 years, > 75 years) and several air pollution variables (PM10, PM2 5, PM10_2 5, CO, O3, NO2,
SO2) in Pittsburgh, PA, during 1989 to 1991 (data on PM25 and PM10_25 were only available for
half of the 3-year study period). Poisson GLM regression was used, including filtering of data
based on cubic B-spline functions to adjust for seasonal trends; models included indicators for
day of week, and temperature was modeled as a V-shape function. Single- and multipollutant
models were run for 0, 1,2, and 3 day lags. Single- and multipollutant nonseasonal models
showed significant positive associations between PM10 and daily mortality, but seasonal models
showed much multicolinearity, masking association of any pollutant with mortality. PM25
and PM10_2 5 were both positively associated with mortality, but the coefficients were unstable in
this data set when stratified by age group and season and no conclusions were drawn on the
relative roles of PM25 and PM10_25. In their conclusions, the authors emphasized issues of
seasonal dependence of correlation among pollutants, multicolinearity among pollutants, and
instability of coefficients for PM25 and PM10_25.
Lipfert et al. (2000a), using data for Philadelphia and the seven-county Philadelphia
metropolitan area from 1992 to 1995, regressed twelve mortality variables (as categorized by
area, age, and cause) on 29 pollution variables (PM components, O3, SO2, NO2, CO, and by
subareas), yielding 348 regression results. Both dependent and explanatory variables were
prefiltered 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
8-51
-------
hot and cold days and day-of-week using stepwise procedure, and the average of current and
previous days' pollution levels were used. Significant associations were reported for a wide
variety of particulate and gaseous pollutants, especially for peak O3. No systematic differences
were seen according to particle size or chemistry. Mortality for one part of the metropolitan area
could be associated with air quality from another, not necessarily neighboring part.
Clyde (1999) employed Bayesian model averaging (BMA) techniques, more often used in
other fields but relatively newly applied in some air pollution epidemiology studies, to
evaluate PM10 effects on total (nonaccidental) mortality among the elderly (> 65 years old) in
the city of Chicago and Cook County, Illinois. The BMA approach (discussed later in Section
8.4.1) provides an approach for taking into account model uncertainty, uses a set of models
rather than one, and each model contributes to the overall BMA outcome in proportion to the
support received from observed data. The Clyde (1999) analyses included 24-h PM10 values
derived from one daily monitoring site and daily averages of every-six-day data from a subset of
six of 20 other monitoring stations during 1985-1990 and a number of other meteorological
variables. Clyde (1999) reported a posterior probability close to 1.0, indicating a very high
probability of a parti culate matter effect, and went on to estimate a 5 to 16% increase in
mortality compared to the average level of PM10. She further indicated that, based on the results
of the model with the linear term for PM10, 95% posterior probability intervals for the expected
decrease in mortality per 10 |ig/m3 PM10 decrement would be 0.25 to 0.82 deaths/day or roughly
91 to 300 deaths per year in the > 65 year old population in Cook County. Clyde (1999) noted,
however, that these were preliminary results that are subject to a number of caveats (e.g., if
the PM10 measurements were not representative of the outdoor exposure of the population, then
the effect may have been over- or under-estimated).
Moolgavkar (2000a) evaluated (using GAM with default convergence criteria) associations
between short-term measures of major air pollutants and daily deaths in three large U.S. urban
areas (Cook Co., IL, encompassing Chicago; Los Angeles Co., CA; and Maricopa Co., AZ,
encompassing Phoenix) during a 9-year period (1987 to 1995). Moolgavkar (2003) reanalyzed
the data for Cook Co. and Los Angeles Co. (but not Maricopa Co.), using GAM with stringent
convergence criteria as well as GLM with natural splines. Ozone was analyzed in the original
8-52
-------
analysis but not in the reanalyses (it was only positive and significant in Cook county in the
original analysis). This section describes the results from the reanalyses. Total nonaccidental
deaths, deaths from cardiovascular disease (CVD) and chronic obstructive lung disease (COPD)
were analyzed in relation to 24-h readings for PM, CO, NO2, and SO2 averaged over all monitors
in a given county. Cerebrovascular mortality was analyzed in the original analysis but not in the
reanalyses (its association with air pollution was weak in the original analysis). The results of
cause-specific mortality analyses are described in a later section. Daily readings were available
for each of the gaseous pollutants in both Cook Co. and Los Angeles Co., as were PM10 values
for Cook Co. However, PM10 and PM2 5 values were only available every sixth day in Los
Angeles Co. PM values were highest in summer in Cook Co. and in the winter and fall in Los
Angeles Co.; whereas the gases (except for O3) were highest in winter in both counties. The PM
indices were moderately correlated (r = 0.30 to 0.73) with CO, NO2, and SO2 in Cook Co. and
Los Angeles Co. Total nonaccidental, CVD, and COPD deaths were all highest during winter in
both counties. Adjusting for temperature and relative humidity effects in separate analyses for
each mortality endpoint for these two counties, varying patterns of results were found (as noted
in Appendix A, Table 8A-1). Moolgavkar (2003) also reported sensitivity of results to different
degrees of freedom (df) for smoothing of temporal trends (30 df and 100 df).
As for Cook Co., PM10 was significantly associated with total nonaccidental mortality at
lag 0 (most significant) and 1 day in GAM models with both 30 df andlOO df for smoothing of
temporal trends, as well as in a GLM model with 100 df for smoothing of temporal trends. The
gaseous pollutants were also significantly associated with total nonaccidental mortality at
various lags (wider lags than PM10), but were most significant at lag 1 day. These associations
did not appear to be sensitive to the extent of smoothing for temporal trends, at least at their most
significant lags. In two pollutant models (results were not shown in tables but described in text),
the PM10 association remained "robust and statistically significant" at lag 0 day; whereas the
coefficients for the gases became nonsignificant. However, at lag 1 day, the PM10 association
became nonsignificant and the gases remained significant. Thus, some extent of "sharing" of the
association is apparent, and whichever pollutant is more strongly associated than the other at that
lag tended to prevail in the two pollutant models in this data set.
8-53
-------
For Los Angeles Co., CO was more significantly associated (positive and significant at lag
0 through 3 days) with mortality than PM10 (positive and significant at lag 2) or PM2 5 (positive
and significant at lag 1). In two pollutant models in which CO and PM indices were included
simultaneously at PM indices = "best" lags, CO remained significant, whereas PM coefficients
became nonsignificant (and negative for cases with 30 df for temporal smoothing). For Los
Angeles data, the PM coefficients appeared to be more sensitive to the choice of the degrees of
freedom than to the default versus stringent convergence criteria. GLM models tended to
produce smaller risk estimates than GAM models. Moolgavkar also reported that these
associations were robust to varying the extent of smoothing for weather covariates.
The results for these two cities do not reflect a common pattern. In Cook Co., all the
pollutants were associated with mortality, and their relative importance varied depending on the
lag day, whereas CO appeared to show the strongest mortality associations in Los Angeles.
Moolgavkar concluded that, considering the substantial differences that can result from different
analytic strategies, no particular numeric estimates were too meaningful, although the patterns of
associations appeared to be robust.
Ostro et al. (2000; reanalyzed Ostro et al., 2003) conducted a study in Coachella Valley,
CA, using (a) PM10 data collected from 1989-1998 and (b) PM25 and PM10_25 data collected
during the last 2.5 years of the study period. Both PM25 and PM10_25 were also estimated for the
earlier remaining years to increase the power of the analyses, but only PM10_25 could be reliably
estimated; so, predicted PM2 5 data were not used. Original analyses used GAMs, with
smoothing functions for time and indicators for day of week. Different lags for temperature,
humidity and dewpoint were tested for use in the models; and then pollutants were added
individually and next in combination. In the reanalyses, more stringent convergence criteria and
natural splines were used, but the reanalyses were only done for cardiovascular mortality.
For such cause-specific mortality, significant associations were found for PM10_2 5 and PM10, but
not for PM25 (possibly due to the low range of PM2 5 concentrations and reduced sample size
for PM2 5 data) and PM risk estimates were higher for multi-day averages. The PM risk
estimates were slightly reduced in the reanalyses using GAM with stringent convergence criteria
8-54
-------
or using GLM; but sensitivity analyses showed that results were not sensitive to alternative
degrees of freedom for temporal trends and temperature.
In another study, total, cardiovascular, and respiratory deaths in Santa Clara Co., CA were
regressed on PM10, PM2 5, PM10_2 5, CoH, nitrate, sulfate, O3, CO, NO2, adjusting for time trend,
season, and minimum and maximum temperature, using a Poisson GAM model (Fairley, 1999;
reanalyzed Fairley, 2003). Reanalyses included use of GAM with stringent convergence criteria,
as well as natural splines and an additional indicator for O3 (daily number of hours exceeding
60 ppb). In the reanalyses, the PM coefficients were either unchanged or only slightly decreased
or increased; and the original findings, including the pattern in two-pollutant models, were
unchanged. PM2 5 and nitrate were most significantly associated with mortality, but significant
associations were reported for all pollutants except PM10_2 5 in single-pollutant models. In two-
and four-pollutant models, PM2 5 or nitrate remained significant for total mortality, but the other
pollutants did not. The PM2 5 risk estimates for respiratory deaths were larger than those for total
or cardiovascular deaths but the associations were only significant for total mortality.
Mar et al. (2000; reanalyzed Mar et al., 2003) evaluated associations between air pollutants
and total (nonaccidental) and cardiovascular deaths in Phoenix for only those who resided in the
zip codes located near the air pollution monitor. GAM Poisson models were used, adjusting for
season, temperature, and relative humidity, and a variety of air pollution variables were used,
including O3, SO2, NO2, CO, TEOM PM10, TEOM PM2 5, TEOM PM10.2 5, DFPSS PM2 5, S, Zn,
Pb, soil, soil-corrected K (KS), nonsoil PM, OC, EC, and TC. Lags 0 to 4 days were evaluated.
Factor analysis was also conducted on chemical components of DFPSS PM25 (Al, Si, S, Ca, Fe,
Zn, Mn, Pb, Br, KS, OC, and EC); and factor scores were included in the mortality analyses.
Reanalyses were done using stringent convergence criteria as well as natural splines only for
cardiovascular mortality. In the reanalyses, small reductions were seen in risk estimates for PM
mass concentration indices using GAM/stringent convergence criteria or GLM/natural splines.
For source factors, there were moderate reductions in risk estimates for the motor vehicle factor
and slight increases for the regional sulfate factor and slight reductions in the coefficients for EC
and OC; but the estimates remained unchanged for the vegetative burning factor. Cardiovascular
mortality was significantly associated with CO, NO2, SO2, PM2 5, PM10, PM10_2 5, OC and EC.
8-55
-------
Vehicular traffic factors and regional sulfate 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. However, soil in PM2 5 was positively and
significantly associated with total mortality during the third year of the study when a WINS
impactor (with sharper cut) was used instead of a cyclone sampler.
In all of the studies discussed above, some statistically significant associations between
mortality and PM indicators, especially PM10 and PM2 5 were found. In multipollutant models,
PM coefficients were often robust to inclusion of gaseous pollutants, but sometimes reduced for
specific co-pollutants (see also the co-pollutant model discussion in Section 8.4).
8.2.2.5 The Role of Particulate Matter Components
Delineation of the roles of specific ambient PM components in contributing to associations
between short-term PM exposures and mortality requires evaluation of several factors, e.g., size,
chemical composition, surface characteristics, and the presence of gaseous co-pollutants. While
possible combinations of these factors can in theory be limitless, the actual data tend to cover
definable ranges of aerosol characteristics and co-pollutant environments due to typical source
characteristics (e.g., fine particles tend to be combustion products in most cities). Newly
available studies conducted in the last few years have begun to provide more extensive
information on the roles of PM components; and their results are discussed below in relation to
three topics: (1) PM particle size (e.g., PM25 versus PM10_25); (2) chemical components; and
(3) source oriented evaluations.
The ability to compare the relative roles of different PM size fractions and various PM
constituents is restricted by the limitations of the available studies. Comparisons nevertheless
can be attempted, using such information as the relative level of significance and/or the strength
of correlation between component estimate and health outcome. The relative significance across
cities/studies is influenced by the sample size and the level of the pollutants. The width of the
confidence band also needs to be taken into account, according more weight for studies with
narrower confidence bands. Caution in interpretation of such information, however, is warranted
8-56
-------
because of potential measurement error and possible high correlations between indices being
compared. Additionally, limitations of single-city studies must be recognized.
8.2.2.5.1 Particulate Matter Particle Size Evaluations
With regard to the relative importance of the fine and coarse fractions of inhalable PM10
particles capable of reaching thoracic regions of the respiratory tract, at the time of the 1996 PM
AQCD only one acute mortality study (Schwartz et al., 1996a) had examined this issue. That
study (which used GAM with default convergence criteria in analyzing Harvard Six-City
Study data) suggested that fine particles (PM2 5), distinctly more so than thoracic coarse-
fraction (PM10_2.5) particles, were associated with daily mortality. Recent reanalyses using GAM
with more stringent convergence criteria have yielded only slightly smaller PM2 5 effect-size
estimates (Schwartz, 2003a). It should also be noted that (a) the Klemm et al. (2000) reanalysis
reconstructed the data and replicated the original analyses (using GAM with default convergence
criteria) and (b) the Klemm and Mason (2003) reanalysis, using GAM with stringent
convergence criteria and GLM with parametric smoothers, also essentially reproduced the
original investigators' results.
Since the 1996 PM AQCD, several new studies have used size-fractionated PM data to
investigate the relative importance of fine (PM25) versus coarse (PM10_25) fraction particles.
Table 8-2 provides synopses of those studies with regard to the relative importance of the two
size fractions, as well as some characteristics of the data. The average levels of PM25 ranged
from about 13 to 30 |ig/m3 in the U.S. cities, but much higher average levels were measured in
Santiago, Chile (64.0 |ig/m3 ). As can be seen in Table 8-2, in the northeastern U.S. cities
(Philadelphia, PA and Detroit, MI), there was more PM2 5 mass than PM10_2 5 mass on the
average; whereas in the western U.S. (Phoenix, AZ; Coachella Valley, CA; Santa Clara County,
CA) the average PM10_2 5 levels were higher than PM2 5 levels. It should be noted that the three
Phoenix studies in Table 8-2 used much the same data set; all used fine and coarse particle data
from EPA's 1995-1997 platform study. Seasonal differences in PM component levels should
also be noted. For example, in Santa Clara County and in Santiago, Chile, winter PM2 5 levels
averaged twice those during summer. The temporal correlation between PM2 5 and PM10_2 5
8-57
-------
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 PM2 s versus
PM,n, s and Comments
Fairley (1999,
2003)*
Santa Clara
County, CA
Ostro et al.
(2000, 2003)*
Coachella
Valley, CA
Mar et al.
(2000, 2003)*
Phoenix, AZ
1995 to 1997
Smith et al.
(2000)
Phoenix, AZ
Clyde et al.
(2000)
Phoenix, AZ
Lippmann
et al. (2000);
Ito, (2003)*
Detroit, MI
1992 to 1994
Lipfert et al.
(2000a)
Philadelphia, PA
1992 to 1995.
PM25 mean= 13;
PM25/PM10 = 0.38;
r=0.51.
PM25 (Palm Springs
and Indio, respectively)
mean= 12.7, 16.8;
PM25/PM10 = 0.43, 0.35;
r= 0.46, 0.28.
PM2 5 (TEOM)
mean= 13.0;
PM25/PM10 = 0.28;
r=0.42.
Not reported, but likely
same as Clyde's or
Mar's data from the
same location.
PM25mean= 13.8;
PM2 5/PM10 = 0.30;
r=0.65.
PM25mean= 18;
PM2 5/PM10 =0.58;
r=0.42.
PM25 mean= 17.3;
PM2 5/PM10 = 0.72.
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 associated (and not
significant) at the lags examined.
Cardiovascular mortality was significantly associated with both
PM2 5 (lags 1, 3, and 4) and PM10.2 5 (lag 0). 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, a statistically significant mortality
association with PM10_2 5 was found, but not with PM2 5.
In models allowing for a threshold, indications of a threshold for
PM2 5 (in the range of 20-25) were 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 the
contributions of Fe, Cu, Zn, and Pb to PM10_2 5 were lowest.
Using Bayesian Model Averaging that incorporates model
selection uncertainty with 29 covariates (lags 0- to 3 -day), the
effect of coarse particles (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.
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.
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 and precisely associated with
cardiovascular mortality than PM10_2 5.
8-58
-------
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
ofPM25toPM10;and
Correlation Between
PM2 5 and PM10.2 5
Results Regarding Relative Importance of PM2 s versus
PM,n, s and Comments
Klemm and
Mason (2000)
Atlanta, GA
PM25mean= 19.9;
PM2 5/PM10 = 0.65
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.
Schwartz
(2003a)
6 U.S. cities
Klemm et al.
(2000);
Klemm and
Mason (2003)*
6 U.S. cities
Chock et al.
(2000)
Pittsburgh, PA
Not specified in
Schwartz (2003a) paper;
but see values below for
same 6 cities.
Mean ranged from 11.3
to29.6;MeanPM10.25
ranged from 6.6 to 16.1;
MeanPM25/PM10 ranged
from 0.50 to 0.66 in the
six cities.
Data distribution not
reported.
PM2 5/PM10 = 0.67
Schwartz (2003a) reanalysis of Schwartz et al. (1996a) Harvard
Six City time-series analyses confirmed original study findings of
significant associations between total mortality and PM2 5 across
the six U.S. cities, but not with PM10_2 5 (except in one city,
Steubenville).
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,
multicolinearity 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 to 1996
PM2 5 mean
PM25/PM10
r=0.37.
PM2 5 mean
PM25/PM10
r=0.52.
= 13.3;
= 0.51;
= 64.0;
= 0.58;
Both PM2 5 and PM10_2 5 were significantly associated with total
nonaccidental 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
smoother and amount of smoothing.
In GLM results for the whole years, PM2 5 and NO2 were more
consistently significantly associated with total nonaccidental
mortality than PM10_2 5.
Note: * next to author name indicates that the study was originally analyzed using GAM models only with
default convergence criteria using at least two nonparametric smoothing terms.
ranged between 0.30 and 0.65. Such differences in ambient PM mix features from season to
season or from location to location complicates assessment of the relative importance of PM25
and PM,
l!0-2.5-
8-59
-------
To facilitate a quantitative overview of the effect size estimates and their corresponding
uncertainties from these studies, the percent excess risks are plotted in Figure 8-5. These
excluded the Smith et al. study (which did not present linear term RRs for PM2 5 and PM10_2 5)
and the Clyde et al. study (for which the model specification did not obtain RRs for PM2 5
and PM10_25 separately). Note that, in most of the original studies, the RRs were computed for
comparable distributional features (e.g., interquartile range, mean, 5th -to-95th percentile, etc.).
However, the increments derived and their absolute values varied across studies; therefore, the
RRs used in deriving the excess risk estimates delineated in Figure 8-5 were re-computed for
consistent increments of 25 |ig/m3 for both PM25 and PM10_25. Note also that re-computing the
RRs per 25 |ig/m3 in some cases changed the relative effect size between PM2 5 and PM10_2 5, but
it did not affect the relative significance.
All of the studies found positive associations between both the fine and coarse PM indices
and increased mortality risk, however, most of the studies did not have large enough sample
sizes to separate out what often appear to be relatively small differences in effect size estimates.
However, three studies do show distinctly larger mortality associations with PM2 5 than for
nonsignificant PM10_2 5 effects. For example, the Klemm et al. (2000) and Klemm and Mason's
(2003) recomputations of the Harvard Six Cities time-series study data reconfirmed the original
Schwartz et al. (1996a) finding that PM2 5 was significantly associated with excess total
mortality, but PM10_2 5 across all cities was not (although the Schwartz [2003a] reanalyses also
reconfirmed their original findings of a statistically significant PM10.25-mortality relationship in
one city, i.e., Steubenville, OH). Similar findings of PM25 being significantly associated with
total mortality were obtained in Santa Clara County (Fairley, 1999; Fairley 2003), and Mar et al.
(2000, 2003) reported much larger PM2 5 associations with cardiovascular mortality than
for PM10_2 5 (although both were statistically significant at p < 0.05). There were several other
studies in which the importance of PM25 and PM10_25 were considered to be similar or, at least,
not distinguishable: Philadelphia, PA (Lipfert et al., 2000a); Detroit, MI (Lippmann et al., 2000;
reanalysis by Ito 2003); Eight Canadian cities (Burnett at al., 2000; reanalysis by Burnett and
Goldberg, 2003); and Santiago, Chile (Cifuentes et al., 2000). Some other studies suggested
8-60
-------
Percent excess death (total unless otherwise noted) per
25 ug/m3 increase in PM2.5 (•) or PMio-2.5 (O)
oo
Klemm and Mason (2003)*
Harvard 6 Cities —
(recomputed)
Burnett and Goldberg (2003)*
o r* *j- r~-*- —
Pittsburgh PA~
Klemm and Mason (2000)
Lipfert et al. (2000a)
Ito (2003)*
Maretal. (2003)*
Fairley (2003)*
Cifuentes et al. (2000)
-2 0 2 4 6 8 10 12 14 16 18
i i i i i i i i i i
... A
Lag 1 day ^
A
Lag 0 day 9^ *^ > a9e less than 7S
Cj T SCIG 75 3nd over
Lag 0 day ^
n
..-.. _ A
Lag 3 day • • — —
Lag 1 day °
i -™ ^ j-,, _....... A
Lag 0 day U
Lag 1+2 day ^^
Figure 8-5. Percent excess risks estimated per 25 ug/m3 increase in PM2 5 or PM10_2 5 from new studies that evaluated
both PM2 5 and PM10_2 5, based on single pollutant (PM only) models. The asterisk next to reference indie
reanalysis of data using GLM with natural splines. Other studies used GLM or OLS.
-------
that PM10_2 5 was more important than PM2 5: Coachella Valley, CA (Ostro et al., 2000 & 2003)
and Phoenix, AZ (Smith et al., 2000, and Clyde et al., 2000).
In the reanalysis (Burnett and Goldberg, 2003) of the Canadian 8-city study (Burnett et al.,
2000), the relative importance of PM2 5 and PM10_2 5 was not clear, with both PM indices being
significant in single pollutant models. In GAM models (stringent convergence criteria) with
LOESS smoothers, PM25 was more significant and showed larger risk estimates than PM10_25.
However, in sensitivity analysis in which varying degrees of freedom for mortality temporal
trends were applied in GLM models, the effect size and significance for these PM indices were
often comparable. The authors commented that PM10_2 5 coefficient was more sensitive to the
extent of temporal smoothing than PM25.
The Lippmann et al. (2000) results and reanalyses (Ito, 2003) for Detroit are also
noteworthy in that additional PM indices were evaluated besides those depicted in Figure 8-5,
and the overall results obtained may be helpful in comparing fine- versus coarse-mode PM
effects. In analyses of 1985 to 1990 data, PM-mortality relative risks and their statistical
significance were generally in descending order: PM10, TSP-SO42 , and TSP-PM10. For the 1992
to 1994 period, relative risks for equivalent distributional increment (e.g., IQR) were comparable
among PM10, PM2 5, and PM10_2 5 for both mortality and hospital admissions categories; and SO42
was more strongly associated with most outcomes than FT. Consideration of the overall pattern
of results led the authors to state that the mass of the smaller size index could explain a
substantial portion of the variation in the larger size indices. In these data, on average, PM2 5
accounted for 60% of PM10 (up to 80% on some days) and PM10 for 66% of TSP mass. The
temporal correlation between TSP and PM2 5 was r = 0.63, and that for PM2 5 and PM10 was
r = 0.90, suggesting that much of the apparent larger particle effects may well be mainly driven
by temporally covarying smaller PM2 5 particles. The stronger associations for sulfates than FT,
suggestive of nonacid fine particle effects, must be caveated by noting the very low FT levels
present (often at or near the detection limit).
Three research groups, using different methods, have utilized the same U.S. EPA research
platform aerometric data to evaluate ambient PM-mortality associations in the Phoenix, AZ area.
Although these groups used somewhat different approaches, there is some consistency across
8-62
-------
their results in that PM10_25 emerged in all three as a likely important predictor of mortality. The
Mar et al. (2000, 2003) analyses evaluated total and cardiovascular mortality among people
residing in zip code areas proximal to the one containing the EPA monitoring platform
yielding PM10, PM2 5, PM10_2 5 and compositional data used in their analyses. In the Mar et al.
(2000 & 2003) analyses, PM10 was significantly associated with total mortality, whereas PM2 5
and PM10_2 5 were positively (but not quite significantly) associated. However, cardiovascular
mortality (CVM) was significantly associated with both PM2 5 and PM10_2 5, as well as being
significantly associated with several source categories (as shown by factor analyses discussed
later). The Smith et al. (2000) analyses related mortality in Phoenix to the EPA PM2 5 data, but
used mortality data from surrounding areas (Tempe, Scottsdale, etc.) within 50 miles of Phoenix
in analyses of PM10_2 5 effects. Based on a linear PM effect, Smith et al. found PM10_2 5 to be
significantly associated with total mortality, but not PM25. However, Smith et al.'s additional
finding that PM25 may have a threshold effect further complicates a simple comparison of the
two size-fractionated mass concentration indices. In the Clyde et al. (2000) analysis, PM-
mortality associations were found only for the geographic area where PM2 5 was considered
uniformly distributed, but the association was stronger for PM10_2 5 than for PM2 5. That is,
whereas the posterior probability for PM25 effect was -0.91, the highly ranked models (based on
the Bayes Information Criterion) consistently included 1-day lagged PM10_25. The PM2 5 in
Phoenix is mostly generated from motor vehicles, whereas PM10_2 5 consists mainly of two types
of particles: (a) crustal particles from natural (wind blown dust) and anthropogenic (construction
and road dust) processes, and (b) organic particles from natural biogenic processes (e.g., the soil-
dwelling Coccidioides immitis fungus in windblown dust, as discussed in Appendix 7B) and
anthropogenic (sewage aeration) processes. The crustal particles may also be contaminated with
metals secondarily deposited over many years as the result of emissions from smelters operating
until recently in the Phoenix area.
In summary, issues regarding the relative importance of PM2 5 and PM10_25 have not yet
been fully resolved. Caution in interpreting size-fraction PM studies is warranted due to
(a) problems with measurement and exposure error (likely higher for PM10_2 5) and (b) the
correlation between the two size fractions. Limitations of single-city studies have also been
8-63
-------
noted. While limited sample sizes typically prevented clear statistical distinction between
relative roles played by PM2 5 and PM1(K25, recent studies show mixed results, with some studies
suggesting coarse particle effects. The relative importance may also vary depending on the
chemical constituents in each size fraction, which may vary from city to city. Nevertheless, a
number of studies published since the 1996 PM AQCD do appear to substantiate associations
between PM25 and increased total and/or CVD mortality. Consistent with the 1996 PM AQCD
findings, effect-size estimates from the new studies generally fall within the range of-1.5 to
6.5% excess total mortality per 25 |ig/m3 PM25. The coarse particle (PM10_25) effect-size
estimates also tend to fall in about the same range, mainly from -0.5 to 6.0%.
Crustal Particle Effects
Since the 1996 PM AQCD, several studies have yielded interesting new information
concerning possible roles of crustal wind-blown particles or crustal particles within the fine
particle fraction (i.e., PM25) in contributing to observed PM-mortality effects.
Schwartz et al. (1999), for example, investigated the association of coarse particle
concentrations with nonaccidental deaths in Spokane, WA, where dust storms elevate coarse PM
concentrations. During the 1990 to 1997 period, 17 dust-storm days were identified. The PM10
levels during those storms averaged 263 |ig/m3, compared to 39 |ig/m3 for the entire period. The
coarse particle domination of PM10 data on those dust-storm days was confirmed by a separate
measurement of PM10 and PMX 0 during a dust storm in August, 1996: the PM10 level was
187 |ig/m3, while PM10 was only 9.5 |ig/m3. The deaths on the day of a dust storm were
contrasted with deaths on control days (n = 95 days in the main analysis and 171 days in the
sensitivity analysis), which are defined as the same day of the year in other years when dust
storms did not occur. The relative risk for dust-storm exposure was estimated using Poisson
regressions, adjusting for temperature, dewpoint, and day of the week. Various sensitivity
analyses considering different seasonal adjustment, year effects, and lags were conducted. The
expected relative risk for these storm days with an increment of 221 |ig/m3 would be about 1.04,
based on PM10 relative risk from past studies, but the estimated RR for high PM10 days was
found to be only 1.00 (CI: 0.95, 1.05) per 50 |ig/m3 PM10 change in this study. Schwartz et al.
8-64
-------
concluded that there was no evidence to suggest that coarse (presumably crustal) particles were
associated with daily mortality.
Ostro et al. (2000 & 2003) also analyzed Coachella Valley, CA data for 1989-1998.
This desert valley, where coarse particles of geologic origin comprise -50-60% of annual-
average PM10 (> 90% during wind episodes throughout the year), includes the cities of Palm
Springs and Indio, CA. Cardiovascular deaths were analyzed using GAM (with stringent
convergence criteria) and GLM Poisson models adjusting for temperature, humidity, day-of-
week, season, and time. Actual PM2 5 and PM10_2 5 data were only available for the last 2.5 years
of the 10-year study period. So, predictive models were developed for estimating PM25
and PM10_2 5 concentrations for earlier years, but the model for PM2 5 was not considered
successful and, therefore, was not used. Thus, a strict comparison of relative strength of risk
estimates for PM2 5 and PM10_2 5 in this data set is difficult. Cardiovascular mortality was reported
to be positively associated with both PM10 and PM10_2 5 at multiple lags between 0 and 2 day lags;
whereas the PM2 5 coefficient was positive only at lag 4 day, based on analyses involving far
fewer observations for PM2 5 (only over a 2 year period versus 10 years for PM10 and PM10_2 5).
These results hint at crustal particle effects possibly being important in this desert situation, but
use of estimated values for PM10_25 lessens the credibility of the reported PM10_25 findings. Also,
the ability to discern more clearly the role of fine particles would likely be improved by analyses
of more years of actual data for PM25.
In two other studies, Laden et al. (2000) and Schwartz (2003b) analyzed Harvard Six-
Cities Study data and Mar et al. (2000) analyzed the Phoenix data to investigate the influence on
daily mortality of crustal particles in PM25 samples. These studies are discussed in more detail
in Section 8.2.2.4.3 on the source-oriented evaluation of PM; and only the basic results regarding
crustal particles are mentioned here. The elemental abundance data, from X-ray fluorescence
(XRF) spectroscopy analysis of daily filters, were analyzed to estimate the concentration of
crustal particles in PM25 using factor analysis. Then the association of mortality with fine
crustal mass was estimated using Poisson regression (regressing mortality on factor scores for
"crustal factor"), adjusting for time trends and weather. No positive association was found
between the fine crustal mass factor and mortality. However, the soil component of PM25 was
8-65
-------
positively and significantly associated with total mortality when only the third year of data
(when a WINS impactor was used instead of a cyclone) was analyzed.
The above results, overall, mostly suggest that crustal particles (coarse or fine) per se are
not likely associated with daily mortality. However, as noted in the previous section, three
analyses of Phoenix, AZ data do suggest that PM10_25 was associated with mortality. The results
from one of the three studies (Smith et al., 2000) indicate that coarse particle-mortality
associations are stronger in spring and summer, when the anthropogenic metal (Fe, Cu, Zn, and
Pb) contribution to PM10_2 5 is lowest as determined by factor analysis. However, during spring
and summer, biogenic processes (e.g., wind-blown pollen fragments, fungal materials,
endotoxins, and glucans) may contribute more to the PM10_2 5 fraction in the Phoenix area,
clouding any attribution of observed PM10_25 effects there to crustal particles alone, per se.
(See the discussion of bioaerosols in Chapter 7 and, also in Section 8.4.3 of this chapter).
Ultraftne Particle Effects
Wichmann et al. (2000) evaluated the attribution of PM effects to specific size fractions,
including both the number concentration (NC) and mass concentration (MC) of particles in a
given size range. To respond to the GAM convergence issues, Stolzel et al. (2003) reanalyzed
the data, using GAM with stringent convergence criteria and GLM with natural splines. The
study was carried out in the former German Democratic Republic city of Erfurt (pop. 200,000)
German. Erfurt was heavily polluted by particles and SO2 in the 1980s, and excess mortality
was attributed to high levels of TSP by Spix et al. (1993). Ambient PM and SO2 concentrations
have markedly dropped since then. The present study provides a more detailed look at potential
health effect associations with ultrafme particles (diameter < 0.1 jim) than earlier studies,
including examination of effects in relation to number counts for fine and ultrafme particles as
well as for their mass. This was made possible by use of the Mobile Aerosol Spectrometer
(MAS), developed by Gessellschaft fur Strahlenforschung (GSF), which measures number and
mass concentrations in three ultrafme size classes (0.01 to 0.1 jim) and three size classes of
larger fine particles (0.1 jim to 2.5 jim). The mass concentration MC001_25 is well correlated with
gravimetric PM2 5, and the number concentration NC0 01_2 5 is well correlated with total particle
8-66
-------
counts from a condensation particle counter (CPC). Mortality data were coded by cause of
death, with some discrimination between underlying causes and prevalent conditions of the
deceased.
In the reanalysis by Stolzel et al. (2003), daily mortality data were fitted using a Poisson
GAM (with stringent convergence criteria) and GLM, with adjustments for weather variables,
time trends, day of week, and particle indices. Weekly data for all of Germany on influenza
and similar diseases were also included in the model. In the original study, two types of models
were fitted; one used the best single-day lag for air pollution and a second the best polynomial
distributed lag (PDL) model for air pollution. Both linear (i.e., raw) and log-transformed
pollution indices were examined. PDL models in the original analysis generally had larger and
more significant PM effects than single-day lag models, but the reanalysis by Stolzel et al.
(2003) focused on single-day lag results only. Therefore, the numerical results in the following
discussion only include the single day lag results from the reanalysis. It should be noted that,
unlike most of the recent reanalyses that have been conducted to address the GAM conversion
issue, the reanalysis results from this study were virtually unchanged from the original results.
Both mass and number concentrations at the size ranges examined were mostly positively
associated with total nonaccidental mortality. The best single-day lags reported were mostly 0 or
1 day lag for mass concentrations and the 4 day lag for number concentrations. For example, the
estimated excess risk for MC001_2 5 at lag 1 day was about 3.9% (CI: 0, 7.7) per 25 |ig/m3. The
corresponding number for smaller fine particles, MC001.10, was 3.5% (CI: -0.4, 7.7). For
number concentration, the estimated excess risk for NC001.2.5 at lag 4 day was about 4.1%
(CI: -0.9, 9.3) per IQR (13,269 particles/cm3). The corresponding number for smaller fine
particles, NCaoi_LO, was 4.6% (CI: -0.3, 9.7) per IQR (12,690 particles/cm3). An examination of
all the results for MC001_2 5 and NC001.0A shown for lags 0 through 5 days indicates that the
associations were mostly positive for these mass and number concentrations, except for the "dip"
around 2 or 3 day lags.
The estimated excess risks are reduced, sometimes drastically, when co-pollutants
(especially SO2 and NO2) are included in a two-pollutant model. This is not surprising, as the
number and mass concentrations of various ultrafine and fine particles in all size ranges are
8-67
-------
rather well correlated with gaseous co-pollutants, except for the intermodal size range MCL0.2.5.
The number correlations range from 0.44 to 0.62 with SO2, from 0.58 to 0.66 with NO2, and
from 0.53 to 0.70 with CO. The mass correlations range from 0.53 to 0.62 with SO2, from
0.48 to 0.60 with NO2, and from 0.56 to 0.62 with CO. The authors found that ultrafme
particles, CO and NO2 form a group of pollutants strongly identified with motor vehicle traffic.
Immediate and delayed effects seemed to be independent in two-pollutant models, with single-
day lags of 0 to 1 days and 4 to 5 days giving 'best fits' to data. The delayed effect of ultrafme
particles was stronger than that for NO2 or CO. The large decreases in excess risk for number
concentration, particularly when NO2 is a co-pollutant with NC0 01.0 b clearly involves a more
complex structure than simple correlation. The large decrease in excess risk when SO2 is a
co-pollutant with MC0 01_2 5 is not readily explained and is discussed in some detail in Wichmann
et al. (2000).
Sulfur dioxide is a strong predictor of excess mortality in this study; and its estimated
effect is little changed when different particle indicators are included in a two-pollutant model.
The authors noted ". . .the [LOESS] smoothed dose response curve showed most of the
association at the left end, below 15 |ig/m3, a level at which effects were considered biologically
implausible. . ." Replacement of sulfur-rich surface coal has reduced mean SO2 levels in Erfurt
from 456 |ig/m3 in 1988 to 16.8 |ig/m3 during 1995 to 1998 and to 6 |ig/m3 in 1998. The
estimated concentration-response functions for SO2 are very different for these time periods,
comparing Spix et al. (1993) versus Wichmann et al. (2000) results. Wichmann et al. concluded
"These inconsistent results for SO2 strongly suggested that SO2 was not the causal agent but an
indicator for something else." The authors offered no specific suggestions as to what the
"something else" might be, but they did finally conclude that their studies from Germany
strongly supported PM air pollution as being more relevant than SO2 to observed mortality
outcomes. However, the HEI committee also did not agree with the investigators' interpretation
that the association of SO2 with mortality was an artifact, given the similar magnitude of effect
sizes between PM and SO2 and the persistence of an SO2 effect in the two-pollutant analyses.
8-68
-------
8.2.2.5.2 Chemical Components
Several new studies from the United States, Canada, and The Netherlands examined
mortality associations with specific chemical components of ambient PM. Table 8-3 shows the
chemical components examined in these studies; the mean concentrations for Coefficient of
Haze (CoH), sulfate, and H+; and indications of those components found to be associated with
increased mortality.
Coefficient of Haze, Elemental Carbon, and Organic Carbon
CoH is highly correlated with elemental carbon (EC) and is often considered as a good PM
index for motor vehicle sources, although other combustion processes such as space heating
likely also contribute to CoH levels. Several studies (Table 8-3) examined CoH; and, in most
cases, positive and significant associations with mortality outcomes were reported. In terms of
relative significance of CoH in comparison to other PM components, CoH was not the clearly
most significant PM component in most of these studies. The average level of CoH in these
studies ranged from 0.24 (Montreal, Quebec) to 0.5 (Santa Clara County, CA) 1000 linear feet.
The correlations between CoH and NO2 or CO in these studies (8 largest Canadian cities; Santa
Clara County, CA) were moderately high (r .0.7 to 0.8) and suggested a likely motor vehicle
contribution. Both EC and OC were significant predictors of cardiovascular mortality in the
Phoenix study; their effect sizes per IQR were comparable to those for PM10, PM2 5, and PM10_2 5.
Also, both EC and OC represented major mass fractions of PM25 (11% and 38%, respectively)
and were correlated highly with PM25 (r = 0.84 and 0.89, respectively). They were also highly
correlated with CO and NO2 (r = 0.8 to 0.9), indicating their associations with an "automobile"
factor. Thus, the CoH and EC/OC results from the Mar et al. (2000 and 2003) study suggest that
PM components from motor vehicle sources are likely associated with mortality. In a recent
study in Montreal, Quebec, by Goldberg et al. (2000 and 2003), CoH appeared to be correlated
with the congestive heart failure mortality (as classified based on medical records) more strongly
than other PM indices such as the visual-range derived extinction coefficient (considered to be
a good indicator of sulfate). However, the main focus of the study was on the role of
cardiorespiratory risk factors for air pollution, and the investigators warned against comparing
8-69
-------
TABLE 8-3. NEWLY AVAILABLE STUDIES OF MORTALITY
RELATIONSHIPS TO PM CHEMICAL COMPONENTS
Author, City
Mean
CoH
(1000ft)
Mean
S04
(ug/m3)
Mean H+
(nmol/m3)
Other PM
Components
Analyzed
Specific PM Components Found
to be Associated with Mortality
(Comments)
Burnett et al. (2000);
Burnett and
Goldberg (2003)*
8 largest Canadian
cities, 1986-1996.
Fairley(1999&
2003)*; Santa Clara
County, CA.
Goldberg et al.
(2000); Goldberg
and Burnett (2003);
Goldberg et al.
(2003)* Montreal,
Quebec, Canada.
1984-1993.
Lipfert et al., (2000a)
Philadelphia, PA.
1992-1995.
Lippmann et al.
(2000); Ito (2003)*
Detroit, MI.
1992-1994.
Klemm and Mason
(2000)
Atlanta, GA
1998-1999
Mar et al. (2000 &
2003)* Phoenix, AZ.
1995-1997.
Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ.
1981-1983.
Hoek (2003)*
the Netherlands.
1986-1994.
0.26
2.6
0.5
0.24
3.3
0.28
5.1
5.2
5.2
12.7
PM10,PM2,,PM10.,,
and 47 trace elements
3.8
(median)
PM10, PM25, PM10.25,
and nitrate
Predicted PM2 5, and
extinction coefficient
(visual- range
derived).
Nephelometry, NH4+,
TSP, PM10, PM25,
and PM10_2 5
PM10,PM2,,
andPM,n.,,
Nitrate, EC, OC,
oxygenated HC, PM10,
PM25,andPM10.25
EC, OC, TC, PM10,
PM25,andPM10.25
PM15, PM25,
cyclohexane-solubles
(CX),
dichloromethane-
solubles (DCM), and
acetone-solubles
(ACE).
PM10,BS, and nitrate
PM10, PM2 5, 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. PM2 5 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, PM2 5, and PM10.2 5 were
more significantly associated with
mortality outcomes than sulfate
orH+.
"Interim" results based on one year
of data. No statistically significant
associations for any pollutants.
Those with t-ratio of at least 1.0
wereH+,PM10, andPM25.
EC, OC, TC, PM10, PM2 5, and
PM10_2 5 were associated with
cardiovascular mortality.
PM15, PM2.5, sulfate, CX, and ACE
were significantly associated with
total and/or cardiovascular mortality
in Newark and/or Camden.
Sulfate, nitrate, and BS were more
consistently associated with total
mortality than was PM10.
*Note: The study was originally analyzed by GAM models only using default convergence criteria and at least two
nonparametric smoothing terms and was recently reanalyzed by GAM using stringent convergence criteria and/or
other non-GAM analyses.
8-70
-------
the relative strength of associations among PM indices, pointing out complications such as likely
error involved in the visual range measurements. Additionally, the estimated PM2 5 values were
predicted from other PM indices, including CoH and extinction coefficient, making it difficult to
compare straightforwardly the relative importance of PM indices.
Sulfate and Hydrogen Ion
Sulfate and H+, markers of acidic components of PM, have been hypothesized to be
especially harmful components of PM (Lippmann et al., 1983; Lippmann and Thurston, 1996).
The newly available studies that examined sulfate are shown in Table 8-3; two of them also
analyzed H+ data. The sulfate concentrations ranged from 1.8 |ig/m3 (Santa Clara County, CA)
to 12.7 |ig/m3 (three NJ cities). Aside from the west versus east coast contrast, the higher levels
observed in the three NJ cities are likely due to their study period coverage of the early 1980s,
when sulfate levels were higher. Sulfate explained 25 to 30% of PM25 mass in eastern U.S. and
Canadian cities, but it was only 14% of PM25 mass in Santa Clara County, CA. The H+ levels
measured in Detroit and Philadelphia were low. The mean H+ concentration for Detroit, MI
(the H+ was actually measured in Windsor, a Canadian city a few miles from downtown Detroit),
8.8 nmol/m3, was low as compared to the reported detection limit of 15.1 nmol/m3 (Brook et al.,
1997) for the measurement system used in the study. Note that the corresponding detection limit
for sulfate was 3.6 nmol/m3 (or 0.34 |ig/m3); and the mean sulfate level for Detroit was
54 nmol/m3 (or 5.2 |ig/m3), so that the signal-to-noise ratio is expected to be higher for sulfate
than for H+. Thus, the ambient levels and possible relative measurement errors for these data
should be considered in interpreting the relative strength of mortality associations in these data.
Sulfate was a statistically significant predictor of mortality, at least in single pollutant
models, in: Santa Clara County, CA; Philadelphia, PA; Newark, NJ; and Camden, NJ, but not in
Elizabeth, NJ; Detroit, MI; or Montreal, CN. However, it should be noted that the relative
significance across the cities is influenced by the sample size (both the daily mean death counts
and number of days available), as well as the range of sulfate levels and should be interpreted
with caution. Figure 8-6 shows the excess risks (± 95% CI) estimated per 5 |ig/m3 increase in
24-h sulfate reported in these studies compared to the reanalysis results of the earlier Six Cities
Study by Klemm and Mason (2003). The largest estimate was seen for Santa Clara County, CA;
8-71
-------
Percent excess death (total non-accidental mortality)
per 5 jjg/m3 increase in sulfate
Klemm and Mason (2003)*
Harvard 6 cities (recomputed)
Fairly (2003)*
Santa Clara Co.
Klemm et al. (2000)
Atlanta, GA
Lipfert et al. (2000a)
Philadelphia, PA
Ito (2003)*
Detroit, Ml
Tsai et al. (2000)
3 NJ cities '
-2
i
0
-4-
2
i
4
i
6
i
8
i
Newark
Camden
Elizabeth
10
Figure 8-6. Excess risks estimated per 5 ug/m3 increase in sulfate, based on U.S. studies
for which both PM2 5 and PM10_2 5 data were available.
but the wide confidence band (possibly due to the small variance of the sulfate, because its levels
were low) should be taken into account. In addition, the sulfate effect in the Santa Clara County
analysis was eliminated once PM2 5 was included in the model, perhaps being indicative of
sulfate mainly serving as a surrogate for fine particles in general there. In any case, more weight
should be accorded to estimates from other studies with narrower confidence bands. In the other
studies, effect size estimates mostly ranged from ~1 to 4% per 5 |ig/m3 increase in 24-h sulfate.
The relative significance of sulfate and H+ compared to other PM components is not
clear in the existing small number of publications. Because each study included different
combinations of co-pollutants that had different extents of correlation with sulfate and because
8-72
-------
multiple mortality outcomes were analyzed, it is difficult to assess the overall importance of
sulfate across the available studies. The fact that the Lippmann et al. (2000) study and the
reanalysis by Ito (2003) found that Detroit, MI data on H+ and sulfate were less significantly
associated with mortality than the size-fractionated PM mass indices may be due to acidic
aerosols levels being mostly below the detection limit in that data. In this case, it appears that
the Detroit PM components show mortality effects even without much acidic input.
In summary, assessment of new study results for individual chemical components of PM
suggest that an array of PM components (mainly fine particle constituents) are associated with
mortality outcomes, including CoH, EC, OC, sulfate, and nitrate. The variations seen with
regard to the relative significance of these PM components across studies may be in part due to
differences in their concentrations from locale to locale. This issue is further discussed below as
part of the assessment of new studies involving source-oriented evaluation of PM components.
8.2.2.5.3 Source-Oriented Evaluations
Several new studies have conducted source-oriented evaluation of PM components.
In these studies, daily concentrations of PM components (i.e., trace elements) and gaseous
co-pollutants were analyzed using factor analysis to estimate daily concentrations due to
underlying source types (e.g., motor vehicle emissions, soil, etc.), which are weighted linear
combinations of associated individual variables. The mortality outcomes were then regressed on
those factors (factor scores) to estimate the effect of source types rather than just individual
variables. These studies differ in terms of specific objectives/focus, the size fractions from
which trace elements were extracted, and the way factor analysis was used (e.g., rotation). The
main findings from these studies regarding the source-types identified (or suggested) and their
associations with mortality outcomes are summarized in Table 8-4.
The Laden et al. (2000) analysis of Harvard Six Cities data for 1979-1988 (reanalyzed by
Schwartz, 2003a) aimed to identify distinct source-related fractions of PM25 and to examine each
fraction's association with mortality. Fifteen elements in the fine fraction samples routinely
found above their detection limits were included in the data analysis. For each of the six cities,
up to 5 common factors were identified from among the 15 elements, using specific rotation
8-73
-------
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 (2003a)*
Harvard Six Cities.
1979-1988.
Mar et al.
(2000 & 2003)*
Phoenix, AZ.
1995-1997.
Tsai et al. (2000).
Newark, Elizabeth,
and Camden, NJ.
1981-1983.
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
PM25 (fromDFPSS) trace elements:
Motor vehicle emissions and re-suspended
road dust: Mn, Fe, Zn, Pb, OC, EC, CO,
andNO2
Soil: Al, Si, andFe
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.
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
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.
*Note: The study was originally analyzed using GAM models only with default convergence criteria using at
least two nonparametric smoothing terms, but was later reanalyzed using more stringent convergence criteria
and/or other approaches.
factor analysis. Using the Procrustes rotation (a type of oblique rotation), the projection of the
single tracer for each factor was maximized. This specification of the tracer element was based
on (a) knowledge from previous source apportionment research; (b) the condition that the
regression of total fine mass on that element must result in a positive coefficient; and (c) the
identifications of additional local source factors that positively contributed to total fine mass
8-74
-------
regression. Three source factors were identified in all six cities: (1) a soil and crustal material
factor with Si as a tracer; (2) a motor vehicle exhaust factor with Pb as a tracer; and (3) a coal
combustion factor with Se as a tracer. City-specific analyses also identified a fuel combustion
factor (V), a salt factor (Cl), and selected metal factors (Ni, Zn, or Mn). In the original analysis
by Laden et al., a GAM Poisson regression model (with default convergence criteria), adjusting
for trend/season, day-of-week, and smooth function of temperature/dewpoint, was used to
estimate impacts of each source type (using absolute factor scores) simultaneously for each city.
In the reanalysis reported by Schwartz (2003a), GAM models with LOESS smoothers were
replaced with penalized splines. Summary estimates across cities were obtained by combining
the city-specific estimates, using inverse-variance weights. The identified factors and their
tracers are listed in Table 8-4. The reanalysis using penalized splines changed somewhat the risk
estimates for source-apportioned mass concentrations in each city compared to those in the
original GAM results (increasing estimates in some cities and reducing them in others), but the
combined estimates across the six cities did not change substantially. The combined estimates
indicated that the largest increase in daily mortality was associated with the mobile source
associated fine mass concentrations, with an excess risk increase of 9.3% (CI: 4.0, 14.9) per
25 |ig/m3 source-apportioned PM25 (average of 0 and 1 day lags). The corresponding value for
the PM2 5 mass apportioned for the coal combustion factor was 2.0% (CI: -0.3, 4.4). The crustal
factor was not associated with mortality (-5.1%; CI: -13.9, 4.6).
Mar et al. (2000) analyzed PM10, PM10_2 5, PM25 measured by two methods, and various
sub-components of PM25 for their associations with total (nonaccidental) and cardiovascular
deaths in Phoenix, AZ during 1995-1997, using both individual PM components and factor
analysis-derived factor scores. In the original analysis, GAM Poisson models (with default
convergence criteria) were used and adjusted for season, temperature, and relative humidity.
In the reanalysis (Mar et al., 2003), GAM models with stringent convergence criteria and GLM
models with natural splines were used. Only cardiovascular mortality was analyzed in the
reanalysis; and the results for that category are summarized here. The evaluated air pollution
variables included O3, SO2, NO2, CO, TEOM PM10, TEOM PM2 5, TEOM PM10.2 5, DFPSS PM2 5,
S, Zn, Pb, soil, soil-corrected K (KS), nonsoil PM, OC, EC, and TC. Lags 0 to 4 days were
8-75
-------
evaluated. A factor analysis conducted on the chemical components of DFPSS PM25 (Al, Si, S,
Ca, Fe, Zn, Mn, Pb, Br, KS, OC, and EC) identified factors for motor vehicle emissions/re-
suspended road dust; soil; vegetative burning; local SO2 sources; and regional sulfate (see
Table 8-4). The results of mortality regression with these factors suggested that the motor
vehicle factor (lag 1 day), vegetative burning factor (3 day lag), and regional sulfate factor
(0 day lag) each had significant positive associations with cardiovascular mortality. The PM2 5
mass was not apportioned to these factors in this study; so information on the excess-deaths
estimate per source-apportioned PM2 5 concentrations was not available. The authors also
analyzed elements from dichot PM10_2 5 samples and identified soil, coarse fraction metals, and
marine influence factors. However, these factors were not analyzed for their associations with
mortality outcomes due to the short measurement period (starting in June 1996 with every third-
day sampling).
It should be noted here that the Smith et al. (2000) analysis of Phoenix data also included
factor analysis on the elements from the coarse fraction and identified essentially the same
factors (the "coarse fraction metals" factor in Mar et al.'s study was called "the anthropogenic
elements" in Smith et al.'s study). While Smith et al. did not relate these factors to mortality
(due to a small sample size), they did show that the anthropogenic elements were low in summer
and spring, when the PM10_2 5 effect was largest. These results suggest that the PM10_2 5 effects
may not necessarily be due to anthropogenic components of the coarse particles, biogenically-
contaminated coarse particles perhaps being key during the warmer months (as noted in
Chapter 7 discussions of bioaerosols).
Tsai et al. (2000) conducted an exploratory analysis of mortality in relation to specific PM
source types for three New Jersey cities (Camden, Newark, and Elizabeth) using factor analysis -
Poisson regression techniques. During the three-year study period (1981 to 1983), extensive
chemical speciation data were available, including nine trace elements, sulfate, and particulate
organic matter. Total (excluding accidents and homicides), cardiovascular, and respiratory
mortality were analyzed. A factor analysis of trace elements and sulfate was first conducted and
identified several major source types: motor vehicle (Pb, CO); geological (Mn, Fe); oil burning
(V, Ni); industrial (Zn, Cu); and sulfate/secondary aerosols (sulfate). In addition to Poisson
8-76
-------
regression of mortality on these factors, an alternative approach was also used, in which the
inhalable particle mass (IPM, D50 < 15 jim) was first regressed on the factor scores of each of the
source types to apportion the PM mass and then the estimated daily PM mass for each source
type was included in Poisson regression, so that RR could be calculated per mass concentration
basis for each PM source type. Oil burning (V, Ni), various industrial sources (Zn, Cd),
motor vehicle (Pb, CO), and secondary aerosols, as well as the individual PM indices IPM,
FPM (D50 < 3.5 |im), and sulfates, were all associated with total and/or cardiorespiratory
mortality in Newark and Camden, but not in Elizabeth. In Camden, the RRs for the source-
oriented PM were higher (1.10) than those for individual PM indices (1.02).
In summary, these source-oriented factor analyses studies suggest that a number of source
types are associated with mortality, including motor vehicle emissions, coal combustion, oil
burning, and vegetative burning. The crustal factor from fine particles was not associated with
mortality in the Harvard Six Cities data. In Phoenix, where coarse particles were reported to be
associated with mortality, the associations between the factors related to coarse particles (soil,
marine influence, and anthropogenic elements) and mortality could not be evaluated due to the
small sample size. Thus, although unresolved issues do remain (mainly due to the lack of
sufficient data), the limited results from the source-oriented evaluation approach (using factor
analysis) thus far seem to implicate ambient fine particles of anthropogenic origin from several
sources as likely being important in contributing to increased mortality risks.
8.2.2.6 New Assessments of Cause-Specific Mortality
Consistent with similar findings described in the 1996 PM AQCD, most of the newly
available studies summarized in Tables 8-1 and 8A-1 that examined nonaccidental total,
circulatory, and respiratory mortality categories (e.g., the NMMAPS analyses reported by
Samet) have continued to find significant PM associations with both cardiovascular and/or
respiratory-cause mortality. Several studies (e.g., Fairley, 1999; his reanalysis, 2003; Wordley
et al., 1997; Prescott et al., 1998) reported estimated PM effects that were generally higher for
respiratory deaths than for circulatory or total deaths.
8-77
-------
The NMMAPS 90-cities analyses not only examined all-cause mortality (excluding
accidents), but also evaluated cardiorespiratory and other remaining causes of deaths (Samet
et al., 2000a,b; reanalysis by Dominici et al., 2002 and 2003). Results were presented for all-
cause, cardiorespiratory, and "other" mortality for lag 0, 1, and 2 days. The investigators
commented that, compared to the result for cardiorespiratory deaths showing 1.6% (CI: 0.8, 2.4)
increase per 50 |ig/m3PM10 in a GLM model (versus 1.1% for total nonaccidental mortality using
GLM), there was less evidence for noncardiorespiratory deaths. However, the estimates for
"other" mortality, although less than half the size of those for cardiorespiratory mortality, were
nevertheless positive, with a fairly high posterior probability (e.g., 0.92 at lag 1 day) that the
overall effects were greater than zero. It should be noted that the "other" (other than
cardiorespiratory) category for underlying cause of mortality may at times include some deaths
influenced by cardiovascular or respiratory causes. For example, Lippmann et al. (2000) noted
that the "other" (noncirculatory and nonrespiratory) mortality in their study showed seasonal
cycles and apparent influenza peaks, suggesting that this series may have also been influenced
by respiratory contributing causes. Thus, interpretation of the observed associations between
PM and broad "specific" categories (e.g., other) of underlying causes of death may not be
straightforward.
As also mentioned earlier in the multicities results section, Schwartz (2003b) reanalyzed
data from Braga et al. (200la) to examine the lag structure of PM10 associations with specific
causes of mortality in ten U.S. cities. The pattern of larger PM10 excess risk estimates for
respiratory categories than for cardiovascular categories found in this study was similar to that in
the Hoek et al. analyses noted above. For example, the combined risk estimates across 10 cities
per 50 |ig/m3 increase in PM10 (2-day mean) were 4.1% (CI: 2.5, 5.6), 7.7% (CI: 4.1, 11.5), and
11.0% (CI: 7, 15.1) for cardiovascular, COPD, and pneumonia, respectively, using GAM with
stringent convergence criteria. These values were even larger for unconstrained distributed lag
models.
Another U.S. study, that of Moolgavkar (2000a), evaluated possible PM effects on cause-
specific mortality across a broad range of lag times (0-5 days) in Cook Co., IL; Los Angeles Co.,
CA; and Maricopa Co., AZ. In addition to total nonaccidental mortality, deaths related to all
8-78
-------
cardiovascular disease (CVD), cerebrovascular disease (CRV), and chronic obstructive lung
disease (COPD) were analyzed in the original study. The data for Cook Co. and Maricopa Co.
were reanalyzed using GAM model with stringent convergence criteria and GLM model with
natural splines (Moolgavkar, 2003). Cerebrovascular disease mortality was not reanalyzed
because there was little evidence of association for PM with this category at any lag in any of the
three counties analyzed. Moolgavkar reported varying patterns of results for PM indices in
evaluations of daily deaths related to CVD and COPD in the two counties. In the Cook Co.
(Chicago) area, the association of PM10 with CVD mortality was statistically significant at a lag
of 3 days based on a single-pollutant analysis and continued to be significantly associated with
CVD deaths with a 3-day lag in two pollutant models including one or another of CO, NO2, SO2,
or O3. In Los Angeles single-pollutant analyses, CVD mortality was significantly associated
with PM10 (2 day lag) and PM2 5 (0 and 1 day lag). Their percent excess risk estimates were up
to twice those for total nonaccidental mortality. In a two-pollutant model with CO (most
strongly positively associated with mortality in Los Angeles Co. among the pollutants), PM10
risk estimates were reduced. However, PM2 5 excess risk estimates in the two-pollutant model
with CO nearly doubled (2.5% per 25 |ig/m3 increase in PM25 to 4.8% using GLM); whereas that
for CO became significantly negative. Obviously, given that CO and PM2 5 were fairly well
correlated (r « 0.58), the estimated associations were most likely confounded between these
two pollutants in this locale. With regard to COPD deaths, PM10 was significantly associated
with COPD mortality (lag 2 days) in Cook Co., but in Los Angeles Co., both PM10 and
(especially) PM2 5 showed erratic associations with COPD mortality at varying lags, alternating
positive and negative (significantly, at lag 3 day) coefficients. The combination of the every
sixth day PM data in Los Angeles (versus daily PM10 in Cook Co.) and relatively small daily
counts for COPD (median = 6/day versus 57/day for CVD) in Los Angeles makes the effective
sample size of COPD mortality analysis small and the results unstable for that county.
The Goldberg et al. (2000; 2001a,b,c,d) study, and its reanalyses (Goldberg et al., 2003;
Goldberg and Burnett, 2003) in Montreal, CN, investigated the role of comorbidity prior to
deaths in PM-mortality associations for various subcategories, including cancer, acute lower
respiratory disease, chronic coronary artery disease, and congestive heart failure (CHF). They
8-79
-------
could classify deaths into these subcategories using medical records from the universal Quebec
Health Insurance Plan (QHIP). This way of classifying deaths would presumably take into
account more detailed information on the disease condition prior to death than the "underlying
cause" in the death records. Thus, the PM-mortality associations could be compared by using
subcategories defined by underlying cause of health (from death records versus those deaths for
which patient records from QHIP could be used to identify the comorbidity conditions). The
Goldberg and Burnett (2003) reanalysis found that total nonaccidental mortality (which was
significantly associated with PM indices in the original report using GAM with default
convergence criteria) was not associated with PM indices in GLM models. They reported that
the associations between PM and nonaccidental mortality were rather sensitive to weather model
specification and did not find significant PM associations with most of the subcategories as
defined from either QHIP or underlying cause. However, they did find significant associations
between CoH, NO2, and SO2 and the CHF deaths as defined from QHIP, but not the CHF deaths
as defined from underlying cause. The association was even stronger in warm seasons. It should
be noted, however, that while the period for this study was relatively long (-10 years) and the
counts for the total nonaccidental deaths were not small (median = 36 deaths per day), the counts
for various subcategories were quite small (e.g., CHF underlying cause mortality mean = 0.75
per day).
Zmirou et al. (1998) presented cause-specific mortality analyses results for 10 of the
12 APHEA European cities (APHEA1). Using Poisson autoregressive models parametrically
adjusting for trend, season, influenza epidemics, and weather, each pollutant's relative risk was
estimated for each city and "meta-analyses" of city-specific estimates were conducted. The
pooled excess risk estimates for cardiovascular mortality were 1.0% (CI: 0.3, 1.7) per
25 |ig/m3increase in BS and 2.0% (CI: 0.5, 3.0) per 50 |ig/m3increase in SO2 in western
European cities. The pooled risk estimates for respiratory mortality in the same cities were
2.0% (CI: 0.8, 3.2) and 2.5% (CI: 1.5, 3.4) for BS and SO2, respectively.
Seeking unique cause-specificity of effects associated with various pollutants has been
difficult because the "cause specific" categories examined are typically rather broad (usually
cardiovascular and respiratory) and overlap and because cardiovascular and respiratory
8-80
-------
conditions tend to occur together. Examinations of more specific cardiovascular and respiratory
subcategories may be necessary to test hypotheses about any specific mechanisms, but smaller
sample sizes for more specific sub-categories may make a meaningful analysis difficult. The
Hoek et al. (2000 and 2001) study and its reanalysis by Hoek (2003) took advantage of a larger
sample size to examine cause-specific mortality. The large sample size, including the whole
population of the Netherlands (mean daily total deaths -330, or more than twice that of
Los Angeles County), allowed examination of specific cardiovascular causes of deaths. The
reanalysis using GAM with stringent convergence criteria as well as GLM with natural splines
either did not change or even increased the effect estimates. Deaths due to heart failure,
arrhythmia, and cerebrovascular causes were more strongly (~2 to 4 times larger excess risks)
associated with air pollution than the overall cardiovascular deaths. The investigators concluded
that specific cardiovascular causes (such as heart failure) were more strongly associated with air
pollution than total cardiovascular mortality, but noted that the largest contribution to the
association between air pollution and cardiovascular mortality was from ischemic heart disease
(about half of all CVD deaths). The analyses of specific respiratory causes, COPD, and
pneumonia yielded even larger risk estimates (e.g., ~6 to 10 times, respectively, larger than that
for overall cardiovascular deaths). Estimated PM10 excess risks per 50 |ig/m3 PM10 (average of
0 through 6 day lags) were 1.2% (CI: 0.2, 2.3), 0.9% (CI: -0.8, 2.7), 2.7% (CI: -4.2, 10.1),
2.4% (CI: -2.3, 7.4), 6.1% (CI: 1, 11.4), and 10.3% (CI: 3.7, 17.2), respectively, for total
nonaccidental, cardiovascular, arrhythmia, heart failure, COPD, and pneumonia, using GAM
models with stringent convergence criteria. Thus, the results from this study with a large
effective sample size also confirm past observations that PM risk estimates for specific causes of
cardiovascular or respiratory mortality can be larger than those estimated for total nonaccidental
mortality.
Another study (Gouveia and Fletcher, 2000), using data from Sao Paulo, Brazil for
1991-1993, evaluated respiratory mortality in children (age < 5 years). The Poisson auto-
regressive model included parametric terms (e.g., quadratic, two-piece linear temperature etc.) to
adjust for weather and temporal trends. Although Gouveia and Fletcher found significant
associations between air pollution and elderly mortality, they did not find statistically significant
8-81
-------
associations between air pollution and child respiratory mortality (the PM10 coefficient was
negative and not significant). However, it should be noted that the average daily respiratory
mortality counts for this study were relatively small (~2.4/day) and the modest period of
observations (3 years) short. Thus, the statistical power of the data was likely less than
desirable, and there may not have been sufficient power to elucidate the range of short-term PM
effects on child respiratory mortality.
Overall, then, the above assessment of newly available studies provides interesting
additional new information with regard to cause-specific mortality related to ambient PM. That
is, a growing number of studies continue to report increased cardiovascular- and respiratory-
related mortality risks as being significantly associated with ambient PM measures at one or
another varying lag times. When specific subcategories of cardiovascular disease were
examined in a large population (e.g., in the Netherlands study by Hoek et al.), some of the
subcategories such as heart failure were more strongly associated with PM and other pollutants
than total cardiovascular mortality. Largest PM effect size estimates are most usually reported
for 0-1 day lags (with some studies also now noting a second peak at 3-4 day lags). A few of the
newer studies also report associations of PM metrics with "other" (i.e., noncardiorespiratory)
causes, as well. However, at least some of these "other" associations may also be due to
seasonal cycles that include relationships to peaks in influenza epidemics that may imply
respiratory complications as a contributing cause to the "other" deaths. Alternately, the "other"
category may include sufficient numbers of deaths due to diabetes or other diseases which may
also involve cardiovascular complications as contributing causes. Varying degrees of robustness
of PM effects are seen in the newer studies, as typified by PM estimates in multiple pollutant
models containing gaseous co-pollutants. That is, some studies show little effect of gaseous
pollutant inclusion on estimated PM effect sizes, some show larger reductions in PM effects to
nonsignificant levels upon such inclusion, and a number also report significant associations of
cardiovascular and respiratory effects with one or more gaseous co-pollutants. Thus, the newer
studies both further substantiate PM effects on cardiovascular- and respiratory-related mortality,
while also pointing toward possible significant contributions of gaseous pollutants to such cause-
8-82
-------
specific mortality. The magnitudes of the PM effect size estimates are consistent with the range
of estimates derived from the few earlier available studies assessed in the 1996 PM AQCD.
8.2.2.7 Salient Points Derived from Assessment of Studies of Short-Term Particulate
Matter Exposure Effects on Mortality
The most salient key points to be extracted from the above discussion of newly available
information on short-term PM exposures relationships to mortality can be summarized as follow:
PM10 effects estimates. Since the 1996 PM AQCD, there have been more than 80 new
time-series PM-mortality analyses published. Estimated mortality relative risks in these studies
are generally positive, statistically significant, and consistent with the previously reported PM-
mortality associations. However, due to the concerns regarding the GAM convergence issue,
quantitative evaluations were made here based only on the studies that either did not use GAM
Poisson model with default convergence criteria or on those studies that have reanalyzed the data
using more stringent convergence criteria and/or used fully parametric approaches.
Of interest are several studies that evaluated multiple cities using consistent data analytical
approaches. The NMMAPS analyses for the largest 90 U.S. cities (Samet et al., 2000a,b;
Dominici et al., 2002 and 2003b) derived a combined nationwide excess risk estimate of about
1.4% (1.1% using GLM) increase in total (nonaccidental) mortality per 50 |ig/m3increase
in PM10. Other well-conducted multicity analyses, as well as various single city analyses,
obtained larger PM10-effect size estimates for total nonaccidental mortality, generally falling in
the range of 2 to 3.5% per 50 |ig/m3increase in PM10. These estimates are consistent with and
overlap the lower end of the range of PM10 risk estimates given in the 1996 PM AQCD.
However, somewhat more geographic heterogeneity is evident among the newer multicity study
results than was the case among the few studies assessed in the 1996 PM AQCD. In the
NMMAPS analysis of the 90 largest U.S. cities data, for example, the risk estimates varied by
U.S. geographic region, with the estimate for the Northeast being the largest (approximately
twice the nation-wide estimate). The observed heterogeneity in the estimated PM risks across
cities/regions could not be explained by city-specific explanatory variables, such as mean levels
of pollution and weather, mortality rate, sociodemographic variables (e.g., median household
income), urbanization, or variables related to measurement error. Notable apparent
8-83
-------
heterogeneity was also seen among effects estimates for PM (and SO2) indices in the multicity
APHEA studies conducted in European cities. In APHEA2, they found that several city-specific
characteristics, such as NO2 levels and warm climate, were important effect modifiers. The issue
of heterogeneity of effect estimates is discussed further in Section 8.4.
Model specification Issue: The investigations of the GAM convergence issue also led to
examination of the sensitivity of the PM risk estimates to different model specifications.
Of particular importance is the reemergence of model specification issues related to control for
weather effects with results of reanalyses highlighting the sensitivity of modeling outcomes to
kinds and numbers of weather-related variables included in base models. Related to this, several
reanalyses also examined the sensitivity of results to varying the degrees of freedom for
smoothing of weather and temporal trends. PM risk estimates were often reduced when more
degrees of freedom were given to model temporal trends. While at present there is no consensus
as to what constitutes an "adequate" extent of smoothing (from an epidemiologic viewpoint), the
overall assessment of PM risk estimates should take into consideration the range of sensitivity of
results to this aspect of model specification.
Confounding and effect modification by other pollutants. Numerous new short-term PM
exposure studies not only continue to report significant associations between various PM indices
and mortality, but also between gaseous pollutants (O3, SO2, NO2, and CO) and mortality.
In most of these studies, simultaneous inclusions of gaseous pollutants in the regression models
did not meaningfully affect the PM-effect size estimates. This was the case for the NMMAPS
90 cities study with regard to the overall combined U.S. regional and nationwide risk estimates
derived for that study. The issue of confounding is discussed further in Section 8.4.
Fine and coarse particle effects. Newly available studies provide generally positive (and
often statistically significant) PM2 5 associations with mortality, with effect size estimates falling
in the range reported in the 1996 PM AQCD. New results from Germany appear to implicate
both ultrafine (nuclei-mode) and accumulation-mode fractions of urban ambient fine PM as
being important contributors to increased mortality risks. As to the relative importance of fine
and coarse particles, in the 1996 PM AQCD there was only one acute mortality study (Schwartz
et al., 1996a) that examined this issue. The results of that study of six U.S. cities suggested that
8-84
-------
fine particles (PM2 5), were associated with daily mortality, but not coarse particles (PM10_2.5),
except for in Steubenville, OH.. Now, eight studies have analyzed both PM25 and PM10_25 for
their associations with mortality. While the results from some of these new studies (e.g., the
Santa Clara County, CA analysis [Fairley, 1999]) did suggest that PM2 5 was more important
than PM10_2 5 in predicting mortality fluctuations, other studies (e.g., Phoenix, AZ analyses
[Clyde et al., 2000; Mar et al., 2000; Smith et al., 2000]) suggest that PM10_2 5 may also be
important in at least some locations. Seasonal dependence of size-related PM component effects
observed in some of the studies complicates interpretations.
Chemical components ofPM. Several new studies have examined the role of specific
chemical components of PM. The studies conducted in U.S., Canadian, and European cities
showed mortality associations with specific fine particle components of PM, including sulfate,
nitrate, and CoH; but their relative importance varied from city to city, likely depending on their
levels (e.g., no clear associations in those cities where H+ and sulfate levels were very low, i.e.,
circa nondetection limits). The results of several studies that investigated the role of crustal
particles, although somewhat mixed, overall do not appear to support associations between
crustal particles and mortality (see also the discussion, below, of source-oriented evaluations).
Source-oriented evaluations. Several studies conducted source-oriented evaluations of PM
components using factor analysis. The results from these studies generally indicated that several
combustion-related fine particle source-types are likely associated with mortality, including
motor vehicle emissions, coal combustion, oil burning, and vegetative burning. The crustal
factor from fine particles was not associated with total nonaccidental mortality in the Harvard
Six Cities data, and the soil (i.e., crustal) factor from fine particles in the Phoenix data was not
associated with cardiovascular mortality. Thus, the results of source-oriented evaluations most
clearly appear to implicate fine particles of anthropogenic origin as being important in
contributing to increased mortality, but not short-term exposures to crustal materials in U.S.
ambient environments.
Cause-specific mortality. Findings for new results concerning cause-specific mortality
comport well with those for total (nonaccidental) mortality, the former showing generally larger
effect size estimates for cardiovascular, respiratory, and/or combined cardiorespiratory excess
8-85
-------
risks than for total mortality risks. An analysis of specific cardiovascular causes in a large
population (The Netherlands) suggested that specific causes of deaths (such as heart failure)
were more strongly associated with ambient PM (and other pollutants) than total cardiovascular
mortality.
Lags. In general, maximum effect sizes for total mortality appear to be obtained with
0-1 day lags, with some studies indicating a second peak for 3-4 days lags. There is also some
evidence that, if effects distributed over multiple lag days are considered, the effect size may be
larger than for any single maximum-effect-size lag day. Lags are discussed further in
Section 8.4.
Threshold. Few new short-term mortality studies explicitly address the issue of thresholds.
One study that analyzed Phoenix, AZ data (Smith et al., 2000) did report some limited evidence
suggestive of a possible threshold for PM25. Also, several different analyses of larger PM10 data
sets across multiple cities (Dominici, et al., 2002; Daniels et al., 2000; and reanalysis by
Dominici et al., 2003a) generally provide only limited, if any, support to indicate a threshold
for PM10 mortality effects. Threshold issues are discussed further in Section 8.4.
8.2.3 Mortality Effects of Long-Term Exposure to Ambient
Particulate Matter
8.2.3.1 Studies Published Prior to the 1996 Particulate Matter Criteria Document
8.2.3.1.1 Aggregate Population Cross-Sectional Chronic Exposure Studies
Mortality effects associated with chronic, long-term exposure to ambient PM have been
evaluated in cross-sectional studies and, more recently, in prospective cohort studies. A number
of older cross-sectional studies from the 1970s provided indications of increased mortality
associated with chronic (annual average) exposures to ambient PM, especially with respect to
fine mass or sulfate (SO42 ) concentrations. These cross-sectional studies were discussed in
detail in Section 12.4.1.2 of the 1996 PM AQCD. However, questions unresolved at the time
regarding the adequacy of statistical adjustments for other potentially important covariates (e.g.,
cigarette smoking, economic status, etc.) across cities tended to limit the degree of confidence
that was placed by the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a) on such
purely "ecological" studies or on quantitative estimates of PM effects derived from them.
8-86
-------
Evidence comparing the toxicities of specific PM components was relatively limited, although
the sulfate and acid components were discussed in detail in the 1986 PM AQCD.
8.2.3.1.2 Prospective Cohort Chronic Exposure Studies
Prospective cohort studies of mortality associated with chronic exposures to air pollution
of outdoor origins have yielded especially valuable insights into the adverse health effects of
long-term PM exposures. Such cohort studies using subject-specific information about relevant
covariates (such as cigarette smoking, occupation, etc.) typically are capable of providing more
certain findings of long-term PM exposure effects than are purely "ecological studies" (Kiinzli
and Tager, 1997). The new, better designed cohort studies, as discussed below, have largely
confirmed the magnitude of PM effect estimates derived from past cross-sectional studies.
The extensive Harvard Six-Cities Study (Dockery et al., 1993) and the initial American
Cancer Society (ACS) Study (Pope et al., 1995) agreed in their findings of statistically
significant positive associations between fine particles and excess mortality, although the ACS
study did not evaluate the possible contributions of other air pollutants. Neither study
considered multipollutant models, although the Six-City study did examine various PM and
gaseous pollutant indices (including total particles, PM25, SO42 , H+, SO2, and O3) and found
that sulfate and PM25 fine particles were most strongly associated with mortality. The excess
RR estimates originally reported for total mortality in the Six-Cities study (and 95 percent
confidence intervals, CI) per increments in PM indicator levels were: Excess RR = 18%
(CI: 6.8%, 32%) for 20 |ig/m3 PM10; excess RR = 13.0% (CI: 4.2%, 23%) for 10 |ig/m3 PM25;
and excess RR = 13.4% (CI: 5.1%, 29%) for 5 |ig/m3 SO42 . The estimates for total mortality
derived from the ACS study were excess RR = 6.6% (CI: 3.5%, 9.8%) for 10 |ig/m3 PM25 and
excess RR 3.5% (CI: 1.9%, 5.1%) for 5 |ig/m3 SO42 . The ACS pollutant RR estimates were
smaller than those from the Six-Cities study, although their 95% confidence intervals overlap.
In some cases in these studies, the life-long cumulative exposure of the study cohorts
included distinctly higher past PM exposures, especially in cities with historically higher PM
levels (e.g., Steubenville, OH); but more current PM measurements were used to help estimate
the chronic PM exposures. In the ACS study, the pollutant exposure estimates were based on
8-87
-------
concentrations at the start of the study (during 1979-1983). In addition, the average age of the
ACS cohort was 56, which could overestimate the pollutant RR estimates and perhaps
underestimate the life-shortening associated with PM associated mortality. Still, although
caution must be exercised regarding use of the reported quantitative risk estimates, the Six-Cities
and ACS prospective cohort studies provided consistent evidence of significant mortality
associations with long-term exposure to ambient PM.
The Six-Cities cohort was preselected by the investigators to be a representative population
for the U.S. midwest / eastern regions of the country heavily-impacted by both coal combustion
and motor vehicle effluents. By contrast, the ACS study cohort was drawn from a large pool of
volunteers who happened to live in communities where several years of fine particle and/or
sulfate ambient air concentration data were available. It is important to note that the ACS had a
relatively small proportion of people with less than high school education (12% versus 28% for
Six-Cities) and, by inference, better diets and access to good health care than an average U.S.
population. To the extent that the mortality impact is lower in the better educated portion of the
population, the mortality experience of the ACS cohort likely provides an underestimate for the
U.S. population as a whole.
In contrast to the Six-Cities and ACS studies, early results reported by Abbey et al. (1991)
and Abbey et al. (1995a) from another prospective cohort study, the Adventist Health Study on
Smog (AHSMOG), reported no significant mortality effects of previous PM exposure in a
relatively young cohort of California nonsmokers. However, these analyses used TSP as the PM
exposure metric, rather than more health-relevant PM metrics such as PM10 or PM2 5, included
fewer subjects than the ACS study, and considered a shorter follow-up time than the Six-Cities
study (ten years versus 15 years for the Six-Cities study). Further, the AHSMOG study included
only nonsmokers (indicated by the Six-Cities Study as having lower pollutant RR's than
smokers), suggesting that a longer follow-up time than considered in the past (10 years) might be
required to have sufficient power to detect significant pollution effects than would be needed in
studies that include smokers (such as the Six-Cities and ACS studies). Thus, greater emphasis
was placed in the 1996 PM AQCD on the results of the Six-Cities and ACS studies.
-------
Overall, the previously available chronic PM exposure studies collectively indicated that
increases in mortality are associated with long-term exposure to ambient airborne particles; and
effect size estimates for total mortality associated with chronic PM exposure indices appeared to
be much larger than those reported from daily mortality PM studies. This suggested that a major
fraction of the reported mortality relative risk estimates associated with chronic PM exposure
likely reflects cumulative PM effects above and beyond those exerted by the sum of acute
exposure events (i.e., assuming that the latter are fully additive over time). The 1996 PM AQCD
(Chapter 12) reached several conclusions concerning four key questions about the prospective
cohort studies, as noted below:
(1) Have potentially important confounding variables been omitted?
"While it is not likely that the prospective cohort studies have overlooked plausible
confounding factors that can account for the large effects attributed to air pollution,
there may be some further adjustments in the estimated magnitude of these effects as
individual and community risk factors are included in the analyses." These include
individual variables such as education, occupational exposure to dust and fumes, and
physical activity, as well as ecological (community) variables such as regional
location, migration, and income distribution. Further refinement of the effects of
smoking status may also prove useful."
(2) Can the most important pollutant species be identified?
"The issue of confounding with co-pollutants has not been resolved for the
prospective cohort studies . . . Analytical strategies that could have allowed greater
separation of air pollutant effects have not yet been applied to the prospective cohort
studies." The ability to separate the effects of different pollutants, each measured as a
long-term average on a community basis, was clearly most limited in the Six Cities
study. The ACS study offered a much larger number of cities, but did not examine
differences attributable to the spatial and temporal differences in the mix of particles
and gaseous pollutants across the cities. The AHSMOG study constructed time- and
location-dependent pollution metrics for most of its participants that might have
allowed such analyses, but no results were reported.
(3) Can the time scales for long-term exposure effects be evaluated?
"Careful review of the published studies indicated a lack of attention to this issue.
Long-term mortality studies have the potential to infer temporal relationships based
on characterization of changes in pollution levels over time. This potential was
greater in the Six Cities and AHSMOG studies because of the greater length of the
historical air pollution data for the cohort [and the availability of air pollution data
throughout the study]. The chronic exposure studies, taken together, suggest that
8-89
-------
there may be increases in mortality in disease categories that are consistent with long-
term exposure to airborne particles, and that at least some fraction of these deaths are
likely to occur between acute exposure episodes. If this interpretation is correct, then
at least some individuals may experience some years of reduction of life as a
consequence of PM exposure."
(4) Is it possible to identify pollutant thresholds that might be helpful in health assessments?
"Model specification searches for thresholds have not been reported for prospective
cohort studies. . . . Measurement error in pollution variables also complicates the
search for potential threshold effects. . . . The problems that complicate threshold
detection in the population-based studies have a somewhat different character for the
long-term studies."
8.2.3.2 New Prospective Cohort Analyses of Mortality Related to Chronic Particulate
Matter Exposures
Considerable further progress has been made towards addressing the above issues. As an
example, extensive reanalyses (Krewski et al., 2000) of the Six-Cities and ACS studies
(sponsored by HEI) indicate that the published findings of the original investigators (Dockery
et al., 1993; Pope et al., 1995) are based on substantially valid data sets and statistical analyses.
The HEI reanalysis project demonstrated that small corrections in input data have very little
effect on the findings and that alternative model specifications further substantiate the robustness
of the originally reported findings. In addition, some of the above key questions have been
further investigated by Krewski et al. (2000) via sensitivity analyses (in effect, new analyses) for
the Six City and ACS studies data sets, including consideration of a much wider range of
confounding variables. Newly published analyses of ACS data for more extended time periods
(Pope et al., 2002) further substantiate original findings and also provide much clearer, stronger
evidence for ambient PM exposure relationships with increased lung cancer risk. Newer
published analyses of AHSMOG data (Abbey et al., 1999; Beeson et al., 1998) also extend the
AHSMOG findings and show some analytic outcomes different from earlier analyses reported
out from the study. Results from the Veterans' Administration- Washington University
(hereafter called "VA") prospective cohort study are also now available (Lipfert et al., 2000b).
8-90
-------
8.2.3.2.1 Health Effects Institute Reanalyses of the Six-Cities and ACS Studies
The overall objective of the HEI "Particle Epidemiology Reanalysis Project" was to
conduct a rigorous and independent assessment of the findings of the Six Cities (Dockery et al.,
1993) and ACS (Pope et al., 1995) studies of air pollution and mortality. The following
description of approach, key results, and conclusions is largely extracted from the Executive
Summary of the HEI final report (Krewski et al., 2000). The HEI-sponsored reanalysis effort
was approached in two steps:
• 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.
• Part II: Sensitivity Analyses. The Reanalysis Team tested the robustness of the original
analyses to alternate risk models and analytic approaches.
The Part I audit of the study population data for both the Six Cities and ACS studies and of
the air quality data in the Six Cities Study revealed that data were of generally high quality with
few exceptions. In both studies, a few errors were found in the data coding for and exclusion of
certain subjects; but when those subjects were included in the analyses, they did not materially
change the results from those originally reported. Because the air quality data used in the ACS
Study could not be audited, a separate air quality database was constructed for the sensitivity
analyses in Part II.
The Reanalysis Team was able to replicate the original results for both studies using the
same data and statistical methods as used by the original investigators, as shown in Table 8-5.
The Reanalysis Team confirmed the original point estimates. For the Six Cities Study, they
reported the excess relative risk of mortality from all causes associated with an increase in fine
particles of 10 |ig/m3 to be 14%, close to the 13% reported by the original investigators. For the
ACS Study, they reported the relative risk of all-cause mortality associated with a 10 |ig/m3
increase in fine particles to be 7.0% in the reanalysis, close to the original 6.6% value.
The Part II sensitivity analysis applied an array of different models and variables to
determine whether the original results would remain robust to different analytic assumptions and
model specifications. The Reanalysis Team first applied the standard Cox model used by the
original investigators and included variables in the model for which data were available from
8-91
-------
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 City b
Six City b
ACS Study0
HEI reanalysis Phase I:
Replication
Six City Reanalysis d
ACS Study Reanalysis d
Indicator
PM25
PM15/10
PM25
PM25
PM15
PM25
PM15
(dichot)
PM15 (SSI)
Mortality Risk per
Total Mortality
Excess Relative Risk (95% CI)
13% (4.2%, 23%)
18% (6.8%, 32%)
6.6% (3.5%, 9.8%)
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%)
Increment in PM a
Cardiopulmonary Mortality
Excess Relative Risk (95% CI)
18% (6.0%, 32%)
e
12% (6.7%, 17%)
19% (6.5%, 33%)
20% (2.9%, 41%)
12% (7.4%, 17%)
7.3% (3.0%, 12%)
5.7% (2.5%, 9.0%)
a 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).
dKrewskietal. (2000).
e Results presented only by smoking category subgroup.
both original studies, but had not been used in the published analyses (e.g., physical activity,
lung function, marital status). The Reanalysis Team also designed models to include interactions
between variables. None of these alternative models produced results that materially altered the
original findings.
Next, for both the Six Cities and ACS studies, the Reanalysis Team investigated the
possible effects of fine particles and sulfate on a range of potentially susceptible subgroups of
the population. These analyses did not find differences in PM-mortality associations among
subgroups based on various personal characteristics (e.g., including gender, marital status,
smoking status, and exposure to occupational dusts and fumes). However, estimated effects of
8-92
-------
fine particles did vary with educational level: the association between an increase in fine
particles and mortality tended to be higher for individuals without a high school education than
for those with more education. The Reanalysis Team postulated that this finding could be
attributable to some unidentified socioeconomic effect modifier. The authors concluded "The
Reanalysis Team found little evidence that questionnaire variables had led to confounding in
either study, thereby strengthening the conclusion that the observed association between fine
particle air pollution and mortality was not the result of a critical covariate that had been
neglected by the Original Investigators." (Krewski et al., 2000, pp. 219-220).
In the ACS Study, the Reanalysis Team tested whether the relationship between ambient
concentrations and mortality was linear. They found some indications of both linear and
nonlinear relationships, depending upon the analytic technique used, suggesting that the shapes
of the concentration-response relationships warrant additional research in the future.
One of the criticisms of both original studies has been that neither analyzed the effects of
change in pollutant levels over time. In the Six Cities Study, for which such data were available,
the Reanalysis Team tested whether effect estimates changed when certain key risk factors
(smoking, body mass index, and air pollution) were allowed to vary over time. In general, the
reanalysis results did not change when smoking and body mass index were allowed to vary over
time. The Reanalysis Team did find for the Six Cities Study, however, that when the general
decline in fine particle levels over the monitoring period was included as a time-dependent
variable, the association between fine particles and all-cause mortality was reduced (Excess
RR = 10.4% (CI: 1.5, 20). This would be expected, because the most polluted cities would likely
have the greatest decline as pollution controls were applied, and it is likely indicative of the
effectiveness of control measures in reducing source emissions importantly contributing to the
toxicity of ambient PM in those cities. Despite this adjustment, the PM25 effect estimate
continued to be positive and statistically significant.
To test the validity of the original ACS air quality data, the Reanalysis Team constructed
and applied its own air quality dataset from available historical data. In particular, sulfate levels
with and without adjustment were found to differ by about 10% for the Six Cities Study. Both
the original ACS Study air quality data and the newly constructed data set contained sulfate
8-93
-------
levels inflated by 50% due to artifactual sulfate. For the Six Cities Study, the relative risks of
mortality were essentially unchanged with adjusted or unadjusted sulfate. For the ACS Study,
adjusting for artifactual sulfate resulted in slightly higher relative risks of all-cause mortality and
from cardiopulmonary disease compared with unadjusted data, while the relative risk of
mortality from lung cancer was lower after the data had been adjusted. Thus, the Reanalysis
Team found essentially the same results as the original Harvard Six-Cities and ACS studies,
even after using independently developed pollution data sets and adjusting for sulfate artifact.
Because of the limited statistical power to conduct most model specification sensitivity
analyses for the Six Cities Study, the Reanalysis Team conducted most of its sensitivity analyses
using only the ACS Study data set that considered 151 cities. When a range of city-level
(ecologic) variables (e.g., population change, measures of income, maximum temperature,
number of hospital beds, water hardness) were included in the analyses, the results generally did
not change. The only exception was that associations with fine particles and sulfate were
reduced when city-level measures of population change or SO2 were included in the model.
A major product of the Reanalysis Project is the determination that both pollutant variables
and mortality appear to be spatially correlated in the ACS Study data set. If not identified and
modeled correctly, spatial correlation could cause substantial errors in both the regression
coefficients and their standard errors. The Reanalysis Team identified several methods for
addressing this, each of which resulted in some reduction in the estimated regression
coefficients. The full implications and interpretations of spatial correlations in these analyses
have not been resolved and were noted to be an important subject for future research.
When the Reanalysis Team sought to take into account both underlying variation from city
to city (random effects) and variation from the spatial correlation between cities, positive
associations were still found between mortality and sulfates or fine particles. Results of various
models, using alternative methods to address spatial autocorrelation and including different
ecologic covariates, found fine particle-mortality associations that ranged from 1.11 to 1.29 (the
RR reported by original investigators was 1.17) per 24.5 |ig/m3 increase in PM2 5. With the
exception of SO2, consideration of other pollutants in these models did not alter the associations
found with sulfates. The authors reported stronger associations for SO2 than for sulfate, which
8-94
-------
suggests that artifactual sulfate was "picking up" some of the SO2 association, perhaps because
the sulfate artifact is in part proportional to the prevailing SO2 concentration (Coutant, 1977).
The Reanalysis Team did not use data adjusted for artifactual sulfate for most alternative
analyses. When they did use adjusted sulfate data, relative risks of mortality from all causes and
cardiopulmonary disease increased. This suggests that more analyses with adjusted sulfate
might result in somewhat higher relative risks for sulfate. The Reanalysis Team concluded:
"it suggests that uncontrolled spatial autocorrelation accounts for 24% to 64% of
the observed relation. Nonetheless, all our models continued to show an association
between elevated risks of mortality and exposure to airborne sulfate" (Krewski et al.,
2000, p. 230).
In summary, the reanalyses generally confirmed the original investigators' findings of
associations between mortality and long-term exposure to PM, while recognizing that increased
mortality may be attributable to more than one ambient air pollution component. Regarding the
validity of the published Harvard Six-Cities and ACS studies, the HEI Reanalysis Report
concluded that "Overall, the reanalyses assured the quality of the original data, replicated the
original results, and tested those results against alternative risk models and analytic approaches
without substantively altering the original findings of an association between indicators of
particulate matter air pollution and mortality."
Villeneuve et al. (2002) used Poisson regression models in further analyses of the Harvard
Six City study cohort to evaluate relationships between fixed-in-time and time-dependent
measures of PM25 and the risk of mortality among adult, Caucasian participants. The RR of
mortality using the Poisson method based upon city-specific exposures that remained constant
during the follow up was 1.31 (CI: 1.12, 1.52), similar to results derived from the Cox model
used in the original analysis. However, the authors report that the RR of mortality due to PM2 5
exposure decreased when time-dependent measures of air pollution were modeled (Table 8-6).
Specifically, when the mean PM25 level within each city during each period of follow-up was
modeled, the RR was 1.16 (CI: 1.02, 1.32). The authors noted that "there were considerable
variations in mortality rates across the calendar periods that were modeled," and that "the
8-95
-------
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 DATA b
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 defined according to 1.19(1.04-1.36) 1.52(1.15-2.00) 1.11(0.95-1.29)
13 calendar periods (no smoothing).a
3 Exposure to PM25 defined according to 1.16 (1.02 - 1.32) 1.43 (1.10 - 1.85) 1.09 (0.93 - 1.26)
13 calendar periods (smoothed).b
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
meanPM25 levels. The calendar periods used were: 1970-1978, 1979, 1981,. .. 1989, and 1990+. PM25
associations with all-cause mortality assessed for male Caucasian participants in Six Cities Study.
Source: Villeneuve et al. (2002).
magnitude of these variations in mortality rates may have dampened any real PM25 effect on
mortality."
Similar results were observed by Villeneuve et al. (2002) irrespective of the exposure
window considered. They used various time-dependent indices that denoted (a) exposures
received in the last two years of follow-up and (b) exposures lagged 3 to 4 and > 5 years. Effect
modification was evaluated by fitting interaction terms that consisted of PM25 exposure and
individual risk factors (body mass index, education, smoking, age, gender, and occupational
exposure to dusts). The significance of this term was formally tested by constructing a
likelihood ratio test statistic. An interaction effect between PM2 5 exposure and age was seen
(p < 0.05), and they therefore presented stratified analysis by age group (< 60, > 60 years).
8-96
-------
For each index of PM25, the RR of all-cause mortality was more pronounced among subjects
< 60 years old. There was no effect modification between PM2 5 and the other individual risk
factors. The RR for PM-associated mortality did not depend on when exposure occurred
in relation to death, possibly because of little variation between the time-dependent city-
specific PM25 exposure indices (r > 0.90) and the fact that the rank ordering of the cities changed
little during follow-up. Villeneuve et al. (2002) concluded that the "attenuated risk of mortality
that was observed with a time-dependent index of PM2 5 is due to the combined influence of city-
specific variations in mortality rates and decreasing levels of air pollution that occurred during
follow-up."
8.2.3.2.2 The A CS Study Extension
Pope et al. (2002) extended the analyses (Pope et al., 1995) and reanalyses (Krewski et al.,
2000) of the ACS CPS-II cohort to include an additional nine years of follow-up data. The 2002
study has a number of advantages over the previous analyses, in that it (a) doubles the follow-up
time from 7 to 16 years and triples the number of deaths; (b) expands the ambient air pollution
data substantially, including two recent years of fine particle data and adding data on gaseous
co-pollutants; (c) improves statistical adjustments for occupational exposure; (d) incorporates
data on dietary covariates believed to be important factors in mortality, including total fat
consumption, and consumption of vegetables, citrus fruit, and high-fiber grains; and (e) uses
recent developments in nonparametric spatial smoothing and random effects statistical models as
input to the Cox proportional hazards model. Each participant was identified with a specific
metropolitan area, and mean pollutant concentrations were calculated for all metropolitan areas
with ambient air monitors in the one to two years prior to enrollment. Ambient pollution during
the follow-up period was extracted from the AIRS data base. There was no network of PM2 5
monitoring in the United States between the early 1980s and the late 1990s. In an attempt to
estimate the concentration during this period, the integrated average of PM25 concentrations
during 1999 to 2000 was averaged with the earlier 1979 to 1983 period. For the 51 cities where
paired data were available, the concentrations of PM2 5 were lower in 1999 to 2000 than in 1979
to 1983 for most cities. Mean PM25 levels for the two periods were highly correlated (r = 0.78),
8-97
-------
and the rank order of the cities by relative pollution levels remained nearly the same. Analyses
based on the early period would likely provide the best estimate of PM2 5-associated risks, as
shown in Figures 8-8 and 8-9. Averages of daily averages of the gaseous pollutants were used
except for O3, where the average daily 1-h maximum was calculated for the whole year and for
the typical peak O3 quarter (July, August, September). Mean sulfate concentrations for 1990
were calculated from archived quartz filters, virtually eliminating the historical sulfate artifact
leading to overestimation of sulfate concentrations.
The Krewski et al. (2000), Burnett et al. (200la), and Pope et al. (2002) studies were
concerned that survival times of participants in nearby locations might not be independent of
each other, due to missing, unmeasured, or mis-measured risk factors or their surrogates that
may be spatially correlated with air pollution, thus violating an important assumption of the Cox
proportional hazards model. Thus, model fitting proceeded in two stages, the first of which was
an adjusted relative risk model with a standard Cox proportional hazards model including
individual-specific covariates and indicator variables for each metropolitan area, but not air
pollutants. In the second stage, the adjusted log(relative risks) were fitted to fine particle
concentrations or other air pollutants by a random effects linear regression model.
Models were estimated separately for each of four mortality (total, cardiopulmonary, lung
cancer, and causes other than cardiopulmonary or lung cancer deaths) endpoints for the entire
follow-up period and for fine particles in three time periods (1979-1983, 1999-2000, and the
average of the mean concentrations in these two periods). The results are shown in Table 8-7.
Figures 8-7, 8-8, and 8-9 show the results displayed in Figures 2, 3, and 5 of Pope et al. (2002).
Figure 8-7 shows that a smooth nonparametric model can be reasonably approximated by a
linear model for all-cause mortality, cardiopulmonary mortality, and other mortality; but the
log(relative risk) model for lung cancer appears to be nonlinear, with a steep linear slope up to
an annual mean concentration of about 13 |ig/m3 and a flatter linear slope at fine particle
concentrations > 13 |ig/m3.
Figure 4 in Pope et al. (2002) shows results for the stratified first-stage models: ages
< 60 and > 69 years are marginally significant for total mortality; ages > 70 years are significant
for cardiopulmonary mortality; and ages 60 to 69 years for lung cancer mortality. Men are at
8-98
-------
TABLE 8-7. SUMMARY OF RESULTS FROM THE EXTENDED ACS STUDY*
Cause of Death
PM2 s, Average Over
1979-1983
PM2 s, Average Over
1999-2000
PM2 5, Average Over
All Seven Years
All causes
Cardiopulmonary
Lung cancer
Other
4.1% (0.8,7.5%)
5.9% (1.5, 10.5%)
8.2% (1.1, 15.8%)
0.8% (-3.0,4.8%)
5.9% (2.0, 9.9%)
7.9% (2.3, 14.0%)
12.7% (4.1,21.9%)
0.9% (-3.4,5.5%)
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 PM2 5 by cause of death associated
with each of the multiyear averages of fine particle concentrations. The multiyear 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).
(
0)
'
0)
O)
o
0.2-
0.1 -
0.0-
-0.1 -
-0.2-
-0.3-
-0.4-
(Ł
0}
]s
0)
O)
0
All Cause
10 15 20
Concentration (|jg/m3)
0.2
0.1
0.0
X -°-1
§" -0.3
-0.4-
Cardiopulmonary
10 15 20
Concentration ([jg/m3)
0.2-
0.1 -
01 -
0.2-
0.3-
0.4-
^
.- ^^—
/S
;S /
' /
/ Lung Cancer
/
oE
-------
Random Effects
1.15-
1.10-
1.05-
1 nn —
i .\j\j —
n QS
1.15-
1.10-
1.05-
.52 1 .00 -
&•
0)
> 0.95
15 1.15-
o
1.10-
1.05-
Inn —
.UU
r\ QJT
u.yo
1.15-
J2
c
g
1
'ra
co
n
^
CD
o:
CD
E =a
5 | .1
Standard Cox
g Models ^
^n °
CD" o JD —
D) g y
r 4
f ^
LLJ
f *)
ro
•"E '
ra a
Q.
X
0 g_|^]
! 1 §
D <
c o
1 ^
Random (with all covariates)
Effects P'us SPatial
(with an „ Smoothing
covariates) c „ g ^
I T 5-S II 11
5 + + + 03 +
O JT3 O) >
»
<
-1-
-••=• i Q.
' i ^
it
All-Cause
J
4
[ ,
f i
•
f 1
[ i
[ i
[ i
f ]
> 4
> <
^ ^ ^
• • """
Cardiopulmonary
<
•
1111111
4
^
T
> ^
> ^ ^ ^ ^
Lung Cancer
1.10-
1.05-
1.00-
M f-
•t
All Other -i
« f- i
>
<
* <
> ^
>
Figure 8-8. Relative risk of total and cause-specific mortality per 10 ug/m3 PM25, derived
for means of 1979-1983 PM2 5 data for various cities, using 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).
8-100
-------
Particles
Inhalable, Coarse,
0,
0)
1.25-
1.20-
1.15-
1.10-
1.05-
1 nn
I .UU ~~
0.95-
0.90-
1.25-
1.20-
1.15-
1.10-
1.05-
1.00-
0.90-
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-
„ l
^ 4
^
>
L
^
^
LO 1 .-
rvi O L
ui 00^ C
_J~ r
^ n _.
^ >
^
^
j
k
^ ^
•<
>
>
4
"""
4
>
> n
4]
1 -1
-L
^
4
t
>
• • -A L- - -
"1 ^
J_ ^
...1
.
^...<
It. .i
_
k 1
4
»
,..L.j
T 4
r _l
, -
t J
"CD <Ą ,
^00 CD C
L CXI 5°°
0 °°. Q1 '
~ D- OT ^
"^ rr* C
301 oo *>•
0~ i 0) r
° § S ^ CNI „ P S
0" ^ 2, S- § § g c
> OT w o 8 o S ^ -
^ ^. 1 * * " J
LI t T ^
4
>
n
3 °
? c
^
>
^
M
^
CO
K
All-Cause
T
T ^
* 4> i
TT .
>
T I T 1
T 4> I
.
4
>
^
>
^
Cardiopulmonary -L
<
>
^^ ^
T ^
k 1 -r- -r-
T 1
' 1 " * " •• t r
1 Lung Cancer | ^
>
4
^
>
>
<
^
.
>
L * * , T _ i
>-I -..r...T..|.f..^..,
^ All Other
4
>
.
>
Figure 8-9. 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).
8-101
-------
significantly higher risk for total and lung cancer mortality than are women, but slightly less so
for cardiopulmonary mortality (although still significant). Log (RR) decreases significantly
from individuals with less than to those with more than a high school education, replicating
findings in Krewski et al. (2000), but with twice the time on study. Including smoking status
showed increased fine particle RR for cardiopulmonary and lung cancer mortality in never-
smokers and least effect in current smokers; however, for total mortality, significant or near-
significant effects occurred in both current and never-smokers, but not former smokers.
The second-stage random effects models on the right side of Figure 8-8 have much wider
confidence intervals than the first-stage models, but are still statistically significant for total,
cardiopulmonary, and lung cancer mortality. Spatial smoothing decreased the magnitude and
significance of the fine particle effect for total mortality. For cardiopulmonary mortality, spatial
smoothing increased the magnitude of the RR and its significance by reducing the width of the
confidence intervals in the "50%-span" and "lowest variance" smoothing methods. For lung
cancer mortality, spatial smoothing little changed the magnitude of the RR, but increased its
significance by reducing the width of confidence intervals in the "50%-span" and "lowest
variance" smoothing methods.
Figure 8-9 shows statistically significant relationships between fine particles and total,
cardiopulmonary, and lung cancer mortality no matter which averaging span was used for PM2 5
and slightly larger effect estimates for the average concentration of the 1979 to 1983 and 1999 to
2000 intervals. PM15 for 1979 to 1983 is significantly associated with cardiopulmonary
mortality and marginally with total mortality; whereas 1987 to 1996 PM15 is not significantly
associated with cardiopulmonary mortality. Coarse particles (PM15_2 5) and TSP are not
significantly associated with any endpoint, but are positively associated with cardiopulmonary
mortality. Sulfate particles are very significantly associated with all endpoints, including
mortality from all other causes, but only marginally for lung cancer mortality using 1990 filters.
Figure 8-9 also shows highly positive significant relationships between SO2 and total,
cardiopulmonary, and other-causes mortality, but a weaker SO2 association with lung cancer
mortality. Only O3 using only the third quarter for 1982 to 1998 showed a marginally significant
relationship with cardiopulmonary mortality, but not the year-round average. The other criteria
8-102
-------
pollutants, CO and NO2, are neither significantly nor positively related to any mortality endpoint,
unlike some findings for acute PM exposure-mortality studies.
This paper confirms that the general pattern of findings for the first seven years of the
study (Pope et al., 1995; Krewski et al., 2000) can be reasonably extrapolated to the patterns that
remain present with twice the length of time on study and three times the number of deaths.
As shown later in Table 8-11, the excess relative risk estimate (95% CI) per 10 |ig/m3 PM25 for
total mortality in the original ACS study (Pope et al., 1995) was 6.6% (CI: 3.6, 9.9); in the ACS
reanalysis (Krewski et al., 2000) it was 7.0% (CI: 3.9, 10); and, in the extended ACS data set
(Pope et al., 2002), it was 4.1% (CI: 0.8, 7.5) using the 1979 to 1983 data and 6.2% (CI: 1.6, 11)
using the average of the 1979 to 1983 and 1999 to 2000 data. The excess relative risk estimate
(95% CI) per 10 |ig/m3 PM2 5 for cardiopulmonary mortality in the original ACS study (Pope
et al., 1995) was 12% (CI: 6.7, 17); in the ACS reanalysis (Krewski et al., 2000), it was 12%
(CI: 7.4, 17); and, in the extended ACS data set (Pope et al., 2002), it was 5.9% (CI: 1.5, 10)
using the 1979 to 1983 data and 9.3% (CI: 3.3, 16) using the average of the 1979 to 1983 and
1999 to 2000 data. Thus, the additional data and statistical analyses reported by Pope et al.
(2002) yield somewhat smaller estimates than the original study (Pope et al., 1995), but are
similar to estimates from the Krewski et al. (2000) reanalysis of the original ACS data set.
The Pope et al. (2002) study also considered the PM risks by subgroup characteristics. The
risks were generally found (although not significantly so) to be somewhat higher for males than
females. The PM2 5 relative risks also tended to be higher for nonsmokers than smokers. This is
consistent with the fact that smokers would have a much higher baseline risk, especially for lung
cancer, and would tend to have lower air pollution-mortality risk when viewed relative to the
much higher smoker baseline risk. PM2 5 mortality relative risks also tended to be higher for
those with less education, which may be due to related socioeconomic factors or, more likely,
to the generally greater inter-state mobility of higher-educated persons. Since the MSA was
assumed unchanged from that at the start of the study, this would tend to weaken the association
for higher education subjects, as the MSA-based exposure information would tend to have less
accuracy in that highly mobile group. This may indicate that the less-educated group RR
estimates may be more indicative of the true PM25 effects (i.e., as their exposure information is
8-103
-------
likely to be more accurate) and, therefore, that the overall study PM2 5 RR estimates (which
include the highly-educated) may be biased somewhat low.
Based on the above patterns of results, the authors drew the following conclusions:
(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.
(2) The estimated fine particle effect on cardiopulmonary 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, 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
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 fact that less-education is a
marker for lower socioeconomic status and, hence, poorer health status and greater
pollution susceptibility. These results may also reflect that the mobility of the less-
educated may provide better estimates of exposure in this study (with no follow up of
address changes) than for the more mobile well-educated. In either case, because this
cohort has 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 than are likely to be found for the general
public.
(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 differ across locations or are 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 could address
the problem. However, if location is associated with air pollution, then the spatial
correlation may be evaluated using nonparametric smoothing methods. No significant
spatial autocorrelation was found after controlling for fine particles. Even after adjusting
for spatial correlation, estimated PM2 5 effects were significant and persisted for
cardiopulmonary and lung cancer mortality and were borderline significant for total
mortality, but with much wider confidence intervals after spatial smoothing.
(4) Fine particles (PM2 5) were associated with elevated total, cardiopulmonary, and lung
cancer mortality risks, but not with other-cause mortality. PM10 for 1987-1996 and PM15
for 1979-1983 were just significantly associated with cardiopulmonary mortality, but
neither PM10_2 5 nor TSP were associated with total or any cause-specific mortality.
All endpoints but lung cancer mortality were very significantly associated with sulfates,
8-104
-------
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. Thus,
neither coarse thoracic particles nor TSP were significantly associated with mortality; nor
were CO and NO2 on a long-term exposure basis.
(5) The concentration-response curves estimated using nonparametric smoothers were all
monotonic and nearly linear (except for lung cancer). However, the shape of the curve
may become nonlinear at much higher concentrations.
(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.
8.2.3.2.3 AHSMOGAnalyses
The Adventist Health Study of Smog (AHSMOG), another major U.S. prospective cohort
study of chronic PM exposure-mortality effects, started with enrollment in 1977 of 6,338
nonsmoking non-Hispanic white Seventh Day Adventist residents of California, ages 27 to
95 years. All had resided for at least 10 years within 5 miles (8 km) of their then-current
residence locations, either within one of the three major California air basins (San Diego,
Los Angeles, or San Francisco), or else were part of a random 10% sample of Adventist Health
Study participants living elsewhere in California. The study has been extensively described and
its initial results reported earlier (Hodgkin et al., 1984; Abbey et al., 1991; Mills et al., 1991).
In the more recent AHSMOG analyses (Abbey et al., 1999), the mortality status of subjects
after -15 years of follow-up (1977-1992) was determined by various tracing methods and 1,628
deaths (989 female, 639 male) were found in the cohort. This 50% percent increase during the
follow-up period (versus previous AHSMOG reports) notably enhances the power of the latest
analyses over past published ones. Of 1,575 deaths from all natural (non-external) causes, 1,029
were cardiopulmonary, 135 were nonmalignant respiratory (ICD9 codes 460-529), and 30 were
lung cancer (ICD9 code 162) deaths. Abbey et al. (1999) also created another death category,
8-105
-------
contributing respiratory causes (CRC), which included any mention of nonmalignant respiratory
disease as an underlying or "contributing cause" on the death certificate. Numerous analyses
were done for the CRC category, due to the large numbers and relative specificity of respiratory
causes as a factor in the deaths. Education was used to index socioeconomic status, rather than
income. Physical activity and occupational exposure to dust were also used as covariates. Cox
proportional hazard models adjusted for a variety of covariates or stratified by sex were used.
The "time" variable used in most of the models was survival time from date of enrollment,
except that age on study was used for lung cancer effects due to the expected lack of short-term
effects. Many covariate adjustments were evaluated, yielding results for all non-external
mortality as shown in Table 8-8.
TABLE 8-8. RELATIVE RISK OF MORTALITY FROM ALL NONEXTERNAL
CAUSES, BY SEX AND AIR POLLUTANT, FOR AN ALTERNATIVE COVARIATE
MODEL IN THE AHSMOG STUDY
Pollution Index
PM10 > 100, cl/yr
PM10 mean
SO4 mean
03 > 100 ppb, b/yr
SO2 mean
Pollution
Increment
30 days/yr
20 ug/rn3
5 ug/rn3
551h/yr(IQR)
3.72 (IQR)
RR
0.958
0.95
0.901
0.9
1
Females
LCL
0.899
0.873
0.785
0.8
0.91
UCL
1.021
1.033
1.034
1.02
1.1
RR
1.082
1.091
1.086
1.14
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).
As for cause-specific mortality analyses of the AHSMOG data, positive and statistically
significant effects on deaths with underlying contributing respiratory causes were also found for
30 day/year > 100 |ig/m3 PM10 (RR =1.14, CI: 1.03, 1.56) in models that included both sexes
and adjustment for age, pack-years of smoking, and BMI. Subsets of the cohort had elevated
8-106
-------
risks: (a) former smokers had higher RRs than never-smokers (RR for PM10 exceedances for
never- smokers was marginally significant by itself); (b) subjects with low intake of antioxidant
vitamins A, C, E had significantly elevated risk of response to PM10, whereas those with
adequate intake did not (suggesting that dietary factors or, possibly, other socioeconomic or life
style factors for which they are a surrogate may be important covariates); and (c) there also
appeared to be a gradient of PM10risk with respect to time spent outdoors, with those who had
spent at least 16 h/wk outside being at greater risk from PM10 exceedances. The extent to which
time spent outdoors is a surrogate for other variables or is a modifying factor reflecting temporal
variation in exposure to ambient air pollution is not clear, e.g., if the males spent much more
time outdoors than the females, outdoor exposure time could be confounded with gender. When
the cardiopulmonary analyses are broken down by gender (Table 8-9), the RRs for female deaths
were generally smaller than that for males, but none of the risks for PM indices or gaseous
pollutants were statistically significant at p < 0.05.
TABLE 8-9. RELATIVE RISK OF MORTALITY FROM CARDIOPULMONARY
CAUSES, BY SEX AND AIR POLLUTANT, FOR AN ALTERNATIVE
COVARIATE MODEL IN THE AHSMOG 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/rn3
5 ug/rn3
551 h/yr (IQR)
10 ppb
3.72 (IQR)
RR
0.929
0.933
0.95
0.88
0.975
1.02
Females
LCL
0.857
0.836
0.793
0.76
0.865
0.9
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
0.92
0.86
UCL
1.162
1.212
1.086
1.29
1.236
1.18
LCL = Lower 95% confidence limit
Source: Abbey et al. (1999).
UCL = Upper 95% confidence limit
8-107
-------
The AHSMOG cancer analyses yielded very mixed results (Table 8-10) for lung cancer
mortality. For example, RR's for lung cancer deaths were statistically significant for males
for PM10 and O3 metrics, but not for females. In contrast, such cancer deaths were significant for
mean NO2only for females (but not for males), but lung cancer metrics for mean SO2were
significant for both males and females. This pattern is not readily interpretable, but is reasonably
attributable to the very small numbers of cancer-related deaths (18 for females and 12 for males),
resulting in wide RR confidence intervals and very imprecise effects estimates.
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
03>100 ppb,
h/yr
O3 mean
SO2 mean
Pollution
Increment
30 days/yr
20 ug/rn3
19.78
(IQR)
551 h/yr
(IQR)
10 ppb
3.72 (IQR)
Females
Smoking
Category RR LCL UCL
All1 1.055 0.657 1.695
All 1.267 0.652 2.463
All 2.81 1.15 6.89
All 1.39 0.53 3.67
never
smoker
past
smoker
All 0.805 0.436 1.486
All 3.01 1.88 4.84
never 2.99 1.66 5.4
smoker
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
1.5
0.994
1.24
UCL
2.617
5.147
3.57
9.69
43.08
12.07
3.453
3.2
'All = both never smokers and past smokers.
LCL = Lower 95% confidence limit. UCL = Upper 95% confidence limit.
Source: Abbey et al. (1999).
8-108
-------
The analyses reported by Abbey et al. (1999) attempted to separate PM10 effects from those
of other pollutants by use of two-pollutant models, but no quantitative findings from such
models were reported. Abbey et al. did mention that the PM10 coefficient for CRC remained
stable or increased when other pollutants were added to the model. Lung cancer mortality
models for males evaluated co-pollutant effects in detail and indicated that NO2 was
nonsignificant in all two-pollutant models, but other pollutant coefficients were stable.
The PM10 and O3 effects remained stable when SO2 was added, suggesting possible independent
effects, but PM10 and O3 effects were hard to separate because these pollutants were highly
correlated in this study. Again, the very small number of lung cancer observations and likely
great imprecision of reported effects estimates markedly limit the weight that should be accorded
to these cancer results.
Other analyses, by Beeson et al. (1998), evaluated essentially the same data as in Abbey
et al. (1999), but focused on lung cancer incidence (1977 to 1992). There were only 20 female
and 16 male lung cancer cases among the 6,338 subjects. Exposure metrics were constructed to
be specifically relevant to cancer, those being the annual average of monthly exposure indices
from January, 1973 through ensuing months but ending 3 years before date of diagnosis (thus
representing a 3-year lag between exposure and diagnosis of lung cancer). The covariates in the
Cox proportional hazards model were pack-years of smoking and education, and the time
variable was attained age. Many additional covariates were evaluated for inclusion, but only
"current use of alcohol" met criteria for inclusion in the final model. Pollutants evaluated
were PM10, SO2, NO2, and O3. No interaction terms with the pollutants proved to be significant,
including outdoor exposure times. The RR estimates for male lung cancer cases were:
(a) positive and statistically significant for all PM10 indicators; (b) positive and mostly
significant for O3 indicators, except for mean O3, number of O3 exceedances > 60 ppb, and in
former smokers; (c) positive and significant for mean SO2, except when restricted to proximate
monitors; and (d) positive but not significant for mean NO2. When analyses are restricted to the
use of air quality data within 32 km of the residences of subjects, the RR for PM10 over the IQR
of 24 |ig/m3 in the full data set is 5.21 (or RR = 1.99 per 10 |ig/m3 PM10). The female RRs were
8-109
-------
all much smaller than for males, their being significant for mean SO2 but not for any indicator
ofPM10orO3.
The AHSMOG investigators also attempted to compare effects of fine versus coarse
particles (McDonnell et al, 2000). For AHSMOG participants living near an airport (n = 3,769),
daily PM2 5 levels were estimated from airport visibility using previously-described methods
(Abbey et al, 1995b). Given the smaller numbers of subjects in these subset analyses, it is not
necessarily surprising that no pollutants were found to be statistically significant, even based on
analysis for the male subset near airports (n = 1266). It is important to caveat that (a) the PM2 5
exposures were estimated from visibility measurements (increasing exposure measurement error)
and yielded a very uneven and clustered distribution of estimated exposures and; (b) the PM10_25
values were calculated from the differencing of PM10 and PM25, likely adding yet even more
measurement error for the coarse particle (PM10_2 5) variable.
8.2.3.2.4 The EPRI- Washington University Veterans' Cohort Mortality Study
Lipfert et al. (2000b) reported preliminary results from large-scale mortality analyses for a
prospective cohort of up to -70,000 men assembled by the U.S. Veterans Administration (VA)
in the mid-1970s at 32 VA clinics. The VA study group was not originally formed to study air
pollution, but was later linked to air pollution data collected separately. The study that led to the
development of this clinical cohort (Veterans Administration Cooperative Study Group on
Antihypertensive Agents, 1970; 1967) was a "landmark" VA cooperative study demonstrating
that anti-hypertensive treatment markedly decreased morbidity and mortality (Perry et al., 1982).
The clinical cohort itself involved actual clinical rather than research settings. Lipfert et al.
(2000b) noted: "This cohort differs from a general male population in being limited to
hypertensive patients and it differs from the cohorts that are randomized into large-scale
multicenter trials since it contains a broad spectrum of subjects including many with various
co-morbidities."
The VA study cohort was male, middle-aged (51 ± 12 years) and included a larger
proportion of African-Americans (35%) than the U.S. population as a whole and a large
percentage of current or former smokers (81%). The cohort was selected at the time of
8-110
-------
recruitment as being mildly to moderately hypertensive, with screening diastolic blood pressure
(DBF) in the range 90 to 114 mm Hg (mean 96, about 7 mm more than the U.S. population
average) and average systolic blood pressure (SBP) of 148 mm Hg. The subjects had all been
healthy enough to be in the U.S. armed forces at one time. As stated by Lipfert et al. (2000b),
"Twelve percent had a pulmonary abnormality on physical examination, 9% were diabetic;
19% had a history of heart disease; 7% had a history of stroke, and 56% had a positive
cardiorenal family history." Contextual socioeconomic variables were also assembled at the
ZIP-code and county levels. The ZIP-code level variables were average education, income, and
racial mix. County-level variables included altitude, average annual heating-degree days,
percentage Hispanic, and socioeconomic indices. Census-tract variables included poverty rate
and racial mix.
Detailed exposure information was obtained by averaging air quality data by year for each
county of residence at the time of entry to the study. County-wide air pollution variables
included TSP, PM10, PM2 5, PM15, PM15.2 5, SO4, O3, CO, and NO2. The VA PM2 5 pollutant data
were derived from the same data set as used in the ACS study; that PM2 5 data set (103 counties
with monitors) was much smaller than the VA TSP data set (1,379 counties). In the 1,379
counties with TSP data, there were 67,537 subjects. For PM10, during the period 1989 to 1996,
there were 59,053 subjects; for PM25 during 1979 to 1981, there were 26,067 subjects; for PM2 5
during 1982 to 1984, there were 29,177; for PM15 during 1979 to 1981, there were 26,067;
for PM15 during 1982 to 1984, there were 29,177; for PM15.25 during 1979 to 1981, there were
26,067; and for PM15.25 during 1982 to 1984, there were 29,177. Lipfert et al. (2000b) stated
"The IP data used here were derived from 103 monitors," and "Matching at the county level
substantially reduces the errors in estimated exposures incurred by averaging across an entire
metropolitan area." Lipfert et al. (2000b) also indicated: "In this study, the mortality risks were
based on the mean concentration of pollutants less estimated background weighted by the
number of subjects in each county. . . . Background is estimated as the mean concentration less
3 standard deviations. In the few cases for which this value was negative (indicating a skewed
distribution) the background was taken as zero."
8-111
-------
Besides considering average exposures over the entire period, three sequential mortality
follow-up periods (1976 to 1981, 1982 to 1988, 1989 to 1996) were also evaluated in separate
statistical analyses that related mortality in each of those periods to air pollution in different
preceding, concurrent, or subsequent periods (i.e., up to 1975, 1975 to 1981, 1982 to 1988, and
1989 to 1986, for TSP in the first three periods, PM10 for the last, and NO2, 95th percentile O3,
and 95th percentile CO for all four periods). Mortality in the above-noted periods was also
evaluated in relation to SO4 in each of the same four periods noted for NO2, O3, and CO, and
to PM25, PM15, and PM15.25 in 1979 to 1981 and 1982 to 1984. Thus, Lipfert et al. (2000b)
stated: "With the baseline and final model, deaths during each of the three most recent exposure
periods were considered separately, yielding up to 12 combinations of exposure and mortality
periods for each pollutant. Associations between concurrent air quality and mortality periods
were considered to relate to acute responses, associations with prior exposures were considered
to be emblematic of initiation of chronic diseases and preexposure mortality associations could
only be indirect (temporality violated by design), that is, noncausal, and the results of
intercorrelation or spurious associations."
Results from the VA study are shown in Table 8-11 for various PM indices. Three caveats
were expressed by Lipfert et al. (2000b): "First, the different pollutants, both among species and
among time period within species, may represent different locations because of missing data.
Second, the relative high fraction of mortality within this cohort may have depleted it of
susceptible individuals in the late periods of follow-up. Finally, all of the personal
characteristics of each subject were determined only at the entry to the study. It is quite likely
that many of those characteristics will have changed during the 21 years of follow-up." Lipfert
et al. (2000b) concluded that this may be reason to regard the results for the 1976 to 1981 period
as the most credible. Within a column of this table, the cohort remains unchanged, but the
pollutant differs; however, since missing data vary by pollutant, there are also small changes in
the population considered. Within a row of the table the pollutant remains constant, but the
cohort is "successively depleted in the passage of time."
In Table 8-11 the column at the far right under the heading "single period" presents
regression results from separate model runs for which mortality for the entire follow-up period
8-112
-------
TABLE 8-11. PARTICULATE MATTER EFFECTS ON MORTALITY BY
EXPOSURE AND MORTALITY PERIOD WITH ECOLOGICAL VARIABLES FOR
THE VETERANS COHORT STUDY EXPRESSED AS EXCESS MORTALITY
TSP
TSP
TSP
PM10
PM25
PM25
PM15-PM25
PM15-PM25
PM15
PM15
Exposure Period
up to 1975
1975-81
1982-88
1989-96
1979-81
1982-84
1979-81
1982-84
1979-81
1982-84
Deaths
1976-1981
-0.351D
0.078°
2.060T
7.060T
-5.28°
0.236T
-4.27C
-11.00T
-3.03°
-4.46T
Deaths
1982-1988
-0.81°
- 0.680 D
1.08°
4.33T
-10.07°
-6.11 c
-1.99D
-7.97°
-3.79°
-5.99C
Deaths
1989-1996
-1.49 °
-2.49°
-0.20D
3.43 c
-75.35 D
-10.78°
-9.20°
-12.64°
-7.65°
-9.73°
Single
Period A
-0.18
0.41
0.94
3.92
0.27
-0.06
0.68
-3.64
0.3
-1.54
A - Mortality for the entire followup period (1976-1996) regressed against each exposure period
C - Concurrent
D - Delayed
T - Temporality violated by design
All excess mortality in units of percent per 10 ug/m3. Bold italic print indicates significant at p < 0.05.
Source: Lipfert et al. (2000b).
was regressed against each exposure period for the purpose of comparison with the segmented
mortality analysis and with previous cohort studies (Dockery et al., 1993; Pope et al., 1995; and
Abbey et al., 1999). The single-period analysis represents an aggregated approach to exposure
when the follow-up periods are short, say a few years; thus, the exposure aggregation error may
be small but the study will have reduced power because of the smaller number of deaths. For
longer follow-up periods, say 10 years or more, it becomes important to consider the timing of
death relative to exposure in order to preclude attributing associated mortality to subsequent
exposure. In the present study, the "indirect" cells of the matrix tend to occur early in the
follow-up period while the "delayed" cells tend to occur later.
8-113
-------
Lipfert et al (2000b) stated that "The use of specific exposure periods improve the
precision of the exposure estimates. Response to PM2 5 and PM15 differ greatly between the
single period and the segmented periods; that is thus a prime example of the value of the
segmented analysis in revealing such details. The single mortality period response without
ecological variables are qualitatively similar to what has been reported before (SO4 > PM2 5
> PM15) but the segmented analysis shows that responses to all of the IP variables are negative,
some significantly so."
Lipfert et al (2000b) also stated that specific attention must be given to significant negative
associations between pollution and mortality, which they indicated may be indicative of
confounding or an incomplete model specification. They also noted "It is possible that the
indirect responses may simply reflect random variation and collinearity among time periods."
For example, the correlation between PM25 concentrations the 1979 to 1981 and 1982 to 1984
exposure periods was 0.69. The study found some responses that were consistent with previous
studies but only in the absence of ecological covariates in the model or when responses were
aggregated across the entire period of follow-up. The results from this study indicate that peak
ozone was the only pollutant with constant positive concurrent response. Lipfert et al. (2000b)
state that these overall findings were the result of a more detailed consideration of exposure
timing.
It should be noted that the preliminary screening models used proportional hazards
regression models (Miller et al., 1994) to identify age, SBP, DBF, BMI (nonlinear), age and race
interaction terms, and present or former smoking as baseline predictors, with one or two
pollution variables added. In the final model using 233 terms (of which 162 were interactions of
categorized SBP, DBF, and BMI variables with age), the most significant non-pollution
variables were SBP, DBF, BMI, and their interactions with age, smoking status, average
education, race, poverty, and height. Also, Lipfert et al. (2000b) noted that the mortality risk
associated with current cigarette smoking (1.43) that they found was lower than reported in other
studies. The most consistently positive effects were found for O3 and NO2 exposures in the
immediately preceding years. This study used peak O3 rather than mean O3, as was done in
some other cohort studies. This may account for the higher O3 and NO2 effects here. While the
8-114
-------
PM analyses considering segmented (shorter) time periods gave differing results (including
significant negative mortality coefficients for some PM metrics), when methods consistent with
past studies were used (i.e., many-year average PM concentrations), similar results were
reported: the authors found that "(t)he single-mortality-period responses without ecological
variables are qualitatively similar to what has been reported before (SO4 > PM2 5 > PM15)." With
ecological variables included, a significant PM effect was that for TSP for 1982 to 1988
exposure for the single period. Overall, the authors concluded that "the implied mortality risks
of long-term exposure to air pollution were found to be sensitive to the details of the regression
model, the time period of exposure, the locations included, and the inclusion of ecological as
well as personal variables."
In a follow-up study of the Veterans' Cohort Study, Lipfert et al. (2003) investigated the
importance of blood pressure (BP) as a covariate in studies of long-term associations between air
quality and mortality. The aims of the article were to summarize quantitative relationships
between BP and mortality, to discuss the available information on associations between air
quality and BP, and to present results of a proportional hazard regression sensitivity analysis for
the Veterans' Cohort. The relationship between BP and air quality was considered by reviewing
the literature, by deleting variables from the Veterans' Study proportional hazards regression
models, and by stratifying the analyses of that cohort by diastolic blood pressure (DBF) level.
The literature review found BP to be an important predictor of survival and found small transient
associations between air quality and BP that may be either positive or negative. The regression
model sensitivity runs indicate that the reported VA model associations with air pollution are
robust to the deletion of the BP variables for the entire cohort. For stratified regressions, the
confidence intervals for the air pollution-mortality associations overlapped for the two DBF
groups. The authors, Lipfert et al. (2003), concluded that there is scant evidence that air
pollution affects blood pressure in either healthy or impaired subjects. They went on to note that
the inclusion of BP variables is not strictly essential to derive valid estimates of air pollution
responses, concluding overall that the associations between air quality and mortality are not
mediated through blood pressure.
8-115
-------
8.2.3.2.5 Relationship of Six Cities, A CS, AHSMOG, and VA Study Findings
This section compares findings from the earlier Six Cities study (Dockery et al., 1993), the
ACS Study (Pope et al., 1995), the HEI reanalyses of the latter two studies, the extension of the
ACS Study (Pope et al., 2002), the more recent AHSMOG mortality analyses (Abbey et al.,
1999; McDonnell et al., 2000) and the VA study (Lipfert et al., 2000b). In comparing
prospective cohort studies, some key issues for consideration are: (1) cohort size and
characteristics; (2) study design; and (3) air quality data used in exposure characterization.
Table 8-12 compares the estimated RR for total, cardiopulmonary, and cancer mortality among
the studies.
The number of subjects in these studies varies greatly: 8,111 subjects in the Six-Cities
Study; 295,223 subjects in the 50 fine particle (PM25) cities and 552,138 subjects in the
151 sulfate cities of the ACS Study; 6,338 in the AHSMOG Study; and 26,000 in the VA Study
for PM2 5. This may partially account for differences among their results.
The Six City and AHSMOG studies were designed specifically as prospective studies to
evaluate long-term effects of air pollution and included concurrent air pollution measurements.
The ACS study was also a prospective study, using air pollution data obtained at about the
approximate time of enrollment but not subsequently (Pope et al., 1995), and it evaluated air
pollution effects among a cohort originally recruited to study factors affecting cancer rates. The
extended ACS study incorporated much more air pollution data, including TSP data back to the
1960s and more recent fine particle data. The VA study was originally designed to evaluate the
efficacy of hypertension treatments in male military veterans with hypertension.
The Six-Cities cohort was pre-selected, by the investigators, to be a representative
population, at least for the region of the country that was (is) heavily impacted by both coal
combustion and motor vehicle effluents. By contrast, the ACS study cohort was drawn from a
large pool of volunteers who happened to live in communities where several years of fine
particle and/or sulfate ambient air concentration data were available. The AHSMOG cohort is
drawn from nonsmoking, non-Hispanic white Seventh Day Adventist residents of California.
The VA cohort also presents a narrow population (only male veterans having a very high
percentage of prior smoking, all of whom were diagnosed as hypertensive). Of these four cohort
8-116
-------
TABLE 8-12. COMPARISON OF EXCESS RELATIVE RISKS OF LONG-TERM
MORTALITY IN THE HARVARD SIX CITIES, ACS, AHSMOG, AND VA STUDIES
Study
Six City 3
Six City New4
ACS5
ACS New6
ACS New
ACS New
ACS New
ACS Extend. 7
ACS Extend.
ACS Extend.
AHSMOG8
AHSMOG9
AHSMOG9
VA10
VA10
VA10
PM1
PM25
PM2,
PM2,
PM25
PM15.2.5
PM10/15Dicot
PM10/15SSI
PM25
1979-1983
PM25
1999-2000
PM2 5 Avg.
PM10/15
PM2,
PM10-PM25
PM25
PM25
PM15.25
PM15.2,
PM15
PM15
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%
0.3% u
-10%12
0.7% u
-2.0% 12
0.7% u
-7.6 12
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%)
NS13
SS14
NS13
NS13
NS13
SS14
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,
(6.5,
(6.7,
(7.4,
(-2.2,
(3.0,
(2.5,
(1.5,
(2.3,
(3.3,
(-7.8;
(-3.0;
(-13,
32%)
33%)
17%)
17%)
3.1%)
12%)
9.0%)
10%)
14%)
16%)
, 10%)
, 55%)
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%)
1 Increments are 10 ^ig/rn3 for PM2 5 and PM10/15.2 5 and 20 ng/m3 for PM10/15.
2 Ex. 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 (standard increment) /range).
3FromDockery et al. (1993); Krewski et al. (2000), Part II, Table 21a, original model.
"From Krewski et al. (2000), Part I, Table 21c.
5 From Krewski et al. (2000), Part I, Table 25a.
6 From Krewski et al. (2000), Part I, Table 25c.
'FromPopeetal. (2002).
8 From Abbey et al. (1999), pooled estimate for males and females.
9 From McDonnell et al. (2000), two-pollutant (fine and coarse) models; males only.
10From Lipfert et al. (2000b), Males only, exposure period 1979-1981 from Table 7. Standard errors not provided.
11 Single period mortality (1976-1996).
12 Mortality from 1982-88.
13 Reported by author to be nonsignificant.
14 Reported to be statistically significant.
8-117
-------
studies, the ACS and Six Cities studies are thusly more broadly representative of U.S.
populations.
The estimated mean risk of cigarette smoking in the VA cohort (RR = 1.43) was smaller
than that of the Six City cohort (RR = 1.59) and the ACS cohort (RR = 2.07 for current
smokers). Some possible differences include the higher proportion of former or current smokers
in the VA cohort (81%) versus 51% in the ACS study and 42 to 53% in the Six City study.
A possibly more important factor may be the difference in education levels, as only 12% of the
ACS participants had less than a high school education vs 28% of the Six City cohort. Education
level was not reported for the VA Cohort; however, education differences may be associated
with smoking behavior (more smokers among the less-educated). The ACS, Six Cities and
AHSMOG investigators used Cox Proportional Hazards models to estimate relationships
between mortality and long-term PM exposure; in the VA study, linear regression models were
used. All incorporated potentially confounding variables, such as body mass index or smoking
history; however, as described previously, the VA study included a large number of covariates
and interaction terms in the models. The VA study also differed from the other three studies in
emphasizing analyses using subsets of air quality and mortality data.
As described in more detail in section 8.4.6.4, the Harvard Six Cities study used
dichotomous samplers to measure fine and coarse fraction particles for approximately seven
years in each city. AHSMOG investigators relied on available PM monitoring data, initially
using TSP, then PM10 data, and more recently including PM2 5 data estimated from airport
visibility measurements. Both the VA and ACS studies used PM2 5 data from the same data set,
the IP Network which consisted of 157 sites. For the VA study, the cohort was between 26,000
and 29,000 for the two exposure periods derived from 103 monitors and for the ACS study, the
cohort was 359,000 in 61 MS As for one exposure period 1979 to 1983. Both studies may have
potentially used up to 61 sites in common, but the VA study breaks the average level into two
periods, whereas the ACS study averages the level across the entire time period. For the cities in
common, the same data were used to relate an exposure estimate to the county in the VA study
and to the Metropolitan Statistical Areas (MSAs) in the ACS study. Thus differences in the base
8-118
-------
for deriving exposure estimates for the subjects may have contributed to possible differences
between the representativeness of exposures used in the county or the MSA.
Section 3.2.5 and Appendix 3A discuss spatial variability in PM25 at multiple sites within
MSAs across the United States for 27 MSAs. MSA sites may include those in one to several
counties. Consistency of PM25 values between multiple sites within individual MSAs used to
derive annual averages for PM2 5 can vary by MSAs. The annual averages for many counties
differ by 1 to 2 |ig/m2 across MSAs. In some counties with several monitors, differences
between individual monitors can range from 1 to 2 |ig/m2 to 4 to 6 |ig/m2, at times, for annual
averages.
It is noteworthy that estimated PM effects observed in the VA study appeared to be more
negative in the later years of the study rather than in the earlier years. As noted earlier, this may
also be due to cohort depletion. The participants in the VA Cohort clearly formed an "at-risk"
population, and the results by Vasan et al. (2001) make more plausible the hypothesis stated by
Lipfert et al. (2000b, p. 62) that". . . . the relatively high fraction of mortality within this cohort
may have depleted it of susceptible individuals in the later periods of follow-up."
The Six Cities study found significant associations of PM25 with total and cardiopulmonary
(but not lung cancer) mortality, but not with coarse particle indicators. In the Krewski et al.
(2000) reanalysis of the ACS study data, significant associations were found for both PM25
and PM15 (excess relative risks of 6.6% for a 10 |ig/m3 increase in PM25 and 4% for a 20 |ig/m3
increase in annual PM10/15). The results most recently reported for the AHSMOG study (Abbey
et al., 1999; McDonnell et al., 2000) used PM10 as its PM mass index and found some significant
associations with total mortality and deaths with contributing respiratory causes, even after
controlling for potentially confounding factors (including other pollutants). In further evaluation
of results found for PM10 among males, McDonnell et al. (2000) reported larger associations
with PM25 than PM10_25 for males in the AHSMOG cohort, though none of the 11 PM2 5
associations reached statistical significance. For the VA study, few statistically significant
associations were found with PM indicators; in fact, some statistically significant negative
associations were reported for some subset analyses. Where significant positive associations
were reported, they were generally for the subset of mortality data from the early years of the
8-119
-------
study. The authors note that these were sometimes analyses using mortality data that preceded
the air quality measurements; it is important to note, however, that the design of these studies
uses available air quality data to characterize long-term pollution concentrations, not as a
measure of latency or lag period in effects.
There is no clear consistency in relationships among PM effect sizes, gender, and smoking
status across these studies. The AHSMOG study cohort is a primarily nonsmoker group while
the VA study cohort had a large proportion of smokers and former smokers in an all-male
population. The ACS results show similar and significant associations with total mortality for
both "never smokers" and "ever smokers", although the ACS cohort may include a substantial
number of long-term former smokers with much lower risk than current smokers. The Six Cities
Study cohort shows the strongest evidence of a higher PM effect in current smokers than in
nonsmokers, with female former smokers having a higher risk than male former smokers. This
study suggests that smoking status may be viewed as an effect modifier for ambient PM, just as
smoking may be a health effect modifier for ambient O3 (Cassino et al., 1999).
When the ACS study results are compared with the AHSMOG study results for SO42
(PM10_2.5 and PM10 were not considered in the ACS study, but were evaluated in ACS reanalyses
[Krewski et al., 2000; Pope et al, 2002]), the total mortality effect sizes per 15 |ig/m3 SO42 for
the males in the AHSMOG population fell between the Six-Cities and the ACS effect-size
estimates for males (RR = 1.28 for AHSMOG male participants; RR = 1.61 for Six-Cities Study
male nonsmokers; and RR =1.10 for never smoker males in the ACS study), and the AHSMOG
study 95% confidence intervals encompass both of those other studies' sulfate RR's.
In considering the results of these studies together, statistically significant associations are
reported between fine particles and mortality in the ACS and Six Cities analyses, inconsistent
but generally positive associations with PM were reported in the AHSMOG analyses, and
distinctly inconsistent results were reported in the VA study. Based on several factors, the larger
study population in the ACS study, the larger air quality data set in the Six Cities study, the more
generally representative study populations used in the Six Cities and ACS studies, and the fact
that these studies have undergone extensive reanalyses - the greatest weight should be placed
on the results of the ACS and Six Cities cohort studies in assessing relationships between
8-120
-------
long-term PM exposure and mortality. The results of these studies, including the reanalyses
results for the Six Cities and ACS studies and the results of the ACS study extension, provide
substantial evidence for positive associations between long-term ambient PM (especially fine
PM) exposure and mortality.
8.2.3.2.6 The S-Plus GAM Convergence Problem and Cohort Studies
The long-term pollution-mortality study results discussed above in this section were
unaffected by the GAM default convergence issue reported by Dominici et al. (2002) and
discussed earlier in this chapter, because they did not use such a model specification. Instead,
the cohort studies of long-term PM exposures used Cox Proportional Hazards models. For
example, in the recent Pope et al. study (2002), the baseline models were random effects Cox
Proportional Hazards models without the inclusion of nonparametric smooths. However, Pope
et al. (2002) did include a nonparametric spatial smooth in the model as part of a more extended
sensitivity analysis to evaluate more aggressive control of spatial differences in mortality. They
found that the estimated pollution-mortality effects were not sensitive to this additional spatial
control, so final reported results did not include the smooth; and this study's results, like those
from other cohort studies discussed above, were unaffected by the S-Plus convergence issue.
8.2.3.3 Studies by Particulate Matter Size-Fraction and Composition
8.2.3.3.1 Six Cities, ACS, AHSMOG and VA Study Results
Ambient PM consists of mixtures that may vary in composition over time and from place
to place. This should logically affect the relative toxicity of PM indexed by mass at different
times or locations. Some semi-individual chronic exposure studies have investigated relative
roles of various PM components in contributing to observed air pollution associations with
mortality. However, only a limited number of the chronic exposure studies have included direct
measurements of chemical-specific constituents of the PM mixes indexed by mass measurements
used in their analyses.
As shown in Table 8-13, the Harvard Six-Cities Study (Dockery et al., 1993) results
indicated that the PM2 5 and SO42 RR associations (as indicated by their respective 95% CIs and
8-121
-------
TABLE 8-13. COMPARISON OF ESTIMATED RELATIVE RISKS FOR
ALL-CAUSE MORTALITY IN SIX U.S. CITIES ASSOCIATED WITH
THE REPORTED INTERCITY RANGE OF CONCENTRATIONS OF
VARIOUS PARTICULATE MATTER METRICS
Concentration Range
PM Species (jig/m3)
so42-
PM2 5 - SO4 2-
PM25
PM15.2.5
TSP-PM15
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).
t-statistics) were more consistent than those for the coarser mass components. Further, the
effects of sulfate and non-sulfate PM2 5 are quite similar. Acid aerosol (FT) exposure was also
considered by Dockery et al. (1993), but only less than one year of measurements collected near
the end of the follow-up period were available in most cities; consequently, the Six-Cities results
were much less conclusive for the acidic component of PM than for the other PM metrics
measured over many years during the study.
Table 8-14 presents comparative PM25 and SO42 results from the ACS study, indicating
that both had substantial, statistically significant effects on all-cause and cardiopulmonary
mortality. On the other hand, the RR for lung cancer was notably larger (and substantially
more significant) for SO42 than PM2 5 (not significant).
The most recent AHSMOG analyses also considered SO42 as a PM index for all health
outcomes studied except lung cancer, but SO42 was not as strongly associated as PM10 with
mortality (as shown in Table 8-8) and was not statistically significant for any mortality category.
Also, extensive results from the VA study were reported in Lipfert et al. (2000b) for
various components: TSP, PM10, PM25, PM15.25, PM15, SO42 . There were no significant positive
effects for any exposure period concurrent or preceding any of the segmented mortality periods
for any PM component, unlike for O3. On the other hand, the SO42 levels during the 1979 to
8-122
-------
TABLE 8-14. COMPARISON OF REPORTED SO/' AND PM2 5 RELATIVE RISKS
FOR VARIOUS MORTALITY CAUSES IN THE AMERICAN CANCER
SOCIETY (ACS) STUDY
Mortality Cause
All Cause
Cardiopulmonary
Lung Cancer
SO42
(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).
1981 and 1982 to 1984 exposure periods were significantly associated with deaths aggregated
across all the segmented follow-up mortality periods (1976 to 1996). The first exposure period
was associated with 4.9% increase and the latter with a 6.7% increase in the aggregated ("single
period" in Lipfert et al 2000b terminology) total, nonaccidental mortality risk per 10 |ig/m3 SO42
increment.
Harvard Six Cities, ACS, and AHSMOG study results are compared in Table 8-15 (total
mortality) and Table 8-16 (cause-specific mortality). Results for the VA study are not shown in
Tables 8-15 and 8-16 as the VA cohort is all male and largely consists of current or former
smokers (81%) and is thusly not comparable to the total or male nonsmoker populations of the
other studies. Also, results for females are not presented, as the overall effects were driven
largely by males (female associations generally being statistically nonsignificant).
Estimates for Six Cities parameters were calculated in two ways: (1) mortality RR for the
most versus least polluted city in Table 3 of Dockery et al. (1993), adjusted to standard
increments; and (2) ecological regression fits in Table 12-18 of U.S. Environmental Protection
Agency (1996a). The Six Cities study of eastern and mid-western U.S. cities suggests a strong
and highly significant relationship for fine particles and sulfates, a slightly weaker but still
highly significant relationship to PM10, and a marginal relationship to PM10_2 5. The ACS study
looked at a broader spatial representation of cities, and found a stronger statistically significant
8-123
-------
TABLE 8-15. COMPARISON OF TOTAL MORTALITY RELATIVE RISK
ESTIMATES AND t-STATISTICS FOR PARTICULATE MATTER COMPONENTS
IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM10 (50 ug/m3)
PM2 5 (25 ug/m3)
S042- (15 ug/m3)
Days/year 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
1.28 a
1.24
1.36a; 1.38b
1.21 a
1.17
1.25
1.50 a; 1.57b
1.35
1.11
1.1
1.28
1.08
1.81 a; 1.56b
1.43 a
t Statistic
2.94 a; 3.27 b
0.81 a
1.61
2.94 a; 3.73 b
0.81 a
4.35
1.96
2.94 a; 3.67 b
0.81 a
5.11
1.59
0.96
2.18
2.94 a'c 1.81 b
0.81 a
aMethod 1 compares Portage versus Steubenville (Table 3, Dockery et al., 1993).
b Method 2 is based on ecologic regression models (Table 12-18, U.S. Environmental Protection Agency, 1996a).
c Method 1 not recommended for PM10_2 5 analysis, due to high concentration in Topeka.
relationship to PM2 5 than to sulfate (no other pollutants were examined). The AHSMOG study
at California sites (where sulfate levels are typically low) found significant effects in males
for PM10 100 |ig/m3 exceedances and a marginal effect of mean PM10, but no PM effects for
females or with sulfates. On balance, the overall results shown in Tables 8-15 and 8-16 suggest
statistically significant relationships between long-term exposures to PM10, PM2 5, and/or sulfates
and excess total and cause-specific cardiopulmonary mortality.
8-124
-------
TABLE 8-16. COMPARISON OF CARDIOPULMONARY MORTALITY RELATIVE
RISK ESTIMATES AND t-STATISTICS FOR PARTICULATE MATTER
COMPONENTS IN THREE PROSPECTIVE COHORT STUDIES
PM Index
PM10 (50 ug/m3)
PM2 5 (25 ug/m3)
S042- (15 ug/m3)
Days/year 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
MaleNon-CRCc
All
All
Male
Male Nonsmoker
All
All
Male
Male Nonsmoker
Male Nonsmoker
MaleNon.-CRCc
Male Nonsmoker
MaleNon.-CRCc
All
Relative Risk
1.744a
1.219
1.537
1.527a
1.317
1.245
1.245
1.743 a
1.19
1.147
1.205
1.279
1.219
1.082
1.188
2.251 a
t Statistic
2.94 a
1.12
2.369
2.94 a
4.699
3.061
1.466
2.94 a
5.47
3.412
2.233
0.072
0.357
1.31
2.37
294 a,b
aMethod 1 compares Portage versus Steubenville (Table 3, Dockery et al., 1993).
b Method 1 not recommended for PM10_2 s analysis due to high concentration in Topeka.
cMale non. - CRC = AHSMOG subjects who died of any contributing nonmalignant respiratory cause.
The prospective cohort long-term PM exposure studies conducted to date collectively
appear to confirm earlier cross-sectional study indications that the fine mass component of PM10
(and usually especially its sulfate constituent) are more strongly correlated with mortality than is
the coarse PM10_25 component. However, the greater precision of PM25 population exposure
measurement (both analytical and spatial) relative to PM10_2 5 makes conclusions regarding their
8-125
-------
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 the use of prospective cohort studies in assessing the long-term exposure effects
of particles and gases is preferred, additional useful information may be derived from ecological
studies. In particular, repeated cross-sectional studies may provide another tool for examining
changes in air-pollution-attributable mortality over time. Lipfert and Morris (2002) carried out
cross-sectional regressions for five time periods using data on mortality, air pollution including
various measures of PM, O3, NO2i SO2, CO, climate, and sociodemographic factors using
county-level data. Data were available for TSP and gaseous co-pollutants as far back as 1960
and for PM2 5, PM15, and SO42 from the inhalable particulate network (IPN). The authors
investigated longitudinal and spatial relations between air pollution and age-specific mortality
using 3- to 5-year subsets of data from 1960 to the end of 1997, with the addition of PM25 data
from 1999.
One of the key features of this study is the presentation of attributable risk estimates for
different age groups across varying time periods. It is important to note that cross-sectional
studies such as this one do not directly investigate temporality or latency of effects. One or more
years of pollution measurements are used as estimates of long-term pollution concentrations for
the communities. Lipfert and Morris (2002) note that PM2 5 data from the two available time
periods (1979 to 1984 and 1999) were well correlated (r = 0.71)2 and, in fact, used the 1999 data
for "back-extrapolation" of data for some of the counties where data were not available in 1979
to 19843. The different time periods may provide data that more or less adequately represent
long-term PM concentrations, and it is more important that the measurements reflect long-term
trends than that the PM concentrations predate the mortality data by any specific time period.
2 Pope et al. (2002) also reported a strong correlation (r = 0.78) between PM2 5 concentrations averaged over 1979 to
1983 and 1999 to 2000 in the extended analysis for the ACS cohort.
3 It appears that the extrapolated data were used in the main analyses. EPA observes that it was difficult to interpret
the methodological discussions in the paper and obtained further information via personal communication with the
author.
8-126
-------
While well-motivated, the use of multiple mortality and pollution time periods clearly reduces
the power of any individual analysis.
The counties included in any given analysis varied by time period based on available data.
Lipfert and Morris (2002) stated that "the number of counties with valid air quality data vary
substantially...by pollutant and over time" and that they attempted to differentiate "among many
air quality variables that differ according to species, timing, and in some cases measurement
technology." They stated, "Environmental monitoring coverage increased substantially during
this period [1970 to 1974] (1,258 counties had TSP data) and air quality began to improve in
major cities in response to emission controls and use of clean fuels." There have also been
changes in the measurement of PM (e.g., TSP, PM10 and PM25) over time as well as changes in
the location of monitors as they were sited for different purposes.
An interesting conclusion drawn by Lipfert and Morris (2002) is that pollution-mortality
relationships vary across age groups, with stronger effects among younger age groups. It is
important, however, to note that the results of this study are presented as attributable risks; with
attributable risk defined as the mortality risk based on the mean concentration of the pollutant
and the mean mortality rate. A problem with attributable risk arises when one compares the
risks of different age groups. In Table 8-17, it can be seen that among the young a higher
percentage of mortality is reduced by reducing air pollution, as reflected in the higher
attributable risks. However, with the more standard presentation of risk per 10 |ig/m3 change in
PM2 5, the risk increases for older age groups.
Lipfert and Morris (2002) also reported that they generally found that risk estimates were
highest for pollution estimates in the earlier time periods and decreased in analyses using more
recent pollution and mortality data. This can be seen in Table 8-17, where statistically
significant associations are reported for all but the youngest age group using 1989 to 1991
mortality data, but with 1995 to 1997 data the associations were smaller and only significant for
one age group. The same pattern can be seen for TSP, PM10 and PM2 5 as summarized in
Figure 7 in the authors' report (Lipfert and Morris, 2002). The authors concluded that mortality
responses to air pollution have been decreasing over time for PM and several other pollutants.
8-127
-------
TABLE 8-17. PERCENT ATTRIBUTABLE RISK OF MORTALITY (from Lipfert
and Morris, 2000) AND RISK ESTIMATES CALCULATED PER 10 jig/m3 PM2 5.
SELECTED TIME PERIODS FOR MORTALITY DATA OF 1989-91 AND 1995-97
WITH PM2 5 DATA FROM THE IP NETWORK DATA (1979-84) FOR COUNTIES
IN THE U.S. WITH IP NETWORK MONITORS.
1989-91 Mortality Data
Lipfert and Morris
Results
Age
15-44
45-64
65-74
75-84
> 85
Attributabl
e RiskA
5.2
7.9*
3.7*
2.0*
2.1*
SE
4.9
2.4
1.2
1
0.8
EPA Estimates
(per 100,000)
Mortality
Rate
70
778
2643
5943
15145
Risk
Estimate
4
62
98
119
316
1995-97 Mortality Data
Lipfert and Morris
Results
Attributabl
eRiskA
4.2
5.4*
1.1
1.5
1.6
SE
3.6
2.7
1.4
1.9
1.4
EPA Estimates
(per 100,000)
Mortality
Rate
74
701
2577
5885
15795
Risk
Estimate8
3
38
27
88
246
*P < 0.05
A adapted from Lipfert and Morris (2002) Tables 7 and 8; converted to %attributable risk per 10 ug/m3.
B risk estimate = (coefficient)(10 ug/m3 PM2 5) per 100,000 over a 3-year period; where coefficient = [(AR)
(mean mortality rate)]/(mean PM2 5 concentration). Mortality rates per 100,000 over a 3-year period from
Table 1 and mean PM2 5 concentration (19.19 ug/m3) from Table 2 of Lipfert and Morris (2002).
In any cross-sectional study, exposure misclassification and confounding are important
issues to consider in interpreting the results. For all studies discussed in this section, exposure is
characterized by pollution concentration averaged for a given geographic region. In cohort
studies, some information is generally available on participants' residence histories but in cross-
sectional studies the exposure level is assigned based on the subjects' location according to the
death record. Between the years 1995 to 2000, approximately 20% of people aged 5 to 64 years
moved to a different state. Older adults were slightly less mobile as a group, with 18.8% of
those 65 and older moving to a different state; the rate was 21.2% for adults aged 65 to 74 and
declined with greater age, though there was some evidence of return migration at advanced ages
of 85 years and older (U.S. Census Bureau, 2003). To address potential confounding, the
authors used county-lev el data on a variety of risk factors, using stepwise regression methods to
8-128
-------
select the best-fitting model, then apparently used the residuals from these models to evaluate
relationships with air pollution concentrations. While an impressive list of variables were
included in these analyses, it must be noted that there can be considerable variation in
socioeconomic or personal risk factors across areas within a given county as well as from county
to county.
The inhalable particle network data used in Lipfert and Morris (2002) is basically the same
air quality data set used in analyses for the ACS and VA study cohorts. A major distinction, of
course, is that individual health data were used in the cohort studies, but only county-level data
for the cross-sectional study. Lipfert and Morris (2002) noted reasonable agreement with Pope
et al. (1995) and Dockery et al. (1993), but observed that the VA study analyses (Lipfert et al.,
2000b) "found apparent beneficial effects of PM25...whereas the present study does not." While
subject to the limitations of cross-sectional analyses, the Lipfert and Morris study reports
associations between fine particles and mortality that are generally similar to those from the
large cohort studies, although it is difficult to compare the quantitative results across studies.
Interpreting the results of the many subset analyses conducted is difficult, but the results of
analyses across time periods would appear to indicate reduced mortality risk with pollution from
more recent time periods.
8.2.3.3.3 Mortality and Chronic Exposure to Traffic-Related Ambient PM
Hoek et al. (2002): Traffic and Mortality in the Netherlands
Hoek et al. (2002) assessed the relationship between traffic-related air pollution and
mortality among participants of the Netherlands Cohort Study on Diet and Cancer (NLCS), an
ongoing study. They investigated a random sample of 5000 middle-aged people (aged 55 to
69 years) from the full cohort of the NLCS study during 1986 to 1994. Long-term exposures to
traffic-related air pollutants were estimated using participants' 1986 home address. In a novel
attempt to take into account within-city variation of air pollution concentrations by quantifying
small-scale spatial variations in air pollution concentrations, the home addresses were geocoded
with a geographic information system (GIS). Long-term average exposure to outdoor air
pollution was calculated as a function of an estimation method that consisted of interpolation of
8-129
-------
regional background stations, the estimation of urban background increases for BS and NO2, and
the characterization of local contributions, with indicator variables also indicating whether a
subject lived within 100 m of a freeway or 50 m of a major city road.
Associations between exposure to air pollution and (cause-specific) mortality were
assessed with Cox's proportional hazards models, with adjustment for potential confounders.
Cardiopulmonary mortality was reported to be associated with living near a major road (relative
risk 1.95, CI: 1.09, 3.52). The relative risk for living near a major road was 1.41 (0.94 , 2.12) for
total deaths. The authors considered the potential role of residual confounding factors, finding
that the unadjusted effects estimates were consistently similar to the effects after adjustment for
confounders, and concluding that residual confounding was very unlikely to account for the
association between living near a major road and mortality. The authors concluded that long-
term exposure to traffic-related air pollution may shorten life expectancy. Hoek et al. (2002)
also noted that living near a major road might also include factors other than air pollution as a
factor for mortality.
The use of estimated concentrations for pollutant levels (as contrasted to quantitatively
measured mass levels such as PM10 or PM2 5) indicates that the Hoek et al. (2002) study was a
hypothesis-generating study, the main value of which is more at providing insights into potential
areas for well-designed hypothesis-testing studies that include more quantitative measurements
of PM10 and/or PM25. Without adequate actual measurements, estimates of exposure are
uncertain. Neither BS nor NO2 measures were likely to have provided adequate quantitative data
for the estimation of PM10 or PM25. For example, Hoek et al. (2001) observed that the contrast
in PM10 concentrations across in the Netherlands is relatively small and that BS concentrations
are not a good proxy for PM10 or PM2 5.
This long-term study is unique in that it attempted to examine within-metropolitan-area
small-scale variations in exposures. However, given that exposure estimates were characterized
by interpolation (based on the measured regional and urban background concentration, as well as
using an indicator variable for living near major roads, during only one year at the beginning of
the study period), much caution is warranted in viewing the study's key results as noted above.
Nevertheless, the overall approach used appears to be a promising one, to the extent that future
8-130
-------
analogous studies employ more refined interpolation of exposure estimates based on more
extensive measured data obtained over longer periods.
8.2.3.4 PM-Mortality Intervention Studies
Although many studies have reported short-term associations between PM indices and
mortality, a largely unaddressed question remains as to the extent to which reductions in ambient
air PM actually lead to reductions in deaths attributable to PM. This question is not only
important in terms of "accountability" from the regulatory point of view, but it is also a scientific
question that challenges the predictive validity of statistical models and their underlying
assumptions used thus far to estimate excess mortality due to ambient PM.
The opportunities to address this question are rare. However, at the time of the 1996
PM AQCD, one situation presented a good opportunity for a PM intervention study—that being
the Utah Valley situation evaluated by Pope. In the Pope (1989) analysis of PM10 and children's
hospital admissions in Utah Valley, the study period contained the 13-month steel mill
closure mentioned earlier (during which time PM10 concentrations averaged 35 |ig/m3 versus
50 |ig/m3 when the mill was opened). Analyses of children's respiratory admissions in Utah
Valley before and after the steel mill closure provided evidence of decreased morbidity resulting
from the lower PM10 concentrations during the mill closure.
Two more recent mortality intervention studies have examined: (1) the impact of a ban on
coal sale in Dublin, Ireland (Clancy et al., 2002); and (2) the impact of a regulation to use fuel
oil with low sulfur content in Hong Kong (Hedley et al., 2002). These regulations were enforced
across very short time frames and, as such, they provided opportunities to observe any change in
mortality rate before and after the intervention.
Clancy et al. (2002) examined the impact of the ban on coal sales that took place in
September 1990 in the city of Dublin, Ireland. They assessed the ban's impact on mortality by
conducting Poisson regression of the standardized mortality rate during 72 months before and
after the ban on coal sales (13 years total study period), adjusting for temperature on the same
day and previous days, mean relative humidity and previous days, day-of-week, respiratory
epidemics, and directly standardized deaths rates in the rest of Ireland. The impact of the ban
8-131
-------
was estimated by an indicator variable of the post-ban period. They also reported means for
black smoke (BS), SO2, temperature and relative humidity before and after the ban by season,
as well as age-standardized deaths rates before and after the ban by seasons. A substantial
reduction (35.6 |ig/m3 reduction, or 70% for all seasons) in BS, especially for winter season
(63.8 |ig/m3 reduction) was observed. The reduction for SO2 was less (34% reduction). The
post-ban means of age-standardized mortality rates were significantly lower for total
(nonaccidental), cardiovascular, and respiratory categories for all seasons combined and
especially for the winter season. In contrast, the mean of the other mortality categories slightly
increased for spring and fall (but decreased for summer). The Poisson regression results with
adjustments for time-varying covariates showed statistically significant (p < 0.05) reductions in
age-standardized mortality rate for total (-5.7% [-7.2, -4.1]), cardiovascular (-10.3% [-12.6,
-8.0]), and respiratory (-15.5% [-19.1, - 11.6]) mortality, but not mortality for other causes
(1.7% [-0.7, 4.2]). The results without adjustments for other time-varying covariates showed
larger reductions.
Clancy et al. compared their mortality reduction estimates to the projected reduction based
on APHEA 1 study (Katsouyanni et al., 1997) results. They noted that the BS mortality
regression coefficient from APHEA 1 results would have translated to only a 2.1% reduction in
total deaths had they been applied to the Dublin data where a BS concentration reduction of
35.6 |ig/m3 was observed, compared to a 5.7% decrease that Clancy and colleagues estimated for
the intervention period in their analysis. They also noted that the actual reduction (-3.2% when
the PM10 average was 15 |ig/m3 lower than the period when the mill was operating) in average
deaths during the steel mill closure in Utah Valley, as noted by Pope et al. (1992), would have
translated to 8.0% had it been applied to the BS reduction in the Dublin data (assuming that BS
approximately = PM10) — about the same as their unadjusted estimate (8.0%). It should be
noted, however, that the reduction estimate in Clancy et al.'s study is the "average" reduction
comparing the two 6-year periods before and after the ban of coal sales. In contrast, most time-
series studies, including APHEA, estimate excess mortality risk in response to a short-term
change, usually for mortality on a single day or a few days. As discussed in Section 8.4.5, there
is some suggestive evidence that risk estimates based on a single- or a few-day exposures may
8-132
-------
underestimate the possible multiday effects. The apparent lack of the evidence for "harvesting"
(see Section 8.4.9.1) further suggests that the excess risk (or reduction) estimates based on the
prevailing time-series study design may not predict longer-term effects. Therefore, comparisons
of estimates of reduction in mortality due to interventions and predicted reductions based on
results of time-series studies are not straightforward; and it may not be surprising that Clancy
et al.'s estimate of mortality reduction was larger than predicted based on PM coefficients
derived from most time-series studies. Clancy et al.'s study nevertheless provides suggestive
evidence that a substantial reduction in PM leads to a significant reduction in mortality.
Hedley et al. (2002) assessed the impact on mortality rate of the restriction on use of low
sulfur (not more than 0.5%) fuel oil implemented in July 1990 in Hong Kong. Changes in trends
in deaths were estimated using Poisson regression of monthly mortality rate between 1985 and
1995, adjusting for trends, seasonal cycles (by sine/cosine terms), temperature, and relative
humidity, with stratification by the two five-year pre- and post-intervention periods. They also
estimated a measure of warm to cool season change in death rates relative to the mean by fitting
monthly deaths as a function of sine and cosine terms for each of the five years after the
intervention and by cause (total, respiratory, cardiovascular, neoplasms, and others) and by age
groups (all ages, age 15 to 64, age 65 years and older). Interestingly, although SO2 did decrease
substantially (-50%), PM10 levels did not change at all after the intervention. Even sulfate
levels, while reported to be lower by -20% for the first 2 years after the intervention, were
unchanged five years after the intervention, apparently due to regional influences. Ozone
showed an increasing trend during the study period. The seasonal mortality analysis results
show that the apparent reduction in seasonal death rate occurred only during the first winter, and
this was followed by a rebound (i.e., higher than expected) in the following winter. This pattern
was seen for total, respiratory, and cardiovascular categories. Based on the Poisson regression
of the monthly mortality data analysis, the average annual trend in death rate significantly
declined after the intervention for all cause (2.1%), respiratory (3.9%), and cardiovascular causes
(2.0%). Hedley et al. also estimated expected average gain in life expectancy per year due to the
lower SO2 level to be 20 days for females and 41 days for males.
8-133
-------
Interpreting Hedley et al.'s results is somewhat complicated by an upward trend noted by
them in mortality across the intervention point, due to increased population size and aging. The
results suggest that such an upward trend is less steep after the introduction of low sulfur fuel.
While their Poisson regression model of monthly deaths does adjust for trend and seasonal
cycles, the regression model does not specifically address the influence of influenza epidemics.
Since the magnitude of influenza epidemics can change from year to year, the included
sine/cosine terms will not necessarily fit the year-to-year variation. This issue also applies to the
analysis of warm to cool season change in death rates. The most prominent feature of the time-
series plot (or the fitted annual cycle of monthly deaths) presented in Hedley et al.'s paper is the
lack of a winter peak for respiratory and all cause mortality during the year immediately
following the intervention. Much could be made out of this lack of a winter peak, but no
discussion of potential impact of (or a lack of) influenza epidemics is provided. These issues
make the interpretation of the estimated decline in upward trend of mortality rate or the apparent
lack of winter peak difficult. In any case, since the intervention did not result in the reduction of
PM (PM10 in this case), this study does not provide direct information on the impact of PM
intervention.
The Clancy et al. and Hedley et al. studies share a similar situation in which regulations
caused a sudden reduction in PM and/or SO2. Both studies estimated reductions in mortality rate
before and after an intervention (6-year periods in Clancy et al. study, and 5-year periods in
Hedley et al. study). Both studies attempted to adjust for unmeasured secular changes in social
or other variables that can affect the trend in mortality rate by direct standardization or in the
regression models. The challenge of these analyses is that, unlike regular time-series mortality
analyses in which only the associations in short-term fluctuations are estimated by filtering out
the longer-wave fluctuations, the parameter that is being estimated is in the longer-wave length
where effective sample size of "events" can be small. For example, the number of influenza
epidemics in these data is "small", and yet their magnitude can vary substantially from year to
year, making their influence on the average statistics of long-wave events possibly large.
Furthermore, because the regular short-term daily time-series studies specifically filter out these
long-wave events, it may not be appropriate to directly compare projected risk reductions based
8-134
-------
on PM risk coefficients derived from the daily time-series studies with estimated mortality
reductions based on these intervention studies. Clearly, there is much uncertainty between
mortality risk estimates derived from daily time-series studies versus those derived from cohort
studies (that may be capturing the very long-term effects). The intervention studies appear to
capture the risk (reduction) in a time scale that is in between these two types of studies.
In summary, the Clancy et al. (2002) intervention study suggests evidence of mortality
reduction in response to reduced levels of PM, whereas Hedley et al.'s intervention study
presents an unusual case, where SO2 levels declined substantially (but PM levels did not) and
the SO2 decrease was paralleled by mortality decrements. As such, these specific intervention
studies are valuable in drawing qualitative conclusions that imply likely causal relationships
underlying the observed mortality decrements occurring in concert with declines in ambient PM
and/or SO2 levels.
8.2.3.5 Salient Points Derived from Analyses of Chronic Particulate Matter Exposure
Mortality Effects
A review of the studies summarized in the previous PM AQCD (U.S. Environmental
Protection Agency, 1996a) indicates that past epidemiologic studies of chronic PM exposures
collectively indicated that increases in mortality are associated with long-term exposure to
airborne particles of ambient origins. The PM effect size estimates for total mortality from these
studies also indicate that a substantial portion of these deaths reflected cumulative PM effects
above and beyond those exerted by acute exposure events.
The HEI-sponsored reanalyses of the ACS and Harvard Six-Cities studies (Krewski et al.,
2000) "replicated the original results, and tested those results against alternative risk models and
analytic approaches without substantively altering the original findings of an association
between indicators of particulate matter air pollution and mortality." Several questions,
including the questions (1 to 4) posed at the outset of this Section (8.2.3) were investigated by
the Krewski et al. (2000) sensitivity analyses for the Six City and ACS studies data sets.
Key results emerging from the HEI reanalyses and other new chronic PM mortality studies are
as follows:
8-135
-------
(1) A much larger number of confounding variables and effects modifiers were considered
in the Reanalysis Study than in the original Six City and ACS studies. The only significant air
pollutant other than PM2 5 and SO4 in the ACS study was SO2, which greatly decreased the PM2 5
and sulfate effects when included as a co-pollutant (Krewski et al., 2000, Part II, Tables 34-38).
A similar reduction in particle effects occurred in any multipollutant model with SO2. The most
important new effects modifier was education. The AHSMOG study also suggested that other
metrics for air pollution, and other personal covariates such as time spent outdoors and
consumption of anti-oxidant vitamins, might be useful. Both individual-level covariates and
ecological-level covariates shown in (Krewski et al., 2000, Part II, Table 33) were evaluated,
including whether or not the observations are independent or spatially correlated.
(2) Specific attribution of excess long-term mortality to any specific particle component or
gaseous pollutant was refined in the reanalysis of the ACS study. Both PM25 and sulfate were
significantly associated with excess total mortality and cardiopulmonary mortality and to about
the same extent whether the air pollution data were mean or median long-term concentrations or
whether based on original investigator or Reanalysis Team data. The association of mortality
with PM15 was much smaller, though still significant; and the associations with the coarse
fraction (PM15.2 5) or TSP were even smaller and not significant. The lung cancer effect was
significant only for sulfate with the original investigator data or for new investigators with
regional sulfate artifact adjustment for the 1980 to 1981 data (Krewski et al., 2000, Part II,
Table 31). Associations of mortality with long-term mean concentrations of criteria gaseous
co-pollutants were generally nonsignificant except for SO2 (Krewski et al., 2000, Part II,
Tables 32, 34-38), which was highly significant, and for cardiopulmonary disease with warm-
season ozone. However, the regional association of SO2 with SO4 and SO2 with PM2 5 was very
high; and the effects of the separate pollutants could not be distinguished. Krewski et al. (2000,
p. 234) concluded that, "Collectively, our reanalyses suggest that mortality may be associated
with more than one component of the complex mix of ambient air pollutants in urban areas of
the United States." In the most recent extension of the ACS study, Pope et al. (2002) confirmed
the strong association with SO2 but found little evidence of effects for long-term exposures to
other gaseous pollutants.
8-136
-------
(3) The extensive temporal data on air pollution concentrations over time in the Six City
Study allowed the Reanalysis Team to evaluate time scales for mortality for long-term exposure
to a much greater extent than was reported in Dockery et al. (1993). The first approach was to
estimate the log-hazard ratio as a function of follow up time using a flexible spline-function
model (Krewski et al., 2000, Part II, Figures 2 and 3). The results for both SO42 and PM2 5
suggest very similar relationships, with larger risk after initial exposure decreasing to 0 after
about 4 or 5 years, and a large increase in risk at about 10 years follow-up time.
The analyses of the ACS Study proceeded somewhat differently, with less temporal data
but many more cities. Flexible spline regression models for PM2 5 and sulfate as function of
estimated cumulative exposure (not defined) were very nonlinear and showed quite different
relationships (Krewski et al., 2000, Part II, Figures 10 and 11). The PM25 relationship shows the
mortality log-hazard ratio increasing up to -15 |ig/m3 and relatively flat above -22 |ig/m3, then
increasing again. The sulfate relationship is almost piecewise linear, with a low near- zero slope
below -11 |ig/m3 and a steep increase above that concentration.
A third approach evaluated several time-dependent PM2 5 exposure indicators in the
Six City Study: (a) constant (at the mean) over the entire follow-up period; (b) annual mean
within each of the 13 years of the study; (c) city-specific mean concentration for the earliest
years of the study (i.e., very long-term effect); (d) exposure estimate in 2 years preceding death;
(e) exposure estimate in 3 to 5 years preceding death; and (f) exposure estimate > 5 years
preceding death. The time-dependent estimates (a-e) for mortality risk are generally similar and
statistically significant (Krewski et al., 2000, Part II, Table 53), with RR of 1.14 to 1.19 per
24.5 |ig/m3 being much lower than the risk of 1.31 estimated for exposure at the constant mean
for the period. Thus, it is highly likely the duration and time patterns of long-term exposure
affect the risk of mortality; and further study of this question (along with that of mortality
displacement from short-term exposures) would improve estimates of life-years lost from PM
exposure.
(4) The Reanalysis Study also advanced our understanding of the shape of the relationship
between mortality and PM. Again using flexible spline modeling, Krewski et al. (2000, Part II,
8-137
-------
Figure 6) found a visually near-linear relationship between all-cause and cardiopulmonary
mortality residuals and mean sulfate concentrations, near-linear between cardiopulmonary
mortality and mean PM2 5, but a somewhat nonlinear relationship between all-cause mortality
residuals and mean PM2 5 concentrations that flattens above -20 |ig/m3. The confidence bands
around the fitted curves are very wide, however, neither requiring a linear relationship nor
precluding a nonlinear relationship if suggested by reanalyses. An investigation of the mortality
relationship for other indicators may be useful in identifying a threshold, if one exists, for
chronic PM exposures.
(5) With regard to the role of various PM constituents in the PM-mortality association,
past cross-sectional studies have generally found the fine particle component, as indicated either
by PM2 5 or sulfates, to be the PM constituent most consistently associated with mortality. While
relative measurement errors of various PM indicators must be further evaluated as a possible
source of bias in these estimate comparisons, the Six-Cities and AHSMOG prospective cohort
studies both indicate that the fine mass components of PM are more strongly associated with
mortality effects of chronic PM exposure than are coarse fraction indicators.
(6) The spatial regression methods suggested that part of the relation between sulfate and
mortality was probably due to some unobserved variable or group of confounding variables.
In particular, they found that the sulfate-associated effect drops from a relative risk of 1.25 with
the Independent Cities Model to 1.19 with the Regional Adjustment Model, but all models
continued to show an association between elevated risks of mortality and exposure to airborne
sulfate.
(7) The newly available (2002) ACS study extension more than doubles the original
follow-up period (now to 16 versus 7 years originally); and it both (a) confirms the original ACS
study findings of significant associations between long-term PM25 exposures and increased
cardiopulmonary mortality risks and (b) provides the strongest evidence to date for increased
lung cancer risk associations with ambient fine particles measured as PM2 5.
8-138
-------
8.3 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE
The effects of ambient PM on morbidity endpoints are assessed below in subsections
focused on: (a) effects of acute ambient PM exposure on cardiovascular morbidity; (b) effects of
short-term PM exposure on the incidence of respiratory and other medical visits and hospital
admissions; and (c) short- and long-term PM exposure effects on lung function and respiratory
symptoms in asthmatics and nonasthmatics.
8.3.1 Cardiovascular Morbidity Effects Associated with Acute Ambient
Particulate Matter Exposure
8.3.1.1 Introduction
Very little information specifically addressing cardiovascular morbidity effects of acute
PM exposure existed at the time of the 1996 PM AQCD. Since then, a significantly expanded
body of literature has emerged, both on the ecologic relationship between ambient particles and
cardiovascular hospital admissions and associations of PM exposures with changes in various
physiological and/or biochemical measures. The latter studies are particularly important in that
they are suggestive of possible mechanisms underlying PM cardiovascular effects. However, it
should be noted that the mechanistic interpretation of the cardiovascular physiology results
observed to date (some of which are conflicting) remain unclear, as discussed in more detail in
Chapter 7.
This section begins with a brief summary of key findings from the 1996 PM AQCD on
acute cardiovascular effects of PM. Next, key new studies are reviewed in the two categories
noted above, i.e., ecologic time-series studies and individual-level studies of physiological
measures of cardiac function and/or biochemical measures in blood as they relate to ambient
pollution. This is followed by discussion of several issues of importance for interpreting the
available data, including identification of potentially susceptible subpopulations, roles of
environmental co-factors such as weather and other air pollutants, temporal lags in the
relationship between exposure and outcome, and the relative importance of various size-
classified PM components (e.g., PM2 5, PM10, PM10_2 5).
8-139
-------
8.3.1.2 Summary of Key Findings on Cardiovascular Morbidity from the 1996
Particulate Matter Air Quality Criteria Document
Just two studies were available for review in the 1996 PM AQCD that provided results for
acute cardiovascular (CVD) morbidity outcomes (Schwartz and Morris, 1995; Burnett et al.,
1995). Both studies were of ecologic time-series design and used standard statistical methods.
Analyzing four years of data on the > 65 year old Medicare population in Detroit, MI, Schwartz
and Morris (1995) reported significant associations between PM10 and ischemic heart disease
admissions, controlling for environmental covariates. Based on analysis of admissions data from
168 hospitals throughout Ontario, Canada, Burnett et al. (1995) reported significant associations
between fine particle sulfate concentrations (as well as other air pollutants) and daily
cardiovascular admissions. The relative risk due to sulfate particles was slightly larger for
respiratory than for cardiovascular hospital admissions. The 1996 PM AQCD concluded on the
basis of these studies that: "There is a suggestion of a relationship to heart disease, but the
results are based on only two studies, and the estimated effects are smaller than those for other
endpoints" (U.S. Environmental Protection Agency, 1996a, p. 12-100). The PM AQCD also
stated that acute effects on CVD admissions had been demonstrated for elderly populations
(i.e., > 65 years), but that insufficient data existed to assess relative effects on younger
populations.
When viewed alongside the more extensive literature on acute CVD mortality that was
available at the time, the evidence from ecologic time-series studies reviewed in the 1996
PM AQCD was consistent with acute health risks of PM being larger for cardiovascular and
respiratory causes than for other causes. Given the tendency for end-stage disease states to
include both respiratory and cardiovascular impairment, and the associated diagnostic overlap
that often exists, it was not possible on the basis of these studies alone to determine which of the
two organ systems, if either, was more critically affected.
8.3.1.3 New Particulate Matter-Cardiovascular Morbidity Studies
8.3.1.3.1 Acute Hospital Admission Studies
Salient methodological features and results of numerous newly available studies that
examine associations between daily measures of ambient PM and daily hospital admissions for
8-140
-------
cardiovascular disease are summarized in Table 8B-1 (see Appendix 8B). As discussed earlier
in Sections 8.1.4 and 8.2.2, many studies published since 1995 used GAM with default
convergence criteria. Several of those studies have been reanalyzed by original investigators
using GAM with more stringent convergence criteria and GLM with parametric smooths, such as
natural splines (NS) or penalized splines (PN). Again, since the extent of possible bias in PM
effect-size estimates caused by the default criteria setting in the GAM models is difficult to
estimate for individual studies, the discussion here focuses mainly on the studies that either did
not use GAM Poisson models or those GAM studies which have been reanalyzed using more
stringent convergence criteria and/or alternative approaches. Newly available U.S. and Canadian
studies on relationships between short-term PM exposure and hospital admissions or emergency
visits that meet these criteria are summarized in Table 8-18, along with a few non-North
American studies. Reanalyses studies are indicated in Table 8-18 by indentation of the reference
citation to the pertinent short communication in the HEI Special Report (HEI, 2003b). The table
is organized by first summarizing single-pollutant (PM only) analyses and then multipollutant
(PM plus one or more co-pollutant) analyses for U.S. and non-U.S. studies.
Of much interest are NMMAPS multicity analyses (Samet et al., 2000a,b; Zanobetti
et al., 2000a), as reanalyzed (Zanobetti and Schwartz, 2003a), which provide evidence for
significant PM10 effects on cardiovascular-related hospital admissions and visits, using a variety
of statistical models. These results are supported by another multicity study (Schwartz, 1999)
which, however, has not been reanalyzed with alternative statistical models. Numerous other
studies, carried out by individual investigators in a variety of locales, present a more varied
picture, especially when gaseous co-pollutants have been analyzed in multipollutant models.
Most CVD hospital admissions studies reported to date have used PM10 as the main particle
measure due to the wide availability of ambient PM10 monitoring data.
Samet et al. (2000a,b) analyzed daily emergency-only CVD hospital admissions in persons
65 and older in relation to PM10 in 14 cities from the NMMAPS multicity study. The cities
included Birmingham, AL; Boulder, CO; Canton, OH; Chicago, IL; Colorado Springs, CO;
Detroit, MI; Minneapolis/St. Paul, MN; Nashville, TN; New Haven, CT; Pittsburgh, PA;
Provo/Orem, UT; Seattle, WA; Spokane, WA; and Youngstown, OH. The range of years
8-141
-------
TABLE 8-18. SUMMARY OF STUDIES OF PM10, PM10 2 5, 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 Lag
with PM Structure
Method
Effect Measures
Standardized to 50 ug/m3
PM10 or 25 ug/m3
™* PIVf **
2.5 1 ^1V110-2.5
U.S. Results Without Co-Pollutants
Samet et al.
(2000a,b)
14 Cities
Total CVD
admissions
> 65 yrs
Zanobetti and Schwartz,
(2003a)
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)
PM10.25: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.25: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.2.5: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.25: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.25: 13(7)
PM10: 31(19)
PM25: 18(11)
PM10.25: 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)
8-142
-------
TABLE 8-18 (cont'd). SUMMARY OF STUDIES OF PM10, PM10 25, OR PM25 EFFECTS
ON TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY VISITS
Reference
Citation, Outcome
Location, etc. Measure
Mean PM
Levels (IQR)
in ug/in3
Co-Pollutants
Analyzed Lag
with PM Structure
Method
Effect Measures
Standardized to 50 ug/m3
PM10 or 25 ug/m3
PM * PM **
rlvl2.5 5 rlvll(l-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
none 0 day
none 0 day
none 0-2 day
avg.
none 0-2 day
avg.
GAM not used
Default GAM
Strict GAM10Mf
GLMNS10Mf
Default GAM
Default GAM
Strict GAM30(lf
Strict GAM10Mf
GLM NS10Mf
Strict GAM30(lf
Strict GAM100(lf
GLM
nspline10Mf
GAM not used
GAM not used
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
CO 1 day
1 day
Oday
Default GAM
Default GAM
Default GAM
Default GAM
Default GAM
Default GAM
8.5% (-0.45-18.3)
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)**
8-143
-------
TABLE 8-18 (cont'd). SUMMARY OF STUDIES OF PM10, PM10 25, OR PM25 EFFECTS
ON TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY VISITS
Reference
Citation, Outcome
Location, etc. Measure
Mean PM
Levels (IQR)
in ug/in3
Co-Pollutants
Analyzed Lag
with PM Structure
Method
Effect Measures
Standardized to 50 ug/m3
PM10 or 25 ug/m3
PM * PM **
rlvl2.5 5 rlvll(l-2.5
U.S. Results With Co-Pollutants (cont'd)
Lippmann et al., 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)
PM10.15: 13(7)
PM10: 41,23
PM10: 35,22
PM10: 35,22
PM10: 44* (26)
PM25: 22* (16)
PM10
PM25
CO 0 day
1 day
Oday
CO, NO2, 0 day
SO2, O3
NO2 0 day
CO
CO 0 day
Default GAM
Default GAM
Default GAM
GAM not used
Default GAM
Strict GAM10Mf
GLMNS10Mf
Default GAM
Default GAM
Strict GAM10Mf
GLMNS10Mf
Strict GAM10Mf
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.
(1999a) 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
PM10.25: 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)
8-144
-------
TABLE 8-18 (cont'd). SUMMARY OF STUDIES OF PM10, PM10 25, OR PM25 EFFECTS
ON TOTAL CVD HOSPITAL ADMISSIONS AND EMERGENCY VISITS
Reference
Citation,
Location, etc.
Outcome
Measure
Mean PM
Levels (IQR)
in (ig/m3
Co-Pollutants
Analyzed
with PM
Lag
Structure Method
Effect Measures
Standardized to 50 ug/m3
PM10 or 25 ug/m3
PM * PM **
rlvl2.5 1 rlvll(l-2.5
Non-U.S. Results With Co-Pollutants
Burnett et al.,
(1997a)
Toronto, Canada
Stieb et al.
(2000)
Saint John,
Canada
Atkinson et al.
(1999a)
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
PM10.2.5: 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, CO
CO, H2S, N02,
O3, SO2, total
reduced sulfur
N02, 03,
SO2, CO
S02, N02,
O3, CO
N02, 03, S02
1-4 day GAM not used
avg.
1-3 day GAM not used
avg.
0 day GAM not used
1-3 day GAM not used
avg.
0-2 day GAM not used
avg.
-1.4% (-12.5, 11.2)
-1.6% (-10.5, 8.2)*
12.1% (-1.9, 28.2)**
PM10 not significant;
no quantitative results
presented
PM10 not significant;
no quantitative results
presented
PM10 effect robust;
no quantitative results
presented
PM10 effect robust;
no quantitative results
presented
* PM25 entries, **PM10.25. All others relate to PM10; *Median.
studied encompassed 1985 to 1994, but this varied by city. Covariates included SO2, NO2, O3,
and CO not analyzed directly as regression covariates; rather, individual cities were analyzed
first by Poisson regression methods on PM10 for lags from 0 to 5 days. An overall PM10 risk
estimate was then computed by taking the inverse-variance weighted mean of the city-specific
risk estimates. The city-specific risk estimates for PM10 were also examined for correlations
with omitted covariates, including other pollutants. No relationship was observed between city-
specific risk estimates and measures of socioeconomic status, including percent living in
poverty, percent non-white, and percent college educated. The overall weighted mean risk
estimate for PM10 was greatest for lag 0 and for the mean of lags 0-1. For example, the mean
risk estimate for the mean of lags 0-1 was a 5.9% (CI: 5.1 - 6.7) increase in CVD admissions per
8-145
-------
50 |ig/m3 PM10. The mean risk was larger in a subgroup of data where PM10 was less than
50 |ig/m3, suggesting the lack of a threshold. A weakness of this study was its failure to report
multipollutant results. The authors argued that confounding by co-pollutants was not present
because the city-specific risk estimates did not correlate with city-specific regressions of PM10
on co-pollutant levels. However, the validity of this method for identifying meaningful
confounding by co-pollutants at the daily time-series level has not been demonstrated. Thus, it is
not possible to conclude from these results alone that the observed PM10 associations were
independent of co-pollutants.
Samet et al. (2000a,b) reported results based on use of GAM LOESS smoothing to control
for time and weather covariates. Data from the 14 city NMMAPs analysis of CVD hospital
admissions were reanalyzed by Zanobetti and Schwartz (2003a) using three alternative control
methods. A small decrease in overall effects was observed as compared with the original study
results. Whereas the original 14 city pooled analysis yielded a 5.9% (CI: 5.1-6.7%) increase in
CVD admissions per 50 |ig/m3 increase in mean lags 0 and 1 day PM10, the reanalysis found
4.95% (3.95 to 5.95, 4.8% (3.55 to 6.0), and 5.0 (4.0 to 5.95) when reanalyzed by GAM with
stringent convergence criteria, GLM with natural spline, and GLM with penalized spline,
respectively. Based on these results, no change is warranted with regard to overall conclusions
for the original published study.
Zanobetti et al. (2000a) reanalyzed a subset of 10 cities from among the 14 evaluated by
Samet et al. (2000a,b). The same basic pattern of results obtained by Samet et al. (2000a,b) were
found, with strongest PM10 associations on lag 0 day, smaller effects on lag 1 and 2, and none at
longer lags. The cross-city weighted mean estimate at 0 day lag was excess risk = 5.6% (CI: 4.7,
6.4) per 50 |ig/m3 PM10 increment. For the 0-1 day lag average, excess CVD risk = 6.2%
(CI: 5.4, 7.0) per 50 |ig/m3 PM10 increment. Effect-size estimates increased when data were
restricted to days with PM10 < 50 |ig/m3. As before, no evidence of gaseous (CO, O3, SO2)
co-pollutant modification of PM effects was seen in the second stage analyses. Again, however,
co-pollutants were not tested as independent explanatory variables in the regression analysis.
Like the larger NMMAPS morbidity analyses reported by Samet et al. (2000a,b), this sub-study
utilized the GAM function in SPlus. These 10 cities were among the 14 cities that Zanobetti and
8-146
-------
Schwartz (2003a) reanalyzed using alternative statistical methods, and the reanalyses results
noted above would thus also apply in general here.
Janssen et al. (2002), in further analyses of the data set examined above by Samet et al.
(2000a,b), evaluated whether differences in prevalence of air conditioning (AC) use and/or the
contribution of different sources to total PM10 emissions could partially explain the observed
variability in exposure-effect relations in the 14 cities. Cities were characterized and analyzed as
either winter- or nonwinter-peaking for the AC analyses. Data on the prevalence of AC from the
1993 American Housing Survey of the United States Census Bureau (1995) were used to
calculate the percentage of homes with central AC for each metropolitan area. Data on PM10
emissions by source category were obtained by county from the U.S. EPA emissions and air
quality data web site (U.S. Environmental Protection Agency, 2000a). In an analysis of all
14 cities, central AC was not strongly associated with PM10 coefficients. However, separate
analysis for nonwinter-peaking and winter-peaking PM10 cities yielded coefficients for CVD-
related hospital admissions that decreased significantly with increased percentage of central AC
for both groups of cities. There were also significant positive relationships between CVD effects
and PM10 percent emissions from highways or from diesel vehicles, suggesting that mobile
source particles may have more potent cardiovascular effects than other particle types. For both
analyses, similar though weaker, patterns were found for hospitalization for COPD and
pneumonia. The authors note that the stronger relationship for hospital admission rates for CVD
over COPD and pneumonia may relate to the 10 times higher CVD hospital admissions rate
(which would result in a more precise estimate). However, no co-pollutant analyses were
reported. The ecologic nature and limited sample size also indicate the need for further study.
Because Janssen et al.'s analysis utilized the GAM function in SPlus, Zanobetti et al. (2003a)
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 remained significant for the
nonwinter PM10-peaking cities. The impact of highway vehicles and diesels on PM10 effect sizes
remained significant, as did oil combustion. However, the effect of air conditioning use on PM10
effect estimates was less pronounced and no longer statistically significant at p < 0.05 for the
8-147
-------
winter PM10-peaking cities using natural splines or penalized splines, in comparison to the
original Janssen et al. GAM analysis.
Schwartz (1999) extended the analytical approach he had used in Tucson (described below)
to eight other U.S. metropolitan areas, limiting analyses to a single county in each location to
enhance the representativeness of the air pollution data. The locations analyzed were Chicago,
IL; Colorado Springs, CO; New Haven, CT; Minneapolis, MN; St. Paul, MN; Seattle, WA;
Spokane, WA; and Tacoma, WA. Again, the analyses focused on total cardiovascular (CVD)
hospital admissions among persons > 65 years old. In univariate regressions, remarkably
consistent PM10 associations with CVD admissions were found across the eight locations, with a
50 |ig/m3 increase in PM10 being associated with 3.6 to 8.6% increases in admissions. The
univariate eight-county pooled PM10 effect was 5.0% (CI: 3.7 to 6.4), similar to the 6.1 % effect
per 50 |ig/m3 observed in the previous Tucson analysis. In a bivariate model that included CO,
the pooled PM10 effect size diminished somewhat to 3.8% (CI: 2.0 to 5.5) and the CO association
with CVD admissions was generally robust to inclusion of PM10 in the model. The Schwartz
1999 paper used GAM LOESS smoothing with default convergence criteria to control for time
and weather covariates. Although no direct reanalyses of this study using alternative statistical
methods have been reported, six of the eight cities included in Schwartz (1999) were included in
the NMMAPS reanalyses (Zanobetti et al., 2003; Zanobetti and Schwartz, 2003a).
Turning to some examples of independent single-city analyses, PM10 associations with
CVD hospitalizations were also examined in a study by Schwartz (1997a), which analyzed three
years of daily data for Tucson, AZ linking total CVD hospital admissions for persons > 65 years
old with PM10, CO, O3, and NO2. Only one site monitored daily PM10, whereas multiple sites did
so for gaseous pollutants (O3, NO2, CO). Both PM10 and CO were independently (i.e., robustly)
associated with CVD-related admissions; but O3 and NO2 were not. The percent effect of a
50 |ig/m3 increase in PM10 changed only slightly from 6.07% (CI: 1.12, 11.27) to 5.22%
(CI: 0.17, 10.54) when CO was included in the model along with PM10. The Schwartz 1997
paper utilized GAM smoothing to control for time and weather covariates. To date, no revised
results have been reported using alternative statistical methods.
8-148
-------
Morris and Naumova (1998) reported results for PM10, as well as for O3, NO2, and SO2, in
an analysis of four years of congestive heart failure data among people > 65 years old in
Chicago, IL. As many as eight monitoring sites were available for calculating daily gaseous
pollutant concentrations; but only one site in Chicago monitored daily PM10. Only same-day
results were presented, based on an initial exploratory analysis showing strongest effects for
same-day pollution exposure (i.e., lag 0). Associations between hospitalizations and PM10 were
observed in univariate regressions (3.9% [1.0, 6.9] per 50 |ig/m3 PM10 increase), but these
diminished somewhat in a multipollutant model (2.0%, [-1.4, 5.4]). Strong, robust associations
were seen between CO and congestive heart failure admissions. These results seem to suggest a
more robust association with CO than with PM10. However, the observed differences might also
be due in part to differential exposure misclassification for PM10 (monitored at one site) as
compared with CO (eight sites). This study did not use GAM functions to control for time and
weather covariates.
In a study designed to compare the effects of multiple PM indices, Lippmann et al. (2000)
analyzed associations between PM10, PM2 5, or PM10_2 5 and various categories of CVD hospital
admissions (only emergency and urgent admissions) among the elderly (65+ years old) in Detroit
on 344 days in the period 1992 to 1994. Whereas no consistent differences were observed in the
relative risks for the alternative PM indices, many of the associations involving PM were
significant: (a) ischemic heart disease (IHD) in relation to PM indices (i.e., 8.9% [0.5, 18.0] per
50 |ig PM10); 10.5% (2.8, 18.9) per 25 |ig/m3 PM10.2.5; and 4.3% (-1.4, 10.4) per 25 |ig/m3 PM25
(all at lag 2 day); and (b) heart failure (i.e., 9.7% [0.2, 20.2] 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 PM25 (the first two at lag 0 day
and the latter at lag 1 day). No associations with dysrythmias were seen however. The PM
effects generally were robust when co-pollutants were added to the model. Results for two-
pollutant models involving CO are given in Table 8-16 above. As discussed earlier with regard
to the Lippmann et al. (2000) mortality findings, it is difficult to discern whether the observed
associations with coarse fraction particles (PM10_2 5) are independently due to such particles or
may possibly be attributed to the moderately correlated fine particle (PM2 5) fraction in Detroit.
In addition, power was limited by the small sample size. Because GAM was used in the
8-149
-------
analyses reported in Lippmann et al. (2000), Ito (2003) has reported reanalyses results for the
Detroit study using GAM with more stringent convergence criteria and GLM with natural
splines. PM effect sizes diminished somewhat (up to 30%) and sometimes lost significance.
However, these changes tended to affect all PM metrics in a similar fashion. Thus, there was no
change in basic conclusions for the original Lippmann et al. (2000) study, i.e., that there was no
evidence for stronger effects for one size fraction versus others. Ito (2003) also noted that study
results were more sensitive to alternative weather models and degree of smoothing (degrees of
freedom used for the smoothing function) than to whether or not GAM, with strict convergence
criteria, was used.
Tolbert et al. (2000a) reported preliminary results for multiple PM indices in relation to
daily hospital emergency department (ED) visits for dysrhythmias (DYS) and all CVD
categories for persons aged 16 years or older, based on analyses of data from 18 of 33
participating hospitals in Atlanta, GA. During Period 1 of the study (1993 to 1998), PM10 from
the EPA AIRS database was reported to be negatively associated with CVD visits. In a
subsequent one-year period (Aug. 1998 to Aug. 1999), when data became available from the
Atlanta PM supersite, positive but nonsignificant associations were seen between CVD and PM10
(RR of 5.1% per 50 |ig/m3 PM10) and PM25 (RR of 6.1% per 25 |ig/m3 PM2 5); and significant
positive associations were seen with certain fine particle components, i.e., elemental carbon
(p < 0.005) and organic carbon (p < 0.02), and CO (p < 0.005). No multipollutant results were
reported. Study power was limited due to the short data record in Period 2.
In an analysis of 1992 to 1995 Los Angeles data, Linn et al. (2000) also found that PM10,
CO, and NO2 were all significantly associated with increased CVD admissions in single-
pollutant models among persons aged 30 years and older. Associations generally appeared to be
stronger for CO than for PM10. No PM10 results were presented with co-pollutants in the model.
Lastly, Moolgavkar (2000b) analyzed PM10, CO, NO2, O3, SO2 and limited PM2 5 data in
relation to daily total cardiovascular (CVD) and total cerebrovascular (CRV) admissions for
persons aged > 65 from three urban counties (Cook, IL; Los Angeles, CA; Maricopa, AZ) during
1987-1995. Of particular note was the availability of PM25 data in LA, though only for every
sixth day. Consistent with most studies, in univariate regressions, PM10 (and PM2 5 in LA) were
8-150
-------
associated at some lags with CVD admissions in Cook and LA counties, but not in Maricopa
county. However, in two-pollutant models in Cook and LA counties, the PM risk estimates
diminished substantially and/or were rendered nonsignificant, whereas co-pollutant (CO orNO2)
risk estimates were less affected. These results suggest that gaseous pollutants, with the
exception of O3, may have been more strongly associated with CVD hospitalizations than were
PM indices. These findings were based on an analysis that used GAM functions for time and
weather controls. Moolgavkar (2003) reported results of a reanalysis using improved GAM
convergence criteria and GLM with natural splines (nspline) and a range of degrees of freedom
(30 versus 100) for the smooth function of time. Results were not very sensitive to the use of
default versus improved GAM or splines (Table 8-18) but did appear to be more sensitive to
degrees of freedom. The nspline results were given only with 100 degrees of freedom. This is
an unusually large number, especially for PM2 5, where data were available only every sixth day
over a nine-year period.
The above analyses of daily PM10 and CO in U.S. cities, overall, indicate that elevated
concentrations of both PM10 and CO may enhance risk of CVD-related morbidity leading to
increased ED visits or hospitalizations. The Lippmann results appear to implicate both PM2 5
and PM10_25 in increased hospital admissions for some categories of CVD among the elderly.
8.3.1.3.2 Studies in Non-U.S. Cities
Four separate analyses of hospitalization data in Canada have been reported by Burnett and
coworkers since 1995 (Burnett et al., 1995, 1997a,b, 1999). A variety of locations, outcomes,
PM exposure metrics, and analytical approaches were used. The first study (Burnett et al.,
1995), reviewed briefly in the 1996 PM AQCD, analyzed six years of data from 168 hospitals in
Ontario, CN. Respiratory and CVD hospital admissions were analyzed in relation to sulfate
and O3 concentrations. Sulfate lagged one day was associated with CVD admissions, with an
effect of 2.8% (CI: 1.8, 3.8) increase per 13 |ig/m3 SO42 without O3 in the model and 3.3%
(CI: 1.7, 4.8) with O3 included. When CVD admissions were split out into sub-categories,
larger associations were seen between sulfates and coronary artery disease and heart failure than
for cardiac dysrhythmias. Sulfate associations with total admissions were larger for the
8-151
-------
elderly > 65 years old (3.5% per 13 |ig/m3) than for those < 65 years old (2.5% per 13 |ig/m3).
There was little evidence for seasonal differences in sulfate associations.
Burnett et al. (1997b) analyzed daily congestive heart failure hospitalizations in relation to
CO and other air pollutants (O3, NO2, SO2, CoH) in ten large Canadian cities as a replication of
an earlier U.S. study by Morris et al. (1995). The Burnett Canadian study expanded upon the
previous work both by its size (11 years of data for each of 10 large cities) and by including a
measure of PM air pollution (coefficient of haze, CoH); whereas no PM data were included in
the earlier Morris et al. study. The Burnett study was restricted to the population > 65 years old.
The authors noted that all pollutants except O3 were correlated, making it difficult to separate out
their effects statistically. CoH, CO, and NO2 measured on the same day as admission (i.e., lag 0)
were all strongly associated with congestive heart failure admissions in univariate models.
In multipollutant models, CO remained a strong predictor, but CoH did not (no gravimetric PM
data used).
The roles played by size-selected gravimetric and chemically-speciated particle metrics as
predictors of CVD hospitalizations were explored in analyses of data from metropolitan Toronto
for the summers of 1992 to 1994 (Burnett et al., 1997a). The analyses used dichotomous
sampler (PM2 5, PM10, and PM10_2 5), H+,and SO42 data collected at a central site as well as O3,
NO2, SO2, CO, and CoH data collected at multiple sites in Toronto. Hospital admissions
categories included total cardiovascular (i.e., the sum of ischemic heart disease, cardiac
dysrhythmias, and heart failure) and total respiratory-related admissions. Model specification
with respect to pollution lags was based on evaluation of all lags and averaging times out to
4 days prior to admission in exploratory analyses and "best" metrics being chosen on the basis of
maximal t-statistics. The relative risks of CVD admissions were positive and generally
statistically significant for all pollutants analyzed in univariate regressions, but especially so
for O3, NO2, CoH, and PM10_2 5 (i.e., regression t-statistics > 3). Associations for gaseous
pollutants were generally robust to inclusion of PM covariates, whereas the PM indices (aside
from CoH) were not robust to inclusion of multiple gaseous pollutants. In particular, PM2 5 was
not a robust predictor of CVD admissions in multipollutant models: whereas an 25 |ig/m3
increase in PM25 was associated with a 7.2% increase (t = 1.8) in CVD admissions in a
8-152
-------
univariate model, the effect was reduced to -1.6% (t = 0.3) in a model that included O3, NO2,
and SO2. CoH, like CO and NO2, is generally thought of as a measure of primary motor-vehicle
emissions during the non-heating season. The authors concluded that "particle mass and
chemistry could not be identified as an independent risk factor for exacerbation of
cardiorespiratory diseases in this study beyond that attributable to climate and gaseous air
pollution."
Burnett et al. (1999) later reported results of a more extensive attempt to explore cause-
specific hospitalizations for persons of all ages in relation to a large suite of gaseous and PM air
pollutant measures, using 15 years of Toronto data. Cardiovascular admissions were split out
into separate categories for analysis: dysrhythmias, heart failure, and ischemic heart disease.
Burnett et al. (1999) selected only those admissions to acute care treatment hospitals that were
considered an emergency or urgent. The analyses also examined several respiratory causes, as
well as cerebrovascular and diseases of the peripheral circulation; the latter categories were
included because they should show PM associations if one mechanism of PM action is related to
increased plasma viscosity, as suggested by Peters et al. (1997a). The PM metrics analyzed
were PM2 5, PM10, and PM10_2 5 estimated from daily TSP and TSP sulfate data, based on a
regression analysis for dichotomous sampling data that were available every sixth day during an
eight-year subset of the full study period. Although some statistically significant associations
with one or another PM metric were found in univariate models, no significant PM associations
were seen with any of the three CVD hospitalization outcomes in multipollutant models.
For example, whereas a 25 |ig/m3 increase in estimated PM25 was associated with a 8.05%
increase (t = 6.08) in ischemic heart disease admissions in a univariate analysis, the PM2 5
association was reduced to 2.25% (n.s.) when NO2 and SO2 were included in the model. The
gaseous pollutants dominated most regressions. There also were no associations between PM
and cerebral or peripheral vascular disease admissions. However, the use of estimated rather
than measured PM components limits interpretation of the reported PM results: that is, use of
estimated PM exposure metrics should, in general, tend to increase exposure measurement error
and thereby tend to decrease effects estimates.
8-153
-------
The Burnett et al. studies provide some of the most extensive results for PM in conjunction
with multiple gaseous pollutants, but the inconsistent use of alternative PM metrics in the
various analyses confuses the picture. A general finding appears to be lack of robustness of
associations between cardiovascular outcomes and PM in multipollutant analyses. This was seen
for CoH in the analysis of 10 Canadian cities (Burnett et al., 1997b), for PM2 5 and PM10 in the
analysis of summer data in Toronto (Burnett et al., 1997a), and for linear combinations of TSP
and sulfates (i.e., estimated PM25, PM10, and PM10_25) in the analysis of 15 years of data in
Toronto (Burnett et al., 1999). One exception was the association reported between CVD
admissions to 168 Ontario hospitals and sulfate concentrations (Burnett et al., 1995), where the
sulfate association was robust to the inclusion of O3. Also, although gravimetric PM variables
were not robust predictors in the Toronto summer analysis, CoH was (Burnett et al., 1997a),
perhaps reflecting the influence of primary motor vehicle emissions. This contrasts, however,
with lack of robustness for CoH in the 10-city analysis (Burnett et al., 1997b).
Stieb et al. (2000) studied all-age acute cardiac emergency room visits in relation to a rich
set of pollution covariates in Saint John, Canada for the period 1992 to 1996. Daily data were
available on PM2 5, PM10, fine fraction hydrogen and sulfate ions, CoH, CO, H2S, NO2, O3, SO2,
and total reduced sulfur. In a multipollutant model, neither PM10 nor PM2 5 were significantly
related to total cardiac ED visits, although O3 and SO2 were.
The APHEAII (Le Tertre et al., 2002) project examined the association between PM10 and
hospital admissions for cardiac causes in eight European cities. They found a significant PM10
effect (0.5%; CI: 0.2, 0.8) on admission for cardiac causes (all ages), as well as for both cardiac
causes (0.7%; CI: 0.4, 1.0) and ischemic heart disease (0.8%; CI: 0.3, 1.2) for people > 65 years
old, the effect per unit of PM10 pollution being half that reported for the U.S. NMMAPS
reanalyses (Zanobetti and Schwartz, 2003a). PM10 did not seem to be confounded by O3 or SO2.
The PM10 effect was reduced when CO was incorporated in the regression model and eliminated
when controlling for NO2. In contrast to PM10, BS was robustly associated with CVD hospital
admissions when co-pollutants were introduced into the model. This led the authors to suggest
that diesel PM may be especially important. GAM functions were used in the original analysis.
8-154
-------
In a reanalysis (Le Tertre et al., 2003) using GAM with stringent convergence criteria and GLM
with either natural or penalized splines, no marked changes from original results were observed.
Several additional non-U.S. studies, mainly in the United Kingdom (UK), have also been
published since the 1996 PM AQCD. Most of these studies evaluated co-pollutant effects along
with those of PM. Interpretation is hindered somewhat, however, by the failure to report
quantitative results for PM10 in the presence of co-pollutants. In univariate models, Atkinson
et al. (1999a) reported PM associations for persons aged < 65 years and for those > 65 years.
Significant associations were reported for both ambient PM10 and black smoke (BS), as well as
all other co-pollutants, with daily admissions for total cardiovascular disease and ischemic heart
disease for 1992 to 1994 in London, UK, using standard time-series regression methods.
In two-pollutant models, the associations with PM10, NO2, SO2, and CO were moderated by the
presence of BS in the model, but the BS association was robust to co-pollutants. Interpretation is
hampered somewhat by the lack of quantitative results for two-pollutant models.
In another U.K. study, associations with PM10, and to a lesser extent BS, SO2, and CO,
were reported for analyses of daily emergency hospital admissions for cardiovascular diseases
from 1992 to 1995 for Edinburgh, UK (Prescott et al., 1998). No associations were observed
for NO2 and O3. Significant PM10 associations for CVD admissions were present only in
persons < 65 years old. The authors reported that the PM10 associations were unaffected by
inclusion of other pollutants; however, results were not shown. On the other hand, no
associations between PM10 and daily ischemic heart disease admissions were observed by
Wordley and colleagues (1997) in an analysis of two years of daily data from Birmingham, UK.
However, PM10 was associated with respiratory admissions and cardiovascular mortality during
the same study period. This inconsistency of results across causes and outcomes is difficult to
interpret, but may relate in part to the relatively short time-series analyzed. The authors stated
that gaseous pollutants did not have significant associations with health outcomes independent of
PM, but no results were presented for models involving gaseous pollutants.
A study in Hong Kong by Wong et al. (1999a) found associations between CVD hospital
admissions and PM10, SO2, NO2, and O3 in univariate models, but did not examine multipollutant
models. In models including PM10 and dichotomous variables for gaseous pollutants (high
8-155
-------
versus low concentration), the PM10 effects remained relatively stable. Ye et al. (2001) analyzed
a 16-year record of daily emergency hospital visits for July and August in Tokyo among persons
> 65 years. In addition to PM10, the study included NO2, O3, SO2, and CO. Models were built
using an objective significance criterion for variable inclusion. NO2 was the only pollutant
found to be significantly associated with angina, cardiac insufficiency, and myocardial infarction
hospital visits.
8.3.1.3.3 Summary of Salient Findings for Acute PM Exposure Effects on CVD
Hospital Admissions
The ecologic time-series studies reviewed here add to a growing body of evidence on acute
CVD morbidity effects of PM and co-pollutants. Two U.S. multicity studies offer the strongest
current evidence for effects of PM10 on acute CVD hospital admissions, but uncertainties
regarding the possible role of co-pollutants in the larger of the two studies hinders interpretation
with respect to independent PM10 effects. Among single-city studies carried out in the U.S. and
elsewhere by a variety of investigators (see Table 8-18), less consistent evidence for PM effects
is seen. Of particular importance are possible roles of gaseous co-pollutants (e.g., CO) as
potential confounders of the PM effect. Among 13 independent studies that included
gravimetrically-measured PM10 and co-pollutants, three reported PM effects that appeared to be
independent of co-pollutants (Schwartz, 1997; Lippmann et al., 2000; Prescott et al., 1998); eight
reported no significant PM10 effects after inclusion of co-pollutants (Morris and Naumova, 1998;
Moolgavkar, 2000b; Tolbert et al., 2000a; Burnett et al., 1997a; Steib et al., 2000; Atkinson
et al., 1999a; Wordley et al. (1997); Morgan et al., 1998; Ye et al., 2001); and two studies were
unclear regarding independent PM effects (Linn et al., 2000; Wong et al., 1999a). In a
quantitative review of published results from 12 studies on airborne particles and CVD hospital
admissions, Morris (2001) noted that adjustment for co-pollutants consistently reduced the PM10
effect, with reductions ranging from 10 to 320% across studies. Thus, although several studies
do appear to provide evidence for PM effects on CVD hospital admissions independent of
co-pollutant effects, a number of other studies examining co-pollutants did not find results
indicative of independent PM10 effects on CVD hospital admissions.
8-156
-------
With respect to particle size, only a handful of studies have examined the relative effects of
different particle indicators (Lippmann et al., 2000; Burnett et al., 1997a; Steib et al., 2000;
Moolgavkar, 2000b). Perhaps due to statistical power issues, no clear picture has emerged as to
particle-size fraction(s) most associated with acute CVD effects.
As discussed above, several studies originally based on statistical analyses involving the
SPlus GAM function have reported new results using alternative statistical methods. The
reanalyses yielded some slightly reduced effect estimates and/or increased confidence intervals
or little or no change resulted in other cases. Thus, based on these new results, the overall
conclusions from the cardiovascular hospitalization studies remain the same.
Because hospitalization likely reflects some of the same pathophysiologic mechanisms that
may be responsible for acute mortality following PM exposure, it is of interest to assess the
coherence between the morbidity results reviewed here and the mortality results reviewed in
Section 8.2.2 (Borja-Aburto et al., 1997, 1998; Braga et al., 2001a; Goldberg et al., 2000;
Gouveia and Fletcher, 2000; Hoek et al., 2001; Kwon et al., 2001; Michelozzi et al., 1998;
Morgan et al., 1998; Ponka et al., 1998; Schwartz et al., 1996a; Simpson et al., 1997; Wordley
et al., 1997; Zeghnoun et al., 2001; Zmirou et al., 1998). The mortality studies reported
significant associations between acute CVD mortality and measures of ambient PM, though the
PM metrics used and the relative risk estimates obtained varied across studies. The PM
measurement methods included gravimetrically analyzed filter samples (TSP, PM10, PM2 5,
PM10_25), beta gauge (particle attenuation of beta radiation), nephelometry (light scattering), and
black smoke (filter reflectance). When tested, PM associations with acute CVD mortality
appeared to be generally more robust to inclusion of gaseous covariates than was the case for
acute hospitalization studies (Borja-Aburto et al., 1997, 1998; Morgan et al., 1998; Wordley
et al., 1997; Zmirou et al., 1998). Three studies (Braga et al., 2001a; Goldberg et al., 2000; Hoek
et al., 2001), as noted in Section 8.2.2, provide data indicating that some specific CVD causes of
mortality (such as heart failure) were more strongly associated with air pollution than total CVD
mortality; but it was noted that ischemic heart disease (which contributes about half of all CVD
deaths) was the strongest contributor to the association between air pollution and cardiovascular
8-157
-------
mortality. The above-noted results for acute CVD mortality are qualitatively consistent with
those reviewed earlier in this section for hospital admissions.
Figure 8-10 illustrates PM10 excess risk estimates for single-pollutant models derived from
selected U.S. studies of PM10 exposure and total CVD hospital admissions, standardized to a
50 |ig/m3 exposure to PM10 as shown in Table 8-16. Results are shown both for studies yielding
pooled outcomes for multiple U.S. cities and for studies of single U.S. cities. The Zanobetti and
Schwartz (2003a) and Samet et al. (2000a) pooled cross-city results for 14 U.S. cities provide the
most precise estimate for relationships of U.S. ambient PM10 exposure to increased risk for CVD
hospitalization. That estimate, and those derived from most other studies seen in Figure 8-10,
generally appear to confirm likely excess risk of CVD-related hospital admissions for U.S. cities
in the range of 3 to 9% per 50 |ig/m3 PM10, especially among the elderly (> 65 year). Other
individual-city results (see Table 8-16) from Detroit are also indicative of excess risk being in
the range of approximately 3.0 and 8.1% per 25 |ig/m3 of PM25 or PM10_25, respectively, for
ischemic heart disease and 6.8% and 4.9% excess risk per 25 |ig/m3 of PM2 5 and PM10_2 5,
respectively, for heart failure. 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 Effect Markers
Several new studies have evaluated longitudinal associations between ambient PM and
cardiovascular effect markers (i.e., physiologic measures of cardiovascular function or
biochemical changes in the blood that may be associated with increased cardiac risks).
In contrast to the ecologic time-series studies discussed above, these studies measure outcomes
and most covariates at the individual level, making it possible to draw conclusions regarding
individual risks, as well as to explore mechanistic hypotheses. Heterogeneity of responses
across individuals, and across subgroups defined on the basis of age, sex, preexisting health
status, etc., also can be assessed, in principle. While exposure assessment remains largely
ecologic (i.e., the entire population is usually assigned the same exposure value on a given day),
exposure is generally well characterized in the small, spatially-clustered study populations. The
8-158
-------
Zannobetti and Schwartz (2003a)
14 US Cities
Moolgavkar (2003)
LA, CA
Moolgavkar (2003)
Cook County
Linn et al. (2000)
LA, CA
Morris & Naumova (1998)
Chicago
Ito (2003)
Detroit
i A i
Total CVD A
CHI- ' * '
HFi A i
ii in. A i
inu1 v '
-10
-5
10
15
20
25
Reconstructed Excess Risk Percentage
per 50 ug/m3 Increase in PM10
Figure 8-10. Acute cardiovascular hospitalizations and PM exposure excess risk estimates
derived from U.S. PM10 studies based on single-pollutant models from GAM
strict convergence criteria reanalyses (2003 studies) or alternative (non-
GAM) original analyses. Both multipollutant 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. IHD = ischemic heart disease.
recent studies fall into two broad classes: (1) those addressing heart rate, cardiac rhythm, blood
pressure, or other cardiac function indicators; and (2) those addressing blood characteristics.
While significant uncertainty still exists regarding the interpretation of results from these new
studies, the varied responses that have been reported to be associated with ambient PM and
co-pollutants are of much interest in regard to mechanistic hypotheses concerning
pathophysiologic processes potentially underlying CVD-related mortality/morbidity effects
discussed in preceding sections.
Cardiac physiology and adverse cardiac events
Alterations in heart rate and/or rhythm are thought to reflect pathophysiologic changes that
may represent possible mechanisms by which ambient PM exposures may exert acute effects on
8-159
-------
human health. Decreased heart rate variability, in particular, has been identified as a predictor of
increased cardiovascular morbidity and mortality. Several independent studies have reported
temporal associations between PM exposures and various measures of heart beat rhythm in
panels of elderly subjects (Liao et al., 1999; Pope et al., 1999a,b,c; Dockery et al., 1999; Peters
et al., 1999a, 2000a; Gold et al. 2000; Creason et al., 2001). Changes in blood pressure may also
reflect increases in CVD risks (Linn et al., 1999; Ibald-Mulli et al., 2001). Finally, one
important new study (Peters et al., 200la) has linked acute (2- and 24-h) ambient PM2 5 and PM10
concentrations with increased risk of myocardial infarction in subsequent hours and days.
Liao et al. (1999) studied 26 elderly subjects (age 65 to 89 years; 73% female) over three
consecutive weeks at a retirement center in metropolitan Baltimore, 18 of whom were classified
as "compromised" based on previous cardiovascular conditions (e.g., hypertension). Daily
6-min resting electrocardiogram (ECG) data were collected, and time intervals between
sequential R-R intervals recorded. A Fourier transform was applied to the R-R interval data
to separate its variance into two major components: low frequency (LF, 0.04-0.15 Hz) and
high frequency (FTP, 0.15-0.40 Hz). The standard deviation of all normal-to-normal (N-N;
also designated R-R) heartbeat intervals (SDNN) was computed as a time-domain outcome
variable. PM2 5 was monitored indoors by TEOM and outdoors by dichotomous sampler.
Outdoor PM25 levels ranged from 8.0 to 32.2 |ig/m3 (mean =16.1 |ig/m3). Regression analyses
controlled for inter-subject differences in average variability, allowing each subject to serve as
his/her own control. Consistent associations were seen between increases in PM2 5 levels (both
indoors and outdoors) and decreases in all three outcome variables (LF, FTP, SDNN), with
associations being stronger for the 18 "compromised" subjects. The short time interval (6 min
per day) of measurement for these parameters hampers interpretation of the possible medical
significance of the reported positive results; longer or several measurements per day would have
allowed for clearer indications of likely underlying perturbation of CV function.
Creason et al. (2001) reported results of a subsequent study using similar methods among
56 elderly residents of a retirement center in Baltimore County, MD. The 11 men and 45 women
ranged in age from 72 to 97 years and were all Caucasian. Associations between ambient PM2 5
and decreased FtRV were not statistically significant at p < 0.05. When two episodic PM25 days
8-160
-------
with rainfall were excluded from the 24-day data set, trends associating decreased HRV
and PM2 5 were present, but did not meet significance at p < 0.05. There was no evidence of
effects among subsets of subjects with compromised health status as observed previously in the
study by Liao et al. (1999). No results were presented for pollutants other than PM25.
Pope and colleagues (1999c), using ambulatory ECG monitoring, studied HRV and PM10
in a panel of six elderly subjects (69 to 89 years, 5/6 male) and one 23-year old male subject, all
compromised by some form of heart disease. SDNN, SDANN, and r-MSSD were used as
measures of HRV based on 48-hr hotter readings. Daily gravimetric PM10 data from three sites
in the study area ranged from -10 |ig/m3 to 130 |ig/m3 during the study, with high levels
occurring only during the first half of the 1.5 month study period. No co-pollutants (e.g., O3,
CO, NO2, etc.) were studied. Regression analyses with subject-specific intercepts were
performed, with and without control for daily barometric pressure and mean heart rate. Same-
day and previous-day ambient PM10 were negatively associated with SDNN and SDANN; and
the results were unaffected by inclusion of covariates. Heart rate, as well as r-MSSD, were both
positively, but less strongly, associated with PM10. The specific heart rate variability findings
(i.e., PM associations with decreased SDANN and SDNN and increased r-MSSD) make it
difficult to interpret the results or their cardiac health significance. The decreased SDANN and
SDNN suggests decreased sympathetic activity, whereas the r-MSSD increase suggests increase
parasympathetic (vagal) input to the heart (which is likely protective in terms of risk of ischemic
related arrhythmia, but might increase the risk of atrial arrhythmia). These specific HRV
findings do not allow clear conclusions as to how PM may be affecting cardiac functioning.
The Pope et al. (1999c) study discussed above was nested within a larger cohort of
90 subjects who participated in a study of heart rate and oxygen saturation in the Utah Valley
(Dockery et al., 1999; Pope et al., 1999b). The investigators hypothesized that decreases in
oxygen saturation might occur as a result of PM exposure and that this could be a risk factor for
adverse cardiac outcomes. The study was carried out in winter months (mid-November through
mid-March), when frequent inversions lead to fine particle episodes. PM10 levels at the three
nearest sites averaged from 35 to 43 |ig/m3 during the study, and daily 24-h levels ranged from
5 to 147 |ig/m3. Two populations were studied: 52 retired Brigham Young University
8-161
-------
faculty/staff and their spouses, and 38 retirement home residents. Oxygen saturation (SpO2) and
heart rate (HR) were measured once or twice daily by an optical sensor applied to a finger.
In regression analyses controlling for inter-individual differences in mean levels, SpO2 was not
associated with PM10, but was highly associated with barometric pressure. In contrast, HR
association with PM10 significantly increased but significantly decreased with barometric
pressure in joint regressions. Including CO in the regressions did not change these basic
findings. This was the first study of this type to examine the interrelationships among
physiologic measures (i.e., SpO2 and HR), barometric pressure, and PM10. The profound
physiological effects of barometric pressure noted here highlight the importance of carefully
controlling for barometric pressure effects in studies of cardiac physiology.
Gold and colleagues (2000) obtained somewhat different results in a study of heart rate
variability among 21 active elderly subjects, aged 53 to 87 years, in a Boston residential
community. Resting, standing, exercising, and recovering ECG measurements were performed
weekly using a standardized protocol on each subject, which involved 25 min/week of
continuous Hotter ECG monitoring. Two time-domain measures were extracted: SDNN
and r-MSSD (see above for definitions). Heart rate also was analyzed as an outcome.
Continuous PM10 and PM2 5 monitoring was conducted by TEOM at a site 6 km from the study
site and PM data were corrected for the loss of semivolatile mass. Data on CO, O3, NO2, SO2,
temperature and relative humidity were available from nearby sites. Outcomes were regressed
on PM2 5 levels in the 0 to 24 h period prior to ECG testing, with and without control for HR and
temperature. As for the other studies discussed above, declines in SDNN were associated
with PM2 5 levels, in this case averaged over 4 h. These associations reached statistical
significance at the p < 0.05 level only when all testing periods (i.e., resting, standing, exercise)
were combined. In contrast to the above studies, both HR and r-MSSD here were negatively
associated with PM2 5 levels (i.e., lower HR and r-MSSD) when PM2 5 was elevated. These
associations were statistically significant overall, as well as for several of the individual testing
periods, and were unaffected by covariate control. Gold et al. (2003) subsequently reported
reanalyses involving temperature with either a GAM function with stringent convergence criteria
or a GLM with natural splines, with no substantial changes in results being reported. The
8-162
-------
negative associations between PM25 and decreases in both HR and r-MSSD are puzzling, given
that decreased HR is indicative of increased parasympathetic tone whereas decreased r-MSSD is
reflective of decreased parasympathetic modulation of heart function. This discrepancy raises
the possibility that one or another or both of the observed outcomes may be due to chance.
Evidence for altered HRV in response to PM2 5 exposures comes from two other recent
studies. Magari et al. (2001) found significant decreases in SDNN of 1.4% (CI: 2.1, 0.6) per
100 ug/m3 3-h mean PM25 in young healthy Boston area boilermakers studied during non-work
periods. Another study of 40 boilermakers (including the 20 studied above) analyzed data
collected during both work and non-work periods (Magari et al., 2002). That study found a
significant 2.7% decrease in SDNN and a 1.0% increase in HR for every 100 |ig/m3 increase in
4-h moving average of estimated PM2 5. The larger effect size for the non-work PM exposure
study may reflect differing health effects of ambient versus occupational PM composition.
These studies are suggestive of PM-related HRV effects in young healthy adults, but use of
estimated PM2 5 based on light scattering precludes firm quantitative interpretation of exposure
levels.
Peters et al. (1999a) reported HR results from a retrospective analysis of data collected as
part of the MONICA (monitoring of trends and determinants in cardiovascular disease) study in
Augsburg, Germany. Analyses focused on 2,681 men and women aged 25 to 64 years who had
valid ECG measurements taken in winter 1984-1985 and again in winter 1987-1988. Ambient
pollution variables included TSP, SO2, and CO. The earlier winter included a 10-day episode
with unusually high levels of SO2 and TSP, but not of CO. Pollution effects were analyzed in
two ways: dichotomously comparing the episode and non-episode periods, and continuously
using regression analysis. However, it is unclear from the report as to what extent the analyses
reflect between-subject versus within-subject effects. A statistically significant increase in mean
heart rate was seen during the episode period versus other periods, controlling for cardiovascular
risk factors and meteorology. Larger effects were seen in women. In single-pollutant regression
analyses, all three pollutants were associated with increased HR.
More recently, Ibald-Mulli et al. (2001) reported similar findings from a study of blood
pressure among 2,607 men and women aged 25 to 64 years in the MONICA study. Systolic
8-163
-------
blood pressure increased on average during an episode of elevated TSP and SO2, but the effect
disappeared after controlling for meteorological parameters (e.g., temperature and barometric
pressure). However, when TSP and SO2 were analyzed as continuous variables, both were
associated with elevated systolic blood pressure, controlling for meteorological variables.
In two-pollutant models, TSP was more robust than SO2, and the TSP association was greater in
subgroups of subjects with elevated blood viscosity and heart rates.
Linn et al. (1999) reported associations between both diastolic and systolic blood pressure
and PM10 in a panel study of 30 Los Angeles residents with severe COPD. The relationship was
not observed when inside-home PM levels were used in the analyses. Also, no relationship was
found between PM levels and heart rate or arrhythmias, based on 48 h of Hotter data.
In a retrospective study, Peters and colleagues (2000a) examined incidence of cardiac
arrhythmias among 100 patients (mean age 62.2 years; 79% male) with implanted cardiovertex
defibrillators followed over a three year period. Shocks from cardiovertex defibrillators are
frequently used for life-threatening arrhythmias but not always (only -65-70% are for life-
threatening arrhythmias). PM2 5 and PM10 were measured in South Boston by the TEOM
method, along with black carbon, O3, CO, temperature and relative humidity; SO2 and NO2 data
were obtained from another site. The 5th percentile, mean, and 95th percentiles of PM10 levels
were 7.8, 19.3, and 37.0 |ig/m3, respectively. The corresponding PM25 values were 4.6, 12.7,
and 26.6 |ig/m3. Logistic regression was used to analyze events in relation to pollution variables,
controlling for between-person differences, seasons, day-of-week, and meteorology in two
subgroups: 33 subjects with at least one arrhythmia event and 6 subjects with 10 or more such
events. In the larger subgroup, only NO2 on the previous day, and the mean NO2 over five days,
were significantly associated with arrhythmia incidence. In patients with 10 or more events,
the NO2 associations were stronger. Also, some of the PM25 and CO lags became significant in
this subgroup. Important caveats regarding this study include the fact that the vast majority of
cardiovertex defibrillator discharges occurred among a small subset (i.e., 6) of the patients.
Also, potentially important variables, e.g., cardiovascular drug usage and anti-arrhythmia drug
changes during follow-up, were not reported.
8-164
-------
An exploratory study of a panel of COPD patients (Brauer et al., 2001) examined several
PM indicators in relation to CVD and respiratory health effects. The very low levels of ambient
particles (PM10 mean =19 |ig/m3) and low variability in these levels plus the small sample size
of 16 limit the conclusions that can be drawn. Still, for cardiovascular endpoints, single-
pollutant models indicated that both systolic and diastolic BP decreased with increasing PM
exposure, but this was not statistically significant. Also, 24-h Hotter monitoring data recorded
on 7 separate days for each individual did not show any heart rate variability changes associated
with PM levels. The size of the ambient PM10 effect estimate for AFEVj was larger than the
effect estimate for ambient PM2 5 and personal PM2 5 but not statistically significant. This initial
effort indicated that ambient PM10 consistently had the largest effect estimates, whereas while
models using personal exposure measurements did not show larger or more consistently positive
effect estimates relative to those models using ambient exposure metrics.
A potentially important study by Peters et al. (200la) reported associations between onset
of myocardial infarction (MI) and ambient PM (either PM10 or PM25) as studied in a cohort of
772 MI patients in Boston, MA. Precise information on the timing of the MI, obtained from
patient interviews, was linked with concurrent air quality data measured at a single Boston site.
A case crossover design enabled each subject to serve as his/her own control. One strength of
this study was its analysis of multiple PM indices and co-pollutants, including real-time PM25,
PM10, the PM10_2 5 difference, black carbon, O3, CO, NO2, and SO2. Only PM2 5 and PM10 were
significantly associated with MI risk in models adjusting for season, meteorological parameters,
and day of week. Both the mean PM25 concentration in the previous two hours and in the 24 h
lagged one day were independently associated with MI, with odds ratios of 1.48 (CI: 1.09, 2.02)
for 25 |ig/m3 and 1.62 (CI: 1.13, 2.34) for 20 |ig/m3, respectively. PM10 associations were
similar. The nonsignificant findings for other pollution metrics should be interpreted in the
context of potentially differing exposure misclassification errors associated with the single
monitoring site.
Checkoway et al. (2000) has reported a Seattle mortality study of PM10 levels and cases of
patients experiencing out-of-hospital sudden cardiac death (SCD). They used a case-crossover
study design in 362 subjects suffering an SCD episode. They evaluated PM levels over the
8-165
-------
5 days preceding SCD and compared those levels to levels recorded in the same month and
during the same days of the week (Mean PM10 level = 31.9 jig/m3). They evaluated lags of 0 to
5 days looking for a correlation, but found no correlation between SCD episodes and PM levels
even after controlling for multiple confounding variables. They reported an estimated relative
risk at a one day lag of 0.87 (CI: 0.74, 1.01). The HEI (2000) review commentary noted that the
authors reported, from their power calculations, that the sample size (362) was not large enough
to either find or rule out a relative risk less than 1.5 and that lack of association with PM in this
study does not imply that other cardiac or cardiovascular disease outcomes are not associated
with PM. These negative findings suggest that PM may not be a risk factor for acute myocardial
infarction in previously healthy individuals, or that the pattern and/or mix of PM exposures in
Seattle, where woodsmoke may be an important component, may convey lesser risk than
observed elsewhere.
The above studies present a range of findings regarding possible effects of PM25 on cardiac
rhythm and other cardiac endpoints. However, the studies offer conflicting results, especially
with regard to HRV findings. Several studies reported PM levels to be associated with decreases
in one or more HR variability measured in elderly subjects with preexisting cardiopulmonary
disease, although increased r-MSSD (a measure of high-frequency HR variability) was found to
be associated with PM elevations in at least one study (Pope et al., 1999a). Several other found
no changes related to PM levels (Creason, et al., 2001) or blood pressure (Brauer et al., 2001).
Some recent studies have also reported effects in healthy elderly and young adult populations.
All those studies which examined HR found associations with PM most being positive
associations; but one (Gold et al., 2000; Gold et al., 2003) reported a negative relationship.
Overall, variations in methods used and discrepancies in results obtained across the studies argue
for caution in drawing any conclusions yet regarding ambient PM effects on heart rate variability
or other ECG measures of cardiovascular parameters.
Viscosity and other blood characteristics
Peters et al. (1997a) state that plasma viscosity, a risk factor for ischemic heart disease, is
affected by fibrinogen and other large asymmetrical plasma proteins, e.g., immunoglobulin M
8-166
-------
and a2-macroglobulin. They noted that, in a cohort study (Woodhouse et al., 1994) of elderly
men and women, fibrinogen levels were strongly related to inflammatory markers, such as
neutrophil count and acute-phase proteins (C-reactive protein and c^-antichymotrypsin) and self-
reported infections. They further noted that another prospective study (Thompson et al., 1995;
Haverkate et al., 1997) showed baseline fibrinogen and C-reactive protein concentrations to be
highly correlated in angina patients and to be independently associated with increased risk of
myocardial infraction.
Support for a mechanistic hypothesis, relating to enhanced blood viscosity, was suggested
by an analysis of plasma viscosity data collected in a population of 3,256 German adults in the
MONICA study (Peters et al., 1997a). Each subject provided one blood sample during October
1984 to June 1985. An episode of unusually high air pollution levels occurred during a 13-day
period while these measurements were being made. Among the 324 persons who provided
blood during the episode, there was a statistically significant elevation in plasma viscosity as
compared with 2,932 persons studied at other times. The odds ratio for plasma viscosity
exceeding the 95th percentile was 3.6 (CI: 1.6, 8.1) among men and 2.3 (CI: 1.0, 5.3) among
women. Analysis of the distribution of blood viscosity data suggested that these findings were
driven by changes in the upper tail of the distribution rather than by a general shift in mean
viscosity, consistent with the likelihood of a susceptible subpopulation.
A prospective cohort study of a subset of male participants from the above-described
Augsburg, Germany MONICA study was reported by Peters et al. (200Ib). Based on a survey
conducted in 1984/85, a sample of 631 randomly selected men (aged 45 to 64 years and free of
cardiovascular disease at entry) were evaluated in a 3-year follow-up that examined relationships
of air pollution to serum C-reactive protein concentrations. C-reactive protein is a sensitive
marker of inflammation, tissue damage, and infections, with acute and chronic infections being
related to coronary events. Inflammation is also related to systemic hypercoagulability and onset
of acute ischemic syndromes. During the 1985 air pollution episode affecting Augsburg and
other areas of Germany, the odds of abnormal increases in serum C-reactive protein (i.e.,
> 90th percentile of pre-episode levels = 5.7 mg/L) tripled; and associated increases in TSP
levels of 26 |ig/m3 (5-day averages) were associated with an odds ratio of 1.37 (CI: 1.08, 1.73)
8-167
-------
for C-reactive protein levels exceeding the 90th percentile levels in two pollutant models that
included SO2 levels. The estimated odds ratio for a 30 |ig/m3 increase in the 5-day mean for SO2
was 1.12(CI:0.92, 1.47).
Other studies have examined blood indices in relation to PM pollution in United Kingdom
cities. Seaton and colleagues (1999) collected sequential blood samples (up to 12) over an
18-month period in 112 subjects (all over age 60) in Belfast and Edinburgh, UK. Blood samples
were analyzed for hemoglobin, packed cell volumes, fibrinogen, blood counts, factor VII,
interleukin-6, and C-reactive protein. In a subset of 60 subjects, plasma albumin also was
measured. PM10 data monitored by TEOM were collected from ambient sites in each city.
Personal exposure estimates for three days preceding each blood draw were derived from
ambient PM data adjusted by time-activity patterns and I/O penetration factors. No co-pollutants
were analyzed. Data were analyzed by analysis of covariance, controlling for city, seasons,
temperature, and between-subject differences. In this relatively small panel study, significant
changes in several blood indices were associated with either ambient or estimated
personal PM10 levels; but all the associations were negative, except for C reactive protein
in relation to ambient PM10.
Prescott et al. (2000) also investigated factors that might increase susceptibility to
PM-related cardiovascular events for a large cohort of 1,592 subjects aged 55 to 74 in
Edinburgh, UK. Baseline measurements of blood fibrinogen and blood and plasma viscosity
were examined as modifiers of PM effects (indexed by BS) on the incidence of fatal and nonfatal
myocardial infarction or stroke. All three blood indices were strong predictors of increased
cardiac event risk; but there was no clear evidence of either a main effect of BS, nor interactions
between BS and blood indices.
In another European study, Pekkanen and colleagues (2000) analyzed plasma fibrinogen
data from a cross-sectional survey of 4,982 male and 2,223 female office workers in relation to
same-day and previous three-day PM10, BS, NO2, CO, SO2, and O3 concentrations. In the full
analysis, NO2 and CO were significantly associated with increased fibrinogen levels. When the
analysis was restricted to the summer season, PM10 and BS, as well as NO2 and CO showed
significant univariate associations.
8-168
-------
Schwartz (2001) later reported even larger analyses of possible PM effects on factors
affecting blood coagulability among a subset of the NHANES III cohort. The NHANES III
cohort was comprised of a stratified random sample of the U.S. population, with oversampling of
minorities (Black and Mexican-Americans represented 30% of the cohort) and of the elderly
(20% of the cohort was > 60 years old versus their being 16% of the U.S. population). The
NHANES III study included evaluations of numerous health and nutritional endpoints conducted
in two phases during 1989 to 1994, each phase sampling -20,000 subjects in 44 communities
and including persons representative of the U.S. population. Analyzing data for first phase
subjects living in 30 urban areas having every-six-day PM10 monitoring (no. of PM10
observations = 1,373) by mixed models (PROC MIXED, SAS), Schwartz (2001) found not only
significant positive associations between PM10 exposures and plasma fibrinogen levels in a
subset of the NHANES III cohort, but also PM10 associations with platelet and white cell counts.
The PM10 associations were robust when O3, NO2, or SO2 were included with PM10 in two-
pollutant models. In univariate models, SO2 was only significant for white cell counts and NO2
with platelet counts and fibrinogen but not O3 with any of the three blood coaguability markers.
Given that CO data were not matched to specific subjects, no CO analyses were done.
Overall, the above findings add some limited support for hypotheses about possible
mechanisms by which PM exposure may be linked to adverse cardiac outcomes. They appear to
most clearly implicate ambient PM as likely contributing to increases in C-reactive protein (a
biological marker of inflammatory responses), blood fibrinogen levels, and blood viscosity, all
of which are thought to be predictive of increased risk for serious cardiac events.
8.3.1.4 Issues in the Interpretation of Acute Cardiovascular Effects Studies
Susceptible subpopulations. Because they lack extensive data on individual subject
characteristics, hospital admissions studies provide only limited information on susceptibility
factors based on stratified analyses. The relative effect sizes for PM-cardiovascular associations
(and respiratory) admissions reported in ecologic time-series studies are generally somewhat
higher than those for total admissions. This provides some limited support for hypothesizing
that acute PM effects operate via cardiopulmonary pathways or that persons with preexisting
8-169
-------
cardiopulmonary disease have greater susceptibility to PM, or both. Although there are some
data from ecologic time-series studies showing larger PM effects on cardiovascular admissions
in adults aged > 65 years versus younger populations, the differences are neither striking nor
consistent. One recent study reported larger CVD hospitalization among persons with current
respiratory infections. The other individual-level studies of cardiophysiologic function assessed
above are suggestive, but do not yet fully confirm that elderly persons with preexisting
cardiovascular or respiratory disease are susceptible to subtle changes in heart rate variability in
association with PM exposures. More data are needed before that conclusion can be drawn with
confidence. Because younger and healthier populations have not yet been much studied, it is not
yet possible to assess the extent to which ambient PM exposures may affect their cardiovascular
health status or whether they are at lower risk for PM-related CVD effects than are the elderly.
Role of other environmental factors. The time-series studies published since 1996 have
generally attempted to control for weather influences. In contrast, 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 function. Thus, the possibility of weather-related
influences in such studies cannot yet be discounted. Also, various co-pollutants have been
analyzed extensively in many recent time-series studies of PM and hospital admissions. In some
studies, certain PM indices clearly have an independent association after controlling for gaseous
co-pollutants. In others, the PM effects are reduced once co-pollutants are added to the model;
but this may be in part due to colinearity between PM indices and co-pollutants and/or gaseous
pollutants (e.g., CO) having independent effects on cardiovascular function.
Temporal patterns of responses following PM exposure. The evidence from recent time-
series studies of CVD admissions suggests rather strongly that PM effects tend to be maximal at
lag 0, with some carryover to lag 1, with little evidence for important effects beyond lag 1.
Relationship of CVD effects to PM size and chemical composition attributes. Insufficient
data exist from the time-series CVD admissions studies or the emerging individual-level studies
8-170
-------
to provide clear guidance as to which ambient PM components, defined on the basis of size or
composition, determine ambient PM CVD effect potency. The epidemiologic studies have been
constrained by limited availability of multiple PM metrics. Where multiple metrics exist, they
often are highly correlated or are of differential quality due to differences in numbers of
monitoring sites and monitoring frequency.
PM effects on blood characteristics related to CVD events. Interesting, though limited,
new evidence has also been derived which is highly suggestive of associations between ambient
PM indices and increased blood viscosity, increased serum C-reactive protein, and increased
blood fibrinogen (all biological markers related to increased risks of serious cardiac events).
However, much more research will be needed to order to both confirm such associations and to
better understand which specific ambient PM species may contribute to them.
8.3.2 Effects of Short-Term Particulate Matter Exposure on the Incidence of
Respiratory-Related Hospital Admissions and Medical Visits
8.3.2.1 Introduction
This section evaluates information on epidemiologic associations of ambient PM exposure
with both respiratory hospital admissions and medical visits. Although hospital admissions
represent one severe morbidity measure evaluated in regard to ambient PM exposures, hospital
emergency department (ED) visits are another notable related outcome. Doctors' visits also
represent yet another related health measure that, although less studied, is still very relevant to
assessing air pollution public health impacts. This category of pollution-affected persons can
represent a large population, one generally not evaluated due to the usual lack of centralized data
records for doctors' visits in the United States.
The section intercompares various studies examining size-related PM mass exposure
measures (e.g., for PM10, PM2 5, etc.) or various PM chemical components vis-a-vis their
associations with such health endpoints, and discusses their respective extents of coherence with
PM associations across related health effects measures. In the following discussion, the main
focus for quantitative intercomparisons is on U.S. and Canadian studies considering PM metrics
that measure mass or a specific mass constituent, i.e., PM10, PM10_2 5, PM2 5, or sulfates (SO42 ).
8-171
-------
Study results for other related PM metrics (e.g., BS) are also considered, but mostly only
qualitatively, primarily with respect to their relative coherence with studies using mass or
composition metrics measured in North America. In order to consider potentially confounding
effects of other coexisting pollutants, study results for various PM metrics are presented both for
(1) when the PM metric is the only pollutant in the model and (2) the case where a second
pollutant (e.g., O3) is also included. Results from models with more than two pollutants included
simultaneously, however, are not used here for quantitative estimates of effect size or statistical
strength, because of increased likelihood of bias and variance inflation due to multicolinearity of
various pollutants (e.g., see Harris, 1975).
8.3.2.2 Summary of Key Respiratory Hospital Admissions Findings from the 1996
Particulate Matter Air Quality Criteria Document
In the 1996 PM AQCD, both COPD and pneumonia hospitalization studies were found to
show moderate, but statistically significant, relative risks in the range of 1.06 to 1.25 (or 6 to
25% excess risk increment) per 50 |ig/m3 PM10 increase or its equivalent. Whereas many
hospitalizations for respiratory illnesses occur in those > 65 years of age, there were also
increased hospitalizations for those < 65 years of age. Several hospitalization studies restricted
their analysis by age group, but did not explicitly examine younger age groups. One exception
noted was Pope (1991), who reported increased hospitalization for Utah Valley children (0 to
5 years) for monthly numbers of admissions in relation to PM10 monthly averages, as opposed to
daily admissions in relation to daily PM levels used in other studies. Studies examining acute
associations between indicators of components of fine particles (e.g., BS; sulfates, SO42 ; and
acidic aerosols, H+) and hospital admissions were reported, too, as showing significant
relationships. While sulfates were especially predictive of respiratory health effects, it was not
clear whether the sulfate-related effects were attributable to their acidity, to the broader effects
of associated combustion-related fine particles, or to other factors.
8.3.2.3 New Respiratory-Related Hospital Admissions Studies
New studies appearing since the 1996 PM AQCD have examined various admissions
categories, including: total respiratory admissions for all ages and by age; asthma for all ages
8-172
-------
and by age; chronic obstructive pulmonary disease (COPD) admissions (usually for patients
> 64 years), and pneumonia admissions (for patients > 64 years). Table 8B-2, Appendix 8B
summarizes salient details regarding the study area, study period, study population, PM indices
considered and their concentrations, methods employed, study results, and "bottom-line" PM
index percent excess risks per standard PM increment (e.g., 50 |ig/m3 for PM10) for the newer
studies.
The percent excess risk (ER) estimates presented in Table 8B-2 are based upon the relative
risks (RRs) provided by the authors, but converted into percent increments per standardized
increments used by the U.S. EPA to facilitate direct intercomparisons of results across studies
(as discussed in Section 8.1). The ER's shown in the table are for the most positively significant
pollutant coefficient; and the maximum lag model is used to provide estimates of potential
pollutant-health effects associations.
Based on information from Dominici et al. (2002) indicating that the default convergence
criteria used in the S-Plus function GAM may not guarantee convergence to the best unbiased
estimate (as discussed earlier), only those studies that used other statistical algorithms or which
have reported reanalyzed S-Plus GAM results are assessed in the text below. However, given
the modest effects of such reanalyses on most study results (i.e., while effect estimates are
modified somewhat, the study conclusions remain largely unchanged), Table 8B-2 includes all
studies and notes those that originally used the S-Plus GAM algorithm as well as those studies
that have since been reanalyzed with more appropriate methods.
Of most pertinence here are those newly available studies that evaluate associations
between one or another ambient PM metric and respiratory hospital admissions in U.S. or
Canadian cities, such as those for PM10 mass concentrations summarized in Table 8-19.
Among numerous new epidemiologic studies of PM10 morbidity, many evaluated relatively
high PM10 levels. However, some did evaluate associations with PM10 concentrations ranging to
rather low levels. Of note is the fact that several investigators have reported associations
between acute PM10 exposures and total respiratory-related hospital admissions for numerous
U.S. cities with annual mean PM10 concentrations extending to below 50 |ig/m3. On this
account, the results of the NMMAPS multicity study (Samet et al., 2000a,b) of PM10 levels and
8-173
-------
TABLE 8-19. 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)
Lippmannetal. COPD PM10 = 45.4
(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 (2003a)
Lippmannetal. Pneumonia PM10 = 45.4
(2000)
Reanalysis by Pneumonia
Ito (2003)
Jacobs etal. (1997) Asthma PM10 = 34
Co-pollutants Day
Measured Lag
S03 —
S02-03-N02-CO 1
0-1
0-1
0-1
0-1
SO2- O3- NO2 3
CO-H+
3
— 0
CO
0
0
S02, 03, N02, 1
CO
0-1
0-1
0-1
0-1
S02, 03, N02, 3
CO, H+
3
03, CO —
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
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, 26.8)
Co Poll: 1.0 (-15, 20)
No Co Poll: 9.6 (-5.3, 26. 8)
No Co Poll: 6.5 (-7.8, 23.0)
No Co Poll: 4.6 (-9.4, 20. 8)
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(8.2,36.3)
Co Poll: 24(8.2,43)
No Co Poll: 21.5(8.3,36)
No Co-Poll: 18.1(5.3,32.5)
No Co-Poll: 18.6(5.6,33.1
6.11 (CI not reported)
Nauenberg and Asthma PM10 = 45
Basu(1999)
Tolbert et al. Asthma PM10 = 39
(2000b)
03
03, NOX
0 Poisson GLM
1 GEE
Sheppard et al. Asthma PM10 = 31 CO, O3, SO2 1 Default GAM
(1999)*
Reanalysis by Sheppard
(2003)
NSGLM
Strict GAM
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)
NS = Natural Spline General Linear Model; PS = Penalized Spline General Additive Model
8-174
-------
hospital admissions by persons > 65 years old in 14 U.S. cities are of particular interest.
As noted in Table 8-19, this study indicates PM10 effects similar to other cities, but with
narrower confidence bands, due to its greater power derived by combining multiple cities in the
same analysis. This allows significant associations to be identified, despite the fact that many of
the cities considered have relatively small populations and that each had mean PM10 below
50 |ig/m3. The cities considered and their respective annual mean/daily maximum PM10
concentrations (in |ig/m3) are Birmingham (34.8/124.8); Boulder (24.4/125.0); Canton
(28.4/94.8); Chicago (36.4/144.7); Colorado Springs (26.9/147.2); Detroit (36.8/133.6);
Minneapolis/St Paul (36.8/133.6); Nashville (31.6/128.0); New Haven (29.3/95.4); Pittsburgh
(36.0/139.3); Provo/Orem (38.9/241.0); Seattle (31.0/145.9); Spokane (45.3/605.8); and
Youngstown (33.1/104.0).
Table 8-20 also shows the results of reanalyzing a number of models considered in original
research with the use of models using more stringent convergence requirements than the original
default option. These reanalyses (Zanobetti and Schwartz, 2003a) show that the effect estimates
decline somewhat, but that the basic direction of effect and conclusions about the significance of
the PM effect on hospital admissions remained unchanged. In their reanalyses, Zanobetti and
Schwartz, (2003a) also considered spline models that are thought to better estimate confidence
intervals around pollutant effect estimates than the original GAM analyses. With the spline
models, confidence intervals usually increased over the original GAM model and the coefficients
also decreased somewhat (similar to GAM with more stringent convergence criteria). As for
possible co-pollutant confounding, it was reported that "In our previous studies we did not find
confounding due to other pollutants. These results are confirmed in this reanalysis by the meta-
regression analyses." Overall, the authors concluded that "the general result is that the
association of PM10 with hospital admissions remains and in most cases is little changed."
Janssen et al. (2002) did further analyses for the Samet et al. (2000a,b) 14-city data set
examining associations for variable prevalence in air-conditioning (AC) and/or contributions of
different sources to total PM10. For COPD and pneumonia, the associations were less
significant, but the pattern of association was similar to that for CVD. The Zanobetti and
8-175
-------
TABLE 8-20. 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)
% CVD
Increase (95% CI)
% COPD % Pneumonia
Increase (95% CI) Increase (95% CI)
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)
1.07 (0.93, 1.22)
0.68
1.17
0.99
1.47
1.32
(0.54, 0.81)
(1.01,1.33)
(0.79,1.19)
(1.18, 1.76)
(0.77, 1.87)
1.44 (1.00, 1.89) 1.57 (1.27, 1.87)
1.46 (1.03, 1.88) 1.31 (1.03, 1.58)
1.98 (1.49,2.47) 1.98 (1.65,2.31)
1.71 (0.95,2.48) 1.98 (1.65,2.31)
2.63 (1.71,3.55) 2.84 (2.21,3.48)
2.21 (1.02,3.41) 1.06 (0.06,2.07)
Original Quadratic
distributed lag
Reanalyzed Quadratic
distributed lag
Unconstrained distributed lag
Fixed effects estimate
Original Random effects
1.18
1.09
1.19
1.07
(0.96,
(0.81,
(0.97,
(0.67,
1.39)
1.38)
1.41)
1.46)
2.49
2.53
2.45
2.88
(1.78, 3.20)
(1.20, 3.88)
(1.75,3.17)
(0.19,5.64)
1.68
1.47
1.9
2.07
(1.25,2.11)
(0.86, 2.09)
(1.46, 2.34)
(0.94, 3.22)
estimate
Reanalyzed Random
effects estimate
1.12 (0.84, 1.40)
2.53 (1.21,3.87)
2.07 (0.94,3.22)
Source: Samet et al. (2000a,b) and Zanobetti and Schwartz (2003a) reanalyses.
Schwartz (2003b) reanalyses also examined these results, and they stated that "We still found a
decreased PM10 effect with increasing percentage of home with central AC."
Moolgavkar (2003) also reanalyzed his earlier GAM analyses of hospital admissions for
chronic obstructive pulmonary disease (Moolgavkar, 2000c) in Los Angeles (Los Angeles
County) and Chicago (Cook County). In his original publication, Moolgavkar found -5.0%
excess risk for COPD hospital admissions among the elderly (64+ years) in Los Angeles to be
significantly related to both PM2 5 and PM10_2 5 in one pollutant models; but the magnitudes of the
risk estimates dropped by more than half to non-statistically significant levels in two-pollutant
8-176
-------
models including CO. However, unlike the meta-regression approach to the multiple pollutant
issue used by Zanobetti and Schwartz (2003a), simultaneous regression of moderately to highly
correlated pollutants can lead to biased pollutant coefficients and commonly results in
diminished effect estimates for some or all of the pollutants considered. In the same study,
similar magnitudes of excess risk (i.e., in the range of ~4 to 7%) were found in one-pollutant
models to be associated with PM2 5 or PM10_25 for other age groups (0-19 years; 20-64 years) in
Los Angeles, as well.
In his reanalyses of these GAM results using the more stringent convergence criteria,
Moolgavkar (2003) combined all three Los Angeles age groups into one analysis, providing
greater power, but also complicating before/after comparisons as to the actual effect of using the
more stringent convergence criteria on the results. In the Cook County analyses, the author
changed other model parameters (i.e., the number of degrees of freedom in the model smooths)
at the same time as implementing more stringent convergence criteria; so direct before/after
comparisons are not possible for Moolgavkar's (2003) Chicago analyses. Moolgavkar noted that
"changes in the convergence criteria and the use of GLM instead of GAM can, but does not
always, have substantial impact on the results of the analyses and their interpretation." He also
concluded: "Given that different analytic strategies can make substantial differences to the
estimates of effects of individual pollutants, I do not believe that these numerical estimates are
too meaningful. Patterns of association appear to be robust, however. For example, in Los
Angeles, with the exception of COPD admissions with which NO2 appears to show the most
robust association, it is clear that CO is the best single index of air pollution associations with
health end points, far better than the mass concentration of either PM10 or of PM25. In Cook
County the results are not so clear-cut, however, any one of the gases is at least as good an index
of air pollution effects on human health as is PM10."
Tolbert et al. (2000b) used generalized estimating equations (GEE), logistic regression, and
Baysian models to evaluate associations between emergency department visits for asthma (by
those < 17 years old) in Atlanta during the summers of 1993 to 1995 (-6000 visits for asthma out
of-130,000 total visits) and several air pollution variables (PM10, O3, total oxides of nitrogen).
Logistic regression models controlling for temporal and demographic variables gave statistically
8-177
-------
significant (p < 0.05) lag 1 day relative risk estimates of 1.04 per 15 |ig/m3 24-h PM10 increment
and 1.04 per 20 ppb increase in maximum 8-h O3 levels. In multipollutant models including
both PM10 and O3, the terms for each became nonsignificant due to high collinearity of the two
variables (r2 = 0.75). The authors interpreted their findings as suggesting positive associations
between pediatric asthma visits and both PM10 and O3. The PM10 effects appeared to be stronger
for concentrations > 20 |ig/m3 than below that 24-h value.
Other U.S. studies finding associations of respiratory-related hospital admissions or
medical visits with PM10 levels extending below 50 |ig/m3 include: Schwartz (1994a) in
Minneapolis/St. Paul, Minnesota; Schwartz et al. (1996b) in Cleveland; Sheppard et al. (1999)
in Seattle; Linn et al. (2000) in Los Angeles; and Nauenberg and Basu (1999) in Los Angeles;
in Minneapolis/St. Paul, MN, but not in Birmingham, AL. The excess risk estimates most
consistently fall in the range of 5 to 25% per 50 |ig/m3 PM10 increment, with those for asthma
visits and hospital admissions often being higher than those for COPD and pneumonia
admissions.
Similar associations between increased respiratory related hospital admissions/medical
visits and low short-term PM10 levels were also reported by various investigators for several
non-U.S. cities. Wordley et al. (1997), for example, reported positive and significant
associations between PM10 (mean = 25.6 |ig/m3, max =131 |ig/m3) and respiratory admissions in
Birmingham, UK using multivariate linear regression methods; and Atkinson et al. (1999a),
using Poisson modeling, reported significant increases in hospital admissions for respiratory
disease to be associated with PM10 (mean = 28.5 |ig/m3) in London, UK. Hagen et al. (2000) and
Prescott et al. (1998) also found positive but nonsignificant associations of hospital admissions
and, PM10 levels in Drammen, Norway (mean = 16.8 |ig/m3) and Edinburgh, Scotland (mean =
20.7 |ig/m3). Admissions in Drammen considered relatively small populations, limiting
statistical power in this study. Petroeschevsky et al. (2001) examined associations between
outdoor air pollution and hospital admissions in Brisbane, Australia during 1987 to 1994 using a
light scattering index (BSP) for fine PM. The levels of PM are quite low in this city, relative to
most U.S. cities, but BSP was positively and significantly associated with total respiratory
admissions, but not for asthma.
8-178
-------
8.3.2.3.1 Particulate Matter Mass Fractions and Composition Comparisons
While PM10 mass has generally been the metric most often used as the particle pollution
index in the U.S. and Canada, some new studies have examined the relative roles of
various PM10 mass fractions (e.g., PM2 5 and PM10_2 5) and chemical constituents (such as SO42)
contributing to PM-respiratory hospital admissions associations. Several new studies (from
among those summarized in Tables 8-21 and 8-22, respectively) report significant associations
of increased respiratory-cause medical visits and/or hospital admissions with ambient PM25
and/or PM10_2 5 ranging to quite low concentrations. These include the Lippmann et al. (2000)
study in Detroit, where all PM metrics (PM10, PM2 5, PM10_2 5, H+) were positively related to
pneumonia and COPD admissions among the elderly (aged 65+ yr) in single pollutant models,
with their RR values for pneumonia generally remaining little changed (but with broader
confidence intervals) in multipollutant models including one or more gaseous pollutant (e.g.,
CO, O3, NO2, SO2). However, for COPD admissions, the effect estimates were reduced and
became nonsignificant in multipollutant models including gaseous co-pollutants. Excess risks
for pneumonia admissions in the one pollutant model using default GAM were 13% (CI: 3.7, 22)
and 12% (CI: -0.6, 24) per 25 |ig/m3 of PM25 and PM10.25, respectively; those for COPD
admissions were 5.5% (CI: -4.7, 17) and 9.3% (CI: -4.2, 25) per 25 |ig/m3 PM2 5 and PM10.2 5,
respectively.
Lippmann et al. (2000) reported weaker associations with sulfate and acidic components
of PM2 5 than with PM25 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. (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."
In order to evaluate the potential influence of the GAM convergence specification on the
results of the original Detroit data analysis, Ito (2003) re-examined associations between PM
8-179
-------
TABLE 8-21. SUMMARY OF UNITED STATES PM2 s RESPIRATORY-RELATED
HOSPITAL ADMISSION STUDIES
Reference
Outcome
Measures
Mean Levels Co-Pollutants
u,g/m3 Measured Lag
Method
Effect Estimate (95% CL)
(% increase per 25 ug/m3)
Lippmann et al.
(2000)
COPD
Reanalysis by COPD
Ito (2003)
Moolgavkar
(2000c)*
COPD
(> 64 yrs)
(median)
Reanalysis by COPD
Moolgavkar (all ages)
(2003)
Lippmann et al. Pneumonia
(2000)
Reanalysis by Pneumonia
Ito (2003)
Sheppard et al. Asthma
(1999)*
Reanalysis by
Sheppard (2003)
Freidman et al. Asthma
(2001)
PM25= 18
SO2, O3, NO2,
CO,H+
PM25 = 22,LA
PM2 5 = 22, LA
CO
PM25=18 S02,03,N02.
CO,H+
PM25 = 16.7 CO, O3, SO2
CO
PM2 5 = 36.7- 03
30.8
3 Default GAM
3 Default GAM
Default GAM
Strict GAM
NSGLM
2 Default GAM
2 Default GAM
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
Strict GAM: 30df 4.69(2.06,7.38)
Strict GAM: lOOdf 2.87(0.53,5.27)
NSGLM: lOOdf 2.59 (-0.29, 5.56)
1 Default GAM
1 Default GAM
Default GAM
Strict GAM
NSGLM
1 Default GAM
Default GAM
Strict GAM
NSGLM
Strict GAM
NSGLM
3 Poisson GEE
day
cum
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.
components and daily mortality/morbidity by using more stringent GAM convergence criteria
and by applying GLM analyses 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 the maximum iteration was
set to 1000 for both the local scoring and back-fitting algorithms. The GLM model specification
approximated the original GAM models. Natural splines were used for smoothing terms.
To model time trend, the same degrees of freedom as the smoothing splines in the GAM models
8-180
-------
TABLE 8-22. SUMMARY OF UNITED STATES PM102 s RESPIRATORY-RELATED
HOSPITAL ADMISSION STUDIES
Reference
Moolgavkar
(2000c )*
Lippmann
et al. (2000)*
Reanalysis by
Lippmann
et al. (2000)*
Reanalysis by
Sheppard
etal. (1999)*
Reanalysis by
(2003)
Outcome Mean Levels Co-Pollutants
Measures Mg/ni3 Measured
COPD —
COPD PM10.2 5 = 12 S02, 03, N02,
CO,H+
Ito (2003)
Pneumonia PM; 0.2 5 = 1 2 SO2, O3 , NO2,
CO,H+
Ito (2003)
Asthma PM10.25 = 16.2 CO, O3, SO2
Sheppard
Lag
3
3
3
1
1
1
1
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
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)
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.
were used, with the default placement of knots. For weather models, to approximate LOESS
smoothing with a span of 0.5 in the GAM model, natural splines with degrees of freedom were
used. Generally, the GAM models with stringent convergence criteria and GLM models resulted
in somewhat smaller estimated relative risks than those reported in the original study, e.g., for
pneumonia admissions in Table 8-23. It was found that the reductions in the estimated relative
risks were not different across the PM indices. Thus, conclusions of the original study about the
relative roles of PM components by size and chemical characteristics remained unaffected.
Lumley and Heagerty (1999) illustrate the effect of reliable variance estimation on data from
hospital admissions for respiratory disease on King County, WA for eight years (1987-94),
together with air pollution and weather information, using estimating equations and weighted
empirical variance estimators. However, their weather controls were relatively crude (i.e.,
8-181
-------
TABLE 8-23. 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)
PM10.2.5 (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).
seasonal dummy variables and linear temperature terms). This study is notable for having
compared sub-micron PM (PMX 0) versus coarse PM1(M 0 and for finding significant hospital
admission associations only with PMX 0. This may suggest that the PM2 5 versus PM10 separation
may not always be sufficient to differentiate submicron fine particle versus coarse-particle
toxicities.
Asthma hospital admission studies in various U.S. communities provide additional
important new data. Of particular note is a study by Sheppard et al. (1999) which evaluated
relationships between measured ambient pollutants (PM10, PM2 5, PM10_2 5, SO2, O3, and CO) and
nonelderly adult (< 65 years of age) hospital admissions for asthma in Seattle, WA. PM and CO
were found to be jointly associated with asthma admissions. The authors noted ". . . .we
observed unexpected associations for CO that dominated the PM effects. Nevertheless, although
there is substantial literature on the effects of CO on the cardiovascular system, there is no
evidence for an effect on the underlying physiology of asthma. CO may be an important
environmental indicator of incomplete combustion, particularly from mobile sources."
8-182
-------
An estimated 4 to 5% increase in the rate of asthma hospital admissions (lagged 1 day) was
reported to be associated with interquartile range changes in PM indices (19 |ig/m3 for PM10,
11.8 |ig/m3 for PM25, and 9.3 |ig/m3 for PM10_25), equivalent to excess risk rates as follows: 13%
(CI: 0.5-23) per 50 |ig/m3 for PM10; 9% (CI: 3-14) per 25 |ig/m3 PM25; 11% (CI: 3-20) per
25 |ig/m3 PM10_25. Also of note for the same region are analyses by the same research team
using similar methods (Norris et al., 1999) which showed associations of low levels of PM25
(mean = 12 |ig/m3) with markedly increased asthma ED, i.e., excess risk = 44.5% (CI: 21.7-71.4)
per 25 |ig/m3 PM2 5.
Sheppard (2003) recently conducted a reanalysis of their nonelderly hospital admissions
data for asthma in Seattle, WA, to evaluate the effect of the fitting procedure on their previously
published analyses. As shown in Figure 8-11, the effect estimates were slightly smaller when
more stringent convergence criteria were used with GAM, and there was an additional small
reduction in the estimates when GLM with natural splines were used instead. The average
reduction in effect estimate between the default and stringent convergence criteria for PM2 5,
PM10, and PM10_25 (coarse) mass averaged 10.7%. The coefficients remained statistically
significant for both PM2 5 and PM10 but not for coarse mass. Confidence intervals were slightly
wider for the GLM model fit. Sheppard concluded that,
"Overall the results did not change meaningfully. There were small reductions in
estimates using the alternate fitting procedures. I also found that the effect of single
imputation (i.e., not adjusting for replacing missing exposure data with an estimate of
its expected value) was to bias the effect estimates slightly upward. In this data set this
bias is of the same order as the bias from using too liberal convergence criteria in the
generalized additive model."
Moolgavkar (2003) also conducted reanalyses of respiratory-related hospital admissions,
but for COPD data for all ages in Los Angeles. Using GAM with strict convergence criteria and
30 degrees of freedom (df), an excess risk estimate of 4.7% (CI: 2.1, 7.4) was obtained per
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.
8-183
-------
CD
W
(0
O o
C CM
OL
a
Q.
CD
to
to
E
T3
0)
O)
(0
.C
o
d>
O)
d>
Ł*
CD
Q.
4
i
i 4
!
i
'!
>
PM10
Lag1
4
^
i
+
4
i
4
~
PM2.5
Lag1
4
:
• 4
:
i
>:
: 4
*!
i
Coarse
Mass
Lag1
<
4
I
I
I
c
Carbon
Monoxide
Lag 3
i
i
i
Jo or Multiple Imputation
Single Imputation
>
<
• <
Sulfur
Dioxide
LagO
4
f <
>
Ozone
Lag 2
Figure 8-11. 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-rff 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).
Burnett et al. (1997a) evaluated the role that the ambient air pollution mix, comprised of
gaseous pollutants and PM indexed by various physical and chemical measures, plays in
exacerbating daily admissions to hospitals for cardiac diseases and for respiratory diseases
8-184
-------
(tracheobronchitis, chronic obstructive lung disease, asthma, and pneumonia). They employed
daily measures of PM25 and PM10_25, aerosol chemistry (sulfates and H+), and gaseous pollutants
(O3, NO2, SO2, CO) collected in Toronto, Ontario, Canada, during the summers of 1992, 1993,
and 1994. Positive associations were observed for all ambient air pollutants for both respiratory
and cardiac diseases. Ozone was the most consistently significant pollutant and least sensitive to
adjustment for other gaseous and particulate measures. The PM associations with respiratory
hospital admissions were significant for: PM10 (RR =1.11 for 50 |ig/m3; CI: 1.05, 1.17); PM2 5
(fine) mass (RR = 1.09 for 25 |ig/m3; CI: 1.03, 1.14);PM10.25 (coarse) mass (RR =1.13 for
25 |ig/m3; CI: 1.05, 1.20); sulfate levels (RR= 1.11 for 155 nmoles/m3 = 15 |ig/m3;
CI: 1.06, 1.17); and H+(RR= 1.40 for 75 nmoles/m3 = 3.6 |ig/m3, as H2SO4;CI: 1.15, 1.70).
After inclusion of O3 in the model, the associations with the respiratory hospital admissions
remained significant for: PM10 (RR= 1.10, CI: 1.04, 1.16); fine mass (RR= 1.06;
CI: 1.01, 1.12); coarse mass (RR= 1.11; CI: 1.04, 1.19); sulfate levels (RR= 1.06; CI: 1.0,
1.12); and H+ (RR = 1.25; CI: 1.03, 1.53), using the same increments. Of the PM metrics
considered here, H+ yielded the highest RR estimate. Regression models that included all
recorded pollutant simultaneously (with high intercorrelations among the pollutants) were also
presented.
A recent study by Lin et al. (2002) used both case-crossover and time-series analyses to
assess the associations between size-fractionated PM and asthma hospitalization among children
6 to 12 years old living in Toronto between 1981 and 1993. The authors used exposures
averaged over periods varying from 1 to 7 days to assess the PM effects on asthma
hospitalization. Estimates of the relative risk of asthma hospitalization were adjusted for daily
weather conditions (maximum and minimum temperatures, and average relative humidity) for an
incremental exposure corresponding to the PM interquartile range. However, direct
measurements of PM components were available only every sixth day in this data set, and 5 out
of every 6 PM data points in the analysis were based on estimated PM25, PM25.10, and PM10 data,
weakening confidence in these input data. Time-series plots of the PM25.10 data showed much
stronger seasonality in the estimated coarse PM data than in the estimated fine PM mass data.
Seasonality was controlled for in the time-series analyses using a 3 month span smooth of the
8-185
-------
data, rather than the more commonly employed one month or less span. Thus, residual
seasonality may have been a factor in this study's PM2 5_10 results. Both bidirectional case-
crossover and time-series analyses revealed that coarse PM (PM10_2 5) averaged over 5-6 days
was significantly associated with asthma hospitalization in both males and females. The
magnitude of this effect appeared to increase with increasing number of days of exposure
averaging for most models, with the relative risk estimates stabilizing at about 6 days. Using a
bidirectional case-crossover analysis, the estimated relative risks were 1.14 (CI: 1.02, 1.28) for
males and 1.18 (CI: 1.02, 1.36) for females, for an increment of 8.4 |ig/m3 in 6-day averages
of PM10_25. The corresponding relative risk estimates were 1.10 and 1.18, respectively, from the
time-series analysis. The effect of PM10_2 5 remained positive after adjustment for the effects of
gaseous pollutants CO, NO2, SO2, and O3. They did not find significant effects of fine PM
(PM2 5) or of thoracic PM (PM10) on asthma hospitalizations, except in the unidirectional case-
cross-over analyses. Seasonal-specific results were not presented. The paper's discussion
ignores previous results by Thurston et al. (1994), which provided results during summers in the
same time range (1986 to 1988) that are in direct conflict with respect to the significance
of PM2 5. That study used daily direct measurements of size fractionated PM in analysis for
those three summers and found significant effects for summertime PM25. Seasonality of data
analysis may therefore be a factor in the differences between these two Toronto hospital
admissions studies regarding the health effects of fine PM. Overall, this new study suggests that
coarse particle mass can also be a risk factor in children's asthma hospital admissions.
There have also been numerous new time-series studies examining associations between
air pollution and respiratory-related hospital admissions in Europe, as summarized in
Appendix 8B, Table 8B-2, but most of these studies relied primarily on black smoke (BS) as
their PM metric. BS is a particle reflectance measure that provides an indicator of PM blackness
and is highly correlated with airborne carbonaceous particle concentrations (Bailey and Clayton,
1982). In the U.S., Coefficient of Haze (CoH) is a metric of particle transmittance that similarly
most directly represents a metric of particle blackness and ambient elemental carbon levels
(Wolff et al., 1983) and has been found to be highly correlated with BS (r = 0.9; Lee et al.,
1972). However, the relationship between airborne carbon and total mass of overall aerosol
8-186
-------
(PM) composition varies over time and from locality to locality, so the BS-mass ratio is less
reliable than the BS-carbon relationship (Bailey and Clayton, 1982). This means that the
BS-mass relationship is likely to be very different between Europe and the United States, largely
due to differences in local PM source characteristics (e.g., percentages of diesel powered motor
vehicles). Therefore, while these European BS-health effects studies may be of qualitative
interest for evaluating the PM-health effects associations, they are not as useful for quantitative
assessment of PM effects relevant to the United States.
Probably the most extensive and useful recent European air pollution health effects
analyses have been conducted as part of the APHEA multicity study, which evaluated
15 European cities from 10 different countries with a total population of over 25 million.
All studies used a standardized data collection and analysis approach, which included
consideration of the same suite of air pollutants (BS, SO2, NO2, SO2, and O3) and the use of time-
series regression addressing seasonal and other long-term patterns; influenza epidemics; day of
the week; holidays; weather; and autocorrelation (Katsouyanni et al., 1996). The general
coherence of the APHEA results with other results gained under different conditions strengthens
the argument for causality in the air pollution-health effects association. In earlier studies, the
general use of the less comparable suspended particle (SPM) measures and BS as PM indicators
in some of the APHEA locations and analyses lessens the quantitative usefulness of such
analyses in evaluating associations between PM and health effects most pertinent to the U.S.
situation. However, Atkinson et al. (2001) report results of PM10 analyses in a study of eight
APHEA cities.
As for other single-city European studies of potential interest here, Hagen et al. (2000)
compared the association of PM10 and co-pollutants with hospital admissions for respiratory
causes in Drammen, Norway during 1994 to 1997. Respiratory admissions averaged only
2.2 per day; so, the power of this analysis is weaker than studies looking at larger populations
and longer time periods. The NMMAPS modeling approach was employed. While a significant
association was found for PM10 as a single pollutant, it became nonsignificant in multiple
pollutant models. In two pollutant models, the associations and effect size of pollutants were
generally diminished, and when all eight pollutants were considered in the model, all pollutants
8-187
-------
became nonsignificant. These results are typical of the problems of analyzing and interpreting
the coefficients of multiple pollutant models when the pollutants are even moderately
intercorrelated over time. A unique aspect of this work was that benzene was considered in this
community to be strongly affected by traffic pollution. In two pollutant models, benzene was
most consistently still associated. The authors conclude that PM is mainly an 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 noted above, medical visits include both hospital ED and doctors' office visits. As in
the past, most newly available morbidity studies in Table 8B-3, Appendix 8B and in Table 8-24
below are of ED visits and their associations with air pollution. These studies collectively
confirm the results provided in the 1996 PM AQCD, indicating a positive and generally
statistically significant association between ambient PM levels and increased respiratory-related
hospital visits.
Of the medical visit and hospital admissions studies since the 1996 PM AQCD, among the
most informative are those that evaluate health effects at relatively low PM concentrations.
As for U.S. studies, Tolbert et al. (2000b) reported a significant PM10 association with pediatric
ED visits in Atlanta where mean PM10 = 39 |ig/m3 and maximum PM10 =105 |ig/m3. The Lipsett
et al. (1997) study of winter air pollution and asthma emergency visits in Santa Clara Co, CA,
may provide insight where one of the principal sources of PM10 is residential wood combustion
(RWC). Their results demonstrate an association between PM10 levels and asthma. Also of
interest, Delfmo et al. (1997a) found significant PM10 and PM2 5 associations for respiratory ED
visits among older adults in Montreal when mean PM10=21.7 |ig/m3 and mean PM25 =
12.2 |ig/m3. Hajat et al. (1999) also reported significant PM10 associations with asthma doctor's
visits for children and young adults in London when mean PM10 = 28.2 |ig/m3 and the PM10 90th
percentile was only 46.4 |ig/m3. Overall, then, several new medical visits studies indicate
8-188
-------
TABLE 8-24. SUMMARY OF UNITED STATES PM10, PM25, AND PM1025 ASTHMA
MEDICAL VISIT STUDIES
Reference
PM10
Choudhuryetal. (1997)
Lipsett et al.
Tolbert et al.
Tolbert et al.
(1997)
(2000b)
(2000a)**
Outcome
Measures
Asthma
Asthma
Asthma
Asthma
Mean
Levels
Oig/m3)
41.5
61.2
38.9
29.1
Co-Pollutants
Measured
Not considered
NO2, O3
03
NO2, O3, CO, SO2
Lag
0
2
1
0-2
Method
GLM
GLM
GEE
GLM
Effect Estimate*
(95% CL)
20.9(11.8,30.8)
34.7 (16, 56.5)
at 20 °C
13.2(1.2,26.7)
8.8 (-8.7, 54.4)
PM25
Tolbert etal. (2000a)** Asthma
19.4 NO2, 03, CO, SO2 0-2 GLM 2.3 (-14.8, 22.7)
Tolbert etal. (2000a)** Asthma 9.39 NO2, 03, CO, SO2 0-2 GLM 21.1 (-18.2, 79.3)
* Effect estimates derived from single-pollutant models.
**Preriminary results based on emergency department visit data from 18 of 33 participating hospitals.
PM-health effects associations at lower PM2 5 and PM10 levels than those from previous
publications.
8.3.2.4.1 Scope of Medical Visit Morbidity Effects
Several newer medical visit studies consider a new endpoint for comparison with ED
visits: visits in the primary care setting. In particular, key studies showing PM associations for
this health outcome include: the study by Hajat et al. (1999) that evaluated the relationship
between air pollution in London, UK; and daily General Practice (GP) doctor consultations for
asthma and other lower respiratory disease (LRD); the study by Choudhury et al. (1997) of
private asthma medical visits in Anchorage, Alaska; and the study by Ostro et al. (1999b) of
daily visits by young children to primary care health clinics in Santiago, Chile for upper or lower
respiratory symptoms.
8-189
-------
While limited in number, the above studies collectively provide new insight into the fact
that there is a broader scope of morbidity associated with PM air pollution exposure than
previously documented. As the authors of the London study note: "There is less information
about the effects of air pollution in general practice consultations but, if they do exist, the public
health impact could be considerable because of their large numbers." Indeed, the London study
of doctors' GP office visits indicates that the effects of air pollution, including PM, can affect
many more people than indicated by hospital admissions alone.
These new studies also provide indications as to the quantitative nature of medical visits
effects, relative to those for hospital admissions. In the London case, comparing the number of
admissions from the authors' earlier study (Anderson et al., 1996) with those for GP visits in the
1999 study (Hajat et al., 1999) indicates that there are -24 asthma GP visits for every asthma
hospital admission in that city. Also, comparing the PM10 coefficients indicates that the all-ages
asthma effect size for the GP visits (although not statistically different) was about 30% larger
than that for hospital admissions. Thus, these new studies suggest that looking at only hospital
admissions and emergency hospital visit effects may greatly underestimate the overall numbers
of respiratory morbidity events due to acute ambient PM exposure.
8.3.2.4.2 Factors Potentially Affecting Respiratory Medical Visit Study Outcomes
Some newly available studies have examined certain factors that might extraneously affect
the outcomes of PM-medical visit studies. Stieb et al. (1998a) examined the occurrence of bias
and random variability in diagnostic classification of air pollution and daily cardiac or
respiratory ED visits, such as for asthma, COPD, respiratory infection, etc. They concluded that
there was no evidence of diagnostic bias in relation to daily air pollution levels. Also, Stieb et al.
(1998b) reported that for a population of adults visiting an emergency department with
cardiorespiratory disease, fixed site sulfate monitors appear to accurately reflect daily variability
in average personal exposure to particulate sulfate, whereas acid exposure was not as well
represented by fixed site monitors. Another study investigated possible confounding of
respiratory visit effects due to pollens and mold spores (Steib et al, 2000). Aeroallergen levels
did not influence the results, similar to asthma panel studies described below in Section 8.3.3.
8-190
-------
In London, Atkinson et al. (1999a) studied the association between the number of daily ED visit
for respiratory complaints and measures of outdoor air pollution for PM10, NO2, SO2 and CO.
They examined different age groups and reported strongest associations for children for visits for
asthma, but were unable to separate PM10 and SO2 effects.
8.3.2.5 Identification of Potential Susceptible Subpopulations
Associations between ambient PM measures and respiratory admissions have been found
for all age groups, but older adults and children generally have been indicated by hospital
admissions studies to exhibit the most consistent PM-health effects associations. As reported in
previous PM AQCDs, numerous studies of older adults (e.g., those 65+ years of age) have
related acute PM exposure with an increased incidence of hospital admissions (e.g., see
Anderson et al, 1998). However, only a limited number have specifically studied children as a
subgroup. Burnett et al. (1994) examined the differences in air pollution-hospital admissions
associations as a function of age in Ontario, reporting that the largest percentage increase in
admissions was found among infants (neonatal and postneonatal, one year or less in age).
Further efforts have aimed at identifying and quantifying air pollution effects among
potentially especially susceptible subpopulations of the general public. Some new studies have
further investigated the hypothesis that the elderly are especially affected by air pollution.
Zanobetti et al. (2000a) examined PM10 associations with hospital admissions for heart and lung
disease in ten U.S. cities, finding an overall association for COPD, pneumonia, and CVD. They
found that these results were not significantly modified by poverty rate or minority status in this
population of Medicare patients. Ye et al. (2001) examined emergency transports to the hospital.
Both PM10 and N02 levels were significantly associated with daily hospital transports for angina,
cardiac insufficiency, myocardial infarction, acute and chronic bronchitis, and pneumonia. The
pollutant effect sizes were generally found to be greater in men than in women, except those for
angina and acute bronchitis, which were the same across genders. Thus, in these various studies,
cardiopulmonary hospital visits and admissions among the elderly were seen to be consistently
associated with PM levels across numerous locales in the U.S. and abroad, generally without
regard to race or income; but sex was sometimes an effect modifier.
8-191
-------
Several new studies of children's morbidity also support the indication of air pollution
effects among children. Pless-Mulloli et al. (2000) evaluated children's respiratory health and
air pollution near opencast coal mining sites in a cohort of nearly 5,000 children aged 1 to
11 years in England. Mean PM levels were not high (mean < 20 |ig/m3 PM10), but statistically
significant PM10 associations were found with respiratory symptoms. A roughly 5% increase of
General Practitioner medical visits was also noted, but was not significant. Ilabaca et al. (1999)
also found an association between levels of fine PM and ED visits for pneumonia and other
respiratory illnesses among children < 15 years old in Santiago, Chile, where the levels of PM25
were very high (mean = 71.3 |ig/m3) during 1995 to 1996. The authors found it difficult to
separate out the effects of various pollutants, but concluded that PM (especially the fine
component) is associated with the risk of these respiratory illnesses. Overall, these new studies
support past assertions that children, and especially neo-natal infants, are especially susceptible
to the health effects of air pollution.
The respiratory-related hospital admissions studies summarized in Appendix 8B reveal that
the PM RR's for all children (e.g., 0 to 18 years old) are not often notably larger than those for
adults, but such comparisons of RR's must adjust for differences in baseline risks for each group.
For example, if hospital admissions per 100,000 per day for young children are double the rate
for adults, then they will have a pollution relative risk (RR) per |ig/m3 that is half that of the
adults given the exact same impact on admissions/100,0007|ig/m3/day. Thus, it is important to
adjust RR's or Excess Risks (ER's) for each different age groups' baseline, but this information
is usually not available (especially regarding the population catchment for each age group in
each study). One of the few indications that is notable when comparing children with other age
group effect estimates in Table 8B-2 is the higher excess risk estimate for infants (i.e., the group
< 1 year of age) in the Gouveia and Fletcher (2000) study, an age group that has estimated risk
estimate roughly twice as large as for other children or adults.
8.3.2.6 Summary of Salient Findings on Acute Particulate Matter Exposure and
Respiratory-Related Hospital Admissions and Medical Visits
The results of new studies discussed above are generally consistent with and supportive of
findings presented in the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996a),
8-192
-------
with regard to ambient PM associations of short-term exposures with respiratory-related hospital
admissions/medical visits. Figure 8-12 summarizes results for maximum excess risk of
respiratory-related hospital admission and visits per 50 |ig/m3 PM10 based on single-pollutant
models for selected U.S. cities. The excess risk estimates fall most consistently in the range of
5 to 20% per 50 |ig/m3 PM10 increments, with those for asthma visits and hospital admissions
generally somewhat higher than for COPD and pneumonia hospital admissions. More limited
new evidence both (a) substantiates increased risk of respiratory-related hospital admissions due
to ambient fine particles (PM2 5, PMLO, etc.) and also (b) points towards such admissions being
associated with ambient coarse particles (PM10_2.5). Excess risk estimates tend to fall in the range
of-5.0 to 15.0% per 25 |ig/m3 PM25 or PM10_25 for overall respiratory admissions or for COPD
admissions, whereas larger estimates are found for asthma admissions.
Tolbertetal. (2000b)
Atlanta
Choudhury et al. (1997)
Anchorage
Sheppard (2003)
Seattle
Nauenberg and Basu (1999)
LA, CA
Zanobetti and Schwartz (2003b)
14 US Cities
Moolgavkar (2003)
Chicago
Moolgavkar (2003)
LA
Ito (2003)
Detroit
Zanobetti and Schwartz (2003b)
14 US Cities
Ito (2003)
Detroit
Asthma Visits
-10 -5
Asthma Hospital Admissions
COPD Hospital Admissions
Pneumonia Hospital Admissions
I I I I I
10 15 20 25 30 35
Figure 8-12. Maximum excess risk of respiratory-related hospital admissions and visits
per 50 ug/m3 PM10 increment in studies of U.S. cities based on single-
pollutant models.
8-193
-------
Various new medical visits studies (including nonhospital physician visits) indicate that the
use of hospital admissions alone can greatly understate the total clinical morbidity effects of air
pollution. Thus, these results support the hypothesis that considering only hospital admissions
and ED visit effects may greatly underestimate the numbers of medical visits occurring in a
population as a result of acute ambient PM exposure. Those groups identified in these morbidity
studies as most strongly affected by PM air pollution are older adults and the very young.
8.3.3 Effects of Particulate Matter Exposure on Lung Function and
Respiratory Symptoms
In the 1996 PM AQCD, the available respiratory studies used a wide variety of designs
examining pulmonary function and respiratory symptoms in relation to ambient concentrations
of PM10. The populations studied included several different subgroups (e.g., children,
asthmatics, etc.); and the models used for analysis varied, but did not include GAM use. The
pulmonary function studies were suggestive of short-term effects resulting from ambient PM
exposure. Peak expiratory flow rates showed decreases in the range of 2 to 5 1/min per 50 |ig/m3
increase in 24-h PM10 or its equivalent, with somewhat larger effects in symptomatic groups,
e.g., asthmatics. Studies using FEVj or FVC as endpoints showed less consistent effects. The
chronic pulmonary function studies, less numerous than the acute studies, had inconclusive
results.
8.3.3.1 Effects of Short-Term Particulate Matter Exposure on Lung Function and
Respiratory Symptoms
The available acute respiratory symptom studies discussed in the 1996 PM AQCD included
several different endpoints, but typically presented results for upper respiratory symptoms, lower
respiratory symptoms, or cough. These respiratory symptom endpoints had similar general
patterns of results. The odds ratios were generally positive, the 95% confidence intervals for
about half of the studies being statistically significant (i.e., the lower bound exceeded 1.0).
The earlier studies of morbidity health outcomes of PM exposure on asthmatics were
limited in terms of conclusions that could be drawn because of the few available studies on
asthmatic subjects. Lebowitz et al. (1987) reported a relationship with TSP exposure and
8-194
-------
productive cough in a panel of 22 asthmatics but not for peak flow or wheeze. Pope et al. (1991)
reported on respiratory symptoms in two panels of Utah Valley asthmatics. The 34 asthmatic
school children panel yielded estimated odd ratios of 1.28 (1.06, 1.56) for lower respiratory
illness (LRI) and the second panel of 21 subjects aged 8 to 72 years for LRI of 1.01 (0.81, 1.27)
for exposure to PM10. Ostro et al. (1991) reported no association for PM2 5 exposure in a panel of
207 adult asthmatics in Denver; but, for a panel of 83 asthmatic children age 7 to 12 years in
central Los Angeles, found a relationship of shortness of breath to O3 and PM10, but could not
separate effects of the two pollutants (Ostro et al., 1995). These few studies did not indicate a
consistent relationship for PM10 exposure and health outcome in asthmatics.
Numerous new studies of short-term PM exposure effects on lung function and respiratory
symptoms published since 1996 were identified by an ongoing Medline search. Most of these
followed a panel of subjects over one or more time periods and evaluated daily lung function
and/or respiratory symptom in relation to changes in ambient PM10, PM10_2 5, and/or PM2 5. Some
used other measures of airborne particles, e.g., ultrafme PM, TSP, BS, and sulfate fraction of
ambient PM. Lung function was usually measured daily, with most studies including forced
expiratory volume (FEV), forced vital capacity (FVC) and peak expiratory flow rate (PEF),
measured both in the morning and afternoon. Various respiratory symptoms were measured,
e.g., cough, phlegm, difficulty breathing, wheeze, and bronchodilator use. Detailed summaries
of these studies are presented in Appendix 8B. Data on physical and chemical aspects of
ambient PM levels (especially for PM10, PM10_2 5, PM2 5, and smaller size fractions) are of
particular interest, as are new studies examining health outcome effects and/or exposure
measures not much studied in the past.
Specific studies were selected for summarization based on the following criteria:
• Peak flow was used as the primary lung function measurement of interest.
• Cough, phlegm, difficulty breathing, wheeze, and bronchodilator use were summarized as
measures of respiratory symptoms when available.
• Quantitative relationships were estimated using PM10, PM2 5, PM10_2 5, and/or smaller PM
as independent variables.
• Analyses used in the study were done such that each individual served as their
own control.
8-195
-------
8.3.3.1.1 Lung Function and Respiratory Symptom Effects in Asthmatic Subjects
Appendix B Tables 8B-4 and 8B-5 summarize salient features of new studies of short-term
PM exposure effects on lung function and respiratory symptoms, respectively, in asthmatic
subjects; and key quantitative results are summarized in Table 8-25 for PM10 and Table 8-26
for PM2 5. The peak flow analyses results for asthmatics tend to show small decrements for PM10
and PM25 as seen in studies by Gielen et al. (1997), Peters et al. (1997c), Romieu et al. (1997),
and Pekkanen et al. (1997).
For PM10, the available point estimates for morning PEF lagged one day showed decreases,
but the majority of the studies were not statistically significant (as per Table 8-25 and as shown
in Figure 8-13 as an example of PEF outcomes). Lag 1 may be more relevant for morning
measurement of asthma outcome from the previous day. The figure presents studies which
provided such data. The results were consistent for both AM and PM peak flow analyses.
Effects using two- to five-day lags averaged about the same as did the zero to one-day lags, but
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. (1997b) 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. The authors reported morning PEF changes of -0.970 (SE = 0.502) l(cm3) for particle
number count (0.032-0.10 in size), -0.901 (0.536) for PM10.32, and-1.13 (SE = 0.478)
for PM10. 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);
and PM10: -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, FEVj, and spirometric PEFR
were inversely, but mostly nonsignificantly-associated with ultrafme 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
8-196
-------
TABLE 8-25. SUMMARY OF QUANTITATIVE PFT CHANGES IN ASTHMATICS PER 50 jig/m3 PM10 INCREMENT
00
Reference Citation, Location, etc.
Outcome Measure
Mean Particulate
Levels (Range) u,g/m3
Co-Pollutants
Measured
Lag Structure
Effect Measures Standardized
to 50 ug/m3 PM10
Asthma Studies
Pekkanen et al. (1997)
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Peters etal. (1997a)
Peters etal. (1997b)
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Pekkanen et al. (1997)
Peters etal. (1996)
Peters etal. (1997a)
Peters etal. (1997b)
Timonen & Pekkanen (1997) Urban
Timonen & Pekkanen (1997)
Suburban
Gielenetal. (1997)
Romieuetal. (1996)
Romieuetal. (1997)
Segalaetal. (1998)
Pekkanen et al. (1997)
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
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)
NO2
Ozone
Ozone
Ozone
SO2, sulfate, FT
SO2, sulfate, FT
Ozone
Ozone
Ozone
Ozone
Ozone
Ozone
NO2
SO2, sulfate, PSA
S02, sulfate, FT
S02, sulfate, FT
NO2, SO2
NO2, SO2
Ozone
Ozone
Ozone
SO2, NO2
NO2
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
-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)
-------
TABLE 8-25 (cont'd). SUMMARY OF QUANTITATIVE PFT CHANGES IN ASTHMATICS
PER 50 ng/m3 PM10 INCREMENT
00
oo
Reference Citation, Location, etc.
Outcome Measure
Mean Particulate
Levels (Range) u,g/m3
Co-Pollutants
Measured
Lag Structure
Effect Measures Standardized
to 50 ug/m3 PM10
Asthma Studies (cont'd)
Peters etal. (1997b)
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 et al. (1998)
Hiltermann etal. (1998)
Vedal etal. (1998)
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 8-26. SUMMARY OF PFT
Reference Citation, L:ocation, etc.
Romieuetal. (1996)
Peters etal. (1997b)
Romieuetal. (1996)
Peters etal. (1997c)
Romieuetal. (1996)
Peters etal. (1997b)
Romieuetal. (1996)
Peters etal. (1997b)
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)
S02, sulfate, FT
N02, S02
NO2, SO2
SO2, sulfate, PSA
SO2, sulfate, FT
N02, S02
N02, S02
Ozone, NO2, SO2
Ozone, NO2, SO2
Ozone, NO2, SO2
None
CHANGES IN ASTHMATICS PER
Mean Particulate Levels
Outcome Measure (Range) u,g/m3
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
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, FT
Ozone
SO2, sulfate, FT
Ozone
S02, sulfate, FT
Ozone
SO2, sulfate, FT
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
25 jig/m3 PM2 5
Lag Structure
1 day
1 day
2 day
1-5 day
Oday
Oday
2 day
1-5 day
-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 PM2 5
-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)
-------
_l Romieu et al. (1997)
(Mexico)
Pekkanen et ai. (1997)
(Finland)
Romieu et al. (1996)
(Mexico)
_, Gielen et al. (1997)
(Netherlands)
-10,0
l I I
-5.0 0,0 5.0
Change in Pulmonary Function, L/min
10.0
Figure 8-13. Illustrative acute pulmonary function change studies of asthmatic children.
Effect of 50 ug/m3 PM10 on morning peak flow lagged one-day.
particles (PMMau_05, PMMC001_25) and PM25_10. The strongest relations were for 14 day mean PM
levels, especially for the smaller particles (PMMCO 01_2 5).
Overall, then, PM10 and PM2 5 both appear to affect lung function in asthmatics, but there is
only limited evidence for a stronger effect of fine versus coarse fraction particles; nor do
ultrafine particles appear to have any notably stronger effect than other larger-diameter fine
particles. Also, of the studies provided, few if any analyses were able to clearly separate out the
effects of PM10 and PM25 from other pollutants.
The effects of PM10 on respiratory symptoms in asthmatics tended to be positive, although
they are somewhat less consistent than PM10 effects on lung function. Most studies showed
increases in cough, phlegm, difficulty breathing, and bronchodilator use, although these
increases were generally not statistically significant for PM10 (see Tables 8-27, 8-28, 8-29, and
8-30; and, for cough as an example, see Figure 8-14). Vedal et al. (1998) reported that
(a) increases in PM10 were associated with increased reporting of cough, phlegm production, and
sore throat and (b) children with diagnosed asthma are more susceptible to the effects than are
other children. Similarly, in the Gielen et al. (1997) study of a panel of children, most of whom
had asthma, low levels of PM increased symptoms and medication use. The Peters et al. (1997b)
8-199
-------
TABLE 8-27. SUMMARY OF ASTHMA PM,n COUGH STUDIES
10
Reference Citation,
Location, etc.
Outcome Measure
Mean Particulate
Levels (Range) jig/m3
Co-Pollutants
Measured
Lag
Structure
Effect Measures Standardized
to 50 jig/m3 PM10
Asthma Studies
Vedaletal. (1998)
Gielenetal. (1997)
Hiltermannetal. (1998)
Peters etal. (1997b)
Peters etal. (1997c)
Romieuetal. (1997)
Romieu etal. (1996)
Vedaletal. (1998)
oo Gielenetal. (1997)
o Segala etal. (1998)
Neukirch et al. (1998)
Romieuetal. (1996)
Romieuetal. (1997)
Ostro etal. (2001)
Hiltermannetal. (1998)
Peters etal. (1997b)
Peters etal. (1997c)
Ostro etal. (2001)
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
19.1 (1, 159)
30.5 (16, 60)
39.7 (16, 98)
55 (?, 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 (?, 71)
47 (29, 73)
102 (47, 360) 1 hr max
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
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
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 8-28. SUMMARY OF ASTHMA PM,n PHLEGM STUDIES
10
00
to
o
Reference Citation,
Location, etc.
Vedaletal. (1998)
Peters etal. (1997c)
Romieuetal. (1997)
Romieuetal. (1996)
Vedaletal. (1998)
Romieuetal. (1997)
Romieuetal. (1996)
Peters etal. (1997c)
TABLE 8-29.
Reference Citation,
Location, etc.
Vedaletal. (1998)
Gielen etal. (1997)
Romieuetal. (1997)
Romieuetal. (1996)
Vedaletal. (1998)
Gielen etal. (1997)
Segala etal. (1998)
Romieuetal. (1997)
Romieuetal. (1996)
Delfinoetal. (1998a)
Outcome
Measure
OR phlegm
OR phlegm
OR phlegm
OR phlegm
OR phlegm
OR phlegm
OR phlegm
OR phlegm
SUMMARY
Outcome
Measure
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
LRI
Mean Particulate Co-Pollutants Lag
Levels (Range) jig/m3 Measured Structure
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)
OF ASTHMA PM10
None
SO2, sulfate, H+
Ozone
Ozone
None
Ozone
Ozone
SO2, sulfate, H+
LOWER RESPIRATORY
Oday
Oday
Oday
Oday
2 day
2 day
2 day
1-5 day
ILLNESS
Mean Particulate Levels Co-Pollutants Lag
(Range) Measured Structure
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)
None
Ozone
Ozone
Ozone
None
Ozone
SO2, NO2
Ozone
Ozone
Ozone
Ozone
Ozone
Oday
Oday
Oday
Oday
2 day
2 day
2 day
2 day
2 day
Oday
Oday
Oday
Effect Measures Standardized
to 50 jig/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)
(LRI) STUDIES
Effect Measures Standardized
to 50 fig/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 8-30. SUMMARY OF ASTHMA PM,n BRONCHODILATOR USE STUDIES
10
00
bo
o
bo
Reference Citation,
Location, etc.
Gielenetal. (1997)
Hiltermann et al. (1998)
Peters etal. (1997c)
Gielenetal. (1997)
Hiltermann etal. (1998)
Peters etal. (1997c)
Outcome Measure
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
OR bronchodilator use
Mean Particulate Co-Pollutants Lag Effect Measures Standardized
Levels (Range) jig/m3 Measured Structure to 50 jig/m3 PM10
30.5 (16, 60) Ozone 0 day
39.7 (16, 98) Ozone, NO2, SO2 0 day
47 (29, 73) SO2, sulfate, H+ 0 day
30.5 (16, 60) Ozone 2 day
39.7 (16, 98) Ozone, NO2, SO2 1-7 day
47 (29, 73) SO2, sulfate, H+ 1-5 day
0.94 (0.59, 1
1.03 (0.93, 1
1.06 (0.88, 1
2.90(1.81,4
1.12(1.00, 1
1.23 (0.96, 1
.50)
.15)
.27)
.66)
.25)
.58)
-------
Gielenetal. (1997)
(Netherlands)
Romieu etal. (1997)
(Mexico)
Peters etal. (1997c)
(Czech Republic)
Vedal etal. (1998)
(Canada)
0.5
1.0 2.0
Odds Ratio for Cough
4.0
8.0
Figure 8-14. Odds ratios with 95% confidence interval for cough per 50-ug/m3 increase in
PM10 for illustrative asthmatic children studies at lag 0.
study of asthmatics examined particle effects by size and found that fine particles were
associated with increases in cough, of which MC 0.01-2.5 was the best predictor.
Delfino et al. (1998a) used an asthma symptom score to evaluate the effects of acute air
pollutant exposures. The 1- and 8-h PM10 maximum concentrations had larger effects than the
24-h mean. Subgroup analyses showed effects of current day PM maxima to be strongest in the
10 more frequently symptomatic children; the odds ratios for symptoms were 2.24 (1.46, 3.46)
for 47 |ig/m3 1-h PM10; 1.82 (1.18, 2.8) for 36 |ig/m3 8-h PM10, and 1.50 (0.80-2.80) for 25 |ig/m3
24-h PM10. Analyses suggested that effects of O3 and PM10 were largely independent. Delfino
et al. (2002) also studied 22 asthmatic children aged 9 to 19 years in March and April 1996.
Associations were evaluated by generalized estimating equations, adjusting for autocorrelation.
The endpoint was symptoms interfering with daily activities. This endpoint was associated
with PM10, NO2, and O3 and there was a positive interaction effect of PM10 and NO2 jointly.
Both of these studies also reported significant associations with fungal spores, but not pollens;
no significant interactions were found between aeroallergens and air pollutants.
Romieu et al. (1996) found children with mild asthma to be more strongly affected by
high ambient levels of PM (mean PM10 = 166.8 |ig/m3) observed in northern Mexico City than
8-203
-------
in a study (Romieu et al., 1997) conducted in a nearby area with lower PM10 levels
(mean PM10 = 54.2 jig/m3). Yu et al. (2000) reported estimates of odds ratios for asthma
symptoms and 10 |ig/m3 increments in PM10 and PM10 values of 1.18 (1.05, 1.33) and 1.09 (1.01,
1.18), respectively. Multipollutant models with CO and SO2 yielded 1.06 (0.95, 1.19) for PM10,
and 1.11 (0.98, 1.26) for PMLO, thus showing a lower value for PM10 and a loss of significance
for both PM10 and PMX 0. The correlation between CO and PMX 0 and PM10 was 0.82 and 0.86.
Ostro et al. (2001) studied a panel of inner-city African American children using a GEE model
with several measures of PM, including PM10 (both 24-h average and 1-h max.) and PM25,
demonstrating positive associations with daily probability of shortness of breath, wheeze, and
cough.
Desqueyroux et al. (2002) studied 60 adult severe asthmatics from November 1995 to
November 1996. Relationships were estimated from generalized estimating equations adjusting
for autocorrelation. Each asthma exacerbation was confirmed by a physician, and each of the
cases were followed for a sufficient length of time to allow investigations of any lagged
associations with air pollution. Statistical analysis that accounted for temporal, meteorological,
and aerobiological variables and some individual characteristics revealed significant associations
between PM10, O3, and incident asthma attacks. Odds Ratio (OR) for an increase of 10 |ig/m3
of PM10 was 1.41 (CI: 1.16, 1.71). PM10 was not related to incident asthma attacks using lags of
1 or 2 days; but PM10 associations for 3, 4, and 5 day lags were significant. PM10 remained
significant even after adjusting for other pollutants including O3, SO2, and NO2.
Just et al. (2002) also studied 82 asthmatic children for 3 months during spring and early
summer in Paris. Relationships were estimated from generalized estimating equations adjusting
for autocorrelation. No significant relationships were found between PM13 and lung function or
respiratory symptoms. For PM2 5 results, see Table 8-31. All showed positive associations
(several being clearly significant at p < 0.05) between PM2 5 and increased cough, phlegm,
orLRI.
Of studies that included two indicators for PM (PM10, PM2 5) in their analyses, the study of
Peters et al. (1997b) found similar effects for the two PM measures, whereas the Romieu et al.
(1996) study found slightly larger effects for PM2 5.
8-204
-------
TABLE 8-31. SUMMARY OF ASTHMA PM?. RESPIRATORY SYMPTOM STUDIES
-2.5
00
to
o
Reference Citation,
Location, etc.
Peters etal. (1997c)
Romieuetal. (1996)
Tiittanen etal. (1999)
Romieuetal. (1996)
Tittanen etal. (1999)
Ostro etal. (2001)
Peters etal. (1997c)
Romieuetal. (1996)
Romieuetal. (1996)
Romieuetal. (1996)
Romieuetal. (1996)
Outcome
Measure
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR cough
OR Phlegm
OR Phlegm
ORLRI
ORLRI
Mean Particulate Levels
(Range) fig/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)
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
Lag
Structure
Oday
Oday
Oday
2 day
2 day
3 day
1-5 day
Oday
2 day
Oday
2 day
Effect Measures Standardized
to 25 jig/m3 PM2 5
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)
-------
Two asthma studies, both in the United States, examined PM indicators by 1 h averages as
well as by 24 h averages. The PM101 h outcome was larger than the 24 h outcome for lower
respiratory illness in one study (Delfino et al., 1998a) but was lower for cough in the other study
(Ostroetal., 2001).
Several of the studies reviewed above (Delfino et al., 1998a, 2002; Ostro et al., 2001;
Yu et al., 2000; Mortimer et al., 2002; Vedal et al., 1998) that were conducted in the
United States and Canada found positive associations between various health endpoints for
asthmatics and ambient PM exposure (indexed by PM10, PM2 5, or PM10_2 5). The endpoints
included PEF decrements, various individual respiratory symptoms, and combinations of
respiratory symptoms. The various endpoints each represent effects on respiratory health.
8.3.3.1.2 Lung Function and Respiratory Symptom Effects in Nonasthmatic Subjects
Results for PM10 peak flow analyses in non-asthmatic studies (summarized in Appendix 8B
Table 8B-6) were inconsistent, with fewer studies reporting results in the same manner as for the
asthmatic studies. Many of the point estimates showed increases rather than decreases (see
Table 8-32). The effects on respiratory symptoms in nonasthmatics (see Appendix 8B
Table 8B-7) were similar to those in asthmatics. Most studies showed that PM10 increases
cough, phlegm, difficulty breathing, and bronchodilator use, although these were generally not
statistically significant (Table 8-33). Vedal et al. (1998) reported no consistent evidence for
adverse health effects in a nonasthmatic control group.
Results of the PM2 5 peak flow and symptom analyses in nonasthmatic studies (see
Appendix 8B Table 8B-8, Table 8-34) were similar to PM10 results discussed above.
Three authors, Schwartz and Neas (2000), Tiittanen et al. (1999) and Neas et al. (1999),
used PM10_25 as a coarse fraction particulate measure (Table 8-35). Schwartz and Neas (2000)
found that PM10_2 5 was significantly related to cough. Tiittanen found that one day lag
of PM10_2 5 was related to morning PEF, but there was no effect on evening PEF. Neas et al.
found no effects of PM10.25 on PEF.
The Schwartz and Neas (2000) reanalyses allows comparison of fine and coarse particle
effects on healthy school children using two pollutant models of fine and coarse PM. CM was
estimated by subtracting PM2A from PM10 data. For reanalysis of the Harvard Six City Diary
8-206
-------
TABLE 8-32. SUMMARY OF NON-ASTHMA PM,n PFT STUDIES
10
Reference Citation,
Location, etc.
Gold etal. (1999)
Tiittanenetal. (1999)
Neas etal. (1999)
Tiittanenetal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Neas etal. (1999)
Harre etal. (1997)
Neas etal. (1999)
oo Schwartz & Neas (2000)
N-> Uniontown
Schwartz & Neas (2000)
State College
Tiittanenetal. (1999)
Tiittanenetal. (1999)
Gold etal. (1999)
Neas etal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Boezen etal. (1999)
Van der Zee etal. (1999)
Van der Zee et al. (1999)
Van der Zee et al. (1999)
Harre etal. (1997)
Outcome
Measure
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
Mean Particulate
Levels (Range) u,g/m3
51 (23, 878)
28 (5, 122)
32
28 (5, 122)
42 (5, 146)
42 (5, 146)
42 (5, 146)
32
(not given)
32
(not given)
(not given)
28 (5, 122)
28 (5, 122)
51 (23, 878)
32
42 (5, 146)
42 (5, 146)
42 (5, 146)
34 (?, 106)
34 (?, 106)
34 (?, 106)
(not given)
Co-Pollutants
Measured
Ozone
NO2, SO2, CO, ozone
Ozone
NO2, SO2, CO, ozone
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
NO2, SO2
NO2, SO2
N02, S02
NO2, SO2, sulfate
NO2, SO2, sulfate
NO2, SO2, sulfate
N02, S02, CO
Lag
Structure
Iday
Oday
1-5 day
1-4 day
Iday
2 day
1-5 day
Oday
1 day
Oday
Oday
Oday
Oday
Oday
Oday
1-5 day
Oday
2 day
1-5 day
Oday
2 day
1-5 day
1 day
Effect Measures Standardized
to 50 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 8-33. SUMMARY OF NON-ASTHMA PM,n RESPIRATORY SYMPTOM STUDIES
10
00
o
oo
Reference Citation,
Location, etc.
Schwartz & Neas (2000)
Boezenetal. (1998)
Van der Zee etal. (1999)
Urban areas
Tiittanen etal. (1999)
Van der Zee etal. (1999)
Urban areas
Van der Zee etal. (1999)
Urban areas
Tiittanen etal. (1999)
Boezenetal. (1998)
Tiittanen 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 Particulate
Levels (Range) jig/m3
(not given)
42 (5, 146)
34 (?, 106)
28 (5, 122)
34 (?, 106)
34 (?, 106)
28 (5, 122)
42 (5, 146)
28 (5, 122)
(not given)
34 (?, 106)
34 (?, 106)
Co-Pollutants
Measured
Sulfate fraction
NO2, SO2
NO2, SO2, sulfate
NO2, SO2, CO, ozone
NO2, SO2, sulfate
NO2, SO2, sulfate
NO2, SO2, CO, ozone
NO2, SO2
NO2, SO2, CO, ozone
Sulfate fraction
NO2, SO2, sulfate
NO2, SO2, sulfate
Lag
Structure
Oday
Oday
Oday
Oday
2 day
1-5 day
1-4 day
Oday
2 day
Oday
Oday
2 day
Effect Measures
Standardized
to 50 mg/m3 PM10
1.20(1.07, 1.35)
1.06(0.93, 1.21)
1.04(0.95, 1.14)
1.00(0.87, 1.16)
0.94 (0.89, 1.06)
0.95(0.80, 1.13)
1.58 (0.87, 2.83)
1.11(0.91, 1.36)
Positive but not significant
0.98 (0.89, 1.08)
1.01 (0.93, 1.10)
-------
TABLE 8-34. SUMMARY OF NON-ASTHMA PM2, RESPIRATORY OUTCOME STUDIES
00
to
o
Reference Citation,
Location, etc.
Gold etal. (1999)
Tiittanenetal. (1999)
Tiittanenetal. (1999)
Neas etal. (1999)
Schwartz & Neas (2000)
Uniontown
Schwartz & Neas (2000)
State College
Tiittanenetal. (1999)
Tiittanenetal. (1999)
Gold etal. (1999)
Neas etal. (1999)
Tiittanenetal. (1999)
Tiittanenetal. (1999)
Schwartz & Neas (2000)
Outcome
Measure
Morning PEFR
Morning PEFR
Morning PEFR
Morning PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
Evening PEFR
OR cough
OR cough
ORLRS
Mean Particulate
Levels (Range) jig/m3
30.3 (9, 69)
24.5 (?, 88)
(not given)
(not given)
30.3 (9, 69)
24.5 (?, 88)
15 (3, 55)
15 (3, 55)
(not given)
Co-Pollutants
Measured
Ozone
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Ozone
Sulfate fraction
Sulfate fraction
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Ozone
Ozone
NO2, SO2, CO, ozone
NO2, SO2, CO, ozone
Sulfate fraction
Lag
Structure
1 day
Oday
1-4 day
1-5 day
Oday
Oday
Oday
Oday
Oday
1-5 day
Oday
2 day
Oday
Effect Measures Standardized
to 25 jig/m3 PM2 5
-0.22 (-0.46, 0.01)
1.11 (-0.64, 2.86)
-1.93 (-7.00, 3.15)
2.64 (-6.56, 11.83)
-1.52 (-2.80, -0.24)
-0.93 (-1.88, 0.01)
0.70 (-0.81, 2.20)
1.52 (-3. 91, 6.94)
-0.10 (-0.43, 0.22)
1.47 (-7.31, 10.22)
1.04 (0.86, 1.20)
1.24(1.02, 1.51)
1.61(1.19,2.14)
-------
TABLE 8-35. SUMMARY OF NON-ASTHMA COARSE FRACTION STUDIES OF RESPIRATORY ENDPOINTS
Reference Citation,
Location, etc.
Tiittanenetal. (1999)
Neasetal. (1999)
Tiittanenetal. (1999)
Tiittanenetal. (1999)
Neasetal. (1999)
Tiittanenetal. (1999)
Neasetal. (1999)
Tiittanenetal. (1999)
Tiittanenetal. (1999)
™ Neasetal. (1999)
° Tiittanenetal. (1999)
Tiittanenetal. (1999)
Tiittanenetal. (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 Particulate
Levels (Range) jig/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
1 day
1 day
2 day
1-4 day
1-5 day
Oday
1 day
2 day
1-4 day
1-5 day
Oday
2 day
1-4 day
Oday
Oday
Effect Measures Standardized
to 25 jig/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)
-------
Study in the two PM pollutant model, they report for cough a PM25 OR of 1.07 (0.90, 1.26; per
15 |ig/m3 increment) and a PM10_2 5 OR of 1.18 (1.04, 1.34); per 8 |ig/m3 increment in contrast to
lower respiratory symptom results of a PM25 OR of 1.29 (1.06, 1.57) and a PM10_25 OR of
1.05 (0.9, 1.23). In the Uniontown reanalysis, peak flow for PM2A (for a 14 |ig/m3 increment)
was-0.91 l/m(-1.14, -1.68) and for PM10.21 (for a 15 |ig/m3 increment) was+1.04 1/m
(-1.32, +3.4). For State College, peak flow for PM21 was -0.56 (-1.13, +0.01), and forPM10.21
it was-0.17 (-2.07,+1.72).
Coull et al. (2001) reanalyzed data from the Pope et al. (1991) study of PM effects on
pulmonary function of children in the Utah Valley, using additive mixed models which allow for
assessment of heterogeneity of response or the source of heterogeneity. These additive models
describe complex covariate effects on each child's peak expiratory flow while allowing for
unexplained population heterogeneity and serial correlation among repeated measurements. The
analyses indicate heterogeneity among that population with regard to PM10 (i.e., specifically that
there are three subjects in the Utah Valley study who exhibited a particularly acute response
to PM10). However the limited demographic data available in the Utah Valley Study does not
explain the heterogeneity in PM sensitivity among the school children population.
Two studies examined multipollutant models. The Jalaludin et al. (2000) analyses used a
multipollutant model that evaluated PM10, O3, and NO2. They found in metropolitan Sydney
that ambient PM10 and O3 concentrations are poorly correlated (r = 0.13). For PEFR the P (SE)
for PM10 only was 0.0045 (0.0125), p = 0.72; and for PM10 and O3, 0.0051 (0.0124), p = 0.68.
Ozone was also unchanged in the one- and two-pollutant models. Gold et al. (1999) attempted to
study the interaction of PM25 and O3 on PEF in Mexico City children (age = 8 to 12 years). The
authors found independent effects of the two pollutants, but the joint effect was slightly less than
the sum of the independent effects.
8.3.3.2 Long-Term Particulate Matter Exposure Effects on Lung Function and
Respiratory Symptoms
8.3.3.2.1 Summary of 1996 Particulate Matter Air Quality Criteria Document Key Findings
In the 1996 PM AQCD, the available long-term PM exposure-respiratory disease studies
were limited in terms of conclusions that could be drawn. At that time, three studies based on a
8-211
-------
similar type of respiratory symptom questionnaire administered at three different times as part of
the Harvard Six-City and 24-City Studies provided data on the relationship of chronic respiratory
disease to PM. All three studies suggest a long-term PM exposure effect on chronic respiratory
disease. The analysis of chronic cough, chest illness and bronchitis tended to be significantly
positive for the earlier surveys described by Ware et al. (1986) and Dockery et al. (1989). Using
a design similar to the earlier one, Dockery et al. (1996) expanded the analyses to include
24 communities in the United States and Canada. Bronchitis was found to be higher (OR = 1.66)
in the community with the highest particle strong acidity when compared with the least polluted
community. Fine parti culate sulfate was also associated with higher reporting of bronchitis
(OR =1.65, 95% CI 1.12,2.42).
Interpretation of such studies requires caution in light of the usual difficulties ascribed to
cross-sectional studies. That is, evaluation of PM effects is based on variations in exposure
determined by a different number of locations. In the first two studies, there were six locations
and, in the third, twenty-four. The results seen in all studies were consistent with a PM gradient,
but it was not readily possible to separate out clear effects of PM from other factors or pollutants
having the same gradient.
Chronic pulmonary function studies by Ware et al. (1986), Dockery et al. (1989), and Neas
et al. (1994) had good monitoring data and well-conducted standardized pulmonary function
testing over many years, but showed no effect for children from airborne particle pollution
indexed by TSP, PM15, PM25 or sulfates. In contrast, the Raizenne et al. (1996) study of U.S.
and Canadian children found significant associations between FEVj and FVC and acidic
particles (FT). Overall, the available studies provided only limited evidence suggestive of
pulmonary lung function decrements being associated with chronic exposure to PM indexed by
various measures (TSP, PM10, sulfates, etc.). However, it was noted that cross-sectional studies
require very large sample sizes to detect differences because they cannot eliminate person to
person variation, which is much larger than the within person variation. There may be so much
noise in the large variability that one cannot observe the exposure - effect signal.
8-212
-------
8.3.3.2.2 New Studies of Respiratory Effects of Long-Term Particulate Matter Exposure
Several studies published since 1996 evaluated effects of long-term PM exposure on
lung function and respiratory illness (see Appendix 8B, Table 8B-8). The new studies
examining PM10 and PM2 5 in the United States include McConnell et al. (1999), Abbey et al.
(1998), Berglund et al. (1999), Peters et al. (1999a,b), and Avol et al. (2001), all of which
examined effects in California cohorts but produced variable results. McConnell et al. (1999)
noted that, as PM10 increased across communities, the bronchitis risk per interquartile range also
increased, results consistent with those reported by Dockery et al. (1996). However, the high
correlation of PM10, acid, and NO2 precludes clear attribution of the McConnell et al. bronchitis
effects specifically to PM alone. Avol et al. (2001) reported that, for 110 children who moved to
other locations, those subjects who moved to areas of lower PM10 showed increased growth in
lung function and subjects who moved to communities with higher PM10 showed slowed lung
function growth.
Gauderman et al. (2000, 2002) presented results from a study that is both a cohort and a
cross-sectional study. This unique design followed two cohorts of southern California children
who were fourth graders in 1993 and 1996 respectively. The cohorts, located in 12 communities,
were followed for 4 years. A three stage model which allowed for individual slopes, within
community covariates, and community-wide air pollution averages, was fitted using SAS Proc
MIXED. Pulmonary function measurements included FVC, FEV1, MMEF, and PEFR, all of
which gave similar results for both PM2 5 and PM10. In the first cohort, PM10 showed a
significant 1.3% decrease in annual growth rates for a 51.5 |ig/m3 difference in PM10. This
difference was only 0.4% in the second cohort; however, the two were not significantly different
from each other. The effect for PM25 was slightly less for a difference of 22.2 |ig/m3. In an
earlier cross-sectional analysis, Peters et al. (1999b) studied the prevalence of respiratory
symptoms in 12 southern California communities in 1993. To estimate the relationship between
symptoms and pollutants a two-stage regression approach was used. The first stage estimated
community-specific rates adjusted for individual covariates. The second stage regressed these
rates on pollutant averages from 1986 to 1990, finding no significant relationships between
respiratory symptoms and average PM10 levels.
8-213
-------
In a non-U.S. PM10 study, Horak et al. (2002) conducted a combined cohort and cross-
sectional study similar in design to that of Gauderman et al. (2000). The cohorts were taken
from 975 school children in 8 communities in lower Austria between 1994 and 1997.
Relationships were estimated from generalized estimating equations adjusting for
autocorrelation, and adjustments were made for sex, atopy, ETS, baseline lung function, height,
and site. Growth in FVC and MEF were significantly related to winter PM10 levels.
Gehring et al. (2002) enrolled 1,756 newborn children in the Munich area. Individual
PM2 5 and NO2 levels were estimated from actual measurements at 40 sites combined with a GIS
predictor model. PM2 5 levels ranged from 11.9 to 21.9 |ig/m3. The incidence (in the first two
years of life) of cough without infection and dry cough at night were related to PM25 levels.
Wheeze, bronchitis, respiratory infections, and runny nose were not related to PM2 5 levels.
Other non-U.S. studies examined PM measures such as TSP and BS in European countries.
In Germany, Heinrich et al. (2000) reported a cross-sectional survey of children, conducted
twice (with the same 971 children included in both surveys). TSP levels decreased between
surveys, as did the prevalence of all respiratory symptoms (including bronchitis). Also, Kramer
et al. (1999) reported a study in six East and West Germany communities, which found
decreasing yearly TSP levels to be related to ever-diagnosed bronchitis from 1991 to 1995.
Lastly, Jedrychowski et al. (1999) reported an association between both BS and SO2 levels in
various areas of Krakow, Poland, and slowed lung function growth (FVC and FEVj).
Leonard! et al. (2000) studied a different health outcome measure as part of the Central
European Air Quality and Respiratory Health (CESAR) study. Blood and serum samples were
collected from school children ages 9 to 11 years in each of 17 communities in Central Europe
(n = 10 to 61 per city). Numbers of lymphocytes increased as PM concentrations increased
across the cities. Regression slopes, adjusted for confounder effects, were larger and statistically
significant for PM2 5, but small and nonsignificant for PM10_2 5. A similar positive relationship
was found between IgG concentration in serum and PM2 5 gradient, but not for PM10 or PM10_2 5.
8-214
-------
8.3.3.2.3 Summary of Long-Term Particulate Matter Exposure Respiratory Effects
The methodology used in the long-term studies varies much more than the methodology in
the short-term studies. Some studies reported highly significant results (related to one or another
ambient PM indicator), whereas others reported no significant results. The cross-sectional
studies are often confounded, in part, by unexplained differences between geographic regions.
The studies that looked for a time trend are also confounded by other conditions that were
changing over time. The newer cohort studies provide the best evidence bearing on chronic PM
exposure effects. The Gauderman et al. (2000, 2002) cohort studies found significant decreases
in lung function growth among southern California school children to be related to PM10 levels,
but the Peters et al. (1999b) cross-sectional study of the children in the Gauderman et al. cohorts
found no relationship between respiratory symptoms and annual average PM10 levels in 12
southern California communities. In addition, the well-conducted cross-sectional studies by
Dockery et al. (1996) and Raizenne et al. (1996), assessed earlier in the 1996 PM AQCD, found
differences in peak flow and bronchitis rates associated with fine particle sulfate and acidity, and
they remain among the more credible available studies of long-term PM exposure effects on
respiratory function symptoms.
8.3.4 Ambient PM Impacts on Fetal and/or Early Postnatal
Development/Mortality
Some older cross-sectional mortality studies reviewed in the 1996 PM AQCD suggested
that the young may represent a susceptible subpopulation for PM-related mortality. Lave
and Seskin (1977), for example, found significant associations of TSP mortality among those
0-14 years of age. Also, Bobak and Leon (1992) studied neonatal (ages < 1 mo) and
postneonatal mortality (ages 1-12 mo) in the Czech Republic and reported significant
associations between PM10 and postneonatal mortality, even after considering other pollutants.
Postneonatal respiratory mortality showed highly significant associations for all pollutants
considered, but only PM10 remained significant in multipollutant models. The exposure duration
was longer than a few days, but shorter than in adult chronic PM exposure prospective cohort
studies. Thus, the few available studies reviewed in the 1996 PM AQCD suggested an
association between ambient PM concentrations and infant mortality, especially among
8-215
-------
postneonatal infants. More recent studies have focused on ambient PM relationships (a) with
intrauterine mortality and morbidity and (b) with early post neonatal infant mortality.
8.3.4.1 PM Effects on Intrauterine Fetal Morbidity/Mortality
During the past decade or so, increasing attention has begun to be focused on the
evaluation of possible effects or prenatal exposures to ambient PM and other "criteria air
pollutants" on fetal growth and development. Concerns about these possible air pollution
impacts, as indexed by measures such as low birth weight (LEW) or preterm births, are
prompted by studies indicating that both preterm births and low birth weights are important
predictors of infant mortality, childhood morbidity, and perhaps even adult morbidity (Barker
et al., 1993; Spinello et al., 1995; Joseph and Kramer, 1996). In evaluating possible pollutant
effects, a number of variables with well-established links to fetal growth and development must
be taken into account, e.g., maternal and paternal weight and height, gestational weight gain,
maternal smoking and alcohol consumption, the infant's sex and racial/ethnic background, etc.
(Kramer, 1987; Berkowitz and Papiernik, 1993; Divon et al., 1994).
In one large-scale U.S. study, Ritz et al (2000) evaluated the effects of ambient PM10, CO,
NO2, SO2, and O3 exposures during pregnancy on the occurrence of preterm births among a
cohort of 97,518 neonates born in the South Coast Air Basin (SoCARB) of California. Pollutant
values measured at the closest of 17 air-monitoring stations (within 2 mi. radius of residential zip
code on birth certificate) were averaged over the entire pregnancy or distinct periods during
pregnancy for each birth in nondesert portions of Los Angeles, San Bernardino, Riverside, and
Orange Counties (which comprise the SoCARB district) during 1989 to 1993. Adjusting for
various factors known to be related to occurrence of premature births (e.g., maternal age, race,
smoking during pregnancy, etc.), the effects of the different air pollutants on preterm birth risk
were analyzed by means of both single-pollutant and multiple pollutant logistic regression
models. A 16% increase in preterm birth risk was estimated per 50 |ig/m3 increment in PM10
concentrations averaged over the first month of pregnancy (RR = 1.20; CI: 1.06, 1.26) and a 20%
increase per 50 |ig/m3 PM10 averaged over 6 weeks prior to birth. The effects sizes varied only
slightly between single and multiple pollutant molds or with adjustments for other risk factors.
8-216
-------
Significant associations were also found for CO, but inclusion of other pollutants or covariates in
the model caused CO effect estimates to fluctuate. For both PM10 and CO, the most precise
effects estimates (with narrowest CI) were found for exposures averaged over 6 weeks prebirth.
In another large U.S. study, Maisonet et al. (2001) evaluated associations between low
birth weight (LEW, < 2,500 g at birth) and daily average PM10, SO2, and CO levels, based on
every sixth day ambient 24-h PM10 monitor readings and 24-h averages of hourly SO2 and
CO readings at community monitoring stations in six Northeastern U.S. cities (Boston, MA;
Hartford, CT; Philadelphia, PA; Pittsburgh, PA; Springfield, MA; Washington, DC). Using air
pollution data from U.S. EPA, average trimester exposures were estimated for PM10, SO2, and
CO based on 3 to 4 PM10, 1 to 4 SO2, and 2 to 4 CO monitors per city. Based on NCHS data
sets, information on numerous covariates (e.g., gestational age, gender, maternal age, maternal
race/ethnicity, maternal prenatal smoking, alcohol consumption, etc.) was included in a logistic
regression model to generate adjusted odds ratios (AOR) and 95% CI for LEW, and then linear
regression models were used to assess reductions in birth weight (in grams) in relation to each
air pollution variable. Ranges of exposure categories were defined for each air pollutant as
percentiles of the exposure distribution (< 25th percentile; 25 to 50th; 50th to 75th; 75th to 95th;
and > 95th). Of 130,465 live singleton births during 1994 to 1996, after exclusions for several
reasons, 89,557 (68.6%) of infants were included in the final analyses. There were no
statistically significant (p < 0.05) associations between LEW and PM10 percentile groups
> 25th versus < 25th percentile or for continuous (10 |ig/m3) PM10 and LEW during any of the
prenatal trimesters among white or African-American infants. However, first trimester
ambient PM10 levels were associated with increased risk for full-term LEW among Hispanics
(AOR 1.36; CI: 1.06, 1.75). Much more consistent increases in AOR for LEW were found for
various trimesters of CO and/or SO2 exposure. The authors concluded that, overall, LWB was
not associated with PM10 exposure during pregnancy.
Rogers et al. (2000) reported results of another, much smaller, population-based case-
control study of possible associations between exposures to ambient TSP and SO2 (using a
combined TSP SO2 index) and risk of having a very low birth weight (VLBW) baby (i.e.,
weighing < 1,500 g at birth) among women residing in Atlanta, Savannah, or other areas in
8-217
-------
Georgia Health Care District 9 during 1986 to 88. Environmental transport models were used to
estimate TSP SO2 exposures at the birth homes of study subjects, and exposures < 9.94 |ig/m3,
the median of TSP and SO2 exposures for the control subjects were used as referent exposures.
The controls included 202 mothers of babies weighing 1,500 g or more at birth for comparison
versus 143 mothers of VLBW babies. A distinct trend suggesting a relationship between
TSP/SO2 exposure and increased risk for VLBW was reported. However, the results, while
suggestive of possible ambient TSP and/or SO2 effects on VLBW at high exposure levels
(frequency distributions for cases and controls began to separate at -55 |ig/m3 TSP SO2), are not
based on direct TSP or SO2 monitoring data and should not be accorded much weight unless
confirmed by other studies using better more directly measured air pollution exposure indices.
With regard to new non-U.S. studies of prenatal PM exposure effects on intrauterine
morbidity/mortality, Dejmek et al. (1999) evaluated possible impacts of ambient PM10 and
PM2 5 exposure (monitored by EPA-developed VAPS methods) during pregnancy on risk
for intrauterine growth retardation (IUGR) in the highly polluted Teplice District of Northern
Bohemia in the Czech Republic during 1993 to 1996. Mean levels of pollutants (PM, NO2, SO2)
were calculated for each month of gestation and three concentration intervals (low, medium,
high) were derived for each pollutant. Preliminary analyses found significant associations of
IUGR with PM10 and SO2 early in pregnancy but not with NO2. Odds ratios for IUGR for PM10
and PM2 5 levels were determined by logistic regressions for each month during gestation, after
adjusting for potential confounding factors (e.g., maternal smoking, alcohol consumption during
pregnancy, etc.). Definition of an IUGR birth was any one for which the birth weight fell below
the 10th percentile by gender and age for live births in the Czech Republic (1992-93). The
ORs for IUGR were significantly related to PM10 during the first month gestation: that is,
as compared to low PM10, the medium level PM10 OR = 1.47 (CI: 0.99, 2.16) and the high
level PM10 OR = 1.85 (CI: 1.29, 2.66). PM25 levels were highly correlated with PM10 (r = 0.98)
and manifested similar patterns (OR = 1.16, CI: 0.08, 0.69 for medium PM2 5 level; OR = 1.68,
CI: 1.18, 2.40 for high PM25 level). These results suggest effects of PM exposures (probably
including fine particles such as sulfates, acid aerosols, and PAHs in the Teplice ambient mix)
early in pregnancy (circa embryo implantation) on fetal growth and development.
8-218
-------
In broader further analyses of the same data set from which the above Czech study results
were derived, Dejmek et al. (2000) evaluated relationships between IUGR and exposures
to PM10, PM2 5 and PAHs during early pregnancy among women residing in the highly polluted
Teplice area or in Prachatice (an area with similarly high PAH but low particle concentrations).
Mean PM10, PM2 5 and c-PAH exposures during 9 gestation months (GMs) were estimated for
each mother and regressed (controlling for several potential covariates, e.g., maternal age,
smoking, alcohol consumption, parental education, etc.) in logistic regression models against
data for all European-origin live births during 1993 to 1998 in Teplice (n = 3,378) and Prachatice
(n = 1,505). The adjusted odds ratio (AOR) for IURG confirmed the previously published
findings noted above for increased risk of IURG being associated with ambient particle
exposures in Teplice during the first GM, but not in Prachatice (the lower particle area).
Adjusted odds ratios calculated for low, medium, and high c-PAH levels (L = < 15, M = 15-30,
M > 30 ng/m3) for fetuses from Teplice during the first GM were 1.60 (CI: 1.06, 2.15) and
2.15 (CI: 2.7, 3.6), respectively, for medium and high c-PAHs versus the low c-PAH exposure
group. Similar associations were reported for the medium and high PAH exposures (during first
GM) in Prachatice even in the presence of low overall particle levels, prompting the authors to
hypothesize a likely important role for PAHs.
8.3.4.2 PM Effects on Postneonatal Infant Mortality
Results suggestive of possible early postnatal PM exposure effects on neonatal infant
mortality have also emerged from some other new studies. Woodruff et al. (1997), for example,
used cross-sectional methods to evaluate possible associations between postneonatal infant
mortality and ambient PM10 pollution in U.S. urban areas. This study involved an analysis of a
cohort of ~4 million infants born during 1989 to 1991 in 86 U.S. metropolitan statistical areas
(MSAs). Data from the National Center for Health Statistics-linked birth/infant death records
were combined at the MSA level with PM10 data from EPA's Aerometric database. Infants were
categorized as having low, medium, or high exposures based on tertiles of PM10 averaged over
the first 2 postnatal months. Relationships between this early neonatal PM10 exposure and total
and cause-specific postneonatal mortality rates (from 1 mo to 1 year of age) were examined
8-219
-------
using logistic regression analyses, adjusting for demographic and environmental factors. Overall
postneonatal mortality rates per 1,000 live births were 3.1 among infants in areas with low PM10
exposures, 3.5 among infants with medium PM10 exposures, and 3.7 among high PM10 exposed
infants. After adjustment for covariates, the OR and 95% confidence intervals for total
postneonatal mortality for the high versus the low exposure group was 1.10(CI: 1.04, 1.16). For
normal birth weight infants, high PM10 exposure was associated with mortality for respiratory
causes (OR = 1.40, CI: 1.05, 1.85) and sudden infant death syndrome (OR = 1.26, CI: 1.14,
1.39). Among low birth weight babies, high PM10 exposure was positively (but not significantly)
associated with mortality from respiratory causes (OR = 1.18, CI: 0.86, 1.61). However, other
pollutants (e.g., CO) were not considered as possible confounders, and this lack of consideration
of other air pollutants as potential confounders in this new study introduces uncertainty in
attribution of observed effects to PM.
Lipfert et al. (2000c) used a modeling approach similar to that of Woodruff et al. (1997),
but used annual-average PM10 air quality data for one year (1990) instead of PM10 averaged over
the first two postnatal months during 1989 to 1991. The quantitative relationship between the
individual risk of infant mortality did not differ among infant categories (by age, by birthweight,
or by cause), but PM10 risks for SIDs deaths were higher for babies of smoking mothers. SO42
was a strong negative predictor of SIDs mortality for all age and birth weight categories. The
authors (a) noted difficulties in ascribing the reported PM10 and SO42 associations to effects of
the PM pollutants per se versus the results possibly reflecting interrelationships between the air
pollution indices, a strong well-established east-west gradient in U.S. SIDS cases, and/or
underlying sociodemographic factors (e.g., the socioeconomic or education level of parents) and
(b) hypothesized that a parallel gradient in use of wood burning in fireplaces or woodstoves and
consequent indoor wood smoke exposure might explain the observed cross-sectional study
results. Lipfert et al. (2000c) also raised questions as to whether the findings of Woodruff et al.
(1977) could be due to confounding factors.
The study by Loomis et al. (1999) of infant mortality in Mexico City during 1993-1995
provides interesting information pointing towards possible fine particle effects on infant
mortality. That is, in Mexico City (where mean 24-h PM2 5 = 27.4 |ig/m3), infant mortality was
8-220
-------
found to be associated with PM2 5, NO2, and O3 in single pollutant GAM Poisson models, but
much less consistently with NO2 and O3 than PM2 5 in multipollutant models. The estimated
excess risk for PM25-related infant mortality lagged 3 to 5 days was 18.2% (CI = 6.4, 30.7) per
25 |ig/m3 PM2 5. The extent to which such a notable increased risk for infant mortality might be
extrapolated to U.S. situations is not clear, however, due to possible differences in prenatal
maternal or early postnatal infant nutritional status.
Bobak and Leon (1999) conducted a matched population-based case-control study covering
all births registered in the Czech Republic from 1989 to 1991 that were linked to death records.
They used conditional logistic regression to estimate the effects of suspended particles and
nitrogen oxides on risk of death in the neonatal and early postneonatal period, controlling for
maternal socioeconomic status and birth weight, birth length, and gestational age. The effects of
all pollutants were strongest in the postneonatal period and specific for respiratory causes. Only
PM showed a consistent association when all pollutants were entered in one model. Thus, in this
study, long-term exposure to PM was the air pollutant metric most strongly associated with
excess postneonatal deaths.
8.3.4.3 Summary of Saliant Points on PM Effects on Fetal and/or Early Postnatal
Development/Mortality
A few older cross-sectional studies reviewed in the 1996 PM AQCD reported findings
suggestive of (a) possible TSP relationship to increased postnatal mortality among U.S. infants,
children, and adolescents (aged 0 to 14 years) and (b) possible associations between early
postnatal mortality among Czech infants (1 to 12 mo). Several more recent studies conducted in
the U.S. have focused on the possible effects of air pollution exposures during pregnancy on the
occurrence of preterm or low birth weight births, both of these being risk factors for a myriad of
later health problems (childhood morbidity/mortality; possible adult morbidity). One study
found results suggestive of prenatal PM10 exposures during the 1 st month of pregnancy or
averaged over 6 weeks prior to birth being associated with increased risk of preterm birth, even
in multipollutant models. However, another large scale U.S. study found little evidence
indicative of prenatal PM10 exposures being related to increased risk of low birth weight,
whereas a new Czech study did find evidence indicative of interuterine growth retardation
8-221
-------
(leading to low birth weight) being related to PM2 5 exposures during the first gestational month.
Similarly, analogously mixed results were reported for some new studies that evaluated ambient
PM relationships to early postnatal mortality among U.S., Czech, and Mexican infants. These
results, overall, highlight the need for more research to elucidate potential ambient PM effects on
fetal development/mortality and for postnatal morbidity/mortality.
8.4 INTERPRETIVE ASSESSMENT OF THE EPIDEMIOLOGIC
EVIDENCE
8.4.1 Introduction
Numerous PM epidemiology studies assessed in the 1996 PM AQCD implicated ambient
PM as a likely contributor to mortality and morbidity effects associated with ambient air
pollution exposures. Since preparation of the 1996 PM AQCD, the epidemiologic evidence
concerning ambient PM-related health effects has vastly expanded. Past regulatory decisions
have been important in the selection of PM indices and evolution of PM epidemiologic literature.
That is, the adoption of PM10 standards in 1987 and of PM25 standards in 1997 have generated
ambient air concentration databases that have made it possible for research to address many
previously unresolved issues regarding potential linkages between airborne PM and human
health; and the newly implemented nationwide network of speciation samplers holds promise for
further advances regarding identification of the most influential specific components of the
ambient air pollution mixture and their sources.
As was discussed in Sections 8.2 and 8.3, numerous new PM epidemiology studies have
evaluated health effects associated with short or long-term ambient PM exposure. Most have
found positive associations (many being statistically significant) between (a) excess risks for
various mortality and/or morbidity endpoints in many U.S. cities and elsewhere and (b) ambient
PM indexed by a variety of ambient community monitoring methods. Some other new studies
have found positive, but nonsignificant associations with PM, and a few have reported negative
(usually nonsignificant) associations with ambient PM and/or more robust gaseous co-pollutant
effects. Several issues and attendant uncertainties continue to be important in assessing and
8-222
-------
interpreting the overall PM epidemiology database and its implications for estimating risks
associated with exposure to ambient PM concentrations in the United States. These include
issues concerning: (1) approaches to model specification to take into account important effect
modifiers (such as weather) and to control for potential confounding of PM effects by
co-pollutants (especially major gaseous pollutants such as O3, CO, NO2, SO2); (2) temporal
relationships between exposure and effect (lags); (3) potential consequences of measurement
error; (4) attribution of various types of health effects to specific PM components (e.g., PM10,
PM10_2 5, PM2 5, ultrafines, sulfates, metals, etc.) or to source-oriented indicators (motor vehicle
emissions, vegetative burning, etc.); (5) the general shape of exposure-response relationship(s)
between PM and/or other pollutants and observed health effects (e.g., potential indications of
thresholds); (6) geographic homogeneity / heterogeneity of PM exposure-health risk
relationships; and (7) implications of PM-related mortality effects (e.g., mortality displacement;
life shortening). All of these issues are of much importance for characterizing and interpreting
ambient PM-health effects associations.
Assessing the above uncertainties in relation to the PM epidemiology data base remains a
challenge. The basic issue is that there are an extremely large number of possible models, any of
which may turn out to give the best statistical "fit" of a given set of data, and only some of which
can be dismissed a priori as biologically or physically illogical or impossible, except that
putative cause clearly cannot follow effect in time. Most of the models for daily time-series
studies are fitted by adjusting for changes over long time intervals and across season, by day of
week, weather, and climate. Many of the temporal and weather variable models have been fitted
to data using semi-parametric methods such as spline functions or local regression smoothers
(LOESS). The goodness of fit of these base models has been evaluated by criteria suitable for
GLM with Poisson or hyper-Poisson responses (number of events) with a log link function,
particularly the Akaike Information Criterion (AIC) and the more conservative Bayes
Information Criterion (BIC) which adjust for the number of parameters estimated from the data.
The Poisson over-dispersion index and the autocorrelation of residuals are also often used.
However, if high correlations between PM and one or more gaseous pollutants emitted from a
8-223
-------
common source (e.g., motor vehicles) exist in a given area, then disentangling their relative
individual partial contributions to observed health effects associations becomes very difficult.
Testing numerous models or model specifications (e.g., including single versus various
combinations of multiple air pollutants, varying numbers and combinations of meteorological
variables, varying numbers of knots in splines, varying degrees of freedom, etc.) can yield a
widely varying array of results; and approaches to selection of "best fit" models and to
characterizing uncertainties associated with modeling outcomes remain controversial issues.
In an effort to address such issues, Dominici et al. (2003) used a uniform approach (same
variables and smoothing functions) in the NMMAPS 90 city study. This approach reduces the
uncertainty associated with multiple testing, but at the cost of possibly not identifying the best
model in each city. Lumley and Sheppard (2000) used different control variables to check the
bias in model identification. They found that the bias was small, but of the same magnitude as
the estimated health impacts. Another approach is to use one set of data for model identification,
and a second set of data for model fitting. Cross validation also shields light on this issue.
Testing many models to identify the model with the best fit can lead to an under-estimation
of uncertainty. Bayesian model averaging, or BMA, used in other fields (e.g., econometrics) as a
formalized means for conducting model search strategies and providing an approach to account
for modeling uncertainties (Hoeting et al., 1999), has begun to be employed in the late 1990's in
air pollution epidemiologic analyses. Among the first and best known applications of the BMA
approach to evaluation of air pollution effects are the studies by Clyde (1999), Clyde (2000), and
Clyde et al. (2000).
As discussed earlier (in Section 8.2.2), Clyde (1999) employed BMA techniques in an
analysis of PM10-mortality relationships among the elderly (> 65 years old) in Chicago and Cook
County, IL during 1985 to 1990, using daily PM10 data from one monitoring site and daily
averages of one-in-six day PM10 data from subsets of 6 of 20 other monitoring sites. Using
principal component analyses involving combinations of meteorological factors and allowing for
both linear models and nonlinear relationships of mortality to PM10 variations, Clyde (1999)
reported a very high overall model averaging probability (near one) for the existence of a
particulate matter effect. Based on the results, she noted that the prediction interval suggests that
8-224
-------
the overall reduction in deaths among the elderly (> 65 years) in Chicago would be 91 to
300 deaths per year for a 10 |ig/m3 decrease in PM. However, Clyde (1999) also noted the
preliminary nature of her Chicago/Cook County analyses and that the results needed to be
caveated (as per discussion in Section 8.2.2).
A more extensive systematic evaluation of model choice was subsequently carried out by
Clyde (2000), using Bayesian Model Averaging for the same Birmingham, AL, data as was
analyzed earlier by Smith et al. (1999). In the Clyde (2000) study, several different calibrated
information criterion priors were tried in which models with large numbers of parameters are
penalized to various degrees. After taking out a baseline trend (estimated using a GLM estimate
with a 30-knot thin-plate smoothing spline), 7,860 models were selected for use in model
averaging. These included lags 0 to 3 days of a daily monitor PM10, an area-wide average PM10
value with the same lags, temperature (daily extremes and average) lagged 0 to 2 days, humidity
(dewpoint, relative humidity min and max, average specific humidity) lagged 0 to 2 days, and
atmospheric pressure, lagged 0 to 2 days. The model choice is sensitive to the specification of
calibrated information criterion priors, in particular disagreeing as to whether different PM10
variables should be included or not. For example, one or another PM10 variable was included in
all the top 25 AIC models, but only in about 1/3 of the top BIC models. The two approaches
yielded relative risk estimates per 100 |ig/m3 PM10 increment of about 1.05 for AIC, with 95%
probability (confidence) intervals of (0.94, 1.17) for the AIC prior, and of 1.1015 (0.99, 1.11) for
the BIC prior. A validation study in which randomly selected data were predicted using the
different priors favored Bayesian model averaging with BIC prior over model selection (picking
the best model) with BIC or any approach with AIC. This type of modeling may represent
another type of multipollutant modeling approach in addition to more typical hypotheses-driven
model construction and interpretation that draws more on external information (e.g., exposure,
dosimetric, toxicologic relationships) in specifying models and interpreting their results.
However, it should be noted that the 95% probability (confidence) intervals for estimates derived
for both the AIC and BIC priors encompass relative risk (RR) estimates derived for PM10-
mortality associations in Birmingham by more widely-used conventional methods. That is, RR
estimates for PM10-elderly mortality in Birmingham of 1.11 derived by Poisson regression GEE
8-225
-------
(Schwartz, 1993) and 1.09 derived by a linear model using square roots of the daily death counts
as the dependant variable (Smith, 1999) fall within the 95% probability intervals derived by the
Clyde (2000) BMA approach.
In most of this document, confidence intervals, or credible intervals for Bayesian analyses,
are reported in order to emphasize that the effect size is not known with certainty, but some
values are more nearly consistent with the data than effect size values outside the interval.
P-values or t-values are implicitly associated with a null hypothesis of no effect. A nominal
significance level of p < 0.05 or 5% (i.e., a 95% confidence interval) is usually used as a guide
for the reader, but P-values should not be used as a rigid decision-making tool. If the observed
confidence intervals were arrived at by a number of prior model specification searches,
eliminating some worse fitting models, the true interval may well be wider.
Given the now extremely large number of published epidemiologic studies of ambient PM
associations with health effects in human populations and the considerably wide diversity in
applications of even similar statistical approaches (e.g., "time-series analyses" for short-term PM
exposure effects), it is neither feasible nor useful here to try to evaluate the methodological
soundness of every individual study. Rather, a four-pronged approach is likely to yield useful
evaluative information: (1) an overall characterization of evident general commonalities (and/or
notable marked differences) among findings from across the body of studies dealing with
particular PM exposure indices and types of health outcomes, looking for convergence (and/or
divergence) of evidence regarding types of effects and effect-sizes attributable to ambient PM
indices across various geographic locations based on various methodologically acceptable
analyses; (2) thorough, critical assessment of newly published multicity analyses of PM effects,
assuming that greater scientific weight is generally ascribable to their results than those of
smaller-sized studies (often of individual cities) yielding presumably less precise effect size
estimates; (3) evaluation of, albeit at times less precise, single city results; and (4) evaluation of
coherence of the findings with other types of pertinent biological information (e.g., exposure,
dosimetry, toxicity, etc.).
8-226
-------
In the sections that follow, issues noted above are critically discussed. First follows a
discussion of the GAM issue and a summary of some key findings emerging from the short
communications and peer-review commentary published by HEI (2003b).
8.4.2 GAM Issue and Reanalyses Studies
As discussed earlier, Dominici et al. (2002) reported that the default convergence criteria
used in the S-Plus function GAM may not guarantee convergence to the best unbiased estimate
in all cases. The actual importance of this effect has begun to be quantified, with the results of
the reanalyses of a number of important studies described in short communications published in
the HEI (2003c) Special Report being especially helpful in this regard. As for the net outcome
of these reanalyses efforts, HEI (2003c) summarizes it well, as follows:
"Overall, the revised analyses using GAM with more stringent convergence criteria and
iterations and GLM-natural splines resulted in lower estimates, but largely confirmed
the effect of exposure to particulate matter on mortality (Burnett and Goldberg, 2003;
Dominici et al., 2003; Katsouyanni et al., 2003; Samoli et al. ,2003; Schwartz, 2003b;
Zanobetti and Schwartz, 2003b) and morbidity, especially for hospitalizations for
cardiovascular and respiratory diseases (Atkinson et al., 2003; Fairley, 2003; Gold et al.,
2003; Hoek, 2003; Ito, 2003; Le Tertre et al., 2003; Ostro et al., 2003; Schwartz, 2003a;
Sheppard, 2003; Zanobetti and Schwartz, 2003a). As in earlier analyses, the effect was
more pronounced among individuals 65 years of age and older (Fairley; Gold et al.;
Goldberg and Burnett; Ito; Le Tertre et al.; Mar et al.; Moolgavkar; Schwartz a). The
impact of various sensitivity analyses, when these were performed, differed across the
studies. No significant impacts were seen in some (Ostro et al.), whereas in others,
alternative modeling of time (Klemm and Mason; Moolgavkar) and weather factors
(Goldberg and Burnett; Ito) resulted in substantial changes."
The following discussion elaborates in more detail the nature and extent of potential
problems in various studies that have used the GAM default algorithm, but which have also had
their analyses redone using alternative methods in order to address this convergence issue.
8.4.2.1 Impact of Using the More Stringent GAM Model on PM Effect Estimates
for Mortality
Many of the reanalysis studies analyzed associations between PM10 and mortality, allowing
evaluation of the impact of the GAM convergence problem on this PM index. Table 8-36 and
Figure 8-15 show the percent excess total nonaccidental mortality (unless noted otherwise) risk
8-227
-------
TABLE 8-36. PM10 EXCESS RISK ESTIMATES FROM REANALYSIS STUDIES
FOR TOTAL NONACCIDENTAL MORTALITY PER 50 jig/m3 INCREASE IN PM10
Study
Multi-Cities Analyses
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)
Single-Cities Analyses
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**
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
**No CI interval given
estimates per 50 |ig/m3 increase in PM10 derived from the reanalysis studies for (1) GAM with
default convergence criteria; (2) GAM with stringent convergence criteria; and, (3) GLM with
natural splines that approximate the original GAM model. The table and figure show results
only from the studies that used all of the three alternative models for PM10. It can be seen that
most, but not all, reanalyses resulted in reductions in PM10 risk estimates when more stringent
convergence criteria were used in GAM models. Using GLM with natural splines resulted in
additional reduction in PM10 risk estimates for most, but not all, cases. The extent of reduction
in PM10 risk estimates seen with use of GAM with more stringent convergence criteria or GLM
8-228
-------
with natural splines was generally proportionally greater for the larger-scale multicity studies
than for the single-city analyses. The decreases in PM10 effect size estimates for the multicity
studies can be seen in Table 8-36 to fall rather evenly across the range of -3 to -33% for the
GAM-stringent and from -18 to -52% for the GLM reanalyses values. In contrast, the single-
city reanalyses yielded PM10 effect size estimates that were generally little changed from the
original estimates (varying by ±10% for 7 of 9 GAM-stringent and for 6 of 9 GLM reanalyses,
with the others decreasing by 17 to 59%). The relative percent reduction is greater for the
studies that had smaller PM10 risk estimates in the original analyses (e.g., NMMAPS U.S.
90 cities analyses). It can also be seen in Figure 8-15 that the extent of reduction in PM10 risk
estimates is smaller than the variability of PM10 risk estimates across the studies. Thus, the
effect of the GAM convergence problem does not appear, in most cases, to be substantial.
Several of the reanalysis reports also analyzed PM25 and PM10_25. Generally, the pattern and
extent of reductions in mortality risk estimates were similar to those for PM10. The results
for PM2 5 and PM10_2 5 mortality risk estimates are compared in a later section.
Dominici et al. (2002) also illustrated that GAM models, even with stringent convergence
criteria, still result in biased (downward) standard errors of regression coefficients. This was the
main reason for the use of GLM with natural splines in the reanalysis studies. As can be seen
from Figure 8-15, the 95% confidence bands are somewhat wider for GLM results than for GAM
results in some, but not all cases. However, the extent of wider confidence bands is not
substantial in most cases (the bias ranged from a few percent to -15% in most cases). It should
be noted that, while a GLM model with natural splines provides correct standard error of
regression coefficient, it is not equivalently as flexible as LOESS or smoothing splines. Unlike
LOESS or smoothing splines, natural splines fit linearly at both ends of the data span. Natural
splines therefore may not be an ideal model option for temperature effects, for which the slopes
are likely nonlinear (especially at the higher end). Goldberg and Burnett (2003), in their
reanalysis of Montreal data, discussed related issues. In their reanalysis, the originally reported
risk estimates of PM indices (CoH, extinction coefficient, predicted PM25, and sulfate) were
greatly attenuated in the GLM model with natural splines. One of the alternative explanations
for these results was that the natural spline does not fit the possibly nonlinear (threshold) effect
8-229
-------
U.S. 90 cities (1)
Dominici et al. (2002)
Harvard 6 cities (01)
Klemm and Mason (2003)
U.S. 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)*
Maretal. (2003)
Detroit, Ml:'85-'90 (1)
Ito (2003)
Detroit, Ml:'92 -'94 (1)
Ito (2003)
The Netherlands (1)
Hoek(2003)
Erfurt, Germany (0)
Stolzel et al. (2003)
% excess deaths per 50 ug/m3 increase in PM10
-2
L
0
2
i
4
i
6
i
8
i
10
i
Mult! 7 city studies
-x-
Single- city studies
-X-
-e-
-Q-
-e-
-x-
Figure 8-15. PM10 excess risk estimates for total nonaccidental 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).
8-230
-------
of temperature as well as nonparametric smoothers. Hoek (2003), in his reanalysis of the
Netherlands data, also showed that, compared to GAM models, GLM/natural spline models
resulted in larger deviance, indicating poorer fits. Thus, there are remaining issues regarding the
trade-off between GAM/nonparametric smoothers and GLM/parametric smoothers. The
GLM/natural splines may produce correct standard errors but cannot guarantee "correct" model
specifications. More recently, Domimici et al. (2003) developed and published a GAM routine
for SPlus that gives correct standard errors, but it was not developed in time to be used for the
GAM reanalysis effects reported on in HEI (2003c).
Three reanalysis reports applied alternative smoothing approaches (e.g., penalized splines)
that, as with GLM/natural splines, did not have the problem of biased standard error. These
studies were: reanalyses of Harvard six cities data by Schwartz (2003a); reanalysis of 10 U.S.
cities data by Schwartz (2003b); and reanalysis of APHEA2 by Katsouyanni et al. (2003).
Generally, as with GLM/natural splines, the use of alternative smoothing approaches resulted in
smaller PM risk estimates than GAM with stringent convergence criteria. In the reanalysis of
APHEA2 study, the PM10 risk estimates from penalized splines were smaller than those from
GAM, but larger than those from natural splines. Three alternative smoothing approaches
(B-splines, penalized splines, and thin-plate splines) used in the reanalysis of Harvard six
cities PM2 5 data resulted in generally smaller risk estimates than those from natural splines.
As was expected, all of these alternative smoothing approaches resulted in standard errors that
were comparable to those from natural splines but larger than those from GAM models.
Several of the GAM reanalysis reports included additional sensitivity analyses which
provided useful information. These sensitivity analyses included examinations of the effect of
changing degrees of freedom for smoothing of temporal trends and weather variables (Dominici
et al. [2002]; Ito [2003]; Klemm and Mason [2003]; Moolgavkar [2003]; and Burnett and
Goldberg [2003]). In these analyses, changing the degrees of freedom for smoothing of
temporal trends or weather effects often resulted in change of PM coefficients to a similar or
even greater extent than those caused by the GAM convergence problem. A distinctly less well
investigated issue is the effect of the use of different weather model specifications (i.e., how
many and which weather variables and their lags are included). In a limited examination of this
8-231
-------
issue in the reanalysis of Detroit data (Ito, 2003), a weather model specification similar to that
used in the NMMAPS U.S. 90 cities study consistently yielded smaller PM10 risk estimates than
a weather model similar to that used in the Harvard six cities study.
In summary, the results from the GAM reanalysis studies indicate that PM risk estimates
from GAM models were often, but not always, reduced when more stringent convergence
criteria were used. However, the extent of the reduction was not substantial in most cases. The
variability of PM risk estimates due to the model specification, including the number of weather
terms and extent of smoothing, is likely larger than the effect of the GAM convergence problem.
The extent of downward bias in standard errors reported for these data (a few percent to -15%)
also appears not to be very substantial, especially when compared to the range of standard errors
across studies due to differences in population size and numbers of days available. Nevertheless,
this chapter mainly considers results of the reanalyzed studies or of other originally published
studies that did not use GAM with default convergence criteria, because the extent of the effect
of this problem is not generally predictable in each individual study.
8.4.2.2 Impact of Using the More Stringent GAM Model on PM Effect Estimates for
Respiratory Hospital Admissions
The NMMAPS multicity study (Samet et al., 2000a,b) of PM10 concentrations and hospital
admissions used the default GAM model specification with multiple smooths. The changes
derived from use of the more stringent GAM convergence criteria are illustrated by the results of
reanalyses by Zanobetti and Schwartz (2003a). Their results indicate that there was only about a
14% decline in the effect estimates associated with use of the more appropriate stringent
convergence requirement. The two estimates were well within the 95% confidence interval of
each other. Also, in comparing the difference in the estimates between each of the six pairs of
estimates by a two-sided z-statistic, all the p-values are > 0.5, indicating that the two
convergence requirements gave insignificant differences in estimates.
To examine the potential influence of the GAM convergence specification on the results of
the original Detroit data analysis by Lippmann et al. (2000), the associations between PM
components and daily mortality/morbidity were re-examined by Ito (2003) using more stringent
convergence criteria, as well as by applying a GLM that approximated the original GAM
8-232
-------
models. Generally, the GAM models with stringent convergence criteria and GLM models
resulted in somewhat smaller estimated relative risks than those reported in the original study,
averaging 17% less for the stringent GAM case versus the default case. For COPD, the decrease
associated with the more stringent convergence criteria was larger (averaging 30%). Overall, for
all types of hospital admissions (including pneumonia, COPD and ischemic heart disease) the
change to the more stringent GAM convergence criteria gave an average decrease of 20 percent,
while a switch to the GLM model specification gave an average 29% decrease in estimated PM
effect size.
As discussed earlier, Sheppard (2003) recently conducted a reanalysis of their non-elderly
hospital admissions data for asthma in Seattle, WA, in order to evaluate the effect of the fitting
procedure on their previously published analyses. A lag of 1 day was used for all PM models.
As shown in Table 8-37, the results were provided in the manuscript to only one significant
figure (to the nearest whole percent), making the calculation of percent changes between models
problematic, since the rounding of the effect estimates are nearly on the order of the size of the
effect estimate changes. However, it can be seen that the pattern of changes in effects estimates
and 95% CI values is very similar to that seen in other studies.
TABLE 8-37. COMPARISON OF MAXIMUM SINGLE DAY LAG EFFECT
ESTIMATES FOR PM2 s, PM10 2 s, 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
Original Default GAM
Model* % Increase/IQR
(95% CI)
PM25
PM2.5.10
PM10
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)
*PM2 5IQR = 11.8 ug/m3; PM2 5.10 IQR = 9.3 ug/m3; PM10IQR = 19 ug/m3.
Source: Derived from Sheppard (2003).
8-233
-------
Further evidence of the relatively small effect of the default convergence criteria issue in
most applications is the recent work by Moolgavkar (2003), in which he reanalyzed his earlier
GAM analyses of hospital admissions for COPD (Moolgavkar, 2000c) for the cities of
Los Angeles (Los Angeles County) and Chicago (Cook County). In his original publication,
Moolgavkar found -5.0% excess risk for COPD hospital admissions among the elderly
(64+ years old) in Los Angeles to be significantly related to both PM2 5 and PM10_2 5 in one
pollutant models. In the same study, similar magnitudes of excess risk (i.e., in the range of ~4 to
7%) were found in one-pollutant models to be associated with PM2 5 or PM10_2 5 for other age
groups (0 to 19 years; 20 to 64 years) in Los Angeles, as well. In his reanalyses of these GAM
results using the more stringent convergence criteria, however, Moolgavkar (2003) combined all
three Los Angeles age groups into one analysis, providing greater power, but also complicating
before/after comparisons as to the actual effect on the results of using the more stringent
convergence criteria. In the case of the Cook County analyses, the author changed other model
parameters (i.e., the number of degrees of freedom in the model smooths) at the same time as
implementing the more stringent convergence criteria, so direct before/after comparisons were
not possible for Moolgavkar's Chicago reanalyses.
Therefore, in order to provide a one-to-one comparison for Los Angeles, the original age-
specific GAM analyses have been pooled using inverse variance weighting and are presented
along with Moolgavkar's (2003) reanalyses results (in terms of a % increase per 10 |ig/m3 mass
increase for both PM2 5 and PM10) in Table 8-38. As shown in that table, the Moolgavkar
Los Angeles results for all-age COPD admissions for the original and the more stringent
convergence criteria GAM cases (using the same degrees of freedom) are very similar, with the
effects estimate either decreasing (for PM2 5) or increasing (for PM10) very slightly. In those
cases where a much larger number of degrees of freedom were used with either the more
stringent GAM model or a natural spline GLM model, larger reductions in effects estimates were
seen as compared to the original GAM model. For the same number of degrees of freedom, the
natural spline model gave either a slightly larger (for PM2 5) or a slightly smaller (for PM10)
effects estimate than the stringent GAM model. Thus, the reanalyses indicate that use of more
stringent GAM convergence criteria results in minimal changes in PM effect size estimates in
8-234
-------
TABLE 8-38. 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 EFFECT ESTIMATES DERIVED FROM REANALYSES USING
MORE STRINGENT CONVERGENCE CRITERIA AND FOR MODELS SMOOTHED
WITH MORE DEGREES OF FREEDOM
Original Default GAM Reanalysis Stringent Reanalysis Stringent Reanalysis Natural
Model* (30 df) GAM (30 df) GAM (100 df) Spline (100 df)
% Increase/10 jig/m3 % Increase/10 jig/m3 % Increase/10 jig/m3 % Increase/10 jig/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).
this case, as compared to those obtained using the default GAM model; whereas the number of
degrees of freedom used with either GAM or GLM models can result in much larger changes in
the PM effect size estimates and broader confidence intervals.
These various reanalyses results therefore confirm that the PM effect estimates generally
do decline somewhat when using the more stringent convergence criteria, as compared to the
default GAM, with the new estimates being well within the confidence intervals of the original
estimates. However, the effect of using more stringent convergence criteria was seen to have
less influence on the effect estimate than investigator-to-investigator variations in model
specifications (e.g., the extent of smoothing). Overall, then the absolute impact was relatively
small and the basic direction of effect and conclusions regarding the significance of the PM
effect on hospital admissions remained unchanged in these analyses when more stringent GAM
convergence criteria were used.
8-235
-------
8.4.2.3 HEI Commentaries
The HEI Special Report (2003a,c) presents the HEI Special Panel's reviews of both the
Revised Analyses of the National Morbidity, Mortality, and Air Pollution Study (NMMAPS),
Part II and the Revised Analyses of Selected Time-Series Studies, which includes short
communication reports presenting results from other revised analyses of original articles and
reports. Beyond looking at the results of reanalyses designed specifically to address problems
associated with the use of default convergence criteria in the S-Plus GAM function, the reviews
also identified issues associated with the sensitivity of study findings to the use of alternative
modeling approaches that some investigators employed in their reanalyses. In general, the
Special Panel concluded that the original PM effects estimates were more sensitive to the
modeling approach used to account for temporal effects and weather variables than to the
convergence criteria used in the GAM model.
A modeling issue of particular importance highlighted by HEI (2003c) is the sensitivity of
all models (e.g., GAM, GLM-natural splines, GLM-penalized splines) to the degrees of freedom
allotted to potentially confounding weather variables and time. The commentary discusses the
trade-off involved in selecting the number of degrees of freedom for time and weather variables,
while recognizing that there remains no altogether satisfactory way to choose the most
appropriate degrees of freedom. For example, in considering the effect of temperature, if the
degrees of freedom in the smoothing function for temperature are overly restricted, some actual
nonlinear effects of temperature would be falsely ascribed to the pollution variable. To avoid
this, the analyst is tempted to afford many degrees of freedom to temperature or other potentially
confounding variables. However, if more degrees of freedom are allotted than needed, such that
the temperature smooth function is more "wiggly" than the true dose response function, then the
result will be a much less efficient estimate of the pollutant effect. This would have the effect of
incorrectly ascribing part of the true pollution effect to the temperature variable, which would
compromise our ability to detect a true but small pollution effect. The commentary notes that
the empirical data cannot determine the optimal trade-off between these conflicting needs, and it
is difficult to use an a priori biological or meteorologic knowledge to determine the optimal
trade-off. Thus, the Special Panel generally recommended further exploration of the sensitivity
8-236
-------
of these studies both to a wider range of alternative degrees of smoothing and to alternative
specifications of weather variables in time-series models.
More specifically, the HEI Special Panel offered the conclusions and recommendations for
NMMAPS and other revised analyses highlighted below:
NMMAPS Revised Analyses
Dominici et al. (2002) conducted a range of revised analyses, applying alternative methods
to correct shortcomings in the S-Plus GAM programming. HEFs Special Panel review (HEI,
2003a) of this revised analyses yielded the following conclusions:
• 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.
• 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.
• 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.
• 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.
• Weather continues to be a potential confounder of concern, such that further work should
be done on modeling weather-related factors.
Revised Analyses for Other Short Communications
Based on its review, the HEI Special Panel (HEI, 2003c) reached the following
conclusions:
• As was the case with the findings of the original studies, the revised findings will continue to
help inform regulatory decisions regarding PM.
• The PM effect persisted in the majority of studies; however, the number of studies showing
an adverse effect of PM was slightly smaller.
8-237
-------
In some of the large number of studies in which the PM effect persisted, the estimates of PM
effect were substantially reduced.
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.
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
conclusions changed in a meaningful way by the use of stricter criteria compared to the
original analyses.
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.
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.
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.
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.
8.4.3 Assessment of Confounding by Co-Pollutants and Adjustments for
Meteorological Variables
8.4.3.1 Introduction to Assessment of Confounding by Co-Pollutants
Airborne particles are found among a complex mixture of atmospheric pollutants, some of
which are widely measured (such as gaseous criteria co-pollutants O3, CO, NO2, SO2) and others
which are not routinely measured. Determining the extent to which observed ambient PM-health
effects associations can be attributed to airborne particles acting alone or in combination with
other air pollutants or may be due to confounding by other pollutants is one of the more difficult
8-238
-------
issues encountered in assessing PM-related epidemiologic evidence. Because (a) many of the
pollutants are closely correlated due to emissions by common sources and dispersion by
common meteorological factors and (b) some are in the pathway of formation of other pollutants
(e.g., NO —* NO2 —* NO3 ~* Particle Mass), it may be difficult to disentangle their effects (as
noted in Section 8.1.1).
It is widely accepted that some PM metrics are associated with health effects, and that PM
has effects independent of the gaseous co-pollutants. The extent to which ambient gaseous
co-pollutants have health effects independent of PM is important in considering the extent to
which health effects attributed to PM may actually be due in part to co-pollutants or to some
other environmental factors, and vice versa. EPA produces Air Quality Criteria Documents for
four gaseous pollutants: CO, NO2, SO2, and O3 (U.S. Environmental Protection Agency, 1982,
1993, 1996b, 2000b). Health effects of the gaseous pollutants exerted independently from PM,
and in some cases jointly with PM, are discussed in those documents. They are also considered
to some extent in this section and elsewhere in this document because they may affect
quantitative assessments of the effects of various PM metrics when these other pollutants are
also present in the atmosphere. The gaseous pollutants may also be of interest as PM effect
modifiers or through interactions with PM.
Co-pollutant models have received a great deal of attention in the last several years
because there exist improved statistical methods for estimating PM effects by analyses of daily
time-series of mortality (Schwartz and Marcus, 1990; Schwartz, 1991) or hospital admissions
(Schwartz, 1994a,b) and/or in prospective cohort studies (Dockery et al., 1993). A number of
studies using such methods have not only found significant positive relationships between
mortality and one or more PM indicators, but also with one or another of the four gaseous
criteria pollutants (O3, NO2, CO, SO2) in daily time-series studies, and between SO2 and
mortality in the reanalyses of two large prospective cohort studies (Krewski et al., 2000). In the
daily time-series studies, the estimated PM effect is relatively stable when the co-pollutant is
included in the model in some cities, whereas the estimated PM effect in other cities changes
substantially when certain co-pollutants are included. In interpreting the results of any of these
8-239
-------
studies, it is reasonable to consider the biological plausibility of a given pollutant being likely to
affect the particular health endpoint.
Some gaseous co-pollutants (e.g., CO, SO2 and NO2) may be acting as indicators of distinct
emission sources and/or as indicators of PM from these sources. Concentrations of such gaseous
co-pollutants may therefore be correlated with total PM mass or even more strongly correlated
with specific PM constituents (due to their emission from a common source). Thus, one or
another specific gaseous co-pollutant may serve as an indicator of the day-to-day variation in the
contribution of a distinct emission source and to the varying concentrations of airborne PM. In a
model with total PM mass, then, a gaseous co-pollutant may well actually be serving as a
surrogate for the source-apportioned contribution to ambient air PM. Or, PM could also act as
an indicator for emission sources or gaseous co-pollutants. It would be interesting to evaluate
models that include both source-relevant particle components and gaseous pollutants derived
from common sources (e.g., those attributable to motor vehicles, coal combustion, oil
combustion, etc.). The closest approach thus far has been Model II in Burnett et al. (2000), a
default GAM analysis.
The role of gaseous pollutants as surrogates for source-apportioned PM may be distinct
from confounding. The true health effect may be independently associated with a particular
ambient PM constituent that may be more or less toxic than the particle mix as a whole. Thus,
a gaseous co-pollutant may give rise to the appearance of confounding in a regression model.
If it were to serve as an indicator of the more toxic particles, the gaseous co-pollutant could
greatly diminish the coefficient for total particle mass. In such a model, the coefficient for total
particle mass would most properly be interpreted as an indicator of the other, less-toxic particles.
8.4.3.2 Statistical Issues in the Use of Multipollutant Models
Multipollutant models may be useful tools for assessing whether the gaseous co-pollutants
may ^potential confounders of PM effects, but cannot determine if in fact they are. Variance
inflation and effect size instability can occur in nonconfounded multipollutant models as well as
in confounded models. Our usual regression diagnostic tools can only determine whether there
is a potential for confounding. In PM epidemiology studies, the gaseous pollutants, except O3,
8-240
-------
frequently have a high degree of positive linear correlation with PM metrics, a condition known
as multicolinearity; therefore, although multicolinearity leading to effect size estimate instability
and variance inflation are necessary conditions for confounding, they are not sufficient in and of
themselves to determine whether confounding exists.
The most commonly used methods include multipollutant models in which both the
putative causal agent (PM) and one or more putative co-pollutants are used to estimate the health
effect of interest. If the effect size estimate for PM is "stable," then it is often assumed that the
effects of confounding are minimal. "Stable" is usually interpreted as meaning that the
magnitude of the estimated effect is similar in models with PM alone and in models with PM and
one or more co-pollutants, and the statistical significance or width of the confidence interval for
the PM effect is similar for all models, with or without co-pollutants. These criteria (usually
unquantified) diagnose confounding in a narrow sense, interpreted as synonymous with
multicolinearity, not as a failure of the study design or other forms of model mis-specification.
Beyond the conceptual issues discussed above that arise in assessing confounding,
a number of technical issues arise in the use of statistical models, as discussed below.
(a) Model mis-specification assumes many forms. The omission of predictive regressors
("underfilling", defined by Chen et al., 2000) may produce biased estimates of the effects of
truly predictive regressors that are included in the model. Inclusion of unnecessary or
nonpredictive regressors along with all truly predictive regressors ("over-fitting") will produce
unbiased estimates of effect, but may increase the estimated standard error of the estimated
effect if it is correlated with other predictors. Omitting a truly predictive regressor while
including a correlated but noncausal variable ("mis-fitting") will attribute the effect of the causal
regressor to the noncausal regressor. Interaction terms are candidates for omitted regressor
variables. It is important to avoid the "mis-fitting" scenario. Assuming that there is a linear
relationship when the true concentration-response function is nonlinear will produce a biased
estimate of the effect size, high or low, at different concentrations. One of the most common
forms of model mis-specification is to use the wrong set of multiday lags, which could produce
any of the consequences described as "under-fitting" (e.g., using single-day lags when a
8-241
-------
multiday or distributed lag model is needed), "over-fitting" (e.g., including a longer span of days
than is needed), or "mis-fitting" (e.g., using a limited set of lags while the effects are in fact
associated with different set of lags). Different PM metrics and gaseous pollutants may have
different lag structures, so that in a multipollutant model, forcing both PM and gases to have the
same lag structure is likely to yield "mis-fitting." Finally, classical exposure measurement errors
(from use of proxy variables) attenuates (biases) effect size estimates under most assumptions
about correlations among regressors and among their measurement errors (Zeger et al., 2000).
(b) Bias: All of the mis-specifications listed in (a) can bias the effect size estimate except
for "over-fitting" and measurement error of Berkson type. The estimates of the standard error of
the effect size estimate under "over-fitting" or Berkson error cases are inflated, however; and
result in broader confidence intervals than would otherwise occur with a more appropriately
specified model and/or one with less Berkson type measurement error.
(c) Effect size standard error estimates are usually sensitive to model mis-specification.
When all truly predictive regressors are added to an "underfit" model, the reduction in
uncertainty is almost always sufficient to be reflected by the standard errors of estimated effect
size being reduced ("variance deflation"). On the other hand, adding correlated noncausal
variables to "over-fitted" or "mis-fitted" models further increases estimated standard errors
("variance inflation"). Variance inflation can occur whenever a covariate is highly correlated
with the regressor variable that is presumably the surrogate for the exposure of interest.
Confounding with the regressor variable can occur only when the covariate is correlated (a) with
the regressor variable proxy for the exposure of interest and (b) with the outcome of interest in
the absence of the exposure of interest.
(d) Mis-specification errors may compound each other. If the underlying concentration-
response function is nonlinear but there is measurement error in the exposure metrics used, then
different subpopulations may actually have greater or smaller risk than assigned by a linear
model. Consider the hypothetical case of a "hockey-stick" model with a threshold. If there was
8-242
-------
no exposure measurement error, then those in the population with measured concentrations
above the threshold would have excess risk, whereas those below would not. However, if
exposures were measured with error, even if the measured concentration were above the
threshold, some people could experience actual exposures that are, in fact, below the threshold
and, therefore, pose no excess risk. Conversely, if the measured (with error) concentration fell
below the threshold, some people could actually experience concentrations above the threshold
and could be at excess risk. The flattening of a nonlinear concentration-response curve by
measurement error is a well known phenomenon detectable by standard methods (Cakmak
etal., 1999).
(e) Whether effect size estimates and their standard errors are really significantly different
among models is a question not usually addressed quantitatively. Some authors report various
goodness-of-fit criteria such as AIC, BIC, deviance, or over-dispersion index (e.g., Chock et al.,
2000; Clyde et al., 2000), but this is not yet so wide-spread as to assist in analyses of secondary
data for use in this document. Variance inflation may also happen with a correctly specified
model when both pollutants are causal and highly correlated, compared to a model in which only
one pollutant is causal and the noncausal pollutant is omitted. The situation where the variance
or standard error decreases when an additional variable is added (variance deflation) suggests
that the model with the covariate is more nearly correct and that the standard errors of all
covariates may decrease. Statistical significance is a concept of limited usefulness in assessing
or comparing results of many models from the same data set. Still, it is a familiar criterion, and
one addressed here by using a nominal two-sided 5% significance level for all tests and 95%
confidence intervals for all estimates, acknowledging their limitations. There is at present no
consensus on what clearly constitutes "stability" of a model estimate effect size, e.g., effect sizes
that differ by no more than 20% (or some other arbitrary number) from the single-pollutant
models. Simple comparison of the overlap of the confidence intervals of the models is not used
because the model estimates use the same data, and the confidence intervals for effect size in
different models are more-or-less correlated. In analyses with missing days of data for different
pollutants, comparisons must also incorporate differences in sample size or degrees of freedom.
8-243
-------
In any case, statistical comparisons alone cannot fully resolve questions about either
conceptual or statistical issues in confounding via considerations about statistical significance.
If the model is mis-specified in any of the numerous ways described above, then effect size
estimates and/or their estimated standard errors are likely biased.
The most commonly used approach to diagnose potential confounding is fitting
multipollutant models and evaluating the stability of the estimated particle effect sizes against
inclusion of co-pollutants. If an additional covariate is added to a baseline model (e.g., with PM
alone) and the model predicts the outcome better with the covariate, then the reduction in
variance (or deviance for generalized linear or additive models [GLM or GAM]) outweighs the
loss of degrees of freedom for variability. Although not always true, it is reasonable to expect a
decrease in the estimated asymptotic standard error of the effect size estimate ("variance
deflation"), but improved goodness-of-fit may not reduce the standard errors of all parameters in
equal proportion because introducing the new covariate modifies the covariate variance-
covariance matrix. The weighted inverse covariance matrix provides an exact estimate for
standard errors in ordinary linear regression models, and approximately so in GLM or GAM.
The effects on other parameter estimates are rarely reported.
"Variance inflation" may occur under several circumstances, including "under-fitting" and
"mis-fitting" in which a truly predictive covariate is omitted or replaced by a correlated proxy,
and "over-fitting" in which a nonpredictive covariate correlated with the PM metric is also
included in the model. The potential for over-fitting can be diagnosed by evaluating the
eigenvalues of the correlation matrix of the predictors, with very small values identifying near-
colinearity. However, the complete covariate correlation matrix is almost never reported,
including all weather variables and nonlinear functions entered separately as covariates.
Nonetheless, even a correlation matrix among all pollutants would be informative. Furthermore,
composite correlation matrices in multicity studies may conceal important differences among the
correlation matrices.
Multipollutant models may be sensitive to multicolinearity (high correlations among
particle and gaseous pollutant concentrations) and to so-called "measurement errors", possibly
associated with spatial variability. Combining multipollutant models across several cities may
8-244
-------
not improve the precision of the mean combined PM effect-size estimate, if the differences
among the cities are as large or larger in the multipollutant models as in the single-pollutant PM
model. Second-stage regressions have been useful in identifying effect modifiers in the
NMMAPS and APHEA 2 studies, but may not, in general, provide a solution to the problem in
that confounding of effects is a within-city phenomenon.
Three promising alternative approaches versus simple reliance on multipollutant modeling
have begun to be used to evaluate more fully the likelihood that exposures to gaseous
co-pollutants can account for the ambient PM-health effects associations now having been
reported in numerous published epidemiology studies. The first is based on evaluation of
personal exposures to particles and gases, as was done for three panels of participants in
Baltimore, MD (Sarnat et al., 2000, 2001). This study (discussed in detail in Chapter 5) directly
addressed the premise that if individuals are not exposed to a potential confounder, then there is
a lower probability that the potential confounder contributes to the observed effect. The Sarnat
results support the conclusion that personal exposure to sulfates, fine particles, and PM10 are well
correlated with their ambient concentrations measured at corresponding fixed sites, but the
correlations are much lower for PM10_2 5, O3, and NO2. There is, however, a great deal of
variation for one of three two-week panels from one season to the next. The sample size was
small (n = 56), but marginally significant associations were detected between personal and
ambient NO2 for the personal-ambient correlation but was much lower than for particles. There
were, however, some residences in which personal and ambient NO2 were highly correlated.
This has been seen when residences are close to a major road, which was the case for several
members in each of the three studied cohorts (i.e., healthy elderly adults, adults with COPD, and
children 9 to 13 years).
Another promising approach is the use of principal component or factor analysis to
determine which combinations of gaseous criteria pollutants and PM size fractions or chemical
constituents together cannot be easily disentangled, and which pollutants are substantially
independent of the linear combinations of the others. For example, the source-oriented factor
analysis study of Mar et al. (2000) produced evidence suggesting independent effects of regional
8-245
-------
sulfate, motor vehicle-related particles, particles from vegetive burning, and PM10_25 for
cardiovascular mortality in Phoenix (as discussed in Section 8.2.2.4.3).
There are also now available some recent examples of a third promising approach, i.e., the
use of so-called "intervention studies." Interesting evidence for ambient PM effects are
beginning to emerge from some such studies, which relate changes (decreases in health risk
outcomes) to decreases in airborne particles due to deliberate reductions in pollutant emissions
from sources that ordinarily contribute to elevated ambient PM levels in a given locale.
As described before (Section 8.2.3.4), some health outcome changes occurred in some studies in
the presence of low levels of ambient gaseous co-pollutants or little change in at least some of
the co-pollutants in the presence of reduced concentrations of PM mass or constituents.
8.4.3.3 Multipollutant Modeling Outcomes
As stated in the introduction to this chapter, ambient PM exists as a component of a
complex air pollution mixture that includes other criteria pollutants, as well as many other
airborne contaminants that may convey risks to health. Particulate matter is of both primary and
secondary origin, and two of the gaseous criteria pollutants (sulfur dioxide and nitrogen dioxide)
contribute to the formation of secondary particles. Because of shared sources, concentrations of
ambient PM, SO2, and NO2 may be correlated to a moderate degree in urban areas. Generally,
concentrations of PM and other monitored pollutants are imperfect measures of personal
exposures and the extent of measurement error likely varies among the pollutants and also
among population subgroups. In interpreting the findings of multipollutant models, there are
several alternative explanations for observed associations that need to be considered based on the
above points, as follows:
• An effect estimated for PM reflects a "true effect" of paniculate matter (causal
interpretation).
• An effect estimated for PM reflects the total effect of the overall air pollution mixture
(PM is an indicator of mixture toxicity).
• An effect estimated for PM reflects confounding (at least to a degree) by another pollutant
(PM effect is confounded).
8-246
-------
• An effect estimated for PM may be modified by levels of other pollutants (there is effect
modification).
• An effect estimated for PM may be an underestimate of the true effect because of the
inclusion in a model of other criteria air pollutants (e.g., SO2, NO2) which are contributors to
the PM levels observed. This latter effect can be interpreted as the estimated effect of PM
on health not mediated by contributions to PM.
As also stated previously, multipollutant modeling has been one commonly-used method
employed for assessing potential confounding by co-pollutants. Figures 8-16 through 8-19
present results derived from multipollutant models in studies that either did not use GAM
originally or were reanalyzed using GLM.
As shown in Figure 8-16, the single-pollutant PM effect size estimates for total mortality
(with PM10, PM25, and PM10_2 5) in most of the studies did not change much across the various
individual co-pollutants and combinations of co-pollutants as they were added into
multipollutant models, e.g., in the multicity studies by Dominici et al. (2003b) and Schwartz
(2003b) or the single-city studies by Ito (2003), Fairley (2003), and Morgan et al. (1998). One
notable exception is the study by Moolgavkar (2003) in Los Angeles Co., in which the PM effect
estimates were substantially reduced with the inclusion of CO in the model. On the other hand,
in the study in Pittsburgh by Chock et al. (2000), the PM10 effect estimates remained little
changed or were somewhat increased with the inclusion of CO and the other co-pollutants.
For cardiovascular mortality and morbidity (Figure 8-17), in many cases the PM effect
estimates were again little changed when various individual and combinations of co-pollutants
were added to the models, although the pattern seems to be somewhat more variable for
cardiovascular-related effects than for total mortality. For example, in Toronto, PM effects
estimates for cardiovascular hospital admissions for all three PM indicators were appreciably
reduced with the inclusion of NO2, but not CO; the inclusion of all four gaseous co-pollutants
showed the most substantial reductions in the PM effect estimates for each indicator (Burnett
et al., 1997a). Ito (2003) presents results for cardiovascular mortality and hospital admissions in
Detroit and, in most cases, PM effect estimates are similar in models with and without
co-pollutants; some variability is seen across these results, however, with the cardiovascular
mortality effect estimates showing a decrease with the inclusion of either CO or NO2, especially
8-247
-------
00
to
-(^
00
o
Q_
V)
4^
O
111
re
o
v>
(A
O
O
X
111
20-
15-
o
E
Ł
o
= 10
5 -
-10-
X PM Only
PM With:
O Os + Os, NO2
A co A 03, co
O NO2 • CO, NO2
D SO2 • Os, SO2
if 03, CO,
NO2, SO2
1 1
i A i
m 1 »•
•1 ! ? ft "
pllffl . „
rivi-io
< 75 years 75 + years
1 II 1
O
• ?i T «'"( '
> A JjU.
A ,, Wy
1 !H|n P 'I "B"
co-) S „ cog. "ra a » — m
51 1° 1 S3 *S iSl *E
•1 * t || 88 | ^S
E S |< -^ 1-
S 5 g° S
^
CO
o
o
CM
3
[
+
'
K
*-
3
i
1
)
A
i
1
co"
0
o
"ra
1
t Germany
CJ ^
"33 t
^ LU
^
PM2i
<75
years
| 1
1
gavkar (2003)
5 Angeles, CA 1>
:hr,/-L at al fjnnni
o
0
^
H
*
_
75 +
years
r~i
*
L
c'
31
3 C
Q c
s I
L
H
?
3
3
M
5
0.
f
" rj OJ vi
o w s
_l -^ ~
O
o
'x
i
c
c
j
-
I
'
I
C
'^
^
? -"
3 E
3o
>
L
(
>
1
^
<
>
5:
4 >
Fairley(2003)
Santa Clara Co., CA
Morgan et.al. (1998)
Sydney, Australia
PMiO-2.5
<75 75 +
years years
r
;
hnrketal (70001
O
H
<
C
J
.
:
c
n r
H
n
9 <
Chock et al. (2000)
Pittsburgh, PA
>[
? *
i\
r
i
5
B
L
-5 - o
Figure 8-16. Excess risk estimates for total nonaccidental mortality in single-pollutant (PM only) and multipollutant
models. PM increments: 50 ug/m3 for PM10 and 25 ug/m3 for PM25 and PM10_2 5. Results presented from
time-series studies that did not use GAM or were reanalyzed using GLM.
*Estimates for multipollutant models in Ito (2003) obtained from the author via personal communication
(November, 2003).
8-248
-------
00
C
0)
0)
b
c
0)
HI
(0
(A
(A
d>
LU
70-
60-
50-
40-
30-
PM10
X PM Only
O
A
D
PM With:
03 +
CO -^f
N02
03, N02
Os, CO,
N02, S02
PM2..
PM10.2.(
Figure 8-17. Excess risk estimates for cardiovascular-related effects, including mortality, hospital admissions, and
changes in biomarkers (e.g., increases in blood parameters or decreases in heart rate variability measures)
in single-pollutant (PM only) and multipollutant models. PM increments: 50 ug/m3 for PM10 and 25 ug/m3
for PM2 5 and PM10_2 5. Results presented from time-series studies that did not use GAM or were reanalyzed
using GLM. IH = ischemic heart disease; HF = heart failure; HR = heart rate; HRV = heart rate
variability.
*Estimates for multipollutant models in Ito (2003) obtained from the author via personal communication
(November, 2003).
-------
00
to
0
"c
o
E
0
0
Q.
0
Q.
j2
"c
0
LJJ
"J5
i||j
ED Visits
PM With:
O 03 * 03, N02
A CO B Benzene
O N02 -±- 03, CO,
D so2 N02's°2
Hospital
I
Mort Pneu
I II I
i
COPD
o
IT
-I Y
O *J
w2
ff< 1
o O
° in" "c? ™
— m r~'
°% Ła 8 =
— Q jŁ o> !?: co ;r
. _ s< ~° i'
•5 — S °> » "S o' ?
I ipus
So io 5 §
m |H
i-
CO
CO
UJ
1 IB i'
^ i
' ,hl*"
1 1
ill, It
Wl'P
o*~ m c
^^ O i- O) D)
1 °2 || "Sr- 1
j+-
^0
Belfast, Ireland
Ostro et a
Santiago,
5 =j = T E *-
^ ° j Ł o °
1— (^ ~~ TO ? O
0 <" f CN
1'
I
|
?
Hajat et al
tt
' '
3
I
U
c"
o
D
C
O
J
,, (
Ql
Y
11
&
Illl COPD
r
o "
° Ł
5 Q !t
* — o
--S c
li i
23
^ i
Tm I
1 iT'rf
^"g g° S.o>" 5"
•*••* "75 n" ^ O) ^ _J
i_ "JJ c ni ^ ^ ^_i
'S ll p W 1
3S 2
1
>
re
±i
Q
U
r
•
j
i
-CO
1 ii" il
o| ow
— iu-*.i?
Hospital
1 1
|Mort||pneu. COPD
1 II 1 ,
I
T
'i't'fft
-s
Q ^_, r.
If ^ 11
2 Q 3
— m
Respiratory Mortality and Morbidity
Figure 8-18. Excess risk estimates for respiratory-related effects, including mortality, hospital admissions and medical
visits in single-pollutant (PM only) and multipollutant models. PM increments: 50 ug/m3 for PM10 and
25 ug/m3 for PM2 5 and PM10_2 5. Results presented from time-series studies that did not use GAM or were
reanalyzed using GLM. Mort = mortality; Pneu = pneumonia; COPD = chronic obstructive pulmonary
disease.
*Estimates for multipollutant models in Ito (2003) obtained from the author via personal communication
(November, 2003).
-------
00
rb
®
i
u
c
(U
Q.
in
m
CO
10
to
u
X
800 -
t
700 -
.
•
600 -
.
500 -
•
400 -
,
t
100 -
JW
1
.
•
200 -
•
100 -
-10 -
X PMOnly
PMWith:
O o3
D S02
O NO2
A CO, S02
• O3.SO2, NC-2
i i.i |u
Asthma
Symptoms
1 1
Asthma
Symptoms
i
Cough
1 I
HII '
11
1
1 Asthma
'•
Schwartz et al. (1994)
6 U.S. Cities
Mortimer et i
(2002 ) 3 U.S. C
aympioms
1
K
(I
1
i
1
*
/
k
D>
ra
Q
0
1
Yu etal. (2000) Delfin
'• Seattle, WA ,
ities
)
1
Of (U
DJ O3 O) O)
n)«> «>B>
">>>>>>>• >>>,"i>,
Q Q D D Q Q Q Q
1 II 1 1 II 1 1 II 1 1 II 1
Asthma
Symptoms Cough LRI
i 1 i 1 i 1 Morning Morning Evening
i i Peak Peak Peak
1 1 Flow Rate Flow Rate Flow Rate
il \, i
, :!Y Y 1 1 I U 1
Yuetal (2000) Romieu et al. (1996) Gold etal. (1999) Romieu etal. (1996)
Seattle WA Mexico City Mexico City Mexico City
Respiratory Symptoms
Figure 8-19. Excess risk estimates for increases in respiratory symptoms or decreases in lung function measures in
single-pollutant (PM only) and multipollutant models. PM increments: 50 ug/m3 for PM10 and 25 ug/m3
for PM2 5 and PM10_2 5. Results presented from time-series studies that did not use GAM or were
reanalyzed using GLM.
-------
for PM10. In Moolgavkar (2003), the inclusion of CO resulted in variable reductions in the PM10
effect estimates for cardiovascular mortality (L. A. Co.) and hospital admissions (Cook Co.),
although the PM10 estimate for hospital admissions in Cook County remained significant.
As held for cardiovascular-related effects, the PM effect estimates for respiratory-related
mortality and morbidity effects also did not show much change in many cases when various
individual and combinations of co-pollutants were added to the models (Figure 8-18). However,
for some endpoints, PM effect estimates are changed substantially with the addition of specific
co-pollutants, most notably with O3 or NO2. For example, in the Toronto study by Burnett
et al.(1997a), PM effect estimates for respiratory hospital admissions for all three PM indicators
are appreciably reduced with the inclusion of NO2, but not O3; and an even larger reduction was
seen with the inclusion of all four gaseous co-pollutants, as was seen in this study for
cardiovascular hospital admissions. Other Canadian studies of respiratory hospital admissions or
medical visits show appreciable reductions in PM10 and/or PM2 5 effects estimates with the
inclusion of O3 (Thurston et al., 1994; Delfmo et al., 1998b). In Detroit (Ito, 2003), the COPD
hospital admissions effect estimates for PM10 and PM10_2 5 were reduced in models with O3, as
was the respiratory mortality effect estimate for PM10_2 5; whereas the PM effect estimates for
pneumonia hospital admissions were either unchanged or somewhat increased for all three
indicators. As for results of studies on respiratory symptoms and lung function changes
(Figure 8-19), the PM effect estimates were generally robust to adjustment for O3, though
somewhat reduced in a study conducted in Alpine, CA (Delfmo et al., 1998b). Effect estimates
for asthma symptoms were also somewhat reduced in models that included both CO and SO2 in
Seattle (Yu et al., 2000) and in models that included O3, SO2, and NO2 in a 3-city study by
Mortimer et al. (2002).
In addition to the above studies, a number of others only qualitatively reported results for
multipollutant models, but did not provide quantitative results and are thus not included in
Figures 8-16 through 8-19. From this group of studies, some reported that PM effect estimates
remained significant with adjustment for gaseous co-pollutants (e.g., Ostro et al., 2003;
Cifuentes et al., 2000; Sunyer and Basagafia, 2001; Lipsett et al., 1997; Desqueyroux et al.,
2002), while others reported more robust associations with gaseous pollutants (e.g., Lipfert et al.,
8-252
-------
2000a; Stieb et al., 2000; Peters et al., 2000a). Beyond the quantitative results presented above,
Moolgavkar (2003) also describes additional results of multipollutant models in which PM
effects may or may not be robust to the inclusion of gaseous co-pollutants, depending on the
specific lag and co-pollutants used. For example, in Cook County, for a 0-day lag, the PM10
coefficient for total mortality remained robust and statistically significant while coefficients for
each of the gases were attenuated and became insignificant, whereas at a 1-day lag, the PM10
coefficient was attenuated and became insignificant, but coefficients for each of the gases were
robust and remained statistically significant. In some other studies, reductions in PM effect
estimates were reported with adjustment for some gaseous pollutants for some, but not all,
endpoints studied (e.g., Kwon et al., 2001; Prescott et al., 1998). Other authors report that it is
difficult to distinguish among effects of closely correlated pollutants (e.g., Linn et al., 2000, for
CO, NO2 and PM10; Atkinson et al., 1999a, for SO2, NO2 and PM10; Pope et al., 1999a, for CO
and PM10).
For many of the studies discussed above, PM and the gaseous co-pollutants are highly
correlated, especially PM with CO, SO2 and NO2; and it is generally the case that where PM
effect estimates were reduced in size with the inclusion of these co-pollutants, the pollutants
were also highly correlated. Among the studies conducted in the United States, O3 was
positively correlated with the PM indices in Detroit (Ito 2003), Atlanta (Tolbert et al., 2000b)
and Cook County, IL (Moolgavkar, 2003), where in some cases PM effects were reduced with
the inclusion of O3. In other locations, such as Santa Clara County, CA (Fairley, 2003) and
Boston (Peters et al., 2000a),O3 was not correlated with PM and PM effect estimates were not
reported to change in multipollutant models with O3. In contrast with many other U.S. areas,
CO and NO2 were not highly correlated with PM indices in Coachella Valley, CA (Ostro et al.,
2003), and the PM effects estimates for there were reported to be robust to inclusion of gaseous
pollutants. It should also be noted that, in a number of the studies where PM was highly
correlated with the gaseous pollutants, the PM effect estimates were not affected by inclusion of
the gaseous co-pollutants in the models.
Overall, then, a number of the recent studies have reported PM effect estimates that are
robust to adjustment for gaseous co-pollutants and, in a number of studies, independent effects
8-253
-------
of the gaseous pollutants were also found. There are also a number of studies showing generally
independent effects of PM but, for certain health outcomes and co-pollutants, the PM effect
estimate is reduced. For example, in analyses of mortality and hospital admissions data in
Detroit, the authors concluded "...the coefficients of PM mass indices often remain significant in
two-pollutant models, but can be reduced, especially by O3; and gaseous pollutants also are
associated with mortality and morbidity outcomes, but cause specificity of associations has not
been consistent."(Lippmann et al. 2000, p. 33; reanalyzed in Ito, 2003). However, some other
authors have concluded that PM effects were not robust to adjustment for gaseous co-pollutants.
One notable example is the analyses of mortality and hospital admissions data in Cook and
Los Angeles Counties, where the author concluded "... in Los Angeles (with the exception of
COPD admissions with which NO2 appeared to show the most robust association) it is clear that
CO was the best single index of air pollution associations with health endpoints, far better than
the mass concentration of either PM10 or PM2 5. In Cook County the results were not so clear cut.
However, any one of the gases was at least as good an index of air pollution effects on human
health as PM10" (Moolgavkar, 2003, p. 198).
In many of these studies, PM with and without added components of gases thusly appears
to be a key putative agent. However, care must be exercised in interpreting such results, taking
into account what is known about the toxicology and clinical studies of the gases. It is often
clear that these gases, at concentrations present or given the nature of the effects, do not carry
sufficient biologic plausibility to substantially affect the results seen. For example, SO2 is
mostly absorbed in upper airways under normal breathing conditions and, although it might
affect airway neural reflexes to contribute to asthma exacerbation at typical U.S. ambient levels,
it is not likely to exert sufficient effects on COPD or CVD to contribute to excess morbidity and
mortality. Further, because of frequent lack of correlation, separating the effects of PM from O3
seems justified on the basis of simply adjusting one for the other. The same may not be said for
some of the other major gaseous pollutants. It is also the case that the most consistent findings
from amidst the diversity of multipollutant evaluation results for different sites is that the PM
signal most often comes through most clearly.
8-254
-------
8.4.3.4 Bioaerosols as Possible Confounders or Effect Modifiers in PM
Epidemiologic Studies
In addition to possible confounding or effect modification by gaseous co-pollutants,
possible confounding or effect modification by bioaerosols needs to be considered in evaluating
ambient PM epidemiologic findings. As discussed in Chapter 7, various airborne bioaerosols
contain allergens that can contribute to upper (oronasal) respiratory tract irritation (hay fever-
type allergic reactions) and/or to more serious lower respiratory tract effects (e.g., acute
inflammation, bronchoconstriction, exacerbation of asthma attack frequency or intensity, etc.).
A number of epidemiology studies have reported significant associations between ambient
air concentrations of fungal spores and asthma symptoms, hospital admissions, or medical visits
for respiratory diseases (Neas et al., 1996; Delfmo et al., 1996; Delfino et al., 1998a; Delfmo
et al., 2002; Ostro et al., 2001; Stieb et al., 2000; Lewis et al., 2000), although not all found
statistically significant associations (e.g., Tolbert et al., 2000b). Significant associations between
pollen count and respiratory health outcomes have also been reported (Moolgavkar et al., 2000;
Stieb et al., 2000; Lewis et al., 2000), but a number of other studies that evaluated such effects
did not find significant associations with pollen (Thurston et al., 1997; Delfino et al., 1998a;
Delfino et al., 2002; Ostro et al., 2001; Tolbert et al., 2000b; Anderson et al.,1998). Where the
studies have included tests for interaction or potential confounding between aeroallergens and
nonbiological air pollutants for these health responses, all studies have indicated that the
aeroallergen and air pollutant effects were independent, or the authors have concluded that
effects were independent because the aeroallergens and pollutants were poorly correlated (Neas
et al., 1996; Delfino et al., 1996; Delfino et al., 1997b; Delfino et al., 1998a; Delfino et al., 2002;
Stieb et al., 2000; Moolgavkar et al., 2000; Anderson et al., 1998; Lewis et al., 2000).
It is important to emphasize, as discussed in Chapters 3 and 7, that ambient air levels of
bioaerosol components (e.g., fungi, fungal spores, pollen, cytoplasmic fragments of pollen,
endotoxin, or glucan components of bacteria cell walls, etc.) are all seasonally elevated during
warmer, more humid months (e.g., April / May through August / September), but are very low
during colder fall / winter months (October / November through March) in the U.S.
8-255
-------
8.4.3.5 Adjustments for Meteorological Variables
As was noted earlier in Section 8.2.2, it was thought at the time of completion of the 1996
PM AQCD that issues related to model specifications used to control for weather effects in daily
time-series analyses of ambient PM relationships to mortality/morbidity had largely been
resolved. However, as also noted earlier, reanalyses of PM studies to address the GAM
convergence criteria issue led to reexamination of the sensitivity of PM risk estimates to
different model specifications and the consequent reemergence of model specification for control
of weather effects as an important issue in interpreting PM epidemiologic analyses. The
reanalyses results highlighted the sensitivity of modeling outcomes to kinds and numbers of
weather-related variables included in base models and, also, the sensitivity of results to varying
degrees of freedom allotted for smoothing of weather and temporal trends.
Putting the issue of controlling for weather effects into a historical perspective, the 1996
PM AQCD noted that various approaches had previously been used to evaluate potential
contributions by weather to mortality or morbidity effects attributed in different studies to
ambient PM exposures. It noted, as one example, one approach that simply qualitatively
compared PM risk estimates derived from cities differing in typical climatic conditions and
classified as "warm" or "cold" based on long-term mean temperatures — an approach that may
be open to a number of questions, e.g., the fact that the "hot/cold" dichotomy does not
adequately consider cities with moderate climates as part of a broader continuum representative
of the actual range of weather conditions encountered in the United States (as is the case, say, for
San Francisco versus New Orleans or Chicago) or the fact that long-term mean temperatures
over several months are not likely an adequate control for more acute weather changes that are
more likely to affect mortality counts on a daily basis. Other approaches included (a) stratifying
mortality events in relation to one or another weather variable and discarding of the most
extreme days (e.g., the highest X % of mean daily temperature days) from analyses of PM
effects; (b) use of dummy variables that classify days as "hot" or "humid" or "hot/humid" days;
or (c) use of rank-ordered temperatures, or mean temperature for groupings of days, etc. — none
of which, it was noted in the PM AQCD, may provide adequate details to detect actual weather-
mortality relationships.
8-256
-------
The above approaches, to some extent, share certain features that attempt in common to
adjust for the generally recognized nonlinearity of weather influences in mortality/morbidity,
especially with regard to control for temperature effects. Results of various studies noted in the
PM AQCD from the late 1980's and early 1990's provide illustrative examples of outcomes
likely reflective of such nonlinearity of temperature-mortality associations. On the one hand, Ito
et al. (1993) found mortality in London to be associated with BS and aerosol acidity levels and
to a much lesser extent to be affected by weather (perhaps not surprising, given London's
relatively moderate marine clime with relatively infrequent temperature extremes). On the other
hand, several other investigators were noted in the 1996 PM AQCD as finding much more
marked influences of weather in locations experiencing temperature extremes; and some
reported findings indicative of synergistic effects of weather and ambient PM pollution and/or
suggestive of weather exerting notably stronger effects on mortality than pollution. Ramlow and
Kuller (1990), for example, found daily mortality to be more strongly related to prior day
average temperatures than any pollution measure in Pittsburgh (Allegheny Co.) PA, and Wyzga
and Lipfert (1996) reported on apparent synergistic relationship between weather and air
pollution, in that days exceeding 85 °F appeared to contribute most to observed TSP-mortality
relationships in Philadelphia, PA. Kunst et al. (1993) and Machenbach et al. (1993) found
temperature extremes in summer and winter to be primarily determinants of mortality in two
Netherlands studies, with the relationship between temperature and mortality being nonlinear
and characterized by a U-shaped temperature curve with minimum mortality rates seen between
10 to 15 °C.
The 1996 PM AQCD went on to note further the advent of some relatively new approaches
to statistical evaluation of potential weather influences in time-series analyses of ambient PM
effects on mortality or morbidity. One approach, typified by Kalkstein (1991) and Kalkstein
et al. (1994), it was noted, proposes that the meteorology of a given locale is defined by discrete,
identifiable situations that represent frequency modes for combinations of weather elements.
Meteorological delineation of synoptic weather patterns or categories that recognize the
existence of such modes can be used to control for weather in statistical analyses. This view
basically holds that the use of mean weather elements (e.g., mean daily temperature) do not
8-257
-------
permit adequate evaluation of, or control for, daily weather extremes. Also, to the extent that
consideration of weather in most PM/mortality studies focuses almost entirely on thermal
(temperature) and less frequently on moisture (humidity) variables, then PM effect models may
encounter potential weather control problems for some cities affected by certain meteorological
phenomena (e.g., stormy situations associated with mid-latitude cyclones) that are not associated
with thermal extremes and yet can be very important contributors to acute mortality (Kalkstein
et al., 1994). These are rarely controlled for in PM/mortality studies, as they cannot be identified
on the basis of temperature and humidity.
Another approach is one that views adjustment for weather-related variables as being
needed to the extent that any empirical adjustment for such variables provides an adequate basis
for removing potential confounding of excess mortality with PM or other air pollutants. The
1996 PM AQCD noted that one of the most completely empirical methods for adjusting daily
time series data for covariates is by use of nonparametric functions, such as LOESS smoothers,
generalized splines, or GAM, as demonstrated in Schwartz (1994b, 1995, 1997b) and Schwartz
and Morris (1995). These may be empirically satisfactory and provide a better fit to data than
synoptic categories, but at the loss of a basis for defining "offensive" weather episodes.
Application of synoptic climatological procedures to control for weather, it was noted, has the
potential to compensate for these difficulties and may add further insight by defining entire sets
of meteorological conditions which can lead to increases in mortality.
Offensive air masses which lead to mortality totals significantly higher than the long-term
baseline have been identified for a number of U.S. cities, as reported by Kalkstein and Tan
(1995). In most cases "moist tropical" air masses were deemed offensive (especially in the
East), but a very oppressive "dry tropical" air mass category was often associated with the
greatest increases in mortality, especially in New York, St. Louis, Philadelphia, and in
southwestern cities (Kalkstein and Tan, 1995). In some cases, daily mortality totals are over
50% above the baseline (World Health Organization, 1996). Such air mass analyses support the
notion that acute mortality increases only after a meteorological threshold is exceeded. This
threshold is not only temperature dependent; it represents an overall meteorological situation
which is highly stressful. It is noteworthy that most cities demonstrate only one or two types of
8-258
-------
offensive air masses which possess meteorological characteristics exceeding this threshold; and
specific types of oppressive weather patterns associated with increased mortality can vary from
city to city and must be defined individually for a given city for use in statistical analyses of PM
effects.
Detailed analysis of synoptic weather pattern effects in a given city may yield additional
information on specific factors that may need to be accounted for in analyzing PM mortality
effects for that city. For example, the "moist tropical" type of air mass in Philadelphia,
possessing the highest daily minimum and maximum temperatures, both brought mortality
increases and was also associated with the greatest standard deviation in mortality of all air
masses evaluated. That is, although many days within the offensive air mass were associated
with high mortality totals, a number of days showed little mortality increase. The greatest daily
mortality totals during moist tropical air mass incursions occurred as part of a lengthy string of
consecutive days of the air mass, especially when minimum temperatures were particularly high.
This type of information may be important when controlling for weather in PM-mortality
analyses.
In a PM study where stressful weather days are removed from the data base, synoptic
categorization provides an efficient means to remove such days. In studies where weather is
stratified based on certain meteorological elements, synoptic categorization allows for a
meteorologically realistic control and may be preferable to the use of arbitrary dummy variables
when identifying meteorological conditions with an elevated mortality risk.
Evaluation of different weather and time trend model specifications for quantifying PM
concentration-response relationships
The study by Pope and Kalkstein (1996), as discussed in the 1996 PM AQCD, provided
detailed evaluation of the effects of several substantially different approaches to modeling PM
concentration-response relationships and the influence of weather variables on ambient PM
effects. The 1996 PM AQCD noted that the original analyses and reanalyses of the Utah Valley
data by Samet et al. (1995) used quintiles of PM10 as the indicator. The reanalyses reported by
Pope and Kalkstein (1996) as Models 1-8 used a linear model for 5-day moving average PM10
and eight different weather models: (1) no adjustment; (2) indicator variables for 20 seasons
8-259
-------
(1985 to 1990); (3) indicators for 20 seasons and indicators for quintiles of temperature and
relative humidity; (4) indicators for 20 seasons and indicators for 19 synoptic weather categories;
(5) linear time trend and indicators for 19 synoptic categories; (6) LOESS smooth of time
(span = 10 percent of days); (7) LOESS smooths of time (span = 10 percent of days),
temperature (span = 50 percent of days), and relative humidity (span = 50 percent of days); and
(8) LOESS smooth of time (10 percent of days) and indicator variables for 19 synoptic
categories. The results (shown in Table 12-36 of the 1996 PM AQCD and reproduced here as
Table 8-39) were relatively insensitive to the form of time trend and adjustment for weather
variables, with RR for total mortality for 50 |ig/m3 increments in PM10 varying only from -1.058
(Model 2) to 1.112 (Model 10), all of them being statistically significant. The pulmonary
mortality models were somewhat more sensitive to the form of the covariate adjustments, with
RR for 50 |ig/m3 ranging from 1.132 (Model 6) to 1.221 (Model 7); Model 2 showed only a
marginally significant PM10 coefficient, the others being significant with one-tailed (Models 3
and 4) or two-tailed tests. The cardiovascular mortality models had RR ranging from 1.076
(Models 3 and 7) to 1.116 (Model 1), with Model 3 being one-tailed significant and all other
models showing a significant PM10 effect on cardiovascular mortality. While the authors
commented that other communities may show greater sensitivity to the statistical methods for
adjusting for time trend and weather, the relative lack of sensitivity of the estimated PM10 effect
over a very wide range of models is noteworthy.
Table 8-39 also shows subset models that correspond to Models 7 and 8. Cold season
models called Models 9 and 11 by Pope and Kalkstein (1996, Table 4) consist of Models 7 and
8, respectively, limited to the months of October to March. Intra-seasonal differences were
adjusted for by LOESS smoothers of time, and daily weather variation either by LOESS
smoothers of temperature and relative humidity (Model 9) or by indicators for synoptic
categories. Total mortality was highly significant in either case (1.070 for Model 9 and 1.059 for
Model 11). Pulmonary mortality was higher (1.145 for Model 9 and 1.120 for Model 11) and
marginally significant. Cardiovascular mortality had a RR = 1.062 in Model 9 (not significant)
but RR = 1.075 (significant) in Model 11. The corresponding Models 10 and 12 for the warm
season (April-September) showed higher RR effects for total and pulmonary mortality, but the
8-260
-------
TABLE 8-39. EFFECTS OF DIFFERENT MODELS FOR WEATHER AND TIME TRENDS
ON MORTALITY IN UTAH VALLEY STUDY
Model
Identity*
Basel
Base II
1
2
3
4
oo 5
to
^1 6
7
8
9
10
11
12
Relative Risk for PM10 50 fig/m3
Time Model
—
—
None
20 seasons
20 seasons
20 seasons
Linear
LOESS
LOESS
LOESS
Cold season, LOESS
Warm season, LOESS
Cold Season, LOESS
Warm season, LOESS
Weather Model
—
—
None
None
Quintile
Synoptic
Synoptic
None
LOESS
Synoptic
LOESS
LOESS
Synoptic
Synoptic
Total Mortality
1.076
1.083
1.074
1.058
1.062
1.068
1.068
1.059
1.077
1.068
1.070
1.112
1.059
1.091
(1.044,
(1.030,
(1.032,
(1.002,
(1.003,
(1.009,
(1.020,
(1.017,
(1.028,
(1.021,
(1.015,
(0.918,
(1.009,
(0.947,
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.109)
.139)
.118)
.118)
.124)
.130)
.118)
.102)
.129)
.117)
.129)
.346)
.111)
.258)
Pulmonary Mortality
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.198(1.035, 1
.215(1.049, 1
.185(1.056, 1
.133 (0.963, 1
.150(0.972, 1
.169(0.988, 1
.183(1.032, 1
.131(1.006, 1
.221 (1.063, 1
.166(1.018, 1
.145(0.981, 1
.529(0.813,2
.120(0.971, 1
.394 (0.794, 2
.386)
.408)
.331)
.333)
.361)
.382)
.356)
.273)
.402)
.335)
.337)
.877)
.291)
.577)
Cardiovascular
1.094(1.019,
1.094 (1.020,
1.116(1.054,
1.081 (1.000,
1.076 (0.992,
1.090 (1.005,
1.100(1.030,
1.085 (1.024,
1.076 (1.006,
1.099 (1.029,
1.062 (0.984,
1.053 (0.789,
1.075 (1.003,
1.024 (0.780,
Mortality
1.174)
1.174)
1.181)
1.169)
1.167)
1.183)
1.175)
1.150)
1.152)
1.173)
1.146)
1.404)
1.153)
1.343)
*Models 1 through 5 were parametric models, and models 6 through 12 were nonparametric GAM models that have not been reanalyzed.
Source: Pope and Kalkstein (1996).
-------
effects were not statistically significant. The lower statistical significance may reflect the
halving of the sample size in these data sets, since the effect size estimates must be similar to
those obtained by averaging the whole-data analyses across the corresponding seasons, with cold
season = fall + winter approximately, and warm season = spring + summer approximately.
Pope and Kalkstein (1996) also showed four nonparametric smooth regression plots
corresponding to Models 1, 6, 7, and 8, respectively. All of the models using a nonparametric
regression for daily mortality on PM10 were approximately linear, showing some suggestion of
nonlinear structure between roughly 60 and 100 |ig/m3 PM10, but in no case indicating a
threshold or consistent flattening of the concentration-response curve at any PM10 level. The
authors noted that a chi-squared test comparing each nonparametric regression model for PM10
with the corresponding linear model showed no statistically significant deviation from linearity.
The 1996 PM AQCD also discussed another study, by Samet et al. (1996), that compared
different methods for estimating modifying effects of different weather models on the
relationship of TSP and SO2 to total mortality in Philadelphia from 1973 to 1980. The models
included the original Schwartz and Dockery (1992) weather specification, a nonparametric
regression model, LOESS smoothing of temperature and dewpoint, and Kalkstein's Temporal
Synoptic Index (TSI) or Spatial Synoptic Category (SSC) models. The first three methods
allowed the weather model to be adjusted so as to provide an optimal prediction of mortality,
whereas the latter two models were based completely on external criteria and the classification
of days by SSC or TSI categories was not adjusted to improve prediction of mortality. The
authors concluded that". . . the association between air quality as measured by either TSP
alone, SO2 alone, or TSP and SO2 together, cannot be explained by replacing the original
Schwartz and Dockery weather model with either a nonparametric regression, LOESS, or by
synoptic categories using either Kalkstein's TSI or SSC systems. In addition, there is little
evidence in the Philadelphia total mortality data to support the hypothesis that the pollution
effects are modified by the type of weather conditions as measured either by TSI or by strata
created from the predicted weather-induced mortalities using the Dockery and Schwartz model
or the LOESS model. . . . We did not find variation of the effect of pollution across categories of
weather." Their results are not shown here.
8-262
-------
The 1996 PM AQCD noted that additional studies systematically evaluating the differential
effects of PM and other pollutants by weather category would be of interest. The Philadelphia
study by Samet et al. (1996) used only TSP and SO2, whereas the Utah Valley study by Pope and
Kalkstein (1996) did not look at the effects of weather as a modifier with other pollutants as well
as PM10. Still, based on the above two major studies extensively evaluating a number of
different approaches to adjust for weather effects (including evaluations using synoptic weather
patterns), it was concluded in the 1996 AQCD that significant PM-mortality associations were
robust and verifiable via a variety of model specifications controlling for weather.
With the identification of GAM-related statistical issues and in view of new insights gained
from reanalyses addressing those issues (as discussed in Section 8.4.2), questions arise with
regard to the potential resiliency of the findings reported and conclusions drawn based on the
Pope and Kalkstein (1996) and Samet et al. (1996) studies. Given the lack of reanalyses
addressing the GAM issues for these studies, it is neither clear as to the extent that use of GAM
strict convergence criteria or alternative models using natural or penalized splines would confirm
the basic results of the nonparametric GAM analyses in those studies nor as to the magnitude of
reduction in PM effect size estimates that would likely occur. Still, based on likely analogy to
reanalyses results for other GAM-related studies discussed in Section 8.4.2, it would not be
unreasonable to assume that similar reanalyses for the nonparametric models used in these two
new studies could generate PM effect estimates reduced by up to about 50% from the originally
published one. Thus, for example, the Pope and Kalkstein (1996) original estimates for total
mortality could be reduced from ~6 to 11% increase in excess deaths per 50 |ig/m3 PM10 to as
low as ~3 to 5.5%, values that comport very well with the range of results obtained for most
other PM10 total mortality studies. On the other hand, it is less possible at this time to project
likely outcomes of sensitivity analyses that would evaluate effects of markedly varying degrees
of smoothing or degrees of freedom used.
As noted in Section 8.4.2, based on the reanalyses results, the HEI (2003c) Special Panel
Commentary concluded that none of the reanalyzed studies' original conclusions were changed
in a meaningful way by use of stricter convergence criteria. However, the sensitivity of
reanalyses outcomes to the choice of weather variables, to the degree of smoothing, and to the
8-263
-------
numbers of degrees of freedom led the panel to note that neither the appropriate degree of
control for time nor appropriate specification of effects of weather in these time series has been
determined. And the Panel went on to recommend that the sensitivity of these studies to a wide
range of alternative degrees of smoothing and alternative specification of weather variables in
time-series models should continue to be evaluated.
Reanalyses conducted by Ito (2003) of PM10- mortality/hospital admissions associations in
Detroit are illustrative of the sensitivity of PM effect size estimates to alternative model
specifications to control for temporal trends and potential weather effects. Sensitivity analyses
provided by Ito (2003) varying the extent of smoothing for temporal trends and using several
different specifications for weather variables (as well as varying degrees of freedom) provided
results as shown in Figures 8-20 and 8-21 for PM10-mortality and PM10-pneumonia admissions
during 1992-1994 in Detroit. One set of analyses in each figure shows the effect of varying the
extent of smoothing for temporal trends, without inclusion of any adjustments for weather
variables; whereas, the other sets of curves indicate the influence of several different weather
model specifications on the magnitude of the PM10 effect estimate. In general, effect estimates
for models controlling only for temporal trends were notably higher (up to 2- to 3-fold) than
those derived from models adjusting for weather effects. However, most investigators, in fact,
do now make some adjustment(s) for weather. Hence, of most crucial importance are the Ito
(2003) results for the weather adjustment models shown in the two figures. Ignoring the
unadjusted-for-weather line, the coefficients for various weather adjustment models for total
mortality (Figure 8-20) tend to converge in a range from about a factor of ~2 to -1.5 difference
as the period of temporal smoothing is shortened. Decreasing the period length for temporal
smoothing may, to some extent, be washing out the effects of weather (temperature and
dewpoint). The effect of temporal smoothing on air pollution effects can be judged by the
decrease in the coefficients as the temporal smoothing period decreases. Coefficients for
pneumonia admissions (Figure 8-21) appear to be more sensitive to adjustments for temporal
trends. Overall, both temporal smoothing and control for weather can decrease the size of
the PM10 effect estimate.
8-264
-------
0.0015 -
CO.
c
,22
iS 0.0010
IV
o
o
O)
eg
0.0005 -
0.0 -
No control 1 Year 6 Months 3 Months 1 Month 2 Weeks
Approximate Period Length Corresponding to Degrees of Freedom Used in Temporal Smoothing
Figure 8-20. PM10 (lag 1 day) coefficient (p) for total mortality, for 1992-1994, as a
function of alternative weather models and varying degrees of freedom for
fitting temporal trends using natural splines. White circle: natural splines
of same-day temperature and same-day dewpoint, both with df= 2; black
circle: natural splines of same-day temperature (df= 2), the average of
temperatures lagged 1 through 3 days (df= 2), and hot-and-humid day
indicator; white triangle: natural splines of the same-day temperature
(df= 6), the average of temperatures lagged 1 through 3 days (df= 6), same-
day dewpoint (df= 3), and the average of dewpoints lagged 1 through 3 days
(df= 3); x : no adjustment for weather.
Source: Ito (2003).
Most investigators would use periods of length 1 to 3 months for temporal smoothing.
Using such periods, the agreement between the coefficients remained within about 50%, except
that there is more variability between the periods of the temporal smoothing for hospital
admissions. Ito, in a personal communication to EPA, indicated that the degree of temporal
smoothing should depend on the size of the population. That is, for large cities or large multicity
studies, more adjustment is likely warranted; but, for smaller cities, statistical power to detect
the PM10 effect can be greatly reduced by over-adjustment for temporal effects. Considering the
unadjusted-for-weather results, over-adjusting for temporal effects may be allowing this curve to
8-265
-------
0,004 •
,i 0.003 -
u
I
o
H. 0,002 -
><
ro
T3
c? 0.001
o.o-
o
No control 1 Year 6 Months 3 Months 1 Month 2 Weeks
Approximate Period Length Corresponding to Degrees of Freedom Used in Temporal Smoothing
Figure 8-21. PM10 (lag 1 day) coefficient (P) for hospital admissions for pneumonia among
the elderly, for 1992-1994, as a function of alternative weather models and
varying degrees of freedom for fitting temporal trends using natural splines.
White circle: natural splines of same-day temperature and same-day
dewpoint, both with df= 2; black circle: natural splines of same-day
temperature (df= 2), the average of temperatures lagged 1 through 3 days
(df= 2); white triangle: natural splines of the same-day temperature (df= 6),
the average of temperatures lagged 1 through 3 days (df= 6), same-day
dewpoint (df= 3), and the average of dewpoints lagged 1 through 3 days
(df= 3); x : no adjustment for weather.
Source: Ito (2003).
increase as more weather effects become negatively, rather than positively, correlated with PM10
values. This conclusion may apply only to Detroit. There, mortality/pneumonia are higher in
winter and PM10 tends to be higher in summer due to summer sulfate in that region. So,
adjusting for seasonal trends removes the opposing cycles and leaves the positive PM10/mortality
associations.
Some investigators (e.g., Schwartz) note that judgment on the part of the investigator is
required to decide the appropriate amount of smoothing for weather and temporal effects and
8-266
-------
that case-crossover analyses can likely be applied with less judgment considerations. However,
case-crossover can be very sensitive to the choice of control periods (e.g., 7 or 14 days before
and after the case period). So, there are also judgment considerations in applying that method as
well. Still, future use of the case-crossover approach may add information to supplement
outcomes from PM time series analyses.
Several studies published in recent years since the 1996 PM AQCD and not discussed
elsewhere in this chapter provide interesting new information bearing on weather-related effects
that should likely be of value in controlling for weather effects in future PM epidemiologic
analyses. First, the study by Smoyer et al (2000a) of summer weather effects on mortality
among the elderly (> 64 yrs old) showed marked increases in mortality among the elderly over a
17-year period in five Southern Ontario metropolitan areas (Toronto, London, Windsor,
Hamilton, Kitchner-Waterloo-Cambridge) on heat-stress days defined as those with an "apparent
temperature" (heat index) above 32 °C. The study illustrates the likely need to consider
combined effects of temperature and humidity/dewpoint rather than evaluating independent
effects of such variables in controlling for weather effects. Also, relative vulnerability to
heat-related weather effects may be increasing for U.S. populations with the aging of
disproportionate numbers of individuals ("baby boomers") into older middle age groups. This
may complicate future epidemiologic/statistical attempts at disentangling potential PM effects
from those of weather.
Another paper by Smoyer et al. (2000b) evaluated the effects of "offensive weather events"
and air pollution (TSP, O3) in Birmingham, AL and Philadelphia, PA and found that in both
cities offensive weather events had a higher impact on mortality than did high concentrations of
TSP or O3. The authors reported that the highest mortality levels occur when the hottest, but not
necessarily the most polluted, air mass occurs over each of the two cities. Still, lesser increases
in mortality were observed to be associated with TSP in Philadelphia during non-offensive
weather situations, and neither TSP nor O3 seemed to have any add-on effect to weather-related
mortality. In contrast, potential interactive effects seem to be implied by the Birmingham
results. That is, the authors noted that although Birmingham's high-mortality (offensive) air
mass is not the most polluted, offensive air mass days with high pollutant concentrations still
8-267
-------
exhibit higher mean mortality than offensive air mass days with low pollution concentrations.
Also different from the Philadelphia results was the lack of associations in Birmingham between
air pollution levels and mortality on non-offensive air mass days. These results appear to
reinforce the need for city-specific evaluation of weather-related effects; and, they also hint at
potential additive or synergistic effects of weather and air pollution, i.e., the joint occurrence of
elevations of TSP and/or O3 together with high heat-index conditions may increase mortality
over levels than would have been associated with each factor alone.
Two other studies by Braga et al. (2001b, 2002) provide additional interesting results
concerning the time course of weather-related deaths of possible importance for development of
model specifications for control of weather effects in PM epidemiology studies. Braga et al.
(200 Ib) modeled daily death counts in a Poisson regression, examining effects of temperature
and humidity out to lags of 3 weeks and controlling for other covariables (i.e., season, day of
week, barometric pressure) using nonparametric smoothing in GLM analyses for 12 U.S. cities
that represented a wide range of typical climatic conditions and geographic regions (e.g.,
Northeast, Midwest, Northwest, South, etc.). Based on distributed lag modeling, the authors
noted that both high and low temperatures were associated with increased total deaths in "cold"
cities; and that the effects of cold temperatures persisted for days, whereas high-temperature
effects were restricted to the day of death or to the immediately preceding day (likely reflecting
harvesting by high temperatures, as also suggested by other patterns of results). In hot cities,
neither hot nor cold temperatures per se had much effect on mortality; however, the effect of hot
temperature varied with the range of summer temperature variations and the use of air
conditioning. The authors noted that such dissimilarities between cities indicate that analyses of
climate change (and presumably, other weather-related effects) should be taken into account
when evaluating regional differences and temperature-associated harvesting.
The Braga et al. (2002) study used similar methods to evaluate weather-related effects on
cause-specific (i.e., respiratory and cardiovascular) deaths in the same 12 U.S. cities. The effects
and associated lag structures for both temperature and humidity were evaluated based on a
distributed lag model. In cold cities, the authors noted that both low and high temperatures were
associated with CVD deaths, with low-temperature effects persisting for days while high-
8-268
-------
temperature effects were restricted to extreme temperatures on the day of death or the day
before. For myocardial infarction (MI) deaths, hot-day effects were twice those of cold-day
effects; but hot day effects were five times lower than cold days for all CVD causes. Harvesting
effects on hot days were suggested by temporary deficits in numbers of deaths a few days later
(not seen after days with cold temperature deaths). In hot cities, neither hot nor cold temperature
days appeared to affect mortality counts for CVD or pneumonia deaths, but lagged effects were
seen for MI or COPD deaths associated with high temperatures (lagged 4-6 and 3-4 days,
respectively). For respiratory deaths, it was noted that wider variations in summer and winter
temperatures were related to larger effects for hot and cold days, respectively. However, no
clear patterns of results were discerned for humidity effects. Overall, these results suggest that
individuals in cities (e.g., Houston, Atlanta, etc.) with generally warmer weather become adapted
to such and cope better with high-temperature extremes. They also suggest potentially varying
lags for different types of weather-related effects (hot versus cold) and for different cause-
specific endpoints - which may be important to consider in model specifications for control of
weather-related effects in future epidemiologic studies of ambient PM effects.
8.4.4 The Question of Lags
The effect of lag selection on resulting models for PM health effects is an important issue
affecting overall interpretation of epidemiologic analyses. Some interesting and highly
informative points related to lag selection can be discerned based on certain newly available
individual study results and on several illustrative examples comparing lag-related results
obtained by different investigators for analyses of PM effects in the same U.S. city.
Using simulated data with parameters similar to a Seattle PM2 5 data series, Lumley and
Sheppard (2000) showed that potential bias resulting from lag selection can be of similar size
to the relative risk estimates from the measured data. The simulations included a data set
where PM2 5 was significantly associated with hospital admissions for asthma and another data
set where there was no such association. The selection rule used was to choose the single-day
lag (between 0 and 6 days) with the largest estimated relative risk. In the "positive control"
model, where there was a known positive association, the bias associated with selecting the lag
8-269
-------
with the largest effect size from a series of lags was negligible. However, in the analyses using
simulated data where no association was present, selecting the lag with the largest effect size
resulted in a positive bias. The mean bias found in this analysis of simulated data was about half
the size of the effect estimate from a previous publication on associations between PM25 and
asthma hospital admissions (from Sheppard et al., 1999), and the authors reported that the
relative risk from the previous study was at the 90th percentile of the bias distribution in this
analysis. In comparisons to real data from Seattle for other years and from Portland, OR (with
similar weather patterns to Seattle), similar bias issues became evident. Thus, if no association
actually exists in the data, this analysis suggests that selecting the largest risk estimate from a
series of lag periods can lead to potential positive bias toward finding an association.
In considering the results of models for a series of lag days, it is important to consider the
pattern of results that is seen across the series of lag periods. If there is an apparent pattern of
results across the different lags, (such as that seen in Figure 8-22 for results obtained by Peters
et al., 200 Ib), then selecting the single-day lag with the largest effect from a series of positive
associations is reasonable, although it is, in fact, likely to underestimate the overall effect size
(since the largest single-lag day results do not fully capture the risk also distributed over adjacent
or other days). The importance of considering the pattern of results is further illustrated by the
study of Sheppard et al. (1999), in which the pollutant effects reported for asthma hospitalization
at specific lag periods were larger than and consistent with estimates obtained for adjacent lags,
thus lending support for selection of particular lag periods for reporting results. In contrast,
analyses reported by Sheppard et al. (1999) for admissions for appendicitis yielded estimates
from adjacent lags that changed abruptly and an overall unstable pattern was consistent with the
pattern expected for a health endpoint not plausibly associated with air pollution.
In the NMMAPS analysis for mortality, a systematic approach across different data sets
was used to investigate the question of lag selection. The Samet et al. (2000b) analysis, and the
reanalysis by Dominici et al. (2002), for the 90 largest U.S. cities provide particularly useful
information on this matter. Figure 8-23 depicts the Dominici et al. (2002) overall pooled results,
showing the posterior distribution of PM10 effects on total mortality for the 90 cities for lag 0, 1,
and 2 days. It can be seen that the effect size estimate for lag 1 day is about twice that for lag 0
8-270
-------
1.6-.
g i.'
+^
n>
QL
v> 1.24-
"D
T3
O
1.0--
O.f
012345
Hours before onset of Ml
1.6-
w 1.4-
o
(0
CD 1-2-
T3
O
1 0 -
n R .
i
<
»
>
<
>
<
i
<
>
012345
Days before onset of Ml
Figure 8-22. Odds ratios (and 95% confidence intervals) for associations between onset of
myocardial infarction (MI) and 25 ug/m3 increase in hourly (upper panel) or
daily 24-h average (bottom panel) VM2S concentrations. Based on univariate
analyses of ambient PM2 5 data measured in South Boston and interviews of
patients with MI in greater Boston area during Jan. 1995 to May 1996.
Source: Peters et. al. (200Ib).
or lag 2 days, although their distributions overlap. The pattern of lagged effects pooled for each
of the seven regions (see Figure 8-5) in the 90 cities study also shows that the lag with the largest
effect was at 1 day, except for Upper Midwest results where the estimated PM10 effect was about
the same for lag 0 and 1 days.
8-271
-------
Reanalyzed Pooled Estimates
-0,2
0.0
0.2
%/10|jg/m3PIVl10
0.8
1.0
Figure 8-23. Marginal posterior distribution for effects of PM10 on all-cause mortality at
lag 0,1, and 2 for the 90 cities. The numbers in the upper right legend are
posterior probabilities that overall effects are greater than 0.
Source: Dominici et al. (2002).
A review of current studies on short-term health effects of air pollution indicates that there
are essentially three different approaches to deal with temporal structure: (1) assume all sites
have the same lag (e.g., 1 day, for a given effect); (2) use the lag or moving average giving the
largest or most significant effect and for each pollutant and endpoint; and (3) use a flexible
distributed lag model, with parameters adjusted to each site. All three approaches apply to
multicity studies, while the last two also apply to single-city studies. The NMMAPS mortality
analyses used the first approach. This approach introduces a consistent response model across
all locations. However, since the cardiovascular, respiratory, or other causes of acute mortality
usually associated with PM are not at all specific, there is little a priori reason to believe that
8-272
-------
they must have the same relation to current or previous PM exposures at different sites. The
obvious advantage of the first approach in dealing with multicity data is its consistency in
summarizing the point estimate. Conversely, a disadvantage to this approach is that effects may
be underestimated in models using a single lag day. A major factor that makes it difficult to
conduct a meta-analysis of existing PM health effects studies is the lack of consistency in how
lag structures were modeled across the studies.
Figures 8-24 through 8-28 depict results obtained for PM-mortality and/or PM-morbidity
associations as found and reported by different investigators for five U.S. cities. In most single-
city air pollution health effects time-series studies, after the basic model (the best model with
weather and seasonal cycles as covariates) was developed, several pollution lags (usually 0 to
3 or 4 days) were individually introduced and the most significant lag(s) were typically chosen
for presentation of modeling results. Among the U.S. and Canadian mortality studies discussed
in Section 8.2, a number of authors tested associations across a series of lag periods, as was
reported in the NMMAPS multicity analysis. These studies reported stronger associations with
shorter lags, with a pattern of results showing larger associations at the 0- and 1-day lag period
that taper off with successive lag days for varying PM indicators (Moolgavkar, 2003; Ostro
et al., 2000, reanalyzed Ostro et al., 2003; Tsai et al., 2000; Burnett et al., 2000, reanalyzed in
Burnett and Goldberg, 2003; Mar et al., 2000, reanalyzed in Mar et al., 2003; Ito and Thurston,
1996). Several studies used only 0- and 1-day lags in the analyses for PM10, PM2 5 and PM10_2 5
(for example, Schwartz et al., 1996a; Lipfert et al., 2000a; Klemm and Mason, 2000) and Chock
et al. (2000) presented results for models in which 0- to 3- day lags for PM10 were included
simultaneously, with stronger effects generally being seen with the 0-day lag period. However,
some research groups selected longer moving average lag periods for PM10 as providing the best
model fit (Pope et al., 1992, 5-day moving average; Styer et al., 1995, 3-day moving average).
In selecting the lag periods to publish, these authors also discussed the pattern of effect size
estimates across the different lag periods, analogous to what was described previously for
Sheppard et al. (1999, reanalyzed in Sheppard et al., 2003).
Among the U.S. and Canadian studies on cardiovascular and respiratory morbidity, there is
somewhat more variability in which lag periods have been selected for the best-fitting models
8-273
-------
30-
20
oo
to
re
E
LU
x.
'&.
a)
o
x
ill
10
-10
- 0.
N 8
§
IS
O t- N CO <* >O
o "5
0 2
^T
•=
o
Total Mortality
CV Mortality
Resp. Mortality
CV Hospital Admissions
Resp. Hospital Admissions
Figure 8-24. Excess risk estimates for associations between various health outcomes and PM10 (50 ug/m3 increment)
from different studies conducted in Cook County, IL. Results presented are for different PM10 lag
periods from single-pollutant models in studies that either did not use GAM or were reanalyzed; GLM
results presented are from reanalyses.
-------
oo
0)
•+•»
(0
?
(/)
LU
^
(/)
E
V>
(/)
0)
o
X
LU
40-
30-
20 -
10-
-10 -
-20
TO
ro
ro
0
1 8
a S S
i- . ;
ro _: n
•* « T
ro o) c
_ro c „ -
_ ° 'J § c
S- ^ a s S c
O (g Q « 1 1 ^ _
oT3 N, ' ' g "ra ro n
i 5- i 5 Ł~^s* r If! !Q |
_ *O CM > 5 -EtnlnTin--
r § r a > ™|s §0 -
roCLJS o &>_.l|| SO,_
>. O) O <5Oooo^,^,^,^,^,^l5
00 0 ( ( S ____S^ -E,
2Ł.~ OT-CMCO-^-IO =w
E i 1 <• ° ^^
o S^dŁ2-S!Łt o
1 < ' 4 < '
A ' ' < • 1
<• i <• ^'^t T T
X 1 T 1 <» <><> ' I*
I 1*1
| t - o
..
1 II II II II
—
o>
•~^
in
ro
OQ
5 S
> •&
! ro
l_ Z
I <
)
1
g
8
•1
~"3
ro >
E ra
5 §*
~S2.
<
5
c
t
i i
»
<
^ i
i I
^
4
i
r
i
<
.?
=
"•
I
^
0
o
N
ro
0)
2
0
4
< ,
•s s
c c
S .E
0] S
Ul
ro
Jx
ro
«
Total Mortality
CV Mortality Resp. Mortality CV Hospital Admissions
Resp. Hospital Admissions
Asthma Symptoms
180
170
160
150
140
130
120
110
100
90
80
70
60
50
30
20
10
-10
-20
-30
-40
-50
-60
-70
-80
-90
Figure 8-25. Excess risk estimates for associations between various health outcomes and PM10 (50 ug/m3 increment)
from studies conducted in Los Angeles County, CA. Results presented are for different PM10 lag
periods from single-pollutant models in studies that either did not use GAM or were reanalyzed; GLM
results presented are from reanalyses.
-------
20
oo
(0
E
*3
0)
LU
^
V)
E
(A
V)
O
O
X
LU
10-
0
-10
o
o
o
o
o
o
w
ffl
« >
N
1
10
(f>
o
o
s
E
o
Q
O
o
-i i
IS
u
w
O
o
ro Q
*s
o
w
Total Mortality
CV Hospital Admissions
Resp. Hospital Admissions
Figure 8-26. Excess risk estimates for associations between various health outcomes and PM10 (50 ug/m3
increment) from studies conducted in Pittsburgh, PA. Results presented for different PM10 lag
periods are from single-pollutant models in studies that either did not use GAM or were
reanalyzed; GLM results presented are from reanalyses.
-------
40:
oo
to
0}
O
X
Hi
30:
I ^
HI
w 10
E
0
-10
s>
-Q
0
O
O M*
»
>. "3" •
ra to
72 >. 5T
§. •§ §
s Ł - a
1 1 1 *
o Ł1
s «•
"5
_o
"i
o
Q
+
i
) ^
"\
Ł
»
<
t
»
re
•c
1
i—
•C; "S
g1 ,S
S3- ^
|
T
1
~
s" 1
15 o
•e s
o
2 S-
> 8
i (.
> a
s
T
e
5T
O)
_(B
"E
•D
(Q
«
1
O
«"
O
o
o
i i
Ą
"O
_ra' ^
•s | ~ •§
•8 -s. I1 r
-, Ł T3 as
1 N 1- <= J2
5* -L ^
ra — -».
ro
T3
5
i
9 .
S
X -e
Q 51
> 2
<
4
»
9
t
»
4
1 I
i
c e
-, 3,
»
<
*
t
t
1
1
Ł!•_
S
Ti §
%l
fll
S Q.
s
o
V)
r~
•cr *3
X «
TJ
•t
Ł
•<
I
3
i JH
•
3_ •
^
1 i
> 4
«"
01
sr
g
5
i i
»,,_., ^^ "
N ^
I"9
ffl O
Ł
O
to
&
-i S
Q"
! 1
1 i
i I
>
<
<
5 §
1
1
CO Q
°o
ra
3
o
(0
4
>\
"O
ffl"
O
E
0)
c
>
»
Total Mortality
CV Hospital Admissions
Resp. Hospital Admissions
Figure 8-27. Excess risk estimates for associations between various health outcomes and PM10 (50 ug/m3
increment) from studies conducted in Detroit, MI. Results presented are for different PM10 lag
periods from single-pollutant models in studies that either did not use GAM or were reanalyzed;
GLM results presented are from reanalyses.
-------
oo
to
-10:
o
uj -20:
-30:
-40:
-50:
60-
n
o
o
0) -5
ra ra
~ •*-*
.2 ra* N
•a -a r
, , " (0
ra S
v? "w" J™
° 5 o
g S »
i >
_l
<
us
T3
S?
o
/•%
G. B
| 51
Ł <Ł ro
B ^
o i? '
CIS Q"
•—a. -
•o g- >
3 i '
c S
«-< QL
N Q
"C
ra
3
W
1 I
, —
0
a
2J-_n
— "E
ra o
(9 Ł
-C 3
1 c
o
w
"a
•*
T:
• ^
' .
>
i_
1 '
»
i
ET
o
o
~ S1
m "°
fu
3
>-
1
- 150
- 140
- 130
1 120
- 110
'- 100
- 90
- 80
70
- 60
- 50
- 40
- 30
- 20
- 10
- -10
'- -20
- -30
'- -40
1 -50
- -60
- -70
1 -80
'- -90
1-100
--110
1_120
--130
1_140
1 150
Total Mortality
CV Hospital Admissions
Resp. Hospital Admissions
Asthma Symptoms
Figure 8-28. Excess risk estimates for associations between various health outcomes and PM10 (50 ug/m3 increment)
from studies conducted in Seattle or King County, WA. Results presented for different PM10 lag
periods are from single-pollutant models in studies that either did not use GAM or were reanalyzed;
GLM results presented are from reanalyses.
-------
than shown for the mortality studies. In Section 8.3.2, it was found that the time-series studies of
cardiovascular hospital admissions or emergency department visits suggest that PM effects are
stronger at lag 0 with some carryover to lag 1; for cardiac physiology studies the results vary,
with strongest associations for some effects seen with 1- to 2-h lag periods (e.g., Peters et al.,
2001a). Sheppard et al. (1999, reanalyzed Sheppard et al., 2003), reported stronger associations
with asthma hospitalization for 1-day lag periods for PM10, PM10_25 and PM25; and Tolbert et al.
(2000a) reported significant associations for asthma hospitalization in children with 1-day
lagged PM10. Lipsett et al. (1997) and Lin et al. (2002) presented results for asthma
hospitalization or emergency department visits which indicate that longer moving average lag
periods (out to 5- to 7-day moving averages) yield larger PM10 or PM10_2 5 effect estimates and that
the estimates are also fairly consistent across the different lag periods. In panel studies for
respiratory symptoms, several research groups also reported larger effect sizes for longer moving
average lag periods, including 2-, 3- and 4-day lags (e.g., Mortimer et al., 2002; Vedal et al.,1998;
Ostro et al., 2001). Again, however, it is noted that authors generally report finding a pattern of
PM-related effects; for example, Yu et al. (2000) reported a consistent pattern of PM results for
asthma symptoms across 0-, 1- and 2-day lags and selected the 1-day lag for further investigation
in multipollutant models.
It should also be noted that if one chooses the most significant single-lag day only, and if
more than one lag day shows positive (significant or otherwise) associations with mortality, then
reporting a RR for only one lag would also underestimate the pollution effects. Schwartz (2000b;
reanalysis 2003b) investigated this issue, using the 10 U.S. cities data where daily PM10 values
were available for 1986-1993. Daily total (nonaccidental) deaths of persons 65 years of age and
older were analyzed. For each city, a GAM Poisson model (with stringent convergence criteria)
and penalized splines adjusting for temperature, dewpoint, barometric pressure, day-of-week,
season, and time were fitted. Effects of distributed lag were examined using two models:
second-degree distributed lag model using lags 0 through 5 days; and unconstrained distributed
lag model using lags 0 through 5 days. The inverse variance weighted averages of the ten cities'
estimates were used to combine results. The results indicated that the effect size estimates for the
quadratic distributed model and unconstrained distributed lag model using GAM were similar:
8-279
-------
6.3% (CI: 4.9-7.8) per 50 |ig/m3 increase for the quadratic distributed lag model, and 5.8%
(CI: 4.4-7.3) for the other model. These risk estimates are about twice as large as the two-day
average (lag 0 and 1 day) estimate (3.4%; CI: 2.6-4.1) obtained in the reanalysis of the original
10 cities study (Schwartz, 2003b). There are indications that such distributed lag estimates are
even larger when cause-specific of deaths are examined (see 10 U.S. cities study description in
section 8.2.2.3).
The Mar et al. (2000, 2003) study of pollutant-mortality associations in Phoenix offers an
interesting insight into lag structure. It is the only study to have everyday data (except for a few
missing days) for PM10, PM2 5, PM10_2 5, NO2, CO, SO2, and PM source category factors. Phoenix
is also different from most cities studied in two important ways. As a high-temperature city,
associations of mortality with high or low temperatures are minimal and hence more easily
controlled for in data analysis. Correlations of PM25 and PM10_25 between four sites in Phoenix
indicate high correlations for both PM25 and PM10_25 (Smith et al., 2000). In addition, the
mortality data were limited in the Mar et al. (2000, 2003) analyses to ZIP code areas around the
sampling site, further reducing the exposure error.
The pollution variables and the lag days for which the associations with cardiovascular
mortality were statistically significant (p <0.05, GLM with natural splines) are: PM10, 0 and
1; PM10_25, 0; PM25, 1; CO, 1 and 4; NO2, 1 and 4; SO2, 4; regional sulfate, 0; motor vehicle and
resuspended dust (MVRD), 1; vegetative burning, 3. It is reasonable that the PM10_2 5 (lag 0) and
the PM25 (lag 1) would both contribute to a PM10 effect on lag days 0 and 1. Thus, to choose
either a lag 0 or lag 1 for PM10 in Phoenix would underestimate the effects of PM10, given that lag
0 would only capture the PM10_2 5 effects and lag 1 would only capture the PM2 5 effects. The Mar
et al. source category analysis shows an association on lag day 1 for the MVRD factor. High
loadings of CO and NO2 on this source factor suggest that the lag 1 associations with CO and NO2
are due to their high correlation with the MVRD factor. Because the correlation of sulfate
with PM10_25 was only 0.13, the association of the regional sulfate source factor on lag 0 may be
considered to be independent of PM10_2 5. There is also an association of the vegetative burning
factor on lag 3. The effects of these two sources are not strong enough to show up as statistically
significant as part of the PM25 effects, although they may contribute to the positive risks on lag 0
8-280
-------
and 3 and might well contribute to the risk determined in a distributed lag analysis. The
associations on lag 4 are interesting, but as yet unexplained. The Mar et al. (2000, 2003) Phoenix
results suggest that different pollutants are associated with cardiovascular mortality at different
lags. It is also possible that different types of mortality may have different lags for different
pollutants. Hence, the use of PM10 and total mortality would integrate across a variety of
pollutant-type of mortality effects. Selection of any one lag day would neglect associations on
other lag days. Thus, a distributed lag model should more correctly capture all associations. Few
data sets exist with the every day data required for a distributed lag analysis. However, those that
do and that have been analyzed for distributed lag (Schwartz, 2003b), show more excess risk
associated with a distributed lag analysis than from any single day analyses.
An additional complication in assessing the shape of a distributed lag is that the apparent
spread of the distributed lag may depend on the pattern of persistence of air pollution (i.e.,
episodes may persist for a few days), which may vary from city to city and from pollutant to
pollutant. If this is the case, fixing the lag across cities or across pollutants may not be ideal, and
may tend to obscure important nuances of lag structures that might provide important clues to
possible different lags between PM exposures and different cause-specific effects.
One consideration for the evaluation of different lag periods is the availability of data.
Where studies have used PM10 measured on an every-sixth-day sampling schedule, as is common
in many U.S. cities, it is not possible to evaluate multiday lag models, such as moving average
models and distributed lag models, that may be likely to have greater biological plausibility.
It should also be noted that, with the every-sixth-day PM data, a different set of days of mortality
series were evaluated at each lag. For example, an every-other-day sampling schedule was used
in the Harvard Six City Study, for which the PM data on a given day has been used as though it
were a 2-day moving average, alternately concurrent with mortality on half the days and lagging
mortality by one day on the other days.
In summary, the NMMAPS 90 cities study indicated that, of the 0, 1, and 2 day PM10 lags
examined, lag 1 day showed the strongest mortality associations. However, other lags are
reported for various mortality and morbidity outcomes from other studies of one or another
individual city.
8-281
-------
8.4.5 Measurement Error: Concepts and Consequences
8.4.5.1 Theoretical Framework for Assessment of Measurement Error
Since the 1996 PM AQCD, much progress has been made in developing conceptual
frameworks to evaluate potential measurement error effects on the estimation of PM health
effects in time-series studies. Several new studies evaluate the extent of bias caused by
measurement errors under scenarios that differ in extent of error variance and in covariance
structure between co-pollutants.
Zidek et al. (1996) investigated, through simulation, the joint effects of multicolinearity and
measurement error in Poisson regression model, with two covariates with varying extent of
relative errors and correlation. Their error model was of classical error form (W = X + U, where
W and X are surrogate and true measurements, respectively, and the error U is normally
distributed). The results illustrated the transfer of effects from the "causal" variable to the
confounder. However, in order for the confounder to have larger effect size than the true
predictor, the correlation between the two covariates had to be very high (r > 0.9), with moderate
error (a > 0.5) for the true predictor and no error for the confounder in their scenarios. The
transfer-of-causality effect was lessened when the confounder also became subject to error.
Another interesting finding was the behavior of the standard errors of the coefficients: when the
correlation between the covariates was high (r = 0.9) and both covariates had no error, the
standard errors for both coefficients were inflated by a factor of 2; but this phenomenon
disappeared when the confounder had error. Thus, multicolinearity influences the significance of
the coefficient of the causal variable only when the confounder is accurately measured.
Marcus and Chapman (1998) also did a mathematical analysis of PM mortality effects in
ordinary least square (OLS) model with the classical error model, under varying extent of error
variance and correlation between two predictor variables. The error was analytical error (e.g.,
discrepancy between co-located monitors). Only positive regression coefficients were found to
be attenuated; and null predictors (zero coefficient) or weak predictors only appeared stronger
than true positive predictors under unusual conditions, i.e.: (1) true predictors must have very
large positive or negative correlation (i.e., r > 0.9); (2) measurement error must be substantial
(i.e., error variance « signal variance); and (3) measurement errors must have high negative
8-282
-------
correlation. They concluded that fine particle health effects are likely underestimated, but the
bias due to analytical measurement error is not large.
Zeger et al. (2000) illustrated implications of the classical error model and the Berkson error
model (i.e., X = W + U) in the context of time-series study design. Their simulation of the
classical error model with two predictors, with various combinations of error variance and
correlation between the predictors/error terms, showed results similar to those reported by Zidek
et al. (1996). Most notably, for the transfer of the effects of one variable to another (i.e., error-
induced confounding) to be large, the two predictors or their errors must to be highly correlated.
Also, for the spurious association of a null predictor to be more significant than the true predictor,
their measurement errors have to be extremely negatively correlated—a condition not yet seen in
actual air pollution data sets.
Zeger et al. (2000) also laid out a comprehensive framework for evaluating effects of
exposure measurement error on estimates of air pollution mortality relative risks in time-series
studies. The error, i.e., the difference between personal exposure and an ambient pollutant
concentration measured at a community monitoring site, was decomposed into three parts:
(1) the error due to having aggregate rather than individual exposure; (2) the difference between
the average personal exposure and the true ambient concentration level; and, (3) the difference
between the true and measured ambient concentration level. By aggregating individual risks to
obtain expected number of deaths, they found the first component of error (the aggregate rather
than individual) to be a Berkson error and, therefore, not a significant contributor to bias in the
estimated risk. The second error component is a classical error and can introduce bias if short-
term associations exist between indoor source contributions and ambient concentration levels.
Some analyses, however, both using experimental data (Mage et al., 1999; Wilson et al., 2000)
and theoretical interpretations and models (Ott et al., 2000) indicate that there is no relationship
between ambient concentrations and nonambient components of personal exposure to PM. Still, a
bias could arise due to differences between personal exposures to ambient PM (indoors plus
outdoors) and ambient concentrations. The third error component is the difference between the
true and the measured ambient concentration. According to Zeger et al., the final term is largely
8-283
-------
of the Berkson type if the average of the available monitors is an unbiased estimate of the true
spatially averaged ambient level.
Using this framework, Zeger et al. (2000) then used PTEAM Riverside, CA data to estimate
the second error component and its influence on estimated risks. The correlation coefficient
between the error (the average population PM10 total exposure minus the ambient PM10
concentration) and the ambient PM10 concentration was estimated to be -0.63. Since this
y*.
correlation is negative, the/?z (the estimated value of the pollution-mortality relative risk in the
regression of mortality on zt, the daily ambient concentration) will tend to underestimate the
vv
coefficient fix that would be obtained in the regression of mortality on xt, the daily average total
personal exposure, in a single-pollutant analysis. Zeger et al. (2000) then assessed the size of the
bias that will result from this exposure misclassification, using daily ambient concentration, zt.
As shown in Equation 9 of Zeger et al. (2000), the daily average total personal exposure, xt, can
be separated into a variable component, Ql zt, dependent on the daily ambient concentration, zt,
and a constant component, 00, independent of the ambient concentration:
xt = 6>0 + 0{ Zt + st (8-5)
where et is an error term.
If the nonambient component of the total personal exposure is independent of the ambient
concentration, as appears to be the case, Equation 9 from Zeger et al. (2000) becomes the
regression analysis equation familiar to exposure analysts (Dockery and Spengler, 1981; Ott
et al., 2000; Wilson et al., 2000). In this case, 00 gives the average nonambient component of the
total personal exposure and Ql gives the ratio of the ambient component of personal exposure to
the ambient concentration. (The ambient component of personal exposure includes exposure to
ambient PM while outdoors and, while indoors, exposure to ambient PM that has infiltrated
indoors.) In this well-known approach to adjust for exposure measurement error, called
^ >S yv
regression calibration (Carroll et al., 1995), the estimate of ftx has the simple form J3X = /3Z I9l •
Thus, for the regression calibration, the value of /3X (based on the total personal exposure) does
8-284
-------
not depend on the total personal exposure but is given by /?z, based on the ambient concentration,
times 0b the ratio of the ambient component of personal exposure to the ambient concentration.
A regression analysis of the PTEAM data gave an estimate Ql = 0.60.
yv.
Equation 9 from Zeger et al. (2000) was used with 90 = 59.95 and 0X = 0.60, estimated from
the PTEAM data, to simulate values of daily average personal exposure, x*t, from the ambient
concentrations, zt, for PM10 in Riverside, CA, 1987-1994. They then compared the mean of the
simulated ftx s, obtained by the series of log-linear regressions of mortality on the simulated x*t,
/»,
with the normal approximation of the likelihood function for the coefficient Pz from the
y, I y,
log-linear regression of mortality directly on zt. The resulting pj Px = 0.59 is very close to
0j = 0.60. Dominici et al. (2000b) provide a more complete analysis of the bias in fiz as an
estimate of fix using the PTEAM Study and four other data sets and a more complete statistical
n ~
model. Their findings were qualitatively similar in that Px was close to fiz /Oj. Thus, it appears
that the bias is very close to 0l3 which depends not on the total personal exposure but only on the
ratio of the ambient component of personal exposure to the ambient concentration.
Zeger et al. (2000), in the analyses described above, also suggested that the error due to the
difference between the average personal exposure and the ambient level (the second error type
described above) is likely the largest source of bias in estimated relative risk. This suggestion at
least partly comes from the comparison of PTEAM data and site-to-site correlation (the third type
of error described above) for PM10 and O3 in 8 US cities. While PM10 and O3 both showed
relatively high site-to-site correlation («0.6-0.9), a similar extent of site-to-site correlation for
other pollutants is not necessarily expected. Ito et al. (2001) estimated site-to-site correlations
(after adjusting for seasonal cycles) for PM10, O3, SO2, NO2, CO, temperature, dewpoint
temperature, and relative humidity, using multiple stations' data from seven central and eastern
states (IL, IN, MI, OH, PA, WV, WI), and found that, in a geographic scale of less 100 miles,
these variables could be categorized into three groups in terms of the extent of correlation:
weather variables (r > 0.9); O3, PM10, NO2 (r: 0.6-0.8); CO and SO2 (r < 0.5). These results
suggest that the contribution from the third component of error, as described in Zeger et al.
(2000), would vary among pollution and weather variables. Furthermore, the contribution from
8-285
-------
the second component of error would also vary among pollutants; i.e., the ratio of ambient
exposure to ambient concentration, called the attenuation coefficient, is expected to be different
for each pollutant. Some ongoing studies are expected to shed light on this issue, but more
information is needed on attenuation coefficients for a variety of pollutants.
With regard to the PM exposure, longitudinal studies (Wallace, 2000; Mage et al., 1999),
show reasonably good correlation (r = 0.6 to 0.9) between ambient PM concentrations and
average population PM exposure, lending support for the use of ambient data as a surrogate for
personal exposure to ambient PM in time-series mortality or morbidity studies. Furthermore, fine
particles are expected to show even better site-to-site correlation than PM10. Wilson and Suh
(1997) examined site-to-site correlation of PM10, PM25, and PM10_25 in Philadelphia and St. Louis,
and found that site-to-site correlations were high (r « 0.9) for PM2 5 but low for PM10_2 5 (r « 0.4),
indicating that fine particles have smaller errors in representing community-wide exposures. This
finding supports Lipfert and Wyzga's (1997) speculation that the stronger mortality associations
for fine than coarse particles found in the Schwartz et al. (1996a) study may be in part due to
larger measurement error for coarse particles.
However, as Lipfert and Wyzga (1997) suggested, the issue is not whether the fine particle
association with mortality is a "false positive", but rather, whether the weaker mortality
association with coarse particles is a "false negative." Carrothers and Evans (2000) also
investigated the joint effects of correlation and relative error, but they specifically addressed the
issue of fine (FP) versus coarse particle (CP) effect, by assuming three levels of relative toxicity
of fine versus coarse particles (PFP / pcp =1,3, and 10) and, then, evaluating the bias,
(B = |E[PF]/ E[pc,]} / (PF / pc}, as a function of FP-CP correlation and relative error associated
with FP and CP. Their results indicate: (1) if the FP and CP have the same toxicity, there is no
bias (i.e., B=l) as long as FP and CP are measured with equal precision; but, if, for example,
FP is measured more precisely than CP, then FP will appear to be more toxic than CP (i.e.,
B > 1); but (2) when FP is more toxic than CP (i.e., Ppp/Pcp = 3 and 10), the equal precision of FP
and CP results in downward bias of FP (B < 1), implying a relative overestimation of the less
toxic CP. That is, to achieve non-bias, FP must be measured more precisely than CP, even more
so as the correlation between FP and CP increases. In applying this model to real data from the
8-286
-------
Harvard Six Cities Study, estimation of spatial variability for Boston was based on external data
and a range of spatial variability for Knoxville (since no spatial data were available for this city).
For Boston, where the estimated FP-CP correlation was low (r = 0.28), the estimated error was
smaller for FP than for CP (0.85 versus 0.65, as correlation between true versus error-added
series), and the observed FP to CP coefficient ratio was high (11), the calculated FP to CP
coefficient ratio was even larger (26) - thus providing evidence against the hypothesis that FP is
absorbing some of the CP coefficient. For Knoxville, where FP-CP correlation was moderate
(0.54), the error for FP was smaller than for CP (0.9 versus 0.75), and the observed FP to CP
coefficient ratio was 1.4, the calculated true FP to CP coefficient ratio was smaller (0.9) than the
observed value. This indicates that the coefficient was overestimated for the better-measured FP,
while the coefficient was underestimated for the poorer-measured CP. Since the amount (and the
direction) of bias depended on several variables (i.e., correlation between FP and CP; the relative
error for FP and CP; and, the underlying true ratio of the FP toxicity to CP toxicity), the authors
concluded "...it is inadequate to state that differences in measurement error among fine and coarse
particles will lead to false negative findings for coarse particles."
Fung and Krewski (1999) conducted a simulation study of measurement error adjustment
methods for Poisson models, using scenarios like those used in the simulation studies that
evaluated implications of joint effects of correlated covariates with measurement error. The
measurement error adjustment methods used were the Regression Calibration (RCAL) method
(Carroll et al., 1995) and the Simulation Extrapolation (SIMEX) method (Cook and Stefanski,
1994). Briefly, the RCAL algorithm consists of: (1) estimation of the regression of X on W
(observed version of X, with error) and Z (covariate without error); (2) replacement of X by its
estimate from (1) and conducting the standard analysis (i.e., regression); and (3) adjustment of the
resulting standard error of coefficient to account for the calibration modeling. The SIMEX
algorithm consists of: (1) addition of successively larger amount of error to the original data;
(2) obtaining naive regression coefficients for each of the error added data sets; and (3) back
extrapolation of the obtained coefficients to the error-free case using a quadratic or other
function. Fung and Krewski examined the cases for: (1) px = 0.25; pz = 0.25; (2) px = 0.0;
PZ = 0.25; (3) px = 0.25; pz = O.O., all with varying level of correlation (-0.8 to 0.8) with and
8-287
-------
without classical additive error, and also considering Berkson type error. The behaviors of naive
estimates were essentially similar to other simulation studies. In most cases with the classical
error, RCAL performed better than SIMEX (which performed comparably when X-Z correlation
was small), recovering underlying coefficients. In the presence of Berkson type error, however,
even RCAL did not recover the underlying coefficients when X-Z correlation was large (> 0.5).
This is the first study to examine the performance of available error adjustment methods that can
be applied to time-series Poisson regression. The authors recommend RCAL over SIMEX, but
did not discuss possible reasons why RCAL performed better than SIMEX in these scenarios; nor
are the reasons clear from information given in the publication. Also, these error adjustment
methods have not been used in real time-series health effects studies and require either replicate
measurements or some knowledge on the nature of the error (i.e., distributional properties,
correlation, etc.).
Another issue that measurement error may affect is the detection of threshold in time-series
studies. Lipfert and Wyzga (1996) suggested that measurement error may obscure the true shape
of the exposure-response curve, and that such error could have flattened the exposure-response
curve to appear linear even when a threshold may exist. However, based on a simulation with
realistic range of exposure error (due to site-to-site correlation), Cakmak et al. (1999) illustrated
that the modern smoothing approach, LOESS, could adequately detect threshold levels
(12.8 |ig/m3, 24.6 |ig/m3, and 34.4 |ig/m3) even with the presence of exposure error.
Other issues related to exposure error that have not been investigated include potential
differential error among subpopulations. If the exposure errors are different between susceptible
population groups (e.g., people with COPD) and the rest of the population, the estimation of bias
may need to take such differences into account. Also, exposure errors may vary from season to
season, due to seasonal differences in the use of indoor emission sources and air exchange rates
due to air conditioning and heating. This may possibly explain reported season-specific effects of
PM and other pollutants. Such season-specific contributions of errors from indoor and outdoor
sources are also expected to be different from pollutant to pollutant.
In summary, studies that examined joint effects of correlation and error suggest that (a) PM
effects are likely underestimated and (b) spurious PM effects (i.e., qualitative bias such as change
8-288
-------
in the sign of coefficient) due to transferring of effects from other covariates require extreme
conditions and are, therefore, unlikely. Also, one simulation study suggests that, under the likely
range of error for PM, it is unlikely that a threshold is ignored by common smoothing methods.
More data are needed to examine exposure errors for other co-pollutants, since their relative error
contributions will influence their relative significance in relative risk estimates.
8.4.5.2 Measurement Error Issues Related to Divergence Between Monitors and to
Monitoring Frequency
The measurement error framework posed in Dominici et al. (2000b) and Zeger et al. (2000)
explicitly assumes that one of the error components has a Berkson error structure. As noted in
(Zeger et al., 2000, p. 421): "This Berkson model is appropriate when z represents a measurable
factor [e.g., measured PM or another pollutant] that is shared by a group of participants whose
individual [true] exposures x might vary because of time-activity patterns. For example, z might
be the spatially averaged ambient level of a pollutant without major indoor sources and x might
be the personal exposures that, when averaged across people, match the ambient level." This
assumption is likely accurate for sulfates, less so for fine particles and for PM10, and almost
certainly incorrect for gases such as CO and NO2 that may vary substantially on an intra-urban
spatial scale with widely distributed local sources.
The usual characterization of longitudinal or temporal pollutant correlation may not
adequately reflect the spatial variation that is the more crucial aspect of association in evaluating
possible Berkson errors. Temporal correlation coefficients, even across large distances (e.g., Ito
et al., 2001) may be due to large-scale weather patterns affecting concentrations of many
pollutants. Local concentrations for some pollutants with strong local sources and low regional
dispersion (especially for CO and NO2, and PM10_2 5 to a lesser extent) may have somewhat
smaller temporal correlations and much greater relative spatial variations than PM. Thus, persons
in a large metropolitan area may have roughly similar levels of PM exposure x on any given day
for which the ambient average PM concentration z is an adequate surrogate, whatever their space-
time activity patterns, residence, or nonresidential microenvironments, whereas the same
individuals may be exposed to systematically higher or lower concentrations of a co-pollutant
than the spatial average of the co-pollutant. This violates the basic assumption of the Berkson
8-289
-------
error model that within each stratum of the measured (spatially averaged) level z, the average
value of the true concentration x is equal to z, i.e.,
E { X Z } = Z, (8-6)
where E{.} is the average or expected value over the population.
There are empirical reasons to believe that if the strata are chosen to be locations within a
metropolitan area, some individuals far from local sources have consistently less exposure than
the average ambient concentration (denoted p) for co-pollutants from local sources such as CO
and NO2, and PM25, whose true exposure (denoted q) depends on the location of the person's
residence or other micro-environment where most exposure occurs. For this group,
E{q p }< p, (8-7)
while others in locations near the local source (such as a busy highway) have systematically
higher exposure, so that
{q p }
> p-
There is a growing body of evidence that adverse health effects are associated with
proximity to a major road or highway (Wjst et al., 1993; Monn, 2001; Roemer and Van Wijnen,
2001). As shown below, there is good reason to believe that intra-city variation (even in PM25) is
substantial within some U.S. cities. If we assume for the sake of argument that concentrations
of PM10 or PM25 are relatively uniformly distributed, then associations of adverse health effects
with proximity to a source cannot be readily attributed to a pollutant such as PM with a uniform
spatial distribution. NO2 is a pollutant often used to illustrate the spatial nonuniformity of the
gaseous co-pollutants. Monn et al. (1997) compared the concentrations of NO2 and PM10 as a
function of curbside distance in a moderately busy urban street in Zurich and found that PM10
8-290
-------
levels decrease only slightly with increasing distance from the roadway, the decrease more likely
being due to decreasing coarse than decreasing fine particle concentrations. The NO2 levels
showed a much stronger seasonal dependence, decreasing rapidly with increasing distance in the
summer and showing little decrease with distance in the winter. However, the belief that PM2 5 is
spatially uniform should also not be accepted uncritically, as recent analyses for 27 U.S. cities
shown in Chapter 3 and Appendix 3 A of this document demonstrate.
The 90th percentile difference (P90) between a pair of sites may provide a useful guide to the
differences between monitor pairs (and by implication, personal exposure to fine particles) that
might be reasonably expected within a metropolitan area. Table 8-40 shows statistics
summarizing the spatial behavior of PM25 concentrations, based on detailed analyses presented in
Appendix 3 A. The mean Pearson correlation coefficient for all site pairs considered in a given
MSA, the average of the annual mean concentrations and the range of annual means at the sites
considered, the average 90th percentile value (P90) of the absolute concentration difference and
the average coefficient of determination (COD) are shown in Table 8-40 for MSAs satisfying data
completeness criteria used for inclusion in Appendix 3 A. Data in Table 8-40 show the ranges in
the metrics (annual means) considered for all the MSAs included in the analyses. Typically, the
range (i.e., the difference between the lowest and highest means for sites in an MSA) for annual
mean concentrations is about one-quarter of the mean values. However, for about 10% of the
time the average intersite difference in concentrations is greater than roughly one-half of the
annual mean concentration, based on the P90 values. This result suggests that substantial
concentration gradients exist on many days across some MSAs. The effects of outlying sites on
the summary statistics were examined for the Atlanta, GA, Washington, DC, Seattle, WA and
Los Angeles MSAs by removing them from the analyses. Their deletion either had no effect (as
in Washington, DC) or a very large effect (as in Seattle, WA). In addition to outlying monitoring
sites, located outside of the main urban air shed, monitoring sites within the urban core can also
enhance the spatial variability in MSAs, as shown for Detroit, MI. As discussed in Chapter 3,
there are a number of factors that contribute to spatial variability in ambient PM25 concentrations
in urban areas.
8-291
-------
TABLE 8-40. SUMMARY STATISTICS SHOWING MEAN SITE-PAIR PEARSON
CORRELATION COEFFICIENTS, ANNUAL MEAN PM2 5 CONCENTRATIONS
(jig/m3), THE RANGE IN ANNUAL MEAN CONCENTRATIONS (jig/m3),
MEAN OF 90th PERCENTILE DIFFERENCES IN CONCENTRATIONS
BETWEEN ALL SITE PAIRS (jig/m3), AND COEFFICIENTS OF DIVERGENCE
(COD) FOR MSAs MEETING SELECTION CRITERIA GIVEN IN APPENDIX 3A.
VALUE IN ( ) REFERS TO NUMBER OF SITES.
Mean Annual Mean Concentration
Correlation (jig/m3)
Eastern U.S.
Philadelphia, PA-NJ (5)
Washington, DC (6)
Washington, DC* (5)
Norfolk, VA (5)
Columbia, SC (4)
Atlanta, GA (7)
Atlanta, GA* (6)
Birmingham, AL (5)
Tampa, FL (4)
Central U.S.
Cleveland, OH (8)
Pittsburgh, PA (11)
Steubenville, OH-WV (5)
Detroit, MI (10)
Detroit, MI** (9)
Grand Rapids, MI (4)
Milwaukee, WI (8)
Chicago, IL (11)
Gary, IN (4)
Louisville, KY (5)
St. Louis, MO-IL(ll)
Baton Rouge, LA (3)
Kansas City, KS-MO (6)
Dallas, TX (7)
Western U.S.
Boise, ID (4)
Salt Lake City, UT (6)
Seattle, WA (5)
Seattle, WA* (4)
Portland, OR (4)
Los Angeles, C A*** (5)
Los Angeles, CA (6)
Riverside, CA (5)
San Diego, CA (4)
0.89
0.84
0.85
0.96
0.92
0.71
0.78
0.83
0.74
0.9
0.81
0.86
0.89
0.92
0.96
0.9
0.89
0.75
0.9
0.83
0.95
0.91
0.94
0.88
0.91
0.62
0.85
0.83
0.61
0.77
0.89
0.79
15.3
14.6
14.6
13.5
15.6
20.2
20.6
20.3
12.4
17.1
17.9
17.7
16.7
15.9
12.7
13.7
17.6
15.8
17.1
17.4
14.3
12.6
12.6
9.5
11.3
8.9
10.4
7.7
17.9
21
27.5
15.8
Range in Annual Means Mean Pm
(Hg/m3) (ug/m3)
2.3
3.4
3.4
0.7
1.8
4.5
3.5
5.3
1.6
6.2
8.2
2.4
6.4
4.7
1.2
1.3
6.1
3.6
2.7
5.6
0.4
2.4
2.2
1.6
5.0
6.1
3.0
2.8
13.9
5.4
5.0
2.4
5.1
7.1
6.1
3.6
4.3
10.3
8.5
11.5
4.3
8.8
11.8
8.1
9
8.2
4.6
4
7.4
7.7
5.1
8.8
2.7
4.2
4.1
6.4
7.9
10.8
6.1
5.3
21.4
12.2
12.3
8.7
Mean
COD
0.12
0.17
0.17
0.08
0.09
0.18
0.15
0.18
0.12
0.17
0.16
0.18
0.17
0.16
0.13
0.12
0.14
0.18
0.12
0.15
0.08
0.13
0.11
0.17
0.21
0.3
0.17
0.19
0.28
0.18
0.17
0.17
* outlying site removed.
* * interior site removed.
*** Results from analysis including site in Lancaster, CA (included in L.A. MSA, but located across mountains to east of downtown LA)..
8-292
-------
The data shown in Table 8-40 can be used to rank different MSAs according to the relative
degree of spatial homogeneity that concentrations exhibit in them. In Table 8-41, MSAs are first
ranked according to the mean Pearson correlation coefficient (r) for site pairs considered in the
MSA, and then they are ranked according to the average P90 for concentration differences.
It appears that, in general, the western MSAs are not as homogeneous as many in the East, but
there are a number of MSAs in the East where PM25 levels are as heterogeneous as in the West.
It can be seen that there are often substantial differences in rankings according to which of the
two parameters, r or P90, is used. This result suggests that concentration gradients can exist in
MSAs whose monitoring sites are highly correlated and that use of correlation coefficients alone
is not enough to characterize spatial variability. Because of incomplete data capture for some
individual monitors on a given day in a particular MSA, when large concentration gradients exist
across that MSA, day-to-day differences in calculated area-wide 24-h average PM levels may not
accurately reflect the day-to-day changes that would be obtained by the full set of monitors.
The above results provide clear evidence that fine particle concentrations may be less
homogenous in at least some MSAs than has been previously assumed. As noted in Chapter 3,
these differences may not be strictly related to the distance between monitors, especially where
topography and sources of primary PM play a role. In many eastern sites, however, particle
distribution may be more substantially governed by regional rather than by local sources.
Considerable heterogeneity much more often exists across monitoring sites in a given MSA or
county for PM10 values and/or for coarse fraction (PM10_2 5) concentrations typically estimated by
differencing between PM10 and PM2 5 readings on a given day in a given MSA. When the
differencing is done between daily averages for PM10 and PM2 5 values derived from sets of
non-collocated operative monitors in an MSA that may vary from day to day, the resulting
estimates of PM10_2 5 can be subject to considerable error, even leading to such anomalies as
negative values for MSA-wide 24-h average coarse-fraction (PM10_2 5) levels on some days.
Estimates of MSA- or county-wide averages for coarse-fraction thoracic particles derived by first
subtracting PM2 5 from PM10 readings from collocated samplers (or the same dichotomous
sampler with PM10 and PM2 5 cutpoints) and then averaging of those values across the MSA or
county should yield more credible estimates of MSA-or county-wide PM10_25 concentrations.
8-293
-------
TABLE 8-41. SUMMARY OF RELATIVE HOMOGENEITY/HETEROGENEITY
CHARACTERISTICS FOR MSAs GIVEN IN TABLE 8-40. RANKINGS ARE MADE
ACCORDING TO THE MEAN PEARSON CORRELATION COEFFICIENT (left side)
AND 90th PERCENTILE DIFFERENCE IN PM2 5 CONCENTRATIONS (right side).
Spatial
Variability1
East
West
East
West
Relatively
Homogenous
Norfolk, VA
Grand Rapids, MI
Baton Rouge, LA
Dallas, TX
Detroit, MI**
Columbia, SC
Kansas City, KS-MO
Cleveland, OH
Milwaukee, WI
Salt Lake City, UT
Baton Rouge, LA
Norfolk, VA
Milwaukee, WI
Dallas, TX
Kansas City, KS-MO
Columbia, SC
Tampa, FL
Grand Rapids, MI
Louisville, KY
Intermediate
Heterogeneous
Very
Heterogeneous
Louisville, KY
Chicago, IL
Detroit, MI
Philadelphia, PA
Steubenville, OH
Washington, DC*
Washington, DC
St. Louis, MO
Birmingham, AL
Pittsburgh, PA
Atlanta, GA*
Gary, IN
Tampa, FL
Atlanta, GA
Philadelphia, PA
Riverside, CA Washington, DC*
Boise, ID Washington, DC
Chicago, IL
Gary, IN
Seattle, WA* Steubenville, OH
Detroit, MI**
Portland, OR Atlanta, GA*
St. Louis, MO
San Diego, CA Cleveland, OH
Detroit, MI
Los Angeles, CA* Atlanta, GA
Birmingham, AL
Pittsburgh, PA
Seattle, WA
Los Angeles, CA***
Portland, OR
Seattle, WA*
Boise, ID
Salt Lake City, UT
San Diego, CA
Seattle, WA
Los Angeles, CA*
Riverside, CA
Los Angeles, CA***
* outlying site removed. ** interior site removed. *** Results from analysis including site in Lancaster, CA.
Homogeneous: r > .90; P90 < 5.0. Intermediate: r= .80 - .89; P90 = 5.1 - 10.0
Heterogeneous: r = .90 - .79; P90 = 10.1 - 20.0; Very Heterogeneous: r < .69; P90 > 20.0
8-294
-------
Some studies have examined potential implications of variations in addressing the role of
spatial siting of monitors and/or the influence of frequency of data collection on the estimating of
PM effects. Ito et al. (1995) evaluated the influence of the use of different monitors with varying
monitoring schedules on associations between total mortality with PM10 in Cook Co., IL and Los
Angeles Co., CA. The authors used data from six PM10 monitors in Cook Co., with one monitor
operating on a daily basis and the remaining monitors operating on a l-in-6 day schedule, and
four sites in Los Angeles Co., all of which operated on a l-in-6 day schedule. The monitoring
sites were located in urban and suburban settings, according to the State's objectives. Three of
the LA sites were located in residential areas and one was located in an area zoned for
commercial use. One of the Cook Co. sites was classified as residential, two as commercial, and
three as industrial. One of the Cook Co. sites was intended to monitor population exposure, three
to monitor maximum concentrations, and two to monitor both maximum concentrations and
population exposure. There was considerable variation among the distribution of PM10
concentrations in both cities, especially at the upper ends of the distributions. The authors tested
for correlation between individual site pairs, located from 4 to 26 miles apart. The sites were all
temporally well correlated in Cook Co., with correlation coefficients of 0.63 to 0.83 for the site
pairs. Site pairs using three of the Los Angeles Co. sites also had high correlation coefficients
ranging from 0.7 to 0.9, but site pairs that included the fourth monitor had correlation coefficients
ranging from 0.36 to 0.47.
For illustrative epidemiological analyses, Ito et al (1995) then used a sinusoidal model to
account for temporal components. Two methods were used for averaging PM10 data across
monitors: (1) averaging data from all sites, as available; and (2) averaging data from all sites
after first filling in missing data with regression analyses using data from other available
monitors. The authors tested associations between mortality and PM10 measurements averaged
across all sites by each of these methods and at each individual monitor for LA Co. and Cook Co.
Similar results were obtained for both counties, in that (after detrending) the strongest
correlations between PM10 and mortality were found for same-day (lag 0) data in each county
and O3 also showed positive associations for up to 2 days lagged with mortality in each county.
Because more sites were available in Cook Co., additional regression analyses were run to
examine the sensitivity of using data from alternative PM10 sites and/or alternate every 6-day
8-295
-------
samples from one PM10 site. In those analyses, the data from the monitor running every day in
Cook Co. were divided into six data sets to study the influence of the l-in-6 day monitoring
schedule on the PM10-mortality associations observed, and such associations were also evaluated
with a l-in-6 day subset from the two average data sets. The results of these analyses are
summarized in Figure 8-29. Similar associations can be seen using the two methods of
averaging PM10 data. Site 2 is the monitor that operated on a daily basis. The PM10-mortality
association using data from this site is very similar in magnitude to that for the average across
all sites ("avg A"), showing the influence of this site's data on the daily average. Using the PM10
average with data filled in for the remaining sites ("avg B"), the PM10-mortality association is
slightly larger but is not significantly different from the association using the first averaging
method. For analyses using l-in-6 day subsets of the averaged PM10 data ("avg A_6" and
"avg B_6"), the associations are slightly larger (but confidence intervals much wider) than
associations using average data for all study days.
In contrast, there is considerable variation in the associations reported for modeling based
on data from one or another individual monitor, i.e., for sites 1 through 6. In addition, the PM10-
mortality associations shown for the six l-in-6 day subsets of data from site 2 vary widely as
well. The risk estimate sizes derived from the different l-in-6 day data from site 2 vary in a range
similar to that shown for the individual PM10 monitoring sites. The authors observed that it is not
clear whether this sensitivity is due to exposure errors in assigning the PM10 values at individual
sites to the area population, or to exposure errors related to an individual monitor (possibly from
local sources), or to both.
To the extent that the use of less-than-every-day monitoring data is a source of uncertainty
for time-series analyses, it is important to note that many (but not all) U.S. and Canadian time-
series epidemiological studies have used every-day monitoring for PM, with the availability of
daily monitoring data often being described as an important study site selection criterion.
However, a few studies have used data from monitors operating less frequently. One of the more
notable examples is the NMMAPS 90 U.S. cities mortality analyses (Samet et al., 2000b;
Dominici et al., 2003b), where every-day PM10 monitoring data were available for just a few of
the cities (e.g., Pittsburgh, Chicago, St. Paul, Seattle), but PM10 data were collected on varying
8-296
-------
1.25
1.20 -
75 ? 1.15 -
WO. 1.10 -
a) Ł 1.05 H
> O)
re ^ 1.00 4-
— o
01 ? 0.95 -
0.90 -
IV V
-------
2000, reanalyzed Ito, 2003) and Santa Clara County, LA (Lipsett et al., 1997) used PM data
derived from varying sampling schedules.
Lipfert et al. (2000a) examined relationships between the areas in which mortality occurred
among residents and the locations of monitoring sites or averages over monitoring sites for
several particle size components and particle metrics. The mortality data were located for
Philadelphia, PA, for three additional suburban Philadelphia counties, and for Camden, NJ and
other New Jersey counties in the Philadelphia - Camden MSA. A single site was used to obtain
data for fine and coarse particles from Harvard School of Public Health monitors. Additional PA
and NJ thoracic particle data were available for 2 to 4 stations and results were averaged for at
least two stations reporting data. The authors concluded that mortality in any part of the region
may be associated with air pollution concentrations or average concentrations in any other part of
the region, whether particles or gases. The authors suggest two interpretations: (a) the
associations of mortality with pollution were random (from carrying out multiple significance
tests) and not causal, or (b) both particles and gaseous pollutants have a broad regional
distribution. They also noted that interpretation (b) may lead to large uncertainties in identifying
which pollutant exposures for the population are primarily responsible for the observed effects.
These data could be studied further to evaluate smaller-scale spatial relationships among health
effects and gases.
Lippmann et al. (2000) evaluated the effects of monitor siting choice using 14 TSP
monitoring stations in Detroit, MI, and nearby Windsor, ON, Canada. The stations operated from
1981-1987 with almost complete data. When a standard log-linear link Poisson regression model
for mortality was fitted to TSP data for each of the 14 sites, the relative risk estimates were
similar for within-site increments of 5th to 95th percentiles, generally highest and positive at lag
day 1 but not statistically significant at p < 0.05, except for site "w" (site 12, south of the urban
center of Wayne County) and nearly significant at sites "f" (west of the city of Detroit), "g"
(south of the city) and "v" (suburban site in northwestern Wayne County, MI, generally "upwind"
of the urban center). However, as the authors noted, all of the reported relative risks were for
site-specific increments, which varied by a factor of about 2.5 across the Wayne County-Windsor
area. When converted to a common increment of 100 |ig/m3 TSP, the largest excess risks were
found when data used in the model were from site "f' (4.5%), "v" (4.2%), or "w" (3.8%), which
8-298
-------
also showed the most significant effects among the 14 monitors. As the authors noted, "... the
distributional increments [used] to calculate relative risk tend to standardize the scale of relative
risks. This actually makes sense in that if there is a concentration gradient of TSP within a city,
and if the various TSP concentrations fluctuate together, then using a site with a low mean TSP
for time-series analysis would result in a larger coefficient. This result does warn against
extrapolating the effects from one city to an other using a raw regression coefficient [excess
relative risk]."
Other recent studies also point out other aspects of intra-urban spatial variation in PM
concentrations. Kinney et al. (2000) noted that, in a study of personal and ambient PM25 and
diesel exhaust particle (DEP) exposure in a dense urban area of New York City, PM2 5 levels
showed only a moderate site-to-site variation (37 to 47 |ig/m3) due, probably, to broader regional
sources of PM25, but elemental carbon (EC) concentrations showed a four-fold range of site-to-
site variations, reflecting greater local variation in EC as a marker for DEP than for PM2 5 in
general.
Several PM health studies for Seattle (King County), WA (e.g., Levy et al., 2001, for
out-of-hospital primary cardiac arrests) found few statistically significant relationships, partly
attributed by the authors to Seattle having a topographically diverse terrain with local "hot spots"
of residential wood burning, especially in winter. Sheppard et al. (2001) explored reasons for
these findings, focusing on adjustments for location by use of a "topographic index" that included
"downstream" normal flow of wood smoke from higher elevations and trapping of wood smoke
in topographic bowls or basins even at higher elevations. They also adjusted for weather using a
"stagnation index" (average number of hours per day with wind speed < 25th percentile of wind
speeds) and temperature, as well as interaction terms for stagnation on hilltop sites and
temperature at suburban wood smoke-exposed valley sites. Adjustments for exposure
measurement error based on methods developed in Sheppard and Damian (2000) and Sheppard
et al. (2001) had little effect on effect size estimates for the case-crossover study (Levy et al.,
2001), but may be useful in other studies where localized pollutant exposures are believed to be
important.
Daniels et al. (2001) evaluated relative sources of variability or heterogeneity in 1996 PM10
monitoring in the Pittsburgh, PA area, a data-rich area having 25 monitors in a -40 by 80 km
8-299
-------
rectangle. The authors found no isotropic spatial dependence after accounting for other sources
of variability, but indications of (a) heterogeneity in variability of small-scale processes over time
and space and (b) heterogeneity in the mean values and covariate effects across sites. Important
covariates included temperature, precipitation, wind speed and direction. The authors concluded
that significant unmeasured processes might be in operation.
8.4.5.3 Measurement Error and the Assessment of Confounding by Co-Pollutants in
Multipollutant Models
The Zeger et al. (2000) discussion may be interpreted as addressing the extent to which the
apparent lack of a PM10_25 effect in models with both fine and coarse particles demonstrates a
"false negative" due to larger measurement error of coarse particle concentrations. However,
a more important question may involve the relative attenuation of estimated effects of PM25
and gaseous co-pollutants, especially those such as CO that are known to be highly correlated
with PM25. Tables 1 and 2 in Zeger et al. (2000) may be particularly relevant here. The evidence
discussed in this chapter supports the hypothesis that PM has adverse health effects, but leaves
open the question as to the extent the gaseous co-pollutants may contribute to the observed effects
as well when their exposure is measured much less accurately than that of the PM metric. If both
the PM metric and the co-pollutant have effects, Table 1 of Zeger et al. (2000) shows that the
co-pollutant effect size estimate may be greatly attenuated and the PM effect size estimate much
less so, depending on the magnitude of correlation between the true PM and gaseous pollutant
exposures and the correlation between their measurement errors. One would expect that PM2 5,
CO, and NO2 would often have a high positive correlation and their "exposure measurement
errors" would also be positively correlated if PM and the gaseous pollutants were positively
correlated due to common activity patterns, weather, and source emissions. In view of the
substantially greater spatial heterogeneity of traffic-generated ambient pollutants such as CO
and NO2 and the relative (though not absolute) regional spatial uniformity of ambient PM25
in some cities (but not others), it then seems reasonably likely that effect size estimates in
multipollutant models are attenuated downward to a greater extent for gaseous co-pollutants than
for PM metrics in some cities. This may explain part of the heterogeneity of findings for
multipollutant models in different cities. Low effect size estimates for the gaseous co-pollutants
8-300
-------
in a multipollutant model should be interpreted cautiously. The representativeness of the
monitoring sites for population exposure of both the particle metrics and gaseous pollutants
should be evaluated as part of the interpretation of the analysis. Indices such as the maximum
90th percentile of the absolute difference in concentrations between pairs of sites and the median
cross-correlation across sites may be useful for characterizing spatial heterogeneity of gaseous
co-pollutants as well as for particles.
8.4.6 Role of Participate Matter Components
In the 1996 PM AQCD, extensive epidemiologic evidence substantiated very well positive
associations between ambient PM10 concentrations and various health indicators, e.g., mortality,
hospital admissions, respiratory symptoms, pulmonary function decrements, etc. Some studies
were also then available which mortality and morbidity associations with various fine particle
indicators (e.g., PM2 5, sulfate, H+, etc.). One mortality study, the Harvard Six Cities analysis
by Schwartz et al. (1996a), evaluated relative contributions of the fine (PM25) versus the
coarse (PM10_25) fraction of PM10, and found, overall, that PM2 5 appeared to be associated more
strongly with mortality effects than PM10_25. A few studies seemed to be indicative of possible
coarse particle effects, e.g., increased asthma risks associated with quite high PM10 concentrations
in a few locations where coarse particles strongly dominated the ambient PM10 mix.
8.4.6.1 Thoracic Particle (PM10) Mortality/Morbidity Effects
Many new studies have reported associations between mortality and PM10, as discussed in
Section 8.2.2.2. Several new PM epidemiology studies which conducted time-series analyses in
multiple cities were noted to be of particular interest, in that they provide evidence of effects
across various geographic locations (using standardized methodologies) and more precise pooled
effect size estimates with narrow confidence bounds, reflecting the typically much stronger power
of such multicity studies over individual-city analyses to estimate a mean effect. Based on pooled
analyses across multiple cities, using GAM stringent convergence criteria, the percent total
(nonaccidental) excess deaths per 50 |ig/m3 PM10 (24-h) increment were estimated in different
8-301
-------
multicity analyses to be: (a) 1.4% in the 90 largest U.S. cities; (b) 3.4% in 10 large U.S. cities;
(c) 3.6% in the 8 largest Canadian cities; and (d) 3.0% in European cities.
As discussed in Section 8.3.1, a substantial body of new results has emerged since the 1996
PM AQCD that evaluates PM10 effects on cardiovascular-related hospital admissions and visits.
Especially notable new evidence has been provided by multicity studies (Samet et al., 2000a,b;
Zanobetti and Schwartz, 2003b) that yield pooled estimates of PM-CVD effects across numerous
U.S. cities and regions. This study found not only significant PM associations, but also
associations with other gaseous pollutants as well, thus hinting at likely independent effects of
certain gases (O3, CO, NO2, SO2) and/or interactive effects with PM. These and other individual-
city studies generally appear to confirm likely excess risk of CVD-related hospital admission
for U.S. cities in the range of 2 to 9% per 50 |ig/m3 PM10, especially among the elderly
(> 65 years old).
In addition, a number of new studies for respiratory-related hospital admissions and medical
visits have reported results that are generally consistent with and supportive of the findings
presented in the 1996 PM AQCD. As summarized in Section 8.3.3, the excess risk estimates fall
most consistently in the range of 5 to 20% per 50 |ig/m3 PM10, with those for asthma visits and
hospital admissions generally somewhat higher than for COPD and pneumonia hospital
admissions.
8.4.6.2 Fine and Coarse Fraction Particle Effects on Mortality
The 1996 PM AQCD included results from a small number of studies in which air quality
measurements of fine and coarse fraction thoracic particles were used. Some more recent
additional studies are now available that have evaluated associations between various health
outcomes and fine and coarse-fraction particles, the key findings of which are discussed below.
Short-term exposure studies
PM-mortality effect estimates from studies in which both PM2 5 and PM10_2 5 were measured
are shown in Figure 8-5 (Section 8.2.2.5; p. 8-58). Among the more important newly available
results are those derived from reanalyses of two major U.S. and Canadian multicity studies that
investigated associations between PM2 5 and PM10_2 5 and total nonaccidental mortality. These
8-302
-------
include (1) the Schwartz (2003a), the Klemm and Mason (2000) and Klemm and Mason (2003)
reanalyses of the Harvard Six Cities data, all confirming the basic original findings by Schwartz
et al. (1996a); and (2) the Burnett et al. (2000) study of the eight largest Canadian cities and
Burnett and Goldberg (2003) reanalysis of that study. These studies found roughly comparable,
statistically significant excess risk estimates of-2% increased total mortality risk per 25 |ig/m3
PM2 5. In the Harvard Six cities reanalyses, as reported for the original study, PM10_2 5 was not
significantly associated with total mortality across the six cities, though a significant association
was reported for one of the cities (Steubenville, OH). Burnett and colleagues reported an
association of-2% increased total mortality risk per 25 |ig/m3 PM10_25, although it was noted that
this association was more sensitive than that with PM2 5 in the reanalyses.
Effect estimates of about the same size for PM2 5 and PM10_2 5 were reported for single-city
analyses conducted in Philadelphia (Lipfert et al., 2000a), Pittsburgh (Chock et al., 2000), and
Detroit (Ito, 2003), as well as in some areas outside the U.S. such as Santiago, Chile (Cifuentes
et al., 2000). Several U.S. and Canadian studies reported larger effect estimates for PM25 than
for PM10_2 5. Of these, Fairley (2003) reported significant associations for PM25 only for Santa
Clara Co., CA; and in the preliminary analyses by Klemm and Mason (2000), associations with
both PM2 5 and PM10_2 5 in Atlanta, GA did not achieve statistical significance. In two western
areas, Coachella Valley, CA (Ostro et al., 2003) and Phoenix, AZ (Clyde et al., 2000),
associations between mortality and PM10_2 5 were reported to be greater than those for PM2 5.
In all studies, positive associations were reported; however, most associations with PM2 5 were
statistically significant while statistically significant associations with PM10_2 5 were reported for
just a few locations.
In addition, a number of new studies reported associations between both PM25 and PM10_25
with mortality from cardiovascular or respiratory causes. For cardiovascular mortality, Mar et al.
(2003) reported significant associations with both PM2 5 and PM10_2 5, although the confidence
intervals for associations with PM2 5 were very broad. Also, for both cardiovascular and
respiratory mortality, effect estimates of about the same size for PM2 5 and PM10_2 5 were reported
in single-city analyses conducted for Philadelphia (Lipfert et al., 2000a), Santa Clara Co. (Fairley,
2003) and Detroit (Ito, 2003). For both PM2 5 and PM10_2 5, the associations were all positive and
8-303
-------
many were statistically significant for cardiovascular mortality, but for respiratory mortality the
confidence intervals were broader and the associations generally did not reach statistical
significance.
While the associations reported for coarse fraction particles are often not statistically
significant, the findings may well reflect actual associations between mortality and PM10_2 5, at
least in some locations. This may most consistently be the case in arid areas, e.g., in the Phoenix
area (e.g., Mar et al., 2000, 2003) and in Santiago, Chile (Cifuentes et al., 2000). Elevations in
coarse fraction particle mortality risks have also been reported for Steubenville, OH, an eastern
U.S. urban area in the Harvard Six City Study (Schwartz et al., 1996a, Schwartz 2003a; Klemm
et al., 2000; Klemm and Mason, 2003). These results may reflect contamination of later-
resuspended coarse particles by metals in fine particles emitted from smelters (Phoenix) or steel
mills (Steubenville) that was earlier deposited on nearby soils.
Three new papers discussed below provide particularly interesting new information on
relationships between short-term fine and coarse particle exposures and total elderly mortality
(age 65 and older), using TEOM data from the same EPA ORD/NERL monitoring site in
Phoenix, AZ to index PM exposures. Each study , most notably, used quite different models but
each found statistically significant relationships between total mortality and coarse PM
(specifically PM10_2 5) and some associations with PM2-5.
Smith et al. (2000), using a three-day running average as the exposure metric, performed
linear regression of the square root of daily mortality on the long-term trend, meteorological and
air pollution variables. Analyses were done with mortality data obtained for the city of Phoenix,
and for a larger regional area. Two mortality variables were used: (a) total (nonaccidental)
deaths for the city of Phoenix regressed against central EPA site PM25 data (assuming PM2 5
levels to be homogeneous in that area) and (b) total mortality for a larger, regional area within
50 miles around Phoenix regressed against central EPA site PM10_2 5 concentrations (assuming
such levels to be homogeneous in areas around Phoenix). Using linear analysis, associations
between mortality and PM10_2 5 were statistically significant for both regions, whereas associations
with PM2 5 were not. When the possibility of a nonlinear response was taken into account, no
evidence was found for a nonlinear concentration-response relationship for PM10_2 5, but for PM2 5
there was evidence suggesting a threshold for effects at 20 to 25 |ig/m3. There was no evidence
8-304
-------
of confounding between fine and coarse fraction PM, suggesting that the two are "essentially
separate pollutants having distinct effects". Smith et al. (2000) also observed a seasonal effect
for PM10_2 5, the effect being statistically significant only during spring and summer. Based on a
principal component analysis of elemental concentrations, crustal elements are highest in spring
and summer and anthropogenic elements lowest, but Smith et al. (2000) observed that the
implication that crustal, rather than anthropogenic elements, were responsible for the observed
relationship with mortality was counterintuitive.
Clyde et al. (2000) used a more conventional model, a Poisson regression of log-
transformed mortality data with linear pollution variables; and also employed Bayesian model
averaging to consider a wide variety of variations in the basic model. They conducted analyses
that related mortality for three regions (the Phoenix metropolitan area; a small subset of zip codes
to give a region with presumably uniform PM2 5 concentrations; and a still smaller zip code-based
region surrounding the monitoring site that was considered to have uniform PM10 concentrations)
to the air pollution data from the central EPA site. Lag periods of 0, 1, 2, or 3 days were
considered. Stronger associations were reported with PM10_2 5 than PM2 5; the association between
total mortality and PM2 5 was found only in the uniform PM2 5 concentration region.
Mar et al. (2000, 2003) used conventional Poisson regression methods, but limited their
mortality analyses to residents living in zip code locations near the EPA monitoring site (an area
in Phoenix termed "uniform PM10 in Clyde et al., 2000). Mar et al. reported modeling data for
lag days 0 to 4. Positive associations with cardiovascular mortality were reported for both PM2 5
and PM10_2 5; associations were significant at a 0-day lag for PM10_2 5 and at a 1-day lag for PM25.
A significant association was also reported with a regional sulfate factor derived from source
apportionment. The low correlation coefficient between sulfur in PM2 5 (measured by XRF) with
PM10_25 (0.13) suggested separate and distinct effects for regional fine particle sulfate and PM10_25.
In summary, the effects estimates from the newly reported studies are generally consistent
with those derived from the earlier 1996 PM AQCD assessment, which reported risk estimates for
excess total (nonaccidental) deaths associated with short-term (e.g.., 24-hour) PM exposures as
generally falling within the range of 1 to 8% increased per 50 |ig/m3 PM10 and 2 to 6% per
25 |ig/m3 PM2 5; the few earlier studies with PM10_2 5 data provided little evidence for associations
with mortality. Many new single-city studies have reported positive associations (many
8-305
-------
statistically significant at p<0.05) between PM25 and mortality, with effect estimates from U.S.
and Canadian studies typically ranging from 1.5 to 6.5% increase per 25 |ig/m3 PM25 for total
mortality. Excess total mortality risks reported to be associated with short-term exposure
to PM10_25 generally fall in the range of 0.2 to 6.0% increase per 25 |ig/m3 PM10_25, though many
of the results are not statistically significant. Cause-specific estimates appear to mainly fall in
the range of 3.0 to 7.0% increase per 25 |ig/m3 PM25 for cardiovascular or combined
cardiorespiratory mortality and 2.0 to 7.0% increase per 25 |ig/m3 PM25 for respiratory mortality
in U.S. cities. Effect size estimates for PM10_25 generally fall in the range of 3.0 to 8.0% for
cardiovascular mortality and 3.0 to 16.0% for respiratory mortality per 25 |ig/m3 PM10_25.
Long-term exposure / mortality risk studies
Evidence for relationships between long-term exposures to fine and coarse fraction particles
and mortality risk is available from extensive analyses using the Six Cities and ACS cohorts
(original analyses, reanalyses and "extended" analyses with additional cohort follow-up) and
from analyses using the AHSMOG and VA study cohorts. As discussed in Section 8.2.3,
emphasis is placed on the results of the Six Cities and ACS prospective cohort studies, based on
several factors - the larger study population in the ACS study, the larger air quality data set in the
Six Cities study, the more generally representative study populations used in the Six Cities and
ACS studies, and the fact that these studies have undergone extensive reanalyses. These
prospective cohort studies have reported statistically significant risk estimates for total mortality
in the range of 14 to 28% per 10 |ig/m3 PM2 5 (annual average) but no significant associations
with PM10_25. While placing emphasis on the results of the Six Cities and ACS studies, it is noted
that larger associations with PM2 5 than with PM10_2 5 were reported for males in the AHSMOG
cohort, though none of the associations reached statistical significance and the effect estimates
for PM2 5 were in the same range as reported for the ACS and Six Cities cohorts. In the VA study,
the results were more inconsistent from the analyses of differing subsets of data.
Significant associations have also been reported between PM2 5 and cardiorespiratory and
lung cancer mortality in the Six Cities and ACS cohort studies, with effect estimate sizes ranging
from about 6 to 23% per 10 |ig/m3 PM25 for cardiorespiratory mortality and from 8 to 21% per
10 |ig/m3 PM25 for lung cancer mortality. Again, no statistically significant associations have
8-306
-------
been reported between long-term exposure to coarse fraction particles and cause-specific
mortality.
Significant associations for total and cardiopulmonary mortality were also reported with
sulfates, an indicator of fine particles. In the reanalyses of the ACS study, Krewski and
colleagues (2000) reported that the associations between total mortality and PM2 5 or sulfates were
unchanged in models including variables on other risk factors such as personal health or
demographic factors. These associations were also robust to the inclusion of gaseous
co-pollutants in the models, with the exception of SO2. However, SO2 emissions are linked with
the formation of sulfates and secondarily-formed fine particles, so it can be difficult to
disentangle their effects.
Past cross-sectional studies have generally found the fine particle component, as indicated
either by PM2 5 or sulfates, to be the PM constituent most consistently associated with mortality.
While relative measurement errors of various PM indicators must be further evaluated as a
possible source of bias in these estimate comparisons, the new evidence from prospective cohort
studies indicates that the fine mass components of PM are more strongly associated with
mortality effects of chronic PM exposure than are coarse fraction indicators.
8.4.6.3 Source-Oriented Analyses of PM and Mortality
Other recent studies on the relation of mortality to particle composition and source (Laden
et al., 2000; Mar et al., 2000; Tsai et al., 2000) suggest that particles from certain sources may
have much higher potential for adverse health effects than others, as shown by source-oriented
evaluations involving factor analyses. For example, Laden et al. (2000) conducted factor
analyses of the elemental composition of PM2 5 for Harvard Six Cities study data for 1979 to
1988. For all six cities combined, the excess risk for daily mortality was estimated to be 9.3%
(CI: 4.0, 14.9) per 25 |ig/m3 PM25 (average of 0 and 1 day lags) increment in a mobile source
factor; 2.0% (CI: -0.3, 4.4) for a coal source factor, and -5.1% (CI: -13.9, 4.6) for a crustal
factor. There was large variation among the cities and suggestion of an association (not
statistically significant) with a fuel oil factor identified by V or Mn.
Mar et al. (2000) applied factor analysis to evaluate mortality in relation to 1995 to 1997
fine PM elemental components and gaseous pollutants (CO, NO2, SO2) in an area of Phoenix, AZ
8-307
-------
close to the air pollution monitors. The PM2 5 constituents included sulfur, Zn, Pb, soil-corrected
potassium, organic and elemental carbon, and a soil component estimated from oxides of Al, Si,
and Fe. Based on models fitted using one pollutant at a time, statistically significant associations
were found between total mortality and PM10, CO (lags 0 and 1), NO2 (lags 0, 1, 3, 4), S
(negative), and soil (negative). Statistically significant associations were also found between
cardiovascular mortality and CO (lags 0 to 4), NO2 (lags 1 and 4), SO2 (lags 3 and 4), PM2 5 (lags
1,3,4), PM10 (lag 0), PM10_2 5 (lag 0), and elemental, organic, or total carbon. Cardiovascular
mortality was significantly related to a vegetative burning factor (high loadings on organic carbon
and soil-corrected potassium), motor vehicle exhaust/resuspended road dust factor (with high
loadings on Mn, Fe, Zn, Pb, OC, EC, CO, and NO2), and a regional sulfate factor (with a high
loading on S). However, total mortality was negatively associated with a soil factor (high
loadings on Al, Fe, Si) and a local SO2 source factor, but was positively associated with the
regional sulfate factor.
Tsai et al. (2000) analyzed daily time-series of total and cardiorespiratory deaths, using
short periods of 1981-1983 data for Newark, Elizabeth, and Camden, NJ. In addition to inhalable
particle mass (PM15) and fine particle mass (PM2 5), the study evaluated data for metals (Pb, Mn,
Fe, Cd, V, Ni, Zn, Cu) and for three fractions of extractable organic matter. Factor analyses were
carried out using the metals, CO, and sulfates. The most significant sources or factors identified
as predictors of daily mortality were oil burning (targets V, Ni), Zn and Cd processing, and
sulfates. Other factors (dust, motor vehicles targeted by Pb and CO, industrial Cu or Fe
processing) were not significant predictors. In Newark, oil burning sources and sulfates were
positive predictors, and Zn/Cd a negative predictor for total mortality. In Camden oil burning and
motor vehicle emissions predicted total mortality, but copper showed a marginal negative
association. Oil burning, motor vehicle emissions, and sulfates were predictors of
cardiorespiratory mortality in Camden. In Elizabeth, resuspended dust indexed by Fe and Mn
showed marginal negative associations with mortality, as did industrial sources traced by Cu.
The set of results from the above factor analyses studies do not yet allow one to identify
with great certainty a clear set of specific high-risk chemical components of PM. Nevertheless,
some commonalities across the studies seem to highlight the likely importance of mobile source
8-308
-------
and other fuel combustion emissions (and apparent lesser importance of crustal particles) as
contributing to increased total or cardiorespiratory mortality.
8.4.6.4 Fine and Coarse Fraction Particle Effects on Morbidity
Short-term exposure / morbidity studies
A body of new studies published since the 1996 PM AQCD provides further evidence
examining ambient PM association with increased human morbidity. At the time of the 1996 PM
AQCD, fine particle morbidity studies were mostly limited to Schwartz et al. (1994), Neas et al.
(1994, 1995); Koenig et al. (1993); Dockery et al. (1996); and Raizenne et al. (1996); and
discussion of coarse particles morbidity effects was also limited to only a few studies (Gordian
et al., 1996; Hefflin et al., 1994). Since the 1996 PM AQCD, several new studies have been
published in which newly available size-fractionated PM data allowed investigation of the effects
of both fine (PM25) and coarse fraction (PM10_25) particles. PM10, fine (FP) and coarse fraction
(CP) particle results are noted below for studies by morbidity outcome areas, as follows:
cardiovascular disease (CVD) hospital admissions (HA's); respiratory medical visits and hospital
admissions; and respiratory symptoms and pulmonary function changes.
Several new U.S. and Canadian studies evaluated fine-mode PM effects on cardiovascular
outcomes. Lippmann et al. (2000) and Ito (2003) report a positive but not a significant
association with PM2 5; and Moolgavkar (2003) reported PM2 5 to be significantly associated with
CVD HA for lag 0 and 1 in Los Angeles. Burnett et al. (1997a) reported that fine particles were
significantly associated with CVD HA in a single pollutant model, but not when gases were
included in multipollutant models for the 8 largest Canadian city data. Stieb et al. (2000) reported
both PM10 and PM2 5 to be associated with CVD emergency department (ED) visits in single
pollutant, but not multipollutant models. Similarly, Morgan et al. (1998) reported that PM2 5
measured by nephelometry was associated with CVD HA for all ages and 65+ years, but not in
the multipollutant model. Tolbert et al. (2000a) reported that coarse particles were significantly
associated with dysrhythmias, whereas PM2 5 was not. Other studies (e.g., Liao et al., 1999;
Creason et al., 2001; Pope et al., 1999b,c) reported associations between increases in PM25 and
several measures of decreased heart rate variability, but Gold et al. (2000) reported a negative
association of PM25 with heart rate and decreased variability in r-MSSD (one heart rate
8-309
-------
variability measure). A study by Peters and colleagues (200la) reported significant temporal
associations between acute (2-h or 24-h) measures of PM25 and myocardial infarction. Overall,
these new studies collectively appear to implicate fine particles, as well as possibly some gaseous
co-pollutants, in cardiovascular morbidity, but the relative contributions of fine particles acting
alone or in combination with gases such as O3, CO, NO2 or SO2 remain to be more clearly
delineated and quantified. The most difficult issue relates to interpretation of reduced PM effect
size and /or statistical significance when co-pollutants derived from the same source(s) as PM are
included in multipollutant models.
Section 8.3.1 also discussed U.S. and Canadian studies that present analyses of coarse
fraction particles (CP) relationships to CVD outcomes. Lippmann et al. (2000) and Ito (2003)
found significant positive associations of PM10_2 5 with ischemic heart disease hospital admissions
in Detroit (RR = 1.08, CI: 1.04, 1.16). Tolbert et al. (2000a) reported significant positive
associations of heart dysrhythmias with CP (p = 0.04) as well as for elemental carbon (p = 0.004),
but these preliminary results must be interpreted with caution until more complete analyses are
carried out and reported. Burnett et al. (1997b) noted that CP was the most robust of the particle
metrics examined to inclusion of gaseous covariates for cardiovascular hospitalization, but
concluded that particle mass and chemistry could not be identified as an independent risk factor
for exacerbation of cardiorespiratory disease in this study. Based on another Canadian study,
Burnett et al. (1999), reported statistically significant associations for CP in univariate models but
not in multipollutant models; but the use of estimated rather than measured PM exposures indices
limits the interpretation of the PM results reported.
The collective evidence reviewed above, in general, appears to suggest excess risks for
CVD-related hospital admissions of ~1 to 10% per 25 |ig/m3 per PM25 or PM10_25 24-h increment.
Section 8.3.2 also discussed new studies of effects of short-term PM10, PM25, and PM10_25
exposure on the incidence of respiratory hospital admissions and medical visits. Several new
U.S. and Canadian studies have yielded particularly interesting results that are also suggestive of
roles of both fine and coarse particles in respiratory-related hospital admissions. In an analysis of
Detroit data, Lippmann et al. (2000) and Ito (2003) found comparable effect size estimates
for PM2 5 and PM10_2 5. That is, the excess risk for pneumonia hospital admissions (in no
co-pollutant model) was 18.6% (CI: 5.6, 33.1) per 50 |ig/m3 PM10, 10% (CI: 1.5, 19.5) per
8-310
-------
25 |ig/m3 PM25 and 11.2% (CI: -0.02, 23.6) per 25 |ig/m3 PM10.25. Because PM25 and PM10.25
were not highly correlated, the observed association between coarse particles and health outcomes
were possibly not confounded by smaller particles. Despite the greater measurement error
associated with PM10_2 5 than with either PM25 and PM10, this indicator of the coarse particles
within the thoracic fraction was associated with some of the outcome measures. The interesting
result is that PM10_2 5 appeared to be a separate factor from other PM metrics. Burnett et al.
(1997b) also reported PM (PM10, PM25, and PM10_25) associations with respiratory hospital
admissions in 10 Canadian cities, even with O3 in the model. Notably, the PM10_2 5 association
was significant (RR =1.13 for 25 |ig/m3; CI: 1.05, 1.20); and inclusion of ozone still yielded a
significant coarse mass RR = 1.11 (CI: 1.04, 1.19). Moolgavkar (2000a, 2003) reported that, in
Los Angeles, both PM10 and PM2 5 yielded both positive and negative associations at different
lags for single pollutant models but not in two pollutant models. Delfmo et al. (1997a) reported
that both PM2 5 and PM10 are positively associated with ED visits for respiratory disease. Morgan
et al. (1998) reported that PM25 estimated from nephelometry yielded a PM25 association with
COPD hospital admissions for 1-h max PM that was more positive than 24-h average PM2 5.
A new study examines PM associations with asthma-related hospital admissions. Sheppard
et al. (1999) and Sheppard (2003) studied relationships between PM metrics that included PM10_2 5
and non-elderly adult hospital admissions for asthma in the greater Seattle area and reported
significant relative risks for PM10, PM2 5 and PM10_2 5 (lagged 1 day). For PM10_2 5, the relative risk
wasl.05(CI: 1.0, 1.14) and for PM2.5, the relative risk 1.07 (1.02, 1.11).
Thus, although PM10 mass has most often been implicated as the PM pollution index
affecting respiratory hospital admissions, the overall collection of new studies reviewed in
Section 8.3.2 appears to suggest relative roles for PM10 and for both fine and coarse thoracic PM
mass fractions, such as PM2 5 and PM10_2 5.
Section 8.3.3 assessed relationships between PM exposure on lung function and respiratory
symptoms. While most data examined PM10 effects, several studies also examined fine and
coarse fraction particle effects.
Several new asthma studies report associations with ambient PM measures. The peak flow
analyses results for asthmatics tend to show small decrements for both PM10 and PM2 5. Several
studies included PM25 and PM10 independently in their analyses of peak flow. Of these,
8-311
-------
Pekkanen et al. (1997) and Romieu et al. (1996) found comparable results for PM25 and PM10 and
the study of Peters et al. (1997c) found slightly larger effects for PM2 5. Of studies that included
both PM10 and PM25 in their analyses of respiratory symptoms, the studies of Peters et al. (1997c)
and found similar effects for the two PM measures. Only the Romieu et al. (1996) study found
slightly larger effects for PM2 5. While the PM associations with adverse health effects among
asthmatics and others are well documented, the type/source(s) of those particles most associated
with adverse health effects among asthmatics are not known at this time. Indeed, the makeup of
PM varies greatly from place to place and over time, depending upon factors such as the sources
that contribute to the pollution and the prevailing atmospheric conditions, affecting particle
formation, coagulation, transformation, and transport.
For nonasthmatics, several studies evaluated PM2 5 effects. Naeher et al. (1999) reported
similar AM PEF decrements for both PM25 and PM10. Neas et al. (1996) reported a
nonsignificant negative association for PEF and PM2 b and Neas et al. (1999) also reported
negative but nonsignificant PEF results. Schwartz and Neas (2000) reported a significant PEF
association with PM2 5, and Tiittanen et al. (1999) also reported negative but nonsignificant
association between PEF and PM2 5. Gold et al. (1999) reported significant PEF associations
with PM2 5. Schwartz and Neas (2000) reported significant PM2 5 effects relative to lower
respiratory symptoms. Tiittanen et al. (1999) showed significant effects for cough and PM2 5 for
a 4-day average.
Nonasthmatics were evaluated in fewer studies for coarse fraction particle effects. Schwartz
and Neas (2000) report that cough was the only response in which coarse fraction particles
appeared to provide an independent contribution to explaining the increased incidence. The
correlation between CP and PM25 was moderate (0.41). Coarse fraction particles had little
association with evening peak flow. Tiittanen et al. (1999) also reported a significant effect
of PM10_25 for cough. Thus, cough may be an appropriate outcome related to coarse fraction
particle effects. However, the limited data base suggests that further study is appropriate. The
report by Zhang, et al. (2000) of an association between coarse fraction particles and the indicator
"runny nose" is noted also.
The above new studies offer much more information than was available in 1996. Effects
were noted for several morbidity endpoints: cardiovascular hospital admissions, respiratory
8-312
-------
hospital admissions and cough. The data from these relatively limited studies are still insufficient
to allow strong conclusions at this time as to which size-related ambient PM components may be
most strongly related to one or another morbidity endpoints. Very preliminarily, however, fine
particles appear to be more strongly implicated in cardiovascular outcomes than are coarse
fraction particles, whereas both seem to impact respiratory endpoints.
Long-term exposure / morbidity studies
Evidence available in the 1996 PM AQCD included cross-sectional studies using data from
a cohort of children in 24 U.S. and Canadian cities (Dockery et al., 1996; Raizenne et al., 1996).
Positive associations were reported with incidence of respiratory illness or symptoms (Dockery
et al., 1996), and negative associations with lung function measurements (Raizenne et al., 1996)
for changes in PM2 b though the associations were stronger and more likely to reach statistical
significance with measurements of aerosol acidity. Results were not presented for coarse fraction
particles.
The best evidence for effects associated with chronic exposure to fine or coarse-fraction
particles are found in the newer studies that combine the features of cross-sectional and cohort
studies. These include several reports from the Southern California children's cohort study
(McConnell et al., 1999; Gauderman et al., 2000, 2002). McConnell et al. (1999) present results
of cross-sectional analyses using pulmonary function measured upon initiation of the children's
cohort, where positive but not statistically significant associations were reported with some
measures of increased respiratory illness in children with asthma; no associations were reported
for children without asthma. Gauderman et al. (2000, 2002) analyzed lung function growth using
spirometry measurements made in 4th and 7th grades for two separate cohorts, each having
been recruited as 4th grade children. Gauderman et al. (2000) reported results for both PM2 5
and PM10_2 5, noting that decreased lung function growth was associated with 10 |ig/m3 changes
for both indices (some but not all associations reached statistical significance). Gauderman et al.
(2002) also later reported decreased lung function growth with PM2 5 (though the associations
were not statistically significant) but did not report results using PM10_2 5 data.
The recent studies suggest that long-term exposure to fine particles is associated with
development of chronic respiratory disease and reduced lung function growth; little evidence is
-------
available on potential effects of exposure to coarse fraction particles. These findings build upon
the information available in the 1996 PM AQCD. As was true then, there are fewer studies of
long-term exposure effects than short-term exposures, but the evidence indicates fine particle
exposures may result in chronic respiratory effects.
Long-term PM exposure and lung cancer
Of particular interest with regard to PM-related cause-specific mortality is growing
evidence linking long-term PM exposure with increased risk of lung cancer mortality. Historical
evidence 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 incidence and mortality have generally indicated positive associations with living in areas
having higher PM exposures despite possible problems with respect to potential exposure and
other risk factor measurement errors. Table 8-42 provides a partial listing of such studies beyond
those discussed below.
Prospective cohort studies offer a potentially more powerful approach to evaluation of
apparent associations between PM exposures and development of lung cancer. The 1996 PM
AQCD (U.S. Environmental Protection Agency, 1996a) summarized three of these more
elaborate studies that carefully evaluated PM air pollution exposure effects on lung cancer using
the prospective cohort design. In the AHSMOG Study, Abbey et al. (1991) followed a cohort of
Seventh Day Adventists, whose extremely low prevalence of smoking and uniform, relatively
healthy dietary patterns reduce the potential for confounding by these factors. Excess lung cancer
incidence was observed in females in relation to both particle (TSP) and O3 exposure after 6 years
follow-up time. Dockery et al. (1993) reported the results of a 14- to 16-year prospective follow-
up of 8,111 adults living in six U.S. cities that evaluated associations between air pollution and
mortality. After controlling for individual differences in age, sex, cigarette smoking, BMI,
education, and occupational exposure, Dockery et al. (1993) found an elevated but nonsignificant
risk for lung cancer mortality (RR = 1.37; CI: 0.81, 2.31) for a difference in PM25 pollution equal
to that of the most polluted versus the least polluted city. Pope et al. (1995) similarly analyzed
8-314
-------
TABLE 8-42. SUMMARY OF PAST ECOLOGIC AND CASE-CONTROL
EPIDEMIOLOGIC STUDIES OF OUTDOOR AIR AND LUNG CANCER
Study Type
Ecologic
Case-Control
Authors
Henderson etal.,
1975
Buffler et al.,
1988
Archer, 1990
Pike etal., 1979
Vena, 1982
Locale
Los Angeles, CA
Houston, TX
Utah
Los Angeles
Buffalo, NY
Exposure Classification
High PAH Areas
TSP by Census Tract
TSP by county
BAP Geo. Areas
TSP Geo. 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
Jedrychowski, Cracow, Poland TSP and SO2 Geo. Areas
etal., 1990
(CI: 1.0-2.9)
1.1 @ TSP > 150 ug/m3
(CI: N/A)
Katsouyanni, Athens, Greece Soot Concentration
etal., 1990 Geo. Areas
Barbone et al., Trieste, Italy High Particle Deposition
1995 Areas
Nyberg et al., Stockholm, High NO2 Areas
2000 Sweden
1.1 @ soot up to
400 ug/m3
(CI: N/A)
1.4 @ > 0.3 g/mVday
(CI: 1.1-1.8)
1.3
(CI: 0.9-1.9)
Source: Derived from Cohen (2000).
PM25 and sulfate (SO42 ) air pollution as predictors of mortality in a prospective study of 7-year
survival data (1982 to 1989) for about 550,000 adult volunteers obtained by the American Cancer
Society (ACS).
Both the ACS and Harvard studies have been subjected to much scrutiny, including an
extensive independent audit and reanalysis of the original data (Krewski et al., 2000) that
confirmed the originally published results. The ACS study controlled for individual differences
in age, sex, race, cigarette smoking, pipe and cigar smoking, exposure to passive cigarette smoke,
occupational exposure, education, BMI, and alcohol use. In the original ACS study, lung cancer
mortality was significantly associated with particulate air pollution when SO42+ was used as the
index, but not when PM2 5 mass was used as the index for a smaller subset of the study population
that resided in metropolitan areas where PM2 5 data were available from the Inhalable Particle (IP)
8-315
-------
Network. Thus, while these prospective cohort studies have also indicated that long-term PM
exposure is associated with an increased cancer risk, the effect estimates were generally not
statistically significant, quite possibly due to inadequate statistical power by these studies at that
time (e.g., due to inadequate population size and/or follow-up time for long-latency cancers).
A recent follow-up analysis of the major ACS study by Pope et al. (2002) responds to a
number of criticisms previously noted for the earlier ACS analysis (Pope et al., 1995) in the 1996
PM AQCD (U.S. Environmental Protection Agency, 1996a). Most notably, the new study
examined other pollutants, had better occupational indices and diet information, and also
addressed possible spatial autocorrelations due to regional location. The recent extension of the
ACS study included -500,000 adult men and women drawn from ACS-CPS-II enrollment and
follow-up during 1982 to 1998. This new analysis of the ACS cohort substantially expands the
prior analysis, including: (1) more than doubling of the follow-up time to 16 years (and more
than tripling of the number of deaths in the analysis); (2) substantially expanded exposure data,
including gaseous co-pollutant data and new PM25data collected in 1999 to 2001; (3) improved
control of occupational exposures; (4) incorporation of dietary variables that account for total fat
consumption, as well as that of vegetables, citrus and high-fiber grains; and (5) utilization of
recent advances in statistical modeling, including incorporation of random effects and
nonparametric spatial smoothing components in the Cox proportional hazards model.
In the extended ACS analysis, long-term exposure to air pollution, and especially to PM25,
was found to be associated with increased annual risk of mortality. With the longer 16-year
follow-up period and improved PM2 5 exposure metrics, this study detected for the first time, a
statistically significant association between living in a city with higher PM2 5 and increased risk of
dying of lung cancer. Each 10 ug/m3 increment in annual average fine PM was associated with a
13% (CI: 4-23%) increase in lung cancer mortality. Coarse particles and gaseous pollutants were
generally not significantly associated with excess lung cancer mortality. SO42 was significantly
associated with mortality and lung cancer deaths in this extended data set, yielding RR's
consistent with (i.e., not significantly different from) the SO42 RR's reported in the previously
published 7-year follow-up (Pope et al, 1995). However, while PM2 5 was specific to the causes
most biologically plausible to be influenced by air pollution in this analysis (i.e., cardiopulmonary
8-316
-------
and cancer), SO42 was significantly associated with every mortality category in this new analysis,
including that for "all-other causes". This suggests that the PM2 5 associations found are more
biologically plausible than the less specific SO42 associations found. The PM2 5 cancer mortality
risk appears greatest for nonsmokers and among those with lower socioeconomic status (as
indicated by lower educational attainment).
The AHSMOG investigators have re-examined the association between long-term PM
exposure and increased risk of both lung cancer incidence and lung cancer mortality in
nonsmokers using longer-term follow-up of this cohort and improved analytical approaches.
Beeson et al. (1998) considered this cohort of some 6,338 nonsmoking, non-Hispanic, white
Californian adults, ages 27 to 95 years, that was followed from 1977 to 1992 for newly diagnosed
cancers. Among the AHSMOG cohort, incident lung cancer in males was positively and
significantly associated with IQR increases for mean concentrations of PM10 (RR = 5.21; CI:
1.94, 13.99). For females in the cohort, incident lung cancer was positively and significantly
associated with increases in SO2 (RR = 2.14; CI: 1.36,3.37) and frequency of PM10 levels above
50 |ig/m3 (RR= 1.21; CI: 0.55, 2.66) and 60 ug/m3 (RR= 1.25; CI: 0.57, 2.71). Thus, increased
risks of incident lung cancer were deemed by the authors to be associated with elevated long-term
ambient concentrations of PM10 and SO2 in both genders. The higher PM10 effect estimate for
cancer in males appeared to be partially due to gender differences in long-term air pollution
exposures. Abbey et al. (1999) also related long-term ambient concentrations of PM10, SO42 ,
SO2, O3, and NO2 to 1977-1992 mortality in the AHSMOG cohort. After adjusting for a wide
array of potentially confounding factors, including occupational and indoor sources of air
pollutants, PM10 showed a strong association with lung cancer deaths in males (PM10 IQR
RR = 2.38; CI: 1.42, 3.97). In this cohort, males spent more time outdoors than females, thus
having higher estimated air pollution exposures than the cohort females. Ozone showed an even
stronger association with lung cancer mortality for males, and SO2 showed strong associations
with lung cancer mortality for both sexes. The authors reported that other pollutants showed
weak or no association with mortality. Therefore, increases in both lung cancer incidence and
lung cancer mortality in the extended follow-up analysis of the AHSMOG study were found to be
8-317
-------
most consistently associated with elevated long-term ambient concentrations of PM10, O3,
and SO2, especially among males.
Overall, these new cohort studies confirm and strengthen the published older ecological and
case-control evidence indicating that living in an area that has experienced higher PM exposures
can cause a significant increase in the RR of lung cancer incidence and associated mortality.
In particular, the new ACS cohort analysis more clearly indicates that living in a city with
higher PM25 levels is associated with an elevated risk of lung cancer amounting to an increase
of some 10 to 15% above the lung cancer mortality risk in a cleaner city.
With regard to specific ambient fine particle constituents that may significantly contribute
to the observed ambient PM-related increases in lung cancer incidence and mortality, PM
components of gasoline and diesel engine exhaust represent one class of hypothesized likely
important contributors. Such mobile source PM typically comprises a noticeable fraction of
ambient fine particles in many urban areas, having been estimated to comprise from ~5 to 30% of
ambient PM25 in some U.S. urban areas (see Chapter 3). These mobile sources are reasonable
candidates as contributors to ambient PM-lung cancer risks, given their being sources of known
cancer-causing agents (e.g., PAHs), as are other coal-combustion and/or woodburning emission
sources (at least during some seasons).
8.4.7 Concentration-Response Relationships for Ambient PM
In the 1996 PM AQCD, the limitations of identifying possible "thresholds" in the
concentration-response relationships in observational studies were discussed, including
difficulties related to the low data density in the lower PM concentration range, the small number
of quantile indicators often used, and the possible influence of measurement error. Also, a
threshold for a population, as opposed to a threshold for an individual, has some conceptual issues
that should be noted. For example, since individual thresholds vary from person to person due to
individual differences in genetic level susceptibility and preexisting disease conditions (and even
can vary from one time to another for a given person), it is extremely difficult mathematically to
demonstrate convincingly that a clear threshold exists in the population studies. This is especially
true if the most sensitive members of a population are unusually sensitive even down to very low
8-318
-------
concentrations. The person-to-person difference in the relationship between personal exposure
and the concentration observed at a monitor may also add to the variability in observed exposure-
response relationships, possibly obscuring otherwise more evident thresholds. Since one cannot
directly measure but can only compute or estimate a population threshold, it would be difficult to
interpret an observed population threshold biologically, without pertinent collateral
dosimetric/toxicologic information. Despite these issues, several PM-related epidemiologic
studies have attempted to address the question of threshold.
Cakmak et al. (1999) investigated methods to detect and estimate threshold levels in time-
series studies. Based on the realistic range of error observed from actual Toronto pollution data
(average site-to-site correlation: 0.90 for O3; 0.76 for CoH; 0.69 for TSP; 0.59 for SO2; 0.58
for NO2; and 0.44 for CO), pollution levels were generated with multiplicative error for six levels
of exposure error (1.0, 0.9, 0.8, 0.72, 0.6, 0.4, site-to-site correlation). Mortality series were
generated with three PM10 threshold levels (12.8 |ig/m3, 24.6 |ig/m3, and 34.4 |ig/m3). LOESS
with a 60% span was used to observe the exposure-response curves for these 18 combinations of
exposure-response relationships with error. A parameter threshold model was also fit using
nonlinear least squares. Both mortality and PM10 data were prefiltered for the influence of
seasonal cycles using LOESS smooth function. The threshold regression models were then fit to
the prefiltered data. Graphical presentations indicate that LOESS adequately detects threshold
under no error, but the thresholds were "smoothed out" under the extreme error scenario. Use of
a parametric threshold model was adequate to give "nearly unbiased" estimates of threshold
concentrations even under the conditions of extreme measurement error, but the uncertainty in the
threshold estimates increased with the degree of error. They concluded, "if threshold exists, it is
highly likely that standard statistical analysis can detect it."
Daniels et al. (2000; reanalysis by Dominici et al., 2003a) tested for presence of a threshold
using data for the largest 20 U.S. cities during 1987 to 1994. In their original analyses, the
authors compared three log-linear GAM regression models: (1) using a linear PM10 term;
(2) using a natural cubic spline of PM10 with knots at 30 and 60 |ig/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 |ig/m3 with 5 |ig/m3 increment. The covariates
8-319
-------
included in these models are similar to those previously used by the same research group (Kelsall
et al., 1997; Samet et al., 2000a,b), including the smoothing function of time, temperature and
dewpoint, and day-of-week indicators. The 2003 reanalysis evaluated total, cardiorespiratory,
and "other" mortality series by means of covariate adjustments made using natural splines in
GLM models. These models were fit for each city separately and, for model (1) and (2), the
combined estimates across cities were then 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 fits among the models, within each city and over all cities, were also
determined using Akaike's Information Criterion (AIC).
As seen in Figure 8-30, the results using the natural spline model showed that, for total and
cardiorespiratory mortality, the exposure-response spline curves for mean lag were roughly
linear, but less so for current and previous day PM10, making it difficult to discern any evident
threshold. However, the curves for mortality from other causes, most clearly increased once PM10
concentrations exceeded 50 |ig/m3. The posterior probabilities for a threshold for PM10 effects on
total and cause-specific mortality groupings are shown in Figure 8-31 (CVDRESP =
cardiorespiratory causes). There appears to be a reasonably likely possibility of a threshold
existing for daily total or CVDRESP mortality at PM10 levels of-15-20 |ig/m3 or below; but the
likelihood of a threshold occurring above -25 |ig/m3 seems to be essentially zero, based on the
latter analyses. The hypothesis of linearity was examined more formally by comparing AIC
values across models, with the results indicating that the linear model was preferred over the
spline and the threshold models. Thus, these findings do not rule out the possibility that linear
models without a threshold may be appropriate for estimating the effects of PM10 on the types of
mortality of main interest. The available information simply does not allow for a clear choice of
"threshold" or "no threshold" over the other.
Smith et al. (1999) analyzed the slope of the PM10-mortality relationship in Birmingham,
AL and in Cook County, IL. Temperature was modeled using piece-wise linear term with a
change point. PM10 data were modeled at lag 0 through 3 and 3-day averages at these lags.
In addition to the linear model, the existence of a threshold was also investigated by using
B-splines and a parametric threshold model with the profile log likelihood evaluated at changing
8-320
-------
0.08
0.06
0.04
o
0)
O>
c
re
O 0.02
•5
I
ao
Total
CVDRESP
Other
0 10 20 30 40 50 60 70
PM10(|jg/m3)
0 10 20 30 40 50 60 70
PM10 (Mg/m3)
0 10 20 30 40 50 60 70
PM10 (Mg/m3)
Figure 8-30. Concentration-response curves for PM10 mortality relationships in 20 largest
U.S. cities (1987-1994), for total (Total) mortality, cardiovascular and
respiratory (CVDRESP) mortality, and other-causes (Other) mortality. 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. (2003a).
Total
L
I
XXXXXXN
KXXXXXXXXXXX^
I
CVDRESP
P/I
Other
xX
T- T- r^i CM
(Mg/m3)
(Mg/m3)
PM10 (|jg/m3)
Figure 8-31. Posterior probabilities of thresholds for each cause-specific mortality and for
mean PM10, 20 largest U.S. cities, 1987-1994. Total = total nonaccidental
mortality; CVDRESP = cardiovascular mortality and respiratory mortality;
Other = mortality from other causes.
Source: Dominici et al. (2003a).
8-321
-------
threshold points. B-splines results suggest that an increasing effect above 80 |ig/m3 for
Birmingham, and above 100 |ig/m3 for Chicago. The threshold model through examination of log
likelihood across the range of threshold levels also suggested similar change points, but not to an
extent that statistically significant distinctions were demonstrated.
The Smith et al. (2000) study of associations between daily total mortality and PM25
and PM10_2 5 in Phoenix, AZ (during 1995-1997) also investigated the possibility of a threshold.
In the linear model, the authors found that mortality was significantly associated with PM10_2 5,
but not with PM25. In modeling possible thresholds, they applied: (1) a piecewise linear model
in which several possible thresholds were specified; and (2) a B-spline (spline with cubic
polynomials) model with 4 knots. Using the piecewise model, there was no indication that there
was a threshold for PM10_2 5. However, for PM2 5, the piecewise model resulted in suggestive
evidence for a threshold, around 20 to 25 |ig/m3. The B-spline results also showed no evidence of
threshold for PM10_2 5, but for PM2 5, a nonlinear curve showed a change in the slope around
20 |ig/m3. A further Bayesian analysis for threshold selection suggested a clear peak in the
posterior density of PM2 5 effects around 22 |ig/m3. These results make it difficult to evaluate the
relative roles of different PM components (in this case, PM25 versus PM10_25). However, the
concentration-response curve for PM25 presented in this publication suggests more of a U- or
V-shaped relationship than the usual "hockey stick" relationship. Such a relationship is, unlike
the temperature-mortality relationship, difficult to interpret biologically. Because the sample size
of this data (3 years) is relatively small, further investigation of this issue using similar methods
but a larger data set is warranted. Other studies evaluate nonlinear relationships using a multicity
meta-smoothing approach based on non- or semi-parametric smoothers rather than on linear
parametric models.
In summary, the results from large multicity studies suggest that there is no strong evidence
of a clear threshold for PM mortality effect; nor is there clear evidence against possible thresholds
for PM-related effects. Some single-city studies do provide some suggestive hints for possible
thresholds, but not in a statistically clear manner. More data need to be examined with alternative
approaches in order to better resolve the issue.
8-322
-------
8.4.8 The Question of Heterogeneity of Particulate Matter Effects Estimates
Approximately 35 then-available acute PM exposure community epidemiologic studies
were assessed in the 1996 PM AQCD as collectively demonstrating increased risks of mortality
being associated with short-term (24-h) PM exposures indexed by various ambient PM
measurement indices (e.g., PM10, PM25, BS, CoH, sulfates, etc.) in many different cities in the
United States and internationally. Much homogeneity appeared to exist across various
geographic locations, with many studies suggesting, for example, increased relative risk (RR)
estimates for total nonaccidental mortality mainly on the order of 1.025 to 1.05 (or 2.5 to 5.0%
excess deaths) per 50 |ig/m3 increase in 24-h PM10, with statistically significant results extending
more broadly in the range of 1.5 to 8.0%. The elderly > 65 years old and those with preexisting
cardiopulmonary conditions had somewhat higher excess risks. One study, the Harvard Six City
Study, also provided estimates of increased RR for total mortality falling in the range of 1.02 to
1.056 (2.0 to 5.6% excess deaths) per 25 |ig/m3 24-h PM25 increment.
8.4.8.1 Evaluation of Heterogeneity in Time-Series Studies
More than 80 new time-series PM-mortality studies assessed earlier in this chapter provide
extensive additional evidence which, qualitatively, largely substantiates significant ambient PM-
mortality relationships, based on 24-h exposures indexed by a wide variety of PM metrics in
many different cities of the United States, in Canada, in Mexico, and elsewhere (in South
America, Europe, Asia, etc.). The newly available effect size estimates from such studies are
reasonably consistent with the ranges derived from the earlier studies reviewed in the 1996 PM
AQCD. For example, newly estimated PM10 effects generally fall in the range of 1.0 to 8.0%
excess deaths per 50 |ig/m3 PM10 increment in 24-h concentration; and new PM2 5 excess death
estimates for short-term exposures generally fall in the range of 2 to 8% per 25 |ig/m3 increment
in 24-h PM2 5 concentration.
However, in contrast to the past appearance of considerable homogeneity among risk
estimates, somewhat greater spatial heterogeneity appears to exist across newly reported study
results, both with regard to PM-mortality and morbidity effects associations. The newly apparent
heterogeneity of findings across locations is perhaps most notable in relation to reports based on
O OO
-323
-------
multiple-city studies in which investigators used the same analytical strategies and models
adjusted for the same or similar co-pollutants and meteorological conditions, raising the
possibility of different findings reflecting real location-specific differences in exposure-response
relationships rather than potential differences in models used, pollutants included in the models,
etc. Some examples of newly reported and well-conducted multiple-city studies include: the
NMMAPS analyses of mortality and morbidity in 90 and 20 U.S. cities (Samet et al., 2000a,b;
Dominici et al., 2000a); the Schwartz (2000b,c) analyses of 10 U.S. cities; the study of eight
largest Canadian cities (Burnett et al., 2000); the study of hospital admissions in eight U.S.
counties (Schwartz, 1999); and the APHEA studies of mortality and morbidity in several
European cities (Katsouyanni et al., 1997; Zmirou et al., 1998).
The large NMMAPS studies of mortality and morbidity in U.S. cities provide important
information about potential U.S. within- and between-region heterogeneity. HEI (2003a), after
examining the NMMAPS GAM reanalyses by Dominici et al. (2002), concluded that while
formal tests of PM effects across cities did not indicate evidence of heterogeneity because of the
individual-city effects standard error being generally large, the power to assess the presence of
heterogeneity was low and, as such, the possibility of heterogeneity still exists.
Some insight into the possible extent of heterogeneity can be gained by close examination
of data from the NMMAPS study (Samet et al., 2000b). Data for excess risk and 95% confidence
intervals were plotted by EPA against the total number of effective observations, measured by the
number of days of PM10 data times the mean number of daily deaths in the community. This
provides a useful measure of the weight that might be assigned to the results, since the
uncertainty of the RR estimate based on a Poisson mean is roughly inversely proportional to this
product. That is, the expected pattern should typically show less spread of estimated excess risk
with increasing death-days of data. A more refined weight index would also include the spread in
the distribution of PM concentrations. The results for NMMAPS, including the GAM reanalyses
results, conform to some extent to the expected pattern. That is, with increasingly more
mortality-days observations, the 95% confidence intervals generally became narrower. However,
the results for relationships between effect size estimates and precision estimates for different
regions vary considerably. In the Northeast, for example, there is some degree of consistency of
8-324
-------
effect size for larger study-size cities, even with moderately wide confidence intervals for those
with log mortality-days > 8 to 9, and all clearly exceed the overall nationwide grand mean. On
the other hand, the smaller study-size Northeast cities (with much wider confidence intervals at
log mortality-days < 8) show much greater variability of effect size estimates and less precision.
As for the estimates derived for cities in other U.S. regions, there is even less consistency
between magnitude of effect size and precision of the estimates, suggesting that other factors may
account for differences in direction and/or size of the risk estimates.
Burnett and Goldberg (2003) also investigated heterogeneity in effects across eight
Canadian cities, and concluded that there was not sufficient evidence to conclude that the PM
association with mortality varies across the eight cities. In the initial analyses using GAM, a
positive estimate of heterogeneity was reported, but in reanalyses using GLM with natural
splines, negative estimates were reported. The authors stated that this reflected reduced variation
in effect size estimates across cities along with increased within-city estimation error in the
reanalysis results. In addition, as discussed in Section 8.2.2.3.3, in the initial analyses using data
from the APHEA cohort, some apparent heterogeneity was found between results for western and
eastern European cities; however, in reanalyses of these results, the distinctions between the
western and eastern cities were less clear. Variables that may potentially influence heterogeneity
of effects were further investigated in the APHEA2 analyses for 29 European cities, with
mean NO2 concentration in the cities (indicator of traffic-related pollution), warmer climate and
low overall mortality rate being associated with increased PM-mortality associations.
Further closer reexamination of results for different areas in the U.S. or elsewhere may
reveal interesting new insights into what factors may account for apparent differences among the
cities within a given region or across regions. Some potential factors include differences in PM
sources or composition, differences in population exposures across cities, and differences in
potentially susceptible groups (e.g., % of population > 65 yr old). The NMMAPS investigators
reported no substantial differences in PM10-mortality associations based on PM2 5 / PM10 ratios or
socioeconomic indicators for the various cities; however, no statistically significant evidence of
heterogeneity was reported for that study. As stated previously, European investigators discussed
several factors that may have influenced heterogeneity in PM-mortality associations across
29 European cities. These included variations in presence of an indicator of traffic-related
8-325
-------
pollution, wanner climate (postulated to be related to better estimation of exposures since people
were more likely to open windows) and low mortality rate (which the authors suggested was due
to the larger number of potentially susceptible people in cities with lower mortality rates). These
findings are consistent with those reported by Janssen et al. (2002, reanalyzed in Zanobetti and
Schwartz (2003a), where PM10-hospitalization associations were greater in areas with less use of
central air conditioning and with greater contributions of PM10 emissions from vehicle emissions
and oil combustion.
8.4.8.2 Comparison of Spatial Relationships in the NMMAPS and Cohort
Reanalyses Studies
Both the NMMAPS and HEI Cohort Reanalyses studies had a sufficiently large number of
U.S. cities to allow considerable resolution of regional PM effects within the "lower 48" states,
but an attempt was made to take this approach to a much more detailed level in the Cohort
Reanalysis studies than in NMMAPS. There were: 88 cities with PM10 effect size estimates in
NMMAPS; 50 cities with PM25 and 151 cities with sulfates in the original Pope et al. (1995) ACS
analyses and in the HEI reanalyses using the original data; and 63 cities with PM2 5 data and
144 cities with sulfate data in the additional analyses done by the HEI Cohort Reanalysis team.
The relatively large number of data points utilized in the HEI reanalyses effort and additional
analyses allowed estimation of surfaces for elevated long-term concentrations of PM25, sulfates,
and SO2 with resolution on a scale of a few tens to hundreds of kilometers.
The patterns for PM2 5 and sulfates are similar, but not identical. In particular, the
modeled PM25 surface (Krewski et al., 2000; Figure 18) had peak levels around Chicago-Gary,
in the eastern Kentucky-Cleveland region, and around Birmingham AL, with elevated but
lower PM25 almost everywhere east of the Mississippi, as well as southern California. This is
similar to the modeled sulfate surface (Krewski et al., 2000; Figure 16), with the absence of a
peak in Birmingham and an emerging sulfate peak in Atlanta. The only area with markedly
elevated SO2 concentrations was the Cleveland - Pittsburgh region. Secondary sulfates in
particles derived from local SO2 appeared more likely to be important in the industrial midwest,
south from the Chicago - Gary region into Ohio, northeastern Kentucky, West Virginia, and
southwest Pennsylvania, possibly related to combustion of high-sulfur fuels.
8-326
-------
The overlay of mortality with air pollution patterns is also of much interest. The spatial
overlay of long-term PM25 and mortality (Krewski et al., 2000; Figure 21) was highest from
southern Ohio to northeastern Kentucky /West Virginia, but also included a significant association
over most of the industrial midwest. This was reflected, in diminished form, by the sulfates
and SO2 maps (Krewski et al., 2000; Figures 19 and 20), where there appeared to be a somewhat
tighter focus of elevated risk in the upper Ohio River Valley area. This suggests that, while SO2
was an important precursor of sulfates in this region, there may also have been some other
(non-sulfur) contributors to associations between PM2 5 and long-term mortality, encompassing
a wide area of the North Central Midwest and noncoastal Mid-Atlantic region.
The apparent differences in PM10 and/or PM2 5 effect sizes across different regions should
not be attributed merely to possible variations in measurement error or other statistical artifact(s).
Some of these differences may reflect: real regional differences in particle composition or
co-pollutant mix; differences in relative human exposures to ambient particles or other gaseous
pollutants; sociodemographic differences (e.g., percent of infants or elderly in regional
population); or other important, as of yet unidentified PM effect modifiers.
8.4.9 Age-Related Differences in PM Effect Estimates
Numerous epidemiological studies have reported health responses to PM and other
pollutants for one or another specific age group. For example, in the U.S., data on hospital
admissions for older people (aged 65 years and older) are available through a national data system
maintained by the Health Care Financing Administration; and, thus, many U.S. hospital
admissions studies have focused on health responses in this age group. Other studies, such as
panel studies for asthma symptoms, have evaluated groups of schoolchildren. In general, such
studies have indicated that both the elderly and children are likely susceptible subpopulations for
PM-related effects (see Sections 8.3.1.4 and 8.3.2.5).
Though less commonly done, possible age-related differences in ambient PM health effects
have been evaluated in certain recently published epidemiological studies that assessed health
responses to air pollution by means of stratified analyses for different age groups within the
population studied. For example, a number of studies have assessed relationships between PM
8-327
-------
and total mortality across all ages, then evaluated possible differences in risk for the subset of
older adults (50+ or 65+ years); and some of these have reported slightly larger effect estimates
for the older age group (e.g., Schwartz et al., 1996a; Styer et al., 1995; Borja-Aburto et al., 1998),
whereas others have found associations that are similar in magnitude or even slightly smaller for
the older age group (e.g., Ostro et al., 1999a, 1995; Castillejos et al., 2000). Also, Chock et al.
(2000) reported associations between PM and total mortality that were not substantially different
for age groups of 0 to 74 and 75+ years.
In other studies of hospital admissions or medical visits for asthma or respiratory disease,
some studies have reported larger effect estimates for children than for adults (e.g., Anderson
et al., 1998; Medina et al., 1997), whereas others have reported effect estimates of generally
similar size across young and adult age groups (e.g., Atkinson et al., 1999b; Hajat et al., 1999;
Wong et al., 1999a) and some studies of respiratory hospital admissions have shown larger effect
sizes for adults (e.g., Prescott et al., 1998). For hospital admissions or medical visits for
cardiovascular diseases, most studies (but not all — e.g., Atkinson et al., 1999a), have reported
somewhat larger effect estimate sizes for older adults (65+ years) than adults in younger age
categories (e.g., Le Tertre et al., 2003; Wong et al., 1999a; Prescott et al., 1998; Morgan et al.,
1998).
The above rather small group of studies does not show striking differences in effect
estimates from analyses across age group strata, but they do tend to support previous findings
that, depending on the specific type of effect under study, older adults and children may be more
susceptible to certain PM- related effects. More specifically, older adults (aged 65+ years) appear
to be most clearly at somewhat higher risk for PM exacerbation of cardiovascular-related disease
effects and , perhaps, tend to experience higher PM-related total (nonaccidental) mortality risk, as
well. On the other hand, more limited evidence points toward children possibly being at
somewhat higher risk for respiratory-related (especially asthma) PM effects than adults.
8.4.10 Implications of Airborne Particle Mortality Effects
The public health burden of mortality associated with exposure to ambient PM depends not
only on the increased risk of death, but also on the amount of life shortening that is attributable to
8-328
-------
those deaths. The 1996 PM AQCD concluded that confident quantitive determination of years of
life lost to ambient PM exposure was not yet possible and life shortening may range from days to
years (U.S. Environmental Protection Agency, 1996a). Now, some newly available analyses
provide further interesting insights with regard to potential life-shortening associated with
ambient PM exposures.
8.4.10.1 Short-Term Exposure and Mortality Displacement
A few studies have investigated the question of "harvesting," a phenomenon in which a
deficit in mortality occurs following days with (pollution-caused) elevated mortality, due to
depletion of the susceptible population pool. This issue is very important in interpreting the
public health implication of the reported short-term PM mortality effects. The 1996 PM AQCD
discussed suggestive evidence observed by Spix et al. (1993) during a period when air pollution
levels were relatively high. Recent studies, however, generally used data from areas with lower,
non-episodic pollution levels.
Schwartz (2000c; reanalysis 2003a) separated time-series air pollution, weather, and
mortality data from Boston, MA, into three components: (1) seasonal and longer fluctuations;
(2) "intermediate" fluctuations; (3) "short-term" fluctuations. By varying the cut-off between the
intermediate and short term, evidence of harvesting was sought. The idea is, for example, if the
extent of harvesting were a matter of a few days, associations between weekly average values of
mortality and air pollution (controlling for seasonal cycles) would not be seen. Schwartz's
reanalysis using natural splines reported reductions in COPD mortality PM2 5 risk estimates for
longer time scale, suggesting that most of the COPD mortality was only displaced by a few
weeks. However, for pneumonia, ischemic heart disease, and all cause mortality, the effect size
increased, as longer time scales were included. For example, the percent increase in
nonaccidental deaths associated with a 25 |ig/m3 increase in PM25 increased from 5.8% (CI: 4.5,
7.3) for thel5-day window to 9.7% (CI: 8.2, 11.2) for the 60-day window. Note, however, that
the 60-day time scale window is in the range of influenza epidemics. Some caution is therefore
needed in interpreting risk estimates in this range.
Zanobetti et al. (2000b) used what they termed "generalized additive distributed lag
models" (penalized splines using an algorithm that did not require back-fitting were used for all
8-329
-------
the smoothing terms) to help quantify mortality displacement in Milan, Italy, 1980 to 1989.
Nonaccidental total deaths were regressed on smooth functions 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. 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 45 th day (maximum examined). The sum of these coefficients was
over three times larger than that for the single-day estimate.
Zanobetti et al. (2000a; reanalysis by Zanobetti and Schwartz, 2003b) also applied the same
concept described above (up to 41 lag days) to 10 cities from APHEA2 to estimate distributed
lag PM10 mortality risks. They applied the covariate adjustment in a GAM model used in
APHEA2 (Katsouyanni et al., 2001) and in reanalysis (Zanobetti and Schwartz, 2003b), they also
used penalized splines in addition to the GAM model with stringent convergence criteria. The
resulting city specific coefficients were pooled in the second-stage model, taking into account
heterogeneity across cities. The estimated shape of the distributed lag pooled across 10 cities
showed a similar pattern to that from Milan data described above, with the second "hump" of
smaller but positive coefficients between approximately 20 to 35 days. The results indicated that,
compared to PM10 risk estimates obtained for the average of lag 0 and 1 days, the distributed lag
estimates up to 40 days were about twice larger in both GAM and penalized splines models. For
example, the combined distributed lag estimates for the 10 cities using penalized splines was
5.6% (CI: 1.5, 9.8), as compared to 2.9% (CI: 1.4, 4.4). It should be noted, however, that the
results for individual cities varied. For example, the estimates for average of lag 0 and 1 days and
the distributed lag model were comparable in Tel Aviv, whereas it was nearly seven times bigger
for distributed lag model in Lodz. Thus, while these results do support the lack of mortality
displacement up to 40-45 day period, the pattern of lagged associations may vary from city
to city.
Two new studies conducted very different analyses, beginning with the assumption that
harvesting is occurring. Both research groups used models to estimate the size of the frail
8-330
-------
populations. In one study, as part of their analysis of PM10-mortality association in Birmingham,
AL and Cook County, IL, Smith et al. (1999) used a latent variable structure fitted through
Bayesian techniques using Monte Carlo sampling. The resulting posterior mean for the size of
the frail population in Chicago was 765 (posterior s.d. = 189). The mean numbers of days lost per
person as a result of 10 |ig/m3 increase in PM10 was estimated to be 0.079 day (posterior
s.d. = 0.032). In the other study, Murray and Nelson (2000) used Kalman filtering to estimate a
hazard function of TSP in a state space model in the Philadelphia mortality data during 1973 to
1990. The model framework, which assumes an 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.
Combinations of TSP, linear temperature, squared temperature, and interaction of TSP and
temperature were considered in six models. The size of at-risk (or frail) population estimated was
about 500 people. Life expectancy was estimated to be reduced by about 2.5 days with TSP
exposure for the roughly 500 at-risk frail individuals in Philadelphia suggesting that the hazard
causing agent makes a difference of only 2.5 days in the at-risk frail population. In both cases,
the estimated size of the frail population is very small with short life expectancy. In these cases,
based on the assumption that harvesting is occurring and only small frail populations are at risk,
life shortening due to PM exposures is estimated to be on the order of just a few days.
Zeger et al. (1999) first illustrated, through simulation, the implication of harvesting for PM
regression coefficients (i.e., mortality relative risk) as observed in a frequency domain. Three
levels of harvesting (3 days, 30 days, and 300 days) were simulated. As expected, the shorter the
harvesting, the larger the PM coefficient in the higher frequency range. However, in the analysis
(and reanalysis by Dominici et al., 2003a) of real data from Philadelphia, regression coefficients
increased toward the lower frequency range, suggesting that the extent of harvesting, if it exists,
is not in the short-term range. Zeger suggested that "harvesting-resistant" regression coefficients
could be obtained by excluding coefficients in the very high frequency range (to eliminate short-
term harvesting) and in the very low frequency range (to eliminate seasonal confounding). Since
the observed frequency domain coefficients in the very high frequency range were smaller than
those in the mid frequency range, eliminating the "short-term harvesting" effects would only
increase the average of those coefficients in the rest of the frequency range.
8-331
-------
Frequency domain analyses are rarely performed in air pollution health effects studies,
except perhaps for spectral analysis (variance decomposition by frequency) to identify seasonal
cycles. Examinations of the correlation by frequency (coherence) and the regression coefficients
by frequency (gain) may be useful in evaluating the potential frequency-dependent relationships
among multiple time series. A few past examples in air pollution health effects studies include:
(1) Shumway et al.'s (1983) analysis of London mortality analysis, in which they observed that
significant coherence occurred beyond two week periodicity (they interpreted this as "pollution
has to persist to affect mortality"); (2) Shumway et al.'s (1988) analysis of Los Angeles mortality
data, in which they also found larger coherence in the lower frequency; (3) Ito's (1990) analysis
of London mortality data in which he observed relatively constant gain (regression coefficient)
for pollutants across the frequency range, except the annual cycle. These results also suggest that
associations and effect size, at least, are not concentrated in the very high frequency range.
Dominici et al. (2003c) also explored associations between air pollution and mortality using
data from the four cities included in NMMAPS that had every-day PM10 measurements, Chicago,
Minneapolis, Pittsburgh, and Seattle. The authors first used discrete Fourier transformation to
decompose the air pollution time series into distinct component series: < 3.5, 3.5-6, 7-13, 14-29,
30-59, and > 60 days. They then calculated associations without decomposition and with each of
the timescale components, with the expectation that under a short-term mortality displacement
scenario, mortality would be mainly associated with air pollution at the short timescales. For
both individual cities and the four cities overall, a pattern was found of larger effects at the longer
timescales and smaller effects at the shorter timescales. For a 1-day lag, the 4-city overall relative
risk of 0.22% increase in cardiovascular and respiratory mortality per 10 |ig/m3 PM10 (CI: -0.02,
0.46) was comparable to that found in the 90-city analyses (Dominici et al., 2003b). At the > 60
day timescale, the authors report relative risks of 1.35% (CI: 0.52, 2.17) per 10 |ig/m3 PM10 for
total mortality, and 1.87% (CI: 0.75, 2.99) per 10 |ig/m3 PM10 for cardiovascular and respiratory
mortality. The authors also investigated the sensitivity of their findings to alternative lag period
choice (optimal lags from 0-6 days selected), adjustment for long-term trends and seasonality
(ranging from 3.5 to 14 degrees of freedom per year for time), and alternative assumptions of
heterogeneity in effects between cities. For all, the authors report that the overall pattern of
results remains similar, though the confidence intervals widened considerably with greater
8-332
-------
heterogeneity. The authors conclude that the results are inconsistent with the short-term mortality
displacement hypothesis.
In a commentary on the previous analyses, Smith (2003) conducted further analyses using
the software developed by Dominici et al. (2003b) and TSP data from Philadelphia. Results are
presented for models including alternative timescales for meteorological factors in the analyses,
with a consistent pattern of results showing larger effect estimates with longer timescales. Smith
(2003) also used the frequency decomposition software with a simulated data set that had an
assumed association between TSP and mortality (1% increase per 10 |ig/m3 TSP). From these
results, the author concluded that it was more difficult to determine time-scale dependency in
response, particularly for the results of the simulated harvesting model, where the effect estimate
appears to be consistent in size across all time scales but the longest (> 60 days). In addition, the
author conducted a simpler analysis using multi-day averaged TSP concentrations, up to a 30-day
average, with the results of different models indicating a peak in the time-dependent TSP effect at
-15 days but with different patterns for longer time scales. The author concluded that
interpretation of results from these models remained as difficult as ever. In response, Dominici
et al. (2003b) agreed that careful interpretation of air pollution-mortality models and
consideration of assumptions is needed. The authors discuss further their results of analyses
using data from the four NMMAPS cities, observing that their "harvesting-resistent" effect
estimates are larger than the "harvesting-prone" estimates, and that these results are consistent
with air pollution effects on all people, not simply the very frail.
Schwartz (2000c), Zanobetti et al. (2000b), Zanobetti et al., (2000a); reanalysis by Zanobetti
and Schwartz, (2003b) and Zeger et al.'s analysis (1999); reanalysis by Dominici et al. (2003a,
2003b) all suggest that the extent of harvesting, if any, is not a matter of only a few days. Other
past studies that used frequency domain analyses are also at least qualitatively in agreement with
the evidence against the short-term only harvesting. Since long wave cycles (> 6 months) need to
be controlled in time-series analyses to avoid seasonal confounding, the extent of harvesting
beyond 6 months periodicity is not possible in time-series study design. Also, influenza
epidemics can possibly confound the PM-mortality associations in the 1 to 3 month periodicity
ranges. Therefore, interpreting PM risk estimates in these "intermediate" time scale also requires
caution.
8O •"> O
-333
-------
In contrast to this group of studies, Smith et al. (1999) and Murray and Nelson (2000)
suggest that the frail population is very small and its life expectancy short, such that PM or any
external stress cannot have more than a few days of life-shortening impacts on this specific
subpopulation. This may, in part, reflect the limitation of the model itself when applied only to a
small frail subpopulation. Thus, there appears to be consistency in results within the similar
models but not across different types of models. Clearly, more research is needed in this area
both in terms of development of a conceptual framework that can be tested with real data, and
applications of these models to more data sets. However, at least in the models that extend the
common time-series modeling, there appears to be no strong evidence to suggest that PM is
shortening life by only a few days.
8.4.10.2 Life-Shortening Estimates Based on Prospective Cohort Study Results
Brunekreef (1997) reviewed available evidence for long-term PM exposure effects on
mortality and, using life table methods, derived a rough preliminary estimate of the reduction in
life expectancy implied by those effect estimates. Based on the results of Dockery et al. (1993)
and Pope et al. (1995), a relative risk of 1.1 per 10 |ig/m3 exposure over 15 years was assumed for
the effect of PM air pollution on men 25 to 75 years old. A 1992 life table for men in the
Netherlands was developed for 10 successive five-year categories that make up the 25 to 75 year
old age range. Life expectancy of a 25 year old was then calculated for this base case and
compared with the calculated life expectancy for the PM-exposed case, in which the death rates
were increased in each age group by a factor of 1.1. A difference of 1.11 years was estimated
between the "exposed" and "clean air" cohorts' overall life expectancy at age 25. Looked at
another way, this implies that the expected lifespan for persons who actually died from air
pollution would be reduced by more than 10 years, because they represent a small percentage of
the entire cohort population. A similar calculation by present EPA authors, based on the 1969-71
life table for U.S. white males, yielded a larger estimated reduction of 1.31 years for the entire
U.S. population's life expectancy at age 25. Thus, these calculations imply that relatively small
differences in long-term exposure to ambient PM may have substantial effects on life expectancy.
However, these "back of the envelope" calculations have not been verified by others and can only
be viewed as providing very rough "ballpark" estimates of potential life-shortening effects of PM.
8-334
-------
They depend heavily on the specific PM risk estimates used and, for example, would likely have
to be adjusted downward to reflect the newer (presumably more credible) lower RR estimates
derived from the Pope et al. (2002) ACS extension study.
8.4.10.3 Potential Effects of Infant Mortality on Life-Shortening Estimates
Deaths among children would logically have the greatest influence on a population's overall
life expectancy, but the Brunekreef (1997) life table calculations did not consider any possible
long-term air pollution exposure effects on the population aged < 25 years. Thus, any premature
mortality that may occur among children due to PM exposure would logically be likely to
increase significantly any overall population life shortening over and above that estimated by
Brunekreef (1997) for long-term PM exposure of adults aged > 25 years. However, as discussed
earlier, only a few older cross-sectional studies and a few more recent studies provide very
limited evidence bearing on the extent to which infants may be among subpopulations affected by
long-term PM exposure. Thus, much more definitive future research is needed before infant
mortality can be considered in generating estimates of potential PM-related life shortening in the
U.S. population.
8.5 SUMMARY OF KEY FINDINGS AND CONCLUSIONS DERIVED
FROM PARTICULATE MATTER EPIDEMIOLOGY STUDIES
Important types of additions to the epidemiologic database beyond that assessed in the 1996
PM AQCD, as evaluated above in this chapter, include:
• Several new multicity studies of mortality and morbidity effects which provide more
precise estimates of PM effect sizes than most smaller-scale individual city studies;
• A large number of new studies of various health endpoints using mass-based indicators
of thoracic particles (e.g., PM10); fine-fraction particles (e.g., PM25 and/or components
such as sulfates, nitrates, H+, and ultrafme particles [PMl 0 and smaller]); and, to a
lesser extent, coarse-fraction particles (e.g., PM10_25 and components such as crustal
particles).
8-335
-------
• Many new studies that reflect consideration of ambient PM as a component of complex
air pollution mixtures and which evaluate the sensitivity of estimated PM effects to the
inclusion of gaseous co-pollutants (e.g., O3, CO, NO2, SO2) and/or various different PM
indicators / components in analytical models;
• New and reanalyzed studies that provide insight into the sensitivity of PM effects to the
use of alternative statistical models and model specifications for addressing weather
and other temporal variables;
• New studies providing insight into various key issues such as alternative lag structures
(e.g., single-day and distributed lags), concentration-response relationships, spatial
heterogeneity of PM effects, measurement error effects (e.g., differential error across
various PM components and/or gaseous co-pollutants);
• Initial studies using new approaches to evaluate the effects of combinations of air
pollutant or mixtures including PM components, based on empirical combinations
(e.g., factor analysis or source profiles);
• New evidence from "found experiments," or so-called "intervention studies" that
evaluate associations between reduced air pollution levels and improvements in health
endpoints;
• Numerous new studies of cardiovascular endpoints, with particular emphasis on
assessment of cardiovascular risk factors as well as symptoms;
• Additional new studies on asthma and other respiratory conditions potentially
exacerbated by ambient PM exposure;
• New and extended studies of long-term PM exposure effects, notably including
analyses of lung cancer associations with long-term exposures to ambient PM;
• New studies of infants and children as a potentially susceptible population; and
• New studies providing insights into the public health impacts of ambient PM
associations with mortality, as well as with other health indices (e.g., physician visits).
Evaluation of the new epidemiologic studies, in conjunction with previously existing ones
involves consideration of several salient aspects of the evidence so as to reach conclusions as to
the likely causal significance of observed associations between ambient PM indicators and
various health endpoints. As discussed in Section 8.1.4, these aspects include what can be
8-336
-------
generally characterized as the strength and consistency of the epidemiologic evidence, as well as
broader aspects of plausibility and coherence that reflect an integration of the epidemiologic
evidence with information derived from other types of studies (e.g., exposure, dosimetry,
toxicology, etc.). Evaluation of the evidence involves an objective appraisal of these salient
aspects, recognizing that they do not lend themselves to the application of simple formulas for
reaching conclusions with a known degree of certainty, but rather involve an exercise in reaching
scientific judgments, taking into account the broad range of views held by the scientific experts
engaged in this review. Conclusions derived from such an appraisal of the epidemiologic
evidence are presented below, with a broader, more integrative synthesis of all relevant
information being presented in Chapter 9.
(1) Thoracic Particles. An extensive body of epidemiology evidence, confirming earlier-
reported associations between short- and long-term exposures (inferred from stationary
air monitor measures) to ambient thoracic particles (typically indexed by PM10) and
mortality /morbidity effects, supports the general conclusion that ambient thoracic
particles, acting alone and/or in combination with gaseous co-pollutants, are likely
causally related to various human health endpoints.
The strength of the evidence across such endpoints includes especially strong evidence
for PM10 associations with total (nonaccidental) mortality. A large majority of relevant mortality
studies show positive PM10 effect estimates, with most all (especially the relatively more precise)
estimates being statistically significant. In particular, several multicity studies in the U.S.,
Canada, and Europe provide strong support for this conclusion, reporting statistically significant
associations with total mortality effect estimates ranging from -1.0 to 3.5% (per 50 |ig/m3
24-h PM10 increment). These estimates are generally within (but toward the lower end of) the
range of PM10 estimates previously reported in the 1996 PM AQCD. It is notable that the effect
estimates from the largest of the multicity studies (for the 90 largest U.S. cities) have also been
shown to be robust to the inclusion of gaseous co-pollutants, and the significance of the effect
estimates has been shown to be robust to the use of alternative statistical models. The multicity
estimates as well as total mortality risk estimates from many individual-city studies, generally
falling in the range of-1.0 to 8.0% per 50 |ig/m3 24-h PM10 increment, also comport well with
results of numerous new studies reporting increased cause-specific cardiovascular- and
8-337
-------
respiratory-related mortality (most statistically significant) and/or cardiovascular and respiratory-
related (most statistically significant) morbidity effects.
(2) Fine-fraction particles. A growing body of epidemiologic evidence both (a) confirms
associations between short- and long-term ambient exposures (inferred from stationary
air monitor measures) to fine-fraction particles (generally indexed by PM2 5) and
various mortality or morbidity endpoint effects and (b) supports the general conclusion
that PM2 5 (or one or more PM2 5 components), acting alone and/or in combination with
gaseous co-pollutants, are likely causally related to observed ambient fine particle-
associated health effects.
The strength of the evidence varies across such endpoints, with relatively stronger
evidence of associations with cardiovascular than respiratory endpoints. As seen in the PM10
studies, a large majority of studies of fine-fraction particles show positive effects estimates,
with most all of the relatively more precise estimates being statistically significant. In
addition, mortality associations with long-term exposures to PM2 5, in conjunction with
evidence of associations with short-term exposures, provide strong evidence in support of a
casual inference. This conclusion is also supported by studies showing associations with
ultrafme particles and other fine-particle components (e.g., sulfates), and by studies showing
associations with air pollution factors linked to key sources of fine-fraction particles (e.g.,
motor vehicles, other oil and/or coal combustion sources, etc).
(3) Coarse-fraction particles. A much more limited body of evidence is suggestive of
associations between short-term (but not long-term) exposures (inferred from stationary
air monitor measures) to ambient coarse-fraction thoracic particles (generally indexed
by PM10_25) and various mortality and morbidity effects observed at times in some
locations. This suggests that PM10_2 5, or some constituent component(s) of PM10_2 5,
may contribute under some circumstances to increased human health risks.
The strength of the evidence varies across endpoints, with somewhat stronger evidence
for coarse-fraction particle associations with morbidity (especially respiratory) endpoints than
for mortality. 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_2 5 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 (e.g., endotoxins, fungi, etc.) during warm months.
8-338
-------
Other studies suggest that particles of crustal origin are generally unlikely to exert notable
health effects under most ambient exposure conditions. Some exceptions may include
situations where crustal particles have come to be heavily contaminated by metals originally
emitted as fine particles from smelting operations but deposited over many years on soils
around smelters, steel mills, etc. (see Item 10, below). Also, in some U.S. cities (especially in
the NW and the SW) where PM10_2 5 tends to be a large fraction of PM10, measurements, coarse
thoracic particles from woodburning are often an important source during at least some
seasons. In such situations, the relationship between hospital admissions and PM10 may be an
indicator of response to coarse thoracic particles from wood burning.
(4) Co-pollutant confounding and effects modification. Much progress has been made in
sorting out contributions of ambient PM10 and its components to observed health effects
relative to other co-pollutants; and, despite continuing uncertainties, the evidence
overall tends to support the above conclusions that ambient PM10 and PM2 5 are most
clearly associated with mortality/morbidity effects, acting either alone or in
combination with other covarying gaseous pollutants, with more limited support with
regard to PM10_2 5.
A major methodological issue affecting epidemiology studies of both short-term and
long-term PM exposure effects relates to use of appropriate approaches for evaluating the
extent to which other air pollutants correlated with ambient PM, including gaseous criteria
pollutants (e.g, O3, NO2, SO2, CO), air toxics, and/or bioaerosols, may confound or modify
PM-related effects estimates. A variety of statistical methods for assessing potential
confounding arising from these associations have been employed. However, no clear
consensus yet exists as to what methods may be most appropriate or adequate for many
specific cases. The inclusion of multiple pollutants often produces statistically unstable
estimates (for PM, at times, and/or for other gaseous co-pollutants), such that this commonly
applied approach has inherent limitations in disentangling the effects of highly correlated
pollutants. Omission of other well correlated, potentially-contributing pollutants, on the other
hand, may incorrectly attribute some of their independent effects to PM or obscure possible
modifying of PM effects by them. Still, progress has been made in evaluating effects of
ambient PM and those of other co-pollutants; and, overall, the new evidence tends to
8-339
-------
substantiate that observed PM effects are at least partly due to ambient PM acting alone or in
the presence of other covarying gaseous pollutants.
(5) Alternative Model Specifications for Meteorological Variables. The results of available
epidemiologic studies, using a variety of approaches to control for weather effects,
appear to demonstrate increased PM-related mortality and morbidity risks beyond those
attributable to weather influences alone. However, there is no clear consensus at this
time with regard to what constitutes appropriate or adequate model specifications to
control for possible weather contributions to those human mortality/morbidity effects
attributed to PM exposure and/or on how best to characterize possible joint (interactive)
effects of weather and ambient PM or other air pollutants.
A wide variety of statistical approaches have been used in attempting to control for
weather effects. Temperature extremes (hot or cold) are well known to cause increased
morbidity and mortality, leading some investigations to simply characterize cities as "hot" or
"cold" (based on annual mean temperatures) and to compare PM effect estimates across such
categories. Others have included temperature and/or humidity as continuous linear variables
in models and then tested for PM or gaseous pollutant effects on remaining risk residuals.
Others have used widely varying model specifications for nonlinear temperature-response
curves, with varying numbers of knot points, types of splines (natural, penalized, etc.), and
numbers of degrees of freedom used in certain models (e.g. GAM). Still others have argued
for and made at least preliminary attempts to use "synoptic weather categories" that define
daily combinations of temperature, humidity, and/or other weather variables as constituting
"offensive weather patterns" associated with increased risk of morbidity/mortality in a given
city, with such "offensive" synoptic patterns varying from city to city in different regions.
Higher temperatures and/or humidity combinations, for example, are required in certain
southern U.S. cities (e.g. New Orleans, Miami, Atlanta, etc.) to reach "heat index" levels
associated with increased risk of heat stroke and/or heat-related deaths than in northern U.S.
cities (e.g. St. Louis, Chicago, New York, etc.). One study tested a large number of
parametric and nonparametric models with different model specifications for weather
variables and found very consistent PM effect size estimates (all statistically significant), even
for those models using synoptic weather patterns in several of the models. It is not clear,
however, as to what extent the PM effect estimates would be reduced in reanalyses of any of
8-340
-------
the original GAM-related model runs in that study. New evidence is also emerging for
possible weather-related modification of air pollution effects (or vice versa), such as results
indicative of more deaths occurring on high temperature/humidity days that also have elevated
PM or O3 levels present than on high heat index days with cleaner air.
(6) Measurement Error. Newly available statistical simulation studies highlight the
importance of considering differential measurement error in assessing and interpreting
epidemiologic findings concerning the magnitude and precision of PM effect estimates,
especially in relation to comparison of the relative strength or robustness of effect
estimates attributed to one or another PM indicator (e.g., PM10, PM2 5, PM10_2 5, etc.) or
comparison of such to gaseous co-pollutant (e.g., O3, NO2, SO2, CO) effect estimates.
The simulation studies indicate that the greater the measurement error associated with
exposure estimates for a given pollutant or indicator, then the less precise the effect size
estimate and the less robust it tends to be in multipollutant models. Of importance, directly
measured PM10 and PM2 5 values likely have less measurement error than PM10_2 5 values
derived by subtracting (differencing) between PM10 and PM2 5 readings (or city-wide averages
of them), especially if obtained from non-collocated PM10 and PM25 monitors at different
locations in a given urban area. Also, gaseous pollutant exposure estimates based on hourly or
daily measurements at many monitoring sites in an urban area are likely subject to less
measurement error than PM10 or PM25 samples obtained on l-in-6 day monitoring schedules at
fewer locations or extrapolated from measures of other PM indicators (e.g. PM10 from TSP
data) or other types of data (e.g., estimating fine particle or PM25 levels based on airport
visibility via use of light extinction calculations). Importantly, available simulation studies
show that "transfer of effects", wherein the effects of one pollutant (e.g., one or another
gaseous co-pollutant) are inappropriately attributed to another (e.g. PM10 or PM2 5) in
multipollutant models, can occur only under very unusual circumstances, e.g., with
simultaneously very high positive or negative correlation (r > .90) between ambient PM
indicators and co-pollutant levels and high negative correlations between their respective
measurement errors, conditions not yet reported for real world data sets.
(7) Alternative Lag Structures. Different PM size components or particles with different
composition or sources may produce effects by different mechanisms manifested at
different lags, and different preexisting health conditions may lead to different delays
between exposure and effect.
8-341
-------
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. It is plausible that effects linked with PM may arise from different responses or
PM-associated diseases with different characteristic lags, for example, that cardiovascular
responses may arise almost immediately after exposure, within zero or one day lags or even
within two hours (Peter et al., 2001a, for myocardial infarction). In contrast, a number of
studies on respiratory symptoms have reported finding larger effect estimates with moving
average lag models (for example, Mortimer et al., 2001). One would then expect to see
different best-fitting lags for different effects, based on potentially different biological
mechanisms as well as individual variability in responses. If various health effects are
substantiated by toxicological evidence as likely occurring at different lag days, so that the
risks for each lag day should be additive, then higher overall risks may exist that than are
implied by maximum estimates for any given single day lag. In that case, multi-day averages
or distributed lag models should be used to project more fully any potential PM-related public
health risks.
(8) Cardiovascular Endpoints. Numerous time series studies indicate that increased
cardiovascular-related mortality and/or morbidity risks are associated with short-term
(< 24-h) exposure to ambient particles (especially PM10 and/or PM2 5).
Cardiovascular mortality risks appear to be increased most strongly (especially for those
> 65 years old) with PM2 5 and occur within short lag times (0-1 day). Morbidity measures,
e.g., cardiovascular hospital admissions and emergency department visits are also positively
(but not as statistically significantly) related to short-term (24-h) PM2 5 exposures. Several
different panel studies (but not all) that evaluated temporal associations between PM
exposures and measures of heart beat rhythm in elderly subjects found results suggestive of
ambient PM exposure being associated with changes in electrocardiographic (ECG) markers
of cardiac function, e.g., altered heart rate variability (HRV), shown in other studies to be
indicators of increased risk for serious cardiovascular outcomes (e.g., heart attacks).
However, conflicting implications of the specific alterations in ECG patterns indicative of
likely predominance of sympathetic versus parasympathetic cardiac control preclude clear
8-342
-------
conclusions. Other studies point toward changes in blood characteristics (e.g., alterations in
C-reactive protein levels, fibrinogen levels, blood viscosity, etc.) related to increased risk of
ischemic heart disease also being associated with ambient PM exposures. However, these
heart rhythm and blood chemistry findings should currently be viewed as providing only very
limited suggestive evidence indicative of potential pathophysiologic alterations contributing to
serious PM-related cardiovascular effects (e.g., myocardial infarction, stroke, death).
(9) Respiratory Endpoints. Notable new evidence now exists which substantiates positive
associations between ambient PM concentrations and (a) increased respiratory-related
hospital admissions, emergency department, and other medical visits; (b) increased
incidence of asthma and other respiratory symptoms; and (c) decrements in pulmonary
functions.
Of much interest are new findings tending to implicate not only fine particle components
but also coarse thoracic (e.g., PM10_25) particles as likely contributing to exacerbation of
various respiratory conditions (e.g., asthma). Also of much interest are emerging new
findings indicative of likely increased occurrence of chronic bronchitis in association with
(especially chronic) PM exposure. New reanalyses or extensions of earlier prospective cohort
studies of long-term ambient PM exposure effects also show substantial evidence for
increased lung cancer risk being associated with such PM exposures, especially exposure to
fine PM or specific fine particles subcomponents (e.g., sulfates) and/or associated precursors
(e.g., S02).
(10) Spatial Heterogeneity of PM Effects. There appears to be greater spatial heterogeneity
in city-specific excess risk estimates for relationships between short-term ambient PM10
concentrations and acute health effects than was previously evident.
The reasons for such variation in effects estimates are not well understood. Factors likely
contributing to the apparent heterogeneity include geographic differences in air pollution
mixtures, composition of ambient PM components, and personal and sociodemographic
factors potentially affecting PM exposure (such as use of air conditioning), as well as
differences in PM mass concentration. For example, the Utah Valley studies 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
8-343
-------
the PM exposure reduction while the steel mill was closed. In contrast, when PM10 and PM2 5
samples were relatively higher in crustal particles during windblown dust episodes in Spokane
and at three central Utah sites than at other times, they were not associated with higher total
mortality during those periods. These differences require more research that may become
more feasible as the PM2 5 sampling network produces air quality data for speciated samples.
Certain classes of ambient particles appear to be distinctly less toxic than others and are
unlikely to exert human health effects at typical ambient exposure concentrations (or perhaps
only under special circumstances). For example, particles of crustal origin, which are
predominately in the coarse fraction, are relatively non-toxic under most circumstances,
compared to combustion-related particles (such as from coal and oil combustion, wood
burning, etc.) However, under some conditions, crustal particles may become sufficiently
toxic to cause human health effects. For example, resuspended crustal particles may be
contaminated with toxic trace elements and other components from previously deposited fine
PM, e.g., metals from smelters (Phoenix) or steel mills (Steubenville, Utah Valley), PAHs
from automobile exhaust, or pesticides from agricultural lands. Fine particles of differing
composition from different sources may also vary in toxic potency and in associated health
risks. 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.
The above reasons suggest that it is inadvisable to pool epidemiology studies involving
different locations, different time periods, different population subgroups, or different health
endpoints, without assessing potential causes and the consequences of these differences.
However, multicity analyses using common data bases, measurement devices, analytical
strategies, and extensive independent external review (as carried out in the NMMAPS and
APHEA studies) are useful. Quantitative meta-analyses of more diverse collections of
independent studies of different cities, varying in methodologies used and/or in data quality or
representativeness, are likely less credible.
(11) Effects of Long-term PM Exposure. Long-term PM exposure durations, on the order of
months to years, are statistically associated with human health effects (indexed by
mortality, development of chronic respiratory disease, and changes in lung function).
8-344
-------
Notable uncertainties remain regarding the magnitude of and mechanisms underlying
chronic health effects of long-term PM exposures and relationships between chronic exposure
effects and acute responses to short-term exposures. Prospective cohort studies providing
mortality risk estimates likely most representative of the general U.S. population report higher
PM effect estimates for mortality associations with chronic long term exposures to PM2 5 and/or
sulfates than with acute short-term exposures to these fine particle indicators. Also, the most
recent extension of the ACS study, more than doubling the original followup time, provides the
strongest evidence to date for increased lung cancer mortality risk being significantly associated
with long-term fine particle exposures. Studies that combine the features of cross-sectional and
cohort studies provide some of the best evidence for noncancer morbidity effects of chronic PM
exposure.
(12) Intervention Studies. Certain epidemiology evidence suggests that relatively sharp
reductions in ambient PM concentrations may reduce a variety of health effects on a
time scale from a few days to a few months.
This has been observed in epidemiology studies of "natural or "found experiments," such as
in the Utah Valley, and by supporting toxicology studies using particle extracts from ambient
community PM10 sampling filters from the Utah Valley. Another study in Dublin, Ireland also
provides evidence for reductions in ambient PM air pollution (measured as British smoke) being
associated with reductions in mortality rates. Another "found experiments" also provide evidence
for decreases in mortality and/or morbidity being associated with notable declines in SO2 as the
result of interventions aimed at reducing air pollution.
(13) Concentration-Response Functions. The results from large multicity studies suggest
that there is no strong evidence of a clear threshold for PM mortality effects. Some
single city studies suggest a hint of a threshold, but not in a statistically clear manner.
More data may need to be examined with alternative approaches (e.g., Smith et al.'s
parametric model), but meanwhile, the use of a linear PM effect models appears to
be appropriate.
Certain statistical simulation analyses have shown that increasing measurement error tends
to flatten PM concentration-response curves somewhat and to increase uncertainty associated
with estimates of potential thresholds (especially under extreme error scenarios). Nevertheless, it
has been concluded that if thresholds exist, standard statistical analyses should be able to detect
8-345
-------
them. Newly available evaluations of the shape of PM-related concentrations-response
relationships provide very limited results suggestive of possible threshold(s) for health effects
associated with low ambient PM concentrations (e.g., at < 15 to 20 |ig/m3 for 24-h PM10
or < 20 to 25 |ig/m3 24-h PM2 5 levels). However, formal statistical tests comparing linear
(no threshold) models versus various nonlinear or threshold models have not shown statistically
significant distinctions between them or clear preference of one over the other. Results of
analyses of NMMAPS data for the 20 largest U.S. cities, that compared a linear model for PM10,
a natural cubic spline model of PM10 with knots at 30 and 60 |ig/m3, and a threshold model with
grid search in 5 |ig/m3 increments across 5 to 200 |ig/m3 PM10 suggested possible thresholds for
daily total or cardiorespiratory mortality at PM10 levels below -15 to 20 |ig/m3, but essentially
zero probability of a threshold above -25 |ig/m3. However, comparing AIC values across the
models suggested that the linear (no-threshold) model would be preferred over the others. Other
single-city analyses were suggestive of possible threshold change points in Birmingham and
Chicago at 80 and > 100 |ig/m3 PM10 but not statistically significantly so. In another single-city
(Phoenix) study using a piecewise linear model or a B-spline model with 4 knots, some evidence
was found to suggest a possible daily total mortality threshold(s) in the range of 20 to 25 |ig/m3
PM2 5, but no evidence was found for threshold(s) for total mortality associations with PM10_2 5
(perhaps reflecting greater measurement error for PM10_2 5 exposure estimates in the analysis).
(14) Public Health Implications. Progress has been made in advancing our understanding of
public health implications of PM mortality and morbidity effects, both in terms of
(a) potential life shortening due to PM exposures and (b) a broader array of morbidity
effects shown to be associated with ambient PM exposures.
Long-term exposures (on the order of years or decades) to thoracic particles in general and
fine-fraction particles in particular appear to be associated with life shortening well beyond that
accounted for by the simple accumulation of the more acute "harvesting" of effects of short-term
PM exposures (on the order of a few days). Investigations of the public health implications of
such long-term PM exposure-mortality effect estimates have been attempted. For example,
preliminary life table calculations using risk estimates from long-term PM2 5 exposure studies
suggest that relatively small differences in long-term exposure to ambient PM can have
substantial effects on life expectancy. To illustrate, based on the initial 1995 ACS study PM-
8-346
-------
mortality risk estimates, a U.S. EPA calculation for the 1969-71 life table for U.S. white males
projected that a chronic exposure increase of 10 |ig/m3 PM25 could be associated with a reduction
of 1.31 years for the entire U.S. population's life expectancy at age 25. However, such
projections must be viewed with caution given their dependence on the specific PM effect-size
estimates used in the calculations. The 1.31 year life expectancy reduction estimate, as an
example, would need to be recalculated to a lower value based on lower (and presumably more
credible) PM mortality risk estimates from the more recent Pope et al. (2002) extension of the
ACS study.
PM-related health effects in infants and children are emerging as an area of more concern
than in the 1996 PM AQCD; and ultimately, such health effects could have very substantial
implications for life expectancy calculations. However, only very limited evidence currently
exists about potential ambient PM relationships with some of the more serious pertinent health
endpoints (low birth weight, preterm birth, neonatal and infant mortality, emergency hospital
admissions, and mortality in older children). Also, 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 a broader array of health endpoints indicate ambient
PM associations with increased nonhospital medical visits (physician visits) and asthma effects.
Such new findings suggest likely much larger health impacts and costs to society due to ambient
PM than just those indexed either by just hospital admissions/visits and/or mortality.
8-347
-------
REFERENCES
Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
Adventists. Environ. Health Perspect. 94: 43-50.
Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
7: 19-34.
Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
2.5 microns in aerodynamic diameter (PM2 5) from airport visibility data in California. J. Exposure Anal.
Environ. Epidemiol. 5: 161-180.
Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med. 158:
289-298.
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
Care Med. 159: 373-382.
Ackermann-Liebrich, U.; Leuenberger, P.; Schwartz, J.; Schindler, C.; Monn, C.; Bolognini, B.; Bongard, J. P.;
Brandli, O.; Domenighetti, G.; Elsasser, S.; Grize, L.; Karrer, W.; Keller, R.; Keller-Wossidlo, H.; Ktinzli, N.;
Martin, B. W.; Medici, T. C.; Perruchoud, A. P.; Schoni, M. H.; Tschopp, J. M.; Villiger, B.; Wuthrich, B.;
Zellweger, J. P.; Zemp, E. (1997) Lung function and long term exposure to air pollutants in Switzerland. Am.
J. Respir. Crit. Care Med. 155: 122-129.
Agocs, M. M.; White, M. C.; Ursicz, G.; Olson, D. R.; Vamos, A. (1997) A longitudinal study of ambient air
pollutants and the lung peak expiratory flow rates among asthmatic children in Hungary. Int. J. Epidemiol.
26: 1272-1280.
Alberdi Odriozola, J. C.; Diaz Jimenez, J.; Montero Rubio, J. C.; MironPerez, I. J.; Pajares Ortiz, M. S.; Ribera
Rodrigues, P. (1998) Air pollution and mortality in Madrid, Spain: a time-series analysis. Int. Arch. Occup.
Environ. Health 71: 543-549.
Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Strachan, D. P. (1996) Air pollution and daily
mortality in London: 1987-92. Br. Med. J. 312: 665-669.
Anderson, H. R.; Spix, C.; Medina, S.; Schouten, J. P.; Castellsague, J.; Rossi, G.; Zmirou, D.; Touloumi, G.;
Wojtyniak, B.; Ponka, A.; Bacharova, L.; Schwartz, J.; Katsouyanni, K. (1997) Air pollution and daily
admissions for chronic obstructive pulmonary disease in 6 European cities: results from the APHEA project.
Eur. Respir. J. 10: 1064-1071.
Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Emberlin, J.; Strachen, D. P. (1998) Air pollution,
pollens, and daily admissions for asthma in London 1987-92. Thorax 53: 842-848.
Anderson, H. R.; Bremner, S. A.; Atkinson, R. W.; Harrison, R. M.; Walters, S. (2001) Paniculate matter and daily
mortality and hospital admissions in the west midlands conurbation of the United Kingdom: associations with
fine and coarse particles, black smoke and sulphate. Occup. Environ. Med. 58: 504-510.
Archer, V. E. (1990) Air pollution and fatal lung disease in three Utah counties. Arch. Environ. Health 45: 325-334.
Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Ponce de Leon, A. (1999a)
Short-term associations between emergency hospital admissions for respiratory and cardiovascular disease and
outdoor air pollution in London. Arch. Environ. Health 54: 398-411.
Atkinson, R. W.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Bremner, S. A.; Ponce de Leon, A. (1999b)
Short-term associations between outdoor air pollution and visits to accident and emergency departments in
London for respiratory complaints. Eur. Respir. J. 13: 257-265.
Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2001) Acute effects of paniculate air pollution on
respiratory admissions: results from APHEA 2 project. Am. J. Respir. Crit. Care Med. 164: 1860-1866.
Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2003) Acute effects of paniculate air pollution on
respiratory admissions. In: Revised analyses of time-series studies of air pollution and health. Special report.
Boston, MA: Health Effects Institute; pp. 81-84. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
8-348
-------
Avol, E. L.; Gauderman, W. I; Tan, S. M; London, S. I; Peters, J. M. (2001) Respiratory effects of relocating to
areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
Awasthi, S.; Click, H. A.; Fletcher, R. H.; Ahmed, N. (1996) Ambient air pollution & respiratory symptoms complex
in preschool children. Indian J. Med. Res. 104: 257-262.
Bailey, D. L. R.; Clayton, P. (1982) The measurement of suspended particle and total carbon concentrations in the
atmosphere using standard smoke shade methods. Atmos. Environ. 16: 2683-2690.
Baldi, L; Tessier, J. F.; Kauffmann, F.; Jacqmin-Gadda, H.; Nejjari, C.; Salamon, R. (1999) Prevalence of asthma and
mean levels of air pollution: results from the French PAARC survey. Eur. Respir. J. 14: 132-138.
Barbone, F.; Bovenzi, M.; Cavalleri, F.; Stanta, G. (1995) Air pollution and lung cancer in Trieste, Italy. Am. J.
Epidemiol. 141: 1161-1169.
Barker, D. J. P.; Gluckman, P. D.; Godfrey, K. M.; Harding, J. E.; Owens, J. A.; Robinson, J. S. (1993) Fetal nutrition
and cardiovascular disease in adult life. Lancet 341: 938-941.
Basu, R.; Samet, J. M. (1999) A review of the epidemiological evidence on health effects of nitrogen dioxide
exposure from gas stoves. J. Environ. Med. 1: 173-187.
Beckett, W. S. (2001) The air pollution detectives. Am. J. Respir. Crit. Care Med. 164: 515-516.
Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms and
pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
Berkowitz, G. S.; Papiernik, E. (1993) Epidemiology of pretermbirth. Epidemiol. Rev. 15: 414-443.
Beyer, U.; Franke, K.; Cyrys, J.; Peters, A.; Heinrich, J.; Wichmann, H. E.; Brunekreef, B. (1998) Air pollution and
respiratory health of children: the PEACE panel study in Hettstedt and Zerbst, Eastern Germany. Eur. Respir.
Rev. 8: 61-69.
Bobak, M.; Leon, D. A. (1992) Air pollution and infant mortality in the Czech Republic, 1986-1988.
Lancet (8826): 1010-1014.
Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
Bobak, M.; Roberts, A. (1997) Heterogeneity of air pollution effects is related to average temperature [letter].
Br. Med. J. 315: 1161.
Boezen, M.; Schouten, J.; Rijcken, B.; Vonk, J.; Gerritsen, J.; VanDerZee, S.; Hoek, G.; Brunekreef, B.; Postma, D.
(1998) Peak expiratory flow variability, bronchial responsiveness, and susceptibility to ambient air pollution in
adults. Am. J. Respir. Crit. Care Med. 158: 1848-1854.
Boezen, H. M.; VanDerZee, S. C.; Postma, D. S.; Vonk, J. M.; Gerritsen, J.; Hoek, G.; Brunekreef, B.; Rijcken, B.;
Schouten, J. P. (1999) Effects of ambient air pollution on upper and lower respiratory symptoms and peak
expiratory flow in children. Lancet 353: 874-878.
Borja-Aburto, V H.; Loomis, D. P.; Bangdiwala, S. I.; Shy, C. M.; Rascon-Pacheco, R. A. (1997) Ozone, suspended
particulates, and daily mortality in Mexico City. Am. J. Epidemiol. 145: 258-268.
Borja-Aburto, V H.; Castillejos, M.; Gold, D. R.; Bierzwinski, S.; Loomis, D. (1998) Mortality and ambient fine
particles in southwest Mexico City, 1993-1995. Environ. Health Perspect. 106: 849-855.
Braga, A. L. F.; Conceicao, G. M. S.; Pereira, L. A. A.; Kishi, H. S.; Pereira, J. C. R.; Andrade, M. F.; Goncalves,
F. L. T.; Saldiva, P. H. N.; Latorre, M. R. D. O. (1999) Air pollution and pediatric respiratory hospital
admissions in Sao Paulo, Brazil. J. Environ. Med. 1: 95-102.
Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2000) Do respiratory epidemics confound the association between air
pollution and daily deaths? Eur. Respir. J. 16: 723-728.
Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (200la) The lag structure between paniculate air pollution and respiratory
and cardiovascular deaths in ten U.S. cities. J. Occup. Environ. Med. 43: 927-933.
Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (200 Ib) The time course of weather-related deaths. Epidemiology
12: 662-667.
Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2002) The effect of weather on respiratory and cardiovascular deaths in
12 U.S. cities. Environ. Health Perspect. 110: 859-864.
Brauer, M.; Ebelt, S. T.; Fisher, T. V.; Brumm, J.; Petkau, A. J.; Vedal, S. (2001) Exposure of chronic obstructive
pulmonary disease patients to particles: respiratory and cardiovascular health effects. J. Exposure Anal.
Environ. Epidemiol. 11: 490-500.
Braun-Fahrlander, C.; Vuille, J. C.; Sennhauser, F. H.; Neu, U.; Ktinzle, T.; Grize, L.; Gassner, M.; Minder, C.;
Schindler, C.; Varonier, H. S.; Wuthrich, B.; SCARPOL team. (1997) Respiratory health and long-term
exposure to air pollutants in Swiss schoolchildren. Am. J. Respir. Crit. Care Med. 155: 1042-1049.
8-349
-------
Bremner, S. A.; Anderson, H. R.; Atkinson, R. W.; McMichael, A. I; Strachan, D. P.; Bland, J. M; Bower, J. S.
(1999) Short term associations between outdoor air pollution and mortality in London 1992-4. Occup. Environ.
Med. 56: 237-244.
Brook, J. R.; Wiebe, A. H.; Woodhouse, S. A.; Audette, C. V.; Dann, T. F.; Callaghan, S.; Piechowski, M.;
Dabek-Zlotorzynska, E.; Dloughy, J. F. (1997) Temporal and spatial relationships in fine particle strong
acidity, sulphate, PM10, andPM25 across multiple Canadian locations. Atmos. Environ. 31: 4223-4236.
Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
Buffler, P. A.; Cooper, S. P.; Stinnett, S.; Contant, C.; Shirts, S.; Hardy, R. J.; Agu, V.; Gehan, B.; Burau, K. (1988)
Air pollution and lung cancer mortality in Harris County, Texas, 1979-1981. Am. J. Epidemiol. 128: 683-699.
Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian cities.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Burnett, R. T.; Dales, R. E.; Raizenne, M. E.; Krewski, D.; Summers, P. W.; Roberts, G. R.; Raad-Young, M.;
Dann, T.; Brook, J. (1994) Effects of low ambient levels of ozone and sulfates on the frequency of respiratory
admissions to Ontario hospitals. Environ. Res. 65: 172-194.
Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
paniculate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
142: 15-22.
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997a) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105: 614-620.
Burnett, R. T.; Dales, R. E.; Brook, J. R.; Raizenne, M. E.; Krewski, D. (1997b) Association between ambient carbon
monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities.
Epidemiology 8: 162-167.
Burnett, R. T.; Brook, J. R.; Yung, W. T.; Dales, R. E.; Krewski, D. (1997c) Association between ozone and
hospitalization for respiratory diseases in 16 Canadian cities. Environ. Res. 72: 24-31.
Burnett, R. T.; Cakmak, S.; Brook, J. R. (1998a) The effect of the urban ambient air pollution mix on daily mortality
rates in 11 Canadian cities. Can. J. Public Health 89: 152-156.
Burnett, R. T.; Cakmak, S.; Raizenne, M. E.; Stieb, D.; Vincent, R.; Krewski, D.; Brook, J. R.; Philips, O.;
Ozkaynak, H. (1998b) The association between ambient carbon monoxide levels and daily mortality in
Toronto, Canada. J. Air Waste Manage. Assoc. 48: 689-700.
Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54:130-139.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between paniculate- and gas-phase components of urban air pollution and daily mortality
in eight Canadian cities. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol.
12(suppl. 4): 15-39.
Burnett, R.; Ma, R.; Jerrett, M.; Goldberg, M. S.; Cakmak, S.; Pope, C. A., Ill; Krewski, D. (2001a) The spatial
association between community air pollution and mortality: a new method of analyzing correlated geographic
cohort data. Environ. Health Perspect. Suppl. 109(3): 375-380.
Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Raizenne, M. E.; Brook, J. R.; Dales, R. E.; Leech, J. A.; Cakmak, S.;
Krewski, D. (200 Ib) Association between ozone and hospitalization for acute respiratory diseases in children
less than 2 years of age. Am. J. Epidemiol. 153: 444-452.
Cakmak, S.; Burnett, R. T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
concentrations for air pollution-related mortality with exposure measurement error. Risk Anal. 19: 487-496.
Calderon-Garciduenas, L.; Mora-Tiscareno, A.; Chung, C. J.; Valencia, G.; Fordham, L. A.; Garcia, R.; Osnaya, N.;
Romero, L.; Acuna, H.; Villarreal-Calderon, A. (2000) Exposure to air pollution is associated with lung
hyperinflation in healthy children and adolescents in southwest Mexico City: a pilot study. Inhalation Toxicol.
12: 537-561.
Carroll, R. J.; Ruppert, D.; Stefanski, L. A. (1995) Measurement error in nonlinear models. London, United Kingdom:
Chapman & Hall. (Cox, D. R.; Hinkley, D. V; Keiding, N.; Reid, N.; Rubin, D. B.; Silverman, B. W., eds.
Monographs on statistics and applied probability: v. 63).
Carrothers, T. J.; Evans, J. S. (2000) Assessing the impact of differential measurement error on estimates of fine
particle mortality. J. Air Waste Manage. Assoc. 50: 65-74.
Cassino, C.; Ito, K.; Bader, L; Ciotoli, C.; Thurston, G.; Reibman, J. (1999) Cigarette smoking and ozone-associated
emergency department use for asthma by adults in New York City. Am. J. Respir. Crit. Care Med.
159: 1773-1779.
8-350
-------
Castillejos, M; Borja-Aburto, V. H.; Dockery, D. W.; Gold, D. R.; Loomis, D. (2000) Airborne coarse particles and
mortality. In: Phalen, R. F., ed. Inhalation toxicology: proceedings of the third colloquium on paniculate air
pollution and human health (first special issue); June, 1999; Durham, NC. Inhalation Toxicol.
12(suppl. 1): 61-72.
Checkoway, H.; Levy, D.; Sheppard, L.; Kaufman, I; Koenig, I; Siscovick, D. (2000) A case-crossover analysis of
fine paniculate matter air pollution and out-of-hospital sudden cardiac arrest. Cambridge, MA: Health Effects
Institute; research report no. 99. Available: http://www.healtheffects.org/Pubs/Checkoway.pdf [18 October,
2004].
Chen, C.; Chock, D. P.; Winkler, S. L. (1999) A simulation study of confounding in generalized linear models for air
pollution epidemiology. Environ. Health Perspect. 107: 217-222.
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
Chew, F. T.; Goh, D. Y. T.; Ooi, B. C.; Saharom, R.; Hui, J. K. S.; Lee, B. W. (1999) Association of ambient
air-pollution levels with acute asthma exacerbation among children in Singapore. Allergy (Copenhagen)
54: 320-329.
Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52: 113-117.
Cifuentes, L. A.; Vega, J.; Kopfer, K.; Lave, L. B. (2000) Effect of the fine fraction of paniculate matter versus the
coarse mass and other pollutants on daily mortality in Santiago, Chile. J. Air Waste Manage. Assoc.
50: 1287-1298.
Clancy, L.; Goodman, P.; Sinclair, H.; Dockery, D. W. (2002) Effect of air pollution control on death rates in Dublin,
Ireland: an intervention study. Lancet 360: 1210-1214.
Clench-Aas, J.; Bartonova, A.; Skjensberg, O. H.; Leegaard, J.; Hagen, L. O.; Gisver, P.; Moseng, J.; Roemer, W.
(1998) Air pollution and respiratory health of children: the PEACE study in Oslo, Norway. Eur. Respir. Rev. 8:
36-43.
Clyde, M. A. (1999) Bayesian model averaging and model search strategies. In: Bernardo, J. M.; Berger, J. O.;
Dawid, A. P.; Smith, A. F. M., eds. Bayesian Statistics 6: proceedings of the Sixth Valencia International
Meeting, June; pp. 157-185. Oxford, UK. Oxford, UK: Clarendon Press.
Clyde, M. (2000) Model uncertainty and health effect studies for paniculate matter. Environmetrics 11: 745-763.
Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
Arizona. Seattle, WA: University of Washington, National Research Center for Statistics and the Environment;
NRCSE technical report series, NRCSE-TRS no. 040. Available:
http://www.nrcse.washington.edu/pdf/trs40_pm.pdf [18 October, 2004].
Cohen, A. J. (2000) Outdoor air pollution and lung cancer. Environ. Health Perspect. 108(suppl. 4): 743-750.
Cohen, A. J.; Pope, C. A., III. (1995) Lung cancer and air pollution. Environ. Health Perspect. Suppl.
103(8): 219-224.
Conceicao, G. M. S.; Miraglia, S. G. E. K.; Kishi, H. S.; Saldiva, P. H. N.; Singer, J. M. (2001) Air pollution and
child mortality: a time-series study in Sao Paulo, Brazil. Environ. Health Perspect. Suppl. 109(3): 347-350.
Cook, J. R.; Stefanski, L. A. (1994) Simulation-extrapolation estimation in parametric measurement error models.
J. Am. Stat. Assoc. 89: 1314-1328.
Coull, B. A.; Schwartz, J.; Wand, M. P. (2001) Respiratory health and air pollution: additive mixed model analyses.
Biostatistics 2: 337-349.
Coutant, R. W. (1977) Effect of environmental variables on collection of atmospheric sulfate. Environ. Sci. Technol.
11:873-878.
Creason, J.; Neas, L.; Walsh, D.; Williams, R.; Sheldon, L.; Liao, D.; Shy, C. (2001) Paniculate matter and heart rate
variability among elderly retirees: the Baltimore 1998 PM study. J. Exposure Anal. Environ. Epidemiol.
11: 116-122.
Cropper, M. L.; Simon, N. B.; Alberini, A.; Arora, S.; Sharma, P. K. (1997) The health benefits of air pollution
control in Delhi. Am. J. Agric. Econ. 79: 1625-1629.
Cuijpers, C. E. J.; Swaen, G. M. H.; Wesseling, G.; Wouters, E. F. M. (1994) Acute respiratory effects of summer
smog in primary school children. Toxicol. Lett. 72: 227-235.
8-351
-------
Dab, W.; Medina, S.; Quenel, P.; Le Moullec, Y.; Le Tertre, A.; Thelot, B.; Monteil, C.; Lameloise, P.; Pirard, P.;
Momas, I.; Ferry, R.; Festy, B. (1996) Short term respiratory health effects of ambient air pollution: results of
the APHEA project in Paris. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on
health: a European approach using epidemiological time series data. J. Epidemiol. Community Health
50(suppl. 1): S42-S46.
Damia, A. D.; Fabregas, M. L.; Tordera, M. P.; Torrero, L. C. (1999) Effects of air pollution and weather conditions
on asthma exacerbation. Respiration 66: 52-58.
Daniels, M.; Dominici, F.; Samet, J. M.; Zeger, S. L. (2000) Estimating paniculate matter-mortality dose-response
curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am. J. Epidemiol.
152: 397-406.
Daniels, M. J.; Lee, Y.-D.; Kaiser, M. (2001) Assessing sources of variability in measurement of ambient paniculate
matter. Environmetrics 12: 547-558.
Dejmek, J.; Selevan, S. G.; Benes, L; Solansky, L; Sram, R. J. (1999) Fetal growth and maternal exposure to
paniculate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
Dejmek, J.; Solansky, L; Benes, L; Lenicek, J.; Sram, R. J. (2000) The impact of polycyclic aromatic hydrocarbons
and fine particles on pregnancy outcome. Environ. Health Perspect. 108: 1159-1164.
Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154: 633-641.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997a) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Matteucci, R. M.; Anderson, P. R.; Koutrakis, P. (1997b)
The effect of outdoor fungal spore concentrations on daily asthma severity. Environ. Health Perspect.
105: 622-635.
Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998a) Symptoms inpediatric asthmatics and air pollution:
differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging time.
Environ. Health Perspect. 106: 751-761.
Delfino, R. J.; Murphy-Moulton, A. M.; Becklake, M. R. (1998b) Emergency room visits for respiratory illnesses
among the elderly in Montreal: association with low level ozone exposure. Environ. Res. 76: 61-11.
Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; McLaren, C. E. (2002) Association of asthma symptoms
with peak paniculate air pollution and effect modification by anti-inflammatory medication use. Environ.
Health Perspect. 110: A607-A617.
Desqueyroux, H.; Pujet, J.-C.; Prosper, M.; Squinazi, F.; Momas, I. (2002) Short-term effects of low-level air
pollution on respiratory health of adults suffering from moderate to severe asthma. Environ. Res. A 89: 29-37.
Diaz, J.; Garcia, R.; Ribera, P.; Alberdi, J. C.; Hernandez, E.; Pajares, M. S.; Otero, A. (1999) Modeling of air
pollution and its relationship with mortality and morbidity in Madrid, Spain. Int. Arch. Occup. Environ. Health
72: 366-376.
Divon, M. Y.; Boldes, R.; McGahan, J. P. (1994) Assessment of intrauterine growth retardation. In: McGahan, J. P.;
Porto, M., eds. Diagnostic obstetrical ultrasound. Philadelphia, PA: J. B. Lippincott; pp. 67-82.
Dockery, D. W.; Spengler, J. D. (1981) Personal exposure to respirable particulates and sulfates. J. AirPollut. Control
Assoc. 31: 153-159.
Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
Dockery, D. W.; Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily mortality: associations with particulates
and acid aerosols. Environ. Res. 59: 362-373.
Dockery, D. W.; Pope, C. A., Ill; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
(1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 329: 1753-1759.
Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
symptoms. Environ. Health Perspect. 104: 500-505.
Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
Health Effects Institute; research report no. 83.
Dominici, F.; Samet, J. M.; Zeger, S. (2000a) Combining evidence on air pollution and daily mortality from the
largest 20 US cities: a hierarchical modeling strategy. J. R. Stat. Soc. A 163: 263-302.
Dominici, F.; Zeger, S. L.; Samet, J. (2000b) A measurement error model for time-series studies of air pollution and
mortality. Biostatistics 1: 157-175.
8-352
-------
Dominici, F.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2002) Air pollution and mortality: estimating regional and
national dose-response relationships. J. Am. Stat. Assoc. 97: 100-111.
Dominici, F.; Daniels, M.; McDermott, A.; Zeger, S. L.; Samet, J. M. (2003a) Shape of the exposure-response
relation and mortality displacement in the NMMAPS database. In: Revised analyses of time-series studies of
air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 91-96. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Dominici, F.; McDermott, A.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2003b) Mortality among residents of 90 cities.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 9-24. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Dominici, F.; McDermott, A.; Zeger, S. L.; Samet, J. M. (2003c) Airborne paniculate matter and mortality: timescale
effects in four US Cities. Am. J. Epidemiol. 157: 1055-1065.
Fairley, D. (1990) The relationship of daily mortality to suspended particulates in Santa Clara county, 1980-86.
Environ. Health Perspect. 89: 159-168.
Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
Perspect. 107: 637-641.
Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 97-106.
Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Forsberg, B.; Segerstedt, B.; Stjernberg, N.; Roemer, W. (1998) Air pollution and respiratory health of children:
the PEACE panel study inUmea, Sweden. Eur. Respir. Rev. 8: 12-19.
Friedman, M. S.; Powell, K. E.; Hutwagner, L.; Graham, L. M.; Teague, W. G. (2001) Impact of changes in
transportation and commuting behaviors during the 1996 summer Olympic games in Atlanta on air quality and
childhood asthma. JAMA J. Am. Med. Assoc. 285: 897-905.
Frischer, T.; Studnicka, M.; Gartner, C.; Tauber, E.; Horak, F.; Veiter, A.; Spengler, J.; Ktihr, J.; Urbanek, R. (1999)
Lung function growth and ambient ozone: a three-year population study in school children. Am. J. Respir. Crit.
Care Med. 160: 390-396.
Fung, K. Y.; Krewski, D. (1999) On measurement error adjustment methods in Poisson regression. Environmetrics
10: 213-224.
Fusco, D.; Forastiere, F.; Michelozzi, P.; Spadea, T.; Ostro, B.; Area, M.; Perucci, C. A. (2001) Air pollution and
hospital admissions for respiratory conditions in Rome, Italy. Eur. Respir. J. 17: 1143-1150.
Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98-MP26.03.
Garcia-Aymerich, J.; Tobias, A.; Anto, J. M.; Sunyer, J. (2000) Air pollution and mortality in a cohort of patients
with chronic obstructive pulmonary disease: a time series analysis. J. Epidemiol. Community Health 54: 73-74.
Garty, B. Z.; Kosman, E.; Ganor, E.; Berger, V; Garty, L.; Wietzen, T.; Waisman, Y.; Mimouni, M.; Waisel, Y.
(1998) Emergency room visits of asthmatic children, relation to air pollution, weather, and airborne allergens.
Ann. Allergy Asthma Immunol. 81: 563-570.
Gauderman, W. J.; Mcconnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.; Rappaport,
E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung function
growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
Med. 166: 76-84.
Gauvin, S.; Zmirou, D.; Pin, I.; Quentin, J.; Balducci, F.; Boudet, C.; Poizeau, D.; Brambilla, C. (1999) Short-term
effect of exposure to suspended paniculate matter (PM10) on the respiratory function of urban asthmatic and
control adults. J. Environ. Med. 1: 71-79.
Gehring, U.; Cyrys, J.; Sedlmeir, G.; Brunekreef, B.; Bellander, T.; Fischer, P.; Bauer, C. P.; Reinhardt, D.;
Wichmann, H. E.; Heinrich, J. (2002) Traffic-related air pollution and respiratory health during the first 2 yrs.
of life. Eur. Respir. J. 19: 690-698.
Gherghinova, M.; Kostyanve, S.; Ivanova, M. (1989) Theoretic values of body plethysmography indices in healthy
children between 7 and 14 years of age. Pediatria 28: 52-58.
Gielen, M. H.; VanDer Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
155: 2105-2108.
-------
Gold, D. R.; Damokosh, A. I.; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
Castillejos, M. (1999) Particulate and ozone pollutant effects on the respiratory function of children in
southwest Mexico City. Epidemiology 10: 8-16.
Gold, D. R.; Litonjua, A.; Schwartz, I; Lovett, E.; Larson, A.; Nearing, B.; Allen, G.; Verrier, M.; Cherry, R.;
Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101: 1267-1273.
Gold, D. R.; Schwartz, I; Litonjua, A.; Verrier, R.; Zanobetti, A. (2003) Ambient pollution and reduced heart rate
variability. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
Health Effects Institute; pp. 107-112. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Goldberg, M. S.; Burnett, R. T. (2003) Revised analysis of the Montreal time-series study. In: Revised analyses of
time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 113-132. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible to
short-term increases in paniculate air pollution: a time-series study in Montreal, Quebec. Cambridge, MA:
Health Effects Institute; research report 97. Available: http://www.healtheffects.org/Pubs/Goldberg.pdf
[18 October, 2004].
Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Brook, J.; Bonvalot, Y.; Tamblyn, R.; Singh, R.; Valois, M.-F.
(200 la) The association between daily mortality and ambient air particle pollution in Montreal, Quebec. 1.
Nonaccidental mortality. Environ. Res. 86: 12-25.
Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Brook, J.; Bonvalot, Y.; Tamblyn, R.; Singh, R.; Valois, M.-F.;
Vincent, R. (200 Ib) The association between daily mortality and ambient air particle pollution in Montreal,
Quebec. 2. Cause-specific mortality. Environ. Res. 86: 26-36.
Goldberg, M. S.; Burnett, R. T.; Bailar, J. C., Ill; Tamblyn, R.; Ernst, P.; Flegel, K.; Brook, J.; Bonvalot, Y.;
Singh, R.; Valois, M.-F.; Vincent, R. (200 Ic) Identification of persons with cardiorespiratory conditions
who are at risk of dying from the acute effects of ambient air particles. Environ. Health Perspect.
Suppl. 109(4): 487-494.
Goldberg, M. S.; Burnett, R. T.; Brook, J.; Bailar, J. C., Ill; Valois, M.-F.; Vincent, R. (2001d) Associations between
daily cause-specific mortality and concentrations of ground-level ozone in Montreal, Quebec. Am. J.
Epidemiol. 154: 817-826.
Goldberg, M. S.; Burnett, R. T.; Valois, M.-F.; Flegel, K.; Bailar, J. C., Ill; Brook, J.; Vincent, R.; Radon, K. (2003)
Associations between ambient air pollution and daily mortality among persons with congestive heart failure.
Environ. Res. 91:8-20.
Gordian, M. E.; Ozkaynak, H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Particulate air pollution and respiratory
disease in Anchorage, Alaska. Environ. Health Perspect. 104: 290-297.
Gouveia, N.; Fletcher, T. (2000) Respiratory diseases in children and outdoor air pollution in Sao Paulo, Brazil:
a time series analysis. Occup. Environ. Med. 57: 477-483.
Grievink, L.; VanderZee, S. C.; Hoek, G.; Boezen, H. M.; Van't Veer, P.; Brunekreef, B. (1999) Modulation of the
acute respiratory effects of winter air pollution by serum and dietary antioxidants: a panel study. Eur. Respir. J.
13: 1439-1446.
Gtintzel, O.; Bollag, U.; Helfenstein, U. (1996) Asthma and exacerbation of chronic bronchitis: sentinel and
environmental data in a time series analysis. Zentralbl. Hyg. Umweltmed. 198: 383-393.
Guo, Y. L.; Lin, Y.-C; Sung, F.-C; Huang, S.-L.; Ko, Y.-C; Lai, J.-S.; Su, H.-J.; Shaw, C.-K.; Lin, R.-S.; Dockery,
D. W. (1999) Climate, traffic-related air pollutants, and asthma prevalence in middle-school children in
Taiwan. Environ. Health Perspect. 107: 1001-1006.
Gwynn, R. C.; Thurston, G. D. (2001) The burden of air pollution: impacts among racial minorities. Environ. Health
Perspect. Suppl. 109(4): 501-506.
Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
Hagen, J. A.; Nafstad, P.; Skrondal, A.; Bjerkly, S.; Magnus, P. (2000) Associations between outdoor air pollutants
and hospitalization for respiratory diseases. Epidemiology 11: 136-140.
Hajat, S.; Haines, A.; Goubet, S. A.; Atkinson, R. W.; Anderson, H. R. (1999) Association of air pollution with daily
GP consultations for asthma and other lower respiratory conditions in London. Thorax 54: 597-605.
Hajat, S.; Haines, A.; Atkinson, R. W.; Bremner, S. A.; Anderson, H. R.; Emberlin, J. (2001) Association between air
pollution and daily consultations with general practitioners for allergic rhinitis in London, United Kingdom.
Am. J. Epidemiol. 153: 704-714.
8-354
-------
Haluszka, I; Pisiewicz, K.; Miczynski, I; Roemer, W.; Tomalak, W. (1998) Air pollution and respiratory health in
children: the PEACE panel study in Krakow, Poland. Eur. Respir. Rev. 8: 94-100.
Harre, E. S. M; Price, P. D.; Ayrey, R. B.; Toop, L. I; Martin, I. R.; Town, G. I. (1997) Respiratory effects of air
pollution in chronic obstructive pulmonary disease: a three month prospective study. Thorax 52: 1040-1044.
Harris, R. J. (1975) A primer of multivariate statistics. New York, NY: Academic Press; pp. 155-224.
Haverkate, F.; Thompson, S. G.; Pyke, S. D. M.; Gallimore, J. R.; Papys, M. B. (1997) Production of C-reactive
protein and risk of coronary events in stable and unstable angina. Lancet 349: 462-466.
Health Effects Institute. (2000) Critique. Health Review Committee. In: Checkoway, H.; Levy, D.; Sheppard, L.;
Kaufman, J.; Koenig, J.; Siscovick, D. A case-crossover analysis of fine paniculate matter air pollution and
out-of-hospital sudden cardiac arrest. Cambridge, MA: Health Effects Institute; research report no. 99;
pp. 29-32. Available: http://www.healtheffects.org/Pubs/Checkoway.pdf [18 October, 2004].
Health Effects Institute. (2002) Reanalysis of National Morbidity, Mortality, and Air Pollution Study (NMMAPS)
[letter with attachments]. Boston, MA: May 30.
Health Effects Institute. (2003a) Revised analyses of the National Morbidity, Mortality, and Air Pollution Study
(NMMAPS), part II. In: Revised analyses of time-series studies of air pollution and health. Special report.
Boston, MA: Health Effects Institute; pp. 9-72. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Health Effects Institute. (2003b) Revised analyses of time-series studies of air pollution and health. Boston, MA:
Health Effects Institute; special report. Available: Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [27 June 2003].
Health Effects Institute. (2003c) Commentary on revised analyses of selected studies. In: Revised analyses of
time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 255-290. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Hedley, A. J.; Wong, C.-M.; Thach, T. Q.; Ma, S.; Lam, T.-H.; Anderson, H. R. (2002) Cardiorespiratory and
all-cause mortality after restrictions on sulphur content of fuel in Hong Kong: an intervention study. Lancet
360: 1646-1652.
Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
Heinrich, J.; Hoelscher, B.; Wjst, M.; Ritz, B.; Cyrys, J.; Wichmann, H.-E. (1999) Respiratory diseases and allergies
in two polluted areas in East Germany. Environ. Health Perspect. 107: 53-62.
Heinrich, J.; Hoelscher, B.; Wichmann, H. E. (2000) Decline of ambient air pollution and respiratory symptoms in
children. Am. J. Respir. Crit. Care Med. 161: 1930-1936.
Heinrich, J.; Hoelscher, B.; Frye, C.; Meyer, L; Pitz, M.; Cyrys, J.; Wjst, M.; Neas, L.; Wichmann, H.-E. (2002)
Improved air quality in reunified Germany and decreases in respiratory symptoms. Epidemiology 13: 394-401.
Henderson, B. E.; Gordon, R. J.; Menck, H.; SooHoo, J.; Martin, S. P.; Pike, M. C. (1975) Lung cancer and air
pollution in southcentral Los Angeles County. Am. J. Epidemiol. 101: 477-488.
Hill, A. B. (1965) The environment and disease: association or causation? Proc. R. Soc. Med. 58: 295-300.
Hiltermann, T. J. N.; de Bruijne, C. R.; Stolk, J.; Zwinderman, A. H.; Spieksma, F. Th. M.; Roemer, W.; Steerenberg,
P. A.; Fischer, P. H.; van Bree, L.; Hiemstra, P. S. (1997) Effects of photochemical air pollution and allergen
exposure on upper respiratory tract inflammation in asthmatics. Am. J. Respir. Crit. Care Med.
156: 1765-1772.
Hiltermann, T. J. N.; Stolk, J.; Van derZee, S. C.; Brunekreef, B.; De Bruijne, C. R.; Fischer, P. H.; Ameling, C. B.;
Sterk, P. J.; Hiemstra, P. S.; Van Bree, L. (1998) Asthma severity and susceptibility to air pollution. Eur.
Respir. J. 11:686-693.
Hoek, G. (2003) Daily mortality and air pollution in The Netherlands. In: Revised analyses of time-series studies of
air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 133-142. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Hoek, G.; Schwartz, J. D.; Groot, B.; Eilers, P. (1997) Effects of ambient paniculate matter and ozone on daily
mortality in Rotterdam, the Netherlands. Arch. Environ. Health 52: 455-463.
Hoek, G.; Dockery, D. W.; Pope, A.; Neas, L.; Roemer, W.; Brunekreef, B. (1998) Association between PM10 and
decrements in peak expiratory flow rates in children: reanalysis of data from five panel studies. Eur. Respir. J.
11: 1307-1311.
Hoek, G.; Brunekreef, B.; Verhoeff, A.; van, Wijnen, J.; Fischer, P. (2000) Daily mortality and air pollution in the
Netherlands. J. Air Waste Manage. Assoc. 50: 1380-1389.
Hoek, G.; Brunekreef, B.; Fischer, P.; Van Wijnen, J. (2001) The association between air pollution and heart failure,
arrhythmia, embolism, thrombosis, and other cardiovascular causes of death in a time series study.
Epidemiology 12: 355-357.
8-355
-------
Hoek, G.; Brunekreef, B.; Goldbohm, S.; Fischer, P.; Van den Brandt, P. A. (2002) Association between mortality
and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet 360: 1203-1209.
Hodgkin, J. E.; Abbey, D. E.; Euler, G. L.; Magie, A. R. (1984) COPD prevalence in nonsmokers in high and low
photochemical air pollution areas. Chest 86: 830-838.
Hoeting, J. A.; Madigan, D.; Raftery, A. E.; Volinsky, C. T. (1999) Bayesian model averaging: a tutorial. Stat. Sci.
14: 382-417.
Hong, Y.-C.; Leem, J.-H.; Ha, E.-H.; Christian!, D. C. (1999) PM10 exposure, gaseous pollutants, and daily mortality
in Inchon, South Korea. Environ. Health Perspect. 107: 873-878.
Horak, F., Jr.; Studnicka, M; Gartner, C.; Spengler, J. D.; Tauber, E.; Urbanek, R.; Veiter, A.; Frischer, T. (2002)
Paniculate matter and lung function growth in children: a 3-yr follow-up study in Austrian schoolchildren. Eur.
Respir. J. 19: 838-845.
Howel, D.; Darnell, R.; Pless-Mulloli, T. (2001) Children's respiratory health and daily paniculate levels in
10 nonurban communities. Environ. Res. 87: 1-9.
Ibald-Mulli, A.; Stieber, J.; Wichmann, H.-E.; Koenig, W.; Peters, A. (2001) Effects of air pollution on blood
pressure: a population-based approach. Am. J. Public Health 91: 571-577.
Ilabaca, M.; Olaeta, L; Campos, E.; Villaire, J.; Tellez-Rojo, M. M.; Romieu, I. (1999) Association between levels of
fine paniculate and emergency visits for pneumonia and other respiratory illnesses among children in Santiago,
Chile. J. Air Waste Manage. Assoc. 49: 154-163.
Ito, K. (1990) An examination of the role of aerosol acidity in historical London, England daily mortality
[dissertation]. Syracuse, NY: New York University. Available from: University Microfilms International,
Ann Arbor, MI; AAD91-13012.
Ito, K. (2003) Associations of paniculate matter components with daily mortality and morbidity in Detroit, Michigan.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Ito, K.; Thurston, G. D. (1996) Daily PM10/mortarity associations: an investigation of at-risk subpopulations.
J. Exposure Anal. Environ. Epidemiol. 6: 79-95.
Ito, K.; Thurston, G. D.; Hayes, C.; Lippmann, M. (1993) Associations of London, England, daily mortality with
paniculate matter, sulfur dioxide, and acidic aerosol pollution. Arch. Environ. Health 48: 213-220.
Ito, K.; Kinney, P.; Thurston, G. D. (1995) Variations in PM-10 concentrations within two metropolitan areas and
their implications for health effects analyses. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the
colloquium on paniculate air pollution and human mortality and morbidity, part II; January 1994; Irvine, CA.
Inhalation Toxicol. 7: 735-745.
Ito, K.; Thurston, G. D.; Nadas, A.; Lippmann, M. (2001) Monitor-to-monitor temporal correlation of air pollution
and weather variables in the North-Central U.S. J. Exposure Anal. Environ. Epidemiol. 11: 21-32.
Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
1983-1992. Environ. Health Perspect. 105: 980-985.
Jalaludin, B. B.; Chey, T.; O'Toole, B. I.; Smith, W. T.; Capon, A. G.; Leeder, S. R. (2000) Acute effects of low
levels of ambient ozone on peak expiratory flow rate in a cohort of Australian children. Int. J. Epidemiol.
29: 549-557.
Jamason, P. F.; Kalkstein, L. S.; Gergen, P. J. (1997) A synoptic evaluation of asthma hospital admissions in
New York City. Am. J. Respir. Crit. Care Med. 156: 1781-1788.
Janssen, N. A. H.; Schwartz, J.; Zanobetti, A.; Suh, H. H. (2002) Air conditioning and source-specific particles as
modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ. Health Perspect.
110:43-49.
Jedrychowski, W.; Flak, E. (1998) Separate and combined effects of the outdoor and indoor air quality on chronic
respiratory symptoms adjusted for allergy among preadolescent children. Int. J. Occup. Med. Environ. Health
11: 19-35.
Jedrychowski, W.; Becher, H.; Wahrendorf, J.; Basa-Cierpialek, Z. (1990) A case-control study of lung cancer with
special reference to the effect of air pollution in Poland. J. Epidemiol. Commun. Health 44: 114-120.
Jedrychowski, W.; Flak, E.; Mroz, E. (1999) The adverse effect of low levels of ambient air pollutants on lung
function growth in preadolescent children. Environ. Health Perspect. 107: 669-674.
Joseph, K. S.; Kramer, M. S. (1996) Review of the evidence onfetal and early childhood antecedents of adult chronic
disease. Epidemiol. Rev. 18: 158-174.
Just, J.; Segala, C.; Sahraoui, F.; Priol, G.; Grimfeld, A.; Neukirch, F. (2002) Short-term health effects of paniculate
and photochemical air pollution in asthmatic children. Eur. Respir. J. 20: 899-906.
8-356
-------
Kalandidi, A.; Gratziou, C.; Katsouyanni, K.; Manalis, N.; Tzala, L.; Pantazopoulou, A.; Efthimiou, M.; Roussos, C.;
Roemer, W. (1998) Air pollution and respiratory health of children: the PEACE panel study in Athens, Greece.
Eur. Respir. Rev. 8: 117-124.
Kalkstein, L. S. (1991) A new approach to evaluate the impact of climate on human mortality. Environ. Health
Perspect. 96: 145-150.
Kalkstein, L. S.; Tan, G. (1995) Human health. In: Strzepek, K. M.; Smith, J. B., eds. As climate changes:
international impacts and implications. New York, NY: Cambridge University Press, 124-145.
Kalkstein, L. S.; Barthel, C. D.; Ye, H.; Smoyer, K.; Cheng, S.; Greene, J. S.; Nichols, M. C. (1994) The differential
impacts of weather and pollution on human mortality. Newark, DE: University of Delaware, Department of
Geography, Center for Climatic Research; November.
Katsouyanni, K.; Touloumi, G. (1998) Causes of regional differences in air pollution effects are being studied further
[letter]. Br. Med. J. 316: 1982.
Katsouyanni, K.; Karakatsani, A.; Messari, L; Touloumi, G.; Hatzakis, A.; Kalandidi, A.; Trichopoulos, D. (1990) Air
pollution and cause specific mortality in Athens. J. Epidemiol. Community Health 44: 321-324.
Katsouyanni, K.; Schwartz, J.; Spix, C.; Touloumi, G.; Zmirou, D.; Zanobetti, A.; Wojtyniak, B.; Vonk, J. M.;
Tobias, A.; Ponka, A.; Medina, S.; Bacharova, L.; Andersen, H. R. (1996) Short term effects of air pollution on
health: a European approach using epidemiology time series data: the APHEA protocol. In: St Leger, S., ed.
The APHEA project. Short term effects of air pollution on health: a European approach using epidemiological
time series data. J. Epidemiol. Community Health 50(suppl. 1): S12-S18.
Katsouyanni, K.; Touloumi, G.; Spix, C.; Schwartz, J.; Balducci, F.; Medina, S.; Rossi, G.; Wojtyniak, B.; Sunyer, J.;
Bacharova, L.; Schouten, J. P.; Ponka, A.; Anderson, H. R. (1997) Short term effects of ambient sulphur
dioxide and paniculate matter on mortality in 12 European cities: results from time series data from the
APHEA project. Br. Med. J. 314: 1658-1663.
Katsouyanni, K.; Touloumi, G.; Samoli, E.; Gryparis, A.; Le Tertre, A.; Monopolis, Y.; Rossi, G.; Zmirou, D.;
Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
Schindler, C.; Schwartz, J. (2001) Confounding and effect modification in the short-term effects of ambient
particles on total mortality: results from 29 European cities within the APHEA2 project. Epidemiology
12:521-531.
Katsouyanni, K.; Touloumi, G.; Samoli, E.; Petasakis, Y.; Analitis, A.; Le Tertre, A.; Rossi, G.; Zmirou, D.;
Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
Schindler, C.; Schwartz, J. (2003) Sensitivity analysis of various models of short-term effects of ambient
particles on total mortality in 29 cities in APHEA2. In: Revised analyses of time-series studies of air pollution
and health. Special report. Boston, MA: Health Effects Institute; pp. 157-164. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Keatinge, W. R.; Donaldson, G. C. (2001) Mortality related to cold and air pollution in London after allowance for
effects of associated weather patterns. Environ. Res. A 86: 209-216.
Keles, N.; Ilicali, O. C.; Deger, K. (1999) Impact of air pollution on prevalence of rhinitis in Istanbul. Arch. Environ.
Health 54: 48-51.
Kelsall, J. E.; Samet, J. M.; Zeger, S. L.; Xu, J. (1997) Air pollution and mortality in Philadelphia, 1974-1988. Am. J.
Epidemiol. 146: 750-762.
Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM10 associations in Los Angeles. In:
Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air pollution and human
mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 59-69.
Kinney, P. L.; Aggarwal, M.; Northridge, M. E.; Janssen, N. A. H.; Shepard, P. (2000) Airborne concentrations of
PM2 5 and diesel exhaust particles on Harlem sidewalks: a community-based pilot study. Environ. Health
Perspect. 108: 213-218.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 165-172. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage. Assoc.
50: 1215-1222.
Koenig, J. Q.; Larson, T. V.; Hanley, Q. S.; Rebolledo, V.; Dumler, K.; Checkoway, H.; Wang, S.-Z.; Lin, D.;
Pierson, W. E. (1993) Pulmonary function changes in children associated with fine paniculate matter. Environ.
Res. 63:26-38.
8-357
-------
Kontos, A. S.; Fassois, S. D.; Deli, M. F. (1999) Short-term effects of air pollution on childhood respiratory illness in
Piraeus, Greece, 1987-1992: nonparametric stochastic dynamic analysis. Environ. Res. 81: 275-296.
Korrick, S. A.; Neas, L. M.; Dockery, D. W.; Gold, D. R.; Allen, G. A.; Hill, L. B.; Kimball, K. D.; Rosner, B. A.;
Speizer, F. E. (1998) Effects of ozone and other pollutants on the pulmonary function of adult hikers. Environ.
Health Perspect. 106: 93-99.
Kostianev, S.; Gerginova, M.; Ivanova, M. (1994) Reference values of lung function parameters in Bulgarian girls
aged 7 to 14 years. Pediatria 33: 30-33.
Kotesovec, F.; Skorkovsky, I; Brynda, I; Peters, A.; Heinrich, J. (2000) Daily mortality and air pollution in northern
Bohemia; different effects for men and women. Cent. Eur. J. Public Health 8: 120-127.
Kramer, M. S. (1987) Intrauterine growth and gestational duration determinants. Pediatrics 80: 502-511.
Kramer, U.; Behrendt, H.; Dolgner, R.; Ranft, U.; Ring, J.; Wilier, H.; Schlipkoter, H.-W. (1999) Airway diseases and
allergies in East and West German children during the first 5 years after reunification: time trends and the
impact of sulphur dioxide and total suspended particles. Int. J. Epidemiol. 28: 865-873.
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M.; White,
W. H. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of paniculate
air pollution and mortality. A special report of the Institute's Particle Epidemiology Reanalysis Project.
Cambridge, MA: Health Effects Institute.
Kunst, A. E.; Looman, C. W. N.; Mackenbach, J. P. (1993) Outdoor air temperature and mortality in the Netherlands:
a time-series analysis. Am. J. Epidemiol. 137: 331-341.
Ktinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared to
ecologic studies. Environ. Health Perspect. 105: 1078-1083.
Ktinzli, N.; Ackermann-Liebrich, U.; Brandli, O.; Tschopp, J. M.; Schindler, C.; Leuenberger, P.; SAPALDIA Team.
(2000) Clinically "small" effects of air pollution on FVC have a large public health impact. Eur. Respir. J.
15: 131-136.
Kwon, H.-J.; Cho, S.-H.; Nyberg, F.; Pershagen, G. (2001) Effects of ambient air pollution on daily mortality in a
cohort of patients with congestive heart failure. Epidemiology 12: 413-419.
Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
Lave, L. B.; Seskin, E. P. (1977) Air pollution and human health. Baltimore, MD: The Johns Hopkins University
Press.
Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2002) Short term effects of paniculate air
pollution on cardiovascular diseases in eight European cities. J. Epidemiol. Community Health 56: 773-779.
Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2003) Short-term effects of paniculate air
pollution on cardiovascular diseases in eight European cities. In: Revised analyses of time-series studies of air
pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 173-176. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Lebowitz, M. D.; Collins, L.; Holberg, C. J. (1987) Time series analyses of respiratory responses to indoor and
outdoor environmental phenomena. Environ. Res. 43: 332-341.
Lee, J.-T.; Schwartz, J. (1999) Reanalysis of the effects of air pollution on daily mortality in Seoul, Korea:
a case-crossover design. Environ. Health Perspect. 107: 633-636.
Lee, J.-T.; Shy, C. M. (1999) Respiratory function as measured by peak expiratory flow rate and PM10:
six communities study. J. Exposure Anal. Environ. Epidemiol. 9: 293-299.
Lee, R. E., Jr.; Caldwell, J. S.; Morgan, G. B. (1972) The evaluation of methods for measuring suspended particulates
in air. Atmos. Environ. 6: 593-622.
Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
Perspect. 107: 149-154.
Lee, J.-T.; Kim, H.; Hong, Y.-C.; Kwon, H.-J.; Schwartz, J.; Christiani, D. C. (2000) Air pollution and daily mortality
in seven major cities of Korea, 1991-1997. Environ. Res. 84: 247-254.
Leonardi, G. S.; Houthuijs, D.; Steerenberg, P. A.; Fletcher, T.; Armstrong, B.; Antova, T. (2000) Immune
biomarkers in relation to exposure to paniculate matter: a cross-sectional survey in 17 cities of central Europe.
In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation Toxicol. 12(suppl. 4): 1-14.
Levy, D. (1998) Fine paniculate air pollution and out-of-hospital mortality in King County, Washington. In: Vostal,
J. J., ed. Health effects of paniculate matter in ambient air. Proceedings of an international conference; 1997;
Prague, Czech Republic. Pittsburgh, PA: Air & Waste Management Association; pp. 262-271. (A&WMA
publication VIP-80).
8-358
-------
Levy, J. I.; Hammitt, J. K.; Spengler, J. D. (2000) Estimating the mortality impacts of paniculate matter: what can be
learned frombetween-study variability? Environ. Health Perspect. 108: 109-117.
Levy, D.; Sheppard, L.; Checkoway, H.; Kaufman, J.; Lumley, T.; Koenig, J.; Siscovick, D. (2001) A case-crossover
analysis of paniculate matter air pollution and out-of-hospital primary cardiac arrest. Epidemiology
12: 193-199.
Lewis, P. R.; Hensley, M. J.; Wlodarczyk, J.; Toneguzzi, R. C.; Westley-Wise, V. J.; Dunn, T.; Calvert, D. (1998)
Outdoor air pollution and children's respiratory symptoms in the steel cities of New South Wales. Med. J. Aust.
169: 459-463.
Lewis, S. A.; Corden, J. M.; Forster, G. E.; Newlands, M. (2000) Combined effects of aerobiological pollutants,
chemical pollutants and meteorological conditions on asthma admissions and A & E attendances in Derbyshire
UK, 1993-96. Clin. Exp. Allergy 30: 1724-1732.
Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
Lin, C. A.; Martins, M. A.; Farhat, S. C. L.; Pope, C. A., Ill; Conceicao, G. M. S.; Anastacio, V. M.; Hatanaka, M.;
Andrade, W. C.; Hamaue, W. R.; Bohm, G. M.; Saldiva, P. H. N. (1999) Air pollution and respiratory illness
of children in Sao Paulo, Brazil. Paediatr. Perinat. Epidemiol. 13: 475-488.
Lin, M.; Chen, Y.; Burnett, R. T.; Villeneuve, P. J.; Krewski, D. (2002) The influence of ambient coarse paniculate
matter on asthma hospitalization in children: case-crossover and time-series analyses. Environ. Health
Perspect. 110: 575-581.
Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D.; Gong, H., Jr. (1996)
Short-term air pollution exposures and responses in Los Angeles area schoolchildren. J. Exposure Anal.
Environ. Epidemiol. 6: 449-472.
Linn, W. S.; Gong, H., Jr.; Clark, K. W.; Anderson, K. R. (1999) Day-to-day paniculate exposures and health
changes in Los Angeles area residents with severe lung disease. J. Air Waste Manage. Assoc.
49:PM108-PM115.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lipfert, F. W.; Morris, S. C. (2002) Temporal and spacial relationships between age-specific mortality and ambient
air quality in the United States: preliminary results for counties, 1960-97. Occup. Environ. Med. 59: 156-174.
Lipfert, F. W.; Wyzga, R. E. (1996) The effects of exposure error on environmental epidemiology. In: Phalen, R. F.;
Mannix, R. C.; Tonini, M. C., eds. The second colloquium on paniculate air pollution & human mortality &
morbidity: report to the California Air Resources Board; May; Park City, UT. Sacramento, CA: State of
California, Air Resources Board; pp. 4-295--4-302; ARE contract no. 95-323. (University of California Air
Pollution Health Effects Laboratory report no. 96-02). Available:
http://www.arb.ca.gov/research/abstracts/95-323.htm#Ch95-323 (24 September 2002).
Lipfert, F. W.; Wyzga, R. E. (1997) Air pollution and mortality: the implications of uncertainties in regression
modeling and exposure measurement. J. Air Waste Manage. Assoc. 47: 517-523.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and
size-classified paniculate matter. J. Air Waste Manage. Assoc.: 1501-1513.
Lipfert, F. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) The Washington
University-EPRI veterans' cohort mortality study: preliminary results. In: Grant, L. D., ed. PM2000: paniculate
matter and health. Inhalation Toxicol. 12(suppl. 4): 41-73.
Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000c) Infant mortality and air pollution: a comprehensive analysis of U.S.
data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
Lipfert, F. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2003) Air pollution, blood
pressure, and their long-term associations with mortality. Inhalation Toxicol. 15: 493-512.
Lippmann, M.; Thurston, G. D. (1996) Sulfate concentrations as an indicator of ambient paniculate matter air
pollution for health risk evaluations. J. Exposure Anal. Environ. Epidemiol. 6: 123-146.
Lippmann, M.; Lioy, P. J.; Leikauf, G.; Green, K. B.; Baxter, D.; Morandi, M.; Pasternack, B. S.; Fife, D.; Speizer,
F. E. (1983) Effects of ozone on the pulmonary function of children. In: Lee, S. D.; Mustafa, M. G.; Mehlman,
M. A., eds. International symposium on the biomedical effects of ozone and related photochemical oxidants;
March 1982; Pinehurst, NC. Princeton, NJ: Princeton Scientific Publishers, Inc.; pp. 423-446. (Advances in
modern environmental toxicology: v. 5).
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report no. 95.
Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
California. Environ. Health Perspect. 105: 216-222.
8-359
-------
Long, W.; Tate, R. B.; Neuman, M; Manfreda, J.; Becker, A. B.; Anthonisen, N. R. (1998) Respiratory symptoms in
a susceptible population due to burning of agricultural residue. Chest 113: 351-357.
Loomis, D.; Castillejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
mortality in Mexico City. Epidemiology 10: 118-123.
Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression. J. R.
Stat. Soc. B 61(part 2): 459-477.
Lumley, T.; Sheppard, L. (2000) Assessing seasonal confounding and model selection bias in air pollution
epidemiology using positive and negative control analyses. Environmetrics 11: 705-717.
Mackenbach, J. P.; Looman, C. W. N.; Kunst, A. E. (1993) Air pollution, lagged effects of temperature, and
mortality: The Netherlands 1979-87. J. Epidemiol. Commun. Health 47: 121-126.
Magari, S. R.; Hauser, R.; Schwartz, J.; Williams, P. L.; Smith, T. J.; Christiani, D. C. (2001) Association of heart
rate variability with occupational and environmental exposure to paniculate air pollution. Circulation
104: 986-991.
Magari, S. R.; Schwartz, J.; Williams, P. L.; Hauser, R.; Smith, T. J.; Christiani, D. C. (2002) The association
between personal measurements of environmental exposure to particulates and heart rate variability.
Epidemiology 13: 305-310.
Mage, D.; Wilson, W.; Hasselblad, V.; Grant, L. (1999) Assessment of human exposure to ambient paniculate matter.
J. Air Waste Manage. Assoc. 49: 174-185.
Maisonet, M.; Bush, T. J.; Correa, A.; Jaakkola, J. J. K. (2001) Relation between ambient air pollution and low birth
weight in the northeastern United States. Environ. Health Perspect. Suppl. 109(3): 351-356.
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003) Air pollution and cardiovascular
mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health. Special
report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Marcus, A. H.; Chapman, R. (1998) Estimating the health effects of fine particles from epidemiology studies: how
serious are problems of measurement error, correlation, and confounding? In: Chow, J.; Koutrakis, P., eds.
PM2 5: a fine particle standard. Volume II: proceedings of an international specialty conference.; January; Long
Beach, CA. Pittsburgh, PA: Air & Waste Management Association; pp. 899-919.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern California
children with asthma. Environ. Health Perspect. 107: 757-760.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Islam, T.; Gauderman, W. J.; Avol, E.; Margolis, H. G.;
Peters, J. M. (2002) Asthma in exercising children exposed to ozone: a cohort study. Lancet 359: 386-391.
McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of mortality
with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers. J. Exposure Anal.
Environ. Epidemiol. 10: 427-436.
McGowan, J. A.; Hider, P. N.; Chacko, E.; Town, G. I. (2002) Paniculate air pollution and hospital admissions in
Christchurch, New Zealand. Aust. N. Z. J. Public Health 26: 23-29.
McGregor, G. R.; Walters, S.; Wordley, J. (1999) Daily hospital respiratory admissions and winter air mass types,
Birmingham, UK. Int. J. Biometeorol. 43: 21-30.
Medina, S.; Le Tertre, A.; Quenel, P.; Le Moullec, Y.; Lameloise, P.; Guzzo, J. C.; Festy, B.; Ferry, R.; Dab, W.
(1997) Air pollution and doctors' house calls: results from the ERPURS system for monitoring the effects of air
pollution on public health in greater Paris, France, 1991-1995. Environ. Res. 75: 73-84.
Metzger, K. B.; Tolbert, P. E.; Klein, M.; Peel, J. L.; Flanders, W. D.; Todd, K. H.; Mulholland, J. A.; Ryan, P. B.;
Frumkin, H. (2004) Ambient air pollution and cardiovascular emergency department visits. Epidemiology
15: 46-56.
Michelozzi, P.; Forastiere, F.; Fusco, D.; Perucci, C. A.; Ostro, B.; Ancona, C.; Pallotti, G. (1998) Air pollution and
daily mortality in Rome, Italy. Occup. Environ. Med. 55: 605-610.
Miller, J. P.; Perry, H. M., Jr.; Rossiter, J. E.; Baty, J. D.; Carmody, S. E.; Sambhi, M. P. (1994) Regional differences
in mortality during 15-year follow-up of 11,936 hypertensive veterans. Hypertension 23: 431-438.
Mills, P. K.; Abbey, D.; Beeson, W. L.; Petersen, F. (1991) Ambient air pollution and cancer in California
Seventh-day Adventists. Arch. Environ. Health 46: 271-280.
Monn, C. (2001) Exposure assessment of air pollutants: a review on spatial heterogeneity and
indoor/outdoor/personal exposure to suspended paniculate matter, nitrogen dioxide and ozone. Atmos.
Environ. 35: 1-32.
8-360
-------
Monn, Ch.; Carabias, V.; Junker, M; Waeber, R.; Karrer, M.; Wanner, H. U. (1997) Small-scale spatial variability of
paniculate matter <10 |_im (PM10) and nitrogen dioxide. Atmos. Environ. 31: 2243-2247.
Moolgavkar, S. H. (2000a) Air Pollution and Mortality in Three U.S. Counties. Environ. Health Perspect.
108: 777-784.
Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J. Air Waste Manage Assoc. 50: 1199-1206.
Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
metropolitan areas in the United States. In: Grant, L. D., ed. PM2000: paniculate matter and health. Inhalation
Toxicol. 12(suppl. 4): 75-90.
Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Moolgavkar, S. H.; Luebeck, E. G. (1996) A critical review of the evidence on paniculate air pollution and mortality.
Epidemiology 7: 420-428.
Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
Moolgavkar, S. H.; Hazelton, W.; Luebeck, G.; Levy, D.; Sheppard, L. (2000) Air pollution, pollens, and admissions
for chronic respiratory disease in King County, Washington. In: Inhalation toxicology: proceedings of the third
colloquium on paniculate air pollution and human health; June, 1999; Durham, NC. Inhalation Toxicol.
12(suppl. 1): 157-171.
Morgan, G.; Corbett, S.; Wlodarczyk, J.; Lewis, P. (1998) Air pollution and daily mortality in Sydney, Australia,
1989 through 1993. Am. J. Public Health 88: 759-764.
Morris, R. D. (2001) Airborne particulates and hospital admissions for cardiovascular disease: a quantitative review
of the evidence. Environ. Health Perspect. 109(suppl. 4): 495-500.
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
Mortimer, K. M.; Redline, S.; Kattan, M.; Wright, E. C.; Kercsmar, C. M. (2001) Are peak flow and symptom
measures good predictors of asthma hospitalizations and unscheduled visits? Pediatr. Pulmonol. 31: 190-197.
Mortimer, K. M.; Neas, L. M.; Dockery, D. W.; Redline, S.; Tager, I. B. (2002) The effect of air pollution on
inner-city children with asthma. Eur. Respir. J. 19: 699-705.
Murray, C. J.; Nelson, C. R. (2000) State-space modeling of the relationship between air quality and mortality. J. Air
Waste Manage. Assoc. 50: 1075-1080.
Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
Healthy women's PEF variations with ambient summer concentrations of PM10, PM2 5, SO42", H+, and O3.
Am. J. Respir. Crit. Care Med. 160: 117-125.
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations and
exposure to air pollution. Public Health Rep. 114: 135-148.
Neas, L. M.; Dockery, D. W.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr.; Speizer, F. E. (1994) Concentration of
indoor paniculate matter as a determinant of respiratory health in children. Am. J. Epidemiol. 139: 1088-1099.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
141: 111-122.
Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
Philadelphia. Environ. Health Perspect. 107: 629-631.
Neukirch, F.; Segala, C.; Le Moullec, Y.; Korobaeff, M.; Aubier, M. (1998) Short-term effects of low-level winter
pollution on respiratory health of asthmatic adults. Arch. Environ. Health 53: 320-328.
Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
Perspect. 107: 489-493.
Norris, G.; Larson, T.; Koenig, J.; Claiborn, C.; Sheppard, L.; Finn, D. (2000) Asthma aggravation, combustion, and
stagnant air. Thorax 55: 466-470.
Nyberg, F.; Gustavsson, P.; Ja'rup, L.; Bellander, T.; Berglind, N.; Jakobsson, R.; Pershagen, G. (2000) Urban air
pollution and lung cancer in Stockholm. Epidemiology 11: 487-495.
8-361
-------
Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
70: 98-104.
Ostro, B. D.; Lipsett, M. I; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols. Am. J.
Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
Ostro, B.; Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1998) The impact of fine paniculate matter in Bangkok,
Thailand. In: Chow, J.; Koutrakis, P., eds. PM2.5: a fine particle standard. Volume II: proceedings of an
international specialty conference; January; Long Beach, CA. Pittsburgh, PA: Air & Waste Management
Association; pp. 939-949. (A&WMA publication VIP-81).
Ostro, B. D.; Hurley, S.; Lipsett, M. J. (1999a) Air pollution and daily mortality in the Coachella Valley, California:
a study of PM10 dominated by coarse particles. Environ. Res. 81: 231-238.
Ostro, B. D.; Eskeland, G. S.; Sanchez, J. M.; Feyzioglu, T. (1999b) Air pollution and health effects: a study of
medical visits among children in Santiago, Chile. Environ. Health Perspect. 107: 69-73.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol.: 10: 412-419.
Ostro, B.; Lipsett, M.; Mann, J.; Braxtron-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
African-American children in Los Angeles. Epidemiology 12: 200-208.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coachella Valley, California.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdftl8 October,
2004].
Osunsanya, T.; Prescott, G.; Seaton, A. (2001) Acute respiratory effects of particles: mass or number? Occup.
Environ. Med. 58: 154-159.
Ott, W.; Wallace, L.; Mage, D. (2000) Predicting paniculate (PM10) personal exposure distributions using a random
component superposition statistical model. J. Air Waste Manage. Assoc. 50: 1390-1406.
Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
Environmental Health; March 25.
Pantazopoulou, A.; Katsouyanni, K.; Kourea-Kremastinou, J.; Trichopoulos, D. (1995) Short-term effects of air
pollution on hospital emergency outpatient visits and admissions in the greater Athens, Greece area. Environ.
Res. 69:31-36.
Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles in
urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
Pekkanen, J.; Brunner, E. J.; Anderson, H. R.; Tiittanen, P.; Atkinson, R. W. (2000) Daily concentrations of air
pollution and plasma fibrinogen in London. Occup. Environ. Med. 57: 818-822.
Penttinen, P.; Timonen, K. L.; Tiittanen, P.; Mirme, A.; Ruuskanen, J.; Pekkanen, J. (2001) Number concentration
and size of particles in urban air: effects on spirometric lung function in adult asthmatic subjects. Environ.
Health Perspect. 109: 319-323.
Pereira, L. A. A.; Loomis, D.; Conceicao, G. M. S.; Braga, A. L. F.; Areas, R. M.; Kishi, H. S.; Singer, J. M.; Bohm,
G. M.; Saldiva, P. H. N. (1998) Association between air pollution and intrauterine mortality in Sao Paulo,
Brazil. Environ. Health Perspect. 106: 325-329.
Perry, H. M., Jr.; Schnaper, H. W.; Meyer, G.; Swatzell, R. (1982) Clinical program for screening and treatment of
hypertension in veterans. J. Natl. Med. Assoc. 74: 433-444.
Peters, A.; Goldstein, I. F.; Beyer, U.; Franke, K.; Heinrich, J.; Dockery, D. W.; Spengler, J. D.; Wichmann, H.-E.
(1996) Acute health effects of exposure to high levels of air pollution in eastern Europe. Am. J. Epidemiol.
144: 570-581.
Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997a) Increased plasma viscosity during an air pollution
episode: a link to mortality? Lancet 349: 1582-1587.
Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997b) Respiratory effects are associated with the
number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997c) Short-term effects of paniculate air pollution on
respiratory morbidity in asthmatic children. Eur. Respir. J. 10: 872-879.
8-362
-------
Peters, A.; Kotesovec, F.; Skorkovsky, J.; Brynda, J.; Heinrich, J. (1999a) Akute Auswirkung der
Schwebstaubkonzentrationen in der Aulknluft auf die Mortalitat - Vergleichsstudie Nordost-Bayern /
Nordbohmen. AbschluBbericht [Acute effects of suspended particle concentrations in the atmosphere on
mortality - a study comparing northeast Bavaria and north Bohemia. Final report]. Bavaria, Federal Republic
of Germany: InstitutfurEpidemiologie; report no. GSF-EP S 1/99.
Peters, J. M; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
Med. 159: 760-767.
Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999c) A study of twelve southern California communities with
differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
159: 768-775.
Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
Monahan, K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology
11: 11-17.
Peters, A.; Skorkovsky, J.; Kotesovec, F.; Brynda, J.; Spix, C.; Wichmann, H. E.; Heinrich, J. (2000b) Associations
between mortality and air pollution in central Europe. Environ. Health Perspect. 108: 283-287.
Peters, A.; Dockery, D. W.; Muller, J. E.; Mittleman, M. A. (2001a) Increased paniculate air pollution and the
triggering of myocardial infarction. Circulation 103: 2810-2815.
Peters, A.; Frohlich, M.; Doring, A.; Immervoll, T.; Wichmann, H.-E.; Hutchinson, W. L.; Pepys, M. B.; Koenig, W.
(200 Ib) Paniculate air pollution is associated with an acute phase response in men: results from the
MONICA-Augsburg Study. Eur. Heart J. 22: 1198-1204.
Petroeschevsky, A.; Simpson, R. W.; Thalib, L.; Rutherford, S. (2001) Associations between outdoor air pollution
and hospital admissions in Brisbane, Australia. Arch. Environ. Health 56: 37-52.
Pike, M. C.; Jing, J. S.; Rosario, I. P.; Henderson, B. E.; Menck, H. R. (1979) Occupation: "explanation" of an
apparent air pollution related localized excess of lung cancer in Los Angeles County. In: Breslow, N. E.;
Whitemore, A. S., eds. Energy and Health: proceedings of a conference sponsored by Siam Institute for
Mathematics and Society; 1978; Aha, Utah. [SIAM-SDVIS conference series 6].
Pless-Mulloli, T.; Howel, D.; King, A.; Stone, I.; Merefield, J.; Bessell, J.; Darnell, R. (2000) Living near opencast
coal mining sites and children's respiratory health. Occup. Environ. Med. 57: 145-151.
Ponce de Leon, A.; Anderson, H. R.; Bland, J. M.; Strachan, D. P.; Bower, J. (1996) Effects of air pollution on daily
hospital admissions for respiratory disease in London between 1987-88 and 1991-92. In: St Leger, S., ed. The
APHEA project. Short term effects of air pollution on health: a European approach using epidemiological time
series data. J. Epidemiol. Community Health 50(suppl. 1): S63-S70.
Ponka, A.; Savela, M.; Virtanen, M. (1998) Mortality and air pollution in Helsinki. Arch. Environ. Health
53:281-286.
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and Cache
Valleys. Arch. Environ. Health 46: 90-97.
Pope, C. A., Ill; Kalkstein, L S. (1996) Synoptic weather modeling and estimates of the exposure-response
relationship between daily mortality and paniculate air pollution. Environ. Health Perspect. 104: 414-420.
Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM10 pollution:
a daily time series analysis. Am. Rev. Respir. Dis. 144: 668-674.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah valley.
Arch. Environ. Health 47: 211-217.
Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J. Respir.
Crit. Care Med. 151: 669-674.
Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
Front. Environ. Health Perspect. 107: 567-573.
Pope, C. A., Ill; Dockery, D. W.; Kanner, R. E.; Vollegas, G. M.; Schwartz, J. (1999b) Oxygen saturation, pulse rate,
and paniculate air pollution: a daily time-series panel study. Am. J. Respir. Crit. Care Med. 159: 365-372.
Pope, C. A., Ill; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
G. M.; Gold, D. R.; Dockery, D. W. (1999c) Heart rate variability associated with paniculate air pollution.
Am. Heart J. 138:890-899.
8-363
-------
Pope, C. A., Ill; Burnett, R. T.; Thun, M. I; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med. Assoc.
287: 1132-1141.
Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
Prescott, G. J.; Lee, R. J.; Cohen, G. R.; Elton, R. A.; Lee, A. J.; Fowkes, F. G. R.; Agius, R. M. (2000) Investigation
of factors which might indicate susceptibility to paniculate air pollution. Occup. Environ. Med. 57: 53-57.
Qian, Z. Chapman, R. S.; Tian, Q.; Chen, Y.; Lioy, P. J.; Zhang, J. (2000) Effects of air pollution on children's
respiratory health in three Chinese cities. Arch. Environ. Health 55: 126-133.
Rahlenbeck, S. I.; Kahl, H. (1996) Air pollution and mortality in East Berlin during the winters of 1981-1989. Int. J.
Epidemiol. 25: 1220-1226.
Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
Perspect. 104: 506-514.
Ramlow, J. M.; Kuller, L. H. (1990) Effects of the summer heat wave of 1988 on daily mortality in Allegheny
County, PA. Public Health Rep. 105: 283-289.
Ritz, B.; Yu, F.; Chapa, G.; Fruin, S. (2000) Effect of air pollution on preterm birth among children born in Southern
California between 1989 and 1993. Epidemiology 11: 502-511.
Roemer, W. H.; Van Wijnen, J. H . (2001) Daily mortality and air pollution along busy streets in Amsterdam,
1987-1998. Epidemiology 12: 649-653.
Roemer, W.; Hoek, G.; Brunekreef, B. (1993) Effect of ambient winter air pollution on respiratory health of children
with chronic respiratory symptoms. Am. Rev. Respir. Dis. 147: 118-124.
Roemer, W.; Hoek, G.; Brunekreef, B.; Haluszka, J.; Kalandidi, A.; Pekkanen, J. (1998) Daily variations in air
pollution and respiratory health in a multicentre study: the PEACE project. Eur. Respir. J. 12: 1354-1361.
Roemer, W.; Hoek, G.; Brunekreef, B.; Clench-Aas, J.; Forsberg, B.; Pekkanen, J.; Schutz, A. (2000) PM10 elemental
composition and acute respiratory health effects in European children (PEACE project). Eur. Respir. J.
15: 553-559.
Rogers, J. F.; Thompson, S. J.; Addy, C. L.; McKeown, R. E.; Cowen, D. J.; Decoufle, P. (2000) Association of very
low birth weight with exposures to environmental sulfur dioxide and total suspended particulates. Am. J.
Epidemiol. 151:602-613.
Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution on
the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
154: 300-307.
Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
Mexico City. Arch. Environ. Health 52: 368-376.
Rooney, C.; McMichael, A. J.; Kovats, R. S.; Coleman, M. P. (1998) Excess mortality in England and Wales, and in
greater London, during the 1995 heatwave. J. Epidemiol. Community Health 52: 482-486.
Rosas, I.; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, J.; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis of
the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
emergency admissions to a hospital in Mexico City. Allergy (Copenhagen) 53: 394-401.
Rossi, G.; Vigotti, M. A.; Zanobetti, A.; Repetto, F.; Gianelle, V.; and Schwartz, J. (1999) Air pollution and
cause-specific mortality in Milan, Italy, 1980-1989. Arch. Environ. Health 54: 158-164.
Rothman, K. J.; Greenland, S., eds. (1998) Modern epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven
Publishers.
Rutherford, S.; Clark, E.; McTainsh, G.; Simpson, R.; Mitchell, C. (1999) Characteristics of rural dust events shown
to impact on asthma severity in Brisbane, Australia. Int. J. Biometeorol. 42: 217-225.
Samet, J. M.; Zeger, S. L.; Berhane, K. (1995) The association of mortality and paniculate air pollution. In:
Paniculate air pollution and daily mortality: replication and validation of selected studies, the phase I report of
the particle epidemiology evaluation project [preprint]. Cambridge, MA: Health Effects Institute; pp. 1-104.
Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1996) Weather, air pollution and mortality in
Philadelphia, 1973-1980, report to the Health Effects Institute on phase IB, Particle Epidemiology Evaluation
Project. Cambridge, MA: Health Effects Institute; review draft.
Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
(2000a) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94.
8-364
-------
Samet, J. M; Dominici, F.; Zeger, S. L.; Schwartz, J.; Dockery, D. W. (2000b) National morbidity, mortality, and air
pollution study. Part I: methods and methodologic issues. Cambridge, MA: Health Effects Institute; research
report no. 94.
Samet, J. M.; Dominici, F.; Curriero, F. C.; Coursac, L; Zeger, S. L. (2000c) Fine paniculate air pollution and
mortality in20 U.S. cities, 1987-1994. N. Engl. J. Med. 343: 1742-1749.
Samoli, E.; Schwartz, J.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Balducci, F.; Medina, S.; Rossi, G.; Sunyer, J.;
Bacharova, L.; Anderson, H. R.; Katsouyanni, K. (2001) Investigating regional differences in short-term
effects of air pollution on daily mortality in the APHEA project: a sensitivity analysis for controlling long-term
trends and seasonality. Environ. Health Perspect. 109: 349-353.
Samoli, E.; Schwartz, J.; Analitis, A.; Petasakis, Y.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Balducci, F.; Medina, S.;
Rossi, G.; Sunyer, J.; Anderson, H. R.; Katsouyanni, K. (2003) Sensitivity analyses of regional differences in
short-term effects of air pollution on daily mortality in APHEA cities. In: Revised analyses of time-series
studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 205-210.
Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
Sarnat, J. A.; Schwartz, J.; Catalano, P. J.; Suh, H. H. (2001) Gaseous pollutants in paniculate matter epidemiology:
confounders or surrogates? Environ. Health Perspect. 109: 1053-1061.
Scarlett, J. F.; Abbott, K. J.; Peacock, J. L.; Strachan, D. P.; Anderson, H. R. (1996) Acute effects of summer air
pollution on respiratory function in primary school children in southern England. Thorax 51:1109-1114.
Schouten, J. P.; Vonk, J. M.; de Graaf, A. (1996) Short term effects of air pollution on emergency hospital admissions
for respiratory disease: results of the APHEA project in two major cities in The Netherlands, 1977-89. In:
St Leger, S., ed. The APHEA project. Short term effects of air pollution on health: a European approach using
epidemiological time series data. J. Epidemiol. Community Health 50(suppl. 1): S22-S29.
Schwartz, J. (1991) Paniculate air pollution and daily mortality in Detroit. Environ. Res. 56: 204-213.
Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137: 1136-1147.
Schwartz, J. (1994a) PM10, ozone, and hospital admissions for the elderly in Minneapolis, MN. Arch. Environ. Health
49: 366-374.
Schwartz, J. (1994b) Air pollution and hospital admissions for the elderly in Detroit, Michigan. Am. J. Respir. Crit.
Care Med. 150: 648-655.
Schwartz, J. (1995) Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory
disease. Thorax 50: 531-538.
Schwartz, J. (1997a) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
8: 371-377.
Schwartz, J. (1997b) Health effects of air pollution from traffic: ozone and paniculate matter. In: Fletcher, T.;
McMichael, A. J., eds. Health at the crossroads: transport policy and urban health. [Proceedings of the]
London School of Hygiene & Tropical Medicine fifth annual public health forum; April, 1995; London,
United Kingdom. Chichester, United Kingdom: John Wiley & Sons Ltd.; pp. 61-85.
Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
10: 17-22.
Schwartz, J. (2000a) Assessing confounding, effect modification, and thresholds in the association between ambient
particles and daily deaths. Environ. Health Perspect. 108: 563-568.
Schwartz, J. (2000b) The distributed lag between air pollution and daily deaths. Epidemiology 11: 320-326.
Schwartz, J. (2000c) Harvesting and long term exposure effects in the relation between air pollution and mortality.
Am. J. Epidemiol. 151: 440-448.
Schwartz, J. (2000d) Daily deaths are associated with combustion particles rather than SO2 in Philadelphia.
Occup. Environ. Med. 57: 692-697.
Schwartz, J. (2001) Air pollution and blood markers of cardiovascular risk. Environ. Health Perspect.
Suppl. 109(3): 405-409.
Schwartz, J. (2003a) Daily deaths associated with air pollution in six US cities and short-term mortality displacement
in Boston. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
Health Effects Institute; pp. 219-226. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Schwartz, J. (2003b) Airborne particles and daily deaths in 10 US cities. In: Revised analyses of time-series studies of
air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 211-218. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
8-365
-------
Schwartz, I; Dockery, D. W. (1992) Increased mortality in Philadelphia associated with daily air pollution
concentrations. Am. Rev. Respir. Dis. 145: 600-604.
Schwartz, J.; Marcus, A. (1990) Mortality and air pollution in London: a time series analysis. Am. J. Epidemiol.
131: 185-194.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan.
Am. J. Epidemiol. 142: 23-35.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology. 11: 6-10.
Schwartz, J.; Zanobetti, A. (2000) Using meta-smoothing to estimate dose-response trends across multiple studies,
with application to air pollution and daily death. Epidemiology 11: 666-672.
Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
Ferris, B. G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
Am. J. Respir. Crit. Care Med. 150: 1234-1242.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J. Air
Waste Manage. Assoc. 46: 927-939.
Schwartz, J.; Spix, C.; Touloumi, G.; Bacharova, L.; Barumamdzadeh, T.; le Tertre, A.; Piekarksi, T.; Ponce de Leon,
A.; Ponka, A.; Rossi, G.; Saez, M.; Schouten, J. P. (1996b) Methodological issues in studies of air pollution
and daily counts of deaths or hospital admissions. In: St Leger, S., ed. The APHEA project. Short term effects
of air pollution on health: a European approach using epidemiological time series data. J. Epidemiol.
Community Health 50(suppl. 1): S3-S11.
Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
Schwartz, J.; Zanobetti, A.; Bateson, T. (2003) Morbidity and mortality among elderly residents of cities with daily
PM measurements. In: Revised analyses of time-series studies of air pollution and health. Special report.
Boston, MA: Health Effects Institute; pp. 25-58. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Seaton, A.; Soutar, A.; Crawford, V.; Elton, R.; McNerlan, S.; Cherrie, J.; Watt, M.; Agius, R.; Stout, R. (1999)
Paniculate air pollution and the blood. Thorax 54: 1027-1032.
Segala, C.; Fauroux, B.; Just, J.; Pascual, L.; Grimfeld, A.; Neukirch, F. (1998) Short-term effect of winter air
pollution on respiratory health of asthmatic children in Paris. Eur. Respir. J. 11: 677-685.
Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18
October, 2004].
Sheppard, L.; Damian, D. (2000) Estimating short-term PM effects accounting for surrogate exposure measurements
from ambient monitors. Environmetrics 11: 675-687.
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on nonelderly
asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Sheppard, L.; Levy, D.; Checkoway, H. (2001) Correcting for the effects of location and atmospheric conditions on
air pollution exposures in a case-crossover study. J. Exposure Anal. Environ. Epidemiol. 11: 86-96.
Shima, M.; Nitta, Y.; Ando, M.; Adachi, M. (2002) Effects of air pollution on the prevalence and incidence of asthma
in children. Arch. Environ. Health 57: 529-535.
Shumway, R. H.; Tai, R. Y.; Tai, L. P.; Pawitan, Y. (1983) Statistical analysis of daily London mortality and
associated weather and pollution effects. Sacramento, CA: California Air Resources Board; contract no.
Al-154-33.
Shumway, R. H.; Azari, A. S.; Pawitan, Y. (1988) Modeling mortality fluctuations in Los Angeles as functions of
pollution and weather effects. Environ. Res. 45: 224-241.
Simpson, R. W.; Williams, G.; Petroeschevsky, A.; Morgan, G.; Rutherford, S. (1997) Associations between outdoor
air pollution and daily mortality in Brisbane, Australia. Arch. Environ. Health 52: 442-454.
Smith, R. L. (2003) Invited commentary: timescale-dependent mortality effects of air pollution. Am. J. Epidemiol.
157: 1066-1070.
Smith, M. A.; Jalaludin, B.; Byles, J. E.; Lim, L.; Leeder, S. R. (1996) Asthma presentations to emergency
departments in western Sydney during the January 2194 bushfires. Int. J. Epidemiol. 25: 1227-1236.
Smith, R. L.; Davis, J. M.; Speckman, P. (1999) Human Health Effects of Environmental Pollution in the
Atmosphere. London, England. New York, NY: Statistics for the Environment 4: Pollution Assessment and
Control. Edited by V. Barrett, A, Stein and K.F. Turkman. Wiley, pp. 91-115.
8-366
-------
Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and coarse
particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
Smoyer, K. E.; Rainham, D. G. C.; Hewko, J. N. (2000a) Heat-stress-related mortality in five cities in southern
Ontario: 1980-1996. Int. J. Biometeorol. 44: 190-197.
Smoyer, K. E.; Kalkstein, L. S.; Greene, J. S.; Ye, H. (2000b) The impacts of weather and pollution on human
mortality in Birmingham, Alabama and Philadelphia, Pennsylvania. Int. J. Climatol. 20: 881-897.
Spinillo, A.; Capuzzo, E.; Egbe, T. O.; Fazzi, E.; Colonna, L.; Nicola, S. (1995) Pregnancies complicated by
idiopathic intrauterine growth retardation: severity of growth failure, neonatal morbidity and two-year infant
neurodevelopmental outcome. J. Reprod. Med. 40: 209-215.
Spix, C.; Heinrich, J.; Dockery, D.; Schwartz, J.; Volksch, G.; Schwinkowski, K.; Collen, C.; Wichmann, H. E.
(1993) Air pollution and daily mortality in Erfurt, East Germany, 1980-1989. Environ. Health Perspect.
101: 518-526.
Spix, C.; Anderson, H. R.; Schwartz, J.; Vigotti, M. A.; LeTertre, A.; Vonk, J. M.; Touloumi, G.; Balducci, F.;
Piekarski, T.; Bacharova, L.; Tobias, A.; Ponka, A.; Katsouyanni, K. (1998) Short-term effects of air pollution
on hospital admissions of respiratory diseases in Europe: a quantitative summary of APHEA study results.
Arch. Environ. Health 53: 54-64.
Stieb, D. M.; Burnett, R. T.; Beveridge, R. C.; Brook, J. R. (1996) Association between ozone and asthma emergency
department visits in Saint John, New Brunswick, Canada. Environ. Health Perspect. 104: 1354-1360.
Stieb, D. M.; Beveridge, R. C.; Rowe, B. H.; Walter, S. D.; Judek, S. (1998a) Assessing diagnostic classification in an
emergency department: implications for daily time series studies of air pollution. Am. J. Epidemiol.
148: 666-670.
Stieb, D. M.; Brook, J. R.; Broder, L; Judek, S.; Burnett, R. T.; Beveridge, R. C. (1998b) Personal exposure of adults
with cardiorespiratory disease to paniculate acid and sulfate in Saint John, New Brunswick, Canada. Appl.
Occup. Environ. Hyg. 13: 461-468.
Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
Stolzel, M.; Peters, A.; Wichmann, H.-E. (2003) Daily mortality and fine and ultrafine particles in Erfurt, Germany.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 231-240. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October,
2004].
Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne paniculate matter on daily
death counts. Environ. Health Perspect. 103: 490-497.
Sunyer, J.; Basagana, X. (2001) Particles, and not gases, are associated with the risk of death in patients with chronic
obstructive pulmonary disease. Int. J. Epidemiol. 30: 1138-1140.
Sunyer, J.; Spix, C.; Quenel, P.; Ponce-de-Leon, A.; Ponka, A.; Barumandzadeh, T.; Touloumi, G.; Bacharova, L.;
Wojtyniak, B.; Vonk, J.; Bisanti, L.; Schwartz, J.; Katsouyanni, K. (1997) Urban air pollution and emergency
admissions for asthma in four European cities: the APHEA project. Thorax 52: 760-765.
Sunyer, J.; Schwartz, J.; Tobias, A.; Macfarlane, D.; Garcia, J.; Anto, J. M. (2000) Patients with chronic obstructive
pulmonary disease are at increased risk of death associated with urban particle air pollution: a case-crossover
analysis. Am. J. Epidemiol. 151: 50-56.
Taggart, S. C. O.; Custovic, A.; Francis, H. C.; Faragher, E. B.; Yates, C. J.; Higgins, B. G.; Woodcock, A. (1996)
Asthmatic bronchial hyperresponsiveness varies with ambient levels of summertime air pollution. Eur. Respir.
J. 9: 1146-1154.
Tan, W. C.; Qiu, D.; Liam, B. L.; Ng, T. P.; Lee, S. H.; Van Eeden, S. F.; D'Yachkova, Y.; Hogg, J. C. (2000) The
human bone marrow response to acute air pollution caused by forest fires. Am. J. Respir. Crit. Care Med.
161: 1213-1217.
Tanaka, H.; Honma, S.; Nishi, M.; Igarashi, T.; Teramoto, S.; Nishio, F.; Abe, S. (1998) Acid fog and hospital visits
for asthma: an epidemiological study. Eur. Respir. J. 11: 1301-1306.
Tellez-Rojo, M. M.; Romieu, I.; Ruiz-Velasco, S.; Lezana, M.-A.; Hernandez-Avila, M. M. (2000) Daily respiratory
mortality andPMlO pollution in Mexico City: importance of considering place of death. Eur. Respir. J.
16: 391-396.
Tenias, J. M.; Ballester, F.; Rivera, M. L. (1998) Association between hospital emergency visits for asthma and air
pollution in Valencia, Spain. Occup. Environ. Med. 55: 541-547.
Thompson, S. G.; Kienast, J.; Pyke, S. D. M.; Haverkate, F.; Van De Loo, J. C. W. (1995) Hemostatic factors and the
risk of myocardial infarction or sudden death in patients with angina pectoris. N. Engl. J. Med. 332: 635-641.
8-367
-------
Thompson, A. I; Shields, M. D.; Patterson, C. C. (2001) Acute asthma exacerbations and air pollutants in children
living in Belfast, Northern Ireland. Arch. Environ. Health 56: 234-241.
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
65: 271-290.
Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
Timonen, K. L.; Pekkanen, J. (1997) Air pollution and respiratory health among children with asthmatic or cough
symptoms. Am. J. Respir. Crit. Care Med. 156: 546-552.
Tobias, A.; Campbell, M. J. (1999) Modelling influenza epidemics in the relation between black smoke and total
mortality. A sensitivity analysis. J. Epidemiol. Community Health 53: 583-584.
Tolbert, P. E.; Klein, M.; Metzger, K. B.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.; Frumkin, H.
(2000a) Interim results of the study of particulates and health in Atlanta (SOPHIA). J. Exposure Anal. Environ.
Epidemiol. 10: 446-460.
Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000b) Air quality and
pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
Touloumi, G.; Katsouyanni, K.; Zmirou, D.; Schwartz, J.; Spix, C.; Ponce de Leon, A.; Tobias, A.; Quennel, P.;
Rabczenko, D.; Bacharova, L.; Bisanti, L.; Vonk, J. M.; Ponka, A. (1997) Short-term effects of ambient
oxidant exposure on mortality: a combined analysis within the APHEA project. Am. J. Epidemiol.
146: 177-185.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (1999) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Berkeley, CA: Lawrence Berkeley National Laboratory,
Environmental Energy Technologies Division; report no. LBNL-43583.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
Turnovska, T.; Kostianev, S. (1999) Effects of reduced air pollution on children's pulmonary function. Cent. Eur. J.
Public Health 7: 77-79.
U.S. Census Bureau. (1995) American housing survey for the United States in 1993. Washington, DC: U.S.
Department of Commerce; current housing reports H150/93. Available:
http://www.census.gov/hhes/www/housing/ahs/nationaldata.html [29 April, 2002].
U.S. Census Bureau. (2003) Internal migration of the older population: 1995 to 2000. Washington, DC: U.S.
Department of Commerce. Available: http://www.census.gov/prod/2003pubs/censr-10.pdfp6 April, 2004].
U.S. Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
U.S. Environmental Protection Agency. (1986) Second addendum to air quality criteria for particulate matter and
sulfur oxides (1982): assessment of newly available health effects information. Research Triangle Park, NC:
Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report
no. EPA-600/8-86-020F. Available from: NTIS, Springfield, VA; PB87-176574.
U.S. Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park,
NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525, and
PB95-124517.
U.S. Environmental Protection Agency. (1996a) Air quality criteria for particulate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
U.S. Environmental Protection Agency. (1996b) Air quality criteria for ozone and related photochemical oxidants.
Research Triangle Park, NC: Office of Research and Development; report nos. EP A/600/AP-93/004aF-cF. 3v.
Available from: NTIS, Springfield, VA; PB96-185582, PB96-185590, andPB96-185608. Available online at:
www.epa.gov/ncea/ozone.htm.
U.S. Environmental Protection Agency. (2000a) Emissions and air quality data. Washington, DC: Technology
Transfer Network. Available: http://www.epa.gov/ttn/naaqs/ozone/areas/ [20 May, 2003].
U.S. Environmental Protection Agency. (2000b) Air quality criteria for carbon monoxide. Research Triangle Park,
NC: National Center for Environmental Assessment; report no. EPA/600/P-99/00 IF. Available:
www.epa.gov/ncea/co/ [2000, October 6].
8-368
-------
U.S. Environmental Protection Agency. (2002) Health assessment document for diesel engine exhaust. Washington,
DC: Office of Research and Development, National Center for Environmental Assessment; report no.
EPA/600/8-90/057F. Available: http://cfpub.epa.gov/ncea/ [22 May, 2003].
Van Der Zee, S. C.; Hoek, G.; Boezen, H. M; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (1999) Acute effects
of urban air pollution on respiratory health of children with and without chronic respiratory symptoms. Occup.
Environ. Med. 56: 802-813.
Van Der Zee, S. C.; Hoek, G.; Boezen, M. H.; Schouten, J. P.; Van Wijnen, J. H.; Brunekreef, B. (2000) Acute effects
of air pollution on respiratory health of 50-70 yr old adults. Eur. Respir. J. 15: 700-709.
Vasan, R. S.; Larson, M. G.; Leip, E. P.; Kannel, W. B.; Levy, D. (2001) Assessment of frequency of progression to
hypertension in non-hypertensive participants in the Framingham Heart Study: a cohort study. Lancet
358: 1682-1686.
Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
Vena, J. E. (1982) Air pollution as a risk factor in lung cancer. Am. J. Epidemiol. 116: 42-56.
Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1967) Effects of treatment on
morbidity in hypertension. Results in patients with diastolic blood pressures averaging 115 through 129 mm
Hg. JAMA J. Am. Med. Assoc. 202: 1028-1034.
Veterans Administration Cooperative Study Group on Antihypertensive Agents. (1970) Effects of treatment on
morbidity in hypertension. II. Results in patients with diastolic blood pressure averaging 90 through 114 mm
Hg. JAMA J. Am. Med. Assoc. 213: 1143-1152.
Vigotti, M. A.; Rossi, G.; Bisanti, L.; Zanobetti, A.; Schwartz, J. (1996) Short term effects of urban air pollution on
respiratory health in Milan, Italy, 1980-89. In: Leger, S., ed. The APHEA project. Short term effects of air
pollution on health: a European approach using epidemiological time series data. J. Epidemiol. Community
Health 50(suppl. 1): S71-S75.
Villeneuve, P. J.; Goldberg, M. S.; Krewski, D.; Burnett, R. T.; Chen, Y. (2002) Fine paniculate air pollution and
all-cause mortality within the Harvard six-cities study: variations in risk by period of exposure. Ann.
Epidemiol. 12: 568-576.
VonKlot, S.; Wolke, G.; Tuch, T.; Heinrich, J.; Dockery, D. W.; Schwartz, J.; Kreyling, W. G.; Wichmann, H. E.;
Peters, A. (2002) Increased asthma medication use in association with ambient fine and ultrafine particles. Eur.
Respir. J. 20: 691-702.
Wallace, L. (2000) Real-time monitoring of particles, PAH, and CO in an occupied townhouse. Appl. Occup.
Environ. Hyg. 15: 39-47.
Wang, B.; Peng, Z.; Zhang, X.; Xu, Y.; Wang, H.; Allen, G.; Wang, L.; Xu, X. (1999) Paniculate matter, sulfur
dioxide, and pulmonary function in never-smoking adults in Chongqing, China. Int. J. Occup. Environ. Health
5: 14-19.
Ward, D. J.; Miller, M. R.; Walters, S.; Harrison, R. M.; Ayres, J. G. (2000) Impact of correcting peak flow for
nonlinear errors on air pollutant effect estimates from a panel study. Eur. Respir. J. 15: 137-140.
Ware, J. H.; Ferris, B. G., Jr.; Dockery, D. W.; Spengler, J. D.; Stram, D. O.; Speizer, F. E. (1986) Effects of ambient
sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev. Respir. Dis.
133: 834-842.
Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W. G.; Heyder, J. (2000) Daily
mortality and fine and ultrafine particles in Erfurt, Germany. Part I: role of particle number and particle mass.
Cambridge, MA: Health Effects Institute; Research Report no. 98.
Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
Wilson, W. E.; Mage, D. T.; Grant, L. D. (2000) Estimating separately personal exposure to ambient and nonambient
paniculate matter for epidemiology and risk assessment: why and how. J. Air Waste Manage. Assoc.
50: 1167-1183.
Wjst, M.; Reitmeir, P.; Dold, S.; Wulff, A.; Nicolai, T.; Von Loeffelholz-Colberg, E. F.; Von Mutius, E. (1993) Road
traffic and adverse effects on respiratory health in children. Br. Med. J. 307: 596-600.
Wolff, G. T.; Stroup, C. M.; Stroup, D. P. (1983) The coefficient of haze as a measure of paniculate elemental
carbon. J. AirPollut. Control Assoc. 33: 746-750.
Wong, T. W.; Lau, T. S.; Yu, T. S.; Neller, A.; Wong, S. L.; Tarn, W.; Pang, S. W. (1999a) Air pollution and hospital
admissions for respiratory and cardiovascular diseases in Hong Kong. Occup. Environ. Med. 56: 679-683.
Wong, C. M.; Hu, Z. G.; Lam, T. H.; Hedley, A. J.; Peters, J. (1999b) Effects of ambient air pollution and
environmental tobacco smoke on respiratory health of non-smoking women in Hong Kong. Int. J. Epidemiol.
28: 859-864.
8-369
-------
Woodhouse, P. R.; Khaw, K. T.; Plummer, M.; Foley, A.; Meade, T. W. (1994) Seasonal variations of plasma
fibrinogen and factor VII activity in the elderly: winter infections and death from cardiovascular disease.
Lancet 343: 435-439.
Woodruff, T. I; Grille, I; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal infant
mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
Wordley, I; Walters, S.; Ayres, J. G. (1997) Short term variations in hospital admissions and mortality and
paniculate air pollution. Occup. Environ. Med. 54: 108-116.
World Health Organization. (1996) Climate change and human health. Geneva, Switzerland: United Nations
Environment Programme.
Wyzga, R. E.; Lipfert, F. W. (1996) Ozone and daily mortality: the ramifications of uncertainties and interactions and
some initial regression. In: Vostal, J. J., ed. Tropospheric ozone: critical issues in the regulatory process,
proceedings of a specialty conference sponsored by the Air & Waste Management Association; May 1994;
Orlando, FL. Pittsburgh, PA: Air & Waste Management Association; pp. 453-487. (A&WMA publication no.
VIP-54).
Xu, Z.; Yu, D.; Jing, L.; Xu, X. (2000) Air pollution and daily mortality in Shenyang, China. Arch. Environ. Health
55: 115-120.
Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
Inhalation Toxicol. 9: 15-29.
Ye, F.; Piver, W. T.; Ando, M.; Portier, C. J. (2001) Effects of temperature and air pollutants on cardiovascular and
respiratory diseases for males and females older than 65 years of age in Tokyo, July and August 1980-1995.
Environ. Health Perspect. 109: 355-359.
Yu, O.; Sheppard, L.; Lumley, T.; Koenig, J. Q.; Shapiro, G. G. (2000) Effects of ambient air pollution on symptoms
of asthma in Seattle-area children enrolled in the CAMP study. Environ. Health Perspect. 108: 1209-1214.
Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality.
J. Occup. Environ. Med. 42: 469-474.
Zanobetti, A.; Schwartz, J. (2001) Are diabetics more susceptible to the health effects of airborne particles? Am. J.
Respir. Crit. Care Med. 164: 831-833.
Zanobetti, A.; Schwartz, J. (2003a) Airborne particles and hospital admissions for heart and lung disease. In: Revised
analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 241-248. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Zanobetti, A.; Schwartz, J. (2003b) Multicity assessment of mortality displacement within the APHEA2 project. In:
Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
Institute; pp. 249-254. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdftl8 October, 2004].
Zanobetti, A.; Schwartz, J.; Dockery, D. W. (2000a) Airborne particles are a risk factor for hospital admissions for
heart and lung disease. Environ. Health Perspect. 108: 1071-1077.
Zanobetti, A.; Wand, M. P.; Schwartz, J.; Ryan, L. M. (2000b) Generalized additive distributed lag models:
quantifying mortality displacement. Biostatistics 1: 279-292.
Zanobetti, A.; Schwartz, J.; Samoli, E.; Gryparis, A.; Touloumi, G.; Peacock, J.; Anderson, R. H.; Le Tertre, A.;
Bobros, J.; Celko, M.; Goren, A.; Forsberg, B.; Michelozzi, P.; Rabczenko, D.; Hoyos, S. P.; Wichmann,
H. E.; Katsouyanni, K. (2003) The temporal pattern of respiratory and heart disease mortality in response to air
pollution. Environ. Health Perspect. Ill: 1188-1193.
Zeger, S. L.; Dominici, F.; Samet, J. (1999) Harvesting-resistant estimates of air pollution effects on mortality.
Epidemiology 10: 171-175.
Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
Perspect. 108: 419-426.
Zeghnoun, A.; Beaudeau, P.; Carrat, F.; Delmas, V.; Boudhabhay, O.; Gayon, F.; Guincetre, D.; Czernichow, P.
(1999) Air pollution and respiratory drug sales in the city of Le Havre, France, 1993-1996. Environ. Res.
81:224-230.
Zeghnoun, A.; Czernichow, P.; Beaudeau, P.; Hautemaniere, A.; Froment, L.; Le Tertre, A.; Quenel, P. (2001)
Short-term effects of air pollution on mortality in the cities of Rouen and Le Havre, France, 1990-1995. Arch.
Environ. Health 56: 327-335.
Zemp, E.; Elsasser, S.; Schindler, C.; Ktinzli, N.; Perruchoud, A. P.; Domenighetti, G.; Medici, T.;
Ackermann-Liebrich, U.; Leuenberger, P.; Monn, C.; Bolognini, G.; Bongard, J.-P.; Brandli, O.; Karrer, W.;
Keller, R.; Schoni, M. H.; Tschopp, J.-M.; Villiger, B.; Zellweger, J.-P.; SAPALDIA Team. (1999) Long-term
ambient air pollution and respiratory symptoms in adults (SAPALDIA study). Am. J. Respir. Crit. Care Med.
159: 1257-1266.
8-370
-------
Zhang, I; Qian, Z.; Kong, L.; Zhou, L.; Yan, L.; Chapman, R. S. (1999) Effects of air pollution on respiratory health
of adults in three Chinese cities. Arch. Environ. Health 54: 373-381.
Zhang, H.; Triche, E.; Leaderer, B. (2000) Model for the analysis of binary time series of respiratory symptoms. Am.
J. Epidemiol. 151: 1206-1215.
Zidek, J. V.; Wong, H.; Le, N. D.; Burnett, R. (1996) Causality, measurement error and multicollinearity in
epidemiology. Environmetrics 7: 441-451.
Zmirou, D.; Schwartz, J.; Saez, M; Zanobetti, A.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Ponce de Leon, A.;
Le Moullec, Y.; Bacharova, L.; Schouten, J.; Ponka, A.; Katsouyanni, K. (1998) Time-series analysis of air
pollution and cause-specific mortality. Epidemiology 9: 495-503.
8-371
-------
APPENDICES 8A AND 8B
SHORT-TERM PARTICULATE MATTER
EXPOSURE—MORTALITY AND
PARTICULATE MATTER-MORBIDITY STUDIES:
SUMMARY TABLES
-------
APPENDIX 8A
SHORT-TERM PM EXPOSURE-MORTALITY STUDIES:
SUMMARY TABLE
8A-1
-------
TABLE 8A-1. 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.
OO
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
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)
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.
Lag 1 day PM10 concentration positively associated with total
mortality in most locations, though estimates ranged from 2.1%
to -0.4% per 10 ug/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.
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).
Daniels et al. (2000).*
The largest U.S. 20 cities,
1987-1994.
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 ug/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.
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.
Dominici et al. (2003a).
Re-analysis of above study.
Re-analysis of above model using GLM/natural splines.
The shapes of concentration-response curves were similar to the
original analysis.
+ = 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
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
>
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. (2003a).
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)
+ = 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
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)
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.
Braga etal. (200la).*
Ten U.S. cities.
Same as Schwartz (2000b).
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.
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.
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 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.
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.
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.
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
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)
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). Re-analysis
of above study.
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 O3 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).
+ = 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
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)
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
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
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
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)
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 1; per 25 ug/m3 PM25 in
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 PM2 5,
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
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)
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_25 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(lag4)= 11.5(0.2, 24.1)per
25 ug/m3
PM10.25 (lag 0 or 2) = 1.3 (-0.6, 3.5) per
25 ug/m3
Cardio deaths:
PM10 (lag 0) = 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 (lag 0) = 2.6 (0.7, 4.5) per
25 ug/m3
Respiratory deaths:
PM10 (lag 3) = -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.
-------
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
>
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)
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).
Fairley (2003). Re-analysis
of above study.
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
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
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
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(ld) = 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 (0 d) = 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)
+ = 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
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)
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)
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)
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.
-------
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.
OO
>
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.
+ = 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
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
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)
+ = 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
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)
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), and PM10. The nephelometer measures
were converted to gravimetric units based on a regression.
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_2 5, and EC.
Cardiovascular mortality was significantly associated with CO,
NO2, SO2, PM25, PM10, PM10_25, OC and EC. Combustion-
related factors and secondary aerosol factors were also
associated with cardiovascular mortality. Soil-related factors, as
well as individual variables that are associated with soil were
negatively associated with total mortality.
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
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)
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
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
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
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. Daily PM10 concentrations were
estimated from the every-sixth-day measures by an
estimation model.
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 PM2 5 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.
+ = 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
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
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 |ig/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 |ig/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.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
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
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.
+ = 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
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
to
Canada
Burnett etal.(1998a).+
11 Canadian cities.
1980-1991.
No PM index data available
on consistent daily basis.
Burnett et al. (2000).*
8 largest Canadian cities.
1986-1996. All city mean
PM1025.9;PM25 13.3;
PM10.25 12.6; sulfate 2.6.
Burnett and Goldberg
(2003).
Re-analysis of above study.
Total non-accidental deaths were linked to gaseous air
pollutants (NO2, O3, SO2, and CO) using GAM Poisson
models adjusting for seasonal cycles, day-of-week, and
weather (selected from spline-smoothed functions of
temperature, dewpoint, relative humidity with 0, 1, and
2 day lags using forward stepwise procedure). Pollution
variables evaluated at 0, 1, 2, and up to 3-day lag averages
thereof. No PM index included in analyses because daily
PM measurements not available. City-specific models
containing all four gaseous pollutants examined. Overall
risks computed by averaging risks across cities.
Total non-accidental deaths linked to PM indices (PM10,
PM25, PM10_25, sulfate, 47 elemental component
concentrations for fine and coarse fractions) and gaseous air
pollutants (NO2, O3, SO2, and CO). Each city's mortality,
pollution, and weather variables separately filtered for
seasonal trends and day-of-week patterns. The residual
series from all the cities then analyzed in a GAM Poisson
model. The weather model was selected from spline-
smoothed functions of temperature, relative humidity, and
maximum change in barometric pressure within a day, with
0 and 1 day lags using forward stepwise procedure.
Pollution effects were examined at lags 0 through 5 days.
To avoid unstable parameter estimates in multi-pollutant
models, principal components were also used as predictors
in the regression models.
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.
NO2 had 4.1% increased risk per mean concentration; O3 had
1.8%; SO2 had 1.4%, and CO had 0.9% in multiple pollutant
regression models. A 0.4% reduction in excess mortality was
attributed to achieving a sulfur content of gasoline of 30 ppm in
five Canadian cities. Daily PM data for fine and coarse mass
and sulfates available on varying (not daily) schedules allowed
ecologic comparison of gaseous pollutant risks by mean fine
particle indicators mass concentrations.
O3 was weakly correlated with other pollutants and other
pollutants were "moderately" correlated with each other (the
highest was r = 0.65 for NO2 and CO). The strongest association
with mortality for all pollutants considered were for 0 or 1 day
lags. 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.
Found suggestion of weak negative
confounding of NO2 and SO2 effects
with fine particles and weak positive
confounding of particle effects with O3.
No quantitative RR or ER estimates
reported for PM indicators.
Percentage increase in daily filtered non-
accidental deaths associated with
increases of 50 ug/m3 PM10 and
25 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
PM2 5, PM10_2 5, and the 4 gaseous
pollutants, 1.9 (0.6, 3.2) for PM25; and
1.2 (-1.3, 3.8)forPM10.25.
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)per25 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.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
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
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. (200 lb)*
Montreal, Quebec.
1984-1993. Predicted PM25
mean= 17.6. CoH (1000ft)
mean = 0.24, sulfate mean
= 3.3.
Goldberg etal. (200 Id).
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
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
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.
+ = 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
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
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.
Samolietal. (2001). *
APHEA 1 cities (see
Katsouyanni (1997). At least
five years between 1980-
1992. The PM levels are the
same as those in Katsouyanni
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).
+ = 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
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
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).
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
(2003b). 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
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
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.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
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
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).
Hoek (2003). Re-analysis of
above study.
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.
Re-analysis of above study using stringent convergence
criteria and natural splines.
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.
Very little change in PM risk coefficients (often slightly
increased) whether GAM with stringent convergence criteria or
GLM./natural splines were used.
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.
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).
Hoeketal. (2001).*
The Netherlands.
1986-1994. PM10 (median
34); BS (median 10).
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.
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).
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
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
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.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE MATTER EXPOSURE MORTALITY EFFECTS STUDIES
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
Europe (cont'd)
Hoeketal. (1997).
+Rotterdam, the
Netherlands, 1983-1991.
TSP (median 42); BS
(median 13).
Kotesovec et al. (2000).
Northern Bohemia, Czech
Republic, 1982-1994.
TSP (121.3).
Zanobetti et al. (2000a).
Milan, Italy. 1980-1989.
TSP mean = 142.
Total mortality (also by age group) was regressed on TSP,
Fe (from TSP filter), BS, O3, SO2, CO, adjusting for
seasonal cycles, day-of-week, influenza, temperature, and
humidity using Poisson GAM model.
Total (excluding accidents and children younger than 1 yr),
cause specific (cardiovascular and cancer), age (65 and less
vs. otherwise), and gender specific mortality series were
examined for their associations with TSP and SO2 using
logistic model, adjusting for seasonal cycles, influenza
epidemics, linear and quadratic temperature terms. Lags 0
through 6 days, as well as a 7 day mean values were
examined.
The focus of this study was to quantify mortality
displacement using 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.
Daily deaths were most consistently associated with TSP. TSP
and O3 effects were "independent" of SO2 and CO. Total iron
(from TSP filter) was associated "less consistently" with
mortality than TSP was. The estimated RRs for PM indices were
higher in warm season than in cold season.
For the total mortality, TSP, but not SO2, was associated. There
were apparent differences in associations were found between
men and women. For example, for age below 65 cardiovascular
mortality was associated with TSP for men but not for women.
TSP was positively associated with mortality up to 13 days,
followed by nearly zero coefficients between 14 and 20 days, and
then followed by smaller but positive coefficients up to the 45th
day (maximum examined). The sum of these coefficients was
over three times larger than that for the single-day estimate.
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).
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.
Anderson et al. (1996).
London, UK, 1987-1992.
BS(15)
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.
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.
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).
+ = 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
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)
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).
P° Rahlenbeck and Kahl
7 (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
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)
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
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
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
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)
+ = 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
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)
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
*7 measurements available at
S^ 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, PM25,
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).
+ = 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
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
oo
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).
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)
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.
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.
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.
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.
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,.
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).
+ = 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
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
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.
-------
TABLE 8A-1 (cont'd). SHORT-TERM PARTICULATE 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.
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.
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,,.
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
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
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.
+ = 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
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
+ = 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.
VO
-------
APPENDIX 8B
PARTICULATE MATTER-MORBIDITY STUDIES:
SUMMARY TABLES
8B-1
-------
Appendix 8B.1
PM-Cardiovascular Admissions Studies
8B-2
-------
TABLE 8B-1. 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
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)
-------
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)
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
Zanobetti and Schwartz (2003a)
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.)
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.
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.
Zanobetti and Schwartz (2003a) 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
-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]
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]
-------
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
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 (2003a) 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)
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
-------
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)
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
-------
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
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
SO; = 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)
PM10.25 (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)
-------
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)
Ito 2003
Detroit (Wayne County), MI
OO
td
i
OO
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)**
-------
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
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
OO
td
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)
-------
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)*
GLMnsplinelOOdf: 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)
-------
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
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)
-------
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
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
95th percentile: 1.5 103 In feet
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)
-------
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
Burnett etal. (1997c)
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: 0-4 d.
6.2 (4.0, 8.4)
5.9(2.8, 9.1) w. gases
H+: 2-4 d.
2.4(0.4,4.5)
0.5 (-1.6, 2.7) w. gases
S04:2-4d.
1.7 (-0.4, 3.9)
-1.6 (-4.4, 1.3) w. gases
PM10: l-4d.
7.7(0.9, 14.8)
-0.9 (-8.3, 7.1)w. gases
PM25 : 2-4 d.
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
-------
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
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)
PM10 est (0-2 d): 9.7(4.2,15.5)
IHD:
FPest(0-2d): 8.1(5.4, 10.8)
CPest(Od): 3.7(1.3,6.3)
PM10est(0-ld): 8.4(5.3,11.5)
-------
TABLE 8B-1 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND CARDIOVASCULAR
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
Stieb et al. (2000)
Saint John, Canada
7/1/92-3/31/96
mean and S.D.:
PM10(ng/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.
-------
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
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. (1999a)
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; for IHD: 24.5 <65, 37.6
>65. Covariates: NO2, O3, SO2, CO, temperature,
relative humidity. Poisson regression using APHEA
methodology; sine and cosine functions for seasonal
control; day of week dummy variables. Lags of 0-3, as
well as corresponding multi-day averages ending on
lag 0, were considered.
Pooled results were reported for the cardiac
admissions results in table format. City-specific
and pooled results were depicted in figures only.
Found a significant effect of PM10 and black
smoke on admissions for cardiac causes (all
ages) and cardiac causes and ischemic heart
disease for people over 65 years with the impact
of PM10 per unit of pollution being half that
found in the United States. PM10 did not seem to
be confounded by O3 or SO2. The effect was
reduced when CO was incorporated in the
regression model and eliminated when
controlling for NO2. There was little evidence of
an impact of particles on hospital admissions for
ischemic heart disease for people below 65 years
or stroke for people over 65 years. The authors
state results were consistent with a role for traffic
exhaust/diesel in Europe.
In single-pollutant models, both PM metrics
showed positive associations with both CVD and
IHD admissions across age groups. In Two-
pollutant models, the BS effect, but not the PM10
effect, was robust. No quantitative results
provided for two-pollutant models. Study does
not support a PM10 effect independent of co-
pollutants.
For a 10 ug/m3 increase in PM10
Cardiac admissions/all ages
0.5% (0.2, 0.8)
Cardiac admissions/over 65 years
0.7% (0.4, 1.0)
Ischemic heart disease/over 65 years
0.8% (0.3, 1.2)
For cardiac admissions for people over 65
years: All the city-specific estimates were
positive with London, Milan, and
Stockholm significant at 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)
-------
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
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
Diaz et al. (1999)
Madrid, Spain
1994-1996
TSP by beta attenuation
Summary statistics not given.
Daily emergency hospital admissions for
cardiovascular disease (ICD9 codes 410-414, 426-429,
434-440) for persons less than 65 years and for
persons 65 or older. Separate analyses presented for
long (1981-1995) and short (1992-1995) series. Mean
hospital admissions for long and short series: <65, 3.5,
3.4; 65+, 8.0, 8.7. Covariates: SO2, NO2, O3, CO,
wind speed, temperature, rainfall. PM10 measured by
TEOM. Stats: Poisson log-linear regression; trend
and seasons controlled by monthly dummy variables
over entire series; day of week dummy variables; min
daily temperature modeled using octile dummies.
Pollutants expressed as cumulative lag 1-3 day moving
Daily hospital admissions for acute ischemic heart
disease (ICD9 codes 410-429) for all ages. Mean
hospitalizations: 25.6/day. Covariates: temperature
and relative humidity. Stats: Linear regression with
day of week and monthly dummy variables, linear
trend term. Lags of 0-3 considered, as well as the
mean of lags 0-2.
Daily emergency hospital admissions for all
cardiovascular causes (ICD9 codes 390-459) for
the Gregorio Maranon University Teaching Hospital.
Mean admissions: 9.8/day. Covariates: SO2, NO2, O3,
temperature, pressure, relative humidity, excess
sunlight. Stats: Box-Jenkins time-series methods used
to remove autocorrelations, followed by cross-
correlation analysis; sine and cosine terms for
seasonality; details unclear.
In long series, neither BS nor NO2 were
associated with CVD admissions in either age
group. In the short series, only 3-day moving
average PM10 was positively and significantly
associated with CVD admissions in single-
pollutant models, and only for persons 65 or
older. BS, SO2, and CO also 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 ug/m3 change PM10
IHD admissions:
PM10 0-dlag:
1.4% (-4.4, 7.2)
PM10 1-dlag:
-1.3% (-7.1, 4.4)
No significant effects of TSP on CVD reported. No quantitative results presented for PM.
-------
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
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, 0-2 d. avg.
CVD:
5-64: 2.5 (-1.5, 6.7)
65+: 4.1(1.3,6.9)
All: 3.0 (0.8, 5.4)
HE (PM10, 0-3 d ave.):
All: 26.4(17.1,36.4)
IHD (PM10, 0-3 d ave.):
All: 3.5 (-0.5, 7.7)
-------
Appendix 8B.2
PM-Respiratory Hospitalization Studies
8B-19
-------
TABLE 8B-2. ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
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
to
o
Reanalysis of Samet et al (2000a) by
Zanobetti and Schwartz (2003a)
Re-analyses of Samet et al. (2000a) 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)
-------
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
to
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)
-------
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
to
to
United States (cont'd)
Gwynn et al. (2000)+
Buffalo, NY (5/88-10/90)
PM10 mn./max. = 24.1/90.8 ng/m3
PM10 IQR = 14.8-29.2 ug/m3
SO4" mn./max. = 2.4/3.9 ug/m3
S04-IQR = 23.5-7.5ng/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 |ig/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)
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)
-------
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)
PM10IQR =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
to
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)
-------
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
to
-------
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
to
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)
-------
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
to
United States (cont'd)
Reanalysis by
Ito (2003)
Lumley and Heagerty (1999)
Seattle (King Cty.), WA (87-94)
Population = NR
PMj 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
Re-analyses of Lippmann et al. (2000) with more
stringent GAM convergence criteria and
alternative models.
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)
NSGLM: 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)
-------
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
to
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)
-------
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
to
OO
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)
-------
TABLE 8B-2 (cont'd). ACUTE PARTICIPATE MATTER EXPOSURE AND RESPIRATORY HOSPITAL
ADMISSIONS STUDIES
Reference/Citation, Location,
Duration, PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
to
VO
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)
StrictGAM: ER=6.5%(1.1,12.0)
NSGLM: ER= 8.7% (3.3,14.4)
With Co-poll:
StrictGAM: 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)
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.
-------
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
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(
25ug/m3PM25andPM(10.25):
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%)
-------
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: 18ng/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
Europe
Atkinson et al. (1999b)
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+ vrs.)
(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)
-------
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
to
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
McGregor et al. (1999)
Birmingham, UK.
Population = NR
Mean PM10 = 30.0 ug/m3
Hagen et al. (2000)+
Drammen, Sweden( 11/94-12/97)
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 564 yr. (mean resp HA = 8.7/day), and for
subjects with multiple HA's.
A synoptic climatological approach used to
investigate linkages between air mass types
(weather situations), PM10, and all respiratory
hospital admissions (mean= 19.2/day) for the
Birmingham area.
Examined PM10, SO2, NO2, VOC's, and O3
associations with respiratory hospital admissions
from one hospital (mean = 2.2/day). Used
Poisson GAM controlling for temperature and
RH (but not their interaction), long-wave and
seasonality, day-of-week, holidays, and
influenza epidemics.
Daily mortality and general admissions to Paris
public hospitals for respiratory causes were
considered (means/day: all resp.=79/d,
asthma=14/d, COPD=12/d). Time series
analysis used linear regression model followed
by a Poisson regression. Epidemics of influenza
A and B, temperature, RH, holidays, day of
week, trend, long-wave variability, and nurses'
strike variables included. No two pollutant
models considered.
The two strongest findings were for
cardiovascular HA's of people aged >64, which
showed a positive association with PM10 as a
mean of the 3 previous days. PM10 was
consistently positively associated with
Respiratory HA's in both age groups, with the
greatest effect size in those >64, especially
among those with >4 HA's during '81-'95.
Weak significances likely contributed to by low
population size.
Study results show distinct differential
responses of respiratory admission rates to the
six winter air mass types. Two of three types of
air masses associated with above- average
admission rates also favor high PM10 levels.
This is suggestive of possible linkage between
weather, air quality, and health.
As a single pollutant, the PM10 effect was of
same order of magnitude as reported in other
studies. The PM10 association decreased when
other pollutants were added to the model.
However, the VOC's showed the strongest
associations.
For the all respiratory causes category, the
authors found "the strongest association was
observed with PM13" for both hospital
admissions and mortality, indicating a coherence
of association across outcomes. Asthma was
significantly correlated with NO2 levels, but not
PM,,.
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)
NR
Respiratory Hospital Admissionsfall ages)
For IQR=50 ug/m3
-Single Pollutant Model:
PM10 (lag 0) ER = 18.3% (CI: -4.2, 46)
-Two Pollutant Model (with O3):
PM10 (lag 0) ER = 18.3% (CI: -4.2, 45.4)
-Two Pollutant Model (with Benzene):
PM10 (lag 0) ER = 6.5% (CI:-14 , 31.8)
For PM13 = 50 ug/m3; BS = 25 ug/m3;
Respiratory HA's (all ages):
PM13 Lag 0 ER = 2.2% (CI: 0.2, 4.3)
BS Lag 0 ER = 1.0% (0.2, 1.8)
COPD HA's (all ages):
PM13Lag2ER= 2.3% (CI: -6.7,2.2)
BSLag2ER= 1.1% (-2.9, 0.6)
Asthma HA's (all ages):
PM13Lg2ER= 1.3% (CI: -4.6,2.2)
BS Lg 0 ER = 1.2% (-0.5, 2.9)
-------
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
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)
-------
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
Europe (cont'd)
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
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
Kontos etal. (1999)
Piraeus, Athens GR (87 - 92)
Population = NR
BS mean =46.5 ng/m3
BS max =200 ug/m3
Association between adult respiratory HA's
(15-64 yr mean =11.3/day, and 65 + yr mean
=8.8/day) and air pollution evaluated, using the
APHEA protocol. Poisson regression used with
control for weather and long term trend, year,
influenza epidemics, and season
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).
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.
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.
Daily hospital admissions for asthma found to
have associations with O3, SO2, NO2, and
particles (BS), but there was lack of consistency
across the age groups in the specific pollutant.
BS association was strongest in the 65 + group,
especially in winter. Pollens not consistently
associated with asthma HA's, sometimes being
positive, sometimes negative. Air pollution
associations with HA's not explained by
airborne pollens in simultaneous regressions,
and there was no consistent pollen-pollutant
interaction.
Pollution found to explain significant portion of
the HA variance. Of pollutants considered, BS
was consistently among most strongly
explanatory pollutants across various reported
analyses.
Young Adult (15-64 yrs.) Resp. HA's
100 ug/m3 increase in TSP
Lag 2 ER = 5% (CI: 0, 10)
Older Adult (65+ yrs.) Resp. HA's
100 |ig/m3 increase in TSP
Lag 1 ER = 5% (CI: -1, 10)
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)
NR
-------
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
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)
-------
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
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 ng/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)
Paris ER = -0.7% (Ig 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)
-------
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
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 PM25 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)
PMio.
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)
-------
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
NO, 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)
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.
-------
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
Latin 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
-------
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
PMi5 lO'-gO^/o = 7.5-44.4 ug/m3
McGowan et al. (2002)
Christchurch, New Zealand
June 1988 through December 1998
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.
The relationship between PM10 and admissions
to hospital with cardiorespiratory illnesses for
both adults and children using a time series
analysis controlling for weather variables
missing PM10 values were interpolated from CO
data from the same period. GAM used with
default criteria.
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.
There was a significant association between
PM10 levels and cardiorespiratory admissions.
For all age groups combined there as a 3.37%
increase in respiratory admissions for each
interquartile value rise in PM10 (interquartile
value 14.8. ug/m3) and a 1.26% rise in cardiac
admissions for each interquartile rise in PM10
(IQR =14.8 ug/m3).
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+vrs.)
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
Tanakaetal. (1998)
StdyPd.: 1/92-12/93
Kushiro, Japan
Pop. = 102 adult asthmatics
PM10 mean = 24.0 ug/m3
PM10 IQR = 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)
-------
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
+ = 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.
-------
Appendix 8B.3
PM-Respiratory Visits Studies
8B-42
-------
TABLE 8B-3. 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
OO
td
United States
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)
-------
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
OO
td
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
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
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.
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.
Stagnation persistence index was strongly
associated with ED visits for asthma in both
cities. Factor analysis indicated that products
of incomplete combustion (especially wood-
smoke related K, OC, EC, and CO) are the air
pollutants driving this association. Multi-
pollutant models run with "stagnation" as the
"co-pollutant" indicated importance of general
air pollution over any single air pollutant index,
but not of the importance of various pollutants
relative to each other.
Both PM10 and O3 positively associated with
asthma ED visits using all three modeling
approaches. In models with both O3 and PM10,
both pollutants become non-significant because
of high collinearity of the variables (r=0.75).
Asthma ED Visits
Single Pollutant Models
Persons<65 years (Spokane)
For PM10IQR = 50 ug/m3
Lag 3 ER = 2.4% (CI: -10.9, 17.6)
Persons<18 years (Seattle)
For PM10 IQR = 50 ug/m3
Lag 3 ER = 56.2% (95 CI: 10.4,121.1)
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
Reference/Citation, Location, Duration,
PM Index/Concentrations
Study Description:
Results and Comments
PM Index, Lag, Excess Risk %,
(95% CI = LCI, UCL) Co-Pollutants
OO
td
United States (cont'd)
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
PM25 soluble metals (ng/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
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
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.
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)
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
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.
Only O3 showed significant associations with
asthma ER visits. However, the crude season
adjustment and linear model (rather than
Poisson) may have adversely affected results.
Also, Beta-gauge PM10 mass index used, rather
than direct gravimetric mass measurements.
NR
-------
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
Delfmo et al. (1997a)
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
Delfmo et al. (1998b)
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
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
Canada (cont'd)
Stieb et al. (2000)+
Saint John, New Brunswick, Canada
7/1/92-3/31/96
mean and S.D.:
PM10(ng/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. (1999a)
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)
-------
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
OO
td
-U
OO
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)
0-14yrs.(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)
-------
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
Europe (cont'd)
Medina etal. (1997)+
Greater Paris 91 - 95
Populations 6.5 MM
Mean PM13 = 25 ug/m3
PM13 min/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)
0-14yrs.(lgO-3): ER = 41.5%(CI: 20,66.8)
15-64yrs.(lg2): ER = 6.3%(CI: -4.6,18.5)
OO
td
-U
VO
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)
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
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)
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.
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.
N/A
Latin America
oo
td
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
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)
Linetal. (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)
-------
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
OO
td
i
(^
to
Australia
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)
B!catt median = 0.25 10 4/m
BscattIQR = 0.18-0.39 10 4/m
Bscatt 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
TSPmean = 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
(mean 10/day) increased as result of bushfire-
generated PM ( Bscatt from nephelometry) in Jan.,
1994 (period 2). Air pollution data included
nephelometry (Bscatt), 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
-------
Appendix 8B.4
Pulmonary Function Studies
8B-53
-------
TABLE 8B-4. SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY FUNCTION
TESTS IN STUDIES OF ASTHMATICS
Effect measures standardized to 50 ug/m3
PM10 (25 ug/m3 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
Canada
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.
OO
td
Vedaletal. (1998)
Port Alberni, BC
PM10 via a Sierra-Anderson dichotomous sampler. PM1(
ranged from 1 to 159 |ig/m3.
Europe
Gielen et al. (1997)
Amsterdam, NL
Mean PM10 level: 30.5 ug/m3 (16, 60.3).
Mean maximum 8 hr O3: 67 ng/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.
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.
Ozone, SO2 and sulfate levels low due to
low vehicle emissions. PM10 associated
with change in peak flow.
The strongest relationships were found
with ozone, although some significant
relationships found with PM10.
Lag 0, PM 10 average
PEE = 0.27 (-0.54, -0.01) per 10 ug/m3
increment
Lag 0, PM10:
Evening PEE = -0.08 (-2.49, 2.42)
Lag 1, PM10:
Morning PEE = 1.38 (-0.58, 3.35)
Lag 2, PM10:
Morning PEE = 0.34 (-1.78, 2.46)
Evening PEE = -1.46 (-3.23, 0.32)
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)
-------
TABLE 8B-4 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY
FUNCTION TESTS IN STUDIES OF ASTHMATICS
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 PMM). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
OO
td
Europe (cont'd)
Peters et al. (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
-------
TABLE 8B-4 (cont'd). SHORT-TERM PARTICIPATE MATTER EXPOSURE EFFECTS ON PULMONARY
FUNCTION TESTS IN STUDIES OF ASTHMATICS
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 PMM). Negative coefficients
for lung function and ORs greater than 1 for
other endpoints suggest PM effects
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
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.
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.
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.
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.
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.
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.
Effects found related to PM10 were less
than those found related to the other
pollutants. The strongest effects were
found with SO,.
Lag 4, PM13:
Morning PEF = -0.62 (-1.52, 0.28)
-------
TABLE 8B-4 (cont'd). 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 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
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.
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
-------
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
Latin America
oo
td
i
u\
oo
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.
Ozone strongly related to changes in
morning PEE as well as PM10.
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.
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)
LagO, PM25:
Evening PEE = -4.27 (-7.12, -0.85)
Lag 2, PM25:
Evening PEE = -2.55 (-7.84, 2.74)
Lag 1, PM10
Morning PEE = -4.70 (-7.65, -1.7)
Lag 2, PM10
Morning PEE = -4.90 (-8.4, -1.5)
Lag 0, PM10:
Evening PEE = -1.32 (-6.82, 4.17)
Lag 2, PM10:
Evening PEE = -0.04 (-4.29, 4.21)
Morning PEE = 2.47 (-1.75, 6.75)
Lag 0, PM10:
Morning PEE = 0.65 (-3.97, 5.32)
-------
Appendix 8B.5
Short-Term PM Exposure Effects on Symptoms in
Asthmatic Individuals
8B-59
-------
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
oo
td
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.
Delfmo et al. (1997b)
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.
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.
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
Pollen not associated with asthma
symptom scores. 12-hr personal O3 but
not ambient O3 related to symptoms.
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.
No significant relationships with PM10.
Delfmo et al. (1998a)
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.
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
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.
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)
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 ng/m3 with a mean of
51.8.
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.
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
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
-------
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)
Vedaletal. (1998)
PM10 measured by Sierra-Anderson dichotomous
sampler
PM10 range: -1 to 159 ug/m3
Port Alberrni
British, Columbia
206 children aged 6 to 13 years, 75 with physician's
diagnosis of asthma. Respiratory symptom data
from diaries, GEE model. Temp., humidity.
PM10 associated with respiratory
symptoms.
LagO
Cough OR = 1.08(1.00, 1.16) per 10 ug/m3
PM,n increments
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.
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.
Strongest relationships found with O3,
although some significant relationships
found with PM10.
Lag 0, Symptoms:
Cough OR = 2.19 (0.77, 6.20)
Bronch. Dial. OR = 0.94 (0.59, 1.50)
Lag 2, Symptoms:
Cough OR = 2.19 (0.47, 10.24)
Bronch. Dial. OR = 2.90 (1.80, 4.66)
-------
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
Europe (cont'd)
Hiltermann et al. (1998)
Leiden, NL
July-Oct 1995.
Ozone, PM10, NO2, SO2, BS
PM10 ranged from 16 to 98 ug/m3 with a mean of 40.
Hiltermann et al. (1997)
The Netherlands
Ozone and PM10
PM10 averaged 40 ug/m3,
Peters etal. (1997b)
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 et al. (1997c)
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)
ag , 10:
Cough OR = 1.32(1.16, 1.50)
Feeling ill OR = 1.20 (1.01, 1.44)
5 Day Mean, PM10:
=
5 Day Mean, PM10:
Cough OR = 1.30 (1.09, 1.55)
Feeling ill OR = 1.47(1.16, 1.86)
LagO, PM25:
Cough OR = 1.19(1.07, 1.33)
Feeling ill OR = 1.24 (1.09, 1.41)
5 Day Mean, PM25:
Cough OR =1.02 (0.91, 1.15)
Feeling ill OR = 1.21 (1.06, 1.38)
Lag 0, Symptoms:
Cough OR =1.0 1(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)
-------
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
Europe (cont'd)
Peters et al. (1997c)
Sokolov, Czech Republic
PM10 one central site. SO4 reported.
Mean PM10: 55 ug/m3, max 177 ug/m3.
SO4-fme: mean 8.8 ug/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.
OO PM was measured by radiometry.
Co PM13 ranged from 9 to 95 ug/m3 with a mean of 34.
00 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
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.
-------
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
Von Klot et al. (2002)
PM25.10, PM10, NO2, SO2, CO, temperature
Desqueyroux et al. (2002)
PM10, O3, SO2, and NO2
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 PM2 5_10
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 PM25.10.
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
8B-65
-------
TABLE 8B-6. 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
United States
Hoeketal. (1998)
(summary paper)
OO
td
ON
a\
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.
Significant decreases in peak flow
found to be related to PM10
increases.
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.
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 PEER (lag 0) = -0.05 (-1.73,
0.63)
PM PEER (lag 1)= -0.64 (-1.73,
0.44)
-------
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
United States (cont'd)
Neasetal. (1999)
Philadelphia, PA
Median PM10 level: 31.6 in SW camps,
27.8 in NE camps (IQR ranges of about 18).
Median 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.
Schwartz and Neas (2000)
Eastern U.S.
PM25 and CM (PM10_25) measured.
Summary levels not given.
Linn et al. (1996)
So. California
NO2 ozone, and PM5 measured.
PM5 was measured using a Marple low volume
sampler PM5 ranged from 1-145 ug/m3 with a mean
of 24.
Panel study of 156 normal children attending YMCA and YWCA
summer camps in greater Philadelphia area in 1993. Children
followed for at most 54 days. Morning and evening deviations of
each child's PEE were analyzed using a mixed-effects model
adjusting for autocorrelation. Covariates included time trend and
temp. Lags not used in the analysis.
Analyses for 1844 school children in grades 2-5 from six urban
areas in eastern U.S. and from separate studies from Uniontown
and State College, PA. Lower resp. symptoms, cough and PEE
used as endpoints. The authors replicated models used in the
original analyses. CM and were used individually and jointly in
the analyses. Sulfate fractions also used in the analyses. Details of
models not given.
Study of 269 school children in Southern California twice daily for
one week in fall, winter and spring for two years. A repeated
measures analysis of covariance was used to fit an autoregressive
model, adjusting for year, season, day of week, and temperature.
Analyses that included sulfate
fraction and O3 separately also
found relationship to decreased
flow. No analyses reported for
multiple pollutant models.
Sulfate fraction was highly
correlated with PM2 5 (0.94), and,
not surprisingly, gave similar
Morning FVC was significantly
decreased as a function of PM5 and
NO,
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)
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.
-------
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
Vedaletal. (1998)
Port Alberni, BC
PM10 via a Sierra-Anderson dichotomous sampler.
PM10 ranged from 1 to 159 ug/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.
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 PEE 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
-------
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
ON
VO
Europe (cont'd)
Kiinzli et al. (2000)
Roemer et al. (2000)
PM10 means for 17 panels ranged 11.2 to 98.8 ug/m3.
SO2, NO2, and elemental content of PM also measured.
Measurement methods were not described.
Scarlett et al. (1996)
PM10, O3, and NO2 measured.
Ackermann-Liebrich et al. (1997) data reanalyzed. Authors
showed that a small change in FVC (-3.14 percent) can result in a
60% increase in number of subjects with FVC less than 80 percent
of predicted.
Combined results from 1208 children divided among 17 panels
studied. Separate results reported by endpoints included symptoms
as reported in a dairy and 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.
The results were for two
hypothetical communities, A and
B.
Daily concentrations of most
elements were not associated with
the health effects.
PM10 was related to changes in
FEV and FVC
PM10 analyses not focus of this
paper.
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)
-------
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
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)
1.5
= -1.93 (-7.00,
= 1.52 (-3.91, 6.94)
-------
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
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.
-------
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
8B-73
-------
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 PM25 (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
Vedaletal. (1998)
Port Alberni, BC
PM10 via a Sierra-Anderson dichotomous sampler.
PM10 ranged from 1 to 159 ug/m3.
Long et al. (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
-------
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
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.
-------
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
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)
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 ug/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.
-------
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
-------
Appendix 8B.8
Long-Term PM Exposure Effects on Respiratory Health Indicators,
Symptoms, and Lung Function
8B-78
-------
TABLE 8B-8. 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
ofPM
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
Berglundetal.. (1999)
California communities
Peters et al. (1999b,c)
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.
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 = -1.09CT(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).
-------
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
ofPM
OO
td
i
OO
o
United States (cont'd)
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).
McConnell et al. (2002)
12 Southern California communities
1994-1997
4-year mean cone. PM10 ug/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.
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).
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.
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.
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.
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)
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.
-------
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
ofPM
OO
td
i
OO
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).
Europe
Ackermann-Liebrich et al. (1997)
Eight Swiss regions
Pollutants: SO2, NO2, TSP, O3, and PM10.
PM was measured with a Harvard impactor.
PM10 ranged from 10 to 53 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 ug/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 ug/m3 with a mean of
21.
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.
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.
PM measures (e.g., particle strong acidity)
associated with FEV and FVC decrement.
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 ng/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.
-------
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
ofPM
OO
td
i
OO
to
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 ug/m3;
PM10 measured October 1993 - March 1994
ranged from 33 to 40; and BS ranged from 26 to
42 ug/m3. PM was measured with a Harvard
impactor.
Heinrich et al. (2000)
Three areas of former E. Germany
Pollution measures: SO2, TSP, and some limited
PM10 data. TSP decreased from 65, 48, and 44
ug/m3 to 43, 39, and 36 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 et al. (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
-------
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
ofPM
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
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
-------
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
ofPM
OO
td
i
OO
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.
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,.
-------
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
ofPM
OO
td
i
OO
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
Latin America
Calderon-Garciduenas et al. (2000)
Southwest Metropolitan Mexico City (SWMMC)
winter of 1997 and summer of 1998.
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
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.
Study of 59 SWMMC children to evaluate relationship
between exposure to ambient pollutants (O3 and PM10)
and chest x-ray abnormalities. Fishers exact test used
to determine significance in a 2x2 task between
hyperinflation and exposure to SWMMC pollutant
atmosphere and to control, low-pollutant city
atmosphere.
Cross-sectional survey of children's health and home
environment between Oct 1993 and Dec 1993
evaluated frequency of respiratory symptoms (night
cough, chest colds, wheeze, and diagnosed asthma).
Covariates included parental education and smoking,
unflued gas heating, indoor cats, age, sex, and maternal
allergy. Logistic regression analysis used allowing for
clustering by GEE methods.
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.
Bilateral symmetric mild lung hyperinflation was
significantly associated with exposure to the
SWMMC air pollution mixture (p>0.0004). This
raises concern for development of chronic disease
outcome in developing lungs.
SO2 was not related to differences in symptom
rates, but adult indoor smoking was.
Night cough OR 1.34 (1.18, 1.53)
Chest colds OR 1.43(1.12, 1.82)
Wheeze OR 1.13 (0.93, 1.38)
Wong et al. (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.
-------
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
ofPM
OO
td
i
OO
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 ng/m3 with a mean
of 69.
Wangetal. (1999)
Chongquing, China
April to July 1995
Dichot samplers used to measure PM2 5.
Mean PM25 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
SE 30.80
t- 1.88
p<0.05
-------
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)
3 China cities (1985-1988)
The 4 year average TSP means were 191,
296, 406, and 1067 ug/m3. SO2andNO2
measurements were also available.
TSP was measured gavimetrically.
A pilot study of 4 districts of 3 Chinese cities 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.
Shima et al. (2002)
8 communities in Japan, a prospective cohort
study (1989-1992).
Respiratory symptoms of 3049 school children were
evaluated by questionnaires every year from the 1st
through the 6th grades. PM10 measured continuously
by beta attenuation.
Incidence rates of asthma were associated
significantly with ambient levels of NO2. PM10 was
also associated but not significantly (OR 2.84; 95%
CI 0.84-9.58).
-------
9. INTEGRATIVE SYNTHESIS
9.1 INTRODUCTION
This chapter integrates key information from the preceding chapters to provide coherent
frameworks for assessment of human health and welfare risks posed by ambient particulate
matter (PM) in the United States. Rather than simply resummarizing information from earlier
chapters, the focus here is on integrating newly available scientific information with that
available from the last review so as to address a set of issues central to EPA's assessment of
scientific information upon which the PM NAAQS review is to be based.
In particular, this chapter provides an updated synthesis of the scientific information that is
intended to facilitate consideration of the key policy-related NAAQS issues to be addressed in
the PM Staff Paper, prepared by EPA's Office of Air Quality Planning and Standards (OAQPS)
staff. These policy-related issues include selection of appropriate indicators, averaging times,
forms, and levels for primary and secondary PM NAAQS in the United States. Ultimately,
EPA's consideration of these issues will be informed not only by the scientific information and
integrative assessment presented here and throughout this document, but also by additional
policy evaluations of scientific and technical information to be included in the PM Staff Paper.
As such, the PM Staff Paper serves to "bridge the gap" between scientific assessments and the
judgments required of the EPA Administrator in deciding whether to retain or revise the existing
PM NAAQS.
While this synthesis focuses on what has been learned since the last PM NAAQS review, it
also highlights important uncertainties that remain and recognizes the value of continuing PM
research efforts in a number of key areas. Although detailed delineation of research needs is
beyond the scope of this document, such recommendations are to be discussed in later PM
research needs documents and/or research plans to be prepared by EPA.
9.1.1 Chapter Organization
As part of this opening introduction, Section 9.1.2 first summarizes important information
on U.S. PM air quality trends and current ambient concentrations, to provide the context for
ensuing discussions of ambient PM characteristics, exposures, and effects.
9-1
-------
In considering PM-related health effects information, Section 9.2 then builds specifically
upon the integrative synthesis presented in Chapter 13 of the 1996 PM AQCD (U.S.
Environmental Protection Agency, 1996). The Section 9.2 synthesis of PM-related health effects
information is organized around five key issues: (1) consideration of fine and coarse thoracic
particles as separate subclasses of PM pollution, taking into account atmospheric science,
exposure, and dosimetric information; (2) assessment of strengths and limitations of the
epidemiological evidence for associations between health effects and fine and coarse thoracic
PM within the mix of ambient air pollutants; (3) integration of epidemiologic and experimental
(e.g., dosimetric and toxicologic) evidence supporting judgments about the extent to which
causal inferences can be made about observed associations between health endpoints and various
indicators or constituents of ambient PM, acting alone and/or in combination with other
pollutants; (4) characterization of susceptible and vulnerable subpopulations potentially at
increased risk for PM-related health effects; and (5) discussion of potential public health impacts
(including newly emerging evidence for adverse cardiovascular effects) of human exposures to
ambient PM in the United States.
Building upon information presented in the 1996 PM AQCD where possible, Section 9.3
addresses the major PM-related welfare effects of importance for decision-making for secondary
standards. This includes drawing upon key findings and conclusions on visibility and climate
effects from Chapter 8 and on damage to manmade materials from Chapter 9 of the 1996
document, as well as consideration of new findings discussed in Chapter 4 of this document.
Since PM-related effects on vegetation and ecosystems were not addressed in the 1996 PM
AQCD, the present discussion is based entirely on findings characterized in Chapter 4 of this
document.
9.1.2 Trends in United States PM Air Quality
PM10, PM25, andPM10_25 Concentrations and Trends
The nationwide average concentration of PM10 decreased from about 28 |ig/m3 to 24 |ig/m3
from 1992 through 2001 (U.S. Environmental Protection Agency, 2003). Most of this decrease
occurred during the first half of that time period. There was considerable variability in the trends
9-2
-------
for various geographic subregions, with the largest decreases being found in the northwest
(-9.6 |ig/m3) and the smallest in the south-central United States (-1.3 |ig/m3). These trends
reflect the continuation of longer-term declines in U.S. PM concentrations. For example, Lipfert
(1998) estimated that total suspended particulate (TSP) concentrations may have declined by
two- to three-fold in urban areas between 1950 and 1980. Data for quantifying nationwide
trends in PM2 5 concentrations are not available over this period. However, it may be surmised
that notable declines in PM2 5 concentrations also likely occurred over the same period. The
consistent reductions in PM10 concentrations found in a wide variety of environments may have
resulted from common controls that affected PM2 5 more strongly than PM10_2 5 particles
(Darlington et al., 1997). These considerations suggest that PM10_25 concentrations likely
decreased to a smaller extent over this period.
Annual mean PM25 concentrations in the United States currently average about 13 |ig/m3,
based on data collected from 1999 through 2001. Such fine particle concentrations can be less
than a few |ig/m3 in many remote areas in the western United States and in many urban areas
immediately after it has rained. However, 24-h PM2 5 concentrations on individual days can also
exceed 100 |ig/m3 at certain locations, especially if there are events such as wild fires or dust
storms. These values indicate a high degree of spatial and temporal variability in PM2 5
concentrations. PM25 concentrations observed in a number of urban areas across the United
States are characterized in Chapter 3; see Section 3.2 and Appendices 3A (for urban areas)
and 3E (for relatively remote areas).
It should be noted that the mean PM2 5 concentrations given above are considerably lower
than those obtained and used in many air pollution-health outcome studies conducted during the
1980s. Lipfert (1998) has estimated that PM2 5 concentrations decreased by about 4 to 5% per
year from 1970 to 1990; some of that change may be attributable to use of different monitoring
methods during earlier versus later years.
The Composition ofPM25 and PM10_25 Particles
Data for PM2 5, PM10_2 5 and PM10 from earlier monitoring studies, spanning the time period
from the late 1970s to the mid 1990s, were presented in Appendix 6A of the 1996 PM AQCD.
The data from such studies were summarized in Appendix 6A as pie charts showing the gross
9-3
-------
composition of the three size fractions for the eastern, central, and western United States. The
chemical composition of particles in the PM25 size range, as determined most recently by the
speciation network in 13 urban areas across the United States during 2001 to 2002 is
summarized in Chapter 3.
As summarized in Chapter 3 (Section 3.2 and Appendix 3B), sulfate and organic carbon
compounds constitute the major identified components of fine-particle the aerosol in the eastern
and central United States. In the western United States, organic compounds and nitrate and/or
sulfate constitute the major identified PM2 5 aerosol components. Other important components
are: elemental carbon, ammonium and crustal materials. Even though total organic compounds
constitute 10 to 70% of PM25, only about 10 to 20% of organic compounds in ambient samples
can be quantified due to analytical limitations resulting largely from the polar nature of some
organic compounds and the presence of oligomeric or polymeric substances (i.e., biopolymers
and humic-like substances). Results of studies characterizing the composition of organic
compounds in ambient particles are summarized in Appendix 3C. The attribution of organic
carbon to primary or secondary sources is still under study in different regions of the country.
Three mechanisms have been identified for the formation of secondary organic components in
ambient PM: (1) condensation of oxidized end- products of photochemical reactions (e.g.,
ketones, aldehydes, organic acids, and hydroperoxides), (2) adsorption of semivolatile organic
compounds (e.g., polycyclic aromatic compounds) onto existing particles, and (3) dissolution of
soluble gases (e.g., aldehydes) that can undergo reactions in particle bound water (PBW).
Primary biological particles (or bioaerosols) are also usually lumped into the broad category of
organic compounds. In addition to soluble organic compounds, soluble oxidants (e.g., H2O2) can
be taken up or formed in PBW, as discussed further below. As will be seen later, these
considerations have implications for the delivery of these and other soluble components to lower
respiratory tract regions.
Trace metals typically constitute a much smaller fraction of PM than the components given
above. Typically, their average combined air concentrations constitute less than 1% of PM25
levels (or on the order of 0.1 |ig/m3 or less), as shown in Appendix 3B. There are exceptions to
this general pattern in industrial cities (e.g., St. Louis, MO), where metals can constitute closer to
2% of PM25. However, maximum concentrations of Fe, typically the most abundant trace metal,
9-4
-------
can be on the order of several tenths of a |ig/m3 in any of the urban areas characterized. Prior to
the phaseout of leaded gasoline, Pb was often found to be the most abundant trace metal in urban
atmospheres (Tables 6A2a-c, PM AQCD 1996) at quarterly-average concentrations in the range
of 0.1 to 1.0 |ig/m3. Currently, ambient air Pb concentrations for most U.S. urban areas are in
the range of several ng/m3. The second most abundant trace metal, Zn, is typically present at
< 0.1 |ig/m3, whereas other transition elements (e.g., Ni, V) are typically < 10 ng/m3. Many
transition elements are currently nondetectable in most U.S. 24-h ambient air filter samples,
using X-ray fluorescence spectrometry.
The composition of PM10_2 5 particles has not been characterized to the same extent as
for PM2 5. In general, the inorganic composition of PM10_2 5 particles is dominated by crustal
particles; and, at times, there is also some evidence of combustion-related PM in some U.S.
locations. Photomicrographs obtained by scanning electron microscopy also indicate that large
numbers of biologic particles, such as pollen spores, are often present among coarse (PM10_25)
ambient air particles. The contributions of organic compounds and elemental carbon to PM10_2 5
particles are poorly known.
9.2 SYNTHESIS OF AVAILABLE INFORMATION ON PM-RELATED
HEALTH EFFECTS
The integrative synthesis of the latest available information on PM-related health effects
poses especially large challenges in view of:
• The unprecedented amount of new information generated since the 1996 PM AQCD,
which adds greatly to the complexity of any integrative assessment;
• Extensive new information available from epidemiologic studies, which reflects much
progress in addressing many research recommendations from the last review, but also
raises new issues or resurfaces issues earlier thought to have been adequately addressed
but which remain important in interpreting the body of epidemiologic evidence and the
characterization of its strengths and limitations;
• Much new information from dosimetric and toxicologic studies, which makes notable
progress toward identifying and exploring potential mechanisms of action and
characteristics of PM that may underlie health effects observed in experimental studies,
but still leaving open many issues to be more fully addressed in the future.
9-5
-------
Thus, despite substantial progress, challenges remain in integrating these different types of
evidence into a coherent synthesis.
As discussed in Section 8.1.4, concepts underlying an integrative assessment of statistical
associations reported in epidemiologic public health studies have been discussed in numerous
publications, from the historic publication by Hill (1965) to the most recent report by the U.S.
Surgeon General on the health consequences of smoking (Centers for Disease Control and
Prevention, 2004). All such discussions recognize that making causal inferences based on such
associations requires expert judgment, and criteria to aid such judgments generally derive from
those put forward earlier by Hill. Such criteria are not intended to serve as a checklist or a set of
rigid rules of evidence, but rather as a means of organizing an evaluation of the evidence to
facilitate reaching such judgments and conclusions. The criteria used in this assessment are
generally consistent with those defined in the Surgeon General's report and include the
following:
• Strength of association., which includes "the magnitude of the association and its
statistical strength."
• Consistency, which refers to the "persistent finding of an association between exposure
and outcome in multiple studies of adequate power, and in different persons, places,
circumstances, and times." This criterion serves to address issues related to potential
confounding, which in this assessment are separately considered in a discussion of the
robustness of the associations to the inclusion of potential confounding factors.
• Temporality, which most simply refers to "the occurrence of a cause before its purported
effect." In this assessment, temporality is more broadly defined to include consideration
of lag periods between exposure and effect.
• Biologic gradient, or concentration-response relationships, which refers to "the finding
of an increment in effect with an increase in the strength of the possible cause. . . ."
• Experiment, which refers to "situations where natural conditions might plausibly be
thought to imitate conditions of a randomized experiment, producing a 'natural
experiment' whose results might have the force of a true experiment."
• Coherence and plausibility, which in combination address the idea "that a proposed
causal relationship not violate known scientific principles, and that it be consistent
with experimentally demonstrated biologic mechanisms and other relevant data . . . ."
(Centers for Disease Control and Prevention, 2004, pp. 21-23).
9-6
-------
Section 9.2 is organized so as to first address the question of whether there is continued
support for considering fine and coarse thoracic PM as separate subclasses of PM based on
atmospheric science, air quality, exposure, and dosimetric information. Next, the strengths and
limitations of epidemiologic evidence are evaluated, taking into account the criteria outlined
above, including the strength and robustness of the reported associations; assessment of the
consistency or general concordance of study results and consideration of potential reasons for
observed differences; information related to lags and concentration-response relationships; and
information from so-called intervention studies of "natural" or "found" experiments. Looking
beyond the epidemiologic evidence, consideration is then also given to toxicological and other
information bearing on the biological plausibility and coherence of the PM-effects associations
observed in the epidemiologic studies to make causal inferences with regard to different
categories of health effects (cardiovascular, respiratory, etc.) and to reach conclusions regarding
the extent to which observed effects can be attributed to ambient fine and coarse thoracic PM,
acting alone and in combination with other pollutants. This is followed by discussion of
evidence regarding various risk factors (e.g., pre-existing disease and age-related factors) to
reach conclusions as to which susceptible and vulnerable subpopulations are most likely to be at
risk for health effects related to fine and coarse thoracic PM. Finally, information on the
magnitude of susceptible subpopulations is discussed, to provide context for the consideration of
potential public health impacts of exposures to ambient fine and coarse thoracic PM in the U.S.
9.2.1 Fine and Coarse Particles as Separate Subclasses of PM Pollution
The question of whether fine and coarse particles should continue to be considered as
separate subclasses of ambient PM is addressed below, drawing upon information and
assessments found primarily in Chapters 2, 3, 5, and 6 of this document related to the physics
and chemistry of particle pollution, the measurement of airborne particles, relationships between
ambient PM concentrations and population exposure, and PM dosimetry. The focus here is on
whether the newly available science in these areas continues to support consideration of fine and
coarse thoracic PM separately in the context of the Agency's periodic review of the PM
NAAQS, and if so, on appropriate indicators for these subclasses of PM.
9-7
-------
The primary focus in the last review was on thoracic particles (with PM10 defined as the
index for regulatory purposes) and on the question of whether fine and coarse thoracic particles
should be addressed by separate standards with different indicators. The 1996 PM AQCD noted
that the PM10 indicator was established as a result of the 1987 PM NAAQS review, which
concluded that the indicator for primary standards should represent those particles small enough
to penetrate to the thoracic region (including the tracheobronchial and pulmonary regions) of the
lower respiratory tract and should generally exclude particles that deposit only in the
extrathoracic region (the latter being particles previously included in the original TSP indicator).
The PM10 cut-point closely matches the definition for thoracic PM given by the American
Conference of Government and Industrial Hygienists (1994), as shown in Chapter 2 (Figure 2-6).
As discussed in the 1996 PM AQCD, the natural division of ambient PM into fine particles
and coarse particles is based on the recognition that "the fine and coarse modes originate
separately, are transformed separately, are removed separately, and are usually chemically
different. . . ." (Whitby, 1978). Consistent with this distinction, the 1996 PM AQCD stated that
the evidence indicates that "it would be appropriate to consider fine and coarse particles as
separate subclasses" of PM pollution. This conclusion was based on various considerations:
• Differences in formation processes and sources of fine and coarse thoracic particles,
as well as differences in chemical and physical properties, atmospheric residence times
and distances transported in the atmosphere;
• Resulting differences in patterns of ambient population exposures to fine and coarse
thoracic particles;
• Evidence from dosimetric studies showing differences in the fractions inhaled,
deposited, and/or retained in various regions of the respiratory tract for fine versus
coarse thoracic particles; and
• Evidence from health studies leading to conclusions that fine particles are more
strongly associated with more serious health effects and that chemical components
likely to have higher relative toxicity occur primarily in the fine fraction.
The evidence available in the last review strongly focused on particle size as the basis for
distinguishing between these essentially different classes of PM pollution. A cut point of 2.5 jim
was chosen for use in a new dichotomous sampler in the mid-1970s, when it was recognized that
within the range of about 1 to 3 |im there was no unambiguous definition of the appropriate cut
9-8
-------
point for the separation of the overlapping fine and coarse particle modes. Subsequent
epidemiologic studies of fine particles available in the last review, e.g., the Harvard Six City
Study and the American Cancer Society (ACS) cohort study, were based on the continued use of
the 2.5 |im cut point. During the last review, EPA gave consideration to a cut point of 1 |im as
an alternative to the 2.5 jim cut point as a basis for a fine particle standard. In so doing, EPA
took into account published size distributions that showed considerable variability in the
intermodal range of about 1 to 3 |im, including, for example, distributions from Philadelphia,
Phoenix, and Los Angeles (shown in Figure 2-9). Very little mass is seen in the intermodal
region in Philadelphia; in Phoenix, the coarse mode can be seen to extend to below 1 jim; and in
Los Angeles, a droplet mode, comprising the upper end of the fine mode, occurs under high
relative humidity conditions (usually associated with very high fine particle concentrations) and
extends above 2.5 jim.
EPA's decision to select a nominal cut-point of 2.5 jim in the last review reflected a
number of considerations. Available epidemiologic studies of fine particles were-based largely
on PM2 5 since PMX had not been widely monitored. Further, while it was recognized that
using PMj as an indicator of fine particles would exclude the tail of the coarse mode in some
locations, in other locations it would miss a portion of the fine PM, especially under high
humidity conditions, that would result in falsely low fine PM measurements on days with some
of the highest fine PM concentrations. The selection of a 2.5 jim cut point reflects the regulatory
importance that was placed on defining an indicator for fine particle standards that would more
completely capture fine particles under all conditions likely to be encountered across the United
States, especially when fine particle concentrations are likely to be high, while recognizing that
some small coarse-mode particles would also be captured by PM2 5 monitoring.
In selecting an indicator for coarse thoracic particles in the last review, EPA concluded that
the available dosimetric evidence continued to support the use of the same nominal upper cut-
point of 10 |im that had previously been selected as the basis for the standards set in 1987.
While recognizing that this cut point is on a part of the size distribution curve where the
concentration is changing rapidly, such that the amount of PM collected is sensitive to small
changes in the effective cut point of the sampler, it still represents the most appropriate cut point
to be used as the basis for an indicator of thoracic particles. EPA's decision in the last review to
9-9
-------
retain PM10 as an indicator for standards to address coarse particles, rather than an indicator that
would generally exclude fine particles (e.g., PM10_2 5), was based largely on the limited
epidemiologic studies and air quality data specifically available for coarse thoracic particles
beyond that which could be inferred or derived from PM10 studies in areas dominated by
coarse particles.1
As a consequence of these decisions made by EPA in the last PM NAAQS review, a
national PM25 monitoring network was established that has provided extensive air quality data
on PM2 5 and, by difference between co-located PM10 and PM25 monitors, more limited data
on PM10_2 5. The availability of such air quality data has prompted the increased use of PM2 5 and,
to a lesser degree, PM10_2 5 as indicators in new epidemiologic studies, as well as increasing focus
on these PM size fractions in other types of studies (exposure, dosimetry, toxicology, etc.).
In considering the distinctions between fine and coarse thoracic particles based on
currently available information, the following discussion builds upon the most salient key
findings from the previous PM NAAQS reviews, while updating and integrating key findings
and conclusions from the newly available studies assessed in earlier chapters of this document.
9.2.1.1 Physics and Chemistry Considerations
Since the last PM NAAQS review, the physical and chemical properties of fine and coarse
particles have become better understood. Nonetheless, the fundamental concept of the natural
division of thoracic particles into somewhat overlapping ranges of fine and coarse particles,
with a minimum in the mass distribution between 1 and 3 jim, as illustrated by the idealized
distribution shown in Figure 9-1, remains unchanged. Improved measurement techniques have
provided additional information that refines the general characterization of particles below
~0.1 jim diameter (i.e., ultrafine particles) from a single mode to a bi-modal structure. Thus,
fine particles are now divided into three modes: a nucleation mode, an Aitken mode, and an
accumulation mode. Nucleation mode applies to newly formed particles that have had little
chance to grow by condensation or coagulation. Aitken mode particles are also recently formed
1 As discussed in Chapter 1, subsequent litigation resulted in the court finding the use of PM10 as an indicator for
coarse-mode particles (in conjunction with PM2 5 standards) to be arbitrary, since PM10 includes all fine particles;
the court remanded this aspect of EPA's 1997 decision to the Agency for further consideration.
9-10
-------
0
1 ' I ' '"I
0.001 0.01
Nucleation Mode
1 ' I ' '"I ' ' ' I ''"I
0.1 1
Particle Diameter, Dp (|jm)
Accumulation Mode
Aitken Mode
V
10
Coarse Mode
100
Fine Particles
Ultrafine Particles
Coarse Particles
Figure 9-1. An idealized size distribution, as might be observed in traffic, showing fine
and coarse particles and the nucleation, Aitken, and accumulation modes that
comprise fine particles. Also shown are the major formation and growth
mechanisms of the four modes of atmospheric particles.
particles that are still actively undergoing coagulation but have grown to larger sizes. The
accumulation mode applies to the final stage, as particles originally formed as nuclei grow to a
point where growth slows down, such that accumulation-mode particles normally do not grow
into the coarse particle size range. However, during conditions of high relative humidity,
hygroscopic accumulation mode particles grow in size, increasing the overlap of fine and
coarse particles. The accumulation mode may split into a hygroscopic droplet mode and a
nonhygroscopic condensation mode. In addition, gas-phase pollutants may dissolve and react in
the particle-bound water (PBW) of hygroscopic particles, e.g., forming more sulfate or nitrate,
9-11
-------
leading to particle growth beyond the original size even after removal of PBW. These three
modes, which comprise fine particles (sometimes called fine-mode particles), are formed
primarily by combustion or chemical reactions of gases that yield products with low saturated
vapor pressures. Fine particles include primary PM (metals, black or elemental carbon, and
organic compounds) and secondary PM (sulfate, nitrate, ammonium and hydrogen ions, and
organic compounds).
The coarse mode refers to particles formed by mechanical breakdown of minerals, crustal
material, and organic debris. The composition includes primary minerals and organic material.
The coarse mode may also include sea salt, nitrate formed from the reaction of nitric acid with
sodium chloride, and sulfate formed from the reaction of sulfur dioxide with basic particles.
The accumulation mode and the coarse mode overlap in the region between 1 and 3 jim (and
occasionally over an even larger range). In this intermodal region, the chemical composition of
individual particles can usually, but not always, allow identification of a source or formation
mechanism and so permit identification of a particle as belonging to the accumulation or coarse
mode.
Since the 1996 PM AQCD, several studies have sought to better characterize particles
present in the intermodal region (e.g., indexed by PM2 5 - PMX) and to assess the importance of
coarse mode particles present in the intermodal region on associations reported in epidemiologic
studies. For example, studies conducted in Phoenix suggested that inclusion of such particles
would not likely affect reported associations with PM2 5. Studies using Salt Lake City data
suggest that coarse particles due to windblown dust are less toxic than PM10 present during
non-windblown dust events. Studies in Spokane show that windblown dust contributes to PM2 5
but not to PMj. Thus, the inclusion of days with high windblown dust events could obscure
associations with fine particles if PM25 were used as the indicator.
Natural processes, such as the suspension of soil dust by wind, produce few particles below
1 |im in diameter. However, studies now suggest that biological material, although originally in
the coarse mode, may deteriorate or fragment and produce particles in the fine-particle size
range. Thus, fragments of pollen, endotoxins, and other biological material may be found in the
fine-particle size range. Progress has also been made in understanding the semivolatile
components of PM (particle-bound water, ammonium nitrate, and semivolatile organic
9-12
-------
compounds) and new techniques have been developed to measure the semivolatile components
of mass, either separately or included with the nonvolatile component. The many organic
compounds formed in the atmospheric reactions of biogenic and anthropogenic hydrocarbons,
including condensible species that form organic particles, are now better understood, and
progress has been made in measurement of carbonaceous particles.
This progress has helped to enhance our understanding of ambient aerosol components and
interrelationships among them that may contribute to ambient PM-related effects. Of much
importance, for example, is emerging new evidence related to the role of PBW and associated
submicron PM constituents serving as vectors by which water soluble gases (e.g., SO2), short-
lived reactive species (e.g., peroxides), and organic species (e.g., formaldehyde) present in
atmospheric aerosol mixes can be delivered in enhanced proportions to lower regions of the
respiratory tract. The importance of nonbiological ambient PM components serving as carriers
or vectors enhancing deposition of bioaerosols (e.g., allergen-laden pollen fragments and
endotoxins) in the lower respiratory tract has also been noted. It is notable that rather direct
evidence has also been obtained which demonstrates adherence of allergen-laden pollen
cytoplasm fragments to diesel particles, providing a likely mechanism by which diesel PM may
act to concentrate bioaerosol materials and to increase their focal accumulation in lower regions
of the respiratory tract.
The 1996 PM AQCD listed properties of fine and coarse particles. Because of the
increasing interest in ultrafine particles and additional information on their properties, this
current document provides new information on the chemical and physical properties of ultrafine
and accumulation-mode fine particles and coarse particles, as shown in Table 9-1. As shown,
ultrafine and accumulation-mode particles share similar formation processes and mechanisms,
sources, and compositions. However, their fate and transport are quite dissimilar. In the
atmosphere, ultrafine particles are removed largely by coagulation with other ultrafine particles
(or accumulation-mode particles) and growth into the accumulation mode. Accumulation-mode
particles are removed largely by serving as cloud-condensation nuclei that form cloud droplets
and rain out, and to a lesser extent, by dry deposition. Coarse particles, however, are removed
from the atmosphere rather rapidly by gravitational settling. With regard to the volume or mass
of ambient PM, accumulation-mode and coarse particles both contribute appreciably in most
9-13
-------
TABLE 9-1. COMPARISON OF AMBIENT PARTICLES,
FINE PARTICLES (ultrafine plus accumulation-mode) AND COARSE PARTICLES
Fine
Ultrafine
Accumulation
Coarse
Formation
Processes:
Formed by:
Composed
of:
Solubility:
Sources:
Combustion, high-temperature
processes, and atmospheric reactions
Nucleation
Condensation
Coagulation
Sulfate
Elemental carbon
Metal compounds
Organic compounds
with very low
saturation vapor
pressure at ambient
temperature
Probably less soluble
than accumulation
mode
Combustion
Atmospheric
transformation of
SO2 and some
organic compounds
High temperature
processes
Atmospheric Minutes to hours
half-life:
Removal
Processes:
Travel
distance:
Grows into
accumulation mode
Diffuses to raindrops
< Ito 10s of km
Condensation
Coagulation
Reactions of gases in or
on particles
Evaporation of fog and cloud
droplets in which gases have
dissolved and reacted
Sulfate, nitrate, ammonium,
and hydrogen ions
Elemental carbon
Large variety of organic
compounds
Metals: compounds of Pb, Cd,
V, Ni, Cu, Zn, Mn, Fe, etc.
Particle-bound water
Largely soluble, hygroscopic,
and deliquescent
Combustion of coal, oil,
gasoline, diesel fuel, wood
Atmospheric transformation
products of NOX, SO2, and
organic compounds,
including biogenic organic
species (e.g., terpenes)
High-temperature processes,
smelters, steel mills, etc.
Days to weeks
Forms cloud droplets and
rains out
Dry deposition
100s to 1000s of km
Break-up of large solids/droplets
Mechanical disruption (crushing,
grinding, abrasion of surfaces)
Evaporation of sprays
Suspension of dusts
Reactions of gases in or on particles
Suspended soil or street dust
Fly ash from uncontrolled combustion
of coal, oil, and wood
Nitrates/chlorides/sulfates from
HNO3/HC1/SO2 reactions with
coarse particles.
Oxides of crustal elements
(Si, Al, Ti, Fe)
CaCO3, CaSO4, NaCl, sea salt
Pollen, mold, fungal spores
Plant and animal fragments
Tire, brake pad, and road wear debris
Largely insoluble and nonhygroscopic
Resuspension of industrial dust and
soil tracked onto roads and streets
Suspension from disturbed soil (e.g.,
farming, mining, unpaved roads)
Construction and demolition
Uncontrolled coal and oil combustion
Ocean spray
Biological sources
Minutes to hours
Dry deposition by fallout
Scavenging by falling rain drops
< 1 to 10s of km (small size tail,
100s to 1000s in dust storms)
Source: Adapted from Wilson and Suh (1997).
9-14
-------
areas, with very little contribution from ultrafine particles. With regard to particle surface area,
however, ultrafine and accumulation-mode particles both contribute appreciably, with very little
contribution from coarse particles. To the extent that inhaled PM may carry chemicals or
reactive species on their surfaces, these smaller size fractions may have an additional dimension
to their toxicity (in terms of surface chemical bioavailability) that is not found with coarse PM.
Ultrafine, accumulation mode, and coarse particles also behave differently with regard to
exposure and dosimetric considerations, as discussed below, as well as in toxicologic and
epidemiologic studies, as discussed in subsequent sections of this chapter.
9.2.1.2 Exposure-Related Considerations
The critical relationship to be considered is that between ambient PM concentrations and
personal exposures to ambient PM (ambient PM refers to that PM measured at a community
monitoring site, or the average over several such sites). It is convenient to consider two aspects
of this relationship. One important aspect is the relationship between the ambient concentration
measured at one or more monitoring sites and the distribution of outdoor concentrations across
an area (e.g., outside homes and other microenvironments). This relationship will depend in part
on the uniformity with which the PM indicator of interest is distributed across the community.
For time-series epidemiologic analyses of associations between 24-h concentrations of ambient
PM and health endpoints, the relevant measurement of this relationship is the day-to-day
correlation of 24-h concentration values at various monitoring sites in the community. For long-
term epidemiologic analyses, the variation in the seasonal or yearly average at various sites in
the community is the relevant parameter. Much new information on the distribution of PM25
and PM10_2 5 concentrations across cities is available from the new monitoring networks and is
presented in detail in Chapter 3. In general, PM25 is more evenly distributed than PM10_25 in
terms of both daily/seasonal/yearly averages and day-to-day correlations, although there are
significant differences among cities. Little is known about the spatial distribution of ultrafine
particle concentrations, except that their concentrations are highest in and near heavy traffic
areas and rapidly fall off with distance from traffic due to coagulation and dispersion. Because
of their rapid growth into the accumulation mode, their concentrations are probably highest near
9-15
-------
sources such as traffic. Thus, they likely have a more heterogeneous distribution across a
community than accumulation-mode particles.
The second aspect is the relationship between the concentration of PM outdoors and the
concentration of that outdoor PM which has infiltrated into the home or other microenvironment,
characterized by an infiltration factor which is a function of particle size. Personal exposure
includes a nonambient component due to PM generated indoors or by personal activities, a
component which does not appear to correlate well with outdoor ambient PM concentrations.
As shown in Figure 9-2, the infiltration factor depends on the air exchange rate, but for a given
ventilation condition, the infiltration rate is high for accumulation-mode particles and decreases
to low levels with decreasing size within the ultrafme range and with increasing size within the
coarse-mode range. Exposure-related relationships for the three particle size classes are
summarized in Table 9-2.
o
ra
LL
C
O
1.1
1.0 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0.0
0.1
\ \
Summer Fall
CD
o
CO •st-
p p
odd
CN co M-
p p p
odd
co •<-
o cl
9 oo
CN
o
uo
co ^ LO
cp ci ci
csi co ^
odd
co
CN
^J- uo CD
co ^r LO
o
I
CD
Particle Diameter (jjm)
Figure 9-2. Geometric mean infiltration factor (indoor/outdoor ratio) for hourly
nighttime, nonsource data for two seasons. Box plots of air exchange
rates are shown as inserts for each plot (Boston, 1998).
Source: Long etal. (2001).
9-16
-------
TABLE 9-2. EXPOSURE-RELATED RELATIONSHIPS FOR
PARTICLE SIZE FRACTIONS
Ultrafine Accumulation-Mode Thoracic-Coarse
Even distribution Probably not Frequently Seldom
across city
Site-to-site correlation Probably low Frequently high Frequently low
Infiltration factor Decreasing from high High Lower than accumulation
(for a given to low as particle size mode and decreasing
exchange rate) decreases from 0.1 \\m with increasing size
In most community time-series studies and long-term cohort studies, the ambient
concentration is used as a surrogate for personal exposure to ambient PM (ambient exposure).
For the ambient concentration to be a satisfactory surrogate, there must be a reasonable
correlation between ambient concentration and ambient exposure, as appears to be the case for
fine particles (PM25). However, because of the lower and more variable infiltration factors for
ultrafine and coarse particles and their less even distribution and lower site-to-site correlations
across the community, it is likely that their ambient concentrations will be a somewhat poorer
surrogate for their ambient exposures than is the case for PM2 5. Nonambient PM may also be
responsible for health effects. However, since the ambient and nonambient components of
personal exposure are independent, the health effects due to nonambient PM exposures generally
will not bias the risk estimated for ambient PM exposures.
9.2.1.3 Dosimetric Considerations
The fraction of inhaled particles that are deposited in the various regions of the lung
depends on the particle size, the breathing route (nasal or oral), the breathing frequency (breaths
per minute), the volume of air inhaled (tidal volume), the anatomy of the respiratory tract of the
individual, and deposition mechanisms (diffusion, sedimentation, impaction) which affect
different-sized particles to varying extents. Because of differing effects of all the above-noted
factors, calculated fractional deposition patterns in various respiratory tract regions can vary
considerably for particles in different size ranges. The fractional depositions in the extrathoracic
9-17
-------
(ET), tracheobronchial (TB), and gas exchange or alveolar (A) regions of the respiratory tract are
shown as a function of particle size in Figure 9-3 for nasal and oral breathing at two levels of
activity (resting and light exercise). Particles in the accumulation-mode size range generally
have very low deposition fractions, especially in the ET and TB regions, that are relatively
insensitive to breathing pattern or exercise. However, for nose breathing the deposition of larger
accumulation-mode particles in the ET region does increase with exertion. Thus, most
accumulation mode particles that enter the lungs are exhaled rather than deposited.
Ultrafine particles generally have much higher fractional depositions than accumulation
mode particles. However, the smaller nucleation-mode (< 0.01 jim) ultrafine particles behave
differently from the larger Aitken-mode (-0.01 to -0.1 jim) ultrafine particles. As particle size
decreases below 0.1 jim, the total deposition of particles increases, and the pattern of deposition
within the respiratory tract slowly moves proximally, i.e., toward the ET region. This shift in the
pattern of deposition is quite obvious for decreases in particle size below 0.01 jim where
A deposition fractions rapidly decline and the ET deposition fractions correspondingly increase.
The TB deposition fraction increases to a maximum near 3 nm. For the Aitken mode particles,
the deposition fraction for the A region increases with exertion whereas in the TB region it
decreases. Deposition fractions in the A region for particles less than -1 jim are relatively
insensitive to route of breathing.
The fractional deposition for coarse particles is even more complex. For both the A and
TB regions, the deposition fraction increases with particle diameter above -1 jim, reaches a peak
before the diameter reaches 10 jim, and then declines. The deposition fractions for the A and TB
regions are lower during nasal breathing because a large fraction of the coarse particles deposit
within the nose. For mouth breathing, the A and TB deposition fractions are higher than during
nasal breathing but not as high as those for the ultrafine mode during mouth breathing. For
mouth breathing, the deposition fractions for both the A and the TB regions are greater for
coarse particles than for accumulation-mode particles. Even for nose breathing, some coarse
particles, of a specific size, will have higher A and TB deposition fractions than accumulation
mode particles.
9-18
-------
c
o
5
u_
o
Q_
0)
Q
A region
• • Oral
<> D Nasal
O Resting
D Exercising
0.0
0.001
0.01 0,1
Particle Diameter (|jm)
10 25
Figure 9-3. Deposition fraction as a function of particle size for nasal and oral breathing
during rest and exercise: (a) extrathoracic (ET), (b) tracheobronchial (TB),
and (c) alveolar (A) regions. Deposition estimates shown here were calculated
with the ICRP model and were also shown in Figures 6-16 and 6-17 along with
similar results from the MPPD model. The estimates below 0.01 um are
uncertain but are shown to indicate trends. Note the different scale for the
ET region.
9-19
-------
In general, given these complex deposition patterns, there are no sharp cut points that
clearly distinguish between particle size ranges with relatively high versus relatively low
fractional deposition rates. For example, in the ET region, particles ranging in size from roughly
0.01 |im on up to ~1 |im (for nasal breathing) to over 3 jim (for oral breathing) exhibit relatively
low fractional deposition rates. For the TB region, relatively low rates are exhibited by particles
ranging in size from roughly 0.05 jim up to ~2 jim (for oral breathing) to over 10 jim (for nasal
breathing). For the A region, relatively low rates are exhibited not only by particles from ~0.1 to
~1 |im but also for particles in the low end of the ultrafme size range and as well as for particles
in the upper end of the coarse-mode range. Thus, while differences in dosimetric properties
continue to support the general division of ambient particles into fine and coarse fractions,
dosimetric considerations now also suggest that further distinctions can be made between
subclasses (ultrafme, accumulation-mode) within the range of fine particles.
9.2.1.4 Summary and Conclusions
The distinctions articulated in the last review between fine and coarse ambient particles
(as indicators of of fundamentally different sources and composition, formation mechanisms,
transport, and fate) remain generally unchanged. However, some important advances have been
made in our understanding of such distinctions, especially with regard to characteristics of
particles below ~0.1 jim in diameter (ultrafme particles). In particular, whereas fine particles
were previously characterized in two modes, they are now characterized in terms of three modes,
with two modes, nucleation and Aiken, observed in the ultrafme particle size range. Distinctions
among these modes allow for more differentiation in characterizing properties of fine particles.
Also, progress has been made in better understanding the size distribution of biological
materials. While previously understood mainly to be present in the coarse particle size range,
newly available information indicates that such particles (e.g., pollen grains, endotoxins) may
fragment or deteriorate into the fine particle size range. This information expands our
understanding of the types of particles that can occur in particular within the intermodal size
range of ~1 to ~3 jim. New information indicates that atmospheric particles may carry
components of biological particles (allergens and endotoxin) to the lower respiratory tract and
9-20
-------
confirms earlier suggestions that water soluble gases can dissolve in particle-bound water and be
delivered in enhanced proportions to the lower respiratory tract.
Data now available from the new national PM2 5 monitoring network and speciation sites
have allowed for better assessments of exposure-related considerations which broaden but do not
fundamentally change our understanding of the substantial differences between fine particles in
the accumulation mode and coarse particles. Relationships between ambient PM concentrations
and personal exposure to ambient PM are now better understood, primarily for fine particles, but
also to a more limited degree for coarse particles. For example, new data reinforce our earlier
understanding that ambient concentrations of fine particle mass (measured as PM25) are typically
more highly correlated and/or are more uniform across community monitors within an urban
area than are coarse particle mass concentrations (measured as PM10_2 5), although in some areas
the differences are much less pronounced than in others. More limited data and knowledge of
the behavior of ultrafme particles suggest that spatial distributions of their concentrations (which
decrease quickly from peak levels around major highways) are more variable than those of
accumulation mode particles. Concentrations of coarse particles (which decrease quickly from
peak levels around primary sources) are also more variable spatially than accumulation-mode
particles. Further, new studies reinforce our earlier understanding that, for a given ventilation
condition, fine particles generally infiltrate indoors much better than do either coarse or ultrafme
particles. Thus, central site ambient concentration measurements are a better surrogate for
population exposure to accumulation-mode fine particles, measured as PM2 5, than for either
coarse or ultrafme particles, although there may be large differences in PM2 5 concentrations
across many urban areas. These observations about the behavior of ultrafme particles and coarse
particles are based on far more limited data, highlighting a need for further research on these
particle size ranges.
Newly available dosimetry information continues to reinforce important distinctions
between fine and coarse particles, and submodes within fine particles, with regard to deposition
patterns within the respiratory tract. In general, while deposition patterns within the major
respiratory tract regions as a function of particle size are complex and dependent in varying
degrees on breathing route and ventilation levels, accumulation-mode particles exhibit distinctly
lower fractional deposition rates in any of the major respiratory tract regions than do ultrafme or
9-21
-------
coarse particles on average. The factional deposition of ultrafine, accumulation-mode, and
coarse thoracic particles in the ET, TB, and A regions show complex variations with increasing
levels of activity, associated increases in breathing rate, and associated increased oral nasal/oral
breathing. Thus, it is difficult to characterize more specific size fractions within the range of
thoracic particles that would clearly delineate ranges of relatively high and relatively low
fractional deposition across all respiratory tract regions.
Overall, then, the above considerations reinforce the recommendation made in the 1996
PM AQCD that fine and coarse particles be considered as separate subclasses of PM pollution.
Advances in our understanding of the characteristics of fine and coarse particles continue to
support the use of particle size as an appropriate basis for distinguishing between these
subclasses, even as progress is being made in understanding their composition. The
considerations that led to the selection of a nominal upper cut point of 2.5 jim in the last review
remain relevant, and lead again to the conclusion that an upper cut point of 2.5 jim remains
appropriate as the basis for a regulatory indicator of fine particles, in conjunction with a
regulatory indicator of coarse particles defined by a nominal lower cut point of 2.5 and a
nominal upper cut point of 10 |im.
9.2.2 Assessment of Epidemiologic Evidence
Based on the PM epidemiologic evidence available at the time, the 1996 PM AQCD,
arrived at the following overall conclusions:
"The evidence for PM-related effects from epidemiologic studies is fairly strong,
with most studies showing increases in mortality, hospital admissions, respiratory
symptoms, and pulmonary function decrements associated with several PM indices.
These epidemiologic findings cannot be wholly attributed to inappropriate or incorrect
statistical methods, misspecification of concentration-effect models, biases in study
design or implementation, measurement errors in health endpoint, pollution exposure,
weather, or other variables, nor confounding of PM effects with effects of other
factors. While the results of the epidemiology studies should be interpreted
cautiously, they nonetheless provide ample reason to be concerned that there are
detectable human health effects attributable to PM at levels below the current
NAAQS." (U.S. Environmental Protection Agency, 1996, p. 13-92).
The 1996 PM AQCD went on to state further that, while the epidemiological studies
indicate increased health risks associated with exposure to PM, alone or in combination with
9-22
-------
other air pollutants, the role of PM as an independent causal factor has not been completely
resolved, based on the available studies using multiple air pollutants as predictors of health
outcomes (U.S. Environmental Protection Agency, 1996, p. 13-92).
In assessing the strengths and limitations of the extensive body of new epidemiologic
evidence of associations between health effects and fine and coarse thoracic PM, information
discussed in this section is drawn primarily from Chapter 8, as well as from Chapter 5 of this
document. The information is considered here in relation to several criteria noted at the outset of
Section 9.2: (1) the strength of reported associations, in terms of magnitude, statistical
significance, and statistical power/precision of effects estimates; (2) the robustness of reported
associations to the use of alternative model specifications, potential confounding by
co-pollutants, and exposure misclassification related to measurement error; (3) the consistency or
general concordance of findings in multiple studies of adequate power, and in different persons,
places, circumstances and times; (4) temporality, in terms of lag periods between exposure and
observed effects; (5) the nature of concentration-response relationships; and (6) information
from so-called natural experiments or intervention studies as to the extent to which reductions in
PM-related air pollution have been observed to be associated with improvements in health
measures. The body of epidemiologic evidence is further considered in the following section in
terms of its coherence within itself and in relation to toxicologic findings derived from
controlled exposure studies which, overall, provide insights on the plausibility of reported PM-
related health effects reflecting causal relationships.
Many recent epidemiologic studies have built upon what was previously known, showing
statistically significant associations of various ambient PM indicators with a variety of
cardiovascular and respiratory health endpoints, including mortality, hospital admissions,
emergency department visits, other medical visits, respiratory illness and symptoms,
physiological or biochemical changes related to the cardiovascular system, and physiologic
changes in pulmonary function. Associations have been consistently observed between short-
term PM exposures to certain PM size fractions and one or more of these endpoints; and long-
term PM exposures have been associated with increased risk of mortality, development of
respiratory disease, and changes in lung function. As summarized in Chapter 8, Appendices 8A
and 8B, epidemiologic studies have been conducted in areas across the U.S. and Canada, as well
9-23
-------
as in Mexico and South America, Europe, Asia and Australia; and various methods have been
used to measure ambient PM concentrations. Considering the evidence from the full body of
epidemiologic studies using various PM indicators, the available findings demonstrate well that
human health outcomes are associated with ambient PM. Discussions in the following sections
focus primarily on studies conducted in the U.S. and Canada using various mass measurements
of thoracic particles (e.g., PM10, PM2 5, PM10_25) and source-oriented PM analyses.
9.2.2.1 Strength of Epidemiologic Associations
As quoted above, the 1996 PM AQCD concluded that the epidemiologic evidence for
cardiorespiratory effects was "fairly strong" considering both magnitude and statistical
significance of results available at that time. At that time, it was recognized that the relative risk
estimates from time-series studies were generally small in magnitude. Since then, however, the
results of recent reanalyses to address GAM-related issues has led to smaller effect estimates in
some cases (as discussed in Chapter 8). In contrast with the marked increase in health effects
observed during historic episodes of very high air pollution levels, relatively small effect
estimates would generally be expected with current ambient PM concentrations in the United
States. The etiology of most air pollution-related health outcomes is multifactorial, and the
impact of ambient air pollution exposure on these outcomes may be small in comparison to that
of other risk factors (e.g., smoking, diet).
9.2.2.1.1 Short- Term Exposure Studies
Many new epidemiologic studies have built upon what was available in the 1996 PM
AQCD. These include several multicity studies that can provide more precise estimates of
effects than individual city studies, offer consistency in data handling and modeling, allow for
systematic evaluation of geographic patterns in effects, and clearly do not suffer from potential
omission of negative findings due to "publication bias." In addition, there are studies of new
health indices (e.g., physician visits) and cardiovascular health outcomes, analyses that provide
insight into the sensitivity of PM effects to alternative statistical modeling, new assessments on
the potential for confounding by gaseous co-pollutants, and new evidence from "found
experiments" that evaluate improvements in health seen with reductions in air pollution levels.
9-24
-------
The results from key United States and Canadian studies on short-term PM exposure for
several commonly-used health outcomes — mortality, hospitalization and medical visits — are
depicted in Figures 9-4 and 9-5. While recognizing that epidemiologic studies of short-term air
pollution exposures have also evaluated other health outcomes (e.g., respiratory symptoms,
cardiovascular health indicators, lung function changes), these figures illustrate results for a few
major health outcome categories commonly used in PM time-series epidemiologic analyses.
It should be noted that, while nearly all effect estimates shown in Figure 9-4 and 9-5 were
derived from single-city time-series analyses, a few (among those with the narrowest 95%
confidence intervals) represent results aggregated across multiple cities. See pertinent references
(indicated by bold font in the figure captions) and discussion of such studies presented in
Chapter 8 for more details on them.
It should also be noted that the results are drawn from studies using one or more of the
three major PM mass indicators (PM10, PM2 5, or PM10_2 5) that either did not use GAM or were
reanalyzed to address GAM-related questions. Single-pollutant (PM only) results are presented
here for purposes of comparison across studies, and it is noted that multipollutant model results
are presented and discussed in Chapter 8 (see especially Section 8.4.3). The results of models
using different lag periods from time-series epidemiologic studies are also presented and
discussed in Chapter 8 (see Section 8.4.4). For each health outcome, the results are presented in
Figures 9-4 and 9-5 in order (from left to right) of decreasing study power, using as an indicator
the product of the number of study days and number of health events per day.
To be consistent with the rest of this document, the effect estimates are presented in
Figures 9-4 and 9-5 using standardized PM increments to allow for comparison across studies.
As described in Section 8.1.1, current air quality data distributions were used to select
increments of 50 |ig/m3 for PM10 and 25 |ig/m3 for both PM2 5 and PM10_2 5 as representative of
realistic high-to-low ranges of concentrations for most U.S. communities. Alternatively, if the
effect estimates were presented per unit mass for each PM indicator, the estimates for PM2 5
and PM10_2 5, relative to those for PM10, would be twice as large as those depicted in these figures.
On a unit mass basis, the effect estimates for both PM2 5 and PM10_2 5 are generally larger than
those for PM10, which is consistent with PM2 5 and PM10_2 5 having independent effects.
9-25
-------
VO
to
a\
50 -
45 -
40 -
0) 35 -
27 1
26|28|
1
.
32 2
' Jf
t
33 3
24
3
29
27
K
'it
31
0
CV Mortality
29
;
24
10
> .
3
19 1
14 , 23
ilflt
I
3
7
18
|
2
I
6 30
1
PM2.5
33
31
32 23
' m
24
30
27
1.
CV Mortality
i Total Mortality
Resp.
Mortality
PMlO-2.5
24
23
31
10
36
I {,.
19
9 14
HI'
T
1
Resp
Mortal!
2
20
3
fli
II
52
t
2
26
Total Mortality
y
Total Mortality
2
32
8
I
92) •
24 1
2
3
33
cv"
4
23
34
I
Resp.
Mortality
Mortality
Figure 9-4.
Excess risk estimates for total nonaccidental, cardiovascular, and respiratory mortality in single-pollutant
models for U.S. and Canadian studies, including aggregate results from two multicity studies (denoted in bold
print below). PM increments: 50 ug/m3 for PM10 and 25 ug/m3 for PM25 and PM10_2 5. Results presented from
time-series studies that did not use GAM or were reanalyzed using GLM.
1. Dominici et al. (2003a), 90 U.S. cities
2. Moolgavkar (2003), Cook County
3. Kinney et al. (1995), Los Angeles
4. Schwartz (2003), Chicago
5. Ito and Thurston (1996), Cook County
6. Schwartz (2003), Pittsburgh
7. Styer et al. (1995), Cook County
8. Schwartz (2003), Detroit
9. Burnett and Goldberg (2003),
8 Canadian cities
10. Moolgavkar (2003), Los Angeles
11. Schwartz (2003), Seattle
12. Schwartz (2003), Minneapolis
13. Klemm and Mason (2003), St. Louis
14. Klemm and Mason (2003), Boston
15. Schwartz (2003), Birmingham
16. Schwartz (2003), New Haven
17. Chock et al. (2000), Pittsburgh (< 75 y.o.)
18. Chock et al. (2000), Pittsburgh (75+ y.o.)
19. Klemm and Mason (2003), Kingston-Harriman
20. Klemm and Mason (2003), Portage
21. Schwartz (2003), Canton
22. Schwartz (2003), Spokane
23. Ito (2003), Detroit
24. Fairley (2003), Santa Clara County
25. Schwartz (2003), Colorado Springs
26. Klemm and Mason (2003), Topeka
27. Tsai et al. (2000), Newark
28. Klemm and Mason (2003), Steubenville
29. Pope et al. (1992), Utah Valley
30. Tsai et al. (2000), Elizabeth
31. Tsai et al (2000), Camden
32. Lipfert et al. (2000), Philadelphia
33. Mar et al. (2003), Phoenix
Ostro et al. (2003), Coachella Valley
Klemm and Mason (2000), Atlanta
36. Ostro et al. (1995), Southern California
-------
VO
K)
70 -
"
60 -
.
.
•
50 -
Q)
re 40 -
^
(A
LLI 30 -
^
t/i
&• 20 -
(A
(A
U
% 1°:
LU
•
-10-
•
.
-20 -
PM,o
Is !
U
ro
o
•a
ro
Q
re
3
U
v>
re
>
o
re
.2 <
E
O >,
i_ . .c
0 -3 3
"> =1 o 0
Ł o c/) c/)
,2 ™ > >
"E o ° •-
ro ~ ~a "a ^ (
"sssl 5
fj?jljl
r
x E
8
1
-'
!j
f
<
Hospital
i »-
= <:
' !
!
<
Admissions
i
3
i
Ł1
o
Ł ."
O Q.
ro w
.- a:
t/j
0
_ a:
re
'z -i ^
c ro i «
o ^
3 ••-•
Is
>-
Q S.
n ro
0 i
c.
c >.
II
C Q. <
1 &•
\\
' .2 5
c a
o a
|Q7
O 1 0^
3 1 r
{H1
ra
L F n
" j
H
1
1
4
i T i
ED
Visit
1 1
CV Diseases
E
'
11
i
o
J
0.
a
o
o
f
'
Hospital
Admissions
E
s s.
E a>
Ł <
Ł•< 16
2 14rc
E. E
0> In
K <<•
9 15
' '
. . < '
1 1
7
ro
i
ro
>
3
3 ro
o o u.
> 7 Q 8
.2 T
•a
ro
O
4 '
1
ii ii
i
4
4
0
PM2.5
"-
3
O
(/)
ro
>
o
T3
O
o
5
E
'
>1
" IT
0
8
<
ED Hospital
<
Visit Admissions
1 1
>
o
ro
c
o
E
^
O
ro
'a.
[/)
0
o:
13
3
0
nc
Q- g
0 Ł
7 <
Q 1-1
Q. n
O
O
4 ".
H
COPD
(
4
i T i
ED Hospital
Visit Admissions
i i
PM10-2.5
1 1 1
S >13
2 .j o
Q. — re
Ł/) 3 «-
Q) 0 '^
-i 7 > jj;
1 ' O 1-
>
i
•i
i
'E
0
a
, a
Ł
i i
<
L
o ra
ro 0 n
i 7 Ł• oŁ
w o e 5:
-2 E =8
18 ^ °- JF 8
C t/) Q.
1> Ł ^ n
3 > o: 8
• Q o ^ re
I i: !T 7 E
• - — CO ^J, ^
8 I 8 Q 2
8 8
< ' 4 •
*
< '° * '
II II
ED Hospital Hospital
Visit Admissions Admissions
|
ii ii i
Resp. Diseases CV Diseases Resp. Diseases CV Diseases Resp. Dis.
Figure 9-5. Excess risk estimates for hospital admissions and emergency department visits for cardiovascular
and respiratory diseases in single-pollutant models from U.S. and Canadian studies, including
aggregate results from one multicity study (as denoted in bold below). PM increments: 50 ug/m3
for PM10 and 25 ug/m3 for PM2 5 and PM10_2 5. Results presented from time-series studies that did
not use GAM or were reanalyzed using GLM. PM effect size estimate (± 95% confidence intervals)
are depicted for the studies listed below.
1. Zanobetti and Schwartz (2003) 7.
U.S. 14 cities 8.
2. Linn et al. (2000), Los Angeles 9.
3. Moolgavkar (2003), Cook County 10.
4. Moolgavkar (2003), Los Angeles 11.
5. Schwartz and Morris (1995), Detroit 12.
6. Morris and Naumova (1998), Chicago
Burnett et al. (1997), Toronto
Ito (2003), Detroit
Stieb et al. (2000), St. John
Schwartz (1994), Detroit
Sheppard (2003), Seattle
Nauenberg and Basu (1999),
Los Angeles
13. Thurston et al. (1994), Toronto
14. Tolbert et al. (2000), Atlanta
15. Lipsett et al. (1997), Santa Clara County
16. Choudhury et al. (1997), Montreal
17. Delfino et al. (1997), Montreal
18. Delfino et al. (1998), Montreal
-------
In Figure 9-4, effect estimates for associations between mortality and PM are grouped both
by PM indicator (PM10, PM2 5, and PM10_2 5) and by mortality category (total nonaccidental,
cardiovascular or cardiorespiratory, and respiratory). Looking across the results with particular
focus on the more precise estimates, some general observations can be made:
• Almost all of the associations between PM10 and total mortality are positive and
over half are statistically significant, including most all of those with more precise
estimates. All associations reported between PM10 and cardiovascular and
respiratory mortality are positive. Most of the cardiovascular mortality associations
are also statistically significant, whereas most of the respiratory associations are
generally larger in size but less precise and not statistically significant; less precision
would be expected since respiratory deaths comprise only a small portion of total
nonaccidental mortality. The more precise effect estimates range from ~1 to 8%
increased risk of mortality per 50 |ig/m3 PM10; for the multeity studies, effect
estimates ranged from -1.0 to 3.5% per 50 |ig/m3 PM10.
• A similar pattern can be seen for PM2 5, though fewer studies are available; and the
effects estimates are generally somewhat less precise and less frequently statistically
significant. In particular, almost all of the PM25 associations with total mortality are
positive, although less than half are statistically significant. All PM25 associations
with cardiovascular and respiratory mortality are positive; and about half of the
cardiovascular associations, but none of the respiratory associations, are statistically
significant. The more precise effect estimates range from about 2 to 6% increased
risk of mortality per 25 |ig/m3 PM25 and ~1 to 3.5% per 25 |ig/m3 PM25 in
multicity studies.
• Still fewer studies have used PM10_2 5 measurements. The effect estimates are almost
all positive and similar in magnitude to those reported for PM2 5 and PM10, but few
reach statistical significance. Measurement error likely contributes to greater
uncertainty, reflected by wider confidence intervals, in effect estimates for PM10_2 5
than for PM2 5 and PM10.
The results for U.S. and Canadian studies are generally consistent with those presented
in Chapter 8 based on all available epidemiologic studies world wide. These results indicate
that there is substantial strength in the epidemiological evidence for association between PM10
and PM2 5 and mortality, especially for total and cardiovascular mortality, but also for respiratory
mortality. For PM10_2 5, the evidence for associations with mortality is more limited and clearly
not as strong, although it is important in interpreting these results to consider issues such as
exposure error (which could cause the calculated effects to be lower and less significant than the
true values).
9-28
-------
In Figure 9-5, the effect estimates presented for associations between morbidity and
ambient PM are grouped by PM indicator (PM10, PM2 5, and PM10_2 5), general health outcome
category (cardiovascular and respiratory), and more specific outcome measures (hospital
admissions and medical visits). Several general observations can be made:
• All associations between PM10 and hospitalization for cardiovascular and respiratory
diseases are positive and most are statistically significant, including all of the more
precise estimates. Almost all PM10 associations with emergency department (ED)
visits for cardiovascular and respiratory diseases are positive, and most respiratory
(but not cardiovascular) associations are statistically significant. The more precise
effect estimates range from about 2 to 6% increased risk per 50 |ig/m3 PM10 for
cardiovascular diseases and 2 to 12% increased risk per 50 |ig/m3 PM10 for respiratory
diseases, with some effect estimates for respiratory medical visits ranging up to about
30% per 50 |ig/m3 PM10.
• For PM2 5, all associations with hospitalization for cardiovascular and respiratory
diseases are positive and many are statistically significant, especially for respiratory
diseases. All PM2 5 associations with ED visits for cardiovascular and respiratory
diseases are positive, and about half are statistically significant. The more precise
effect estimates range from about 1 to 10% increased risk per 25 |ig/m3 PM25
for cardiovascular diseases, and 1 to 12% increased risk per 25 |ig/m3 PM2 5 for
respiratory diseases.
• Associations between PM10_2 5 and hospitalization for cardiovascular and respiratory
diseases are positive, and the effect estimates are of the same general magnitude as
for PM10 and PM2 5. In general, as was the case for mortality, the confidence intervals
for the PM10_2 5 estimates are broader than those for associations with PM10 or PM2 5 and
some, but not all, of the associations are statistically significant.
• For all PM indicators, associations with medical visits tend to be less precise
than those for hospital admissions. As was noted in Section 8.3.2.4, many of the
medical/physician visits effect estimates are larger in magnitude than those for
hospital admissions.
These figures include effect estimates from both recent studies and those available in the
previous PM NAAQS review, and it can be seen that the results fall within similar ranges.
For example, in the 1996 PM AQCD, a 50 |ig/m3 increase in PM10 was associated with a 2.5 to
5% increase in mortality risk, and results of the Six Cities analysis showed a 3% excess risk per
25 |ig/m3 PM2 5. The effect estimates from the more recent mortality studies, especially those
with greater statistical power, can be seen to fall in the same ranges. It is expected that results of
9-29
-------
multicity studies would more accurately reflect the magnitude of PM-health associations, and it
is important to note that the effect estimates from the new multicity studies, such as NMMAPS,
are at the low end of the effect estimate ranges; these results are, however, statistically
significant and the precision of the multicity estimates is notably greater than for single-city
studies.
For both mortality and morbidity outcomes, many more epidemiologic studies have
used PM10 than have used PM2 5 and PM10_2 5 measurements, since there is a much more
extensive set of air quality monitoring data available for PM10. The few studies that have tested
multipollutant models that include both PM2 5 and PM10_2 5 have reported that the two PM size
fractions have independent outcomes (e.g., Lippmann et al., 2000; Ito, 2003). Thus, it is difficult
to assess the implications of associations with PM10 for effects of fine and coarse fraction
particles, since they are likely to be representing separate effects of PM25 and PM10_25.
An association reported with PM10 is not simply a sum of the effects from the two particle size
classes, but may represent the influence of either fine or coarse fraction particles, or some
combination of the two, depending on the type of effect. PM10, to the extent it is correlated with
PM10 and PM10_25, will capture some of the effects of both PM25 and PM10_25. However, since
the correlation will be less than 1.0, the size of effects captured by PM10 will be less than that
found for PM2 5 and PM10_2 5 for a given concentration increment (e.g., per 10 |ig/m3).
As discussed in more detail in Section 8.2.2.5 and summarized in Table 9-3, various PM
components or characteristics, including ultrafine particles, have been associated with various
health outcomes. In general, evidence for associations have been reported for most components
that have been studied; and at least one new study has reported associations between ultrafine
particles and mortality or respiratory morbidity. However, many PM components are correlated
with each other and also with PM mass, making it difficult to distinguish effects of the various
components. Also, different PM components or characteristics would be expected to be more
closely linked with different health outcomes.
One new approach used to evaluate effects associated with various PM components is to
conduct a source apportionment analysis of the composition data base and to use the resulting
daily source factors as surrogates for exposure in an epidemiologic analysis. Motor vehicles, or
9-30
-------
TABLE 9-3. PARTICULATE MATTER CHARACTERISTICS, COMPONENTS, OR
SOURCE CATEGORIES SHOWN TO BE ASSOCIATED WITH MORTALITY IN
U.S., CANADIAN, OR EUROPEAN EPIDEMIOLOGIC STUDIES u
PM Size Fractions
Ions/Elements
Carbon/Organic Fractions Source Categories (Tracers)3
Mass fractions:
TSP
PM10
Thoracic coarse PM
(e.g., PM10.2.5)
Fine PM (e.g., PM2 5)
Ultrafine PM (PM^)
Particle number
Sulfate (SO4=)
Nitrate (NO3")
Ammonium (NH4-)
Transition metals
(e.g., Cd, Cu, Fe,
Ni, Mn, Zn)
Other toxic metals
(e.g., Pb)
TC (Total Carbon)
BC (Black Carbon)
EC (Elemental Carbon)
COH (Coefficient of Haze)
OC (Organic Carbon)
CX (Cyclohexene-
extractable Carbon)
Strong Acid (H+) Organic PM compounds
Motor Vehicles (CO, Pb)
Motor Vehicles plus
resuspended road dust
(CO, NO2, EC, OC, Mn,
Fe, Zn, Pb)
Fuel oil combustion (Ni, V)
Coal burning (Se)
Sulfate or regional sulfate (S)
Industrial (Zn, Cd)
1 Components measured in PM2 5 unless otherwise specified.
2 Organic PM compounds extracted by three techniques.
3 Source: Laden et al. (2000); reanalyzed in Schwartz et al. (2003); Mar et al. (2000, 2003); Tsai et al. (2000).
more precisely particles associated with vehicular traffic, stand out clearly as a source category
associated with mortality in all three studies that used this approach2. A regional sulfate source
category was also identified as being associated with mortality in all three studies (although
regional sulfate may be acting as a surrogate for PM2 5, given the high correlation between the
two); however, particles of crustal origin in PM2 5 were not significantly associated with
mortality. Also, associations were reported with an oil combustion factor and a source category
related to vegetative burning. These studies suggest that many different chemical components of
fine particles and a variety of different types of source categories are all associated with, and
probably contribute to, mortality, either independently or in combinations.
Multivariate techniques such as factor analyses and principal component analyses were used with speciation data
to determine PM contributions from source categories (Section 8.2.2.5.3, Table 8-4). For example, factors used as
indicators of particles from motor vehicle emissions in studies using older air quality data were Pb (Laden et al.,
2000; Schwartz et al., 2003) or Pb and CO (Tsai et al., 2000), but in a study with more recent air quality data, the
source category included several metals, OC, EC, CO and NO2 (Mar et al., 2000, 2003).
9-31
-------
One key research question that has not been addressed in epidemiologic studies is the
relationship between sources or composition of thoracic coarse particles and health outcomes.
The studies described above used source apportionment based on components of fine particles
or PM15 in an area dominated by fine particles. Some limited information is available from air
quality analyses that may help inform the assessment of epidemiological evidence for thoracic
coarse fraction particles; however, no studies are available to indicate potential differences in
thoracic coarse particle composition in relation to morbidity (for which there appears to be more
epidemiologic evidence suggesting coarse particle effects). As summarized in Chapter 3, crustal
material is an important contributor to thoracic coarse particles. Based on studies described
above, crustal components of PM2 5 have not been found to be important contributors to
associations with mortality, although it is possible that particles of crustal origin may contribute
to morbidity. Thoracic coarse fraction particles also include substantial contributions from
metals and biological constituents, both of which may be linked to adverse health outcomes.
In summary, considering results from studies conducted both within and outside the U.S.
and Canada, the epidemiological evidence is strong for associations between PM10 and PM2 5 and
mortality, especially for total and cardiovascular mortality. The magnitudes of the associations
are relatively small, especially for the multicity studies. However, there is a pattern of positive
and often statistically significant associations across studies for cardiovascular and respiratory
health outcomes, including mortality and hospitalization and medical visits for cardiovascular
and respiratory diseases, with PM10 and PM2 5. The few available PM10_2 5 studies also provide
some evidence for associations between hospitalization for cardiovascular and respiratory
diseases with PM10_2 5. PM10_2 5-hospitalization effect estimates were similar in magnitude to
those for PM10 and PM2 5, but with less precision. For PM10_2 5, the evidence for associations with
mortality is more limited; the magnitude of the effect estimates is very similar to those for PM2 5
and PM10, but in terms of precision, the evidence is not as strong. While there is some new
epidemiological evidence suggesting possible associations between health outcomes and
ultrafine particles and other fine particle components and sources, the data are as yet too sparse
to characterize the relative toxicities of these various components or indicators of fine particles
for different health outcomes.
9-32
-------
9.2.2.1.2 Long- Term Exposure Studies
In the 1996 PM AQCD, results of prospective cohort studies linked long-term exposure to
fine particles and mortality, and there was limited evidence indicating that long-term PM
exposure was linked with chronic respiratory morbidity, such as the development of bronchitis.
More recent long-term exposure studies have built upon these findings and provide further
evidence for associations with both mortality and respiratory morbidity.
A series of analyses using data from the ACS cohort have shown significant associations of
total and cardiopulmonary mortality with fine particles or sulfates, and the most recent analyses
have also reported significant associations with lung cancer mortality. The Six Cities study
found significant associations of total and cardiopulmonary (but not lung cancer) mortality with
PM2 5, but not with coarse particle indicators. The results most recently reported for the
Adventist Health Study on Smog (AHSMOG), reported some significant associations between
PM10 and total mortality and deaths with contributing respiratory causes. In further investigation
of the results found for PM10 among males, the associations with PM25 had larger effect
estimates than those for PM10_2 5 for males in the AHSMOG cohort, although none reached
statistical significance. For the VA study, indices of long-term exposures to PM10, PM25, or
PM10_2 5 were not associated with mortality.
Based on several factors - the larger study population in the ACS study, better
characterization of exposure in the Six Cities study, the more generally representative study
populations used in the Six Cities and ACS studies, and the fact that these studies have
undergone extensive reanalyses - the greatest weight is placed here on the results of the ACS
and Six Cities cohort studies in assessing relationships between long-term PM2 5 exposure and
mortality. The results of these studies, including the reanalyses results for the Six Cities and
ACS studies and the results of the ACS study extension, provide substantial evidence for
positive associations between long-term ambient (especially fine) PM exposure and mortality.
For morbidity, results of studies in a cohort of children in Southern California have built
upon the limited evidence available in 1996 PM AQCD to indicate that long-term exposure to
fine particles is associated with development of chronic respiratory disease and reduced lung
function growth. Long-term exposure to PM2 5 was associated with significant decreases in
lung function growth among a cohort of Southern California school children, but the earlier
9-33
-------
cross-sectional analysis for the same cohort found no relationship between respiratory symptoms
and annual average PM10 levels. These findings support the results of the cross-sectional study
in 24 U.S. and Canadian cities from the!996 PM AQCD, in which long-term PM exposure was
associated with some effects on respiratory function changes and respiratory illness.
As was true in the 1996 PM AQCD, it is more difficult to assess strength of evidence for
long-term exposure studies, since there are fewer studies available. For mortality, reanalyses
and extended analyses of cohort studies provide strong evidence for the link between mortality
and long-term exposure to fine particles; however, the available studies have provided no
evidence for associations between long-term exposure to coarse fraction particles and mortality.
In addition, prospective cohort and cross-sectional analyses have reported associations between
respiratory morbidity and PM10, and sometimes also PM2 5, providing fairly strong evidence for
effects of long-term fine particle exposures on respiratory morbidity. The morbidity studies
have not generally included PM10_2 5 data; so no conclusions can be drawn regarding long-term
exposure to coarse fraction particles and morbidity. Nor can any conclusions yet be drawn
regarding possible effects of long-term exposures to ultrafine particles, given the lack of
relevant data.
9.2.2.2 Robustness of Epidemiologic Associations
Many epidemiologic studies have also included assessment of whether the associations
were robust to such factors as model specification and potential confounding by co-pollutants.
Another factor that is relevant to robustness of epidemiologic findings is exposure error.
Chapter 8 includes detailed discussions on each of these topics, and the following discussion
focuses on the extent to which the current epidemiologic findings can be considered robust.
9.2.2.2.1 Model Specification
The 1996 PM AQCD included considerable discussion of issues regarding model
specification for time-series epidemiologic studies, including results of reanalyses using several
data sets, with a special focus on the large data set available from Philadelphia, PA. In this set of
reanalyses, results reported with the use of alternative modeling strategies were not substantially
different from the original investigators' findings. Also, at the time of completion of the 1996
9-34
-------
PM AQCD, it appeared that issues related to model specifications used to control for weather
effects in daily time-series analyses of ambient PM relationships to mortality/morbidity had
largely been resolved. Based on two major studies extensively evaluating a number of different
approaches to adjust for weather effects (including evaluations using synoptic weather patterns),
it was concluded that significant PM-mortality associations were robust and verifiable via a
variety of model specifications controlling for weather.
More recently, the influence of using default parameters and too low a standard error in a
widely used software package for GAM on epidemiologic study results has been investigated
and, in this process, the question of appropriate adjustment for weather, temporal trends, and
other covariates in time-series models was reopened. Numerous study findings were reanalyzed
to test the effect of using more stringent convergence criteria in the GAM program, as well as
alternative modeling methods such as GLM while provide correct standard errors. The results
from the GAM reanalysis studies indicate that PM risk estimates from GAM models were often,
but not always, reduced when more stringent convergence criteria were used, although the extent
of the reduction was not substantial in most cases. Also, the extent of downward bias in standard
errors reported for these data (a few percent to -15%) appears not to be very substantial,
especially when compared to the range of standard errors across studies due to differences in
population size and numbers of observation days available.
Thus, as stated in the HEI reanalysis report (Health Effects Institute, 2003), revised
analyses using GAM with more stringent convergence criteria or GLM with natural splines and
the use of alternative modeling strategies tended to reduce the PM effect estimate size for all PM
indices, but did not change the overall findings and qualitative conclusions of epidemiologic
studies showing associations between PM and both mortality and morbidity. In general, PM
effect estimates were more sensitive to different controls for time-varying (e.g., weather or
seasonal) effects. In some studies, with use of different methods or degrees of control for
temporal variables, PM effects estimates were largely unchanged, whereas in several other
studies the changes were enough to alter study conclusions. While it is clear that there will not
be one "correct" model or approach for covariate adjustment, further research can help inform
modeling strategies to adjust for temporal trends and weather variables in time-series
epidemiology studies.
9-35
-------
9.2.2.2.2 Assessment of Confounding by Co-Pollutants
Airborne particles are found among a complex mixture of atmospheric pollutants, only
some of which are widely measured (e.g., gaseous criteria co-pollutants O3, CO, NO2, SO2).
Because many of the pollutants are closely correlated due to emission by common sources and
dispersion by common meteorological factors, and some are in the pathway of formation of other
pollutants (e.g., NO —* NO2 —* HNO3 —* particulate nitrates), it is generally difficult to
disentangle their effects. In addition, as described in Section 8.1.3.2, co-pollutants could
possibly act as effect modifiers; for example, exposure to one pollutant could result in greater
sensitivity to another pollutant. Potential effect modification between pollutants has been
investigated in some toxicological or controlled human exposure studies (Section 7.9.3), but
little evidence is available from epidemiologic studies to characterize any such effects.
The potential for co-pollutant confounding in the epidemiologic time-series studies was
assessed in some detail in Section 8.4.3. Multipollutant modeling is the most common method
used to test for potential confounding in epidemiologic studies; however, interpretation of the
results of multipollutant models is complicated by the correlations that often exist between air
pollutants. In interpreting the results of any of these studies, it is important to consider factors
such as the biological plausibility of associations between the pollutants and health outcomes, as
well as questions related to model specification and exposure error. For example, some new
studies described in Section 5.3.3.4 have reported that ambient PM25 concentrations are well
correlated with personal PM2 5 exposure measurements; in contrast, that is generally not the case
for O3, SO2, and NO2.
Multipollutant modeling results for associations of a range of health outcomes with PM10,
PM25, and PM10_25 and gaseous pollutants in single-city studies are presented in Section 8.4.3
(Figures 8-16 through 8-19). For most studies, there was little change in coefficients for all three
indices between single-pollutant and multipollutant models; however, in some cases, the PM
effect estimate was markedly reduced in size and lost statistical significance in models that
included one or more gaseous pollutants. Key results are also available from the NMMAPS
evaluation of associations across many U.S. cities with varying climates and mixes of pollutants;
the NMMAPS associations between PM10 and both mortality and morbidity were not changed
with adjustment for gaseous pollutant concentrations. Thus, for the most part, effect estimates
9-36
-------
for the three PM indices were not substantially changed when gaseous co-pollutants were
included in the models. Often, PM and the gaseous co-pollutants were highly correlated,
especially for fine particles and CO, SO2 and NO2, and it was generally the case that high
correlations existed between pollutants where PM effect estimates were reduced in size with the
inclusion of gaseous co-pollutants.
In the prospective cohort and cross-sectional studies, the potential for confounding by
co-pollutants has been assessed in some studies of mortality, but little studied for morbidity.
The reanalysis of data from the ACS cohort indicated that the relationships with fine particles
and sulfates were reduced in size in co-pollutant models including SO2, but not the other gaseous
pollutants. SO2 is a precursor for fine particle sulfates, thus complicating the interpretation of
multipollutant models (i.e., making it difficult to distinguish effects due to fine particles or
to SO2 for this study). The authors concluded that their results suggested that mortality may be
associated with more than one component of the ambient air pollutant mix and that there were
robust associations of mortality with fine particles and sulfates.
In summary, ambient PM exposure usually is accompanied by exposure to many other
pollutants, and PM itself is composed of numerous physical/chemical components. Assessment
of the health outcomes attributable to ambient PM and its constituents within an already subtle
total air pollution effect is, therefore, very challenging, even with well-designed studies. Indeed,
statistical partitioning of separate pollutant effects is not likely to characterize fully the effects
that actually depend on simultaneous exposure to multiple air pollutants. Overall, the newly
available epidemiologic evidence, especially for the more numerous time-series studies,
substantiates that associations for various PM indices with mortality or morbidity are robust
to confounding by co-pollutants.
9.2.2.2.3 Exposure Error
Numerous analyses of the potential influence of measurement error on time-series
epidemiologic study results are discussed in Section 8.4.5. One consideration in comparing
epidemiologic findings for different pollutants is the relative precision with which the pollutants
are measured. If two pollutants have effects and there is correlation between both the pollutants,
the effect estimate of the pollutant that is less precisely measured may be attenuated when the
9-37
-------
pollutants are considered in a model together. One would expect that PM2 5, CO, and NO2 would
often have a high positive correlation due to common activity patterns, weather, and source
emissions. PM10_2 5 is generally less precisely measured than PM2 5, but the two are not generally
highly correlated. Several recent studies have focused on this question, and reported that for
most situations, it is unlikely that differential measurement error would result in shifting
apparent effects from one pollutant to another. The most extreme case, complete transfer of
apparently causal effects from one pollutant to another, required very high correlation between
the covariates, no error in measurement of the "false" covariate and moderate error in
measurement of the "true" predictor. The results of these analyses indicate that it is unlikely that
effects attributed to PM (generally focusing on PM10 or PM2 5) are falsely transferred from other
less-precisely measured pollutants.
Another facet of exposure error is the degree to which the measurements made at
community monitoring sites reflect population exposures to ambient PM. As discussed in
Section 5.2, further analysis of the PTEAM study has shown that ambient PM10 concentrations
were well correlated with temporal changes in personal exposures to ambient PM10. However, it
should be noted that (a) the spatial variability across a city is generally much greater for PM10_2 5
than for PM2 5 and (b) there may still be substantial variability in PM2 5 concentrations across
some urban areas, as discussed in Chapter 3. In addition, the infiltration factors for PM10_25 and
a number of gases (e.g., O3, SO2) are lower and more variable than that of PM25, likely leading
to a lower correlation between ambient concentration (used as an exposure surrogate in
community time-series studies) and personal exposure to the ambient pollutant for PM10_2 5 and
these gases. Also, studies which included subjects limited to those living near the air monitoring
site (and, therefore, presumably have lower exposure error due to spatial heterogeneity of PM25
concentrations) tend to yield higher effect estimates. Thus, exposure studies indicate that
particle measurements at central monitoring sites are better indicators of personal exposures to
ambient PM2 5 than PM10_2 5 for time-series studies. Exposure relationships for PM10_2 5 have been
less well studied, but exposure error and measurement error would be expected to have greater
influence for associations with PM10_2 5 than for PM2 5; this likely contributes to larger confidence
intervals around PM10_2 5 effect estimates.
9-38
-------
9.2.2.3 Consistency of Findings Across Epidemiologic Studies
In the 1996 PM AQCD, it was observed that PM was associated with mortality and
morbidity in studies conducted in numerous locations in the United States as well as in other
countries. The expanded body of studies available in this review includes studies conducted in a
wider range of locations; as described above, many of those studies, especially those with greater
statistical power, report statistically significant associations. Magnitudes and significance levels
of observed air pollution-related effects estimates would be expected to vary somewhat from
place to place, if the observed epidemiologic associations denote actual effects, because
(a) not only would the complex mixture of PM vary from place to place, but also (b) affected
populations may differ in characteristics that increase susceptibility to air pollution health
effects, and (c) areas may differ in factors that affect population exposures to ambient pollutants.
Multicity studies conducted in the United States, Canada, and Europe have included
quantitative assessments of heterogeneity in PM effect estimates. The city-specific and
regional PM10-mortality associations presented in NMMAPS results suggested greater variability
in effect estimates than had been observed in the studies available in the 1996 PM AQCD.
However, statistical analyses indicated that there was no significant heterogeneity in mortality
effect estimates for 90 U.S. cities (Samet et al., 2000; Dominici et al., 2003). For eight Canadian
cities, no evidence of heterogeneity was reported in the initial analysis, but in reanalysis to
address GAM issues, there appeared to be greater heterogeneity in the PM-mortality associations
(Burnett and Goldberg, 2003). Finally, initial analyses of mortality associations for 29 European
cities indicated differences between eastern and western cities, but these differences were less
clear with reanalysis to address GAM questions (Katsouyanni et al., 2003).
There are a number of reasons to expect variation in PM-health outcome associations for
different geographic regions. Regional differences can include differences in PM sources or
composition, differences in population exposures, and differences in potentially susceptible
groups. In the European multicity study, APHEA, PM-mortality associations were found to be
larger in areas with higher average NO2 levels (considered an indicator of traffic pollution), and
warmer climates (possibly due to more open windows resulting in better exposure estimation).
In NMMAPS, no apparent associations were found between PM-mortality associations and
9-39
-------
either/or PM2 5/PM10 ratios or socioeconomic indicators, but there was also no statistically
significant measure of heterogeneity in this study. However, for hospital admissions in the
NMMAPS, the PM10 admissions associations were greater in areas with less use of central air
conditioning (possibly an indicator of increased exposure to ambient pollutants) and with larger
contributions of PM10 emissions from vehicle emissions and oil combustion.
Variability in PM concentrations across study areas could influence epidemiologic study
results. For larger metropolitan areas, including monitors in outlying areas may bias the
exposure estimate and reduce the correlation between the averaged concentration and the true
population exposure. From among those U.S. cities in which epidemiological studies have been
conducted, areas with more uniformity in PM2 5 concentrations include Chicago and Detroit,
whereas areas with more spatial variability include Seattle and Los Angeles. There are a number
of factors that could influence spatial variability of PM concentrations, including topography,
location of major PM sources, and weather patterns. Greater spatial variability in PM levels
would be expected to increase exposure error, potentially affecting epidemiologic study results
in those areas.
One factor unrelated to geographic location that would likely affect the consistency of
results across studies is the amount of data available for analysis. For time-series studies, the
number of days with measurements is one important indicator of study size, or statistical power.
In Figure 9-4, the PM-mortality associations are plotted in order of decreasing statistical power,
using the product of daily death rate and number of PM measurement days as the indicator.
For single-city mortality studies, the number of PM measurement days ranges from about 150
(Tsai et al., 2000) to over 2,000 days (e.g., Ito and Thurston, 1996). Multicity studies included
ranges of about 500 to 900 days for eight Canadian cities (Burnett and Goldberg, 2003), about
200 to 3,000 days for 90 U.S. cities (Dominici et al., 2003), and 1,500 to 3,000 days for 10 U.S.
cities (Schwartz, 2003). For several studies, more data are available for PM10 than for PM2 5
or PM10_25; Fairley (2003), as an example, used a data set with approximately 800 days of PM10
measurements and 400 days for PM25 and PM10_25. In the 1996 PM AQCD, studies conducted in
the United States had about 300 to 4,000 days of PM measurements, and a clear correlation
between t-ratio and number of monitoring days could be seen (Figure 12-17, Table 12-25).
9-40
-------
Similarly, Figures 9-4 and 9-5 show a tendency for larger studies to have more consistent effect
estimates that are more likely statistically significant. A number of the newer studies, however,
particularly those using PM2 5 and PM10_2 5 data, are somewhat smaller in size than those
available in the 1996 PM AQCD. This would be expected to result in decreased precision and
more variability in effect estimate size for the smaller studies.
In addition, while many single-city epidemiologic studies have used availability of
everyday monitoring data as a criterion for selecting study locations, a number of the newer
studies have used PM2 5 and PM10_2 5 data measured every sixth day. Beyond limiting the number
of days of data available, the use of l-in-6 day data may also complicate time-series analyses.
As discussed in Section 8.4.5.2, one analysis of data from Chicago first used data from an
everyday monitor and then six l-in-6 day data sets were created from the same data.
Whereas the resulting analysis using all the daily data showed clearly statistically significant
positive PM10-mortality associations, the results for the analyses using l-in-6 day data sets were
quite inconsistent. Hence, the use of air quality data with many missing days adds uncertainty
to results for PM-health outcome associations.
Thus, there are numerous reasons to expect study results to vary across cities, based on
different topographies, distribution of sources or emissions, mixes of pollutants, and population
characteristics. The new multicity studies have provided some initial evidence for some of these
factors that may affect the magnitude or significance of PM-health associations. Effect estimates
reported for sets of studies of the same location generally fall well within the range of the
confidence intervals of all studies. With multicity studies, statistical tests for heterogeneity
among effect estimates have been conducted with inconsistent findings. It seems likely that
some apparent variation in effect estimate size is simply statistical variability, while factors
such as differential exposures based on housing characteristics or population characteristics may
contribute to real variations in effects between locations. Overall, the epidemiological
studies indicate that, in numerous locations across the United States, there are associations
between PM10 and PM2 5 and mortality from cardiorespiratory diseases and between PM10, PM2 5
and PM10_2 5 and hospitalization for respiratory or cardiovascular causes.
9-41
-------
9.2.2.4 Temporality and the Question of Lags
As discussed in Section 8.4.4, differing lag periods are likely appropriate for different
health outcome-pollutant associations. For example, the time-series studies of cardiovascular
hospital admissions or emergency department visits suggest that PM effects for all PM indices
are stronger with same day PM concentrations, with some effects also linked with PM from the
previous day. In a few studies of cardiac physiological changes, the strongest associations were
reported for some outcomes with 1- to 2-hour lag periods, indicating that for certain health
outcomes very short-term fluctuations in pollution are most important. In panel studies of
respiratory symptoms and in several studies of asthma hospitalization or emergency department
visits, longer moving average lag periods (up to 5- to 7-day moving averages) yielded larger PM
effect estimates, suggesting that these health responses may have a longer and more extended
latency period than indicated by single day lag analyses.
Where results are presented for a series of lag days, it is important to consider the pattern
of results that is seen across the series of lag periods. When there is a pattern of effects across
lag periods, selecting any one of the single-day lag effect estimates is likely to underestimate the
overall effect size, since the largest single-lag day results do not fully capture the risk also
distributed over adjacent days. Even if there is a jumbled pattern of results across the different
lags, then the single-day lag with the largest effect is likely biased low. In these cases,
a distributed lag model should more correctly capture the effect size.
Studies of long-term exposure have included less evaluation of temporal relationships
between PM exposure and health outcome. The prospective cohort studies have used air quality
measurements made over a period of years as an indicator of long-term exposure to air pollution.
The associations reported in these studies are for relationships with PM across various levels of
exposure, not as a measure of latency of effect. However, some new studies have included some
assessment of temporal relationships between PM exposure and mortality. In the reanalysis of
the Six City Study, the decline in fine particle levels over the monitoring period was included as
a time-dependent variable, to assess the effect of changing PM concentrations over time on the
association with mortality. The association between total mortality and fine particles was
reduced in size, though still statistically significant, as compared with the model not allowing for
9-42
-------
temporal change in pollution level. This is likely indicative of the effectiveness of control
measures in reducing source emissions importantly contributing to the toxicity of ambient
particles in cities where PM levels were substantially decreased over time.
The VA study analysis tested associations between different subsets of long-term exposure
to pollution and mortality data. While the associations found between PM (PM10, PM2 5, and
coarse fraction particles) and mortality varied and were often negative and generally not
statistically significant, it was observed that the associations were larger and more likely to be
statistically significant with the air quality data from the earliest time periods, as well as the
average across all data. Further study is needed to evaluate the relationship between health
outcomes and long-term PM exposure where PM concentrations are changing over time.
In summary, for time-series studies, it is likely that the most appropriate lag period for a
study will vary depending on the health outcome and the specific pollutant under study. Where
effects are found for a series of lag periods, the effect estimate for any one lag period will likely
underestimate the effect size and a distributed lag model will more accurately characterize the
effect estimate size. Caution should be used in selecting results for single lag periods if the
pattern of results across lag periods is highly variable. For effects associated with chronic
exposure, less is known about the importance of different time windows for exposure, and some
recent studies indicate that further investigation is needed.
9.2.2.5 Concentration-Response Relationships
In the 1996 PM AQCD, the limitations of identifying possible "thresholds" in the
concentration-response relationships in observational studies were discussed. It was observed
that detection of threshold levels in population-based epidemiological studies would be very
difficult on several bases, including difficulties related to the low data density in the lower PM
concentration range, the small number of quantile indicators often used, and the possible
influence of measurement error. Few studies had quantitatively assessed the form of PM-effect
concentration-response functions and the potential for a threshold level.
A threshold for a population, as opposed to a threshold for an individual, has some
conceptual issues that should be noted. For example, since individual thresholds vary from
9-43
-------
person to person due to individual differences in genetic-level susceptibility and pre-existing
disease conditions (and even can vary from one time to another for a given person), it is
extremely difficult mathematically to demonstrate convincingly that a clear threshold exists in
the population studies. This is especially true if the most sensitive members of a population are
unusually sensitive even down to very low concentrations. The person-to-person difference in
the relationship between personal exposure to PM of ambient origin and the concentration
observed at a monitor may also add to the variability in observed exposure-response
relationships, possibly obscuring otherwise more evident thresholds. Since one cannot directly
measure but can only compute or estimate a population threshold, it would be difficult to
interpret an observed population threshold biologically, without pertinent collateral
dosimetric/toxicologic information.
Recognizing these difficulties, several epidemiologic studies have evaluated potential
thresholds in time-series mortality analyses. Analyses using NMMAPS data for 90 U.S. cities
showed that, for total and cardiorespiratory mortality, it was difficult to discern any threshold
level for PM10, with statistical tests indicating that a linear concentration-response model was
preferred over the spline and the threshold models. In this study, the likelihood of a threshold
occurring above 24-h PM10 levels of-25 |ig/m3 seems to be essentially zero (see Figure 8-31);
but there was increasing probability of a threshold occurring at levels below 25 |ig/m3. In some
single-city analyses, there were indications of potential population thresholds for associations
between mortality and 24-h PM10 in the range of 80 |ig/m3 to 100 |ig/m3 and, with 24-h PM2 5,
in the range of 20-25 |ig/m3. However, other single-city analyses reported no evidence of a
threshold for PM-mortality associations for PM10, PM2 5 or PM10_2 5. One group of researchers
who did not find evidence for thresholds in the PM10-mortality relationship using various
statistical methods observed it to be "highly likely" that the statistical methods could detect a
threshold if a threshold existed in this population-based study.
In summary, the available evidence does not either support or refute the existence of
thresholds for the effects of PM on mortality across the range of concentrations in the studies.
In the multicity and most single-city studies, statistical tests comparing linear and various
nonlinear or threshold models have not shown statistically significant distinctions between them.
9-44
-------
Where potential threshold levels have been suggested in single-city studies, they are at fairly low
levels. In epidemiological analyses of complex health endpoints (such as total nonaccidental
mortality) in populations where individuals differ in susceptibility and exposures, it is likely to
be extremely difficult to detect threshold levels. Also, it must be recognized that dose-response
relationships observed for one or another PM indicator (e.g., PM10) do not necessarily apply to
other PM indicators (e.g., PM2 5 or PM10_2 5).
9.2.2.6 Natural Experiment Studies
Although many studies have reported short-term associations between PM indices and
mortality, a largely unaddressed question remains as to the extent to which reductions in ambient
PM actually lead to reductions in health effects attributable to PM. This question is not only
important in terms of "accountability" from the regulatory point of view, but it is also a scientific
question that challenges the predictive validity of statistical models and their underlying
assumptions used thus far to estimate excess mortality due to ambient PM.
The opportunities to address this question are rare. However, at the time of the 1996 PM
AQCD, results were available from epidemiologic studies of a "natural" or "found experiment"
in the Utah Valley, where respiratory hospital admissions were found to decrease during the time
a major PM source was closed. Newly available controlled human exposure and animal
toxicology studies, using particle extracts from ambient community PM10 sampling filters from
the Utah Valley, have also shown reduced effects with exposure to extracts of particles collected
during the time period when the source was not operating. A recent epidemiologic study in
Dublin, Ireland also provides evidence for reductions in ambient PM (measured as British
smoke) being associated with reductions in mortality rates. Other "found experiments" also
provide evidence for decreases in mortality and/or morbidity being associated with notable
declines in different indices of PM (and/or gases such as SO2) as the result of interventions
aimed at reducing air pollution.
By providing evidence for improvement in community health following reduction in air
pollutant emissions, these studies add further support to the results of the hundreds of other
epidemiologic studies linking ambient PM exposure to an array of health effects. Such studies
showing improvements in health with reductions in emissions of ambient PM and/or gaseous
9-45
-------
co-pollutants provide strong evidence that reducing emissions of PM and gaseous pollutants
has beneficial public health impacts.
9.2.2.7 Summary and Conclusions
Epidemiological evidence can help to inform judgments about causality. The present
discussion evaluated the epidemiologic evidence in relation to the first five criteria listed in the
beginning of Section 9.2, including key considerations with regard to criteria such as the strength
(magnitude, precision) and robustness of reported associations. Information related to last of the
six criteria (coherence and biological plausibility of the evidence) is discussed in the following
section.
Overall, there is strong epidemiological evidence linking (a) short-term (hours, days)
exposures to PM2 5 with cardiovascular and respiratory mortality and morbidity, and (b) long-
term (years, decades) PM2 5 exposure with cardiovascular and lung cancer mortality and
respiratory morbidity. The associations between PM2 5 and these various health endpoints are
positive and often statistically significant. There are fewer studies available for PM10_2 5 and the
magnitude of the effect estimates for associations with mortality and morbidity effects
(especially respiratory morbidity) is similar to that for PM2 5, but the lesser precision reduces the
strength of the evidence for PM10_2 5 effects. Little evidence is available to allow conclusions to
be drawn about long-term PM10_2 5 exposures and morbidity. There is also extensive and
convincing evidence for associations between short-term exposures to PM10 and both mortality
and morbidity, as was reported in the previous review.
With regard to the robustness of the associations, research questions remain on modeling
issues with time-varying variables, but extensive reanalyses conducted for both time-series and
prospective cohort studies provide further support for the results of the original analyses.
Recent reanalyses of a number of time-series studies found that results were little changed with
adjustments to deal with GAM-related issues, but some results were sensitive to different
adjustment for time-varying factors such as weather. However, the recent studies, using a
variety of approaches to control for weather effects, still appear to demonstrate increased
PM-related mortality and morbidity risks beyond those attributable to weather influences alone.
9-46
-------
Much progress has been made in sorting out contributions of ambient PM10 and its
components to observed health effects relative to other co-pollutants. Despite continuing
uncertainties, the evidence overall tends to support the above conclusions that ambient PM10
and PM2 5 are most clearly associated with mortality/morbidity effects, acting either alone or in
combination with other covarying gaseous pollutants, with more limited support being available
with regard to PM10_2 5. Likely contributing to this is the fact that greater measurement error
associated with exposure estimates for a given pollutant or indicator will result in less precise
effect size estimates that are less robust in multipollutant models. Of importance here, directly
measured PM10 and PM2 5 values likely have less measurement error than PM10_2 5 values
derived by subtracting PM2 5 values from PM10 concentrations, especially if obtained from
non-collocated PM10 and PM2 5 monitors at different locations in a given urban area. Thus, the
current paucity of statistical significance for a pattern of positive associations with PM10_25 may
reflect measurement imprecision, not necessarily lack of effects.
Focusing on the studies with the most precision, it can be concluded that there is much
consistency in epidemiological evidence regarding associations between short-term and long-
term exposures to fine particles and cardiopulmonary mortality and morbidity. For coarse
fraction particles, there is also some consistency in effect estimates for hospitalization for
cardiovascular and respiratory causes based on the few studies available for several locations
across the United States. Some variability in effect estimate size can be seen across locations,
especially in the recent multicity studies. Factors likely contributing to this variability include
geographic differences in air pollution mixtures, composition of ambient PM components, and
personal and sociodemographic factors potentially affecting PM exposure (e.g., use of air
conditioning), as well as differences in PM mass concentration.
Temporality, or the occurrence of the health outcome following the exposure, has been
found to hold well for the time-series epidemiological studies. The length of the lag period for
exposure and effect varies for different health outcomes, with short acute exposures of an hour
or more seemingly more important for some cardiovascular health endpoints and longer average
exposures or distributed lag exposure windows being more closely associated with other health
endpoints. For long-term exposures, the existing studies generally use spatial variation in
9-47
-------
concentrations to estimate exposure changes, thus temporality is not directly tested, but some
new evidence suggests that changes in pollutant mix over time may influence relationships with
health effects.
In conclusion, the epidemiological evidence continues to support likely causal associations
between PM2 5 and PM10 and both mortality and morbidity from cardiovascular and respiratory
diseases, based on an assessment of strength, robustness, and consistency in results. For PM10_25,
less evidence is available, but the studies using short-term exposures have reported results that
are of the same magnitude as those for PM10 and PM2 5, though less often statistically significant
and thus having less strength, and the associations are generally robust to alternative modeling
strategies or consideration of potential confounding by co-pollutants. This evidence is
suggestive of associations for morbidity with short-term changes in PM10_25. Epidemiologic
studies suggest no evidence for clear thresholds in PM-mortality relationships within the range
of ambient PM concentrations observed in these studies. Important new results from source
apportionment studies and found experiments indicating that reductions in PM and other air
pollutants result in improvements in community health lend support to the results of the other
epidemiological studies.
9.2.3 Integration of Experimental and Epidemiologic Evidence
In more broadly assessing the extent to which the overall body of evidence supports the
attribution of observed health effects to exposure to fine and coarse thoracic PM and related
chemical constituents, one needs to look beyond just epidemiologic evidence to consider the
implications of newly available dosimetric, toxicologic, and other evidence as well. More
specifically, the following assessment (a) evaluates information pertaining to the biological
plausibility of the types of health outcome associations observed in the epidemiologic studies,
taking into account toxicologic findings and potential mechanisms of action; and (b) considers
information about the coherence of the overall body of evidence relevant to PM-related health
outcomes supporting conclusions regarding attribution of observed effects to ambient fine or
coarse thoracic PM and related chemical constituents, acting alone and/or in combination with
other pollutants.
9-48
-------
The 1996 PM AQCD highlighted several key findings and conclusions concerning
attribution of observed health outcomes to specific ambient PM size fractions or chemical
compounds:
• "The likelihood of ambient fine mode particles being significant contributors to
PM-related mortality and morbidity among [the] elderly population is bolstered by:
(1) the more uniform distribution of fine particles across urban areas. . . ;
(2) the penetration of ambient particles to indoor environments. . . ; and (3) the longer
residence time of ambient fine particles in indoor air, enhancing the probability of
indoor exposure to ambient fine particles more so than for indoor exposure to ambient
coarse particles."
• The PM indices that have been "most consistently associated with health endpoints are
fine particles (indexed by BS, COH, and PM2 5), inhalable particles (PM10 or PM15),
and sulfate (S04=)," whereas "[l]ess consistent relationships have been observed for TSP,
strong acidity (H+), and coarse PM (PM10_25). . . . [and] none of these indices can
completely be ruled out as a biologically relevant indicator of PM exposure."
• "Based on current evidence from epidemiologic, controlled human, human occupational,
and laboratory animal studies, no conclusions can be reached regarding the specific
chemical components of PM10that may have the strongest biologic activity." Further,
none of the various subclasses of PM [e.g., acid aerosols, bioaerosols, metals (including
transition metals), and insoluble ultrafme particles] that have been considered "can be
specifically implicated as the sole or even primary cause of specific morbidity and
mortality effects." (U.S. Environmental Protection Agency, 1996, p. 13-93)
Hence, although at the time of the 1996 PM AQCD, the epidemiologic evidence was viewed as
substantiating well PM10 or PM2 5 associations with human mortality and morbidity, uncertainties
remained with regard to (a) the contribution of specific PM constituents to PM toxicity and
(b) the biological plausibility of the reported effects and/or the mechanisms of action underlying
them.
Since the 1996 PM AQCD evaluations, progress has been made in (1) further
substantiating and expanding epidemiologic findings indicative of ambient PM-health effect
associations, (2) identifying possible constituents contributing to observed effects, and
(3) obtaining evidence bearing on the biological plausibility of observed effects and possible
mechanisms of action involved. Efforts to interpret the overall meaning of the epidemiologic
findings and to evaluate their biological plausibility and pertinent mechanisms of action are
9-49
-------
complicated by the fact that ambient PM exists as a component of a complex air pollution
mixture that includes gaseous air pollutants. This section addresses these complexities by
considering the extent of coherence observed among findings reported for specific PM
components identified in epidemiologic studies for specific health outcome categories
(cardiovascular; respiratory; lung cancer; fetal and infant development/mortality) and related
toxicologic links to biologic changes observed in controlled exposure human, animal, and
in vitro studies. Hypothesized potential mechanisms of actions and other supporting pieces of
evidence are also summarized.
As discussed at the outset of Section 9.2, several criteria were listed as useful in evaluating
scientific evidence as supporting conclusions regarding potential causal relationships between
two variables. In addition to those criteria addressed in the preceding discussion of PM-related
epidemiologic evidence, it is important to take into account still other information or criteria
which combine consideration of biological plausibility and coherence, so as to help ensure:
"that a proposed causal relationship not violate known scientific principles, and that it be
consistent with experimentally demonstrated biologic mechanisms and other relevant data. . . ."
For the purposes of this assessment: the ensuing discussion of plausibility and coherence
considers both: (a) the extent to which the available epidemiologic evidence (of adequate
power) shows associations in the same location (urban area) with a range of logically linked
health endpoints (i.e., endpoints within a "pyramid of effects" ranging from the most severe
outcome, mortality, to physiological changes in the cardiovascular or respiratory systems, e.g.,
altered fibrinogen levels or lung function changes); and (b) the extent to which the available
toxicologic evidence and mechanistic information provide support for the plausibility of the
array of observed epidemiologic associations likely reflecting causal relationships.
Before embarking on the plausibility and coherence discussions in Section 9.2.3.2 for each
of several health endpoints (cardiovascular, respiratory, etc.) evaluated as likely being impacted
by ambient PM, the next section first provides important background information on three cross-
cutting issues that help to place the ensuing discussions in context.
9-50
-------
9.2.3.1 Background on Cross-Cutting Issues
Background information on several cross-cutting issues is provided here to help place the
ensuing discussions in perspective. First, important considerations related to strengths and
weaknesses of experimental approaches used to study PM health effects are summarized along
with relevant caveats. Next, interspecies dosimetric comparisons are discussed and
representative examples are provided for extrapolation of PM exposure/doses between rats and
humans. Lastly, information on inhaled particles as carriers of other toxic agents and consequent
potential implications are discussed.
9.2.3.1.1 Approaches to Experimental Evaluation ofPM Health Effects
As discussed in Chapter 7, various experimental approaches have been used to evaluate
PM health effects, including: studies of human volunteers exposed to PM under controlled
conditions; in vivo studies of laboratory animals including nonhuman primates, dogs, and rodent
species; and in vitro studies of tissue, cellular, genetic, and biochemical systems. A variety of
exposure conditions have been used, including: whole body, mouth-only, and nose-only
inhalation exposures to concentrated ambient particles (CAPs) or laboratory-generated particles;
intratracheal, intrapulmonary, and intranasal instillation; and in vitro exposures to test materials
in solution or suspension. These approaches have been used mainly to test hypotheses regarding
the role of PM in producing the types of health effects identified by PM-related epidemiologic
studies. Thus, most new toxicological studies have mainly addressed the question of biologic
plausibility of epidemiologically-demonstrated effects and mechanisms of action, rather than
attempting to delineate dose-response relationships.
Reflecting this, most of the toxicology studies have generally used exposure concentrations
or doses that are relatively high compared to concentrations commonly observed in ambient air.
One consideration underlying the use of such experimental exposure concentrations is the fact
that healthy animals have most typically been used in many controlled-exposure toxicology
studies, whereas epidemiologic findings often reflect ambient pollutant effects on compromised
humans (e.g., those with one or another chronic disease) or other susceptible groups at increased
risk due to other factors. Implicit in using relatively high concentrations in experimental studies
of healthy subjects is the assumption that increasing the dose makes up for compromised
9-51
-------
tissue/organ functions that may contribute to observed ambient PM effects, but this may not
be so.
Recognizing this, there has been growing attention to the development and use of
compromised animal models thought to mimic important characteristics contributing to
increased human susceptibility to ambient PM effects. One example is the use of monocrotaline
(MCT)-treated rats, in which the MCT-induced pulmonary vasculitis/hypertension is thought to
render them at possible increased risk for PM effects. A limitation of this model is that
pathology is induced acutely in rats to model a chronic illness in humans. Another example is a
compromised animal model of chronic bronchitis (induced by repeated, prolonged exposure
to SO2 before exposure to PM). Partial coronary artery occlusion is yet another example of a
compromised animal model, evaluated for increased cardiovascular risk. Possible PM
exacerbation of respiratory infections has also been evaluated in animals intratracheally exposed
to various bacteria. There is a need to search for relevant new animal models to better simulate
human pathophysiology in PM exposure studies.
Given the relatively high concentrations used, caution is needed in attempting to interpret
and extrapolate effects seen in these studies to provide insight into the biological plausibility and
mechanisms of action underlying effects seen in humans under "real-world" exposure
conditions. Some reported responses may only be seen at the higher concentrations (more
typical of occupational exposures) and not necessarily at (usually much lower) ambient particle
exposure levels. On the other hand, differences between humans and rodents with regard to the
inhalability, deposition, clearance, and retention profiles for PM (see Chapter 6 for details) could
in some instances make doses to some specific respiratory tract tissues from relatively high
experimental exposures relatively similar to doses from human ambient exposures.
Since the 1996 PM AQCD, the effects of controlled exposures to ambient PM have been
evaluated by use of urban air particles (UAP) collected from ambient samplers (e.g., impactors,
diffusion denuders, etc.) and, more recently, by the use of aerosol concentrators. In the first type
of study, particles from ambient air samplers are collected on filters or other media, then stored
for varying time periods (hours to years or even decades) before later being resuspended in an
aqueous medium and used in inhalation, instillation, or in vitro studies. Depending on the
storage conditions for the filters (e.g., whether or not kept refrigerated or in the dark) varying
9-52
-------
amounts of some originally collected materials (including highly biologically active semivolatile
compounds) may be lost and their possible effects missed in UAP studies. Also of potential
concern is the fact that prolonged storage of filters under inappropriate conditions may lead not
only to volatilization of semivolatile material, but also to chemical alteration of reactive
compounds and/or possible growth of mold, bacterial contamination, etc.
Particle concentrators allow exposure under controlled conditions of animals or humans by
inhalation to concentrated "real-world" ambient particles (CAPs) at levels higher than typical
ambient PM concentrations. However, CAPs studies cannot closely control the mass
concentration and day-to-day variability in ambient particle composition; and they have
sometimes lacked detailed characterization of variations in chemical composition from one
CAPs exposure to another. Because the composition of CAPs varies across both time and
location, thorough physical-chemical characterization is needed (but not always done or
reported) to facilitate comparison of results between studies or even among exposures within
studies, so as to better link specific particle composition to effects. Another limitation is the fact
that concentrators used in many of the studies assessed here lose concentrating efficiency below
0.3 |im, and do not concentrate ambient particles in the ultrafme range < 0.1 jim. Thus, it is
possible that portions of potentially important combustion-generated particles (e.g, from diesel,
gasoline vehicle, wood smoke, coal smoke, etc.) were present only at ambient (not higher
concentrated) levels in most of the CAPs studies assessed here; and many other potentially toxic
co-components (e.g., SO2, O3, peroxides, etc.) of the ambient aerosols may not have been
concentrated or were excluded from the CAPs exposure mix as well. Newer versions of CAPs
concentrators being used in ongoing research do allow for concentrating of particles < 0.1 jim
and for exposure to gaseous co-pollutants present in the ambient air along with the particles
being concentrated. These improvements should enable CAPs exposures in ongoing and/or
future research studies to more fully reflect potentially important interactive effects of overall
"real-world" aerosol mixes.
Controlled human and laboratory animal exposures to particulate material obtained from
combustion-source bag house filters or other combustion-source collection devices have also
been used to evaluate the in vitro and in vivo respiratory toxicity of complex combustion-related
PM. Residual oil fly ash (ROFA) collected from large industrial sources (e.g., oil-fired power
9-53
-------
plants) has been extensively used, and, less often, domestic oil furnace ash (DOFA) or coal fly
ash (CFA). The major disadvantage associated with the use of such materials derives from
questions about the potential relevance of results obtained in understanding ambient PM
exposure effects. Before extensive implementation of air pollution controls, ambient U.S. air
contained mixtures of PM species (at higher than current concentrations) analogous to those in
many of the source samples used in toxicologic studies during the past decade or so. However,
it is unlikely that high concentrations of certain materials that typify such samples would be
found or approached in ambient air PM samples from community monitoring sites in the United
States, Canada, and much of western Europe that generated the aerometric data (collected during
the past 20 to 30 years) that were used to estimate PM exposures in most PM epidemiology
studies assessed here. Very high concentrations of metals (especially Ni and V, for example)
typify most ROFA samples, and experimental exposures to such materials have generally
resulted in exposures and doses orders of magnitude (100s of times) higher than for usual
concentrations of such metals in ambient PM measured routinely since the 1970s at community
monitoring sites across the United States. Thus, significant issues arise concerning the extent to
which the effects of high concentrations of ROFA or other combustion-source particle mixes can
be extrapolated to help interpret ambient air PM effects. However, these studies provide some
insight into the relative toxicity of contributing sources or specific PM components.
Analogous issues arise with evaluation of the toxicity of PM emitted from mobile source
combustion devices, e.g., diesel and gasoline vehicle engines. Complex combustion-related
mixtures in such mobile source emissions include many different types of particles and gaseous
compounds in high concentrations that are not necessarily representative of ambient PM derived
from such sources after passage through particle traps, catalytic converters, exhaust pipes, etc.
For example, ultrafine particles emitted from gasoline and diesel engines are reduced in numbers
and concentrations as they agglomerate to form larger, accumulation-mode particles as they cool
in passing through exhaust systems and/or as they undergo further physical and chemical
transformation as they "age" in ambient air. Further complicating evaluation of the toxicity of
mobile source emission components is: (1) the difficulty in separating out toxic effects
attributable to particles versus those of gaseous components in automotive exhausts; and (2) the
changing nature of those exhaust mixes as a function of variations in engine operating mode
9-54
-------
(e.g., cold start versus warm start or "light" versus "heavy" load operation, etc.) and changes in
engine technology (e.g., "old diesels" versus "new diesels").
The in vivo and in vitro PM exposure studies have almost exclusively used PM10 or PM2 5
as particle size cutoffs for studying the effects of ambient PM. Collection and study of particles
in these size fractions has been made easier by widespread availability of ambient sampling
equipment for PM10 and PM2 5. However, other important size fractions, such as the coarse
fraction (PM10_2 5) and PMX 0, have largely been ignored; and only limited toxicology data are
now available to assess effects of these particle sizes. Similarly, relatively little research has
addressed mechanisms by which organic compounds may contribute to ambient PM-related
effects.
9.2.3.1.2 Interspecies Comparisons of Experimental Results
Much of the new toxicologic data assessed in Chapter 7 and discussed here was derived
from either: (a) in vivo exposures of human subjects or laboratory animals via inhalation
exposures or instillation of PM materials; or (b) in vitro exposures of various (mostly respiratory
tract) cells or tissues to diverse types of PM.
Of the three common experimental approaches for studying PM health effects, inhalation
studies provide the most realistic exposure scenarios and physiologically best mimic biological
reactions to ambient PM. However, because they are expensive, typically require large samples,
are time consuming, and require specialized equipment and personnel, they are often
supplemented by instillation and in vitro studies. Instillation studies, in which particles
suspended in a carrier such as physiological saline are applied to the airways, have certain
advantages over in vitro studies. The exposed cells have normal attachments to basement
membranes and adjacent cells, circulatory support, surrounding cells and normal endocrine,
exocrine, and neuronal relationships. Although the TB region is most heavily dosed in such
studies, alveolar regions can also be exposed via instillation techniques. In vitro studies using
live cells are cost-effective, allow for precise dose delivery, and provide a useful avenue by
which to conduct rapid PM mechanistic and comparative toxicity studies. Often, initial
information on the likely mechanisms of action of particles is obtained through in vitro
9-55
-------
techniques. For PM toxicologic studies, dose selection is important to avoid overwhelming
normal defense mechanisms.
As already noted, the experimental exposure conditions used in these studies are typically
different from those experienced through inhalation of airborne PM by human populations in
ambient environments. To help place the lexicologically relevant concentrations/doses into
context in relation to ambient conditions, EPA carried out illustrative dosimetric/extrapolation
modeling analyses (described in Appendix 7A) to provide comparisons between the high
exposure doses typically used in toxicological studies and doses more typical of human
exposures under ambient conditions. Building upon advances in dosimetric modeling discussed
in Chapter 6, Appendix 7A provides analyses of relationships between rat and human lung doses
predicted for various exposure scenarios ranging from ambient PM exposures to PM instillations
into the lung. As noted in Appendix 7A, establishing firm linkages between exposure and dose
requires consideration of particle characteristics and biological normalizing factors. These
analyses and interpretation of their results provide context for exposure concentrations used and
toxicological results assessed here.
It is difficult to compare particle deposition and clearance among different inhalation and
instillation studies because of differences in experimental methods and in quantification of
particle deposition and clearance. In brief, inhalation may result in deposition within the ET
region, the extent of which depends on the size of the particles used; but intratracheal instillation
bypasses this portion of the respiratory tract and delivers particles directly into the TB tree.
Inhalation generally results in a fairly homogeneous distribution of particles throughout the
lungs, relative to instillation, which is typified by heterogeneous distribution and high focal
levels of particles. This disparity in distribution likely impacts clearance pathways, dose to cells
and tissues, and systemic absorption. This is reflected, for example, by particle burdens within
macrophages, those from animals inhaling particles being burdened more homogeneously and
those with instilled particles showing some populations of cells with heavy burdens and others
with no particles. Also, some studies have found greater percent retention of instilled than
inhaled particles, at least up to 30 days postexposure, while others report similar clearance rates.
Exposure method, thus, clearly influences dose distribution; and, possibly clearance, thus
necessitating much caution in interpreting results from instillation studies.
9-56
-------
In many studies, both toxicologic and epidemiologic, health endpoints are presented and
analyzed as a function of exposure concentration. However, it is generally accepted that the
dose to target cells or tissues, rather than exposure concentration per se, is responsible for
adverse responses. Experimental exposure concentrations can be estimated that should result in
the same tissue dose in a rat as received by a human exposed to various levels of ambient PM as
a function of dose metric, normalizing factor, and level of human exertion. As no single dose
metric nor normalizing factor appears to be appropriate for all situations, numerous potential
exposure scenarios were considered in Appendix 7A. Optimally, the dose metrics and
normalizing factors should be based on the biological mechanisms mediating an effect. For
some effects, the mass of soluble PM depositing in a region of the lung may be an appropriate
dose metric. For example, an appropriate normalizing factor for soluble PM could be the surface
area of the airways for irritants, whereas body mass might be more suitable when considering
systemic effects. The parameters chosen can dramatically affect the rat exposure concentration
estimated to be required to provide a normalized dose equivalent to that occurring in a human, as
illustrated in Tables 7A-7a through 7A-9b of Appendix 7A.
Representative dosimetric calculations provided in Appendix 7A indicate that higher PM
concentration exposures in rats than in humans are needed under certain conditions in order to
achieve nominally similar acute doses per lung tissue surface area in exercising humans exposed
to ambient PM while undergoing moderate to high exertion. However, for resting or light
exertion situations, lower rat exposure concentrations are adequate to produce equivalent lung
tissue doses. Also, given that rats clear PM much faster than humans, Appendix 7A dosimetric
modeling predicted that much higher exposure concentrations in the rat are required to simulate
the retained burden of poorly soluble particles which builds up over years of human ambient PM
exposure. In resuspended PM, used in some inhalation studies, the smaller particles found in the
accumulation and Aitken modes of the original atmospheric aerosol are aggregated onto (or into)
larger particles and are not fully disaggregated during the resuspension process. Thus, for dose
metrics based on particle surface area or number, very high exposure concentration of
resuspended PM for rats would be required to provide a dose equivalent to that received by
humans exposed to ambient atmospheric aerosols.
9-57
-------
The dose to the lung can be estimated for both animal and human inhalation studies. These
analyses make it possible to compare biological responses as a function of dose rather than just
exposure. Equal lung doses should not be assumed in comparing studies, even if PM mass
concentrations, animal species, and exposure times are identical. Differences in the aerosol size
distributions to which animals are exposed also affect dose delivered or retained. For example,
in an Appendix 7A comparison of several CAPs studies, one study was estimated to have
1.7 times the alveolar dose of another study despite a 10% lower exposure concentration. Thus,
to make accurate estimates of dose, it is essential to have accurate and complete information
regarding exposure conditions, i.e., not only concentration and duration of exposure, but also the
aerosol size distribution and the level of exertion (and hence breathing rates) for exposed
subjects.
It was obviously not feasible, given the complexity involved, to attempt extrapolation
modeling for more than a few illustrative health endpoints from among those evaluated in the
vast array of studies assessed in Chapter 7. Such calculations require knowledge about the
characteristics associated with the particles, the exposed subject and the environmental exposure
scenario. Hence, each study can present a unique dosimetric analysis. In most cases, it is useful
to know the relationship between the surface doses in instillation studies and realistic local
surface doses that could occur in humans. However, providing some illustrative modeling
results here should be of value in helping to provide a context by which to gauge the potential
relevance of experimental results for ambient human exposure conditions.
PMNInflux as a Marker for Lung Inflammation
Various types of particulate materials have been shown to cause inflammation of the lung
by migration of polymorphonuclear (PMN) cells (predominantly neutrophils) into the airways,
as discussed in Chapter 7 and summarized below. Alveolar macrophages (AMs) and PMN cells,
constitute an important lung defense mechanism triggered by invasion of PM, bacteria, and some
other foreign matter. The PMN cells, once in the lung, ingest PM and may then degranulate,
forming hydrogen peroxide and superoxide anions. Excessive quantities of PM in the lung can
cause the lysosomal enzymes in PMN cells to enter the extracellular fluid, creating further
9-58
-------
inflammatory responses. Also, PMN cells produce thromboxanes, leukotrienes, and
prostaglandins.
Three new studies discussed in Chapter 7 and Appendix 7A provide data on PMN cell
increases following CAPs exposure. Analysis of PMN data from exposures of rats and humans
to CAPs in these studies confirm that rats exhibit analogous effects on PMN cells as seen in
humans in response to CAPs exposures, although the varying composition of the CAPs materials
from day-to-day or location-to-location and other considerations do not allow confident
evaluation of whether healthy humans may be more or less susceptible to the inflammatory
effects of CAPs than are rats based on currently available data.
Inhibition of Phagocytosis by PM Exposure
Phagocytosis is a form of endocytosis wherein bacteria, dead tissue, or other foreign
material (e.g., inhaled ambient particles) are engulfed by phagocytic cells such as AMs, PMN
cells, or monocytes (MO) as part of normal lung defense mechanisms. Increased numbers of
these cells in lung tissue are an indicator of normal mobilization of lung defenses in response to
infection or deposition of inhaled particles. Inhibition of phagocytosis signals interference with
lung defense mechanisms. If an AM is overwhelmed by the amount or toxicity of ingested
material, that material may be released along with the AM's digestive enzymes onto the alveolar
surface and the numbers of AM or their phagocytic activities may decrease. Several in vitro
studies discussed in Chapter 7 have shown that, in certain instances, one or another type of PM
has caused an inhibition of phagocytosis. As with other endpoints affected by PM, this
inhibitory effect is determined by the size and composition of the specific particle mixes tested.
Comparison in Chapter 7 of in vitro rodent and human data evaluating inhibition of
phagocytosis suggest some important species differences. Human AMs demonstrated inhibition
of phagocytosis at 0.05 ng/cell (Utah Valley PM) and 0.2 to 0.5 ng/cell (UAP and ROFA).
A mouse AM cell line showed inhibition of phagocytosis at concentrations of 0.013 to
0.025 ng/cell of TiO2 and carbon black. However, hamster AMs showed no inhibition of
phagocytosis at doses up to 0.04 ng/cell CAPs and 0.4 ng/cell ROFA. Differences in inhibition
may be attributed to interspecies variability in AM capacity, wherein rodent AMs are smaller,
have less capacity for phagocytosis, and appear to be inhibited at a lower burden of PM per cell
9-59
-------
than human AM. It must be noted that these in vitro exposures are at extremely high doses,
exposing each cell to tens to hundreds of particles at physiologically improbable levels unlikely
to be experienced as the result of human exposures to current U.S. ambient air PM (except
possibly, under very extreme conditions).
9.2.3.1.3 Inhaled Particles as Carriers of Other Toxic Agents
In Chapter 2, it was noted that, although water vapor is not considered an air pollutant
per se, particle bound water (PBW) may serve as a carrier for other toxic pollutants. Wilson
(1995) proposed that water-soluble gases that are usually removed by deposition to wet surfaces
in the upper (ET) regions of the respiratory tract may be dissolved in PBW and be carried into
lower regions (TB, A) of the respiratory tract. Thus, PBW could be a vector by which certain
toxic gases commonly found in polluted air masses may be delivered in enhanced proportions to
deep lung regions, including water-soluble gases such as: oxidants (e.g., H2O2, organic
peroxides); acid gases (e.g., SO2, HC1, HONO, formic acid); and polar organic species.
Kao and Friedlander (1995) also noted that many short-lived chemical species in the gas or
particle phase (such as free radicals) in ambient aerosols may not still be present in sampled
materials when analyzed hours to weeks (or even longer) after collection on filters and being
stored. Also, the unmeasured reactive but metastable species may be much more biochemically
active than the resulting stable components collected or remaining on analyzed filters. They also
noted that since inhalation toxicology studies often do not include the potential for metastable
species and reactive intermediaries to be present, then such studies could greatly underestimate
the effects seen in field or epidemiological studies. Friedlander and Yeh (1998) further noted
that atmospheric submicron (< 1.0 jim) aerosols contain short-lived reaction intermediaries (e.g.,
hydrogen peroxides and other peroxides) formed in clouds and rain water. Also, they indicated
that: (1) hydrogen peroxide particle phase concentrations fall in a toxicologic range capable of
eliciting biochemical effects on respiratory tract airway epithelial cells; (2) this may help to
explain epidemiologic results indicating health effects to be associated with sulfate or other fine
particle aerosols; and (3) such aerosols may be surrogate indicators for hydrogen peroxide or
other species.
9-60
-------
Certain physical modeling of gas-particle-mucus heat and mass transport in human airways
suggests that very soluble gases (e.g., H2O2, formaldehyde) may be largely evaporated from
particles < 0.1 jim diameter before reaching A regions of the lung, but particles > -0.3 jim can
efficiently carry such gases into the air exchange region of the lung. Also one new toxicological
study discussed in Chapter 7 (Section 7.9.1) evaluated whether certain commonly present
hygroscopic components of ambient PM can transport H2O2 into the lower lung and thereby
exert or enhance toxic effects. More specifically, rats exposed by inhalation to combinations
of (NH4) SO4 (0.3 to 0.4 MMD) and H2O2 exhibited enhanced biochemical effects that were
interpreted by the authors as showing that biological effects of inhaled PM are augmented by
coexposure to sulfate and peroxide, including altered production of cytokine mediators by
alveolar macrophages.
The information summarized above has important implications for interpreting and
understanding epidemiologic and experimental toxicology results discussed in ensuing sections
of this chapter. Also, of much importance is dosimetric information discussed in Chapter 6
which indicates that hygroscopicity affects particle deposition patterns in the respiratory tract,
such that under high humidity conditions one can expect increased deposition of small
nucleation (< 0.1 jim) ultrafine particles and larger accumulation-mode (> 0.5 jim) particles, the
latter of which are able to grow to exceed 1.0 jim and both of which would contain enhanced
amounts of PBW and other toxic agents (e.g., SO2, peroxide, aldehydes, etc.). Also, to some
extent, growth of ultrafine and accumulation mode particles under high humidity conditions
would likely enhance particle "hot spot" deposition at airway branching points and increase PM
doses to lung tissue at those points. Enhanced deposition and tissue doses would likely
exacerbate PM respiratory effects in particularly susceptible population groups, e.g., asthmatics,
COPD patients, and others with severe cardiopulmonary conditions.
In addition to recognition that particle-bound water may serve as a carrier of other toxic
agents, there is growing recognition that bioaerosols likely have the potential to contribute to
some ambient PM effects, in part, via their serving as carriers of toxic agents or via their
attaching to and being carried by nonbiological particles. Bioaerosols, from sources such as
plants, fungi, and microorganisms, range in size from 0.01 jim to > 20 jim. Although they
9-61
-------
typically only comprise a small fraction of ambient PM, they likely contribute to some types of
ambient PM-related health effects.
Intact pollen grains from plants, trees and grasses are most abundant during warm/humid
spring/summer months. When they deposit in upper airways, they induce allergic rhinitis.
However, allergen-laden cytoplasmic fragments (-0.1 to 0.4 jim in size) of pollen grains (which
rupture under high moisture conditions) can enter the deep lung, where they can exacerbate
asthma. Binding of allergen-laden pollen cytoplasmic fragments to ambient fine particles, e.g.,
diesel particulate matter (DPM) has also been observed; and synergistic interactions between
pollen debris and other ambient PM, e.g., the polycyclic hydrocarbon component of diesel
exhaust, may be a mechanism that increases incidence of asthma morbidity and mortality.
Pollen granules can also act as vectors for binding of other bioaerosols (e.g., endotoxins, fungi or
fungal fragments, glucans) and thereby enhance their inhalation and deposition in the respiratory
tract, as well.
Fungal spores and fungal fragments are among the largest and most consistently present
bioaerosols found outdoors (levels being higher during warm/humid months). Certain molds and
other fungi cause allergic rhinitis and asthma, which is highly dependent on seasonal variations
in concentration. Exposures have been linked in epidemiologic studies to asthma hospitalization
and death. Some proliferate very effectively on wet cellulose materials (at times posing serious
indoor contamination problems), thus raising the possibility that airborne cellulose-containing
plant debris (which otherwise may be non-toxic when dry) may serve as effective vectors for
proliferation of fungi and their delivery into the lung under high humidity ambient conditions.
Exposures to other soil-dwelling fungi found in contaminated airborne soil particles entrained in
windblown dust stirred up by natural or anthropogenic activities has been linked to increased
risk of serious respiratory infections (Valley Fever) in endemic areas of California and the
southwestern United States (see Appendix 7B).
Bacteria and viruses are also significant bioaerosols. Bacteria have endotoxins in their
outer cell membrane, which trigger production of cytokines and a cascade of inflammation.
Ambient airborne levels of endotoxins vary with seasons (being higher in warm/humid periods
and low in colder months). Another cell wall component of bacteria and fungi, (H3)-P-D-
glucan, has also been shown to cause respiratory inflammation.
9-62
-------
Based on the above, it appears that certain ambient bioaerosols (e.g., pollen, fungi,
endotoxins, glucans) that become abundant during warm/humid weather have the potential to
contribute to seasonal increases in PM-associated risk during spring/summer months in many
U.S. areas, but not during colder winter months. In addition, the copresence of nonbiological
particles, serving as vectors concentrating such bioaerosols and enhancing their delivery into the
deep lung, also appears likely to be important. For example, airborne endotoxins have been
shown to be associated with both fine and coarse thoracic ambient particles (albeit higher with
the coarse PM); and cytoplasmic pollen fragments have been found attached to airborne diesel
particles. It thusly appears that airborne anthropogenic particles (both in fine and coarse size
ranges) as well as naturally-generated biological particles likely enhance the risk for bioaerosols-
stimulated effects.
9.2.3.2 Biological Plausibility and Coherence of Evidence for Different Health
Endpoint Categories
This section is organized to integrate epidemiologic, toxicologic, and mechanistic
information for each of four major categories of health endpoints, i.e., (a) cardiovascular;
(b) respiratory; (c) lung cancer; and (d) fetal/infant development and mortality, purported to be
associated epidemiologically with either short- or long-term ambient PM exposures. Each
subsection concisely summarizes pertinent key information and then arrives at conclusions as to
the plausibility of effects being reasonably attributable to fine and coarse thoracic particles
and/or subcomponents.
9.2.3.2.1 Cardiovascular-Related Health Endpoints
As noted in Section 9.2.2, a number of epidemiologic studies (a) show associations
between short-term and/or chronic ambient PM exposures and increases in cardiac-related deaths
and/or morbidity indicators and (b) indicate that the risk of PM-related cardiac effects may be as
great or greater than those attributed to respiratory causes (see Chapter 8). Hypothesized
mechanisms thought to be involved in cardiovascular responses to PM exposure (as discussed in
Chapter 7) include: (a) effects on autonomic nervous system control of cardiovascular functions
and (b) pathophysiologic effects on certain blood chemistry parameters involved in control of
9-63
-------
blood clotting or otherwise impacting cardiovascular integrity. It should be noted that most new
PM cardiovascular effects research has focused on exposure to fine particles; little evidence is
available on the effects of coarse fraction particles.
With regard to autonomic nervous system control, the heart receives both parasympathetic
and sympathetic inputs that decrease or increase heart rate, respectively. Vasoconstriction,
possibly due to release of endothelin elicited by PM, could cause increased blood pressure and
its detection by baroreceptors. Parasympathetic neural input may then be increased to the heart,
slowing heart rate and decreasing cardiac output (which is sensed by aortic and carotid
chemoreceptors). These, in turn, may stimulate a sympathetic response, manifested by increased
heart rate and contractile force, thus increasing cardiac output. This arrhythmogenesis and
altered cardiac output in either direction can be life-threatening to susceptible individuals.
Pathophysiological changes in cardiac function can be detected by electrocardiographic
(ECG) recordings, with certain ECG parameters (e.g., heart rate variability or HRV) now often
being used as indicators of PM-induced cardiac effects. HRV is a reflection of the overall
autonomic control of the heart and can be divided into time and frequency measures. Frequency
measures of variability help to resolve parasympathetic and sympathetic influences on the heart
better than do time domain measurements. Under some circumstances (as discussed in
Chapter 7), HRV provides insight into sympathetic nervous activity, but more commonly it is a
good measure of parasympathetic modulation. Heart rate variability can be used to judge the
relative influences of sympathetic and parasympathetic forces on the heart, but short-term
spectral parameters (i.e., measures averaged over five minute intervals) can vary as much as
4-fold during the course of a 1-h period. Despite the inherent variability of short-term HRV
measures during routine daily activity, long-term measures show excellent day-to-day
reproducibility. Given the inherent variability in the minute-to-minute spectral measurements,
much care is required in the design of studies using HRV techniques and in interpretation of
HRV results. Still, studies utilizing measures of HRV can provide insight into relationships
between perturbations of the internal or external environment and subsequent changes in the
modulation of autonomic neural input to the heart.
Using both time and frequency domain parameters, HRV has been studied as a marker of
medical prognosis in human clinical populations, most frequently in coronary artery disease
9-64
-------
populations, particularly in the post-myocardial infarction (post-Mi) population. Those variables
most closely correlated with parasympathetic tone appear to have the strongest predictive value
in heart disease populations. The altered HRV itself is not the causative agent, but rather altered
HRV (including changes in HRV associated with exposure to PM) is simply a marker for
enhanced risk of serious cardiac events (e.g., arrhythmia, sudden cardiac death).
Another route by which PM could exert deleterious cardiovascular effects may involve
ambient PM effects on endothelial function. PM exposure may affect blood coagulation through
endothelial injury that results in platelet activation. This then could initiate a cascade of effects
(e.g., platelet activation and/or aggregation, increased blood fibrinogen and fibrin formation
modulated by Factor VII, etc.), leading to increased formation of blood clots. Or PM taken up
into the systemic circulation could possibly affect clot lysing events that normally terminate the
blood coagulation cascade. Newly available studies have measured various blood substances to
evaluate possible PM-induced effects on blood coagulation. Another significant effect of PM
exposure could be vascular inflammation, which induces release of C-reactive proteins and
cytokines that may cause further inflammatory responses which, on a chronic basis, can lead to
atherosclerosis. In narrowed coronary arteries, clots formed by the aforementioned cascade may
block blood flow, resulting in acute myocardial infarction.
Small prothrombotic changes in blood coagulation parameters in a large population can
have substantial effects on the incidence and prevalence of cardiovascular disease events
(Di Minno and Mancini, 1990; Braunwald, 1997; Lowe et al., 1997). Altered coagulation, for
example, could increase heart attack risk through formation of clots on atherosclerotic plaques in
coronary arteries that cut off blood supply to the myocardium or induce ischemic strokes via
clots forming or lodging in the carotid arteries and blocking blood flow to the brain. Also,
evidence exists for formation of small thrombi being common in persons with atherosclerosis
(Meade et al., 1993); and whether such thrombi lead to more serious effects (heart attack, stroke)
depends in part on the balance between thrombogenic factors underlying blood clot formation
and fibrinolytic factors that lyse clots. Increased sympathetic activity is thought to cause
prothrombotic changes in blood coagulation parameters such that even small, homeostatic
modulations of coagulation within a normal range could translate into significant increased risk
for heart attack.
9-65
-------
Another possible effect of PM exposure could be plasma extravasation from post-capillary
venules. The mechanisms by which this occurs are thought to include the release of peptides
(such as neurokinin A, substance P, and calcitonin-gene-related peptide) from unmyelinated
sensory nerves near to or on the blood vessels. These peptides bind to receptors on the
endothelial cells of vessels and create gaps, allowing leakage of plasma, which is one component
of neurogenic inflammation (Piedimonte et al., 1992; Baluk et al., 1992).
Thus, alterations in cardiovascular functions due to PM exposure could be signaled by
small PM-related (a) changes in blood coagulation cascade indicators, e.g., altered blood platelet,
fibrinogen, or Factor VII levels or decreased tissue plasminogen activator (TPA) levels;
(b) increased C-reactive protein or cytokines, possibly contributing to increased atherosclerosis
plaque formation and/or blood coagulation; (c) increased blood pressure; and/or (d) certain
alterations in heart rate, heart rate variability, or other ECG indicators indicative of shifts in
parasympathetic/sympathetic neural inputs to the heart or other underlying cardiac pertubations.
These alterations, while not likely to have significant impact in healthy individuals, may be
deleterious in susceptible individuals with underlying cardiopulmonary disease.
Coherence Between Epidemiologic and Experimental Evidence for Cardiovascular Effects
Considering first the evidence from epidemiologic studies conducted within a given
location (e.g., the same urban area), recent studies have reported associations for PM with both
mortality and hospital admissions for cardiovascular diseases in several U.S. cities. For
example, in Chicago (Figure 8-24), associations were reported between PM10 and cardiovascular
mortality and cardiovascular hospital admissions. In Los Angeles (Figure 8-25), associations
were found between PM10 and cardiovascular mortality, cardiovascular hospital admissions, and
admissions for specific categories of cardiovascular disease (e.g., myocardial infarction,
congestive heart failure, cardiac arrhythmia, cerebrovascular and occlusive stroke); some studies
included associations with PM25. In addition, one recent study in a group of Los Angeles
residents with COPD reported associations between PM10 and diastolic and systolic blood
pressure, although no associations were reported for heart rhythm measures (Section 8.3.1.3.4).
In Detroit (Figure 8-27), as well, associations were seen between PM10 and cardiovascular
mortality, cardiovascular hospital admissions and admissions for specific categories of
9-66
-------
cardiovascular disease (including ischemic heart disease, heart failure, dysrrhythmia, stroke);
associations were also reported with PM2 5 and PM10_2 5.
More broadly, the fuller array of epidemiologic studies shows associations between various
ambient PM indices and a range of cardiovascular health outcomes, from mortality and
hospitalization for various cardiovascular diseases to recent evidence for associations with
incidence of myocardial infarctions and physiological or biochemical indicators of
cardiovascular health. The newer evidence includes epidemiologic panel studies reporting
changes in blood characteristics (e.g., increased fibrinogen or C-reactive protein levels) related
to increased risk of ischemic heart disease, and some indications of changes related to heart
rhythm, including cardiac arrhythmia or changes in HRV that may be linked with more serious
cardiac effects. While further research is needed to more firmly establish and understand links
between particles and these more subtle endpoints, the newly available results provide
suggestive evidence for a chain of endpoints linked to potential mechanisms for cardiac effects.
These more recent studies have mainly found associations for PM10 and PM2 5; only one study
included PM10_25 and reported no associations with HRV changes (Section 8.3.1.3.4). As for
lags seen epidemiologically between PM exposure and observed effects, acute short-term
(< 24-h) exposures to ambient PM appear to exert cardiovascular/systemic effects rather quickly,
with peak lags of 0-1 days being generally seen, and one study reporting myocardial infarction
increases even as early as 2 h post exposure.
There were few toxicologic studies assessed in the 1996 PM AQCD that evaluated
cardiovascular system effects of exposures to particulate matter. Since 1996, numerous studies
have now become available that evaluated cardiovascular effects of exposures (via inhalation or
instillation) of ambient PM, constituent components, complex mixtures from PM emission
sources and/or exposures to single PM substances or binary/ternary combinations of particles of
varying chemical composition. Whereas earlier studies tended to focus on healthy animals, the
more recent studies have, in addition, begun to focus on evaluation of PM effects in animal
models of disease states thought to mimic aspects of pathophysiologic states experienced by
compromised humans at increased risk for PM effects.
A growing number of studies have used extracts of collected/stored ambient PM or real-
time concentrated ambient particles (CAPs) drawn from various airsheds (e.g., Boston, New
9-67
-------
York City, etc.) to evaluate cardiovascular and other systemic effects of PM. Most of this
research has focused, again, on fine particles. A number of new animal studies have also used
residual oil fly ash (ROF A) as one type of combustion source particle mix and others have used
other combustion source materials, e.g., domestic oil fly ash (DOFA), coal fly ash (CFA), or
diesel exhaust (DE).
The ensuing discussion focuses mainly on those toxicology studies using ambient or near-
ambient PM concentrations thought to be most relevant to ambient PM exposure situations in the
United States. Controlled human exposure studies have yielded some limited evidence for
ambient PM effects on cardiac physiological function (as indexed by ECG readings) or systemic
endpoints (as indexed by vasopressor control, blood coagulation control, etc.) linked to more
serious cardiovascular events. Blood coagulation effects of inhaled PM were also observed
with CAPs, UAP, and ROF A. Probably of most note, two controlled human exposure CAPs
studies found evidence that ambient levels (-50 to 300 |ig/m3) of inhaled PM25 can produce
biochemical changes (increased fibrinogen) in blood suggestive of PM-related increased risk for
prothrombotic effects. Also, blood fibrinogen levels increased in both normal and compromised
dogs at 69 to 828 |ig/m3; and decreased Factor VII levels were observed by other investigators in
humans with 2-h CAPs exposure at -174 |ig/m3, perhaps reflecting that enzyme being consumed
in an ongoing coagulation process. On the other hand, the same and many other human and
animal studies did not find significant changes in other factors (e.g., increased platelets or their
aggregation) related to blood coagulation control. Additional other studies have shown no
cardiovascular effects in rats and dogs with CAPs exposures of 3-360 |ig/m3.
One excellent example of linkage between cardiovascular results from epidemiological
and toxicological studies is provided by a series of studies conducted in Boston. Recent
epidemiological studies have linked daily or hourly changes in PM2 5 with several cardiovascular
health outcomes: incidence of myocardial infarction was increased in association with PM
exposures 2 hours prior to the health event; increases in recorded discharges from implanted
cardiovertex defibrillators (an indicator of cardiac arrhythmia) were positively associated with
daily PM2 5 concentrations; and decreases in HRV measures were reported (a) in young healthy
boilermakers to be associated with personal PM2 5 measurements and (b) in elderly residents of a
retirement community with ambientPM25 (Section 8.3.1.3.4). Results of toxicological studies in
9-68
-------
Boston using PM2 5 CAPs exposures in dogs are suggestive of changes in cardiac rhythm with
PM2 5 mass and changes in blood parameters with certain PM25 components (Table 7-1). These
findings in both humans and animals, using the same general mix of particles and co-pollutants,
are suggestive of changes in cardiac rhythm and changes in blood parameters. Also, in addition
to the epidemiologic studies conducted in Los Angeles, results from controlled human inhalation
exposures (for 2 h) of healthy adult volunteers to Los Angeles PM25 CAPs suggest effects on
some cardiovascular outcomes (decreased Factor VII blood levels and some cardiac symptoms),
but not with other cardiovascular indicator measures (such as changes in blood fibrinogen levels)
(Table 7-1). More rigorous characterizations of dose-response relationships with
environmentally relevant levels and species of PM will be necessary to evaluate more fully
cardiovascular risks posed by ambient PM exposures.
Limited new evidence is available regarding the effects of different components or
attributes of particles on the cardiovascular system. Recent epidemiological studies reported
slight increases in blood viscosity with ultrafine particle exposures. Little or no evidence is
available on cardiovascular effects of PM components, such as sulfates or acid aerosols. Particle
constituents such as transition metals (e.g., Ni, V, Zn, Fe) have been shown to cause cell injury
and inflammatory responses in toxicologic studies, that may possibly be linked with
cardiovascular health outcomes. Since particles are complex mixtures, studies using factor
analysis or source apportionment methods may be more relevant than studies of individual
components, and the few studies available to date have linked cardiovascular mortality with
several fine particle source categories (Table 9-3).
More limited evidence is available on cardiovascular effects of long-term exposure to
particles. Epidemiologic studies indicate associations between fine particles and mortality from
cardiovascular diseases, although no evidence is currently available regarding long-term PM
exposure and cardiovascular morbidity. Toxicologic studies have not yet been conducted to
investigate potential cardiovascular effects with chronic PM exposures.
Beyond the evidence of coherence and plausibility described above, it is useful to consider
salient hypotheses that have been proposed to account for PM-related effects. The most salient
hypotheses proposed to account for cardiovascular effects of PM are: alterations in coagulability
(Seaton et al., 1995; Sjogren, 1997); cytokine effects on heart tissue (Killingsworth et al., 1997);
9-69
-------
perturbations in both conductive and hypoxemic arrythmogenic mechanisms (Watkinson et al.,
1998; Campen et al., 2000); altered endothelin levels (Vincent et al., 2001); and activation of
neural reflexes (Veronesi and Oortgiesen, 2001). Only limited progress has been made in
obtaining evidence bearing on such hypotheses; and, to date, the strongest evidence found thus
far most clearly supports the plausibility of the first mechanism being involved. Both
epidemiologic and toxicologic studies (largely using fine particles) have found evidence of
ambient PM effects on blood fibrinogen and/or other measures indicative of increased blood
coagulability within 2 to 24 hours following short-term (< 24 h) exposures to ambient or near-
ambient concentrations of urban PM aerosols, with most evidence being available for fine
particles. Much future research using controlled exposures to PM of laboratory animals and
human subjects will be needed, however, to test further such mechanistic hypotheses so as to
more fully understand pathways by which low concentrations of inhaled ambient PM may be
able to produce life-threatening cardiovascular/systemic changes, and a particular research need
is for more research on cardiovascular effects of coarse fraction particles.
9.2.3.2.2 Respiratory-Related Health Endpoints
As noted in Section 9.2.2, a number of epidemiologic studies show associations between
short-term and/or chronic ambient PM exposure and respiratory effects ranging from respiratory-
related mortality to hospitalization or medical visits for respiratory diseases to increased
respiratory symptoms or decreased lung function. Respiratory system effects of PM may be
exerted via several different types of mechanisms of action, including involving direct
pulmonary effects and others secondary to lung injury. Direct pulmonary effects include lung
injury and inflammation; increased airway reactivity and exacerbation of asthma; and increased
susceptibility to infection.
Numerous toxicological studies point towards lung injury and inflammation being
associated with exposure of lung tissue to complex combustion-related PM materials. Important
evidence pointing towards ambient PM causing lung injury and inflammation derives from the
study of ambient PM (PM10 and TSP) materials on filter extracts collected from community air
monitors before, during the temporary closing of a steel mill in Utah Valley, and after its
reopening. Studies in animals and human volunteers reported greater lung inflammatory
9-70
-------
responses with exposure to materials obtained before and after the temporary closing versus that
collected during the plant closing. Further analyses suggested that the metal constituents of
particles may be important contributors to the pulmonary toxicity observed in these studies. Rats
with SO2-induced bronchitis and monocrotaline-treated rats have been reported to have a greater
inflammatory response to concentrated ambient PM than normal rats. The toxicologic studies
suggest that exacerbation of respiratory disease by ambient PM may be caused in part by lung
injury and inflammation.
Toxicologic studies have also indicated that PM exposure can affect pulmonary defense
responses to microbial agents. Studies using combustion related particles, albeit at high doses,
have shown effects such as increased inflammatory responses or mortality rate from respiratory
infections, compared with animals exposed to infectious agents without PM exposure (as
discussed in Section 7.5.4).
Finally, PM exposure may result in increased airway reactivity and exacerbation of asthma.
The strongest evidence supporting this hypothesis is from studies on DPM, an example of fine
PM. Diesel particulate matter has been shown to increase production of antigen-specific IgE in
mice and humans (summarized in Section 7.2.1.2).
Coherence Between Epidemiologic and Experimental Evidence for Respiratory Effects
Recent time series epidemiologic studies have reported associations between short-term
(24-h) PM exposures for various indices and respiratory-related mortality and hospital
admissions for respiratory diseases in cities such as Chicago, Los Angeles and Detroit. These
studies, and others in Seattle and Pittsburgh, have also reported associations between PM and
hospitalization or emergency department visits for asthma, pneumonia and COPD, as well as
physicians visits for respiratory diseases. In addition, new evidence exists for ambient PM
associations with reductions in pulmonary function and/or increased respiratory symptoms,
especially of note in relation to asthmatic or other chronic lung disease individuals. Respiratory
effects typically exhibit somewhat longer and more extended lag periods, from 1 to 2 days on out
to a week or so after PM exposure, than do cardiovascular effects.
Some epidemiologic studies also indicate associations between long-term (years to
decades) exposures to ambient PM (especially fine particles) and mortality due to
9-71
-------
cardiopulmonary causes, although other recent studies indicated that such fine PM associations
may be more strongly linked to cardiovascular diseases than respiratory disease. Long-term
exposure to PM has also been found to be associated with potential development of chronic
respiratory diseases and reductions in lung function.
The respiratory effects of PM with varying physical and chemical characteristics have been
extensively studied for more than 30 years using a wide range of techniques and with exposure
durations ranging from brief periods to months. The most extensively studied materials have
been sulfates and acid aerosols formed as secondary pollutants in the atmosphere. Fly ash from
coal-fired power plants or other coal-combustion sources has been less extensively studied.
Controlled exposures to crustal materials, e.g., those in Mt. St. Helens volcanic ash, have also
been studied. The toxicological data available today provide little basis for concluding that these
types of specific PM constituents have substantial respiratory effects at current U.S. ambient
levels of exposure. Recently, ROFA, a very specific kind of PM derived from oil combustion,
has been studied extensively and found to produce a range of respiratory effects, especially lung
inflammation, mainly attributable to its very high metal content that is several orders of
magnitude (100s of times) higher than ambient concentrations typically found in U.S.
ambient air.
Probably of more direct relevance for present purposes, other recent studies evaluating
controlled human exposures to CAPs from diverse locations (e.g., Boston, New York City ,
Los Angeles, Toronto, and Chapel Hill, NC) have found little or no effects on pulmonary
function or respiratory symptoms in healthy human adults acutely exposed (for 2 h) by
inhalation to CAPs at concentrations that ranged from about 25 up to about 300 |ig/m3. Some
indications of mild lung inflammation were reported with such exposures in some of the studies,
but not others. Analogous controlled exposures to CAPs of rats, hamsters, and dogs at
concentrations varying across a range of-100 to 1000 |ig/m3 for 1-6 h/day for 1 to 3 days
yielded similar minimal effects on respiratory functions, but did yield some signs of mild
inflammation in normal healthy animals and somewhat enhanced indications of lung
inflammation in at least one compromised animal model of chronic bronchitis. Follow-up
evaluations have produced new evidence implicating transition metal components of ambient
PM from diverse locations and of ROFA as inducing inflammatory responses. Another
9-72
-------
inhalation study found indications of some impairment of lung immune defense functions and
exacerbation of bacterial infection with an acute (3 h) exposure of rats to New York City CAPs
(at 100-350 |ig/m3). Although concentration ranges were reported in the above inhalation
studies, it is difficult to discern the actual lowest concentrations at which effects were observed.
Also, CAPs, UAPs, and ROFA have all been used in in vitro experiments to demonstrate
effects and explore mechanisms whereby PM causes effects. Approximately 0.02 to 0.2 ng
PM/cell is the concentration range where in vitro effects (e.g., cytokine production, inhibition of
phagocytosis, and oxidant formation) were observed, though these doses are extremely high and
are unlikely to be approached with exposures to ambient levels of PM currently found in U.S.
airsheds (except, possibly, under unusual circumstances, e.g., exposure to dense smoke from
forest fires).
A set of epidemiologic, toxicologic and controlled human exposure studies on effects of
particles from the Utah Valley area has linked PM10 with respiratory system effects or, more
specifically, lung inflammation. A special feature of these studies was the closure of a steel mill,
a major source of PM emissions in the area, for a 13-month period. An epidemiologic study
reported that respiratory hospital admissions for children were reduced during the period when
the source was not operating (see Chapter 8, Section 8.2.3.4). New toxicologic and human
studies then used extracts of ambient particles collected on filters from ambient PM10 monitors
operating during the time periods before, during and after steel mill closure. Intratracheal
instillation of particle extracts in both human volunteers and animals resulted in greater lung
inflammatory responses for materials obtained before and after the plant closure period (as
discussed in Chapter 7, Section 7.3.1.2). The health responses were indicative of inflammatory
changes in the lung, including increased levels of neutrophils, protein and inflammatory
cytokines. However, consideration of dosimetric analyses (see Appendix 7A) indicates that the
bolus instillation doses of particles used in these experiments resulted in a single-dose exposure
comparable to the cumulative dose that would result from extended (for 6-9 weeks) continuous
exposure to the higher-end of the range of concentrations of PM10 that the community might
have experienced during wintertime inversions in the Utah Valley. In vitro studies using a
human airway epithelial cell line and primary rat airway epithelial cells also showed evidence
for inflammatory responses, such as increases in cytokine levels, indicators of oxidative response
9-73
-------
in alveolar macrophages and some evidence of cytotoxicity (see Section 7.4.2). Additional
evaluations indicate that the in vivo and in vitro inflammatory responses observed were
attributable to elevated metal content present in the particle extracts during the periods when the
steel mill was operating. This body of evidence provides coherent links between results of
community epidemiologic studies reporting increases in respiratory hospitalization with
toxicologic evidence of respiratory inflammation in humans and animals.
Some evidence is also available on respiratory effects of different components or attributes
of particles, especially fine particles. A few epidemiologic studies have reported associations
between ultrafine particles (measured as particle number) with respiratory symptoms or
decreased lung function. In addition, toxicologic studies have used various types of ultrafine
particles (e.g., carbon black), and reported greater inflammatory responses than those seen with
fine particle mass for the same type of particles. The relative importance of differing
composition or surface area for these effects remains to be determined.
Fine particulate sulfates and acid aerosols have been associated with respiratory
hospitalization, symptoms or and decreased pulmonary function in epidemiological studies, in
both short-term and long-term exposure studies. Toxicological studies, however, have reported
pulmonary or inflammatory effects with acid aerosol or sulfate exposures only at fairly high
concentrations (hundreds of |ig/m3). It appears likely that, in the epidemiological studies,
sulfates are serving as an indicator of particle mixtures or sources of particles. Toxicological
studies have also reported that transition metals in or on fine particles (e.g., Ni, V, Zn, Fe) cause
cell injury and inflammatory responses and, so, they may contribute to associations with
respiratory health outcomes reported in epidemiological studies. Also, recent studies also show
that diesel exhaust particles may exacerbate allergic responses to inhaled antigens.
As summarized above (Section 9.2.3.3.1) and discussed in more detail in Chapter 7
(Section 7.3.6) biological constituents of particles (e.g., fungal spores, plant and insect
fragments, airborne bacteria) have been clearly linked with allergic, pulmonary or inflammatory
responses in toxicological studies, and with respiratory symptoms or lung function changes in
epidemiologic studies. Though the 1996 PM AQCD had concluded that bioaerosols at ambient
levels were unlikely to account for PM-related health effects, more recent findings suggest that
biogenic materials in ambient air may be attached to either natural or anthropogenic particles and
9-74
-------
be carried by them into the deep lung and concentrated at "hot spots", where enhanced doses to
tissue may produce exacerbation of lung inflammatory and allergic responses to of the
bioaerosols.
For the most part, information regarding components of particles has come from studies of
fine particles. Some of these components, particularly biogenic material and metals, can also be
important components of coarse fraction particles. More research involving the systematic
conduct of studies of potential respiratory effects of major components of PM commonly found
in different size fractions in the United States is needed, in recognition that PM of different
composition and from different sources can vary markedly in its potency for producing
respiratory toxicity. Of particular importance are studies that more systematically evaluate
mixtures of ambient constituents found in various airsheds, including short-lived species, e.g.,
peroxides.
9.2.3.2.3 Lung Cancer
Historical evidence linking cancer with PM exposures includes epidemiological 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. Numerous past ecological
and case-control studies of PM and lung cancer have generally found lung cancer relative risks
greater than 1.0 to be associated with living in areas having higher PM exposures despite
possible problems with respect to potential measurement errors for exposure and other risk
factors. The 1996 PM AQCD (Section 8.4.6.4) further noted certain recently published
prospective cohort study results (e.g., those from the ACS study) which found positive, but not
statistically significant, associations between PM2 5 and lung cancer mortality—leading to a
bottom line conclusion in that document that insufficient evidence then existed by which to link
ambient PM exposures to increased risk of lung cancer.
More recent epidemiologic studies published since the 1996 PM AQCD have expanded
upon and extended the earlier findings, including both (a) reported significant associations
between long-term exposure to fine particles and lung cancer mortality in further analyses of
data for the ACS and AHSMOG cohorts, and (b) suggestive evidence for PM-related increases
in lung cancer incidence in analyses using AHSMOG cohort data (see Section 8.4.6.4,
9-75
-------
Tables 8-10 and 8-12). The 2002 extended ACS analysis included additional mortality data from
that cohort, inclusion of more recent air quality data , incorporation of statistical modeling
advances, and additional data on potential confounders (such as dietary information); and it
showed a 13% increase in lung cancer mortality per 10 |ig/m3 increase in PM25. The AHSMOG
analysis also included follow-up data from the AHSMOG cohort and yielded a significant
association between PM10 and lung cancer mortality in males, but not in females. In further
follow-up of the results in males only, positive but not statistically significant associations were
reported with PM2 5 and PM10_2 5, and the authors observed that the association was larger in
magnitude for PM2 5 than for PM10_2 5.
Toxicological studies have shown extensive evidence that certain types of particles are
mutagenic or otherwise genotoxic in various types of bioassays, and several recent in vivo and
in vitro studies have suggested that ambient particles are mutagenic. These latter studies have
included exposures to ambient particles collected in Los Angeles, urban areas of Germany, and
high traffic areas in the Netherlands (Section 7.8.1). In Germany, PM25 extracts were found to
have more mutagenicity than extracts of PM10 samples. Also, evidence of mutagenicity has been
reported in studies using exposures to emissions from wood/biomass burning, coal combustion,
and gasoline and diesel engine exhaust. Some of these studies identified polyaromatic
hydrocarbons (PAHs) as well as some gaseous components of the emissions as being more
mutagenic than other portions (Sections 7.8.2, 7.8.3). Such results appear to provide
experimental evidence that adds some degree of plausibility for the reported epidemiologic
findings of ambient PM associations with increased risk of lung cancer. However, this should be
caveated by noting that some of the bioassay results were not indicative of particularly strong
mutagenic responses to the PM sample extracts or components tested, nor is there necessarily a
high degree of correspondence between some of the types of in-vitro genotoxic effects observed
and demonstrated tumorigenic/cancer-causing potential across a broad array of different types of
gaseous and particulate compounds tested over many years.
Thus, recent epidemiological studies support an association between long-term exposure to
fine particles and lung cancer mortality; and the new toxicological studies provide credible
evidence for the biological plausibility of these associations.
9-76
-------
9.2.3.2.4 Fetal and Infant Development/Mortality
A few older cross-sectional studies reviewed in the 1996 PM AQCD reported findings
suggestive of (a) possible TSP relationship to increased postnatal mortality among U.S. infants,
children, and adolescents (aged 0-14 years) and (b) possible associations between early postnatal
mortality among Czech infants (1-12 months). Several more recent studies conducted in the
U.S. have focused on the possible effects of air pollution exposures during pregnancy on the
occurrence of preterm or low birth weight births, both of these being risk factors for a myriad of
later health problems (childhood morbidity/mortality; possible adult morbidity). One study
found results suggestive of prenatal PM10 exposures during the 1 st month of pregnancy or
averaged over 6 weeks prior to birth being associated with increased risk of preterm birth, even
in multipollutant models. However, another large scale U.S. study found little evidence
indicative of prenatal PM10 exposures being related to increased risk of low birth weight,
whereas a new Czech study did find evidence indicative of interuterine growth retardation
(leading to low birth weight) being related to PM2 5 exposures during the first gestational month.
Similarly, analogously mixed results were reported for some new studies that evaluated ambient
PM relationships to early postnatal mortality among U.S., Czech, and Mexican infants. These
results, overall, highlight the need for more research to elucidate potential ambient PM effects on
fetal development/mortality and for postnatal morbidity/mortality.
9.2.3.3 Summary and Conclusions
Consideration of the plausibility and coherence of PM-related effects involves the
integration of the epidemiologic evidence with information derived from other types of studies
(e.g., exposure, dosimetry, toxicology). As discussed in Section 9.2.2, consideration of
epidemiologic evidence alone gave evidence supporting causality for associations between PM
and a range of cardiovascular and respiratory health outcomes. In this section, evidence from
both epidemiologic and toxicologic studies for health outcomes that are logically linked together
was considered.
Epidemiologic studies have reported associations between ambient PM and cardiovascular
effects across a range of endpoints, from cardiovascular mortality to more subtle effects, such as
9-77
-------
changes in electrocardiographic markers of cardiac function, including evidence of cardiac
arrhythmia and altered heart rate variability, or changes in blood characteristics (e.g., alterations
in C-reactive protein levels, fibrinogen levels, blood viscosity, etc.) related to increased risk of
ischemic heart disease. The new epidemiological findings for physiological changes suggest
links to mechanistic pathways that could result in observed cardiovascular morbidity or
mortality, but as described earlier, there are caveats to be considered in the interpretation of these
studies. Important new evidence is available from toxicologic studies that builds support for
plausibility of associations between particles, especially fine particles (or constituents) with
physiological endpoints indicative of increased risk of ischemic heart disease, development or
exacerbation of atherosclerosis or changes in cardiac rhythm. While many research questions
remain, the convergence of evidence related to cardiac health from epidemiologic and
toxicologic studies indicates both coherence and plausibility in this body of evidence.
For respiratory effects, notable new evidence from epidemiological studies substantiates
positive associations between ambient PM concentrations and not only respiratory mortality, but
(a) increased respiratory-related hospital admissions, emergency department, and other medical
visits; (b) increased incidence of asthma and other respiratory symptoms; and (c) decrements in
pulmonary functions. Of much interest are new findings tending to implicate not only fine
particle components but also coarse thoracic (e.g., PM10_25) particles as likely contributing to
exacerbation of various respiratory conditions (e.g., asthma). Also of much interest are
emerging new findings indicative of likely increased occurrence of chronic bronchitis in
association with (especially chronic) PM exposure. The biological pathways underlying such
effects can include inflammatory responses, increased airway responsiveness or altered
responses to infectious agents. Toxicological studies have provided evidence that supports
plausible biological pathways for respiratory effects of fine particles; little evidence is yet
available on coarse fraction particles.
New epidemiological reanalyses or extensions of earlier prospective cohort studies of long-
term ambient PM exposure also show substantial evidence for increased lung cancer risk being
associated with such PM exposures, especially exposure to fine PM or specific fine particles
subcomponents (e.g., sulfates) and/or associated precursors (e.g., SO2). Toxicological evidence
9-78
-------
of mutagenicity or genotoxicity in ambient or combustion-related particles supports the
plausibility of a relationship between fine particles and lung cancer mortality.
PM-related health effects in infants and children are emerging as an area of more concern
than in the 1996 PM AQCD; and ultimately, such health effects could have very substantial
implications for life expectancy calculations. However, only very limited evidence currently
exists about potential ambient PM relationships with some of the more serious pertinent health
endpoints (low birth weight, preterm birth, neonatal and infant mortality, emergency hospital
admissions, and mortality in older children). Most studies have used PM10 or other measures of
thoracic particles; little evidence is available regarding PM2 5 or PM10_2 5. Also, 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.
Taken together, new evidence from mechanistic studies suggesting plausible biological
response pathways, and the extensive body of epidemiology evidence on associations between
short- and long-term exposures to ambient thoracic particles (typically indexed by PM10) and a
range of health effects, supports the general conclusion that ambient thoracic particles, acting
alone and/or in combination with gaseous co-pollutants, are likely causally related to
cardiovascular and respiratory mortality and morbidity. A growing body of evidence both from
epidemiologic and toxicologic studies also supports the general conclusion that PM2 5 (or one or
more PM2 5 components), acting alone and/or in combination with gaseous co-pollutants, are
likely causally related to cardiovascular and respiratory mortality and morbidity. The strength of
the evidence varies across such endpoints, with relatively stronger evidence of associations with
cardiovascular than respiratory endpoints, potentially due to reduced statistical power where
respiratory outcomes are seen less frequently than cardiovascular outcomes. In addition,
mortality associations with long-term exposures to PM2 5, in conjunction with evidence of
associations with short-term exposures, provide strong evidence in support of a casual inference.
A much more limited body of evidence is suggestive of associations between short-term
(but not long-term) exposures to ambient coarse-fraction thoracic particles (generally indexed
by PM10_2 5) and various mortality and morbidity effects observed at times in some locations.
9-79
-------
This suggests that PM10_25, or some constituent component(s) of PM10_25, may contribute under
some circumstances to increased human health risks. The strength of the evidence varies across
endpoints, with somewhat stronger evidence for coarse-fraction particle associations with
morbidity (especially respiratory) endpoints than for mortality. Reasons for differences among
findings on coarse-particle health effects reported for different cities are still poorly understood,
and much remains to be learned about the contribution of different sources or thoracic coarse
particle components to different health outcomes. Reduced precision for PM10_2 5 effect estimates
may be heavily influenced by the increased error in PM10_2 5 measurements obtained by
subtraction, and exposure error related to greater spatial variability and reduced penetration
indoors, as compared with PM2 5.
There is also important new information highlighting potentially crucial roles that particle-
bound water plays in serving as a carrier or vector by which other toxic agents (e.g., SO2,
peroxides, aldehydes) can be accumulated within inhalable PM and delivered in enhanced
quantities into the deep lung. Water-soluble gases, which would be removed by deposition to
wet surfaces in the upper respiratory system during inhalation, could dissolve in particle-bound
water and be carried with the particles into the deep lung. Of much concern, particle-bound
water appears to be a means by which dissolved hydrogen peroxide and other short-lived
reactive oxygen species can be carried into lower respiratory tract regions and contribute to the
induction of inflammatory responses. Also, certain other toxic species (e.g., nitric oxide [NO],
nitrogen dioxide [NO2], benzene, PAHs, nitro-PAHs, a variety of allergens) may be absorbed
onto solid particles and carried into the lungs. Thus, ambient particles may play important roles
not only in inducing direct health impacts of their constituent components but also in facilitating
delivery of toxic gaseous pollutants or bioagents into the lung and may, thereby, serve as
important mediators of health effects caused by the overall air pollutant mix.
The increased availability of certain bioaerosol materials (e.g., allergen-laded pollen
fragments) in small (0.1 to 0.4 micrometers) fine particle sizes that deposit in TB and A regions
of the lung (where they can exacerbate asthma effects) is also now recognized. Such bioagents
have been found attached to nonbiologic particles of anthropogenic origin, as well as to natural
biologic particles, which may serve to concentrate the biologic agents and enhance their delivery
9-80
-------
into TB and A regions of the lung and to exacerbate consequent inflammatory and allergic
asthmatic responses.
9.2.4 Potentially Susceptible and Vulnerable Subpopulations
The term susceptibility generally encompasses innate or acquired factors that make
individuals more likely to experience effects with exposure to pollutants. Genetic or
developmental factors can lead to innate susceptibility, while acquired susceptibility may result
from age, from disease, or personal risk factors such as smoking, diet, or exercise; personal risk
factors such as smoking, diet, or exercise habits are also associated with the development of
heart and lung diseases. In addition, new attention has been paid to the concept of some
population groups having increased vulnerability to pollution-related effects due to factors
including socioeconomic status (e.g., reduced access to health care) or particularly elevated
exposure levels.
The 1996 PM AQCD included only a relatively limited discussion of susceptible
population groups potentially at increased risk for ambient PM effects, noting:
"There is considerable agreement among different studies that the elderly are
particularly susceptible to effects from both short-term and long-term exposures to
PM, especially if they have underlying respiratory or cardiac disease. . . Children,
especially those with respiratory diseases, may also be susceptible to pulmonary
function decrements associated with exposure to PM or acid aerosols." (U.S.
Environmental Protection Agency, 1996, p. 13-92)
New studies appearing since the 1996 PM AQCD provide additional evidence that
substantiates the above-named groups as likely being at increased risk for ambient PM-related
morbidity or mortality effects; and the evidence related to preexisting disease, age groups, and
genetic susceptibility is summarized below. In addition, recent studies have explored potential
new risk factors related to potential increased vulnerability for certain population groups, and
evidence regarding factors such as socioeconomic status or exposure status are also discussed
below.
9-81
-------
9.2.4.1 Preexisting Disease as a Risk Factor
A number of time-series epidemiologic studies have reported increased risk in study
subsets of individuals with preexisting heart or lung diseases. Several studies have suggested
that people with diabetes are also susceptible to PM effects, possibly due to cardiovascular
complications associated with diabetes. One study reported large relative risk estimates for total
mortality in people with preexisting COPD living in Barcelona. Another study in Montreal
showed larger effect sizes for total mortality in persons with cancer, diabetes, lower respiratory
disease, cardiovascular disease, coronary artery disease, and congestive heart failure.
In addition, an European study found significant effects in the subset of adults who had
bronchial hyperreactivity or increased peak flow variability; and a Canadian study reported
greater effects in a subset of children who had asthma.
Toxicologists have used several animal models of cardiopulmonary disease to evaluate PM
susceptibility aspects. Such animal models include rats with monocrotaline-induced pulmonary
vasculitis/hypertension, experimental coronary artery dockage, apo-E atherosclerosis-prone
mice, SO2-induced chronic bronchitis, spontaneously hypertensive rats, and animals infected
with various viral or bacterial agents. As summarized in Section 7.5.1 of Chapter 7, increased
magnitude or frequency of effects have been reported with PM exposure for these groups of
animals relative to healthy animals. In addition, toxicologists have also studied effects of
particles, including diesel exhaust particles, in animals with heightened allergic sensitivity and
via in vitro studies (summarized in Section 7.5.2). Overall, the results from newly available
toxicological studies provide evidence suggestive of enhanced susceptibility to inhaled PM in
"compromised" hosts.
The underlying biology of lung diseases might also lead to heightened sensitivity to PM,
but this attribute of disease remains hypothetical in the context of PM (see Section 7.4.9 of
Chapter 7). The functional linkages with the cardiac system for maintenance of adequate gas
exchange and fluid balance notwithstanding, the role of inflammation in the diseased respiratory
tract (airways and alveoli) could be an important factor. There is sufficient basic biological data
to hypothesize that the exudated fluids in the airspaces may either interact differently with
deposited PM (e.g., to generate oxidants), to augment injury, or to predispose the lung (e.g.,
sensitize receptors) so as to enhance the response to a stereotypic PM stimulus through otherwise
9-82
-------
normal pathways. Less appreciated is the loss of reserve (functional or biochemical), wherein
the susceptible individual may be incapable of sufficient compensation (e.g., antioxidant
responses). Any of these or related mechanisms may contribute to increased "susceptibility" and
may indeed be a common factor possibly contributing to increased risk for various susceptible
groups.
Studies with humans that might reveal more specific data have been limited both ethically,
as well as by the absence of or limitations associated with biomarkers of response (such as
interpretation of ECG indicators of cardiac function and disease). Measures of blood-gas
saturation and lung function appear not to be sufficiently revealing or sensitive to mild
physiologic changes in those with moderate disease conditions who might be amenable to
participation in laboratory studies. In the field, assessing the degree of underlying disease and
how that relates to responsiveness of these biomarkers is unclear. However, subjects with COPD
and asthma have been studied under controlled conditions with inert aerosols for the purpose of
assessing distribution of PM within the lung, and it is now quite clear that airways disease leads
to very heterogeneous distribution of PM deposited within the lung. Studies have shown up to
10-fold higher than normal deposition at airway bifurcations, thus creating "hot-spots" that may
well have biologic implications, especially if the individual already has diminished function or
other debilitations due to the underlying disease, even cardiovascular disease (CVD).
9.2.4.2 Age-Related At-Risk Population Groups: the Elderly, Infants, and Children
The very young and the very old apparently constitute two other groups thought to be
especially at risk for ambient PM air pollution health effects. Numerous epidemiological studies
have reported health responses to PM and other pollutants for one or another specific age group.
These studies, as summarized in Section 8.4.9 of Chapter 8, tend to support previous
findings that, depending on the effect under study, older adults and children may be more
susceptible to certain PM-related effects. More specifically, older adults (aged 65+ years)
appear to be most clearly at somewhat higher risk for PM exacerbation of cardiovascular-related
disease effects and, perhaps, tend to experience higher PM-related total (nonaccidental) mortality
risk, as well. On the other hand, more limited evidence points to children possibly being at
somewhat higher risk for respiratory-related (especially asthma) PM effects than adults. Some
9-83
-------
newly emerging studies provide suggestive evidence for increased neonatal mortality and
adverse birth outcomes being associated with ambient PM exposures, but other new studies have
found contradicting results.
A major factor in increased susceptibility to air pollution is the presence of a preexisting
illness and susceptibility related to age group may well be closely linked with the potential for
preexisting cardiopulmonary diseases. Cardiopulmonary diseases more common to the elderly
contribute to the potentially higher risk within older age groups. Also, some recent studies have
reported evidence suggestive of associations between ambient PM exposure and effects on total
development or neonatal mortality (see Section 8.3.4 of Chapter 8). Although infection as a risk
factor for PM has already been noted, it is important to emphasize that there are clear age
differences in both the incidence and type of infections across age groups. Young children have
the highest rates of respiratory illnesses related to infection (notably respiratory syncytial virus),
while adults are affected by other infectious agents such as influenza that may also lend
increased susceptibility to PM effects. Data to address fully the importance of these differences
is incomplete, but some of the newly available toxicological studies provide evidence for
ambient PM exposures affecting lung defense mechanisms so as to exacerbate preexisting
respiratory infections.
In addition to their higher incidences of preexisting respiratory conditions, several other
factors may render children and infants more susceptible or vulnerable to PM exposures,
including more time spent outdoors, greater activity levels and ventilation, higher doses per body
weight and lung surface area, and the potential for irreversible effects on the developing lung.
The amount of air inhaled per kilogram body weight decreases dramatically with increasing age,
due in part to ventilation differences (in cubic meters per kilogram a day) of a 10-year-old being
roughly twice that of a 30-year-old person, even without the consideration of activity level.
Child-adult dosage disparities are even greater when viewed on a per lung surface-area basis.
As to potential lung developmental impacts of PM, there exist both experimental and
epidemiologic data, which although limited, suggest that the early post-neonatal period of lung
development is a time of high susceptibility for lung damage by environmental toxicants.
In experimental animals, for example, elevated neonatal susceptibility to lung-targeted toxicants
has been reported at doses "well below the no-effects level for adults" (Plopper and Fanucchi,
9-84
-------
2000); and acute injury to the lung during early postnatal development may impair normal repair
processes, such as down-regulation of cellular proliferation (Smiley-Jewel et al., 2000, Fanucchi
et al., 2000).
9.2.4.3 Genetic Susceptibility
A key issue in understanding adverse health effects of inhaled ambient PM is the
identification of which classes of individuals are susceptible to PM. Although factors such as
age and health status have been studied in both epidemiology and toxicology studies, some
investigators have begun to examine the importance of genetic susceptibility in the response to
inhaled particles because of evidence that genetic factors play a role in the response to inhaled
pollutant gases. To accomplish this goal, toxicologists typically have sought to detect interstrain
differences in responses to particles in rodents; little evidence is available from epidemiological
studies at this time. The small group of newly-available toxicological studies have begun to
demonstrate that genetic susceptibility can play a role in the response to inhaled particles
(Section 7.5.2); for example, one research group has found a genetically-based difference in
susceptibility to lung injury induced by instilled ROFA, using several strains of rats with varying
genetic characteristics.
9.2.4.4 Gender
There are appear to be some gender differences in the regional efficiency of deposition as
well as the deposition rate of particles. These differences derive from differences between males
and females in body size, conductive airway size, and ventilatory parameters. Females have a
somewhat greater deposition of coarse mode particles in the ET and TB regions, but lower
deposition in the A region. This gender effect appears to be particle-size dependent, showing a
greater fractional deposition in females for very small ultrafme and large coarse thoracic
particles. Total fractional lung deposition for 0.04 and 0.06 jam particles also appears to be
somewhat greater in females than males but only negligibly so for particles in the size range
0.8 to 1.0 |im. As the particle size increases (3 to 5 jim), total fractional deposition increases in
females. While deposition appears to be more localized in females than males, deposition rate
appears to be greater in males.
9-85
-------
Little evidence is available from toxicology studies regarding gender differences in
susceptibility to pollution effects. In the epidemiology studies that have included stratified
analyses based on gender, there is no clear pattern of increased vulnerability for either males or
females. Results from two of the prospective cohort studies evaluating long-term PM exposure
effects have reported greater mortality (and cancer) risks in males than in females (e.g., Abbey
et al., 1999; Pope et al., 2002), but a number of studies using long-term and short-term PM
exposures report no clear pattern of differences in effects across genders (e.g., Linn et al., 2000;
Ostro et al., 2001; Dockery et al., 1996; Raizenne et al., 1996; Krewski et al., 2000). Where
differences in effects between males and females were reported in the time-series studies, they
were generally not significantly different and the findings were not consistent. For example,
from PM10-mortality studies conducted in Chicago, Styer et al. (1995) report larger effect
estimates for men, but Ito and Thurston (1996) report larger effect estimates for women. Thus,
insufficient evidence exists overall to allow for any clear conclusions to be drawn as to potential
gender differences with regard to PM health effects.
9.2.4.5 Factors Related to Enhanced Vulnerability
Epidemiological studies of long-term PM exposures have suggested that there is effect
modification of PM-mortality associations due to socioeconomic factors. In the ACS and Six
Cities cohort analyses on mortality risk with long-term exposure to PM2 5, there was clear
evidence of effect modification (though not confounding) by education level, with greater effects
being reported in the cohort subgroups with lower education levels (Krewski et al., 2000; Pope
et al., 2002).
Among the studies of short-term PM exposure (Chapter 8, Appendices 8A and 8B), the
evidence is more mixed regarding potential influence of socioeconomic status on PM-related
health risks. No evidence of effect modification for PM10-mortality associations in 10 U.S. cities
was found using four measures of social or economic status: greater percent of population living
in poverty status; higher unemployment rate; greater percent of population with college degrees;
or greater percent of the population being nonwhite. Similarly, in a study of hospital admissions
in 10 U.S. cities, none of the four measures of social or economic status mentioned above
significantly modified the relationship between PM10 and hospitalization for COPD or
9-86
-------
pneumonia. However, for CVD admissions, PM10 effect estimates were greater in communities
with greater percentages of the population being unemployed, nonwhite, or living in poverty.
This may be a result of increased exposure, increased prevalence of predisposing diseases, or
other factors. Also, one study in Atlanta found race (black versus white) and insurance
(Medicaid versus non-Medicaid) to be effect modifiers for emergency department admissions for
asthma in children, but no associations with interaction terms for these factors and PM10 or
ozone. Another study analyzed associations between hospitalization for asthma with PM10 and
ozone in Los Angeles for subsets of patients who were uninsured, insured by MediCal, or had
other insurance. Significant associations with PM10 were reported only for the subset of patients
using MediCal, not for the privately insured or uninsured; the authors speculate that the small
sample size for uninsured patients may have precluded detection of an effect. However, a
Seattle study reported no effect estimate differences for asthma hospitalization in children (< 18
years) when comparing the inner city area with the rest of Seattle.
Vulnerability to PM-related effects may also be increased in populations experiencing
enhanced exposure to ambient aerosols in comparison to other groups. In some cases, e.g.,
proximity to roadways or other PM sources, there may be overlap with other factors (e.g.,
socioeconomic statuses).
As summarized in Chapter 8, in several reports from the Southern California children's
study, larger effect estimates for reduced lung function or increased respiratory illness with
long-term exposure to PM and other pollutants were reported for the subset of children spending
a larger amount of time outdoors. Also, using data from 14 U.S. cities, other investigators found
that effect estimates between PM10 and hospitalization for CVD and COPD increased with less
air conditioning use in homes (such use being an indicator of decreased exposure due to less
penetration of particles into the home). Increased vulnerability to the effects of pollution may
come from living near a source of PM and other pollutants, such as a major roadway. Numerous
recent studies have linked adverse health effects with indicators of traffic-related pollution and
with residences near a major road.
In addition to the above factors contributing to increased vulnerability, exercise may also
increase the potential health risks of inhaled particles, because exercise increases the rate of
oxygen consumption and changes ventilatory parameters affecting airflow rate and breathing
9-87
-------
patterns. The switch from nose breathing to mouth breathing, which occurs as exercise intensity
increases, leads to an increase in fractional deposition of ultrafme and coarse thoracic particles in
the tracheobronchial and alveolar regions. The higher breathing rate and larger tidal volume lead
to a greater amount of deposition. Total lung deposition rate may be 3 to 4 times greater during
exercise. The more rapid breathing of children also leads to a greater amount of deposition.
9.2.4.6 Summary and Conclusions
The existence of heart and lung disease is clearly linked with increased susceptibility to
effects from PM exposure, based on epidemiological and toxicological studies and dosimetric
evidence. The epidemiological evidence of susceptibility is primarily from studies of short-term
exposure. Long-term exposure studies have suggested that PM exposure may result in chronic
respiratory disease or decreased lung function growth, thus there is the potential that chronic PM
exposure can also increase susceptibility to acute changes in PM. More recent studies also
support considering older adults and children, including possibly infants, as susceptible groups,
recognizing that there is likely overlap between age categories and preexistence of
cardiopulmonary diseases. Some new evidence from toxicologic studies indicates that there may
be populations who are genetically predisposed to PM-related effects. In addition, beyond
consideration of innate or acquired factors related to susceptibility, some population groups can
be considered to be more vulnerable to PM-related effects due to factors such as socioeconomic
status or residing near roadways or other sources.
9.2.5 Potential Public Health Impacts in the United States
The 1996 PM AQCD highlighted the then considerable uncertainty related to estimating
public health impact of ambient PM exposure, stating:
"Efforts to quantify the number of deaths attributable to, and the years of
life lost to, ambient PM exposure are currently subject to much uncertainty."
(U.S. Environmental Protection Agency, 1996, p. 13-87). Nonetheless,
while "PM-related increases in individual health risks are small," they are
"likely significant from an overall public health perspective because of the
large numbers of individuals in susceptible risk groups that are exposed to
ambient PM." (U.S. Environmental Protection Agency, 1996, p. 1-21)
-------
9.2.5.1 Magnitude of Susceptible Groups
As summarized in Section 9.2.4, numerous U.S. population groups may be identified as
having increased susceptibility or vulnerability to adverse health effects from PM. Considering
together the subpopulations of persons with preexisting cardiopulmonary disease, older adults,
children, people of lower socioeconomic status and those with higher potential exposure levels
as potentially susceptible or vulnerable, it is clear that the impact of PM on public health could
be very extensive.
One consideration in the assessment of potential public health impacts is the size of various
population groups that may be at increased risk for health effects associated with PM-related air
pollution exposure. Table 9-4 summarizes information on the prevalence of chronic respiratory
and circulatory conditions and diabetes in the U.S. population in 2000. It can be seen that people
with preexisting cardiopulmonary disease constitute a fairly large proportion of the population,
with tens of millions of people included in each disease category. For circulatory conditions,
approximately 22 million people, or 11% of the U.S. adult population, have received a diagnosis
of heart disease. Approximately 20% of the U.S. adult population has hypertension, with 6%
reporting diagnoses of coronary heart disease. For respiratory conditions, approximately 9% of
U.S. adults (and 11% of children) have been diagnosed with asthma, and 6% of adults diagnosed
with conditions included in COPD. Table 9-5 provides further information on the number of
various specific respiratory conditions per 100 persons by age among the U.S. population during
the mid-1990s. In addition, approximately 6% of the U.S. adult population has diabetes. Both
cardiovascular conditions and diabetes are more common among older age groups, while asthma
prevalence is higher in children.
In addition, as discussed previously, subpopulations based on age group or socioeconomic
status would also comprise substantial segments of the population that may be potentially more
vulnerable to PM-related health impacts. Based on U.S. census data from 2000, about 26% of
people in the U.S. are under 18 years of age, and 12% are 65 years of age or older. From among
commonly-used indicators of socioeconomic status, about 12% of individuals and 9% of families
are below the poverty level, and 20% of the U.S. population does not have a high school or
higher level of education. Hence, large proportions of the U.S. population are included in groups
that are thought likely to be at increased risk for ambient PM-related health effects.
9-89
-------
TABLE 9-4. PREVALENCE OF SELECTED CARDIORESPIRATORY DISORDERS BY AGE GROUP AND BY
GEOGRAPHIC REGION, 2000 (reported as percent or numbers of cases in millions)
Chronic Condition/Disease
Respiratory conditions
Asthma
Asthma (< 18 years)*
COPD.
Chronic bronchitis
Emphysema
Circulatory conditions
All heart disease
Coronary heart disease
Hypertension
Stroke
Diabetes
Age
Adults (18+)* 18-44 45-64 65-74 75+
Number
(x 10 ) % % % % %
18.7 9.3 9.8 8.7 8.7 8.1
8.92* 12.4*
9.36 4.6 3.6 5.5 6.4 6.6
3.13 1.6 0.2 1.9 4.7 5.9
21.99 10.9 4.2 12.5 26.4 35
11.23 5.6 0.7 6.6 17.3 22.7
39.21 19.5 6.4 27.3 46.3 51.5
4.36 2.2 0.3 2.1 6.5 10.5
11.86 5.9 1.9 8.4 15.9 13.4
Regional
NE MW S W
% % % %
8.9 9.3 9 10.3
3.9 4.6 5.4 4.1
1 1.7 2 1.2
10.4 11.5 11.5 9.5
5.1 5.3 6.3 5
17.9 18.8 21.6 18.1
1.6 2.1 2.6 2.1
5.5 5.6 6.4 5.9
Source: Pleis et al. (2003).
*A11 data are for adults except asthma prevalence data for children under 18 years of age, responding to "ever told had asthma"; source for data on children is
Blackwell et al. (2003).
-------
TABLE 9-5. 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
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
44.3
4.3
*1.7
*2.4
18-24
Years
86
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
23.3
3.8
*2.0
*1.1
Years and
45-64
Years
55.9
16.4
7.5
26.1
3.5
*0.9
*1.5
Over
65 Years
and Over
49
15.7
6.1
18.6
*4.4
*3.8
*0.5
Source: Adams et al. (1999).
The health statistics data also illustrate what is known as the "pyramid" of effects. At the
top of the pyramid, there are approximately 2.5 millions deaths from all causes per year in the
U.S. population, with about 900,000 deaths due to cardiovascular diseases and 100,000 from
chronic lower respiratory diseases (Arias et al., 2003). For measures of cardiovascular disease
morbidity, there are approximately 6 million hospital discharges per year (Hall and DeFrances,
2003), nearly 5 million emergency department visits (McCaig et al., 2004), to over 70 million
ambulatory care visits for circulatory system disorders (Cherry et al., 2003). For chronic
respiratory health diseases, there are over 3 million hospital discharges for respiratory diseases
(Hall and DeFrances, 2003), nearly 13 million emergency department visits (McCaig et al.,
2004), over 200 million ambulatory care visits per year for respiratory conditions (Cherry et al.,
2003) and an estimated 700 million restricted activity days per year due to respiratory conditions
(Adams et al., 1999). Combining small risk estimates with relatively large baseline estimates of
health outcomes can result in quite large public health impacts. Thus, even a small percentage
reduction in PM health impacts on cardiorespiratory-related diseases would reflect a large
number of avoided cases.
9-91
-------
Another key input for public health impact assessment is the range of concentration-
response functions for various health outcomes. As described in Chapter 8, epidemiological
studies have reported associations between short-term exposure to PM, especially PM10
and PM2 5, with: mortality, hospitalization and medical visits for cardiovascular and respiratory
diseases; changes in heartbeat rhythm or electrocardiographic markers of cardiac function;
incidence of myocardial infarction; changes in blood characteristics (e.g., C-reactive protein,
fibrinogen levels); incidence of respiratory symptoms; and reduced lung function. As discussed
previously, the fewer studies using PM10_2 5, measurements have reported evidence for
associations with hospitalization for cardiovascular and respiratory causes, and increased
respiratory symptoms, and suggestions of associations with cardiopulmonary mortality.
Associations with long-term exposure to fine particles have been reported for cardiovascular and
lung cancer mortality, increased incidence of respiratory disease and decreased lung function and
lung function growth. The magnitude of the concentration-response function, measured or
anticipated change in air concentration, and size of population group are three major components
of a public health impact assessment.
Of concentration-response functions for PM-related effects, it can generally be said that the
effect estimates are small in magnitude. In historical episodes with very high air pollution
levels, risks on the order of 4-fold (400%) increases in mortality were estimated, but much
smaller risk estimates have been reported from recent studies at current pollution levels.
Risk estimates from long-term exposure studies are often larger in magnitude than those
for the same health outcome associated with short-term PM exposure. These estimates can
reflect different responses — long-term exposure perhaps being linked with development of
disease and short-term exposure with acute exacerbation of existing conditions — but there may
also be some overlap in the effect estimates. Relative risk estimates for total mortality from the
prospective cohort studies fall in the range of 7 to 13% increase per 10 |ig/m3 increase in PM25;
there are no significant associations with long-term exposure to PM10_2 5. Risk estimates from the
short-term exposure studies are considerably smaller in magnitude, on the order of 2 to 6%
increase in mortality per 25 |ig/m3 increase in PM2 5 and PM10_2 5. Time-series studies using
9-92
-------
distributed lag periods over more extended time periods (e.g., 40-60 days) partially bridge
these results.
Effect estimates for morbidity responses to short-term changes in PM tend to be larger in
magnitude than those for mortality; those for hospitalization generally range from 4-10%
increases for cardiovascular diseases and 5-15% increases for respiratory diseases per 25 |ig/m3
increase in PM2 5 and PM10_2 5. From the more recent studies on visits to the emergency
department or physicians' offices for respiratory conditions, effect estimate sizes have been
somewhat larger, ranging up to about 35% per 25|ig/m3 increase in PM2 5.
Other important considerations for public health impact assessment that have been
discussed previously include questions about the concentration-response function form and
potential identification of threshold levels, and attribution of risks for the varying health
outcomes to PM, sources or components of particles, and co-pollutants. Taken together,
however, it can be concluded that small incremental risks for large groups of the U.S. population
would result in large public health impact estimates.
9.2.5.2 Impact on Life Expectancy
Conceptually, ambient PM exposures may be associated with both the long-term
development of underlying health problems ("frailty") and with the short-term variations in
timing of mortality among a susceptible population with some underlying health condition
(Kiinzli et al., 2001). New evidence from toxicological studies have provided insights into
potential mechanisms for PM-related health effects, but this evidence is not sufficient to allow
direct conclusions to be drawn regarding specific effects linked with short-term or long-term PM
exposures. Epidemiologic studies of the mortality effects of short-term exposure to PM using
time-series studies can only capture PM's association with short-term variations in mortality
and, therefore, must systematically underestimate the proportion of total mortality attributable
toPM.
Finally, as discussed in Section 8.4.10 of Chapter 8, there appears to be no strong evidence
to suggest that PM exposures are shortening life by only a few days (i.e., for "harvesting").
To the contrary, the 1996 PM AQCD noted that results from the Harvard Six City Study
9-93
-------
suggested that long-term exposure to PM25 was associated with ~2 year loss of life expectancy.
More recent investigations of the public health implications of effect estimates for long-term PM
exposures were also reviewed in Chapter 8. Using results from prospective cohort studies of
mortality in adults, it has been estimated that loss of population life expectancy may be
substantial (on the order of a year or so) with long-term exposure to PM; however, further
research is needed on this question. Further research is also needed to build upon currently only
very limited evidence about potential PM-related health endpoints in infants and children, which
may ultimately significantly increase estimates of the extent of life shortening due to PM-related
premature mortality.
It is also useful to highlight the newer results of the extension of the ACS study analyses
(that include more years of participant follow-up and address previous criticisms of the earlier
ACS analyses), which indicate that long-term ambient PM exposures are associated with
increased risk of lung cancer. That increased risk appears to be in about the same range as that
seen for a nonsmoker residing with a smoker, with any consequent life-shortening due to lung
cancer.
Lastly, new epidemiologic studies of a broader array of health endpoints indicate ambient
PM associations with increased nonhospital medical visits (physician visits) and asthma effects.
Such new findings suggest likely much larger health impacts and costs to society due to ambient
PM than just those indexed either by just hospital admissions/visits and/or mortality.
9.3 SYNTHESIS OF AVAILABLE INFORMATION ON PM-RELATED
WELFARE EFFECTS
The synthesis of available information on PM-related welfare effects presented in this
section focuses on four types of effects, i.e., PM-related effects on: visibility, vegetation and
ecosystems, climate change processes, and man-made materials. The resulting synthesis of
information and conclusions are intended to provide the scientific bases for options to be
considered by the EPA Administrator as to whether currently available scientific information
supports retention or revision of existing secondary PM NAAQS.
9-94
-------
9.3.1 Airborne Particle Effects on Visibility
The following discussion of the effects of airborne particles on visibility is drawn primarily
from information in Chapter 4 of this document, which itself is supplementary to several other
significant reviews of the science of visibility. These reviews include reports of the National
Acid Precipitation Assessment Program (1991, 1998), the National Research Council (1993)
report on Protecting Visibility in National Parks and Wilderness Areas, and U.S. EPA's Interim
Findings on the Status of Visibility Research (U.S. Environmental Protection Agency, 1995).
The focus here is on characterizing: (a) how ambient PM (in particular ambient fine PM)
impairs visibility, and (b) how the public perceives and values improvements in visibility,
especially in urban areas.
9.3.1.1 Relationships Between Ambient PM and Visibility
The role of ambient PM in impairing visibility has long been well understood, as was
recognized in the 1996 PM AQCD as follows:
"The relationships between air quality and visibility are well understood.
Ambient fine particles are the major cause of visibility impairment.
Significant scientific evidence exists showing that reducing fine particle
concentrations will improve visibility." (U.S. EPA, 1996, p. 1-18).
Airborne particles degrade visibility by scattering and absorbing light. These optical
properties can be well characterized in terms of a light extinction coefficient, which is the
fractional attenuation of light per unit distance. The extinction coefficient produced by a given
distribution of particle sizes and compositions is strictly proportional to the particle mass
concentration. The efficiency with which different particles attenuate light is a function of
particle size, with fine particles in the accumulation mode being much more important in causing
visibility impairment than coarse-mode particles. Thus, it is fine-particle mass concentrations
that tend to drive extinction coefficients in polluted air.
The spatial and temporal variability in the observed extinction coefficient per measured
mass of PM25 is mainly due to the effects of particle-bound water, which varies with relative
9-95
-------
humidity and is removed by drying when ambient PM2 5 mass concentrations are measured using
the Federal reference method. In arid regions such as the Southwest, where this effect is
minimized, observed ratios of extinction coefficients to PM25 mass concentrations are generally
low and exhibit little variability (approximately in the range of 2 to somewhat greater than
3 m2/g). In more humid areas such as the East, observed efficiencies are generally higher and
more variable (ranging from about 4 to 5 m2/g under moderate humidity conditions up to 7 m2/g
or more high under humidity conditions).
The overall effect of increasing humidity on light scattering by particles was quantified
nearly 20 years ago, but current research is greatly increasing the detailed understanding of the
response of aerosol particles to changing humidities and the relationship of this response to the
chemical composition of the particles. Humidity effects generally become important at relative
humidities between 60 and 70%, and increase particle-related light scattering by a factor of 2 at
-85% relative humidity. Light scattering by particles increases rapidly with relative humidity
when the humidity exceeds 90%.
As discussed in Chapter 4, a number of studies available since the last review have resulted
in refinements both (a) in the algorithms and related parameters used to calculate light extinction
based on particle properties and (b) in related measurement methods and monitoring
instrumentation. For example, a few studies have focused on better characterizing the
hygroscopic properties of particles, with a particular focus on organic compounds and mixtures
associated with different sources (e.g., Cocker et al., 2001; Chughtai et al., 1999; Hemming and
Seinfeld, 2001). More broadly, Malm (2000) used data from a special study at the Great Smoky
Mountain National Park to compare the performance of a number of models for calculating light
extinction and found that significant model improvement could be obtained by including the
degree of sulfate ammoniation in the model, so as to better estimate the ambient aerosol water
content. These studies have served primarily to reinforce and refine our understanding of how
airborne particles affect visibility.
Effects to address visibility impairment have historically been focused on rural areas,
particularly in national parks and wilderness areas (i.e., Federal Class I areas). Visibility in such
areas varies substantially between eastern and western sites in the U.S., with the haziest days in
9-96
-------
the rural West typically being roughly equivalent to the clearest days in the East. The largest
monitoring network that measures both visibility and aerosol conditions is the Interagency
Monitoring of Protected Visual Environments (IMPROVE) network, formed in 1987 as a
collaborative effort between Federal, regional, and state entities responsible for visibility
protection in such areas. This network has been used in visibility-related research, including the
advancement of visibility monitoring instrumentation and analysis techniques and source
attribution field studies. This network and related research have provided substantial support to
regulatory programs established to protect Federal Class I areas from local and regional sources
of visibility impairment.
More recent attention has been given to addressing visibility impairment in urban areas, as
well. Such efforts can now draw upon data available from the new national monitoring networks
designed to assess PM2 5 concentrations and composition in urban areas across the country that
have been deployed in conjunction with establishment in 1997 of the PM25 NAAQS.
In addition, higher resolution visibility data are now becoming available from the Automated
Surface Observing System (ASOS) monitoring network in operation at airports across the U.S.
These and other sources of visibility and ambient fine particle data provide important
information that helps to facilitate the characterization of relationships between ambient PM and
visibility especially in urban areas.
In addition to empirically derived relationships between ambient PM and visibility
measurements, photographic modeling techniques that have been refined in recent years are
useful in portraying changes in visibility specifically due to changes in ambient PM levels.
For example, the WinHaze system developed by Molenar et al. (1994) has been used to simulate
changes in visibility as a function of changes in air quality for both rural and urban areas. This
modeling system can produce a simulated photograph that accurately depicts a cloud-free scene
as it would appear to a human observer. Such photographic representations have facilitated the
evaluation of how the public perceives and values improvements in visibility in a number of
urban areas, as discussed below.
9-97
-------
9.3.1.2 Public Perception and Valuation of Visibility Improvements
The Clean Air Act (Section 169A) establishes a national visibility goal to "remedy existing
impairment" and prevent future impairment in national parks and wilderness areas across the
United States, and requires that long-term strategies be put in place to make reasonable progress
toward this national goal. The 1990 Amendments to the Act (Section 169B) place additional
emphasis on improving visibility, leading to EPA's promulgation of a Regional Haze Rule in
1999 that establishes specific goals for improving visibility in such areas. These national goals
and regulations provide clear evidence of the value society places on visibility improvements
that add to enjoyment of scenic vistas.
More specific information about how the public perceives and values improvements in
visibility in rural and urban areas comes from both economic studies and from local and/or state
initiatives in a number of areas to adopt local visibility goals and standards. As summarized in
Chapter 4, there is an extensive scientific literature on the theory and application of economic
valuation methods. Such studies have estimated the value of visibility improvements in the
range of billions of dollars annually, for example, in analyses of visibility improvements in
national parks in the Southwest (e.g., Chestnut and Rowe, 1990) and in an analysis of benefits to
residents in the eastern U.S. due to visibility improvements associated with the Federal acid rain
program (Chestnut and Dennis, 1997). Results vary across studies and uncertainties remain
about specific dollar values estimated. Local initiatives over the past few years, for example in
the Denver, CO and Phoenix, AZ areas, also provide important information about public
perceptions and attitudes about visibility impairments, including what is adverse, although
uncertainty would be involved in extending the public value judgments implied by these
examples to other areas.
More specifically, the initiative in Denver began with a series of visibility-related studies
in the 1970's through the 1980's, leading to the adoption of a visibility standard for the city of
Denver in 1990. This standard is based on a light extinction level of 0.076 km"1, averaged over
four daylight hours, reflecting the short-term nature of the perception of changes in visibility
conditions. This standard is equivalent to a visual range of approximately 50 km and reflects
citizen judgments about acceptable and unacceptable levels of visual air quality. In Phoenix,
9-98
-------
a study conducted between 1988 and 1990 led to establishment of a Blue Sky Index, which
focuses on days in which the visual range, averaged over six daylight hours, is 40 km or more.
This target is based on a method very similar to that used in Denver for obtaining citizen's
judgments as to acceptable levels of visual air quality. While in practice these standard target
values are exceeded many times per year in these areas, they reflect a reasonable degree of
consistency in the outcome of the approach used to characterize the value that citizens in these
two urban areas place on visual air quality. In addition, similar "acceptable" and "not
acceptable" threshold determinations, convergent on a minimal visual range of 40 to 60 km,
have also been identified in visibility standards in the Lake Tahoe area, the lower Fraser Valley
in British Columbia, CN, and the State of Vermont. In areas across the United States, visual
ranges of 40 to 60 km are approximately associated with PM25 concentrations ranging from
< 10 |ig/m3 up to about 20 |ig/m3.
9.3.1.3 Summary and Conclusions
Impairment of visibility in rural and urban areas is directly related to ambient
concentrations of fine particles, as modulated by particle composition, size, and hygroscopic
characteristics, and by relative humidity. Refinements in algorithms that relate these factors to
light extinction, and thus, to visual range, as well as the availability of much expanded databases
of PM2 5 concentrations and related compositional information and higher resolution visibility
data all contribute to the ability to develop improved characterizations of relationships between
ambient fine particle concentrations and visibility impairment.
Various local initiatives to address visibility impairment have demonstrated the usefulness
of approaches now being used to evaluate public perceptions and attitudes about visibility
impairment and public judgments as to the importance of standards to improve visibility relative
to current conditions. Various such initiatives, conducted in areas with notable scenic vistas
(e.g., Denver, CO, Phoenix, AZ, Lake Tahoe, CA, and State of Vermont), have resulted in local
standards that reflect what might be referred to as "adverse thresholds" associated with a
minimum visual range of approximately 40 to 60 km. These various local standards take into
account that visibility impairment is an instantaneous effect of ambient PM25 concentrations and
9-99
-------
that the public primarily values enhanced visibility during daylight hours. These considerations
are reflected in local standards that are based on sub-daily averaging times (e.g., 4 to 6 hours),
typically averaged across midday hours. This general convergence of visual range values and
averaging times that have been determined to be acceptable to the public in a number of such
locations suggests that these values and averaging times are relevant for consideration in
assessing the need for a national secondary standard to protect visibility in such areas.
9.3.2 Effects of Ambient PM on Vegetation and Ecosystems
9.3.2.1 Direct and Indirect Effects of Ambient PM
The direct and indirect effects of deposited ambient PM can span the full range, scale and
properties of biological organization listed under Biotic condition (Chapter 4) and can vary
widely depending on the (1) sensitivity of each ecosystem and/or its component biota (biotic
receptors) to a given concentration and chemical composition (acid/base, trace metal or
nutrients, e.g., nitrates or sulfates) of PM components; (2) the pre-existing buffering capacity of
the soils and/or waters (streams, rivers, ponds, and lakes, estuaries and ocean); (3) the magnitude
(ambient concentration and deposition velocity), mode, and meteorology of the deposition; and
(4) other site-specific features (e.g., terrain, hydrology, climate, land use, etc.). The ability of an
ecosystem to maintain integrity in the presence of the different chemical constituents in
deposited aerosols is a direct function of the sensitivity level of the ecosystem to the different
PM constituents and to the ability of the ecosystem components to ameliorate the effects that can
result. Changes in structural patterns and the functioning of ecological processes must be scaled
in both time and space and propagated to the more complex levels of community interaction to
produce observable ecosystem changes.
Direct effects result when PM is deposited onto sensitive receptors. Such effects can be
either chemical and/or physical; and they have been observed largely downwind of point
sources. These effects were the usually the result of dust from limestone quarries and cement
kilns or heavy metals from iron and lead smelting factories (Chapter 4). Because these effects
tend to be very limited in scope, they do not warrant the level of attention given the more
widespread indirect, ecosystem-level, effects discussed below.
9-100
-------
Indirect effects of major concern (such as nitrogen saturation, acidification, and
eutrophication) are mediated via the soil or aquatic environment and have the potential of
degrading ecosystem functioning by altering species diversity, structure, and sustainability of
ecosystems to the detriment of animals and plant life, so that ecosystems provide fewer benefits
and services for humans (Moomaw, 2002).
Ecosystem effects within the U.S. span the range from remote to urban. Most of the
ecosystem impacts of PM that have been reported occurred at nonurban sites and, as such,
nonurban ecosystems are the primary focus of the discussion that follows in subsequent
subsections. In briefly considering urban ecosystems here, it is recognized that despite a large
body of knowledge on concentrations and chemical reactions of air pollutants in cities, there has
been little work on the rates of atmospheric deposition to urban ecosystems. However, urban
ecosystems are likely to be subjected to large rates of deposition of anthropogenic pollutants
(Lovett et al., 2000). Decades of research on urban air quality indicate that cities are often
sources of nitrogen oxides, sulfur oxides, and dust, among many other pollutants. Some of these
air pollutants are major plant nutrients (e.g., nitrogen, sulfur, and phosphorus) and may be
affecting nutrient cycles in plant-dominated areas in and around cities. Though the effects of
urban PM, as such, appear not to have been sufficiently measured at this time, the deployment of
new PM2 5 speciated urban monitors and concern about urban visibility impairment could lead to
additional information relevant to assessing PM effects on urban ecosystems.
9.3.2.2 Major Ecosystem Stressors
In order for any specific chemical constituent of ambient PM to impact ecosystems, it must
first be removed from the atmosphere through deposition. Deposition can occur in three modes:
wet, dry, or occult. The factors that influence the magnitude and mode of particle deposition are
numerous and complex and depend in part on particle size, shape, chemistry, atmospheric
conditions (e.g., relative humidity, wind speed) and ecosystem surface features (e.g., elevation,
complexity of terrain, land over type, etc.). National deposition monitoring networks routinely
measure total wet or dry deposition of certain compounds. Data from these networks
demonstrate that nitrogen and sulfur compounds are being deposited onto soils and aquatic
9-101
-------
ecosystems in sufficient amounts to impact ecosystems at local, regional and national scales.
Though the ambient PM contribution to total wet or dry deposition has rarely been characterized
and the percentages of nitrogen and sulfur containing compounds in PM vary spatially and
temporally, nitrates and sulfates make up a substantial portion of the chemical composition of
PM. Therefore, the components of PM that are considered of greatest environmental
significance are nitrates, sulfates and the associated hydrogen (H+) ion (Chapter 4).
9.3.2.2.1 Nitrogen
Nitrogen is required by all organisms as it is a major constituent of the nucleic acids that
determine the genetic character of all living things and the enzyme proteins that drive the
metabolic machinery of every living cell (Galloway, 1998; Galloway and Cowling, 2002). It has
long been recognized as the nutrient most important for plant metabolism and, to a large extent it
governs the utilization of phosphorus, potassium, and other nutrients. Typically, the availability
of biologically active nitrogen controls net primary productivity and, possibly, the
decomposition rate of plant litter. Plants usually obtain nitrogen directly from the soil by
absorbing NH4 + or NO3 through their roots and by foliar absorption through the leaves, or it is
formed in their roots by symbiotic organisms (e.g., bacteria and free-living blue-green algae).
The wide-ranging pathways by which nitrogen cycles through various environmental reservoirs
are illustrated in Figure 9-6.
Nitrogen in nature can be divided into two groups: nonreactive (N2) and reactive (Nr).
Molecular nitrogen (N2), though the most abundant element in the Earth's atmosphere, is not
available to more than 99% of living organisms unless converted into reactive forms
(Galloway et al., 2003). Reactive Nr includes the inorganic reduced forms of nitrogen (e.g.,
ammonia [NH3] and ammonium [NH4+]), inorganic oxidized forms (e.g., nitrogen oxide [NOX],
nitric acid [HNO3], nitrous oxide [N2O], and nitrate [NO3"]) , and the organic compounds (e.g.,
urea, amine, proteins, and nucleic acids)]) (Galloway and Cowling, 2002).
Due mainly to three anthropogenically-driven activities, anthropogenic Nr creation now
exceeds the rate of natural terrestrial Nr creation and its conversion back to N2by denitrification
(Galloway and Cowling, 2002). Thus, increase in global Nr is the result of three main causes:
9-102
-------
Food
Production
People
(Food; Fiber)
Human Activities
"org
The Nigrogen
Cascade
Indicates denitrification potential
Agroecosystem Effects
Animal
^g^4
N2O
Figure 9-6. Illustration of the nitrogen cascade showing the movement of the human-
produced reactive nitrogen (Nr) as it cycles through the various environment
reservoirs in the atmosphere, terrestrial ecosystems, and aquatic ecosystems.
Source: Galloway et al. (2003).
(1) widespread cultivation of legumes, rice and other crops that promote conversion of N2 to
organic nitrogen through biological nitrogen fixation (BNF); (2) combustion of fossil fuels
which converts both atmospheric N2 and fossil N to reactive NOX; and (3) the Haber-Bosch
process, developed in 1913, which converts nonreactive N2 to reactive NH3 mainly for use as
fertilizers to sustain food production and some industrial activities (Galloway and Cowling,
2002; Galloway et al., 2003). As a result, Nr is now accumulating in the atmosphere and
terrestrial and aquatic ecosystems on all spatial scales - local, regional and global (Galloway and
Cowling, 2002; Galloway et al., 2003).
9-103
-------
Nitrogen oxides (a compound of Nr) is the only ambient air criteria pollutant that has not
decreased since the passage of the Clean Air Act. Despite decreases in emissions from fossil
fuel burning industries, emissions from automobiles have increased approximately 10% since
1970 due to greater total miles driven (Howarth et al., 2002). Nitrogen oxides emissions from
fuel burning increased exponentially from 1940 until the 1970s, leveled off after the passage in
of the Clean Air Act in 1970, and stabilized at approximately 7 Tg NOX /yr in the late 1990s.
Contemporary emissions of NOX in the U.S. from fossil fuel burning are nearly two-thirds the
rate of Nr releases from the use of inorganic fertilizers and comprise 30% of the global
emissions of NOX from fossil fuel combustion. Some NOX emissions are converted/transformed
into a portion of ambient air PM (particulate nitrate) and are deposited onto sensitive
ecosystems.
Environmental Effects ofNr
The term "nitrogen cascade" refers to the sequential transfers and transformations of Nr
molecules as they move from one environmental system or reservoir (atmosphere, biosphere,
hydrosphere) to another and the multiple linkages that develop among the different ecological
components. Because of these linkages, the addition of anthropogenic Nr alters a wide range of
biogeochemical processes and exchanges as it moves among the different environmental
reservoirs, with the consequences becoming magnified through time (Figure 4-15; Galloway and
Cowling, 2002; Galloway et al., 2003). These changes in the nitrogen cycle are contributing to
both beneficial and detrimental effects to the health and welfare of humans and ecosystems
(Rabalais, 2002; van Egmond et al., 2002; Galloway, 1998).
Some of the detrimental effects resulting from increased inputs of atmospheric Nr (e.g.,
particulate nitrates) include: (1) increases in productivity of Nr-limited forests and grasslands
followed by decreases wherever increase in atmospheric deposition of Nr significantly exceeds
critical thresholds; Nr additions have also been shown to decrease biodiversity in many natural
habitats (Aber et al., 1995); (2) formation of O3 and ozone-induced injury to crops, forests, and
natural ecosystems and the resulting predisposition to attack by pathogens and insects;
(3) nitrogen saturation of soils in forests and other natural ecosystems, leading to shifts in
community composition and leaching of Nr into streams, lakes and rivers; (4) eutrophication,
9-104
-------
hypoxia, loss of biodiversity, and habitat degradation in coastal ecosystems, now considered the
biggest pollution problem in coastal waters (Rabalais, 2002); (5) acidification and loss of
biodiversity in lakes and streams in many regions of the world when associated with sulfur
(Vitousek et al., 1997); and (6) alteration of ecosystem processes through changes in the
functioning and species composition of beneficial soil organisms (Galloway and Cowling 2002).
Indirect effects of Nr on societal values include: (1) increases in fine PM resulting in
regional hazes that decrease visibility at scenic rural and urban vistas and airports; (2) depletion
of stratospheric ozone by N2O emissions which can in turn affect ecosystems and human health;
(3) global climate change induced by emissions of N2O; and (4) formation of acidic deposition
when in association with sulfate (Galloway et al., 2003).
Large uncertainties, however, still exist concerning the rates of Nr accumulation in the
various environmental reservoirs which limits our ability to determine the temporal and spatial
distribution of environmental effects for a given input of Nr. These uncertainties are of great
significance because of the sequential nature of Nr effects on environmental processes. Reactive
nitrogen does not cascade at the same rate through all environmental systems. The only way to
eliminate Nr accumulation and stop the cascade is to convert Nr back to nonreactive N2
(Galloway et al., 2003).
Nitrogen Saturation and Ecosystem Response
A major environmental concern is nitrogen saturation of soils. Nitrogen saturation occurs
when chronic additions of Nr (including nitrate deposition from ambient PM) to soil (nitrogen
loading) exceeds the capacity of plants and soil microorganisms to utilize and retain nitrogen
(Aber et al., 1989, 1998; Garner 1994; U.S. Environmental Protection Agency, 1993). Nitrogen
saturation implies that some resource other than nitrogen is now limiting biotic functions. The
appearance of nitrate in soil solution (leaching) is an early symptom of excess Nr accumulation.
Nitrogen saturation does not occur at a specific point in time, but is a set of gradually
developing critical changes in ecosystem processes which represent the integrated response of a
system to increased nitrogen availability over time (Aber, 1992).
Not all vegetation or ecosystems react in the same manner to Nr deposition. Responses
vary depending on numerous factors, including soil composition and the length of time Nr
9-105
-------
deposition has been occurring. For example, ecosystems comprised of older, mature forests with
high stores of soil Nr and low carbon/nitrogen (C:N) ratios receiving high Nr deposition are
prone to Nr saturation (Fenn et al., 1998).
Variations in the response of different forest types ecosystems across the eastern and the
western United States to differing amounts of nitrate deposition illustrate this point (Chapter 4,
Table 4-14). Although soils of most North American forest ecosystems are nitrogen limited,
some exhibit severe symptoms of nitrogen saturation (See Figure 4-17; Chapter 4 (Aber et al.,
1989). In the east, these include the Great Smoky Mountains National Park (3.1 to 26.6 kg
N ha"1 yr) (Johnson and Lindberg, 1992); the Fernow Experimental Forest, WV (15 to 20 kg
N ha'1 yr) (Gilliam et al., 1996); Whitetop Mountain, VA (32 kg N ha1 yr); the Catskill
Mountains in southeastern NY (10.2 kg N ha"1 yr); and the Adirondack Mountains of
northeastern NY (9.3 kg N ha"1 yr) (see Table 4-14).
In the west, wildland ecosystems within the South Coast Air Basin of California receive
the highest Nr deposition in the United States (Fenn et al., 1998; 2003). The areas receiving the
greatest deposition are the south-facing slopes of the San Gabriel Mountains and the western and
southern edges of the San Bernardino Mountains where deposition ranges from 23.3 to 30 kg
N ha"1 per yr. Deposition in the low- and mid-elevation chaparral and mixed conifer forests
ranges from 20 to 45 kg Nr ha"1 per yr in the most exposed areas. However, when fog occurs in
late summer with unusually high NO3" and NH4+ concentrations, deposition values can be higher
than 90 kg Nr ha"1 yr (Fenn et al., 2003). The forests in the southwestern Sierra Nevada of
Central California receive 6-11 kg N ha"1 yr as throughfall (Fenn et al; 1998). Nr deposition
since the 1980s has resulted in saturation in the high-elevation Front Range in northern Colorado
where deposition values currently range from 8 to 10 kg Nr ha"1 yr (Bowman and Steltzer, 1998;
Bowman, 2000; Baron et al., 2000) (Chapter 4, Table 4-14.)
On the other hand, the Harvard Forest hardwood stand in Massachusetts has absorbed over
900 kg Nr ha"1 without significant nitrate leaching during a nitrogen amendment study of 8 years.
However, leaching losses were high in Harvard pine sites suggesting that deciduous forests may
have a greater capacity for Nr retention (Fenn et al., 1998). Magill et al. (2000) suggest that the
sharp contrasts that exist between hardwood and pine forests indicate that the mosaic of
9-106
-------
community types across the landscape must be considered when determining regional scale
response to Nr deposition.
Increases in soil Nr can also play a selective role in ecosystems, by affecting competition
among species that result in changes in biodiversity, i.e., community composition. In general,
plants adapted to living in an environment of low nitrogen availability will be replaced by
nitrophilic plants which are capable of using increased amounts of Nr, because they have a
competitive advantage when nitrogen becomes more readily available (Fenn et al., 1998).
Several long-term fertilization studies have observed these effects. For example, fertilization
and nitrogen gradient experiments at Mount Ascutney, VT suggest that nitrogen saturation may
lead to the slow-growing, slow nitrogen-cycling spruce-fir forest stands being replaced by fast-
growing deciduous forests that cycle nitrogen rapidly. Similarly, experimental studies of the
effects of Nr deposition over a 12-year period on Minnesota grasslands dominated by native
warm-season grasses observed the shift to low-diversity mixtures dominated by cool-season
grasses at all but the lowest rates of Nr addition (Wedin and Tilman, 1996). The shift to low-
diversity mixtures was associated with the decrease in biomass carbon to Nr (C:N) ratios,
increased Nr mineralization, increased soil nitrate, high nitrogen losses, and low carbon storage
(Wedin and Tilman, 1996).
The mutualistic relationship between plant roots, fungi, and microbes is critical for the
growth of the organisms involved. The rhizosphere, the soil that surrounds and is influenced by
plant roots is an important region of nutrient dynamics. Bacteria are essential components of the
nitrogen and sulfur cycles while fungi in association with plant roots form mycorrhizae that are
essential in the uptake of mineral nutrients. The action of bacteria make N, S, Ca, P, Mg, K
available for plant growth while mycorrhizae are of special importance in the uptake of N and P
(Section 4.3.3; Wall and Moore, 1999; Rovira and Davy, 1974). Changes in soil Nr influence
the mycorrhizal-plant relationship. Mycorrhizal fungal diversity is associated with
above-ground plant biodiversity, ecosystem variability, and productivity (Wall and Moore,
1999). During nitrogen saturation, soil microbial communities change from being fungal, and
dominated by mycorrhizae, to being dominated by bacteria. The decline in the coastal sage
scrub species can be directly linked to the decline of the arbuscular mycorrhizal community
(Edgerton-Warburton and Allen, 2000; Allen et al., 1998; Padgett et al., 1999).
9-107
-------
Nitrate Effects on Aquatic Habitats
Aquatic ecosystems (streams, rivers, lakes, estuaries or oceans) receive increased nitrogen
inputs either from direct atmospheric deposition (including nitrogen-containing particles),
surface runoff, or leaching from saturated soils into ground or surface waters. The primary
pathways of Nr loss from forest ecosystems are hydrological transport beyond the rooting zone
into groundwater or stream water, or surface flows of organic nitrogen as nitrate and Nr loss
associated with soil erosion (Fenn et al., 1998). Based on data from a number of hydrologic,
edaphic, and plant indicators, the mixed conifer forest and chaparral watershed with high smog
exposure in the Los Angeles Air Basin exhibited the highest stream water NO3" concentrations in
wilderness areas of North America (Bytnerowicz and Fenn, 1996; Fenn et al., 1998). High
nitrate concentrations have also been observed in streams draining watersheds in the Great
Smoky Mountains National Park in Tennessee and North Carolina (Fenn et al., 1998).
Estuaries are among the most intensely fertilized systems on Earth (Fenn et al., 1998).
They receive far greater nutrient inputs than other systems. For example, atmospheric Nr
deposition into soils in watershed areas feeding into estuarine sound complexes (e.g.,
Chesapeake Bay, the Pamlico Sound of North Carolina) contribute to excess Nr flows that also
include runoff from agricultural practices or other uses (e.g., fertilization of lawns or gardens).
Especially during and after heavy rainfall events such as hurricanes, massive influxes of nitrogen
into watersheds and sounds can lead to dramatic decreases of oxygen in water and increases in
algae blooms that can cause extensive fish kills and damage to commercial fish and sea food
harvesting (Paerl et al., 2001).
9.3.2.2.2 Acidification from PM Deposition
Acidic deposition is composed of ions, gases, and particles derived from the precursor
gaseous emissions of sulfur dioxide (SO2), nitrogen oxides (NOX), ammonia (NH3) and
particulate emissions of other acidifying compounds. It connects air pollution to diverse
terrestrial and aquatic ecosystems and alters the interactions of the (FT) and many elements (e.g.,
S, N, Ca, Mg, Al, and Hg) (Driscoll et al., 2001). Linked also to the nitrogen cascade (see
Figure 4-15), acid precipitation is a critical environmental stress that affects forest landscapes
and aquatic ecosystems in North America, Europe, and Asia (Driscoll et al., 2001).
9-108
-------
Acid deposition and acidification of soils can lead to high Al-to-nutrient ratios that limit
plant uptake of essential nutrients, such as Ca and Mg. Calcium is essential in the formation of
wood and the maintenance of the primary plant tissues necessary for tree growth (Shortle and
Smith, 1988), and tree species can be adversely affected if altered Ca/Al ratios impair Ca or Mg
uptake. A region-wide increase in Ca above expected levels followed by decreasing changes in
wood Ca suggests that Ca mobilization began possibly 30 to 40 years ago and has been followed
by decreased accumulation in wood, presumably associated with decreasing Ca availability in
soil (Chapter 4; Bondietti and McLaughlin, 1992).
9.3.2.3 Characterization of PM-Related Ecosystem Stressors
The critical loads concept has been used in Europe for estimating the amounts of pollutants
that sensitive ecosystems can absorb on a sustained basis without experiencing measurable
degradation (Lokke et al., 1996). The estimation of ecosystem critical loads requires an
understanding of how an ecosystem will respond to different loading rates in the long term and
can be of special value for ecosystems receiving chronic deposition of Nr and sulfur
independently and as acid deposition when in combination. Time scales must be considered
when selecting and evaluating ecosystems response(s) to changes in atmospheric deposition.
Indicators of ecosystems at risk of nitrogen saturation should include those that can be identified
when nitrogen availability exceeds biotic demand. The cardinal indicator of nitrogen saturation
in all ecosystem types is increased and prolonged NO3" loss below the main rooting zone in
stream water (Fenn and Poth, 1998). A paucity of baseline data makes it difficult to determine
the time scale for critical loading of most U.S. ecosystems because nitrogen deposition began so
many years ago. Though atmospheric sources of Nr, including ambient PM, are clearly
contributing to the overall excess nitrogen load/burden entering ecosystems annually,
insufficient data are available at this time to quantify the contribution of ambient PM to total
Nr or acidic deposition as its role varies both temporally and spatially along with a number of
other factors.
9-109
-------
9.3.2.4 Summary and Conclusions
A number of ecosystem-level conditions (e.g., nitrogen saturation, terrestrial and aquatic
acidification, coastal eutrophication) that can lead to negative impacts on human health and
welfare have been associated with chronic, long-term exposure of ecosystems to elevated inputs
of compounds containing Nr, sulfur and/or associated hydrogen ions. Some percentage of total
ecosystem inputs of these chemicals is contributed by deposition of atmospheric particles,
although the percentage greatly varies temporally and geographically and has not generally been
well quantified. Unfortunately, our ability to relate ambient concentrations of PM to ecosystem
response is hampered by a number of significant data gaps and uncertainties.
First, U.S. monitoring networks have only recently begun to measure speciated PM.
Historically, measurements were focused only on a particular size fraction such as PM10 and,
more recently, PM25. An exception to this is the IMPROVE network, which collects speciated
measurements. Additionally, except for the IMPROVE and some CASTNet sites, much of the
PM monitoring effort has focused on urban or near urban exposures, rather than on those in
sensitive ecosystems. Thus, the lack of a long-term, historic database of annual speciated PM
deposition rates precludes establishing relationships between PM deposition (exposure) and
ecosystem response at this time.
A second source of uncertainty lies in predicting deposition velocities based on ambient
concentrations of PM. There are a multitude of factors that influence the amounts of PM that get
deposited from the air onto sensitive receptors, including the mode of deposition, e.g., wet, dry,
and occult (cloud and fog deposition), windspeed, surface roughness/stickiness, elevation,
particle characteristics (e.g., size, shape, chemical composition, etc.) relative humidity, etc.
Therefore, modeled deposition rates, used in the absence of monitored data, can be highly
uncertain.
Third, each ecosystem has developed within a context framed by the topography,
underlying bedrock, soils, climate, meteorology, hydrologic regime, natural and land use history,
species associations that co-occur at that location (i.e., soil organisms, plants, etc.), and
success!onal stage, making it unique from all others. Because of this variety, and insufficient
baseline data on each of these features for most ecosystems, it is currently impossible to
extrapolate with much confidence any effect from one ecosystem to another, or to predict an
9-110
-------
appropriate "critical load." Thus, for example, a given PM deposition rate or load of nitrates in
one ecosystem may produce entirely different responses than the same deposition rate at another
location.
Finally, related in part to the complexity and unique set of characteristics belonging to each
ecosystem as discussed above, there remain large uncertainties associated with the length of
residence time of Nr in a particular ecosystem component or reservoir, and thus, its impact on
the ecosystem as it moves through the various levels of the N cascade. As additional PM
speciated air quality and deposition monitoring data become available, there is much room for
fruitful research into the areas of uncertainty identified above.
9.3.3 Relationships Between Atmospheric PM and Climate Change Processes
With regard to the role of ambient PM in affecting climate change-related processes, the
1996 PMAQCD stated:
"Particles [primarily fine particles] suspended in the atmosphere affect the earth's
energy budget and thus exert an impact on climate: (a) directly by increasing
the reflection of solar radiation by cloud-free portions of the atmosphere, and
(b) indirectly by affecting cloud microphysical properties in ways that increase
the brightness and stability of clouds." Since aerosol lifetimes are much shorter
than the time required for global mixing, "aerosol radiative effects are most likely
to exert their influence on a regional rather than on a global basis." (U.S.
Environmental Protection Agency, 1996, p. 1-19, 1-21)
The same physical processes (i.e., light scattering and absorption) responsible for visibility
degradation are also responsible for airborne particle effects on transmission of solar visible and
ultraviolet radiation. Scattering of solar radiation back to space and absorption of solar radiation
determine the effects of an aerosol layer on solar radiation. Atmospheric particles greatly
complicate projections of future trends in global warming processes because of emissions of
greenhouse gases; consequent increases in global mean temperature; resulting changes in
regional and local weather patterns; and mainly deleterious (but some beneficial) location-
specific human health and environmental effects. Available evidence, ranging from satellite to
in situ measurements of aerosol effects on incoming solar radiation and cloud properties, is
strongly indicative of an important role in climate for aerosols, but this role is still poorly
9-111
-------
quantified. No significant advances have been made since the 1996 PM AQCD in reducing the
uncertainties assigned to forcing estimates for aerosol-related forcing, especially for black
carbon-containing aerosol. The IPCC characterizes the scientific understanding of greenhouse
gas-related forcing as "high" in contrast to that for aerosol, which it describes as "low" to
"very low."
In addition to direct climate effects through the scattering and absorption of solar radiation,
particles also exert indirect effects on climate by serving as cloud condensation nuclei, thus
affecting the abundance and vertical distribution of clouds. The direct and indirect effects of
particles appear to have significantly offset global warming effects caused by the buildup of
greenhouse gases on a globally averaged basis. However, because the lifetime of particles is
much shorter than that required for complete mixing within the Northern Hemisphere, the
climate effects of particles generally are felt much less homogeneously than are the effects of
long-lived greenhouse gases.
Quantification of the effect of anthropogenic aerosol on hydrological cycles requires more
information than is presently available regarding ecosystems responses to reduced solar radiation
and other changes occurring in the climate system. However, several global-scale studies
indicate that aerosol cooling alone can slow down the hydrological cycle, while cooling plus the
nucleation of additional cloud droplets can dramatically reduce precipitation rates.
Any effort to model the impacts of local alterations in particle concentrations on projected
global climate change or consequent local and regional weather patterns would be subject to
considerable uncertainty.
Atmospheric particles also complicate estimation of potential future impacts on human
health and the environment projected as possible to occur because of increased transmission of
solar ultraviolet-B radiation (UV-B) through the Earth's atmosphere, secondary to stratospheric
ozone depletion due to anthropogenic emissions of chlorofluorocarbons (CFCs), halons, and
certain other gases. The transmission of solar UV-B radiation is strongly affected by
atmospheric particles. Measured attenuations of UV-B under hazy conditions range up to 37%
of the incoming solar radiation. Measurements relating variations in PM mass directly to UV-B
transmission are lacking. Particles also can affect the rates of photochemical reactions occurring
in the atmosphere, e.g., those involved in catalyzing tropospheric ozone formation. Depending
9-112
-------
on the amount of absorbing substances in the particles, photolysis rates either can be increased or
decreased. Thus, atmospheric particle effects on UV-B radiation, which vary depending on size
and composition of particles, can differ substantially over different geographic areas and from
season to season over the same area. Any projection of effects of location-specific airborne PM
alterations on increased atmospheric transmission of solar UV radiation (and associated potential
human health or environmental effects) due to stratospheric ozone-depletion would, therefore,
also be subject to considerable uncertainty.
9.3.4 Effects of Ambient PM on Man-Made Materials
The 1996 PM AQCD arrived at the following key findings and conclusions related to PM
effects on man-made materials:
"Particle exposure results in the soiling of painted surfaces and other building
materials, increasing the cleaning frequency for exposed surfaces and possibly
reducing their useful lifetimes." (U.S. EPA, 1996, p. 1-19) Damage to materials
can result from the deposition of acid aerosols and the dissolution of acid forming
gases on metal surfaces, increasing the corrosion of metals; "exposure to acid forming
gases may also limit the life expectancy of paints and may damage various building
stones and cement products beyond that resulting from natural weathering processes."
(U.S. Environmental Protection Agency, 1996, p. 1-20).
As noted in the 1996 PM AQCD and restated in Chapter 4 (Section 4.4), building materials
(metals, stones, cements, and paints) undergo natural weathering processes from exposure to
environmental elements (wind, moisture, temperature fluctuations, sun light, etc.). Metals form
a protective film of oxidized metal (e.g., rust) that slows environmentally induced corrosion.
On the other hand, the natural process of metal corrosion from exposure to natural environmental
elements is enhanced by exposure to anthropogenic pollutants, in particular SO2 or other acidic
substances, that render the protective film less effective. For example, dry deposition of SO2
enhances the effects of environmental elements on calcereous stones (limestone, marble, and
cement) by converting calcium carbonate (calcite) to calcium sulfate dihydrate (gypsum). The
rate of deterioration is determined by the SO2 concentration, the deposition rate, and the stone's
permeability and moisture content; however, the extent of the damage to stones produced by the
pollutant species above and beyond that from the natural weathering processes is uncertain.
9-113
-------
Sulfur dioxide also has been found to limit the life expectancy of paints by causing discoloration
and loss of gloss and thickness of the paint film layer.
As also highlighted in the 1996 PM AQCD, the soiling of painted surfaces and other
building materials is a significant detrimental effect of PM pollution. Soiling changes the
reflectance of a material from opaque and decreases the transmission of light through transparent
materials; it is also a degradation process that requires remediation by cleaning or washing and,
depending on the soiled surface, repainting. Available data indicate that airborne particles can
result in increased cleaning frequency of exposed surfaces and may decrease the usefulness of
soiled materials. Attempts have been made to quantify the pollutant exposures at which
materials damage and soiling have been observed; but, to date, insufficient data are available to
advance our knowledge regarding perception thresholds with respect to pollutant concentration,
particle size, and chemical composition.
9-114
-------
REFERENCES
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. I; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
Care Med. 159: 373-382.
Aber, J. D. (1992) Nitrogen cycling and nitrogen saturation in temperate forest ecosystems. Trends Ecol. Evol.
7: 220-224.
Aber, J. D.; Nadelhoffer, K. J.; Steudler, P.; Melillo, J. M. (1989) Nitrogen saturation in northern forest ecosystems:
excess nitrogen from fossil fuel combustion may stress the biosphere. Bioscience 39: 378-386.
Aber, J. D.; Magill, A.; McNulty, S. G.; Boone, R. D.; Nadelhoffer, K. J.; Downs, M.; Hallett, R. (1995) Forest
biogeochemistry and primary production altered by nitrogen saturation. Water Air Soil Pollut. 85: 1665-1670.
Aber, J.; McDowell, W.; Nadelhoffer, K.; Magill, A.; Berntson, G.; Kamakea, M.; McNulty, S.; Currie, W.; Rustad,
L.; Fernandez, I. (1998) Nitrogen saturation in temperate forest ecosystems. BioScience 48: 921-934.
Adams, P. F.; Hendershot, G. E.; Marano, M. A. (1999) Current estimates from the National Health Interview
Survey, 1996. Hyattsville, MD: U.S. Department of Health and Human Services, Public Health Service,
National Center for Health Statistics; DHHS publication no. (PHS) 99-1528. (Vital and health statistics: v. 10,
data from the National Health Survey, no. 200). Available:
http://www.cdc.gov/nchs/products/pubs/pubd/series/srlO/pre-200/pre-200.htm [12 March, 2001].
Allen, E. B.; Padgett, P. E.; Bytnerowicz, A.; Minich, R. (1998) Nitrogen deposition effects on coastal sage
vegetation of southern California. USDA Forest Service Gen. Tech. Rep. PSW-GTR-166, pp. 131-139.
American Conference of Governmental Industrial Hygienists. (1994) Appendix D: particle size-selective sampling
criteria for airborne paniculate matter. In: 1994-1995 Threshold limit values for chemical substances and
physical agents and biological exposure indices. Cincinnati, OH: American Conference of Governmental
Industrial Hygienists; pp. 43-46.
Arias, E.; Anderson, R. N.; Kung, H.-S.; Murphy, S. L.; Kochanek, K. D. (2003) Deaths: final data for 2001.
Hyattsville, MD: U.S. Department of Health & Human Services, National Center for Health Statistics; DHHS
publication no. (PHS) 2003-1120. (National vital statistics reports: v. 52, no. 3). Available:
http://www.cdc.gov/nchs/data/nvsr/nvsr52/nvsr52_03.pdf (9 June, 2004).
Baluk, P.; Nadel, J. A.; McDonald, D. M. (1992) Substance P-immunoreactive sensory axons in the rat respiratory
tract: a quantitative study of their distribution and role in neurogenic inflammation. J. Compar. Neurol.
319:586-598.
Baron, J. S.; Rueth, H. M.; Wolfe, A. M.; Nydick, K. R.; Allstott, E. J.; Minear, J. T.; Moraska, B. (2000) Ecosystem
responses to nitrogen deposition in the Colorado front range. Ecosystems 3: 352-368.
Blackwell, D. L.; Vickerie, J. L.; Wondimu, E. A. (2003) Summary health statistics for U.S. children: National
Health Interview Study, 2000. Hyattsville, MD: U.S. Department of Health & Human Services, National
Center for Health Statistics. (Vital and health statistics, series 10, no. 213). Available:
http://www.cdc.gov/nchs/data/series/sr_10/srlO_213.pdf [15 June, 2004].
Bondietti, E. A.; McLaughlin, S. B. (1992) Evidence of historical influences of acidic deposition on wood and soil
chemistry. In: Johnson, D. W.; Lindberg, S. E., eds. Atmospheric deposition and forest nutrient cycling:
a synthesis of the integrated forest study. New York, NY: Springer-Verlag: pp. 358-377. (Billings, W. D.;
Golley, F.; Lange, O. L.; Olson, J. S.; Remmert, H., eds. Ecological studies analysis and synthesis: v. 91).
Bowman, W. D. (2000) Biotic controls over ecosystem response to environmental change in alpine tundra of the
Rocky Mountains. Ambio 29: 396-400.
Bowman, W. D.; Steltzer, H. (1998) Positive feedbacks to anthropogenic nitrogen deposition in Rocky Mountain
alpine tundra. Ambio 27: 514-517.
Braunwald, E. (1997) Cardiovascular medicine at the turn of the millennium: triumphs, concerns, and opportunities.
N. Engl. J. Med. 337: 1360-1369.
Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian
cities. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
Health Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105: 614-620.
Bytnerowicz, A.; Fenn, M. E. (1996) Nitrogen deposition in California forests: a review. Environ. Pollut.
92: 127-146.
9-115
-------
Campen, M. I; Costa, D. L.; Watkinson, W. P. (2000) Cardiac and thermoregulatory toxicity of residual oil fly ash
in cardiopulmonary-compromised rats. In: Phalen, R. F., ed. Inhalation toxicology: proceedings of the third
colloquium on paniculate air pollution and human health (second special issue); June, 1999; Durham, NC.
Inhalation Toxicol. 12(suppl. 2): 7-22.
Centers for Disease Control and Prevention. (2004) The health consequences of smoking: a report of the Surgeon
General. Atlanta, GA: U.S. Department of Health and Human Services, National Center for Chronic Disease
Prevention and Health Promotion, Office on Smoking and Health. Available:
http://www.cdc.gov/tobacco/sgr/sgr_2004/chapters.htm (18 August, 2004).
Cherry, D. K.; Burt, C. W.; Woodwell, D. A. (2003) National ambulatory medical care survey: 2001 summary.
Hyattsville, MD: U.S. Department of Health & Human Resources, National Center for Health Statistics;
DHHS publication no. (PHS) 2003-1250. (Advance data from vital and health statistics: no. 337). Available:
http://www.cdc.gov/nchs/data/ad/ad337.pdf (9 June, 2004).
Chestnut, L. G.; Dennis, R. L. (1997) Economic benefits of improvements invisibility: acid rain. Provisions of the
1990 Clean Air Act Amendments. J. Air Waste Manage. Assoc. 47: 395-402.
Chestnut, L. G.; Rowe, R. D. (1990) Preservation values for visibility protection at the national parks (draft final
report). Research Triangle Park, NC: U.S. Environmental Protection Agency, Office of Air Quality Planning
and Standards, Economic Analysis Branch.
Chock, D. P.; Winkler, S. L.; Chen, C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52: 113-117.
Chughtai, A. R.; Williams, G. R.; Atteya, M. M. O.; Miller, N. J.; Smith, D. M. (1999) Carbonaceous particle
hydration. Atmos. Environ. 33: 2679-2687.
Cocker, D. R., Ill; Whitlock, N. E.; Flagen, R. C.; Seinfield, J. H. (2001) Hygroscopic properties of Pasadena,
California aerosol. Aerosol Sci. Technol. 35: 637-647.
Darlington, T. L.; Kahlbaum, D. F.; Heuss, J. M.; Wolff, G. T. (1997) Analysis of PM10 trends in the United States
from 1988 through 1995. J. Air Waste Manage. Assoc. 47: 1070-1078.
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
Delfino, R. J.; Murphy-Moulton, A. M.; Becklake, M. R. (1998) Emergency room visits for respiratory illnesses
among the elderly in Montreal: association with low level ozone exposure. Environ. Res. 76: 67-77.
Di Minno, G.; Mancini, M. (1990) Measuring plasma fibrinogen to predict stroke and mycardial infarction.
Arteriosclerosis 10: 1-7.
Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
symptoms. Environ. Health Perspect. 104: 500-505.
Dominici, F.; McDermott, A.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2003) Mortality among residents of 90 cities.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 9-24. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Driscoll, C. T.; Lawrence, G. B.; Bulger, A. J.; Butler, T. J.; Cronan, C. S.; Eagar, C.; Lambert, K. F.; Likens, G. E.;
Stoddard, J. L.; Weathers, K. C. (2001) Acidic deposition in the northeastern United States: sources and
inputs, ecosystem effects, and management strategies. BioScience 51: 180-198.
Edgerton-Warburton, L. M.; Allen, E. B. (2000) Shifts in arbuscular mycorrhizal communities along an
anthropogenic nitrogen deposition gradient. Ecol. Appl. 10: 484-496.
Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 97-106. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Fanucchi, M. V.; Wong, V. J.; Hinds, D.; Tarkington, B. K.; Van Winkle, L. S.; Evans, M. J.; Plopper, C. G. (2000)
Repeated episodes of exposure to ozone alters postnatal development of distal conducting airways in infant
rhesus monkeys. Am. J. Respir. Crit. Care Med. 161: A615.
Fenn, M. E.; Poth, M. A. (1998) Indicators of nitrogen status in California forests. U.S.D.A. Forest Service Gen.
Tech. Rep. PSW-GTR-166.
Fenn, M. E.; Poth, M. A.; Aber, J. D.; Baron, J. S.; Bormann, B. T.; Johnson, D. W.; Lemly, A. D.; McNulty, S. G.;
Ryan, D. F.; Stottlemyer, R. (1998) Nitrogen excess in North American ecosystems: predisposing factors,
ecosystem responses, and management strategies. Ecol. Appl. 8: 706-733.
9-116
-------
Perm, M. E.; Haeuber, R.; Tonnesen, G. S.; Baron, J. S.; Grossman-Clarke, S.; Hope, D.; Jaffe, D. A.; Copeland, S.;
Geiser, L.; Rueth, H. M.; Sickman, J. O. (2003) Nitrogen emissions, deposition, and monitoring in the
western United States. Bioscience 53: 391-403.
Friedlander, S. K.; Yeh, E. K. (1998) The submicron atmospheric aerosol as a carrier of reactive chemical species:
case of peroxides. Appl. Occup. Environ. Hyg. 13: 416-420.
Galloway, J. N. (1998) The global nitrogen cycle: changes and consequences. Environ. Pollut. 102(suppl. 1): 15-24.
Galloway, J. N.; Cowling, E. B. (2002) Reactive nitrogen and the world: 200 years of change. Ambio 31: 64-71.
Galloway, J. N.; Aber, J. D.; Erisman, J. W.; Seitzinger, S. P.; Howarth, R. W.; Cowling, E. B.; Cosby, B. J. (2003)
The nitrogen cascade. Bioscience 53: 341-356.
Garner, J. H. B. (1994) Nitrogen oxides, plant metabolism, and forest ecosystem response. In: Alscher, R. G.;
Wellburn, A. R., eds. Plant responses to the gaseous environment: molecular, metabolic and physiological
aspects, [3rd international symposium on air pollutants and plant metabolism]; June 1992; Blacksburg, VA.
London, United Kingdom: Chapman & Hall; pp. 301-314.
Gilliam, F. S.; Adams, M. B.; Yurish, B. M. (1996) Ecosystem nutrient responses to chronic nitrogen inputs at
Fernow Experimental Forest, West Virginia. Can. J. For. Res. 26: 196-205.
Hall, M. J.; DeFrances, C. J. (2003) 2001 national hospital discharge survey. Hyattsville, MD: U.S. Department of
Health & Human Resources, National Center for Health Statistics; DHHS publication no. (PHS) 2003-1250.
(Advance data from vital and health statistics: no. 332). Available:
http://www.cdc.gov/nchs/data/ad/ad332.pdf (9 June, 2004).
Health Effects Institute. (2003) Revised analyses of time-series studies of air pollution and health. Boston, MA:
Health Effects Institute; special report. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[27 June 2003].
Hemming, B. L.; Seinfeld, J. H. (2001) On the hygroscopic behavior of atmospheric organic aerosols. Ind. Eng.
Chem. Res. 40: 4162-4171.
Hill, A. B. (1965) The environment and disease: association or causation? Proc. R. Soc. Med. 58: 295-300.
Howarth, R. W.; Boyer, E. W.; Pabich, W. J.; Galloway, J. N. (2002) Nitrogen use in the United States from
1961-2000 and potential future trends. Ambio 31: 88-96.
Ito, K. (2003) Associations of paniculate matter components with daily mortality and morbidity in Detroit,
Michigan. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[12 May, 2004].
Ito, K.; Thurston, G. D. (1996) Daily PM10/mortality associations: an investigation of at-risk subpopulations.
J. Exposure Anal. Environ. Epidemiol. 6: 79-95.
Johnson, D. W.; Lindberg, S. E. (1992) Nitrogen chemistry, deposition, and cycling in forests. In: Johnson, D. W.;
Lindberg, S. E., eds. Atmospheric deposition and forest nutrient cycling: a synthesis of the integrated forest
study. New York, NY: Springer-Verlag, Inc.; pp. 150-213. (Billings, W. D.; Golley, F.; Lange, O. L.; Olson,
J. S.; Remmert, H., eds. Ecological studies: analysis and synthesis: v. 91).
Kao, A. S.; Friedlander, S. K. (1995) Temporal variations of paniculate air pollution: a marker for free radical
dosage and adverse health effects? Inhalation Toxicol. 7: 149-156.
Katsouyanni, K.; Touloumi, G.; Samoli, E.; Petasakis, Y.; Analitis, A.; Le Tertre, A.; Rossi, G.; Zmirou, D.;
Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
Schindler, C.; Schwartz, J. (2003) Sensitivity analysis of various models of short-term effects of ambient
particles on total mortality in 29 cities in APHEA2. In: Revised analyses of time-series studies of air pollution
and health. Special report. Boston, MA: Health Effects Institute; pp. 157-164. Available:
http://www.healtheffects.org/news.htm [16 May, 2003].
Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
(1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly ash
inhalation. Inhalation Toxicol. 9: 541-565.
Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM10 associations in Los Angeles.
In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air pollution and human
mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 59-69.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol Research and Inhalation Epidemiological Study (ARIES): air
quality and daily mortality statistical modeling-interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 165-172. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
9-117
-------
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J.; Jerrett, M.; Abrahamowicz, M; White,
W. H. (2000) Reanalysis of the Harvard Six Cities study and the American Cancer Society study of
paniculate air pollution and mortality: a special report of the Institute's Particle Epidemiology Reanalysis
Project. Cambridge, MA: Health Effects Institute.
Ktinzli, N.; Medina, S.; Kaiser, R.; Quenel, P.; Horak, F., Jr.; Studnicka, M. (2001) Assessment of deaths
attributable to air pollution: should we use risk estimates based on time series or on cohort studies? Am. J.
Epidemiol. 153: 1050-1055.
Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108: 941-947.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lipfert, F. W. (1998) Trends in airborne paniculate matter in the United States. Appl. Occup. Environ. Hyg.
13: 370-384.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000) Daily mortality in the Philadelphia metropolitan area and
size-classified paniculate matter. J. Air Waste Manage. Assoc. 50: 1501-1513.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report no.
95.
Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
California. Environ. Health Perspect. 105: 216-222.
Lekke, H.; Bak, J.; Falkengren-Grerup, U.; Finlay, R. D.; Ilvesniemi, H.; Nygaard, P. H.; Starr, M. (1996) Critical
loads of acidic deposition for forest soils: is the current approach adequate. Ambio 25: 510-516.
Long, C. M.; Suh, H. H.; Catalano, P. J.; Koutrakis, P. (2001) Using time- and size-resolved paniculate data to
quantify indoor penetration and deposition behavior. Environ. Sci. Technol. 35: 2089-2099.
Lovett, G. M.; Traynor, M. M.; Pouyat, R. V.; Carreiro, M. M.; Zhu, W.-X.; Baxter, J. W. (2000) Atmospheric
deposition to oak forest along an urban-rural gradient. Environ. Sci. Technol. 34: 4294-4300.
Lowe, G. D. O.; Lee, A. J.; Rumley, A.; Price, J. F.; Fowkes, F. G. R. (1997) Blood viscosity and rtsk of
cardiovascular events: the Edinburgh Artery Study. Br. J. Haematol. 96: 168-173.
Magill, A. H.; Aber, J. D.; Berntson, G. M.; McDowell, W. H.; Nadelhoffer, K. J.; Melillo, J. M.; Steudler, P. (2000)
Long-term nitrogen additions and nitrogen saturation in two temperate forests. Ecosystems 3: 238-253.
Malm, W. C. (2000) Spatial and seasonal patterns and temporal variability of haze and its constituents in the United
States. Report III. Fort Collins, CO: Cooperative Institute for Research in the Atmosphere, Colorado State
University. Available: http://vista.cira.colostate.edu/improve/Publications/Reports/2000/2000.htm
[22 March, 2002].
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108: 347-353.
Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003) Air pollution and cardiovascular
mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
http://www.healtheffects.org/news.htm [16 May, 2003].
McCaig, L. F.; Burt, C. W. (2004) National hospital ambulatory medical care survey: 2002 emergency department
summary. Hyattsville, MD: U.S. Department of Health & Human Resources, Centers for Disease Control and
Prevention, National Center for Health Statistics; DHHS publication no. (PHS) 2004-1250. (Advance data
from vital and health statistics: no. 340). Available: http://www.cdc.gov/nchs/data/ad/ad340.pdf (9 June,
2004).
Meade, T. W.; Ruddock, V.; Stirling, Y.; Chakrabarti, R.; Miller, G. J. (1993) Fibrinolytic activity, clotting factors,
and long-term incidence of ischaemic heart disease in the Northwick Park Heart Study. Lancet
342: 1076-1079.
Molenar, J. V.; Malm, W. C.; Johnson, C. E. (1994) Visual air quality simulation techniques. Atmos. Environ.
28: 1055-1063.
Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Moomaw, W. R. (2002) Energy, industry and nitrogen: strategies for decreasing reactive nitrogen emissions. Ambio
31: 184-189.
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
9-118
-------
National Acid Precipitation Assessment Program. (1991) National Acid Precipitation Assessment Program 1990
integrated assessment report. Washington, DC: National Acid Precipitation Assessment Program.
National Acid Precipitation Assessment Program. (1998) NAPAP biennial report to Congress. Washington, DC:
National Acid Precipitation Assessment Program.
National Research Council. (1993) Protecting visibility in national parks and wilderness areas. Washington, DC:
National Academy Press. 3v.
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
and exposure to air pollution. Public Health Rep. 114: 135-148.
Ostro, B. D.; Lipsett, M. I; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
Ostro, B.; Lipsett, M.; Mann, J.; Braxton-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
African-American children in Los Angeles. Epidemiology 12: 200-208.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coachella Valley,
California. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
Health Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Padgett, P. E.; Allen, E. B.; Bytnerowicz, A.; Minich, R. A. (1999) Changes in soil inorganic nitrogen as related to
atmospheric nitrogenous pollutants in southern California. Atmos. Environ. 33: 769-781.
Paerl, H. W.; Bales, J. D.; Ausley, L. W.; Buzzelli, C. P.; Crowder, L. B.; Eby, L. A.; Go, M.; Peierls, B. L.;
Richardson, T. L.; Ramus, J. S. (2001) Ecosystem impacts of three sequential hurricanes (Dennis, Floyd, and
Irene) on the United States' largest lagoonal estuary, Pamlico Sound, NC. Proc. Nat. Acad. Sci. U. S. A.
98:5655-5611.
Piedimonte, G.; Hoffman, J. I. E.; Husseini, W. K.; Hiser, W. L.; Nadel, J. A. (1992) Effect of neuropeptides
released from sensory nerves on blood flow in the rat airway microcirculation. J. Appl. Physiol.
72: 1563-1570.
Pleis, J. R.; Benson, V.; Schiller, J. S. (2003) Summary health statistics for U.S. adults: National Health Interview
Study, 2000. Hyattsville, MD: U.S. Department of Health & Human Services, National Center for Health
Statistics. (Vital and health statistics, series 10, no. 215). Available:
http://www.cdc.gov/nchs/data/series/sr_10/srlO_215.pdf [15 June, 2004].
Plopper, C. G.; Fanucchi, M. V. (2000) Do urban environmental pollutants exacerbate childhood lung diseases?
Environ. Health Perspect. 108: A252-A253.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah valley. Arch.
Environ. Health 47: 211-217.
Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. JAMA J. Am. Med.
Assoc. 287: 1132-1141.
Rabalais, N. N. (2002) Nitrogen in aquatic ecosystems. Ambio 31: 102-112.
Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
Perspect. 104: 506-514.
Rovira, A. D.; Davey, C. B. (1974) Biology of the rhizosphere. In: Carson, E. W., ed. The plant root and its
environment: proceedings of an institute; July, 1971; Blacksburg, VA. Charlottesville, VA: University Press
of Virginia; pp. 153-204.
Samet, J. M.; Zeger, S. L.; Dominici, F.; Curriero, F.; Coursac, L; Dockery, D. W.; Schwartz, J.; Zanobetti, A.
(2000) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute; research report no. 94, part II.
Schwartz, J. (1994) Air pollution and hospital admissions for the elderly in Detroit, Michigan. Am. J. Respir. Crit.
Care Med. 150: 648-655.
Schwartz, J. (2003) Airborne particles and daily deaths in 10 US cities. In: Revised analyses of time-series studies of
air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 211-218. Available:
http://www.healtheffects.org/news.htm [16 May, 2003].
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
Michigan. Am. J. Epidemiol. 142: 23-35.
9-119
-------
Schwartz, I; Zanobetti, A.; Bateson, T. (2003) Morbidity and mortality among elderly residents of cities with daily
PM measurements. In: Revised analyses of time-series studies of air pollution and health. Special report.
Boston, MA: Health Effects Institute; pp. 25-58. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[20 June, 2004].
Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Paniculate air pollution and acute health effects. Lancet
(8943): 176-178.
Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Shortle, W. C.; Smith, K. T. (1988) Aluminum-induced calcium deficiency syndrome in declining red spruce.
Science (Washington, DC) 240: 1017-1018.
Sjogren, B. (1997) Occupational exposure to dust: inflammation and ischaemic heart disease. Occup. Environ. Med.
54: 466-469.
Smiley-Jewell, S. M.; Liu, F. J.; Weir, A. J.; Plopper, C. G. (2000) Acute injury to differentiating Clara cells in
neonatal rabbits results in age-related failure of bronchiolar epithelial repair. Toxicol. Pathol. 28: 267-276.
Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
John, Canada. J. Exposure Anal. Environ. Epidemiol. 10: 461-477.
Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne paniculate matter on daily
death counts. Environ. Health Perspect. 103: 490-497.
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
65: 271-290.
Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000) Air quality and
pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
U.S. Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park,
NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525, and
PB95-124517.
U.S. Environmental Protection Agency. (1995) Interim findings on the status of visibility research. Research
Triangle Park, NC: Office of Research and Development; report no. EPA/600/R-95/021.
U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
U.S. Environmental Protection Agency. (2003) National air quality and emissions trends report. 2003 special studies
edition. Research Triangle Park, NC: Office of Air Quality Standards; Emissions Monitoring and Analysis
Division; report no. EPA454/R-03-005. Available: http://www.epa.gov/air/airtrends/aqtrnd03/toc.html
(27 August, 2004).
Van Egmond, K.; Bresser, T.; Bouwman, L. (2002) The European nitrogen case. Ambio 31: 72-78.
Veronesi, B.; Oortgiesen, M. (2001) Neurogenic inflammation and paniculate matter (PM) air pollutants.
Neurotoxicology 22: 795-810.
Vincent, R.; Kumarathasan, P.; Goegan, P.; Bjarnason, S. G.; Guenette, J.; Berube, D.; Adamson, I. Y.;
Desjardins, S.; Burnett, R. T.; Miller, F. J.; Battistini, B. (2001) Inhalation toxicology of urban ambient
paniculate matter: acute cardiovascular effect in rats. Boston, MA: Health Effects Institute; research report
no. 104. Available: http://www.healtheffects.org/Pubs/Vincent.pdf [29 January 02].
Vitousek, P. M.; Mooney, H. A.; Lubchenco, J.; Melillo, J. M. (1997) Human domination of Earth's ecosystems.
Science (Washington, DC) 277: 494-499.
Wall, D. H.; Moore, J. C. (1999) Interactions underground: soil biodiversity, mutualism, and ecosystem processes.
BioScience 49: 109-117.
Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil fly
ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
Wedin, D. A.; Tilman, D. (1996) Influence of nitrogen loading and species composition on the carbon balance of
grasslands. Science (Washington, DC) 274: 1720-1723.
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
Wilson, W. E. (1995) Aerosol exposure, physics, and chemistry. Inhalation Toxicol. 7: 769-772.
9-120
-------
Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
Zanobetti, A.; Schwartz, J. (2003) Airborne particles and hospital admissions for heart and lung disease. In: Revised
analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
Institute; pp. 241-248. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
9-121
-------
United States
Environmental Protection
Agency
Please make all necessary changes in the below label,
detach copy or copy, and return to the address in the upper
left-hand corner.
If you do not wish to receive these reports CHECK HERE |_
detach copy or copy, and return to the address in the
upper left-hand corner.
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
National Center for
Environmental Assessment
Research Triangle Park, NC 27711
Official Business
Penalty for Private Use
$300
EPA/600/P-99/002bF
October 2004
------- |