United States Office of Research EPA 600/P-99/002b
Environmental Protection and Development October 1999
Agency Washington, DC 20460 External Review Draft
Air Quality Criteria for
Particulate Matter
Volume II
Notice
This document is a preliminary draft. It has not been formally
released by EPA and should not at this stage be construed to
represent Agency policy. It is being circulated for comment on
its technical accuracy and policy implications.
-------
EPA 600/P-99/002b
October 1999
External Review Draft
Air Quality Criteria for
Particulate Matter
Volume II
Notice
This document is a preliminary draft. It has not been formally
released by EPA and should not at this stage be construed to
represent Agency policy. It is being circulated for comment on its
technical accuracy and policy implications.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
Disclaimer
This document is an external review draft for review purposes only and does not constitute
U.S. Environmental Protection Agency policy. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
October 1999 Il-ii DRAFT-DO NOT QUOTE OR CITE
-------
Preface
National Ambient Air Quality Standards (NAAQS) are promulgated by the United States
Environmental Protection Agency (U.S. 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 may reasonably be expected to endanger public health or
welfare; (2) to issue air quality criteria for them which 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 and 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-yrs) review and revise, as appropriate, the criteria
and NAAQS for a given listed pollutant or class of pollutants.
The original U.S. 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 |um) 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 Particulate 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 |ug/m3, 24-h; 50 |ug/m3, annual ave.) were
retained in modified form and new standards (65 |ug/m3, 24-h; 15 |ug/m3, annual ave.) for
particles < 2.5 |um (PM25) were promulgated in July 1997.
This First External Review Draft of revised Air Quality Criteria for Particulate Matter
assesses new scientific information that has become available since early 1996 through mid-
1999. Extensive additional pertinent information is expected to be published during the next 6 to
9 months (including results from a vastly expanded U.S. EPA PM Research program and from
other Federal and State Agencies, as well as other partners in the general scientific community)
and, as such, the findings and conclusions presented in this draft document must be considered
only provisional at this time. The present draft is being released for public comment and review
by the Clean Air Scientific Advisory Committee (CASAC) mainly to obtain comments on the
October 1999 II-iii DRAFT-DO NOT QUOTE OR CITE
-------
organization and structure of the document, the issues addressed, and the approaches employed
in assessing and interpreting the thus far available new information on PM exposures and effects.
Public comments and CASAC review recommendations will be taken into account, along with
newly available information published or accepted for peer-reviewed publication by April/May
2000, in making further revisions to this document for incorporation into a Second External
Review Draft. That draft is expected to be released in June 2000 for further public comment and
CASAC review (September 2000) in time for final revisions to be completed by December
2000). Evaluations contained in the present document will be drawn upon to provide inputs to
associated PM Staff Paper analyses prepared by EPA's Office of Air Quality Planning and
Standards (OAQPS) to pose options for consideration by the EPA Administrator with regard to
proposal and, ultimately, promulgation by July 2000 of decisions on potential retention or
revision of the current PM NAAQS.
This document was prepared and reviewed by experts from Federal and State government
agencies, academia, industry, and NGO's for use by EPA in support of decision making on
potential public health and environmental risks of ambient PM. It describes the nature, sources,
distribution, measurement, and concentrations of PM in both the outdoor (ambient) and indoor
environments and evaluates the latest data on the health effects in exposed human populations, as
well as environmental effects on: vegetation and ecosystems; visibility and climate; manmade
materials; and associated economic impacts. Although not intended to be an exhaustive literature
review, this document is intended to assess all pertinent literature through mid-1999.
The National Center for Environmental Assessment - Research Triangle Park, NC
(NCEA-RTP) acknowledges the contributions provided by authors, contributors, and reviewers
and the diligence of its staff and contractors in the preparation of this document.
October 1999 Il-iv DRAFT-DO NOT QUOTE OR CITE
-------
Air Quality Criteria for Particulate Matter
VOLUME I
1. EXECUTIVE SUMMARY 1-1
2. INTRODUCTION 2-1
3. PHYSICS, CHEMISTRY, AND MEASUREMENT OF
PARTICULATE MATTER 3-1
4. CONCENTRATIONS, SOURCES, AND EMISSIONS OF
ATMOSPHERIC PARTICLES 4-1
APPENDIX 4A: Composition of Particulate Matter Source
Emissions 4A-1
5. HUMAN EXPOSURE TO AMBIENT PARTICULATE MATTER:
RELATIONS TO CONCENTRATIONS OF AMBIENT AND
NON-AMBIENT PARTICULATE MATTER AND OTHER AIR
POLLUTANTS 5-1
APPENDIX 5A: Nomenclature 5A-1
VOLUME II
6. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM
AMBIENT PARTICULATE MATTER 6-1
7. DOSIMETRY AND TOXICOLOGY OF PARTICULATE
MATTER 7-1
8. INTEGRATIVE SYNTHESIS OF KEY POINTS: PARTICULATE
MATTER EXPOSURE, DOSIMETRY, AND HEALTH RISKS 8-1
VOLUME III
9. ENVIRONMENTAL EFFECTS OF PARTICULATE MATTER 9-1
October 1999 II-v DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
Page
List of Tables II-xiii
List of Figures II-xix
Authors, Contributors, and Reviewers II-xxi
U.S. Environmental Protection Agency Project Team for Development of
Air Quality Criteria for Particulate Matter II-xxvii
6. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM AMBIENT
PARTICULATE MATTER 6-1
6.1 INTRODUCTION 6-1
6.2 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE 6-5
6.2.1 Short-Term Effects on Lung Function and Respiratory Symptoms 6-6
6.2.1.1 Short-Term Effects on Lung Function And Respiratory
Symptoms in Asthmatics 6-8
6.2.1.2 Short-Term Effects on Lung Function and Respiratory
Symptoms in Non-Asthmatics 6-21
6.2.1.3 Discussion of Co-Pollutant Studies 6-26
6.2.2 Long-Term Exposure Effects on Lung Function and Respiratory
Symptoms 6-30
6.2.3 Effects of Short-Term PM Exposure on the Incidence of Respiratory
Medical Visits and Hospital Admissions 6-34
6.2.3.1 Introduction 6-34
6.2.3.2 Summary and Implications of Studies Assessed in the
1996 PM AQCD 6-35
6.2.3.3 Key New Respiratory Medical Visits Studies 6-35
6.2.3.4 Key New Hospital Admissions Studies 6-42
6.2.3.5 Syntheses of Comparable Hospital Admissions PM10 and
SO4= Studies 6-51
6.2.3.6 Overall Conclusions 6-55
6.2.4 Cardiovascular Effects Associated with Acute Ambient PM Exposure . . 6-56
6.2.4.1 Introduction 6-56
6.2.4.2 Summary of Conclusions from 1996 PM AQCD 6-56
6.2.4.3 Review of New Studies 6-58
6.2.4.4 Individual-Level Studies of Cardiovascular Physiology 6-63
6.3 MORTALITY EFFECTS OF PARTICULATE MATTER EXPOSURE 6-65
6.3.1 Introduction 6-65
6.3.2 Mortality Effects of Short-Term Particulate Matter Exposure 6-66
6.3.2.1 Summary of 1996 PM Criteria Document Findings on
Unresolved Issues 6-66
6.3.2.2 New Multi-City Studies 6-69
6.3.2.3 New Results From Individual City Studies 6-73
6.3.2.4 New Studies on the Temporal Structure of Short-Term
Effects 6-83
October 1999 II-vii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
6.3.2.5 New Assessments of Confounding 6-93
6.3.2.6 New Assessments of Cause-Specific Mortality 6-98
6.3.2.7 New Assessment of Methodological Issues 6-100
6.3.2.8 Summary of Newly Available Information 6-101
6.3.3 Human Mortality and Long-Term Exposure to PM of Ambient
Origin 6-103
6.3.3.1 Studies Published Prior to the 1996 Particulate Matter
Criteria Document 6-103
6.3.3.2 Prospective Cohort Studies of Chronic Exposure Published
Since the Last Particulate Matter Criteria Document 6-107
6.3.3.3 Relationship of AHSMOG to Six Cities and ACS Study
Findings 6-117
6.3.3.4 Population-Based Mortality Studies in Children 6-122
6.3.3.5 Studies by Particulate Matter Size-Fraction and
Composition 6-123
6.3.3.6 Shortening-of-Life Associated with Long-Term Exposure
to Particulate Matter of Ambient Origins 6-125
6.3.3.7 Effects of Exposure to Multiple Pollutants 6-126
6.3.3.8 Discussion 6-129
6.3.3.9 Conclusions 6-131
6.4 DISCUSSION OF EPIDEMIOLOGY FINDINGS 6-132
6.4.1 Overview of Section 6-132
6.4.2 Relationship of Ambient Concentrations to Specific Diseases and
Age Groups 6-135
6.4.2.1 Asthma Studies 6-135
6.4.2.2 Other Non-Asthma Studies 6-139
6.4.2.3 Cardiovascular Effects of Ambient PM Exposure 6-139
6.4.2.4 Issues in the Interpretation of Acute Cardiovascular
Effects Studies 6-140
6.4.3 Consistency of Health Effects for Short-Term and Long-Term
Exposure 6-142
6.4.4 Susceptible Populations 6-146
6 A A.I Summary of Previous Criteria Document 6-146
6.4.4.2 Children as a Susceptible Subpopulation 6-147
6.4.5 Consistency of Mortality and Morbidity Effects (Coherence) 6-155
6.4.6 Effects of PM Size Distribution and Composition 6-157
6.4.6.1 Summary of Previous 1996 PM AQCD 6-157
6.4.6.2 Assessment of Effects of PM25 and Gaseous Co-Pollutants . . . 6-159
6.4.6.3 Factors and Components Including PM 6-180
6.4.6.4 Chemical Components of PM25 and PM10 6-181
October 1999 II-viii DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
6.4.7 Effects of Exposure Estimation and Model Specification Errors in
Epidemiology Studies 6-182
6.4.7.1 Measurement Errors in Air Pollution Exposure Surrogates:
General Issues 6-183
6.4.7.2 New Theoretical Assessments of Consequences of
Measurement Error 6-185
6.4.7.3 Concentration-Response Relationships 6-191
6.4.7.4 Methodology Issues in Modeling Time Series Studies 6-193
6.4.7.5 General Issues in Modeling Prospective Cohort Studies 6-195
6.4.7.6 Effects of Co-Pollutants on Estimated PM Effects 6-195
6.4.8 Synthesis of Information from Multiple Studies: Meta-Analyses 6-197
6.4.8.1 General Issues 6-197
6.4.8.2 Other Recent Research Syntheses and Reviews 6-201
6.5 THE USE OF EPIDEMIOLOGY STUDIES FOR CAUSAL INFERENCES
ABOUT PM HEALTH EFFECTS 6-201
6.5.1 Causal Inference and Preventive Intervention 6-201
6.5.2 Biological Plausibility in Causal Inference 6-204
6.5.3 Natural Experiments, Quasi-Experiments, and Causal Inference 6-205
6.6 CONCLUSIONS AND DISCUSSION 6-206
6.6.1 Conclusions 6-206
6.6.2 PM Health Effects May Occur in Any Size Fraction; Some Health
Effects May Also Be Absent From Some Mass Fractions Under
Some Circumstances 6-208
6.6.3 PM Health Effects at Different Time Scales 6-211
6.6.4 Alternative Hypotheses for Adverse Health Effects 6-212
REFERENCES 6-216
7. DOSIMETRY AND TOXICOLOGY OF PARTICULATE MATTER 7-1
7.1 INTRODUCTION 7-1
7.2 PARTICLE DOSIMETRY 7-2
7.2.1 Introduction 7-2
7.2.1.1 Size Characterization of Inhaled Particles 7-4
7.2.1.2 Structure of the Respiratory Tract 7-5
7.2.2 Factors Controlling Dose 7-5
7.2.3 Particle Deposition 7-6
7.2.3.1 Mechanisms of Deposition 7-6
7.2.3.2 Deposition Patterns in the Human Respiratory Tract 7-8
7.2.3.3 Biological Factors Modifying Deposition 7-13
7.2.3.4 Interspecies Patterns of Deposition 7-20
7.2.4 Particle Clearance and Translocation 7-22
7.2.4.1 Mechanisms and Pathways of Clearance 7-22
October 1999 Il-ix DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
7.2.4.2 Clearance Kinetics 7-27
7.2.4.3 Interspecies Patterns of Clearance 7-31
7.2.4.4 Biological Factors Modifying Clearance 7-32
7.2.5 Particle Overload 7-35
7.2.6 Comparison of Deposition and Clearance Patterns of Particles
Administered by Inhalation and Intratracheal Instillation 7-35
7.2.7 Modeling the Disposition of Particles in the Respiratory Tract 7-38
7.2.7.1 Modeling Deposition and Clearance 7-38
7.2.7.2 Models To Estimate Retained Dose 7-42
7.3 TOXICOLOGY OF PARTICULATE MATTER 7-44
7.3.1 Summary of Previous Criteria Document 7-44
7.3.1.1 Acid Aerosols 7-44
7.3.1.2 Metals 7-46
7.3.1.3 Ultrafme Particles 7-46
7.3.1.4 Bioaerosols 7-47
7.3.1.5 "Other Particulate Matter" 7-47
7.3.2 Respiratory Effects of Particles 7-48
7.3.3 Effects in Healthy Humans 7-49
7.3.3.1 Human Acid Aerosol Exposure Studies 7-49
7.3.4 Effects in Healthy Animals 7-53
7.3.4.1 Ambient Particles 7-53
7.3.4.2 Coal Fly Ash or Residual Oil Fly Ash 7-61
7.3.4.3 In Vitro Exposures 7-66
7.3.5 Susceptibility to the Effects of PM Exposure 7-78
7.3.5.1 Effects of PM on Cardiopulmonary Compromised Hosts 7-78
7.3.5.2 Effect of PM on Allergic Hosts 7-80
7.3.5.3 Resistance to Infectious Disease 7-86
7.4 CARDIOVASCULAR AND OTHER SYSTEMIC RESPONSES TO PM 7-88
7.5 RESPONSES TO PM AND GASEOUS POLLUTANT MIXTURES 7-91
7.6 MECHANISMS OF PM TOXICITY AND PATHOPHYSIOLOGY 7-97
7.6.1 Introduction 7-97
7.6.2 Soluble Metals and Reactive Oxygen Intermediates 7-97
7.6.3 Intracellular Signaling Mechanisms 7-100
7.6.4 The Role of Particle Size and Surface Area 7-102
7.6.5 Summary 7-104
REFERENCES 7-105
8. INTEGRATIVE SYNTHESIS OF KEY POINTS: PM EXPOSURE,
DOSIMETRY, AND HEALTH RISKS 8-1
8.1 INTRODUCTION 8-1
October 1999 II-x DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
8.2 AIRBORNE PARTICLES: DISTINCTIONS BETWEEN FINE AND
COARSE PARTICLES AS SEPARATE POLLUTANT SUBCLASSES 8-1
8.2.1 Size Distinctions 8-3
8.2.2 Formation Mechanisms 8-5
8.2.3 Chemical Composition 8-7
8.2.3.1 Fine-Mode Particulate Matter 8-7
8.2.3.2 Coarse-Mode Particulate Matter 8-8
8.2.4 Atmospheric Behavior 8-9
8.2.5 Sources 8-9
8.2.6 Community and Personal Ambient PM Concentrations Exposure
Relationships 8-11
8.3 FACTORS AFFECTING DOSIMETRY 8-14
8.3.1 Factors Determining Deposition and Clearance 8-14
8.3.2 Factors Determining Toxicant-Target Interactions and Response 8-18
8.3.3 Construction of Exposure-Dose-Response Continuum for PM 8-20
8.4 EXPANDING EPIDEMIOLOGIC INFORMATION ON HEALTH
EFFECTS OF PARTICULATE MATTER 8-21
8.4.1 Introduction 8-21
8.4.2 Strengths and Limitations in the Newly Available Daily
Time-Series Studies 8-27
8.4.3 Combining Results from Hospital Admissions Studies 8-28
8.4.4 Strengths and Limitations of Prospective Cohort Studies 8-30
8.4.5 Evaluating the Coherence of the New Studies 8-33
8.4.6 Evaluating the Plausibility of Inferences about the Relationships
Between Human Health and Ambient PM Concentrations 8-38
8.4.7 Assessing the Extent to Which Adverse Health Effects Are
Attributable to PM Size Fractions or Components, or Other
Environmental Factors 8-39
8.4.7.1 Introduction 8-39
8.4.7.2 Epidemiology Evidence Suggesting That Crustal Particles
Are Less Harmful to Human Health 8-40
8.4.7.3 Toxicology Evidence Suggesting That Crustal Particles
Are Less Harmful to Human Health 8-47
8.4.8 Quantifying the Relationships Between Ambient PM Concentrations
and Adverse Health Effects in Susceptible Subpopulations at
Different Time Scales 8-47
8.5 TOXICOLOGIC EVALUATION OF PATHOPHYSIOLOGIC EFFECTS
OF PM CONSTITUENTS AND MECHANISMS OF ACTION 8-50
8.5.1 Possible Mechanisms of PM-Induced Injury 8-50
8.5.2 Ultrafme Particles 8-53
8.5.3 Bioaerosols 8-54
October 1999 Il-xi DRAFT-DO NOT QUOTE OR CITE
-------
Table of Contents
(cont'd)
8.5.4 Metals 8-55
8.5.5 Concentrated Ambient Particles 8-56
8.5.6 Summary 8-56
REFERENCES 8-58
October 1999 II-xii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
Number Page
6-1 Alternative Versions of Hypothesis 1 That May Affect the Synthesis of
Epidemiology Studies 6-2
6-2 Effect of 50 //g/m3 PM10 on Evening Peak Flow in Asthmatics Lagged
Zero or One Day 6-15
6-2A Effect of 50 //g/m3 PM10 on Morning Peak Flow in Asthmatics Lagged
Zero or One Day 6-16
6-3 Effect of 50 //g/m3 PM10 on Evening Peak Flow in Asthmatics Lagged
Two to Five Days 6-16
6-4 Effect of 25 //g/m3 PM2 5 on Evening Peak Flow in Asthmatics Lagged
Zero or One Day 6-17
6-5 Effect of 25 //g/m3 PM2 5 on Evening Peak Flow in Asthmatics Lagged
Two to Five Days 6-17
6-6 Effect of 50 //g/m3 PM10 on Cough in Asthmatics Lagged Zero or One Day .... 6-17
6-7 Effect of 50 //g/m3 PM10 on Cough in Asthmatics Lagged Two to Five Days ... 6-18
6-8 Effect of 50 //g/m3 PM10 on Phlegm in Asthmatics Lagged Zero or One Day ... 6-18
6-9 Effect of 50 //g/m3 PM10 on Phlegm in Asthmatics Lagged Two to Five Days ... 6-19
6-10 Effect of 50 //g/m3 PM10 on Difficulty in Breathing in Asthmatics
Lagged Zero or One Day 6-19
6-11 Effect of 50 //g/m3 PM10 on Difficulty in Breathing in Asthmatics
Lagged Two to Five Days 6-20
6-12 Effect of 50 //g/m3 PM10 on Bronchodilator Use in Asthmatics Lagged
Zero or One Day 6-20
6-13 Effect of 50 //g/m3 PM10 on Bronchodilator Use in Asthmatics Lagged
Two to Five Days 6-20
6-14 Effect of 50 //g/m3 PM10 on Peak Flow in Non-Asthmatics Lagged
Zero or One Day 6-24
October 1999 II-xiii DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
6-15 Effect of 50 //g/m3 PM10 on Cough in Non-Asthmatics Lagged Zero or
One Day 6-25
6-16 Effect of 50 //g/m3 PM10 on Lower Respiratory Illness in Non-Asthmatics
Lagged Zero or One Day 6-25
6-17 Effect of 50 //g/m3 PM10 on Upper Respiratory Illness in Non-Asthmatics
Lagged Zero or One Day 6-26
6-18 Other Asthmatic Panel Studies 6-27
6-19 Other Non-Asthmatic Panel Studies 6-29
6-20 Comparable Respiratory Hospital Admissions PM10 Studies 6-52
6-21 Comparable Respiratory Hospital Admissions SO4 studies 6-53
6-22 Synthesis of Comparable Time-Series Hospital Admissions Studies'
Estimates of Relative Risk Due to PM10 Exposure 6-54
6-23 Synthesis of Comparable Time-Series Hospital Admissions Studies'
Estimates of Relative Risk Due to SO4= Exposure 6-54
6-24 Summaries of Recently Published Single-City PM Time-Series Studies 6-74
6-25 Relative Risk of Mortality from Contributing Non-Malignant Respiratory
Causes, for 30 Days Per Year with PM10 > 100 //g/m3 6-109
6-26 Relative Risk of Mortality from Contributing Non-Malignant Respiratory
Causes, by Sex and Air Pollutant, with Alternative Covariate Model 6-110
6-27 Relative Risk of Mortality From all Non-External Causes, by Sex and
Air Pollutant, for an Alternative Covariate Model 6-111
6-28 Relative Risk of Mortality From Cardiopulmonary Causes, by Sex and
Air Pollutant, for an Alternative Covariate Model 6-111
6-29 Relative Risk of Mortality from Lung Cancer, by Sex and Air Pollutant,
for an Alternative Covariate Model 6-112
October 1999 II-xiv DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
6-30 Relative Risk of Lung Cancer Incidence in Males, by Air Pollutant,
for Adventist Health Study 6-115
6-31 Relative Risk of Lung Cancer Incidence in Females, by Air Pollutant,
for Adventist Health Study 6-116
6-32 Relative Risk of Total Mortality in Three Prospective Cohort Studies,
by Sex and Smoking Status 6-118
6-33 Relative Risk of Cardiopulmonary Mortality in Three Prospective
Cohort Studies, by Sex and Smoking Status 6-119
6-34 Relative Risk of Lung Cancer Mortality in Three Prospective Cohort
Studies, by Sex and Smoking Status 6-120
6-35 Comparison of Estimated Relative Risks for All-Cause Mortality in
Six U.S. Cities Associated with the Reported Inter-City Range of
Concentrations of Various PM Metrics 6-124
6-36 Comparison of Reported SO4= and PM25 Relative Risks for Various
Mortality Causes in the ACS Study 6-124
6-37 Comparison of Total Mortality Relative Risk Estimates and t-Statistics
for PM Components in Three Prospective Cohort Studies 6-128
6-38 Comparison of Cardiopulmonary Mortality Relative Risk Estimates and
t-Statistics for PM Components in Three Prospective Cohort Studies 6-129
6-39 Recent PM Studies of Pulmonary Function Tests or Acute Respiratory
Symptoms in School-Age Children, Generally Using Panel Studies 6-149
6-40 Recent PM Studies of Emergency Department Visits, Hospital Admissions,
or Doctor's Visits in Children, Attributable to Short-Term PM Exposure 6-151
6-41 Neonatal, Infant, and Child Mortality Attributable to Short-Term
PM Exposure 6-152
6-42 Recent PM Studies of Pulmonary Function Tests or Respiratory Symptoms
in School-Age Children Attributable to Long-Term PM Exposure 6-153
October 1999 II-xv DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
6-43 Other Neonatal and Infant Effects Attributable to Longer Term
PM Exposure 6-155
6-44 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-161
6-45 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-162
6-46 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-164
6-47 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-165
6-48 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-166
6-49 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-167
6-50 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-168
6-51 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-169
6-52 Sensitivity of PM Relative Risk Estimate to Co-pollutant Models 6-171
6-53 Sensitivity of PM Relative Risk Estimate to Co-pollutants Models 6-173
6-54 Sensitivity of PM Relative Risk Estimate for Total Mortality to
Co-pollutant Models 6-174
6-55 Sensitivity of PM Relative Risk Estimate for Infant Mortality to
Co-pollutant Models 6-177
6-56 Effects of Including One or More Gaseous Co-pollutants on PM25
Relative Risk Estimates for Hospital Admissions and Mortality in
Toronto, Santa Clara County, and Southwest Mexico City 6-179
6-57 Conclusions About Alternative Hypotheses That May Affect the Synthesis
of Epidemiology Studies 6-213
7-1 Overview of Respiratory Tract Particle Clearance and Translocation
Mechanisms 7-23
7-2 Respiratory Effects of Particulate Matter 7-54
October 1999 II-xvi DRAFT-DO NOT QUOTE OR CITE
-------
List of Tables
(cont'd)
Number Page
7-3 Effects of Particulate Matter Ex Vivo 7-67
7-4 Immunological Effects of Particulate Matter 7-82
7-5 Cardiovascular Effects and Other Systemic Effects of Particulate Matter 7-89
7-6 Respiratory and Cardiovascular Effects of Mixtures 7-92
7-7 Numbers and Surface Areas of Monodisperse Particles of Unit Density
of Different Sizes at a Mass Concentration of 10 //g/m3 7-102
8-1 Comparison of Physical and Chemical Properties of Ambient Particles:
Fine Mode (Nuclei Mode plus Accumulation Mode) and Coarse Mode 8-2
8-2 Constituents and Major Sources of Atmospheric Particles 8-10
8-3 Model Simulation Considering Regional Deposition for Fine and Coarse
PM10 for Mass and Number 8-17
8-4 Fine and Coarse Mode: Exposure, Deposition, Epidemiology, and
Biological Effects 8-22
8-5 Relative Risk of Total Mortality in Three Prospective Cohort Studies,
by Sex and Smoking Status 8-31
8-6 Relative Risk of Cardiopulmonary Mortality in Three Prospective
Cohort Studies, by Sex and Smoking Status 8-32
8-7 Relative Risk of Lung Cancer Mortality in Three Prospective Cohort
Studies, by Sex and Smoking Status 8-33
October 1999 II-xvii DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
Number Page
6-1 Relationship between the effect estimates from Katsouyanni et al. (1997)
and average temperature 6-72
6-2 Frequency distribution of the lag day for which PM RRs were computed
in 33 studies 6-84
6-3 Relationship of ambient PM25 to relative risk of mortality and to deviations
of FEVj from its expected (standard) value in the Harvard Six Cities study .... 6-158
6-4 Relationship of two health endpoints to PM25 in the Six Cities study 6-158
6-5 Relative risk of total mortality from PM25 in southwest Mexico City as a
function of PM lag or moving average, with 95% confidence limits 6-176
6-6 Relative risk of infant mortality from PM25 in southwest Mexico City as a
function of PM lag or moving average, with 95% confidence limits 6-178
6-7 Ambient PM concentration isopleths and monitoring sites in a hypothetical
urban area 6-184
7-1 Total deposition data in humans as a function of particle size 7-9
7-2 Major physical clearance pathways for particles deposited in the
extrathoracic region and tracheobronchial tree 7-24
7-3 Diagram of known and suspected clearance pathways for poorly soluble
particles depositing in the alveolar region 7-26
8-1 Volume size distribution, measured in traffic, showing fine-mode and
coarse-mode particles and the nuclei and accumulation modes within
the fine-particle mode 8-4
8-2 Human respiratory tract PM deposition fraction and PM10 or PM2 5 sampler
collection versus mass median aerodynamic diameter with geometric
standard deviation, og = 1.8 8-16
8-3 Distribution of coarse, accumulation, and nuclei or ultrafine mode
particles by three characteristics: volume, surface area, and number
for the grand average continental size distribution 8-19
8-4 Fort Ord, CA, dust episode: fine and coarse particles 8-42
October 1999 II-xix DRAFT-DO NOT QUOTE OR CITE
-------
List of Figures
(cont'd)
Number Page
8-5 Spokane dust storm episodes 8-43
8-6 During normal wind conditions, accumulation mode particles dominate
PM25 8-45
October 1999 II-xx DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
CHAPTER 6. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS FROM
AMBIENT PARTICULATE MATTER
Principal Authors
Dr. Vic Hasselblad—29 Autumn Woods Drive, Durham, NC 27713
Dr. Kazuhiko Ito—New York University Medical Center, Institute of Environmental Medicine,
Long Meadow Road, Tuxedo, NY 10987
Dr. Patrick Kinney, Columbia University, 60 Haven Avenue, B-l, Room 119
New York, NY 10032
Dr. Dennis J. Kotchmar—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Allan Marcus—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. George Thurston—New York University Medical Center, Institute of Environmental
Medicine, Long Meadow Road, Tuxedo, NY 10987
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 K1A OL2
Dr. Raymond Carroll—Texas A & M University, Department of Statistics, College Station, TX
77843-3143
Dr. Robert Chapman—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Steven Colome—Integrated Environmental Services, 5319 University Drive, #430
Irvine, CA 92612
Dr. Ralph Delfino—University of California at Irvine, Epidemiology Division, Department of
Medicine, University of California at Irvine, Irvine, CA 92717
October 1999 II-xxi DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Dr. Douglas Dockery—Harvard School of Public Health, 665 Huntington Avenue, 1-1414
Boston, MA 02115
Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Paul Humphreys—University of Virginia, Department of Philosophy
Charlottesville, VA 22901
Dr. Fred Lipfert—23 Carll Court, Northport, NY 11768
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. David Mage—National Center for Environmental Assessment (MD-52), 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. Isabelle Romieu—Centers for Disease Control (CDC), 4770 Bufford Hwy, NE
Atlanta, GA 30341
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, N.C. 27695
Dr. Duncan Thomas—University of Southern California, Preventative Medicine Department
1540 Alcazar Street, CH-220, Los Angeles, CA 90033-9987
Dr. Clarice Weinberg—National Institute of Environmental Health Sciences, P.O. Box 12233
Research Triangle Park, NC 27709
October 1999 II-xxii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Dr. William Wilson—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Vanessa Vu—Office of Research and Development, U.S. Environmental Protection Agency
(8601), Waterside Mall, 401 M St. S.W., Washington, DC 20460
CHAPTER 7. DOSIMETRYAND TOXICOLOGY OF PARTICULATE MATTER
Principal Authors
Dr. Lung Chi Chen—New York University School of Medicine, Nelson Institute of
Environmetnal Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Terry Gordon—New York University Medical Center, Department of Environmental
Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Dr. Richard Schlesinger—New York University School of Medicine, Department of
Environmental Medicine, 57 Old Forge Road, Tuxedo, NY 10987
Contributors and Reviewers
Dr. Dan Costa—National Health and Environmental Effects Research Laboratory (MD-82)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Robert Devlin—National Health and Environmental Effects Research Laboratory (MD-58)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Kevin Dreher—National Health and Environmental Effects Research Laboratory (MD-82)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Andrew Ohio—National Health and Environmental Effects Research Laboratory (MD-58D)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Judith Graham—National Exposure Research Laboratory (MD-75)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
October 1999 II-xxiii DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
Dr. Chong Kim—National Health and Environmental Effects Research Laboratory (MD-58B)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Hillel Koren—National Health and Environmental Effects Research Laboratory (MD-58A)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Ted Martonen—National Health and Environmental Effects Research Laboratory (MD-74)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Jim Samet—National Health and Environmental Effects Research Laboratory (MD-58)
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Watkinson—National Health and Environmental Effects Research Laboratory
(MD-82), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William Bennett—University of North Carolina at Chapel Hill, Campus Box 7310
Chapel Hill, NC 37599
Dr. Mark Frampton—University of Rochester, 601 Elmwood Avenue, Box 692, Rochester, NY
14642
Dr. John Godleski—421 Conant Road, Weston, MA 02493
Dr. Gunter Oberdorster—University of Rochester, Department of Environmental Medicine
Rochester, NY 14642
Dr. Kent Pinkerton—University of California, ITEH, One Shields Avenue, Davis, CA 95616
Dr. Peter J.A. Rombout—National Institute of Public Health and Environmental Hygiene,
Department of Inhalation Toxicology, P.O. Box 1, NL-3720 BA Bilthoven, The Netherlands
Dr. Vanessa Vu—Office of Research and Development, U.S. Environmental Protection Agency
(8601), Waterside Mall, 401 M St. S.W., Washington, DC 20460
October 1999 II-xxiv DRAFT-DO NOT QUOTE OR CITE
-------
Authors, Contributors, and Reviewers
(cont'd)
CHAPTERS. INTEGRATIVE SYNTHESIS OF KEY POINTS: PM EXPOSURE,
DOSIMETRY, AND HEALTH RISKS
Principal Authors
Dr. Lawrence J. Folinsbee—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Dennis J. Kotchmar—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. David Mage—National Center for Environmental Assessment (MD-52), U.S. Environmental
Protection Agency, Research Triangle Park, NC 27711
Dr. Allan Marcus—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Joseph P. Pinto—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. William E. Wilson—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Contributors and Reviewers
Dr. Robert Chapman—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. William Ewald—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lester D. Grant—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. James McGrath—National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Vanessa Vu—Office of Research and Development, U.S. Environmental Protection Agency
(8601), Waterside Mall, 401 M St. S.W., Washington, DC 20460
October 1999 II-xxv DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
Scientific Staff
Dr. Lester D. Grant—Director, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. Randy Brady—Deputy, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Dr. Lawrence J. Folinsbee—Health Coordinator, Chief, Environmental Media Assessment
Group, National Center for Environmental Assessment (MD-52), U.S. Environmental Protection
Agency, Research Triangle Park, NC 27711
Dr. William E. Wilson—Air Quality Coordinator, Physical Scientist, National Center for
Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle
Park,NC 27711
Dr. Dennis J. Kotchmar—Project Manager, Medical Officer, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Dr. Robert Chapman—Technical Consultant, Medical Officer, National Center for
Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle
Park,NC 27711
Ms. Beverly Comfort—Health Scientist, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. William Ewald—Technical Project Officer, Health Scientist, National Center for
Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle
Park,NC 27711
Dr. David Mage—Technical Project Officer, Physical Scientist, National Center for
Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle
Park,NC 27711
Dr. Allan Marcus—Technical Project Officer, Statistician, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
October 1999 II-xxvii DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Dr. James McGrath—Technical Project Officer, Visiting Senior Health Scientist, National
Center for Environmental Assessment (MD-52), U.S. Environmental Protection Agency,
Research Triangle Park, NC 27711
Dr. Joseph P. Pinto—Technical Project Officer, Physical Scientist, National Center for
Environmental Assessment, U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Technical Support Staff
Mr. Douglas B. Fennell—Technical Information Specialist, National Center for Environmental
Assessment (MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC
27711
Ms. Emily R. Lee—Management Analyst, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Diane H. Ray—Program Analyst, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Ms. Eleanor Speh—Office Manager, Environmental Media Assessment Branch, National Center
for Environmental Assessment (MD-52), U.S. Environmental Protection Agency, Research
Triangle Park, NC 27711
Ms. Donna Wicker—Administrative Officer, National Center for Environmental Assessment
(MD-52), U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Mr. Richard Wilson—Clerk, National Center for Environmental Assessment (MD-52),
U.S. Environmental Protection Agency, Research Triangle Park, NC 27711
Document Production Staff
Mr. John R. Barton—Document Processing Coordinator
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Ms. Yvonne Harrison—Word Processor
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
October 1999 II-xxviii DRAFT-DO NOT QUOTE OR CITE
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
PROJECT TEAM FOR DEVELOPMENT OF AIR QUALITY CRITERIA
FOR PARTICULATE MATTER
(cont'd)
Ms. Bettye Kirkland—Word Processor
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Mr. David E. Leonhard—Graphic Artist
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Ms. Carolyn T. Perry—Word Processor
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Ms. Veda E. Williams—Graphic Artist
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Technical Reference Staff
Mr. R. David Belton—Reference Specialist
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Mr. John Bennett—Technical Information Specialist
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Mr. William Hardman—Reference Retrieval and Database Entry Clerk
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Ms. Sandra L. Hughey—Technical Information Specialist
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
Mr. Jian Ping Yu—Reference Retrieval and Database Entry Clerk
OAO Corporation, Chapel Hill-Nelson Highway, Beta Building, Suite 210, Durham, NC 27713
October 1999 II-xxix DRAFT-DO NOT QUOTE OR CITE
-------
i 6. EPIDEMIOLOGY OF HUMAN HEALTH EFFECTS
2 FROM AMBIENT PARTICULATE MATTER
3
4
5 6.1 INTRODUCTION
6 Epidemiology studies linking community ambient PM concentrations to adverse health
7 effects played an important role in the 1996 PM Air Quality Criteria Document (PM AQCD), and
8 continue to play an important role. Those studies are indicative of measurable excesses in
9 pulmonary function decrements, respiratory symptoms, hospital and emergency department
10 admissions, and mortality being associated with ambient levels of PM25, PM10, and other
11 indicators of PM exposure. The more recent epidemiologic studies reviewed in this chapter
12 generally identify more cities and extend the earlier findings. Therefore, the main emphasis in
13 this chapter has shifted from a detailed discussion of the findings in the individual studies (as
14 contained in the 1996 PM AQCD) to a greater emphasis here on integrating and interpreting the
15 findings in the body of evidence provided by the newer studies, as well as those reviewed in
16 1996.
17 Several hypotheses may be proposed, based on the earlier evidence:
18 Hypothesis 0: Exposure to ambient PM at current levels cannot cause adverse health
19 effects in susceptible sub-populations, even in the presence of other environmental factors such
20 as weather conditions or the presence of other air pollutants;
21 Hypothesis 1: Exposure to ambient PM or some component at current levels is associated
22 with adverse health effects in some susceptible sub-population.
23 Hypothesis 1 has many alternative forms. Those of greatest interest in this chapter concern
24 the circumstances under which the adverse health effects may be manifested, typically the
25 occurrence of adverse health effects in association with either a specific form or component of
26 ambient PM, or with specific environmental co-factors (such as weather or co-pollutants) along
27 with PM exposure. These circumstances can markedly affect the approaches taken to the
28 synthesis of epidemiology studies from different sites. Table 6-1 illustrates several of the
29 possible variants of Hypothesis 1:
30
October 1999 6-1 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-1. ALTERNATIVE VERSIONS OF HYPOTHESIS 1 THAT MAY
AFFECT THE SYNTHESIS OF EPIDEMIOLOGY STUDIES
Alternative Hypotheses
Adverse Health Effects
Depend Only on Ambient
PM, Independent of
Co-Factors
Adverse Health Effects
Depend on Ambient PM
Concentrations as Well as
Co-Factors
Adverse Health Effects
Depend Only on Ambient
PM Size Range and
Concentration
Adverse Health Effects
Depend on PM With Specific
Physical Properties or
Composition, and on
Concentration
Adverse health effects from
ambient PM at a given
concentration are the same in
all sites with the same PM
size range
Adverse health effects from
PM are different at sites
where PM has different
physical properties or
composition with the same
PM size range and
concentration.
Adverse health effects from
PM are different in sites
where PM has different
co-factors with the same PM
size range and concentration.
Adverse health effects from
PM are different in sites
where PM has different
physical properties,
composition, or co-factors,
even at the same PM size
range concentrations.
1 If row 1, column 1 in Table 6-1 occurs, it is feasible to combine information from all
2 epidemiology studies of a common design concerning a specific health effect of PM at all sites.
3 The estimated PM effect depends only on the size range in the study, and is assumed to be
4 independent of the PM chemical components or physical characteristics within that size range,
5 and independent of the levels of co-pollutants or the occurrence of certain weather variable
6 ranges with which the PM levels are associated. The implicit assumption of a single
7 quantitatively similar effect of PM of any composition at all sites appears to be present in a
8 number of published research syntheses or meta-analyses. This is important in evaluating the
9 consistency of the quantitative effects (concentration-response) across different studies or
10 different sites. While some studies may show a positive and statistically significant adverse
11 health effect of the PM index being evaluated, others may show statistically non-significant
12 effects. The different findings may suggest a substantive reason for grouping sites by PM
13 components or by environmental co-factors. Statistically positive or non-positive PM health
14 effects may be attributable to the presence or absence of a particular PM component, or to the
October 1999
6-2
DRAFT-DO NOT QUOTE OR CITE
-------
1 presence or absence of a specific environmental co-factor, and may help to define a quantitative
2 relationship to the component or the environmental co-factor.
3 If row 1, column 2 applies, then it may be informative to group sites with some
4 commonality, such as the presence of significant specific sources, specific mixtures of
5 co-pollutants, common weather patterns, or even common demographic factors that affect
6 exposure and toxicity. Examples referred to below include location ('eastern' U.S. versus
7 'western' U.S., 'central-eastern Europe' versus 'western' Europe) and source types (gasoline
8 combustion mobile sources, diesel combustion sources, fossil fuel power plants, metal smelters
9 or factories, crustal particles, sea salt particles, other organic particle sources, biotoxins, etc.,
10 as discussed in Chapters 5 and 7).
11 Row 2 considers the possibility that site-to-site or study-to-study differences in outcome
12 may depend on properties of the ambient PM itself, independent of or in addition to its mass
13 concentration. These may include chemical composition (acids, metals, oxidants) or physical
14 properties (number concentration or surface area) of the PM, or even of gaseous pollutants
15 adhering to particles. In the row 2, column 2 case, it might only be appropriate to evaluate sites
16 with common distributions of air pollution, weather, and population.
17 The epidemiology studies presented here should be regarded in combination with the
18 ambient concentration studies in Chapter 4, the human exposure studies in Chapter 5, and the
19 toxicology studies in Chapter 7. The contribution of the epidemiology studies is to identify
20 whether specific adverse health effects occur in susceptible human populations that are likely to
21 have been exposed to relevant levels of ambient PM. Chapter 8 provides a more detailed causal
22 synthesis indicating (a) that personal exposure to ambient PM does occur and (b) that the
23 biological effects may be similar to those observed in free-living human populations.
24
25 Types of Epidemiology Studies Reviewed
26 A concise definition of the various types of epidemiology studies used here is given in the
27 1996 PM AQCD (U.S. Environmental Protection Agency, 1996) and in most epidemiology texts;
28 it is not repeated here. Briefly, the epidemiology studies are divided into morbidity studies and
29 mortality studies. The morbidity studies include a wide range of health endpoints, such as
30 changes in pulmonary function tests (PFT), reports of respiratory symptoms, self-medication in
31 asthmatics, medical visits, low birthweight infants, and hospitalization. Mortality studies from
October 1999 6-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 many causes have provided the most unambiguous evidence of a clearly adverse endpoint, and
2 are sufficiently numerous to be discussed in detail.
3 The most commonly used study designs are cross-sectional, panel, time series, and
4 prospective cohort studies. Cross-sectional studies evaluate subjects at a 'point' in time, where
5 measurements of health status, pollution exposure, and individual covariates are observed
6 simultaneously. When summary statistics are used to compare either health outcomes or
7 exposure indices for different populations, the study is called an ecological or semi-ecological
8 study. Panel studies follow the health outcomes of the same individuals over several time points.
9 Prospective cohort studies follow a group of initially recruited individuals over a long period of
10 time, where some events (such as the date of death of a subject) may be determined individually
11 with great precision. To contrast these study designs, individuals in a prospective cohort study
12 may have their vital status monitored over a long period of time, whereas symptoms may be
13 recorded daily only in the first few months of the study and PFT measurements made initially and
14 after 3, 6, and 12 years of follow-up (as in the Harvard Six Cities cohort). Time-series studies, as
15 referred to in this document, usually involve aggregate-level outcomes, such as the daily number
16 of deaths in a community, or the daily count of hospital admissions or emergency department
17 visits, over a large number of days. Time-series studies require different analytical strategies
18 than the other study designs. A few recent analyses have examined case-control designs.
19 Studies with individual-level outcome data, covariates, and PM exposure indices would be
20 preferred, whatever the design. Individual-level exposure data are the most commonly missing
21 component. Community-level air pollution concentrations are usually substituted as surrogate
22 indices of population exposures, and the evidence presented in Chapter 5 suggests that exposure
23 to PM10 and PM25 of ambient origin, and to sulfates, is adequately characterized by
24 community-level measurements. Substantial efforts to develop individual long-term exposure
25 indices for a prospective cohort study have been reported by Abbey et al. (1991, 1995, 1999;
26 Beeson et al., 1998) for the California Adventist Health Study of Smog (AHSMOG).
27 The strengths and weaknesses of the study designs have been discussed in some detail in
28 (U.S. Environmental Protection Agency, 1996). In general, prospective cohort and panel studies
29 would be preferred because of the individual-level information. However, time series studies
30 allow valid inferences about short-term responses to changes in environmental factors in exposed
October 1999 6-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 populations, and are considered very informative, provided that the environmental factors (such
2 as co-pollutants) are not too highly correlated with the PM exposure index.
3
4 Selection of Studies for Detailed Review
5 Numerous papers on PM epidemiology have been published since the 1996 PM AQCD.
6 Those papers selected here as being most clearly relevant to this NAAQS review are described in
7 greater detail in the text of this chapter, and the others are included in tables where appropriate.
8 Some of the criteria for selecting relevant literature for text discussion include consideration as to
9 whether a given study:
10 1. Presents PM indices previously considered: PM10, fine or coarse PM10 fractions;
11 2. Presents analyses with informative new PM indices such as nitrates or ultrafines;
12 3. Presents information on health endpoints not previously considered;
13 4. Presents new information on multiple pollutant analyses;
14 5. Presents new information on long-term effects, mortality displacement;
15 6. Presents new information on health effects of specific PM constituents.
16 The present chapter is organized as follows. After this brief introduction, Section 6.2
17 examines the case for causal inference based on studies of morbidity as a health endpoint.
18 Section 6.3 examines the case for causal inference based on studies of PM effects on mortality,
19 the most clearly adverse effect. Section 6.4 discusses a variety of issues related to the inferences
20 that can be drawn from the studies reviewed in Sections 6.2 and 6.3. Section 6.5 briefly reviews
21 some of the general and methodological issues in inferring causal relationships from the large
22 number of epidemiology studies reviewed here. The overall findings of this Chapter are then
23 summarized in Section 6.6.
24
25
26 6.2 MORBIDITY EFFECTS OF PARTICULATE MATTER EXPOSURE
27 This morbidity section is presented in sub sections, dealing with: (a) short-term PM
28 exposure effects on lung function and respiratory symptoms in asthmatics and non-asthmatics;
29 (b) long-term PM exposure effects on lung function and respiratory symptoms; (c) effects of
30 short-term PM exposure on the incidence of respiratory and other medical visits and hospital
31 admissions, and (d) effects of ambient PM exposure on acute cardiovascular morbidity.
October 1999 6-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 For consistency with the prior PM Criteria Document (U.S. Environmental Protection Agency,
2 1996), the pollutant increments utilized here to report Relative Risks (RR's) or Odds Ratios for
3 various health effects are: for PM10, 50 //g/m3; for PM2 5,25 //g/m3; for SO4=, 155 nmoles/m3
4 (=15 //g/m3); and, for H+, 75 nmoles/m3 (=3.6 //g/m3, if as H2SO4).
5
6 6.2.1 Short Term Effects on Lung Function and Respiratory Symptoms
7 In the 1996 PM AQCD, the available respiratory disease studies used a wide variety of
8 designs examining pulmonary function and respiratory symptoms in relation to PM10. The
9 models for analysis varied and the populations included several different subgroups. Pulmonary
10 function studies were suggestive of short term effects resulting from particulate exposure. Peak
11 expiratory flow rates showed decreases in the range of 2 to 5 1/min resulting from an increase of
12 50 //g/m3 in PM10 or its equivalent, with somewhat larger effects in symptomatic groups such as
13 asthmatics. Studies using FEVj or FVC as endpoints showed less consistent effects. The chronic
14 pulmonary function studies were less numerous than the acute studies and the results were
15 inconclusive.
16 The available acute respiratory symptom studies included several different endpoints, but
17 typically presented results for: (1) upper respiratory symptoms, (2) lower respiratory symptoms,
18 or (3) cough. These three respiratory symptom endpoints had similar general patterns of results.
19 The odds ratios were generally positive, the 95% confidence intervals for about half of the
20 studies were statistically significant (i.e., the lower bound exceeded 1.0). As part of the Six
21 Cities studies, three analyses done for different time periods suggested a chronic effect of PM
22 exposure on respiratory disease. Chronic cough, chest illness, and bronchitis showed positive
23 associations with PM for the earlier surveys. One study was strongly suggestive of an effect on
24 bronchitis from acidic particles or from other PM.
25 The earlier studies of morbidity health outcomes of PM10 exposure on asthmatics were
26 limited in terms of conclusions that could be drawn because of the few available studies on
27 asthmatic subjects. Lebowitz et al. (1987) reported a relationship with TSP exposure and
28 productive cough in a panel of 22 asthmatics but not for peak flow or wheeze. Pope et al. (1991)
29 studied respiratory symptoms in two panels of asthmatics in the Utah Valley. The 34 asthmatic
30 school children panel yielded estimated odd ratios of 1.28 (1.06, 1.56) for lower respiratory
31 illness and the second panel of 21 subjects aged 8 to 72 for LRI of 1.01 (0.81, 1.27) for exposure
October 1999 6-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 to PM10. Ostro et al. (1991) reported no association for PM25 exposure in a panel of 207 adult
2 asthmatics in Denver; but, for a panel of 83 asthmatic children age 7 to 12 in central
3 Los Angeles, reported a relationship of shortness of breath to O3 and PM10 but could not separate
4 the effect of the two pollutants. These few studies did not indicate a consistent relationship for
5 PM10 exposure and health outcome in asthmatics.
6 Several new studies of short term PM exposure effects on lung function and respiratory
7 symptoms have been published since early 1996. Most of these studies followed a panel of
8 subjects over one or more periods. Daily lung function and/or respiratory symptoms were
9 associated with changes in ambient PM10 and/or PM2 5. Lung function was usually measured
10 daily with many studies including forced expiratory volume (FEV), forced vital capacity (FVC)
11 and peak expiratory flow rate (PEF). Some analyses included both morning and afternoon
12 measurements. A large variety of respiratory symptoms were measured, including cough,
13 phlegm, difficulty breathing, wheeze, and bronchodilator use. Finally, several measures of
14 particulate matter were used including PM10, PM2 5, TSP, British Smoke (BS), and sulfate
15 fraction of ambient PM.
16 These various studies are discussed in the following text and tables. Studies providing
17 quantitative information can often add more understanding of possible relationships between PM
18 exposure and health outcomes. Data on physical and chemical aspects of ambient particulate
19 levels, especially for PM10 and PM25 and smaller size fractions are a focus of discussion.
20 Additionally, new studies are discussed that present potential new insights or that examine health
21 outcome effects and/or exposure measures not studied as much in the past.
22 This section is organized around discussion of PM effects on lung function, respiratory
23 symptoms, and related pulmonary outcomes. The acute studies are split into two groups: panels
24 of asthmatics and panels of non-asthmatics. The lung function studies are not split from the
25 respiratory symptom studies because many of the studies included both endpoints. To facilitate a
26 quantitative synthesis of outcomes, studies are placed into homogeneous groupings and their
27 results are presented as shown. Note that some unique study results need to be discussed in that
28 they may examine an aspect which no other study has, such as number of particles or 1-h and 8-h
29 PM averages.
30 The following section contains a quantitative synthesis of several studies. The combination
31 of differing endpoints and differing lag-times in the analyses resulted in a large number of
October 1999 6-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 potential analyses for each study. Authors may have tended to choose and report results for those
2 models that gave the strongest effects. In an attempt to present these studies in an organized
3 manner, specific analyses were selected based on the following criteria (which also may
4 minimize potential bias of reported results):
5 (1) Peak flow was used as the primary lung function measurement of interest. While FEVj
6 would be a good measure, peak flow is most often measured in these panel studies.
7 (2) Only cough, phlegm, difficulty breathing, wheeze, and bronchodilator use were summarized
8 as measures of respiratory symptoms.
9 (3) Summaries focused on those studies presenting results in terms of PM10 and PM25 and
10 smaller PM.
11 (4) The analyses were also restricted to include a short term lag (zero or one day) a longer term
12 lag (two to five day), and a moving average analysis. If both zero and one day lag analyses
13 were presented, the zero day lag analysis was selected for all but the AM peak flow
14 measurements, and here for longer term lags, the measure which came closest to being an
15 average of two to five days was selected.
16 (5) Studies were included only if they modeled individual responses. A few studies modeled
17 group rates and this flaw is also known as the "panel data problem" (Neuhaus and
18 Kalbfieisch, 1998).
19 Whenever three or more studies of a similar endpoint were available, the results were
20 combined using a random effects model (Hedges and Olkin, 1985). Tests for homogeneity are
21 also reported.
22 A few of the analyses included more than one pollutant in the model at the same time.
23 However, while the number of such studies was too small to allow for any meaningful
24 conclusions, these results are discussed due to the importance of considering co-pollutants. The
25 summary from this section reflects the above organization.
26
27 6.2.1.1 Short Term Effects on Lung Function and Respiratory Symptoms In Asthmatics
28 Ostro et al. (1995) followed 87 African-American children, ages 7-12 years with confirmed
29 asthma for at least 6 weeks. Four subjects were dropped because of concerns about the accuracy
30 of responses, leaving 83 subjects. Most subjects lived in central and south-central Los Angeles,
31 CA. Analyses were done using "daily reporting of respiratory symptoms including cough,
October 1999 6-8 DRAFT-DO NOT QUOTE OR CITE
-------
1 shortness of breath, and wheeze" as the dependent variables and the pollutants of TSP, sulfates,
2 nitrates, ozone, SO2, NO2, and PM10 as the independent variables. PM10 from 3 downtown LA
3 sites had a mean of 56 //g/m3 (range 19 to 101 //g/m3). General logistic regression models were
4 used with generalized estimating equation (GEE) corrections for autocorrelation. Significant
5 relationships were found between shortness of breath and PM10 or ozone, with symptoms
6 estimated to increase about 9% per each 10 //g/m3 increase in PM10. The authors examined other
7 symptoms and found no significant associations, but results were not reported.
8 Peters et al. (1997a) studied 89 children, aged 6 to 14 years, with asthma in Sokolov, Czech
9 Republic. The subjects kept diaries and measured peak flow for seven months during the winter
10 of 1991-92. Aerometric measurements included PM10, SO2, TSP, sulfate, and particle strong
11 acidity. PM10 was measured at one central site, with a mean of 55 //g/m3 and a max of
12 171 //g/m3. The analysis was done using linear regression for the pulmonary function data and
13 logistic regression for binary outcomes. First order autocorrelations were observed and corrected
14 for using polynomial distributed lag structures. Only weak associations were reported between
15 the measures of particulate pollution and lung function or respiratory symptoms. Although the
16 magnitudes of effect were still weak, there were associations for both morning and evening PEF.
17 This was a wintertime study in an area with relatively high SO2 (median 46 //g/m3) and so, not
18 unexpectedly, in comparison to PM10, sulfate effects on PEF and symptoms were similar and
19 slightly larger for some models.
20 In a further analysis, Peters et al. (1997b) compared children with mild asthma who were
21 either taking p-agonist medications (31 subjects) or not taking them (51 subjects) during the
22 winter of 1991-92 in Sokolov, Czech Republic. Those taking such medications had more severe
23 asthma than those not taking them. For the relationship between PEF and 5-day mean sulfate
24 (interquartile range of 6.5 //g/m3), effects were larger for the medicated subjects (-5.62, 95% CI
25 -9.93 to -1.30 L/min) as compared with unmedicated subjects (-1.35, 95% CI -3.69 to
26 0.99 L/min). Effects of the same day sulfate were small and non significant.
27 Gielen et al. (1997) studied 61 children aged, 7 to 13 years, living in Amsterdam, The
28 Netherlands during the summer of 1995. Seventy-seven percent of the children were taking
29 asthma medication and the others were hospitalized for respiratory problems. Peak flow
30 measurements were taken twice daily and respiratory symptoms were recorded by the parents in a
31 diary. PM10 was measured at one city site with a mean of 30.5 //g/m3. Associations of air
October 1999 6-9 DRAFT-DO NOT QUOTE OR CITE
-------
1 pollution were evaluated using time series analyses. The analyses adjusted for pollen counts,
2 time trend, and day of week. The studies found relationships with ozone and PM10. Stronger
3 associations were found for black smoke (BS) than for PM10, in relation to PEF, symptoms and
4 bronchodilator use. The authors hypothesized that BS may be a better surrogate for fine particles
5 emitted by diesel engines or for other chemicals that may be the causal components in PM.
6 Romieu et al. (1997) studied 65 children with mild asthma aged 5-13 years living in the
7 southwest area of Mexico City, Mexico. During the study period, maximum daily 1-h ozone
8 ranged from 40 to 390 ppb (mean 196 ppb SD = 78 ppb) and PM10 daily average ranged from
9 12 to 126 //g/m3 (x ="54.2 //g/m3). Pollutant measurement were made at a local site. Morning
10 and evening peak flow measurements were made and respiratory symptoms were recorded by the
11 parents in a daily diary. Peak flow measurements were standardized for each person and a model
12 was fitted using GEE methods. The model included terms for minimum temperature.
13 An autoregressive logistic regression model using GEE methods was used to analyze the
14 presence of respiratory symptoms. The strongest relationships were found between ozone and
15 the respiratory symptoms.
16 Peters et al. (1996) studied two mild/moderate asthma panels in Erfurt and Weiman,
17 Germany and in Sokolov, Czech Republic from September 1990 thru June 1992 for health
18 outcomes in relation to pollutant exposure. During that period, TSP was measured at 3 central
19 sites. From January 10 through June 1992, PM10 was also measured. The panels consisted of
20 102 adults and 155 children aged 7 to 15 years. Mean levels of PM10 in Erfurt, for the winter
21 1991-92 was 64 //g/m3. The panelists recorded daily symptom, medication intake and PEF.
22 A linear regression analysis was conducted. The dominate air pollutant was SO2. An increase of
23 52 //g/m3 PM10 was associated with a 0.43% decrease in evening PEF for children with asthma.
24 Because of the small observed effects and the pollutants being highly correlated, separation of
25 contributions of individual air pollutants was difficult.
26 Three studies attempted to relate lung function or respiratory symptoms to particles smaller
27 than PM2 5. Peters et al. (1997c) studied 27 non-smoking adult asthmatics living in Erfurt,
28 Germany during the winter season 1991-92. The study measured particulate fractions over a
29 range of sizes from ultrafine (<0.1 //m in diameter) to fine (0.1 to 2.5 //m), including PM10 at one
30 site. A 5-day mean level of 60 //g/m3 PM10 was observed. Morning and evening peak flow were
31 measured and a diary was used to record the presence of cough. An autoregressive model was
October 1999 6-10 DRAFT-DO NOT QUOTE OR CITE
-------
1 used to analyze the deviations in the individual peak flow values. The model included terms for
2 time trend, temperature, humidity, and wind speed and direction. The strongest effects on peak
3 flow were found with the ultrafine particles although the confidence intervals were significantly
4 overlapping. The symptom information was analyzed using multiple logistic regression analysis.
5 The authors reported no association between PM and phlegm or dyspnea, but estimates and
6 standard errors were not given. The Peters et al. (1997c) study is unique for two reasons:
7 (1) they studied the size distribution of particles in the range 0.01 to 2.5 //m, and (2) they
8 examined the number of particles. They report that the health effects of 5 day means of the
9 number of ultrafine particles were larger than those of the mass of fine particles, and that the size
10 distribution of ambient particles helps elucidate the properties of ambient aerosol responses for
11 health effects.
12 Pekkanen et al. (1997) studied 39 asthmatic children, aged 7-12 years, living in Kuopio,
13 Finland for 57 days in early 1994. Changes in peak flow measurements were analyzed using a
14 linear first-order autoregressive model. The study measured particulate fractions at a local site
15 over a range of sizes from ultrafine to fine, including PM10. Particulate measurements included
16 both particles <0.03 //m in diameter, 0.03 to 0.1, 0.1 to 0.32, and 0.32 to 1.0 //m in diameter.
17 The number of particles was also determined by size. The mean PM10 level was 18 //g/m3.
18 Decrements in peak flow were found to be related with all measures of particulate matter after
19 adjusting for minimum temperature. The results were quite variable across zero to 3 day lags,
20 showing no particular pattern across size. Results for two day lags tended to be larger than other
21 lags. Similar results were found for evening PEFR except that the one day lags tended to show a
22 stronger relationship. In contrast to the findings of Peters et al. (1997c) discussed above, PM10
23 was more consistently associated with PEF across the different lags, and gave the only models
24 that were statistically significant (with the exception of BS). The authors note that the different
25 particle size fractions were highly intercorrelated and that future studies should aim at obtaining
26 data where these intercorrelations are lower.
27 Timonen and Pekkanen (1997) studied 74 asthmatic children and 95 children with dry
28 cough ages 7-12 in Kuopio, Finland, during the winter of 1994. PM10 levels were a mean of
29 18 //g/m3, 25 to 75 percentile, 10-23 //g/m3. Linear regression analyses with autoregressive
30 parameters for PEF data and logistic regression models for binary symptom data were performed.
October 1999 6-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 A significant association for morning PEF and PM10 was found, but not for evening PEF.
2 No clear or consistent trends in association between air pollution and symptoms was observed.
3 Tiittanen et al. (1999) studied 49 children with chronic respiratory symptoms (age
4 8-13 years) for six weeks in the spring of 1995. Particulate measurements included both ultrafine
5 particles (<0.1 //m in diameter) and fine particles (0.1 to 1.0 //m in diameter). No consistent
6 differences in effects by particle diameter on morning and evening PEFR were found. Similarly,
7 no consistent differences in effects by particle diameter on the incidence of cough were found.
8 Delfino et al. (1998) examined the relationship of adverse asthma symptoms (bothersome
9 or interfered with daily activities or sleep) to O3 and PM10 in a southern California community in
10 the air inversion zone (1200-2100 ft) with high O3 and low PM (R = 0.3). The region was
11 initially chosen for study to examine effects of high ambient O3 with less co-pollutant
12 confounding from PM. A panel of 25 asthmatics ages 9-17 were followed daily, August through
13 October, 1995 (N=l,759 person-days excluding 1 subject without symptoms). The highest
14 24-hour PM10 mean was only 54 //g/m3, in contrast to the median of 1-hr maximums (56 //g/m3).
15 Longitudinal regression analyses utilized the GEE model controlling for autocorrelation, day of
16 week, outdoor fungi and weather. Asthma symptoms were significantly associated with both
17 outdoor O3 and PM10 in single pollutant- and co-regressions, with 1-hr and 8-hr maximum PM10
18 having larger effects than the 24-hr mean. This aspect of this study reporting particle effects
19 from 1-hr and 8-hr maximum PM10 as compared with the standard metric of 24-hr means makes
20 these results unique. The author notes that particle effects may be determined by factors not
21 entirely dependent on mass and/or 24-hr averages and thus may miss important short-term
22 excursions, during peak exposure periods.
23 Vedal et al. (1998) studied 206 children (aged 6 to 13 years) living in Port Alberni, BC.
24 The authors chose this town of 30,000 on Vancouver Island to maximize the spatial relationship
25 of a central air monitor to subject activities and to minimize the influence of co-pollutants such
26 as ozone, SO2 or acid aerosol, which were low in the region. PM10 levels ranged from 0.2 to
27 159.0 //g/m3 (median 22.1 //g/m3). Major sources of PM in the region were a pulp and paper
28 mill, and residential wood burning. Seventy-five children had physician-diagnosed asthma,
29 57 had an exercise-induced fall in FEV1? 18 children with airway obstruction, and 56 children
30 without any symptoms. Peak flow was measured twice daily and respiratory symptom data were
31 obtained from diaries. An autoregressive model was fitted to the data using GEE methods.
October 1999 6-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 Covariates included temperature, humidity, and precipitation. In general, PM10 was associated
2 with changes in both peak flow and respiratory symptoms. The objective of the study was to
3 compare the acute effects of inhalable particles on PEF and respiratory symptoms in asthmatic
4 versus nonasthmatic children. Even though levels of PM10 were low (only 1.2% of days
5 > 100 //g/m3), PEF was inversely, and cough positively, associated with PM10 among children
6 with diagnosed asthma, but not among the other groups of children.
7 Segala et al. (1998) studied children, aged 7-15 years, living in Paris, France. The study
8 was separated into substudies of 43 mild asthmatics and 43 moderate asthmatics, followed from
9 November 15, 1992 to May 9, 1993. Peak flow was measured three times a day and respiratory
10 symptoms and as-needed-bronchodilator use were recorded daily in a diary by parents. The study
11 measured SO2, NO2, PM13 (instead of PM10) at four stations, and British smoke. A PM13 mean
12 level of 34.2 //g/m3 was reported with a range of 8.8 to 95.0 //g/m3. Covariates in the model
13 included temperature and humidity. An autoregressive model was fitted to the data using GEE
14 methods. There were no significant associations for PM13 and prevalent symptoms in mild
15 asthmatics. For an increase of 50 //g/m3 the odds ratio for p-agonist inhaler use in moderate
16 asthmatics ranged from 3 to 5 for PM13 lags 0 to 3 days and all were statistically significant.
17 A subpopulation analysis of the 21 mild asthmatic subjects not taking regularly scheduled
18 corticosteroids or p-agonists showed a significant effect of lag 4 PM13 on 2-transformed morning
19 PEF. Effects were less related to PM10 than those found related to the other pollutants. Only
20 selected results from selected panels were given.
21 Neukirch et al. (1998) studied the effect of particulate air pollution on 40 non-smoking
22 mild to moderate adult asthmatics in Paris. The study was conducted from September to
23 December, 1992, and was analyzed using group rates. Generalized estimating equations were
24 used to adjust for autocorrelation in the data. The study found some relationships between PM13
25 and nocturnal cough, shortness of breath, and peak flow. Only selected results were given,
26 making the study difficult to evaluate.
27 Romieu et al. (1996) studied 71 children with mild asthma aged 5-7 years living in the
28 northern area of Mexico City, Mexico. Morning and evening peak flow measurements were
29 made and respiratory symptoms were recorded by the parents in a daily diary. During the study
30 period, maximum daily one hour ozone ranged from 40 to 370 ppb (mean 190 ppb,
31 SD = 80 ppb). The 24 hour average PM10 levels measured at one site ranged from 29 to
October 1999 6-13 DRAFT-DO NOT QUOTE OR CITE
-------
1 363 //g/m3 (mean 166.8 //g/m3, SD 72.8 //g/m3). For 53 percent of the study days, PM10 levels
2 exceeded 150 //g/m3. Peak flow measurements were standardized for each person and a model
3 was fitted using GEE methods. The model included terms for minimum temperature. Peak flow
4 was found to be strongly related to PM10. This greater effect may be due to relatively higher
5 levels of pollution in Mexico City and the fact that none of the subjects were on medication.
6 An autoregressive logistic regression model was used to analyze the presence of respiratory
7 symptoms. An increase of 20 //g/m3 PM10 was related to an 8 percent increase in lower
8 respiratory illness.
9 Hiltermann et al. (1998) studied 270 adult asthmatic patients from an out-patient clinic in
10 Leiden, The Netherlands, during July 3 to October 6, 1995. Peak flow was measured twice daily
11 and respiratory symptom data were obtained from diaries. An autoregressive model was fitted to
12 group prevalence of outcomes rather than individual repeated measures. Covariates included
13 temperature and day of week. PM10, ozone, and NO2 were associated with increases in
14 respiratory symptoms. During the study, PM10 levels measured at one central site never exceeded
15 98 //g/m3 (mean 39 //g/m3). Shortness of breath and nocturnal asthma were weakly associated
16 with PM10. The results of this paper were not included in the analysis tables presented later
17 because individual responses were not modeled.
18 The Pollution Effects on Asthmatic Children in Europe (PEACE) study developed
19 methodology for assessing the relationship between short-term changes in air pollution and in
20 acute changes in the health status of children with chronic respiratory symptoms (Roemer et al.,
21 1998). Children with chronic respiratory symptoms (i.e., a positive answer to one of several
22 selected questions) were selected into the panels. The symptom with one of the larger selection
23 percentages was dry cough (range over sample of study communities 29 to 92% [22/75; 84/91]
24 with most values over 50%). The symptom that would most typify selection of asthmatics was
25 doctor-diagnosed asthma (2 to 59% [1/63; 43/72] with most about 20%) (Kotesovec et al., 1998;
26 Clench-Aas et al., 1998; Haluszka et al., 1998; Kalandidi et al., 1998; Forsberg et al., 1998;
27 Beyer et al., 1998). Thus, while asthmatics were included in the subject pool, the overall panels
28 by city tended to have a small percent of asthmatics. The group as a whole did not characterize
29 effects on asthmatics as much as those with chronic respiratory disease, especially cough. The
30 PEACE study was a multi-center study of PM10, BS, SO2, and NO2 on respiratory health of
31 children with chronic respiratory symptoms (Roemer et al., 1998). Mean PM10 levels measured
October 1999 6-14 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
at locals sites ranged from 11.2 to 98.8 //g/m3 over the 28 sites. 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). These studies modeled
group rates and are an example of the panel data problem mentioned earlier.
Roemer et al. (1998, 1999) summarized the results for asthmatic patients in the PEACE
studies. Phlegm was not related to PM10 levels for lags of zero, one or two days. Furthermore,
no relationship was found with the seven day mean. The results for lower respiratory disease
were similar to phlegm.
The above studies are summarized in Tables 6-2 thru 6-13. These tables examine peak
flow, symptoms, and medication use for 50 //g/m3 for PM10 and 25 //g/m3 for PM25. The tables
are split by zero to one-day lags and two-to 5-day lags. Also included in the tables (except for
those with only 2 studies) are results of meta-analyses of the combined results of all the studies
summarized in a given table, using an Empirical Bayes Model Chi-square for Homogenity.
6-2.
EFFECT OF 50 /ttg/m3 PM10 ON EVENING PEAK FLOW (L/MIN)
IN ASTHMATICS LAGGED ZERO OR ONE DAY
Study
Gielen et al.
(1997)
Romieu et al.
(1997)
Peters et al.
(1997c)
Pekkanen et al.
(1997)
Peters et al.
(1997a)
Romieu et al.
(1996)
COMBINED
Description
61 children aged 7 to 13 years living in
Amsterdam, The Netherlands
65 children aged 5-13 years living south of
Mexico City
27 non-smoking adults living in Erfurt, Germany
39 children aged 7-12 years living in Kuopio,
Finland
89 children aged 6-14 living in Sokolov, Czech
Republic
71 children aged 5-7 years living north of
Mexico City
Meta-analysis using an Empirical Bayes Model
Chi-square for homogeneity = 2.89, p = .716
Change in
PFR per
50 Mg/m3
-0.30
-1.55
-0.37
-0.35
-0.92
-2.95
-0.82*
Standard
Error of
Change
0.99
2.11
0.74
2.02
0.53
1.52
0.38
95%
Confidence
Interval
-2.24, 1.64
-5.69, 2.59
-1.82, 1.08
-4.31,3.61
-1.96,0.12
-5.93, 0.03
-1.57, -0.07
*p < 0.05
October 1999
6-15
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-2A. EFFECT OF 50 ^ug/m3 PM10 ON MORNING PEAK FLOW (L/MIN)
IN ASTHMATICS LAGGED ZERO OR ONE DAY
Study
Romieu et al.
(1997)
Peters et al.
(1997c)
Pekkanen et al.
(1997)
Peters et al.
(1997a)
Timonen and
Pekkanen (1997)
Timonen and
Pekkanen (1997)
Romieu et al.
(1996)
COMBINED
*p < 0.05
**p<0.01
Change in PFR Standard Error
Description per 50 Mg/m3 of Change
65 children aged 5-13 years living south of -0.65
Mexico City
27 non-smoking adults living in Erfurt, -1.30**
Germany
39 children aged 7-12 years living in -2.71
Kuopio, Finland
89 children aged 6-14 living in Sokolov, -0.84*
Czech Republic
45 urban children age 7-12 years living in 2.93
Kuopio, Finland
40 suburban children age 7-12 years living -5.55
near Kuopio, Finland
71 children aged 5-7 years living north of -2.95
Mexico City
Meta-analysis using an Empirical Bayes -1.11**
Model Chi-square for homogeneity = 9.33,
p = .156
2.37
0.54
1.97
0.40
2.07
2.87
1.52
0.32
TABLE 6-3. EFFECT OF 50 //g/m3 PM10 ON EVENING PEAK FLOW
Study
Gielenetal. (1997)
Romieu et al. (1997)
Peters et al. (1997c)
Pekkanen et al.
(1997)
Peters et al. (1997a)
Segalaetal. (1998)
Romieu et al. (1996)
COMBINED
IN ASTHMATICS LAGGED TWO TO FIVE
Change in PFR
Description per 50 //g/m3
61 children aged 7 to 13 years living in Amsterdam, -2.32
The Netherlands
65 children aged 5-13 years living south of Mexico -0.04
City
27 non-smoking adults living in Erfurt, Germany -2.3 1 *
39 children aged 7-12 years living in Kuopio, Finland 0. 14
89 children aged 6-14 living in Sokolov, Czech -1.34
Republic
21 asthmatic children aged 7-15 years living in Paris, -0.62
France
71 children aged 5-7 years living north of Mexico City -3.65*
Meta-analysis using an Empirical Bayes Model Chi- -1.21**
square for homogeneity = 2.89, p = .717
DAYS
Standard
Error
1.55
2.17
1.13
3.63
0.76
0.46
1.81
0.39
95%
Confidence
Interval
-5.30, 3.99
-2.36, -0.24
-6.57, 1.15
-1.62, -0.06
-1.13,6.99
-11.18,0.08
-5.93, 0.03
-1.74, -0.48
(L/MIN)
95% Confidence
Interval
-5.36, 0.72
-4.29, 4.23
-4.53, -0.10
-6.98, 7.26
-2.83,0.15
-1.52,0.28
-7.20, -0.10
-1.98, -0.44
*p<0.05
**p<0.01
October 1999
6-16
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-4. EFFECT OF 25 //g/m3 PM2 5 ON EVENING PEAK FLOW IN
ASTHMATICS LAGGED ZERO OR ONE DAY
Study
Peters et al.
(1997c)
Romieu et al.
(1996)
TABLE
Description
27 non-smoking adults living in Erfurt,
Germany
71 children aged 5-7 years living north of
Mexico City
Change in
PFR per
25 Mg/m3
-1.38
-1.98
6-5. EFFECT OF 25 /ttg/m3 PM25 ON EVENING
Standard
95%
Error of Confidence
Change
0.71
2.39
PEAK FLOW
Interval
-2.77,0.01
-6.66, 2.70
IN
ASTHMATICS LAGGED TWO TO FIVE DAYS
Study
Peters et al.
(1997c)
Romieu et al.
(1996)
*p<0.01
TABLE 6-6.
Study
Gielen et al.
(1997)
Romieu et al.
(1997)
Peters et al.
(1997c)
Romieu et al.
(1996)
Peters et al.
(1997a)
Vedaletal. (1998)
COMBINED
Description
27 non-smoking adults living in Erfurt,
Germany
71 children aged 5-7 years living north of
Mexico City
Change in PFR
per 25 //g/m3
-2.18*
-2.55
Standard 95%
Error
0.82 -3
2.70 -7
Confidence
Interval
.79, -0.57
.84,2.74
EFFECT OF 50 /ttg/m3 PM10 ON COUGH IN ASTHMATICS LAGGED
ZERO OR ONE
Description
61 children aged 7 to 13 years living in
Amsterdam, The Netherlands
65 children aged 5-13 years living south of
Mexico City
27 non-smoking adults living in Erfurt, Germany
71 children aged 5-7 years living north of
Mexico City
89 children aged 6-14 living in Sokolov, Czech
Republic
206 children aged 6 to 1 3 years living in Port
Alberni, British Columbia
Meta-analysis using an Empirical Bayes Model
Chi-square for homogeneity = 32.71, p < .001
DAY
Odds Ratio for
Event per
50 Mg/m3
2.19
1.21
1.01
1.21*
1.01
1.40
1.12*
Standard Error
of Log-Odds
Ratio
0.531
0.046
0.026
0.047
0.032
0.150
0.046
95%
Confidence
Interval
0.77, 6.20
1.11, 1.32
0.96, 1.06
1.10, 1.33
0.95, 1.08
1.04, 1.88
1.03, 1.23
*p<0.05
October 1999
6-17
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-7. EFFECT OF 50 //g/m3 PM10 ON COUGH IN ASTHMATICS LAGGED
TWO TO FIVE DAYS
Study
Gielen et al.
(1997)
Romieu et al.
(1997)
Peters et al.
(1997c)
Romieu et al.
(1996)
Peters et al.
(1997a)
Vedal et al.
(1998)
COMBINED
*p < 0.05
**p<0.01
TABLE 6-8.
Study
Romieu et al.
(1997)
Peters et al.
(1997c)
Romieu et al.
(1996)
Peters et al.
(1997a)
Vedal et al.
(1998)
COMBINED
Description
61 children aged 7 to 13 years living in
Amsterdam, The Netherlands
65 children aged 5-13 years living south of
Mexico City
27 non-smoking adults living in Erfurt,
Germany
71 children aged 5-7 years living north of
Mexico City
89 children aged 6-14 living in Sokolov,
Czech Republic
206 children aged 6 to 13 years living in
Port Alberni, British Columbia
Meta-analysis using an Empirical Bayes
Model Chi-square for homogeneity
= 16.17, p = . 006
Odds Ratio for
Event per
50 Mg/m3
2.19
1.21*
1.08
1.27*
1.10*
1.40*
1.15**
Standard
Error of Log-
Odds Ratio
0.787
0.047
0.026
0.086
0.031
0.108
0.032
EFFECT OF 50 ^g/m3 PM10 ON PHLEGM IN ASTHMATICS
ZERO OR ONE DAY
Description
65 children aged 5-13 years living south of
Mexico City
27 non-smoking adults living in Erfurt,
Germany
71 children aged 5-7 years living north of
Mexico City
89 children aged 6-14 living in Sokolov,
Czech Republic
206 children aged 6 to 13 years living in
Port Alberni, British Columbia
Meta-analysis using an Empirical Bayes
Model Chi-square for homogeneity = 0.80,
p = .938
Odds Ratio for
Event per
50 ,ug/m3
1.10
1.10*
1.10
1.13*
1.28
1.11**
Standard
Error of Log-
Odds Ratio
0.095
0.037
0.085
0.043
0.200
0.026
95%
Confidence
Interval
0.47, 10.24
1.10, 1.33
1.03, 1.14
1.07, 1.50
1.03, 1.17
1.08, 1.23
LAGGED
95%
Confidence
Interval
0.91, 1.33
1.02,1.18
0.93, 1.30
1.04, 1.23
0.86, 1.89
1.06, 1.17
*p<0.05
**p<0.01
October 1999
6-18
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-9. EFFECT OF 50 ^ug/m3 PM10 ON PHLEGM IN ASTHMATICS LAGGED
TWO TO FIVE DAYS
Study
Romieu et al.
(1997)
Peters et al.
(1997c)
Romieu et al.
(1996)
Peters et al.
(1997a)
Vedal et al.
(1998)
COMBINED
*p < 0.05
**p<0.01
TABLE
Study
Romieu et al.
(1997)
Ostro et al.
(1995)
Romieu et al.
(1996)
COMBINED
Description
65 children aged 5-13 years living south of
Mexico City
27 non-smoking adults living in Erfurt,
Germany
71 children aged 5-7 years living north of
Mexico City
89 children aged 6-14 living in Sokolov,
Czech Republic
206 children aged 6 to 13 years living in
Port Alberni, British Columbia
Meta-analysis using an Empirical Bayes
Model Chi-square for homogeneity = 0.80,
p = .938
Odds Ratio for
Event per
50 ptg/m3
1.00
1.05
1.05
1.12*
1.40*
1.09**
Standard
Error of Log-
Odds Ratio
0.074
0.096
0.096
0.037
0.156
0.031
95%
Confidence
Interval
0.86, 1.16
0.87, 1.27
0.87, 1.27
1.04, 1.20
1.03, 1.90
1.03, 1.16
6-10. EFFECT OF 50 ^g/m3 PM10 ON DIFFICULTY IN BREATHING IN
ASTHMATICS LAGGED ZERO OR ONE DAY
Description
65 children aged 5-13 years living south of
Mexico City
83 African- American children living in
central Los Angeles
71 children aged 5-7 years living north of
Mexico City
Meta-analysis using an Empirical Bayes
Model Chi-square for homogeneity = 0.80,
p = .938
Odds Ratio for
Event per
50 Mg/m3
1.18*
1.71**
1.05
1.18
Standard
Error of Log-
Odds Ratio
0.046
0.180
0.096
0.043
95%
Confidence
Interval
1.08, 1.29
1.20,2.43
0.87, 1.27
1.08, 1.28
*p < 0.05
**p<0.01
October 1999
6-19
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-11. EFFECT OF 50 //g/m3 PM10 ON DIFFICULTY IN BREATHING IN
ASTHMATICS LAGGED TWO TO FIVE DAYS
Study
Romieu et al.
(1997)
Romieu et al.
(1996)
Description
65 children aged 5-13 years living
south of Mexico City
71 children aged 5-7 years living
north of Mexico City
Odds Ratio for
Event per 50 //g/m3
1.21
1.05
Standard Error of
Log-Odds Ratio
0.058
0.150
95% Confidence
Interval
1.08,1.36
0.78, 1.41
TABLE 6-12. EFFECT OF 50 ^g/m3 PM10 ON BRONCHODILATOR USE IN
ASTHMATICS LAGGED ZERO OR ONE DAY
Odds Ratio for Standard 95%
Event per Error of Log- Confidence
Study Description 50 //g/m3 Odds Ratio Interval
Gielen et al.
(1997)
Peters et al.
(1997c)
61 children aged 7 to 13 years living in
Amsterdam, The Netherlands
27 non-smoking adults living in Erfurt,
Germany
0.94
1.06
0.237
0.094
0.59, 1.50
0.88, 1.27
TABLE 6-13. EFFECT OF 50 ^ug/m3 PM10 ON BRONCHODILATOR USE IN
ASTHMATICS LAGGED TWO TO FIVE DAYS
Study
Gielen et al.
(1997)
Peters et al.
(1997c)
Description
61 children aged 7 to 13 years living in
Amsterdam, The Netherlands
27 non-smoking adults living in Erfurt,
Germany
Odds Ratio for
Event per
50 ,ug/m3
2.90*
1.23
Standard
Error of Log-
Odds Ratio
0.242
0.128
95%
Confidence
Interval
1.80,4.66
0.96, 1.58
*p<0.01
1 The results of the peak flow analyses consistently show small decrements for both PM10
2 and PM2 5. The results were shown for both morning (AM) and evening (PM) peak flow The
3 effects using two to five day lags averaged about the same as did the zero to one day lags, but the
October 1999
6-20
DRAFT-DO NOT QUOTE OR CITE
-------
1 effects seen were slightly less consistent. None of the studies provided analyses which were able
2 to separate out the effects of PM10 and PM25 from other pollutants, nor were they able to
3 distinguish the effects between PM10 and PM2 5.
4 The effects on respiratory symptoms also tended to be positive although they were much
5 less consistent. Most studies showed increases in cough, phlegm, difficulty in breathing, and
6 bronchodilator use, although these increases were generally not statistically significant.
7 Bronchodilator use was the only endpoint that appeared to be more strongly related to the longer
8 lag times, but this result is based on three studies.
9
10 6.2.1.2 Short Term Effects on Lung Function and Respiratory Symptoms in
11 Non-Asthmatics
12 Roemer et al. (1993) studied acute respiratory symptoms in a panel of 73 Dutch children
13 with chronic respiratory symptoms in the winter of 1990-91 living in Wageningen and
14 Bennekom, The Netherlands. Exposure measurements included SO2, NO2, and PM10. PM10
15 measured at a central site exceeded 110 //g/m3 on six days over the 3 month study period. Daily
16 measurements of peak flow were made twice a day. A diary was used to measure the occurrence
17 of acute respiratory symptoms and medication use. A time series analysis was performed using
18 the SAS procedure, AUTOREG, using the Yule-Walker estimation method. Both morning and
19 evening peak flow measurements were marginally significant in their relationship to PM10, BS
20 and SO2 levels. PM10 was also associated with increased bronchodilator use.
21 Hoek and Brunekreef (1994) studied 1079 children living in four non-industrial
22 communities in The Netherlands. The study was conducted during the three winters of 1987-88,
23 1988-89, and 1989-90. Pollutants measured included SO2, NO2, PM10, sulfate fraction, nitrate
24 fraction, and acid aerosol. PM10 levels were low (mean 45, range 14-126 //g/m3). A first order
25 autoregressive model, which contained temperature as a covariate, found a weak association
26 between most pollutants and peak flow but no relationship with respiratory symptoms.
27 Hoek et al. (1998) summarized and reanalyzed results from several other studies reported in
28 the literature, including those on asymptomatic children in the Utah Valley of Utah (Pope et al.,
29 1991), children in Bennekom, The Netherlands (Roemer et al., 1993), children in Uniontown, PA
30 (Neas et al., 1995), and children in State College, PA (Neas et al., 1996). The point of the
31 reanalysis was to show an alternative method for summarizing peak flow studies, that is, relative
October 1999 6-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 odds for a substantial 10% decrease in peak flow. Hoek et al. (1998) conducted a different type
2 of analysis on peak flow data, the results of which become much more readily interpretable from
3 a clinical point of view: most PEF studies have looked at changes in population mean PEF in
4 response to air pollution findings, effects on the order of a few percentage points over a relevant
5 pollution concentration range. The small changes seen are within the circadian variation
6 everybody experiences every day, which is not associated with the symptomatology. Hoek et al.
7 (1998) shows that these small changes in the population mean PEF are accompanied by
8 significantly increased percentages of subjects experiencing large PEF changes (of more than
9 10% or more than 20% of their habitual level). Such relatively large changes can be associated
10 with symptoms and with the need to use bronchodilators, which makes this analysis more
11 coherent with findings of increased respiratory symptoms and relief medication use also found to
12 be associated with air pollution. Significant decreases in peak flow were found to be related to
13 increases in PM10 for models using data from the five panels. The analyses were done using a
14 first-order autoregressive model with adjustments for time trend and ambient temperature.
15 Co-pollutant models showed effects of ozone and PM10 to be largely independent.
16 Schwartz et al. (1994) also summarized results from several studies, including the "Six
17 Cities study" (Ferris et al., 1979) and the asymptomatic patients from the Utah Valley study
18 (Pope et al., 1992). Autoregressive logistic models were fitted using GEE methods. The
19 symptoms of cough, lower respiratory disease, cough, and upper respiratory disease were found
20 to be associated with PM10, SO2, and ozone.
21 Roemer et al. (1998, 1999) summarized the results for patients selected for cough in the
22 PEACE studies. Phlegm was not related to PM10 levels for lags of zero, one or two days.
23 Furthermore, no relationship was found with the seven day mean. The results for lower
24 respiratory disease were similar to phlegm.
25 Linn et al. (1998) report the outcome of a study of 30 volunteer Los Angeles area residents
26 with severe chronic obstructive pulmonary disease (COPD), relating pollutant levels (PM10,
27 PM25, O3, NO2) to health outcomes (blood pressure, lung function, arterial blood oxygen
28 saturation). The authors report that LA area monitoring stations appeared to give meaningful
29 estimates of PM exposures outdoors at the homes of the COPD subjects studied with respect to
30 temporal as well as spatial variations but note that on the whole, their findings provide only weak
31 support that personal exposures track ambient background PM levels. They found the following:
October 1999 6-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 (1) daily mean diastolic blood pressure increased significantly with same-day or previous day
2 24 hr-mean PM10 at the nearest monitoring stations, (2) no significant relationships with blood
3 pressure were found for outside home PM2 5, (3) some analyses of lung function versus
4 subject-oriented PM measurements showed statistically significant negative relationships, but
5 were usually eliminated by excluding 2 or 3 "outlying" subjects, and (4) arterial blood oxygen
6 saturation showed no significant relationships. The authors suggest caution in interpreting the
7 study.
8 Harre et al. (1997) studied 40 subjects aged over 55 years with COPD living in
9 Christchurch, New Zealand. The study was conducted during the winter of 1994. Subjects
10 completed diaries twice daily as well as their peak flow measurement. Pollutants measured
11 included SO2, NO2, PM10, and CO. Local PM10 measures exceeded 120 //g/m3 five times during
12 the study period. A log-linear regression model with adjustment for first order autocorrelation
13 was used to analyze the peak flow data and a Poisson regression model was used to analyze the
14 symptom data. Few significant associations between the health endpoints and the pollutants
15 were found.
16 Boezen et al. (1999) studied 632 children aged 7 to 11 years of age during three winters
17 (1992-95) in The Netherlands. The analyses were performed on two subpopulations: the
18 36 percent of children with no bronchial hyperresponsiveness and total IGE of 60 kU/L or less,
19 and the remaining 64 percent without. Lung function was measured as the dichotomous variable:
20 a decrease of 10 percent or more. Upper and lower respiratory symptoms were also measured.
21 The PM10 readings ranged from 4.7 to 145.6 //g/m3. A logistic regression model was used to
22 analyze the data. Autocorrelation was adjusted for using the AR macro (SAS version 6.12).
23 No consistent relationships were found between the health endpoints and PM10 levels.
24 Korrick et al. (1998) studied the effect of short-term changes in pollution on adult hikers on
25 Mt. Washington, NH. Ozone levels were measured at two sights near the top of the mountain
26 and PM2 5 was measured near the base of the mountain. Both linear and non-linear regression
27 models were used to assess the effect of pollution on lung function. Logistic regression was used
28 to assess the effect on the odds of having a decline of greater than 10 percent in lung function.
29 No estimates were given for the effect of PM25 but it was stated that PM25 did not affect the
30 coefficient for ozone.
October 1999 6-23 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
Naeher et al. (1999) studied 473 non-smoking women (ages 19 to 43 years) in Virginia over
the 1995 and 1996 summers. Subjects took peak flow measurements twice a day for two weeks
each summer. A regional monitoring site measured PM25, PM10, sulfate, strong acid (H+), hourly
ozone, and meteorological variables. The data were analyzed using a mixed model assuming a
linear regression term for each subject. These terms were assumed to be random and normally
distributed and were estimated using SAS proc MIXED. Morning changes in peak flow were
related to current day H+ and PM2 5. Ozone was related to changed in evening peak flow.
Tables 6-14 through 6-17 examine outcomes for studies of non-asthmatics. Again, results
of meta-analyses of combined results from the studies summarized in each table are also
provided.
TABLE 6-14. EFFECT OF 50 ^ug/m3 PM10 ON PEAK FLOW (L/MIN)
IN NON-ASTHMATICS LAGGED ZERO OR ONE DAY
Study
Hoek and
Brunkreef(1994)
Hoek etal. (1998)
Hoek etal. (1998)
Hoek etal. (1998)
Hoek etal. (1998)
Harre etal. (1997)
COMBINED
Description
73 children aged 6-12 years with
respiratory symptoms in the Netherlands
39 asymptomatic children in the Utah
Valley
67 children in Bennekom, the Netherlands
83 children in Uniontown, PA
108 children in State College, PA
40 adults aged 55+ years with COPD living
in Christchurch, New Zealand
Meta-analysis using an Empirical Bayes
Model Chi-square for homogeneity = 1.12,
p = 0.952.
Change in
PFR per
50 //g/m3
-0.42*
-0.33*
-0.45
-0.95
-0.15
-0.86
-0.38*
Standard
Error of
Change
0.15
0.11
1.05
1.60
1.45
0.75
0.09
95%
Confidence
Interval
-0.71 ,-0.1260
-0.55, -0.11
-2.51, 1.61
-4.09,2.19
-2.99, 2.69
-2.33,0.61
-0.56, -0.20
*p < 0.05
October 1999
6-24
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-15. EFFECT OF 50 //g/m3 PM10 ON COUGH IN NON-ASTHMATICS
LAGGED ZERO OR ONE DAY
Odds Ratio Standard Error 95%
for Event per of Log-Odds Confidence
Study Description 50 //g/m3 Ratio Interval
Hoek and Brunkreef
(1994)
Schwartz et al. (1994)
Schwartz et al. (1994)
COMBINED
73 children aged 6-12 years with 0.94
respiratory symptoms in the
Netherlands
Six Cities Study 1.39*
3 9 asymptomatic children in the Utah 1.36*
Valley
Meta-analysis using an Empirical 1.31*
Bayes Model Chi-square for
homogeneity =, p = .
0.255 0.57, 1.55
0.145 1.05, 1.85
0.120 1.07, 1.72
0.088 1.10,1.56
*p<0.05
TABLE 6-16. EFFECT OF 50
PM ON LOWER RESPIRATORY
10
ILLNESS IN NON-ASTHMATICS LAGGED ZERO OR ONE DAY
Study
Hoek and
Brunkreef
(1994)
Schwartz et al.
(1994)
Schwartz et al.
(1994)
Boezen et al.
(1999)
COMBINED
Odds Ratio for
Event per
Description 50 ptg/m3
73 children aged 6-12 years with 1 .00
respiratory symptoms in the
Netherlands
Six Cities Study 2.03*
39 asymptomatic children in the Utah 1.21
Valley
167 children aged 7-11 years living in 1 .02
The Netherlands
Meta-analysis using an Empirical 1.211
Bayes Model Chi-square for
homogeneity = 10.23, p = 0.017
Standard 95%
Error of Log- Confidence
Odds Ratio Interval
0.172 0.71, 1.40
0.206 1.36,3.04
0.187 0.84, 1.75
0.100 0.84, 1.24
0.128 0.94, 1.56
*p < 0.05
October 1999
6-25
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-17. EFFECT OF 50 //g/m3 PM10 ON UPPER RESPIRATORY
ILLNESS IN NON-ASTHMATICS LAGGED ZERO OR ONE DAY
Study
Hoek and
Brunkreef
(1994)
Schwartz et al.
(1994)
Schwartz et al.
(1994)
Boezen et al.
(1999)
COMBINED
Odds Ratio for
Event per
Description 50 ,ug/m3
73 children aged 6-12 years with 1 .00
respiratory symptoms in the
Netherlands
Six Cities Study 1.39
39 asymptomatic children in the Utah 1.03
Valley
167 children aged 7-11 years living in 1.01
The Netherlands
Meta-analysis using an Empirical 1 .02
Bayes Model Chi-square for
homogeneity = 2.76, p = .251
Standard Error 95%
of Log-Odds Confidence
Ratio Interval
0.089 0.84,1.19
0.186 0.97,2.00
0.166 0.74,1.43
0.033 0.94, 1.07
0.031 0.96, 1.08
1 The results of the peak flow analyses consistently show small decrements for increases in
2 PM10. The results are similar to those found for asthmatics. There were no studies that gave
3 results for PM2 5, and no studies gave results for longer lag times.
4 Studies not meeting the criteria for discussion above are summarized in Tables 6-18 and
5 6-19. Many excellent studies are included in these tables without further discussion. These
6 studies provide supporting evidence for the conclusions reached from the studies selected for text
7 discussion.
9
10
11
12
13
14
15
16
6.2.1.3 Discussion of Co-Pollutant Studies
A small number of short-term PM exposure respiratory studies considered multiple
pollutants in the same model. These are described individually.
Delfino et al. (1998) found that the presence of ozone in a model with PM10 reduced
slightly the effect of PM10 on asthma symptoms. However, all terms that were significant
without ozone in the model remained significant with ozone added to the model.
October 1999
6-26
DRAFT-DO NOT QUOTE OR CITE
-------
o
a
o
cr
fD
TABLE 6-18. OTHER ASTHMATIC PANEL STUDIES
Study
Design
Model Type
Other
Pollutants
Other Covariates
Results
H
o
o
2
o
H
o
H
W
O
^
O
HH
H
W
Thurston Three 5-day summer camps
et al. (1997) with 166 asthmatic children
conducted in 1991, 1992,
1993 measuring symptoms
and change in lung function
(morning to evening)
Agocs et al. A panel of 60 asthmatic
(1997) children was followed for
two months in Budapest,
Hungary
Giintzel et al. An asthma reporting system
(1996) was us ed in connection with
pollutant monitoring in
Switzerland from the fall of
1988 to the fall of 1990
Taggart et al. A panel of 38 adult
(1996) asthmatics were followed
from July 17 to September
22, 1993 in northern
England
Delfinoetal. A panel of 12 asthmatic
(1996) children with symptomatic
asthma living in San Diego,
CA were followed during
the early fall of 1993.
Linear regression for lung
function, Poisson regression
for symptoms and
bronchodilator use in
relation to sulfate, with
random subject effects.
A mixed model relating TSP
to the morning and evening
PEFR measurements was
used
A Box-Jenkins ARIMA time
series model was used to
relate asthma to TSP
A generalized linear model
was used to relate pollutants
to bronchial
hyperresponsiveness
A random effects model was
fitted to ordinal symptom
scores and bronchodilator
use in relation to 24-hour
PM25.
Ozone, H+ Pollen, daily maximum
temperature,
SO,
Ozone,
SO2, NO2
time trend, day of
week, temperature,
humidity
Temperature
SO2, NO2 Temperature
Temperature, relative
humidity, fungal
spores, and day of
week.
Sulfate and ozone were related to
both respiratory symptoms and
bronchodilator use.
No significant relationships with
TSP were found
No significant relationships were
found
Small effects were seen in relation
to NO2 and black smoke
Symptoms and bronchodilator use
were associated with 12-hour
personal ozone measurements, but
not stationary outdoor monitor data
on 1 -hour maximum ozone or
24-hour PM2 5. Fungal spores were
associated with symptoms and
inhaler use.
-------
o
a
o
cr
fD
TABLE 6-18 (cont'd). OTHER ASTHMATIC PANEL STUDIES
Study
Design
Model Type
Other
Pollutants
Other Covariates
Results
Define et al. A panel of 9 adults and
(1997) 13 children were followed
during the late spring of
1994 in a 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 PEF in relation to
24-hour PM10.
Temperature, relative
humidity, fungal
spores, day of week
Although PM10 never exceeded
51 Mg/m3, bronchodilator use was
significantly associated with PM10
(0.152 inhaler puffs/10 Mg/m3;
SE 0.064). Fungal spores were
associated with all respiratory
outcomes.
Hiltermann Sixty outpatient asthmatics
et al. (1997) were examined for nasal
inflammatory parameters in
The Netherlands from July 3
to October 6, 1995.
The association of log
transformed inflammatory
parameters to 24-h PM10
were analyzed for using a
linear regression model.
Mugwort-pollen
An association of inflammatory
parameters in nasal lavage of
patients with intermittent to severe
persistent asthma with ambient
ozone and allergen exposure was
observed, but not with exposure to
PM,n.
oo
H
o
O
2
O
H
o
H
W
O
&
O
HH
H
W
-------
o
o
r+
O
cr
o
TABLE 6-19. OTHER NON-ASTHMATIC PANEL STUDIES
Study
Design
Model Type
Other Pollutants
Other Covariates
Results
Spektoretal. (1991)
Studnicka et al. (1995)
Scarlett et al. (1996)
Monthly time series analysis of
pulmonary function
Three panels of 16 to 19 day
duration measured at a summer
camp
154 school children had
pulmonary function measured
daily for 3 1 days
Not given
SO2, O3
Height, weight
Linear regression of lung function H+, sulfate, ammonia, Temperature,
allowing for repeated measures ozone humidity, pollen
Separate autoregressive models for
each child were pooled
PMi0, ozone, NO2
Pollen, machine,
operator, time of day,
time trend
Pulmonary function related to PMU
Pulmonary function related to H+
but not to PMi0
PMi0 was related to changes in FEV
and FVC
s_/
£
Tl
H
o
o
o
H
O
c
o
H
W
O
Gordian et al. (1996)
Cuijpers et al. (1994)
Awasthi et al. (1996)
Miyao et al. (1993)
Long et al. (1998)
Boezen et al. (1998)
Linn et al. (1996)
Outpatient visits for upper
respiratory symptoms were related
to ambient PMjo levels
Summer episode study in
Maastrucht, The Netherlands PMi0
measured.
A cohort of 664 preschool children
were followed for two weeks each
in northern India
Japanese national health insurance
records were correlated with
suspended particulate matter
428 participants with mild airway
obstruction in a health study were
surveyed during a pollution
episode
75 symptomatic and asymptomatic
adults near Amsterdam were
surveyed during the winter of
1993-1994 for three months
269 school children in Southern
California were surveyed twice
daily for one week in the fall,
winter and spring for two years
An autoregressive Poisson model was CO
fitted to detrended pollution and
meteorological data
Paired t tests were used for
pulmonary function tests, methods
for respiratory symptoms not given
Ordinary least squares was used to
relate a respiratory symptom complex
to suspended particulate matter
Pearson correlation
Gender specific odds ratios of
symptoms were calculated for
differing PM10 levels using the
Breslow-Day test
An autoregressive logistic model was
used to relate PMi0 to respiratory
symptoms, cough, and phlegm.
A repeated measures analysis of
covariance was used to fit an
autoregressive model
SO2, NO2, BS, ozone,
H+
SO2, nitrates
SO2, NO2
TSP, VOC
Weekday, temperature
none in model
Coal, wood, kerosene
cedar and Cyprus
pollen
NO2, ozone
Daily minimum
temperature, time
trend, day of week
Year, season, day of
week, temperature
Upper respiratory symptoms were
associated with increased PMi0
levels
Small decreases in lung function
were found
A significant regression coefficient
between PM and symptoms was
found
No significant correlation with
particulate matter was found
Cough, wheezing, chest tightness,
and shortness of breath were all
increased during the episode
No relationship was found with
pulmonary function. Some
significant relationships with
respiratory disease were found in
subpopulations
Morning FVC was significantly
decreased as a function of PM5 and
NO2
O
HH
H
W
-------
1 Timonen and Pekkanen (1997) found a relationship between PM10 and decreases in
2 morning peak flow for lags of two days or for a four day mean. When NO2 was added to the
3 model, the coefficients for PM10 remained essentially unchanged.
4 Romieu et al. (1997) found a significant relationship between peak flow and respiratory
5 symptoms as dependent variables and ozone and PM10 as independent variables in asthmatic
6 children in Mexico City. When both ozone and PM10 were included in the model, the
7 significance of ozone increased. PM10 was not significant in this study and it remained so after
8 the inclusion of ozone. Romieu et al. (1996) in a separate study found an effect of PM25 on peak
9 flow. This effect remained after the inclusion of ozone in the model.
10 Studnika et al. (1995) observed decreases in FEVj as a function of H+ and PM10 levels in
11 children participating in a summer camp. The coefficients for PM10 remained relatively constant
12 with the inclusion of H+ in the model.
13 Gold et al. (1999) found that both PM25 and ozone were related to changes in morning peak
14 flow. The combined effect of the two pollutants was somewhat larger than the effect of each
15 pollutant individually in the model, but was less than the sum of the two separate effects.
16 Although these studies are not definitive about the effects of multiple pollutants, it does
17 appear that the pollutants considered in these models act somewhat independent of each other.
18 There were no cases where the statistical significance was eliminated by the inclusion of
19 co-pollutants.
20
21 6.2.2 Long-Term Exposure Effects on Lung Function and Respiratory
22 Symptoms
23 A small number of studies have been published recently which looked at the effects of
24 long-term PM exposure on lung function and respiratory illness. Some of these did not relate
25 effects directly to PM10 or PM2 5.
26 Dockery et al. (1996) examined respiratory health effects among 13,369 white children,
27 aged 8 to 12 years, from 24 communities in the United States and Canada. A two-stage logistic
28 regression model was used to adjust for potential confounding from gender, history of allergies,
29 parental asthma, parental education, and smoking in the home. The city-specific range in the
30 annual pollutant means from local monitors were 14.9 //g/m3 for PM2 j and 17.3 //g/m3 for PM10.
October 1999 6-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 Although the endpoint of bronchitis was significantly related to fine particulate sulfates, no
2 endpoints were related to PM10 levels.
3 Abbey et al. (1998) studied associations between lung function measures collected in 1997
4 and estimated 20-year exposure to respirable particulates, suspended sulfates, ozone, and PM10 in
5 1391 non smoking seventy-day adventist, throughout CA. Increased frequency of days where
6 PM10 exceeded 100 //g/m3 was associated with a decrement in FEV in males whose parents had
7 asthma, bronchitis, emphysema, or hay fever. No other effects were seen in any other subgroups.
8 Braun-Fahrlander et al. (1997) studied the impact of long-term exposure to air pollution on
9 respiratory symptoms and illnesses in a cross-sectional study of school children (aged 6 to
10 15 years), living in ten different communities in Switzerland. Air pollution measurements
11 included PM10, NO2, SO2, and ozone. Local PM10 measurements yielded annual PM10 means in
12 the ten communities ranging from 10 to 33 //g/m3. Symptoms were analyzed using a logistic
13 regression model that included the covariates of family history of respiratory and allergic
14 diseases, number of siblings, parental education, indoor fuels, passive smoking, and others. The
15 endpoints of chronic cough, bronchitis, wheeze and conjunctivitis symptoms were all related to
16 the various pollutants. The collinearity of the pollutants prevented any causal separation.
17 Ackermann-Liebrich et al. (1997) studied the long-term effect of air pollution in a
18 cross-sectional population-based sample of Swiss adults aged 18 to 60 years. Air pollutants
19 monitored by local authorities in each of eight regions included SO2, NO2, TSP, ozone, and PM10.
20 Annual PM10 means ranged from 10.1 to 33.4 //g/m3. A random sample of 2,500 adults in each
21 region was drawn from registries of the local inhabitants. The natural logarithm of FVC and
22 FEVj was regressed against the natural logarithms of height, weight, age, gender, atopic status,
23 and pollutant variables. The estimated regression coefficient for PM10 versus FVC was
24 -0.0345 with 95 percent confidence intervals of-0.0407 to-0.0283. The corresponding value for
25 FEVj was -0.0160 with 95 percent confidence interval of-0.0225 to -0.0095. Thus, a 10 //g/m3
26 increase in PM10 was estimated to lead to an estimated 3.4 percent decrease in FVC and
27 1.6 percent decrease in FEV].
28 Von Mutius et al. (1995) studied 9 to 11 year old school children living in Leipzig, East
29 Germany. Parents filled out a respiratory disease questionnaire. The presence of respiratory
30 symptoms was confirmed by a physician. After controlling for parental education, passive smoke
31 exposure, temperature, and humidity, the odds ratio for developing upper respiratory illness was
October 1999 6-31 DRAFT-DO NOT QUOTE OR CITE
-------
1 1.62 with 95 percent confidence limits of 1.08 to 2.45. PM was measured by beta-absorption.
2 The same effects were seen when the other pollutants were used as the independent variable.
3 Raizenne et al. (1996) examined the health effects of exposure to acidic air pollution
4 among children living in 24 communities in the United States and Canada. Parents of the
5 children (aged 8 to 12) completed a self-administered questionnaire. Pulmonary function
6 measurements were taken at the end of the pollution monitoring period. These measurements
7 were analyzed using a two-stage regression analysis that adjusted for age, gender, height, and
8 weight. Although decreases in lung function were highly related to particle strong acidity, they
9 were also related to PM10. The particle strong acidity range over the study areas was
10 43.4nmol/m3.
11 To study whether chronic effects of ozone exposure can be observed in an unselected
12 cohort of 1,150 children, Frischer et al. (1999) prospectively followed a cohort of primary school
13 children in Austrian counties by repeated measurements of lung function between January 1994
14 and December 1996. This unique study observed during the 3 yr time period small but consistent
15 decrements in lung function in a cohort of children with ambient ozone exposure. They report
16 that for PM10, a positive effect was seen for winter exposure, which was completely confounded
17 by temperature. Tager (1999) comments that in this study the data indicated that summertime
18 PM10 was associated with decreased growth of MEF50 (maximum expiratory flow at 50% of vital
19 capacity). However, the more important comment of Tager (1999) is that cross-sectional studies
20 are fraught with too many methodological problems, and should be abandoned as a design
21 strategy and that the approach taken by Frischer et al. (1999) offers a practical study design that
22 should be emulated. Such shorter term prospective studies of young children will help develop a
23 more substantive database through which more accurate estimates of effects of ambient air
24 pollution on lung function growth can be obtained.
25 Lewis et al. (1998) studied 3,023 primary school children in New South Wales. Particulate
26 and sulfur dioxide measures were collected from January 1993 to December 1993. Children in
27 each of the nine areas were chosen to be within 3 km of a monitoring station. Night cough, chest
28 colds, and wheeze were found to be related to PM10 levels.
29 Peters et al. (1999a) studied 12 southern California communities with differing levels of
30 pollution. In each community, 150 students in grades 4 and 7 were enrolled. The study was
31 conducted in 1993, but pollution estimates were based on the years 1986-1990. Several
October 1999 6-32 DRAFT-DO NOT QUOTE OR CITE
-------
1 respiratory illness rates were related to PM10 levels including the presence of asthma, bronchitis,
2 cough, or wheeze. None of the symptoms were significantly related to PM10. Peters et al.
3 (1999b) also studied the effect of PM10 on lung function in the same population. Forced vital
4 capacity and mid-maximal flow rate were negatively related to PM10 levels, but FEV] and peak
5 flow were not. All results tended to show a negative effect of PM10.
6 Three other studies related long-term PM10 exposure to PM10. Jammes et al. (1998)
7 compared lung function at two sites (suburbs and downtown Marseilles) and found some
8 differences. Zemp et al. (1999) studied the effect of PM10 on respiratory symptoms in eight
9 communities in Switzerland and found that PM10 was related to differences in rates of chronic
10 phlegm, breathlessness, and dyspnea in never smokers. Results for ex-smokers and current
11 smokers tended to show smaller effects. McDonnell et al. (1999) looked at a cohort of
12 3,091 non-smokers in southern California (AHSMOG Study). Most of the analyses were done
13 using ozone as the pollutant of interest, and the inclusion of PM10 in the analyses had little effect
14 on the coefficients for ozone.
15 Goren et al. (1999) studied school children (ages 7 to 13 years) who resided in two
16 communities where one community had quarries and a cement factory as PM sources, while the
17 other community was geographically near but divided from the sources by two valleys and two
18 hills. Pollutants measured included only TSP and, for a limited time, PM10. Both pulmonary
19 function measurements and respiratory symptom data were obtained by standard methods in the
20 spring of 1995. While geographically close the socioeconomics status of the populations differed
21 significantly and was thus controlled for the logistic regression analyses. Approximately 16% of
22 PM10 levels exceeded 150 //g/m3 in the source-exposed community. TSP levels in the adjacent
23 community were about 30 to 60% of those in the source-exposed community. The source
24 community had a PEF decrement of 97.03% predicted, while the adjacent community was at
25 99.80% with a p = 0.0326 for the difference. While not significant, positive OR's were found for
26 the symptoms such as a cough without cold.
27 Other new studies examined long term pollutant effects but did not use PM10 or PM2 5 as the
28 PM measure (De Kok et al., 1996; Chen et al., 1998; and Zejda et al., 1996).
29 Overall, the results of the chronic studies are not consistent. Some studies show effects for
30 some endpoints, but other studies fail to find the same effects. It is generally more difficult to
31 find a gradient in long term exposures, whereas short term studies need only find an area with
October 1999 6-33 DRAFT-DO NOT QUOTE OR CITE
-------
1 occasional high exposures. For this reason it is not surprising that the studies show less
2 consistency than the acute exposure studies.
3
4 6.2.3 Effects of Short-Term PM Exposure on the Incidence of Respiratory
5 Medical Visits and Hospital Admissions
6 6.2.3.1 Introduction
7 Potentially the most severe morbidity measure evaluated with regard to PM exposure is
8 hospitalization. Hospital emergency department (ED) visits represent a somewhat less severe,
9 but related, outcome that has also been studied in relation to air pollution. In addition, doctors
10 visits also represent a related health measure that, although even less studied, is also relevant to
11 those who also suffer severe health effects, but captures a different population than ED visits:
12 those who choose to visit a private doctor, rather than attend an emergency department. This
13 latter category of pollution affected persons can represent a large population, yet one that is
14 largely unobserved, due to the usual lack of centralized data regarding doctors' visits. Indeed, we
15 are often limited in such epidemiologic studies to "looking under the lamp post" for effects in
16 data that are routinely collected, but which may miss many cases of adverse effects that are just
17 as severe, but are not included in the routine health care record keeping system at hospitals.
18 This section evaluates present knowledge regarding the epidemiologic associations of
19 hospitalizations and medical visits with ambient PM exposure. It intercompares various PM
20 components vis-a-vis their relative associations with adverse health effects, as well as their
21 respective extents of coherence in the PM associations exhibited across related measures of
22 adverse health effects. In the following discussions, the focus for quantitative intercomparisons
23 is on studies and results considering PM metrics that quantitatively measure mass or a specific
24 mass constituent, i.e.: PM10, PM2 5, sulfates (SO4=), or acidic aerosols (H+). Study results for
25 other related PM metrics (e.g., Black Smoke; BS) are also considered, but only qualitatively,
26 primarily with respect to their coherence or lack of coherence with studies using the mass or
27 composition metrics measured in North America. In order to consider the potentially
28 confounding effects of other co-existing pollutants, study results for the various PM metrics are
29 presented both for: (1) the case when the PM metric is the only pollutant in the model; and,
30 (2) the case where a second pollutant, ozone, is also included in the model (as ozone has been
31 shown by most past studies to be the most important co-pollutant affecting respiratory
October 1999 6-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 admissions). Results from models with more than two pollutants included simultaneously are
2 not used for quantitative estimates of coefficient size or statistical strength, due to increased
3 likelihood of bias and variance inflation due to multi-collinearity of various pollutants (e.g., see
4 Harris, 1975).
5
6 6.2.3.2 Summary and Implications of Studies Assessed in the 1996 PM AQCD
7 In the 1996 PM AQCD, it was found that both COPD and pneumonia hospitalization
8 studies showed moderate, but statistically significant, relative risks in the range of 1.06 to
9 1.25 per increase of 50 //g/m3 in PM10 or its equivalent. There was a suggestion of a relationship
10 between ambient PM10 and heart disease admissions, but the estimated effects were smaller than
11 the effects for other endpoints. While a substantial number of hospitalizations for respiratory
12 illnesses occur in those >65 years of age, there are also numerous hospitalizations for those under
13 65 years of age. Several of the hospitalization studies restricted their analysis by age of the
14 individuals, but did not explicitly examine younger age groups. One exception noted was Pope
15 (1991), who reported an increase in hospitalization for Utah Valley children (aged 0 to 5) for
16 monthly numbers of admissions in relation to PM10 monthly averages, as opposed to daily
17 admissions in relation to daily PM levels used in other studies.
18 Studies examining acute associations between indicators of components of fine particles
19 (e.g., British smoke, BS; sulfates, SO4=; and acidic aerosols, H+) and hospital admissions were
20 also reported as finding significant relationships. While sulfates were especially predictive of
21 health effects, it was not clear whether the sulfate-related effects were attributable to their acidity,
22 or to the broader effects of associated combustion-related fine particles, in general.
23
24 6.2.3.3 Key New Respiratory Medical Visits Studies
25 As discussed above, medical visits include both hospital emergency department (ED) visits
26 and doctors' office visits. As in the past CD's, most of the available morbidity studies discussed
27 below are of ED visits and their associations with air pollution, but new studies look at the
28 primary care setting such as a study conducted in Paris, France which looks at doctors' visits to
29 patients in that city. This provides a new insight into the scope of air pollution's effects on
30 severe morbidity.
October 1999 6-35 DRAFT-DO NOT QUOTE OR CITE
-------
1 Delfino and colleagues (1997) examined the association between air pollution and
2 emergency department (ED) visits in Montreal, Canada from June through September 1992 and
3 1993. Air pollutants measured included: O3, PM10, PM25, the SO4= fraction of PM25, and aerosol
4 strong acidity (H+). Temporal trends, autocorrelation, and weather were controlled for in
5 time-series regressions. For 1992, no significant associations with ER visits were found, but
6 33% of the particulate data were missing. For 1993, only H+ was significant for children less
7 than 2 years of age, despite very low levels of aerosol acidity (mean effect of 4 nmol/m3= 5.0%:
8 CI = 0.4-9.6%). There were no significant associations between air pollution and ED visits for
9 persons 2-64 yrs. of age. For patients over 64 yrs. of age, 1-h maximum O3, PM10, PM25, and
10 SO4= were all positively associated with respiratory visits (p < 0.02). Effects of particles in the
11 older adults were smaller than for O3, with mean effect sizes of 16% (CI = 4-28%), 12%
12 (CI = 2-21%) and 6% (CI = 1-12%) for PM10, PM25, and SO4= , respectively. Ozone and PM10
13 levels never exceeded 67 ppb and 51 //g/m3, respectively.
14 Delfino et al. (1998) examined the relationship between the number of daily ED visits for
15 respiratory illnesses and outdoor air pollution in Montreal, Quebec (June-August, 1989-1990).
16 Air pollutants measured included 1- and 8-h maximum ozone (O3) and estimated PM2 5. Seasonal
17 and day-of-week trends, autocorrelation, temperature, and relative humidity were addressed
18 in-time series regressions. Although O3 levels never exceeded the U.S. National Ambient Air
19 Quality Standard (NAAQS) of 120 ppb (maximum day, 106 ppb), statistically significant
20 (P < 0.01) relationships were found between respiratory ER visits for patients over the age of
21 64 with both 1- and 8-h maximum O3 measured 1 day prior to the ER visit day during the
22 1989 summer. There was an association between respiratory ER visits for the elderly and
23 estimated PM25 lagged 1 day (0.1 visit///g/m3 PM25, P < 0.07), but this was found to be
24 confounded by both temperature and O3. Direct PM2 5 measurements were only available every
25 sixth day, and the estimated PM25 was derived from daily measurements of Coefficient of Haze
26 (CoH), O3, NOX, barometric pressure, and RH corrected light extinction coefficient (Bext) derived
27 from visibility data. The authors noted that, because of this, their estimated PM25 metric "should
28 obviously be viewed as an indicator of the level of photochemical smog rather than as an
29 accurate measurement of PM25". The authors concluded that their findings confirmed the
30 impression that, while air quality standards may protect the respiratory health of the general
31 population, this is not the case for especially susceptible subgroups such as the elderly.
October 1999 6-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 Lipsett et al. (1997) examined whether there was a relationship between ambient air
2 pollution in Santa Clara County, CA and emergency room visits for asthma during the winters of
3 1988-89 through 1991-1992. ED records from three acute-care hospitals were abstracted to
4 compile counts of daily visits for asthma and for a control diagnosis (gastroenteritis) for 3-mo
5 periods during each winter. Air monitoring data included daily CoH and PM10, as well as hourly
6 NO2 and O3 concentrations. Daily CoH measurements were used to predict values for missing
7 days of PM10 to develop a complete PM10 time series. Daily data were also obtained for
8 temperature, precipitation, and relative humidity. In time-series analyses using Poisson
9 regression, consistent relationships were found between ER visits for asthma and PM10.
10 Same-day NO2 was also associated with asthma ED visits, while ozone was not. Because there
11 was a significant interaction between PM10 and minimum temperature in this data set, estimates
12 of relative risks (RR's) for PM10-associated asthma ED visits were temperature-dependent.
13 A 60 //g/m3 change in PM10 (2-day lag) corresponded to RR's of 1.43 (95% CI = 1.18-1.69) at
14 20 degrees F, representing the low end of the temperature distribution, 1.27 (95% CI = 1.13-1.42)
15 at 30 degrees F, and 1.11 (95% CI = 1.03-1.19) at 41 degrees F, the mean of the observed
16 minimum temperature. ED visits for gastroenteritis were not significantly associated with any
17 pollutant variable. Several sensitivity analyses, including use of robust regressions and of
18 nonparametric methods for fitting time trends and temperature effects in the data, supported these
19 findings. The authors conclude that these results demonstrate an association between ambient
20 wintertime PM10 and exacerbations of asthma in an area where one of the principal sources of
21 PM10 is residential wood combustion.
22 Pantazopoulou et al. (1995) examined short-term exposure effects of air pollution on
23 morbidity in the greater Athens area during 1988. Data were collected on the daily number of
24 emergency outpatient visits and admissions for cardiac and respiratory causes (diagnoses at time
25 of admission) to all major hospitals. Measurements of air pollution included values for BS, CO,
26 and NO2. Statistical analysis was done using multiple linear regression models controlling for
27 potential confounding effects of meteorological and chronological variables, separately for winter
28 and summer. The daily number of emergency visits was related positively with the levels of the
29 various air pollutants, but this association reached the nominal level of statistical significance
30 only for NO2 in winter. The number of emergency admissions for cardiac and respiratory causes
31 was related to a statistically significant degree with all indices of air pollution during the winter.
October 1999 6-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 Two studies examine medical visits of children in Brazil and Chile. Lin et al. (1999) report
2 the association between air pollution and pediatric respiratory emergency visits (RESP) in
3 Sao Paulo, Brazil, from May 1991 to April 1993. The average often centrally located stations
4 were used to provide daily city-wide levels of SO2, CO, PM10, O3, and NO2. Poisson regression
5 was employed on 5-day lagged moving average based on the outcome of preliminary analyses.
6 Based on the results of this regression model, they reported a statistically significant association
7 between PM10 and pediatric emergency admission due to respiratory diseases even while
8 controlling for season, weather, and day of week (1.040; 95% CI 1.034-1.046). The inclusion of
9 other pollutants in the model did not substantially modify the estimated coefficients of PM10,
10 suggesting that the pollutant has an independent effect on RESP (1.052, 95% CI 1.042-1.063).
11 Further, quintile analysis of unadjusted PM10 data appears to have revealed a general
12 concentration-response relationship. The author states that this outcome represents a serious
13 public health problem for the children of Sao Paulo - noting that a large increase in the counts of
14 respiratory emergency visits; more than 20% - can be observed for the most polluted days.
15 In Santiago, Chile, Ostro et al. (1999) studied cohorts of children 3-15 years of age and
16 below age 2 for daily medical visits to primary health care clinics for either upper or lower
17 respiratory symptoms in relation to PM10 and ozone measurements. For children under age 2,
18 a 50 //g/m3 change in PM10 was associated with a 4 to 12% increase in lower respiratory
19 symptoms while the older group (3 to 15 years) ranged from 3 to 9%. The average of four
20 stationary monitors located downtown within a 12 km2 quadrilateral was used to obtain the daily
21 concentrations of PM10 and O3. The daily mean PM10 level was 108.6 //g/m3 (min-max 18.5 to
22 380 //g/m3). The magnitude of the effect of one pollutant was not impacted by the inclusion of a
23 second pollutant in the model. Analyses of lag periods appeared to be most significant for single
24 day lags and cumulative exposure over a 5-day period generated the strongest effect.
25 Stieb et al. (1996) examined the relationship of asthma emergency department (ED) visits
26 to daily concentrations of air pollutants in Saint John, New Brunswick, CN. Data on ED visits
27 with a presenting complaint of asthma (n = 1987) were abstracted for the period 1984-92
28 (May-September). Air pollution variables included: O3, SO2, NO2, SO4=, and TSP. Weather
29 variables included temperature, humidex, dew point, and relative humidity. Daily ED visit
30 frequencies were prefiltered to remove day-of-week and long wave trend effects, and filtered
31 values were regressed on air pollution and weather variables for the same day and the 3 previous
October 1999 6-38 DRAFT-DO NOT QUOTE OR CITE
-------
1 days. A positive, statistically significant (p < 0.05) association was observed between O3 and
2 asthma ED visits 2 days later. The O3 effect was not significantly influenced by the addition of
3 weather or other pollutant variables into the model or by exclusion of repeat ED visits. However,
4 given the limited number of sampling days available for SO4= and TSP (measured only once
5 every sixth day of this study), the authors concluded "a particulate effect could not be ruled out".
6 Other recent medical visit studies for asthma include the following: Norris et al. (1999)
7 examined emergency department visits by children (under the age of 18) to six hospitals for
8 asthma and daily air pollutant data to include PM2 5 levels in the inner city of Seattle, Washington
9 over 15 months during 1995 and 1996. A semiparametric poisson regression model yielded
10 significant association for PM25 and CO. A change of 11 //g/m3 in PM25 was associated with a
11 relative rate of 1.15 (95% CI 1.08 - 1.23). Yang et al. (1997) studied asthma emergency room
12 visits and air pollution (CO, O3, PM10) in Reno, Nevada for the period 1992-1994. Total asthma
13 visits were found to increase 33.7%, (95% CI range 6.0 - 61.5%) for every 100 ppb in the O3
14 level. No association, for CO or PM levels with asthma ER visits were found. PM10byp-
15 method had a mean of 33.6 with a min/max of 2.17 to 157.32 //g/m3 and 6-day Hi-Vol yielded a
16 mean of 38.01, min 10.2 to max 119.17. Ozone average was 51.01 range 16 to 100 ppb. Daily
17 average emergency visits for asthma averaged 1.75. In a bivariate analyses none of the three air
18 pollutants was statistically significantly associated with ER asthma visits. The concentration of
19 PM10 of CO, and O3 were correlated. Choudhury et al. (1997) examined insurance claims data in
20 Anchorage, Alaska for medical visits for asthma, bronchitis, and upper respiratory infection over
21 a three year period in relation to 24-hr PM10 concentration. No other pollutants were examined.
22 They concluded that fall is positively associated with asthma visits, but not winter, and suggest
23 that PM10 levels are less during the winter when the ground is covered with snow and ice.
24 Garty et al. (1998) report a study in Israel of emergency room visits for acute asthma
25 attacks of asthmatic children age 1 to 18 years for January 1 to December 31, 1993 in relation to
26 pollutant levels. No significant correlation with concentration of particulate was observed. The
27 authors comment that airborne particle levels (50% with size less than 10 //m fluctuated around a
28 rather constant level without outstanding peaks or troughs. An exceptionally high incidence of
29 ER visits of asthmatic children, observed during September, coincided with the beginning of the
30 school year, possibly due to an increase in the number of viral illnesses. The correlation between
31 ER visits and pollutants increased significantly when the September peak was excluded.
October 1999 6-39 DRAFT-DO NOT QUOTE OR CITE
-------
1 Tenias et al. (1998) studied the association between hospital emergency visits for asthma
2 and air pollution (black smoke, SO2, NO2, O3) in Valencia, Spain during the period 1994-95
3 using the APHEA analysis approach that takes into account potential confounding factors. For
4 an increase in smoke of 10 //g/m3 at lag 0 over the whole period a relative risk of 1.025 (95%
5 CI.O. 981 to 1.072) was found, with higher rates in cold months versus warm months. Based on
6 a three year series analysis during the cold months, when PM levels are highest, the relative risk
7 estimated for an increase of 10 //g/m3 (lag 0) was 1.049 (95% CI 1.008 to 1.093).
8 In London, Atkinson et al. (1999) studied the association between the number at daily visits
9 to accident and emergency departments for respiratory complaints and measures of outdoor air
10 pollution to include PM10, NO2, SO2 and CO. They examined different age groups and reported
11 the strongest association for children for visits for asthma, but were unable to separate the effects
12 of PM10 and SO2. Pollen levels did not influence the results.
13 Two studies examined methodological aspects of medical visit studies. Stieb et al. (1998a)
14 examined the occurrence of bias and random variability in diagnostic classification of air
15 pollution and daily cardiac respiratory emergency department visits such as asthma, COPD
16 respiratory infection and cardiac. They concluded that there was no evidence of diagnostic bias
17 in relation to daily air pollution levels. Stieb et al. (1998b) report that in a population of adults
18 visiting an emergency department with cardiac respiratory disease, fixed site sulfate monitors
19 appear to accurately reflect daily variability in average personal exposure to particulate sulfate,
20 while particulate acid was not as well represented by fixed site monitors.
21 Minshew and Towle (1999) and Mulla et al. (1998) report health outcomes related to
22 wildfires in Florida during June and July 1998. The frequency of selected conditions reported by
23 hospital emergency rooms in Volusia and Flagler counties were reported for the same period of
24 days in 1997 June 1 - July 6 to compare for the same time period during the wildfires in 1998.
25 For example, asthma reports increased from 77 to 147; bronchitis from 28 to 65; and shortness of
26 breath from 68 to 90, but no specific pollutant data were examined.
27 As mentioned at the outset of this section, there is also information from new studies that
28 look at primary care setting doctor visits. Medina et al. (1997) examined short-term relationships
29 between doctors' house calls and urban air pollution in Greater Paris for the period 1991-95.
30 Poisson regressions were employed with nonparametric smoothing functions controlled for time
31 trend, seasonal patterns, pollen counts, influenza epidemics, day-of-week, holidays, and weather.
October 1999 6-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 The relationship between asthma house calls and air pollution was stronger for children.
2 A relative risk of 1.32 [95% confidence interval (CI) = 1.17-1.47)] was observed for an increase
3 from the 5th to the 95th percentile (7-51 //g/m3) in daily concentrations of black smoke (BS).
4 The risks for 24-hr sulfur dioxide, nitrogen dioxide, and PM13 levels were in the same range.
5 Cardiovascular conditions were concluded to show weaker associations than angina
6 pectoris/myocardial infarction. Eye conditions were exclusively related to ozone (relative risk of
7 1.17 (95% CI 1.02-1.33) for the BS range 7-51 //g/m3). In two-pollutant models of asthma house
8 calls that included BS with, successively, SO2, NO2, and O3, only BS and O3 effects remained
9 stable. These results suggest that the widely documented air pollutant associations noted for
10 hospital emergency department visits are also applicable to a wider population consulting their
11 physician, rather than an emergency department.
12 Another published study looking at effects of air pollution on health in the primary care
13 setting was conducted Hajat et al. (1999), who evaluated the relationship between daily General
14 Practice (GP) doctor consultations for asthma and other lower respiratory disease (LRD) and air
15 pollution in London, UK. Time-series analysis of daily numbers of GP consultations was
16 performed, controlling for time trends, season factors, day of week, influenza, weather, pollen
17 levels, and serial correlation. Consultation data were available for between 268,718 and
18 295,740 registered patients from 45-47 London practices contributing to the General Practice
19 Research Database during 1992-94. Positive associations, weakly significant and consistent
20 across lags, were observed between asthma consultations and NO2 and CO in children, and PM10
21 in adults, and between other LRD consultations and SO2 in children. The authors concluded that
22 there are associations between air pollution and daily concentrations for asthma and other lower
23 respiratory disease in London. In children, the authors identified the most important pollutants to
24 be NO2, CO, and SO2. In adults, however, the authors concluded that the only consistent
25 association was with PM10 (30 //g/m3 RR=1.09; 95% CI=1.04-1.15). Moreover, across all of the
26 various age, cause, and season categories considered in this research, PM10 was the pollutant
27 most coherent in giving positive pollutant RR estimates for both asthma and other LRD (11 of
28 12 categories positive) in the single pollutant models considered.
29 Of these new severe morbidity studies, the two studies of doctors' visits are most
30 informative. As the authors of the London study note: "There is less information about the
31 effects of air pollution in general practice consultations but, if they do exist, the public health
October 1999 6-41 DRAFT-DO NOT QUOTE OR CITE
-------
1 impact could be considerable because of their large numbers." Indeed, the Paris doctors' house
2 calls and the London doctors' GP office visits studies both indicate that the effects of air
3 pollution, including PM, can affect many more people than indicated by hospital admissions
4 alone. In the London case, comparing the number of admissions from the authors' earlier study
5 (Anderson et al, 1996) with those for GP visits in the 1999 study (Hajat et al, 1999) indicates
6 that there are approximately 24 asthma GP visits for every asthma admission in that city.
7 In addition, a comparison of the PM10 coefficients indicates that the all ages asthma effect size
8 for the GP visits (although not statistically different) was approximately thirty percent larger than
9 that for hospital admissions. Similarly, the number of doctors' house calls for asthma
10 approximated 45/day in Paris (based on an average 9 asthma house calls in the SOS-Medocina
11 data base, representing 20% of the total; Medina et al., [1997]), versus an average 14 asthma
12 admissions/day (Dab et al., 1996), or a factor of 3 more doctors' house calls than hospital
13 admissions. Moreover, the RR for Paris asthma doctors' house calls was substantially higher
14 than asthma admissions (RR=1.32 for 43 //g/m3 BS for house calls vs. RR=1.04 per 100 //g/m3
15 BS for hospital admissions). Thus, these two new studies are coherent in supporting the
16 hypothesis that looking at only hospital admissions and emergency hospital visit effects can
17 greatly underestimate the numbers of severe respiratory morbidity events in a population due to
18 acute ambient PM exposure.
19
20 6.2.3.4 Key New Hospital Admissions Studies
21 PM Mass and Composition Studies: Several new studies have evaluated PM associations
22 with respiratory hospital admissions. These studies have examined various outcomes, including:
23 total respiratory admissions for all ages; asthma for all ages; total respiratory admissions by age;
24 Chronic Obstructive Pulmonary Disease (COPD) admissions (usually for patients > 64 yrs.), and;
25 Pneumonia admissions (for patients > 64 yrs.). While particle mass is the metric most often
26 employed as the particle pollution index in the U.S. and Canada, some of these new studies have
27 begun to examine the relative roles of various PM constituents, such as SO4=, in the PM-
28 respiratory hospital admissions association.
29 Burnett et al. (1997) evaluated the role that the ambient air pollution mix, comprised of
30 gaseous pollutants and various physical and chemical measures of particulate matter, plays in
31 exacerbating daily admissions to hospitals for cardiac diseases and for respiratory diseases
October 1999 6-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 (tracheobronchitis, chronic obstructive long disease, asthma, and pneumonia). They employed
2 daily measures of fine and coarse particulate mass, aerosol chemistry (sulfates and acidity), and
3 gaseous pollution (ozone, nitrogen dioxide, sulfur dioxide, and carbon monoxide) collected in
4 Toronto, Ontario, Canada, in the summers of 1992, 1993, and 1994. After adjusting the
5 admission time series for long-term temporal trends, seasonal variations, the effects of short-term
6 epidemics, day-of-week effects, and ambient temperature and dew point temperature, positive
7 associations were observed for all ambient air pollutants for both respiratory and cardiac
8 diseases. Ozone was the most consistently significant pollutant, and least sensitive to adjustment
9 for other gaseous and particulate measures. The PM associations with the respiratory hospital
10 admissions were significant for PM10 (RR=1.1 Ifor A=50 //g/m3; CI=1.05-1.18), PM25 (RR=1.09
11 for A=25 //g/m3; CI=1.03-1.14), PM10_25 (coarse) mass (RR=1.13 for A=25 //g/m3; CI=1.05-
12 1.21), sulfate levels (RR=1.11 for A=155 nmoles/m3=15 //g/m3; CI=1.06-1.17), and aerosol
13 acidity (RR=1.40 for A=75 nmoles/m3= 3.6 //g/m3, if as H2SO4; CI=1.15-1.70).
14 The study's authors dismissed these various PM associations with adverse health effects
15 based on subsequent regression models that included all recorded pollutants simultaneously, but
16 those simultaneous coefficients and confidence intervals may be somewhat suspect, given high
17 intercorrelations that can exist across the various pollutant coefficients in such a many-pollutant
18 model. The two-pollutant model results are probably more useful in assessing pollutant
19 interactions. The authors also concluded that the individual particle mass and chemistry could
20 not be separated as independent risk factors in the exacerbation of cardio-respiratory diseases in
21 this study. However, even after the simultaneous inclusion of ozone in the model, the
22 associations with the respiratory hospital admissions were still significant for PM10 (RR=1.10;
23 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),
24 sulfate levels (RR=1.06; CI=1.0-1.12), and aerosol acidity (RR=1.25; CI=1.03-1.53), using the
25 same increments. Of the PM metrics considered here, aerosol acidity yields the highest RR
26 estimate, despite having the lowest mass concentration increment.
27 Schwartz (1996) sought to replicate previous PM10 health effects findings in a location
28 where SO2 concentrations were relatively low, and the correlation between both airborne
29 particles and ozone with temperature was considerably less than in previous studies. Daily
30 counts of admissions to all hospitals in Spokane, WA for respiratory disease (ICD9 codes
31 460-519) for persons age 65 years and older were studied. Daily concentrations of PM10 and O3
October 1999 6-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 were each averaged across all monitors in the city. SO2 concentrations in Spokane were so low
2 that monitoring had been discontinued. Daily respiratory admission counts were regressed on
3 daily average temperature and humidity, day of the week indicators, and air pollution. Long
4 wavelength patterns were addressed using a nonparametric smooth function of day of study. The
5 U-shaped dependence of admissions on temperature and/or humidity was addressed using
6 nonparametric smooth functions of weather variables. Both same-day PM10 and two-day lagged
7 O3 were associated with increased risk of respiratory hospital admissions (RR = 1.085; 95%
8 CI = 1.036-1.136 for a 50-//g/m3 increase in PM10, and RR = 1.244; 95% CI = 1.002-1.544 for a
9 50-//g/m3 increase in peak-hour ozone). The PM10 association was insensitive to alternative
10 methods of control for weather, including exclusion of extreme temperature days and control for
11 temperature on multiple days. The O3 results were more sensitive to the approach for weather
12 control. The author noted that the magnitude of these PM10 effects, in a location where SO2 was
13 low and where the correlation between PM10 and temperature was close to zero, was similar to
14 that reported in other locations in the eastern United States and Europe, where confounding by
15 weather and SO2 might be a more substantial concern.
16 Dab et al. (1996) considered daily hospital admissions to public hospitals due to respiratory
17 causes in Paris, France during 1987-92. Air pollution was monitored by measurement of black
18 smoke (BS) (15 monitoring stations), sulfur dioxide (SO2), nitrogen dioxide (NO2), particulate
19 matter less than 13 micrometers in diameter (PM13), and ozone (O3). The statistical analysis was
20 based on a time series procedure using linear regression modeling followed by a Poisson
21 regression. Meteorological variables, epidemics of influenza, and strikes of medical staff were
22 included in the models. An increase in the mean daily concentration of PM13 of 100 //g/m3
23 increased the risk of hospital admissions due to all respiratory diseases by 4.5%
24 (CI = 1.004-1.087), and the BS association was similar (4.1% per 100 //g/m3; CI = 1.007-1.075).
25 SO2 levels consistently influenced hospital admissions for all respiratory diseases, chronic
26 obstructive pulmonary disease, and asthma, but no multiple pollutant models were presented.
27 Asthma was significantly correlated with NO2 levels, but not PM13. For the all respiratory causes
28 category, the authors found "the strongest association was observed with PM13" for both hospital
29 admissions and mortality, indicating a coherence of association.
30 Moolgavkar et al. (1997) investigated the association between air pollution and hospital
31 admissions for chronic obstructive pulmonary disease and pneumonia among the elderly in
October 1999 6-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 Minneapolis-St. Paul, MN, and Birmingham, AL, during January 1, 1986 to December 31, 1991.
2 Pollutants included were PM10, SO2, NO2, O3, and CO in Minneapolis-St. Paul, and PM10, O3,
3 and CO in Birmingham. After adjusting for temperature, day of week, season, and temporal
4 trends, a positive but non-significant association between air pollution and hospital admissions
5 for respiratory causes in Birmingham was found. In contrast, air pollution was significantly
6 associated with hospital admissions for respiratory causes in Minneapolis-St. Paul. Among the
7 individual pollutants, O3 was most strongly associated with admissions (t~4.4), and this
8 association was robust in the sense that it was little affected by the simultaneous consideration of
9 other pollutants. PM10, SO2, and NO2 were also associated with hospital admissions (but not
10 CO), although none were singled out by the authors as being more important than the others.
11 In the Minneapolis-St. Paul analysis, PM10 was significantly and positively associated with total
12 daily COPD and Pneumonia admissions among the elderly, even after the simultaneous inclusion
13 of O3. When four pollutants were included in the model (PM10, SO2, O3,and NO2), all pollutants
14 remained positively associated with hospital admissions, but only O3 remained statistically
15 significant. However, the usefulness of significance tests in such many-pollutant models is
16 suspect, given the intercollinearities of the various pollutants over time.
17 Asthma hospital admission studies conducted in various communities provide additional
18 new data. A unique study by Sheppard et al. (1999) evaluated relationships between measured
19 ambient pollutants (PM10, PM25, PM10_25, SO2, O3 and CO) in Seattle Washington and non
20 elderly (<65 years of age) hospital admission, with a principal diagnoses of asthma. Daily
21 hospital admission to local hospitals for area residents for 1987-94 were regressed on the
22 pollutants in a poisson regression model with control for time trends, seasonal variations, and
23 temperature related weather effects. They report an estimated 4-5% increase in the rate of
24 asthma hospital admission associated with an interquarterly range change in PM (19 //g/m3
25 PM10,11.8 //g/m3 PM25, and 9.3 //g/m3 PM10_25) lagged 1 day with relative rates as follows: for
26 PM10, 1.05 (95% CI = 1.02 - 1.08) for PM2 5, 1.04 (95% CI = 1.02 - 1.07) and for PM10_2 5
27 1.04 (95% CI = 1.01 - 1.07). The highest increase in risk was in the spring and fall season.
28 PM and CO were found to be jointly associated with asthma admission. Course particle mass,
29 PM10_2 5 was calculated as the difference between weighed PM10 and PM2 5 values. Daily
30 admission for asthma averaged 2.7.
October 1999 6-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 Lumley and Heagerty (1999) illustrate the effect of reliable variance estimation on data
2 from hospital admissions for respiratory disease on King County, WA for eight years (1987-94),
3 together with air pollution and weather information. However, their weather controls were
4 relatively crude (i.e. seasonal dummy variables and linear temperature terms). They concluded
5 that valid inference is possible in regression models for correlated data when the true dependence
6 structure is unknown. This study is notable for having compared sub-micron PM (PMj 0) versus
7 coarse PM1(M 0, finding significant hospital admission associations only with PMj 0.
8 Jamason et al. (1997) study weather/asthma relationships in the New York Standard
9 Metropolitan Statistical Area (SMSA) using a synoptic climatological methodology. This
10 procedure relate homogenous air masses to daily counts of overnight asthma hospital admission.
11 Air pollution is reported as having, little impact on asthma during fall and winter. During spring
12 and summer the high risk categories are association with high concentration of various pollutant
13 (i.e., PM10, SO2, NO2). In a London study, airborne pollen did not confound the analysis of air
14 pollution to (including black smoke) and daily admissions for asthma during the time period
15 1987-1992. (Anderson et al, 1998).
16 Rosas et al. (1998) report a statistical analysis of the relationships between emergency
17 admissions for asthma to a hospital in Mexico City and daily average airborne concentrations of
18 pollen, fungal spores, air pollutants (O3, NO2, SO2, and PM10) and weather factors. The analysis
19 used environmental data averaged over the day of admission and the 2 previous days. However,
20 long-wave influences were not addressed, except for the division of the data by season, which
21 likely resulted in uncontrolled confounding. Perhaps as a result, there were few statistical
22 associations between asthma admissions and air pollutants for the three age groups studied
23 (children under 15 years, adults, and seniors [adults over 59 years]) in either season.
24 Morgan et al. (1998a) studied air pollution and hospital admissions in Sydney, Australia,
25 from 1990 to 1994, using Poisson regression that allows for overdispersion and autocorrelation.
26 A light scattering index was used as the PM metric. An increase in daily maximum 1-hour
27 particulate concentration (lag 0) from the 10th to the 90th percentile was associated with an
28 increase of 3.0% (P = 0.08) in COPD admissions. Daily new particulate concentration (lag 0)
29 from the 10th to the 90th percentile were associated with a significant increase (2.8%) in heart
30 disease admissions in the elderly (65+ years). The estimates of the effects of particulate or
31 COPD and heart disease admissions were reduced in the multiple pollutant models.
October 1999 6-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 Gwynn et al. (1998) considered a two-and-a-half year period (May 1988-Oct. 1990) in the
2 Buffalo, NY region in a time-series analysis of daily mortality and hospital admissions for the
3 total, respiratory, and circulatory categories. Pollutants considered included: PM10, H+, SO4=
4 CoH, O3, CO, SO2, and NO2. The H+and SO4= concentrations were determined from daily PM25
5 samples that were, unfortunately, not analyzed for mass (in order to avoid possible acid
6 neutralization). Various modeling techniques were applied to control for confounding of effect
7 estimates due to seasonality, weather and day-of-week effects. They found multiple significant
8 pollutant-health effect associations, the most significant being between SO4= and respiratory
9 hospital admissions. When calculated in terms of the increments employed across studies in this
10 report, the various PM RR's were: H+ RR=1.06, 95% C.I.=1.03-1.09 (for A=75 nmoles/m3 =
11 3.6 //g/m3, if as H2SO4); SO4= RR=1.06, 95% C.I.=1.03-1.09 (for A=155 nmoles/m3=15 //g/m3);
12 and, PM10 RR=1.11, 95% C.I.=1.05-1.18(for A=50 //g/m3). These associations were not
13 significantly affected by the inclusion of gaseous co-pollutants into the regression model. Thus,
14 all PM components considered except CoH were found to be associated with increased hospital
15 admissions, but H+, SO4= and O3 had the most coherent associations with respiratory admissions.
16 Braga et al. (1999) studied hospital admission for children under 13 years of age in
17 Sao Paula, Brazil in 112 hospitals from October 1992 to October 1993 in relation to daily levels
18 of PM10, O3, SO2, CO, and NO2. The mean levels of PM10 observed of 70 //g/m3 were
19 associated with an increase of 12% in respiratory admissions. High degrees of interdependence
20 among co-pollutants (independent variables) were observed (PM10-SO2: 0.73; PM10-CO: 0.6;
21 PM10-NO2: 0.53). A three pollutant model (PM10, O3, CO) yielded coefficients that decreased in
22 magnitude and loss of statistical significance. The number of hospital admissions in this study
23 do not reflect the actual health demand observed in the study period, but the number of
24 procedures paid by the state. Thus, the total number of people who could be affected by
25 pollutants remain unknown.
26 Wordley et al. (1997) examined the presence and magnitude of any relation between short-
27 term variations in ambient concentrations of PM10 and hospital admissions and mortality in
28 Birmingham, UK. Air pollution data were taken from a national network monitoring station
29 between 1 April 1992 and 31 March 1994. Daily total hospital admissions for the same period
30 for asthma, bronchitis, pneumonia, chronic obstructive pulmonary disease (COPD), acute
31 ischaemic heart disease, acute cerebro-vascular disease, all respiratory conditions, and all
October 1999 6-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 circulatory conditions were obtained from the West Midlands Regional Health Authority.
2 Multiple linear regression models were constructed after adjusting for confounding factors (day
3 of week, month, linear trend, relative humidity, and temperature). Relative risks of admission at
4 various thresholds of PM10 were calculated with the model by comparing the risk of admission
5 over the threshold with the mean risk of admission over the whole period. Potential public health
6 benefits at various thresholds were also calculated with the model to predict the number of
7 admissions that could be avoided if, on each day that PM10 had exceeded that threshold, it had
8 instead been kept at the threshold level. Significant associations were found between all
9 respiratory admissions, cerebro-vascular admissions, and bronchitis admissions and PM10 on the
10 same day. Pneumonia, all respiratory admissions, and asthma admissions were significantly
11 associated with the mean PM10 values for the past three days. The effect of a 10 //g/m3 rise in
12 PM10 was estimated to represent a 2.4% increase in respiratory admissions, a 2.1% increase in
13 cerebro-vascular admissions, and a 1.1% increase in all causes mortality. Neither regression
14 results for other pollutants, nor multiple pollutant models, were presented. Other air pollutants
15 were reportedly examined, but, according to the authors, "these did not have a significant
16 association with health outcomes independent from that of PM10.
17 Jacobs et al. (1997) report that increases in rice straw burn acreage were shown to be
18 associated with hospital admissions for asthma in Butte County, C A during a decade of
19 observation. However, rice burning was not correlated with criteria pollutants (i.e., PM10 did not
20 show a statistically significant elevation for risk to admissions with asthma).
21 The results of these new PM mass studies are generally consistent with and supportive of
22 the studies presented in the last PM AQCD (U.S. Environmental Protection Agency, 1996)
23 showing significant associations between increased risk of hospital admissions and ambient PM
24 exposure, indexed by various PM metrics.
25
26 The APHEA Black Smoke Studies
27 There have been a large number of new time-series studies examining the air pollution-
28 hospital admissions association in Europe, but many of these studies have relied primarily on
29 Black Smoke (BS) as their PM metric. BS is mainly an indicator of particulate carbon
30 concentration in the atmosphere, but the relationship between airborne carbon and total mass of
31 PM overall aerosol composition varies from locality to locality, and is likely very different
October 1999 6-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 between Europe and the U.S., due largely to differences in local PM source characteristics (e.g.,
2 between gas and diesel powered automobiles). Therefore, while these European BS-health
3 effects studies are of qualitative use in evaluating the PM-health effects association, they are not
4 as useful for quantitative assessment of the health effects of PM in North America.
5 The most recent European air pollution health effects studies have been conducted as part
6 of the "Air Pollution and Health, a European Approach" (APHEA) study which considers
7 15 European cities from 10 different countries with a total population of over 25 million.
8 All studies used a standardized data collection and analysis approach which included:
9 consideration of the same suite of air pollutants (BS, SO2, NO2, SO2, and O3) and the use of
10 time-series regression addressing: seasonal and other long-term patterns; influenza epidemics;
11 day of the week; holidays; weather; and, autocorrelation (Katsouyanni et al., 1996).
12 Anderson et. al (1997) investigated the short-term effects of air pollution on hospital
13 admissions for chronic obstructive pulmonary disease (COPD) in the APHEA cities. For all
14 ages, the relative risks and 95% confidence limits (95% CL) for a 50 //g/m3 increase in daily
15 mean level of pollutant (lagged 1-3 days) were: SO2 RR=1.02 (0.98, 1.06); BS RR=1.04 (1.01,
16 1.06); TSP RR=1.02 (1.00, 1.05), NO2 RR=1.02 (1.00, 1.05); and, O3 (8 h) 1.04 (1.02, 1.07).
17 Models estimating effects of multiple pollutants simultaneously were not considered. The results
18 for particles and ozone are broadly consistent with those from North America, though the
19 coefficients for BS and TSP are substantially smaller.
20 Schouten et al. (1996) examined short-term effects of air pollution on emergency hospital
21 admissions for respiratory disease in two APHEA cities in the Netherlands during the period
22 1977-89. Black smoke did not show any clear associations with admissions in Amsterdam, while
23 in Rotterdam it was positively but not significantly related to the number of admissions. The
24 authors concluded that the results show that the relation between short-term air pollution and
25 emergency hospital admissions is not always consistent at these rather low levels of air pollution
26 (BS-24 h values averaged 14 //g/m3; range 1-84 //g/m3).
27 Spix et al. (1998) considered several years of hospital admissions data for all respiratory
28 causes ( ICD 460-519) from five West European APHEA cities (i.e., London, Amsterdam,
29 Rotterdam, Paris, Milan). The age groups studied were 15-64 yr (i.e., younger adults) and
30 65 + yr (older adults). The air pollutants studied were SO2, particles (i.e., BS or total suspended
31 particles), O3, and NO2. The most consistent and strongest finding was a significant increase in
October 1999 6-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 daily admissions for respiratory diseases (adults and elderly) with elevated levels of O3. This
2 finding was stronger in the elderly, had a rather immediate effect (same or next day), and was
3 homogeneous over cities. For PM, a meta-analysis of the various cities' results for TSP found no
4 overall significant associations in these groups for this outcome, while the daily mean of BS was
5 significant for young adults (50 //g/m3 RR=1.028; CI 1.006-1.051), but not for the elderly
6 (50 //g/m3 RR=1.020; CI 0.996-1.046). In by-season analyses, it was found that the BS
7 association in younger adults was significant in the cold season only (50 //g/m3 RR=1.04;
8 CI 1.02-1.07), while the PM-admissions association was significant for older adults in the warm
9 season (50 //g/m3 RR=1.07; CI 1.00-1.15). The authors found that the ozone results were in good
10 agreement with the results of U.S. studies, but no conclusion related to an overall particle effect
11 could be drawn based on these BS and TSP results.
12 Sunyer and colleagues (1997) analyzed urban air pollution and emergency admissions for
13 asthma during 1986-92 in the APHEA cities of Barcelona, Helsinki, Paris and London. Daily
14 counts of asthma admissions in adults (aged 15-64 years) and children (< 15 years) were
15 considered. Daily admissions for asthma increased significantly with increasing ambient levels
16 of NO2 in adults, and with NO2 and SO2 in children. For a 50 //g/m3 increase in BS, a
17 consistently positive (i.e., RR>1.0) but overall non-significant association was observed in all
18 cities for both children (RR=1.03, 95% CI 0.99-1.08) and adults (RR=1.02, 95% CI 0.99-1.06).
19 The authors indicated that their findings of less significant PM-health effects associations than
20 found elsewhere could be explained by the use of black smoke as the indicator, since presumably
21 the proportion of unmeasured biologically active non-black particles varies on a day-to-day basis.
22 The general coherence of the above-described APHEA results with other results gained
23 under different conditions strengthens the argument for causality in the air pollution-health
24 effects association. Unfortunately, the general use of the less comparable TSP and BS as PM
25 indicators in these studies diminishes the quantitative usefulness of these analyses in evaluating
26 the associations between PM and health effects elsewhere, such as in the U.S.
27
28 Low Birth Weight Studies
29 Several studies examine low birth rate and related endpoints. Bobak and Leon (1998)
30 conducted an ecological study of TSP levels and stillbirth and low birth weight (<2,500 g) in the
31 Czech Republic for 1986-88. The stillbirth rate was not associated with any air pollutants. In a
October 1999 6-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 model with all pollutants, only SO2 was significantly related to low birth weights. The author
2 notes the biological mechanisms for such an effect are not clear. Ritz and Yu (1999) report that
3 the relation between CO and low birth rate appeared more pronounced after adjustment for
4 concurrent exposure to NO2, PM10, and ozone. Between 1986 and 1991, Wang et al. (1997)
5 studied in Beijing, China, low birth weight due to pollutants (TSP, SO2) exposure during the
6 third trimester of pregnancy. The adjusted odds ratio for low birth rate (<2,500 g) was
7 1.10 (95% CI; 1.05-1.14) for every 100 //g/m3 increase in TSP. Biological mechanisms whereby
8 air pollution may be associated with low birth weights remain to be elucidated. Xu et al. (1995)
9 studied the effects of TSP and SO2 levels on preterm delivery in Beijing, China, from early
10 pregnancy until delivery in 1988. The analyses yielded a significant dose-dependant association
11 between gestational age and SO2 and TSP levels. The adjusted odds for preterm delivery for
12 each 100 //g/m3 increase in TSP was 1.10 (95% CI = 1.01-1.20). The authors concluded that
13 high levels of TSP and SO2 or a more complex pollution mixture appear to contribute to excess
14 risk of preterm delivery in this population.
15
16 6.2.3.5 Syntheses of Comparable Hospital Admissions PM10 and SO4= Studies
17 Among the studies discussed above and those summarized in the previous PM AQCD
18 (U.S. Environmental Protection Agency, 1996), those judged most useful for the quantitative
19 evaluation of the size and significance of the PM10 association with emergency hospital
20 admissions (i.e., studies that considered comparable age and cause categories) are listed in
21 Table 6-20. The available studies for SO4= are similarly listed in Table 6-21. ToofewPM25
22 studies were available to develop such a table for PM25, but SO4= is a strong indicator of PM25
23 concentrations in many locales. Also noted in each table for each study listed are the reported
24 RR estimates (and their respective 95% confidence intervals) for the various health outcome
25 categories most commonly considered by the various studies (e.g., all age total respiratory
26 admissions).
27 Tables 6-22 and 6-23 present the results of a mathematical synthesis of the various studies
28 listed in Tables 6-20 and 6-21 for hospital admissions and PM10 and SO4=, respectively. These
29 are selected because there are three or more studies with results for these outcomes and
30 pollutants, warranting an overall synthesis of results. In these syntheses, a single overall
31 combined estimate is calculated from a weighted aggregation of the available studies for each
October 1999 6-51 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-20. COMPARABLE RESPIRATORY HOSPITAL ADMISSIONS
PM,n STUDIES
Study
Thurston et al. (1994)
Dab et al. (1 996 )
Burnett et al. (I997a)
Wordley et al. (1 997)
Gwynn et al. (1998)
Thurston et al. (1994)
Dab et al. (1 996)
Wordley et al. (1997)
Schwartz (1995)
Schwartz (1995)
Schwartz (1995)
Moolgavkar et al.
(1997)
Moolgavkar et al.
(1997)
Schwartz (1996)
Schwartz (1993)
Schwartz (I994a)
Schwartz (I994b)
Schwartz (1995)
Schwartz (1993)
Schwartz (I994a)
Schwartz (I994b)
Area/Period
Toronto, Ontario
summers (86-88)
Paris, France
(1987-92)
Toronto, Ontario
(1992-94)
Birmingham,
England (92-94)
Buffalo, NY
(1988-90)
Toronto, Ontario
summers (86-88)
Paris, France
(1987-92)
Birmingham,
England
(1992-1994)
Spokane, WA
(1988-90)
New Haven, CT
(1988-90)
Tacoma, WA
(1988-90)
Minneapolis,MN
(1986-91)
Birmingham, AL
(1986-91)
Spokane, WA
1988-90
Birmingham, AL
(1986-89)
Minneapolis-St. Paul,
MN
(1986-89)
Detroit, MI
(1986-89)
Spokane, WA
(1988-90)
Birmingham, AL
(1986-89)
Minn.-St. Paul, MN
(1986-89)
Detroit, MI
(1986-891
Single Pollutant Models
50 Aig/m3
Outcome RR 195 u95
Respiratory 1.23 1.02 1.44
(all ages)
Respiratory 1.04 1.00 1.08
(all ages)
Respiratory 1.11 1.05 1.18
(all ages)
Respiratory 1.12 1.06 1.19
(all ages)
Respiratory 1.11 1.04 1.190
(all ages)
Asthma 1.14 0.94 1.34
(all ages)
Asthma 0.99 0.95 1.03
(all ages)
Asthma 1.17 1.04 1.36
(all ages)
Respiratory 1.08 1.04 1.14
>65yrs
Respiratory 1.06 1.00 1.13
>65yrs
Respiratory 1.10 1.03 1.17
>65yrs
Respiratory 1.09 1.05 1.13
>65yrs
Respiratory
>65yrs
Pneumonia 1.05 0.99 1.13
>65yrs
Pneumonia 1.09 1.03 1.15
>65yrs
Pneumonia 1.08 1.015 1.15
>65yrs
Pneumonia
>65yrs
COPD 1.17 1.08 1.27
>65yrs
COPD>65yrs 1.13 1.04 1.22
COPD>65yrs 1.10 0.99 1.23
COPD >65yrs
Two Pollutant Models
(with O3)
50 Aig/m3
RR 195 u95
1.22 0.99 1.37
1.10 1.04 1.16
1.14 1.04 1.24
1.02 0.96 1.26
1.07 1.01 1.14
1.11 1.02 1.20
1.07 1.03 1.11
1.03 0.99 1.07
1.08 1.01 1.15
1.06 1.02 1.10
1.25 1.10 1.43
1.10 1.05 1.17
October 1999
6-52
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-21. COMPARABLE RESPIRATORY HOSPITAL ADMISSIONS
SO4 STUDIES
Single Pollutant Models
Study
Burnett et al.
(1995)
Thurston et al.
(1992)
Thurston et al.
(1992)
Thurston et al.
(1994)
Gwynn et al.
(1998)
Burnett et al.
(1997)
Thurston et al.
(1992)
Thurston et al.
(1992)
Thurston et al.
(1994)
Burnett et al.
(1995)
Bates and Sizto
(1987)
Burnett et al.
(1995)
Burnett et al.
(1995)
Burnett et al.
(1995)
Area/Period
Ontario, Canada
(83-88)
Buffalo, NY
Summers
(88-89)
New York, NY
Summers
(88-89)
Toronto, Canada
Summers
(86-88)
V /
Buffalo, NY
(1988-90)
Toronto, Canada
(1992-94)
New York, NY
Summers
(88-89)
Buffalo, NY
Summers
(88-89)
Toronto, Canada
Summers
(92-94)
Ontario, Canada
(1983-88)
Ontario, Canada
(1983-88)
Ontario, Canada
(1983-88)
Ontario, Canada
(1983-88)
Ontario, Canada
(1983-88)
Outcome
Respiratory
(all ages)
Respiratory
(all ages)
Respiratory
(all ages)
Respiratory
(all ages)
Respiratory
(all ages)
Respiratory
(all ages)
Asthma
(all ages)
Asthma
(all ages)
Asthma
(all ages)
Asthma
(all ages)
Respiratory
<15yrs
Respiratory
15-64yrs
Respiratory
>64yrs
COPD
(all ages)
RR
1.04
1.07
1.01
1.11
1.06
1.11
1.02
1.02
1.11
1.04
1.03
1.04
1.04
1.06
15 Mg/m3
195
1.03
1.02
1.00
1.00
1.03
1.06
1.00
0.98
0.99
1.02
0.99
1.02
1.02
1.02
u95
1.06
1.11
1.02
1.23
1.09
1.17
1.03
1.15
1.23
1.07
1.05
1.06
1.07
1.08
Two Pollutant Models
(with O3)
15 Mg/m3
RR 195 u95
1.03 1.02 1.05
1.07 0.99 1.15
1.14 1.08 1.21
1.06 1.00 1.12
1.04 0.91 1.18
* Pollutant increment (in //g/m3) used to calculate the presented RR.
October 1999
6-53
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-22. SYNTHESIS OF COMPARABLE TIME-SERIES
HOSPITAL ADMISSIONS STUDIES' ESTIMATES OF RELATIVE
RISK DUE TO PM10 EXPOSURE
(per 50 j
Hospital
Admissions
Category
Respiratory
(all ages)
Asthma
(all ages)
Respiratory
(>64yrs.)
Pneumonia
(>65 yrs.)
COPD
(>65yrs.)
Number
of Studies
5
3
4
4
4
Single
Pooled
RR
1.10
1.07
1.08
1.08
1.14
Pollutant
Pooled
L95%
1.05
0.95
1.06
1.04
1.08
Model
Pooled
U95%RR
1.15
1.21
1.11
1.12
1.19
Two
Pooled
RR
1.12
1.02
1.07
1.07
1.15
Pollutant Model
Pooled
L95%
1.07
0.96
1.04
1.03
1.03
(with O 3)
Pooled
U95% RR
1.17
1.25
1.10
1.10
1.31
TABLE 6-23. SYNTHESIS OF COMPARABLE TIME-SERIES
HOSPITAL
ADMISSIONS STUDIES' ESTIMATES OF RELATIVE
RISK DUE TO SO4= EXPOSURE
Hospital
Admissions
Category
Respiratory
(all ages)
Asthma
(all ages)
Number
of Studies
6
4
Single
Pooled
RR
1.06
1.03
(per
Pollutant
Pooled
L95%
1.03
1.01
15 ^ug/m3)
Model
Pooled
U95%RR
10.9
1.05
Two
Pooled
RR
1.05
N/A
Pollutant Model
Pooled
L95%
1.03
N/A
(with O 3)
Pooled
U95% RR
1.07
N/A
N/A = No studies available that provide results for this category.
1 pollutant and health outcome. To obtain these combined estimates, we used a two-stage random
2 effects model approach, as suggested by DerSimonian and Laird (1986) to take into account the
3 among-studies variance.
October 1999
6-54
DRAFT-DO NOT QUOTE OR CITE
-------
1 The syntheses in Tables 6-22 and 6-23 indicate reasonably consistent RR effect sizes for
2 PM10 and SO4= (i.e., within their respective confidence intervals) across admissions categories.
3 However, the aggregate PM10 RR's are not significant for the asthma category, while there is a
4 suggestion that the PM10 RR may be larger for the elderly with COPD than for other respiratory
5 admissions categories. For SO4=, both asthma and total respiratory admissions have aggregate
6 RR estimates that are significant. Comparing the aggregate PM10 and SO4= results in Tables 6-22
7 and 6-23 collectively suggests that, although the differences between these two components'
8 RR's are not statistically significant, the aggregate PM10 respiratory admissions RR effect sizes in
9 these tables are approximately double those for SO4=. However, the concentration increment
10 employed for PM10 in these calculations (50 //g/m3) is more than three times the SO4=
11 concentration increment employed (15 //g/m3), which suggests that the SO4= effect size may be
12 larger than for PM10 overall, when viewed on an equal weight basis.
13
14 6.2.3.6 Overall Conclusions
15 The results of these new PM mass studies are generally consistent with and supportive of
16 the studies presented in the previous PM AQCD (U.S. Environmental Protection Agency, 1996).
17 Moreover, mathematical syntheses of multiple hospital admissions studies for the various age
18 and disease categories (including relevant studies from the previous CD) were conducted as part
19 of this new review. Overall, it was found that significant and reasonably consistent RR effect
20 sizes (i.e., within their respective confidence intervals) were generally found across admissions
21 categories for both PM10 and SO4=. As discussed by Hill (1965), such coherence across outcomes
22 and among multiple studies conducted in different places by different investigators are supportive
23 of the conclusion that these associations are caused by PM mass or a closely related pollutant
24 correlate.
25 Hospital admissions studies considering multiple PM components were also evaluated in
26 order to try to assess the relative roles of the various components in the reported PM-health
27 effects associations (in those studies where multiple PM components were considered). These
28 results indicated that sulfates and acidic aerosols were often among the PM metrics most strongly
29 associated with respiratory morbidity.
30 The doctors' visits studies in Paris (Medina et al., 1997) and London (Hajat et al., 1999)
31 indicate that the use of hospital admissions alone can understate the total severe morbidity effects
October 1999 6-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 of air pollution. In both Paris and London, the number of doctors' visits amount to many times
2 the number of hospital admissions. Moreover, the Paris Black Smoke RR for asthma doctors
3 visits was actually much higher than that for asthma hospital admissions (doctors' visits
4 RR=1.74 for 100 //g/m3, versus hospital admissions RR=1.04). Thus, these results support the
5 hypothesis that considering only hospital admissions and emergency hospital visit effects may
6 greatly underestimate the numbers of medical visits occurring in a population as a result of acute
7 ambient PM exposure.
8
9 6.2.4 Cardiovascular Effects Associated with Acute Ambient PM Exposure
10 6.2.4.1 Introduction
11 Very little information specifically addressing acute cardiovascular morbidity effects of PM
12 existed at the time of the 1996 PM AQCD. While the literature still remains relatively sparse, an
13 important new body of data now exists that both extends the available quantitative information
14 on the ecologic relationship between ambient pollution and hospital admissions and which, more
15 intriguingly, illuminates some of the physiological changes that may occur on the mechanistic
16 pathway leading from PM exposure to adverse cardiac outcomes.
17 This section begins with a brief summary of the conclusions that were reached in the 1996
18 PM AQCD regarding acute cardiovascular impacts of PM. Next, new studies falling into two
19 general classes are reviewed: ecologic time series studies of daily hospitalizations in relation to
20 ambient PM and other pollutants; and individual-level studies of temporal changes in
21 physiological measures of cardiac function as they relate to ambient pollution. This review is
22 followed by a discussion of several issues that are important in interpreting the available data,
23 including the identification of potentially susceptible sub-populations, the roles of environmental
24 co-factors such as weather and other air pollutants, temporal lags in the relationship between
25 exposure and outcome, and the relative importance of various size-classified PM sub-
26 components (e.g., PM25, PM10, PM10_25). In each case the extent of the available research data
27 base bearing on the issue is noted, and the current state of knowledge summarized.
28
29 6.2.4.2 Summary of Conclusions from 1996 PM AQCD
30 Just two studies were available for review in the 1996 PM AQCD that provided data on
31 acute cardiovascular morbidity outcomes (Schwartz and Morris, 1995; Burnett et al., 1995).
October 1999 6-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 Both studies were of ecologic time series design using standard statistical methods. Analyzing
2 four years of data on the > 65 year old Medicare population in Detroit, MI, Schwartz and Morris
3 (1995) reported significant associations between ischemic heart disease admissions and PM10,
4 controlling for environmental covariates. Based on an analysis of admissions data from
5 168 hospitals throughout Ontario, Canada, Burnett and colleagues (1995) reported significant
6 associations between particle sulfate concentrations, as well as other air pollutants, and daily
7 cardiovascular admissions. The relative risk due to sulfate particles was slightly larger for
8 respiratory than for cardiovascular hospital admissions. The AQCD concluded on the basis of
9 these studies that "There is a suggestion of a relationship to heart disease, but the results are
10 based on only two studies and the estimated effects are smaller than those for other endpoints."
11 (U.S. Environmental Protection Agency, 1996, p. 12-100). The AQCD went on to state that
12 acute impacts on CVD admissions had been demonstrated for elderly populations (i.e., > 65), but
13 that insufficient data existed to assess relative impacts on younger populations.
14 Also relevant to an evaluation of the acute impacts of particles on cardiovascular endpoints
15 are insights gained from time series studies of daily mortality, which, aside from the outcome
16 studied, are nearly identical in design and analysis to time series studies of hospitalizations. It is
17 also probable that acute effects of air pollution on cardiovascular hospitalizations and mortality
18 follow qualitatively similar etiologic mechanisms.
19 Several acute mortality studies reviewed in the 1996 AQCD analyzed cause-specific deaths
20 (usually total cardiovascular and total respiratory) in relation to ambient particle concentrations.
21 The AQCD noted that, in general, cause-specific analyses "reported higher estimated relative
22 risks for respiratory and cardiovascular categories than for total or other categories" (U.S.
23 Environmental Protection Agency, 1996, p. 12-349). It was noted that these findings were
24 consistent with analyses of case reports from historic air pollution episodes, like the December,
25 1952 London episode, in which the mortality impacts were greatest among the elderly and those
26 with pre-existing respiratory and/or cardiovascular disease. A comparative analysis of age- and
27 cause-specific mortality effects of particles in modern-day Philadelphia with those observed in
28 the 1952 London episode concluded that the patterns of mortality were largely consistent, once
29 the order of magnitude difference in exposure levels was taken into account (Schwartz, 1994c,d).
30 Viewed as a group, the acute morbidity and mortality studies reviewed in the 1996 AQCD
31 were thus consistent with the notion that acute health risks of PM are larger for cardiovascular
October 1999 6-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 and respiratory causes than for other causes. Given the tendency for end-stage disease states to
2 include both respiratory and cardiovascular impairment, and the associated diagnostic overlap
3 that often exists, it was not possible on the basis of these studies alone to determine which organ
4 system, if any, was most critically impacted.
5
6 6.2.4.3 Review of New Studies
7 Acute Hospitalization Studies
8 Four separate analyses of hospitalization data in Canada have been reported since 1995
9 (Burnett et al., 1995; 1997a,b; 1999). A variety of locations, outcomes, PM exposure metrics,
10 and analytical approaches were utilized in these four studies, which hinders somewhat our ability
11 to draw broad, cross-cutting conclusions from the full group of studies.
12 The first study, reviewed briefly in the 1996 AQCD, analyzed six years of data from
13 168 hospitals in Ontario, CN. Cardiovascular and respiratory hospital admissions were analyzed
14 in relation to sulfate and ozone concentrations. Sulfate lagged one day was associated with
15 cardiovascular admissions, with a percent effect of 2.8 (CI 1.8-3.8) per 13 //g/m3 without ozone
16 in the model and 3.3 (CI 1.7-4.8) with ozone included. When cardiovascular admissions were
17 split out into sub-categories, larger associations were observed between sulfates and coronary
18 artery disease and heart failure than for cardiac dysrhythmias. Sulfate associations with total
19 admissions were larger for the sub-population > 65 (3.5% per 13 //g/m3) than for those < 65 years
20 old (2.5% per 13 //g/m3). There was little evidence for seasonal differences in sulfate
21 associations.
22 Burnett et al. (1997a) analyzed daily congestive heart failure hospitalizations in relation to
23 carbon monoxide and other air pollutants (ozone, NO2, SO2, COH) in ten large Canadian cities as
24 a replication of an earlier U.S. study by Morris et al. (1995). The Canadian study expanded upon
25 the previous work both by its size (11 years data in each of 10 large cities) and also by including
26 a measure of particulate matter air pollution (coefficient of haze); no particle data were included
27 in the Morris et al. study. The study was restricted to the population > 65 years old. The authors
28 noted that all pollutants except ozone were correlated, making it difficult to statistically separate
29 them. COH, CO, and NO2 measured on the same day as admission (i.e., lag 0) were all strongly
30 associated with congestive heart failure admissions in univariate models. However, in
October 1999 6-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 multi-pollutant models, CO remained a strong predictor whereas COH did not. No gravimetric
2 particle data were included in this analysis.
3 The roles played by size-selected gravimetric and chemically speciated particle metrics as
4 predictors of cardiovascular hospitalizations were explored in an analysis of data from
5 metropolitan Toronto for the summers of 1992-1994 (Burnett et al., 1997b). The analysis
6 utilized dichotomous sampler (PM2 5, PM10, and PM10_2 5) and hydrogen ion and sulfate data
7 collected at a central site as well as ozone, NO2, SO2, CO, and COH data collected at multiple
8 sites in Toronto. Hospital admissions categories included total cardiovascular (i.e., the sum of
9 ischemic heart disease, cardiac dysrhythmias, and heart failure) and total respiratory. Model
10 specification with respect to pollution lags was completely data-driven, with all lags and
11 averaging times out to 4 days prior to admission evaluated in exploratory analyses, and "best"
12 metrics chosen on the basis of maximal t-statistics.
13 The relative risks of cardiovascular admissions were positive and generally statistically
14 significant for all pollutants analyzed in univariate regressions, but were especially significant for
15 ozone, NO2, COH, and PM10_25 (i.e., regression t-statistics > 3). Associations involving gaseous
16 pollutants were generally robust to inclusion of PM covariates, whereas the PM covariates, aside
17 from COH, were not robust to inclusion of multiple gaseous pollutants. In particular, PM25 was
18 not a robust predictor of cardiovascular admissions in multi-pollutant models. Whereas an
19 11 //g/m3 increase in PM25 was associated with a 3.1 percent increase (t=l.8) in cardiovascular
20 admissions in a univariate model, the percent effect was reduced to -0.7 (t=0.3) in a model that
21 included ozone, NO2, and SO2. COH, like CO and NO2, is generally thought of as a measure of
22 primary motor-vehicle emissions during the non-heating season. The authors concluded that
23 "particle mass and chemistry could not be identified as an independent risk factor for
24 exacerbation of cardiorespiratory diseases in this study beyond that attributable to climate and
25 gaseous air pollution."
26 Burnett et al. (1999) reported results of an ambitious attempt to explore cause-specific
27 hospitalizations for persons of all ages in relation to a large suite of gaseous and particulate air
28 pollutants using 15 years of data from Toronto. Cardiovascular admissions were split out into
29 separate categories for analysis, including dysrhythmias, heart failure, and ischemic heart disease.
30 The analysis also examined several respiratory causes, as well as cerebral vascular diseases and
31 diseases of the peripheral circulation; the latter categories were included because they should
October 1999 6-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 show associations with PM if the mechanism of action is related to increases in plasma viscosity,
2 as suggested by Peters et al. (1997d) (see study review below). The PM metrics analyzed were
3 PM2 5, PM10, and PM10_2 5 that had been estimated from daily TSP and TSP sulfate data based on a
4 regression analysis on dichotomous sampling data that were available every sixth day during an
5 eight-year subset of the full study period. This use of estimated rather than measured PM
6 components limits the interpretation of the PM results reported here. Model specification for
7 lags was again data-driven based on maximal t-statistics.
8 In multi-pollutant models, there were no significant PM associations with any of the three
9 cardiovascular hospitalization outcomes. For example, while an 18 //g/m3 increase in estimated
10 PM25 was associated with a 5.7 percent increase (t-statistic = 6.08) in ischemic heart disease
11 admissions in a univariate analysis, the PM25 association was reduced to 1.6 percent (n.s.) when
12 NO2 and SO2 were included in the model. The gaseous pollutants generally dominated most
13 regressions. There also were no associations between PM and cerebral or peripheral vascular
14 disease admissions. While these PM findings do not speak to the issue of the relative roles of
15 various size-classified PM components (because all the PM data were estimated from TSP and
16 sulfates in ways that were not made explicit), they do suggest that a linear combination of TSP
17 and sulfate concentrations does not have a strong independent association with cardiovascular
18 admissions when a full range of gaseous pollutants are also modeled. In this sense, these results
19 are generally consistent with those obtained from the summer Toronto analysis reviewed above
20 (Burnett et al, 1997b).
21 The Burnett et al. studies represent the most extensive body of results for PM in
22 conjunction with multiple gaseous pollutants. While the inconsistent use of alternative PM
23 metrics in the various analyses confuses the picture somewhat, a general finding is of lack of
24 robustness of associations between cardiovascular outcomes and PM. This was seen for COH in
25 the analysis of 10 Canadian cities (Burnett et al., 1997a), for PM25 and PM10 in the analysis of
26 summer data in Toronto (Burnett et al., 1997b), and for linear combinations of TSP and sulfates
27 (i.e., estimated PM25, PM10, and PM10_25) in the analysis of 15 years of data in Toronto (Burnett
28 et al., 1999). One exception was the associations reported between cardiovascular admissions to
29 168 Ontario hospitals and sulfate concentrations (Burnett et al., 1995), where the sulfate
30 association was robust to the inclusion of ozone. However, that study did not include CO in the
31 regressions, a gaseous pollutant often associated with cardiovascular outcomes in several studies.
October 1999 6-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 Thus, it is possible that the sulfate results of Burnett et al. (1995) represent confounding by
2 omitted pollutants. Also, although gravimetric PM variables were not robust predictors in the
3 Toronto summer analysis, CoH was a robust predictor (Burnett et al., 1997b), perhaps
4 representing the role of primary motor vehicle emissions. This contrasts however with CoH's
5 lack of robustness in the 10-city analysis (Burnett et al., 1997a).
6 No associations between PM10 and daily ischemic heart disease admissions were observed
7 by Wordley and colleagues (1997) in an analysis of two years of daily data from Birmingham,
8 UK. On the other hand, PM10 was associated with respiratory admissions and cardiovascular
9 mortality during the study period. The inconsistency of results across causes and outcomes is
10 difficult to interpret, but may relate in part to the relatively short time series analyzed. The
11 authors stated that gaseous pollutants did not have significant associations with health outcomes
12 independent of PM; however, no results were presented from models involving gaseous
13 pollutants.
14 Associations with PM10 were reported by Morris and Naumova (1998) in their analysis of
15 four years of congestive heart failure data among people > 65 years old in Chicago, IL. While
16 the analysis was directed primarily at evaluating modification by temperature of CO effects on
17 congestive heart failure admissions (building on previous results of Morris et al., 1995), results
18 were also presented for PM10, as well as for ozone, NO2, and SO2. As many as eight monitoring
19 sites were available for calculating daily gaseous pollutant concentrations; however only one site
20 in Chicago monitored daily PM10. Only same-day results were presented based on an initial
21 exploratory analysis showing strongest effects at for same-day pollution exposure (i.e., lag 0).
22 Strong and robust associations were reported between congestive heart failure admissions and
23 CO. Associations between hospitalizations and PM10 were observed in univariate regressions
24 (RR = 1.04; C.I. 1.01-1.07), but these diminished somewhat in a multi-pollutant model
25 (RR = 1.02; C.I. 0.99-1.06). Although these results suggest a more robust association with CO
26 than with PM10, the observed differences may be explained by differential exposure
27 misclassification for PM10 (monitored at one site) as compared with CO (eight sites). Thus,
28 no firm conclusions regarding PM/gaseous confounding are warranted on the basis of this study
29 alone.
30 PM10 associations with cardiovascular hospitalizations were also examined in two studies
31 by Schwartz (1997, 1999a). The 1997 study analyzed three years of daily data for Tucson, AZ
October 1999 6-61 DRAFT-DO NOT QUOTE OR CITE
-------
1 linking total cardiovascular hospital admissions for persons > 65 years old with PM10, CO,
2 ozone, and NO2. As was the case in Chicago, only one site was monitored daily PM10 while
3 multiple sites were available for gaseous pollutants. Both PM10 and CO were independently (i.e.,
4 robustly) associated with admissions while ozone and NO2 were not. The percent effect of a
5 23 //g/m3 increase in PM10 changed only slightly from 2.75 (CI 0.52-4.04) to 2.37 (CI 0.08-4.72)
6 when CO was included in the model along with PM10.
7 Schwartz (1999a) extended the analytical approach used in Tucson to an additional eight
8 U.S. metropolitan areas, limiting the analysis to a single county in each location to enhance the
9 representativeness of the air pollution data. The locations analyzed were: Chicago, IL, Colorado
10 Springs, CO, Minneapolis, MN, New Haven, CT, St. Paul, MN, Seattle, WA, Spokane, WA, and
11 Tacoma, WA. Again, the analysis focused on total cardiovascular hospital admissions among
12 persons >65 yr old. In univariate regressions, remarkably consistent PM10 associations with
13 cardiovascular admissions were observed across the eight locations, with a 25 //g/m3 increase in
14 PM10 associated with between 1.8 and 4.2 percent increases in admissions. The univariate eight-
15 county pooled PM10 effect was 2.48 percent (CI 1.81-3.14), similar to the 2.99 percent effect per
16 25 //g/m3 observed in the previous Tucson analysis. In a bivariate model that included CO, the
17 pooled PM10 effect size diminished somewhat to 1.86 percent (CI 1.01-2.71). As was the case in
18 previous studies (Morris and Naumova, 1998; Schwartz, 1997), the association between
19 cardiovascular admissions and CO was robust to the inclusion of PM10 in the model.
20 The PM10 effects were positive in all locations, and statistically significant in New Haven,
21 Chicago, St. Paul, Spokane, Tacoma, and Tucson, but not in Colorado Springs, Minneapolis, and
22 Seattle. CO effects were also positive in all locations, and significant in 7 of 9. The PM10 effect
23 in Minneapolis (2.03%, CI -1.87% to 6.09%) was much smaller than in the immediately adjacent
24 city of St. Paul (4.19%, CI 1.44% to 7.00%) although not significantly different. The CO effects
25 show the opposite relationship, larger in Minneapolis (4.09%, CI 1.59% to 6.65%) than in
26 St. Paul (0.74%, CI -1.84% to 3.39%). CO and PM10 are not strongly correlated in these 'Twin
27 Cities', however, 0.244 in Minneapolis and 0.113 in St. Paul, so that it is unlikely that
28 collinearity between PM10 and CO accounted for the differences. The PM10 effects in Seattle and
29 Tacoma are more similar (1.77% vs. 2.63%), but the CO effects are different (4.22%, CI 2.44%
30 to 6.02% in Seattle, vs. 1.84%, CI 0.24 to 3.46% in Tacoma), even though the correlation
31 between PM10 and CO is high in both cities (0.642 in Seattle, 0.676 in Tacoma).
October 1999 6-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz (1999a) argues that the low correlation between the effect size estimates and the
2 PM10-vs.-CO or PM10-versus-co-pollutant correlation coefficients is evidence that there is little
3 confounding across pollutants. However, since the health effects of air pollution may have
4 different seasonal dependence and may involve complex multivariate relationships among
5 weather and pollution in the different cities, further examination of this hypothesis may be
6 required.
7 Also relevant to the present review of associations between acute cardiovascular morbidity
8 and PM are nine recent studies of acute cardiovascular mortality (Borja-Aburto et al., 1997,
9 1998; Michelozzi et al., 1998; Morgan et al., 1998b; Ponka et al., 1998; Schwartz et al., 1996;
10 Simpson et al., 1997; Wordley et al., 1997; Zmirou et al., 1998). Acute mortality can be viewed
11 as a more severe manifestation of the same pathophysiologic mechanism, if any, that is
12 responsible for acute hospital admissions following PM exposure. All nine studies reported
13 significant associations between acute cardiovascular mortality and measures of ambient PM,
14 though the PM metrics utilized and the relative risk estimates varied across studies. PM
15 measurement methods included gravimetrically analyzed filter samples (TSP, PM10, PM2 5,
16 PM10_2 5), beta gauge (particle attenuation of beta radiation), nephelometry (light scattering), and
17 black smoke (filter reflectance). Where tested, PM associations appeared to be generally more
18 robust to inclusion of gaseous covariates than was the case for acute hospitalization studies
19 (Borja-Aburto et al., 1997, 1998; Morgan et al., 1998b; Wordley et al., 1997; Zmirou et al.,
20 1998). One study which examined multiple alternative PM metrics reported strongest
21 associations with PM25 and no associations for PM10_25 and hydrogen ion. These results for acute
22 cardiovascular mortality are qualitatively consistent with those reviewed above for hospital
23 admissions.
24
25 6.2.4.4 Individual-Level Studies of Cardiovascular Physiology
26 Several very recent studies carried out by a variety of groups have reported longitudinal
27 associations between physiologic measures of cardiovascular function and ambient PM
28 concentrations. These studies possess several important advantages over the ecologic time-series
29 studies discussed above, including the measurement of physiologic outcomes on an individual
30 basis that potentially may yield profound insights into mechanisms, as well as improved
31 assessment of individual exposures and covariates. Such studies have the capability to assess
October 1999 6-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 heterogeneity of responses across individuals and to investigate variations in susceptibility as a
2 function of individual factors such as age and pre-existing health status. Though the populations
3 studied to date have for the most part been relatively small, the effects that have observed in
4 association with PM exposures, especially those related to alterations in the balance of
5 sympathetic and parasympathetic control of the heart, offer the first compelling insights into
6 possible pathophysiologic mechanisms for PM effects on cardiovascular ill-health in humans.
7 Three independent studies have recently reported associations between repeated measures
8 of PM and changes in heart-rate variability in small panels of elderly subjects (Pope et al.,
9 1999a; Liao et al., 1999; Gold et al., 1998, 1999). Liao and colleagues (1999) studied 26 elderly
10 subjects (age range: 65-89 years) over three consecutive weeks at a retirement center in
11 metropolitan Baltimore. 18 subjects were classified as "compromised" on the basis of previous
12 cardiovascular conditions including hypertension. Daily six-minute resting heart rate data were
13 collected during which the time between sequential R-R intervals were recorded. A Fourier
14 transform was applied to the R-R interval data to enable separation of variability (i.e., variance)
15 into two major components: low frequency and high frequency for separate analysis. PM25 was
16 monitored daily both indoors and outdoors using standard methods. Regression analyses
17 controlled for inter-subject differences in average variability, in effect allowing each subject to
18 serve as his/her own control. Statistically significant associations were observed between
19 decreases in both high and low frequency heart rate variations and PM25 concentrations measured
20 indoors or outdoors. Associations were stronger for the 18 subjects with compromised
21 cardiovascular health.
22 Pope and colleagues (1999a) reported similar findings in a panel of six elderly subjects
23 (plus one younger subject suffering from Crohn's disease) selected from a larger group of
24 90 subjects who participated in a study of heart rate and oxygen saturation in the Utah Valley
25 (Pope et al., 1999b). The six elderly subjects ranged in age from 69 to 89 years and had histories
26 of cardiopulmonary disease. Subjects carried Holter monitors for up to 48 hours during different
27 weeks that varied in ambient PM10 concentrations. N-N heartbeat intervals were recorded and
28 used to calculate several measures of heart rate variability in the time domain. PM10 data were
29 obtained from three sites in the study area. Regression analysis with subject-specific intercepts
30 was performed, controlling for daily barometric pressure. Two HR measures related primarily to
31 long-term HR variability were negatively associated with same-day ambient PM10. Heart rate, as
October 1999 6-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 well as a measure related primarily to short-term HR variability, were both positively, but less
2 strongly, associated with PM10. An effect on HR was also observed in the larger cohort of
3 90 subjects (Pope et al, 1999b).
4 Gold and colleagues (1998, 1999) reported decreases in HRV among 21 active elderly
5 subjects aged 53-87 years in association with PM25 measured in the two hours prior to physical
6 exam.
7 Although differences in the processing and analysis of HR data (e.g., frequency- vs. time-
8 domain analyses) make it difficult to quantitatively compare results of the above studies, they
9 imply that PM, or some associated co-pollutant, is associated with alterations in the normal
10 balance of sympathetic and parasympathetic control of HR variability in susceptible populations
11 - i.e., the elderly and infirm. Depressions of HR variability have been associated with adverse
12 cardiac outcomes in prospective studies (Neas, 1999).
13
14
15 6.3 MORTALITY EFFECTS OF PARTICIPATE MATTER EXPOSURE
16 6.3.1 Introduction
17 The relationship of PM and other air pollutants to excess mortality has been intensively
18 studied and has played an important role in previous health risk assessments (U.S. Environmental
19 Protection Agency, 1986, 1996). Mortality is the most severe adverse health endpoint, and in
20 some ways the easiest to study. Excellent death records are maintained at every level of
21 government in almost all nations, and records are made available to academic investigators.
22 Furthermore, from a narrowly technical point of view, individual deaths are more amenable to
23 statistical analyses, since individual deaths from natural causes (typically respiratory and
24 cardiovascular diagnoses) are statistically independent except in rare extremely infectious
25 instances. Individual deaths are also non-recurring events, unlike hospital admissions or
26 respiratory symptoms.
27 In this section recent findings are evaluated for the two most important epidemiology
28 designs by which mortality is studied. Time series mortality studies are discussed in
29 Section 6.3.2, and prospective cohort studies in Section 6.3.3. The time series studies are most
30 appropriate to assessing acute responses to short-term PM exposure, although some recent work
October 1999 6-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 (not yet published) suggests that the time series data sets can be used to examine responses to
2 exposures over a longer time scale. Time series studies use a community-level response, the total
3 number of deaths each day, by age or by cause of death. The prospective cohort studies provide a
4 useful complement to the time series studies. These studies use individual health records, with
5 adjustments of survival lifetimes or hazard rates adjusted for individual risk factors, but so far
6 have provided less information about the role of exposures of extended duration.
7
8 6.3.2 Mortality Effects of Short-Term Particulate Matter Exposure
9 6.3.2.1 Summary of 1996 PM Criteria Document Findings on Unresolved Issues
10 The time-series mortality studies reviewed in the 1996 and past criteria documents
11 provided strong evidence that ambient air pollution was associated with increases in daily
12 mortality. The 1996 AQCD summarized 37 PM-mortality time-series studies (of which 13 were
13 PM10 studies) published between 1988 and 1996. The available information from these studies
14 were consistent with the hypothesis that PM is a causal agent in the mortality impacts of air
15 pollution. The PM10 relative risk estimates derived from the PM10 studies reviewed in the 1996
16 AQCD suggested that an increase of 50 //g/m3 in the 24-hr average of PM10 is associated with an
17 increased risk of premature mortality of the order of RR = 1.025 to 1.05 in the general population
18 (total deaths minus accident/injury). Higher relative risks are indicated for the elderly and for
19 those with pre-existing respiratory conditions.
20 While a large number of studies reported PM-mortality associations, there were several
21 important issues that needed to be addressed in interpreting those relative risks. The 1996
22 AQCD extensively discussed most critical issues including: (1) seasonal confounding and effect
23 modification; (2) confounding by weather; (3) confounding by co-pollutants; (4) measurement
24 error; (5) functional form and threshold; (6) harvesting and life shortening. As the issues related
25 to model specification became further clarified, increasing numbers of recent studies have
26 addressed most of the critical issues, and some of the issues were at least partially resolved, while
27 others required further investigations and additional data. The following paragraphs briefly
28 summarize the status of these issues at the time of the 1996 AQCD publication.
29 One of the most important components in time-series model specification was the
30 adjustment for seasonal cycles and other longer-term temporal trends. Not adequately adjusting
31 for these temporal trends could result in biased RRs. Residual over-dispersion and
October 1999 6-66 DRAFT-DO NOT QUOTE OR CITE
-------
1 autocorrelation also result from inadequate control for these temporal trends. Modern smoothing
2 methods allow efficient fits of these temporal trends, and minimize the statistical problems
3 although some issues of statistical methodology may require additional research. Most recent
4 studies have controlled for these seasonal and other temporal trends, and it was unlikely that
5 inadequate control for these trends seriously biased the estimated PM coefficients. The effect
6 modification by season has been examined in several studies. Season specific analyses are often
7 not feasible in small sizes (due to marginally significant PM effect size), but some studies (e.g.,
8 Samet et al., 1996; Moolgavkar and Luebeck, 1996) suggested that estimated PM coefficients
9 varied from season to season. However, it is not certain whether these results are real seasonal
10 effect modification, or due to the varying extent of correlation between PM and co-pollutants or
11 weather variables in each season.
12 While most available studies included some function of weather variables and some
13 reported sensitivity of PM coefficients to weather model specification, some speculated that
14 inadequate weather model specifications may still have ascribed the residual weather effects to
15 PM. Two PM studies (for Philadelphia, Samet et al. [1996, 1998]; for Utah Valley, Pope and
16 Kalkstein [1996]) included collaboration with a meteorologist who modeled weather effects
17 using synoptic weather categories. These studies reported that estimated PM effects were similar
18 if weather effects on mortality were fitted by synoptic weather variables or by other models. The
19 results also indicated that the synoptic weather model did not provide better model fits in
20 predicting mortality when than other weather model specifications used in past PM-mortality
21 studies. Thus, these results suggested that the reported PM effects were not explained by
22 inadequate modeling of weather effects.
23 Many PM studies have considered at least one co-pollutant in the mortality regression, but
24 increasing number of studies examined several multiple pollutants. In most cases, when PM
25 indices were significant in single pollutant models, an addition of a co-pollutant diminished the
26 PM effect size somewhat, but did not eliminate the associations. When multiple pollutant
27 models were analyzed by season, the PM coefficients were less stable, again, possibly due to
28 PM's varying correlation with co-pollutants among season, or due to smaller sample sizes in
29 seasonal subsets. In many studies, PM indices were more significant than other pollutants in
30 single and multiple pollutant models. Thus, it was concluded that PM-mortality associations
31 were not seriously distorted by co-pollutants. However, interpretation of relative significance of
October 1999 6-67 DRAFT-DO NOT QUOTE OR CITE
-------
1 each pollutant in mortality regression as relative causal strength was difficult because of lack of
2 quantitative information on relative exposure measurement/characterization errors among air
3 pollutants, as discussed below.
4 Measurement errors can influence the size and significance of air pollution coefficients in
5 time-series regression analyses. This issue is also important in assessing confounding among the
6 multiple pollutants, as varying extent of such error among the pollutants would also influence
7 corresponding relative significance. The 1996 AQCD discussed several types of such exposure
8 measurement/characterization error including site-to-site variability and site-to-person variability.
9 These errors are thought to bias the estimated PM coefficients downward in most cases.
10 However, there was not sufficient quantitative information available at the time to allow an
11 estimation of such bias.
12 The 1996 AQCD also reviewed evidence for threshold and functional form of short-term
13 PM mortality associations. Several studies indicated that montonic associations were seen below
14 the PM standards. It was considered difficult, however, to statistically identify a threshold from
15 available data because of low data density at lower concentrations, potential influence of
16 measurement error and adjustments for other covariates. Thus, the use of relative risk (rate ratio)
17 derived from the log-linear Poisson models was considered adequate.
18 The extent of prematurity of death (i.e., mortality displacement, or harvesting) in
19 PM-mortality association has important public health policy implications. At the time of the
20 1996 criteria review, only a few studies had investigated this issue. While one of the studies
21 suggested that the extent of such prematurity may be only a few days, this finding may not be
22 generalizable because the RR estimate was obtained only for identifiable PM episodes. There
23 was not sufficient evidence to suggest the extent of prematurity for non-episodic periods, from
24 which most of the recent PM relative risks were derived.
25 Only a limited number of PM-mortality studies analyzed fine particles and chemically
26 specific components of PM. The Harvard Six-Cities Study (Schwartz et al., 1996) analyzed size-
27 fractionated PM (PM25, PM10/15, and PM10/15_25) and PM chemical components (sulfates and H+).
28 The results suggested that PM2 5 was most significantly associated with mortality among the
29 components of PM. While H+was not significantly associated with mortality in this and earlier
30 analyses (Dockery et al., 1992), the smaller sample size for H+than for other PM components
31 made a direct comparison difficult. The 1996 AQCD also discussed that the mortality
October 1999 6-68 DRAFT-DO NOT QUOTE OR CITE
-------
1 associations with BS or COH reported in earlier studies in Europe and the U.S. during the 1950s
2 to 1970s most likely reflected contributions from fine particles, as those PM indices had low 50%
3 cut-off diameters (~ 4.5//m). Furthermore, there are respiratory morbidity studies that showed
4 associations between hospital admissions/visits with components of PM in the fine particle
5 range. Thus, the U.S. EPA concluded that there was adequate evidence to suggest that fine
6 particles play especially important roles in observed PM mortality effects.
7 Summarizing status of the issues raised in the 1996 AQCD: (1) The observed PM effects
8 are unlikely to be seriously biased by inadequate statistical modeling (e.g., control for
9 seasonality); (2) The observed PM effects are unlikely to be significantly confounded by weather;
10 (3) The observed PM effects may be to some extent confounded or modified by co-pollutants,
11 and such extent may vary from season to season; (3) Determining the extent of confounding and
12 effect modification by co-pollutants requires knowledge of relative exposure measurement
13 characterization error among pollutants; there was insufficient information on this;
14 (4) No evidence for a threshold for PM-mortality associations was reported. Statistically
15 identifying a threshold from existing data was also considered difficult, if not impossible;
16 (5) Some limited evidence for harvesting, a few days of life-shortening, was reported for episodic
17 periods. No study was conducted to investigate harvesting in non-episodic US data;
18 (6) A limited number of studies suggested a causal role of fine particles in PM-mortality
19 associations, but in the light of historical data, biological plausibility, and the results from
20 morbidity studies, a greater role of fine particles than coarse particles was suggested.
21 The following sections assess results from the studies that have been published since the
22 1996 AQCD. Because a large number of PM-health effects studies are currently being conducted
23 and numerous PM-related publications are expected after this draft is completed (but before
24 review of the final draft), the current assessment of the evidence here is incomplete. First, the
25 results from several studies that analyzed data from multiple cities are reviewed. Then, rather
26 than simply documenting the synopses of individual studies, the newly available results are
27 discussed with regard to the previously unresolved issues noted above.
28
29 6.3.2.2 New Multi-City Studies
30 Several studies conducted time-series analyses in multiple cities. The major advantage of
31 these studies over the meta-analyses of multiple "independent" studies is the consistency in data
October 1999 6-69 DRAFT-DO NOT QUOTE OR CITE
-------
1 handling and model specifications, thus eliminating variation due to study design. Also, many of
2 the cities included in these studies were the ones for which no time-series analyses had been
3 conducted. Therefore, unlike regular meta-analysis, they likely do not suffer from omission of
4 negative studies caused by publication bias. Furthermore, any heterogeneity of air pollution
5 effects can be systematically evaluated in multiple-city analyses. Thus, the results from multi-
6 city studies can provide most valuable evidence regarding the consistency and heterogeneity, if
7 any, of PM health effects.
8
9 APHEA Studies
10 The Air Pollution and Health: a European Approach (APHEA) project is a coordinated
11 multi-center study of the short-term effects of air pollution on mortality and hospital admissions
12 using data from 15 European cities, with a wide range of geographic, sociodemographic,
13 climatic, and air quality patterns. The obvious strength of this approach is to be able to evaluate
14 potential effect modifiers in a consistent manner. It should be noted that PM indices measured in
15 those cities were mostly British Smoke, with exception of Paris, Lyon (PM13), Barcelona (BS and
16 TSP), Bratislava, Cologne, and Milan (TSP). There have been three papers published that
17 presented either a meta-analysis or a pooled summary estimates of these multi-city mortality
18 results: (1) Katsouyanni et al. (1997) SO2 and PM results from 12 cities; (2) Touloumi et al.
19 (1997) ambient oxidants (O3 and NO2) results from six cities; (3) Zmirou et al. (1998) cause-
20 specific mortality results from 10 cities. The following briefly discuss each paper's findings.
21 Katsouyanni et al. (1997) SO2 and PM results from 12 cities. The cities included were:
22 Athens, Barcelona, Bratislava, Cracow, Cologne, Lodz, London, Lyons, Milan, Paris, Poznan,
23 and Wroclaw. In western European cities it was found that an increase of 50 //g/m3 in SO2 or BS
24 was associated with a 3% (95% confidence interval 2% to 4%) increase in daily mortality and the
25 corresponding figure for PM10 (they used conversion: PM10 = TSP*0.55) was 2% (1% to 3%).
26 In central eastern European cities the increase in mortality associated with a 50 //g/m3 change in
27 SO2 was 0.8% (-0.1% to 2.4%) and in BS (per a 50 //g/m3 change) 0.6% (0.1% to 1.1%).
28 Cumulative effects of prolonged (two to four days) exposure to air pollutants resulted in
29 estimates comparable with the one-day effects. The effects of both SO2 and BS were stronger
30 during the summer and were mutually independent. Regarding the contrast between the western
31 and central eastern Europe results, the authors speculated that this could be due to: difference in
October 1999 6-70 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposure representativeness, difference in pollution toxicity or mix, difference in proportion of
2 sensitive sub-population, and model fit for seasonal control. Bobak and Roberts (1997)
3 commented that the heterogeneity between eastern and western Europe could be explained by the
4 difference in mean temperature. They plotted the estimated RR for SO2 versus mean temperature
5 (Figure 6-1), and showed that Spearman correlation for SO2 was 0.85 (0.72 for BS). However, in
6 response to this explanation, Katsouyanni and Touloumi (1998) mentioned that they had
7 examined the source of heterogeneity, and found that other factors could apparently explain the
8 difference in estimates as well as or even better than temperature (e.g., the correlation between
9 age standardized mortality and the effect estimates was -0.92), and in light of other potential
10 explanations, they considered it premature to ascribe the difference in effect estimates to
11 temperature. They are conducting additional analyses to fully investigate the effect modifiers in
12 APHEA 2.
13 Touloumi et al. (1997) ambient oxidants results from six cities. Touloumi et al.
14 summarized the results of the short-term effects of ambient oxidants (O3 and NO2) on daily
15 deaths from all causes (excluding accidents). These studies are discussed here to provide a basis
16 of comparison with estimated SO2 or BS effects in the APHEA cities. Six cities in Central and
17 Western Europe provided data on daily deaths and NO2 and/or O3 levels. Poisson autoregressive
18 models allowing for overdispersion were fitted. Fixed effects models were used to pool the
19 individual regression coefficients when there was no evidence of heterogeneity among the cities
20 and random effects models otherwise. Significant positive associations were found between
21 daily deaths and both NO2 and O3. Increases of 50 //g/m3 in NO2 (1-hour maximum) or O3
22 (1-hour maximum) were associated with a 1.3% (95% CI 0.9-1.8) and 2.9% (95% CI 1.0-4.9)
23 increase in the daily mortality, respectively. There was a tendency for larger effects of NO2 in
24 cities with higher levels of BS. The pooled estimate for the O3 effect was only slightly reduced,
25 but the coefficient for NO2 was reduced by half (but remained significant) when BS was
26 included in the model. The authors speculated that the short-term effects of NO2 on mortality
27 may be confounded by other vehicle-derived pollutants.
28 Zmirou et al. (1998) cause-specific mortality results from 10 cities. This analysis
29 presented cause-specific mortality results for APHEA cities. Again, using Poisson
30 autoregressive models adjusting for trend, season, influenza epidemics, and weather, each
31 pollutant's relative risk was estimated in each city, and "meta-analyses" of city specific estimates
October 1999 6-71 DRAFT-DO NOT QUOTE OR CITE
-------
S 1.08
•
Spearman M3,85; P
o -"' 9 ° London
i
Cologne
Barcelona
D"
* Lodz
o Wroclaw
Bratislava
0
4 6 8 10 12
Average winter temperature (*G)
Figure 6-1. Relationship between the effect estimates from Katsouyanni et al. (1997) and
average temperature. Source: Bobak and Roberts (1997).
1
1
3
4
5
6
7
were conducted. The pooled RRs for cardiovascular mortality were 1.02 (95% CI: 1.01-1.04) per
50 //g/m3 increase in BS and 1.04 (95% CI: 1.01-1.06) per 50 //g/m3 increase in SO2 in western
European cities. The pooled RRs for respiratory mortality in western European cities were
1.04 (95% CI: 1.02-1.07) and 1.05 (95% CI: 1.03-1.07) for BS and SO2, respectively. However,
these associations were not found in the central European cities. Again, the investigators point
out the potential explanation for the difference between the western and central European cities:
smaller fraction of elderly population and likely larger exposure representativeness error in the
October 1999
6-72
DRAFT-DO NOT QUOTE OR CITE
-------
1 central European cities. The lack of consistency in NO2 - mortality associations was also
2 mentioned.
3 Urban Air Pollution Mix and Daily Mortality in 11 Canadian Cities (Burnett et al,
4 1998). The number of daily deaths for non-accidental causes were obtained in 11 cities from
5 1980 to 1991 and linked to concentrations of ambient gaseous air pollutants using relative risk
6 regression models for longitudinal count data. No PM index was included in their analyses
7 because daily PM measurements were not available. NO2 had the largest effect on mortality with
8 a 4.1% increased risk (p < 0.01), followed by O3 at 1.8% (p < 0.01), SO2 at 1.4% (p < 0.01), and
9 CO at 0.9% (p = 0.04) in multiple pollutant regression models. A 0.4% reduction in excess
10 mortality was attributed to achieving a sulfur content of gasoline of 30 ppm in five Canadian
11 cities. They compared the previously estimated percentage reduction in deaths due to PM2 5 and
12 sulfates (computed by the Canadian Health and Economics Assessment Panel based on results
13 from Harvard 6-city time-series analysis), and noted that the reductions in risk due to reduction in
14 concentrations of the mix of CO, SO2, and NO2 averaged among the five cities were 12 times
15 greater than that for sulfate and 19 times greater than for PM25. However, because the estimates
16 from PM were not based on the Canadian data, and model specification could have made
17 difference in risk estimates, a direct comparison between the risk reduction estimates for PM and
18 the gaseous pollutants may not be adequate.
19
20 6.3.2.3 New Results from Individual City Studies
21 Studies in individual cities can provide more detailed information on specific PM
22 components or source types. Identification of the chemical properties or size range of PM
23 components that are responsible for the reported PM effects would be very valuable for
24 understanding the biological mechanism of PM effects. Multiple PM components are rarely
25 measured simultaneously, but several new studies investigated this issue. Also, several studies
26 examined size specific component of PM.
27 Table 6-24 summarizes newly available individual studies (excludes studies that were
28 discussed in the 1996 PM CD) and lists name of the city, study period, the type of PM indices
29 used and their mean level, basic study description (mortality categories, covariates included, type
30 of regression, etc.), main results, the "representative" RR for the PM index used, the lag reported,
31 and the reference. While our main interest in this document is the PM effects observed in U.S.
October 1999 6-73 DRAFT-DO NOT QUOTE OR CITE
-------
o
o
r+
O
cr
o
l-l
•O
•O
•O
TABLE 6-24. SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Study Description
Results and Comments
PM RR for
total deaths
PM Lags
Reference
H
o
O
2
O
H
o
H
W
O
&
O
HH
H
W
U.S. Cities
Philadelphia, PA
1973-1988
TSP (68)
Philadelphia, PA
1974-1988
TSP (67)
Spokane, WA
1989-1995
PM10:
(dust-storm
days: 2 63;
"control" days:
42)
Ogden, Salt
Lake City, and
Provo/Orem, UT
1985-1995
PM10(32for
Ogden; 41 for
SLC; 38 for
P/O)
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 GAM
models.
Total, cardiovascular, respiratory, and by-
age mortality regressed on TSP, SO2, NO2,
O3, and CO, adjusting for temporal trends
and weather, using GAM.
Effects of high concentration of coarse
crustal particles was investigated by
comparing the 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 were
investigated in three metropolitan areas in
Utah's Wasatch Front using GAM Poisson
model and adjusting for seasonality,
temperature, humidity, and barometric
pressure. The 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.
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.
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.
No association was found between the
mortality and dust storm days on the
same day or the following day.
Salt Lake City, 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 Salt Lake City and
Provo/Orem.
Ranged from 1 day lag
n o/
~ U 70
(winter) to ~
4% (summer)
1.1% (0,2.1) 0 day lag
per 35//g/m3
increase in
TSP
0% (-19, 22)
for dust
storm days.
0 day lag
(lagged days
also reported
to have no
associations)
Moolgavkar
and Leubeck
(1996)
Kelsall et al.
(1997)
Schwartz
etal.
(1999)
12% (4.5,
20); 2.3% (0,
4.7); 1.9%
(-2.1,6.0)
per 50,ug/m3
PM10 for 0
day lag
0 day lag, but
the RRs for
longer
averaging
days
increased for
Salt Lake City
and
Provo/Orem
Pope et al.
(1999b)
-------
o
o
r+
O
cr
o
l-l
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
H
O
O
2
O
H
o
H
W
O
&
O
HH
H
W
City/Year/PM
(mean)
Dallas, TX
1990-1994
PM10 (25)
Study Description
Total, respiratory, cardiovascular, cancer,
and remaining non-accidental deaths were
related to PM10, O3, NO2, SO2, and CO,
adjusting for temperature, dewpoint, day-
of-week, and seasonal cycles (trigonometric
terms) using Poisson regression.
Results and Comments
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.
PM RR for
total deaths
-0.6% (-2.2,
1.1) per 8.3
Mg/m3 PM10
PM Lags
0 day lag
(other lags
also reported
to have no
associations)
Reference
Gamble
(1998)
King County,
WA
1990-1994
PM10 (30);
nephelometer
(0.59 bsp unit)
Santa Clara
County, CA
1989-1996
PM25(13);
PM10 (34);
PM10.2.5(ll);
CoH (0.5 unit)
; NO3 (3.0);
SO4(1.8)
Buffalo, NY
1988-1990
PM10 (24);
CoH (0.2
/1000ft);
SO4= (62
nmoles/m3)
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 pan fine particles),
SO2, and CO, adjusting for day-of-week,
month of the year, temperature and
dewpoint, using Poisson regression.
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]).
Total, circulatory, and respiratory mortality
and unscheduled hospital admissions were
analyzed for their associations with H+,
SO4=, PM10, CoH, O3, CO, SO2, andNO2,
adjusting for seasonal cycles, day-of-week,
temperature, humidity, using Poisson and
negative binomial GAM models.
Nephelometer data were not associated with
mortality. Cause-specific death analyses
suggest PM associations with ischemic heart
disease deaths. Mortality associations with
SO2 and CO were not mentioned. The mean
daily death counts were small (e.g., 7.7 for
total; 1.6 for ischemic heart disease). This
is an apparently preliminary analysis.
PM2 5 and nitrate were most significantly
associated with mortality, but all the
pollutants (except PM10_2 5) were
significantly associated in single poll.
models. In 2 and 4 poll, models with PM2 5
or nitrate, other pollutants were not
significant. The RRs for respiratory deaths
were always larger than those for total or
cardiovascular deaths. The difference in risk
between season was not significant for
PM25. The 1980-1986 results were similar,
except that CoH was very significantly
associated with mortality.
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.
1.4% (-1.3,
4.2) per
10,ug/m3
PM10
9% in one
poll, model;
10-13% in 2
poll, model;
13% in 4-
poll. model,
per 28
Mg/rn3
increase in
PM25
2.2% (0.5,
4.0) per 9.6
g/m3 PM10
avg. of 2 to 4
day lag
Levy (1998)
0 day lag
Fairley
(1999)
2 day lag
Gwynn et al.
(1999)
-------
o
o
r+
O
cr
o
l-l
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Study Description
Results and Comments
PMRR for total
deaths PM Lags
Reference
H
O
O
2
o
H
o
H
W
O
&
O
HH
H
W
Canada
Toronto
1970-1991
TSP (80);
CoH (0.42
/1000ft)
Toronto
1980-1994
TSP (60);
CoH (0.42);
S04= (9.2
Mg/m3);
PM10 (30,
estimated);
PM25 (18,
estimated)
Mexico City
Mexico-City
1990-1992
TSP (median:
204)
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.
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 GAM
Poisson model.
Total, respiratory, cardiovascular, and age-
specific (age >= 65) deaths were related to
O3, TSP, and CO, adjusting for minimum
temperature (temperature also fitted
seasonal cycles) using Poisson models.
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 a factor
with high loadings for NO2, CoH, and
CO, apparently representing automobile
factor. This factor was a significant
predictor for total, cancer,
cardiovascular, respiratory, and
pneumonia deaths.
Essentially all the pollutants were
significant predictors of total deaths in
single pollutant models, but in two
pollutant models with CO, most
pollutants' estimated RRs were reduced
(all the PM indices remained significant).
Based on the results from the co-
pollutant models and various stepwise
regressions, the authors noted that the
effects of the complex mixture of air
pollutants could be "almost completely
explained by the levels of CO and TSP".
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.
2.8% per
0 day lag
TSP
Ozkaynak
etal. (1996)
2.0% (0.7, 3.3)
per 88 ptg/m3
TSP;
2.9% (1.5, 4.4)
per 42 ptg/m3
PM10; 4.2% (2.9,
5.6)per22,ug/m3
0 day lag for
TSP and
PM10; Avg.
ofOandl
day for CoH
and PM25
Burnett et al.
(1998)
6% (3.3, 8.3) per
100,ug/m3for
total deaths
0 day lag
Borja-Aburto
etal. (1997)
-------
o
o
r+
O
cr
o
l-l
•O
•O
•O
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Mexico-City
1993-1995
PM25
(mean:27)
Mexico-City
1993-1995
Study Description
Total, respiratory, cardiovascular, other
deaths, and age-specific (age >= 65)
deaths were related to PM2 5, O3, and NO2,
adjusting for 3 -day lagged temperature
and periodic cycles, using Poisson GAM
model.
Infant mortality (avg. ~ 3/day) related to
PM2 5, O3, and NO2, adjusting for
Results and Comments
PM2 5, O3, and NO2 were associated with
mortality with different lag/averaging
periods (1 and 4 day lags; 1-2 avg.; 1-5
avg., respectively). PM25 associations
were most consistently significant. SO2
was available, but not analyzed because of
its "low" levels.
Excess infant mortality was associated
with PM2 5, but also with NO2, and O3 in
PMRR for total
deaths
1.3% and 1.4%
(0.2, 2.5) per
100//g/m3for
total deaths for
0 and 4 day,
respectively
6.9% (2.5, 11.3)
per 10 Mg/m3
PM Lags
0 day and
4 day lag
Avg. 3-5
lag days
Reference
Borja-Aburto
etal. (1998)
Loomis et al.
(1999)
PM2 5 (mean:
27.4,
temperature and smoothed time, using
Poisson GAM models.
the same average/lags. NO2, and O3
associations were less consistent in multi-
pollutant models.
H
o
O
2
O
H
o
H
W
O
&
O
HH
H
W
London, UK
1987-1992
BS(15)
London, UK
1992-1994
BS(13)
PM10 (29)
Total, cardiovascular, and respiratory
mortality series were regressed on BS, O3,
NO2, and SO2, adjusting for seasonal
cycles, day-of-week, influenza, holidays,
temperature, humidity, and autocorrelation
using Poisson model.
Total, cardiovascular, and respiratory (by
age) mortality series were regressed on
PM10, BS, O3, NO2, CO, and SO2,
adjusting for seasonal cycles, day-of-
week, influenza, holidays, temperature,
humidity, and autocorrelation using
Poisson model.
Both 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.
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
PMjQ and BS were similar for the same
distributional increment.
1.7% (0.8, 2.6)
per 14//g/m3
increase (10% to
90%) in BS
1 day lag
1.2% (0.0, 2.4) 1 day lag
per!6//g/m3 forBS
increase (10% to
90%) in BS
0.8% (-0.6, 2^.2)
per 31 yUg/m
increase (10% to
90%) in BS
Anderson
etal. (1996)
Bremner et al.
(1999)
-------
o
o
r+
O
cr
o
l-l
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Study Description
Results and Comments
PM RR for
total deaths
PM Lags Reference
oo
H
O
O
2
O
H
o
H
W
O
&
O
HH
H
W
England and
Wales, and
Greater London,
UK
PM10 (56, during
the worst heat
wave; 39, July-
August mean)
Edinburgh, UK
1981-1995
PM10(21,by
TEOM only for
1992-1995)
BS (8.7)
Birmingham,
UK
1992-1994
PM10 (apparently
beta-attenuation,
26)
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).
Both mortality (total, cardiovascular, and
respiratory) and emergency hospital
admissions (cardiovascular and
respiratory), in two age groups (<65 and
>= 65), were analyzed for their
associations with PM10, BS, SO2, NO2, O3,
and CO, using Poisson regression
adjusting for seasonal cycles, day-of-
week, temperature, and wind speed.
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.
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.
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.
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 //g/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.
2.6%
increase
for PM10 in
Greater
London
during heat
wave
NA
Rooney
etal. (1998)
1.5% (0.5,
2.5) per
10//g/m3
increase in
BS for all
cause
mortality
in age 65+
group
1.1% (-0.1,
2.1) per
10//g/m3
PM,n
Avg. of
1-3 day
lags
1 day lag
Prescott
etal. (1998)
Wordley
etal. (1997)
-------
October 1
•o
•o
ON
•o
O
^
Tj
H
o
0
2;
o
*^s
H
O
c
0
H
W
O
^
o
H
W
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY
City/Year/PM
(mean)
Rotterdam, the
Netherlands
1983-1991
TSP (median
42);
BS (median 13)
East Berlin
1981-1989
"SP" (beta
attenuation, 97)
Helsinki,
Finland
1987-1993
TSP (median
64); PM10
(median 28)
Madrid, Spain
1986-1992
"TSP" (beta
attenuation,
47 for average
of 2 stations)
Study Description
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 mortality (as well as deviations from
long-wave cycles) was regressed on SP and
SO2, adjusting for day-of-week, month, year,
temperature, and relative humidity, using
OLS, with options to log-transform
pollution, and w/ and w/o days with
pollution above 150 //g/m3.
Total and cardiovascular deaths, for age
groups < 65 and 65 +, were related to PM10,
TSP, SO2, NO2, and O3, using Poisson
model adjusting for temperature, relative
humidity, day-of-week, temporal patterns,
holiday and influenza epidemics.
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.
Results and Comments
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.
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.
No pollutant was significantly associated with
mortality from all causes or from
cardiovascular in age group (65+). Only in
age less than 65 group, PM10 was associated
with total and cardiovascular deaths with
4 and 5 day lags, respectively. The
"significant" lags were rather "spiky". O3
was also associated with cardiovascular
mortality in age under 65 group with
inconsistent signs and late and spiky lags
(negative on day 5 and positive on day 6).
TSP (1-day lag) and SO2 (3-day lagged) were
independently associated with mortality.
PM TIME-SERIES STUDIES
PM RR for
total deaths
5% (1,9) per
91 Mg/m3
TSP
6.1% per
100//g/m3
"SP"
3.5% (1.1,
5.9) per
10,ug/m3
PM10
4. 8% (1.8,
7.7) per
100//g/m3
TSP
PM Lags Reference
1 day lag Hoek et al.
(1997)
2 day lag Rahlenbeck
and Kahl
(1996)
4 day lag Ponka et al.
(other (1998)
lags
negative
or zero)
1 day lag Alberdi
Odriozola
etal. (1998)
-------
o
o
r+
O
cr
o
l-l
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Study Description
Results and Comments
PM RR for
total deaths
PM Lags Reference
oo
o
o
O
H
O
O
H
W
O
O
HH
H
W
Rome, Italy
1992-1995
TSP ("PM13"
beta
attenuation,
84)
Milan, Italy
1980-1989
TSP ("PM13"
beta
attenuation,
142)
South
America
Sao Paulo,
Brazil
1991-1992
PM10 (beta-
attenuation,
65)
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.
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 GAM
Poisson model.
Associations between intrauterine mortality
and PM10, NO2, SO2, CO, and O3 were
investigated using Poisson regression
adjusting for season and weather.
Association between ambient CO and
carboxyhemoglobin of blood sampled from
the umbilical cord of non-smoking pregnant
mothers were investigated in separate time
period.
PM13 and NO2 were most consistently 0.38% (0.09, 0 day lag Michelozzi,
associated with mortality. CO and O3 0.68) per et al. (1998)
coefficients were positive, SO2 10 ptg/m3 PM13
coefficients negative. RR estimates higher
in the warmer season. RRs similar for
in- and out-of hospital deaths.
All three pollutants were associated with 3.3% (2.4, 0 day lag Rossi etal.
all cause mortality. Cause-specific 4.3) per (1999)
analysis was conducted for TSP only.
Respiratory infection and heart failure TSP
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.
NO2, SO2, and CO were individually 4.1 % (-1.8, 0 day lag Pereira et al.
significant predictor of the intrauterine 10.4) per (?) (1998)
mortality. NO2 was most significant in 50 ptg/m3
multi-pollutant model. PM10 and O3 were PM10 for
not significantly associated with the intrauterine
mortality. There was an association mortality
between the ambient CO levels and
carboxyhemoglobin of blood sampled
from the umbilical cords.
-------
o
o
r+
O
cr
o
l-l
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Study Description
Results and Comments
PM RR for
total deaths
PM Lags Reference
oo
H
O
O
2
O
H
o
H
W
O
&
O
HH
H
W
Australia
Brisbane
1987-1993
PM10 (27, not used
in analysis)
Nephelometer (0.26
bscat/104m, size
range: 0.01-2 pan).
Sydney
1989-1993
Nephelometer (0.30
bscat/104m). Site-
specific conversion:
PM25~9;
PM10~18
Asia
Delhi, India
1991-1994
TSP (375)
Total, cardiovascular, and respiratory
deaths (also by age group) were related
to PM (nephelometer), O3, SO2, and
NO2, adjusting for seasonal cycles, day-
of-week, temperature, dewpoint,
holidays, and influenza, using Poisson
GEE to adjust for autocorrelation.
Season-specific (warm and cold)
analyses were also conducted.
Total, cardiovascular, and respiratory
deaths were related to PM
(nephelometer), O3, and NO2, adjusting
for seasonal cycles, day-of-week,
temperature, dewpoint, holidays, and
influenza, using Poisson GEE to adjust
for autocorrelation.
Total (by age group), respiratory,
cardiovascular deaths were related to
TSP, SO2, and NOX, using GEE Poisson
model (to control for autocorrelation),
adjusting for seasonal cycles
(trigonometric terms), temperature, and
humidity. 70% of all deaths occur
before age 65 (in US, 70% of deaths
occur after age 65).
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.
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.
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.
0.9% (0.3,
1.5) per 0.1
bscat/104m
nephelometer
increment
0 day lag
Simpson
etal. (1997)
2.6% (0.9,
4.2) per 14
//g/m3 PM2 5
or 28 //g/m3
PM10(10%
to 90%)
Avg. of 0
and 1 day
lags
Morgan
etal. (1998b)
2.3%
(significant at
0.05, but SE
of estimate
not reported)
per 100//g/m3
2 day lag
Cropper
etal. (1997)
-------
October ".
"vO
•O
•vO
ON
i
oo
K>
TABLE 6-24 (cont'd). SUMMARIES OF RECENTLY PUBLISHED SINGLE-CITY PM TIME-SERIES STUDIES
City/Year/PM
(mean)
Bangkok, Thailand
1992-1995
PM10 (beta
attenuation, 65)
Seoul and Ulsan,
Korea
1991-1995
TSP (beta
attenuation, 93 for
Seoul and 72 for
Ulsan)
Study Description
Total, cardiovascular, respiratory deaths
were examined for their associations
with PM10 (separate measurements
showed that -50% of PM10 was
PM2 5),using Poisson GAM model
adjusting for seasonal cycles, day-of-
week, temperature, humidity.
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.
Results and Comments
All the mortality series were associated with
PM10 at various lags. The effects appear
across all age groups. No other pollutants
were examined.
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.
PM RR for
total deaths
1%(0.5, 1.6)
per 1 0 Mg/m3
PM10
5.1% (3.1,
7.2) for
Seoul , and -
0.1% (-3. 9,
3.9) for
Ulsan, per
100//g/m3
TSP
PM Lags
3 day lag
(0 and 2
day lags
also
significant)
Avg. of 0,
I,and2
day lags
Reference
Ostro et al.
(1998)
Lee et al.
(1999)
H
o
o
2
o
H
o
H
W
O
&
O
HH
H
W
-------
1 cities, the results from other countries are certainly useful in evaluating the consistency and
2 source-type specificity, if any, of PM effects. Not all of the studies listed in Table 6-24 will be
3 discussed in detail here, but instead, the studies will be discussed in the context of issues.
4
5 6.3.2.4 New Studies on the Temporal Structure of Short-Term Effects
6 Consistency of Lags for Short-Term PM Exposure
7 Table 6-24 lists the lags for which the PM-mortality associations were reported. In most
8 studies, after the basic model (the best model with weather and seasonal cycles as covariates) is
9 developed, several pollution lags (usually 0 to 3 or 4 days) are individually introduced, and most
10 significant lag(s) is chosen for the RR calculation. While this practice may bias the chance of
11 finding a significant association, without a firm biological reason to establish a fixed
12 pre-determined lag, it appears reasonable. Due to likely individual variability in response to air
13 pollution, the apparent lags of effects observed for aggregated population counts are expected to
14 be "distributed" (i.e., symmetric or skewed bell-shape). The "most significant lag" in such
15 distributed lags is also expected to statistically fluctuate. It should also be noted that if one
16 chooses the most significant single lag day only, and if more than one lag day show positive
17 (significant or otherwise) associations with mortality, then reporting a RR for only one lag would
18 also underestimate the pollution effects. Some studies did consider several multiple-day
19 averaging of exposure variables to capture such multi-day effects, but this practice is not a
20 prevailing one.
21 An additional complication in assessing the shape of distributed lag is that the apparent
22 spread of the distributed lag may depend on the pattern of the persistence of air pollution (i.e.,
23 episodes may persist for a few days), which may vary from city to city. Also, it is possible that
24 the extent of lag and its spread may vary depending on the cause of death. For example, Rossi
25 et al. (1999) report that, in their analysis of TSP-cause specific mortality in Milan, Italy, the lags
26 varied for different cause of death (i.e., same day for respiratory infections and heart failure,
27 3-4 days for myocardial infarction and COPD). Thus, the lag for the total mortality may exhibit
28 mixed lags (weighted by the frequency of deaths in each cause). A somewhat unusual example,
29 from this perspective was reported from a recent Mexico City study (Borja-Aburto et al., 1998)
30 in which they found significant PM2 5-total mortality associations for same day and 4-day lag, but
31 not in the intervening 2 to 3 days (percent increase per 10 //g/m3 were 1.34, -0.16, 0.41, 0.43,
October 1999 6-83 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
1.36, 0.99, for 0 through 5 day lags, respectively). The authors hypothesize that, "This
phenomenon is consistent with both a harvesting of highly susceptible persons on the day of
exposure to high pollution levels and a lagged increase in mortality due to delayed effects of
reduction of pulmonary defenses, cardiovascular complications, or other homeostatic changes
among less-compromised individuals". However, the 4-day lagged effects are certainly not the
most frequently reported lag.
Figure 6-2 shows the distribution of the reported lags from Table 6-24 as well as from
Table 12-25 in the 1996 AQCD. It can be seen that the same day and 1-day lag are the most
frequently reported lags. This is also consistent with the immediate effects observed in the
1952 London Smog episode.
CD
GO
—
o
15n
10-
5 -
0 J
LXXXXXXXI
0
Lag (day)
Figure 6-2. Frequency distribution of the lag day for which PM RRs were computed in
33 studies (from Table 6-24 and Table 12-25 in the 1996 PM CD).
Multiple-day averaged lags were omitted.
October 1999
6-84
DRAFT-DO NOT QUOTE OR CITE
-------
1 New Assessments of Mortality Displacement
1 There have been a few studies that investigated the question of "harvesting", a phenomenon
3 in which a deficit in mortality results following the days with (pollution-caused) elevated
4 mortality, due to the depletion of susceptible population pool. The issue is important in
5 interpreting the public health implication of the reported short-term PM mortality effects. In the
6 1996 AQCD, suggestive evidence was observed (Spix et al., 1993) during a period when the air
7 pollution levels were relatively high. Recent studies generally used data from areas with lower,
8 non-episodic pollution levels.
9 Schwartz (1999b) separated time-series air pollution, weather, and mortality data from
10 Boston, MA, into three components: (1) seasonal and longer fluctuations; (2) "intermediate"
11 fluctuations; (3) "short-term" fluctuations. By varying the cut-off between the intermediate and
12 short term, he sought the evidence of harvesting. The idea is, for example, if the extent of
13 harvesting were a matter of a few days, then associations between weekly average values of
14 mortality and air pollution (controlling for seasonal cycles) would not be seen. He reported that,
15 for COPD, there was evidence that most of the mortality was only displaced by a few months; for
16 pneumonia, heart attacks, and all cause mortality, the effect size increased as the longer time
17 scales were included. The percent increase in deaths associated with a 10 //g/m3 increase in
18 PM25 increased from 2.1% (95%CI: 1.5, 4.3) to 3.75% (95%CI: 3.2, 4.3).
19 Zeger et al. (1999) first illustrated, through simulation, the implication of harvesting for PM
20 regression coefficients (i.e., mortality relative risk) as observed in frequency domain. Three
21 levels of harvesting, 3 days, 30 days, and 300 days, were simulated. As expected, the shorter the
22 harvesting, the larger the PM coefficient in the higher frequency range. However, in the real data
23 from Philadelphia, the regression coefficients increased toward the lower frequency range,
24 suggesting that the extent of harvesting, if it exists, is not in the short-term range. Zeger et al.
25 suggested that "harvesting-resistant" regression coefficients can be obtained by excluding the
26 coefficients in the very high frequency range (to eliminate short-term harvesting) and in the very
27 low frequency range (to eliminate seasonal confounding). Since the observed frequency domain
28 coefficients in the very high frequency range were smaller than those in the mid frequency range,
29 eliminating the "short-term harvesting" effects would only increase the average of the
30 coefficients in the rest of the frequency range.
October 1999 6-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 Frequency domain analyses are rarely performed in air pollution health effects studies,
2 perhaps except the spectra analysis (variance decomposition by frequency) to identify seasonal
3 cycles. Examinations of the correlation by frequency (coherence) and the regression coefficients
4 by frequency (gain) may be useful in evaluating the potentially frequency-dependent
5 relationships among multiple time series. A few past examples in air pollution health effects
6 studies include: (1) Shumway et al. (1983) analysis of London mortality analysis in which they
7 observed that significant coherence occurred beyond two week periodicity (they interpreted this
8 as "pollution has to persist to affect mortality); (2) Shumway et al. (1988) analysis of Los Angels
9 mortality data in which they also found larger coherence in the lower frequency; (3) Ito (1990)
10 analysis of London mortality data in which he observed relatively constant gain (regression
11 coefficient) for pollutants across the frequency range, except the annual cycle. These results also
12 suggest that associations and effect size are, at least, not concentrated in the very high frequency
13 range.
14 Both Schwartz (1999b) and Zeger et al. (1999) analyses suggest that the extent of
15 harvesting, if any, is not a matter of few days. Other past frequency domain studies are also at
16 least qualitatively in agreement with the evidence against the short-term only harvesting. Since
17 very long wave cycles (> 6 months) need to be controlled in time-series analyses, it is not
18 possible to estimate the extent of harvesting beyond 6 months periodicity in a time-series study
19 design. While these studies suggest that observed short-term associations are not simply due to
20 short-term harvesting, more data are needed to quantify prematurity of deaths.
21
22 Santa Clara County, CA. Fairley (1999) conducted a time-series analysis of mortality-air
23 pollution relationship in Santa Clara County, CA for years 1989-1996. His previous analysis of
24 this locale (Fairley, 1990) showed an association between Coefficient of Haze (CoH) and
25 mortality in the same County for 1980-1986 period. Fairley provides useful information
26 regarding the type of air pollution in the study area. In contrast to Eastern or Midwestern cities,
27 SO2 levels are so low (<10 ppb) that it is no longer measured. Consequently, sulfate is low, and it
28 represents only 5% of PM2 5 (in contrast to up to 45% in Eastern U.S.). Also, unlike Eastern
29 cities where fine particles are high during summer due to sulfate levels, in Santa Clara County,
30 fine particles are much higher in winter (25 //g/m3 in winter vs. 10 //g/m3 during the rest of the
31 year) due to contributions from wood burning and ammonium nitrate.
October 1999 6-86 DRAFT-DO NOT QUOTE OR CITE
-------
1 Total, cardiovascular, and respiratory deaths were regressed on PM10, PM2 5, PM10_2 5, CoH,
2 nitrate, sulfate, O3, CO, NO2, adjusting for trend, season, and min and max temperature, using
3 GAM Poisson models. Season-specific analysis was also conducted. The same approach was
4 also used to re-analyze 1980-1986 data (previously analyzed by Fairley, 1990). PM25 and nitrate
5 were most significantly associated with mortality, but all the pollutants (except PM10_2 5) were
6 significantly associated in single pollutant models. In multiple pollutant models with PM2 5 or
7 nitrate, other pollutants were not significantly associated with mortality. The RRs for respiratory
8 deaths were always larger than those for total or cardiovascular deaths. The difference in risk
9 between season was not significant for PM25. The 1980-1986 results were similar, except that
10 CoH was a very significantly associated with mortality, consistent with the results from the
11 author's 1990 analysis.
12 This study presents the first evidence of the mortality effects of directly measured PM2 5 in
13 the West Coast. While other studies of PM in the West Coast (PM10 in Los Angeles [Kinney
14 et al., 1995]; visibility-derived PM25 in San Bernardino and Riverside Counties [Ostro, 1996])
15 indicated larger effect estimates in summer, Fairley's result indicates that the estimated PM2 5
16 coefficients were relatively constant across season. This may be in part due to the difference in
17 air pollution mix between the Los Angeles Metropolitan area and Santa Clara County (i.e.,
18 San Francisco Bay area). Fairley's results also indicated that the coarse fraction of PM10
19 (PM10_25) was not a significant predictor of mortality, consistent with Schwartz et al. (1999)
20 findings from six-cities study.
21
22 Mexico City Studies. There have been three time-series mortality studies in Mexico that
23 examined PM indices: (1) TSP-mortality study for years 1990-1993 (Loomis et al., 1996;
24 Borja-Aburto et al., 1997); (2) PM2 5-mortality (total, cardiovascular, and respiratory) for years
25 1993-1995 (Borja-Aburto et al., 1999a); and (3) PM2 5-infant (children less than 1 year of age)
26 mortality study 1993-1995 (Borja-Aburto et al., 1999b). The TSP study (the study focus was on
27 O3) considered O3, SO2, and CO as co-pollutants, and the PM25 studies considered O3 and NO2.
28 In the PM2 5 studies, SO2 was available, but not analyzed because the concentration was
29 "comparable to the those in the cities wit lowest levels". These studies employed Poisson
30 models with adjustment for temperature and long-term trends, but the TSP study employed the
31 iteratively weighted and filtered least-square method to control for autocorrelation and
October 1999 6-87 DRAFT-DO NOT QUOTE OR CITE
-------
1 over-dispersion, while the PM2 5 studies used generalized additive models to model mortality as a
2 smooth function of time, which should also remove autocorrelation and over-dispersion.
3 In the TSP results (Loomis et al., 1996), the RRs for total mortality using single pollutant
4 models were: 1.049 (95% CI 1.030, 1.067) per 100 //g/m3 increase in TSP; 1.029 (95% CI 1.015,
5 1.044) per an increase of 100 ppb in one-hour maximum O3; and, 1.075 (95% CI 0.984, 1.062)
6 per 100-ppb increase for SO2. CO was only weakly associated with mortality, and was not
7 considered further in multiple pollutant models. When all three pollutants were considered
8 simultaneously, only TSP remained associated with mortality, indicating excess mortality of 5%
9 per 100 //g/m3 increase [RR = 1.052, 95% CI 1.034, 1.072]. Excess mortality was larger and
10 more significant for persons over 65 years of age. Addition of SO2 to the TSP model did not
11 change the TSP coefficient. Air pollution levels in Mexico City were much higher than U.S.
12 levels. For example, the 25th percentile of daily 1-hr maximum ozone was 122 ppb; the median
13 for TSP, SO2, CO, and O3 (all daily mean) were 204 (//g/m3), 53 (ppb), 5.8 (ppm), and 54 (ppb),
14 respectively. The authors concluded: ".... it is difficult to attribute the observed effects to a
15 single pollutant. The technical feasibility and scientific validity of isolating the effect of single
16 pollutants in such complex mixtures requires further research and careful consideration".
17 In the PM2 5- mortality (total, cardiovascular, and respiratory) analysis, PM2 5, O3, and NO2
18 were associated with mortality with different lag/averaging periods (1 and 4 day lags; 1-2 d avg.;
19 1-5 d avg., respectively). PM25 associations were most consistently significant. As mentioned
20 previously, the authors interpret this pattern of 1 and 4 day lag associations as the immediate
21 harvesting of highly susceptible people and delayed effects on less compromised individuals
22 (2 and 3 day lags were not significant). EPA estimates, based on this result, in a single pollutant
23 model, a 25 //g/m3 increase in PM25 was associated with a 3.5 percent increase in total mortality
24 both on the current day and four days after exposure (95% CI=0.5-6.3 percent). Inclusion of O3
25 and NO2 in a three-pollutant model somewhat increased the estimated PM25 effect: 4.2 percent
26 (95% CI= 0.6-7.9). A 10 ppb increase in the mean of 1 and 2 day lagged O3 was associated with
27 a 1.8 percent increase in cardiovascular disease.
28 In the PM2 5 - infant mortality analysis, excess infant mortality was associated with PM2 5,
29 but also with NO2, and O3 in the same lag pattern (3 to 5 day lag). NO2, and O3 associations were
30 less consistent in multi-pollutant models. The RR calculated for infant mortality per 10 //g/m3
October 1999 6-88 DRAFT-DO NOT QUOTE OR CITE
-------
1 increase in PM25 (average of 3 to 5 day lags) was 6.9% (2.5, 11.3). In three pollutant models, it
2 was 6.3% (-0.5, 13.2).
3 These Mexico City studies collectively suggest significant mortality associations with PM
4 indices, but there are some notable differences. In the TSP analyses, the authors found "no
5 independent effect of O3", while in the PM25 study, the O3-cardiovascular mortality association
6 remained significant in two and three pollutant models. This may be, in part, due to the smaller
7 number of sample days available for multi-pollutant models in the TSP analysis (n=211 days,
8 compared to n>800 days for PM25 analysis). The weaker associations for gaseous pollutants in
9 these studies may also be partly explained by poorer spatial uniformity for gaseous pollutants, as
10 the spatial correlation reported in the TSP analysis indicate better site-to-site correlation for TSP
11 (r ~ 0.85) than for gaseous pollutants (r ~ 0.5).
12
13 Northeastern United States/Eastern Canada: Summer Haze and Automobile
14 Toronto, Ontario and Buffalo, NY are relatively close in distance and both experience the
15 same regional summer haze pollution, which contains O3 and acid aerosols/sulfate. The studies
16 from these two locales will be discussed and contrasted in the following paragraphs.
17
18 Toronto, Ontario. The main focus of the Burnett et al. (1998) study was on CO effects on
19 mortality in metropolitan Toronto during 1980-1994, but their analysis also considered NO2, O3,
20 SO2, TSP, CoH, SO4=, and estimated PM10 and PM25 After adjusting for day-of-week,
21 nonparametric smooth function of day of study, and weather variables, all of these pollutants,
22 except O3, were significantly positively associated with nonaccidental mortality in one pollutant
23 models. In two pollutant models with CO as co-pollutant, these pollutants' coefficients, as well
24 as CO's, were reduced, although all PM indices' coefficients remained significant. The
25 correlation between CO and other pollutants ranged from -0.23 (O3) to 0.56 (CoH), but the
26 smallest positive correlation was with TSP (0.19). CO and TSP were selected for inclusion in the
27 final model using the stepwise procedures. In the two pollutant model with TSP and CO, the
28 excess risk for TSP was 1.5% (0.2, 2.8) per 88//g/m3 increase (5th to 95th percentile range). The
29 authors mention that the vast majority (86%) of emissions for CO in Toronto are from vehicular
30 sources, while only a small fraction of TSP (21%) was attributed to vehicles. Thus, air pollution
31 from vehicular sources, CO in particular, was suggested as a cause of increased mortality in this
October 1999 6-89 DRAFT-DO NOT QUOTE OR CITE
-------
1 locale. This is consistent with the results from earlier Ozkaynak et al. (1996) analysis of Toronto
2 data 1970-1991 (discussed in Section 6.3.2.9).
3 While various PM indices were considered in this analysis, PM10 and PM2 5 were available
4 only every-6th-day during 1984-1990 (total of 272 days during the 15 year study period). Since
5 the missing PM10 and PM25 values were imputed using daily values of TSP sulfates (which may
6 suffer from artifact due to the TSP glass fiber filter), TSP, and CoH, it is difficult to compare the
7 estimated PM10 and PM2 5 effects with those of other PM components. However, in the
8 unimputed data, sulfates were strong predictors of PM2 5 (R2=0.77), and TSP was a weak
9 predictor of PM25 (R2=0.22), a moderate predictor of PM10 (R2=0.50), and a stronger predictor of
10 PMjQ.2.5 (R2=0.63). Thus, the estimated PM10 and PM2 5 may have adequately represented daily
11 fluctuations of the thoracic and fine components. The RRs for these estimated PM10 and PM2 5
12 were significant in single pollutant models and with CO in the model.
13
14 Buffalo, NY. Gwynn et al. (1998) analyzed a two and a half year record of daily H+ and
15 SO4= measurements collected in the Buffalo, NY region. Their analysis of respiratory,
16 circulatory, and total daily mortality and hospital admissions also considered PM10, CoH, O3, CO,
17 SO2, and NO2. Poisson and negative binomial regression models were employed, adjusting for
18 seasonality, weather, and day-of-week. For total mortality, all the PM components were
19 significantly associated, with H+ being the most significant and CoH the least significant
20 predictors. The gaseous pollutants were mostly weakly associated with total mortality. The
21 effect size estimated for respiratory mortality with inter-quartile-ranges of these PM components
22 were 2 to 3 times larger than those for total mortality (except CoH), but were less significant due
23 to the large standard error of coefficients (from the small daily counts). Parallel analyses of
24 hospital admission and mortality data in this study allowed an examination of "coherence", or the
25 consistency check for causality suggested by Bates (1992). Gwynn et al. noted that H+, SO4=, and
26 O3 showed the most coherent associations with respiratory hospital admissions and respiratory
27 mortality (RRs for respiratory mortality, H+: RR=1.55 per 346 nanomole/m3, 95% CI=1.16-2.07;
28 SO4=: RR=1.24 per 329 nanomoles/m3, 95% CI=1.05-1.47; O3: RR=1.16 per 61 ppb daily
29 average, 95% CI=1.02-1.33, calculated for maximum - mean increment), lending support to the
30 theory of a "summer haze effect" (Bates and Sizto, 1987).
October 1999 6-90 DRAFT-DO NOT QUOTE OR CITE
-------
1 The Toronto results (Burnett et al., 1998) suggest the importance of motor vehicle related
2 primary pollutants, although all the PM components were significant predictors of mortality in
3 both single and two pollutant models. The secondary regional pollutant, O3 (which was the only
4 pollutant that was negatively correlated with CO), was not associated with mortality in Toronto.
5 In contrast, in Buffalo, it was secondary "summer haze mix", O3, SO4=, and H+, that showed the
6 most "coherent" effects in mortality and morbidity. Automobile related primary pollutants, CO,
7 CoH, and NO2 were not as strongly associated with mortality as the "summer haze mix" in
8 Buffalo. Since Buffalo and Toronto are geographically close, the levels of "summer haze mix"
9 were relatively comparable. One possible reason for the apparent discrepancy between the
10 results from these two cities may be explained by the relative contributions of the primary
11 (automobile) pollution and the regional summer haze mix. In fact, the daily average CO level in
12 Toronto (1.2 ppm) was 70% higher than that in Buffalo (0.7 ppm). Also, another primary
13 vehicle-related pollutant, CoH, was twice as high in Toronto as in Buffalo (0.42 vs. 0.2,
14 respectively, 1,000 linear feet). It is also possible that the relative spatial representativeness of
15 sites for primary pollutants used in these cities may have been different. Measurements of
16 primary pollutants are likely more influenced by strong local source impact than the regional
17 secondary pollutants, and therefore, the location of monitor is more crucial.
18 In studies where multiple PM components were examined, in most cases, all the PM
19 components were significantly associated with mortality. Unless mutually exclusive size
20 fractionated PM components (i.e., PM25 vs. PM10_25) are examined, as in Schwartz et al. (1996)
21 6 city time-series analysis, establishing size dependency of PM effects remains difficult.
22 Examination of the role of acid aerosols on mortality in the U.S. has been difficult partly
23 because the number of days available for acid aerosol measurements were smaller than that for
24 other PM components, as was the case for the six city time-series data. Gwynn et al. (1999)
25 analysis of Buffalo data used comparable sample sizes for all PM components, and their results
26 are suggestive of the role of acidic particles, but distinguishing the individual pollutant effects of
27 the "summer haze mix" was not possible. These results suggest that, while it is difficult to
28 identify "responsible" PM components, identification of a group of pollutants that represents a
29 certain source type or pollution mix is useful.
30
October 1999 6-91 DRAFT-DO NOT QUOTE OR CITE
-------
1 New Studies on Crustal Particle Effects
1 In the 1996 AQCD, the only study that analyzed both fine and coarse particles (Schwartz
3 et al., 1996) at that time suggested that fine particles (PM25), but not coarse particles (PM10_25)
4 were associated with daily mortality. Since then, a few studies investigated the effects of coarse
5 particles, as identified as crustal wind-blown particles, or crustal particles within fine particles.
6 Schwartz et al. (1999) investigated the association of coarse particle concentrations with
7 non-accidental deaths in Spokane, Washington, where dust storms elevate coarse particle
8 concentration. During the 1990-1997 period, 17 dust storm days were identified. The average
9 PM10 levels during those storms were 263 //g/m3, compared to 39 //g/m3 for the entire period.
10 The coarse particle domination of PM10 data on those dust storm days was confirmed by a
11 separate measurement of PM10 and PMj during a dust storm in August, 1996: the PM10 level was
12 187 //g/m3, while PMj level was only 9.5 //g/m3. The deaths on the day of a dust storm were
13 contrasted with deaths on control days (n=95 days in the main analysis and 171 days in the
14 sensitivity analysis), which are defined as the same day of the year in other years when dust
15 storms did not occur. The relative risk for dust storm exposure was estimated using Poisson
16 regressions, adjusting for temperature, dewpoint, and day of the week. Various sensitivity
17 analyses considering different seasonal adjustment, year effects, and lags, were conducted. The
18 expected relative risk for these storm days with an increment of 221 //g/m3 would be about 1.20,
19 based on PM10 relative risk from past studies, but the estimated RR for high PM10 days was
20 1.00 (95% CI=0.81-1.22). Schwartz et al. concluded that there was no evidence to suggest that
21 coarse particles in the Spokane summer dust storms were associated with daily mortality.
22 Pope et al. (1999b) investigated PM10-mortality associations in three metropolitan areas
23 (Ogden, Salt Lake City, and Provo/Orem) in Utah's Wasatch Front mountain region during
24 1985-1995 period. While the three metropolitan areas shared common weather pattern, pollution
25 levels and patterns among the three areas were different due to different emission sources. They
26 ingeniously utilized the index of air stagnation, a clearing index (the National Weather Service
27 computes this index from temperature, moisture and wind), to identify and screen obvious
28 windblown dust days, as clearly identified as high PM10 days on the days with low air stagnation
29 index. They found that Salt Lake City experienced substantially more episodes of wind-blown
30 dusts. They therefore conducted Poisson regression of mortality series using both unscreened
31 and screened PM10 data. The effects of screening was most apparent in Salt Lake City results.
October 1999 6-92 DRAFT-DO NOT QUOTE OR CITE
-------
1 After screening, the RRs per 10 //g/m3 increase in PM10 for mortality in three metropolitan areas
2 were 1.6% (0.3 - 2.9), 0.8% (0.3 - 1.3), and 1.0% (0.2 -1.8) for Ogden, Salt Lake City, and
3 Provo/Orem, respectively. These results suggest that the pollution episodes of wind-blown
4 (crustal-derived) dusts were less associated with mortality than were the episodes of
5 (presumably) combustion-related particles.
6 Laden et al. (1999) analyzed the Harvard Six-Cities data to investigate the role of crustal
7 particles in fine particles on daily mortality. The elemental abundance data (from X-ray
8 fluorescence spectroscopy analysis of daily filters) were analyzed to estimate the concentration of
9 crustal particles using factor analysis. Then, they estimated city-specific association of mortality
10 with fine crustal mass using Poisson regression (regressing mortality on factor scores for "crustal
11 factor"), adjusting for time trends and weather. They found no associations between fine crustal
12 mass factor and mortality. Details could not be reviewed because this was an abstract.
13 These results suggest that crustal particles (coarse or fine) are not associated with daily
14 mortality in these cities.
15
16 6.3.2.5 New Assessments of Confounding
17 Assessment of Co-Pollutant Confounding
18 As discussed above, the issue of potential confounding by weather was extensively
19 examined in two studies as reviewed in the 1996 AQCD, and was considered essentially
20 resolved. Therefore, discussion of confounding in this section is focused on potential
21 confounding among pollutants. Evaluating the extent of confounding of multiple pollution
22 effects in time-series studies can be complicated by differences in model specification (e.g.,
23 choice of lags). The following example of Philadelphia analyses conducted by two groups
24 illustrates the complexity of this issue.
25 Moolgavkar and Luebeck (1996) speculated that many of the past PM-mortality studies
26 suffered from "serious" deficiencies in their control of the confounding effects of other
27 pollutants. As a consequence, they argued, the small risks reported to be associated with the
28 particulate component of air pollution could be attributed to residual confounding by
29 co-pollutants. They conducted a new analysis of mortality in Philadelphia (1973-1988) that
30 considered four pollutants simultaneously, as well as seasonal effects, to illustrate this point.
31 Their findings are qualitatively similar to Samet et al. (1996) (or Kelsall et al. [1997], which
October 1999 6-93 DRAFT-DO NOT QUOTE OR CITE
-------
1 presented essentially the same results) in which the Philadelphia data for the study period
2 1974-1988 were analyzed also by season and with simultaneous inclusion of multiple pollutants.
3 The Samet et al. (1996) study (Kelsall et al, 1997) was extensively discussed in the 1996 PM CD
4 (Section 12.6), and therefore will be discussed only in comparison to Moolgavkar-Luebeck
5 results here. Both studies reported that pollution effects varied by season, and TSP coefficients
6 diminished when other pollutants were simultaneously included.
7 The notable differences in findings between Samet's group and Moolgavkar-Luebeck
8 included: (1) NO2 in Samet et al.'s study was mostly negatively associated (except summer) with
9 mortality, while in Moolgavkar-Luebeck study, NO2 was mostly positively associated (except
10 winter); (2) O3 in Samet et al.'s study was positively associated with mortality across seasons
11 (weakest in the summer), while in Moolgavkar-Luebeck study, O3 was positively associated with
12 mortality only in the summer. The difference may have been due to the absence of CO in
13 Moolgavkar-Luebeck analysis, or the difference in the optimum lags chosen for pollutants
14 (in Samet et al. study, concurrent day levels were used for all the pollutants except CO;
15 in Moolgavkar-Luebeck study, one-day lag was used for all pollutants except NO2). Thus, there
16 are some differences between the two groups of investigator's results from essentially the same
17 data. Moolgavkar-Luebeck concluded that "..it is not possible with the present evidence to show
18 a convincing correlation between particulate air pollution and mortality", while Samet's group
19 concluded "...These analyses confirm the association between TSP and mortality found in
20 previous studies in Philadelphia and the association is robust to consideration of other
21 pollutants".
22 Analyses of one city's data by different researchers may produce conflicting results, but
23 these discrepancies can in part result from instability of regression coefficients due to collinearity
24 of co-pollutants, as well as model specification choice. The collinearity problem may be further
25 complicated by different seasonal patterns of concentrations for each pollutant, including
26 differential changes in distribution shape for each pollution, changes in temporal correlations for
27 each pollutant, and changes in the matrix of correlations of pollutants with each other (and with
28 weather variables) across season. PM indices can contain both primary and secondary particles,
29 whose seasonal patterns may vary from city to city. There may be regional, local city-to-city, or
30 even within city difference in the relative impact of source types. The issue of relative exposure
October 1999 6-94 DRAFT-DO NOT QUOTE OR CITE
-------
1 error among co-pollutants, which further complicates the confounding problem, is discussed in
2 the measurement error section.
3 Thus, an evaluation of apparently inconsistent results from a city or a few cities analyzed
4 using different model specifications, without quantitative information on city specific
5 characteristics, is unlikely to yield useful information to resolve the issue of confounding.
6 By analyzing multiple cities' data, a more consistent pattern may emerge, although difference in
7 approach may still result in inconsistent multi-city results by different researchers. Thus, a more
8 definitive discussion of confounding by co-pollutants awaits a large multi-city studies (e.g.,
9 Samet's 100 city study) that are underway and expected to produce results soon.
10
11 Simulation Analysis of Confounding. Since no single model specification can be
12 "correct" in addressing confounding effects of co-pollutants, discrepancies in results among
13 studies, even for the same data set, are expected. While any assessment of relative
14 "adequateness" of these alternative model specifications is difficult with observational data, the
15 implication of "inadequate" model specifications may be studied through simulations using
16 synthetic data in which the "correct" model is known. Chen et al. (1999) conducted such
17 simulations using synthetic data set in which the causal variables are known, and the effects of
18 model misspecifcation were studied in the presence of two variables (xj and x2), with varying
19 level of correlation, in a Poisson model. They considered three situations: (1) model underfit, in
20 which mortality was generated with both Xj and x2, but regressed only on Xj; (2) model over/it, in
21 which mortality was generated with only x1? but regressed on both Xj and x2; (3) model misfit, in
22 which mortality was generated with either Xj or x2 but regressed on the other variable. They
23 observed that the confounding of covariates in an overfitted model does not bias the estimated
24 coefficients but reduces their significance; and that the effect of model underfit or misfit leads to
25 not only erroneous estimated coefficients but also erroneous significance. Chen et al. (1999),
26 based on these observations, suggested that "models which use only one or two air quality
27 variables, such as PM10 and SO2, are probably unreliable, and that model containing several
28 correlated and toxic or potentially toxic air quality variables should also be investigated...".
29 While conceptually useful, this simulation study ignored one factor that is crucial in evaluating
30 the implication of confounding, the relative error. For example, including several correlated
31 pollutants in regression model may lead to erroneous inference unless one considers the relative
October 1999 6-95 DRAFT-DO NOT QUOTE OR CITE
-------
1 error associated with each of the pollutants. Several simulation studies that considered such
2 relative errors are discussed in Section 6.4.6.4.
3
4 Alternative Approaches to Deal with Confounding and Address Source-Type Specific Effects:
5 Use of Factor A natysis
6 In time-series analyses of the acute effects of PM, the usual approach to deal with gaseous
7 co-pollutantss is to treat them as confounders and to simply include them simultaneously in
8 regression models. There has even been a suggestion, as mentioned above, based on a simulation
9 analysis of synthetic data, that "several" correlated pollutants should be included in regression
10 models (Chen et al., 1999). This prevailing approach can not only lead to misleading
11 conclusions in "identifying" a specific : "causal" pollutant (e.g., when pollutants have a varying
12 extent of exposure error), but also ignores the potential combined effects of PM and gaseous
13 co-pollutants (e.g., when PM adsorbs SO2 and carries it deeper in the airways, as shown by
14 Amdur and Chen, 1989).
15 Another potential problem of the simultaneous inclusion of PM and gaseous pollutants is
16 that the gaseous pollutant in question may be coming from the same source, or the PM
17 constituent may be derived from the gaseous pollutants. For example, SO2 can be converted to
18 sulfate, which is a PM constituent. Since a confounder cannot be an intermediate step in the
19 causal pathway (Rothman and Greenland, 1998), strictly speaking, SO2 does not qualify as a
20 confounder of PM, except in a situation where the PM is known to be solely of secondary origin
21 (transported aerosols), and SO2 is solely from local origin. Furthermore, any reduction in
22 emission of a gaseous pollutant may also result in reducing the level of PM. In such a case, the
23 inference drawn from the results of simultaneous regression may be misleading, because the
24 relative risk for PM is based on the assumption that the covariates could be kept unchanged while
25 the PM level changes.
26 Alternative approaches are needed to address the above noted weakness in the general
27 practice of effect estimation using simple simultaneous regressions. There have been a few
28 alternative approaches tried in recent years to estimate the effects of air pollution. For example,
29 Ozkaynak et al. (1996) analyzed 21 years of mortality and air pollution data in Toronto, Canada.
30 In addition to the usual simultaneous inclusion of multiple pollutants in mortality regression, they
31 also conducted a factor analysis of all the air pollution and weather variables including TSP, SO2,
October 1999 6-96 DRAFT-DO NOT QUOTE OR CITE
-------
1 Coefficient-of-Haze (CoH), NO2, O3, CO, relative humidity and temperature. The factor with the
2 largest variance contribution (~50%) had the highest factor loadings for CO, CoH, and NO2,
3 which they considered as representative of motor vehicle emissions, since this pollution grouping
4 was also consistent with the emission inventory information for that city. They then regressed
5 mortality on the factor scores (a linear combination of standardized scores for the covariates),
6 after filtering out seasonal cycles and adjusting for temperature and day-of-week effects. The
7 estimated excess mortality from motor vehicle pollution ranged from 1 to 6%, depending on the
8 outcomes.
9 Another recent example of the application of factor analysis is an analysis of the Harvard
10 Six-Cities data to investigate the role of crustal particles in fine particles on daily mortality
11 (Laden et al., 1999). They used elemental abundance data (obtained from X-ray fluorescence
12 spectroscopy analysis of daily filters) to estimate the concentration of crustal particles using
13 factor analysis. Then, they estimated city-specific association of mortality with fine crustal mass
14 by Poisson regression, adjusting for time trends and weather. They found no associations
15 between fine crustal mass factor and mortality.
16 Daisey et al. (1999) conducted an exploratory analysis of mortality in relation to specific
17 PM source types for three New Jersey cities (Camden, Newark, and Elizabeth) using factor
18 analysis - Poisson regression technique. During the three year study period (1981-1983),
19 extensive chemical speciation data were available including nine trace elements, sulfate,
20 particulate organic matter. Total (excluding accidents and homicides), cardiovascular and
21 respiratory mortality were analyzed. Daisey et al. first conducted a factor analysis of trace
22 elements and sulfate, identifying major source types: automobile (Pb, CO); geological (Mn, Fe);
23 oil burning (V, Ni); industrial (Zn, Cu); and sulfate/secondary aerosols (sulfate). In addition to
24 Poisson regression of mortality on these factors, they also used an alternative approach in which
25 the inhalable particle mass (IPM, D50 < 15 //m) was first regressed on the factor scores of each of
26 the source types to apportion the PM mass, then the estimated daily PM mass for each source
27 type was included in Poisson regression, so that RR could be calculated per mass concentration
28 basis for each PM source types. They found that oil burning (V, Ni), various industrial sources
29 (Zn, Cd), motor vehicle (Pb, CO), and the secondary aerosols as well as individual PM indices
30 IPM, FPM (D50 < 3.5 //m), and sulfates were associated with total and/or cardio-respiratory
October 1999 6-97 DRAFT-DO NOT QUOTE OR CITE
-------
1 mortality in Newark and Camden, but not in Elizabeth. In Camden, the RRs for the
2 source-oriented PM were higher (~ 1.10) than those for individual PM indices (~1.02).
3 Factor analyses have been routinely used in air pollution source apportionment field, but its
4 application to short-term health effects analyses is relatively new. It may be a useful alternative
5 approach for a source-oriented evaluation of the combined effects of fine particles and gaseous
6 co-pollutants. The advantages of the use of factor analysis approach include: (1) it allows an
7 examination of association between a health outcome and a group(s) of pollutants that vary
8 together (due to the same source type); (2) use of independent or orthogonal factors or
9 components in a regression model may allow much more stable estimates of the effects of groups
10 of pollutants that occur together; (3) it may reduce inflated uncertainty on effect size estimates
11 associated with including individual highly correlated pollution variables. The potential
12 drawback of the factor analysis approach is that the resulting factor that may represent a
13 source-type well may not necessarily have the variation that is relevant for the health outcome.
14 There are also additional issues in interpreting the results from the analyses that utilize factor
15 analysis, including the "interpretability" of resulting factors as derived from a common source,
16 and technical issues such as the choice of rotation of factors. While potentially useful, some
17 issues still need to be investigated.
18
19 6.3.2.6 New Assessments of Cause-Specific Mortality
20 In most of the new studies that examined nonaccidental total, circulatory, and respiratory
21 mortality categories (Borja-Aburto et al, 1997; Wordley et al, 1997; Borja-Aburto et al, 1998;
22 Gwynn et al., 1998; Ostro et al., 1998; Prescott et al., 1998), estimated PM effects were generally
23 higher for respiratory deaths than for circulatory or total deaths, consistent with the same findings
24 in the 1996PM CD.
25 Evaluation of the cause-specificity of various pollution and weather variables' mortality
26 associations may be helpful in checking consistency with criteria for causality. A review of the
27 newly available studies and previous studies do not necessarily provide consistent patterns of
28 cause specificity that can distinguish one pollutant from others. For example, in Toronto analysis
29 by Burnett et al. (1998), they reported that, although the estimated CO RR was higher for cardiac
30 death category, "a clear positive association" was also observed for non-cardiac categories.
31 However, since the presumed mechanism for CO (i.e., binding to carboxyhemoglobin) may result
October 1999 6-98 DRAFT-DO NOT QUOTE OR CITE
-------
1 in impaired oxygen delivery to the peripheral tissues, leading to other complications that may not
2 be necessarily cardiac, the apparent lack of cause-specificity for CO in this case may not be easily
3 interpreted as confounded associations.
4 Seeking unique cause-specificity of various pollutants has been also difficult because the
5 "cause specific" categories examined are rather broad (usually cardiovascular and respiratory),
6 and overlap or co-existence of cardiovascular and respiratory conditions is expected.
7 Examinations of more specific cardiovascular and respiratory sub-categories may be necessary to
8 test hypotheses on any specific mechanism, but smaller sample sizes for more specific
9 sub-categories may make a meaningful analysis difficult. The study by Rossi et al. (1999),
10 however, examined associations between TSP and detailed cardio-vascular and respiratory
11 cause-specific mortality in Milan, Italy for years 1980-1989. They found a significant association
12 for respiratory infections (11% increase per 100 //g/m3 increase in TSP; 95%CI: 5, 17) and for
13 heart failure (7%; 95%CI: 3, 11), both on the same day TSP. The associations with myocardial
14 infarction (10%; 95%CI: 3,18) and COPD (12%; 95%CI: 6, 17) were found for the average of
15 3 and 4 day TSP levels. They noted the difference in lags between temperature effects (i.e., cold
16 temperature at lag 1 day for respiratory infections; hot temperature at lag 0 for heart failure and
17 myocardial infarction) and air pollution (TSP) effects. The immediate hot temperature effects
18 and the lagged cold temperature effects for total and cardiovascular mortality have been reported
19 in many of the past studies (e.g., Philadelphia, Chicago), but investigations of the differences in
20 lags of PM effects for specific cardiovascular or respiratory categories have rarely been
21 conducted in time-series mortality studies.
22 Some of recent PM studies did examine more specific type of deaths, such as intrauterine
23 mortality (Pereira et al., 1998) and post neonatal mortality (Woodruff et al., 1997; Bobak and
24 Leon, 1998). In the case of intrauterine mortality, PM10 was not a significant predictor, but CO's
25 association was supported by the association between increased carboxyhemoglobin in fetal
26 blood and ambient CO levels on the day of delivery measured in a separate study. The Woodruff
27 et al. study used logistic regressions (adjusting for demographic and environmental factors) to
28 examine relationship between exposure to PM10 in the first two month of life and the chance of
29 dying from specific causes of death between 1 month and 1 year of age using cohort data of about
30 4 million infants. They found associations between PM10 and deaths from respiratory causes, as
31 well as sudden infant death syndrome. Bobak and Leon (1998) also reported associations
October 1999 6-99 DRAFT-DO NOT QUOTE OR CITE
-------
1 between air pollution (TSP, NO2, SO2) and respiratory causes for post neonatal period using a
2 matched case-control study design. While these are not time-series studies, the presumed
3 exposure-effect period is not "long-term". These and other study design may be potentially
4 useful to investigate more specific cause or type of deaths that are difficult to analyze in
5 time-series study design.
6
7 6.3.2.7 New Assessment of Methodological Issues
8 Methodological issues in time-series analyses of air pollution-mortality association were
9 discussed extensively in the 1996 AQCD. Since then, increasing numbers of researchers have
10 been utilizing essentially the same Poisson regression approach: (1) model seasonal cycles and
11 other temporal trends using smoothing functions of time; (2) model weather effects using
12 smoothing functions of temperature, humidity, and/or their interaction at various lags; (3) after
13 adjustment for these confounding factors, enter various lags (and averaging periods) of air
14 pollutant, and report results for all the lags, and/or report results for the lags that resulted in the
15 highest significance; (4) repeat (3) with other pollutants in the model; (5) conduct sensitivity
16 analyses using alternative weather model specifications. Seasonal cycles and weather effects are
17 often modeled using Generalized Additive Models (GAM). As the modeling temporal trends
18 became more efficient using the GAM models, it became clearer that the residual over-dispersion
19 and autocorrelation can be essentially eliminated. Also, more researchers appear to rely on
20 Akaike's Information Criteria (AIC) or on the more conservative Bayes (Schwarz) Information
21 Criterion (BIC) to choose between models when epidemiological reasoning does not favor one
22 over the other. Similar estimates may be obtained by other techniques, such as the Liang-Zeger
23 Generalized estimating Equation (GEE) method described in (Samet et al., 1995) that deals with
24 autocorrelated time series. While these techniques do not necessarily eliminate inadequate model
25 specifications, they do help "standardize" the approaches that researchers can take, reducing the
26 inconsistency in model specification among studies.
27 Differences in results among investigators using the same or similar data sets are more
28 likely to be associated with other differences in model-building strategy, not with statistical
29 methodology. These often include: (1) choice of the range of lags and averaging periods of
30 pollution included; (2) smoothing spans used for modeling temporal trends and weather effects;
31 (3) the increment used to calculate relative risks; and, (4) choice to detrend pollution variables.
October 1999 6-100 DRAFT-DO NOT QUOTE OR CITE
-------
1 The choice of lag can lead to inconsistent results even for the same data. The choice of the
2 combination of lags multiply as the number of co-pollutants in the model increases. In the case
3 of temperature effects, it has been repeatedly observed that the heat effects tend to be immediate
4 (0 or 1 day lag), while cold effects tend to lag longer (2 to 4 days). For air pollutants, however,
5 reported lags are less consistent.
6 The smoothing span for temporal trends can be determined based on epidemiological
7 reasons (for example, to eliminate influenza epidemics) or to optimize goodness of fit using AIC
8 or BIC criteria. The effects of temporal smoothing choices on the estimated effect size of
9 pollution variables may be substantial, but is not reported by many investigators. The span for
10 weather effects is usually determined through data exploration. Characterizing PM and
11 co-pollutant effect size by RR across the inter-quartile range for all the co-pollutants may be
12 problematic when co-pollutants have inconsistent distributional properties, such as different
13 within-season ranges and between-season ranges. While these issues may appear rather minor,
14 in practice, they appear to make substantial differences in reported effects and interpretations.
15
16 6.3.2.8 Summary of Newly Available Information
17 • Since the 1996 PM AQCD, thus far, there have been more than 30 new time-series
18 PM-mortality analyses, several of which investigated multiple cities using consistent data
19 analytical approaches. PM relative risks estimated for daily mortality in these studies are
20 generally positive, statistically significant, and consistent with the previously reported
21 PM-mortality associations. However, several studies also showed significant associations
22 between mortality and gaseous pollutants, such as CO and O3. Since a large number of
23 studies, including U.S. multi-city studies, are expected to be published in the next several
24 months, a quantitative summary of PM and other pollutants' effects will not be attempted at
25 this time.
26 • The multi-city study conducted in European cities showed generally consistent associations
27 between mortality and both SO2 and PM indices in western European cities, but not in central
28 and eastern European cities. The pooled estimate of PM10 - mortality relative risks calculated
29 for western European cities were roughly comparable to the estimates from US data. The
30 contrast between western and central eastern Europe results was speculated to be due to:
October 1999 6-101 DRAFT-DO NOT QUOTE OR CITE
-------
1 difference in exposure representativeness, difference in pollution toxicity or mix, difference in
2 proportion of sensitive sub-population, or climate.
3 • Several new studies are available regarding the role of size and chemistry in PM-mortality
4 associations. In Santa Clara County, CA, PM25, as well as nitrate, were significantly
5 associated with mortality. The studies conducted in US and Canadian cities also showed
6 mortality associations with specific fine particle components of PM including H+, SO4=,
7 as well as PM2 5. CoH, which likely reflects motor vehicle related carbon particles, was
8 significantly associated with mortality in Toronto, Canada, where its level is relatively high,
9 but not in Buffalo, NY, where its level was lower (50% of that in Toronto). Seeking the
10 effects of a group of source types, rather than any individual component, may be useful for
11 inferring causes of adverse health effects. An association between PM2 5 and mortality was
12 also reported in Mexico City. Several studies that investigated the role of crustal and coarse
13 particles suggest that crustal particles, coarse or fine, are unlikely to be associated with
14 mortality.
15 • A few studies conducted simulation analyses of effects of measurement errors on the
16 estimated PM mortality effects. These studies suggest that PM effects are more likely
17 underestimated than overestimated, and that spurious PM effects (i.e., qualitative bias such as
18 change in the sign of the coefficient) due to transferring of effects from other covariates
19 require extreme conditions, and are therefore unlikely. The error due to the difference
20 between the average personal exposure and the ambient level are likely the major source of
21 bias in estimated relative risk. One study also suggested that apparent linear
22 exposure-response curves are unlikely to be artifacts of measurement error.
23 • Newly available simulation and empirical analyses suggest that the extent of harvesting, if it
24 exists, is not in the short-term (i.e., ~ 3 days) range. These new results, combined with the
25 results from a few past studies, suggest that the PM-mortality risk estimates are not heavily
26 influenced by displacement of mortality in the very short-term period. More analyses are
27 needed to replicate these findings in order to evaluate whether they apply to other cities than
28 Philadelphia and Boston.
29 • An increasing number of studies have considered co-pollutants in their analyses. While PM
30 indices remained significant in most of these multi-pollutant analyses, the relative significance
31 of mortality associations among co-pollutants varied from study to study. Several studies
October 1999 6-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 suggested that the relative significance of mortality associations may be partly explained by
2 differences in the spatial representativeness of the monitors from which the exposure-related
3 data were derived. The apparent difference among cities may also be explained by the
4 difference in relative impact of source types (e.g., regional pollutants vs more local
5 automobile related pollution). More systematic and quantitative evaluation of these factors
6 using multi-city data could explain the apparent discrepancy in individual study results.
7
8 6.3.3 Human Mortality and Long-Term Exposure to PM of Ambient Origin
9 6.3.3.1 Studies Published Prior to the 1996 Particulate Matter Criteria Document
10 Aggregate Population Cross-Sectional Chronic Exposure Studies
11 Mortality effects associated with chronic, long-term exposure to particulate matter (PM) air
12 pollution of outdoor origins have been assessed in cross-sectional studies and, more recently, in
13 prospective cohort studies. A number of older cross-sectional studies from the 1970s provided
14 indications of increased mortality associated with chronic (annual average) exposures to ambient
15 PM, especially with respect to fine mass or sulfate (SO4=) concentrations. However, questions
16 unresolved at that time regarding the adequacy of statistical adjustments for other potentially
17 important covariates (e.g., cigarette smoking, economic status, etc.) across cities tended to limit
18 the degree of confidence that was placed by the 1996 PM AQCD (U.S. Environmental Protection
19 Agency, 1996) on such purely "ecological" studies or on quantitative estimates of PM effects
20 derived from these studies. Evidence comparing the toxicities of specific PM components was
21 relatively limited. The sulfate and acid components had already been discussed in detail in the
22 previous PM AQCD (U.S. Environmental Protection Agency, 1986).
23 Lippmann (1989) hypothesized that the acidic portion of the fine particle aerosol (i.e., H+)
24 was an important contributor to the adverse health effects of PM. Ozkaynak and Thurston (1987)
25 applied source apportionment methods to the IP Network data, finding that fine particles from
26 coal combustion and from the metals industry were more important contributors to the
27 PM-mortality association than other PM mass contributors (e.g., soil particles). Lipfert (1984)
28 examined the 1980 U.S. mortality-PM data set using much more heavily specified models,
29 finding that PM2 5 was the strongest particulate variable in linear models (with Mn, a possible
30 tracer for metals industry emissions, also approaching significance).
31
October 1999 6-103 DRAFT-DO NOT QUOTE OR CITE
-------
1 Semi-Individual (Prospective Cohort) Chronic Exposure Studies
2 Semi-individual cohort studies have used subject-specific information about relevant
3 covariates (such as cigarette smoking, occupation, etc.), providing more certain findings of long-
4 term PM exposure effects than past purely "ecological studies" (Kiinzli and Tager, 1997). At the
5 same time, these better designed cohort studies have largely confirmed the magnitude of the
6 effect estimates from past cross-sectional study results, renewing interest and confidence in their
7 findings.
8 Prospective cohort semi-individual studies of mortality associated with chronic exposures
9 to air pollution of outdoor origins have yielded especially valuable insights into the adverse
10 health effects of long-term PM exposures. The extensive Harvard 6-Cities Study (Dockery et al.,
11 1993) and the American Cancer Society (ACS) Study (Pope et al., 1995) agreed in their findings
12 of statistically significant positive associations between fine particles and excess mortality,
13 although the ACS study did not evaluate the possible contributions of other air pollutants.
14 Neither study considered multi-pollutant models, although the 6-City study did examine various
15 gaseous and particulate matter pollutants (including total particles, PM2 5, SO4= H+ SO2, and
16 ozone), finding that sulfate and PM25 fine particles were best associated with mortality. The RR
17 estimates for total mortality in the 6-Cities study (and 95 percent confidence intervals, CI) per
18 increments in PM indicator levels were: RR=1.42 (CI=1.16-2.01) for 50 //g/m3 PM10, RR=1.31
19 (CI=1.11-1.68) for 25//g/m3PM25, and RR=1.46(CI=1.16-2.16) for 15//g/m3 SO4=. The
20 estimates for total mortality derived from the ACS study were RR= 1.17 (CI= 1.09-1.26) for
21 25 //g/m3 PM2 5, and 1.11 (CI= 1.06-1.16) for 15 //g/m3 SO4=. The ACS pollutant RR estimates
22 are smaller than those from the 6-Cities study, although their 95% continence intervals overlap.
23 In some cases in these studies, the life-long cumulative exposure of the study cohorts included
24 distinctly higher past PM exposures, especially in the cities with historically higher PM
25 concentrations such as Steubenville, OH; but more current PM measurements were used to
26 estimate the chronic PM exposures. In the ACS study, the pollutant exposure estimates were
27 based upon concentrations at the start of the study (during 1979-1983. Also, the average age of
28 the ACS cohort was 56, which could overestimate the pollutant RR estimates, and might
29 underestimate the life-shortening associated with PM associated mortality. Thus, while caution
30 must be exercised regarding the use of the reported quantitative risk estimates, the 6-Cities and
October 1999 6-104 DRAFT-DO NOT QUOTE OR CITE
-------
1 ACS semi-individual studies provided consistent evidence of a significant mortality association
2 with long-term exposure to PM of ambient origins.
3 In contrast to the 6-Cities and ACS studies, early results from the Adventist Health Study
4 on Smog (AHSMOG) of California nonsmokers by Abbey et al. (1991) and Abbey et al. (1995)
5 found no significant mortality effects of previous PM exposure in a relatively young cohort.
6 However, these analyses used TSP as the PM exposure metric, rather than more health relevant
7 PM metrics such as PM10 or PM2 5, included fewer subjects than the ACS study, and considered a
8 shorter follow-up time than the 6-Cities study (ten years vs. 15 years for the 6-Cities study).
9 Moreover, the AHSMOG study included only non-smokers, who the 6-cities studies indicate to
10 have lower pollutant RR's than smokers, suggesting that a longer follow-up time than considered
11 in the past (10 years) might be required to have sufficient power to detect significant pollution
12 effects than is required in studies that include smokers, such as the 6-Cities and ACS studies.
13 Thus, to date, greater emphasis has been placed on the 6-Cities and ACS studies.
14 Overall, these past chronic exposure studies collectively indicate that there are increases in
15 mortality that are associated with long-term exposure to airborne particles of ambient origins.
16 These estimates of long-term PM exposure effect size for total mortality (e.g., RR=1.17 for
17 25 //g/m3 PM2 5 from the ACS study) are much larger than those reported from daily mortality
18 PM studies (e.g., multi-study pooled RR=1.044 per 50 //g/m3 PM10, from Schwartz, 1997). Thus,
19 even the upper limit estimate of the long-term implications of the reported daily mortality effects
20 (i.e., assuming that they are fully additive over time) falls well below the chronic exposure study
21 mortality effect estimates. This suggests that a major fraction of the reported mortality relative
22 risk estimates associated with chronic PM exposure reflect cumulative PM impacts above and
23 beyond those that could be exerted by the sum of acute exposure events.
24 The 1996 PM AQCD reached several conclusions concerning four key questions about the
25 prospective cohort studies. Relevant sections from Ch. 12 (pp. 180-182) of the 1996 document
26 are quoted directly:
27
28 1. Have potentially important confounding variables been omitted?
29 "While it is not likely that the prospective cohort studies have overlooked plausible
30 confounding factors that can account for the large effects attributed to air pollution, there may be
31 some further adjustments in the estimated magnitude of these effects as individual and
October 1999 6-105 DRAFT-DO NOT QUOTE OR CITE
-------
1 community risk factors are included in the analyses." These include individual variables such as
2 education, occupational exposure to dust and fumes, and physical activity, as well as ecological
3 (community) variables such as regional location, migration, and income distribution. Further
4 refinement of the effects of smoking status may also prove useful.
5
6 2. Can the most important pollutant species be identified?
7 "The issue of confounding with co-pollutants has not been resolved for the prospective
8 cohort studies. ... Analytical strategies that could have allowed greater separation of air pollutant
9 effects have not yet been applied to the prospective cohort studies." The ability to separate the
10 effects of different pollutants, each measured as a long-term average on a community basis, was
11 clearly most limited in the Six Cities study. The ACS study offered a much larger number of
12 cities, but did not examine differences attributable to the (spatial and temporal) differences in the
13 mix of particles and gaseous pollutants across the cities. The AHSMOG study constructed time-
14 and location-dependent pollution metrics for most of its subjects that might have allowed such
15 analyses, but no results were reported, then or subsequently.
16
17 3. Can the time scales for long-term exposure effects be evaluated?
18 "Careful review of the published studies indicated a lack of attention to this issue.
19 Long-term mortality studies have the potential to infer temporal relationships based on
20 characterization of changes in pollution levels over time." This potential was greater in the Six
21 Cities and AHSMOG studies because of the greater length of the historical air pollution data for
22 the cohort.
23
24 4. Is it possible to identify pollutant thresholds that might be helpful in health assessments?
25 "Model specification searches for thresholds have not been reported for prospective cohort
26 studies." The time scale of an air pollution exposure metric for which a threshold is being sought
27 is a key element in a model specification search.
28 "The chronic exposure studies, taken together, suggest that there may be increases in
29 mortality in disease categories that are consistent with long-term exposure to airborne particles,
30 and that at least some fraction of these deaths are likely to occur between acute exposure
October 1999 6-106 DRAFT-DO NOT QUOTE OR CITE
-------
1 episodes. If this interpretation is correct, then at least some individuals may experience some
2 years of reduction of life as a consequence of PM exposure." (P 12-368).
3 Many of these issues remain unresolved at this time. Extensive reanalyses of the Six Cities
4 and ACS studies are underway to evaluate these questions, under the sponsorship of the Health
5 Effects Institute. Preliminary public presentations (Health Effects Institute, 1999) suggest that
6 the published findings of the original investigators (Dockery et al., 1993; Pope et al., 1995) are
7 based on substantially valid data sets and statistical analyses, and that small corrections in input
8 data have very little effect on the findings. Additional investigations to evaluate the effects of
9 alternative model specifications are in progress, and may be available in time for the final draft of
10 this document.
11 Recently published analyses of the AHSMOG study (Abbey et al., 1999; Beeson et al.,
12 1998) considerably extend the earlier findings of the investigators, and also show some
13 differences from earlier studies. Of particular interest are their findings in relation to lung
14 cancer. These are discussed below in some detail. Additional studies suggest possible effects of
15 sub-chronic PM exposures on infant mortality (Woodruff et al., 1997; Bobak and Leon, 1998),
16 and these are also included below in this discussion of long-term PM exposure effects on
17 mortality.
18
19 6.3.3.2 Prospective Cohort Studies of Chronic Exposure Published Since the Last
20 Particulate Matter Criteria Document
21 Abbey et al. (1999)
22 The Adventist Health Study of Smog (AHSMOG) enrolled 6,338 non-smoking
23 non-Hispanic white Seventh Day Adventist residents of California, ages 27 to 95 years, in 1977.
24 The participants had resided for at least 10 years within 5 miles (8 km) of their then-current
25 residence locations. Subjects lived either within the 3 major California air basins (San Diego,
26 Los Angeles, or San Francisco), or else were part of a random 10% sample of Adventist Health
27 Study participants in the rest of California. The study has been extensively described elsewhere
28 (Hodgkin et al., 1984; Abbey et al., 1991; Mills et al., 1991). Mortality status of the subjects
29 after ca. 15-years of follow-up (1977-1992) was determined by a variety of tracing methods,
30 finding 1,628 deaths (989 female, 639 male) in the cohort. There were 1,575 deaths from all
31 natural (non-external) causes, of which 1,029 were cardiopulmonary deaths, 135 were
October 1999 6-107 DRAFT-DO NOT QUOTE OR CITE
-------
1 non-malignant respiratory deaths (ICD9 codes 460-529), and 30 were lung cancer deaths
2 (ICD9 code 162). Abbey et al. also created an additional death category, contributing respiratory
3 causes (CRC). CRC included any mention of nonmalignant respiratory death as either an
4 underlying cause or a contributing cause on the death certificate coded by an exposure-blinded
5 nosologist (the other groups listed only underlying causes), with 410 deaths (246 female and
6 164 male). A large number of analyses were done for the CRC category, due to the large
7 numbers and relative specificity of respiratory causes as a factor in the deaths.
8 Education was used as an index of socio-economic status, rather than income. Physical
9 activity and occupational exposure to dust were also used as covariates. Migration is not a major
10 concern in this residentially stable cohort.
11 A number of exposure indicators were used: mean values of PM10 (imputed from TSP in
12 the earlier years of the study), SO4=, SO2, O3, and NO2; and "threshold" indicators, days per year
13 with PM10 > 100 //g/m3, and hours per year with O3 > 100 ppb. The "standard" increments used
14 for PM10 and SO4 in these tables are the same as described above for the short-term mortality
15 studies, 50 //g/m3 for PM10 and 15 //g/m3 for SO4, and 30 days per year for exceedances of PM10
16 above 100 //g/m3. The mean values for PM10 and SO4 during the study period were 51 and
17 7.2 //g/m3 respectively, and 31 days per year for PM10 exceedances over 100 //g/m3. The means
18 were much larger than the inter-quartile ranges (IQR) of 24 and 3.0 //g/m3. IQR is the increment
19 used for other variables. RR and confidence limits using IQR from (Abbey et al., 1999) are
20 shown to 2 decimal places, those estimated for standard increments are shown to 3 decimal
21 places.
22 Cox proportional hazard models adjusted for a variety of covariates, or stratified by sex,
23 were used in the models. The "time" variable used in most of the models was survival time from
24 date of enrollment, except that age on study was used for lung cancer effects due to the expected
25 lack of short-term effects. A large number of covariate adjustments were evaluated, as shown in
26 Table 6-25 and described by Abbey et al. (1999).
27 The CRC estimates of RR from 30 days per year with PM10 > 100 //g/m3 for males and
28 females combined are shown in Table-25. Positive and statistically significant effects are found
29 for almost all models that include age, pack-years of smoking, and body-mass index (BMI)
30 categories as covariates. Subsets of the cohort also often had elevated risks. Former smokers
31 had higher relative risks than never-smokers (RR for PM10 exceedances for never-smokers was
October 1999 6-108 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-25. RELATIVE RISK OF MORTALITY FROM CONTRIBUTING
NON-MALIGNANT RESPIRATORY CAUSES, FOR
30 DAYS PER YEAR WITH PM10 > 100 //g/m3
PM Covariate Model RR RRLCL RRUCL
BASE (age, sex)
BASE + pack-years
BASE + pack-years + body-mass-index cats.
BASE + pack-years + body-mass-index cats.+ exercise cats.
STANDARD (age, pack-y., y. lived with smoker, occup., educ., BMI)
STANDARD w. PM10 (100) over last 4 years only
STANDARD, subset for former smokers
STANDARD, subset for never smokers
STANDARD, subset for low anti-oxidant vitamin intake
STANDARD, subset for high anti-oxidant vitamin intake
STANDARD, subset for < 4 h/wk outdoors
STANDARD, subset for 4-16 h/wk outdoors
STANDARD, subset for 16+ h/wk outdoors
STANDARD, subset for reported respiratory symptoms
1.069
1.096
1.122
1.122
1.122
1.102
1.155
1.116
1.175
1.055
1.048
1.122
1.207
1.321
0.978
1.000
1.022
1.017
1.017
1.001
0.937
0.999
1.008
0.917
0.896
0.928
1.015
1.079
1.168
1.201
1.233
1.239
1.239
1.214
1.424
1.246
1.370
1.214
1.227
1.358
1.436
1.616
Source: Abbey et al. (1999).
1 marginally significant by itself, in spite of the reduced sample size). Subjects with low intake of
2 anti-oxidant vitamins A, C, E had significantly elevated risk of response to PM10 whereas those
3 with adequate intake did not, suggesting that dietary factors (or possibly other socio-economic or
4 life style factors for which they are a surrogate) may be important covariates in other studies.
5 There also appears to be a gradient of PM10 risk with respect to time spent outdoors, with
6 individuals who had spent at least 16 hours per week outside at distinctly elevated risk from PM10
7 exceedances. The extent to which time spent outdoors is a surrogate for other variables or is a
8 modifying factor reflecting temporal variation in exposure to ambient air pollution is not certain.
9 For example, males spend about twice as much time outdoors as females, so that outdoor
10 exposure time is confounded with gender.
October 1999 6-109 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
A considerably different picture is shown when the analyses are broken down by gender.
Table 6-26 shows much lower RR for female CRC deaths for all co-pollutants, with all female
RR positive, but not statistically significant. The CRC for males remains significant only for
PM10 exceedances, but not for other air pollution metrics. The PM10 exceedance effect for CRC
for both sexes is roughly the average of that for males and females.
TABLE 6-26. RELATIVE RISK OF MORTALITY FROM CONTRIBUTING
NON-MALIGNANT RESPIRATORY CAUSES, BY SEX AND
AIR POLLUTANT, WITH ALTERNATIVE COVARIATE MODEL
Pollution Index
PM10>100, d/yr
PM10 mean
SO4 mean
O3>100 ppb, h/yr
30
50
15
Pollution Incr.
days/yr
Mg/m3
Mg/m3
551 h/yr (IQR)
RR
1.069
1.219
1.105
1.01
Females
RRLCL
0.
0.
0.
0.
,936
,739
,396
,77
RRUCL
1
2
o
3
i
.220
.011
.086
.33
RR
1.188
1.537
1.219
1.20
Males
RRLCL
1.
0.
0.
0.
,030
,879
,411
,88
RRUCL
1.370
2.688
3.619
1.64
Source: Abbey et al. (1999).
1 Personal monitoring was not conducted on this part of the cohort, and other factors such as
2 occupational exposure for which the questionnaire was not adequate may also account for male
3 vs. female differences, along with gender differences in the amount of time spent outdoors.
4 Finally, it is not surprising that individuals reporting respiratory symptoms in 1977 may be at
5 greater risk to PM10 or other environmental insults presumably involved in subsequent CRC
6 deaths, and prior health status may also be gender-related.
7 Table 6-27 shows much lower RR for female non-external deaths for all co-pollutants, with
8 no female RR positive nor statistically significant. Deaths from non-external causes for males
9 remains statistically significant for PM10 exceedances, but not for other air pollution metrics.
10 However, the RR estimates for males for other air pollutant metrics are relatively large.
11 Table 6-28 shows much lower RR for female cardio-pulmonary deaths for all co-pollutants,
12 with only the female RR for mean SO2 positive and none statistically significant. Deaths from
October 1999
6-110
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-27. RELATIVE RISK OF MORTALITY FROM ALL
NON-EXTERNAL CAUSES, BY SEX AND AIR POLLUTANT,
FOR AN ALTERNATIVE COVARIATE MODEL
Pollution Index
PM10>100, d/yr
PM10 mean
SO4 mean
O3>100 ppb, h/yr
SO2 mean
Pollution Incr.
30 days/yr
50 ptg/m3
15 Mg/m3
551 h/yr (IQR)
3.72 (IQR)
RR
0.958
0.879
0.732
0.90
1.00
Females
RRLCL
0.899
0.713
0.484
0.80
0.91
RRUCL
1.021
1.085
1.105
1.02
1.10
RR
1.082
1.242
1.279
1.140
1.05
Males
RRLCL
1.008
0.955
0.774
0.98
0.94
RRUCL
1.162
1.616
2.116
1.32
1.18
Source: Abbey et al. (1999).
TABLE 6-28. RELATIVE RISK OF MORTALITY FROM
CARDIOPULMONARY CAUSES, BY SEX AND AIR POLLUTANT,
FOR AN ALTERNATIVE COVARIATE MODEL
Pollution Index
PM10>100, d/yr
PM10 mean
SO4 mean
O3>100 ppb, h/yr
O3 mean
SO2 mean
Pollution Incr.
30 days/yr
50 ,ug/m3
15 ,ug/m3
551 h/yr (IQR)
10 ppb
3.72 (IQR)
RR
0.929
0.841
0.857
0.88
0.975
1.02
Females
RRLCL
0.857
0.639
0.498
0.76
0.865
0.90
RRUCL
1.007
1.107
1.475
1.02
1.099
1.15
RR
1.062
1.219
1.279
1.06
1.066
1.01
Males
RRLCL
0.971
0.862
0.002
0.87
0.920
0.86
RRUCL
1.162
1.616
1018
1.29
1.236
1.18
Source: Abbey et al. (1999).
1 cardiopulmonary causes for males is no longer statistically significant for PM10 exceedances, nor
2 for other air pollution metrics. However, the RR estimates for males for air pollutant metrics are
3 relatively large.
4 Table 6-29 shows lower RR for female lung cancer deaths for all co-pollutants, but some
5 female RR are positive and statistically significant: mean NO2, mean SO2 for all women and for
October 1999
6-111
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-29. RELATIVE RISK OF MORTALITY FROM LUNG CANCER, BY SEX
AND AIR POLLUTANT, FOR AN ALTERNATIVE COVARIATE MODEL
Pollution
Index
PM10>100, d/yr
PM10 mean
NO2 mean
O3>100ppb, h/yr
O3 mean
SO2 mean
Pollution
Incr.
30 days/yr
50 Mg/m3
19.78 (IQR)
551 h/yr (IQR)
lOppb
3.72 (IQR)
Smoking
Category
All1
All
All
All
never smoker
past smoker
All
All
never smokers
Females
RR RR LCL
1.055 0.657
1.808 0.343
2.81 1.15
1.39 0.53
0.805 0.436
3.01 1.88
2.99 1.66
RRUCL
1.695
9.519
6.89
3.67
1.486
4.84
5.40
RR
1.831
12.385
1.82
4.19
6.94
4.25
1.853
1.99
Males
RRLCL
1.281
2.552
0.93
1.81
1.12
1.50
0.994
1.24
RRUCL
2.617
60.107
3.57
9.69
43.08
12.07
3.453
3.20
Source: Abbey et al. (1999).
Note 1: All = both never smokers and past smokers.
1 female never-smokers. Deaths from lung cancer for males remains statistically significant for all
2 air pollution metrics except mean NO2, and is nearly significant for mean O3. The RR estimates
3 for males for all air pollutant metrics are relatively large. However, the confidence intervals are
4 wide, due to the small numbers of lung cancer deaths (18 for females and 12 for males).
5 Lung cancer effects are significant for males for PM10 and O3 metrics, but not for females.
6 Lung cancer metrics for mean SO2 are significant for both males and females. Lung cancer
7 deaths are significant for mean NO2 for females, but not for males. This pattern is not readily
8 interpretable but may be attributable to the small numbers of deaths.
9 The CRC effects identified in this study are significant when both females and males are
10 included, but not when female and male subjects are analyzed separately, except for male
11 subjects with PM10 exceeding 100 //g/m3. The effects of PM10 exceeding 100 //g/m3 on mortality
12 from all non-external causes are also statistically significant for males, but not for other air
13 pollutants, and not for females. Separate analyses by gender may be more appropriate in order to
October 1999 6-112 DRAFT-DO NOT QUOTE OR CITE
-------
1 maintain the assumption of "proportional hazards" necessary for the validity of Cox method of
2 analysis.
3 In general, this study suggests a pattern of mortality from diverse causes in males, but
4 provides little evidence for female mortality from these causes. The male causes are primarily
5 associated with exposures to PM10, especially PM10 > 100 //g/m3. Other air pollutants are
6 associated with lung cancer deaths in females as well as males.
7 The analyses reported here attempted to separate PM10 effects from those of the other
8 pollutants by use of two-pollutant models, but none of the quantitative findings from these
9 models were reported. The text mentions that the PM10 coefficient for CRC remained stable or
10 increased when other pollutants were added to the model. Lung cancer models for males were
11 evaluated for co-pollutant effects in detail. NO2 remained nonsignificant in all two-pollutant
12 models, and the other pollutant coefficients were stable in magnitude. The PM10 and O3 effects
13 remained stable when SO2 was added, suggesting that their effects are independent. However,
14 the effects of PM10 and O3 were hard to separate because these pollutants were highly correlated
15 in this study. When both exceedances PM10 > 100 //g/m3 and O3 > 100 ppb were used in the
16 model, both RR were reduced in magnitude, but the O3 exceedance RR remained more
17 significant than the RR for the PM10 exceedance. The possibility that the finding of a significant
18 PM10 effect is partially attributable to correlation with other pollutants such as O3 cannot be
19 completely precluded. The finding of an O3 effect on lung cancer is not readily explained, and
20 the finding of a PM10 effect on lung cancer (consistent with the ACS and Six Cities studies) is
21 more plausible.
22 The SO2 coefficient for lung cancer in females remained stable in two-pollutant models
23 when PM10 and O3 exceedances were included. This suggests that the significance of the SO2
24 effect for females may not be an artifact attributable to collinearity with these co-pollutants.
25
26 Beeson et al. (1998)
27 This study uses essentially the same data as in (Abbey et al., 1999), but concentrates on
28 lung cancer incidence (1977-1992) as an endpoint. There were only 20 female cases and 16 male
29 cases of lung cancer among the 6,338 subjects. The exposure metrics were constructed to be
30 specifically relevant to cancer, being the annual average of the monthly exposure indices from
31 January, 1973, through the following months, but ending 3 years before the date of diagnosis of
October 1999 6-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 the case. This allows a 3-year lag between exposure and diagnosis of lung cancer, allowing for a
2 latency period. Therefore, statistical indices for exposure have somewhat different statistics than
3 in (Abbey et al., 1999), such as the IQR and mean. In spite of improved treatment of lung cancer
4 over the last decades, this disease still has a rather pessimistic prognosis compared to other
5 diagnoses of long-term mortality, so is discussed along with the mortality analyses.
6 The covariates in the Cox proportional hazards model were pack-years of smoking and
7 education, and the time variable was attained age. A number of additional covariates were
8 evaluated for inclusion in the model, but only 'current use of alcohol' met the criteria for
9 inclusion in the final model. Individual pollutants evaluated were PM10, SO2, NO2, and O3.
10 No interaction terms with the pollutants proved to be significant, including outdoor exposure
11 times. Gender-specific relative risk estimates were reported for the various risk factors.
12 Results are shown in Table 6-30 for males and Table 6-31 for females. Standard
13 increments were used for PM10 mean (50 //g/m3) and exceedances of PM10 > 100 //g/m3 (30 d/y).
14 RR and confidence limits using IQR from (Beeson et al., 1998) are shown to 2 decimal places,
15 those estimated for standard increments are shown to 3 decimal places.
16 The male RR for lung cancer are positive and are statistically significant for all PM10
17 indicators. Male RR reported are positive and predominantly significant for O3 indicators, except
18 for mean O3, number of O3 exceedances > 60 ppb, and in former smokers. Reported RR for
19 mean SO2 are positive and significant except when restricted to proximate monitors. RR for
20 mean NO2 is positive but not significant.
21 The very high RR for mean PM10 for males (31.1) may be attributable to the small number
22 of cases (N = 16) and the large standard increment (50 //g/m3) used. When data are restricted to
23 subjects with at least 80 percent A/B quality data (within 32 km of the residence), the RR is
24 reduced to 9.26 over 50 //g/m3. The RR over the IQR of 24 //g/m3 in the full data set is 5.21, so
25 that the use of the IQR may be more appropriate for the exposure in long-term studies. The
26 female RR reported are much smaller, not being statistically significant for any indicator of PM10
27 or O3, and statistically significant only for mean SO2.
28 Extensive multi-pollutant analyses were carried out, but the results are not described in
29 detail. Regression coefficients for PM10 and SO2 were not reduced when O3 or NO2 were added
30 to the single-pollutant models for males. The regression coefficients for the two-pollutant model
31 with PM10 and SO2 remained highly positive and significant, which the authors suggest may be
October 1999 6-114 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-30. RELATIVE RISK OF LUNG CANCER INCIDENCE IN MALES, BY
AIR POLLUTANT, FOR ADVENTIST HEALTH STUDY
Pollution Index
PM10>40 ,ug/m3
PM10>50 Mg/m3
PM10>60 Mg/m3
PM10>80 ,ug/m3
PM10>100 //g/m3
PM10 mean
SO2 mean
NO2 mean
O3>60 ppb
O3>80 ppb
O3>100 ppb
O3>120 ppb
O3>150 ppb
O3 mean
PM10>100 Mg/m3
O3>100 ppb
O3>100 ppb
PM10>100 Mg/m3
O3>100 ppb
SO2 mean
PM10 mean
SO2 mean
Pollution Incr.
139 d/y (IQR)
149 d/y (IQR)
132 d/y (IQR)
78 d/y (IQR)
30 d/y
50 Mg/m3
3. 7 ppb
2.0 ppb
935 h/y
756 h/y
556 h/y
367 h/y
185 h/y
2.1 ppb
30 d/y
556 h/y
556 h/y
30 d/y
556 h/y
3. 7 ppb
50 ptg/m3
3. 7 ppb
Covariate Model or Sub-Group
standard
standard
standard
standard
standard
standard
standard
standard
standard
standard
standard
standard
standard
standard
never smokers
never smokers
past smokers
high population density
high population density
high population density
> 80% data from monitors within
20 miles of residence
> 80% data from monitors within
20 miles of residence
RR
4.50
4.96
4.72
3.43
2.127
31.147
2.66
1.45
2.14
2.96
3.56
3.75
3.61
2.23
2.102
4.48
2.15
2.865
10.18
3.22
9.256
2.18
RRLCL
1.31
1.54
1.69
1.71
1.454
3.978
1.62
0.67
0.82
1.09
1.35
1.55
1.78
0.79
1.325
1.25
0.42
1.794
2.44
1.87
1.135
0.92
RRUCL
15.44
16.00
13.18
6.88
3.112
243.85
4.39
3.14
5.62
8.04
9.42
9.90
7.35
6.34
3.335
16.04
10.89
4.574
42.45
5.54
75.516
5.20
Source: Beeson et al. (1998).
October 1999
6-115
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-31. RELATIVE RISK OF LUNG CANCER INCIDENCE IN FEMALES, BY
AIR POLLUTANT, FOR ADVENTIST HEALTH STUDY
Pollution Index
PM10>50 ,ug/m3
PM10>60 Mg/m3
SO2 mean
O3>100 ppb
PM10>100 //g/m3
SO2 mean
PM10 mean
SO2 mean
Pollution Incr.
149 d/y (IQR)
132d/y(IQR)
3. 7 ppb
556 h/y
30 d/y
3. 7 ppb
50 Mg/m3
3. 7 ppb
Covariate Model or Sub-Group
standard
standard
standard
standard
high population density
high population density
> 80% data from monitors
within 20 miles
> 80% data from monitors
within 20 miles
RR
1.21
1.25
2.14
0.94
1.089
2.11
2.425
2.52
RRLCL
0.55
0.57
1.36
0.41
0.726
1.32
0.310
1.19
RRUCL
2.66
2.71
3.37
2.16
1.633
3.38
19.004
5.33
Source: Beeson et al. (1998).
1 associated with independent effects of PM10 and SO2 on lung cancer incidence. PM10 was more
2 strongly correlated with lung cancer in males than the other pollutants. For females, the SO2
3 coefficient remained significant when co-pollutants were added one at a time, and was the air
4 pollutant most strongly associated with lung cancer in females.
5
6 Relationship to Earlier AHSMOG Studies
1 The results of the preceding two studies are somewhat different than those of earlier studies
8 using the same cohort. Abbey et al. (1991) reported completely non-significant relationships
9 between total ('all natural causes') mortality and air pollution. The RR for 1000 h/y of
10 TSP > 200 //g/m3 was 0.99 (CI 0.87-1.13), and for 500 h/y of O3 > 100 ppb was
11 1.00 (CI 0.89-1.12), after 10 years of follow-up.
12 Abbey et al. (1991) report no statistically significant increases in all malignant neoplasms
13 for males attributable to air pollution. The RR for 1000 h/y of TSP > 200 //g/m3 was
14 0.96 (CI 0.68-1.36), and for 500 h/y of O3 > 100 ppb was 1.09 (CI 0.80-1.47), after 10 years of
15 follow-up. However, there was a statistically significant increase in all malignant neoplasms in
16 females. The RR for females attributed to 1000 h/y of TSP > 200 //g/m3 was 1.37 (CI 1.05-1.80).
October 1999 6-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 Neoplasms in females attributed to 500 h/y > 100 ppb were much less significant with
2 RR= 1.03 (CI 0.81-1.32).
3 There were only 17 cases of respiratory cancer in this study, so while the estimated RR
4 were large (1.72 for TSP and 2.25 for O3), the confidence intervals were also very wide
5 (CI 0.81-3.65 for TSP and 0.96-5.31 for O3 effects). Findings of significant effects of TSP on
6 respiratory symptoms but not lung cancer may be attributed to the longer latency time of cancers.
7 Additional results reported by Mills et al. (1991) further elaborated on site-specific cancers
8 among females, but only breast cancers (RR = 1.51) and 'other' cancers (RR = 1.65) showed a
9 marginal relationship (P = 0.10) to TSP exceedances in the 10-year follow-up. These results are
10 also summarized in (Abbey et al., 1995).
11
12 6.3.3.3 Relationship of AHSMOG to Six Cities and ACS Study Findings
13 The results of the recent AHSMOG mortality studies (Abbey et al., 1999) are compared
14 with the earlier Six Cities study (Dockery et al., 1993) and ACS study (Pope et al., 1995) to a
15 greater level of detail than in the 1996 PM AQCD. Tables 6-32, 6-33, and 6-34 compare the
16 estimated RR for total, cardiopulmonary, and lung cancer mortality respectively among the
17 studies. The PM indices used are the mean PM10 concentration for the Six Cities and AHSMOG
18 studies (increment 50 //g/m3), and the mean PM25 and SO4 concentrations (increments 25 and
19 15 //g/m3 respectively) for the ACS study. The comparisons for the Six Cities and ACS studies
20 have been translated from published RR for the most polluted vs. least polluted city for PM10,
21 PM25, and SO4. Results are shown by sex and smoking status. The AHSMOG subjects are
22 classified as 'non-smokers', although some former smokers are included. The ACS study
23 combines past and current smokers into an 'ever smoker' category, although long-term past
24 smokers are at much lower risk than current smokers (U.S. Department of Health and Human
25 Services, 1990). The number of subjects in these studies varies greatly, partially accounting for
26 differences 8,111 subjects in the Six Cities study, compared to 295,223 subjects in the 50 fine
27 particle cities and 552,138 subjects in the 151 sulfate cities of the ACS study.
28 Table 6-32 shows relative risks for total mortality at comparable standard increments. RR
29 is generally highest for the Six Cities study. The AHSMOG study found a much smaller RR for
30 women than did the other studies, whereas the effect for males was similar to non-smokers in the
31 ACS study and marginally significant. RR among the three studies varied substantially with
October 1999 6-117 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-32. RELATIVE RISK (RR) OF TOTAL MORTALITY IN THREE
PROSPECTIVE COHORT STUDIES, BY SEX AND SMOKING STATUS
Sex Smoking Status
F NON-SMOKER
PAST
PAST +
CURRENT
CURRENT
M NON-SMOKER
PAST
PAST +
CURRENT
CURRENT
Study
Six Cities
ACS
AHSMOG
Six Cities
ACS
Six Cities
Six Cities
ACS
AHSMOG
Six Cities
ACS
Six Cities
PM
Index
PM10
PM2.5
S04
PM10
PM10
PM2.5
S04
PM10
PM10
PM2.5
S04
PM10
PM10
PM2.5
S04
PM,n
PM
Inc.
50
25
15
50
50
25
15
50
50
25
15
50
50
25
15
50
RR
1.280
1.215
1.147
0.879
1.999
1.102
1.104
1.442
1.568
1.245
1.104
1.242
1.611
1.164
1.104
1.858
RRLCL
0.704
1.020
1.045
0.713
0.704
0.898
0.977
0.719
0.674
1.000
0.977
0.955
0.930
1.051
1.037
1.090
RRUCL
2.345
1.440
1.261
1.085
5.632
1.338
1.240
3.166
3.678
1.554
1.247
1.616
2.825
1.297
1.176
3.166
Sources: Dockery et al. (1993); Pope et al. (1995); Abbey et al. (1999).
1
2
3
4
5
6
7
sex and smoking categories. Six of the 16 independent analyses showed significant positive RR
(LCL > 1.0), but subsetting the data allowed less power to detect effects than the whole data sets
would have allowed. Neither of the AHSMOG RR were significant using the mean as the PM10
index, but another PM10 index (exceedances over 100 //g/m3) was significant for males.
Table 6-33 shows relative risks for cardiopulmonary mortality at comparable standard
increments. RR is highest for the Six Cities study, which did not report separate effects by sex
and smoking status. The AHSMOG study found a much smaller cardiopulmonary RR for
October 1999
6-118
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-33. RELATIVE RISK (RR) OF CARDIOPULMONARY MORTALITY IN
THREE PROSPECTIVE COHORT STUDIES, BY SEX AND SMOKING STATUS
Sex Smoking Status
F NON-
SMOKERS
PAST +
CURRENT
M NON-
SMOKERS
PAST +
CURRENT
F+M ALL
Study
ACS
AHSMOG
AHSMOG -
CRC
ACS
ACS
AHSMOG
AHSMOG -
CRC
ACS
Six Cities
PM
Index
PM25
S04
PM10
PM10
PM25
S04
PM25
S04
PM10
PM10
PM25
S04
PMin
PM
Inc.
25
15
50
50
25
15
25
15
50
50
25
15
50
RR
1.585
1.316
0.841
1.219
1.276
1.219
1.245
1.205
1.219
1.537
1.235
1.126
1.744
RRLCL
1.235
1.147
0.639
0.739
0.918
1.008
0.929
1.023
0.862
0.879
1.061
1.037
1.202
RR
UCL
2.039
1.518
1.107
2.011
1.760
1.465
1.668
1.412
1.616
2.688
1.440
1.233
2.501
Sources: Dockery et al. (1993); Pope et al. (1995); Abbey et al. (1999).
1
2
3
4
5
6
7
women than did the other studies. However, the RR for male non-smokers was much more
similar to the ACS studies than for female non-smokers. RR for the AHSMOG endpoint CRC
('contributing respiratory causes') was more similar to the ACS findings for women, but higher
in men, although the confidence intervals are very wide. Seven of 13 of the independent analyses
showed significant positive RR (LCL > 1.0). The AHSMOG cardiopulmonary RR using mean
PM10 were not significant. However, the 100 //g/m3 exceedance index for males was nearly so.
Table 6-34 shows relative risks for lung cancer mortality at comparable standard
increments. RR was highest for males in the AHSMOG study, and statistically significant. The
AHSMOG study also found a larger RR for women than did the other studies. The only other
October 1999
6-119
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-34. RELATIVE RISK (RR) OF LUNG CANCER MORTALITY IN THREE
PROSPECTIVE COHORT STUDIES, BY SEX AND SMOKING STATUS
Sex Smoking Status
F NON-
SMOKERS
PAST +
CURRENT
M NON-
SMOKERS
PAST +
CURRENT
F+M ALL
Study
ACS
AHSMOG
ACS
ACS
AHSMOG
ACS
Six Cities
ACS
PM
Index
PM2.5
S04
PM10
PM2.5
S04
PM2.5
S04
PM10
PM2.5
S04
PM10
PM2.5
S04
PM
Inc.
25
15
50
25
15
25
15
50
25
15
50
25
15
RR
0.644
1.432
1.808
0.949
1.074
0.483
1.261
12.385
1.123
1.316
1.744
1.031
1.261
RRLCL
0.203
0.731
0.343
0.563
0.781
0.086
0.501
2.552
0.827
1.104
0.689
0.796
1.082
RRUCL
2.091
2.800
9.519
1.595
1.479
2.714
3.190
60.107
1.533
1.577
4.390
1.338
1.465
Sources: Dockery et al. (1993); Pope et al. (1995); Abbey et al. (1999).
1 significant finding for lung cancer was in past and current male smokers in the ACS 151-city
2 sulfate study. This pattern is in some contrast with the findings of the other studies, and merits
3 further analysis.
4 There is no obvious statistically significant relationship between PM effect sizes, gender,
5 and smoking status across these studies. The AHSMOG studies show no statistically significant
6 relationships between PM10 and total mortality or cardiovascular mortality for either sex, and
7 only for male lung cancer incidence and lung cancer deaths in a predominantly non-smoking
8 sample. The ACS results, in contrast, show similar and significant associations with total
9 mortality for both "never smokers" and "ever smokers", although the ACS cohort may include a
10 substantial number of long-term former smokers with much lower risk than current smokers.
October 1999
6-120
DRAFT-DO NOT QUOTE OR CITE
-------
1 The Six Cities study cohort shows the strongest evidence of a higher PM effect in current
2 smokers than in non-smokers, with female former smokers having a higher risk than male former
3 smokers. This study suggests that smoking status may be viewed as an "effect modifier" for
4 ambient PM, just as smoking may be a health effect modifier for ambient ozone (Cassino et al.,
5 1999).
6 In summary, the recent AHSMOG studies find positive but non-significant excesses of total
7 and cardiopulmonary mortality associated with mean PM10 exposure in males, but not in females.
8 The male RR in AHSMOG are similar to those in male non-smokers in the ACS study. There is
9 a large and statistically significant RR for lung cancer in males in the AHSMOG cohort, which is
10 much larger than in the other studies. While all of these studies find some associations with
11 mean PM10 or PM2 5, there are differences among specific endpoints and subgroups that remain to
12 be resolved.
13 It is interesting to note, in relation to the above discussion, that a comparison of the 6-Cities
14 Study non-smoker RRs with the 6-Cities results in Table 6-32 for smokers indicates that larger
15 and more significant effects of ambient PM pollution are found for smokers than non-smokers.
16 This suggests that smoking is an effect modifier that increases the adverse effects of ambient
17 pollution. This trend is consistent with air pollution effect causality, as smokers represent a
18 compromised population, logically more likely to be adversely affected by air pollution. This
19 may also explain why the reported AHSMOG study RRs are generally not significant, in contrast
20 with the overall 6-Cities study (but consistent with the 6-Cities nonsmoker results), as there are
21 no identified smokers among the AHSMOG study group to "drive up" the overall significance of
22 the air pollution effect. This again indicates that more years of follow-up may be required to see
23 any statistically significant total mortality effects in both the AHSMOG and 6-Cities studies'
24 non-smoking populations.
25 When the ACS study results are compared with the AHSMOG study results for SO4=
26 (as PM10 was not considered in the ACS study), the total mortality effect sizes per 15 //g/m3 SO4=
27 for the males in the AHSMOG population are seen to fall between the 6-Cities and the ACS
28 effect estimates for males: RR= 1.28 for AHSMOG male participants; RR= 1.61 for 6-Cities
29 Study male non-smokers, and; RR=1.10 for never smoker males in the ACS study. The
30 AHSMOG study 95% confidence intervals encompass both of those other studies' sulfate RRs.
31
October 1999 6-121 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.3.3.4 Population-Based Mortality Studies in Children
2 The older cross-sectional mortality studies suggest that the very young may represent an
3 especially susceptible sub-population. Lave and Seskin (1977) found mortality among those
4 0-14 years of age to be significantly associated with TSP. More recently, Bobak and Leon (1992)
5 studied neonatal (ages less than one month) and post-neonatal mortality (ages 1-12 months) in
6 the Czech Republic, finding significant and robust associations between post-neonatal mortality
7 and PM10, even after considering other pollutants. Post-neonatal respiratory mortality showed
8 highly significant associations for all pollutants considered, but only PM10 remained significant
9 in simultaneous regressions. The exposure duration was longer than a few days, but shorter than
10 in the adult prospective cohort studies. Thus, the limited available studies reviewed in the last
11 PM Criteria documented were highly suggestive of an association between ambient PM
12 concentrations and infant mortality, especially among post-neonatal infants.
13 Since the 1996 PM AQCD, Woodruff et al. (1997) used cross-sectional methods to
14 follow-up on the reported post-neonatal mortality association with ambient PM10 pollution in a
15 U.S. population. This study involved an analysis of a cohort consisting of approximately
16 4 million infants born between 1989 and 1991 in 86 metropolitan statistical areas (MSAs). Data
17 from the National Center for Health Statistics-linked birth/infant death records were combined at
18 the MSA level with measurements of PM10 from the EPA's Aerometric data base. Infants were
19 categorized as having high, medium, or low exposures based on tertiles of PM10 averaged over
20 the first 2 postnatal months. Total and cause-specific postneonatal mortality rates were examined
21 using logistic regression to control for demographic and environmental factors. Overall
22 postneonatal mortality rates per 1,000 live births were 3.1 among infants with low PM10
23 exposures, 3.5 among infants with medium PM10 exposures, and 3.7 among highly exposed
24 infants. After adjustment for other covariates, the odds ratio (OR) and 95% confidence intervals
25 for total postneonatal mortality for the high exposure versus the low exposure group was
26 1.10 (CI=1.04-1.16). In normal birth weight infants, high PM10 exposure was associated with
27 mortality for respiratory causes (OR = 1.40, CI=1.05-1.85) and sudden infant death syndrome
28 (OR = 1.26, CI=1.14-1.39). Among low birth weight babies, high PM10 exposure was associated,
29 but not significantly, with mortality from respiratory causes (OR = 1.18, CI=0.86-1.61).
30 However, other pollutants, such as O3, were not considered as possible confounders. This study
31 confirms past study results indicating that outdoor PM air pollution is associated with risk of
October 1999 6-122 DRAFT-DO NOT QUOTE OR CITE
-------
1 postneonatal mortality (e.g., Bobak and Leon, 1992), but the lack of consideration of other air
2 pollutants in this new study reduces the certainty that PM is the specific causal outdoor air
3 pollutant in this case.
4
5 6.3.3.5 Studies by Particulate Matter Size-Fraction and Composition
6 Particulate matter mass varies widely over time and from place to place in composition, and
7 this should affect the toxicity of that mass. Unfortunately, only a limited number of the chronic
8 exposure studies have included direct measurements of the chemical-specific constituents of the
9 PM mass.
10 The semi-individual studies also investigated the relative roles of various PM components
11 in the air pollution association with mortality. As shown in Table 6-35, the Harvard 6-Cities
12 study (Dockery et al., 1993) results indicated that the PM25 and SO4= RR associations
13 (as indicated by their respective 95% CI's and t-statistics) were stronger than those for the
14 coarser mass components. However, the effects of sulfate and non-sulfate PM25 are indicated to
15 be quite similar. Acid aerosol (H+) exposure was also considered by Dockery et al. (1993), but
16 only less than one year of measurements collected near the end of the follow-up period were
17 available in most cities, so the 6-Cities results were much less conclusive for the acidic
18 component of PM than for these other PM metrics (that, in contrast, were measured over many
19 years during the study). The 6-Cities study also yielded total mortality RR estimates for the
20 reported range across those cities of PM25 and SO4= concentrations that, although not statistically
21 different, were roughly double the analogous RR's for the TSP-PM15 and PM15_25 mass
22 components.
23 Table 6-36 presents comparative PM25 and SO4= results from the ACS study that indicate
24 that, although the RR differences were not statistically significant across pollutants, the SO4= RRs
25 were in every case more strongly significant than those for the PM2 5 across the various mortality
26 cause classifications considered, especially for lung cancer (SO4= t=2.92 vs. t=0.38 for PM25).
27 The most recent AHSMOG study analysis (Abbey et al., 1999) employed PM10 as its PM
28 mass index, finding significant associations with total and by-cause mortality, even after
29 controlling for potentially confounding factors (including other pollutants). This analysis also
30 considered SO4= as a PM index for all health outcomes studied except lung cancer, but SO4= was
31 not as strongly associated as PM10 with mortality, and was not found to be statistically significant
October 1999 6-123 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-35. COMPARISON OF ESTIMATED RELATIVE RISKS
FOR ALL-CAUSE MORTALITY IN SIX U.S. CITIES ASSOCIATED
WITH THE REPORTED INTER-CITY RANGE OF CONCENTRATIONS
OF VARIOUS PM METRICS
(Dockery et al., 1993; U.S. Environmental Protection Agency, 1996)
PM Species
S04=
PM25 - SO4=
PM25
PM15.2.5
TSP-PM,,
Concentration Range
(Mg/m3)
8.5
8.4
18.6
9.7
27.5
Relative Risk RR Relative Risk
Estimate 95% CI t-Statistic
1.29
1.24
1.27
1.19
1.12
(1.06-1.56)
(1.16-1.32)
(1.06-1.51)
(0.91-1.55)
(0.88-1.43)
3.67
8.79
3.73
1.81
1.31
TABLE 6-36. COMPARISON OF REPORTED SO4= AND PM25 RELATIVE
RISKS FOR VARIOUS MORTALITY CAUSES IN THE ACS STUDY
(POPE et al., 1995)
Mortality Cause
All Cause
Cardiopulmonary
Lung Cancer
SO4=
(Range = 19.9,ug/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 ptg/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
1 for any mortality category. The significant mortality associations found for PM10 contrasts with
2 previously published AHSMOG study PM analyses that found weaker mortality associations
3 with TSP (Abbey et al., 1991). Although the longer follow-up time in this new analysis may
4 have also contributed, the greater strength of association by PM10 vs. TSP is consistent with the
5 Harvard Six-City study results presented in Table 6-35, as well as with the Ozkaynak and
6 Thurston (1987) cross-sectional comparisons of mortality associations with the various PM
7 fractions.
October 1 999
6- 1 24
DRAFT-DO NOT QUOTE OR CITE
-------
1 Overall, the semi-individual studies conducted to-date collectively confirm the cross-
2 sectional study indications that, as opposed to the more coarse mass fractions, the fine mass
3 component of PM (and sometimes including its acidic sulfate constituent) are strongly correlated
4 with mortality.
5
6 6.3.3.6 Shortening-of-Life Associated With Long-Term Exposure to Particulate Matter of
7 Ambient Origins
8 The public health burden of mortality associated with exposure to ambient PM depends not
9 only on the increased risk of death, but also on the length of life shortening that is attributable to
10 those deaths. However, the 1996 PM AQCD concluded that confident qualitative determination
11 of years of life lost to ambient PM exposure is not yet possible; life shortening may range from
12 days to years (U.S. Environmental Protection Agency, 1996). Fortunately, a new analysis has
13 now provided a first estimate of the life-shortening associated with chronic PM exposure.
14
15 Life-Shortening Estimates Based on Semi-Individual Cohort Study Results
16 Brunekreef (1997) reviewed the available evidence of the mortality effects of long-term
17 exposure to particulate matter air pollution and, using life table methods, derived an estimate of
18 the reduction in life expectancy that is associated with those effect estimates. Based on the
19 results of Pope et al. (1995) and Dockery et al. (1993), a relative risk of 1.1 per 10 //g/m3
20 exposure over 15 years was assumed for the effect of particulate matter air pollution on men
21 25-75 years of age. A 1992 life table for men in the Netherlands was developed for
22 10 successive five-year categories that make up the 25-75 year old age range. Life expectancy of
23 a 25 year old was then calculated for this base case, and then compared with the calculated life
24 expectancy for the PM exposed case where the death rates were increased in each age group by a
25 factor of 1.1. A difference of 1.11 years was found between the "exposed" and "clean air"
26 cohorts' overall life expectancy at age 25. Looked at another way, this would imply that the
27 expectation of the lifespan of the persons who actually died from air pollution was reduced by
28 more than 10 years, since they represent a small percentage of the entire cohort population.
29 A similar calculation by the authors for the 1969-71 life table for U.S. white males yielded an
30 even larger reduction of 1.31 years for the entire population's life expectancy at age 25. Thus,
October 1999 6-125 DRAFT-DO NOT QUOTE OR CITE
-------
1 this study's table calculations imply that relatively small differences in long-term exposure to
2 particulate matter of outdoor origins can have substantial effects on life expectancy.
3
4 Potential Effects of Infant Mortality on Life-Shortening Estimates
5 Deaths among children can logically have the greatest influence on a population's overall
6 life expectancy, but the Brunekreef (1997) life table calculations did not consider any possible
7 effects of long-term air pollution exposure on the population aged less than 25 years.
8 As discussed above, some of the older cross-sectional studies and the more recent studies by
9 Bobak and Leon (1992) and Woodruff et al. (1997) suggest that infants may be among the
10 sub-populations that are especially affected by long-term PM exposure. Thus, although it is
11 difficult to quantify, any premature PM associated mortality that occurs among children due to
12 long-term PM exposure, as suggested by these studies, would significantly increase the overall
13 population life shortening over and above that estimated by Brunekreef (1997) for long-term PM
14 exposure to adults 25 years and older.
15
16 6.3.3.7 Effects of Exposure to Multiple Pollutants
17 Little is known about the effects of exposure to multiple pollutants in prospective cohort
18 studies. While qualitative results of multiple-pollutant analyses have been reported for the
19 AHSMOG study (Abbey et al., 1999; Beeson et al., 1998), the numerical values and the
20 correlations of the regression coefficients have not been published. The AHSMOG study allows
21 detailed evaluation of multiple-pollutant effects because each subject has an individual exposure
22 "history" for PM and gaseous pollutants, based on his /her residence history and occupation.
23 This allows some degree of inter-individual variation of exposures, even among individuals
24 living in the same census tract. However, it must be acknowledged that there is still substantial
25 correlation of pollutant concentration values because of the relatively small number of
26 community air monitors from which the assigned exposure histories have been derived.
27 Many of the results from multiple-pollutant analyses in the AHSMOG study have been
28 reported in terms of levels of exceedance frequencies, such as the number of days per year with
29 PM10 > 100 //g/m3, the number of hours per year with ozone > 100 ppb, and so on. These are
30 extremely useful indices, but not readily comparable to analyses reported by other investigators.
31 The construction of comparable indices appears feasible for the Six Cities study, since PM and
October 1999 6-126 DRAFT-DO NOT QUOTE OR CITE
-------
1 gaseous co-pollutant measurements were made daily or every 2nd day during the course of the
2 multi-year study. However, assignment of individual exposures would not result in variability
3 among individuals in each of the Six Cities, unless additional air pollution information were
4 available for the same communities from which differential exposure histories could be
5 constructed. Otherwise, only 5 'degrees of freedom' would be available for statistical analyses.
6 It is unlikely that comparable indices of high-level exposure exceedance could be
7 constructed for subjects in the ACS study, since the PM indices were observed only for a few
8 years, mostly prior to the interval over which vital status was ascertained. Additional aerometric
9 data would be needed for multi-pollutant analyses. If appropriate data for other air pollutants
10 could be obtained for all cities, there could be up to 49 degrees of freedom in the fine-particle
11 cities and 150 degrees of freedom in the sulfate cities from which multi-pollutant or other
12 ecological analyses could be carried out, even without subject-specific exposure indices.
13 Analyses to address similar questions for the Six Cities and ACS studies are now being
14 carried out by investigators at the University of Ottawa as part of a project sponsored by the
15 Health Effects Institute. Results will be included when they become available.
16 Single-pollutant results about PM components are informative, however, as shown in
17 Table 6-37 for total mortality, and in Table 6-38 for cardiopulmonary causes. The t-statistics are
18 compared for studies where appropriate: mean PM10, PM10_25, PM25, and sulfate for the Six Cities
19 (Dockery et al, 1993); mean PM25 and sulfate for ACS (Pope et al, 1995); mean PM10 and
20 sulfate, and PM10 exceedances of 100 //g/m3 for AHSMOG (Abbey et al., 1999).
21 Estimates for Six Cities parameters were calculated in two ways: (1) mortality RR for most
22 versus least polluted city in (Table 3, Dockery et al., 1993) adjusted to standard increments;
23 (2) ecological regression fits in (Table 12-18, U.S. Environmental Protection Agency, 1996).
24 The eastern and mid-western Six Cities suggest a strong and highly significant relationship for
25 fine particles and sulfates, a slightly weaker but still highly significant relationship to PM10, and a
26 marginal relationship to PM10_25. The ACS study looked at a broader spatial representation of
27 cities, and found a stronger statistically significant relationship to PM25 than to sulfate (no other
28 pollutants were examined).
29 The AHSMOG study at California sites where sulfate levels are typically low found
30 significant effects in males for PM10 exceedances, and a marginal effect of mean PM10, but no
31 PM effects for females or for sulfates. These results are consistent in suggesting a statistically
October 1999 6-127 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-37. COMPARISON OF TOTAL MORTALITY RELATIVE RISK
ESTIMATES AND T-STATISTICS FOR PM COMPONENTS IN THREE
PROSPECTIVE COHORT STUDIES
PM Index Study
PM10 (50 Mg/m3) Six Cities
AHSMOG
PM2 5 (25 Mg/m3) Six Cities
ACS (50 cities)
SO4= ( 1 5 Mg/m3) Six Cities
ACS (151 cities)
AHSMOG
Days/y with PM10>100 (30 AHSMOG
days)
PM10.2 5 (25 ,ug/m3 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
1.504(1); 1.530(2)
1.280(1)
1.242
1.364(1); 1.379(2)
1.207(1)
1.174
1.245
1.504(1); 1.567(2)
1.359
1.111
1.104
1.279
1.082
1.814(1); 1.560(2)
1.434(1)
t Statistic
2.94(1);
3.27 (2)
0.81(1)
1.616
2.94(1);
3.73 (2)
0.81(1)
4.35
1.960
2.94(1);
3.67 (2)
0.81(1)
5.107
1.586
0.960
2.183
2.94(1,3);
1.816(2)
0.81 (1)
(1) Method 1 compares Portage vs. Steubenville (Table 3, Dockery et al., 1993).
(2) Method 2 is based on ecologic regression models (Table 12-18, U.S. Environmental Protection Agency,
1996).
(3) Method 1 not recommended for PM10-2.5 analysis due to high concentration in Topeka.
1 significant relationship of PM10 to excess mortality, less so for cardiopulmonary causes than for
2 contributing respiratory causes, as shown in Table 6.3.3.7.2. The findings to date are more
3 ambivalent about the role of sulfates, which tend to be much higher at eastern sites than at
4 western sites. The AHSMOG investigators are currently investigating effects of fine and coarse
5 particles.
October 1999
6-128
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-38. COMPARISON OF CARDIOPULMONARY MORTALITY RELATIVE
RISK ESTIMATES AND T-STATISTICS FOR PM COMPONENTS IN THREE
PROSPECTIVE COHORT STUDIES. 'MALE NON. - CRC' IDENTIFIES SUBJECTS
WHO DIED OF ANY CONTRIBUTING NON-MALIGNANT RESPIRATORY CAUSE
IN THE AHSMOG STUDY
PM Index
PM10(50Mg/m3)
PM2.5 (25 Mg/m3)
S04=(15,ug/m3)
Days/y with
PM10>100 (30 days)
PM10.2.5 (25 Mg/m3
Study
Six Cities
AHSMOG
Six Cities
ACS (50 cities)
Six Cities
ACS (151 cities)
AHSMOG
AHSMOG
Six Cities
Subgroup
All
Male Nonsmoker
Male Non. - CRC
All
All
Male
Male Nonsmoker
All
All
Male
Male Nonsmoker
Male Nonsmoker
Male Non. - CRC
Male Nonsmoker
Male Non. - CRC
All
Relative Risk
1.744(1)
1.219
1.537
1.527(1)
1.317
1.245
1.245
1.743 (1)
1.190
1.147
1.205
1.279
1.219
1.082
1.188
2.251 (1)
t Statistic
2.94(1)
1.120
2.369
2.94(1)
4.699
3.061
1.466
2.94(1)
5.470
3.412
2.233
0.072
0.357
1.310
2.370
2.94(1,2)
(1) Method 1 compares Portage vs. Steubenville (Table 3, Dockery et al., 1993).
(2) Method 1 not recommended for PM10-2.5 analysis due to high concentration in Topeka.
1 6.3.3.8 Discussion
2 Gamble (1998) presents a critique of the earlier prospective cohort mortality studies,
3 structured around a substantially modified and altered version of Austin Bradford Hills's (1965)
4 nine considerations for inferring causality from epidemiology studies. He ignores Hill's caveats
5 and reservations about using the nine points as a check-list or set of criteria. The relevance of
6 community stationary air monitoring concentrations as surrogates for individual exposure to PM
7 of ambient origin was discussed at length in Chapter 5 of this document, and in the 1996 PM
October 1999 6-129 DRAFT-DO NOT QUOTE OR CITE
-------
1 AQCD. Individual differences in exposure to PM of non-ambient origin (such as cigarette
2 smoke, cooking, and resuspended house dust) are sometimes important contributors to total PM
3 exposure, but appear to be largely independent of exposure to PM of ambient origin. Gamble's
4 (1998) emphasis on total personal exposure appears to be misplaced, since the toxicity of
5 ambient PM may be quite different than the toxicity of non-ambient PM from indoor sources,
6 occupations, and personal activities such as smoking tobacco and simple comparison of TWA
7 daily PM concentrations may not adequately characterize differences in health effects from
8 short-term (much less long-term) PM exposures from particles with different chemical
9 properties. There is considerable direct observational evidence that short exposures to PM under
10 certain conditions, as in London during the 1952 fog episode, can greatly increase mortality and
11 sickness from cardio respiratory causes. The evidence presented earlier suggests some degree of
12 coherence for respiratory and cardiovascular effects in susceptible populations: the elderly,
13 children, and asthmatics. There is now also some evidence suggesting large PM mortality at a
14 "midrange" of exposure time scales, roughly 30 to 120 days of exposure (Schwartz, 1999b; Zeger
15 et al., 1998), as well as short-term effects over a few days, increasing the plausibility that there
16 may be even larger effects over longer time-scales. Evidence from the well-conducted
17 AHSMOG study (Beeson et al., 1998; Abbey et al., 1999) supports the earlier studies, with some
18 differences (little or no excess mortality in the female cohort; little cardiovascular effect,
19 considerable effects for lung cancer and for death with underlying respiratory causes). The
20 AHSMOG study used individualized exposure metrics based on the subjects' residence and work
21 history, and could therefore evaluate a number of multi-pollutant models, thus dealing with some
22 of Gamble's (1998) concerns. Further analyses of data from Dockery et al. (1993) and Pope et al.
23 (1995) being carried out under HEI sponsorship may also resolve some of Gamble's (1998)
24 concerns.
25 Valberg and Watson (1998) question PM health effects, but their paper focuses upon
26 raising possible alternative hypotheses that might confound the PM- health effects association.
27 The three alternative pathways the authors propose as possible are: (1) weather conditions;
28 (2) human behavior, and 3) indoor air pollution. No direct examples are provided of cases where
29 published PM- health effects were actually explainable by these factors, but arguments are made,
30 using largely anecdotal evidence, as to how these factors might be related to both health and daily
31 variations in PM, and there by might explain away the PM- health effects relationships. As in
October 1999 6-130 DRAFT-DO NOT QUOTE OR CITE
-------
1 Gamble (1998), the authors do not distinguish between total personal exposure to all PM vs. total
2 personal exposure to PM of outdoor origins only the latter of which is relevant to outdoor air
3 pollution epidemiology. The authors conclude that their analysis "supports the importance of
4 considering the factors that drive outdoor PM fluctuations in the first place to determine if the
5 observed mortality/morbidity changes may be caused by pathways other than the hypothetical
6 frank toxicity from inhalation of ambient PM."
7
8 6.3.3.9 Conclusions
9 A review of the studies summarized in the previous PM AQCD (U.S. Environmental
10 Protection Agency, 1996) indicates that past epidemiologic studies of chronic PM exposures
11 collectively indicate increases in mortality to be associated with long-term exposure to airborne
12 particles of ambient origins. The PM effect size estimates for total mortality from these studies
13 also indicate that a substantial portion of these deaths reflected cumulative PM impacts above
14 and beyond those exerted by acute exposure events.
15 The new ASHMOG study (Abbey et al., 1999) provides all-cause mortality RR estimates
16 for adult males that are quantitatively and qualitatively consistent with prior semi-individual
17 studies, especially the similarly designed 6-Cities study. Extensive new by-gender, by-cause, and
18 multiple pollutant sensitivity analyses, as well as a more comprehensive analyses of numerous
19 potentially uncontrolled factors in this study (such as of the effects of variations in the time spent
20 outdoors) provide important new evidence that is largely supportive of the mortality associations
21 with PM of ambient origins previously reported by the 6-Cities and ACS studies.
22 Published cross-sectional studies collectively also indicate that older adults and infants are
23 the age groups most affected by PM from ambient origins, while both the cross-sectional and
24 semi-individual studies indicate that those deaths involving respiratory disease (either malignant
25 or not) are especially associated with exposure to PM air pollution. These results are biologically
26 plausible and consistent with a causal relationship between mortality and exposure to PM of
27 outdoor origins.
28 With regard to the role of various PM constituents in the PM-mortality association, cross-
29 sectional studies have generally found that the fine particle component, as indicated either by
30 PM25 or sulfates, was the PM constituent most consistently associated with mortality.
31 In addition, the Six-cities prospective semi-individual study also indicates that the fine mass
October 1999 6-131 DRAFT-DO NOT QUOTE OR CITE
-------
1 components of PM are more strongly associated with the mortality effects of PM than the coarse
2 PM components.
3 Recent investigations of the public health implications of these PM mortality effect
4 estimates were also reviewed. Life table calculations by Brunekreef (1997) found that relatively
5 small differences in long-term exposure to airborne particulate matter of ambient origin can have
6 substantial effects on life expectancy. For example, a calculation for the 1969-71 life table for
7 U.S. white males indicated that a chronic exposure increase of 10 //g/m3 PM was associated with
8 a reduction of 1.31 years for the entire population's life expectancy at age 25. Also, new
9 evidence of infant mortality associations with PM exposure (Bobak and Leon, 1992; Woodruff
10 et al., 1997) may have the implication that life shortening in the entire population from long-term
11 PM exposure could well be significantly larger than that estimated by Brunekreef (1997).
12
13
14 6.4 DISCUSSION OF EPIDEMIOLOGY FINDINGS
15 6.4.1 Overview of Section
16 The purpose of this section is to evaluate the results of the epidemiology studies reported
17 above, and particularly to assess factors impacting the validity of the reported results (whether
18 positive or negative). These factors take a number of forms. Section 6.4.2 discusses the
19 specificity of the adverse health effects attributed to ambient PM exposure with respect to
20 diagnosed outcomes. Section 6.4.3 evaluates the consistency of morbidity attributable to short-
21 term and long-term ambient PM exposures. Section 6.4.4 similarly evaluates the consistency of
22 mortality attributable to short-term and long-term ambient PM exposures. Section 6.4.5 discusses
23 the coherence of the findings for different susceptible subpopulations.
24 The specificity of health effects attributed to PM size fraction is discussed in Section 6.4.6.
25 A variety of methodological issues are discussed in Section 6.4.7. Section 6.4.8 discusses some
26 of the approaches that have been used in synthesis of results from different cities. Finally,
27 Section 6.4.9 discusses the attribution of health effects to ambient PM alone, to ambient PM in
28 the presence of other air pollutants, and to PM as an index or co-indicator of a mixture of
29 pollutants.
30
October 1999 6-132 DRAFT-DO NOT QUOTE OR CITE
-------
1 Evaluation of the Quantitative Consistency of Effect Size
2 The quantitative consistency of effect size across different studies lends credibility to the
3 hypothesis that the partial contribution of a causal factor to the effect is the same under differing
4 conditions. The appropriate metric for determining consistency depends on the health endpoint.
5 For binary outcomes (e.g., FEVj < 85% of normal in a subject, versus normal), an odds ratio over
6 a standard PM increment is used here. For counts of discrete events in a population (e.g., daily
7 number of hospital admissions), a relative risk over a standard PM increment is used here.
8 For continuous endpoints (e.g., FEVj as liters or as percent of expected value), either the
9 percentage decrease or absolute decrease over a standard PM increment is used here. Many of
10 the statistical analyses are appropriately carried out on a logarithmic scale, and the results could
11 have been reported on a log-odds-ratio or log-relative-risk scale (or as excess risk). We prefer to
12 report results as OR or RR, which we believe more clearly shows the relatively modest health
13 risks for individuals or urban populations associated with current levels of PM (compared to the
14 London smog episode of 1952, say), although the aggregate population risk may be substantial.
15 This also shows that normally small biases of 2 or 3 percent attributable to methodology or study
16 design may seriously alter the conclusions of any study. For this reason, findings of similar
17 effects across studies as well as findings of differences between effect size estimates should be
18 interpreted cautiously.
19 Certain factors facilitate the comparison of studies:
20 (1) similarity in study design (subject selection or recruitment, retention, follow-up);
21 (2) similarity in data analysis (methodology, criteria for covariate adjustment, smoothing
22 or detrending);
23 (3) similarity in air pollution measurements (averaging times or lags, spatial averaging,
24 representativeness of monitors, instrumentation);
25 (4) similarity in environmental factors that may modify PM health effects.
26
27 Evaluation of the Statistical Significance of Effect Size
28 Statistical significance plays an important role in assessment of study results, but we
29 believe that it should not play a dominant role, particularly in the synthesis of results from
30 multiple studies. Even if each study were regarded as an independent replicate of an exactly
31 identical design, it would be normal to find some studies with statistically significant findings of
October 1999 6-133 DRAFT-DO NOT QUOTE OR CITE
-------
1 adverse health effects, some with findings of adverse health effects that are not statistically
2 significant, and some studies with findings of beneficial effects (including those that are
3 statistically significant). For this reason, causal conclusions about adverse health effects
4 associated with PM or other environmental factors are strengthened when a number of
5 independent studies find statistically significant adverse health effects, and when the majority of
6 studies find adverse rather than beneficial effects. However, far more than simple 'vote
7 counting' is required to interpret a body of evidence with mixed findings (see Section 6.4.8 on
8 research synthesis).
9 In reality, any simple description of a specific study finding as 'positive' or 'negative' is
10 likely to be misleading, in view of the large number of model selection and data analysis
11 strategies that may be applied. It is hardly surprising that different analyses of the same data set
12 may find greater or smaller statistical significance associated with a given PM effect size
13 estimate. It is often possible to separate more adequate analyses from less adequate based on
14 generally accepted strategies for data analysis, but some element of judgement is inevitably
15 involved. For this reason, the results of different analyses are often presented in different forms,
16 to assist the reader in drawing an independent conclusion.
17 Results are usually presented in two forms: (1) 95 percent confidence intervals of OR or
18 RR for a standard PM increment, to express uncertainty (these can also be used to test statistical
19 significance by comparing the lower confidence limit with 1.0); (2) a t-statistic for the PM or
20 co pollutant regression coefficient in a generalized linear model, which allows a consistent
21 single-digit plus single-decimal format for easy visual comparison, although P-values and other
22 equivalent representations could be used. Any equivalent representation of statistical
23 significance would be equally subject to the same criticism, that it combines elements of sample
24 size, effect size, internal uncertainty (standard error), and model selection. We find t-statistics
25 particular convenient for evaluating the sensitivity of alternative PM and co-pollutant models
26 fitted to the same data sets.
27
28
29
30
October 1999 6-134 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.4.2 Relationship of Ambient Concentrations to Specific Diseases and
2 Age Groups
3 New studies have expanded the overall database for morbidity outcomes in relation to
4 ambient PM concentration. A brief summary of age groups is followed by discussion of asthma,
5 non-asthma, and finally, cardiovascular disease.
6 Most studies have been designated to examine age groups with the largest potential yield,
7 the elderly (>65 years) in hospitalized studies, and children in panel studies of respiratory
8 symptoms and pulmonary function outcomes. Since the subjects examined in most panel studies
9 were a uniform age group such as 8 to 12 years of age, no difference by age can be drawn. Those
10 studies with adults in the panels did not examine the data by age. Some age-related outcomes are
11 available from emergency department (ED) visits and hospital admission studies. Delfino et al.
12 (1997) report no significant associations between air pollution and ED respiratory visits for
13 persons 2 to 6 years of age, but positive association for patients over 64 years of age. Medina
14 et al. (1997) reported that the relationship between asthma visits and air pollution was strong for
15 children. Spix et al. (1998) studied TSP and BS and hospital admissions data by older (>65) and
16 younger (15 to 64 years) people which yielded mixed results. Sunyer et al. (1997) report
17 nonsignificant outcomes for BS and age group 15 to 64 and <15 years. However, the majority of
18 hospital admission studies report positive outcomes for the age group >65 years of age.
19 Comparative analyses by age group have not been done, usually. Morbidity studies provide
20 limited insight into potential age differences for outcome because age is usually part of the study
21 design, to produce a more homogeneous study group with larger effects, such as age >65.
22
23 6.4.2.1 Asthma Studies
24 The asthma panel studies discussed in Section 6.2.2 were conducted by over 10 research
25 teams in various locations world-wide. As a group the studies examine health outcome effects
26 that are similar such as PEFR. Each study characterizes the clinical-symptomatic aspects in a
27 community of a sample of asthmatics, mainly children age 5 to 16 years of age observed in their
28 natural setting. Their asthma is "ideally" being treated to achieve the therapeutic purpose of
29 keeping them symptom free, with "normal" pulmonary function rates, normal activity levels, and
30 to prevent recurrent exacerbations of asthma (National Institutes of Health, 1991). They are mild
31 to moderate asthmatics. Characterization of their asthma is by symptom, pulmonary function and
October 1999 6-135 DRAFT-DO NOT QUOTE OR CITE
-------
1 medication use. In the new severity classification (National Institutes of Health, 1997) these
2 panel studies would be classified to include mild intermittent to mild persistent. As a group they
3 may differ from asthmatics are examined in studies of hospitalization or doctor visits for acute
4 asthmatic episodes who may have more severe asthma.
5 Severity of asthma is important in that these panel studies are examining primarily mild and
6 moderate asthmatics, but still those who require medication for the condition seen to demonstrate
7 enough disease or reaction to stimulus to be the appropriate group to study. The results for the
8 study may represent only that segment of the asthma population with these levels of disease.
9 Physician visits and hospitalization for asthma are better outcome measures for those with
10 moderate or severe asthma (Schwartz et al., 1993; Gordian et al., 1996; Tseng et al., 1992).
11
12 Study Characteristics
13 Most of these studies used a single ambient monitoring site to characterize PM ambient
14 exposure estimates. For PM10 and especially for PM2 5, exposure derived from such sites
15 (Janssen et al., 1997; U.S. Environmental Protection Agency, 1996; Wilson et al., 1998; Mage
16 et al., 1999) provide an estimate of PM ambient exposure for a panel of subjects such as
17 asthmatics. The study authors provide limited discussion of exposure estimates. While the
18 opportunity for measurement error is an important consideration in this circumstance, the major
19 potential effect is to weaken a possible association. Some of the studies used monitors that were
20 located in more residential areas in the community and thus may provide a better exposure
21 estimate for PM, others used multiple monitors and one examined personal exposure.
22 Most studies reported ambient PM10 results. PM25 was examined in two studies. Other
23 ambient PM measures (Black Smoke (BS) and SO4) were also used. For these studies, mean
24 PM10 levels range from a low of 13 //g/m3 in Finland to a high of 167 //g/m3 in Mexico City.
25 This Mexico City level is over 3 times more than each of the other levels and is unique compared
26 to the others. Related 95% CI for these means or ranges show one day maximums above
27 100 //g/m3 in four studies with two of these above 150 //g/m3. The studies thus provide a
28 measure of differentiated levels of PM in the range of US cities. All the studies controlled for
29 temperature and several controlled for relative humidity.
30 Most panel studies are analyzed using a design which takes advantage of the repeated
31 measures. Linear models are often used for lung function and logistic models are used for
October 1999 6-136 DRAFT-DO NOT QUOTE OR CITE
-------
1 dichotomous outcomes. Meteorological variables are used as covariates. Perhaps the most
2 critical choice in the model is the choice of the lag of the pollution variable. The use of
3 medication is also an important confounder. Study subject number various from 12 to 164. Most
4 of the studies had cohort samples over 50, and all had gathered adequate subject-day data to
5 provide sufficient power for the conducted analysis. All the studies reported a positive outcome
6 for either O3 or PM.
7 Three asthma panel studies (Gielen et al, 1997; Peters et al, 1997b; Delfino et al, 1998)
8 used symptom-scoring approaches as opposed to measuring, the presence or absence of cough,
9 wheeze, sputum production (phlegm), shortness of breath and dyspnea (chest tightness). Since
10 the complex of symptoms and their clinical expression varies between asthmatic subjects,
11 outcome misclassification would be less in the symptom scoring approaches as opposed to
12 symptom specific approach.
13 Presenting lag periods with the strongest associations introduces potential bias since the
14 biological basis for lag structure may be related to effect. No biological basis for lag periods is
15 known, but some hypotheses can be proposed. Acute asthmatic reaction can occur 4-6 hours
16 after exposure and thus 0-day lag may be more appropriate than 1-day lags for that reaction.
17 Lag 1 may be more relevant for morning measurement of asthma outcome from the day before
18 and longer term lag, (i.e., 2-5 days may represent the outcome of an inflammatory mechanism).
19 There is too little information to predetermine the appropriate lag period. Additional research is
20 needed.
21 A qualitative summary of these studies examining ambient PM10 exposure on asthmatic
22 health outcomes indicate that as a group, the majority of studies report a positive odds ratio for
23 the relationship with almost half having 95% CI above 1.00. This is looking at the endpoints one
24 at a time. Viewing all the indicators together within a study may be a better test of the
25 relationship for an asthma attack. Examining all the studies as a group quantitatively describes a
26 stronger relationship.
27 The available group of studies examined PM10 and other measures, but did not study
28 outcomes for coarse, PM10_25. One study (Romieu et al., 1996) examining both PM10 and PM25
29 reported that PM2 5 had a larger effect.
30 Peters et al. (1997c) is unique for two reasons: (1) they studied the size distribution of the
31 particles in the range 0.01 to 2.5 //m, and (2) examined the number of particles. They report that
October 1999 6-137 DRAFT-DO NOT QUOTE OR CITE
-------
1 the health effects of 5 day means of the number of ultra fine particles were larger than those of
2 the mass of the fine particles. In contrast Pekkanen et al. (1997) also examined a range of PM
3 sizes but PM10 was more consistently associated with PEF. Mean PM10 levels were 18 //g/m3.
4 Delfino et al. (1998) is unique in that the report effects with 1-hr and 8-hr maximum PM10 having
5 larger effects than the 24 hr mean.
6 Several new panel studies of asthmatic effects in relation to ambient PM10 concentration
7 have been published since the 1996 PM AQCD. Some new studies of emergency department
8 visits and hospitalizations for asthma have been published. The panel studies examine mild
9 asthmatics while the hospitalization studies examine those with asthma attacks severe enough to
10 result in hospitalization.
11 As a group, the results for the asthma panels (see Table 6-2 thru 6-5) of the peak flow
12 analysis consistently show small decrements for both PM10 and PM25. The effects using 2 to
13 5 day lags averaged about the same as did the zero to one day lags. The effects on respiratory
14 symptoms in asthmatics (see Tables 6-5 thru 6-12) also tended to be positive. Most studies
15 showed increases in cough, phlegm, difficulty breathing, and bronchodilator use. The only
16 endpoint more strongly related to longer lag times was bronchodilator use which was observed in
17 three studies. The peak flow decrements and respiratory symptoms are indicators for asthma
18 episodes.
19 The group of studies examining emergency department visits and hospitalization for asthma
20 were fewer in number for these studying ambient PM10 as compared to other ambient PM
21 measures (TSP, BS). Lipsett et al. (1997) reported consistent relationships between ER visits for
22 asthma and wintertime ambient PM10 in an area where one of the principal sources of ambient
23 PM10 is residential wood combustion. Medina et al. (1997) report a relationship between asthma
24 visits and daily concentrations of black smoke. Dab et al. (1996) report that asthma hospital
25 admissions were significantly correlated with NO2 levels, but not ambient PM13. Wordley et al.
26 (1997) report significant associations with mean ambient PM10 values for the past three days and
27 asthma admissions. Sunyer et al. (1997) report, for BS and emergency admissions for asthma,
28 a consistently positive but overall nonsignificant association in all cities they studied for both
29 children and adults.
30
October 1999 6-138 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.4.2.2 Other Non-Asthma Studies
2 Few new studies examine non-asthma outcomes. One review study, by Hoek et al. (1998),
3 yields results of peak flow analysis that consistently show small decrements for increases in
4 ambient PM10. The results of limited chronic studies are not consistent. Specific endpoints
5 examined are bronchitis, COPD, and pneumonia in hospital admission studies. The results of
6 these new PM studies are generally consistent with, and supportive of, those examined in the last
7 AQCD (U.S. Environmental Protection Agency, 1996).
8
9 6.4.2.3 Cardiovascular Effects of Ambient PM Exposure
10 About 75% of all U.S. deaths occur in persons at least 65 years old, and of these nearly
11 40% are for cardiac causes (nearly 45% if deaths from cerebrovascular causes are counted).
12 Thus, if ambient PM exposure indeed produces increased total mortality in the elderly, it would
13 seem possible that cardiovascular deaths may be involved.
14 Seaton et al. (1995) hypothesized that inhalation of very fine ambient particles might
15 provoke alveolar inflammation, with release of chemical mediators. In some susceptible
16 individuals, such mediators might induce both aggravation of underlying lung disease and
17 increase in blood coagulability. These mechanisms could serve to explain the observed
18 association of ambient PM exposure with increased cardiovascular deaths in the elderly.
19 Interestingly, in the European MONICA study, Peters et al. (1996) observed a short-term increase
20 in blood viscosity in men and women, in association with an episode of high TSP and SO2 levels
21 in Augsburg, Germany. This observation lends some weight to the hypothesis of Seaton et al.,
22 especially since blood viscosity has been independently associated with risk of a first myocardial
23 infarction and with incidence of coronary heart disease. At the same time, size-specific PM
24 measurements were not available in Peters et al., and effects of TSP and SO2 could not be
25 statistically separated. Also, high ambient CO levels were associated with increased blood
26 viscosity in women. Also, it is not known whether an increase in mortality or other adverse
27 health effect was temporally associated with the observed increase in blood viscosity in
28 Augsburg.
29
30
31
October 1999 6-139 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.4.2.4 Issues in the Interpretation Of Acute Cardiovascular Effects Studies
2 Three recent analyses of daily PM10 and CO in U.S. cities suggest that elevated
3 concentrations of both PM10 and CO may enhance risk of cardiovascular morbidity leading to
4 acute hospitalizations (Morris andNaumova, 1998; Schwartz, 1997, 1999a). Schwartz (1999a)
5 argued that independent effects of both pollutants are biologically plausible. For CO, the
6 argument is based on the well-established effects of CO on oxygen transport by hemoglobin,
7 albeit at relatively high concentrations. In the case of PM10, Schwartz' plausibility argument
8 draws upon the emerging literature which has demonstrated: effects of ambient PM on
9 pulmonary inflammation in laboratory animals and human volunteers (Gilmour et al., 1996;
10 Salvi et al., 1997); toxicity of transition metals carried by combustion-generated particles (Costa
11 and Dreher, 1999); effects on cardiac dysfunction in animals with pre-existing cardiopulmonary
12 disease (Godleski et al., 1996; Watkinson et al., 1998); and new epidemiologic evidence of
13 associations between ambient PM and physiologic changes in cardiac function (Pope et al.,
14 1999a,b; Liao et al., 1999; Peters et al., 1998; Gold et al., 1998) and plasma viscosity (Peters
15 et al., 1997d) in humans. While a great deal more research is needed to confirm the hypothesized
16 linkages among these new findings, these arguments provide an initial foundation for inferring
17 that ambient levels of PM may play a causal role in cardiovascular illness.
18 Another mechanistic hypothesis, relating to enhanced blood viscosity, is suggested in a
19 recent analysis of plasma viscosity data collected over time in a population of 3256 German
20 adults in the MONICA study (Peters et al., 1997d). Each subject provided one blood sample
21 over the period from October 1984 to June 1985. An episode of unusually high air pollution
22 concentrations occurred over a 13 day period while these measurements were being collect. The
23 authors reported that, among the 324 persons who provided blood during the episode, there was a
24 statistically significant elevation in plasma viscosity as compared with the 2932 persons studied
25 at other times. The odds ratio for plasma viscosity exceeding the 95th percentile was
26 3.6 (CI 1.6-8.1) among men and 2.3 (CI 1.0-5.3) among women. Analysis of the distribution of
27 blood viscosity data suggested that these findings were driven by changes in the upper tail of the
28 distribution rather than by a general shift in mean viscosity. This is consistent with the presence
29 of a susceptible sub-population of individuals.
30 Because they lack data on individual subject characteristics, ecologic time series studies
31 provide only limited information on susceptibility factors based on stratified analyses. The
October 1999 6-140 DRAFT-DO NOT QUOTE OR CITE
-------
1 relative impact of PM on cardiovascular (and respiratory) admissions reported in ecologic time
2 series studies are generally somewhat higher than those reported for total admissions. This
3 provides some limited support for the hypothesis that the acute effects of PM operate via
4 cardiopulmonary pathways or that persons with pre-existing cardiopulmonary disease have
5 greater susceptibility to PM, or both. Although there is some data from the ecologic time series
6 studies showing larger relative impacts of PM on cardiovascular admissions in adults 65 and over
7 as compared with younger populations, the differences are neither striking nor consistent.
8 Individual-level studies of cardiophysiologic function suggest that elderly persons with
9 pre-existing cardiopulmonary disease are susceptible to subtle changes in HRV in association
10 with PM exposures. However, because younger and healthier populations have not yet been
11 assessed, it is not possible to delineate clearly the extent to which the elderly have increased
12 susceptibility, although this does represent a reasonable working hypothesis.
13 It is unlikely that all acute cardiovascular effects can be attributed to weather rather than air
14 pollution. The ecologic time series studies published since 1996 have controlled adequately for
15 weather influences. Some recent studies (Morris and Naumova, 1998) have extensively explored
16 the bivariate effects of weather and CO, and deemed it unlikely that residual confounding by
17 weather is responsible for the PM associations observed. Also, Pope et al. (1999a) evaluated the
18 role of barometric pressure and other meteorological factors in the individual-level studies of
19 cardiac function, but many of these factors have not yet been studied in conjunction with air
20 pollution. Thus, the possibility of confounding by weather, although unlikely, cannot be entirely
21 discounted at present.
22 Co-pollutants have been analyzed rather extensively in many of the recent time-series
23 studies of hospital admissions and PM. In some studies, PM clearly carries an independent
24 association after controlling for gaseous co-pollutants. In others, the PM "effects" are markedly
25 reduced once co-pollutants are added to the model. Among the gaseous criteria pollutants,
26 carbon monoxide has emerged as the one most consistently associated with cardiovascular
27 hospitalizations. The CO effects are generally robust in multi-pollutant models; however, in
28 spite of this, the EPA CO AQCD (U.S. Environmental Protection Agency, 1999) did not find
29 compelling toxicological evidence to support a finding of biological plausibility for the reported
30 epidemiologic observations of CO effects at the current very low ambient CO concentrations
31 typically seen in the United States.
October 1999 6-141 DRAFT-DO NOT QUOTE OR CITE
-------
1 The temporal patterns of cardiovascular response appear more consistent than for many
2 respiratory responses. The evidence from recent time series studies of CVD admissions suggests
3 rather strongly that PM effects, if any, are maximal at lag 0, with some carryover to lag 1. There
4 is little evidence for important effects beyond lag 1.
5 The characterization of PM attributes associated with acute CVD is incomplete.
6 Insufficient data exist from the time series CVD admissions literature or from the emerging
7 individual-level studies to provide guidance as to which PM attributes, defined either on the basis
8 of size or composition, determine potency. The epidemiologic studies published to date have
9 been constrained by the limited availability of multiple PM metrics. Where multiple metrics
10 exist, they often are of differential quality due to differences in numbers of monitoring sites and
11 in monitoring frequency. Until more extensive and consistent data become available for
12 epidemiologic research, the question of PM size and composition, as they relate to acute CVD
13 impacts, will remain epidemiologically unanswerable.
14
15 6.4.3 Consistency of Health Effects for Short-Term and Long-Term
16 Exposure
17 Morbidity health effect outcome measures are demonstrated for ambient PM exposure in
18 asthmatic panel studies for respiratory symptoms and pulmonary function decrements. New
19 studies that look at primary care settings of doctor visits show effects related to PM levels.
20 Effects are also reported for increased hospitalization rates for ambient PM exposure for COPD,
21 asthma, and cardiovascular disease. Hospitalization can be a more severe outcome for an
22 asthmatic than the increase in symptoms or medication use observed in panel studies. These
23 effects may be related to the disease status of the individual in the studies group. The effects are
24 consistent between studies. They are consistent at different locations, and times. Lipfert and
25 Wyzga (1998) and Wyzga and Lipfert (1998) note a lack of coherence among endpoints in
26 long-term studies, in part, because of the lack of long-term studies of hospitalization rates. The
27 studied health effects of ambient PM (i.e., lung function decrements, symptoms, hospitalization
28 utilization, death) may not occur on the same mechanistic pathway and thus there is no reason to
29 expect a quantitative consistency across outcomes.
30 It is not necessarily the case that there is uniformity in effects seen, such as deaths due to
31 respiratory causes with different time scales for ambient PM exposures. For example, PM
October 1999 6-142 DRAFT-DO NOT QUOTE OR CITE
-------
1 exposures on a short time scale are reported as associated with excess cardiopulmonary deaths
2 and respiratory hospital admissions. At this time, however, it is not clear that such deaths or
3 hospital admissions in association with high ambient PM concentrations represent a long-term
4 difference in the mortality rate, or a displacement of events that would have occurred a few days
5 later without the high PM levels. The prospective cohort mortality studies do not yet allow
6 assessment as to whether the occurrence and frequency of deaths from short-term exposures are
7 consistent with a larger rate of occurrence of deaths that is associated with longer-term
8 exposures, on a scale of months or years.
9 There is as yet little basis (conceptual, experimental, or mathematical) that allows direct
10 quantitative linkage of endpoints between mortality attributable to short ambient PM exposures
11 ("acute") and mortality attributable to longer ambient PM exposures ("chronic" or
12 "sub-chronic"). Exposure indices are more easily compared at different time scales when
13 endpoints are identified (Evans et al., 1984). One hypothesis for discussion is that an individual
14 with high susceptibility to ambient PM at a given moment (e.g., elderly, with an acute respiratory
15 infection as well as COPD or other serious pre-existing conditions) may succumb to a moderately
16 elevated ambient PM exposure or to some coincident cause, when that individual would have
17 survived the same PM exposure if it had occurred during a time of lesser susceptibility to
18 ambient PM.
19 The 1996 PM AQCD reported that excess daily mortality associated with a large PM
20 increment (50 //g/m3 for PM10; or 25 //g/m3 for PM2 5) was only about 1.05, whereas the excess
21 risk associated with 20 //g/m3 PM25 between Portage and Steubenville was 1.26 in the Six Cities
22 study, and a comparable PM25 increment yielded a mortality RR of 1.17 in (Pope et al., 1995) for
23 50 U.S. cities. This is consistent with (but does not prove) the hypothesis that there is a
24 substantial baseline risk from long-term exposure to PM that is much larger than the
25 accumulation of mortality from elevated PM episodes.
26 The recent analyses of the long-term AHSMOG study continue to show serious adverse
27 health effects associated with ambient PM10 exposure for which a substantially greater level of
28 individualized ambient PM10 information is available, but also demonstrates some differences
29 with the earlier Six Cities and ACS studies (Dockery et al., 1993; Pope et al., 1995). Statistically
30 significant increases in lung cancer incidence (Beeson et al., 1998), and statistically significant
31 increases in lung cancer deaths and in deaths associated with any contributing respiratory causes
October 1999 6-143 DRAFT-DO NOT QUOTE OR CITE
-------
1 (CRC) were found in AHSMOG males, but not females. The results were generally very robust
2 to different confounder specifications, population subsets, and inclusion of co-pollutants, and
3 were larger for and more significant PM exceedance indices (number of days per year with PM10
4 greater than a cut point, typically 100 //g/m3) than with the mean PM10 concentration. However,
5 PM10 was estimated from TSP rather than measured in the earlier part of the study.
6 Using the same mean PM10 increment of 50 //g/m3, total mortality attributable to long-term
7 ambient PM10 RR was similar to that of the ACS study for PM2 5 for male nonsmokers (1.24) and
8 smaller than that for the Six Cities study (1.57), albeit only significant for the ACS study
9 (Table 6-31). The AHSMOG RR for females (Table 6-31) is smaller and non-significant (0.88),
10 whereas the ACS RR for female non-smokers is significant and only somewhat smaller than the
11 male RR (1.22 in the 50-city PM2 5 study, 1.15 in the 151 -city SO4 study) and 1.28 in the
12 Six Cities.
13 The AHSMOG findings for cardiopulmonary mortality attributable to long-term ambient
14 PM10 are positive for males, but not statistically significant, whereas the ACS findings are
15 significant for female nonsmokers in both studies and in male nonsmokers for the 151-city study
16 (Table 6-32). However, the male RR in AHSMOG (1.22 for cardiopulmonary deaths, 1.54 for
17 CRC deaths) is similar to that of ACS male non-smokers (1.24 for the 50-city study, 1.21 for the
18 151-city study) and smaller than that for all Six Cities subjects (1.74, includes smokers and
19 non-smokers). The ACS female non-smokers have RR of 1.58 and 1.32 respectively, both
20 significant, compared to 0.84 in AHSMOG.
21 Lung cancer mortality attributable to long-term ambient PM10 is not significant for females
22 in any of the female studies (Table 6.3.3.3.3), nor for male nonsmokers in ACS, but is significant
23 for male nonmokers in AHSMOG and male smokers in ACS 151-city. Lung cancer mortality
24 attributable to long-term ambient PM10 is not significant for both genders in the ACS and
25 Six Cities studies, but would be more significant if combined in AHSMOG.
26 The AHSMOG studies also found indications of adverse effects associated with other air
27 pollutants, particularly high concentrations of ozone. Including co-pollutants in PM10 models
28 with exceedance days as the indicator tended to slightly attenuate the PM10 effect, but usually left
29 it significant. This may suggest that long-term PM10 exposures may be better modeled by a
30 non-linear concentration-response function, where data such as in AHSMOG allow this analysis.
October 1999 6-144 DRAFT-DO NOT QUOTE OR CITE
-------
1 Some SO4 analyses were carried out by Abbey et al. (1999). No statistically significant
2 results were found, unlike in the ACS and Six Cities studies. However, SO4 levels tend to be
3 lower in California than in many Eastern cities, and the SO4/PM2 5 ratio is much different.
4 Several recent analyses have started to bridge the gap between the short-term and long-term
5 mortality studies. A relatively transparent approach has been applied by Schwartz (1999b) to the
6 Boston daily mortality data, using the Six Cities data base. This approach, discussed in
7 Section 6.3.2, is to filter out long time trends from the data using a LOESS smoother with a
8 120 window, then to use weighted moving averages of mortality, PM2 5, and meteorological
9 covariables over longer averaging times. This extends the daily time series studies, predicting
10 daily counts from predictors in the preceding few days, to averages of mortality compared to
11 averages of pollution and weather variables over longer periods of time. The windows evaluated
12 by Schwartz (1999b) were 0, 15, 30, 45, and 60 days (weighted moving averages over 1,31,61,
13 91, 121 days resp.). RR for some endpoints, such as COPD mortality, showed relatively little
14 change from 0 and 15 d windows, suggesting that these may be predominantly determined by
15 short-term events. RR for pneumonia, ischemic heart disease, and all cause mortality showed
16 substantial and systematic increases from 0 d to 60 d windows, suggesting that there are mortality
17 effects attributable to prolonged PM25 and weather exposures that greatly exceed a few days.
18 The excess mortality over a 60-d window can be nearly twice as large as those over a 0 d
19 window, thereby providing a substantial degree of plausibility for even larger differences for
20 multi-year mortality studies which do not require seasonal and secular detrending.
21 A different approach was used by Zeger et al. (1999), who decomposed total mortality and
22 TSP time series into Fourier components with distinct time scales, roughly year, season, month,
23 week, day. The decomposition allows each daily term to be represented as the sum of its
24 components. By discarding long-term and very short-term (i.e. "Harvesting") frequencies, a
25 "harvesting-resistant" estimator may be calculated. The analyses in (Zeger et al., 1999) suggest
26 that only a small part of the variation in mortality time series may be attributed to short-term
27 events, such as daily fluctuations in TSP and weather. A large part of the total variation appears
28 to be associated with time scales of 5 to 100 days. Lag times of 0 to 3 days had little effect.
29 While Zeger et al. (1999) establish the existence of a mid-frequency signal (scale of
30 months), they do not establish the absence of a harvesting effect, since the "harvesting-resistant"
31 estimator only allows detection of the longer-term signals that are not simply harmonics of
October 1999 6-145 DRAFT-DO NOT QUOTE OR CITE
-------
1 high-frequency signals. It is entirely feasible that short-term harvesting does occur, along with
2 excess mortality attributable to accumulated ambient PM exposures on a scale of months or
3 longer, as suggested by the prospective cohort studies.
4 The daily time-series studies in fact contain a great deal of information about long-term
5 effects. The results of detrending the daily time-series produces a series with substantial
6 variations in relative risk, which are sometimes shown graphically, but never used to evaluate
7 adverse health effects associated with long-term exposure. Quantitative assessment of the large
8 seasonal and secular variations, often larger than the short-term RR, would be of interest.
9 As yet there is little methodology for quantifying the relationship between short-term and
10 long-term PM epidemiology studies. McMichael et al. (1998) have pointed out some difficulties
11 in assuming that long-term exposure effects can be estimated accurately by summing up day-to-
12 day increments from short-term exposure. The long-term issue requires further study, because
13 extensive life-shortening is an extremely serious adverse health effect.
14 Effects from extremely short-term exposures to ambient PM have not yet received much
15 attention. A number of hypotheses have been raised (Michaels et al., 1998, 1999), but data are
16 extremely scarce because of the relative absence of short-term PM monitoring. An important
17 recent study on asthmatics (Delfino et al., 1998) suggests substantially stronger PM10 effects in
18 asthmatic children in response to 8-hr events than to 24-hr concentrations.
19
20 6.4.4 Susceptible Populations
21 6.4.4.1 Summary Of Previous Criteria Document
22 The 1996 PM AQCD identified several potentially susceptible sub-populations: (1) the
23 elderly (age 65+), particularly those with pre-existing respiratory conditions such as COPD, or
24 previous cardiovascular disease; (2) asthmatics, possibly including both adults and children.
25 Recent research summarized in sections 6.2 and 6.3 often report higher relative risks or odds
26 ratios for adverse effects in these groups associated with ambient PM exposure. More recent
27 studies, summarized in Sections 6.2.4 and discussed in Section 6.4.2.3, find a variety of
28 cardiovascular responses in the elderly associated with ambient PM exposure. Therefore, these
29 findings will not be further discussed, as they tend to confirm the findings in the 1996 AQCD
30 (U.S. Environmental Protection Agency, 1996).
October 1999 6-146 DRAFT-DO NOT QUOTE OR CITE
-------
1 The most important addition to the last document is in the area of children's health. Many
2 recent studies have looked at larger and more general populations of children, in a variety of
3 locations. The findings in (U.S. Environmental Protection Agency, 1996) were somewhat mixed
4 about effects in children, particularly in healthy children. A number of recent studies have shown
5 serious endpoints, including physician visits, hospital admissions, and even mortality associated
6 with various ambient PM indices. The evidence on children's health effects from previous
7 sections is summarized here.
8
9 6.4.4.2 Children as a Susceptible Subpopulation
10 New publications provide a much stronger basis for identifying children as a susceptible
11 sub-population. General overviews of child health air pollution studies include (Raizenne et al.,
12 1998; Romieu, 1998). A quantitative summary of PM effects in many recent studies is presented
13 in (Anderson et al., 1998; esp. Table 7), complementing materials presented in Section 6.2 of this
14 document. Recent evidence suggests a variety of effects: (1) reduced pulmonary function or
15 increased respiratory symptoms in asthmatics and non-asthmatics; (2) an increased incidence of
16 doctor's visits, emergency department visits, and hospital admissions for respiratory symptoms,
17 including asthma, upper and lower respiratory infection (URI, LRI) associated with short-term
18 PM exposures; (3) Increased infant and child mortality associated with short-term PM exposures;
19 (4) Reduced pulmonary function or increased respiratory symptoms associated with longer-term
20 PM exposures; (5) Increased infant mortality, intrauterine growth reduction, or preterm delivery
21 associated with long-term PM exposures. Unfortunately, the wide variety of endpoints and PM
22 indices complicates the estimation of composite effects.
23 Most studies have carried out only single-pollutant analyses. The most commonly used
24 indices are PM10, TSP, and BS, although significant effects are sometimes reported for PM25,
25 SO4, H+. Column 4, labeled "PM Effects" in each of the tables in this section briefly describes
26 the statistical significance of the PM index or indices used in the analyses, with effect sizes
27 reported earlier in Sections 6.2 and 6.3. Column 5, labeled "Pollutants", lists all of the pollutants
28 that were used or were available for use in the analyses. Sensitivity of the PM effect to
29 co-pollutant models is also shown in Column 4, where reported.
30
October 1999 6-147 DRAFT-DO NOT QUOTE OR CITE
-------
1 Reduced Pulmonary Function or Increased Respiratory Symptoms in Asthmatics and
2 Non-Asthmatics
3 The studies are discussed in detail in Section 6.2.2. The findings for children are
4 summarized in Table 6-39. The tables are generally organized so as to characterize populations
5 in the U.S. and Canada first, then Mexico and other western hemisphere countries, Europe, and
6 Asia, roughly in order of relevance to U.S. population demographics. There is a somewhat
7 patchy pattern of symptom effects in asthmatics, particularly in U.S. studies in California.
8 Delfino's findings of significant effects for 1-h and 8-h PM10 exposures is noteworthy. Many
9 asthmatics self-medicate with bronchodilators, which may also be a useful indicator of
10 respiratory distress in these subjects. Cough, phlegm, and LRI are also sometimes found.
11 A number of investigators have found statistically significant peak expiratory flow reduction
12 (PEFR) associated with PM10 and other indices, and sometimes significant reduction in FEVj
13 andFVC.
14 There is an overall indication that respiratory symptoms in children are exacerbated by
15 exposure to airborne particles, and may be more serious in asthmatics or other susceptible
16 groups. There appear to be large numbers of children who may be susceptible to more-or-less
17 serious adverse health effects from brief exposures to airborne particles leading to exacerbation
18 of asthmatic responses. The investigations summarized here generally focus on respiratory
19 endpoints in school-age children.
20
21 Increased Incidence of Doctor's Visits, Emergency Department Visits, Hospital Admissions
22 Table 6-40 summarizes studies described in more detail in Sections 6.2.3 and 6.2.4.
23 Several of the studies using PM10 as an index are statistically significant, suggesting that serious
24 symptomatic responses to PM requiring explicit medical intervention may occur in a number of
25 locations. However, some of these associations became statistically non-significant when
26 gaseous co-pollutants were included in the model, including O3, SO2, NO2, CO. This suggests
27 that specific local conditions, possibly related to the co-pollutant mixture, may play a role in the
28 effects of PM on children. Norris et al. (1999) is the only U.S. study using an index of fine or
29 ultra-fine particles.
30
October 1999 6-148 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-39. RECENT PM STUDIES OF PULMONARY FUNCTION TESTS
OR ACUTE RESPIRATORY SYMPTOMS IN SCHOOL-AGE CHILDREN,
GENERALLY USING PANEL STUDIES
Study
Ostroetal. (1995)
Los Angeles, CA
Delfino et al.
( 1998) Alpine, CA
Delfino et al.
( 1997) Alpine C A
Delfino et al.
(1996), San Diego,
CA
Linnetal. (1996)
southern CA
Thurston et al.
(1997) NY summer
camps
Hoeketal. (1998)
re-analyses of other
studies in the U.S.
and the
Netherlands
Romieu et al.
(1997)
Mexico City
Romieu et al.
(1996)
Mexico City
Gold etal. (1999)
Mexico City
Peters etal. (1996)
Erfurt and Weimar,
Germany; Sokolov,
Cz.
Tittanen et al.
(1999)Kuopio
Finland
Endpoint
Asthma symptoms
for at least six
weeks
Bothersome
asthma symptoms
Symptom score,
bronchodilator use
Symptom scores,
broncho-dilator use
Pulmonary
function
lung function,
symptoms, dilator
use
PEF, large changes
related to
symptoms
PEF, respiratory
symptoms
PEF, respiratory
symptoms
PEF, respiratory
symptoms
PEF, daily
symptoms,
medication
PEF, cough
Ages
(years) PM Effects
7-12 Shortness of
breath risk, 9%
per 10 //g/m3
PM10
9-17 Symptoms signif.
1-h, 8-hPM10,
24-h less signif.
PM10 signif.
dilator use
Signif. O3
personal monitor,
N.S. SAM O3,
PM25
Morning FVC
signif. PM5?, NO2
PM10N.S., SO4,
O3 rel. symptoms,
dilator use
Utah Signif. PEFR,
Valley Cough
Bennekom PEFR N.S. .
Uniontown PEFR N.S.
State PEFR N.S.
College
5-13 N.S. for PM.
Strongest effect
w. O3
5-7 PM10 signif.
PEFR, LRI
8-11 PM2 5, O3 signif.
PEFR, phlegm
7-15 0.43% decrease
with 52 ptg/m3
PM,o
8-13 N.S. for PM
Pollutants Remarks (N)
PM10, TSP, African-American
SO4,NO3,O3, (N = 83)
SO2, NO2
PM10, O3 Panel of
(others low) asthmatics
(N = 25) higher
elevation
PM10, 03 Asthmatics
(N=13)
PM2 5, O3 Asthmatics
(N=12)
PM5?, NO2 School children
(N = 269)
PM10, SO4, H+, PM10, TSP, SO4,
O3, H+, SO2
PM10 N = 39
N-=67
N = 83
N=108
PM10, O3 Mild asthmatics
(N = 65)
PM10, O3 Mild asthma
(N = 71)
PM10, PM2 5, O3 School children
(N = 40)
PM10, TSP, SO2 Mild and
moderate
asthmatics
(N=155)
PM0.1,PM1.0 Chronic resp.
symptoms
(N = 49)
October 1999
6-149
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-39 (cont'd). RECENT PM STUDIES OF PULMONARY FUNCTION TESTS
OR ACUTE RESPIRATORY SYMPTOMS IN SCHOOL-AGE CHILDREN,
GENERALLY USING PANEL STUDIES
Study
Pekkanen et al.
(1997)Kuopio,
Finland
Timonen and
Pekkonen(1997)
Kuopio, Finland
Segalaetal. (1998)
Paris, France
Gielenetal. (1997)
Amsterdam, the
Netherlands
PEACE studies,
the Netherlands
Roemer et al.
(1993) the
Netherlands
Hoek and
Brunekreef(1994)
the Netherlands
Boezen et al.
(1999) the
Netherlands
Agocsetal. (1997)
Budapest HU
Scarlett et al.
(1996)
Studnicka et al.
(1995)
Endpoint
PEF
PEF, respiratory
symptoms
PEF, respiratory
symptoms,
bronchodilator use
PEF, respiratory
symptoms
changes in
respiratory
symptoms
PEF, respiratory
symptoms,
medication
PEF, respiratory
symptoms
PEF, respiratory
symptoms
PEFR
Pulmonary
function
Pulmonary
function
Ages
(years) PM Effects
7-12 PEFR signif. for
difft. PM
measures, w. difft.
lags
7-12 Morning PEF and
PM10 signif, not
evening; N.S.
symptoms
7-15 PM13N.S. in
mild asthma,
signif. inhaler in
moderates.
PM13< other
7-13 PEF, symptoms,
dilator use > for
BSthanPM10
Assessed by
group rates;
N.S.PM
6-12 PM10 signif.
broncho -dilator,
marg. Signif.
PEFR
Weak assoc.
PEFR, N.S.
symptoms
7-11 PM10N.S.
TSP N.S.
PM10 signif. FEV,
FVC
Pulmonary
function related to
H+, notPM10
Pollutants
PM0.03,
PMO.l,
PM0.32, PM1,
PM10
PM10
PM13,SO2,
NO2
PM10, BS, O3
PM10, BS, SO2,
NO2
PM10, BS, SO2
PM10, S04,
NO3, H+, SO2,
NO2
PM10
TSP, SO2
PM10,03,N02
PM10, S04, H+,
NH3, O3
Remarks (N)
Asthmatic
(N = 39)
Asthmatic
(N = 74) or dry
cough (N = 95)
Mild (N = 43),
moderate
(N = 43)
asthmatics
Meds. (N = 47) or
hospitalized (N =
14)
Chronic resp.
disease, usu.
Dry cough
Chronic
respiratory
symptoms
(N = 73)
Non-asthmatics
(N=1079)
N = 632 w and
w/o bronchial
hyperresp.
Asthmatic
children
(N = 60)
School children
(N= 154)
Three panels of
children at
summer camp.
Note abbreviations: EF, peak expiratory flow; PEFR, reduction in PEF; N.S., not statistically significant (two-tailed,
P = 0.05).
October 1999
6-150
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-40. RECENT PM STUDIES OF EMERGENCY DEPARTMENT VISITS
(EDV), HOSPITAL ADMISSIONS, OR DOCTOR'S VISITS IN CHILDREN,
ATTRIBUTABLE TO SHORT-TERM PM EXPOSURE
Study
Norris et al.
(1999) Seattle
WA
Delfino et al.
(1997) Montreal
PQ
Rosas et al.
(1998) Mexico
City
Lin etal. (1999)
Sao Paulo, Brazil
Braga et al.
(1999)
Sao Paulo, Brazil
Ostro et al.
(1999) Santiago,
Chile
Garty et al.
(1998) Israel
Atkinson et al.
(1999) London
UK
Medina et al.
(1997) Paris,
France
Hajat et al.
(1999) London
U.K.
Sunyer et al.
(1997) Barcelona
Helsinki
London, Paris
Endpoint
EDV for asthma
EDV, 1992-
1993
emergency
admissions for
asthma
Respiratory
emergency visits
Hospital
admissions
Medical visit for
LRI, URI
EDV for asthma
EDV for
respiratory
complaints
Doctor's house
calls
GP visits for
asthma, LRI
emergency
admissions for
asthma
Ages
(years) PM Effects
0-17 PM10 signif. all
hosp., It-scatter
each
0-1 H+ signif. only
1993
0-15 PM10N.S.
0-12 PM10 signif. w. and
w/o co-pollutants
0-12 PM10 signif, not
w. O3, CO
0-2 LRI 4-12%
3-15 LRI 3-9%
1-18 PM10N.S.
0-14 PM10 signif. total
resp., asthma
0-14 Asthma signif. BS.
0-14 PM10N.S., BS
signif. LRI
0-14 BS positive, N.S.
NO2 and SO2
signif.
Pollutants
PM10, light scatter,
CO, SO2, NO2
PM10,PM25, SO4,
H+,03
PM10, TSP, O3,
SO2, NO2
PM10, 03, S02,
NO2, CO
PM10,S02,N02,
CO
PM10, 03
PM10, O3, SO2,
NO2
PM10, BS, O3, SO2,
NO2, CO
PM13,BS, SO2,
NO2, O3
PM10, BS, 03, S02,
NO2, CO
BS, NO2, SO2
Remarks (N)
PM! index from
light scattering
grass, fungal
spores signif.
LRI, URI,
wheezing w.
co-pollutants
N=1076
N.S. in2-poll.
models w. SO2,
NO2
Similar RR for
PM13, SO2,NO2
October 1999
6-151
DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Increased Infant and Child Mortality Associated with Short-Term PM Exposures
Table 6-41 shows the results of four recent studies, none in the U.S., in which excess
mortality was associated with PM. Significant mortality was reported in three of the four studies,
using PM25 exposure for infants in Mexico City (Loomis et al., 1999), TSP exposure for school-
age children (but not younger children) in Delhi (Cropper et al., 1997), and PM10 exposure for a
composite group of children 0-14 years in Bangkok (Ostro et al., 1999). Pereira et al. (1998) did
not find excess stillbirths associated with PM10 in Sao Paulo. These studies are highly diverse in
terms of age group, location, and environment. As with adult mortality, there are no known
biological mechanisms that specifically account for excess child mortality from short exposures
to PM at levels found in these Latin American and Asian countries. However, the studies
suggest that short-term PM exposure in general may cause deaths of some children in certain
urban environments. The mortality findings are consistent with findings of less serious health
effects from short-term particle exposure, represented by respiratory endpoints from lung
function deficits through respiratory symptoms, medical encounters, and death, that may affect
substantial numbers of children.
TABLE 6-41. NEONATAL, INFANT, AND CHILD MORTALITY ATTRIBUTABLE
TO SHORT-TERM PM EXPOSURE
Study
Loomis et al.
(1999) Mexico
City
Pereira et al.
(1998)
Sao Paulo, Brazil
Cropper et al.
(1997) Delhi,
India
Mortality
Total
Intrauterine
Total,
cardiovascular,
respiratory
Ages
0-11 mo
Od
0-4 yr
5-14 yr
PM Effects
PM2 5 signif. w and
w/o copollutant
PM10N.S.
TSP N.S. for total
mort.
TSP signif. for total
mort
Pollutants Remarks (N)
PM2 5, O3, NO2
PM10, O3, SO2,
NO2, CO
TSP, SO2, NOX Similar RR in
both age groups,
s.e. not given
Ostro et al. Total,
(1998) Bangkok, cardiovascular,
Thailand respiratory
0-5 yr PM10 signif. all PM10, PM2
October 1999
6-152
DRAFT-DO NOT QUOTE OR CITE
-------
1 Pulmonary Function or Respiratory Symptoms Associated with Longer-Term PM Exposures
2 Table 6-42 shows a number of studies related to longer-term PM exposures. All of these
3 studies involve school-age children, and many use PM10 as an index. PM10 is not always
4 significantly associated with adverse health effects, although other indicators sometimes are
5 (SO4, H+). There are no known biological mechanisms by which elevated PM exposure over
6 long periods of time may be associated with increased risk of respiratory symptoms or decreased
7 pulmonary function in children, although such mechanisms may exist. However, the
8 epidemiology findings from different sites are too diverse to allow simple conclusions about
9 underlying causes.
10
11
TABLE 6-42.
RECENT PM
STUDIES OF
PULMONARY FUNCTION
TESTS OR
RESPIRATORY SYMPTOMS IN SCHOOL-AGE CHILDREN ATTRIBUTABLE
TO LONG-TERM
Study
Dockery et al.
(1996) 24 U.S.,
Canad.
Communities
Braun-Fahrlander
(1997)
10 Swiss comm.
von Mutius
(1995) Leipzig,
Ger.
Raizenne et al.
(1996) 24 U.S.,
Canad.
Communities
Lewis et al.
(1998)N.S.W.
Australia
Peters et al.
(1999a,b) 12 So.
CA communities
Endpoint
Various
Various
URI
Pulmonary
function
Night cough,
chest colds,
wheeze
Asthma,
bronchitis, cough,
wheeze, lung
function
Ages (years)
8-12
6-15
9-11
8-12
8-10
Grades 4, 7
(9, 12 years)
PM EXPOSURE
PM Effects
SO4 signif.
bronchitis; PM10
N.S. any endpoint
Cough, bronchitis,
wheeze signif.
pollut.
PM signif. winter
Strong signif. H+,
signif. PM10
PM10 signif. chest
colds, night cough.
N.S. wheeze.
PM10 signif. FVC,
MMFRN.S. FEV!,
symptoms, PEFR,
Pollutants
PM10,PM2.1
SO4, H+, SO
03
PM10, 03,
NO2, SO2
PM beta, SO
NO2
PM10,PM2.1
SO4, H+, SO
03
PM10, S02
Remarks (N)
;
2>
2, (N=1854)
;
2>
(N = 3023)
(N=150
each in
grades 4, 7)
October 1999
6-153
DRAFT-DO NOT QUOTE OR CITE
-------
1 Increased Infant Mortality, Intrauterine Growth Reduction, or Preterm Delivery
1 Finally, a number of recent studies associated with children less than one year old is
3 pointing toward the possibility of adverse consequences to the mother, fetus, and infant, from
4 prolonged PM exposure during and shortly after pregnancy. Some of the studies shown in
5 Table 6-43 were discussed in Section 6.3.3. There appears to be a possible relationship between
6 preterm birth (< 37 weeks gestational age) or low birth weight (< 2,500 g) and PM exposure in
7 several locations. A significant relationship with PM10 and PM2 5 was found in Teplice, Czech
8 Republic (Dejmek et al, 1999), but not with PM10 in Los Angeles (Ritz and Yu, 1999). Bobak
9 and Leon (1999) did not find a relationship of low birth weight to TSP. There was a significant
10 risk of low birth weight and preterm delivery in Beijing (Xu et al., 1995; Wang et al., 1997)
11 associated with TSP, but SO2 was the only co-pollutant available. However, low birth weight is
12 known to be an important risk factor for infant mortality, so that the findings of excess mortality
13 in U.S. and Czech infants (Woodruff et al., 1997; Bobak and Leon, 1992) is consistent with many
14 of the other findings on intrauterine growth reduction (IUGR).
15 Several issues still require resolution. Dejmek et al. (1999) characterize IUGR as
16 low-weight-for-gestational-age, whereas others use a fixed weight for full-term infants (37 to
17 44 weeks) without adjusting for gestational age. Dejmek et al. (1999) also find the average PM
18 during the first month of pregnancy as the index of fetal exposure, whereas Xu et al. (1995),
19 Wang et al. (1997), and Ritz and Yu (1999) use final trimester averages. This is difficult to
20 interpret in terms of underlying biological mechanisms. In spite of these methodological
21 differences, there appears to be an identifiable PM risk to the fetus and infant.
22 The findings of adverse health effects in infants or young children associated with air
23 pollution exposures of one month to one year is of particular importance in interpreting findings
24 of long-term effects in adults. The duration of exposure in infants and young children is much
25 than that used to characterize adult exposure in long-term prospective cohort studies. While the
26 biological mechanisms for PM-related health effects in children are not necessarily the same as in
27 adults, the child studies do suggest less-than-chronic exposures may be harmful to adults as well
28 as to children. This hypothesis merits further investigation.
29
October 1999 6-154 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-43. OTHER NEONATAL AND INFANT EFFECTS ATTRIBUTABLE TO
LONGER TERM PM EXPOSURE
Study
Woodruff et al.
(1997)
Bobak and Leon
(1992) Czech.
Repub.
Bobak and Leon
(1999) Czech.
Rep.
Dejmek et al.
(1999)Teplice,
Czech. Rep.
Ritz and Yu
(1999) Los
Angeles, CA
Wang et al.
(1997) Beijing,
PRC
Xuetal. (1995)
Beijing, PRC
Effects
Total infant
mortality, SIDS,
resp.
Total infant
mortality, respir.
mort.
Low birth wt.
Stillbirth
Intrauterine
growth reduction
Low birth weight
(adj . Gest age)
Low birth weight
Preterm
gestational age
Ages
1-11 mo
0+ d neonatal
post-neonatal
post, respir.
Od
Od
Od
Od.
Od
PM Effects
PM10 signif.
total, SIDS,
respir. NEW
TSP-10N.S.
TSP signif.
TSP signif.
TSPN.S.
First month
PM25 > 37, PM10
> 40 signif.
Last trimester
PM10N.S.
TSP signif.
increases risk of
LEW
TSP signif. lag
5-10 days
Pollutants
PM10
TSP-10, SO2,
NOX
TSP, SO2,
NOX
PM10,PM25,
SO2, NOX,
PAH
PM10, 03,
NO2, CO
TSP, SO2 in
third
trimester
TSP, SO2
Remarks (N)
PM10 avg over
2 mos.
Ecologic study;
TSP indexed as
90th percentile
30-d avg PM per
month of
pregnancy
CO signif, more
signif. with
co-pollutants
SO2 also signif.
Small reduc.
mn.wt.
SO2 also signif.
1 6.4.5 Consistency of Mortality and Morbidity Effects (Coherence)
2 The criterion of coherence was emphasized by Bates (1992) and it was discussed in detail
3 by U.S. Environmental Protection Agency (1996; Section 12.6.4.3). It consists of the assessment
4 of the entire body of epidemiology data, as well as supporting medical and toxicological data, for
5 consistency in a variety of health outcomes by its repeated observation in different populations of
6 individuals, under different circumstances of duration and level of ambient PM concentration,
7 and in different places. The adverse health effects associated with PM are: (1) lung function
8 decrements; (2) respiratory symptoms, or exacerbation of symptoms requiring bronchodilator
9 therapy; (3) hospital admissions for respiratory and cardiovascular causes; (4) emergency medical
10 visits; and (5) death from cardiopulmonary causes. None of the currently available time series
11 studies are based on a temporal sequence of these outcomes in single individuals. Panel studies
October 1999
6-155
DRAFT-DO NOT QUOTE OR CITE
-------
1 of respiratory symptoms look at the repeated occurrence of symptoms in individuals, but not at
2 the progression of, for example, repeated respiratory symptoms into hospital admissions, or
3 repeated hospital admissions into cardiopulmonary mortality. Indeed, the extent to which this
4 progression occurs in individuals is uncertain. While some studies are currently underway that
5 will examine large public health data bases, no preliminary results have been published. It is
6 therefore necessary to look at indirect indicators of the quantitative consistency at a group level.
7 For example, in asthmatic panel studies, mild asthmatics rarely go to the hospital.
8 Asthmatic symptoms may be controlled by medication use, or may be sufficiently mild that
9 changes in pulmonary function or occurrence of symptoms can be detected and self-treated, but
10 not so severe that they lead to a hospital or emergency department visit. Moderate to severe
11 asthmatics usually are not assessed in panel studies, but these are the subjects more likely to go to
12 a hospital or emergency department for asthmatic episodes. Ambient PM concentrations may not
13 be associated with a progression from mild to severe asthmatic symptoms, since mild and severe
14 asthmatics are distinct sub-groups.
15 It is by no means self-evident that the numbers of events on some appropriate baseline of
16 time and reference group will follow a sequence (1) > (2) > (3) > (4). There are many causes
17 other than PM for each of these endpoints. Long-term lung function decreases with age and
18 chronic respiratory illness. It is affected by cigarette smoking behavior and exposure to
19 occupational air pollution. A number of studies have found an association of mortality with
20 reduced ventilatory function (Strachan, 1992; Higgins and Keller, 1970). Strachan (1992) notes
21 that little is known about the constitutional or environmental determinants of ventilatory function
22 decline. Longitudinal studies such as the Harvard Six Cities study could be used to evaluate a
23 hypothetical causal pathway. Ambient PM concentration —>• PM exposure —>• FEVj decrease
24 —>• death in individuals. The decline of FEVj may be either a precursor of ambient-PM-induced
25 health effects, or an independent factor in susceptibility to air pollution effects leading to hospital
26 respiratory admissions or mortality.
27 The relationship of hospital admissions and mortality in independent studies has been
28 studied. Hospital admissions may be affected by health status as well as by environmental
29 factors. For an individual, the relationship of prior hospital admissions to mortality is uncertain.
30 In general non-environmental studies (Seneff et al., 1995), hospitalization in the preceding few
31 months or year is a good predictor of subsequent hospital admissions or death. The major risk
October 1999 6-156 DRAFT-DO NOT QUOTE OR CITE
-------
1 factor for subsequent death was the development and severity of non-respiratory organ system
2 dysfunction. It is possible that medical intervention is provided to the most seriously ill
3 individuals, but if these interventions reduce the likelihood of death associated with elevated
4 ambient PM concentration and exposure, then many of the deaths attributed to ambient PM may
5 occur in a less frequently hospitalized population. Quantitative consistency would be suggested
6 if there were more cause-specific hospital admissions than deaths from respiratory or
7 cardiovascular causes as described in U.S. Environmental Protection Agency (1996).
8
9 An Ecological Assessment
10 In the absence of appropriate multi-endpoint cohort studies, some evaluations looking at
11 group-level outcomes may provide some insight. Figure 6-3 shows three indicators for the
12 Harvard Six Cities, using data from Dockery et al. (1993); Ferris et al. (1979) . The vertical axis
13 is the relative risk of mortality from (Dockery et al., 1993), the x-axis is the mean PM25
14 concentration for each city from the same paper, and the y-axis is the mean deviation from
15 expected FEVj (not necessarily in the same subjects in each city) (Ferris et al., 1979). The close
16 linear relationship of mortality RR to PM25 was noted in U.S. Environmental Protection Agency
17 (1996). The 2-dimensional relationships to PM2 5 are shown in Figure 6-4. The relationship of
18 mean FEV] to PM2 5 shows a general trend of greater FEVj decrements with increasing PM2 5, but
19 the relationship is weaker and appears nonlinear. As did the cohort in Strachan (1992), the
20 cohort in Dockery et al. (1993) should show a relationship between pulmonary function and
21 mortality unrelated to ambient PM, although this has not yet been evaluated. Further evaluation
22 using other ambient PM indices such as PM2 5 is needed.
23
24 6.4.6 Effects of PM Size Distribution and Composition
25 6.4.6.1 Summary of Previous 1996 PMAQCD
26 Most of the information about size-fractionated PM health effects in U.S. Environmental
27 Protection Agency (1996) was based on PM10, with supporting studies using BS, SO4, or BS.
28 The only daily time series study on mortality (Schwartz et al., 1996) evaluated the single-
29 pollutant effects of PM25, PM10_25, and PM10 in the Harvard Six Cities study. While
30 two-pollutant models with both PM2 5 and PM10_2 5 were fitted, the only results reported were that,
October 1999 6-157 DRAFT-DO NOT QUOTE OR CITE
-------
20
PM2.5
Figure 6-3. Relationship of ambient PM2 5 to relative risk of mortality and to deviations of
FEVj from its expected (standard) value in the Harvard Six Cities study
(Ferris et al., 1979; Dockery et al., 1993).
fe 1.2H
"S
J3£
CO
b:
£ 1.1-
o:
1.0
/
StubenvilleO
St. Louis
10
15
20
PM
(a)
25
25
30
LU
LL
CD
Q.
X
LU
I
03
13
Q
-0.10-
-0.15-
-0.20-
-0.25
O Portage
Topeka
St. Louis
10 15
20
PM
(b)
25 30
2.5
Figure 6-4. Relationship of two health endpoints to PM2 5 in the Six Cities study.
(a) strong linear relationship between relative risk of mortality and mean
PM25; (b) nonlinear or weakly linear relationship between mean FEVj
decrement and PM25. Data from (Dockery et al., 1993; Ferris et al., 1979).
October 1999
6-158
DRAFT-DO NOT QUOTE OR CITE
-------
1 "The estimated effects of the 5th to the 95th percentile increase for PM25 (5.8%, CI 4.3% to 7.4%
2 [excess mortality per 38.8 //g/m3]) was unchanged while the estimate for CM [coarse matter,
3 PM10_2 5] (-0.6%, 95% CI -2.2% to 1.0% [excess mortality per 29.1 //g/m3]) was essentially zero."
4 Significant or marginally significant PM2 5 effects on mortality were identified in most of the
5 cities, and very significant effects in the combined analysis. Only one city suggested a significant
6 effect of PM10_25.
7 Long-term prospective cohort studies used single-pollutant PM2 5, PM10_2 5, PM10, SO4 or H+
8 (Dockery et al, 1993) or PM25 and SO4 (Pope et al, 1995). Neither study evaluated effects of
9 gaseous co-pollutants.
10 The studies cited in Sections 6.2 and 6.3 include a number of studies in which PM25 was
11 observed, or estimated using site-specific data. A few studies have even included PMj 0 as an
12 indicator of the effects of ultrafine particles. These studies are tabulated below in extensive
13 detail because of the importance of characterizing the health effects of fine particles. Many of
14 the new PM25 studies include one or more gaseous co-pollutants, and (Fairley, 1999) included
15 both PM25 and PM10_25, the nitrate (NO3) component of PM10, and gaseous co-pollutants. The
16 new studies thus greatly expand the information available compared to U.S. Environmental
17 Protection Agency (1996).
18
19 6.4.6.2 Assessment of Effects of PM2 5 and Gaseous Co-Pollutants
20 Several recent studies in Santa Clara County, Toronto, and Mexico City have been reported
21 in sufficient detail to allow evaluation of the sensitivity of PM25 health effects estimates to
22 co-pollutants and lag times or moving averages used in the model. These analyses generally use
23 Poisson regression models for daily mortality counts or hospital admissions, non-parametric
24 smoothers for temporal detrending, and adjustments for temperature and other meteorological
25 variables, thus are likely to provide adequate control of non-pollution factors that affect health
26 endpoints. The hospital admissions studies in Toronto are discussed first (Burnett et al., 1997b,
27 1999), then the mortality studies in Toronto (Burnett et al., 1998) and in Mexico City
28 (Borja-Abuto et al., 1998; Loomis et al., 1999).
29
October 1999 6-159 DRAFT-DO NOT QUOTE OR CITE
-------
1 Hospital Admissions in Toronto
1 Burnett et al, 1997 Study
3 There is an extensive reporting of results for 1-, 2-, and 4-pollutant models for summer
4 hospital admissions for respiratory causes (Table 6-44) and cardiovascular causes (Table 6-45)
5 separately. Table 6-44 shows the respiratory admissions relative risks (RR) and confidence
6 limits (LCL, UCL) for three PM indices (estimated PM2 5 or EPM2 5, estimated PM10 or EPM10,
7 and estimated coarse fraction PM10_25 or ECF), and for four co-pollutants (O3, NO2, SO2, CO).
8 All of the single-pollutant models from Table 2, 2-pollutant models with one of the PM indices
9 from Tables 4 and 5, and 4-pollutant models in Table 6 in (Burnett et al., 1997) are shown. The
10 sensitivity of the PM and co-pollutant predictors of daily respiratory admissions may be assessed
11 by use of the t-statistics in the last 5 columns. Although t-statistics are less useful for
12 comparisons across studies, they are extremely helpful for within-study assessments shown here.
13 The PM25 t-statistic for respiratory admissions shows slight attenuation when CO is
14 included in the model, moderate attenuation from single-pollutant estimates when O3 and SO2 are
15 included in the model, but major attenuation when NO2 or 3 co-pollutants are included. There
16 appears to be a substantial collinearity involving PM25 (averaged 1+2+3+4 days) and NO2
17 (averaged 0+1+2+3+4 d). There is little collinearity of PM25 with CO (averaged 2+3+4 d).
18 The PM10 t-statistic for respiratory admissions shows slight attenuation when CO and O3
19 are included in the model, moderate attenuation from single-pollutant estimates when SO2 is
20 included in the model, but major attenuation when NO2 or 3 co-pollutants are included. There
21 appears to be a substantial collinearity involving PM10 (averaged 0+1+2+3 d) and NO2 (averaged
22 0+1+2+3+4 d). There is little collinearity of PM10 with CO or O3 in summer.
23 The PM10_2 5 t-statistic for respiratory admissions shows slight attenuation when CO or O3
24 are included in the model, moderate attenuation from single-pollutant estimates when SO2 is
25 included in the model, but major attenuation when NO2 or 3 co-pollutants are included. There
26 appears to be a substantial collinearity involving PM10_2 5 (averaged 0+1+2+3+4 d) and NO2
27 (averaged 0+1+2+3+4 d). There is little collinearity of PM10_25 with CO or O3 in summer.
28 The PM2 5 t-statistic for cardiovascular admissions shows slight attenuation when CO is
29 included in the model, moderate-to-large attenuation from single-pollutant estimates when O3,
30 NO2, SO2 or are included in the model, but major attenuation when 3 co-pollutants are included.
31
October 1999 6-160 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-44. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS (IN BURNETT ET AL., 1997b; TABLES 2, 4, 6)
ENDPOINT: Hospital admissions for respiratory causes.
SITE: Toronto, Canada, summers.
PM Index
EPM2 5 (25) Moving
average lag
days 1+2+3+4
EPM10 (50) Moving
average lag
days 0+1+2+3
E. Coarse fraction
(25) Moving average
lag days 0+1+2+3+4
PM
RR
1.086
1.062
1.030
1.062
1.081
0.998
1.109
1.098
1.021
1.079
1.106
1.014
1.127
1.110
1.048
1.098
1.121
1.037
Relative
LCL
1.034
1.010
0.972
1.003
1.025
0.954
1.045
1.036
0.945
1.005
1.035
0.940
1.052
1.038
0.957
1.016
1.041
0.950
Risk
UCL
1.141
1.118
1.091
1.125
1.141
1.043
1.177
1.164
1.104
1.159
1.181
1.094
1.207
1.187
1.149
1.188
1.208
1.133
PM
3.3
2.3
1.0
2.1
2.8
-0.1
3.4
3.2
0.5
2.1
3.0
0.4
3.4
3.0
1.0
2.4
3.0
0.8
t-statistics for RR
O3 NO2 SO2 CO
4.6
3.1
1.9
0.5
4.7 2.3 1.6
5.0
2.8
1.7
0.5
4.7 1.8 1.5
4.9
3.0
2.0
0.7
4.7 1.7 1.5
Other PM indices: CoH, SCV, H+
Note: PM25, avg. lags 1-4 days; PM10, avg. lags 0-3 d; CP, avg. lags 0-4 d; SO4, avg. lags 1-4 d; H+, avg. lags
0-1 d; CoH, dayt. avg. lags 0-4 d; O3, dayt. avg. lags 1-3 d; NO2, dayt. avg. lags 0-4 d; SO2, max. avg.
lags 0-3 d; CO, max. avg. lags 2-4 days.
1 There appears to be a substantial collinearity involving PM2 5 with NO2, SO2, and possibly O3.
2 There is little collinearity of PM25 with CO (averaged 2+3+4 d).
October 1999
6-161
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-45. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1997b; TABLES 2, 5, 6)
ENDPOINT: Hospital admissions for cardiovascular causes.
SITE: Toronto, Canada, summers.
PM Index
EPM2.5 (25)
Moving average
lag days 1+2+3+4
EPM10 (50)
Moving average
lag days 0+1 +2+3
E. Coarse fraction
(25) Moving
average lag
days 0+1+2+3+4
RR
1.072
1.032
1.039
1.046
1.065
0.984
1.121
1.091
1.036
1.072
1.109
0.986
1.205
1.192
1.139
1.168
1.198
1.121
PM Relative Risk
LCL
0.994
0.953
0.960
0.966
0.983
0.895
1.014
0.986
0.922
0.964
1.001
0.875
1.082
1.073
0.992
1.037
1.072
0.981
UCL
1.156
1.117
1.125
1.133
1.154
1.082
1.1238
1.206
1.163
1.192
1.229
1.112
1.341
1.325
1.308
1.317
1.340
1.282
PM
1.8
0.8
1.0
1.1
1.5
-0.3
2.2
1.7
0.6
1.3
2.0
-0.2
3.4
3.3
1.8
2.6
3.2
1.7
t-statistics for RR
O3 NO2 SO2 CO
3.5
2.7
2.2
0.8
3.7 1.7 1.6
3.6
2.3
1.9
0.7
3.8 1.6 1.6
3.7
1.4
1.4
0.5
3.8 0.6 1.4
Note: Other PM indices CoH, SCV, H+
Note: PM25, avg. lags 1-4 days; PM10, avg. lags 0-3 d; CP, avg. lags 0-4 d; SO4, avg. lags 1-4 d; H+, avg. lags
0-1 d; CoH, dayt. avg. lags 0-4 d; O3, dayt. avg. lags 1-3 d; NO2, dayt. avg. lags 0-4 d; SO2, max. avg.
lags 0-3 d; CO, max. avg. lags 2-4 days.
1 The PM10 t-statistic for cardiovascular admissions shows slight attenuation when CO and
2 O3 are included in the model, moderate attenuation from single-pollutant estimates when SO2 is
3 included in the model, but major attenuation when NO2 or 3 co-pollutants are included. There
October 1999
6-162
DRAFT-DO NOT QUOTE OR CITE
-------
1 appears to be a substantial collinearity involving PM10 (averaged 0+1+2+3 d) and NO2 (averaged
2 0+1+2+3+4 d). There is little collinearity of PM10 with CO or O3 in summer.
3 The PM10_2 5 t-statistic for respiratory admissions shows slight attenuation when CO, O3 or
4 SO2 are included in the model, but major attenuation when NO2 or all 4 co-pollutants are
5 included. There appears to be a substantial collinearity involving PM10_2 5 (averaged
6 0+1+2+3+4 d) and NO2 (averaged 0+1+2+3+4 d). There is little collinearity of PM10_25 with CO
7 or O3 in summer.
8 The averaging times used for PM indices and for the gaseous pollutants are all much longer
9 than is found in many other studies. Results for shorter averaging times would also be of
10 interest, but were not reported. The longer averaging times may have corresponded to summer
11 episodes, or may have been necessary in order to find statistically significant relationships in the
12 relatively short time series data sets (summers, 1992-1994).
13 Inferences that may be drawn from the data set are limited by the use of estimated PM2 5
14 and PM10 data. This introduces an unknown measurement error into these PM indices. Other
15 measured PM indices are also available in the paper (H+, SO4, Coefficient of Haze).
16 The PM25 estimates achieve statistical significance for respiratory admissions, but not for
17 cardiovascular admissions, and even then, are not significant when NO2 is included as a
18 co-pollutant. Conversely, the coarse fraction PM10_25 estimates are highly significant predictors
19 for both respiratory and cardiovascular admissions, except when NO2 is included as a
20 co-pollutant. Even then, ECF is positive and marginally significant (0.05
-------
TABLE 6-46. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1999; TABLES 3, 4, 5)
ENDPOINT: Hospital admissions for respiratory infections.
SITE: Toronto, Canada.
PM Relative Risk
t-statistics for RR
Pollutant
Lags, d
averaged
RR
LCL
UCL
PM
O3
NO2
NO2 + SO2 + O3
+ EPM2 5
NO2 + SO2 + O3
NO2 + SO2 + O3
+ ECF
Table 5: NO2 +
O3 + EPM2 5
1.098
1.113
1.034
1.010
1.166
1.227
3.03
2.16
.44
2.94
3.03
1.018 0.915 1.133
1.121 1.066 1.178
4.46
3.80 3.31
SO2
1.04
1.56
0.33 3.63 3.34 3.06
CO
EPM2 5 (25)
EPM10 (50)
ECF (25)
03
NO2
SO2
CO
NO2 + SO2 + O3
0+1+2 1.108 1.072 1.145 6.09
0+1+2 1.112 1.074 1.152 5.96
0+1+2 1.093 1.046 1.142 4.00
1+2 4.29
0 5.53
0+1+2 5.04
0 4.25
4.04 3.38 3.43
1 circulation, which are not shown here. Tables 6-46 through 6-51 show the admissions relative
2 risks (RR) and confidence limits (LCL, UCL) for three PM indices (estimated PM2 5 or EPM2 5,
3 estimated PM10 or EPM10, and estimated coarse fraction PM10_2 5 or ECF), and for four
4 co-pollutants (O3, NO2, SO2, CO), along with the optimal moving average or lag times for each
5 endpoint and each pollutant. All of the single-pollutant models from Table 3, and multipollutant
6 models from Tables 4 and 5 in (Burnett et al., 1999) are shown. Burnett et al. (1999) drew
7 conclusions based on their Table 5 models. The Table 4 models all use the "optimal" model with
8 gaseous pollutants as their starting point, with each PM index added to this model. The
9 sensitivity of the PM and co-pollutant predictors of daily respiratory admissions may be assessed
October 1999
6-164
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-47. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1999; TABLES 3, 4, 5)
ENDPOINT: Hospital admissions for asthma.
SITE: Toronto, Canada.
PM Relative Risk
Lags,
Pollutant d averaged RR LCL UCL
EPM25(25) 0+1+2 1.064 1.024 1.106
EPM10(50) 0+1+2 1.089 1.037 1.144
ECF (25) 2+3+4 1.111 1.058 1.166
O3 1+2+3
NO2 0
SO2 2+3+4
CO 0
CO + SO2 + O3
CO + SO2 + O3 1.027 0.965 1.093
+ EPM2 5
CO + SO2 + O3 1.031 0.937 1.135
CO + SO2 + O3 1.170 1.041 1.314
+ ECF
Table 5: CO+ 1.179 1.036 1.342
O3 + ECF
t-statistics for RR
PM O3 NO2 SO2 CO
3.22
3.39
4.20
4.63
2.37
1.76
3.92
4.56 1.53 3.72
0.85 4.40 1.13 3.15
0.63 4.20 1.27 3.32
2.63 3.49 0.25 3.84
3.04 3.48 3.86
1 by use of the t-statistics in the last 5 columns. Burnett et al. (1999) presented coded t-statistics
2 for co-pollutant models, so only single-digit t values may be inferred.
3 The PM t-statistics for respiratory infections in Table 6-46 show substantial attenuation
4 when NO2, SO2, and O3 are included in the models. The estimated coarse fraction loses almost
5 all effect when gaseous pollutants are included. However, when SO2 is omitted as not
6 consistently significant, then the Table 5 model with EPM2 5 is less attenuated, and appears to be
7 more significant than when SO2 is included is included in the Table 4 co-pollutant model. Thus,
8 there appears to be a significant PM2 5 effect on respiratory infection admissions that is not
9 wholly attributable to gaseous pollutants.
October 1999
6-165
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-48. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1999; TABLES 3, 4, 5)
ENDPOINT: Hospital admissions for obstructive lung disease.
SITE: Toronto, Canada.
Pollutant
EPM2 5 (25)
EPM,o (50)
ECF (25)
03
NO2
SO2
CO
CO + O3
CO + O3 + EPM2 5
CO + O3 + EPM10
CO + O3 + ECF
Table 5: CO + O3 +
ECF
PM
Lags, d
averaged RR
1+2 1.048
2 1.069
2+3+4 1.128
2+3+4
1
2
0
1.046
1.077
1.172
1.172
Relative Risk
LCL UCL
0.998 1.100
1.013 1.128
1.049 1.213
0.986 1.110
0.978 1.186
0.952 1.443
0.952 1.443
t-statistics for RR
PM O3 NO2 SO2 CO
1.89
2.44
3.26
4.23
1.07
0.05
1.48
3.80 1.50
1.28 3.69 1.22
1.59 3.41 1.39
1.90 2.74 1.52
1.90 2.74 1.52
1 The PM t-statistics for asthma admissions in Table 6-47 show substantial attenuation when
2 CO, SO2, and O3 are included in the models. Only the estimated coarse fraction retains statistical
3 significance when gaseous pollutants are included. However, when SO2 is omitted as not
4 consistently significant, then the Table 5 ECF effect is slightly larger, and significance of the
5 coarse fraction is greater than the Table 4 co-pollutant model. Thus, there appears to be a
6 significant PM10_2 5 effect on asthma admissions that is not wholly attributable to gaseous
7 pollutants.
8 The PM t-statistics for obstructive lung disease in Table 6-48 show substantial attenuation
9 when CO and O3 are included in the models. The estimated PM2 5, PM10, and coarse fractions
10 lose their statistically significant effects on admissions for obstructive lung disease when gaseous
11 pollutants are included, although the coarse fraction is still marginally significant.
October 1999
6-166
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-49. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1999; TABLES 3, 4, 5)
ENDPOINT: Hospital admissions for heart failure.
SITE: Toronto, Canada.
PM Relative Risk
Lags, d
Pollutant averaged RR LCL UCL
EPM25(25) 0+1+2 1.066 1.025 1.108
EPM10(50) 0+1+2 1.097 1.042 1.155
ECF (25) 0+1+2 1.079 1.023 1.314
O3 1+2
NO2 0
SO2 0
CO 0+1
CO + NO2
CO + NO2+ 1.019 0.959 1.082
EPM25
CO + NO2+ 1.050 0.954 1.155
EPM10
CO + NO2+ 1.071 0.956 1.201
ECF
Table 5:
CO + NO2
t-statistics for RR
PM O3 NO2 SO2 CO
3.20
3.51
2.79
1.42
6.33
3.85
5.71
3.44 2.08
0.60 3.33 1.83
1.00 3.18 1.84
1.18 3.11 2.04
3 2
1 The PM t-statistics for heart failure admissions in Table 6-49 show substantial attenuation
2 when NO2 and CO are included in the models. The PM25, PM10, and estimated coarse fractions
3 lose even marginal statistically significant effects on admissions for heart failure when these
4 gaseous pollutants are included.
5 The very significant single PM t-statistics for ischemic heart disease admissions in
6 Table 6-50 show almost complete attenuation when NO2 and SO2 are included in the models.
7 The NO2 and SO2 effects are somewhat attenuated, but remain statistically significant when PM10
8 only 1 of the 3 respiratory endpoints (admissions for asthma) retains a clearly significant effect of
9 PM10_25, after specific co-pollutants are included in the models. None of the cardiovascular or
October 1999
6-167
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-50. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1999; TABLES 3, 4, 5)
ENDPOINT: Hospital admissions for ischemic heart disease.
SITE: Toronto, Canada.
Pollutant
EPM2 5 (25)
EPM10 (50)
ECF (25)
03
NO2
SO2
CO
NO2 + SO2
NO2 + SO2 + EPM2 5
NO2 + SO2 + EPM10
NO2 + SO2 + ECF
Table 5: NO2 + SO2
PM Relative Risk t-statistics for RR
Lags, d
averaged RR LCL UCL PM O3 NO2 SO2 CO
0+1+2 1.080 1.054 1.105 6.08
0+1+2 1.084 1.053 1.115 5.55
2+3+4 1.037 1.013 1.062 3.02
1+2+3 0.99
0 8.40
2+3+4 6.13
0 6.46
6.10 2.07
1.031 0.985 1.080 1.32 5.45 1.20
0.996 0.967 1.025 -0.3 5.56 1.98
0.985 0.871 1.112 -0.2 5.94 2.03
6.10 2.07
1 circulatory endpoints appear to show clearly significant effects for any PM index when specific
2 co-pollutants are included in the model. Marginally significant effects of the coarse fraction for
3 COPD and PM10 for dysrhythmia are also noteworthy in this context.
4 The differences between the previous study (Burnett et al., 1997a) and this study require
5 further elaboration. The early study looked only at summer outcomes, for combined respiratory
6 causes and combined cardiovascular causes, rather than year-round effects for more specific
7 causes. Combining the specific causes, and looking for seasonal effects, might have found more
8 similar results in this study.
9 Differences in PM and co-pollutant moving averages might also account for different
10 findings. The PM25 moving average in (Burnett et al., 1997) was 1+2+3+4 days, whereas the
11 moving averages for the significant effect for PM25 in Table 6-46 was 0+1+2 days, the
October 1999
6-168
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-51. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT ET AL., 1999; TABLES 3, 4, 5)
ENDPOINT: Hospital admissions for dysrhythmias.
SITE: Toronto, Canada.
PM Relative Risk t-statistics for RR
1
2
3
4
5
6
7
8
Lags, d
Pollutant averaged RR LCL UCL PM O3
EPM25(25) 0 1.061 1.019 1.104 2.91
EPM10(50) 0 1.084 1.029 1.142 3.03
ECF (25) 0 1.051 0.998 1.108 1.88
O3 2+3+4 1.71
NO2 0+1+2
SO2 0
CO 0+1
CO + O3 1.58
CO + O3+ 1.048 0.985 1.115 1.49 1.63
EPM25
CO + O3+ 1.089 0.991 1.197 1.77 1.58
EPM10
CO + O3 + ECF 1.072 0.961 1.195 1.24 1.53
Table 5: CO+ 1.048 0.985 1.115 1.49 1.63
O3+EPM25
non-significant effects for PM25 in Tables 6-47 through 6-51 were 0+1+2,
NO2 SO2 CO
1.73
1.43
3.60
3.52
2.50
2.58
3.29
2.50
1+2, and 0 days. The
PM10_25 moving average in (Burnett et al., 1997a) was 0+1+2+3+4 days, whereas the moving
averages for the significant effect for PM2 5 in Table 6-47 was 2+3+4 days,
and the coarse PM fraction are included. Including PM2 5 reduces the NO2
SO2 effect becomes non-significant.
the non-significant
effect slightly, and the
The PM t-statistics for dysrhythmias in Table 6-5 1 show substantial attenuation when CO
and O3 are included in the models. The estimated PM2 5, PM10, and coarse
their effect when gaseous pollutants are included. However, there appears
fractions lose much of
to be a positive
October 1999
6-169
DRAFT-DO NOT QUOTE OR CITE
-------
1 marginally significant PM10 effect on admissions for dysrhythmias that is not wholly attributable
2 to gaseous pollutants.
3 The results of this study provide a rather mixed picture, Only 1 of the 3 respiratory
4 endpoints (admissions for respiratory infections) retains a clearly significant effect of PM25, and
5 effects for PM25 in Tables 6-46 and 6-48 through 6-51 were 0+1+2, 2+3+4, and 0 days. The
6 co-pollutant moving averages and lag times were much more diverse across the health endpoints
7 than the PM moving averages. It is not possible to generalize these findings into a single
8 statement about the potential effect of gaseous pollutants as confounders of PM health effects for
9 various endpoints at various time lags.
10
11 Mortality Studies
12 Fairley (1999) in Santa Clara County, California
13 This paper evaluates the effects of PM25 and its co-pollutants in considerable detail, using
14 data for 1989-1996, updating earlier studies (Fairley, 1990, 1994) discussed in (U.S.
15 Environmental Protection Agency, 1996). Santa Clara County (SCC) is a major metropolitan area
16 centered on the city of San Jose, just south of the San Francisco MSA. While particle
17 concentrations from mobile sources, wood smoke and other stationary sources are relatively high
18 during the winter, SO2 concentrations are so low that they are no longer monitored in SCC.
19 However, PM25, PM10_25, and PM10 were measured on an every-sixth-day schedule at the SCC
20 San Jose 4th Street site from 1990. The analyses did not used imputed PM data, so that the
21 sample sizes are only N = 408 for fine and coarse fractions from a dichot sampler, N = 823 for
22 PM10 from dichot and hi-vol samplers, N = 523 for sulfate and 534 for nitrate (NO3) samples
23 from PM10. The dichotomous sampling at this site also allowed collection of PM10 nitrate and
24 sulfate species data. Daily measurements of CoH were also available for comparison with the
25 1980-1986 study (Fairley, 1990). Poisson regression models were fitted to daily mortality data,
26 as well as to cardiovascular and respiratory mortality data, but extensive assessments for
27 sensitivity to co-pollutant models are only reported for total mortality. Daily measurements were
28 available for gaseous co-pollutants. Because of the sparse PM measurements, only concurrent-
29 day and previous-day effects are considered.
30 The results are shown in Table 6-52. All of the PM2 5 coefficients with gaseous
31 co-pollutants are statistically very significant, and practically unchanged in magnitude, as are the
October 1999 6-170 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-52. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANT MODELS IN (FAIRLEY, 1999; TABLE 4)
ENDPOINT: Mortality
SITE: Santa Clara County, California, 1989-1996.
PM Index
PM25
25 Mg/m3
PM10
(50)
CF (25)
S04(15)
NO3
PM
RR
1.085
1.091
1.108
1.097
1.098
0.999
1.090
1.100
1.117
1.080
1.045
1.328
Relative
LCL
1.034
1.036
1.049
1.035
1.021
0.913
1.031
1.043
1.049
1.035
0.933
1.059
Risk
UCL
1.138
1.148
1.171
1.163
1.180
1.095
1.153
1.160
1.190
1.127
1.171
1.665
t-statistics for RR
PM O3 NO2 SO2 CO CoH NO3 SO4 CF
3.3
3.3 1.1
3.7 -1.2
3.1 -0.8
2.5 -0.2
-0.0 1.9
3.0 0.0
3.5 -0.9
3.4 1.4 -1.0 -0.1
3.6
0.8
2.5
3.4
3.5 1.5
3.3 -0.5
3.0 0.4
3.4 2.0 -0.9 1.6
Note: All pollutants lag 0 except for CoH, CO, or NO 2 (lag 1 where they fit better). Ozone is 8-hr average.
1 PM2 5 coefficients with CoH, SO4, and PM10_2 5, and these co-pollutant effects are greatly
2 attenuated in size and significance when PM2 5 is included. This suggests that PM2 5 has an effect
3 on mortality that is largely independent of all co-pollutants in the model except for the thoracic
4 nitrate fraction (NO3) in PM10. The thoracic nitrate fraction appears slightly more significant as a
5 predictor of total mortality than does PM25. Further elaboration of the possible role of PM2 5 and
October 1999
6-171
DRAFT-DO NOT QUOTE OR CITE
-------
1 nitrate in SCC would be of interest, such as the mortality effects of the non-nitrate thoracic
2 particles (PM10-NO3) and the non-sulfate fine particles (PM2 5-SO4). Cause-specific mortality
3 effects with co-pollutants included in the model would also be of interest.
4
5 Burnett et al (1998) in Toronto, Canada
6 Although this paper emphasizes the role of CO in mortality, a great deal of information
7 about PM effects is also provided. The paper emphasizes short-term exposure lags, either
8 same-day or previous day (0 or 1) or their average. The results are shown in Table 6-53. There
9 is an extensive reporting of results for 1- and 2-pollutant models for total mortality.
10 Table 6-53 shows the relative risks (RR) and confidence limits (LCL, UCL) for five PM indices
11 (estimated PM2 5 or EPM2 5, estimated PM10 or EPM10, sulfates, TSP, and coefficient of haze or
12 CoH). No results are reported for the estimated coarse fraction PM10_2 5, a major omission. The
13 t-statistics are given for models with and without CO as a co-pollutant for the five PM indicators,
14 and for three gaseous pollutants (O3, NO2, SO2).
15 All five PM indicators have a statistically significant relationship to mortality, although the
16 significance is greatly reduced by inclusion of CO as a co-pollutant. None of the gaseous
17 pollutants are statistically significant with inclusion of CO. The statistical significance of CO is
18 only slightly diminished by the inclusion of a PM indicator, with the exception of CoH
19 (an indicator of elemental fine carbon that is often highly correlated with CO). In every case,
20 CO has greater statistical significance than the PM index, although both are significant.
21 The two most significant particle indicators of mortality are CoH and EPM25, which have
22 the same t-statistic of 3.5 with CO as a co-pollutant. A more thorough co-pollutant assessment
23 of PM25 for Toronto, with particular attention to NO2 and O3 effects, seasonal effects, and
24 specific causes of death, would be of interest.
25
26 Borja-Abuto et al. (1998) in Southwest Mexico City, Mexico
27 The authors carried out a detailed assessment of non-accidental mortality for the period
28 1 January 1993-31 July 1995 in a populous section of Mexico City with historically high ozone
29 levels, but lower levels of PM than in heavily industrialized areas. In addition to total mortality,
30 the authors also examined mortality for age > 65 years, mortality from respiratory causes, from
31 cardiovascular causes, and from other causes. The results are shown in Table 6-54, for a PM2 5
October 1999 6-172 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-53. SENSITIVITY OF PM RELATIVE RISK ESTIMATE TO
CO-POLLUTANTS MODELS IN (BURNETT, ET. AL., 1998: TABLE 2)
ENDPOINT: Total mortality from all natural causes.
SITE: Toronto, Canada.
Lag or
PM Index or average
Pollutant (days)
CO 0+1
EPM25 0+1
25 yUg/m3
EPM10 0
50 yUg/m3
SO4= 0
15 yUg/m3
LagOd
TSP 0
100 Mg/m
CoH 0+1
(0.6 per kft)
03 1
NO2 0
SO2 0
PM Relative Risk
RR LCL UCL
1.048 1.033 1.064
1.030 1.014 1.047
1.035 1.018 1.053
1.021 1.004 1.039
1.027 1.012 1.043
1.020 1.005 1.035
1.023 1.008 1.038
1.017 1.002 1.032
1.059 1.043 1.075
1.035 1.016 1.055
t-statistics for RR
CORR
(1.4
PM O3 NO2 SO2 CO ppm)
8.4 1.070
6.0
3.5 6.3 1.056
3.9
2.4 7.1 1.063
3.4
2.8 7.9 1.066
3.1
2.3 6.7 1.056
7.5
3.5 4.3 1.043
0.7
1.4 8.7 1.072
3.2
0.7 8.0 1.067
2.6
1.5 8.0 1.067
Note: Multiple-pollutant models have 2 or more t-values in the same row, for each pollutant.
Note: Average of lags 0 and 1 days used for CO, CoH. EPM 2 5. Lag 0 (same day) used for EPM10, TSP,
NO2, SO2. Lag 1 used for O3.
SO4,
1 lag of 4 days (largest effect), with co-pollutants O3 (average 1+2 d) and NO2 (average
2 1+2+3+4+5 d). We will examine the choices of lag structures later. The authors reported results
3 of single-pollutant models, and all 2- and 3-pollutant models with PM25.
October 1999
6-173
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-54. SENSITIVITY OF PM RELATIVE RISK ESTIMATE FOR
TOTAL MORTALITY TO CO-POLLUTANT MODELS
(BORJA-ABUTO ET AL., 1998, TABLE 5)
PM INDEX: PM25, lag 4 days.
SITE: Southwest Mexico City, Mexico
Endpoint
Total mortality
from all natural
causes
Age > 65
Respiratory
Cardiovascular
Other natural
causes
PM
(25
RR
1.034
1.036
1.034
1.043
1.040
1.041
1.046
1.058
1.064
1.067
1.051
1.043
1.056
1.056
1.043
1.088
1.020
1.021
1.018
1.022
2 5 Relative
Mg/m3, lag
LCL
1.005
1.006
0.997
1.006
1.001
1.000
0.997
1.008
0.972
0.973
0.939
0.930
1.000
0.999
0.930
1.017
0.983
0.984
0.971
0.975
Risk
4d)
UCL
1.064
1.067
1.071
1.080
1.080
1.082
1.096
1.109
1.160
1.166
1.170
1.163
1.113
1.116
1.163
1.162
1.057
1.060
1.065
1.069
t-statistics for RR
PM
2.3
2.4
1.8
2.3
2.0
2.0
1.8
2.3
1.4
1.4
0.9
0.7
2.0
1.9
0.7
2.4
1.0
1.1
0.7
0.9
03
0.6
0.9
1.0
1.3
0.9
1.1
-0.5
-0.6
-0.7
2.0
1.8
2.0
0.5
0.2
0.2
NO2 SO2 CO
1.1
0.0
-0.0
0.7
-0.4
-0.6
0.9
0.4
1.0
0.9
-0.3
-0.2
0.7
0.1
-0.0
Note: PM25 lag 4 days; O3 mean of lags 1 and 2 days; NO2 mean of lags 1 through 5 days.
October 1999
6-174
DRAFT-DO NOT QUOTE OR CITE
-------
1 The t-statistics for PM2 5 were generally stable across co-pollutant models within each
2 endpoint. For total mortality, PM2 5 had a significant effect when O3, and both O3 and NO2 were
3 included. When NO2 was the only co-pollutant, the PM2 5 effect was of about the same
4 magnitude, but only marginally significant.
5 For elderly mortality, PM2 5 had a significant effect when O3, and both O3 and NO2 were
6 included. When NO2 was the only co-pollutant, the PM2 5 effect was of about the same
7 magnitude, but only marginally significant. The relative risks were somewhat higher than for
8 total mortality, but showed a similar pattern.
9 Neither PM2 5 nor the co-pollutants had a significant effect on respiratory mortality.
10 However, the PM25 relative risk was higher than all-cause mortality. The lack of statistical
11 significance may reflect the small number of deaths in this category.
12 For cardiovascular mortality, PM2 5 had a larger and even more significant effect when both
13 O3 and NO2 were included. When O3 or NO2 was the only co-pollutant in the model, the PM25
14 effect was of about the same magnitude, but not significant with NO2 and marginally significant
15 with O3. The O3 effects were stable and significant or marginal for cardiovascular mortality.
16 This was the only endpoint for which O3 effects were significant, with or without PM2 5 as a
17 co-pollutant.
18 The PM2 5 relative risks for other natural causes were lower than for respiratory or
19 cardiovascular causes. None of the pollutants had a significant effect on other-cause mortality.
20 The time lags and moving averages selected are somewhat atypical. The authors present
21 information on risk estimates at various lags. Figure 6-5 shows a pattern of relative risks, with
22 confidence limits, that is not readily interpretable. Positive, statistically significant RR of
23 roughly similar magnitude are shown at lag days 0 and 4, and for the moving average of
24 1+2+3+4+5 days. The other lags do not show significant effects. Other results may have been
25 obtained by other choices.
26 In summary, this study suggests that PM2 5 is associated with moderately elevated risks of
27 cardiovascular mortality, and total mortality, to an extent that cannot be attributed to the gaseous
28 pollutants O3 and NO2. The effects of CO and other co-pollutants remains to be evaluated. The
29 lag structure may have been chosen to maximize the PM2 5 effect and other forms might be
3 0 worthy of investigation.
31
October 1999 6-175 DRAFT-DO NOT QUOTE OR CITE
-------
1.05
LO
CM
Q_
CO
E
LO
CM
CH
O
CO
a: 1,00
UJ
UJ
a:
0,95
v 95% UPPER C.L,
A 95% LOWER C,L,
O RELATIVE RISK
0
5
LAG FOR PM
Figure 6-5. Relative risk of total mortality from PM2 5 in southwest Mexico City as a
function of PM lag or moving average, with 95% confidence limits.
Source: Borja-Abuto etal. (1998).
1 Loomis et al. (1999) Infant Mortality in Southwest Mexico City, Mexico
1 This study parallels that of Borja-Abuto et al. (1998) for mortality in children less than one
3 year of age. The results are shown in Table 6-55, for a PM2 5 moving average of 3+4+5 days
4 (largest effect), with co-pollutants O3 (average 2+3 d) and NO2 (average 3+4+5 d). We will
5 examine the choices of lag structures later. The authors reported results of single-pollutant
6 models, and all 2- and 3-pollutant models with PM2 5.
7 The PM2 5 relative risk is statistically very significant with no co-pollutants, but is
8 somewhat attenuated by including O3 (still significant), NO2, and both NO2 and O3 (marginally
9 significant) as co-pollutants. Neither of the gaseous pollutants is a significant predictor of infant
10 mortality when PM25 is included as a co-pollutant. The results are only moderately robust to the
11 inclusion of gaseous co-pollutants.
October 1999
6-176
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-55. SENSITIVITY OF PM RELATIVE RISK ESTIMATE FOR INFANT
MORTALITY TO CO-POLLUTANT MODELS (LOOMIS ET AL., 1999, TABLE 5)
PM INDEX: PM25, mean of lag days 3+4+5
SITE: Southwest Mexico City, Mexico
PM25
(25 W
RR
1.181
1.163
1.154
1.165
Relative Risk
g/m3, mean of lag days
LCL
1.063
1.034
0.981
0.987
3+4+5)
UCL
1.306
1.302
1.345
1.362
t-statistics for RR
PM O3 NO2 SO2 CO
3.2
2.6 0.8
1.8 0.4
1.9 0.8 0.1
1.7
2.5
Note: PM25, mean of lags 3, 4, 5 days; O3, mean of lags 2 and 3 days; NO2, mean of lags 3, 4, 5 days.
1 The time lags and moving averages selected are somewhat atypical. The authors present
2 information on risk estimates at various lags. Figure 6-6 shows a pattern of relative risks, with
3 confidence limits, that is reasonably interpretable. Positive, statistically significant RR are
4 shown for the 7 moving averages on the right side, ranging from single days (lags 3, 4, 5) up to
5 2+3+4+5 days. The other lags do not show significant effects. The two best choices are
6 averages of 3+4+5 and 2+3+4+5 days. These findings are reasonably consistent with the
7 long-delayed effects in the total mortality study of Borja-Abuto et al. (1998).
9 Summary: Sensitivity of PM25 Effect Estimates to Inclusion of Gaseous Co-Pollutants
10 The findings of this discussion are shown in Table 6-56. It is often the case that inclusion
11 of gaseous co-pollutants attenuates the estimated PM25 effect size and statistical significance, but
12 the attenuation is occasionally modest or negligible, and the effect remains statistically
13 significant. Examples include: Hospital admissions for respiratory infection (Burnett et al.,
14 1999); total mortality (Burnett et al., 1998; Fairley, 1999); total mortality, age > 65 mortality and
15 cardiovascular mortality (Borja-Abuto et al., 1998); and marginally, infant mortality (Loomis
16 etal, 1999).
October 1999
6-177
DRAFT-DO NOT QUOTE OR CITE
-------
1.4
1.3
1.2
CD
- 1 1
CO '• '
1.0
0.9
i i i i i i i i i i i i i i i i r
Q 01 I I I I I I I I I I I I I I I I I
LAG
RR
UCRR
LCRR
Figure 6-6. Relative risk of infant mortality from PM2 5 in southwest Mexico City as a
function of PM lag or moving average, with 95 percent confidence limits.
Source: Loomis etal. (1999).
1 These form a body of evidence that the effects are moderately robust to the inclusion of certain
2 co-pollutants. This evidence applies to 5 or 6 of 16 independent endpoints in the 6 studies.
3 However, the biological meaning of the 6 significant or marginally significant robust PM2 5
4 effects is limited by two factors: (1) differences in the gaseous co-pollutants used as covariates
5 across the studies, and (2) differences in PM2 5 and co-pollutant averages used as predictors of
6 effects. The multipollutant model used in (Burnett et al., 1997) includes O3, NO2, and SO2,
7 whereas only CO is used as a co-pollutant for PM2 5 in Burnett et al. (1998). All four co-
8 pollutants are considered in Burnett et al. (1999), but different models are reported for different
9 endpoints, some with different moving average lags as well. Fairley (1999) evaluates many
October 1999
6-178
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-56. EFFECTS OF INCLUDING ONE OR MORE GASEOUS CO-POLLUTANTS ON PM RELATIVE RISK
o
r+
O
a'
CT
VO
VO
VO
ON
i
VO
a
H
a
o
o
H
0
0
H
W
O
O
H
W
ESTIMATES FOR HOSPITAL ADMISSIONS AND MORTALITY IN TORONTO, SANTA CLARA COUNTY,
AND SOUTHWEST MEXICO CITY
Study
Burnett et al. (1997)
Burnett et al. (1999)
hospital admissions in
Toronto
Burnett et al. (1998)
Fairley ( 1999) Santa
Clara County
Borja-Abuto et al.
(1998) Mexico City
Loomis et al. (1999)
Mexico City
End-point
Respiratory Admission
Cardiovasc.
Admissions
Respiratory
Infection
Asthma
Obstructive Lung Dis.
Heart failure
Ischemic Heart Dis.
Dysrhythmia
Total mortality
Total mortality
Total mortality
Age> 65 mortality
Respiratory mortality
Cardiovasc. Mortality
Other mortality
Infant mortality
PM2
Avg. lag d.
for PM2 5 t
1+2+3+4 3.3
1+2+3+4 1.8
0+1+2 6.1
0+1+2 3.2
1+2 1.9
0+1+2 3.2
0+1+2 6.1
0 2.9
0+1 6.0
0 3.3
4 2.3
4 2.0
4 1.4
4 2.0
4 1.0
3+4+5 3.2
s only
RR
1.086
1.072
1.108
1.064
1.048
1.066
1.080
1.061
1.048
1.085
1.034
1.040
1.064
1.056
1.020
1.181
With 2+ Gaseous Co-Pollutants
With 1 Gaseous Co-Pollutant
min. t RR Co-Poll. t RR
1.0 1.030 NO2 -0.1 0.998
0.8 1.032 O3 -0.3 0.984
4.5 1.085
0.8 1.027
1.3 1.031
0.6 1.019
1.3 1.031
1.5 1.048
3.5 1.030 CO
3.1 1.097 CO 3.4 1.117
1.8 1.034 NO2 2.3 1.043
1.8 1.046 NO2 2.3 1.058
0.9 1.051 NO2 0.7 1.043
0.7 1.043 NO2 2.4 1.088
0.7 1.018 NO2 0.9 1.022
1.8 1.154 NO2 1.9 1.165
Co-Poll.
NO2 O3 SO2
NO2O3 SO2
NO2O3
O3 SO2 CO
O3CO
NO2CO
NO2 SO2
O3CO
CO NO2 O3
NO2 O3
NO2O3
NO2O3
NO2O3
NO2O3
NO2O3
Note: t-statistics > 1.96 are generally taken as statistically significant.
-------
1 co-pollutants, including a model with CO, NO2, and O3. The Mexico City analyses (Borja-Abuto
2 et al., 1998; Loomis et al., 1999) use only NO2 and O3 as co-pollutants, and all include longer
3 lags (4 d and 3+4+5 d) than the Toronto analyses (0+1 d, 0+1+2 d). Thus, the evidence suggests
4 that there may be adverse health effects associated with PM2 5, but does not suggest that a single
5 mechanism or time structure applies to all locations.
6
7 6.4.6.3 Factors and Components Including PM
8 Many of the concerns about attribution of health effects to ambient is that ambient PM is
9 closely associated with co-pollutants and weather, an intrinsic causal association for which
10 standard epidemiologic methods of covariate adjustment may not be adequate. Certain sources
11 produce PM of specific size or composition, as well as gaseous pollutants. Weather affects
12 ambient concentrations of particles and gases, affects the rate of formation of secondary particle
13 components, and may affect human exposure patterns. Several new approaches have been
14 developed that directly confront these issues: (1) construct common factors involving PM,
15 co-pollutants, and weather variables, and use these common factors as nominal predictors of
16 health effects; (2) Use elemental or other compositional components of PM as markers of sources
17 from which the PM was derived. Both approaches are potentially useful.
18 The first approach is illustrated by a study of daily mortality in Toronto, Canada, carried
19 out by Ozkaynak et al. (1996). Factor analysis methods were applied to construct composite
20 indicators as linear combinations of relative humidity, temperature, CoH, TSP, SO2, CO, NO2,
21 and maximum O3. Results are given in Sec. 6.3.3. Relationships between total and
22 cardiovascular mortality were much stronger for composite factors 1, 4, and 5 than the
23 relationships for individual components. Factor 1 was interpreted as an automobile factor,
24 Factor 4 as a TSP factor, Factor 5 as a Temperature factor. Interestingly, Factor 3 (Ozone)
25 predicted excess total mortality, but not cardiovascular mortality. Factor 1 involved mainly CoH
26 and NO2, with lesser weights on CO, TSP and SO2. The method used by Ozkaynak et al. (1996)
27 produced plausible and interpretable combinations of the original variables.
28 An alternative method is described briefly by Laden et al. (1999) in an abstract. This
29 method uses elemental components of PM25 and a priori characterizations of sources by
30 elemental profiles to estimate the contributions attributable to "crustal" materials (Al and Si), and
October 1999 6-180 DRAFT-DO NOT QUOTE OR CITE
-------
1 so on. Laden et al. (1999) found that little excess mortality in the Harvard Six Cities could be
2 attributed to the "crustal" component of PM25.
3 The preliminary findings of these methods suggest possible direction for improved
4 characterization of ambient PM components with differential toxicity, whether defined by size,
5 source, or composition. The Ozkaynak et al. (1996) approach uses components of air pollution
6 and weather. The approach of Laden et al. (1999) uses elemental components of PM25. The
7 findings of Laden et al. (1999) are also discussed in Section 8.4.
8 These methods may also be advantageous in comparing results across cities, using factors.
9 The correlations between air pollutants may differ substantially from one city to another.
10 Therefore, cross-city comparisons using only two pollutant correlations at a time, such as in the
11 Schwartz (1999a) eight-cities assessment, may not adequately characterize effects such as
12 hospital admissions or mortality in multi-pollutant models. Comparisons of factors from
13 different cities' correlation matrix or covariance matrix decompositions may provide useful
14 information, particularly if site-specific source profiles are appropriate.
15
16 6.4.6.4 Chemical Components of PM25 and PM10
17 The new studies continue to identify sulfates (and presumably acidity) as predictive of
18 adverse health effects in some cases, such as SO4= in the Burnett et al. (1998) and Fairley (1999)
19 time-series mortality studies, and not predictive in other situations, such as SO4= in Abbey et al.
20 (1999). It is possible that sulfate acidity is a major factor in causing health effects under certain
21 conditions, but not always. The conditions under which adverse effects occur may require
22 conditions external to PM (such as weather or co-pollutant concentrations), or conditions specific
23 to the particles (size range, concentration of transition metals, organic carbon, and so on).
24 The study by Fairley et al. (1999) also finds a strong effect of PM10 nitrates on mortality.
25 The nitrates may also occur in coarse particles, but PM10_25 was not predictive of mortality,
26 whereas the PM2 5 and NO3 effects were of approximately equal strength in predicting mortality.
27 This suggests nitrate acidity in PM2 5 may be an important factor in locations such as Santa Clara
28 County, where SO2 and sulfate levels are very low, NO2 and nitrate levels relatively high.
29 Toxicology studies of PM with high nitrate acidity would clarify this question.
30 There is increasing interest in evaluating the effects of iron or other transition metals
31 associated with PM. There are few epidemiology studies in which ambient metal concentrations
October 1999 6-181 DRAFT-DO NOT QUOTE OR CITE
-------
1 are used explicitly as predictors of health endpoints. Dusseldorp et al. (1994) identified adverse
2 effects associated with Fe. Other studies in progress will evaluate human health endpoints in
3 relation to metals. This is of particular interest in locations such the Utah Valley, where the
4 major point source of particle emissions was a steel mill, and in other metal processing
5 communities. The question then arises as the effects of different physical and chemical types of
6 particles, including mixtures of metal, metals and acids, and mineral matrix embedding of metals
7 within particles.
8 There is also important new epidemiology evidence that certain types of particles may be
9 much less toxic than other types of particles. Papers by Schwartz et al. (1999), Pope et al.
10 (1999b), Ostro et al. (1999), and Laden et al. (1999) all suggest that particles of crustal or
11 geological origin may be much less toxic than other particles more typical of urban combustion
12 products. These ambient concentrations of crustal particles are particularly high in the coarse
13 PM10_25 fraction during wind-blown dust episodes, and may even be a major component of PM25
14 in the "intermodal" PM25_j 0 fraction during such episodes. Under these conditions, the
15 concentration of ambient particles in a certain size class may not be a unique indicator of risk.
16 While high concentrations of acid sulfates, metals, and other potentially toxic components of
17 PM25 in eastern U.S. cities are presumed to be a health risk, high concentrations of fine particles
18 may cause fewer or less serious health effects under other conditions. It is not possible at this
19 time to provide a general definition of the conditions under which crustal particles are not
20 harmful to human health, and non-crustal particles from different sources are harmful to health.
21
22 6.4.7 Effects of Exposure Estimation and Model Specification Errors in
23 Epidemiology Studies
24 The interpretation of statistical associations between ambient PM and adverse health effects
25 reviewed in Sections 6.2 and 6.3 depends on the appropriate attribution of effects to ambient PM
26 exposures and to other environmental factors. Some questions have been raised about the
27 adequacy of the air pollution concentration measurements to carry; this inferential burden. It is
28 known that the measurements of ambient air pollution may have instrumental errors, spatial
29 variability, and temporal uncertainties that affect statistical and causal inferences. Individual
30 variation in personal exposure to ambient PM provide a different kind of "error". These are
31 called "measurement errors", and include a wide variety of issues that are discussed in this
October 1999 6-182 DRAFT-DO NOT QUOTE OR CITE
-------
1 section. Incorrect specification of the statistical model also constitutes a "characterization error",
2 and the combination of measurement errors and characterization errors may be important.
3 A number of authors have raised questions about the effects of mis-specification of the
4 statistical models used in various epidemiology studies. We discuss several of these issues here,
5 with regard to particular aspects of the data analyses in the studies. The modeling failures most
6 often cited as relevant to daily time series or prospective cohort epidemiology studies are:
7 (1) PM indicators and indicators of other air pollutants measured at a central monitor, do not
8 adequately characterize the exposure of the population, or of individuals in the study group
9 (2) important covariates are not included in the model, or are replaced by error-prone surrogates;
10 (3) the functional relationship of the exposure-response (ER) function or the distributional
11 model is incorrectly specified.
12
13 6.4.7.1 Measurement Errors in Air Pollution Exposure Surrogates: General Issues
14 The term "measurement error" has been applied to a number of distinct concepts, and it is
15 useful to distinguish several kinds of "error" to which the term might apply: (a) analytical or
16 instrument errors, apart from exposure imputation; (b) spatial errors arising from assignment of a
17 measurement at a stationary air monitor (SAM) as an index of exposure to ambient air pollution
18 among all members of the population; (c) assignment of a time- averaged value to an individual
19 who is exposed to ambient air pollution over a space-time variable range; (d) assignment of the
20 same air pollution exposure index to individuals with different patterns of biological and
21 behavioral intake of ambient air. The differences are sketched in Figure 6-7. The first error
22 (Figure 6-7a) is that two so-called "identical" instruments (denoted A and A') located at the same
23 place ("collocated") will almost surely give different values of the ambient air pollution
24 concentration over any given time interval. Instruments that use different measurement
25 techniques or methodologies are expected to show systematic as well as random differences.
26 A different form of exposure measurement error is suggested in Figure 6-7b. Here we show
27 measurements at different locations (A, B, C, D) in a hypothetical region, along with
28 concentration isopleths to suggest spatial differences at various locations. Any combination of
29 these measurements to produce a single regional value introduces an error in the spatial
30 interpolation of the pollutant concentration field to a given point, for example, to the average
31 ambient concentration at the subject's residence (denoted PR). Figure 6-7c suggests an even
October 1999 6-183 DRAFT-DO NOT QUOTE OR CITE
-------
Co-located
monitors at
A and A1
(a)
D
Monitors located
at A, B, C, D
(b)
D
Personal exposure
in a space-time
trajectory P
(c)
Figure 6-7. Ambient PM concentration isopleths and monitoring sites in a hypothetical
urban area. Figure (a): two colocated monitors. Figure (b): Four regional
monitors. Figure (c): personal exposure trajectory of a subject with a
residence at PR and a moving personal exposure monitor PEM.
October 1999
6-184
DRAFT-DO NOT QUOTE OR CITE
-------
1 more complicated and realistic situation, in which both the individual and the spatial
2 concentration move over time so that a hypothetical personal monitor (denoted PEM) on the
3 individual would measure a space-and-time-weighted exposure averaged concentration. We do
4 not show an example of (d), in which two otherwise hypothetical individuals with different levels
5 of physical activity might inhale different amounts of pollutants, even if they had the same
6 space-time trajectory as PEM.
7 The statistical consequences of using imprecisely measured indicators of air pollution
8 exposure have long been known. Suppose that X is the true individual exposure to pollution of
9 ambient origin, measured without error (a) along trajectory PEM. However, this is not observed,
10 and cannot be observed with perfect accuracy. All that is observed is an error-prone surrogate
11 measurement denoted W. W could be any of: the central monitor A; the average M over several
12 sites (for example, M = (A + B + C + D)/4 ); the residential monitor PR; or the personal monitor
13 PEM. Each choice induces a different statistical model for an observed outcome.
14 Since X can't be measured accurately, all we have is a surrogate measurement W, where W
15 is one of the observed values A, M, PR, or PEM. No hospital admissions or mortality studies
16 have used PEM. Classical measurement error models have assumed that the measurement error
17 in X (denoted U) is additive and normally distributed, so that W = X + U. Type (a) errors might
18 be of this form, although a more complicated structure such as a combination of additive or
19 multiplicative errors with unequal variance seems more likely. Studies in progress should soon
20 clarify the nature and distribution of these errors.
21 The exposure measurement errors of type (b) or type (c) have a different statistical
22 structure. The true exposure X is not observed. However, all individuals are assigned the same
23 value W (again, A or M). The linear error model in this case is defined by X = W + U. This is
24 sometimes called Berkson error structure. Both Berkson and classical errors may be present.
25
26 6.4.7.2 New Theoretical Assessments of Consequences of Measurement Error
27 Since the 1996 PM AQCD (U.S. Environmental Protection Agency, 1996) there have been
28 some advances in the conceptual framework to investigate the effects of measurement error on
29 PM health effects estimated in time-series studies. These studies evaluated the extent of bias
30 caused by measurement errors under a number of scenarios with various error variances and
31 covariance structure between co-pollutants.
October 1999 6-185 DRAFT-DO NOT QUOTE OR CITE
-------
1 Zidek et al. (1996) investigated the joint effects of multicollinearity and measurement error
2 in Poisson regression models with two covariates, through simulations over error size and
3 correlation. Their error model was of classical error form (W=X+U where W and X are
4 surrogate and true measurements, respectively, and the error U is normally distributed). The
5 results illustrated the transfer of effects from the "causal" variable to the confounder. However,
6 for the confounder to have larger coefficients than the true predictor, the correlation between the
7 two covariates had to be large (r = 0.9), with moderate error (o > 0.5 ) for the true predictor, and
8 no error for the confounder in their scenarios. The transfer-of-causality effect was mitigated
9 when the confounder also became subject to error. Another interesting finding that Zidek et al.
10 reported is the behavior of the standard errors of these coefficients: (1) when the correlation
11 between the covariates are high (r = 0.9), and both covariates have no error, the standard errors
12 for both coefficients were inflated by factor of 2; (2) however, this phenomenon disappeared
13 when the confounder had error. Thus, the multicollinearity influences the significance of the
14 coefficient of the causal variable only when the confounder is accurately measured.
15 Zeger et al. (1999) illustrated the implication of the classical error model and the Berkson
16 error model in the context of time-series study design. Their simulation of the classical error
17 model with two predictors, with various combination of error variance and correlation between
18 the predictors/error terms, showed results similar to those reported by Zidek et al. (1996). Most
19 notably, for the transfer of the effects of one variable to the other (i.e., error-induced
20 confounding) to be large, the two predictors or their errors need to be substantially correlated.
21 Also, for the spurious association of a null predictor to be more significant than the true
22 predictor, their measurement errors have to be extremely negatively correlated. Zeger et al. also
23 laid out a comprehensive framework for evaluating the effects of exposure measurement error on
24 estimates of air pollution mortality relative risks in time-series studies. The error, the difference
25 between personal exposure and the central station's measurement z, was decomposed into three
26 components: (1) the error due to having aggregate rather than individual exposure; (2) the
27 difference between the average personal exposure and the true ambient level; and, (3) the
28 difference between the true and measured ambient level. By aggregating individual risks to
29 obtain expected number of deaths, they showed that the first component of error (the aggregate
30 rather than individual, or, the Berkson error) is not a significant contributor to bias in the
31 estimated risk. The second error component is a classical error and can introduce bias if there are
October 1999 6-186 DRAFT-DO NOT QUOTE OR CITE
-------
1 short-term associations between indoor source contributions and ambient levels. The third error
2 component is also of the classical error type, and includes spatial and instrumental error. Using
3 this framework, the investigators then used PTEAM Riverside, CA data to estimate the second
4 error component and its influence on estimated risks. The correlation between the average
5 population PM10 exposure and the ambient PM10level was estimated to be 0.58. By simulating
6 the average personal exposure from the measured ambient PM10 levels using this relationship,
7 they compared the estimated mortality relative risks in a log-linear model for the measured
8 ambient PM10 data and the simulated average population PM10 exposure for Riverside, CA,
9 during the period 1987-1994. They found that the relative risk (1.20% per 10 //g/m3 increase in
10 PM10 95% CI: -0.38 - 2.77%) estimated using the simulated population average exposure was
11 larger but less precise than the relative risk (0.78%; 95% CI: -0.17- 1.75%) estimated using the
12 ambient data.
13 Zeger et al., in the analyses described above, also suggested that the error due to the
14 difference between the average personal exposure and the ambient level (the second type
15 described above) are likely the largest source of bias in estimated relative risk. This suggestion
16 at least partly comes from the comparison of PTEAM data and site-to-site correlation (the third
17 type of error described above) for PM10 and O3 in 8 US cities. While PM10 and O3 both showed
18 relatively high site-to-site correlation (~0.6-0.9), we do not necessarily expect similar extent of
19 site-to-site correlation for other pollutants. Ito et al. (1998) estimated site-to-site correlation
20 (after adjusted for seasonal cycles) for PM10, O3, SO2, NO2, CO, temperature, dewpoint
21 temperature, and relative humidity using multiple stations' data from seven Central and eastern
22 states (IL, IN, MI, OH, PA, WV, WI), and found that, in a geographic scale of less 100 miles,
23 these variables could be categorized into three groups in terms of the extent of correlation:
24 (1) weather variables (r > 0.9); (2) O3, PM10, NO2 (r: 0.6 - 0.8); CO and SO2 (r < 0.5). These
25 results suggest that the contribution from the third component of error as described in Zeger et al.
26 would vary among pollution and weather variables. Furthermore, the contribution from the
27 second component of error would also vary among pollutants, as the contributions from indoor
28 sources are expected to be different for each pollutants. Thus, more data on personal exposures
29 for multiple pollutants is needed in order to assess the effects of these errors. Some of the
30 ongoing exposure studies are expected to shed some light on this issue.
October 1999 6-187 DRAFT-DO NOT QUOTE OR CITE
-------
1 With regard to the PM exposure, the longitudinal part of the PTEAM data discussed above,
2 as well as other studies (e.g., Tamura et al., 1996) show reasonable correlation (r = 0.6 - 0.9)
3 between ambient and average population PM exposure, lending support for the use of ambient
4 data as a surrogate for personal exposure to outdoor PM in time-series mortality or morbidity
5 studies, under conditions similar to these non-smoking Japanese households. Furthermore, fine
6 particles are expected to show even better correlation. Wilson and Suh (1997) examined site-to-
7 site correlation of PM10, PM25, and PM10_25 in Philadelphia and St. Louis, and found that site-to-
8 site correlations for PM2 5 were high (r ~0.9) , but low for PM10_2 5 (r ~ 0.4), indicating that fine
9 particles have smaller errors in representing community-wide exposures. This finding supports
10 Lipfert and Wyzga's (1997) speculation that the stronger mortality associations for fine particles
11 than coarse particles found in Schwartz et al.'s (1996) study may be due to larger measurement
12 error for coarse particles.
13 However, as Lipfert and Wyzga (1997) suggested, the issue is not whether the fine particle
14 association with mortality is a "false positive", but rather, whether the weaker mortality
15 association with coarse particles is a "false negative". Carrothers and Evans (1999) also
16 investigated the joint effects of correlation and relative error, but they specifically addressed the
17 issue of fine (FP) versus coarse particle (CP) effects, by assuming three levels of relative toxicity
18 of fine versus coarse particles (Ppp / PCP =1,3, and 10), and then evaluating the bias, (B = |E[PF]
19 / E[PC,]} / (PF / Pc}, as a function of FP-CP correlation and relative error associated with FP and
20 CP. Their results indicate: (1) if the FP and CP have the same toxicity, there is no bias (i.e.,
21 B=l) as long as FP and CP are measured with equal precision, but, if, for example, FP is
22 measured more precisely than CP, then FP will appear to be more toxic than CP (i.e, B > 1);
23 (2) when FP is more toxic than CP (i.e., PFP / PCP = 3 and 10), however, the equal precision of
24 FP and CP results in downward bias of FP (B < 1), implying an relative overestimation of the
25 less toxic CP. That is, to achieve non-bias, FP must be measured more precisely than CP, even
26 more so as the correlation between FP and CP increases. They also applied this model to real
27 data from the Harvard Six-Cities Study, in particular, the data from Boston and Knoxville.
28 Estimation of spatial variability for Boston was based on an external data, and a range of spatial
29 variability for Knoxville (since there was no spatial data available for this city). For Boston
30 (where estimated FP-CP correlation was low (r = 0.28), estimated error was smaller for FP than
31 for CP (0.85 versus 0.65, as correlation between true versus error-added series), and the observed
October 1999 6-188 DRAFT-DO NOT QUOTE OR CITE
-------
1 FP to CP coefficient ratio was high (11)), the calculated FP to CP coefficient ratio was even
2 larger (26), providing evidence against the hypothesis that FP is absorbing some of the
3 coefficient of CP. For Knoxville, where FP-CP correlation was moderate (0.54), the error for FP
4 was smaller than for CP (0.9 versus 0.75), and the observed FP to CP coefficient ratio was 1.4),
5 the calculated true FP to CP coefficient ratio was smaller (0.9) than the observed value,
6 indicating that the coefficient was overestimated for the better-measured FP, while the coefficient
7 was underestimated for the worse-measured CP. Since the amount (and the direction) of bias
8 depended on several variables (i.e., correlation between FP and CP; the relative error for FP and
9 CP; and, the underlying true ratio of the FP toxicity to CP toxicity), the authors concluded "...for
10 instance, it is inadequate to state that differences in measurement error among fine and coarse
11 particles will lead to false negative findings for coarse particles".
12 Fung and Krewski (1999) conducted a simulation study of measurement error adjustment
13 methods for Poisson models, using scenarios similar to those used in the simulation studies that
14 investigated implication of joint effects of correlated covariates with measurement error. The
15 measurement error adjustment methods employed in this analysis were Regression Calibration
16 (RCAL) method (Carroll et al, 1995) and Simulation Extrapolation (SIMEX) method (Cook and
17 Stefanski, 1995). The RCAL algorithm consists of: (1) estimation of the regression of X on W
18 (observed version of X, with error) and Z (covariate without error); (2) replacement of X by its
19 estimate from (1), and conducting the standard analysis (i.e., regression); and (3) adjustment of
20 the resulting standard error of coefficient to account for the calibration modeling. SIMEX
21 algorithm consists of: (1) addition of successively larger amount of error to the original data;
22 (2) obtaining naive regression coefficients for each of the error added data sets; and, (3) back
23 extrapolation of the obtained coefficients to the error-free case using a quadratic or other
24 function. Fung and Krewski examined the cases for : (1) Px = 0.25; Pz = 0.25; (2) Px = 0.0;
25 pz = 0.25; (3) px = 0.25; pz = 0.0, all with varying level of correlation (-0.8 to 0.8) with and
26 without classical additive error, and also considering Berkson type error. The behaviors of naive
27 estimates were essentially similar to other simulation studies. In most cases with the classical
28 error, RCAL performed better than SIMEX (which performed comparably when X-Z correlation
29 was small), recovering underlying coefficients. In the presence of Berkson type error, however,
30 even RCAL did not recover the underlying coefficients when X-Z correlation was large (> 0.5).
31 This is the first study to examine the performance of available error adjustment methods that can
October 1999 6-189 DRAFT-DO NOT QUOTE OR CITE
-------
1 be applied to time-series Poisson regression. The authors recommend RCAL over SIMEX.
2 Possible reasons why RCAL performed better than SIMEX in these scenarios were not discussed,
3 nor are they clear from the information given in the publication. There has not been a study to
4 apply these error adjustment methods in real time-series health effects studies. These
5 methodologies require either replicate measurements or some knowledge on the nature of error
6 (i.e., distributional properties, correlation, etc.). Since the information regarding the nature of
7 error is still being collected at this time, it may take some time before applications of these
8 methods become practical.
9 Another issue that measurement error may affect is the detection of threshold in time-series
10 studies. Lipfert and Wyzga (1996) suggested that measurement error may obscure the true shape
11 of the exposure response curve, and that such error could make the concentration-response curve
12 to appear linear even when a threshold may exist. Burnett et al. (1999) investigated methods to
13 detect and estimate threshold levels in time series studies. Based on the realistic range of error
14 observed from actual Toronto pollution data (average site-to-site correlation: 0.90 for O3;
15 0.76 for CoH; 0.69 for TSP; 0.59 for SO2; 0.58 for NO2; and 0.44 for CO), pollution levels were
16 generated with multiplicative error for six levels of exposure error (1.0, 0.9, 0.8, 0.72, 0.6, 0.4,
17 site-to-site correlation). Mortality series were generated with three PM10 threshold levels
18 (12.8 //g/m3, 24.6 //g/m3, and 34.4 //g/m3). LOESS with a 60% span was used to observe the
19 concentration-response curves for these 18 combinations of concentration-response relationships
20 with error. A parameter threshold model was also fit using non-linear least squares. Graphical
21 presentations indicate that LOESS adequately detects threshold under no error, but the thresholds
22 were "smoothed out" under the extreme error scenario. Parametric threshold model were
23 adequate to give "nearly unbiased" estimates of threshold concentrations even under the
24 conditions of extreme measurement error, but the uncertainty in the threshold estimates increased
25 with the degree of error. They concluded that "if threshold exists, it is highly likely that standard
26 statistical analysis can detect it".
27 Other issues related to exposure error that have not been investigated include potential
28 differential error among subpopulations. If the exposure errors are different between susceptible
29 population (i.e., people with COPD) and the rest of the population, need to take into account for
30 such difference. Also, the exposure errors may vary from season to season, due to the seasonal
31 differences in the use of indoor emission sources and air exchange rate due to air conditioning
October 1999 6-190 DRAFT-DO NOT QUOTE OR CITE
-------
1 and heating. This may possibly explain the reported season specific effects of PM and other
2 pollutants. Such season specific contributions of errors from indoor and outdoor sources are also
3 expected to be different from pollutant to pollutant.
4 In summary, the studies that examined joint effects of correlation and error suggest that PM
5 effects are likely underestimated, and that spurious PM effects (i.e., qualitative bias such as
6 change in the sign of coefficient) due to transferring of effects from other covariates require
7 extreme conditions, and therefore unlikely. A simulation study suggests that the under the likely
8 range of error for PM, it is unlikely that a threshold is ignored by the common smoothing
9 methods. More data is needed to examine the exposure errors for other pollutants since their
10 relative error contributions and correlations with PM will influence their relative significance in
11 relative risk estimates.
12
13 6.4.7.3 Concentration-Response Relationships
14 New research allows some conclusions about methodology for the shape of the
15 concentration-response function to be used in PM epidemiology studies. Lipfert and Wyzga
16 (1995), Lipfert (1997), and Carroll and Galindo (1998) have raised questions about the role of
17 measurement error in attenuating and linearizing intrinsically non-linear relationships in air
18 pollution epidemiology studies. Piecewise linear or "threshold" models in ambient PM studies
19 have been most often evaluated, but similar issues may affect other models, such as the use of
20 non-parametric smoothers introduced by Schwartz (1994). A few studies compare the goodness
21 of fit of non-parametric models and linear models, but these have not been adjusted for the
22 effects of measurement error in ambient PM and other predictors with which PM is correlated.
23 Schwartz (1998) discusses the threshold issue in detail, plotting a non-parametric smooth
24 fit of total mortality versus PM25 in the Boston component of the Harvard Six Cities study for
25 which (log) linear models were fitted in earlier (Schwartz et al., 1996; Schwartz, 1999b).
26 Schwartz (1996), Pope et al. (1999a) and Ostro et al. (1999) also extensively explore the change
27 in the linear slope when higher PM25 or PM10 concentration days are deleted in daily time series
28 mortality studies. They find that the linear regression coefficient in a log-linear model is either
29 little changed or increases somewhat when PM concentrations are restricted to lower values.
30 Schwartz (1998) interprets this as evidence against a threshold in the lower range of
31 concentrations. Effects of measurement errors in co-pollutants are not evaluated in these papers.
October 1999 6-191 DRAFT-DO NOT QUOTE OR CITE
-------
1 Cakmak et al. (1999) suggest the following two-step approach (discussed in Section 6.4.6)
2 using mortality data for Toronto. First, examine a non-parametric smoothed fit to the data and
3 attempt to identify thresholds or other non-linearities graphically. Secondly, fit a (log)linear
4 model where indicated, and both linear and threshold models where a threshold is suggested.
5 Simulations showed that standard Poisson methods using a threshold model had a high success
6 rate in detecting threshold models with moderately large thresholds versus linear models with
7 zero threshold. Cakmak et al. (1999) conclude that "... standard statistical methods of regression
8 modeling and diagnostic plots of the association between air pollution and daily mortality rates
9 are adequate to detect the presence of a threshold if it exists. Using these methods, the
10 investigator is likely to correctly identify the form of the association even in the presence of
11 exposure measurement error." It would be useful to extend this work to multiple correlated
12 predictors with measurement errors.
13 The parameters estimated in the time-series models are based on population-level
14 responses, such as daily hospital admissions or deaths. The parameters in the prospective cohort
15 studies are based on individual deaths in relation to risk factors, such as ambient PM, smoking
16 status, and BMI. A threshold parameter for an individual is interpretable, whereas a population-
17 level threshold is difficult to interpret if individual thresholds differ over a wide range, or vary
18 over time. If the "population threshold" is a "non-concept", inconsistent findings about
19 thresholds in different time-series studies is readily understood.
20 Nonlinear concentration-response functions using log(ambient air pollutant concentration)
21 for the pollution predictors is recommended in many analyses by Lipfert and Wyzga (1997,
22 1999). In cases where preliminary graphical assessment suggests a concentration-response
23 function that bends downward with increasing concentration, this might be attempted. It is
24 obviously necessary to restrict this logarithmic model at some lower bound, since it implies that
25 RR = 0 when any concentration = 0 (from the log-log model log [RR] = b * log
26 [concentration] + ...).
27 In summary, it is not possible to confidently assert that the concentration-response model
28 for PM indices in time series or long-term studies is (log)linear, threshold, other nonlinear, or
29 different from site to site. New methods being developed may allow resolution of this issue.
30
October 1999 6-192 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.4.7.4 Methodology Issues in Modeling Time Series Studies
2 There has been a considerable convergence of thought on the most appropriate methods for
3 modeling daily time series epidemiology studies. Many authors use generalized estimating
4 equation (GEE) models to control for autocorrelation (Zeger and Liang, 1986; Samet et al.,
5 1995). The methods now most commonly used in recent studies follow some variation of the
6 following structure:
7 • The daily mortality or hospital admissions counts are smoothed by a nonparametric smoother,
8 in lieu of filtering the counts or other approaches. This amounts to calculating a regression
9 function S(t) as a function of the number Y(t) of events on day t, using either local regression
10 functions (LOESS) or spline functions. The smoothing span for loess smoothers is usually
11 5% to 10% of the length of the time series. The spline function is usually chosen to maximize
12 some criterion such as AIC (Akaike Information Criterion) or BIC (Bayes Information
13 Criterion), and the typical span is 30 to 90 days per degree of freedom. Other
14 non-environmental terms, such as the day of the week or the season, may also be included at
15 this stage. Such functions are implemented in the S-Plus statistical package (Mathsoft, 1998),
16 although similar analyses can be done using other packages. S-Plus has become the
17 instrument of choice for most modelers. The most basic Poisson regression model is specified
18 by E{Y(t)} = exp(S(t)), where S(t) is the smoothed or detrended mortality at time t. S-Plus
19 allows use of non-parametric smoothers in generalized linear models (GLM) that are
20 appropriate for Poisson or over-dispersed Poisson data.
21 • Various weather variable specifications are evaluated. The variables frequently include:
22 mean, minimum, or maximum daily temperature; mean dewpoint temperature; relative or
23 specific humidity. Barometric pressure or changes in barometric pressure are used less often.
24 These variables are evaluated for several lag functions (days t, t-1, t-2, etc.) or moving
25 averages. Parametric functions are sometimes fit, typically low-order polynomials or
26 piecewise linear functions, to account for the fact that weather-attributable deaths increase at
27 very high temperatures and at very low temperatures. Newer analyses more often use
28 non-parametric smoothers as well. The next level of model using generalized additive
29 (Poisson) model (GAM) for Y(t) is now specified by its mean E (Y(t)} = exp( S(t) + W(t)),
30 where W(t) is a linear combination of terms involving the weather variables and their
31 interactions (for example, 'hot and humid' weather). Adjustable parameter coefficients of the
October 1999 6-193 DRAFT-DO NOT QUOTE OR CITE
-------
1 smoothers or polynomials in S(t) and W(t) are implicit in these models, and the S(t) term is
2 re-estimated. This next level model may include model selection so as to minimize deviance
3 (variance) or index of dispersion of the fitted model, or to maximize AIC or BIC.
4 • Various air pollution variable specifications are evaluated. The variables frequently include:
5 mean or maximum concentrations of gaseous pollutants measured hourly; daily PM
6 concentrations, or mean hourly concentration of PM if measured more frequently (e.g., by
7 TEOM®). These variables are evaluated for several lag functions (days t, t-1, t-2, etc.) or
8 moving averages. Parametric functions are sometimes fit, typically linear or piecewise linear
9 functions, to account for a possible 'threshold' in the data. Newer analyses sometimes use
10 non-parametric smoothers of air pollution as well. The generalized additive (Poisson) model
11 (GAM) for Y(t) is now specified by its mean E{Y(t)} = exp( S(t) + W(t) + P(t) ), where P(t) is
12 a linear combination of terms involving the air pollution variables. This is a higher order
13 model, usually with re-estimation of S(t) and W(t). Selection of air pollution variables may be
14 done at this stage, or - if the number of candidates is not too large - results of several models
15 are reported. It is not always clear that the models reported are "best" in any statistically
16 objective sense. While any mathematical model in epidemiology can only be an
17 approximation to reality, some models will fit a given data set better than other models, even
18 though the best-fitting models are not necessarily "true". While EPA does not subscribe to
19 "the tyranny of the optimal" (a remark attributed to Prof. John Tukey), there is often little
20 information provided in published papers about the goodness of fit of the selected model(s)
21 relative to other candidate models. The models reported may be the best of the models that
22 include some statistically significant PM index, but there may be other models that are
23 "good". The "good" models may or may not include some PM index. More information on
24 goodness-of-fit would benefit an objective comparison of the model(s) reported in the study.
25 • The final stage consists of detailed sensitivity analyses with respect to alternative model
26 specifications. If season was not included earlier, season and seasonal interactions may be
27 evaluated at this stage. Data-specific estimates may be carried out, such as excluding all days
28 with PM greater than some cut-point, or estimating a PM effect in different subsets (strata) of
29 values of a co-pollutant.
30 Unless explicitly stated otherwise, the studies reviewed here have used this approach.
31
October 1999 6-194 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.4.7.5 General Issues in Modeling Prospective Cohort Studies
2 The prospective cohort studies have generally been modeled using the Cox proportional
3 hazards regression model for discrete outcomes such as death, or various regression and repeated
4 measures analyses for continuous outcomes, such as pulmonary function decrements. The studies
5 have used a variety of covariates specific to the individual, including age, sex, various indices of
6 current or past smoking behavior, other indices of potential health problems or conditions such as
7 body-mass index, and some indices of current or past alcohol use where available. These studies
8 do not use measured individual indicators of air pollution, so may be regarded as "semi-
9 individual" studies of air pollution risk. The ambient air pollution indices are surrogates for
10 exposure that vary in terms of their ability to characterize individual exposure to air pollution of
11 ambient origin. Sulfates and PM25 of ambient origin are good indicators of personal exposure to
12 sulfates (which have no significant indoor sources), and to PM25 of ambient origin, because
13 PM25 has a reasonably uniform regional distribution, and it has a high penetration rate into
14 indoor locations (Wilson and Suh, 1997). In locations with significant local sources of coarse
15 particles, PM10 may include a substantial contribution from coarse particles, in which case
16 ambient PM10 may not be as good an index of exposure to personal PM10 of ambient origin.
17 Recent analyses of the AHSMOG study (Beeson et al, 1998; Abbey et al., 1999) use individually
18 estimated air pollution exposure studies.
19 Other prospective study designs should be considered. Multi-level mixed models allow for
20 random effects of community not accounted for by the measured ambient pollution levels and the
21 individual covariates. The USC/CARB study (Peters et al., 1999a,b) illustrates use of these
22 models.
23 Questions may always be raised about the selection of the subjects in the study, and about
24 the adequacy of the adjustments for individual risk factors. EPA's assessment is that these were
25 generally adequate, although extensive independent re-analyses of the data in (Dockery et al.,
26 1993) and (Pope et al., 1995) are being carried out under the sponsorship of HEI. It is hoped that
27 the new analyses will clarify the role of many of the issues raised here.
28
29 6.4.7.6 Effects of Co-Pollutants on Estimated PM Effects
30 One of the more important areas of uncertainty is in the assessment of the effects of
31 co-pollutants on the estimated ambient PM effect. As noted in Section 6.3.2, the effect of
October 1999 6-195 DRAFT-DO NOT QUOTE OR CITE
-------
1 inclusion of co-pollutants in the time series mortality studies is to reduce the estimated effect
2 attributable to ambient PM in some cases (often found for SO2 in eastern U.S. cities), to increase
3 the estimated effect attributed to PM (found for NO2 at some sites), or to have little effect on the
4 PM coefficient. The expected effect, based on analogies with effects of collinearity and
5 measurement error in ordinary linear regression models, would be to find an attenuated estimate
6 of ambient PM effect when positively correlated co-pollutants with small measurement error are
7 included in the model. There is a general awareness that omission of co-pollutants that may have
8 mortality effects and inclusion of co-pollutants that may have no mortality effect can bias the
9 estimated PM effect.
10 A variety of methods have been proposed that may allow more complete assessment of the
11 separability of ambient PM effects from those of other pollutants where this is possible, and
12 better identification of those cases where separation is not possible. Much could be learned by
13 use of principal components of ambient air pollution concentrations, as discussed in
14 Section 6.4.6.3. Section 6.4.6.2 identifies a number of studies in which PM25 effects were little
15 attenuated by co-pollutants.
16 One issue that has received remarkably little attention is the possible interaction of PM and
17 co-pollutants as effect modifiers. This is simply included in log-linear Poisson regression models
18 by use of linear interaction terms, such as PM*CO and so on. Although "interactions" are
19 sometimes identified as elements in an analysis, no results for linear co-pollutant interactions
20 appear to have been published. Interactions of PM indices and weather or season indices are
21 sometimes reported.
22 The only published co-pollutant interaction model appears to be in (Samet et al., 1996),
23 reproduced by the U.S. Environmental Protection Agency (1996a). They presented three-
24 dimensional plots of non-parametric surfaces, showing expected excess mortality as a function of
25 TSP and SO2, by season. The joint effect of TSP and SO2 appeared to vary by season. Although
26 software for these analyses is readily available, no subsequent investigations have been carried
27 out.
28 A variety of analytical methods are available for exploring co-pollutant effects in
29 time-series studies, and in prospective cohort studies where there are enough sites to produce
30 enough variation among PM and co-pollutant correlations that their effects can be separated.
October 1999 6-196 DRAFT-DO NOT QUOTE OR CITE
-------
1 These methods have not yet been widely used, and many important questions about the separate
2 effects of specific pollutants in an urban air mixture remain unresolved.
3
4 6.4.8 Synthesis of Information from Multiple Studies: Meta-Analyses
5 6.4.8.1 General Issues
6 There is now an extensive literature on methods for summarizing results of diverse research
7 studies into a single evaluation or synthesis, this being one of the most important aims of this
8 document. Some earlier documents provided a qualitative synthesis, or a "study of studies".
9 Much of the emphasis in the recent literature has been on methods for providing quantitative
10 summaries, often called "meta-analysis". However, there is not yet any consensus on how this
11 should be done. There is some question as to whether quantitative summaries can be overly
12 concise, implying greater precision and confidence in the quantitative summary than may be
13 justified (Bailar, 1997; Weed, 1997a). The basic question is: do meta-analytic summaries of
14 epidemiology studies provide findings of effect that can be subsequently verified using more
15 definitive studies, such as interventions, chamber experiments, or large-scale randomized clinical
16 trials?
17 The possibilities for verifying causal inferences about health effects from ambient air
18 pollution in human populations are very limited, due to the large-scale nature of any
19 interventions, the variety and complexity of atmospheric processes, and the general difficulty of
20 defining appropriate control groups. Clinical random trials are frequently restricted to highly
21 selected groups (an analogy with chamber studies), often excluding susceptible subjects who are
22 of the greatest interest in environmental epidemiology studies.
23 Absent one or two "natural experiments", the best alternative is in the comparison of
24 individual and population-based rates of adverse health effects in relation to differences in air
25 pollution mixtures at different locations. At this time, there are few such multi-city comparisons
26 with sufficient commonality in design and analysis that comparisons of effect sizes are believed
27 to be meaningful, notably the half-dozen APHEA studies. Even among the APHEA studies,
28 effects of differences in air pollutant indices, adjustments for season and weather, and differences
29 in study populations pose complications for their interpretation.
30 Extensive meta-analyses of asthmatic responses are reported in Section 6.2, along with
31 some published syntheses of hospital admissions studies (Schwartz, 1999a). Some APHEA
October 1999 6-197 DRAFT-DO NOT QUOTE OR CITE
-------
1 meta-analyses are also described in Sections 6.2 and 6.3. No attempt was made at this time to
2 synthesize mortality studies in this Chapter, since extensive studies sponsored by the Health
3 Effects Institute (HEI) are currently underway for 20 U.S. cities, soon to be expanded to
4 100 cities, using sophisticated new statistical methods.
5 There is an extensive literature on meta-analysis, much of it in the biomedical fields.
6 Specific applications to environmental epidemiology are given by (Blair et al., 1995; Wong and
7 Raabe, 1996). Although not necessarily representing EPA's views on meta-analyses, the criteria
8 in (Blair et al., 1995) provide a useful and relevant basis for evaluating meta-analyses.
9 Methodological issues in research synthesis require some discussion. Several critical questions
10 in any research synthesis are:
11 (1) Are the summarized studies sufficiently comparable in characterization of effect,
12 design, and analysis that a composite numeric estimate of the ambient PM effect has a
13 meaningful interpretation across all of the studies being summarized?
14 (2) Are the summarized studies sufficiently comparable in characterization of effect,
15 design, and analysis that a composite numeric estimate of the uncertainty of the estimated
16 ambient PM effect is operationally meaningful: do we know what the 'between-study' variation
17 indicates?
18 (3) Does the methodology allow for identification of distinctive sub-groups of studies
19 (such as western Europe vs. eastern Europe in the APHEA meta-analyses), should the data
20 suggest that more than one sub-group is involved?
21 (4) Do the studies synthesized include all studies available? If not, are the study selection
22 criteria unbiased with respect to effect?
23 The general procedures in use allow the summary estimate to be a "weighted" combination
24 of estimates from the individual studies, where the "weights" depend on statistical uncertainty or
25 likelihood of the individual studies and on differences of their estimates from a central estimate
26 or weighted average. Many sophisticated extensions of this basic idea have been proposed,
27 including methods for estimating the effects of including hypothetical unpublished studies with
28 less significant or non-significant findings compared to the published studies from which the
29 estimates were assembled (Givens et al., 1997). Studies with negative or non-significant effects
30 may be less likely to be published than studies with significant positive findings, or may be
31 longer delayed before publication. Likelihood-based methods and Bayesian methods have also
October 1999 6-198 DRAFT-DO NOT QUOTE OR CITE
-------
1 been proposed for reviewing PM studies (Hasselblad et al., 1992; Clyde, 1999). However, no
2 single method is generally accepted although simple "vote counting" is clearly inadequate.
3 Any assessment of the value of specific methods for meta-analysis would probably require
4 knowing the "right" answer from a definitive evaluation carried out separately from the studies
5 being summarized.
6 Under some circumstances, one may have greater confidence in a research synthesis.
7 Sometimes a synthesis or meta-analysis is based on studies by a single investigator or
8 investigative team using virtually identical methods on each individual study; or, at times,
9 a meta-analysis may be based on studies by different investigators using a generally similar
10 analytical strategy, such as in the APHEA studies. Conversely, all studies in the group of
11 analyses may have the same biases.
12 Assessment of the uncertainty of the estimated effect introduces other problems. Most
13 estimates of ambient PM effect size depend on the regression model that was fitted to the data set
14 in the individual study. These models include adjustments for a wide variety of covariates, with
15 choices about which covariates to use, the shape or form of the adjustment for each covariate, the
16 time-lag structure of the model for time series data, the method for smoothing mortality or
17 morbidity counts, and so on. Uncertainty about the estimate of an ambient PM effect from the
18 study clearly depends on the model that was finally selected in obtaining the PM effect estimate,
19 but the uncertainty introduced in the model selection process is hardly ever explicitly included in
20 assessing uncertainty of a composite estimate of effect. Published summaries of models rarely
21 provide enough information to allow any assessment of the uncertainty in selection of the "best"
22 model for estimation.
23 Finally, in any research synthesis, there may be a substantive basis for dividing estimates
24 into subgroups based on location, inclusion or exclusion of common groups of covariates, or
25 other group-level differences. These can be treated by hierarchical regression models or by
26 empirical clustering methods. However, such approaches are not widely used.
27 In summary, no clear consensus currently exists regarding explicit criteria for assessing
28 meta-analyses. Some issues in evaluating research reviews are discussed by Weed (1997b). The
29 criteria that he suggests for evaluating review papers should also be applicable, with reasonable
30 modifications, to the assessment of meta-analyses: These include:
31 (1) Explicit statement of the purpose of the review [meta-analysis];
October 1999 6-199 DRAFT-DO NOT QUOTE OR CITE
-------
1 (2) Literature search methods objective and subjective criteria for inclusion or exclusion
2 of studies;
3 (3) Explicit statement of criteria for evaluating validity or quality of the studies;
4 (4) Methods for summarizing evidence;
5 (5) Criteria used in the review for drawing causal or preventive inferences;
6 The purposes of a review may include: making research recommendations; making causal
7 scientific inferences; making recommendations for public health preventive action or
8 intervention. For example, the purposes of this Chapter are to review the evidence relating
9 adverse health effects in humans with environmental exposure to ambient PM, to draw
10 conclusions about the role of ambient PM concentrations in causing the adverse effects, and to
11 quantify the relationship between ambient PM and health effects where possible.
12 Literature searches were carried out in support of the substantive reviews in Sections 6.2
13 and 6.3. In general, this Chapter includes mainly studies published since the last PM Criteria
14 Document was completed early in 1996. Earlier studies are included when there is reason to
15 expand the discussion over that in U.S. Environmental Protection Agency (1996) so as to
16 emphasize the interpretation of the more recent studies. The material selected for review here
17 includes: publications in peer-reviewed journals, books, and conference proceedings; peer-
18 reviewed publications that are in press or accepted for publication; abstracts of papers; Ph.D.
19 dissertations; and in some cases, peer-reviewed technical reports where there is an institutional
20 mechanism for publishing these reports in lieu of journal publication. However, some
21 consideration has also been given to unpublished material that has a high probability of being
22 published by the time of ultimate completion of this document. Clearly, much additional
23 material submitted for publication and accepted for publication or published is expected to be
24 included in the next external review draft of this Criteria Document expected to be released in
25 mid 2000 (incorporating revisions in response to public comment and CAS AC review of the
26 present external review draft, as well as new information accepted by mid-2000 for publication).
27 A number of criteria have been used for placing greater reliance on the results of some
28 studies than others in this draft document. The paramount criteria are:
29 (1) Studies in the United States, or in Canadian populations with air mixtures and
30 demographics similar to those of the United States;
October 1999 6-200 DRAFT-DO NOT QUOTE OR CITE
-------
1 (2) Study populations are clearly identified (for example, by counties of residence or death;
2 by recruitment and exclusion criteria in prospective cohort studies);
3 (3) Health endpoints are clearly identified (for example, by ICD codes);
4 (4) Statistical analyses are described with a sufficiently detailed description of covariate
5 adjustments, ambient PM and co-pollutant exposure indices, effect size and uncertainty, and
6 goodness of fit of the model, such that the informed reader can carry out assessments of study
7 quality and validity, and perform secondary statistical analyses for comparison and synthesis.
8
9 6.4.8.2 Other Recent Research Syntheses and Reviews
10 A review of the findings in many short-term mortality studies was published by Gamble
11 and Lewis (1996). A number of methodological and conceptual issues raised by the Gamble
12 review were evaluated in U.S. Environmental Protection Agency (1997). A critique of the
13 findings in the prospective cohort studies was also published in (Gamble, 1998). Earlier
14 integrative syntheses for mortality were published in Lipfert and Wyzga (1995) and Moolgavkar
15 and Luebeck (1996), and for all health endpoints by Vedal (1997), but these syntheses did not
16 include quantitative meta-analyses. A large-scale meta-analysis was published by Zmirou et al.
17 (1997), mostly involving older studies reviewed in U.S. Environmental Protection Agency
18 (1996). Some investigators such as Schwartz (1999a) have assessed multiple sites in a single
19 publication, looking at hospital admissions in eight cities. Meta-analyses for the APHEA study
20 included Anderson et al. (1997), Katsouyanni et al. (1997), Spix et al. (1998), and Zmirou et al.
21 (1998). Results from 14 cities in the PEACE study are given by Roemer et al. (1998).
22
23
24 6.5 THE USE OF EPIDEMIOLOGY STUDIES FOR CAUSAL
25 INFERENCES ABOUT PM HEALTH EFFECTS
26 6.5.1 Causal Inference and Preventive Intervention
27 Causal inference methodology should play an important role in evaluating the effectiveness
28 of proposed interventions, such as the effects of changes in regulatory standards for ambient PM.
29 Little research has so far been done, and perhaps little can be done using the regression methods
30 that are found in most published papers, that would clarify the predicted effect of reductions in
31 ambient PM on public health. There is an implicit steady-state assumption that if the PM
October 1999 6-201 DRAFT-DO NOT QUOTE OR CITE
-------
1 concentrations in two communities differ by a certain amount, reducing the ambient PM in the
2 community with the higher level to that in the community with the lower level would - sooner or
3 later - reduce the excess of ambient-PM-attributable health effects in the higher community to
4 that in the lower. One of the strongest empirical bases for this expectation is provided by the
5 natural experiment that was carried out in the Utah Valley, UT, where it was shown that there
6 were substantial reductions in a wide range of adverse health effects associated with the
7 13-month closure of a major PM-emitting source, with consequent reductions in ambient PM
8 concentrations (Pope, 1996).
9 PM concentrations have decreased greatly in many cities in the preceding decades. In cities
10 such as London that have virtually eliminated coal smoke as an air pollutant, there have not been
11 any recurrences of smog episodes, such as those of December, 1952 or December, 1962.
12 However, concentrations of many other air pollutants have also decreased during that period of
13 time, so that the effect of banning coal in smoke-free zones can be reasonably, but not wholly,
14 attributed to this intervention. Other major changes have occurred in the population
15 demographics, health care system, transportation, industry, and housing of Greater London.
16 There have even been increases in some sources of air pollution, such as diesel trucks and buses.
17 It is therefore difficult to compare the effects of banning coal in smoke-free zones with the
18 effects of not doing so, "everything else being equal". However, the matter of "absolute proof
19 notwithstanding, it would be also unreasonable to assume that banning coal smoke had nothing
20 to do with the virtual elimination of "black smoke" health episodes, and consequent reduction of
21 adverse health effects at lower concentrations, no matter that lower current ambient PM
22 concentrations make it much more difficult to detect relationships between ambient PM excess
23 mortality or morbidity.
24 Technical methods for dealing with these issues may require further development. It would
25 be necessary to evaluate joint exposure-age-time relationships in order to understand the change
26 in death rates in London and other cities where changes in PM took place. It is also possible that
27 susceptibility to ambient PM increases with age, so the appropriate model for RR may not be a
28 constant-hazard-rate log-linear model where the predictor of effect is the mean ambient PM
29 concentration over the last X years, as in the prospective cohort studies of Dockery et al. (1993),
30 Pope et al. (1995), Beeson et al. (1998), Abbey et al. (1999). An age-weighted cumulative
31 exposure index would likely be more appropriate, and could (in principle) be constructed for
October 1999 6-202 DRAFT-DO NOT QUOTE OR CITE
-------
1 Harvard Six Cities and AHSMOG cohorts, if not for London. This would allow assessment of
2 alternative hypotheses about increased RR in the elderly, for example, increasing cumulative
3 ambient PM exposure with age versus increasing sensitivity or susceptibility with age. Some
4 approaches are described by Thomas (1988), and Thomas et al. (1992) for other applications in
5 environmental epidemiology.
6 It is important to note that "natural experiments" differ from planned experiments in that it
7 may not be possible to control all of the relevant factors. Some factors were probably not
8 affected by the shut-down of the steel mill, such as the range of meteorological conditions during
9 the duration of the shut-down. Other factors, such as the levels of some co-pollutants produced
10 by the plant, probably changed in a manner correlated with the changes in emitted PM levels.
11 However, the study was carried out in a single community, with both baseline pre-shutdown and
12 post-shutdown information available. The evidence must therefore be regarded as providing
13 stronger information about the causal role of ambient PM reductions than do most observational
14 studies.
15 In order for regression models to capture the effect of reducing ambient PM when other air
16 pollutants also play a role in human health, it would be necessary to model and characterize the
17 reductions (or possibly increases) in all other pollutants attendant to a reduction in ambient PM,
18 including PM precursors that could form secondary particles (see Chapter 3). There is a
19 theoretical basis for such modeling (Pearl, 1995). The underlying concept is that it is necessary
20 to evaluate the consequences of not only setting ambient PM to a specific level, but also
21 modifying the values or distribution of values of all of the PM causal ancestors or predecessors
22 that are implied by specifying the ambient PM value. This has so far received scant attention in
23 discussions of ambient PM by forced changes or interventions. Weed (1995) summarizes grade-
24 of-evidence criteria used in preventive inference. He points out that data-and-judgment based
25 causal inferences need not necessarily govern judgment-based preventive inferences. Other
26 applications, such as regulatory decision making, bear the same relation to causal inference for
27 ambient PM health effects that medical decision-making in clinical practice bears to causal
28 inference in medical science. The scientific search for valid causal inferences are expected to
29 help to inform the preventive (public health) inferences. The concern of this document is
30 scientific causal inference, where evidence exists to carry it out.
31
October 1999 6-203 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.5.2 Biological Plausibility in Causal Inference
2 The notion of "biological plausibility" is important for several reasons. The first is that it
3 clearly affects the likelihood of publication of an epidemiology study For example, an editor of
4 the New England Journal of Medicine recently wrote that publication may be warranted for large
5 effects that "do not make biologic sense" (Angell, 1990, p. 824). PM-related health effects are
6 clearly do make "sense", in view of the relationship between excess deaths and respiratory illness
7 in the historic London, Donora, and Meuse Valley episodes. The issue now is whether
8 reasonable analogues of the illnesses and causes of sudden death observed during these episodes
9 can be expected to occur when ambient PM concentrations and related exposures to PM are more
10 than an order of magnitude lower. The publication of a large number of PM epidemiology
11 studies suggest that this remains a subject of concern, even when no specific biologic mechanism
12 is identified with the putative ambient PM-attributable deaths or illnesses.
13 Weed and Hursting (1998, p. 416) discuss three commonly held viewpoints about what
14 constitutes biological plausibility:
15 (1) Viewpoint 1 - "a biologically plausible association is one for which a reasonable
16 mechanism can be hypothesized, but for which no biologic evidence may exist;"
17 (2) Viewpoint 2 - "simply suggesting a mechanism for a factor-cancer association is
18 insufficient. Evidence supporting the proposed mechanism is also necessary;"
19 (3) Viewpoint 3 - "an association is considered biologically plausible if there is sufficient
20 evidence to show how the factor influences a known disease mechanism", (p. 416; authors'
21 italics). The factor in this instance is the concentration of PM of ambient origin in the
22 community.
23 An example of the first viewpoint might be to hypothesize that airborne particles cause
24 inflammatory reactions in the lungs, exacerbating susceptibility to respiratory diseases. A much
25 more extensive discussion of this hypothesis and others was given by Seaton (1996), who cited
26 studies in laboratory animals as supportive of this hypothesis. Some scientists might require
27 proof beyond the animal experiments, which involved artificial particles such as Teflon
28 microspheres, and require evidence of analogous effects using actual particles recovered from the
29 ambient air. An extensive discussion of biological plausibility at this secondary level is given by
30 Schwartz et al. (1999), who argues that differences in PM effects in eight U.S. counties may be
October 1999 6-204 DRAFT-DO NOT QUOTE OR CITE
-------
1 attributable to cardiovascular effects of PM and CO in animals and in humans. This also
2 illustrates the second viewpoint about biological plausibility.
3 The presence of urban air particles or other biological, physical, or molecular markers in
4 human tissue, along with in vivo damage of the tissue, is likely to be necessary for scientists who
5 hold the third point of view. This might be provided by specimens of diseased tissue with
6 airborne particles present in proximity to sites of damage, or by DNA adducts of components of
7 airborne particles in conjunction with DNA damage. A recent study provides such evidence for
8 polycyclic aromatic hydrocarbons (Whyatt et al., 1998) in mothers and newborns in Poland.
9 In our opinion, biological plausibility plays an important role in the acceptance of
10 epidemiology data as providing causal evidence. The studies themselves may or may not provide
11 proof of causality because some other criterion (for example, only modest strength of
12 association) is not satisfied in the mind of the reviewer. We tend to agree with Hill (1965) that
13 biological plausibility is not required in order to accept the results of a consistent set of
14 epidemiology studies, provided that the findings do not strongly disagree with commonly
15 (currently) held beliefs about biological mechanisms. Specific causal mechanisms are discussed
16 in Chapter 7.
17
18 6.5.3 Natural Experiments, Quasi-Experiments, and Causal Inference
19 Most scientists would hold that experimental evidence is the most definitive of all scientific
20 methods. This requires direct intervention in the system being studied, varying the factor of
21 interest, and holding all other factors constant, so that conclusions may be drawn about the
22 experimental factor "everything else being equal".
23 The closest example of this is in the observations in Utah Valley, UT, during the
24 1986-1989 period, when the main source of PM was shut down for 13 months during that period.
25 This constitutes a "natural experiment" rather than a planned intervention, but certain aspects of
26 the Utah Valley studies allow fairly strong conclusions to be drawn. A comprehensive summary
27 of earlier studies is given by Pope (1996). The most difficult issues arise in connection with
28 time-dependent confounders or covariates associated with ambient PM: everything else rarely is
29 "equal", and one would expect that other pollutants with which emitted PM is associated (SO2,
30 NOX, and other combustion products) would also change in a corresponding manner. The Utah
31 Valley studies were evaluated by U.S. Environmental Protection Agency (1996) as showing
October 1999 6-205 DRAFT-DO NOT QUOTE OR CITE
-------
1 likely particle effects in the presence of low concentrations of other co-pollutants. Also, Vedal
2 (1997) writes, "The Utah Valley, for example, experiences only very low concentrations of SO2,
3 and because the particle concentrations occur mainly in the winter, as opposed to the East where
4 the high particle concentrations are largely a summer-time phenomenon, also very low O3
5 concentrations (Pope, 1996). Very low concentrations of acid aerosol in the Utah Valley have
6 also been documented. ... Because the effects are related to particle increases in the wintertime,
7 rather than during the summertime as in the East, it has also been reasonably argued that it is
8 unlikely some effect of meteorology not adequately accounted for in the analyses is responsible
9 for what appears to be the effect of particles." Possible effects of CO, which also tends to be
10 elevated during the winter, were not explicitly evaluated.
11 Other explanations have been advanced for the Utah Valley findings, such as two-year
12 cycles in the rate of respiratory syncytial virus (Lyon et al., 1996), or other short-term variations
13 in respiratory illness (Lamm et al., 1994). Another possibility is that the Utah Valley population
14 changed during the steel mill strike, but there is little evidence of this. However, these arguments
15 have been refuted by comparisons with nearby locations (Pope et al., 1991; Pope, 1996). On the
16 whole, the Utah Valley studies may be interpreted as a natural experiment in which other causes
17 of adverse health effects with cardiopulmonary symptoms were effectively "controlled" by the
18 context, and the largest source of variation was in the large changes in ambient PM levels during
19 the interval.
20
21
22 6.6 CONCLUSIONS AND DISCUSSION
23 6.6.1 Conclusions
24 The epidemiology evidence about the health effects of ambient PM has expanded greatly
25 since the 1996 PM AQCD. The major enhancements in information cover the following areas:
26 (1) Additional studies of health endpoints using ambient PM10 and closely related mass
27 concentration indices such as PM13 and PM7;
28 (2) New studies on a variety of health endpoints for which information on the ambient
29 coarse PM fraction (PM10_25), the ambient fine particle fraction (PM25), and even
October 1999 6-206 DRAFT-DO NOT QUOTE OR CITE
-------
1 ambient ultrafine particle mass concentrations (PM1 and smaller) were observed or
2 estimated from site-specific calibrations;
3 (3) New studies on some health endpoints in which the relationship of health endpoints to
4 ambient particle number concentrations were evaluated;
5 (4) Additional studies which evaluated the sensitivity of estimated PM effects to the
6 inclusion of gaseous co-pollutants in the model;
7 (5) Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures
8 including PM components, based on empirical combinations (factor analysis) or source
9 profiles;
10 (6) New studies of infants and children as a potentially susceptible population;
11 (7) Additional studies on cardiovascular endpoints associated with PM exposures.
12 (8) Studies on asthma and other respiratory conditions exacerbated by PM exposure.
13 This additional information has not yet resolved some of the key issues in PM air pollution
14 epidemiology. Within each group of health effects studies relating to short-term and long-term
15 PM exposure, the findings of statistically significant associations of ambient PM concentrations
16 with mortality or morbidity in some studies are often accompanied by other studies finding
17 PM-health effects associations that are not statistically significant (these should not be
18 inaccurately characterized as 'negative' findings). In fact, reported results are rarely "negative"
19 (indicating a beneficial affect associated with PM exposure), and the negative results reported are
20 very rarely statistically significant. This does not appear simply to reflect any bias in favor of
21 publication of only positive and significant results in peer-reviewed journals. In order to try to
22 capture the flavor of the full range of newly emerging information, some papers have been
23 included here in this PM AQCD draft review based on published abstracts or non-peer-reviewed
24 conference proceedings when the information presented (whether 'positive' or 'negative') is
25 particularly relevant to inference about the role of PM in producing adverse health effects.
26 Overall, based on the information thus far reviewed in this draft document, it appears
27 warranted to conclude that the multiplicity of findings about PM health effects suggest that
28 exposure to ambient PM at current concentrations may cause serious adverse health effects, but
29 that the quantitative magnitude of the effects depends on several environmental and biological
30 factors whose role is not yet known. That is, current levels of ambient PM may be harmful to
October 1999 6-207 DRAFT-DO NOT QUOTE OR CITE
-------
1 human health, but not necessarily equally harmful everywhere or at all times. The most
2 important factors are discussed in the following sections.
3
4 6.6.2 PM Health Effects May Occur In Any Size Fraction; Some Health
5 Effects May Also Be Absent From Some Mass Fractions Under Some
6 Circumstances
7 There are some studies in which health effects were estimated in a large number of
8 locations, using generally very similar data bases and ambient PM exposure indices, including
9 the APHEA studies in 12 European cities, Canadian studies in 13 or 16 cities, the Six (U.S.)
10 Cities studies, and the HEI-sponsored NMMAPS studies in 8-, 20-, and 100 U.S. cities. Results
11 of the NMMAPS studies are still pending. Results of the studies in the other cities have been
12 discussed herein. Interpretations of the relatively modest effect sizes reported (odds ratio or
13 relative risk < 2, often < 1.2) are subject to the possibility that differences may be due to small
14 biases from confounding, from model mis-specification and measurement error, or model
15 selection strategy.
16 The APHEA studies suffer from the use of different PM indices in different locations. The
17 most nearly comparable findings use: (a) PM10 in London (Anderson et al., 1996; Ponce de Leon
18 et al., 1996); (b) PM7 in Lyon (Zmirou et al., 1996); and (c) PM13 in Paris (Dab et al., 1996).
19 Time-series mortality studies found large PM effects for Lyon, much smaller effects for Paris and
20 Cologne. Most other APHEA studies used BS as the PM index. PM effects indexed by BS were
21 much lower in Central-Eastern European cities than in many of the Western European cities, as
22 shown in Section 6.3.2. Various hypotheses have been proposed for this, such as demographic
23 differences in susceptibility, and inadequate adjustments for weather and seasonality. However,
24 the modeling strategy was the same for all the cities, so if inadequate in one region, may be
25 inadequate in all. Perhaps most importantly, it is not clear to what extent the subject BS
26 measurements can be credibly interpreted as mass concentration estimates, given the lack of any
27 documented necessary site-specific calibrations of BS reflectance readings versus co-located
28 gravimetric measurements (see discussion of the need for this in previous U.S. EPA PM
29 AQCDs). Even the comparison across the three western European cities with measurements
30 nominally comparable to PM10 are difficult, with the extent to which estimates of PM-related
October 1999 6-208 DRAFT-DO NOT QUOTE OR CITE
-------
1 total mortality may differ across all three not being clear, especially given mortality differences
2 attributed to different co-pollutant mixtures or climate (as discussed below).
3 Differences across the Harvard Six (eastern and mid-western U.S.) Cities are also of
4 interest in evaluating excess mortality attributable to fine and coarse fractions of PM10. Even
5 though Topeka, KS had a relatively high mean concentration of PM10_25, there was little evidence
6 of excess mortality associated with either the coarse or fine fractions of PM10. It is likely that the
7 coarse fraction is usually dominated by crustal particles in Topeka, and even the fine fraction
8 (from a high concentration of crustal particles smaller than 2.5 //g), to a larger extent than in the
9 other cities.
10 Supporting evidence for the hypothesis that there is relatively little excess mortality
11 associated with coarse particles of crustal origin is provided by recent studies in the Wasatch
12 Front (Salt Lake City, Ogden, and Utah Valley, UT) by Pope et al. (1999b), in Spokane, WA, by
13 Schwartz et al. (1999), and in the Coachella Valley of California (Ostro et al., 1999). These
14 Western U.S. studies show that, by removing from the analyses days in which PM10
15 concentrations appear to be dominated by wind-blown dust of presumably crustal origin, there
16 appears to be a stronger association between PM10 and mortality during normal wind conditions
17 when air pollution in these metropolitan areas is dominated by urban sources. The excess
18 mortality for the Utah Valley (Provo-Orem) was about 8% per 50 mcg/m3 PM10, compared with
19 5% for Ogden and 4% for Salt Lake City. These findings were somewhat higher than earlier
20 studies for Salt Lake City (Styer et al., 1995) and Utah Valley (Pope et al., 1992; Pope, 1996).
21 The local particle sources in the Utah Valley are likely to be enriched in transition metals such as
22 iron, especially during the operation of the steel mill there, possibly accounting for the greater
23 excess mortality attributed to PM10 in Utah Valley. A similar assessment for Topeka would also
24 be of interest.
25 Conversely, Steubenville, OH showed the strongest relationship between PM10_2 5 and
26 mortality among the Harvard Six Cities (Schwartz et al., 1996). Steubenville had a very high
27 mean PM2 5 concentration and a very high correlation between the fine and coarse fractions
28 (0.69). It is therefore possible that some of the excess mortality attributed to the coarse fraction
29 in Steubenville may be due to 'large small particles', i.e., accumulation-mode or fine-mode
30 particles larger than 2.5 //g. Other time series studies find some association between coarse
31 particles and excess total mortality in the Coachella Valley of California (Ostro et al., 1999), and
October 1999 6-209 DRAFT-DO NOT QUOTE OR CITE
-------
1 between excess hospital admissions and mortality in Toronto (Burnett et al., 1998, 1999),
2 although it is possible that the ambient coarse PM fraction in these locations may include a high
3 concentration of coarse particles of non-crustal origin (such as agriculture and wood-burning)
4 that cause adverse health effects. That is, there may well be circumstances in which coarse
5 particles at current concentrations can be harmful to humans. There is some evidence that this
6 may not always apply to coarse particles (or possibly even smaller particles) of crustal origin.
7 Section 6.4.6.2 discusses evidence suggesting that fine particles contribute to excess
8 mortality in infants as well as in adults in Mexico City (Borja-Abuto et al., 1998; Loomis et al.,
9 1999). No comparisons with coarse particles or with PM10 have been published. There is also
10 some indication that PM25 is associated with hospital admissions in Toronto (Burnett et al.,
11 1997b, 1998), and with excess mortality (Burnett et al., 1998) even after adjusting for CO. Other
12 indices such as CoH (elemental carbon particles) and CO may be more strongly related to
13 adverse health effects, but PM25 (and to a lesser extent, PM10_25) appear to retain their
14 significance even when CO is included as a co-pollutant. A seasonal decomposition of fine
15 versus coarse particle effects would be of interest.
16 Recent European studies (Peters et al., 1998) also suggest that ultrafine particles are also
17 significantly associated with adverse health effects in humans. Although the number
18 concentration of ultrafine particles appears to be a better predictor than the mass concentration in
19 their studies, the latter is also associated with health effects.
20 Finally, size-differentiated health effects of particles may also be associated with chemical
21 components or physical properties of the particles, possibly related to their sources. The role of
22 acidity, sulfates, metals, or other components (alone or in conjunction) in causing human health
23 effects remains to be better clarified. The relationship of both sulfate and non-sulfate
24 components of fine particles to excess mortality in the Harvard Six Cities long-term cohort
25 mortality study (Dockery et al., 1993) was shown in the prior PM AQCD (U.S. Environmental
26 Protection Agency, 1996) to be roughly similar. Comparison of sulfate and non-sulfate FP
27 effects, and crustal versus non-crustal FP effects, would be useful in time series studies as well as
28 cohort studies in which both pollutants were available.
29
30
31
October 1999 6-210 DRAFT-DO NOT QUOTE OR CITE
-------
1 6.6.3 PM Health Effects at Different Time Scales
2 Recently published studies (Schwartz 1999b; Zeger et al., 1999) suggest that a substantial
3 part of excess mortality attributable to ambient PM in Boston and Philadelphia may be found in
4 responses that occur from ambient PM concentrations averaged over intervals of about 30 to
5 120 days. These so-called "mid-term" effects may occur in addition to the wide variety of
6 responses that occur on the same day as the ambient PM exposure, or within a few days
7 thereafter. Schwartz (1999b) finds ambient PM25 effects increase RR nearly two-fold for
8 pneumonia and total mortality when a 60-day window is used, compared to a 0-day window at
9 lags 0 to 3 days. This may explain part of the larger RR in the prospective cohort studies of
10 Dockery et al. (1993) and Pope et al. (1995) compared to the daily time-series mortality studies.
11 The recent reanalyses of the AHSMOG cohort by Beeson et al. (1998) and Abbey et al.
12 (1999) find significant effects of PM10 on total mortality, mortality with any contributing
13 non-malignant respiratory causes (CRC), or lung cancer incidence and mortality associated with
14 long exposure to PM10 in males, but not in females. There are also no significant effects
15 associated with long-term S04= or SO2 exposure, which was probably very low in California
16 during most of this time frame, but some relationships with ozone. The most significant PM10
17 associations are found with the average number of days per year in which PM10 exceeded a
18 specific cut-point, such as 100 //g/m3, not with the long-term mean PM10 concentration for each
19 individual. No significant excess mortality was found for cardiopulmonary effects. These
20 findings tend to support the conclusion of the U.S. Environmental Protection Agency (1996) PM
21 AQCD that long-term PM exposures can also be harmful, but with some noteworthy differences
22 about gender and diagnosis of death cause that remain to be resolved. Reanalyses of the Harvard
23 Six Cities and ACS cohort data commissioned by HEI may also modify conclusions about the
24 earlier studies.
25 Key issues in the interpretation of the prospective cohort studies are:
26 (a) Effects of spatial confounding for S04=, and possibly even PM2 5, due to local variation;
27 (b) Lack of assessment of co-pollutant models for comparison;
28 (c) Lack of consideration of longer-term exposures preceding the study interval, or of alternative
29 exposure metrics over different time periods preceding and during the study period;
30 (d) Lack of evaluation of non-linear relationships with PM10 (although the results from use of
31 exceedance indicators in the AHSMOG study suggests these may exist);
October 1999 6-211 DRAFT-DO NOT QUOTE OR CITE
-------
1 (e) The absence of exposure-time-age studies that would better link short-term, mid-term, and
2 long-term PM exposures to specific health endpoints, or to the progression of endpoints from less
3 adverse to more adverse, with numeric estimates of disability and life-shortening;
4 (f) Potential confounding associated with personal or demographic risk factors;
5
6 6.6.4 Alternative Hypotheses for Adverse Health Effects
7 In Section 6.1, some alternative hypotheses for PM health effects were proposed. The
8 foregoing evaluation of recent PM epidemiology studies has allowed some tentative conclusions
9 to be drawn about the plausibility of various alternatives expressed in Table 6.1. These tentative
10 conclusions are shown in Table 6-57. The foregoing discussion suggests a somewhat pessimistic
11 prognosis for meaningful large-scale formal (mathematical) synthesis of PM epidemiology
12 studies, except possibly in sub-categories with similar health endpoints, similar PM
13 characterization, and similar environmental co-factors. There is clear evidence that some
14 categories of PM such as ambient PM2 5 and PM10 may be associated with a number of health
15 effects outcomes under some conditions (perhaps many); that health effects may be associated
16 with the ambient PM coarse fraction under some (but not all) conditions; and that these health
17 effects may vary from place to place and from time to time, depending on a variety of
18 incompletely identified factors. It is further possible that indoor-source particles have different
19 potential for causing health effects than particles from ambient PM exposure. The overall picture
20 is somewhat more diverse than in the previous PM AQCD (in spite of the considerable quantity
21 of new information), because the range of outcomes has expanded considerably. However, a
22 sufficiently large number of additional new findings of health effects being associated with
23 exposure to ambient PM and other atmospheric pollutants (in competently executed independent
24 studies carried out by a large number of investigators around the world) have been reported to
25 justify public health concern that ambient PM exposure at current levels may be a health hazard
26 to susceptible individuals and to at least provisionally conclude that the newer studies support the
27 1996 PM AQCD key judgement that the observed PM-morbidity/mortality associations likely
28 represent causal relationships.
29 Of particular interest are new findings and insights derived from new hospital admission
30 studies using various PM mass metrics (PM10, PM2 5, etc.) The results of the new PM mass
31 studies are generally consistent with and supportive of the studies presented in the 1996 PM
October 1999 6-212 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 6-57. TENTATIVE CONCLUSIONS ABOUT ALTERNATIVE
HYPOTHESES THAT MAY AFFECT THE SYNTHESIS OF
EPIDEMIOLOGY STUDIES
Alternative Hypotheses
Adverse Health Effects Depend
Only on Ambient PM,
Independent of Co-Factors
Adverse Health Effects
Depend on Ambient PM
as well as Co-factors
Adverse health effects
depend only on
ambient PM size range
Adverse health effects
depend on ambient PM
with specific physical
properties or
composition
Not likely. Adverse health
effects from coarse particles
may occur at some sites, not
others.
Possible. Adverse health
effects from ambient PM of a
given size may be different in
sites where PM has different
physical properties or
composition with same PM size
range.
Possible. Adverse health effects
from ambient PM are different in
sites where ambient PM has
different co-factors with same
PM range.
Probable. Adverse health effects
from ambient PM are different in
sites where PM has different
physical properties, composition,
or co-factors, even in the same
ambient PM size range
1 AQCD (U. S. Environmental Protection Agency, 1996). Moreover, mathematical syntheses of
2 multiple hospital admissions studies for the various age and disease categories (including
3 relevant studies from the previous AQCD conducted as part of this new review) generally found
4 overall significant and reasonably consistent RR effect sizes (i.e., within their respective
5 confidence intervals) across admissions categories for both PM10 and SO4=. As discussed by Hill
6 (1965), such coherence across outcomes and among multiple studies conducted in different
7 places by different investigators are supportive of the conclusion that these associations are
8 caused by PM mass or a closely related pollutant correlate. Hospital admissions studies
9 considering multiple PM components were also evaluated in order to assess the relative roles of
10 the various components in the reported PM-health effects associations (in those studies where
11 multiple PM components were considered). These results indicated that sulfates and acidic
12 aerosols are often among the PM metrics most strongly associated with respiratory morbidity.
13 Studies of doctors' visits in some cities indicate that the use of hospital admissions alone
14 can understate the total severe morbidity effects of air pollution. For example, in both Paris and
15 London, the number of doctors' visits amount to many times the number of hospital admissions.
October 1999
6-213
DRAFT-DO NOT QUOTE OR CITE
-------
1 Moreover, the Paris Black Smoke RR for asthma doctors visits was actually much higher than
2 that for asthma hospital admissions (doctors' visits RR=1.74 for 100 //g/m3, versus hospital
3 admissions RR=1.04). These results suggest that considering only hospital admissions and
4 emergency hospital visit effects may greatly underestimate the numbers of medical visits
5 occurring in a population due to acute particulate matter air pollution exposure.
6 Also of much interest are new panel studies that report associations between ambient PM
7 concentrations and worsening of asthma conditions. Most studies reported PM10 results, and
8 PM25 was evaluated in two studies (none used PM10_25 measures of the coarse fraction PM10).
9 Some employed symptoms-scoring approaches and/or asthma medication usage as indicators of
10 asthma exacerbations. A qualitative summary of those studies examining ambient PM10
11 exposure on asthmatic health outcomes indicate that, as a group, the majority of studies report a
12 positive odds ratio for the relationship, with almost half having 95% CI above 1.00, looking at
13 the endpoints one at a time. Viewing all the indicators together within a study may be a better
14 test of the relationship for an asthma attack. Examining all the studies as a group quantitatively
15 describes a stronger relationship. One study examining both PM10 and PM2 5 reported that PM2 5
16 had a larger effect. Another unique study, which (1) evaluated the size distribution of particles in
17 the range 0.001 to 2.5 //m and (2) examined the number of particles, found that the health effects
18 of 5 day means of the number of ultrafine particles were larger than those for the mass of the fine
19 particles. In contrast, another study also examined a range of PM sizes but found PM10 to be
20 more consistently associated with PEF changes, and another study reported that 1-hr and 8-hr
21 maximum PM10 had larger effects than the 24 hr mean. Other newer studies also report
22 significant associations of increased emergency room visits or hospital admissions for asthma
23 with various ambient PM indices. Collectively, these studies point toward exacerbation of
24 asthma likely being related to ambient PM exposures; but they do not yet allow for clear-cut
25 attribution of such effects being more closely related to specific PM size fractions or composition
26 or for clear delineation of quantitative contributions of ambient PM acting along versus in
27 combination with ozone (at least in some studies where significant ozone associations with
28 asthma-related health outcomes were also found). Lastly, it should be noted that, overall, the
29 newer epidemiologic analyses of PM-cardiovascular effect relationships also appear to be
30 generally supportive of previous conclusions in the 1996 PM AQCD highlighting such
31 relationships as likely being causative and of likely public health concern. Again, however, it is
October 1999 6-214 DRAFT-DO NOT QUOTE OR CITE
-------
1 still difficult to delineate quantitatively the proportion of risk attributable to: ambient PM acting
2 alone; PM acting in combination with other co-pollutants; other specific co-pollutants, per se; or
3 overall ambient mix of pollutants.
October 1999 6-215 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
3 suspended participates and oxidants as related to incidence of chronic disease in California Seventh-Day
4 Adventists. Environ. Health Perspect. 94: 43-50.
5 Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995) Long-term
6 ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
7 nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
8 particulate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
9 7: 19-34.
10 Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
11 particulate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med.
12 158:289-298.
13 Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
14 Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
15 Care Med. 159:373-382.
16 Ackermann-Liebrich, U.; Leuenberger, P.; Schwartz, J.; Schindler, C.; Monn, C.; Bolognini, B.; Bongard, J. P.;
17 Brandli, O.; Domenighetti, G.; Elsasser, S.; Grize, L.; Karrer, W.; Keller, R.; Keller-Wossidlo, H.;
18 Kiinzli, N.; Martin, B. W.; Medici, T. C.; Perruchoud, A. P.; Schoni, M. H.; Tschopp, J. M.; Villiger, B.;
19 Wuthrich, B.; Zellweger, J. P.; Zemp, E. (1997) Lung function and long term exposure to air pollutants in
20 Switzerland. Am. J. Respir. Crit. Care Med. 155: 122-129.
21 Agocs, M. M.; Rudnai, P.; Etzel, R. A. (1992) Respiratory disease surveillance in Hungary. Morb. Mortal.
22 Wkly. Rep. 41: 1-8.
23 Agocs, M. M.; White, M. C.; Ursicz, G.; Olson, D. R.; Vamos, A. (1997) A longitudinal study of ambient air
24 pollutants and the lung peak expiratory flow rates among asthmatic children in Hungary. Int. J. Epidemiol.
25 26: 1272-1280.
26 Alberdi Odriozola, J. C.; Diaz Jimenez, J.; Montero Rubio, J. C.; Miron Perez, I. J.; Pajares Ortiz, M. S.; Ribera
27 Rodrigues, P. (1998) Air pollution and mortality in Madrid, Spain: a time-series analysis. Int. Arch. Occup.
28 Environ. Health 71: 543-549.
29 Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects. In:
30 Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC. Environ.
31 Health Perspect. 79: 147-150.
32 Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Strachan, D. P. (1996) Air pollution and daily
33 mortality in London: 1987-92. Br. Med. J. 312: 665-669.
34 Anderson, H. R.; Spix, C.; Medina, S.; Schouten, J. P.; Castellsague, J.; Rossi, G.; Zmirou, D.; Touloumi, G.;
35 Wojtyniak, B.; Ponka, A.; Bacharova, L.; Schwartz, J.; Katsouyanni, K. (1997) Air pollution and daily
36 admissions for chronic obstructive pulmonary disease in 6 European cities: results from the APHEA project.
37 Eur. Respir. J. 10: 1064-1071.
38 Anderson, H. R.; Ponce de Leon, A.; Bland, J. M.; Bower, J. S.; Emberlin, J.; Strachen, D. P. (1998) Air pollution,
39 pollens, and daily admissions for asthma in London 1987-92. Thorax 53: 842-848.
40 Angell, M. (1990) The interpretation of epidemiologic studies. N. Engl. J. Med. 323: 823-825.
41 Atkinson, R. W.; Anderson, H. R.; Strachan, D. P.; Bland, J. M.; Bremner, S. A.; Ponce de Leon, A. (1999)
42 Short-term associations between outdoor air pollution and visits to accident and emergency departments in
43 London for respiratory complaints. Eur. Respir. J. 13: 257-265.
44 Awasthi, S.; Glick, H. A.; Fletcher, R. H.; Ahmed, N. (1996) Ambient air pollution & respiratory symptoms
45 complex in preschool children. Indian J. Med. Res. 104: 257-262.
46 Bailar, J. C., III. (1997) The promise and problems of meta-analysis [letter]. N. Engl. J. Med. 337: 559-561.
47 Barton, K. (1958) The influence of dust on atmospheric corrosion of metals. Werkst. Korros. 8/9: 547-549.
48 Bates, D. V. (1992) Health indices of the adverse effects of air pollution: the question of coherence. Environ. Res.
49 59:336-349.
50 Bates, D. V.; Sizto, R. (1987) Air pollution and hospital admissions in southern Ontario: the acid summer haze
51 effect. Environ. Res. 43: 317-331.
52 Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
53 lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
October 1999 6-216 DRAFT-DO NOT QUOTE OR CITE
-------
1 Beyer, U.; Franke, K.; Cyrys, J.; Peters, A.; Heinrich, J.; Wichmann, H. E.; Brunekreef, B. (1998) Air pollution and
2 respiratory health of children: the PEACE panel study in Hettstedt and Zerbst, Eastern Germany. Eur. Respir.
3 Rev. 8: 52, 61-69.
4 Blair, A.; Burg, J.; Foran, J.; Gibb, H.; Greenland, S.; Morris, R.; Raabe, G.; Savitz, D.; Teta, J.; Wartenberg, D.;
5 Wong, O.; Zimmerman, R. (1995) Guidelines for application of meta-analysis in environmental
6 epidemiology. Regul. Toxicol. Pharmacol. 22: 189-197.
7 Bobak, M.; Leon, D. A. (1992) Air pollution and infant mortality in the Czech Republic, 1986-1988. Lancet
8 (8826): 1010-1014.
9 Bobak, M.; Leon, D. (1998) Air pollution and infant mortality: the effects are specific for respiratory causes in
10 postneonatal period. Epidemiology 9: S58.
11 Bobak, M.; Leon, D. A. (1999) Pregnancy outcomes and outdoor air pollution: an ecological study in districts of the
12 Czech Republic 1986-8. Occup. Environ. Med. 56: 539-543.
13 Bobak, M.; Roberts, A. (1997) Heterogeneity of air pollution effects is related to average temperature [letter].
14 Br. Med. J. 315: 1161.
15 Boezen, M.; Schouten, J.; Rijcken, B.; Vonk, J.; Gerritsen, J.; Van Der Zee, S.; Hoek, G.; Brunekreef, B.;
16 Postma, D. (1998) Peak expiratory flow variability, bronchial responsiveness, and susceptibility to ambient
17 air pollution in adults. Am. J. Respir. Crit. Care Med. 158: 1848-1854.
18 Boezen, H. M.; Van Der Zee, S. C.; Postma, D. S.; Vonk, J. M.; Gerritsen, J.; Hoek, G.; Brunekreef, B.;
19 Rijcken, B.; Schouten, J. P. (1999) Effects of ambient air pollution on upper and lower respiratory symptoms
20 and peak expiratory flow in children. Lancet 353: 874-878.
21 Borja-Aburto, V. H.; Loomis, D. P.; Bangdiwala, S. I.; Shy, C. M.; Rascon-Pacheco, R. A. (1997) Ozone,
22 suspended particulates, and daily mortality in Mexico City. Am. J. Epidemiol. 145: 258-268.
23 Borja-Aburto, V. H.; Castillejos, M.; Gold, D. R.; Bierzwinski, S.; Loomis, D. (1998) Mortality and ambient fine
24 particles in southwest Mexico City, 1993-1995. Environ. Health Perspect. 106: 849-855.
25 Borja-Aburto et al. (1999a).
26 Borja-Aburto et al. (1999b).
27 Braga, A. L. F.; Concei5ao, G. M. S.; Pereira, L. A. A.; Kishi, H. S.; Pereira, J. C. R.; Andrade, M. F.; Gon5alves,
28 F. L. T.; Saldiva, P. H. N.; Latorre, M. R. D. O. (1999) Air pollution and pediatric respiratory hospital
29 admissions in Sao Paulo, Brazil.
30 Braun-Fahrlander, C.; Vuille, J. C.; Sennhauser, F. H.; Neu, U.; Kiinzle, T.; Grize, L.; Gassner, M.; Minder, C.;
31 Schindler, C.; Varonier, H. S.; Wuthrich, B.; SCARPOL team. (1997) Respiratory health and long-term
32 exposure to air pollutants in Swiss schoolchildren. Am. J. Respir. Crit. Care Med. 155: 1042-1049.
33 Bremner, S. A.; Anderson, H. R.; Atkinson, R. W.; McMichael, A. J.; Strachan, D. P.; Bland, J. M.; Bower, J. S.
34 (1999) Short term associations between outdoor air pollution and mortality in London 1992-4. Occup.
35 Environ. Med. 56: 237-244.
36 Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
37 Burnett, R. T.; Dales, R.; Krewski, D.; Vincent, R.; Dann, T.; Brook, J. R. (1995) Associations between ambient
38 particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am. J. Epidemiol.
39 142: 15-22.
40 Burnett, R. T.; Dales, R. E.; Brook, J. R.; Raizenne, M. E.; Krewski, D. (1997a) Association between ambient
41 carbon monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities.
42 Epidemiology 8: 162-167.
43 Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997b) The role of particulate size and chemistry in the
44 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
45 Environ. Health Perspect. 105: 614-620.
46 Burnett, R. T.; Cakmak, S.; Brook, J. R. (1998) The effect of the urban ambient air pollution mix on daily mortality
47 rates in 11 Canadian cities. Can. J. Public Health 89: 152-156.
48 Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of particulate and gaseous air
49 pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
50 Burney, P. (1999) Air pollution and asthma: the dog that doesn't always bark [letter]. Lancet 353: 859-860.
51 Cakmak, S.; Burnett, R.; Krewski, D. (1998) Adjusting for temporal variation in the analysis of parallel time series
52 of health and environmental variables. J. Exposure Anal. Environ. Epidemiol. 8: 129-144.
53 Cakmak, S.; Burnett, R.T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
54 concentrations for air pollution-related mortality with exposure measurement error Risk Anal. 19: 487-496.
October 1999 6-217 DRAFT-DO NOT QUOTE OR CITE
-------
1 Calvin, J. A.; Carroll, R. J. (1998) Effects of Berkson measurement errors in estimating PM exposure-related health
2 effects. Cincinnati, OH: U.S. Environmental Protection Agency; order no. 8C-R369-NASX.
3 Carroll, R. J.; Galindo, C. D. (1998) Measurement error and validating complex models. Environ. Health Perspect:
4 in press.
5 Carroll, R. J.; Ruppert, D.; Stefanski, L. A. (1995) Measurement error in nonlinear models. London, United
6 Kingdom: Chapman & Hall. (Cox, D. R.; Hinkley, D. V.; Keiding, N.; Reid, N.; Rubin, D. B.; Silverman,
7 B. W., eds. Monographs on statistics and applied probability: v. 63).
8
9 Carrothers, T. J.; Evans, J. S. (1999) Assessing the impact of differential measurement error on estimates of fine
10 particle mortality. J. Air Waste Manage. Assoc.: accepted.
11 Cassino, C.; Ito, K.; Bader, L; Ciotoli, C.; Thurston, G.; Reibman, J. (1999) Cigarette smoking and ozone-associated
12 emergency department use for asthma by adults in New York City. Am. J. Respir. Crit. Care Med.
13 159: 1773-1779.
14 Chen, P.-C.; Lai, Y.-M.; Wang, J.-D.; Yang, C.-Y.; Hwang, J.-S.; Kuo, H.-W.; Huang, S.-L.; Chan, C.-C. (1998)
15 Adverse effect of air pollution on respiratory health of primary school children in Taiwan. Environ. Health
16 Perspect. 106:331-335.
17 Chen, C.; Chock, D. P.; Winkler, S. L. (1999) A simulation study of confounding in generalized linear models for
18 air pollution epidemiology. Environ. Health Perspect. 107: 217-222.
19 Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM 10 air
20 pollution. Arch. Environ. Health 52:113-117.
21 Clench-Aas, J.; Bartonova, A.; Skj0nsberg, O. H.; Leegaard, J.; Hagen, L. O.; Giasver, P.; Moseng, J.; Roemer, W.
22 (1998) Air pollution and respiratory health of children: the PEACE study in Oslo, Norway. Eur. Respir. Rev.
23 8: 36-43.
24 Clyde, M. A. (1999) Bayesian model averaging and model search strategies. In: Bernardo, J. M.; Berger, J. O.;
25 Dawid, A. P.; Smith, A. F. M., eds. Bayesian Statistics 6. Oxford University Press.
26 Cook and Stefanski (1995).
27 Coutant, R. W. (1977) Effect of environmental variables on collection of atmospheric sulfate. Environ. Sci.
28 Technol. 11: 873-878.
29 Costa, D. L.; Dreher, K. L. (1999) What do we need to know about airborne particles to make effective risk
30 management decisions? A toxicology perspective. Hum. Ecol. RiskAssess. 5: 481-491.
31 Cropper, M. L.; Simon, N. B.; Alberini, A.; Arora, S.; Sharma, P. K. (1997) The health benefits of air pollution
32 control in Delhi. Am. J. Agric. Econ. 79: 1625-1629.
33 Cuijpers, C. E. J.; Swaen, G. M. H.; Wesseling, G.; Wouters, E. F. M. (1994) Acute respiratory effects of summer
34 smog in primary school children. Toxicol. Lett. 72: 227-235.
35 Dab, W.; Medina, S.; Quenel, P.; Le Moullec, Y.; Le Tertre, A.; Thelot, B.; Monteil, C.; Lameloise, P.; Pirard, P.;
36 Momas, I.; Ferry, R.; Festy, B. (1996) Short term respiratory health effects of ambient air pollution: results of
37 the APHEA project in Paris. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on
38 health: a European approach using epidemiological time series data. J. Epidemiol. Commun. Health
39 50(suppl. 1): S42-S46.
40 Daisey, J. M.; Tsai, F. C.; Apte, M. G. (1999) An exploratory analysis of the relationship between adverse health
41 effects of airborne particles and their chemical composition. Presented at: 3rd colloquium on particulate air
42 pollution & human health; May; Durham, NC.
43 Davis, D. L.; Ahmed, A. K.; Von Schirnding, Y.; Corvalon, C.; Saldiva, P. H. N.; Carreira, J.; Bandhari, R.;
44 White, R.; Wu, C.; Xue, Z. (1999) Risks to children from urban air pollution [penultimate draft].
45 World Resources Institute, World Health Organization, University of Sao Paulo; April 16.
46 Dejmek, J.; Selevan, S. G.; Benes, L; Solansky, I.; Sram, R. J. (1999) Fetal growth and maternal exposure to
47 particulate matter during pregnancy. Environ. Health Perspect. 107: 475-480.
48 De Kok, M. E.; Mertens, P. L. J. M.; Cuijpers, C. E. J.; Swaen, G. M. H.; Wesseling, G. J.; Broer, J.; Sturmans, F.;
49 Wouters, E. F. M. (1996) The rate of respiratory symptoms among primary school children in two Dutch
50 regions. Eur. J. Pediatr. 155: 506-511.
51 Delfmo, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
52 relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154: 633-641.
53 Delfmo, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
54 on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
55 155: 568-576.
October 1999 6-218 DRAFT-DO NOT QUOTE OR CITE
-------
1 Delfino, R. J.; Murphy-Moulton, A. M.; Becklake, M. R. (1998) Emergency room visits for respiratory illnesses
2 among the elderly in Montreal: association with low level ozone exposure. Environ. Res. 76: 61-11.
3 DerSimonian, R.; Laird, N. (1986) Meta-analysis in clinical trials. Controlled Clin. Trials 7: 177-188.
4 Dockery, D. W.; Brunekreef, B. (1996) Longitudinal studies of air pollution effects on lung function. Am. J. Respir.
5 Crit. Care Med. 154(suppl.): S250-S256.
6 Dockery, D. W.; Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily mortality: associations with particulates
7 and acid aerosols. Environ. Res. 59: 362-373.
8 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.
9 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 329:
10 1753-1759.
11 Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
12 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
13 symptoms. Environ. Health Perspect. 104: 500-505.
14 Dusseldorp, A.; Kruize, H.; Brunekreef, B.; Hofschreuder, P.; de Meer, G.; van Oudvorst, A. B. (1994)
15 Associations of PM10 and airborne iron with respiratory health of adults living near a steel factory. Am. J.
16 Respir. Crit. Care Med. 152: 1932-1939.
17 El Khouri Miraglia, S. G.; De Souza Conceisao, G. M.; Saldiva, P. H. N.; Strambi, O. (1997) Analysis of the impact
18 of fuel consumption on mortality rates in Sao Paulo. In: Sucharov, L.; Bidini, G., eds. Urban transport and
19 the environment for the 21st century III: proceedings of the third international conference on urban transport
20 and the environment for the 21st century; September; Aquasparta, Italy. Southampton, United Kingdom:
21 Computational Mechanics Publications; pp. 435-444.
22 Eskeland, G. S. (1997) Air pollution requires multipollutant analysis: the case of Santiago, Chile. Am. J. Agric.
23 Econ. 79: 1636-1641.
24 Evans, J. S.; Kinney, P. L.; Koehler, J. L.; Cooper, D. W. (1984) The relationship between cross-sectional and time
25 series studies. J. Air Pollut. Control Assoc. 34: 551-553.
26 Fairley, D. (1990) The relationship of daily mortality to suspended particulates in Santa Clara county, 1980-86.
27 Environ. Health Perspect. 89: 159-168.
28 Fairley, D. (1994) Mortality and particulate exposure in Santa Clara County, CA 1980-86. Santa Clara, CA: Bay
29 Area Air Quality Management District.
30 Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
31 Perspect. 107:637-641.
32 Ferris, B. G., Jr.; Speizer, F. E.; Spengler, J. D.; Dockery, D.; Bishop, Y. M. M.; Wolfson, M.; Humble, C. (1979)
33 Effects of sulfur oxides and respirable particles on human health: methodology and demography of
34 populations in study. Am. Rev. Respir. Dis. 120: 767-779.
35 Forsberg, B.; Segerstedt, B.; Stjernberg, N.; Roemer, W. (1998) Air pollution and respiratory health of children:
36 the PEACE panel study in Umea, Sweden. Eur. Respir. Rev. 8: 12-19.
37 Frischer, T.; Studnicka, M.; Gartner, C.; Tauber, E.; Horak, F.; Veiter, A.; Spengler, J.; Kuhr, J.; Urbanek, R.
38 (1999) Lung function growth and ambient ozone: a three-year population study in school children. Am. J.
39 Respir. Crit. Care Med. 160: 390-396.
40 Fung, K. Y.; Krewski, D. (1999) On measurement error adjustment methods in Poisson regression. Environmetrics
41 10:213-224.
42 Gamble, J. F. (1998) PM2 5 and mortality in long-term prospective cohort studies: cause-effect or statistical
43 associations? Environ. Health Perspect. 106: 535-549.
44 Gamble, J. F.; Lewis, R. J. (1996) Health and respirable particulate (PM 10) air pollution: a causal or statistical
45 association? Environ. Health Perspect. 104: 838-850.
46 Garty, B. Z.; Kosman, E.; Ganor, E.; Berger, V.; Garty, L.; Wietzen, T.; Waisman, Y.; Mimouni, M.; Waisel, Y.
47 (1998) Emergency room visits of asthmatic children, relation to air pollution, weather, and airborne allergens.
48 Ann. Allergy Asthma Immunol. 81: 563-570.
49 Gielen, M. H.; Van Der Zee, S. C.; Van Wijnen, J. H.; Van Steen, C. J.; Brunekreef, B. (1997) Acute effects of
50 summer air pollution on respiratory health of asthmatic children. Am. J. Respir. Crit. Care Med.
51 155:2105-2108.
52 Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
53 effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
54 53: 817-822.
October 1999 6-219 DRAFT-DO NOT QUOTE OR CITE
-------
1 Givens, G. H.; Smith, D. D.; Tweedie, R. L. (1997) Publication bias in meta-analysis: a Bayesian data-augmentation
2 approach to account for issues exemplified in the passive smoking debate. Stat. Sci. 12: 221-240.
3 Godleski, J. J.; Sioutas, C.; Katler, M.; Koutrakis, P. (1996) Death from inhalation of concentrated ambient air
4 particles in animal models of pulmonary disease. Am. J. Respir. Crit. Care Med. 153: A15.
5 Gold, A.; Litonius, J.; Schwartz, M.; Verrier, R.; Milstein, A.; Larson, E.; Lovett, B. (1998) Cardiovascular
6 vulnerability to particulate pollution. Am. J. Respir. Crit. Care Med. 157: A261.
7 Gold, D. R.; Damokosh, A. L; Pope, C. A., Ill; Dockery, D. W.; McDonnell, W. F.; Serrano, P.; Retama, A.;
8 Castillejos, M. (1999) Particulate and ozone pollutant effects on the respiratory function of children in
9 southwest Mexico City. Epidemiology 10: 8-16.
10 Gordian, M. E.; Ozkaynak, H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Particulate air pollution and respiratory
11 disease in Anchorage, Alaska. Environ. Health Perspect. 104: 290-297.
12 Goren, A.; Hellmann, S.; Gabbay, Y.; Brenner, S. (1999) Respiratory problems associated with exposure to airborne
13 particles in the community. Arch. Environ. Health 54: 165-171.
14 Gouveia, N.; Fletcher, T. (1998) Some evidence of the 'harvesting effect' of air pollution. Epidemiology 9: S59.
15 Giintzel, O.; Bollag, U.; Helfenstein, U. (1996) Asthma and exacerbation of chronic bronchitis: sentinel and
16 environmental data in a time series analysis. Zentralbl. Hyg. Umweltmed. 198: 383-393.
17 Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (1999) A time-series analysis of acidic PM and daily mortality and
18 morbidity in the Buffalo, NY region. Environ. Health Perspect. submitted.
19 Hajat, S.; Haines, A.; Goubet, S. A.; Atkinson, R. W.; Anderson, H. R. (1999) Association of air pollution with
20 daily GP consultations for asthma and other lower respiratory conditions in London. Thorax 54: 597-605.
21 Haluszka, J.; Pisiewicz, K.; Miczynski, J.; Roemer, W.; Tomalak, W. (1998) Air pollution and respiratory health in
22 children: the PEACE panel study in Krakow, Poland. Eur. Respir. Rev. 8: 94-100.
23 Harre, E. S. M.; Price, P. D.; Ayrey, R. B.; Toop, L. J.; Martin, I. R.; Town, G. I. (1997) Respiratory effects of air
24 pollution in chronic obstructive pulmonary disease: a three month prospective study. Thorax 52: 1040-1044.
25 Harris, R. R. (1975) A primer of multivariate statistics. New York, NY: Academic Press.
26 Hasselblad, V.; Eddy, D. M.; Kotchmar, D. J. (1992) Synthesis of environmental evidence: nitrogen dioxide
27 epidemiology studies. J. Air Waste Manage. Assoc. 42: 662-671.
28 Health Effects Institute. (1999) [Presentation at the Health Effects Institute annual meeting; May]. Health Effects
29 Institute: unpublished.
30 Hedges, L.; Olkin, I. (1985) Statistical methods for meta-analysis. Orlando, FL: Academic Press.
31 Higgins, M. W.; Keller, J. B. (1970) Predictors of mortality in the adult population of Tecumseh. Arch. Environ.
32 Health 21: 418-424.
33 Hill, A. B. (1965) The environment and disease: association or causation? Proc. R. Soc. Med. 58: 295-300.
34 Hiltermann, T. J. N.; de Bruijne, C. R.; Stolk, J.; Zwinderman, A. H.; Spieksma, F. Th. M.; Roemer, W.;
35 Steerenberg, P. A.; Fischer, P. H.; van Bree, L.; Hiemstra, P. S. (1997) Effects of photochemical air pollution
36 and allergen exposure on upper respiratory tract inflammation in asthmatics. Am. J. Respir. Crit.
37 Hiltermann, T. J. N.; Stolk, J.; Van der Zee, S. C.; Brunekreef, B.; De Bruijne, C. R.; Fischer, P. H.; Ameling,
38 C. B.; Sterk, P. J.; Hiemstra, P. S.; Van Bree, L. (1998) Asthma severity and susceptibility to air pollution.
39 Eur. Respir. J. 11:686-693.
40 Hodgkin, J. E.; Abbey, D. E.; Euler, G. L.; Magie, A. R. (1984) COPD prevalence in nonsmokers in high and low
41 photochemical air pollution areas. Chest 86: 830-838.
42 Hoek, G.; Brunekreef, B. (1994) Effects of low-level winter air pollution concentrations on respiratory health of
43 Dutch children. Environ. Res. 64: 136-150.
44 Hoek, G.; Schwartz, J. D.; Groot, B.; Eilers, P. (1997) Effects of ambient particulate matter and ozone on daily
45 mortality in Rotterdam, the Netherlands. Arch. Environ. Health 52: 455-463.
46 Hoek, G.; Dockery, D. W.; Pope, A.; Neas, L.; Roemer, W.; Brunekreef, B. (1998) Association between PM 10 and
47 decrements in peak expiratory flow rates in children: reanalysis of data from five panel studies. Eur. Respir.
48 1.11:1307-1311.
49 Hoek, G.; Verhoeff, A.; Fischer, P.; Van Wijnen, J.; Brunekreef, B. (1998) Daily mortality in the Netherlands and
50 outdoor air pollution. In: Abstracts of the 1998 international conference of the American Thoracic Society;
51 April; Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A879.
52 Ito, K. (1990) An examination of the role of aerosol acidity in historical London, England daily mortality
53 [dissertation]. Syracuse, NY: New York University. Available from: University Microfilms International,
54 Ann Arbor, MI; AAD91-13012.
October 1999 6-220 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ito, K.; Nadas, A.; Burnett, R. T.; Lippmann, M. (1997) Particulate matter components and respiratory mortality in
2 Detroit, MI. In: Abstracts: 1997 International Conference; May; San Francisco, CA. New York, NY:
3 American Thoracic Society. [American journal of respiratory and critical care medicine, v. 155, no. 4, pt. 2,
4 suppl.].
5 Ito, K.; Thurston, G. D.; Nadas, A.; Lippmann, M. (1998) Multiple pollutants' spatio-temporal exposure
6 characterization errors. In: Measurement of toxic and related air pollutants: proceedings of a specialty
7 conference, volume I; September; Gary, NC; pp. 53-63. (A&WMA publication VIP-85).
8 Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
9 1983-1992. Environ. Health Perspect. 105: 980-985.
10 Jamason, P. F.; Kalkstein, L. S.; Gergen, P. J. (1997) A synoptic evaluation of asthma hospital admissions in
11 New York City. Am. J. Respir. Crit. Care Med. 156: 1781-1788.
12 Jammes, Y.; Delpierre, S.; Delvolgo, M. J.; Humbert-Tena, C.; Burnet, H. (1998) Long-term exposure of adults to
13 outdoor air pollution is associated with increased airway obstruction and higher prevalence of bronchial
14 hyperresponsiveness. Arch. Environ. Health 53: 372-377.
15 Janssen, N. A. H.; Hoek, G.; Harssema, H.; Brunekreef, B. (1997) Childhood exposure to PM10: relation between
16 personal, classroom, and outdoor concentrations. Occup. Environ. Med. 54: 888-894.
17 Kalandidi, A.; Gratziou, C.; Katsouyanni, K.; Manalis, N.; Tzala, L.; Pantazopoulou, A.; Efthimiou, M.;
18 Roussos, C.; Roemer, W. (1998) Air pollution and respiratory health of children: the PEACE panel study in
19 Athens, Greece. Eur. Respir. Rev. 8: 117-124.
20 Katsouyanni, K.; Touloumi, G. (1998) Causes of regional differences in air pollution effects are being studied
21 further [letter]. Br. Med. J. 316: 1982.
22 Katsouyanni, K.; Schwartz, J.; Spix, C.; Touloumi, G.; Zmirou, D.; Zanobetti, A.; Wojtyniak, B.; Vonk, J. M.;
23 Tobias, A.; Ponka, A.; Medina, S.; Bacharova, L.; Andersen, H. R. (1996) Short term effects of air pollution
24 on health: a European approach using epidemiology time series data: the APHEA protocol. In: St Leger, S.,
25 ed. The APHEA project. Short term effects of air pollution on health: a European approach using
26 epidemiological time series data. J. Epidemiol. Commun. Health 50(suppl. 1): S12-S18.
27 Katsouyanni, K.; Touloumi, G.; Spix, C.; Schwartz, J.; Balducci, F.; Medina, S.; Rossi, G.; Wojtyniak, B.;
28 Sunyer, J.; Bacharova, L.; Schouten, J. P.; Ponka, A.; Anderson, H. R. (1997) Short term effects of ambient
29 sulphur dioxide and particulate matter on mortality in 12 European cities: results from time series data from
30 the APHEA project. Br. Med. J. 314: 1658-1663.
31 Kelsall, J. E.; Samet, J. M.; Zeger, S. L.; Xu, J. (1997) Air pollution and mortality in Philadelphia, 1974-1988.
32 Am. J. Epidemiol. 146: 750-762.
33 Kinney, P. L.; Ito, K; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM 10 associations in Los Angeles.
34 In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on particulate air pollution and human
35 mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol. 7: 59-69.
36 Korrick, S. A.; Neas, L. M.; Dockery, D. W.; Gold, D. R.; Allen, G. A.; Hill, L. B.; Kimball, K. D.; Rosner, B. A.;
37 Speizer, F. E. (1998) Effects of ozone and other pollutants on the pulmonary function of adult hikers.
38 Environ. Health Perspect. 106: 93-99.
39 Kotesovec, F.; Vitnerova, N.; Leixner, M.; Benes, I.; Skorkovsky, J.; Roemer, W. (1998) Air pollution and
40 respiratory health of children: the PEACE panel study in Terplice, Czech Republic. Eur. Respir. Rev.
41 8:70-77.
42 Krieger, N. (1994) Epidemiology and the web of causation: has anyone seen the spider? Soc. Sci. Med.
43 39:887-903.
44 Kiinzli, N.; Tager, I. B. (1997) The semi-individual study in air pollution epidemiology: a valid design as compared
45 to ecologic studies. Environ. Health Perspect. 105: 1078-1083.
46 Laden, F.; Neas, L. M.; Schwartz, J. (1999) The association of crystal particles in the fine particulate fraction with
47 daily mortality in six cities. Am. J. Respir. Crit. Care Med. 159: A332.
48 Lamm, S. H.; Hall, T. A.; Engel, A.; Rueter, F. H.; White, L. D. (1994) PM 10 particulates: are they the major
49 determinant of pediatric respiratory admissions in Utah County, Utah (1985-1989). In: Dodgson, J.;
50 McCallum, R. I., eds. Inhaled particles VII: proceedings of an international symposium on inhaled particles
51 organized by the British Occupational Hygiene Society; September 1991; Edinburgh, United Kingdom. Ann.
52 Occup. Hyg. 38(suppl. 1): 969-972.
53 Lave, L. B.; Seskin, E. P. (1970) Air pollution and human health: the quantitative effect, with an estimate of the
54 dollar benefit of pollution abatement, is considered. Science (Washington, DC) 169: 723-733.
October 1999 6-221 DRAFT-DO NOT QUOTE OR CITE
-------
1 Lave, L. B.; Seskin, E. P. (1977) Air pollution and human health. Baltimore, MD: The Johns Hopkins University
2 Press.
3 Leaderer, B. P.; Naeher, L.; Jankun, T.; Balenger, K.; Holford, T. R.; Toth, C.; Sullivan, J.; Wolfson, J. M.;
4 Koutrakis, P. (1999) Indoor, outdoor, and regional summer and winter concentrations of PM 10, PM25, SO42",
5 H+, NH4+, NO3", NH3, and nitrous acid in homes with and without kerosene space heaters. Environ. Health
6 Perspect. 107:223-231.
7 Lebowitz, M. D.; Collins, L.; Holberg, C. J. (1987) Time series analyses of respiratory responses to indoor and
8 outdoor environmental phenomena. Environ. Res. 43: 332-341.
9 Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
10 Perspect. 107: 149-154.
11 LeLorier, J.; Gregoire, G.; Benhaddad, A.; Lapierre, J.; Derderian, F. (1997) Discrepancies between meta-analyses
12 and subsequent large randomized, controlled trials. N. Engl. J. Med. 337: 536-542.
13 Levy, D. (1998) Fine particulate air pollution and out-of-hospital mortality in King County, Washington. In: Health
14 effects of particulate matter in ambient air: proceedings of the international symposium on the health effects
15 of particulate matter; Prague, Czechoslovakia. Pittsburgh, PA: Air & Waste Management Association,
16 VIP-80; pp. 262-271.
17 Lewis, P. R.; Hensley, M. J.; Wlodarczyk, J.; Toneguzzi, R. C.; Westley-Wise, V. J.; Dunn, T.; Calvert, D. (1998)
18 Outdoor air pollution and children's respiratory symptoms in the steel cities of New South Wales. Med. J.
19 Aust. 169: 459-463.
20 Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of particulate air
21 pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
22 Lin, C. A.; Martins, M. A.; Farhat, S. C. L.; Pope, C. A., Ill; Conceisao, G. M. S.; Anastacio, V. M.; Hatanaka, M.;
23 Andrade, W. C.; Hamaue, W. R.; Bohm, G. M.; Saldiva, P. H. N. (1999) Air pollution and respiratory illness
24 of children in Sao Paulo, Brazil.
25 Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D.; Gong, H., Jr. (1996)
26 Short-term air pollution exposures and responses in Los Angeles area schoolchildren. J. Exposure Anal.
27 Environ. Epidemiol. 6: 449-472.
28 Linn, W. S.; Gong, H., Jr.; Clark, K. W.; Anderson, K. R. (1998) Day-to-day particuate exposures and health
29 changes in Los Angeles area residents with severe lung disease. In: Chow, J.; Koutrakis, P., eds. PM 2 5: a fine
30 particle standard: proceedings of an international specialty conference; January; Long Beach, CA., v. 2.
31 Pittsburgh, PA: Air & Waste Management Association; pp. 851-861. (AWMA specialty conference;
32 VIP-81).
33 Lipfert, F. W. (1978) The association of human mortality with air pollution: statistical analyses by region, by age,
34 and by cause of death. Mantua, NJ: Eureka Publications.
35 Lipfert, F. W. (1984) Air pollution and mortality: specification searches using SMSA-based data. J. Environ. Econ.
36 Manage. 11:208-243.
37 Lipfert, F. W. (1988) Exposure to acidic sulfates in the atmosphere: review and assessment. Palo Alto, CA:
3 8 Electric Power Research Institute; final report no. EPRI EA-6150.
39 Lipfert (1997).
40 Lipfert, F. W.; Wyzga, R. E. (1995) Air pollution and mortality: issues and uncertainties. J. Air Waste Manage.
41 Assoc. 45: 949-966.
42 Lipfert and Wyzga (1996).
43 Lipfert, F. W.; Wyzga, R. E. (1997) Air pollution and mortality: the implications of uncertainties in regression
44 modeling and exposure measurement. J. Air Waste Manage. Assoc. 47: 517-523.
45 Lipfert, F. W.; Wyzga, R. E. (1999) Statistical considerations in determining the health significance of constituents
46 of airborne particulate matter. J. Air Waste Manage. Assoc.: accepted.
47 Lippmann, M. (1989) Background on health effects of acid aerosols. In: Symposium on the health effects of acid
48 aerosols; October 1987; Research Triangle Park, NC. Environ. Health Perspect. 79: 3-6.
49 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
50 California. Environ. Health Perspect. 105: 216-222.
51 Long, W.; Tate, R. B.; Neuman, M.; Manfreda, J.; Becker, A. B.; Anthonisen, N. R. (1998) Respiratory symptoms
52 in a susceptible population due to burning of agricultural residue. Chest 113: 351-357.
53 Loomis, D. P.; Borja-Aburto, V. H.; Bangdiwala, S. I.; Shy, C. M. (1996) Ozone exposure and daily mortality in
54 Mexico City: a time-series analysis. Cambridge, MA: Health Effects Institute; research report no. 75.
October 1999 6-222 DRAFT-DO NOT QUOTE OR CITE
-------
1 Loomis, D.; Castilejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
2 mortality in Mexico City. Epidemiology 10: 118-123.
3 Lumley, T.; Heagerty, P. (1999) Weighted empirical adaptive variance estimators for correlated data regression.
4 J. Royal Stat. Soc. Series B 61 (part 2): 459-477.
5 Lyon, J. L.; Mori, M.; Gao, R. (1995) Is there a causal association between excess mortality and exposure to PM 10
6 air pollution? Additional analyses by location, year, season and cause of death. In: Phalen, R. F.; Bates,
7 D. V., eds. Proceedings of the colloquium on particulate air pollution and human mortality and morbidity,
8 part II; January 1994; Irvine, CA. Inhalation Toxicol. 7: 603-614.
9 Lyon etal. (1996).
10 Mage, D. T.; Buckley, T. J. (1995) The relationship between personal exposures and ambient concentrations of
11 particulate matter. Presented at: 88th annual meeting of the Air & Waste Management Association; June;
12 San Antonio, TX. Pittsburgh, PA: Air & Waste Management Association; paper no. 95-MP18.01.
13 Mage, D. T.; Wilson, W.; Hasselblad, V.; Grant, L. (1999) Assessment of human exposure to ambient particulate
14 matter. J. Air Waste Manage. Assoc.: in press.
15 Marcus, A. H.; Chapman, R. (1998) Estimating the health effects of fine particles from epidemiology studies: how
16 serious are problems of measurement error, correlation, and confounding? In: Chow, J.; Koutrakis, P., eds.
17 PM2.5: a fine particle standard. Volume II: proceedings of an international specialty conference.; January;
18 Long Beach, CA. Pittsburgh, PA: Air & Waste Management Association; pp. 899-919.
19 Mathsoft, Inc. (1998) S-plus, getting started with S-plus 4.5. Seattle, WA: Mathsoft, Inc., Data Analysis Products
20 Division.
21 McDonnell, W. F.; Abbey, D. E.; Nishino, N.; Lebowitz, M. D. (1999) Long-term ambient ozone concentration and
22 the incidence of asthma in nonsmoking adults: the ahsmog study. Environ. Res. 80: 110-121.
23 McMichael, A. J.; Anderson, H. R.; Brunekreef, B.; Cohen, A. J. (1998) Inappropriate use of daily mortality
24 analyses to estimate longer-term mortality effects of air pollution. Int. J. Epidemiol. 27: 450-453.
25 Medina, S.; Le Tertre, A.; Quenel, P.; Le Moullec, Y.; Lameloise, P.; Guzzo, J. C.; Festy, B.; Ferry, R.; Dab, W.
26 (1997) Air pollution and doctors' house calls: results from the ERPURS system for monitoring the effects of
27 air pollution on public health in greater Paris, France, 1991-1995. Environ. Res. 75: 73-84.
28 Michaels etal. (1998).
29 Michaels etal. (1999).
30 Michelozzi, P.; Forastiere, F.; Fusco, D.; Perucci, C. A.; Ostro, B.; Ancona, C.; Pallotti, G. (1998) Air pollution and
31 daily mortality in Rome, Italy. Occup. Environ. Med. 55: 605-610.
32 Mills, P. K.; Abbey, D.; Beeson, W. L.; Petersen, F. (1991) Ambient air pollution and cancer in California
33 Seventh-day Adventists. Arch. Environ. Health 46: 271-280.
34 Minshew, P.; Towle, J. (1999) The 1998 wildfires in central Florida: Volusia County's own armageddon. J. Environ.
35 Health 61: 22-26.
36 Miyao, M.; Furuta, M.; Ozawa, K.; Kondo, T.-A.; Sakakibara, H.; Ishihara, S.; Yamanaka, K. (1993) Morbidity of
37 allergic rhinitis based on the National Health Insurance records of Japan. Tohoku J. Exp. Med. 169: 345-350.
38 Moolgavkar, S. H.; Luebeck, E. G. (1996) A critical review of the evidence on particulate air pollution and
39 mortality. Epidemiology 7: 420-428.
40 Moolgavkar, S. H.; Luebeck, E. G.; Anderson, E. L. (1997) Air pollution and hospital admissions for respiratory
41 causes in Minneapolis-St. Paul and Birmingham. Epidemiology 8: 364-370.
42 Morgan, G.; Corbett, S.; Wlodarczyk, J. (1998a) Air pollution and hospital admissions in Sydney, Australia, 1990 to
43 1994. Am. J. Public Health 88: 1761-1766.
44 Morgan, G.; Corbett, S.; Wlodarczyk, J.; Lewis, P. (1998b) Air pollution and daily mortality in Sydney, Australia,
45 1989 through 1993. Am. J. Public Health 88: 759-764.
46 Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
47 evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
48 Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
49 heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
50 Mulla, Z. D.; Sorensen, B.; Fuss, M. (1998) Active surveillance of morbidity during the wildfires in central Florida.
51 Fla. J. Environ. Health 163: 14-17.
52 Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
53 Healthy women's PEF variations with ambient summer concentrations of PM 10, PN2 5, SO42", H+, and O3.
54 Am. J. Respir. Crit. Care Med. 160: 117-125.
October 1999 6-223 DRAFT-DO NOT QUOTE OR CITE
-------
1 National Institutes of Health. (1991) Guidelines for the diagnosis and management of asthma. Bethesda, MD: U.S.
2 Department of Health and Human Services, National Heart, Lung, and Blood Institute, National Asthma
3 Education Program; publication no. 91-3042.
4 National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
5 Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
6 publication no. 97-4051.
7 National Research Council. (1980) Trace-element geochemistry of coal resource development related to
8 environmental quality and health. Washington, DC: National Academy Press.
9 Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
10 pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
11 141:111-122.
12 Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and
13 other determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
14 Neas, L. M.; Schwartz, J.; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
15 Philadelphia. Environ. Health Perspect. 107: 629-631.
16 Neuhaus, J. M.; Kalbfleisch, J. D. (1998) Between- and within-cluster covariate effects in the analysis of clustered
17 data. Biometrics 54: 638-645.
18 Neukirch, F.; Segala, C.; Le Moullec, Y.; Korobaeff, M.; Aubier, M. (1998) Short-term effects of low-level winter
19 pollution on respiratory health of asthmatic adults. Arch. Environ. Health 53: 320-328.
20 Nishino, N.; Abbey, D. E.; Burchette, R. J. (1997) Long-term ambient concentrations of PM10 and mortality in
21 nonsmokers—the AHSMOG study. In: The ninth annual conference of the International Society for
22 Environmental Epidemiology; August; Taipei, Taiwan. Epidemiology 8: S58
23 Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Shappard, L.; Stout, J. W. (1999) An association
24 between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
25 Perspect. 107: 489-493.
26 Ostro (1996).
27 Ostro, B.; Chestnut, L. (1998) Assessing the health benefits of reducing particulate matter air pollution in the
28 United States. Environ. Res. 76: 94-106.
29 Ostro, B.; Lipsett, M. (1998) Air pollution and mortality in the Coachella Valley, California: an examination of
30 effects of PM10 dominated by coarse particles. In: Abstracts of the 1998 international conference of the
31 American Thoracic Society; April; Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A260.
32 Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
33 Am. J. Public Health 81: 694-702.
34 Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
35 exacerbations among African-American children in Los Angeles. In: Phalen, R. F.; Bates, D. V., eds.
36 Proceedings of the colloquium on particulate air pollution and human mortality and morbidity, part II;
37 January 1994; Irvine, CA. Inhalation Toxicol. 7: 711-722.
38 Ostro, B.; Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1998) The impact of fine particulate matter in Bangkok,
39 Thailand. In: Chow, J.; Koutrakis, P., eds. PM2.5: a fine particle standard: proceedings of an international
40 specialty conference; January; Long Beach, CA., v. 2. Pittsburgh, PA: Air & Waste Management
41 Association; pp. 939-949. (AWMA specialty conference VIP-81).
42 Ostro, B. D.; Eskeland, G. S.; Sanchez, J. M.; Feyzioglu, T. (1999) Air pollution and health effects: a study of
43 medical visits among children in Santiago, Chile. Environ. Health Perspect. 107: 69-73.
44 Ozkaynak, H.; Thurston, G. D. (1987) Associations between 1980 U.S. mortality rates and alternative measures of
45 airborne particle concentration. Risk Anal. 7: 449-461.
46 Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
47 pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
48 Environmental Health; March 25.
49 Pantazopoulou, A.; Katsouyanni, K.; Kourea-Kremastinou, J.; Trichopoulos, D. (1995) Short-term effects of air
50 pollution on hospital emergency outpatient visits and admissions in the greater Athens, Greece area.
51 Environ. Res. 69:31-36.
52 Pearl, J. (1995) Causal diagrams for empirical research. Biometrika 82: 669-688.
53 Pekkanen, J.; Timonen, K. L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles
54 in urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74: 24-33.
October 1999 6-224 DRAFT-DO NOT QUOTE OR CITE
-------
1 Pereira, L. A. A.; Loomis, D.; Concei5ao, G. M. S.; Braga, A. L. F.; Areas, R. M.; Kishi, H. S.; Singer, J. M.;
2 Bohm, G. M.; Saldiva, P. H. N. (1998) Association between air pollution and intrauterine mortality in
3 Sao Paulo, Brazil. Environ. Health Perspect. 106: 325-329.
4 Peters, A.; Goldstein, I. F.; Beyer, U.; Franke, K.; Heinrich, J.; Dockery, D. W.; Spengler, J. D.; Wichmann, H.-E.
5 (1996) Acute health effects of exposure to high levels of air pollution in eastern Europe. Am. J. Epidemiol.
6 144: 570-581.
7 Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997a) Short-term effects of particulate air pollution on
8 respiratory morbidity in asthmatic children. Europ. Respir. J. 10: 872-879.
9 Peters, A.; Dockery, D. W.; Heinrich, J.; Wichmann, H. E. (1997b) Medication use modifies the health effects of
10 particulate sulfate air pollution in children with asthma. Environ. Health Perspect. 105: 430-435.
11 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997c) Respiratory effects are associated with the
12 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
13 Peters, A.; Doring, A.; Wichmann, H.-E.; Koenig, W. (1997d) Increased plasma viscosity during an air pollution
14 episode: a link to mortality? Lancet 349: 1582-1587.
15 Peters etal. (1998).
16 Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
17 Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern Californic communities with
18 differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit. Care
19 Med. 159: 760-767.
20 Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
21 Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
22 differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
23 159:768-775.
24 Ponce de Leon, A.; Anderson, H. R.; Bland, J. M.; Strachan, D. P.; Bower, J. (1996) Effects of air pollution on
25 daily hospital admissions for respiratory disease in London between 1987-88 and 1991-92. In: St Leger, S.,
26 ed. The APHEA project. Short term effects of air pollution on health: a European approach using
27 epidemiological time series data. J. Epidemiol. Commun. Health 50(suppl. 1): S63-S70.
28 Ponka, A.; Savela, M.; Virtanen, M. (1998) Mortality and air pollution in Helsinki. Arch. Environ. Health
29 53:281-286.
30 Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM 10 pollution in Utah, Salt Lake, and
31 Cache Valleys. Arch. Environ. Health 46: 90-97.
32 Pope, C. A., III. (1996) Particulate pollution and health: a review of the Utah Valley experience. J. Exposure Anal.
33 Environ. Epidemiol. 6: 23-34.
34 Pope, C. A., Ill; Kalkstein, L S. (1996) Synoptic weather modeling and estimates of the exposure-response
35 relationship between daily mortality and particulate air pollution. Environ. Health Perspect. 104: 414-420.
36 Pope, C. A., Ill; Dockery, D. W.; Spengler, J. D.; Raizenne, M. E. (1991) Respiratory health and PM 10 pollution:
37 a daily time series analysis. Am. Rev. Respir. Dis. 144: 668-674.
38 Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM 10 pollution in Utah valley. Arch.
39 Environ. Health 47: 211-217.
40 Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
41 (1995) Particulate air pollution as a predictor of mortality in a prospective study of U. S. adults. Am. J.
42 Respir. Crit. Care Med. 151: 669-674.
43 Pope, C. A., Ill; Dockery, D. W.; Kanner, R. E.; Vollegas, G. M.; Schwartz, J. (1999a) Oxygen saturation, pulse
44 rate, and particulate air pollution: a daily time-series panel study. Am. J. Respir. Crit. Care Med.
45 159: 365-372.
46 Pope, C. A.; Hill, R. W.; Villegas, G. M. (1999b) Particulate air pollution and daily mortality on Utah's Wasatch
47 Front. Environ. Health Perspect.: submitted.
48 Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
49 cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
50 Rahlenbeck, S. I.; Kahl, H. (1996) Air pollution and mortality in East Berlin during the winters of 1981-1989. Int. J.
51 Epidemiol. 25: 1220-1226.
52 Raizenne, M.; Neas, L. M.; Damokosh, A. L; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer,
53 F. E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ.
54 Health Perspect. 104: 506-514.
October 1999 6-225 DRAFT-DO NOT QUOTE OR CITE
-------
1 Raizenne, M.; Dales, R.; Burnett, R. (1998) Air pollution exposures and children's health. Can. J. Public Health
2 May-June: S43-S48.
3 Ritz, B.; Yu, F. (1999) The effect of ambient carbon monoxide on low birth weight among children born in southern
4 California between 1989 and 1993. Environ. Health Perspect. 107: 17-25.
5 Roemer, W.; Hoek, G.; Brunekreef, B. (1993) Effect of ambient winter air pollution on respiratory health of
6 children with chronic respiratory symptoms. Am. Rev. Respir. Dis. 147: 118-124.
7 Roemer, W.; Hoek, G.; Brunekreef, B.; Haluszka, J.; Kalandidi, A.; Pekkanen, J. (1998) Daily variations in air
8 pollution and respiratory health in a multicentre study: the PEACE project. Eur. Respir. J. 12: 1354-1361.
9 Roemer, W.; Clench-Aas, J.; Englert, N.; Hoek, G.; Katsouyanni, K.; Pekkanen, J.; Brunekreef, B. (1999)
10 Inhomogeneity in response to air pollution in European children (PEACE project). Occup. Environ. Med.
11 56:86-92.
12 Romieu(1998).
13 Romieu, I.; Meneses, F.; Ruiz, S.; Sienra, J. J.; Huerta, J.; White, M. C.; Etzel, R. A. (1996) Effects of air pollution
14 on the respiratory health of asthmatic children living in Mexico City. Am. J. Respir. Crit. Care Med.
15 154: 300-307.
16 Romieu, I.; Meneses, F.; Ruiz, S.; Huerta, J.; Sienra, J. J.; White, M.; Etzel, R.; Hernandez, M. (1997) Effects of
17 intermittent ozone exposure on peak expiratory flow and respiratory symptoms among asthmatic children in
18 Mexico City. Arch. Environ. Health 52: 368-376.
19 Rooney, C.; McMichael, A. J.; Kovats, R. S.; Coleman, M. P. (1998) Excess mortality in England and Wales, and in
20 greater London, during the 1995 heatwave. J. Epidemiol. Commun. Health 52: 482-486.
21 Rosas, I.; McCartney, H. A.; Payne, R. W.; Calderon, C.; Lacey, J.; Chapela, R.; Ruiz-Velazco, S. (1998) Analysis
22 of the relationships between environmental factors (aeroallergens, air pollution, and weather) and asthma
23 emergency admissions to a hospital in Mexico City. Allergy 53:3 94-401.
24 Rossi, G.; Vigotti, M. A.; Zanobetti, A.; Repetto, F.; Gianelle, V.; Schwartz, J. (1999) Air pollution and
25 cause-specific mortality in Milan, Italy, 1980-1989. Arch. Environ. Health 54: 158-164.
26 Roth, H. D.; et al. (1998) A comparison of the association between air pollution and mortality or hospital
27 admissions in Prague, Czech Republic with the effects of particulate matter in Birmingham and Philadelphia,
28 USA. In: Health effects of particulate matter in ambient air: proceedings of the international symposium on
29 the health effects of particulate matter; Prague, Czechoslovakia. Pittsburgh, PA: Air & Waste Management
30 Association, VIP-80; pp. 248-256.
31 Rothman, K. J.; Greenland, S., eds. (1998) Modern epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven
32 Publishers.
33 Salvietal. (1997).
34 Samet, J. M.; Zeger, S. L.; Berhane, K. (1995) The association of mortality and particulate air pollution.
35 In: Particulate air pollution and daily mortality: replication and validation of selected studies, the phase I
36 report of the particle epidemiology evaluation project [preprint]. Cambridge, MA: Health Effects Institute;
37 pp. 1-104.
38 Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J. (1996) Air pollution and mortality in Philadelphia, 1974-1988,
39 report to the Health Effects Institute on phase IB: Particle Epidemiology Evaluation Project. Cambridge, MA:
40 Health Effects Institute; accepted.
41 Samet, J.; Zeger, S.; Kelsall, J.; Xu, J.; Kalkstein, L. (1998) Does weather confound of modify the association of
42 particulate air pollution with mortality? An analysis of the Philadelphia data, 1973-1980. Environ. Res.
43 77:9-19.
44 Scarlett, J. F.; Abbott, K. J.; Peacock, J. L.; Strachan, D. P.; Anderson, H. R. (1996) Acute effects of summer air
45 pollution on respiratory function in primary school children in southern England. Thorax 51: 1109-1114.
46 Schmitzberger, R.; Rhomberg, K.; Biichele, H.; Puchegger, R.; Schmitzberger-Natzmer, D.; Kemmler, G.;
47 Panosch, B. (1993) Effects of air pollution on the respiratory tract of children. Pediatr. Pulmonol. 15: 68-74.
48 Schouten, J. P.; Vonk, J. M.; de Graaf, A. (1996) Short term effects of air pollution on emergency hospital
49 admissions for respiratory disease: results of the APHEA project in two major cities in The Netherlands,
50 1977-89. In: St Leger, S., ed. The APHEA project. Short term effects of air pollution on health: a European
51 approach using epidemiological time series data. J. Epidemiol. Commun. Health 50(suppl. 1): S22-S29.
52 Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137: 1136-1147.
53 Schwartz, J. (1994a) PM10, ozone, and hospital admissions for the elderly in Minneapolis, MN. Arch. Environ.
54 Health 49: 366-374.
October 1999 6-226 DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz, J. (1994b) Air pollution and hospital admissions for the elderly in Detroit, Michigan. Am. J. Respir. Crit.
2 Care Med. 150: 648-655.
3 Schwartz, J. (1994c) What are people dying of on high air pollution days? Environ. Res. 64: 26-35.
4 Schwartz, J. (1994d) Air pollution and daily mortality: a review and meta analysis. Environ. Res. 64: 36-52.
5 Schwartz, J. (1994e) Nonparametric smoothing in the analysis of air pollution and respiratory illness. Can. J. Stat.
6 22: 1-17.
7 Schwartz, J. (1995) Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory
8 disease. Thorax 50: 531-538.
9 Schwartz, J. (1996) Air pollution and hospital admissions for respiratory disease. Epidemiology 7: 20-28.
10 Schwartz, J. (1997) Health effects of air pollution from traffic: ozone and particulate matter. In: Fletcher, T.;
11 McMichael, A. J., eds. Health at the crossroads: transport policy and urban health. [Proceedings of the]
12 London School of Hygiene & Tropical Medicine fifth annual public health forum; April, 1995; London,
13 United Kingdom. Chichester, United Kingdom: John Wiley & Sons, Ltd.; pp. 61-85.
14 Schwartz, J. (1998) Harvesting and long term exposure effects in the relationship between air pollution and
15 mortality. In: Abstracts of the 1998 international conference of the American Thoracic Society; April;
16 Chicago, IL. Am. J. Respir. Crit. Care Med. 157: A879.
17 Schwartz, J. (1999a) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
18 10: 17-22.
19 Schwartz, J. (1999b) Harvesting and long term exposure effects in the relationship between air pollution and
20 mortality. Am. J. Epidemiol.: in press.
21 Schwartz, J.; Marcus, A. (1990) Mortality and air pollution in London: atime series analysis. Am. J. Epidemiol.
22 131: 185-194.
23 Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
24 Michigan. Am. J. Epidemiol. 142: 23-35.
25 Schwartz, J.; Slater, D.; Larson, T. V.; Pierson, W. E.; Koenig, J. Q. (1993) Particulate air pollution and hospital
26 emergency room visits for asthma in Seattle. Am. Rev. Respir. Dis. 147: 826-831.
27 Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
28 Ferris, B. G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
29 Am. J. Respir. Crit. Care Med. 150: 1234-1242.
30 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles?
31 J. Air Waste Manage. Assoc. 46: 927-939.
32 Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
33 concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
34 Seaton, A. (1996) Particles in the air: the enigma of urban air pollution. J. R. Soc. Med. 89: 604-607.
35 Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Particulate air pollution and acute health effects.
36 Lancet (8943): 176-178.
37 Segala, C.; Fauroux, B.; Just, J.; Pascual, L.; Grimfeld, A.; Neukirch, F. (1998) Short-term effect of winter air
38 pollution on respiratory health of asthmatic children in Paris. Eur. Respir. J. 11: 677-685.
39 Seneff, M. G.; Wagner, D. P.; Wagner, R. P.; Zimmerman, J. E.; Knaus, W. A. (1995) Hospital and 1-year survival
40 of patients admitted to intensive care units with acute exacerbation of chronic obstructive pulmonary disease.
41 JAMA J. Am. Med. Assoc. 274: 1852-1857.
42 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
43 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
44 Shumway, R. H.; Tai, R. Y.; Tai, L. P.; Pawitan, Y. (1983) Statistical analysis of daily London mortality and
45 associated weather and pollution effects. Sacramento, CA: California Air Resources Board; contract no.
46 Al-154-33.
47 Shumway, R. H.; Azari, A. S.; Pawitan, Y. (1988) Modeling mortality fluctuations in Los Angeles as functions of
48 pollution and weather effects. Environ. Res. 45: 224-241.
49 Simpson, R. W.; Williams, G.; Petroeschevsky, A.; Morgan, G.; Rutherford, S. (1997) Associations between
50 outdoor air pollution and daily mortality in Brisbane, Australia. Arch. Environ. Health 52: 442-454.
51 Spektor, D. M.; Hofmeister, V. A.; Artaxo, P.; Brague, J. A. P.; Echelar, F.; Nogueira, D. P.; Hayes, C.; Thurston,
52 G. D.; Lippmann, M. (1991) Effects of heavy industrial pollution on respiratory function in the children of
53 Cubatao, Brazil: a preliminary report. Environ. Health Perspect. 94: 51-54.
October 1999 6-227 DRAFT-DO NOT QUOTE OR CITE
-------
1 Spix, C.; Heinrich, J.; Dockery, D.; Schwartz, J.; Volksch, G.; Schwinkowski, K.; Collen, C.; Wichmann, H. E.
2 (1993) Air pollution and daily mortality in Erfurt, East Germany, 1980-1989. Environ. Health Perspect. 101: 518-526.
3 Spix, C.; Anderson, H. R.; Schwartz, J.; Vigotti, M. A.; LeTertre, A.; Vonk, J. M.; Touloumi, G.; Balducci, F.;
4 Piekarski, T.; Bacharova, L.; Tobias, A.; Ponka, A.; Katsouyanni, K. (1998) Short-term effects of air
5 pollution on hospital admissions of respiratory diseases in Europe: a quantitative summary of APHEA study
6 results. Arch. Environ. Health 53: 54-64.
7 Stern, B. R.; Raizenne, M. E.; Burnett, R. T.; Jones, L.; Kearney, J.; Franklin, C. A. (1994) Air pollution and
8 childhood respiratory health: exposure to sulfate and ozone in 10 Canadian rural communities. Environ. Res.
9 66: 125-142.
10 Stieb, D. M.; Burnett, R. T.; Beveridge, R. C.; Brook, J. R. (1996) Association between ozone and asthma
11 emergency department visits in Saint John, New Brunswick, Canada. Environ. Health Perspect.
12 104: 1354-1360.
13 Stieb, D. M.; Beveridge, R. C.; Rowe, B. H.; Walter, S. D.; Judek, S. (1998a) Assessing diagnostic classification in
14 an emergency department: implications for daily time series studies of air pollution. Am. J. Epidemiol.
15 148: 666-670.
16 Stieb, D. M.; Brook, J. R.; Broder, L; Judek, S.; Burnett, R. T.; Beveridge, R. C. (1998b) Personal exposure of
17 adults with cardiorespiratory disease to particulate acid and sulfate in Saint John, New Brunswick, Canada.
18 Appl. Occup. Environ. Hyg. 13: 461-468.
19 Strachan, D. P. (1992) Ventilatory function, height, and mortality among lifelong non-smokers. J. Epidemiol.
20 Community Health 46: 66-70.
21 Studnicka, M. J.; Frischer, T.; Meinert, R.; Studnicka-Benke, A.; Hajek, K.; Spengler, J. D.; Neumann, M. G.
22 (1995) Acidic particles and lung function in children: a summer camp study in the Austrian Alps. Am. J.
23 Respir. Crit. Care Med. 151: 423-430.
24 Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne particulate matter on daily
25 death counts. Environ. Health Perspect. 103: 490-497.
26 Sunyer, J.; Spix, C.; Quenel, P.; Ponce-de-Leon, A.; Ponka, A.; Barumandzadeh, T.; Touloumi, G.; Bacharova, L.;
27 Wojtyniak, B.; Vonk, J.; Bisanti, L.; Schwartz, J.; Katsouyanni, K. (1997) Urban air pollution and emergency
28 admissions for asthma in four European cities: the APHEA project. Thorax 52: 760-765.
29 Susser, M. (1973) Causal thinking in the health sciences: concepts and strategies of epidemiology. New York, NY:
3 0 Oxford University Press.
31 Susser, M. (1986) The logic of Sir Karl Popper and the practice of epidemiology. Am. J. Epidemiol. 124: 711-718.
32 Susser, M. (1991) What is a cause and how do we know one? A grammar for pragmatic epidemiology. Am. J.
33 Epidemiol. 133: 635-648.
34 Tager, I. B. (1999) Air pollution and lung function growth. Is it ozone? [editorial]. Am. J. Respir. Crit. Care Med.
35 160: 387-389.
36 Taggart, S. C. O.; Custovic, A.; Francis, H. C.; Faragher, E. B.; Yates, C. J.; Higgins, B. G.; Woodcock, A. (1996)
37 Asthmatic bronchial hyperresponsiveness varies with ambient levels of summertime air pollution. Eur.
38 Respir. J. 9: 1146-1154.
39 Tamura, K; Ando, M.; Sagai, M.; Matsumoto, Y. (1996) Personal exposure assessment of suspended particulate
40 matter and nitrogen dioxide in Tokyo. Onogawa, Tsukuba, Japan: National Institute for Environmental
41 Studies, Environmental Health Sciences Division.
42 Tenias, J. M.; Ballester, F.; Rivera, M. L. (1998) Association between hospital emergency visits for asthma and air
43 pollution in Valencia, Spain. Occup. Environ. Med. 55: 541-547.
44 Thomas (1988).
45 Thomas etal. (1992).
46 Thurston, G. D.; Ito, K.; Kinney, P. L.; Lippmann, M. (1992) A multi-year study of air pollution and respiratory
47 hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers.
48 J. Exposure Anal. Environ. Epidemiol. 2: 429-450.
49 Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
50 summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
51 65:271-290.
52 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
53 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
54 Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine particulate air pollution,
55 resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
October 1999 6-228 DRAFT-DO NOT QUOTE OR CITE
-------
1 Timonen, K. L.; Pekkanen, J. (1997) Air pollution and respiratory health among children with asthmatic or cough
2 symptoms. Am. J. Respir. Crit. Care Med. 156: 546-552.
3 Tseng, R. Y. M.; Li, C. K.; Spinks, J. A. (1992) Particulate air pollution and hospitalization for asthma. Ann.
4 Allergy 68: 425-432.
5 Touloumi, G.; Katsouyanni, K.; Zmirou, D.; Schwartz, J.; Spix, C.; Ponce de Leon, A.; Tobias, A.; Quennel, P.;
6 Rabczenko, D.; Bacharova, L.; Bisanti, L.; Vonk, J. M.; Ponka, A. (1997) Short-term effects of ambient
7 oxidant exposure on mortality: a combined analysis within the APHEA project. Am. J. Epidemiol.
8 146: 177-185.
9 U.S. Department of Health and Human Services. (1990) The health benefits of smoking cessation: a report of the
10 surgeon general. Rockville, MD: U.S. Public Health Service, U.S. Centers for Disease Control, Office on
11 Smoking and Health; DHHS publication no. (CDC) 90-8416.
12 U.S. Environmental Protection Agency. (1986) Second addendum to air quality criteria for particulate matter and
13 sulfur oxides (1982): assessment of newly available health effects information. Research Triangle Park, NC:
14 Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; EPA report
15 no. EPA-600/8-86-020F. Available from: NTIS, Springfield, VA; PB87-176574.
16 U.S. Environmental Protection Agency. (1996) Air quality criteria for particulate matter. Research Triangle Park,
17 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001 aF-cF. 3v.
18 Available from: NTIS, Springfield, VA; PB96-168224.
19 U.S. Environmental Protection Agency. (1997) Responses to significant comments on the 1996 proposed rule on the
20 national ambient air quality standards for particulate matter (December 13, 1996; 61 FR 65638). Research
21 Triangle Park, NC: National Center for Environmental Assessment; docket number A-95-54; July.
22 ECAO-CD-92-0671; STD-081.
23 U.S. Environmental Protection Agency. (1999) Air quality criteria for CO second external review draft (October
24 1999).
25 Valberg, P. A.; Watson, A. Y. (1998) Alternative hypotheses linking outdoor particulate matter with daily morbidity
26 and mortality. Inhalation Toxicol. 10: 641-662.
27 Vedal, S. (1997) Ambient particles and health: lines that divide. J. Air Waste Manage. Assoc. 47: 551-581.
28 Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
29 nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
30 Von Mutius, E.; Sherrill, D. L.; Fritzsch, C.; Martinez, F. D.; Lebowitz, M. D. (1995) Air pollution and upper
31 respiratory symptoms in children from East Germany. Eur. Respir. J. 8: 723-728.
32 Wang, X.; Ding, H.; Ryan, L.; Xu, X. (1997) Association between air pollution and low birth weight: a
33 community-based study. Environ. Health Perspect. 105: 514-520.
34 Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
35 fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
36 Weed, D. L. (1995) Causal and preventive inference. In: Greenwald, P.; Kramer, B. S.; Weed, D. L., eds.
37 Cancer prevention and control. New York, NY: Marcel Dekker, Inc.; pp. 285-302.
38 Weed, D. L. (1997a) Meta-analysis under the microscope. J. Natl. Cancer Inst. 89: 904-905.
39 Weed, D. L. (1997b) Methodologic guidelines for review papers. J. Natl. Cancer Inst. 89: 6-7.
40 Weed, D. L.; Hursting, S. D. (1998) Biologic plausibility in causal inference: current method and practice. Am. J.
41 Epidemiol. 147: 415-425.
42 Whyatt, R. M.; Santella, R. M.; Jedrychowski, W.; Garte, S. J.; Bell, D. A.; Ottman, R.; Gladek-Yarborough, A.;
43 Cosma, G.; Young, T.-L.; Cooper, T. B.; Randall, M. C.; Manchester, D. K.; Perera, F. P. (1998)
44 Relationship between ambient air pollution and DNA damage in Polish mothers and newborns. Environ.
45 Health Perspect. 106: 821-826.
46 Wilkins, E. T. (1954) Air pollution aspects of the London Fog of December 1952. Q. J. R. Meteorol. Soc.
47 80:267-271.
48 Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
49 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
50 Wilson etal. (1998).
51 Wong and Raabe (1996).
52 Woodruff, T. J.; Grillo, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal
53 infant mortality and particulate air pollution in the United States. Environ. Health Perspect. 105: 608-612.
54 Wordley, J.; Walters, S.; Ayres, J. G. (1997) Short term variations in hospital admissions and mortality and
55 particulate air pollution. Occup. Environ. Med. 54: 108-116.
October 1999 6-229 DRAFT-DO NOT QUOTE OR CITE
-------
1 Wyzga, R. E.; Lipfert, F. W. (1998) Particulate air pollution and public health: estimated mean responses by particle
2 size and composition? In: Health effects of particulate matter in ambient air: proceedings of the international
3 symposium on the health effects of particulate matter; Prague, Czechoslovakia. Pittsburgh, PA: Air & Waste
4 Management Association, VIP-80; pp. 61-72.
5 Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
6 Inhalation Toxicol. 9: 15-29.
7 Xu, X.; Ding, H.; Wang, X. (1995) Acute effects of total suspended particles and sulfur dioxides on preterm
8 delivery: a community-based cohort study. Arch. Environ. Health 50: 407-415.
9 Zeger, S. L.; Liang, K.-Y. (1986) Longitudinal data analysis for discrete and continuous outcomes. Biometrics
10 42: 121-130.
11 Zeger etal. (1998).
12 Zeger, S. L.; Dominici, F.; Samet, J. (1999) Harvesting-resistant estimates of air pollution effects on mortality.
13 Epidemiology 10: 171-175.
14 Zejda, J. E.; Skiba, M.; Mensink, I. (1996) Lung function in children of upper Silesian industrial zone, Poland:
15 results of the cross-sectional study in two towns of different ambient air pollution levels. Cent. Eur. J. Public
16 Health 4: 252-256.
17 Zejda, J. E.; Skiba, M.; Orawiec, A.; Dybowska, T.; Cimander, B. (1996) Respiratory symptoms in children of
18 Upper Silesia, Poland: cross-sectional study in two towns of different air pollution levels. Eur. J. Epidemiol.
19 12:115-120.
20 Zemp, E.; Elsasser, S.; Schindler, C.; Kunzli, N.; Perruchoud, A. P.; Domenighetti, G.; Medici, T.;
21 Ackermann-Liebrich, U.; Leuenberger, P.; Monn, C.; Bolognini, G.; Bongard, J.-P.; Brandli, O.; Karrer, W.;
22 Keller, R.; Schoni, M. H.; Tschopp, J.-M.; Villiger, B.; Zellweger, J.-P.; SAPALDIA Team. (1999)
23 Long-term ambient air pollution and respiratory symptoms in adults (SAPALDIA study). Am. J. Respir. Crit.
24 Care Med. 159: 1257-1266.
25 Zidek, J. V.; Wong, H.; Le, N. D.; Burnett, R. (1996) Causality, measurement error and multicollinearity in
26 epidemiology. Environmetrics 7: 441-451.
27 Zmirou, D.; Barumandzadeh, T.; Balducci, F.; Ritter, P.; Laham, G.; Ghilardi, J.-P. (1996) Short term effects of air
28 pollution on mortality in the city of Lyon, France, 1985-90. In: St Leger, S., ed. The APHEA project. Short
29 term effects of air pollution on health: a European approach using epidemiological time series data. J.
30 Epidemiol. Commun. Health 50(suppl. 1): S30-S35.
31 Zmirou, D.; Balducci, F.; Dechenaux, J.; Piras, A.; Filippi, F.; Benoit-Guyod, J.-L. (1997) Meta-analyse et fonctions
32 dose-reponse des effets respiratoires de la pollution atmospherique [Meta-analysis and dose-reponse
33 functions of respiratory effects of air pollution]. Rev. Epidemiol. Sante Publique 45: 293-304.
34 Zmirou, D.; Schwartz, J.; Saez, M.; Zanobetti, A.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Ponce de Leon, A.;
35 Le Moullec, Y.; Bacharova, L.; Schouten, J.; Ponka, A.; Katsouyanni, K. (1998) Time-series analysis of air
36 pollution and cause-specific mortality. Epidemiology 9: 495-503.
October 1999 6-230 DRAFT-DO NOT QUOTE OR CITE
-------
i 7. DOSIMETRY AND TOXICOLOGY OF
2 PARTICULATE MATTER
3
4
5 7.1 INTRODUCTION
6 Epidemiological studies strongly implicate respirable particles in increased morbidity and
7 mortality in the general population. The mechanisms by which particulate matter (PM) may
8 cause adverse responses leading to severe health consequences is a matter of extensive
9 investigation. This chapter summarizes recent literature examining particle dosimetry; the
10 toxicological responses of susceptible animal models; pulmonary and systemic responses of
11 healthy and susceptible humans; and the effects of particles on ex vivo systems of cells and/or
12 cellular constituents. This chapter focuses primarily on material published since 1995. Earlier
13 published studies have been summarized in U. S. Environmental Protection Agency (1996a).
14 The reader should refer to chapters 10 and 11 of the 1996 document for more information.
15 The first section of the chapter deals with human and animal particle dosimetry - the study
16 of the deposition, translocation, clearance, and retention of particles within the respiratory tract
17 and adjacent tissues. Although the physical principles governing deposition of particles have not
18 changed since the publication of the previous Air Quality Criteria for Particulate Matter (U.S.
19 Environmental Protection Agency, 1996a), there is an improved understanding of the role of
20 certain biological determinants of the deposition/clearance process, including body size
21 (especially in relation to respiratory tract dimensions), respiratory flow and volume, and nose
22 versus mouth breathing. In addition, the understanding of regional dosimetry has expanded.
23 Studies addressing the toxicology of particles or their constituent chemicals on humans,
24 animals, and ex vivo systems is covered in the next major section of the chapter. There are
25 extensive new toxicological studies on combustion particles, such as oil fly ash. Of particular
26 interest are a small, but growing, number of new studies on the toxicology of concentrated
27 ambient air particles (CAPS) (Sioutas et al., 1995; Gordon et al., 1998), especially in susceptible
28 animal models. The aim of these studies is not only to understand the responses to CAPS but
29 also to provide potentially useful dose-response information for risk assessment.
October 1999 7-1 DRAFT-DO NOT QUOTE OR CITE
-------
1 Because the primary purpose of this document is to assess the health effects of particles in
2 humans, human controlled exposure and dosimetry studies are extremely valuable. Such studies
3 avoid uncertainties associated with extrapolation of biochemical, dosimetric, and physiological
4 responses from animals to humans and extrapolating target tissue doses from animals to humans.
5 Both animal and human controlled exposures avoid the uncertainties associated with exposure
6 assessment in field and epidemiological studies. Furthermore, sensitive humans, such as those
7 with cardiovascular or respiratory disease, can also be directly studied. Nevertheless, the number
8 and types of responses that can ethically be evaluated in human volunteers is limited. Such
9 studies are generally limited to acute or very short term exposure scenarios.
10 Animal toxicology studies are particularly useful for long term or chronic exposures and for
11 the invasive examination of response mechanisms. Animal models of disease can be useful in
12 providing information about responses of sensitive humans. Even though animal models cannot
13 duplicate human disease, certain features of human diseases can be simulated, providing valuable
14 insight into mechanisms of response. Several new studies have examined the effects of inhaled
15 or instilled particles on cardiovascular responses, a potentially important pathway for acute
16 effects. When only small amounts of PM test material are available, intratracheal instillation
17 studies can be useful to examine mechanisms of response. In in vitro studies, animal and human
18 cells and cellular components can be exposed to particles, soluble extracts of particles, or
19 individual chemical constituents of particles. Such studies are often useful in examining cellular
20 and biochemical mechanisms of action within specific cell types, with the recognition that cells
21 may behave differently in tissues or within the whole organism, especially at dose levels that
22 occur with actual ambient inhalation exposures.
23
24
25 7.2 PARTICLE DOSIMETRY
26 7.2.1 Introduction
27 A basic tenet of toxicology is that the dose delivered to the target site, rather than the
28 external exposure, is the proximal cause of any biological response. Characterization of the
29 exposure-dose-response continuum for PM requires the elucidation and understanding of the
30 mechanistic determinants of inhaled particle dose, which is dependent initially upon the
October 1999 7-2 DRAFT-DO NOT QUOTE OR CITE
-------
1 deposition of particles within the airways of the respiratory tract. Particle deposition refers to the
2 removal of particles from their airborne state due to their aerodynamic and thermodynamic
3 behavior in air. Once particles have deposited onto the surfaces of the respiratory tract, some
4 may undergo various transformations and others will not, but subsequently all will be subjected
5 to either absorptive or non-absorptive particulate removal processes. This may result in their
6 removal from airway surfaces as well as their removal to a greater or lesser degree from the
7 respiratory tract as a whole. Particulate matter translocated from initial deposition sites is said to
8 have undergone clearance. Clearance of deposited particles depends on the initial deposition site,
9 physicochemical properties of the particles, and on translocation mechanisms. Retained particle
10 burdens are determined by the dynamic relationship between deposition and clearance
11 mechanisms.
12 This section is concerned with important particle characteristics related to dosimetric
13 considerations and with mechanisms of particle deposition, clearance and retention in the
14 respiratory tract. It summarizes basic concepts as presented in the 1996 Air Quality Criteria
15 Document on Particulate Matter (U.S. Environmental Protection Agency, 1996a) and updates the
16 state of the science based upon new literature on particle deposition, clearance and retention
17 appearing since the publication of the 1996 document.
18 The dose of inhaled particles deposited and retained in the respiratory tract is governed by
19 the exposure concentration, by respiratory tract anatomy and physiology, and by physicochemical
20 properties of the particle (e.g., initial inhaled particle size, distribution, hygroscopicity,
21 solubility). Anatomic and physiologic factors influencing particle deposition and retention are
22 discussed in depth in the 1996 Criteria Document. Anatomical factors include: nasal, oral and
23 pharyngeal anatomy; size and shape of laryngeal opening; tracheal anatomy; bronchial anatomy;
24 mucus distribution; alveolar anatomy. Physiological factors include breathing pattern
25 characteristics such as respiratory rate; air flow velocities, breathing volumes; air distribution
26 among and within lobes; air-mixing characteristics; breath holding. The physicochemical
27 properties of particles as they relate to dosimetry are summarized in Section 7.2.1.1, while
28 concepts related to specific deposition and clearance mechanisms are presented in subsequent
29 sections.
30
31
October 1999 7-3 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.2.1.1 Size Characterization of Inhaled Particles
2 Information about particle size distribution is important in the evaluation of effective
3 inhaled dose. This section summarizes particle attributes requiring characterization and provides
4 general definitions important in understanding particle fate within the respiratory tract.
5 Particles exist in the atmosphere as aerosols, which are suspensions of finely dispersed
6 solid or liquid particles in the air. Because aerosols can consist of almost any material, their
7 description in simple geometric terms can be misleading unless important factors relating to
8 constituent particle size, shape, and density are considered. While the size of particles within
9 aerosols can be described based upon actual physical measurements, such as obtained with a
10 microscope, in many cases it is better to use some equivalent diameter in place of the physical
11 diameter. The most commonly used metric is aerodynamic equivalent diameter (AED), whereby
12 particles of differing geometric size, shape and density are compared aerodynamically with the
13 instability behavior of particles that are unit density (1 gm/cm3) spheres. The aerodynamic
14 behavior of unit density spherical particles constitutes a useful standard by which many types of
15 particles can be compared in terms of certain deposition mechanisms.
16 It is important to note that aerosols present in natural and work environments have
17 polydisperse size distributions. This means that the constituent particles within an aerosol have a
18 range of sizes, and are more appropriately described in terms of a size distribution parameter.
19 The lognormal distribution, i.e., the logarithms of particle diameter are normally distributed, can
20 be used for describing size distributions of most aerosols. In linear form, the logarithmic mean is
21 the median of the distribution, and the metric of variability around this central tendency is the
22 geometric standard deviation, og. The og, a dimensionless term, is the ratio of the 84th (or 16th)
23 percentile particle size to the 50th percentile size. Thus, the only two parameters needed to
24 describe a log normal distribution of aerosol particle sizes are the median diameter and the
25 geometric standard deviation. However, the actual size distribution may be obtained in various
26 ways. For example, when a distribution is described by counting particles, the median is called
27 the count median diameter (CMD). On the other hand, the median of a distribution based upon
28 particle mass in an aerosol is the mass median diameter (MMD). When using aerodynamic
29 diameters, a term that is frequently encountered is mass median aerodynamic diameter (MMAD),
30 which refers to the median of the distribution of mass with respect to aerodynamic equivalent
31 diameter. Most of the discussion in this chapter will focus on MMAD because it is the most
October 1999 7-4 DRAFT-DO NOT QUOTE OR CITE
-------
1 commonly used measure of aerosol distribution. However, alternative size distributions should
2 be used for particles below a certain size, namely -0.5 //m, which should be based upon actual
3 size, since aerodynamic properties become less important. One such metric is
4 thermodynamic-equivalent size, which is the diameter of a spherical particle that has the same
5 diffusion coefficient in air as the particle of interest.
6
7 7.2.1.2 Structure of the Respiratory Tract
8 Details of respiratory tract structure are provided in the previous Criteria Document
9 (U.S. Environmental Protection Agency, 1996a), and only a brief synopsis is presented here.
10 For dosimetry purposes, the respiratory tract can be divided into three regions on the basis of
11 structure, size, and function: extrathoracic (ET), tracheobronchial (TB) and alveolar (A).
12 The ET region consists of head airways (i.e., nasal or oral passages) through the larynx, and
13 represents the areas through which inhaled air first passes. In humans, inhalation can occur
14 through the nose or mouth (or both, known as oronasal breathing). However, most animals
15 commonly used in respiratory toxicological studies are obligate nose breathers.
16 From the ET region, inspired air enters the TB region at the trachea. From the level of the
17 trachea, the conducting airways then undergo branching for a number of generations.
18 The terminal bronchiole is the most peripheral of the distal conducting airways and these lead,
19 in humans, to the respiratory bronchioles, alveolar ducts, alveolar sacs and alveoli, all of which
20 comprise the A region. All of the conducting airways, except the trachea and portions of the
21 mainstem bronchi, are surrounded by parenchymal tissue. This is composed primarily of the
22 alveolated structures of the A region and associated blood and lymphatic vessels. It should be
23 noted that these respiratory tract regions are comprised of various cell types, and that there are
24 distinct differences in the cellular composition of the ET, TB and A regions. While a discussion
25 of cellular structure of the respiratory tract is beyond the scope of this section, details may be
26 found in a number of sources (e.g., Crystal et al., 1997).
27
28 7.2.2 Factors Controlling Dose
29 Characterization of the exposure-dose-response continuum is the fundamental objective of
30 any dose-response assessment. This section reviews the major factors controlling the fate of
31 inhaled particles as discussed in detail in the previous Criteria Document (U.S. Environmental
October 1999 7-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 Protection Agency, 1996a), and provides an update as new information on particle fate has
2 become available. It must be emphasized that dissection of the factors that control inhaled dose
3 into discrete topics masks the dynamic and interdependent nature of the intact respiratory system.
4 For example, although deposition is discussed separately from clearance mechanisms, retention
5 (i.e., the actual amount of particles found in the respiratory tract at any point in time) is
6 determined by the relative rates of deposition and clearance. Retention and the toxicologic
7 properties of the inhaled agent are related to the magnitude of the pharmacologic, physiologic, or
8 pathologic response. Therefore, assessment of the overall dosimetry and toxic response requires
9 integration of these various factors.
10 Inasmuch as particles which are too massive to be inhaled occur in the environmental air,
11 the description "inhalability" is used to denote the overall spectrum of particle sizes which are
12 potentially capable of entering the respiratory tract. Inhalability can be defined as the ratio of the
13 number concentration of particles of a certain aerodynamic diameter that are inspired through the
14 nose or mouth to the number concentration of the same diameter particle present in an inspired
15 volume of ambient air (International Commission on Radiological Protection, 1994). In general,
16 for humans, unit density particles > 100 //m diameter have a low probability of entering the
17 mouth or nose in still air. However, there is no sharp cutoff to zero probability. Furthermore,
18 there is no lower limit to inhalability as long as the particle exceeds a critical size where the
19 aggregation of atomic or molecular units is stable enough to endow it with "particulate"
20 properties, in contrast to those of free ions or gas molecules.
21
22 7.2.3 Particle Deposition
23 7.2.3.1 Mechanisms of Deposition
24 Particles may deposit within the respiratory tract by five mechanisms: inertial impaction,
25 sedimentation, diffusion, electrostatic precipitation and interception.
26 Sudden changes in airstream direction and velocity cause particles to fail to follow the
27 streamlines of airflow. As a consequence, the particles contact, or impact, onto airway surfaces.
28 The ET and upper TB airways are characterized by high air velocities and sharp directional
29 changes and, thus, dominate as sites of inertial impaction. Impaction is a significant deposition
30 mechanism for particles larger than 1 //m AED.
October 1999 7-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 All aerosol particles are continuously influenced by gravity, but particles with an AED
2 > 0.5 //m are affected to the greatest extent. A particle will acquire a terminal settling velocity
3 when a balance is achieved between the acceleration of gravity acting on the particle and the
4 viscous resistance of the air, and it is this settling which takes it into contact with airway
5 surfaces. Both sedimentation and inertial impaction can influence the deposition of particles
6 within the same size range. These deposition processes act together in the ET and TB regions,
7 with inertial impaction dominating in the upper airways and gravitational settling becoming
8 increasingly dominant in the lower conducting airways, especially for the largest particles which
9 can penetrate into the smaller bronchial airways.
10 As particle diameters become <1 //m, the particles are increasingly subjected to diffusive
11 deposition due to random bombardment by air molecules, which results in contact with airway
12 surfaces. The root mean square displacement that a particle experiences in a unit of time along a
13 given cartesian coordinate is a measure of its diffusivity. The density of a particle is unimportant
14 in determining particle diffusivity. Thus, instead of having an aerodynamic equivalent size,
15 diffusive particles of different shapes can be related to the diffusivity of a thermodynamic
16 equivalent size based on spherical particles.
17 The particle size region around 0.3-0.5 //m is frequently described as consisting of particles
18 which are small enough to be minimally influenced by impaction or sedimentation and large
19 enough to be minimally influenced by diffusion. Such particles are the most persistent in inhaled
20 air and undergo the lowest extent of deposition in the respiratory tract.
21 Interception is deposition by physical contact with airway surfaces. The interception
22 potential of any particle depends on its physical size, and fibers are the chief concern in relation
23 to the interception process. Their aerodynamic size is determined predominantly by their
24 diameter, rather than their length.
25 Electrostatic precipitation is deposition related to particle charge. The minimum charge an
26 aerosol particle can have is zero, when it is electrically neutral. This condition is rarely achieved
27 because of the random charging of aerosol particles by air ions. Aerosol particles will acquire
28 charges from these ions by collisions with them due to their random thermal motion.
29 Furthermore, many laboratory generated aerosols are charged. Such aerosols will lose their
30 charge slowly as they attract oppositely charged ions. An equilibrium state of these competing
31 processes is eventually achieved. This Boltzmann equilibrium represents the charge distribution
October 1999 7-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 of an aerosol in charge equilibrium with bipolar ions. The minimum amount of charge is very
2 small, with a statistical probability that some particles within the aerosol will have no charge and
3 others will have one or more charges.
4 The electrical charge on some particles may result in an enhanced deposition over what
5 would be expected from size alone. This is due to image charges induced on the surface of the
6 airway by these particles or to space-charge effects whereby repulsion of particles containing like
7 charges results in increased migration toward the airway wall. The effect of charge on deposition
8 is inversely proportional to particle size and airflow rate. This type of deposition is probably
9 small compared to the effects of turbulence and other deposition mechanisms, and has generally
10 been considered to be a minor contributor to overall particle deposition. However, a recent study
11 employing hollow airway casts of the human tracheobronchial tree assessed deposition of
12 ultrafine (0.02 //m) and fine (0.125 //m) particles; the deposition of singly charged particles was
13 found to be 5-6 times that of particles having no charge, and 2-3 times that of particles at
14 Boltzmann equilibrium (Cohen et al., 1998). This suggests that electrostatic precipitation may,
15 in fact, be a significant deposition mechanism for ultrafine, and some fine, particles within the
16 TB region.
17
18 7.2.3.2 Deposition Patterns in the Human Respiratory Tract
19 Knowledge of sites where particles of different sizes deposit in the respiratory tract and the
20 amount of deposition is necessary for understanding and interpreting the health effects associated
21 with exposure to particles. Particles deposited in the various regions are subjected to large
22 differences in clearance mechanisms and pathways and, consequently, retention times. This
23 section summarizes concepts of particle deposition in humans and laboratory animals as reported
24 in U.S. Environmental Protection Agency (1996a), and provides additional information based
25 upon studies published since the release of this document.
26
27 Total Respiratory Tract Deposition
28 Total human respiratory tract deposition, as a function of particle size, is depicted in
29 Figure 7-1. These data were obtained by various investigators using different sizes of spherical
30 test particles in healthy male adults under different ventilation conditions; the large standard
31 deviations reflect interindividual viability of deposition efficiencies. Deposition with nose
October 1999 7-8 DRAFT-DO NOT QUOTE OR CITE
-------
100
80
£ 60
o
'.^
I
a 40
20
0
-^ O
I
O Human (Oral)
• Human (Nasal)
I
I
0.01
0.1 1.0
Particle Diameter (urn)
10
Figure 7-1. Total deposition data (percentage deposition of amount inhaled) in humans
as a function of particle size. All values are means with standard deviations
when available. Particle diameters are aerodynamic (MMAD) for those
^ 0.5 //m.
Source: Schlesinger (1988).
1 breathing is generally higher than that with mouth breathing due to the superior filtration
2 capabilities of the nasal passages. For particles with aerodynamic diameters greater than 1 //m,
3 deposition is governed by impaction and sedimentation and it increases with increasing AED.
4 When AED is >10 //m, almost all inhaled particles are deposited. As the particle size decreases
5 from -0.5 //m, diffusional deposition becomes dominant and total deposition depends more upon
6 the actual physical diameter of the particle, with decreasing particle diameter leading to an
7 increase in total deposition. Total deposition shows a minimum for particle diameters in the
October 1999
7-9
DRAFT-DO NOT QUOTE OR CITE
-------
1 range of 0.3 - 0.5 //m, where, as noted above, neither sedimentation, impaction or diffusion
2 deposition are very effective.
3 A property of some ambient particulate species that affects deposition is hygroscopicity, the
4 propensity of a material for taking up and retaining moisture under certain conditions of humidity
5 and temperature. Such particles can increase in size in the humid air within the respiratory tract
6 and when inhaled will deposit according to their hydrated size rather than their initial size. The
7 implications of hygroscopic growth upon deposition has been extensively reviewed by Morrow
8 (1986) and Hiller (1991). In general, compared to nonhygroscopic particles of the same initial
9 size, the deposition of hygroscopic aerosols in different regions of the lung may be higher or
10 lower depending upon the initial size. Thus, for particles with initial sizes larger than ~ 0.5 //m,
11 the influence of hygroscopicity is to increase total deposition, whereas for smaller ones total
12 deposition is decreased.
13
14 ETRegion
15 The fraction of inhaled particles depositing in the ET region is quite variable, depending on
16 particle size, flow rate, breathing frequency and whether breathing is through the nose or the
17 mouth. Mouth breathing bypasses much of the filtration capabilities of the nasal airways, leading
18 to increased deposition in the lungs (TB and A regions). The ET region is clearly the site of first
19 contact with particles in the inhaled air, and acts as a "prefilter" for the lungs.
20 Since release of the previous Criteria Document, a number of studies have explored ET
21 deposition with in vivo studies, as well as in both physical and mathematical model systems.
22 In one new study, the relative distribution of particle deposition between the oral and nasal
23 passages was assessed during inhalation by use of a physical model (silicone rubber) of the
24 human upper respiratory system, extending from the nostrils and mouth through the main bronchi
25 (Lennon et al., 1998). Monodisperse particles ranging in size from 0.3 - 2.5 //m were used at
26 various flow rates ranging from 15-50 1/min. Total deposition was assessed, as was regional
27 deposition in the oral passages, lower oropharynx-trachea, nasal passages, and
28 nasopharynx-trachea. Deposition within the nasal passages was found to agree with available
29 data obtained from a human inhalation study (Heyder and Rudolf, 1977), being proportional to
30 particle size, density and inspiratory flow rate. It was also found that for oral inhalation, the
31 relative distribution between the oral cavity and the oropharynx-trachea was similar, while for
October 1999 7-10 DRAFT-DO NOT QUOTE OR CITE
-------
1 nasal inhalation, the nasal passages contained most of the particles deposited in the model, with
2 only about 10% depositing in the nasopharynx-trachea section. Furthermore, the deposition
3 efficiency of the nasopharynx-trachea region was greater than that of the oropharynx-trachea
4 region. For simulated oronasal breathing, deposition in the ET region depended primarily upon
5 particle size rather than fiowrate. For all flows and for all breathing modes, total deposition in
6 the ET region increased as particle diameter increased. Such information on deposition patterns
7 in the ET region is useful in refining empirical deposition models.
8 Deposition within the nasal passages was further evaluated by Kesavanathan and Swift
9 (1998), who examined the deposition of 1-10 //m particles in the nasal passages of normal adults
10 under an inhalation regime in which the particles were drawn through the nose and out through
11 the mouth at flow rates ranging from 15-35 1/min. At any particle size, deposition increased
12 with increasing flow rate, while at any flow rate, deposition increased with increasing particle
13 size. In addition, as was shown experimentally by Lennon et al. (1998) under oronasal breathing
14 conditions, deposition of 0.3 to 2.5 um particles within the nasal passages was significantly
15 greater than within the oral passages, and nasal inhalation resulted in greater total deposition in
16 the model than did oral inhalation. These results are consistent with other studies discussed in
17 the previous Criteria Document and with the dominance of impaction deposition within the ET
18 region.
19 For ultrafine particles (d < 0.1 //m), deposition in the ET region is controlled by diffusion,
20 which depends only on the particle's geometric diameter. Prior to 1996, ET deposition for this
21 particle size range had not been studied extensively in humans, and this remains the case. In the
22 earlier document, the only data available for ET deposition of ultrafine particles were from cast
23 studies. More recently, deposition in the ET was examined using mathematical modeling. Three
24 dimensional numerical simulations of flow and particle diffusion in the human upper respiratory
25 tract, which included the nasal region, oral region, larynx and first two generations of bronchi,
26 were performed by Yu et al. (1998). Deposition of particles ranging from 0.001 -0.1 //m in these
27 different regions was calculated under inspiratory and expiratory flow conditions. Deposition
28 efficiencies in the total model were lower on expiration than inspiration, although the former
29 were quite high. Nasal deposition of ultrafine particles can be quite high. For example, it
30 accounted for up to 54% of total deposition in the model system for 0.001 //m particles (total
31 deposition efficiency in the mathematical model was 75% [of amount entering] for this size
October 1999 7-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 particle). With oral breathing, deposition efficiency was estimated at 48% (of amount entering)
2 (Yuetal, 1998).
3 Swift and Strong (1996) examined the deposition of ultrafine particles, ranging in size from
4 0.053-0.062 //m, in the nasal passages of normal adults during constant inspiratory flows of
5 6-22 1/min. The results are consistent with results noted in studies above, namely that the nasal
6 passages are highly efficient collectors for ultrafine particles. In this case, fractional deposition
7 ranged from 94 - 99% (of amount inhaled). There was found to be only a weak dependence of
8 deposition on flow rate, which contrasts with results noted above (Lennon et al., 1998) for
9 particles >0.3 //m but is consistent with diffusion as the deposition mechanism for ultrafines.
10 Cheng et al. (1997) examined oral airway deposition in a replicate cast of the human nasal
11 cavity, oral cavity and laryngeal-tracheal sections. Particle sizes ranged from 0.005 - 0.150 //m,
12 and constant inspiratory and expiratory flow rates of 7.5 - 30 1/min were used. They noted that
13 the deposition fractions within the oral cavity were essentially the same as that in the
14 laryngeal-tracheal sections for all particle sizes and fiowrates. They ascribed this to the balance
15 between flow turbulence and residence time in these two regions. Svartengren et al. (1995)
16 examined the effect of changes in external resistance on oropharyngeal particle deposition in
17 asthmatics. Under control mouthpiece breathing conditions, the median deposition as a
18 percentage of inhaled particles in the mouth and throat was 20% (mean = 33%; range 12-84%).
19 Although the mean deposition fell to 22% with added resistance, the median value remained at
20 20% (range 13-47%). Fiberoptic examination of the larynx revealed that there was a trend for
21 increased mouth and throat deposition associated with laryngeal narrowing. Katz et al. (1999)
22 suggest, on the basis of mathematical model calculations, that turbulence may play a key role in
23 enhancing particle deposition in the larynx and trachea.
24 The results of all of the above studies support the previously known ability of the ET
25 region, and especially the nasal passages, to act as an efficient filter for inhaled particles.
26 Ultrafine particles <0.01 //m also have significant deposition within the ET region and this
27 region, therefore, serves as an important filter for these particles as well as for larger ones,
28 potentially reducing the amount of particles within some size ranges which are available for
29 deposition in the TB and A regions.
30
31
October 1999 7-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 TB and A Regions
1 Particles which do not deposit in the ET region enter the lung, but their regional deposition
3 in the lung cannot be precisely measured. Much of the available regional deposition data have
4 been obtained from experiments with radioactive labeled poorly soluble particles. These have
5 been described in U.S. Environmental Protection Agency (1996a) and there are no new regional
6 data available since the publication of this document which would amend the descriptions of
7 regional deposition patterns as presented in that document.
8
9 Local Distribution of Deposition
10 Airway structure and its associated air flow patterns are exceedingly complex and
11 ventilation distribution of air in different parts of the lung is uneven. Thus, it is expected that
12 particle deposition patterns within the ET, TB, and A regions would be highly nonuniform. This
13 was discussed in detail. Basically, using deposition data from living subjects as well as from
14 mathematical and physical models, enhanced deposition has been shown to occur in the nasal
15 passages, trachea and at branching points in the TB and A regions. Recently, Churg and Vedal
16 (1996) examined retention of particles on carinal ridges and tubular sections of airways from
17 lungs obtained at necropsy. Results indicated significant enhancement of particle deposits on
18 carinal ridges through the segmental bronchi; the ratios were similar in all airway generations
19 examined.
20 Deposition "hot spots" at airway bifurcations have undergone additional analyses using
21 mathematical modeling techniques. Using calculated deposition sites, a number of studies
22 showed a strong correlation between secondary flow patterns and deposition sites and density for
23 large (10 //m) particles, as well as for ultrafine particles (0.01 //m) (Heistracher and Hofmann,
24 1997; Hofmann et al., 1996). This supports experimental work, noted in U.S. Environmental
25 Protection Agency (1996a) indicating that, like larger particles, ultrafine particles also show
26 enhanced deposition at airway branch points, even in the upper tracheobronchial tree.
27
28 7.2.3.3 Biological Factors Modifying Deposition
29 Experimental deposition data in humans are commonly derived using healthy adult
30 Caucasian males. Various factors can act to alter deposition patterns from those obtained in this
31 group. Evaluation of these factors is important to help understand potentially susceptible
October 1999 7-13 DRAFT-DO NOT QUOTE OR CITE
-------
1 subpopulations, since differences in response may be due to dosimetry differences as well as to
2 differences in sensitivity to a pollutant. The effects of different biological factors on deposition
3 were discussed in U.S. Environmental Protection Agency (1996a) and are summarized below
4 with additional information obtained from more recent studies.
5
6 Gender
7 Males and females differ in body size and ventilatory parameters, so it is expected that
8 there would be gender differences in deposition. Using particles in the 2.5 to 7.5 //m size range
9 Pritchard et al. (1986) indicated that, for comparable particle sizes and inspiratory flow rates,
10 females had higher ET and TB deposition and smaller A deposition than did males. The ratio of
11 A deposition to total thoracic deposition in females was also found to be smaller. These
12 differences were attributed to gender differences in airway size.
13 In a recent study (Bennett et al., 1996), the total respiratory tract deposition of 2 //m
14 particles was examined in adult males and females aged 18-80 years who breathed with a
15 normal resting pattern. Deposition was assessed in terms of a deposition fraction, which was the
16 difference between the amount of particles inhaled and exhaled during oral breathing. While
17 there was a tendency for a greater deposition fraction in females compared to males, because
18 males had greater minute ventilation, the deposition rate (i.e., deposition per unit time) was
19 greater in males than in females.
20 Kim and Hu (1998) assessed regional deposition patterns in healthy adult males and
21 females using sebacate aerosols with median aerodynamic sizes of 1, 3 and 5 //m and a bolus
22 delivery technique, which involved controlled breathing. The total deposition in the lungs was
23 similar for both genders with the smaller particle, but was greater in women for the 3 and 5 //m
24 particles regardless of the inhalation flow rate used; this difference ranged from 9-31%, with
25 higher values associated with higher flow rates. The pattern of deposition was similar for both
26 genders, although females showed enhanced deposition peaks for all three particle sizes. The
27 volumetric depth location of these peaks was similar in both genders, but was found to shift from
28 peripheral (increased volumetric depth) to proximal (shallow volumetric depth) regions of the
29 lung with increasing particle size. Thus, deposition appeared to be more localized in the lungs of
30 females compared to those of males. Local deposition of 1 //m particles was somewhat flow
October 1999 7-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 dependent, but for larger (5 //m) particles was largely independent of flow (flows did not include
2 those that would be typical of exercise).
3
4 Age
5 Airway structure and respiratory conditions vary with age, and these variations may alter
6 the deposition pattern of inhaled particles. The limited experimental studies reported in the
7 earlier Criteria Document indicated results ranging from no clear dependence of total deposition
8 on age to slightly higher deposition in children than adults. Potential deposition differences
9 between children and adults were assessed to a greater extent using mathematical models. These
10 indicated that if the entire respiratory tract and a complete breathing cycle at normal rate are
11 considered, that ET deposition in children would generally be higher than that in adults, but that
12 TB and A regional deposition in children may be either higher or lower than the adult, depending
13 upon particle size (Xu and Yu, 1986). Enhanced deposition in the TB region would occur for
14 particles <5 //m in children (Xu and Yu, 1986; Hofmann et al, 1989a).
15 Cheng et al. (1995) examined deposition of ultrafine particles in replica casts of the nasal
16 airways of children aged 1.5 to 4 yrs. Particle sizes ranged from 0.0046 to 0.2 //m, and both
17 inspiratory and expiratory flowrates were used (3-16 1/min). Deposition efficiency was found to
18 decrease with increasing age for a given particle size and fiowrate.
19 Oldham et al. (1997) examined the deposition of monodisperse particles having diameters
20 of 1, 5, 10 and 15 //m in hollow airway models which were designed to represent the trachea and
21 the first few bronchial airway generations of an adult, a 7 yr old child and a 4 yr old child. They
22 noted that in most cases, the total deposition efficiency was greater in the child-size models than
23 in the adult model.
24 Bennett et al. (1997a) analyzed the regional deposition of 4.5 //m, poorly soluble (Fe2O3)
25 particles in children and in adults with mild cystic fibrosis (CF), but who likely had normal upper
26 airway anatomy, such that intra- and extra- thoracic deposition would be similar to that in healthy
27 adults. The mean age of the children was 13.8 yr and adults were 29.1 yr. ET deposition, as a
28 percentage of total respiratory tract deposition, was higher by about 50% in children compared to
29 CF and healthy adults (30.7% vs 20.1% vs 16.0% respectively). There was an age dependence of
30 ET deposition in the children, in that the percentage ET deposition tended to be higher at a
31 younger age; the younger group (<14 yr) had almost twice the percentage ET deposition of the
October 1999 7-15 DRAFT-DO NOT QUOTE OR CITE
-------
1 older group (>14 yr). Additional analyses showed an inverse correlation of extrathoracic
2 deposition with body height (Bennett et al., 1997a). There was no significant difference in lung
3 or total respiratory tract deposition between the children and adults. Since ET deposition was
4 age-dependent and total deposition was not, this suggests that the ET region does a more
5 effective job in children of filtering out the particles that would otherwise reach the TB region.
6 However, since the lungs of children are smaller than those of adults, children may still have
7 comparable deposition per unit surface area as would adults.
8 Bennett and Zeman (1998) measured the deposition of monodisperse 2 //m (MMAD) wax
9 particles in children aged 7-14 yr and adolescents aged 14-18 yr for comparison to that in adults
10 (19-35 yr). Each subject inhaled the particles by following their previously determined
11 individual spontaneous resting breathing pattern. Deposition was assessed by measuring the
12 amount of particles inhaled and exhaled. There was no age-related difference in deposition
13 within the children group. There was also no significant difference in deposition between the
14 children and adolescents, between the children and adults, or between the adolescents and adults.
15 However, the investigators noted that since the children had smaller lungs and higher minute
16 volumes relative to lung size, they would likely receive greater doses of particles per lung surface
17 area compared to adults. Furthermore, deposition in children did vary with tidal volume,
18 increasing with increasing volume to a greater extent than was seen in adults. These additional
19 studies still do not provide unequivocal evidence for significant differences in deposition
20 between adults and children, even when considering differences in lung surface area. However,
21 it should be noted that differences in levels of activity between adults and children are likely to
22 play a fairly large role in age-related differences in deposition patterns of ambient particles.
23 Children generally have higher activity levels during the day, and higher associated minute
24 ventilation per lung size. This will contribute to a greater inhaled dose of particles. Activity
25 levels in relationship to exposure are discussed more fully in Chapter 5.
26 Another subpopulation of potential concern related to susceptibility to inhaled particles is
27 the elderly. In the study of Bennett et al. (1996) described above, in which the total respiratory
28 tract deposition of 2 //m particles was examined in people aged 18 to 80, the deposition fraction
29 in the lungs in people with normal lung function was found to be independent of age, depending
30 solely upon breathing pattern and airway resistance.
31
October 1999 7-16 DRAFT-DO NOT QUOTE OR CITE
-------
1 Respiratory Tract Disease
2 The presence of respiratory tract disease can affect airway structure and ventilatory
3 parameters, thus altering deposition compared to that in healthy individuals. The effect of airway
4 diseases on deposition has been studied extensively (U.S. Environmental Protection Agency,
5 1996a). Studies described therein had shown that people with chronic obstructive pulmonary
6 disease (COPD) had deposition patterns that were very heterogeneous with differences in
7 regional deposition compared to normals. People with asthma and obstructive pulmonary disease
8 tended to have greater TB deposition than did healthy people. Furthermore, there tended to be an
9 inverse relationship between bronchconstriction and the extent of deposition in the A region,
10 while total respiratory tract deposition generally increased with increasing level of airway
11 obstruction. The described studies were performed during controlled breathing, where all
12 subjects breathed with the same tidal volume and respiratory rate. However, while resting tidal
13 volume is similar or elevated in people with COPD compared to normals, the former tend to
14 breathe at a faster rate, resulting in higher than normal tidal peak flow and resting minute
15 ventilation. Thus, some of the reported differences in the deposition of particles could have been
16 due to increased fractional deposition with each breath. While the extent to which lung
17 deposition may change with respect to particle size, breathing pattern, and disease status in
18 people with COPD is still unclear, some recent studies have attempted to provide additional
19 insight into this issue.
20 Bennett et al. (1997b) measured the fractional deposition of insoluble 2 //m particles in
21 people with severe to moderate COPD (mix of emphysema and chronic bronchitis, mean age
22 62 yr) and compared this to healthy older adults (mean age 67 yr) under conditions where the
23 subjects breathed using their individual resting breathing pattern as well as a controlled breathing
24 pattern. People with COPD tended to breathe with elevated tidal volume and at a faster rate than
25 people with healthy lungs, resulting in about 50% higher resting minute ventilation. Total
26 respiratory tract deposition was assessed in terms of deposition fraction, a measure of the amount
27 deposited based upon measures of aerosol inhaled and amount exhaled, and deposition rate, the
28 particles deposited per unit time. Under typical breathing conditions, people with COPD had
29 about 50% greater deposition fraction than did age-matched healthy adults. Due to the elevation
30 in minute ventilation, people with COPD had average deposition rates about 2.5 times that of
31 healthy adults. Similar to previously reviewed studies (U.S. Environmental Protection Agency,
October 1999 7-17 DRAFT-DO NOT QUOTE OR CITE
-------
1 1996a), these investigators observed an increase in deposition with an increase in airway
2 resistance, suggesting that, at rest, COPD resulted in increased deposition of fine particles in
3 proportion to the severity of airway disease. The investigators also reported a decrease in
4 deposition with increasing mean effective airspace diameter; this suggested that the enhanced
5 deposition was associated more with the chronic bronchitic component of COPD than with the
6 emphysematous component of the disease. Greater deposition was noted with the natural
7 breathing compared to the fixed pattern.
8 Kim and Kang (1997) measured lung deposition of 1 //m particles inhaled via the mouth by
9 healthy adults (mean age 27 yr) and by those with various degrees of airway obstruction, namely
10 smokers (mean age 27 yr), smokers with small airway disease (SAD; mean age 37 yr), asthmatics
11 (mean age 48 yr) and patients with COPD (mean age 61 yr) breathing under the same controlled
12 pattern. Deposition fraction was obtained as in the study of Bennett et al. (1997b), described
13 above. There was a marked increase in deposition in people with COPD. Deposition was 16%,
14 49%, 59% and 103% greater in smokers, smokers with SAD, asthmatics and people with COPD,
15 respectively, than healthy adults. Deposition in COPD patients was significantly greater than that
16 associated with either SAD or asthma; there was no significant difference in deposition between
17 people with SAD and asthma. Deposition fraction was found to be correlated with percent
18 predicted forced expiratory volume (FEVj) and forced expiratory flow (FEF25-75%). Kohlhaufl
19 et al. (1999) also showed increased deposition of fine particles (0.9 //m) in women with
20 bronchial hyperresponsiveness.
21 Thus, the data base related to particle deposition and lung disease suggests that total lung
22 deposition is generally increased with obstructed airways, regardless of deposition distribution
23 between the TB and A regions. Airflow distribution is very uneven in COPD due to the irregular
24 pattern of obstruction, and there can be closure of small airways. In this situation, a part of the
25 lung is inaccessible and particles can penetrate deeper into other ventilated regions. Thus,
26 deposition can be enhanced locally in regions of active ventilation, particularly in the A region.
27
28 Anatomical Variability
29 As indicated above, variations in anatomical parameters between genders and between
30 healthy people and those with obstructive lung disease can affect deposition patterns. However,
31 previous analyses have generally overlooked the effect upon deposition of normal interindividual
October 1999 7-18 DRAFT-DO NOT QUOTE OR CITE
-------
1 variability in airway structure in healthy individuals. This is an important consideration in
2 dosimetry modeling, which is often based upon a single idealized structure. Studies available
3 since 1996 have attempted to assess the influence of such variation in respiratory tract structure
4 upon deposition patterns.
5 The ET region is the first to contact inhaled particles, and therefore deposition within this
6 region would reduce the amount of particles available for deposition in the lungs. Variations in
7 relative deposition within the ET region will, therefore, propagate through the rest of the
8 respiratory tract, creating differences in calculated doses from individual to individual.
9 A number of studies have examined the influence of variations in airway geometry upon
10 deposition in the ET region.
11 Cheng et al. (1996) examined nasal airway deposition in healthy adults using particles
12 ranging in size from 0.004-0.15 //m at two constant inspiratory flow rates, 167 and 33 ml/sec.
13 Deposition was evaluated in relation to measures of nasal geometry as determined by magnetic
14 resonance imaging and acoustic rhinometry. They noted that interindividual variability in
15 deposition was correlated with the wide variation of nasal dimensions, in that greater surface
16 area, smaller cross-sectional area and increasing complexity of airway shape were all associated
17 with enhanced deposition.
18 Using a regression analysis of data on nasal airway deposition derived from Cheng et al.
19 (1996), Guilmette et al. (1997) noted that the deposition efficiency within this region was highly
20 correlated with both nasal airway surface area and volume; this indicated that airway size and
21 shape factors were important in explaining the intraindividual variability noted in experimental
22 studies of human nasal airway aerosol deposition. Thus, much of the variability in measured
23 deposition among people was due to differences in the size and shape of airway regions.
24 Kesavanathan and Swift (1998) also evaluated the influence of geometry in affecting
25 deposition in the nasal passages of normal adults from two ethnic groups. Mathematical
26 modeling of the results indicated that the shape of the nostril affected particle deposition in the
27 nasal passages, but that there still remained large intersubject variations in deposition when this
28 was accounted for, and which was likely due to geometric variability in the mid and posterior
29 regions of the nasal passages.
30 Bennett et al. (1998) studied the role of anatomic dead space (ADS) in particle deposition
31 and retention in bronchial airways using an aerosol bolus technique. They found that the
October 1999 7-19 DRAFT-DO NOT QUOTE OR CITE
-------
1 fractional deposition was dependant on the subject's ADS and that a significant number of
2 particles were retained beyond 24 h. This finding of prolonged retention of insoluble particles in
3 the airways substantiates the findings of Scheuch et al. (1995) and Stahlhofen et al. (1986a).
4 Bennett et al. (1999) also found a lung volume-dependant asymmetric distribution of particles
5 between the left and right lung; the leftright ratio was increased at increased percent of total lung
6 capacity (at 70% TLC, L:R was 1.60).
7
8 7.2.3.4 Interspecies Patterns of Deposition
9 The various species used in inhalation toxicology studies that serve as the basis for
10 dose-response assessment may not receive identical doses in a comparable respiratory tract
11 region (i.e., ET, TB, or A) when exposed to the same aerosol at the same inhaled concentration.
12 Such interspecies differences are important because the adverse toxic effect is often related to the
13 quantitative pattern of deposition within the respiratory tract as well as to the exposure
14 concentration; this pattern determines not only the initial respiratory tract tissue dose but also the
15 specific pathways by which deposited material is cleared and redistributed (Schlesinger, 1985).
16 Differences in patterns of deposition between humans and animals have been summarized (U.S.
17 Environmental Protection Agency, 1996a; Schlesinger et al., 1997). Such differences in initial
18 deposition must be considered when relating biological responses obtained in laboratory animal
19 studies to effects in humans.
20 Some recent studies have addressed various aspects of interspecies differences in
21 deposition using mathematical modeling approaches. Hofmann et al. (1996) compared
22 deposition between rat and human lungs using three-dimensional asymmetric bifurcation models
23 and mathematical procedures for obtaining air flow and particle trajectories. Deposition in
24 segmental bronchi and terminal bronchioles was evaluated under both inspiration and expiration,
25 at particle sizes of 0.01, 1 and 10 //m (which covered the range of deposition mechanisms from
26 diffusion to impaction). Total deposition efficiencies of all particles in the upper and lower
27 airway bifurcations were comparable in magnitude for both rat and human. However, the
28 investigators noted that penetration probabilities from preceding airways must be considered.
29 When considering the higher penetration probability in the human lung, the resulting bronchial
30 deposition fractions were generally higher in human than rat. For all particle sizes, deposition at
October 1999 7-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 rat bronchial bifurcations was less enhanced on the carinas compared to that found in human
2 airways.
3 Hofmann et al. (1996) attempted to account for interspecies differences in branching
4 patterns in deposition analyses. Numerical simulations of three-dimensional particle deposition
5 patterns within selected (species-specific) bronchial bifurcations indicated that morphologic
6 asymmetry was a major determinant of the heterogeneity of local deposition patterns. They noted
7 that many interspecies deposition calculations used morphometry which was described by
8 deterministic lung models, i.e., the number of airways in each airway generation adopts a
9 constant value and all airways in a given generation have identical lengths and diameters. Such
10 models cannot account for variability and branching asymmetry of airways in the lungs. Thus,
11 their study employed computations which used stochastic morphometric models of human and
12 rat lungs (Koblinger and Hofmann, 1985, 1988; Hofmann et al., 1989b) and evaluated regional
13 and local particle deposition. Stochastic models of lung structure describe, in mathematical
14 terms, the inherent asymmetry and variability of the airway system, including diameter, length
15 and angle. They are based upon statistical analyses of actual morphometric analyses of lungs.
16 The model also incorporated breathing patterns for humans and rats. The dependence of
17 deposition on particle size was found to be similar in both rats and humans, with a deposition
18 minima in the size range of 0.1 to 1 //m for both total deposition and deposition within the TB
19 region. This was not found to occur in the A region, where a deposition maximum occurred at
20 about 0.02-0.03 //m in both species followed by a decline, and then another maximum between
21 3 and 5 //m. The deposition decrease in the A region at the smallest and largest sizes was due to
22 the filtering efficiency of upstream airways. While deposition patterns were qualitatively similar
23 in rat and human, total respiratory tract and TB deposition in the human lung appeared to be
24 consistently higher than in the rat. Alveolar region deposition fraction in humans was lower than
25 in the rat over the size range of 0.001 to 10 //m. Furthermore, both species showed a similar
26 pattern of dependence of deposition on flow rate.
27 The above model also assessed local deposition. In both human and rat, deposition of
28 0.001 //m and 10 //m particles was highest in the upper bronchial airways, while 0.1 and 1 //m
29 particles showed higher deposition in more peripheral airways, namely the bronchiolar airways
30 in rat and the respiratory bronchioles in humans. Deposition was variable within any branching
31 generation due to differences in airway dimensions, and regional and total deposition also
October 1999 7-21 DRAFT-DO NOT QUOTE OR CITE
-------
1 exhibited intrasubject variations. Airway geometric differences between rats and humans were
2 reflected in deposition. Due to the greater branching asymmetry in rats, prior to about generation
3 12, each generation showed deposition maxima at two particle sizes, reflecting deposition in
4 major and minor daughters. These geometric differences became reduced with depth into the
5 lung; beyond generation 12, these two maxima were no longer seen.
6 The probability of any biological effect in humans or animals depends on deposition and
7 retention of particles as well as the underlying dose-response relationship. Interspecies
8 dosimetric extrapolation must consider differences in deposition, clearance and dose response.
9 Thus, even similar deposition patterns may not result in similar effects in different species, since
10 dose is also affected by clearance mechanisms and species sensitivity. In addition, the total
11 number of particles deposited in the lung may not be the most relevant dose metric to compare
12 species. For example, it may be the number of deposited particles per unit surface area that
13 determines response. More specifically, even if deposition is similar in rat and human, there
14 would be a higher deposition density in the rat due to the smaller surface area of rat lung. Thus,
15 species specific differences in deposition density should be considered when health effects
16 observed in laboratory animals are being evaluated in terms of the human situation.
17
18 7.2.4 Particle Clearance and Translocation
19 7.2.4.1 Mechanisms and Pathways of Clearance
20 Particles that deposit upon airway surfaces may be cleared from the respiratory tract
21 completely, or may be translocated to other sites within this system, by various regionally distinct
22 processes. These clearance mechanisms, which are outlined in Table 7-1, can be categorized as
23 either absorptive (i.e., dissolution) or nonabsorptive (i.e., transport of intact particles) and may
24 occur simultaneously or with temporal variations. It should be mentioned that particle solubility
25 in terms of clearance refers to solubility within the respiratory tract fluids and cells. Thus, an
26 "insoluble" particle is considered to be one whose rate of clearance by dissolution is insignificant
27 compared to its rate of clearance as an intact particle. For the most part, all deposited particles
28 are subject to clearance by the same mechanisms, with their ultimate fate a function of deposition
29 site, physicochemical properties (including any toxicity), and sometimes deposited mass or
30 number concentration. Clearance routes from the various regions of the respiratory tract have
October 1999 7-22 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-1. OVERVIEW OF RESPIRATORY TRACT PARTICLE CLEARANCE
AND TRANSLOCATION MECHANISMS
Extrathoracic region (ET)
Mucociliary transport
Sneezing
Nose wiping and blowing
Dissolution (for "soluble" particles) and absorption into blood
Tracheobronchial region (TB)
Mucociliary transport
Endocytosis by macrophages/epithelial cells
Coughing
Dissolution (for "soluble" particles) and absorption into blood
Alveolar region (A)
Macrophages, epithelial cells
Interstitial
Dissolution for "soluble" and "insoluble" particles (intra-and extracellular) and absorption into blood/lymph
Source: Schlesinger (1995).
1 been discussed in detail (U.S. Environmental Protection Agency, 1996a; Schlesinger et al.,
2 1997). They are schematically shown in Figure 7-2, and will be reviewed only briefly.
3
4 ET Region
5 The clearance of insoluble particles deposited in the posterior portions of the nasal passages
6 occurs via mucociliary transport, with the general flow of mucus towards the nasopharynx.
7 Mucus flow in the most anterior portion of the nasal passages is forward, clearing deposited
8 particles to the vestibular region where removal is by sneezing, wiping, or blowing.
9 Soluble material deposited on the nasal epithelium is accessible to underlying cells via
10 diffusion through the mucus. Dissolved substances may be subsequently translocated into the
11 bloodstream. The nasal passages have a rich vasculature, and uptake into the blood from this
12 region may occur rapidly.
13 Clearance of poorly soluble particles deposited in the oral passages is by coughing and
14 expectoration or by swallowing into the gastrointestinal tract. Soluble particles are likely to be
15 rapidly absorbed after deposition.
October 1999 7-23 DRAFT-DO NOT QUOTE OR CITE
-------
Nasal Passages
Anterior
Posterior
Extrinsic Clearance
1
Pharynx
c
Tracheobronchial Tree
Figure 7-2. Major physical clearance pathways for particles deposited in the extrathoracic
region and tracheobronchial tree.
Source: U.S. Environmental Protection Agency (1996a)
1
2
3
4
5
6
7
8
9
10
11
TB Region
Poorly soluble particles deposited within the TB region are cleared by mucociliary
transport towards the oropharynx, followed by swallowing. Poorly soluble particles may also
traverse the epithelium by endocytotic processes, entering the peribronchial region, be engulfed
via phagocytosis by airway macrophages, which can then move cephalad on 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. There is, however, evidence that even
some soluble particles may be cleared by mucociliary transport (Bennett and Ilowite, 1989;
Matsui et al, 1998).
October 1999
7-24
DRAFT-DO NOT QUOTE OR CITE
-------
1 A Region
1 Clearance from the A region occurs via a number of mechanisms and pathways. Particle
3 removal by macrophages comprises the main nonabsorptive clearance process in this region.
4 These cells, which reside on the epithelium, phagocytize and transport deposited material which
5 they contact by random motion or via directed migration under the influence of chemotactic
6 factors.
7 While alveolar macrophages normally comprise up to about 5% of the total alveolar cells in
8 healthy, non-smoking humans and other mammals, the actual cell count may be altered by
9 particle loading. The magnitude of any increase in cell number is related to the number of
10 deposited particles rather than to total deposition by weight. Thus, equivalent masses of an
11 identically deposited substance would not produce the same response if particle sizes differed,
12 and the deposition of smaller particles would tend to result in a greater elevation in macrophage
13 number than would deposition of larger particles.
14 Particle-laden macrophages may be cleared from the A region along a number of pathways.
15 As noted in Figure 7-3, this includes: cephalad transport via the mucociliary system after the cells
16 reach the distal terminus of the mucus blanket; movement within the interstitium to a lymphatic
17 channel; or perhaps traversing of the alveolar-capillary endothelium, directly entering the
18 bloodstream. Particles within the lymphatic system may be translocated to tracheobronchial
19 lymph nodes, which can become reservoirs of retained material. Particles subsequently reaching
20 the post-nodal lymphatic circulation will enter the blood. Once in the systemic circulation, these
21 particles, or transmigrated macrophages, can travel to extrapulmonary organs. Deposited
22 particles which are not ingested by alveolar macrophages may enter the interstitium, where they
23 are subject to phagocytosis by resident interstitial macrophages, and may travel to perivenous,
24 peribronchiolar or subpleural sites, where they become trapped, increasing particle burden. The
25 migration and grouping of particles and macrophages within the lungs can lead to the
26 redistribution of initially diffuse deposits into focal aggregates. Some particles or components
27 can bind to epithelial cell membranes or macromolecules, or other cell components, delaying
28 clearance from the lungs.
29 Churg and Brauer (1997), examined lung autopsy tissue from 10 never smokers from
30 Vancouver, Canada. They noted that the geometric mean particle diameter (GMPD) in lung
31 parenchymal tissue was 0.38 //m (og = 2.4). Ultrafines were less than 5% of the total retained
October 1999 7-25 DRAFT-DO NOT QUOTE OR CITE
-------
Deposited Particle
Phagocytosis by
Alveolar Macrophages
T
Movement within
Alveolar Lumen
Bronchiolar / Bronchial
Lumen *
Mucociliary Blanket
I
Gl Tract
Endocytosis by
->-Type I Alveolar
Epithelial Cells
Passage Through
Alveolar Epithelium
Interstitium
Blood
A
Lymphatic Channels
Lymph Nodes
Passage through
Pulmonary Capillary
Endothelium
A A
Phagocytosis by
Interstitial
Macrophages
Figure 7-3. Diagram of known and suspected clearance pathways for poorly soluble
particles depositing in the alveolar region.
Source: Modified from Schlesinger et al. (1997).
1 PM. Metal particles had a GMPD of 0.17//m and silicates 0.49//m. Ninety-six percent of
2 retained PM was less than 2.5 //m. This observation suggests that PM <2.5 //m may be of
3 appropriate concern for chronic PM effects.
4 Clearance by the absorptive mechanism involves dissolution in the alveolar surface fluid,
5 followed by transport through the epithelium and into the interstitium, and diffusion into the
6 lymph or blood. Although factors affecting the dissolution of deposited particles are poorly
7 understood, solubility is influenced by the particle's surface to volume ratio and other properties,
8 such as hydrophilicity and lipophilicity (Mercer, 1967; Morrow, 1973; Patten, 1996). Thus, as
9 noted, materials generally considered to be relatively insoluble may still have high dissolution
10 rates and short dissolution half-times if the particle size is small.
October 1999
7-26
DRAFT-DO NOT QUOTE OR CITE
-------
1 Some deposited particles may undergo dissolution in the acidic milieu of the
2 phagolysosomes after ingestion by macrophages, and such intracellular dissolution may be the
3 initial step in translocation from the lungs for these particles and for material associated with
4 these particles (Kreyling, 1992; Lundborg et al., 1985). Following dissolution, the material can
5 be absorbed into the blood. Dissolved materials may then leave the lungs at rates which are more
6 rapid than would be expected based upon an "expected" normal dissolution rate in lung fluid.
7
8 7.2.4.2 Clearance Kinetics
9 The kinetics of clearance has been reviewed in U. S. Environmental Protection Agency
10 (1996a) and in a number of monographs (e.g., Schlesinger et al., 1997). It will be discussed
11 briefly. The actual time frame over which clearance occurs affects the cumulative dose delivered
12 to the respiratory tract, as well as delivered to extrapulmonary organs.
13
14 ET Region
15 Mucus flow rates in the posterior nasal passages are highly nonuniform, but the median rate
16 in a healthy adult human is about 5 mm/min, resulting in a mean anterior to posterior transport
17 time of about 10-20 min for poorly soluble particles (Rutland and Cole, 1981; Stanley et al.,
18 1985). Particles deposited in the anterior portion of the nasal passages are cleared more slowly
19 by mucus transport, and are usually more effectively removed by sneezing, wiping, or nose
20 blowing (Fry and Black, 1973; Morrow, 1977).
21
22 TB Region
23 Mucus transport in the tracheobronchial tree occurs at different rates in different local
24 regions; the velocity of movement is fastest in the trachea, and it becomes progressively slower
25 in more distal airways. In healthy non-smoking humans, using non-invasive procedures and no
26 anesthesia, average tracheal mucus transport rates have been measured at 4.3 to 5.7 mm/min
27 (Yeates et al., 1975, 1981; Foster et al., 1980; Leikauf et al., 1981, 1984), while that in the main
28 bronchi has been measured at -2.4 mm/min (Foster et al., 1980). Estimates for human medium
29 bronchi range between 0.2-1.3 mm/min, while those in the most distal ciliated airways range
30 down to 0.001 mm/min (Morrow et al., 1967; Cuddihy and Yeh, 1988; Yeates and Aspin, 1978).
October 1999 7-27 DRAFT-DO NOT QUOTE OR CITE
-------
1 The total duration of bronchial clearance, or some other time parameter, is often used as an
2 index of mucociliary kinetics. While clearance from the TB region is generally rapid, there is
3 experimental evidence, discussed in U.S. Environmental Protection Agency (1996a) that a
4 fraction of material deposited in the TB region is retained much longer than the 24 h commonly
5 used as the outer range of clearance time for particles within this region (Stahlhofen et al.,
6 1986a,b; Scheuch and Stahlhofen, 1988; Smaldone et al., 1988). Some recent studies (Bennett
7 et al., 1998) described below continue to support the concept that TB regional clearance consists
8 of both a fast and a slow component.
9 Falk et al. (1997) studied clearance in healthy adults using monodisperse Teflon particles
10 (6.2 //m) inhaled at two flow rates. A considerable fraction (about 50%) of particles deposited in
11 small airways had not cleared within 24 h following exposure. These particles cleared with a
12 half time of 50 days. While the deposition sites of the particles were not confirmed
13 experimentally, calculations suggested these to be in the smaller ciliated airways. Camner et al.
14 (1997) also noted that clearance from the TB region was incomplete by 24 h post exposure, and
15 suggested that this may be due to incomplete clearance from bronchioles. Healthy adults inhaled
16 teflon particles (6, 8 and 10 //m) under low flow rates to maximize deposition in the small
17 ciliated airways. The investigators noted a decrease in 24 h retention with increasing particle
18 size, indicating a shift with increasing size toward either a smaller retained fraction, deposition
19 more proximally in the respiratory tract, or both. They calculated that a large fraction, perhaps as
20 high as 75%, of particles depositing in generations 12-16 was still retained at 24 h post-exposure.
21 The underlying sites and mechanisms of long-term TB retention in the smaller airways are
22 not known. Some proposals were presented in U.S. Environmental Protection Agency (1996a).
23 This slow clearing tracheobronchial compartment may be associated with those bronchioles
24 <1 mm in diameter (Lay et al., 1995). Based upon a study in which an adrenergic agonist was
25 used to stimulate mucus transport, so as to examine the role of mucociliary transport in the
26 bronchioles, it was found that clearance from the smaller airways was not influenced by the drug,
27 suggesting to the investigators that mechanisms other than mucociliary transport contributed to
28 clearance from this region (Svartengren et al., 1998). However, the issue of retention of large
29 fractions of tracheobronchial deposit is not resolved.
30 Long-term TB retention patterns are not uniform. There is an enhancement at bifurcation
31 regions (Radford and Martell, 1977; Henshaw and Fews, 1984; Cohen et al., 1988), the likely
October 1999 7-28 DRAFT-DO NOT QUOTE OR CITE
-------
1 result of both greater deposition and less effective mucus clearance within these areas. Thus,
2 doses calculated based upon uniform surface retention density may be misleading, especially if
3 the material is, lexicologically, slow acting.
4
5 A Region
6 Particles deposited in the A region generally are retained longer than those deposited in
7 airways cleared by mucociliary transport. There are limited data on alveolar clearance rates in
8 humans. Within any species, reported clearance rates vary widely due, in part, to different
9 properties of the particles used in the various studies. Furthermore, some chronic experimental
10 studies have employed high concentrations of poorly soluble particles, which may have interfered
11 with normal clearance mechanisms, resulting in clearance rates different from those that would
12 typically occur at lower exposure levels. Prolonged exposure to high particle concentrations is
13 associated with what is termed particle "overload". This is discussed in greater detail in
14 Section 7.2.5.
15 There are numerous pathways of A region clearance, and the utilization of these may
16 depend upon the nature of the particles being cleared. Little is known concerning relative rates
17 along specific pathways. Thus, generalizations about clearance kinetics are difficult to make.
18 Nevertheless, A region clearance is usually described as a multiphasic process, each phase
19 considered to represent removal by a different mechanism or pathway, and often characterized by
20 increased retention half-times following exposure.
21 The initial uptake of deposited particles by alveolar macrophages is very rapid, and
22 generally occurs within 24 h of deposition (Lehnert and Morrow, 1985; Naumann and
23 Schlesinger, 1986; Lay et al., 1998). The time for clearance of particle-laden alveolar
24 macrophages via the mucociliary system depends upon the site of uptake relative to the distal
25 terminus of the mucus blanket at the bronchiolar level. Furthermore, clearance pathways, and
26 subsequent kinetics, may depend to some extent upon particle size. For example, some smaller
27 ultrafine particles (perhaps < 0.02 //m) may be less effectively phagocytosed than are larger ones
28 (Oberdorster, 1993).
29 Uningested particles may penetrate into the interstitium within a few hours following
30 deposition. This transepithelial passage seems to increase as particle loading increases,
31 especially to a level above the "overload" point for increasing macrophage number (Ferin, 1977;
October 1999 7-29 DRAFT-DO NOT QUOTE OR CITE
-------
1 Adamson and Bowden, 1981). It may also be particle size dependent, since insoluble ultrafine
2 particles (<0.1 //m diameter) of low intrinsic toxicity show increased access to and greater
3 lymphatic uptake than do larger ones of the same material (Oberdorster et al., 1992). However,
4 ultrafine particles of different materials may not enter the interstitium to the same extent.
5 Similarly, a depression of phagocytic activity, a reduction in macrophage ability to migrate to
6 sites of deposition (Madl et al., 1998), or the deposition of large numbers of ultrafine particles
7 may increase the number of free particles in the alveoli, perhaps enhancing removal by other
8 routes. In any case, free particles may reach the lymph nodes, perhaps within a few days after
9 deposition (Lehnert et al., 1988; Harmsen et al., 1985), although this route is not certain and may
10 be species dependent. Furthermore, the extent of lymphatic uptake of particles may depend upon
11 the effectiveness of other clearance pathways, in that lymphatic translocation probably increases
12 when phagocytic activity of alveolar macrophages is decreased. This may be a factor in lung
13 overload. However, it seems that the deposited mass or number of particles must exceed some
14 threshold below which increases in loading do not affect translocation rate to the lymph nodes
15 (Ferin and Feldstein, 1978; LaBelle and Brieger, 1961). In addition, the rate of translocation to
16 the lymphatic system may be somewhat particle size dependent. Although no human data are
17 available, translocation of latex particles to the lymph nodes of rats was greater for 0.5 to 2 //m
18 particles than for 5 and 9 //m particles (Takahashi et al., 1992), and smaller particles within the
19 3 to 15 //m size range were found to be translocated at faster rates than were larger sizes (Snipes
20 and Clem, 1981). On the other hand, translocation to the lymph nodes was similar for both
21 0.4 //m barium sulfate or 0.02 //m gold colloid particles (Takahashi et al., 1987). It seems that
22 particles < 2 //m clear to the lymphatic system at a rate independent of size, and it is particles of
23 this size, rather than those > 5 //m, that would have significant deposition within the A region
24 following inhalation. In any case, the normal rate of translocation to the lymphatic system is
25 quite slow and elimination from the lymph nodes is even slower, with half-times estimated in
26 tens of years (Roy, 1989).
27 Soluble particles depositing in the A region may be rapidly cleared via absorption through
28 the epithelial surface into the blood. Actual rates depend upon the size of the particle (i.e., solute
29 size), with smaller molecular weight solutes clearing faster than larger ones. Absorption may be
30 considered as a two stage process, with the first stage dissociation of the deposited particles into
31 material that can be absorbed into the circulation (dissolution) and the second stage the uptake of
October 1999 7-30 DRAFT-DO NOT QUOTE OR CITE
-------
1 this material. Each of these stages may be time-dependent. The rate of dissolution depends upon
2 a number of factors, including particle surface area and chemical structure. A portion of the
3 dissolved material may be absorbed more slowly due to binding to respiratory tract components.
4 Accordingly, there is a very wide range for absorption rates depending upon the physicochemical
5 properties of the material deposited.
6
7 7.2.4.3 Interspecies Patterns of Clearance
8 The inability to study the retention of certain materials in humans for direct risk assessment
9 requires use of laboratory animals. Since dosimetry depends upon clearance rates and routes,
10 adequate toxicologic assessment necessitates that clearance kinetics in these animals be related to
11 those in humans. The basic mechanisms and overall patterns of clearance from the respiratory
12 tract are similar in humans and most other mammals. However, regional clearance rates can
13 show substantial variation between species, even for similar particles deposited under
14 comparable exposure conditions. This has been extensively reviewed (U.S. Environmental
15 Protection Agency, 1996a; Schlesinger et al, 1997; Snipes et al, 1989).
16 In general, there are species-dependent rate constants for various clearance pathways.
17 Differences in regional and total clearance rates between some species are a reflection of
18 differences in mechanical clearance processes. For example, the relative proportion of particles
19 cleared from the A region in the short and longer term phases differs between laboratory rodents
20 and larger mammals, with a greater percentage cleared in the faster phase in rodents. A recent
21 study (Oberdorster et al., 1997) showed interstrain differences in mice and rats in the handling of
22 particles by alveolar macrophages. Macrophages of B6C3F1 mice could not phagocytize 10 //m
23 particles, but those of C57 black/6Jmice did. In addition, the nonphagocytized 10 //m particles
24 were efficiently eliminated from the alveolar region, while previous work in rats found that these
25 large particles, after uptake by macrophages, were persistently retained (Snipes and Clem, 1981;
26 Oberdorster et al., 1992). The end result of interspecies differences in clearance for
27 consideration in assessing particle dosimetry is that the retention of deposited particles can differ
28 between species, and this may result in differences in response to similar particulate exposure
29 atmospheres.
30 Hsieh and Yu (1998) summarized the existing data on pulmonary clearance of inhaled,
31 poorly soluble particles in the rat, mouse, guinea pig, dog, monkey and human. Clearance at
October 1999 7-31 DRAFT-DO NOT QUOTE OR CITE
-------
1 different initial lung burdens, ranging from 0.001 - 10 mg particles/g lung, was analyzed using a
2 two phase exponential decay function. Two clearance phases in the alveolar region, namely fast
3 and slow, were associated with mechanical clearance along two pathways, the former with the
4 mucociliary system and the latter with the lymph nodes. Rats and mice were noted to be fast
5 clearers compared to the other species. Increasing the initial lung burden resulted in an
6 increasing mass fraction of particles cleared by the slower phase. As lung burden increased
7 beyond 1 mg particles/g lung, the fraction cleared by the slow phase increased to almost 100%
8 for all species. However, the rate for the fast phase was similar in all species and did not change
9 with increasing lung burden of particles, while the rate for the slow phase decreased with
10 increasing lung burden. At elevated burdens, the "overload" effect on clearance rate was greater
11 in rats than in humans, an observation consistent with previous findings (Snipes, 1989).
12
13 7.2.4.4 Biological Factors Modifying Clearance
14 A number of factors have been assessed in terms of modulation of normal clearance
15 patterns. These include aging, gender, workload, disease and irritant inhalation, and have been
16 discussed in detail previously (U.S. Environmental Protection Agency, 1996a).
17
18 Age
19 Studies previously described (U.S. Environmental Protection Agency, 1996a) indicated
20 there appeared to be no clear evidence for any age-related differences in clearance from the
21 respiratory tract, either from child to adult or adult to elderly. Studies of mucociliary function
22 have shown either no changes or some slowing in mucous clearance function with age after
23 maturity, but at a rate which would be unlikely to significantly affect overall clearance kinetics.
24
25 Gender
26 Previous studies (U.S. Environmental Protection Agency, 1996a) indicated no gender
27 related differences in nasal mucociliary clearance rates in children (Passali and Bianchini
28 Ciampoli, 1985) nor in tracheal transport rates in adults (Yeates et al., 1975).
29
30
31
October 1999 7-32 DRAFT-DO NOT QUOTE OR CITE
-------
1 Physical Activity
1 The effect of increased physical activity upon mucociliary clearance is unresolved, with
3 previously discussed studies (U.S. Environmental Protection Agency, 1996a) indicating either no
4 effect or an increased clearance rate with exercise. There are no data concerning changes in
5 A region clearance with increased activity levels. Breathing with an increased tidal volume was
6 noted to increase the rate of particle clearance from the A region, and this was suggested to be
7 due to distension related evacuation of surfactant into proximal airways, resulting in a facilitated
8 movement of particle-laden macrophages or uningested particles due to the accelerated motion of
9 the alveolar fluid film (John et al, 1994).
10
11 Respiratory Tract Disease
12 Various respiratory tract diseases are associated with clearance alterations. The
13 examination of clearance in individuals with lung disease requires careful interpretation of
14 results, since differences in deposition of particles used to assess clearance function may occur
15 between normal individuals and those with respiratory disease; this would directly impact upon
16 the measured clearance rates, especially in the tracheobronchial tree. Earlier studies reported in
17 U.S. Environmental Protection Agency (1996a) noted findings of slower nasal mucociliary
18 clearance in humans with chronic sinusitis, bronchiectasis, rhinitis, or cystic fibrosis, and slowed
19 bronchial mucus transport associated with bronchial carcinoma, chronic bronchitis, asthma and
20 various acute respiratory infections. However, a recent study by Svartengren et al. (1996a)
21 concluded, based upon deposition and clearance patterns, that particles cleared equally
22 effectively from the small ciliated airways of healthy humans and those with mild to moderate
23 asthma. However, this similarity was ascribed to effective therapy for the asthmatics.
24 In another study, Svartengren et al. (1996b) examined clearance from the TB region in
25 adults with chronic bronchitis who inhaled 6 //m Teflon particles. Based upon calculations,
26 particle deposition was assumed to be in small ciliated airways at low flow and in larger airways
27 at higher flow. The results were compared to that obtained in healthy subjects from other
28 studies. At low flow (resulting in small airway deposition), a larger fraction of particles was
29 retained over 72 h in people with chronic bronchitis compared to healthy subjects, indicating that
30 clearance due to spontaneous cough could not fully compensate for impaired mucociliary
31 transport in small airways. For larger airways, patients with chronic bronchitis cleared a larger
October 1999 7-33 DRAFT-DO NOT QUOTE OR CITE
-------
1 fraction of the deposited particles over 72 h than did healthy subjects, but this was reportedly due
2 to differences in deposition resulting from airway obstruction.
3 An important method of clearance from the tracheobronchial region, under some
4 circumstances, is cough. While cough is generally a reaction to an inhaled stimulus, in some
5 individuals with respiratory disease, spontaneous coughing also serves to clear the upper
6 bronchial airways of deposited substances by dislodging mucus from the airway surface. Recent
7 studies confirm that this mechanism likely plays a significant role in clearance for people with
8 mucus hypersecretion, at least for the region of the upper bronchial tree affected by cough, and
9 for a wide range of deposited particle sizes (0.5-5 //m) (Toms et al., 1997; Groth et al., 1997).
10 There appears to be a general trend towards an association between the extent, i.e., number, of
11 spontaneous coughs and the rate of particle clearance, with faster clearance associated with a
12 greater number of coughs (Groth et al., 1997). Thus, recent evidence continues to support cough
13 as an adjunct to mucociliary movement in the removal of particles from the lungs of individuals
14 with COPD. However, some recent evidence suggests that, like mucociliary function, cough
15 clearance may become depressed with worsening airway disease. Noone et al. (1999) found that
16 the efficacy of clearance via cough in patients with primary ciliary dyskinesia, who rely on
17 coughing for clearance due to immotile cilia, correlated with lung function (FEV1), in that
18 decreased cough clearance was associated with decreased percentage of predicted FEV1.
19 Earlier reported studies (U.S. Environmental Protection Agency, 1996a) indicated that rates
20 of A region particle clearance were reduced in humans with chronic obstructive lung disease and
21 in laboratory animals with viral infections, while the viability and functional activity of
22 macrophages was impaired in human asthmatics and in animals with viral induced lung
23 infections. However, any modification of functional properties of macrophages appears to be
24 injury specific, in that they reflect the nature and anatomic pattern of disease.
25 A factor which may affect clearance of particles is the integrity of the epithelial surface
26 lining of the lungs. Damage or injury to the epithelium may result from disease or from the
27 inhalation of chemical irritants. Earlier studies performed with particle instillation had shown
28 that alveolar epithelial damage at the time of deposition in mice resulted in increased
29 translocation of inert carbon to pulmonary interstitial macrophages (Adamson and Hedgecock,
30 1995). A similar response was observed in a recent assessment (Adamson and Prieditis, 1998),
31 whereby silica (<0.3 //m) was instilled into a lung having alveolar epithelial damage, as
October 1999 7-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 evidenced by increased permeability; retained particles were noted to reach the interstitium and
2 lymph nodes, thus increasing the translocation to the interstitium and increasing the extent of
3 activation of interstitial macrophages.
4
5 7.2.5 Particle Overload
6 Experimental studies using some laboratory rodents have employed high exposure
7 concentrations of relatively nontoxic, poorly soluble particles. These particle loads interfered
8 with normal clearance mechanisms, producing clearance rates different from those that would
9 occur at lower exposure levels. Prolonged exposure to high particle concentrations is associated
10 with a phenomenon that has been termed particle "overload", defined as the overwhelming of
11 macrophage-mediated clearance by the deposition of particles at a rate that exceeds the capacity
12 of that clearance pathway. It has been hypothesized that in the rat, overload will begin when
13 deposition approaches 1 mg particles/g lung tissue (Morrow, 1988). Overload is a nonspecific
14 effect noted in experimental studies using many different kinds of poorly soluble particles and
15 results in A region clearance slowing or stasis, with an associated chronic inflammation and
16 aggregation of macrophages in the lungs and increased translocation of particles into the
17 interstitium.
18 The relevance of lung overload to humans, and even to species other than laboratory rats,
19 remains unclear. While it is likely to be of little relevance for most "real world" ambient
20 exposures of humans, it may be of concern in interpreting some long-term experimental exposure
21 data and, perhaps, also for human occupational exposures. In addition, the relevance to humans
22 is clouded by the suggestion that macrophage-mediated clearance is normally slower and perhaps
23 of less relative importance in overall clearance in humans than in rats (Morrow, 1994), and that
24 there can be significant differences in macrophage loading between species.
25
26 7.2.6 Comparison of Deposition and Clearance Patterns of Particles
27 Administered by Inhalation and Intratracheal Instillation
28 The most relevant exposure route to evaluate the toxicity of particulate matter is inhalation.
29 However, many studies delivered particles by intratracheal instillation. This latter technique has
30 been used since it is easy to perform, and requires significantly less effort, cost and amount of
31 test material than does inhalation and can deliver a known, exact dose of a toxicant to the lungs.
October 1999 7-35 DRAFT-DO NOT QUOTE OR CITE
-------
1 Since particle disposition is a determinant of dose, it is important to compare deposition and
2 clearance of particles delivered by these two routes. However, in most instillation studies, the
3 effect of this route of administration upon particle deposition and clearance per se was not
4 examined. While these parameters were evaluated in some studies, it is very difficult to compare
5 particle deposition/clearance between different inhalation and instillation studies due to
6 differences in experimental procedures and in the manner by which particle deposition/clearance
7 was quantitated. Nevertheless, a few studies directly compared the two exposure techniques
8 upon deposition/clearance, and the results are summarized in this section.
9 The pattern of initial regional deposition is strongly influenced by the exposure technique
10 used. Furthermore, the patterns within specific respiratory tract regions are also influenced in
11 this regard. Depending upon particle size, inhalation results in varying degrees of deposition
12 within the upper respiratory tract, a region which is completely bypassed by instillation. Thus,
13 differences in amount of particles deposited in the lower airways will occur between the two
14 procedures (Leong et al., 1986; Oberdorster et al., 1980). Furthermore, inhalation tends to result
15 in greater variability in particle burden among animals than does instillation (Domes and
16 Valberg, 1992).
17 The specific exposure procedure also influences the intrapulmonary distribution of
18 particles. This would potentially affect routes and rates of ultimate clearance from the lungs and
19 dose delivered to specific sites within the respiratory tract or to extrapulmonary organs.
20 Intratracheal instillation tends to disperse particles fairly evenly within the tracheobronchial tree,
21 but can result in inhomogeneous distribution in the alveolar region, while inhalation tends to
22 produce a more homogeneous distribution throughout the major conducting airways as well as
23 the alveolar region (Leong et al., 1986, 1998). In addition, inhalation produced more uniform
24 distribution of particles within a specific lobe than did instillation which produced focally
25 increased particle burdens compared to inhalation (Pritchard et al., 1985; Driscoll et al., 1990,
26 1991). In one study, the intralobular distribution (i.e., homogeneity of distribution) with
27 instillation was about four fold less homogeneous than with inhalation (Pritchard et al., 1985).
28 Thus, inhalation results in a randomized distribution of particles within the lungs, while
29 intratracheal instillation produces an inhomogeneous distribution within the lungs. The
30 periphery of the lung receives little particle load and most of the particles are found in regions
31 that have a short path length from the major airways. Furthermore, inhalation results in greater
October 1999 7-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 deposition in apical areas of the lungs and less in basal areas, while intratracheal instillation
2 results in less apical than basal deposition (Brain et al., 1976). Some of the differences between
3 these two exposure techniques may be due to the positioning of the animal during the instillation
4 process, which was not the same as that during inhalation exposure. But, in general, instillation
5 produces less uniform deposition than inhalation, although the inhomogeneity of distribution of a
6 single instillation exposure could be reduced by multiple instillation exposures. Instillation
7 results in heavier and more centralized particle deposition, while particles delivered by inhalation
8 are more evenly and more widely distributed throughout the lungs.
9 Comparison of the kinetics of clearance of particles administered by instillation or
10 inhalation have shown similarities (Oberdorster et al., 1980, 1997; Dahl et al., 1983; Drew et al.,
11 1987), as well as differences (Pritchard et al., 1985; Muller et al., 1989), in rates for different
12 clearance phases, dependent upon the exposure technique used. However, some of the
13 differences in kinetics may be explained by differences in the initial sites of deposition (Driscoll
14 etal, 1990, 1991).
15 There are some data to allow assessment of differences in the pathways by which particles
16 delivered by different techniques may be cleared. For example, while particles delivered via
17 either intratracheal instillation or inhalation were transported via peribronchial lymphatics to
18 bronchus-associated lymphoid tissue (BALT), only those delivered by inhalation reached the
19 pleura (likely via pleural lymphatics) in sufficient amounts to produce a biological response
20 (granuloma) (Henderson et al., 1995). Furthermore, particles depositing within the lungs may be
21 subject to systemic absorption, the extent of which may depend upon the clearance rates and
22 routes. Thus, Chui et al. (1988) noted that the systemic absorption of particles delivered via
23 inhalation was somewhat greater than that for particles delivered by intratracheal instillation; the
24 half-time for elimination from blood was over twice as long for instillation as for inhalation,
25 suggesting that the route of administration can affect the rate and extent of systemic
26 bioavailability of inhaled particles.
27 One of the major pathways of clearance involves particle uptake and removal via
28 pulmonary macrophages. Domes and Valberg (1992) noted that inhalation resulted in a lower
29 percentage of particles recovered in lavaged cells and a more even distribution of particles among
30 macrophages. More individual cells received measurable amounts of particles via inhalation than
31 via intratracheal instillation, whereas with the latter, many cells received little or no particles
October 1999 7-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 while others received very high burdens. The distribution among macrophages was more
2 homogeneous with inhalation than with instillation. Furthermore, with intratracheal instillation,
3 macrophages at the lung periphery contained few if any particles, while cells in the regions of
4 highest deposition were overloaded, reflecting the inhomogeneity of particle distribution when
5 particles are administered via instillation (Pritchard et al., 1985). Thus, the route of exposure
6 influenced the particle distribution in the macrophage population and could, by assumption,
7 influence clearance pathways and clearance kinetics.
8 In conclusion, inhalation may result in deposition within the upper respiratory tract, the
9 extent of which depends upon the size of the particles used. Of course, intratracheal instillation
10 bypasses this portion of the respiratory tract and delivers particles directly to the tracheobronchial
11 tree. While some studies indicate that short (0-2d) and long (100 - 300 d post exposure) phases
12 of clearance of insoluble particles delivered either by inhalation or intratracheal instillation are
13 similar, other studies indicate that the percentage retention of particles delivered by instillation is
14 greater than that for inhalation, at least up to 30d post exposure. Thus, there is some
15 inconsistency in this regard. Perhaps the most consistent conclusion regarding differences
16 between inhalation and intratracheal instillation is related to the intrapulmonary distribution of
17 particles. Inhalation generally results in a fairly homogeneous distribution of particles
18 throughout the lungs. On the other hand, instillation results in an inhomogeneous distribution,
19 especially within the alveolar region, and focally high concentrations of particles. The bulk of
20 instilled material penetrates beyond the major tracheobronchial airways, but the lung periphery is
21 often virtually devoid of particles. This difference is reflected in particle burdens within
22 macrophages, with those from animals inhaling particles being burdened more homogeneously
23 and those from animals with instilled particles showing some populations of cells with no
24 particles and others with heavy burdens. This difference reflects upon clearance pathways, dose
25 to cells and tissues and systemic absorption. Exposure method, thus, clearly influences dose
26 distribution.
27
28 7.2.7 Modeling the Disposition of Particles in the Respiratory Tract
29 7.2.7.1 Modeling Deposition and Clearance
30 The biologic effects of inhaled particles are a function of their disposition. This, in turn,
31 depends on their patterns of both deposition and clearance. Removal of deposited materials
October 1999 7-38 DRAFT-DO NOT QUOTE OR CITE
-------
1 involves the competing processes of macrophage - mediated clearance and
2 dissolution-absorption. Over the years, mathematical models for predicting deposition, clearance
3 and, ultimately, retention of particles in the respiratory tract have been developed. Such models
4 help interpret experimental data and can be used to make predictions of deposition for cases
5 where data are not available.
6 A review of various mathematical deposition models was given by Morrow and Yu (1993)
7 and in U.S. Environmental Protection Agency (1996a). There are three major elements involved
8 in mathematical modeling. First, a structural model of the airways must be specified in
9 mathematical terms. Second, deposition efficiency in each airway must be derived for each of
10 the various deposition mechanisms. Finally, a computational procedure must be developed to
11 account for the transport and deposition of the particles in the airways. As noted earlier, most
12 models are deterministic, in that particle deposition probabilities are calculated using anatomical
13 and airflow information on an airway generation by airway generation basis. Other models are
14 stochastic, whereby modeling is performed using individual particle trajectories and finite
15 element simulations of airflow.
16 Recent reports involve modeling the deposition of ultrafine particles and deposition at
17 airway bifurcations. Zhang and Martonen (1997) used a mathematical model to simulate
18 diffusion deposition of ultrafine particles in the human upper tracheobronchial tree, and
19 compared the results to those in a hollow cast obtained by Cohen et al. (1990). The model was in
20 good agreement with experimental data. Zhang and Martonen (1997) studied the inertial
21 deposition of particles in symmetric three-dimensional models of airway bifurcations,
22 mathematically examining effects of geometry and flow. They developed equations for use in
23 predicting deposition based upon Stokes numbers, Reynolds numbers and bifurcation angles for
24 specific inflows.
25 Models for deposition, clearance, and dosimetry of the respiratory tract of humans have
26 been available for the past four decades. The International Commission on Radiological
27 Protection (ICRP) has recommended three different mathematical models during this time period
28 (International Commission on Radiological Protection, 1960, 1979, 1994). These models make
29 it possible to calculate the mass deposition and retention in different parts of the respiratory tract
30 and provide, if needed, mathematical descriptions of the translocation of portions of the
31 deposited material to other organs and tissues beyond the respiratory tract.
October 1999 7-39 DRAFT-DO NOT QUOTE OR CITE
-------
1 Another respiratory tract dosimetry model was developed, concurrently with the new ICRP
2 model, by the National Council on Radiation Protection and Measurements (NCRP) (1997).
3 As with the ICRP model (International Commission on Radiological Protection, 1994), the new
4 NCRP model addresses (1) inhalability of particles, (2) revised sub-regions of the respiratory
5 tract, (3) dissolution-absorption as an important aspect of the model, and (4) body size and age.
6 The NCRP model defines the respiratory tract in terms of a naso-oro-pharyngo-laryngeal (NOPL)
7 region, a tracheobronchial (TB) region, a pulmonary (P) region, and lung-associated lymph nodes
8 (LN). Deposition and clearance are calculated separately for each of these regions. As with the
9 1994 ICRP model, inhalability of aerosol particles is considered, and deposition in the various
10 regions of the respiratory tract is modeled using methods that relate to mechanisms of inertial
11 impaction, sedimentation, and diffusion.
12 Fractional deposition in the NOPL region was developed from empirical relationships
13 between particle diameter and air flow rate. Deposition in the TB and P regions were projected
14 from model calculations based upon geometric or aerodynamic particle diameter and physical
15 deposition mechanisms such as impaction, sedimentation, diffusion and interception. Deposition
16 in the TB and P regions used the lung model of Yeh and Schum (1980), with a method of
17 calculation similar to that of Findeisen (1935) and Landahl (1950). This method was modified so
18 as to be able to accomodate an adjustment of lung volume and substitution of realistic deposition
19 equations. These calculations were based upon air flow information and idealized morphometry,
20 using a typical pathway model. (Comparison of regional deposition fraction predictions between
21 the NCRP and ICRP models was given in U.S. Environmental Protection Agency [1996a]).
22 Inhalability was defined as per the American Conference of Governmental Industrial Hygenists
23 (1985) definition. Breathing frequency, tidal volume and functional residual capacity are the
24 ventilatory factors used to model deposition. These were related to body weight and to three
25 levels of physical activity, namely low activity, light exertion and heavy exertion.
26 Clearance from all regions of the respiratory tract was considered to result from
27 competitive mechanical and absorptive mechanisms. Mechanical clearance in the NOPL and TB
28 regions was considered to result from mucociliary transport. This was represented in the model
29 as a series of escalators moving towards the glottis and where each airway had an effective
30 clearance velocity. Clearance from the P region was represented by fractional daily clearance
31 rates to the TB region, the pulmonary LN region and the blood. A fundamental assumption in
October 1999 7-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 the model was that the rates for absorption into blood were the same in all regions of the
2 respiratory tract; the rates of dissolution-absorption of particles and their constituents were
3 derived from clearance data primarily from laboratory animals. The effect of body growth on
4 particle deposition was also considered in the model, but particle clearance rates were assumed to
5 be independent of age. Some consideration for compromised individuals was incorporated into
6 the model by altering rates (compared to normal) for the NOPL and TB regions.
7 Mathematical deposition models for deposition in a number of nonhuman species have
8 been developed and discussed previously (U.S. Environmental Protection Agency, 1996a).
9 Despite difficulties, modeling studies in laboratory animals remain a useful step in extrapolating
10 exposure-dose-response relationships from laboratory animals to human. Some additional work
11 on modeling deposition in animals has been reported, but it merely expands upon work and
12 approaches already reported (U.S. Environmental Protection Agency, 1996a).
13 Respiratory-tract clearance begins immediately upon deposition of inhaled particles. Given
14 sufficient time, the deposited particles may be completely removed by these clearance processes.
15 However, single inhalation exposures may be the exception rather than the rule. It is generally
16 accepted that repeated or chronic exposures are common for environmental aerosols. As a result
17 of such exposures, accumulation of particles may occur. Chronic exposures produce respiratory
18 tract burdens of inhaled particles that continue to increase with time until the rate of deposition is
19 balanced by the rate of clearance. This is defined as the "equilibrium respiratory tract burden".
20 It is important to evaluate these accumulation patterns, especially when assessing ambient
21 chronic exposures, because they dictate what the equilibrium respiratory tract burdens of inhaled
22 particles will be for a specified exposure atmosphere. Equivalent concentrations can be defined
23 as "species-dependent concentrations of airborne particles which, when chronically inhaled,
24 produce equal lung deposits of inhaled particles per gram of lung during a specified exposure
25 period" (Schlesinger et al., 1997). Available data and approaches to evaluate exposure
26 atmospheres that produce similar respiratory tract burdens in laboratory animals and humans
27 have been discussed in detail in the previous criteria document.
28 Several laboratory animal models have been developed to help interpret results from
29 specific studies that involved chronic inhalation exposures to non-radioactive particles (Wolff
30 et al., 1987; Strom et al., 1988; Stober et al., 1994). These models were adapted to data from
31 studies involving high level chronic inhalation exposures in which massive lung burdens of low
October 1999 7-41 DRAFT-DO NOT QUOTE OR CITE
-------
1 toxicity, poorly soluble particles were accumulated but the models have not been adapted to
2 chronic exposures to low concentrations of aerosols in which particle overload does not occur.
3
4 7.2.7.2 Models To Estimate Retained Dose
5 Models have routinely been used to express retained dose in terms of temporal patterns for
6 alveolar retention of acutely inhaled materials. Available information for a variety of
7 mammalian species and humans can be used to predict deposition patterns in the respiratory tract
8 for inhalable aerosols with reasonable degrees of accuracy. Additionally, alveolar clearance data
9 for mammalian species commonly used in inhalation studies are available from numerous
10 experiments that involved small amounts of inhaled radioactive particles.
11 A very important factor in using models to predict retention patterns in laboratory animals
12 or humans is the dissolution-absorption rate of the inhaled material. Factors that affect the
13 dissolution of materials or the leaching of their constituents in physiological fluids, and the
14 subsequent absorption of these constituents, are not fully understood. Solubility is known to be
15 influenced by the surface-to-volume ratio and other surface properties of particles (Mercer, 1967;
16 Morrow, 1973). The rates at which dissolution and absorption processes occur are influenced by
17 factors that include the chemical composition of the material. Temperature history of materials is
18 an important consideration for some metal oxides. For example, in controlled laboratory
19 environments, the solubility of oxides usually decreases when the oxides are produced at high
20 temperatures, which generally results in compact particles having small surface-to-volume ratios.
21 It is sometimes possible to accurately predict dissolution-absorption characteristics of materials
22 based on physical/chemical considerations. However, predictions for in vivo
23 dissolution-absorption rates for most materials, especially if they contain multivalent cations or
24 anions, should be confirmed experimentally.
25 Phagocytic cells, primarily macrophages, clearly play a role in dissolution-absorption of
26 particles retained in the respiratory tract (Kreyling, 1992). Some particles dissolve within the
27 phagosomes due to the acidic milieu in those organelles (Lundborg et al., 1984, 1985), but the
28 dissolved material may remain associated with the phagosomes or other organelles in the
29 macrophage rather than diffuse out of the macrophage to be absorbed and transported elsewhere
30 (Cuddihy, 1984). This same phenomenon has been reported for organic materials. For example,
31 covalent binding of benzo(a)pyrene or metabolites to cellular macromolecules resulted in an
October 1999 7-42 DRAFT-DO NOT QUOTE OR CITE
-------
1 increased alveolar retention time for that compound after inhalation exposures of rats (Medinsky
2 and Kampcik, 1985). Understanding these phenomena and recognizing species similarities and
3 differences are important for evaluating alveolar retention and clearance processes and
4 interpreting results of inhalation studies.
5 Dissolution-absorption of materials in the respiratory tract is clearly dependent on the
6 chemical and physical attributes of the material. While it is possible to predict rates of
7 dissolution-absorption, it is prudent to experimentally determine this important clearance
8 parameter. It is important to understand the impact of this clearance process for the lung, TLNs,
9 and other body organs that might receive particles, or their constituents that enter the circulatory
10 system from the lung.
11 Insufficient data were available to adequately model long-term retention of particles
12 deposited in the conducting airways of any mammalian species at the time of the previous
13 document, and this remains the case. Additional research must be done to provide the
14 information needed to properly evaluate retention of particles in conducting airways.
15 However, a number of earlier studies discussed in the previous document and in
16 Section 7.2.2.2 herein noted that some particles were retained for relatively long times in the
17 upper respiratory tract and tracheobronchial regions, effectively contradicting the general
18 conclusion that almost all inhaled particles that deposit in the TB region clear within hours or
19 days. These studies have demonstrated that variable portions of the particles that deposit in, or
20 are cleared through, the TB region are retained with half-times on the order of weeks or months.
21 Long-term retention and clearance patterns for particles that deposit in the head airways and TB
22 region must continue to be thoroughly evaluated because of the implications of this information
23 for respiratory tract dosimetry and risk assessment.
24 Model projections are possible for the A region using the cumulative information in the
25 scientific literature relevant to deposition, retention, and clearance of inhaled particles.
26 Clearance parameters for six laboratory animal species were summarized in U.S. Environmental
27 Protection Agency (1996a). Recently, Nikula et al. (1997) evaluated results when rats were
28 exposed to high levels of either diesel soot or coal dust. While the amount of retained material
29 was similar in both species, the rats retained a greater portion in the lumens of the alveolar ducts
30 and alveoli than did monkeys, while the monkeys retained a greater portion of the material in the
31 interstitium than did rats. The investigators concluded that intrapulmonary retention patterns in
October 1999 7-43 DRAFT-DO NOT QUOTE OR CITE
-------
1 one species may not be predictive of those in another species at high levels of exposure, but this
2 may not be the case at lower levels.
3
4
5 7.3 TOXICOLOGY OF PARTICULATE MATTER
6 Due to the lack of data on actual ambient PM toxicology at the time the previous CD was
7 compiled, the respiratory effects of PM were organized into specific chemical components of
8 ambient PM or model "surrogate" particles, e.g. acid aerosols, metals, ultrafine particles,
9 bioaerosols, and "other particle matter". The following section summarizes the conclusions of
10 the 1996 CD for each of these components.
11 There are many new studies of combustion related particles. The reasons for this increased
12 interest in combustion particles are that these particles are typically the dominant sources
13 represented in the fine fraction of PM. The combined evaluation of the health effects and the
14 physico-chemical properties of combustion PM will be useful in relating emission sources with
15 health effects.
16
17 7.3.1 Summary of Previous Criteria Document
18 7.3.1.1 Acid Aerosols
19 EPA previously concluded that healthy subjects experienced no decrements in lung
20 function and only mild lower respiratory symptoms following single exposures to sulfuric acid
21 aerosol at exposure concentrations in the mg/m3 range, even with exercise and the use of acidic
22 oral rinses to minimize neutralization by oral ammonia (U.S. Environmental Protection Agency,
23 1996a). Acid aerosols do alter mucociliary clearance in healthy subjects at lower concentrations,
24 with effects dependent on exposure concentration and the region of the lung being studied.
25 Asthmatic subjects appear to be more sensitive than healthy subjects to the effects of acid
26 aerosols on lung function, but the effective concentration differs widely among studies.
27 Adolescent asthmatics may be more sensitive than adults, and may experience small decrements
28 in lung function in response to H2SO4 at exposure levels only slightly above peak ambient levels.
29 In a very limited number of studies, the elderly and individuals with chronic obstructive
October 1999 7-44 DRAFT-DO NOT QUOTE OR CITE
-------
1 pulmonary disease do not appear to be particularly susceptible to the effects of acid aerosols on
2 lung function.
3 The adverse respiratory effects of acid aerosols in humans were supported by laboratory
4 animal studies of H2SO4 and other acidic sulfates. The available evidence indicates that the
5 observed responses to these are likely due to H+ rather than to SO4=. Acidic sulfates exert their
6 action throughout the respiratory tract, with the response and location of effect dependent upon
7 particle size and mass and number concentration.
8 Both acute and chronic exposure to H2SO4 at well below lethal levels can produce
9 functional changes in the respiratory tract of animals. Acute exposure will alter pulmonary
10 function, largely due to bronchoconstrictive action. However, attempts to produce changes in
11 airway resistance in healthy animals at levels below 1,000 //g/m3 have been largely unsuccessful,
12 except in the guinea pig. In general, the smaller size droplets (submicron) were more effective in
13 altering pulmonary function, especially at low concentrations. Very low concentrations
14 (< 100 //g/m3) of acid-coated ultrafine particles are associated with lung function and diffusion
15 decrements, as well as changes in airway responsiveness. Chronic exposure to H2SO4 is also
16 associated with alterations in pulmonary function (e.g., changes in the distribution of ventilation
17 and in respiratory rate in monkeys). But, in these cases, the effective concentrations are
18 >500 //g/m3. Airway hyperresponsiveness has been induced with repeated exposures to
19 250 //g/m3 H2SO4 in rabbits, and has been suggested to occur following single exposures at
20 75 //g/m3.
21 Lung defenses, such as resistance to bacterial infection, may be altered by acute exposure to
22 concentrations of H2SO4 around 1,000 //g/m3. However, the bronchial mucociliary clearance
23 system is very sensitive to inhaled acids; fairly low levels of H2SO4 produce alterations in
24 mucociliary transport rates in healthy animals. The lowest level shown to have such an effect,
25 100 //g/m3 with repeated exposures in rabbits, is well below that which results in other
26 physiological changes in most experimental animals. Furthermore, exposures to somewhat
27 higher levels that also alter clearance have been associated with various morphometric changes in
28 the bronchial tree that are suggestive of hypersecretion.
29 Limited data also suggest that exposure to acid aerosols may affect the functioning of AMs.
30 The lowest level examined in this regard to date is 500 //g/m3 H2SO4. Alveolar region particle
31 clearance is affected by repeated H2SO4 exposures to as low as 125 //g/m3.
October 1999 7-45 DRAFT-DO NOT QUOTE OR CITE
-------
1 U.S. Environmental Protection Agency (1996a) also reported results of studies which
2 examined potential interactions of acid sulfates with other air pollutants. Such interactions may
3 be antagonistic, additive, or synergistic. Evidence for interactive effects may depend upon the
4 sequence of exposure as well as on the endpoint examined. Low levels of H2SO4 (100 //g/m3)
5 have been shown to react synergistically with O3 in simultaneous exposures using biochemical
6 endpoints. In this case, the H2SO4 enhanced the damage due to the O3. The most realistic
7 exposures were to multicomponent atmospheres, but the results of these studies are often
8 difficult to assess due to chemical interactions of components and a resultant lack of precise
9 control over the composition of the exposure environment.
10
11 7.3.1.2 Metals
12 Data from occupational studies and laboratory animal studies indicate that acute exposures
13 to very high levels (hundreds of//g/m3 or more) or chronic exposures to lower levels (up to
14 15 //g/m3 albeit high compared to ambient levels) of metallic particulates can have an effect on
15 the respiratory tract. However, it was concluded that the metals at concentrations present in the
16 ambient atmosphere (1 to 14 //g/m3) were not likely to have a significant acute effect in healthy
17 individuals. These metals include arsenic, cadmium, copper, vanadium, iron, and zinc. Other
18 metals found at concentrations less than 0.5 //g/m3 were not reviewed in the previous CD.
19
20 7.3.1.3 Ultrafme Particles
21 There were only limited data available from human studies or laboratory animal studies on
22 ultrafine aerosols at the time of the release of the previous CD. In vitro studies have shown that
23 ultrafine particles have the capacity to cause injury to cells of the respiratory tract. High levels of
24 ultrafine particles, as metal or polymer "fume," are associated with toxic respiratory responses in
25 humans and other mammals. Such exposures are associated with cough, dyspnea, pulmonary
26 edema, and acute inflammation. At concentrations less than 50 //g/m3, freshly generated
27 insoluble ultrafine teflon polymer fume particles can be severely toxic to the lung. However it
28 was not clear what role in the observed effects was played by fume gases which adhered to the
29 particles. Thus it was not clear at the time of the previous review (U.S. Environmental
30 Protection Agency, 1996a) what role, if any, ambient ultrafine particles may play in PM-induced
31 mortality/morbidity.
October 1999 7-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.3.1.4 Bioaerosols
2 Ambient bioaerosols include fungal spores, pollen, bacteria, viruses, endotoxins, and plant
3 and animal debris. Such biological aerosols can produce various health effects including:
4 infections, hypersensitivity, and toxicoses. Bioaerosols present in the ambient environment have
5 the potential to cause disease in humans under certain conditions. However, it was concluded that
6 bioaerosols, at the concentrations present in the ambient environment, would not account for the
7 observed effects of particulate matter on human mortality and morbidity reported in PM
8 epidemiological studies. Moreover, bioaerosols generally represent a rather small fraction of the
9 measured urban ambient PM mass and are typically present even at lower concentrations during
10 the winter months when notable ambient PM effects have been demonstrated. Bioaerosols tend
11 to be in the coarse fraction of PM, but some bioaerosols are found in the fine fraction.
12
13 7.3.1.5 "Other Particulate Matter"
14 Toxicologic studies of other particulate matter species were discussed in the previous CD.
15 These studies included exposures to fly ash, volcanic ash, coal dust, carbon black, TiO2, and
16 miscellaneous other particles, either alone or in mixture. Some of the particles discussed were
17 considered to be models of "nuisance" or "inert" dusts (i.e., those having low intrinsic toxicity)
18 and were used in instillation studies to delineate nonspecific particle effects from effects of
19 known toxicants. A number of studies on "Other PM" examined effects of up to 50,000 //g/m3 of
20 respirable particles with inherently low toxicity. While there was no mortality, some mild
21 pulmonary function changes after exposure to 5,000 to 10,000 //g/m3 of inert particles were
22 observed in rats and guinea pigs. Lung morphology studies revealed focal inflammatory
23 responses, some epithelial hyperplasia, and fibrotic responses after exposure to >5,000 //g/m3.
24 Changes in macrophage clearance after exposure to > 10,000 //g/m3 were equivocal (no
25 infectivity effects). In studies of mixtures of particles and other pollutants, effects were variable
26 depending on the toxicity of the associated pollutant. In humans, co-exposure to carbon particles
27 appeared to increase responses to formaldehyde but not to acid aerosol. None of the "other"
28 particles mentioned above are present in ambient air in more than trace quantities. Thus, it was
29 concluded that the relevance of any of these studies to ambient particulate standard setting may
30 be extremely limited.
31
October 1999 7-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.3.2 Respiratory Effects of Particles
2 The following section (7.3.3) assesses results of exposure to various types of PM in
3 humans. Section 7.3.4 discusses controlled animal toxicology studies as well as in vitro studies
4 using animal or human respiratory cells. It focuses on those studies published since the 1996 Air
5 Quality Criteria Document for Particulate Matter (U.S. Environmental Protection Agency,
6 1996a).
7 The biological responses occurring in the respiratory tract following controlled PM
8 inhalation encompass a continuum of changes, including changes in pulmonary function,
9 pulmonary inflammation, and systemic effects. The observed responses are dependent on the
10 physicochemical characteristics of the PM, the total exposure and the health status of the host.
11 However, many of the responses are usually seen only at higher level exposures characteristic of
12 occupational and laboratory animal studies and not at (typically much lower) ambient particle
13 concentrations.
14 Particulate matter is a broad term that encompasses thousands of chemical species, many of
15 which have not been investigated in controlled laboratory animal or human studies. However, a
16 full discussion of all types of particles that have been studied is beyond the scope of this chapter.
17 Thus, specific criteria were used to select topics for presentation. High priority was placed on
18 studies that may: (1) elucidate health effects of major common constituents of ambient PM
19 and/or (2) contribute to enhanced understanding of the epidemiological studies (e.g., use of
20 ambient particles, "surrogate" particles, or particles with low inherent toxicity that may cause
21 effects due to their physicochemical characteristics, such as their size and composition).
22 Diesel exhaust particles (DPM) generally fit the criteria; but, because they are described
23 elsewhere in great detail (U. S. Environmental Protection Agency, 1999; Health Effects Institute,
24 1995), they are not covered in this chapter except in the discussions of their immunological
25 effects. Particles with high inherent toxicity, such as silica and asbestos, that are of concern
26 primarily because of occupational exposure, are also excluded from this chapter and are
27 discussed in detail elsewhere (U.S. Environmental Protection Agency, 1996b; Gift and Faust,
28 1997). Most of the laboratory animal studies summarized here have used high particulate mass
29 concentrations administered by inhalation, compared to ambient levels, even when laboratory
30 animal-to-human dosimetric differences or high doses by intratracheal instillation are considered.
October 1999 7-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 This raises a question about the relevance, for example, of a rat study at 5,000 //g/m3 in terms of
2 direct extrapolation to humans in ambient exposure scenarios.
3 As mentioned earlier, the data available in the previous Criteria Document were from
4 studies that investigated the respiratory effects of specific components of ambient PM or
5 surrogate particles. More recently, pulmonary effects upon controlled exposures to ambient PM
6 have been investigated by the use of aerosol concentrators (Sioutas et al., 1995; Gordon et al.,
7 1998). These concentrators are capable of exposing animals or humans to PM concentrations
8 that are up to 90-fold higher than ambient PM levels and have been used to investigate the effects
9 of ambient PM in normal and compromised animals and humans.
10
11 7.3.3 Effects in Healthy Humans
12 7.3.3.1 Human Acid Aerosol Exposure Studies
13 There have been extensive studies of the effects of controlled exposures to aqueous acid
14 aerosols on various aspects of lung function in humans. Many of these studies were reviewed in
15 U.S. Environmental Protection Agency (1996a) and in the Acid Aerosol Issue Paper (U.S.
16 Environmental Protection Agency, 1989). Methodology and measurement methods for
17 controlled human exposure studies have been reviewed elsewhere (Folinsbee et al., 1997).
18 These studies have illustrated that aqueous acidic aerosols have minimal effects on
19 symptoms and mechanical lung function in young healthy adult volunteers at concentrations as
20 high as 2000 //g/m3. The findings include minimal changes in lung function accompanied by
21 only mild lower respiratory symptoms. However at concentrations as low as 100 //g/m3, acid
22 aerosols can alter mucociliary clearance. Brief exposures (< 1 h) to low concentrations
23 (-100 //g/m3) may accelerate clearance while longer (multihour) exposures to higher
24 concentrations (> 100 //g/m3) can cause depression of clearance. Gulp et al. (1995) examined the
25 effects of inhaled acid aerosols on mucus secretion and composition; they found no significant
26 changes in mucin glycoproteins, collected via B AL at 18 h postexposure, as a result of an acute
27 2 h exposure to approximately 1000 //g/m3.
28 Asthmatic subjects appear to be more sensitive to the effects of acidic aerosols on
29 mechanical lung function. Responses have been reported in adolescent asthmatics at
30 concentrations as low as 68 //g/m3 and modest bronchoconstriction has been seen in adult
31 asthmatics exposed to concentrations >400 //g/m3. The effects of acidic aerosols are most likely
October 1999 7-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 related to decreases in local airway surface pH as a result of acid deposition. Airway ammonia
2 can neutralize a portion of the inhaled acid aerosol and the airway surface fluids have the
3 capacity to buffer the deposited acid. The variability of the response to acid aerosols among
4 asthmatics is at least partially explained by the variability of the disease itself; variability in acid
5 neutralization by oral/airway ammonia may also be a factor.
6 The inhalation of submicron acid particles appears to have greater effects on spirometry
7 and airway resistance than does inhalation of large particles characteristic of fog droplets
8 (10-20 //m). Leduc et al. (1995) exposed two groups of asthmatics to acid "fog" droplets
9 (7-9 //m MMAD; sulfuric acid or ammonia sulfate aerosol) and found no evidence of
10 bronchoconstriction or change in airway responsiveness to methacholine. Although there is
11 some evidence of increased airway responsiveness after acid aerosol exposure (Utell et al.,
12 1983a) as well as enhanced responses to ozone after acid aerosol inhalation in asthmatic subjects
13 (Linn et al., 1994; Frampton et al., 1995) the weight of evidence indicates that low concentrations
14 of acid aerosols (<200 //g/m3) typically do not tend to change airway responsiveness.
15 Acid aerosol exposure in humans (1000 //g/m3) did not result in airway inflammation
16 (Frampton et al., 1992) and there was no evidence of altered macrophage host defenses. Zelikoff
17 et al. (1997) compared the responses of rabbits and humans exposed to similar concentrations of
18 acid aerosol (-1000 //g/m3, 0.8 to 0.9 //m MMAD, 3 h duration). For both rabbits and humans
19 there was no evidence of PMN infiltration into the lung and no change in BAL protein level,
20 although there was an increase in LDH in rabbits but not in humans. Macrophages showed less
21 antimicrobial activity in rabbits; insufficient data were available for humans. Macrophage
22 phagocytic activity was slightly reduced in rabbits but not in humans. Superoxide production by
23 macrophages was somewhat depressed in both species.
24
25 Human Exposure to Other Particles
26 Only limited controlled human exposure studies have been performed with particles other
27 than acid aerosols. Metal fume fever has long been recognized in workers occupationally
28 exposed to metal fume (Mueller and Seger, 1985). The associated fever appears to be related to
29 pulmonary inflammation subsequent to inhalation of metal particles, most commonly zinc oxide.
30 In BAL samples taken from welders, marked elevation of PMNs was noted (Blanc et al., 1991).
31 This inflammation was associated with increased TNF-cc and IL-8, which may be responsible for
October 1999 7-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 the chemotaxis for PMNs (Blanc et al., 1993). Controlled exposure studies to high
2 concentrations of two different metal fumes, MgO and ZnO, demonstrate the differences in
3 response based on particle metal composition (Kushner et al., 1997). Up to 6400 mg/m3* min
4 cumulative dose (100-200 mg/m3 for 45 min; 99% of particles <1.8 //m and 29% <0.1 //m) of
5 MgO had no effect on lung function (spirometry, DLCO), symptoms of metal fume fever, or
6 changes in inflammatory mediators or cells recovered by BAL. However, lower concentrations
7 of ZnO fume (165-1110 mg/rn3- min; 15-120 min at 3-37 mg/m3; MMD 0.17 //m) induced a
8 neutrophilic inflammatory response in the airways 20 h post-exposure. Lavage fluid PMNs,
9 TNF-cc, and IL-8 were increased by ZnO exposure. However, the concentrations used in these
10 exposure studies exceed ambient levels by more than 1000-fold. The absence of a response to an
11 almost 10-fold higher concentration of MgO compared with ZnO indicates that metal
12 composition is an important factor in response to inhaled PM. Fine et al. (1997) have shown
13 elevated body temperature (metal fume fever) and increased levels of plasma IL-6 (from 2.9 to
14 6.4 pg/ml) in naive subjects exposed to the 8h TLV concentration of ZnO of 5 mg/m3 for 2 h.
15 Several metals have been shown to stimulate cytokine release in cultured human pulmonary
16 cells including zinc, chromium, cobalt, and vanadium. Boiler makers, exposed occupationally to
17 approximately 400-500 //g/m3 of fuel oil ash, showed acute nasal inflammatory responses
18 characterized by increased PMNs and elevated IL-8 which were associated with vanadium levels
19 (increased about 9-fold) in the upper airway (Woodin et al., 1998). Irsigler et al. (1999) reported
20 that V2O5 can induce asthma and bronchial hyperreactivity in exposed workers. A comparison of
21 autopsy cases in Mexico City from the 1950s with the 1980s indicated substantially higher levels
22 of (5-20 fold) Cd, Co, Cu, Ni, and Pb in lung tissue from the 1980s (Fortoul et al., 1996). Similar
23 studies have examined metal content in human blood and lung tissue (Tsuchiyama et al., 1997;
24 Osman et al., 1998). These data support the hypothesis that certain particulate metals may play a
25 role in the effects of inhaled ambient PM.
26 Because iron is the most abundant of the elements, which are capable of catalyzing oxidant
27 generation, and present in ambient urban particles, Lay and colleagues examined the cellular and
28 biochemical response of human subjects instilled with iron (III) oxide via the intrapulmonary
29 route (Lay et al., 1998). Subjects underwent bronchoalveolar lavage at 1 to 91 days after
30 instillation of 2.6 //m diameter iron oxide particles. The investigators found the greatest iron
31 oxide-induced inflammatory response in the alveolar fraction of the lavage fluid, although a
October 1999 7-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 significant increase in neutrophils was also observed in the bronchial fraction. The peak
2 response for all cellular and biochemical changes occurred at 1 day post-instillation. The same
3 iron oxide preparation, which contained a small amount of soluble iron, instilled in rats produced
4 similar pulmonary changes. Instillation of rats with 2 iron oxide preparations that contained no
5 soluble iron did not produce injury or inflammation, thus suggesting that soluble iron was
6 responsible for the observed intrapulmonary changes. It is not clear, however, whether the dose
7 of iron oxide delivered acutely to the lingular subsegment of the human subjects (approximately
8 5 mg or 2.1 x 108 particles) would be relevant to deposition of iron oxide particles at the
9 concentrations of iron present in ambient urban air (generally less than 1 //g/m3).
10
11 Human Exposures to Endotoxin
12 Endotoxin exposure in pig farmers is associated with the farmers' large annual decline in
13 FEVj (mean of 73 ml/yr), which is about 2-3 times more rapid than in healthy adults (Vogelzang
14 et al., 1998). Michel et al. (1997) examined the dose-response relationship to inhaled
15 lipopolysaccharide (LPS: the purified derivative of endotoxin) in normal healthy volunteers
16 exposed to 0, 0.5, 5, and 50 //g of LPS. Inhalation of 5 or 50 //g of LPS resulted in increased
17 PMNs in blood and sputum samples. At the higher concentration, a slight (3%) but not
18 significant decrease in FEVj was observed. Cormier et al. (1998) reported an approximate 10%
19 decline in FEVj and an increase in methacholine airway responsiveness after a 5 h exposure
20 inside a swine containment building. This exposure induced significant neutrophilic
21 inflammation in both the nose and the lung. Although these exposures are massive compared to
22 endotoxin levels in ambient PM in U.S. cities, these studies serve to illustrate the effects of
23 endotoxin and associated bioaerosol material in healthy non-sensitized individuals.
24 Adverse health effects have been observed after occupational exposure to complex aerosols
25 containing endotoxin at concentrations relevant to ambient levels. Zock et al. (1998) reported a
26 decline in FEVj (~3%) across a shift in a potato processing plant with up to 56 endotoxin units
27 (EU)/m3 in the air. Rose et al. (1998) reported a high incidence (65%) of BAL lymphocytes in
28 lifeguards working at a swimming pool where endotoxin levels in the air were on the order of
29 28 EU/m3. While these latter two studies may point towards pulmonary changes at low
30 concentrations of airborne endotoxin, it is not possible to rule out the contribution of other agents
31 in these complex organic aerosols.
October 1999 7-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.3.4 Effects in Healthy Animals
2 7.3.4.1 Ambient Particles
3 The majority of the in vivo exposures to ambient particles have utilized intratracheal
4 instillation techniques. A discussions on the pros and cons of this technique are covered in
5 Section 7.2.6 and these issues have also been reviewed (Oberdorster et al., 1997; Osier and
6 Oberdorster, 1997). The doses used in these instillation studies are generally high relative to
7 ambient concentrations, even when laboratory animal-to-human dosimetric differences are
8 considered. Therefore, in terms of direct extrapolation to humans in ambient exposure scenarios,
9 greater importance should be place on inhalation studies. Table 7-2 outlines studies in which
10 various biological endpoints were measured following intratracheal instillation of ambient PM,
11 complex combustion related PM, as well as laboratory derived surrogate PM.
12 In most of these studies, PM samples were collected on filters, resuspended in a vehicle
13 (usually saline) and a small volume of the suspension was intratracheally instilled into the
14 animals. Various inflammatory indices were evaluated at several time points post-instillation.
15 In some studies, the responses in animals to the soluble (leachate) or insoluble components were
16 measured. Costa and Dreher (1997) investigated the health effects of PM samples from three
17 emission sources (two oil and one coal fly ash) and four ambient airsheds (St Louis, MO;
18 Washington, B.C.; Dusseldorf, Germany; and Ottawa, Canada). PM was administered to rats by
19 intratracheal instillation in doses that were either equivalent in total mass or in total mass of
20 metal. Increases in neutrophils and eosinophils were noted at 24 h for all types of emission
21 source and for ambient PM. Biomarkers of permeability (total protein, albumin) and cellular
22 injury (LDH) were also increased. The results of this study indicate that the lung dose of
23 bioavailable transition metal, not instilled PM mass, was the primary determinant of the acute
24 inflammatory response. Kennedy et al. (1998) observed a dose-dependent inflammation (i.e.,
25 increase in protein and PMN in lavage fluid, proliferation of bronchiolar epithelium, and
26 intra-alveolar hemorrhage) in rats instilled with particles (TSP) collected in Provo, Utah.
27 Treatment with anti-cytokine-induced neutrophil-chemoattractant (CINC) antibody, at the same
28 time as TSP, blocked the airway inflammation suggesting CINC's role in the PMN-induced
29 responses. In addition, by comparing the responses of cultured BEAS-2B cells exposed to Provo
30 particles or its components (copper, lead, zinc and iron), these authors suggested that copper ions
31 may contribute to the biological effects.
October 1999 7-53 DRAFT-DO NOT QUOTE OR CITE
-------
o
r-K
o
a'
3
VO
VO
VO
Species, Gender,
Strain, Age, etc.
M Syrian golden
hamsters
90-125g.
Particle
Kuwaiti Oil
Fire particles
Urban
particles from
St. Louis, MO
Exposure
Technique Concentration
Intratracheal 0.15, 0.75, and 3.75
Instillation mg/lOOg
Particle Size (//m)
Oil fire particles: ^3.5 ,um,
10 days of 24 h samples
(Apr 30 to May 9, 1991),
in Ahmadi, Kuwait
Exposure Duration
Sacrificed 1 and 7
days post
instillation
Effect of Particles
Increases in PMN, AM, albumin, LDH,
myeloperoxidase, and P— N-
acetylglucosaminidase;
acute toxicity of the particles found in
the smoke from the Kuwaiti oil fires is
comparable to that of urban particles.
Reference
Brain et al.
(1998)
NMRI mouse
H
O
O
2
O
H
O
c
o
H
W
O
&
O
HH
H
W
CFA
CMP
WC
Intratracheal
instillation
NHBE cells
ROFA
In vitro
CMP: 20 ,ug arsenic/kg; N/A
or CMP
100 mg particles/kg;
WC alone (100 mg/kg),
CFA alone (100 mg/kg,
i.e., 20 //g arsenic/kg),
CMP mixed with WC
(CMP, 13.6 mg/kg, i.e.,
20 //g arsenic/kg; WC,
86.4 mg/kg) and
Ca3(AsO4)2 mixed with
WC (20 ,tig arsenic/kg;
WC, 100 mg/kg)
0, 50, 200
1, 5, 30 days post Mild inflammation for WC; Broeckaert
treatment, lavage Ca3(AsO4)2 caused significant et al. (1997)
for total protein inflammation;
content, CMP caused severe but transient
inflammatory cell inflammation;
number and type, CFA caused persistent alveolitis;
and TNF-a Cytokine production was upregulated
production in WC- and Ca3(AsO4) treated animals
particle retention after 6 and 30 d, respectively;
A 90% inhibition of TNF-a production
was still observed at day 30 after
administration of CMP and CFA;
a significant fraction persisted (10-15%
of the arsenic administered) in the lung
of CMP- and CFA-treated mice at day
30. Suppression of TNF-a production
is dependent upon the slow elimination
of the particles and their metal content
from the lung
Analysis at 2 and Increase in expression of the cytokines Carter et al.
24 h post IL-6, IL-8 and TNF-a; (1997)
Inhibition by DMTU or deferoxamine
Male S-D rats
200-225 g.
control and SO2-
treated
Concentrated
Ambient
Particles
(Boston)
(CAPS)
Harvard/EPA
Fine Particle
Concentrator
Animals
restrained in
chamber
206,733, 607 jj.g/m3 for 0.18 urn
days 1-3; 29 °C, 59% 5g = 2.9
RH
5h/day for 3 days PEF and TV increased in CAPS
exposed animals. Increased protein
and % neutrophils and lymphocytes in
lavage fluid after CAPS exposure.
Responses were greater in SO2-
bronchitis animals. No changes in
LDH. No deaths occurred.
Clarke et al.
(1999)
-------
o
o
*— K
o
cr
CT
VO
*o
VO
TABLE 7-2 (cont'd). RESPIRATORY EFFECTS OF PARTICIPATE MATTER
Species, Gender,
Strain, Age, etc. Particle
Rats, M, SD, Emission
60 days old source PM
MCT Ambient
(60/mg/kg), i.p airshed PM
ROFA
Exposure
Technique Concentration
Intratracheal Total mass:
instillation 2.5 mg/rat
Total transition
metal: 46 ,ug/rat
Particle Size (,um) Exposure Duration Effect of Particles
Emission PM: Analysis at 24 & Increases in PMNs, albumin, LDH,
1.78-4. 17 ,um 96 h following PMN, and eosinophils following
instillation. exposure to emission and ambient
Ambient PM: 3.27-4.09 /J,m particles;
Induction of injury by emission and
ambient PM samples is determined
primarily by constituent metals and
their bioavailability;
MCT-ROFA show enhanced
neutrophilic inflammation;
MCT-ROFA increase in mortality;
More and worsened dysrhythmias.
Reference
Costa and Dreher
(1997)
o
Tl
H
O
O
2
o
H
o
H
M
O
*>
O
I-H
H
M
WISTARmale
rats.
Bor: WISW
strain).
Coal oil Fly
Ash
Inhalation
(chamber)
0, 11,32,
103 mg/m3
1.9-2.6 ^m
(5g=1.6-1.8)
6h/day,
5d/week,
4 weeks.
At the highest cone., type II cell
proliferation and mild fibrosis occurred
and increased perivascular lymphocytes
were seen. The main changes at the
lowest concentration were particle
accumulation in AM and mediastinal
lymph nodes. Lymphoid hyperplasia
observed at all concentrations. Effects
increased with exposure duration.
Dormans et al.
(1999)
Rat, M, SD, ROFA
60 day old
C57Bl/6Jmice PTFE
TiO2
Intratracheal 8.33 mg/ml
instillation 0.3 ml/rat
Inhalation PTFE:
1.25, 2.5, 5 x 10s
particles/cc
TiO2-F: 10 mg/m3
NiO: 5 mg/m3
Ni3S2: 0.5 mg/m3
1.95,umMMAD
PTFE: 18nm
TiO2-F: 200 nm
TiO2-D: lOnm
Analysis at 24 and
96 h
30 min or 6 h/d,
5d/w, 6 months
Increased PMNs, protein, LDH at both
time points
Effects on the epithelium are due to
direct interactions with particles, not a
result of macrophage-derived mediators,
and suggest a more significant role in
the overall pulmonary response than
previously suspected;
type II cell growth factor production
may be significant in the pathogenesis
of pulmonary fibrosis
Dreher et al.
(1997)
Finkelstein et al.
(1997)
-------
o
o
r+
O
VO
VO
VO
TABLE 7-2 (cont'd). RESPIRATORY EFFECTS OF PARTICULATE MATTER
ON
O
Tl
H
O
O
2
o
H
o
H
M
O
*>
O
I-H
H
M
Species, Gender,
Strain, Age, etc.
Rat, M, SD.
60 day old
Female
Balb/cJ
mice 7-15
weeks.
Human
Rat, M, SD
S-D rats
60 days
Rats
Rat
Particle
Two ROFA
Samples
Rl had 2x
saline-leachable
sulfate, Ni, and
V and 40x Fe as
R2;
R2had31x
higher Zn
ROFA
Colloidal iron
oxide
ROFA
Provo, UT
TSP filters (10
yr old) soluble
and insoluble
extracts.
CAP
PTFE Fumes
Exposure Technique
Intratracheal
instillation
Intratracheal
instillation
Bronchial
instillation
Intratracheal
instillation
Intratracheal
instillation
nose-only
inhalation
Whole body
inhalation
Concentration Particle Size (,um)
2.5 mg in 0.3 ml Rl:1.88,um,
MMAD
R2: 2.03 //m,
MMAD
60 tAg in 50 /A < 2.5
(dose 3mg/kg)
5 mg in 10 ml. 2.6/^m
500 /^g/animal 3.6/^m
1 00-1 000 Mg of N/A
PM extract m
0.5 ml saline.
110-350 N/A
Mg/m3
1, 2.5 or 5 x 10s 18 nm
particles/cm3
Exposure
Duration Effect of Particles
Analysis at 4 days 4 of the 24 animals treated with R2 or R2s
(supernatant) died; none in Rl s treated
animals;
more AM, PMN, eosinophils protein, and
LDH in R2 and R2s animals;
more focal alveolar lesions, thickened
alveolar septae, hyperplasia of type II cells,
alveolar fibrosis in R2 and R2s animals;
Baseline pulmonary function and airway
hyperreactivity were worse in R2 and R2s
groups
N/A ROFA caused increases in eosinophils, IL-4
and IL-5 and airway responsiveness in
ovalbumin-sensitized and challenged mice.
1, 2, 4 days after L-ferritin increased after iron oxide particle
instillation exposure;
Transferrin was decreased, both lactoferrin
and transferrin receptor were increased
Analyzed 4 and Ferritin and transferrin were elevated;
96 h post greatest increase in ferritin, lactoferrin,
exposure transferrin occurred 24 h post exposure
24 h Inflammation (PMN) and lavage fluid protein
was greater with the soluble fraction
containing more metal (Zn, Fe, Cu).
3 h Increased peripheral blood neutrophils and
decreased lymphocytes heart rate increased
10-20 BPM after PM exposure.
1 5 min Increased PMN, mRNA of MnSOD and MT,
analysis 4 h post IL- 1 a, IL- 1 P, IL-6, MIP-2, TNF-a
exposure mRNA of MT and IL-6 expressed around all
airways and interstitial regions;
PMN expressed IL-6, MT, TNF-a; AM and
epithelial cells were actively involved
Reference
Gavett et al.
(1997)
Gavett et al.
(1999)
Ghio et al.
(1998a)
Ghio et al.
(1998b)
Ghio et al.
(1999b)
Gordon et al.
(1998)
Johnston et al.
(1996)
-------
o
r+
O
cr
CT
VO
VO
VO
Species, Gender,
Strain, Age, etc. Particle
Mice,C57BL/6J, PTFE Fumes
8 wk and 8 mo
old
Exposure
Technique
Whole body
inhalation
Concentration
1, 2.5 or 5 x 10s
particles/cm3
Particle Size (,um) Exposure Duration
1 8 nm 30 min exposure
Analysis 6 h
following exposure
Effect of Particles
Increased PMN, lymphocytes and protein
levels in old mice over young mice;
Increased TNF-a mRNA in old mice over
young mice;
No difference in LDH and
P-Glucuronidase
Reference
Johnston et al.
(1998)
o
Tl
H
O
O
2
o
H
o
H
M
O
*>
O
I-H
H
M
Rat, M, SD, 60
day old
SD rats
Human Bronchial
Epithelial
(BEAS-2B) cells
ROFA
TSP collected in
Provo, Utah
Intratracheal
instillation
Intratracheal
instillation
Rats, M, SD and
F-344 rats (60
days old)
Rats, M, SD,
WIS and F-344
rats (60 days old)
ROFA
ROFA
Intratracheal
instillation
Intratracheal
instillation
1.0 mg in 0.5 ml
saline
TSP filter samples
(36.5 mg/ml) agitated
in deionized H2O2 for
96 h; centrifuged at
I,200gfor30min;
lyophylized, and
resuspended in
deionized H2O2 or
saline
8.3 mg/kg
8.3 mg/kg
1.95/^m
N/A (TSP samples,
comprised 50 to
60% PM10)
1.95/^m
1.95/^m
Analysis at 24 h Increased PMNs, protein
Sacrificed at 24 h Provo particles caused cytokine-induced
neutrophil chemoattractant-dependent
inflammation of rat lungs;
Provo particles stimulated IL-6 and IL-8
production, increased IL-8 mRNA and
ICAM-1 in BEAS-2B cells, and
stimulated IL-8 secretion in primary
cultures of BEAS-2B cells;
Cytokine secretion was preceded by
activation of NF-KB and was reduced by
SOD, DBF, orNAC;
Quantities of Cu2+ found in Provo
particles replicated the effects
Sacrificed at 24 h increase in neutrophils in both SD and
F-344 rats;
A time-dependent increase in eosinophils
occurred in SD rats but not in F-344 rats.
Sacrificed at 24 h inflammatory cell infiltration as well as
alveolar, airway, and interstitial
thickening in all three rat strains;
a sporadic incidence of focal alveolar
fibrosis in SD rats, but not in WIS and
F-344 rats;
Fn mRNA isoforms EIIIA(+) were
upregulated in SD and WIS rats but not in
F-344 rats. Fn mRNA expression by
macrophage and alveolar and airway
epithelium and within fibrotic areas in SD
rats;
increased presence of Fn EIIIA(+) protein
in the areas of fibrotic injury and basally
to the airway epithelium.
Kadiiska et al.
(1997)
Kennedy et al.
(1998)
Kodavanti et al.
(1996)
Kodavanti et al.
(1997)
-------
o
o
r+
O
VO
VO
VO
TABLE 7-2 (cont'd). RESPIRATORY EFFECTS OF PARTICULATE MATTER
oo
o
Tl
H
O
O
2
o
H
o
H
M
O
*>
O
I-H
H
M
Species, Gender,
Strain, Age, etc.
Rats, M, SD and
F-344 rats (60
days old)
60 day male
S-D rats
treated with
monocrotaline
60 mg/kg
Brown Norway
rat
Human 27M, 7F
20-36 yr.
Fischer 344 rats.
(25g)
Rats Wis (HAN
strain),
Particle
10 ROFA
water and . 1 M
HC1 leachable
As, Be, Cd, Co,
Cr, Cu, Fe, Mn,
Ni, Pb, V, Zn, S
ROFA
ROFA
Fe203
Fe203
Ambient PM
Edinburgh; CB,
CB Ultrafine
(UCB)
Exposure
Technique
Intratracheal
instillation
Intratracheal
instillation (IT);
Nose-only
inhalation (IN)
Intratracheal
instillation
Intrapulmonary
instillation
Intratracheal
instillation
Intratracheal
instillation
Concentration Particle Size (,um)
0.833,3.33,8.3 1.99- 2.59 ,um
mg/kg MMAD
0, 0.83, 3.3 mg/kg; Source: 1.95 /j,m;
15mg/m3 N/A
200 Mg N/A
100/^g
3 x 108microspheres 2.6/^m
in 10ml saline.
7.7xl07 2.6 ^m
microspheres in 5 ml
saline
50-125 ,ug in 0.2 ml PMi0
CB = (200-500 nm)
UCB = 20 nm
Exposure Duration
Effect of Particles
Sacrificed at 24 h ROFA induced increases in BAL
protein and LDH, but not PMN were
associated with water-leachable total
metal, Ni, Fe, and S;
PMN was correlated with V;
Chemiluminescence signals in vitro
(AM) were greatest with ROFA
containing soluble V and less with Ni
plus V.
24-96 h ; Both IT and M rats showed
inflammatory responses (IL-6, MIP-2,
6h/day for 3 days inflammatory cells, etc.). 58% of IT
rats exposed to ROFA died within
96 h. No mortality occurred in the IN
rats. ROFA exacerbated lung lesions
(edema, inflammatory cells, alveolar
thickening).
N/A ROFA enhanced the response to house
dust mite (HDM) antigen challenge.
Eosinophil numbers, LDH, BAL
protein, and IL-10 were increased with
ROFA + HDM versus HDM alone.
N/A Transient inflammation induced
initially (neutrophils, protein, LDH, IL-
8) was resolved by 4 days post-
instillation.
N/A Transient inflammation at 1 day post
instillation.
Sacrificed at 6 h Increased PMN, protein and LDH
following PMi0;
greater response with ultrafine CB but
not CB;
Reference
Kodavanti et al.
(1998a)
Kodavanti et al.
(1999)
Lambert et al.
(1999)
Lay et al.
(1998)
Li et al.
(1996, 1997)
decreased GSH level in BAL; free
radical activity (deplete supercoil
DNA);
Leukocytes from treated animals
produced greater NO and TNF
-------
VO
O
Tl
H
O
O
2
O
H
Species, Gender,
Strain, Age, etc. Particle
NMRI mice; MnO2
Mouse peritoneal
macrophage
60 day S-D rats, Florida ROFA;
male Domestic oil fly
ash
Male SD Diesel,
rat SiO2,
(200g) Carbon black
Rat , M, F344, TiO2
175-225g
Rat, M. F344, TiO2
175-225 g
Exposure
Technique
Intratracheal
instillation;
In vitro
Intratracheal
instillation
Intratracheal
instillation
Intratracheal
inhalation and
Intratracheal
instillation
Intratracheal
inhalation and
Intratracheal
instillation
Concentration
0.037,0.12,0.75,
2.5 mg/animal
1, 000 IA in 0.5 ml
1 mg in 0.4 ml.
Inhalation at
125 ,ug/m3
Instillation at:
500 ,ug for fine
750 ,ug for ultrafine
Inhalation at
125 Mg/m3
Instillation at:
500 ,ug for fine
750 [j.g for ultrafine
Particle Size (,um)
surface area of
0.16,0.5; 17,
62 m2/g
DEP Collected as
TSP - disaggregated
in solution by
sonication (20 nm);
SiO2 (7 nm);
Carbon Black
Fine: 250 nm
Ultrafine: 21 nm
Fine: 250 nm
Ultrafine: 21 nm
Exposure Duration
Sacrificed at 5 days
1 5 min to 24 h
Sacrificed at 2, 7,
21,42, and 84 days
post-instillation
inhalation
exposure, 2 h;
sacrificed at 0, 1 3,
and 7 days
postexposure for
both techniques
inhalation
exposure, 2 h;
sacrificed at 0, 1 3,
and 7 days
postexposure for
both techniques
Effect of Particles
LDH, protein and cellular recruitment
increased with increasing surface area;
Freshly ground particles had enhanced
cytotoxicity
ROFA induced production of
acetaldehyde with a peak at about 2 h.
No acetaldehyde was seen in plasma at
any time. DOFA increased
acetaldehyde as did V and Fe.
Amorphous SiO2 increased
permeability, neutrophillic
inflammation. Carbon black and DEP
translocated to interstitum and lymph
nodes by 12 weeks.
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
Lison et al.
(1997)
Madden et al.,
(1999)
Murphy et al.,
1998
Osier and
Oberdorster
(1997)
Osier et al.
(1997)
O
H
M
O
O
I-H
H
M
-------
o
o
r+
O
VO
VO
VO
TABLE 7-2 (cont'd). RESPIRATORY EFFECTS OF PARTICULATE MATTER
Species, Gender,
Strain, Age, etc.
Rats
Particle
NaVO3
VOSO4
V2OS
Exposure
Technique
Intratracheal
instillation
Concentration
21 or210,ugV/kg
(NaVO3, VOSO4
soluble)
42 or 420 ,ug V/kg
(V2OS) less soluble
Exposure Duration
Particle Size (,um)
N/A Ih or 10 days
following
instillation
Effect of Particles
PMN influx was greatest following
VOSO4, lowest for V2OS.
VOSO4 induced inflammation
persisted longest;
MIP-2 and KC (CXC chemokines)
Reference
Pierce et al.
(1996)
were rapidly induced as early as 1 h
post instillation and persisted for 48 h;
Soluble V induced greater chemokines
mRNA expression than insoluble V
AM have the highest expression level;
ON
O
o
£
Tj
H
O
O
AM - Alveolar Macrophage
CAP - Concentrated Ambient Particles
CB - Carbon Black
CFA-Coal Fly Ash
CMP - Copper Smelter Dust
DBF - Deferoxamine or Desferrioxamine
DMTU - Dimethylthiourea
DNA - Deoxyribonucleic Acid
DPM - Diesel Particulate Matter
ERK - Extracellular Receptor Kinases
F-344 - Fischer 344
Fe2O3 - Iron Oxide
Fn - Fibronectin
G-6-PDH - Glucose-6-Phosphate Dehydrogenase
IL-6; 11-8 - Interleukin 6,8
JNK - Jun Hj-Terminal Kinases
LCL - Luminol Enhanced Chemiluminescence
LDH - Lactate Dehydrogenase
LPS - Liopopolysaccharide (Endotoxin)
MAPK - Mitogen-Activated Protein Kinases
MCT - Monocrotaline
MMAD - Mass Median Aerodynamic Diameter
mRNA - Messenger Ribonucleic Acid
NF-KB - Nuclear factor-kappa B
NHBE - Normal Human Broncial Epithelial Cells
PHS - Prostaglandin H Synthetase
PMA - Phorbol Myristate Acetate
PMN - Neutrophil
PTFE - Polytetrafluoroethylene
ROFA - Residual Oil Fly Ash
RTE - Rat Tracheal Epithelial Cells
SD - Sprague Dawley
SiO2 - Silicon Dioxide
SOD - Superoxide Dismutase
SR - Scavenger-Type Receptors
Ti O2 - Titanuim Dioxide
TNF - Tumor Necrosis Factor
TSP - Total Suspended Paticles
UAP - Urban Air Particles
UF - Ultrafme
WC - Tungsten Carbide
Wis - Wistar
o
H
o
H
W
O
O
I-H
H
W
-------
1 Instillation of ambient PM10 collected in Edinburgh, Scotland, also caused pulmonary
2 injury and inflammation in rats (Li et al., 1996, 1997). Six hours after intratracheal instillation of
3 PM10 (in 0.2 ml saline containing 50 to 125//g of particles), an influx of PMN, and increased
4 epithelial permeability, total protein, and LDH in BAL were observed. An even greater
5 inflammatory response was observed after instillation of ultrafine carbon black (125//g, 20 nm)
6 but not after fine carbon black (200-250 nm) particles. A decrease in reduced glutathione (GSH)
7 in the BAL suggested that PM10 is capable of inducing the production of oxidants. In addition,
8 BAL leukocytes recovered from rats treated with PM10 produced greater amounts of nitric oxide
9 and tumor necrosis factor alpha (TNF-cc) than control animals. These data provide evidence that
10 instillation of PM10 causes lung injury, which may be mediated through oxidant generation and
11 cytokines.
12 Brain et al. (1998) examined the effects of particles that resulted from the Kuwaiti oil fires
13 in 1991. Hamsters were intratracheally instilled with particles (0.15, 0.75, and 3.75 mg per 100 g
14 animal) collected in Ahmadi, a residential area of Kuwait. The response of hamsters instilled
15 with particles from Ahmadi was compared to the response of hamsters instilled with particles
16 collected in St. Louis, MO. When compared to hamsters instilled with St. Louis particles,
17 hamsters instilled with Ahmadi particles had between 1.4 and 2.2-fold more neutrophils in their
18 BAL fluid. However, Ahmadi-treated hamsters had a lesser decrease in macrophage number and
19 a smaller increase in LDH activity. There were no significant differences in albumin and
20 p-N-acetylglucosaminidase levels (the latter indicative of damage to macrophages or
21 neutrophils). These results showed that on an equal mass basis, the acute toxicity of the Ahmadi
22 combustion particles was similar to that of urban particles collected in the United States.
23 In summary, intratracheal instillation of ambient particles induced an inflammatory
24 response in the lungs, perhaps mediated through cytokines, and the responses appears to be
25 dependent on the bioavailable transition metals and subsequent oxidant production, rather than
26 on instilled PM mass.
27
28 7.3.4.2 Coal Fly Ash or Residual Oil Fly Ash (ROFA)
29 Many studies investigating the response of animals to particle exposures have used ROFA
30 as a surrogate for ambient particles. ROFA has a high content of water soluble sulfate and
31 metals. As described previously (U.S. Environmental Protection Agency, 1996a), intratracheal
October 1999 7-61 DRAFT-DO NOT QUOTE OR CITE
-------
1 instillation of high doses of ROFA suspension generally produced severe inflammation, an
2 indicator of pulmonary injury which included recruitment of neutrophils, eosinophils, and
3 monocytes into the airway. The biological effects of ROFA have been shown to depend on
4 aqueous leachable chemical constituents of the particles. Dreher et al. (1997) have shown that a
5 leachate prepared from ROFA, containing predominantly Fe, Ni, V, Ca, Mg, and sulfate,
6 produced similar lung injury to that induced by the complete ROFA suspension. Depletion of Fe,
7 Ni, and V from the ROFA leachate eliminated its pulmonary toxicity. Correspondingly, minimal
8 lung injury was observed in animals exposed to saline-washed ROFA particles. A surrogate
9 transition metal sulfate solution containing Fe, V, and Ni largely reproduced the lung injury
10 induced by ROFA. Interestingly, ferric sulfate and vanadium sulfate antogonized the pulmonary
11 toxicity of nickel sulfate. Interactions between different metals and the acidity of PM were found
12 to influence the severity and kinetics of lung injury induced by ROFA and its soluble transition
13 metals.
14 To further confirm the role of soluble metal components in the toxicity of emission source
15 particles, Gavett et al. (1997) investigated the effects of two ROFA samples of equivalent
16 diameters, but having different metal and sulfate content, on pulmonary responses in
17 Sprague-Dawley rats. ROFA sample 1 (Rl) (the same emission particles used by Dreher et al.
18 [1997]) had approximately twice as much saline-leachable sulfate, nickel, and vanadium, and
19 40 times as much iron as ROFA sample 2 (R2); while R2 had a 31-fold higher zinc content. Four
20 groups of rats received intratracheal instillations with suspensions of 2.5 mg R2 in 0.3 ml saline,
21 the supernatant of R2 (R2s), the supernatant of 2.5 mg Rl (Rls), or saline only. By 4 days after
22 instillation, 4 of 24 rats treated with R2s or R2 had died. None of those treated with Rls or
23 saline died. Pathological indices such as alveolitis, early fibrotic changes and perivascular
24 edema, were greater in both R2 groups. In surviving rats, baseline pulmonary function
25 parameters and airway hyperreactivity to acetylcholine were significantly worse in R2 and R2s
26 groups than in the Rls groups. Other than BAL neutrophils, which were significantly higher in
27 the R2 and R2s groups, no other inflammatory cells (macrophages, eosinophils, or lymphocytes)
28 or biochemical parameters of lung injury were significantly different between the R2 and R2s
29 groups and the Rls group. They also observed an increase in focal alveolar lesions, thickened
30 alveolar septae, hyperplasia of type II cells and fibrosis in rats instilled with ROFA (2.5 mg) and
31 examined 4 days later. Although soluble forms of zinc had been found in guinea pigs to produce
October 1999 7-62 DRAFT-DO NOT QUOTE OR CITE
-------
1 a greater pulmonary response than other sulfated metals (Amdur et al., 1978), and although the
2 level of zinc was 30 fold greater in R2 than Rl, the precise mechanisms by which zinc may
3 induce such responses are unknown. Nevertheless, these results show that the composition of
4 soluble metals and sulfate leached from ROFA, a type of emission source particle, is critical in
5 the development of airway hyperractivity and lung injury.
6 In an effort to determine the role of reactive oxygen species in the in vivo toxicity of
7 ROFA, Dye et al. (1997) treated rats with an intraperitoneal injection of saline or
8 dimethylthiourea (DMTU) (500 mg/kg), followed 30 min later by intratracheal instillation of
9 either acidic saline (pH = 3.3) or an acidified suspension of ROFA (500 //g/0.3 ml). The
10 systemic administration of DMTU impeded development of the cellular inflammatory response
11 to ROFA, but did not ameliorate biochemical alterations in BAL fluid. In addition, it is
12 suggested that systemic administration of DMTU resulted either in scavenging or diminished
13 production of key reactive oxygen species. This in turn ameliorated cellular redox changes and
14 thus diminished the "effector cell" response to ROFA (i.e., decreased cytokine production by
15 airway epithelial cells, alveolar macrophages, or lymphocytes).
16 The response of different strains of rats to ROFA instillation was investigated by Kodavanti
17 et al. (1996). Male Sprague Dawley (SD) and Fischer-344 (F-344) rats were intratracheally
18 instilled with saline or ROFA particles. ROFA exposure produced an increase in BAL
19 neutrophils in both SD and F-344 rats. A time-dependent increase in eosinophils occurred in SD
20 rats but not in F-344 rats. In a subsequent study (Kodavanti et al., 1997a), male SD, Wistar
21 (WIS), and F-344 rats (60 days old) were exposed to saline or ROFA (8.3 mg/kg) by intratracheal
22 instillation and examined for up to 12 wk. Histology indicated focal areas of lung damage
23 showing inflammatory cell infiltration as well as alveolar, airway, and interstitial thickening in
24 all three rat strains during the week following exposure. Trichrome staining of the lung sections
25 indicated a sporadic incidence of focal alveolar fibrosis at 1, 3, and 12 wk in SD rats, whereas
26 WIS and F-344 rats showed only a modest increase in trichrome staining in the septal areas.
27 One of the isoforms of fibronectin (Fn) mRNA was upregulated in ROFA-exposed SD and WIS
28 rats but not in F-344 rats. These studies indicate that there is a rat strain variation in
29 ROFA-induced fibrosis and associated Fn expression.
30 To determine which of the three dominant metals of ROFA were associated with the
31 development of lung lesions and how it related to the kind of inflammatory response and
October 1999 7-63 DRAFT-DO NOT QUOTE OR CITE
-------
1 proinflammatory cytokine expression, SD rats were intratracheally instilled with either ROFA or
2 metal sulfates (iron sulfate, vanadium sulfate, and nickel sulfate), individually or in combination
3 of all three (at a dose equivalent to one ROFA instillate) Kodavanti et al., 1997b). Kinetics of
4 pulmonary injury and cytokine expression profile over 96 h indicated that injury induced by Ni
5 was greater, had a delayed onset and was persistent over 96 h. However, V caused less injury
6 and early (3 h) cytokine gene expression. Iron was associated with mild lung lesions and early
7 cytokine gene expression. Within 96 hours lung lesions consolidated to fibrosis in the case of V,
8 the metal mixture, and ROFA (ROFA>metal mixture>V). In the case of Ni, lesions were
9 associated with continued alveolar edema, hemorrhage, and inflammatory cell influx but did not
10 consolidate to fibrosis in 96 h. Thus, of all three predominant metals, Ni accounted for the
11 majority of the ROFA toxicity.
12 To further investigate the response to ROFA with differing metal and sulfate composition,
13 male SD rats (60 days old) were exposed to ten different ROFA samples collected at various sites
14 within a power plant (Kodavanti et al., 1998a). Animals received intratracheal instillations of
15 either saline or a saline suspension of whole ROFA (< 3.0 //m MMAD) at three concentrations
16 (0.833, 3.33, or 8.33 mg/kg). ROFA-induced increases in BAL fluid protein and LDH, but not
17 neutrophilic inflammation, were associated with its water-leachable total metal, Ni, Fe, and
18 sulfate content. However, the neutrophilic response following ROFA exposure was positively
19 correlated with its water-leachable V content. Modest lung injury was observed with ROFA
20 samples that contained the smallest amounts of water-leachable metals. The ability of ROFA to
21 induce oxidative burst in AM was determined in vitro using a chemiluminescence (CL) assay.
22 Alveolar macrophage CL signals in vitro were greatest with ROFA containing primarily soluble
23 V and were less with ROFA containing Ni plus V. These results showed that ROFA-induced
24 PMN influx appeared to be associated with its water-leachable V content; however, protein
25 leakage appeared to be associated with water-leachable Ni content. ROFA-induced in vitro
26 activation of AM was highest with ROFA containing leachable V but not with Ni plus V,
27 suggesting that the potency and the mechanism of pulmonary injury may differ between
28 emissions containing V and Ni.
29 Kadiiska et al. (1997) instilled rats with either saline or 500 //g of ROFA in saline.
30 Neutrophils accounted for the majority of cells recovered by lavage 24 h following instillation of
31 ROFA. An increased BAL fluid protein concentration was indicative of increased epithelial
October 1999 7-64 DRAFT-DO NOT QUOTE OR CITE
-------
1 permeability subsequent to injury. Both the influx of neutrophils and the increase in lavage
2 protein concentration were associated with the soluble, rather than insoluble component of the
3 ROFA. Twenty-four hours following instillation, rats were intraperitoneally injected with an
4 aqueous solution of cc-(4-pyridyl l-oxide)-N-fer£-butylnitrone (POBN). One hour later, the
5 Electron Spin Resonance (ESR) spectrum of POBN radical adducts was determined in lung lipid
6 extracts. Relative to the ESR spectrum after saline instillation, there were significant increases in
7 the intensities of the signals after exposure to total ROFA, the soluble component of ROFA, the
8 synthetic ROFA (a mixture of vanadyl, nickel and ferric sulfate, reflecting their concentrations in
9 the ROFA), vanadyl sulfate (VOSO4), and ferric sulfate (Fe2(SO4)3). However, there were no
10 significant differences in the signal intensities of the spectra associated with exposure to the
11 insoluble component of the ROFA and nickel sulfate (NiSO4). The authors concluded that ESR
12 analysis of lung tissue demonstrates in vivo free radical production and that the generation of free
13 radicals appears to be associated with soluble metals in the oil fly ash.
14 Broeckaert et al. (1997) investigated the effect of intratracheally instilled coal fly ash (CFA)
15 and copper smelter dust (CMP) on the lung integrity and on the ex vivo release of TNF-cc by
16 alveolar phagocytes. Female Naval Medical Research Institute (NMRI) mice were instilled with
17 different particles normalized for the arsenic content (20 //g/kg body weight, i.e., 600 ng
18 arsenic/mouse) and the particle load (100 mg/kg body weight, i.e., 3 mg/mouse). Mice received
19 tungsten carbide (WC) alone (100 mg/kg), CFA alone (100 mg/kg, i.e., 20 //g arsenic/kg), CMP
20 mixed with WC (CMP, 13.6 mg/kg, i.e., 20 //g arsenic/kg; WC, 86.4 mg/kg) and Ca3(AsO4)2
21 mixed with WC (20 //g arsenic/kg; WC, 100 mg/kg). Additional mice were studied to evaluate
22 particle retention by measuring total arsenic retention in the lung over time. Instillation of WC
23 induced a mild and transient (day 1) inflammatory reaction characterized by an increase of BAL
24 fluid protein and an influx of PMNs into the alveolar compartment. Compared to WC,
25 Ca3(AsO4)2 produced a significant increase of BAL protein. CMP particles caused a severe but
26 transient inflammatory reaction, while a persisting alveolitis (30 d) was observed after treatment
27 with CFA. Compared to control saline, a marked inhibition of TNF-cc release in AMs ex vivo
28 was observed in response to lipopolysaccharide (LPS) in all groups at day 1. Cytokine
29 production was upregulated in WC- and Ca3(AsO4)2- treated animals after 6 and 30 d,
30 respectively. However, a 90% inhibition of TNF-cc production was still observed at 30 d after
31 administration of CMP and CFA. Although arsenic was cleared from the lung tissue 6 d after
October 1999 7-65 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ca3(AsO4)2 administration, a significant fraction persisted (10-15% of the arsenic administered)
2 in the lung of CMP- and CFA-treated mice at day 30. It is possible that suppression of TNF-cc
3 production is dependent upon the slow elimination of the particles and their metal content from
4 the lung.
5 In summary, intratracheally injected ROFA produced similar or greater inflammatory
6 response than that produced by ambient particles at similar high doses. Although the water
7 soluble metals in ROFA appear to play a role in the inflammatory response, the precise
8 mechanisms of their effect and the complex interactions are not yet clear.
9
10 7.3.4.3 In Vitro Exposures
11 In vitro exposure is a useful technique when only limited quantities of the test material are
12 available. Exposing respiratory cells to particles in vitro not only reduces the amount of material
13 needed for the experiments but also provides an opportunity to investigate the mechanisms of
14 particle toxicity. In addition, in vitro exposure allows the examination of the response to
15 particles in only one or two cell types. Limitations of in vitro studies include difficulty in
16 dose-response and mechanistic extrapolation. Furthermore, use of "non-cytotoxic" doses in
17 in vitro studies does not necessarily imply that the doses are relevant to in vivo exposure
18 situations. However, in vitro studies do not provide an approach to identify potential cellular and
19 molecular mechanisms by which PM mediates health effects. These mechanisms can then be
20 evaluated in vivo. In vitro/ex vivo studies are summarized in Table 7-3.
21
22 Ambient Particles
23 To investigate the mechanisms of the in vivo effects of PM, Kennedy et al. (1998) used
24 cultured BEAS-2B cells to study the effect of ambient particles (TSP) collected in Provo, Utah.
25 Similar to ROFA, Provo particles stimulated IL-6 and IL-8 production as well as increased
26 IL-8 mRNA and enhanced expression of intercellular adhesion molecule-1 (ICAM-1) in these
27 cells. Provo particles also stimulated IL-8 secretion in primary cultures of human bronchial
28 epithelial cells. Cytokine secretion was preceded by activation of nuclear factor kappa B (NF-KB)
29 and was reduced by treatment with superoxide dismutase (SOD), Deferoxamine (DBF), or
30 N-acetylcysteine. The addition of similar quantities of Cu2+ as found in the Provo extract
October 1999 7-66 DRAFT-DO NOT QUOTE OR CITE
-------
ON
H
6
o
2
o
H
o
H
W
O
*>
O
HH
H
W
Species, Cell type,
etc.
Human bronchial
epithelial cells
Asthmatic (ASTH)
Nonasthmatic
(NONA)
Human bronchial
epithelial cells
(smokers)
Human and
rat alveolar
macrophages
Human AM and
blood monocytes
Rat alveolar
macrophages
Particle or Exposure
Constituent Technique
Diesel In vitro
Exhaust
Particles
DEP In vitro
4 Urban air In vitro exposure;
particles 2 x 10s cells
ROFA exposed for 2 h
Diesel
Volcanic
ash
Silica
urban air In vitro
particles;
St Louis
SRM 1648;
Washington
, DC SRM
1649;
Ottawa,
Canada
EHC-93
PMi0 In vitro
Mexico City
1993;
voleanic
ash.
(MSHA)
Concentration
10-100 jig/ml
10-100 jig/ml
Urban and diesel:
12,27, 111,333,
1000/^g/ml
SiO2 and TiO2: 4, 12,
35, 167 ,ug/ml
Fe2O3: 1:1,3:1;
10:1 particles/cell
ratio
33 or 100 ,ug/ml
1-100 ,ug/ml
Particle Size
(,um)
0.4 jim
0.4 jim
Urban particles:
0.3-0.4 ,um
Diesel: 0.3 ,um
ROFA: 0.5 ,um
Volcanic ash:
1.8/^m
Silica: 05-10 ,um
TiO2: <5 /j,m
Latex: 3.8 ,um
0.2 to 0.7 ,um
<10,um
Exposure Duration Effect of Particles
2, 4, 6, 24 h DEP caused no gross cellular damage. Ciliary
beat frequency was attentuated at all doses.
DEP caused IL-8 release at lower dose in
ASTH than NONA. Higher concentrations of
DEP suppressed IL-8, GM-CSF, RANTES in
ASTH cells.
24 h DEP attenuated ciliary beating. Release of IL-
8, protein, GM-CSF, SICAM-1 increased after
DEP exposure.
2 h for cytotoxicity UAP-induced cytokine production (TNF, IL-6)
16-18 h for in AM of both species and is not related to
cytokine assay respiratory burst or transition metals; but may
Chemiluminescenc be related to LPS (blocked by polymyxin B
e at 30 minutes but not DBF)
ROFA induced strong chemiluminescence but
had weak effects on TNF production.
3 h, 6 h, Phagocytosis was inhibited by UAP at 1 8 h.
or 1 8-20 h UAP caused decreased expression of
P2-integrins involved in antigen presentation
and phagocytosis.
24 h PMi0 stimulated alveolar macrophages to
induce upregulation of PDGF « receptor on
myofiboroblasts. Endotioxin and metal
components of PMi0 stimulate release of IL-P.
This is a possible mechanism for
PMi0-induced airway remodeling.
Reference
Bayram et al.
(1998a)
Bayram et al.
(1998b)
Becker et al.
(1996)
Becker and
Soukup (1998)
Bonner et al.
(1998)
-------
o
r+
O
cr
CT
VO
VO
VO
Species, Cell
type, etc.
Supercoiled
DNA
Particle or
Constituent
PMi0 from
Edinburgh,
Scotland
Exposure
Technique
In vitro
Particle Size
Concentration (/^m)
996.2±181.8 PMi0
Mg/filter in
100 //I
Exposure
Duration Effect of Particles
8 h PMi0 caused damage to DNA; mediated by hydroxyl
radicals (inhibited by mannitol) and iron (DBF);
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)
ON
oo
Rat AM
UAP
DPM
In vitro
Primary
cultures of RTE
ROFA
In vitro
50 to DPM: 1.1 - 1.3 ,um 2 h exposure; Dose dependent increase in TNF-a, IL-6, CINC, MIP- Dong et al.
200 ,ug/ml UAP: St Louis, MO supernatant 2 gene expression by urban particles but not with (1996)
between 1974 and collected 18 h DPM;
1976 in abaghouse, post Cytokine production were not related to ROS;
sieved through Cytokine production can be inhibited by polymyxin B;
200-mesh (125 ,um) EPS was detected on UAP but not DPM;
Endotoxin is responsible for the cytokine gene
expression induced by UAP in AM
5,10, Same as Dreher Analysis at Particle induced epithelial-cell detachment and lytic Dye et al.
20 ,ug/cm2 et al. (1997) 6 and cell injury; (1997)
24 hours alterations in the permeability of the cultured RTE cell
layer;
Increase in LDH, G-6-PDH, gluathione reductase,
glutathione S-transferase;
Mechanism of ROFA-induced RTE cytotoxicity
involves the development of an oxidative burden.
\J
3>
Tl
H
1
o
o
h->
z;
H
0
H
W
o
^~
O
w
Peripheral
blood
monocytes
Rat AM
NHBE
BEAS-2B
Organic extract of
TSP, Italy
ROFA, iron sulfate,
nickel sulfate,
vanadyl sulfate
Latex particles with
metal complexed on
the surface
ROFA
In vitro 42.5 /^g
extract/m3
(acetone)
In vitro 0.01-1.0
(0.7 x 106 cells/ml) mg/ml
In vitro 5 - 200 ,ug/ml
N/A, Collected from 2 h
high volume
sampler (60 m3/h)
3.6,umMMAD Up to
400 minutes
3.6/^m 2 and 24 h
Superoxide anion generation was inhibited at a
particulate concentration of 0.17 mg/ml when
stimulated with PMA; 50% increase in LDH;
disintegration of plasma membrane
Increase chemiluminescence, inhibited by DBF and
hydroxyl radical scavengers;
solutions of metal sulfates and metal-complexed latex
particles similarly elevated chemiluminescence in a
dose-and time-dependent manner.
mRNA for ferritin did not change; ferritin protein
increase;
mRNA for transferrin receptor decreased, mRNA for
lactoferrin increased;
Transferrin decreased whereas lactoferrin increased;
Deferoxamine alone increased lactoferrin mRNA
Fabiani et al.
(1997)
Ohio et al.
(1997a)
Ohio et al.
(1998c)
-------
o
r+
O
cr
CT
VO
VO
VO
Species, Cell Particle or Exposure Particle Size
type, etc. Constituent Technique Concentration (/^m) Exposure Duration Effect of Particles
BEAS-2B oil fly ash In vitro lOO/^g/ml N/A
respiratory
epithelial cells.
~lh Lactoferrin binding with PM metal occurred within
5 min. V and Fe (m) but not Ni bound to the
lactoferrin receptor.
Reference
Ghio et al.
(1999a)
ON
O
Tl
H
6
O
2
o
H
O
c
o
H
W
O
&
O
I-H
H
W
BEAS-2B
0X174 RF1
DNA
Provo, UT
TSP soluble and
insoluble extract
PMi0 from
Edinburgh, Scotland
In vitro
In vitro
TSP
3.7,
24 h
8h
Hamster AM
ROFA or CAP
In vitro
0, 25, 50, 100,
CAP: 0.1 - 2.5/^m (from
Harvard concentrator)
TiO2: 1 ,um
Hamster AM
CAP, ROFA, and
their water-soluble
and particulate
fractions
In vitro
0-200 mg/ml
CAP = 0.125
ROFA= 1.0
30 min incubation;
Analysis
immediately
following
30 min
Water soluble fraction caused greater release of IL- Ohio et al.
8 than insoluble fraction. The effect was blocked (1999b)
by deferoxamine and presumably due to metals (Fe,
Cu, Zn, Pb).
Significant free radical activity on degrading Gilmour et al.
supercoiled DNA; (1996)
Mainly due to hydroxyl radicals (inhibited by
mannitol);
Fe involvement (DEF-B conferred protection);
More Fe3+ was released compared to Fe2+,
especially at pH 4.6 than at 7.2
Dose dependent increase in AM oxidant stress with Goldsmith et al.
both ROFA and CAP (1997)
Increase in particle uptake;
Mac-type SR mediate a substantial proportion of
AM binding;
particle-associated components (e.g., transition
metals) are likely to mediate intracellular oxidant
stress and proinflammatory activation
ROFA and CAPs (water soluble components) Goldsmith et al.
caused increases in DCFH oxidation; (1998)
CAPs samples and components showed substantial
day-to-day variability in their oxidant effects;
ROFA increased MIP-2 and TNF-a production in
AM and can be inhibitable by NAC.
AM's from
female CD rats.
Human PMN
vanadyl chloride In vitro
sodium metavanadate
Aqueous and organic In vitro
extracts of TSP in
Dusseldorf and
Duisburg, Germany
10-1000 ,um
metavanadate
0.42 - 0.78 mg
dust/ml
N/A 30 min
Collected by high volume Up to 35 min
sampler, 90% < 5 ,um,
50% < Ijiun, maximum at
0.3 - 0.45 ,um
Extracted using water and
then dichloromethane to
yield aqueous and organic
extracts
Metavanadate caused increased production of ROS.
The LOEL was 50 ,uM
PM extract alone significantly stimulated the
production and release of ROS in resting but not in
zymosan-stimulated PMN. The effects of the PM
extracts were inhibited by SOD, catalase and
sodium azide (NaN3);
Zymosan-induced LCL is inhibited by both types of
extracts, but aqueous extracts have a stronger
inhibitory effect.
Grabowski et al.
(1999)
Hitzfeld et al.
(1997)
-------
O
TI
H
6
O
2
O
H
O
c
o
H
W
O
&
O
I-H
H
W
Species, Cell
type, etc.
Human AM
BEAS-2B, airway
epithelial cells
Male (Wistar)
Rat lung
macrophages
Human blood
monocytes and
neutrophils
(PMN)
Human airway
epithelium
derived cell lines
BEAS-2B
(S6-subclone)
Human airway
epithelium
derived cell line
BEAS 2B
Human airway
epithelium
derived cell line
BEAS
Particle or
Constituent
UAP
(#1648, 1649)
Volcanic ash
ROFA
ROFA
urban dust SRM
1649, TiO2,
Quartz
ambient air
particles, carbon
black; oil fly
ash, coal fly ash
ROFA
ROFA
ROFA
Synthetic ROFA
(soluble Ni, Fe,
andV)
Exposure
Technique Concentration
In vitro 0,25,100
200 Mg/ml
In vitro 0, 0.5, 2.0 mg in
10ml
In vitro 0-100 Mg in 1 ml
In vitro 100/^gin
0.2ml
In vitro 0,6,12,25
50 Mg/ml
In vitro 2, 20, 60 Mg/cm2
In vitro ROFA: 0 - 200
Mg/ml
Synthetic ROFA
(100Mg/ml):
Ni, 64 MM
Fe, 63 MM
V, 370 MM
Particle Size Exposure
(Mm) Duration
Volume median 24 h
diameter:
ROFA 1.1 ^m
#1648: 1.4 /^m
#1649: 1.1
volcanic ash 2.3 Mm
N/A Ih
N/A 18 h
N/A 40 min.
1.96 Mm land 24 h
1.96 Mm 24 h
exposure
ROFA: 1 .96 Mm Up to 24 h
Synthetic ROFA: N/A
(soluble)
Effect of Particles
ROFA highly toxic; Urban PM toxic at 200//g/ml;
ROFA produced significant apoptosis as low as 25 Mg/mlj
UAP produced apoptosis at 100 Mg/nil;
UAP and ROFA also affect AM phenotype: increase
immune stimulatory while decrease immune suppressor
phenotype
ROFA induced production of acetaldehyde in dose-
dependant fashion
Cytotoxicity ranking was quartz > SRM 1649 > TiO2,
based on cellular ATP decrease and LDH, acid
phosphatase, and p-glucuronidase release
ROS generation measured by LCL increased in PMN, was
correlated with Si, Fe, Mn, Ti, Co content by not V, Cr,
Ni, and Cu. Deferoxamine did not affect LCL suggesting
that transition metals are not related to LCL.
Activation of IL-6 gene by NF-KB activation and binding to
specific sequences in promoter of IL-6 gene;
Inhibition of NF-KB activation by DBF and NAC;
Increase in PGE2, IL-6, TNF, and IL-8;
Activation NF-KB may be a critical first step in the
inflammatory cascade following exposure to ROFA
particles.
Epithelial cells exposed to ROFA for 24 h secreted
substantially increased amounts of the PHS products
prostaglandins E2 and F2o;
ROFA-induced increase in prostaglandin synthesis was
correlated with a marked increase in PHS activity
Tyrosine phosphatase activity, which was known to be
inhibited by vanadium ions, was markedly diminished after
ROFA treatment;
ROFA exposure induces vanadium ion-mediated inhibition
of tyrosine phosphatase activity, leading to accumulation of
protein phosphotyrosines in BEAS cells.
Reference
Holian et al.
(1998)
Madden et al.
(1999)
Nadeau et al.
(1996)
Prahalad et al.
(1999)
Quay et al.
(1998)
Samet et al.
(1996)
Samet et al.
(1997)
-------
o
r+
O
cr
CT
VO
VO
Species, Cell
type, etc.
Human airway
epithelium
derived cell lines
BEAS-2B
A .TiAjAjAjj / —*} yvuiii l*7« A^A1 A1 AI*X_> A kj v^A1 A,rm.AVj
Particle or Exposure Particle Size
Constituent Technique Concentration (/^m)
Particle In vitro 500 /M of As, F, Cr N/A (soluble)
components As, (III), Cu, V, Zn
Cr, Cu, Fe, Ni,
V, and Zn
L A v> u A_/,rm. A ijj iTAj-m. A i ijj AX ijj^-v TAT v^
Exposure
Duration Effect of Particles
20 min Noncytotoxic concentrations of As, V, and Zn induced
6 and 24 h a rapid phosphorylation of MAPK in BEAS cells;
activity assays confirmed marked activation of ERK,
JNK, and P38 in BEAS cells exposed to As, V, and Zn.
Cr and Cu exposure resulted in a relatively small
activation of MAPK, whereas Fe and Ni did not
activate MAPK under these conditions;
the transcription factors c-Jun and ATF-2, substrates of
JNK and P38, respectively, were markedly
phosphorylated in BEAS cells treated with As, Cr, Cu,
V, and Zn;
acute exposure to As, V, or Zn that activated MAPK
was sufficient to induce a subsequent increase in IL-8
protein expression in BEAS cells.
Reference
Samet et al.
(1998)
O
Tl
H
6
O
2
o
H
O
c
o
H
W
O
&
O
I-H
H
W
A549
0X174 RFI
DNA
Rat (Wistar) AM
RAM cells (a rat
AM cell line)
A549
A549
RLE-6TN cells
(type II like cell
line)
Urban particles: In vitro
SRM 1648, St.
Louis, MO
SRM 1649,
Washington, DC
TiO2 In vitro
ROFA, a-quartz, In vitro
TiO2
TiO2, Fe2O3, In vitro
CAPs, and the
fibrogenic
particle a-quartz
PM2 s, In vitro
Burlington, VT;
Fine/ultrafine
TiO2
1 mg/ml for Fe
mobilization assay
20, 50, 80 ,ug/ml
1 mg/ml
TiO2 [40 Mg/mL],
Fe2O3 [100/^g/mL],
a-quartz
[200 Mg/mL], or
CAPs [40 Mg/mL]
1, 2.5, 5, 10 ,ug/ml
SRM 1648: 50% < Up to 25 h
10 ,um
SRM 1649: 30% <
10 ,um
N/A 4h
N/A 60 min
N/A 24 h
PM25: 39 nm 24 and 48 h
Fine TiO2: 159 nm exposure
UF TiO2: 37 nm
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.
Opsonization of TiO2 with surfactant components
resulted in a modest increase in AM uptake compared
with that of unopsonized TiO2;
surfactant components increase AM phagocytosis of
particles
Exposure of A549 cells to ROFA, a-quartz, but not
TiO2, caused increased IL-8 production in TNF-a
primed cells in a concentration-dependent manner
TiO2 > Fe2O3 > a-quartz > CAP in particle binding;
binding of particle was found to be calcium-dependent
for TiO2 and Fe2O3, while a-quartz binding was
calcium-independent;
scavenger receptor, mediate particulate binding;
a-quartz, but not TiO2 or CAP, caused a dose-
dependent production of IL-8;
Increases in c-jun kinase activity, levels of
phosphorylated c-Jun immunoreactive protein, and
transcriptional activation of activator protein- 1-
dependent gene expression; elevation in number of
cells incorporating 5'-bromodeoxyuridine
Smith and Aust
(1997)
Stringer and
Kobzik
(1996)
Stringer and
Kobzik (1998)
Stringer et al.
(1996)
Timblin et al.
(1998)
-------
o
o
r+
O
VO
VO
VO
TABLE 7-3 (cont'd). EFFECTS OF PARTICULATE MATTER EX VIVO
Species, Cell
type, etc.
BEAS-2B
human bronchial
epithelial cells
Particle or
Constituent
ROFA
Birmingham,
AL. 188mg/gr
ofVO
Exposure
Technique Concentration
In vitro 100/^g/ml
Particle Size
(//m) Exposure Duration Effect of Particles
N/A 2-6 h ROFA caused increased intracellular Ca++, IL-6,
IL-8, TNF-a through activation of capsicin- and
pH-sensitive receptors
Reference
Veronesi et al.
(1999)
AM - Alveolar Macrophage
CAP - Concentrated Ambient Particles
CB - Carbon Black
CFA-Coal Fly Ash
CMP - Copper Smelter Dust
DBF - Deferoxamine or Desferrioxamine
DMTU - Dimethylthiourea
DNA - Deoxyribonucleic Acid
DPM - Diesel Particulate Matter
ERK - Extracellular Receptor Kinases
F-344 - Fischer 344
Fe2O3 - Iron Oxide
Fn - Fibronectin
G-6-PDH - Glucose-6-Phosphate
Dehydrogenase
IL-6; 11-8 - Interleukin 6,8
INK - Jun H2-Terminal Kinases
LCL - Luminol Enhanced Chemiluminescence
LDH - Lactate Dehydrogenase
LPS - Liopopolysaccharide (Endotoxin)
MAPK - Mitogen-Activated Protein Kinases
MCT - Monocrotaline
MMAD - Mass Median Aerodynamic Diameter
mRNA - Messenger Ribonucleic Acid
NF-KB - Nuclear factor-kappa B
NHBE - Normal Human Broncial Epithelial Cells
PHS - Prostaglandin H Synthetase
PMA - Phorbol Myristate Acetate
PMN - Neutrophil
PTFE - Polytetrafluoroethylene
ROFA - Residual Oil Fly Ash
RTE - Rat Tracheal Epithelial Cells
SD - Sprague Dawley
SIC AM-1 - Soluble Intercellular Adhesion Molecule-1
SiO2 - Silicon Dioxide
SOD - Superoxide Dismutase
SR - Scavenger-Type Receptors
Ti O2 - Titanuim Dioxide
TNF - Tumor Necrosis Factor
TSP - Total Suspended Paticles
UAP - Urban Air Particles
UF - Ultrafine
WC - Tungsten Carbide
Wis - Wistar
o
£
Tj
H
6
O
o
H
O
O
H
W
O
O
I-H
H
W
-------
1 replicated the biological effects observed with particles alone. When normal constituents of
2 airway lining fluid (mucin or ceruloplasmin) were added to BEAS cells, particulate-induced
3 secretion of IL-8 was modified. Mucin reduced IL-8 secretion, while ceruloplasmin significantly
4 increased IL-8 secretion and activation of NF-KB. The authors suggest that copper ions may
5 cause some of the biologic effects of inhaled PM in the Provo region and may provide an
6 explanation for the sensitivity of asthmatics to Provo PM, seen in epidemiologic studies.
7 Goldsmith et al. (1998) investigated intracellular oxidant production in normal hamster
8 AMs upon in vitro exposure to concentrated ambient particles (CAPs), ROFA, and their
9 water-soluble and particulate fractions. ROFA and CAPs caused increases in dichlorofiuorescin
10 (DCFH) oxidation, a fluorescent measure of intracellular reactive oxygen species (ROS)
11 production, comparable to the positive control, phorbol myristate acetate (PMA). The
12 water-soluble component of both CAPs and ROFA significantly increased AM oxidant
13 production over negative control. CAPs samples and components showed substantial day-to-day
14 variability in their oxidant effects. Metal chelation by DBF (1 mM) caused significant inhibition
15 of PM-induced AM oxidant production. ROFA exposure resulted in increased macrophage
16 inflammatory protein-2 (MIP-2) mRNA in AMs and in increased TNF-cc production by the
17 monocyte-macrophage cell line, RAW 264.7. TNF-cc production was inhibitable by the
18 antioxidant, N-acetylcysteine (NAC). The data suggest that metal components of urban air PM
19 can significantly contribute to the ability of particles to cause oxidant stress and cytokine
20 production in AMs.
21 Fabiani et al. (1997) investigated the response of peripheral blood monocytes to extracts of
22 ambient particles. Particles in 1 m3 of air were collected on glass fiber filters and were extracted
23 with acetone for 18 h. At a particulate concentration of 0.17 mg/mL, the production of
24 superoxide (O2~) in peripheral blood monocytes was reduced to 22% and 40% of the control
25 values when the cells were stimulated with PMA and Zymosan, respectively. Concomitantly,
26 there was a release of LDH into the supernatant (50% of the total cellular LDH activity),
27 indicating that a large proportion of cells were damaged by the treatment with the PM extract,
28 and some cytosolic components were released from the cells. Giemsa staining of the treated
29 monocytes revealed the presence of many cells with a dispersed cytosol; the nucleus, although
30 not destroyed, had a different shape. This study suggested that the airborne particulate matter has
31 a toxic effect that induces the disintegration of the plasma membrane. Cytosolic factors (proteins
October 1999 7-73 DRAFT-DO NOT QUOTE OR CITE
-------
1 and coenzymes) necessary for O2 production leak from the cells and O2 generation is therefore
2 reduced. It remains to be determined whether this phenomenon also occurs in vivo.
3 Becker and Soukup (1998) examined the effect of urban air particles (UAP) on human
4 alveolar macophages and blood monocytes in vitro. UAP decreased expression of receptors
5 important for phagocytosis of opsonized microbes and for interaction with extracellular matrix
6 components. These responses suggest that clearance of microbes from the lung could be
7 impaired by exposure to UAP.
8 The effects of water soluble as well as organic components of ambient PM were
9 investigated by exposing human PMN to PM extracts (Hitzfeld et al., 1997). PM was collected
10 with high volume samplers in two German cities, Dusseldorf and Duisburg; these sites have high
11 traffic and high industrial emissions, respectively. The collected particles were extracted using
12 water and then dichloromethane, resulting in an aqueous and an organic extract. The production
13 of ROS was determined using luminol-enhanced chemiluminescence (LCL) of resting and
14 zymosan-stimulated PMN. Organic, but not aqueous, extracts of PM alone significantly
15 stimulated the production and release of ROS in resting human PMN. The effects of the PM
16 extracts were inhibited by SOD, catalase and sodium azide (NaN3). Zymosan-induced LCL was,
17 however, diminished by coincubation with PM extracts, with the aqueous extracts having a
18 stronger inhibitory effect. The mineralogical and chemical characterization of the PM showed
19 that the particles were small (>90% <5//m) and consisted mainly of quartz and other silicates
20 with no difference between the two sites. These authors speculated that PAHs in the organic
21 extract and metals in the aqueous extract may be responsible for the biologic effects.
22 Mechanisms other than oxidative stress have been proposed to explain cytokine production
23 by epithelial cells. Dong et al. (1996) suggest that activation of cytokine gene expression and
24 secretion in rat AM treated with UAP (collected in St. Louis) is due to the presence of endotoxin
25 on the particles. Cytokine production (TNF-cc, IL-1,11-6, CINC, and MIP-2) was increased in
26 macrophages following treatment with 50 to 200 //g/mL of urban PM. The authors suggest that
27 cytokine expression was not related to ROS because antioxidants, such as catalase, had no effect
28 on TNF-cc secretion. However the collection system for the UAP did not exclude large particles
29 and biological materials. Given previous observations that LPS stimulates cytokine secretion in
30 AM (Yoo et al., 1995), the authors suggest that LPS is responsible for the observed cytokine
31 responses. The UAP-induced TNF-cc secretion was completely abrogated by treatment with
October 1999 7-74 DRAFT-DO NOT QUOTE OR CITE
-------
1 polymyxin B, an antibiotic that blocks LPS-associated activities (Morrison and Jacobs, 1976).
2 Polymixin B is also an inhibitor of protein kinase C (Cave and Apstein, 1996) and an antagonist
3 of calmodulin (Hegemann et al., 1991) and may alter cellular responses by mechanisms other
4 than inhibition of LPS. Thus, polymixin B may inhibit PM cellular responses by mechanisms
5 other than LPS inhibition. Extrapolation of these in vitro results to a potential role for endotoxin
6 in the adverse effects of ambient PM must be done with caution because the investigators could
7 not exclude the possibility that the presence of endotoxin with the PM was not due to inadvertent
8 contamination during the year long collection process or from the handling of the particles.
9 Urban PM10 collected from north, south, and central regions of Mexico City was used with
10 SD rat AM to examine PM effects on platelet derived growth factor (PDGF) receptors on lung
11 myofibroblasts (Bonner et al., 1998). Mexico City PM10 (but not volcanic ash) stimulated
12 secretion of upregulatory factors for the PDGF a receptor, possibly via IL-1 p. In the presence of
13 an endotoxin-neutralizing protein, the Mexico City PM10 effect on PDGF was partially blocked,
14 suggesting that LPS was partially responsible for the effect of the PM10 on macrophages.
15 In addition, both LPS and vanadium (both present in the PM10) acted directly on lung
16 myofibroblasts. However, the V levels in Mexico City PM10 were probably not high enough to
17 exert an independent effect. The authors concluded that PM10 exposure may lead to airway
18 remodeling by enhancing myofibroblast replication and chemotaxis.
19
20 ROFA
21 Several in vitro methodologies have been employed to better examine the role of oxidative
22 stress in particle-induced lung injury. Carter et al. (1997) exposed normal human bronchial
23 epithelial (NHBE) cells for either 2 or 24 h to 0, 5, 50, or 200 //g/mL ROFA. NHBE cells
24 exposed to ROFA produced significant amounts of IL-8, IL-6 and TNF, as well as mRNAs
25 coding for these cytokines. Increases in cytokine production, but not m-RNA expression, were
26 dose-dependent. The cytokine production was inhibited by the addition of either the metal
27 chelator, DBF, or the free radical scavenger, DMTU. In addition, the authors reported that
28 V containing compounds, but not Fe or Ni sulfates, mimicked the effects of ROFA. The authors
29 suggest that certain metals associated with ROFA may be responsible for the production and
30 release of inflammatory mediators by the respiratory tract epithelium and that these mediators
31 may contribute to the toxic effects of PM. Furthermore, the finding that the free radical
October 1999 7-75 DRAFT-DO NOT QUOTE OR CITE
-------
1 scavenger DMTU effectively prevents the induction of IL-6 and IL-8 mRNA expression by
2 ROFA is consistent with an oxidant-dependent mechanism.
3 Dye et al. (1997) exposed primary cultures of rat tracheal epithelial (RTE) cells to
4 suspensions of ROFA (5, 10, or 20 //g/cm2) for 24 h. ROFA exposure resulted in cytotoxicity
5 and detachment of cells from the collagen matrix, along with altered permeability of the RTE cell
6 layer. In addition ROFA exposure caused the cellular glutathione levels to decrease, producing a
7 condition of oxidative stress. Treatment with buthionine sulfoxamine, an irreversible inhibitor of
8 the rate-limiting enzyme involved in glutathione synthesis (y-glutamyl cysteine synthetase)
9 augmented ROFA-induced cytotoxicity. While treatment with DMTU inhibited ROFA-induced
10 LDH release and permeability changes in a dose-dependent manner, inhibition of nitric oxide
11 synthesis had no effect on ROFA toxicity. The authors suggested that ROFA-induced RTE cell
12 injury may be mediated by hydroxyl-radical-like ROS (scavenged by DMTU) that are generated
13 via non-nitric oxide pathways.
14 Using a human airway epithelial cell line (BEAS-2B), Quay et al. (1998) observed that
15 ROFA stimulated a time- and dose-dependent increase in IL-6 mRNA, which was preceded by
16 the activation of nuclear proteins binding to the NF-KB sequence motif in the IL-6 promoter
17 region. Transient transfection of BEAS-2B cells with the 5' promoter region of the IL-6 gene
18 linked to a luciferase reporter gene confirmed that NF-KB binding is necessary for the
19 transcription of IL-6 mRNA. The IL-6 response was inhibited by the metal chelator DBF and the
20 free radical scavenger NAC, suggesting that the activation of NF-KB may be mediated through
21 ROS generated by transition metals found in ROFA. These data suggest that activation of
22 NF-KB may therefore be a critical first step in the inflammatory cascade following exposure to
23 ROFA particles.
24 Prostaglandin metabolism in cultured human airway epithelial cells (BEAS-2B and NHBE)
25 exposed to ROFA was investigated by Samet et al. (1996). Epithelial cells exposed to ROFA for
26 24 h secreted substantially increased amounts of prostaglandins E2 and F2cc. The ROFA-induced
27 increase in prostaglandin synthesis was correlated with a marked increase in activity of the
28 PHS-2 form of prostaglandin H synthase. Increases in PHS2 mRNA were evident by 2 h of
29 continuous ROFA exposure. In contrast, expression of the PHS1 form of the enzyme was not
30 affected by ROFA treatment of airway epithelial cells. These data show that exposure to ROFA
October 1999 7-76 DRAFT-DO NOT QUOTE OR CITE
-------
1 induces PHS2 expression, leading to increased prostaglandin synthesis in cultured human airway
2 epithelial cells.
3 Samet et al. (1997) also investigated the effect of ROFA on protein tyrosine phosphate
4 metabolism in BEAS-2B cells. Non cytotoxic levels of ROFA induced a significant dose- and
5 time-dependent increase in protein tyrosine phosphate, an important regulator of signal
6 transduction leading to cell growth and proliferation. ROFA-induced increases in protein
7 phosphotyrosines were associated with its soluble fraction and were mimicked by vanadyl
8 [V(rV)]- and vanadate [V(V)]-containing solutions. Tyrosine phosphatase activity is known to
9 be inhibited by V. Tyrosine kinase activity was unaffected. Ferrous, ferric, and nickel (II) ion
10 solutions failed to increase phosphotyrosine levels. It is possible that ROFA exposure induces
11 V ion-mediated inhibition of tyrosine phosphatase activity, leading to accumulation of protein
12 phosphotyrosines in BE AS cells. These findings demonstrate that ROFA exposure disrupts
13 protein tyrosine phosphate homeostasis in BEAS cells.
14 Stringer and Kobzik (1998) observed that "primed" lung epithelial cells exhibited enhanced
15 cytokine responses to PM. Compared to normal cells, exposure of TNF-a-primed A549 cells to
16 ROFA or a-quartz caused increased IL-8 production in a concentration-dependent manner for
17 particle concentrations ranging from 0-200 //g/ml. Addition of the antioxidant NAC (1.0 mM)
18 decreased ROFA and cc-quartz -mediated IL-8 production by approximately 50% in both normal
19 and TNF-a-primed A549 cells. Exposure of A549 cells to ROFA caused an increase in oxidant
20 levels that could be inhibited by NAC. These data suggest that (1) lung epithelial cells primed by
21 inflammatory mediators show increased cytokine production after exposure to PM, and
22 (2) oxidant stress is an important mechanism for this response.
23 Becker et al. (1996) suggested that the response elicited by oil fly ash (OFA) or UAP may
24 not be due solely to metal components. In this study, AM were tested for a chemiluminescence
25 response (ROS generation) induced by the particles, as well as for IL-6 and TNF-a production.
26 The OFA dose ranged from 10-1000 //g per 2-3 x 105 AM. Acute cytotoxicity of OFA and SiO2
27 was observed above 100 //g in both human and rat AMs (LDH release at 2 h). Diesel particulate
28 matter (DPM) was found to be nontoxic even at the highest dose. However, after 20 h of
29 coincubation, UAP concentrations >167 //g/ml were also somewhat cytotoxic. Subcytotoxic
30 concentrations of OFA induced a strong immediate chemiluminescence response by AMs.
31 A small but significant chemiluminescence response was also induced by two out of three UAP
October 1999 7-77 DRAFT-DO NOT QUOTE OR CITE
-------
1 tested, but no chemiluminescence was generated in response to DPM. The magnitude of
2 particle-induced chemiluminescence was not predictive of a cytokine response by either human
3 or rat AMs. TNF-cc and IL-6 production was strongly induced by UAP, but not DPM or OFA,
4 over a range of noncytotoxic concentrations of particles. The AM cytokine response to UAP was
5 partly inhibitable by polymyxin B, but not by the iron chelator DBF, indicating that endotoxin,
6 but not transitional iron, induced IL-6 production in the UAP preparations.
7 In summary, exposure of lung cells to non-cytotoxic concentrations of ambient PM or
8 ROFA leads to increased production of cytokines and the effects may be mediated, at least in
9 part, through production of ROS. Day-to-day variations in the components of PM which are
10 critical to eliciting the response are also suggested. The involvement of organic components in
11 ambient PM as well as ROFA was also suggested in some studies.
12
13 7.3.5 Susceptibility to the Effects of PM Exposure
14 Susceptibility of an individual to adverse health effects of PM can vary depending upon a
15 variety of host factors such as age, nutritional status, physiological activity profile, genetic
16 predisposition, or preexistent disease. The potential for pre-existent disease to alter adverse
17 response to toxicant exposure is widely acknowledged but poorly understood. Due to inherent
18 variability and ethical concerns associated with using diseased subjects in clinical research
19 studies, a solid database on human susceptibilities is lacking. For more control over both host
20 and environmental variables, animal models are often used. However, care must be taken in
21 extrapolation from animal models of human disease to humans. Rodent models of human
22 disease, their use in toxicology and the criteria for judging their appropriateness as well as their
23 limitations have been reviewed (Kodavanti et al., 1998b; Kodavanti and Costa, 1999).
24
25 7.3.5.1 Effects of PM on Cardiopulmonary Compromised Hosts
26 Epidemiological studies suggest there may be subsegments of the population that are
27 especially susceptible to effects from inhaled particles (see Chapter 6). The elderly with chronic
28 cardiopulmonary disease, those with pneumonia and possibly other lung infections, and those
29 with asthma (at any age) appear to be at higher risk than healthy people of similar age. Most
30 toxicology studies have used healthy adult animals. Few studies have examined effects of
31 ambient particles in compromised host models. Costa and Dreher (1997) used a rat model of
October 1999 7-78 DRAFT-DO NOT QUOTE OR CITE
-------
1 cardiopulmonary disease to explore the question of susceptibility and the possible mechanisms
2 by which PM effects are potentiated. Rats with advanced monocrotaline (MCT)-induced
3 pulmonary vasculitis/hypertension were given intratracheal instillations of ROFA (0, 0.25,
4 1.0 and 2.5 mg/rat). The MCT-treated animals had a marked neutrophilic inflammation. In the
5 context of this inflammation, ROFA induced a 4-5 fold increase in BAL PMNs. There was an
6 increased mortality at 96 h that was ROFA-dose dependent. The results of this study indicate
7 that PM enhanced the neutrophilic inflammation and mortality in MCT-treated animals.
8 The manner in which treatment with MCT can alter the response of rats to inhaled particles
9 was examined by Madl and colleagues (1998). Rats were exposed to fluorescent colored
10 microspheres (1 //m) 2 weeks after treatment with MCT. In vivo and in vitro phagocytosis of the
11 microspheres was altered in the MCT-treated rats in comparison with control animals. Fewer
12 microspheres were phagocytized in vivo by alveolar macrophages and there was a concomitant
13 increase in free microspheres overlaying the epithelium at airway bifurcations. The decrease in
14 in vivo phagocytosis was not accompanied by a similar decrease in vitro. Macrophage
15 chemotaxis, however, was significantly impaired in MCT-treated rats compared with control rats.
16 Thus, MCT-treatment appeared to impair particle clearance from the lungs via inhibition of
17 macrophage chemotaxis.
18 The sulfur dioxide (SO2)-induced model of chronic bronchitis has also been used to
19 examine the potential interaction of PM with pre-existing lung disease. Clarke and colleagues
20 pretreated rats for 6 weeks with air or 170 ppm SO2 for 5 hours/day and 5 days/week (Clarke
21 et al., 1999). Exposure to concentrated air particles for 5 hours/day for 3 days at an average
22 concentration of 515 //g/m3 produced changes in pulmonary function as evidenced by significant
23 increases in tidal volume in both air- and SO2-pretreated rats. Exposure to concentrated ambient
24 PM also produced significant changes in cellular and biochemical markers in lavage fluid.
25 In comparison to control animal values, protein was increased approximately 3-fold in
26 SO2-pretreated animals exposed to concentrated ambient PM. Lavage fluid neutrophils and
27 lymphocytes were significantly increased in both pretreatment groups of rats exposed to
28 concentrated ambient PM, with greater increases in both cell types in the SO2-pretreated rats.
29 Thus, exposure to concentrated ambient PM produced adverse changes in the respiratory system
30 in both normal rats and in a rat model of chronic bronchitis. Although the exposure
31 concentrations were much higher than those encountered in urban centers in North America,
October 1999 7-79 DRAFT-DO NOT QUOTE OR CITE
-------
1 these findings are important because few studies have used actual ambient urban PM in
2 inhalation exposures.
3 In a study to investigate the influence of age on susceptibility to PM, Johnston et al. (1998)
4 exposed 8 week old mice (young) and 18 month old mice (old) to polytetrafluoroethylene fumes
5 (PTFE) (0, 10, 25, and 50 //g/m3) for 30 min. Lung lavage endpoints (PMN, protein, LDH, and
6 p-glucuronidase) as well as lung tissue mRNA levels for various cytokines, metallothionein and
7 for Mn superoxide dismutase were measured 6 hours following exposure. Protein, lymphocyte,
8 PMN, and TNF-cc mRNA levels were increased in older mice when compared to younger mice.
9 These findings suggest that the inflammatory response to PTFE fumes is altered with age, being
10 greater in the older animals. Although Teflon particles are not a valid surrogate for ambient
11 ultrafine particles, this study did provide evidence to support the hypothesis that particle-induced
12 pulmonary inflammation is different between the young and old organism.
13 In summary, although these studies are just emerging and are only now being replicated or
14 followed more thoroughly to investigate the mechanisms, they do provide evidence of enhanced
15 susceptibility to inhaled PM in "compromised" hosts.
16
17 7.3.5.2 Effect of PM on Allergic Hosts
18 Relatively little is known about the effects of inhaled particles on humoral (antibody) or
19 cell-mediated immunity. Alterations in the response to a specific antigenic challenge have been
20 observed in animal models at high concentrations of acid sulfate aerosols (above 1,000 //g/m3)
21 (Pinto et al., 1979; Kitabatake et al., 1979; Fujimaki et al., 1992). Several studies have reported
22 an enhanced response to non-specific bronchoprovocation agents, such as acetylcholine and
23 histamine, after exposure to inhaled particles. This non-specific airway hyperresponsiveness,
24 a central feature of asthma, occurs in animals and human subjects exposed to sulfuric acid under
25 controlled conditions (Gearhart and Schlesinger, 1986; Utell et al., 1983b). Although, its
26 relevance to specific allergic responses in the airways of atopic individuals is unclear, it
27 demonstrates that the airways of asthmatics may become sensitized to either specific or
28 non-specific triggers that could result in increases in asthma severity and asthma-related hospital
29 admissions (Peters et al., 1997; Jacobs et al., 1997; Lipsett et al., 1997).
30
31
October 1999 7-80 DRAFT-DO NOT QUOTE OR CITE
-------
1 Nel et al. (1998) have suggested that the rise in the U.S. prevalence rate for allergic rhinitis
2 (5% in the 1950s to about 20% in the 1980s) may be related to increased DPM, in addition to
3 other combustion related PM. Combustion particles may also serve as carrier particles for
4 allergens (Knox et al., 1997).
5 A number of in vivo and in vitro studies have demonstrated that DPM can alter the immune
6 response to challenge with specific antigens and suggest that DPM may act as an adjuvant.
7 These studies have shown that treatment with DPM enhances the secretion of antigen-specific
8 IgE in mice (Takano et al., 1997) and in the nasal cavity of human subjects (Diaz-Sanchez et al.,
9 1996, 1997; Ohtoshi et al., 1998). Because IgE levels play a major role in allergic asthma
10 (Wheatley and Platts-Mills, 1996), upregulation of its production could lead to an increased
11 response to inhaled antigen in particle-exposed individuals (See Table 7-4).
12 Only a small number of studies have examined the mechanisms underlying the
13 enhancement of allergic asthma by ambient urban particles. Ohtoshi et al. (1998) reported that a
14 coarse size-fraction of resuspended ambient PM, collected in Tokyo, induced the production of
15 granulocyte macrophage colony stimulating factor (GMCSF), an upregulator of dendritic cell
16 maturation and lymphocyte function, in human airway epithelial cells in vitro. In addition to
17 increased GMCSF, epithelial cell supernatants contained increased IL-8 levels when incubated
18 with DPM, a principal component of ambient particles collected in Tokyo. Although, the sizes
19 of the two types of particles used in this study were not comparable, the results suggest that
20 ambient PM, or at least the DPM component of ambient PM, can upregulate the immune
21 response to inhaled antigen through GMCSF production. Similarly, Tokano et al. (1998)
22 reported airway inflammation, airway hyperresponsiveness, and increased GM-GSF and IL-5 in
23 diesel exhaust exposed mice.
24 Gavett et al. (1999) have investigated the effects of ROFA (intratracheal instillation) in
25 ovalbumin (OVA) sensitized and challenged mice. ROFA induced inflammatory and
26 physiological responses in the OVA mice that were related to increases in Th2 cytokines (IL-4,
27 IL-5). ROFA induced greater than additive increases in eosinophil numbers and in airway
28 responsiveness to methacholine.
29 Goldsmith et al. (1999) have compared the effect of inhalation of concentrated ambient PM
30 for 6 hours/day for 3 days versus the effect of a single exposure to a ROFA leachate aerosol on
31 the airway responsiveness to methacholine in OVA-sensitized mice. Daily exposure to ROFA
October 1999 7-81 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 7-4. IMMUNOLOGICAL EFFECTS OF PARTICIPATE MATTER
o
r+
O
&
CD
H^
H^
*o
>o
1 ^\
>i^
Species, Gender,
Strain Age, or
Body Weight
Rats, Fischer 344,
males, 10 weeks
Human subjects,
with and without
seasonal allergies
i.r»A»A_;ij / T. iiTJ-LTiV
Exposure
Particle Technique
ammonium Nose-only
metavanadate inhalation,
aqueous
aerosol
DPM 200 fA aerosol
delivered
directly into
each naris.
n V^AJV^VJAX_>,rm.J
Mass
Concentration
(//g) or (Mg/m3)
2000 A^g/m3 as
vanadium
150,agof
particles per
naris
LJ iJJi1 i1 iJJX_> i kj V^i1 A ^-iAXi AX_> \J 1
Particle
Characteristics Exposure
and Size (//m) Duration
0.32 /zm 8 hours/day for
aqueous aerosol 4 days
not specified Single treatment
with vehicle or
particles
L/^-i i iJJ IT-Lrm. i i iJJAX
Respiratory Effects of
Inhaled Particles
Increased PMN, PAM, LDH, and
protein in BAL. Immune
competence of PAM also inhibited.
Enhanced expression of IL-4, IL-6,
and IL-13, cytokines known to be
B-cell proliferation factors; increases
in several other cytokines mRNA;
increased IgE in nasal lavage fluid.
Reference
Cohen et al.
(1996)
Diaz-Sanchez
etal. (1996)
•-J
oo
**s
$
Tl
H
O
O
o
H
o
H
W
O
&
O
HH
H
W
Human subjects
with allergic rhinitis
and positive skin
test to ragweed
DPM with or
without
ragweed
antigen
200 jA aerosol
delivered
directly into
each naris
ISO^gof
particles per
naris
not specified
Single treatment
with ragweed
antigen followed
by a single
treatment with
antigen and
particles
Human airway
epithelial cells
Mice, ICR and
WAV, male,
4 weeks
UAP, DPM
DPM
In vitro
Tracheal
instillation
2.5 to
2500 /zg/ml
UAP and 10 to
100//g/ml
DPM
100or200,ag
per instillation
UAP = 7 to
10 fj,m;
DPM = mean
diameter of
0.4^m
not specified
Treated in vitro
for up to 48 h
Instilled
once/week for
5 to 1 6 weeks
Combined treatment with ragweed Diaz-Sanchez
antigen and particles resulted in et al. (1997)
significantly greater antigen-specific
IgE and IgG4. Combined challenge
also decreased expression of
Thl-type cytokines and increased
expression of IL-4, IL-5, IL-10, and
IL-13.
Increased release of both GMCSF Ohtoshi et al.
and IL-8 by epithelial cells treated (1998)
with DPM, whereas only GMCSF
was released by epithelial cells
treated with UAP.
Repeated instillation of DPM Sagai et al.
produced goblet cell hyperplasia, (1996)
eosinophilia, airway constriction,
and airway hyperresponsiveness to
acetylcholine. These changes were
largely blocked by treatment with
PEG-derivitized superoxide
dismutase.
-------
o
r+
O
&
CD
H^
>o
>o
>o
Species, Gender,
Strain Age, or
Body Weight
Mice, ICR, male, 6
to 7 weeks
Particle
DPM with or
without
ovalbumin
Exposure
Technique
Intratracheal
instillation
Mass
Concentration
100 Mgper
instillation
Particle
Characteristics
and Size (/^m)
mean diameter of
Exposure
Duration
Instilled
once/week
for 6 or
9 weeks
Respiratory Effects of
Inhaled Particles Reference
Enhanced production of antigen- Takano et al
specific IgG and IgE in combined f 1997)
DPM/ovalbumin animals compared to
ovalbumin-alone animals. A similar
increase in inflammatory cells and
cytokines in lavage fluid of animals
treated with DPM/ovalbumin.
Transformed IgE
producing human B
cell line
organic extract
ofPAH-DPM
in vitro
0.1 ng/ml
PAH-DPM
dichloromethane
extraction of
large chain
aggregates with
individual
particles <0.1 /^m
Treated in PAH-DPM increased IgE production Tsien et al.
vitro for and altered the levels of epsilon (1997)
72 hours mRNA variants, but did not increase
cytokine production.
•-J
oo
**s
$
Tl
H
O
O
DPM - Diesel Exhaust Particles
UAP - Urban Air Particles
IL-4, IL-5, 1L-6, IL-8, IL-10, IL-13 - Interleukins
PEG - Polyethylene Glycol
IgE, IgG4 - Immunoglobulins
Th-1 Tl Helper Cell
GMCSF - Granulocyte Macrophage Colony Stimulating Factor
o
H
o
H
W
O
O
HH
H
W
-------
1 leachate aerosols significantly enhanced the airway hyperresponsiveness in OVA-sensitized
2 mice. Importantly, exposure to concentrated ambient PM (average concentration of 787 //g/m3)
3 had no effect on airway responsiveness in 6 separate experiments. Thus, the effect of the ROFA
4 leachate aerosols on the induction of airway hyperresponsiveness in allergic mice was
5 significantly different than that of a high concentration of concentrated ambient PM. Although
6 airway responsiveness was examined at only one post-exposure time point, these findings
7 suggest that a great deal of caution should be used in interpreting the results of studies using
8 ROFA particles or leachates in the attempt to investigate the biologic plausibility of the adverse
9 health effects of PM.
10 Hamada et al. (1999) have examined the effect of ROFA leachate aerosol in a neonatal
11 mouse model of allergic asthma. OVA-sensitized, neonatal mice developed airway
12 hyperresponsiveness, eosinophilia, and elevated serum anti-ovalbumin IgE after a challenge with
13 inhaled OVA. Exposure to the ROFA leachate aerosol had no marked effect on the airway
14 responsiveness to inhaled methacholine in non-sensitized mice, but did enhance the airway
15 hyperresponsiveness to methylcholine produced in OVA-sensitized mice. No other interactive
16 effects of ROFA exposure with OVA were observed. Lambert et al. (1999) also examined the
17 effect of ROFA on a rodent model of pulmonary allergy. Rats were instilled intratracheally with
18 200 or 1,000 //g ROFA 3 days prior to sensitization with house dust mite antigen. HDM
19 sensitization after 1000 //g ROFA produced increased eosinophils, LDH, BAL protein, and IL-10
20 relative to HDM alone. The immediate bronchoconstrictive and associated antigen-specific IgE
21 response to a subsequent antigen challenge was increased in the ROFA-treated group in
22 comparison with the control group. Together, these studies suggest the components of ROFA
23 can augment the immune response to antigen.
24 Several other studies have examined in greater detail the contribution to allergic asthma of
25 the particle component and the organic fraction of DPM. Tsien et al. (1997) treated transformed
26 IgE-producing human B lymphocytes in vitro with the organic extract of DPM. The organic
27 phase extraction had no effect on cytokine production but did increase IgE production.
28 Moreover, these experiments determined that DPM appeared to be acting on cells already
29 committed to IgE production, thus suggesting a mechanism by which the organic fraction of
30 combustion particles can directly affect B cells and influence human allergic asthma.
October 1999 7-84 DRAFT-DO NOT QUOTE OR CITE
-------
1 Cultured epithelial cells from atopic asthmatics show a greater response to diesel
2 particulate matter (DPM) exposure when compared with cells from non-atopic non-asthmatics.
3 IL-8, GM-CSF, and soluble ICAM-1 increased in response to DPM at a concentration of
4 10 //g/ml DPM (Bayram et al., 1998a,b). This study suggests that particles could modulate
5 airway disease through their actions on airway epithelial cells. This study also suggests that
6 bronchial epithelial cells from asthmatics are different from those of nonasthmatics in regard to
7 their mediator release in response to DPM.
8 Sagai and colleagues (1996) repeatedly instilled mice with DPM for up to 16 weeks and
9 found increased numbers of eosinophils, goblet cell hyperplasia, and non-specific airway
10 hyperresponsiveness, changes which are central features of chronic asthma (National Institutes of
11 Health, 1997). Takano et al. (1997) extended this line of research and examined the effect of
12 repeated instillation of DPM on the specific response to antigen (OVA) in mice. They observed
13 that antigen-specific IgE and IgG levels were significantly greater in mice repeatedly instilled
14 with both DPM and OVA. Because this upregulation in antigen-specific immunoglobulin
15 production was not accompanied by an increase in inflammatory cells or cytokines in lavage
16 fluid, it would suggest that, in vivo, DPM may act directly on immune system cells, as described
17 in the work by Tsien et al. (1997).
18 Diaz-Sanchez and colleagues (1996) have continued to study the mechanism of
19 DPM-induced upregulation of allergic response in the nasal cavity of human subjects. In one
20 study, a 200 //I aerosol bolus containing 0.15 mg of DPM was delivered into each naris of
21 subjects with or without seasonal allergies. In addition to increases in IgE in nasal lavage fluid
22 (NAL), they found an enhanced production of IL-4, IL-6, and IL-13, cytokines known to be
23 B cell proliferation factors. The levels of several other cytokines were also increased, suggesting
24 a general inflammatory response to a nasal challenge with DPM. In a following study, these
25 investigators delivered ragweed antigen, alone or in combination with DPM, on two occasions, to
26 human subjects with both allergic rhinitis and positive skin tests to ragweed (Diaz-Sanchez et al.
27 1997). They found that the combined challenge with ragweed antigen and DPM produced
28 significantly greater antigen-specific IgE and IgG4 in NAL. A peak response was seen at 96 h
29 postexposure. The combined treatment also induced expression of IL-4, IL-5, IL-10, and IL-13,
30 with a concomittant decrease in expression of Thl-type cytokines. Although the treatments were
31 not randomized (antigen alone was given first to each subject), the investigators reported that
October 1999 7-85 DRAFT-DO NOT QUOTE OR CITE
-------
1 pilot work showed no interactive effect of repeated antigen challenge on cellular and biochemical
2 markers in NAL. DPM also resulted in the nasal influx of eosinophils, granulocytes, monocytes,
3 and lymphocytes as well as production of various inflammatory mediators. The combined DPM
4 plus ragweed exposure did not increase the rhinitis symptoms beyond those of ragweed alone.
5 Blomberg et al. (1998) observed a 10-fold increase in NAL fluid ascorbate concentration
6 after a 1 h exposure to diluted diesel exhaust (300 //g/m3 particles and 1.6 ppm NO2). However,
7 there were no effects on BAL ascorbate levels. Rudell et al. (1990) had previously shown
8 increased BAL neutrophils in nonsmoking subjects exposed to 100 //g/m3 of DPM in diesel
9 exhaust (gases were present). Thus, diesel exhaust (particles and gases) can produce an enhanced
10 response to antigenic material in the nasal cavity. Extrapolation of these findings, of enhanced
11 allergic response in the nose, to the lung, would suggest that ambient combustion particles
12 containing DPM may have significant effects on allergic asthma. These studies provide
13 biological plausibility for the exacerbation of allergic asthma associated with episodic exposure
14 to PM. Although DPM may make up only a fraction of the mass of urban PM, because of their
15 small size, DPM may represent a significant fraction of the ultrafine particle mode in urban air,
16 especially in cities and countries that rely heavily on diesel-powered vehicles.
17
18 7.3.5.3 Resistance to Infectious Disease
19 The development of an infectious disease requires both the presence of the appropriate
20 pathogen, as well as host susceptibility to the pathogen. There are numerous specific and
21 non-specific anti-microbial host defenses against microbes, and the ability of inhaled particles to
22 modify resistance to bacterial infection could result from a decreased ability to clear or kill
23 microbes. Rodent infectivity models have frequently been used to examine the effect of inhaled
24 particles on host defense and infectivity. Mice or rats are challenged with a bacterial or viral load
25 either before or after exposure to the particles (or gas) of interest; mortality rate, survival time, or
26 bacterial clearance are then examined. A number of studies which have used the infectivity
27 model to assess the effect of inhaled PM were discussed previously (U.S. Environmental
28 Protection Agency, 1982, 1989, 1996a). In general, acute exposure to sulfuric acid aerosols at
29 concentrations up to 5,000 //g/m3 were not very effective in enhancing mortality in a
30 bacterially-mediated murine model. In rabbits, however, sulfuric acid aerosols altered
31 antimicrobial defenses after exposure for 2 hours/day for 4 days to 750 //g/m3 (Zelikoff et al.,
October 1999 7-86 DRAFT-DO NOT QUOTE OR CITE
-------
1 1994). Acute or short term repeated exposures to high concentrations of relatively inert particles
2 have produced conflicting results. Carbon black (10,000 //g/m3) was found to have no effect on
3 susceptibility to bacterial infection (Jakab, 1993), while a very high concentration of TiO2
4 decreased the clearance of microbes and the bacterial response of lymphocytes isolated from
5 mediastinal lymph nodes (Gilmour et al., 1989a,b). In addition, exposure to DPM has been
6 shown to enhance the susceptibility of mice to the lethal effects of some, but not all, microbial
7 agents (Hatch et al., 1985; Hahon et al., 1985). Thus, the pulmonary response to microbial
8 agents has been shown to be altered at relatively high particle concentrations in animal models.
9 Moreover, these effects appear to be highly dependent on the microbial challenge and the test
10 animal studied. Pritchard et al. (1996) observed in rats exposed to particles with a high
11 concentration of metals (e.g., ROFA), that the increased mortality rate after streptococcus
12 infection was associated with the amount of metal in the PM.
13 Despite the reported association between ambient PM and deaths due to pneumonia
14 (Schwartz, 1994), there are few recent studies which have examined the mechanisms which may
15 be responsible for the effect of PM on infectivity. In one study, Cohen and colleagues (1997)
16 examined the effect of inhaled V on immunocompetence. Healthy rats were repeatedly exposed
17 to 2 mg/m3 V, as ammonium metavanadate, and then instilled with polyinosinic-polycytidilic
18 acid (poly I:C), a double-stranded polyribonucleotide which acts a potent immunomodulator.
19 Induction of increases in lavage fluid protein and neutrophils was greater in animals pre-exposed
20 to V. Similarly, IL-6 and interferon-gamma were increased in V-exposed animals. Alveolar
21 macrophage function, as determined by zymosan-stimulated superoxide anion production and by
22 phagocytosis of latex particles, was depressed to a greater degree after poly I:C instillation in
23 V-exposed rats as compared to filtered air-exposed rats. These findings provide evidence that
24 inhaled V, a trace metal found in combustion particles and shown to be toxic in vivo in studies
25 using instilled or inhaled ROFA (Dreher et al., 1997), has the potential to inhibit the pulmonary
26 response to microbial agents. It must be taken into consideration that these effects were found at
27 very high occupational exposure concentrations of V, and as with many studies, care must be
28 taken in extrapolating the results to the ambient exposure of healthy individuals or those with
29 pre-existing cardiopulmonary disease to trace concentrations (approximately 3 orders of
30 magnitude lower concentration) of metals in ambient PM.
31
October 1999 7-87 DRAFT-DO NOT QUOTE OR CITE
-------
1 7.4 CARDIOVASCULAR AND OTHER SYSTEMIC RESPONSES TO PM
2 A small number of epidemiology studies have demonstrated that increases in
3 cardiac-related deaths are associated with exposure to PM (U.S. Environmental Protection
4 Agency, 1996a) and that PM-related cardiac deaths appear to be as great or greater than those
5 attributed to respiratory causes. The toxicological consequences of inhaled particles on the
6 cardiovascular system had not been extensively investigated prior to 1996. Since then (see
7 Table 7-5) Costa and colleagues (Costa and Dreher, 1997) have demonstrated that tracheal
8 instillation of ambient particles can increase or accelerate death related to monocrotaline
9 administration in rats. These deaths did not occur with all types of ambient particles tested.
10 Some dusts, such as volcanic ash from Mount St. Helens, were relatively inert, while other
11 ambient dusts, including those from urban sites, were toxic. These observations suggest that
12 particle composition plays an important role in the adverse health effects associated with episodic
13 exposure to ambient PM, despite the 'general particle' effect attributed to the epidemiological
14 associations of ambient PM exposure and increased mortality in many regions of the U.S. (i.e.,
15 regions with varying particle composition). Work which examines the role of inherent
16 susceptibility to the adverse effects of PM in compromised animal models provides a potentially
17 important link to epidemiological observations.
18 Killingsworth and colleagues (1997) used a fuel oil fly ash to examine the adverse effects
19 of a model urban particle in an animal model (monocrotaline-MCT) of cardiorespiratory disease;
20 MCT causes pulmonary vascular inflammation and hypertension. They observed 42% mortality
21 in MCT-treated rats exposed to approximately 580 //g/m3 fly ash for 6 hours/day for
22 3 consecutive days. These rats showed a massive neutrophillic inflammation from MCT
23 treatment which was further exacerbated by fly ash inhalation. Deaths did not occur in
24 MCT-treated rats exposed to filtered air or in saline-treated rats exposed to fly ash. The increase
25 in deaths in the MCT/fly ash group was accompanied by an increase in neutrophils in lavage
26 fluid and an increased immunostaining of MIP-2 in the heart and lungs of the MCT/fly ash
27 animals. Cardiac immunohistochemical analysis indicated increased MIP-2 in cardiac
28 macrophages. The fly ash-induced deaths did not result from a change in pulmonary arterial
29 pressure; the cause of death was not identified. In a similar experimental model, Watkinson et al.
30 (1998) examined the effect of intratracheally instilled ROFA on ECG measurements in control
31 and MCT-treated rats. As reported by Killingsworth et al. (1997), an increase in spontaneous
October 1999 7-88 DRAFT-DO NOT QUOTE OR CITE
-------
a
o
cr
CD
i-j
>O
>o
>o
lAJVLJli /-3. ^AKiflU
Species, Gender,
Strain Age, or
Body Weight Particle
Rat, Sprague- ROFA
Dawley, males,
monocrotaline-
treated (MCT)
'VAa^ui^AK jLrrin^ia AI
Mass
Exposure Concentration
Technique (Mg) or (Mg/m3)
Instillation 0, 250, 1000, or
2500 Mg in
0.3 ml saline
MU \JltltLK. ai
Particle
Characteristics
Size (Mm); a g
1.95 Mm MMAD
8g = 2.19
: a i JLIVII^ j^r j
Exposure
Duration
Monitored for
96 hours after
instillation of
ROFA particles
P-ti^ia \Jf rAK.ll^\JLiAltL 1V1A1 1JLK
Cardiovascular Effects Reference
Dose-related increases in the incidence and Watkinson et al.
duration of serious arrhythmic events in (1998)
normal rats. Incidence and severity of
arrythmias were greatly increased in the
monocrotaline-treated rats. Deaths were
seen at each instillation level in
monocrotaline-treated rats only (6/12 died
after MCT + ROFA).
•-J
I
oo
Rats FOFA Inhalation 580±110Mg/m3 MMAD 2.06 Mm 6 h/day for Death occurred only in MCT rats exposed
monocrotaline- og=1.57 3 days to ROFA. Neutrophils in lavage fluid was
treated significantly increased in MCT rats exposed
to ROFA versus filtered air. MIP-2 mRNA
expression in lavage cells was induced in
normal animals exposed to fly ash.
Rabbit, New colloidal Instillation 2 ml of 1% <1 Mm Examined for Colloidal carbon stimulated the release of
Zealand White, carbon colloidal carbon 24 to 192 h BRDU-labeled PMNs from the bone
female, 1.8 to (20 mg) after instillation marrow. The supernatant of alveolar
2.4 kg macrophages treated with colloidal carbon
in vitro also stimulated the release of PMNs
from bone marrow, likely via cytokines.
Killingsworth et al.
(1997)
Terashima et al.
(1997)
>
'rl
H
6
0
2J
0
H
/O
O
_j
j
w
o
o
I-H
H
W
Rats, M, SD, Emission Intratracheal Total mass: Emission PM:
60 days old source PM instillation 2.5 mg/rat 1.78-4. 17 Mm
MCT Ambient
(60/mg/kg), i.p airshed PM Total transition Ambient PM:
and healthy. ROFA metal: 46 Mg/rat 3.27-4.09 Mm
ROFA - Residual Oil Fly Ash
FOFA - Fuel Oil Fly Ash
MMAD - Mass Median Aerodynamic Diameter
MIP-2 - Macrophage Inflammatory Protein-2
BRDU - 5' Bromo 2' Deoxyuridine
MCT - Monocrotaline treated
Analysis at 24 ROFA alone induced some mild
& 96 h arrhythumas;
following MCT-ROFA showed enhanced neutrophilic
instillation. inflammation;
MCT-ROFA animals showed more
numerous arrhythmias including S-T
segment inversions and A-V block.
Costa and Dreher
(1997)
-------
1 deaths occurred only in MCT- treated rats treated with ROFA. Deaths were observed at both the
2 250 and 2,500 //g dose levels. Watkinson et al. (1998) also observed a dose-related increase in
3 the incidence and duration of serious arrhythmic events in control animals exposed to ROFA
4 particles. They examined the effects of ROFA (0, 0.25, 1.0, 2.5 mg ROFA in 0.3 ml saline) in
5 healthy rats and rats treated with MCT, twelve days after MCT treatment, ROFA instilled
6 animals were studied for 4 days. The incidence and duration of serious arrhythmic events were
7 related to the dose of ROFA. Healthy animals treated with ROFA suffered no deaths, but
8 MCT-treated rats had 1, 2, and 3 deaths in the low, medium, and high dose groups. This study
9 suggests that ROFA PM may be implicated in conductive and hypoxemic arrhythmias associated
10 with the cardiac-related deaths.
11 Kodavanti et al. (1999) exposed MCT-pretreated rats to ROFA by either intratracheal
12 instillation (0.88 or 3.33 mg/kg) or nose-only inhalation (15 mg/m3, 6 h/d for 3 consecutive
13 days). As reported in Watkinson et al. (1998), intratracheal instillation of ROFA in
14 MCT-pretreated rats resulted in 58% mortality whereas no mortality occurred in MCT-pretreated
15 rats exposed to ROFA by inhalation exposure. No mortality occurred in healthy rats exposed to
16 ROFA or in MCT-pretreated rats exposed to clean air. Despite the fact that mortality was not
17 associated with ROFA inhalation exposure of MCT-pretreated rats, exacerbation of lung lesions
18 and pulmonary inflammatory cytokine gene expression was clearly evident. No deaths have also
19 been observed by Gordon et al. (1998) in MCT-treated rats exposed for a single 3 hour exposure
20 or for 6 hours/day for 3 days to concentrated ambient PM (CAP).
21 Gordon and colleagues (1998) have reported systemic effects in animals exposed to inhaled
22 CAP. Increases in peripheral blood neutrophils were observed in control and MCT-treated rats at
23 3 h, but not 24 h, after exposure to 150 to 400 //g/m3 CAP. This effect, likely a result of vascular
24 demargination, did not appear to be dose-related and did not occur on all exposure days, thus,
25 suggesting that day-to-day changes in particle composition may play an important role in the
26 systemic effects of inhaled particles. However, Terashima et al. (1997) has instilled rabbits with
27 20 mg colloidal carbon, a relatively inert particle (<1 //m), and observed a stimulation of the
28 release of 5'-bromo-2'deoxyuridine (BrdU)-labeled PMNs from the bone marrow at 2 to 3 days
29 after instillation. Because the instilled supernatant from rabbit AM treated in vitro with colloidal
30 carbon also stimulated the release of PMNs from the bone marrow, they hypothesized that
31 cytokines released from activated macrophages could be responsible for this systemic effect.
October 1999 7-90 DRAFT-DO NOT QUOTE OR CITE
-------
1 In summary, controlled animal studies have provided initial evidence that inhaled or
2 instilled particles can have systemic, especially cardiovascular, effects. In the case of
3 MCT-treated rats, these effects can be lethal. Understanding the pathways by which very small
4 concentrations of inhaled ambient PM can produce systemic, life-threatening changes, however,
5 is far from clear. Among the hypotheses that have been proposed to account for the
6 non-pulmonary effects of PM are activation of neural reflexes, cytokine effects on heart tissue
7 (Killingsworth et al., 1997), alterations in coagulability (Seaton et al., 1995; Sjogren, 1997), and
8 perturbations in homeostatic processes such as heart rate or heart rate variability (Watkinson
9 et al., 1998). A great deal of research using controlled exposures of animal and human subjects
10 to PM will be necessary to test mechanistic hypotheses generated to date, as well as those which
11 are likely to be proposed in the future.
12
13
14 7.5 RESPONSES TO PM AND GASEOUS POLLUTANT MIXTURES
15 Ambient PM itself is a mixture of varying size and composition. The following discussion
16 examines effects of mixtures of ambient PM, or PM surrogates, with gaseous pollutants.
17 Ambient PM co-exists in indoor and outdoor air with a number of co-pollutant gases, including
18 ozone, sulfur dioxide, oxides of nitrogen, and carbon monoxide. Toxicological interactions
19 between PM and gaseous co-pollutants may be antagonistic, additive, or synergistic (Mauderly,
20 1993). The presence and nature of any interaction appears to depend upon the size and
21 concentration of pollutants in the mixture, exposure duration, and the endpoint being examined.
22 It is not possible to predict a priori from the presence of certain pollutants whether any
23 interaction will occur and, if there is interaction, whether it will be synergistic, additive, or
24 antagonistic (Table 7-6).
25 Mechanisms responsible for the various forms of interaction are speculative. In terms of
26 potential health effects, the greatest hazard from pollutant interaction is the possibility of synergy
27 between particles and gases, especially if effects occur at concentrations at which no effects
28 occur when individual constituents are inhaled. Various physical and chemical mechanisms may
29 underly synergism. For example, physical adsorption or absorption of some material on a
30 particle could result in transport to more sensitive sites, or sites where this material would not
31 normally be deposited in toxic amounts. This physical process may explain the interaction found
October 1999 7-91 DRAFT-DO NOT QUOTE OR CITE
-------
o
o
r+
O
TABLE 7-6. RESPIRATORY AND CARDIOVASCULAR EFFECTS OF MIXTURES
Tl
H
b
o
z
o
H
o
H
W
O
Species,
Gender, Strain
Age, or Body
Weight
Lambs
Pigeons
(Columba
livia)
Canine
Rats
Mice, Swiss,
female,
5 weeks old
Rats, Sprague-
Dawley, male,
250-300 g
Gases and
Particles
Ambient NOx,
SO2, CO, and
PM
Ambient gases
and particles
Ambient gases
and particles
O3 and
resuspended
urban PM
SO2 and
carbon
H2SO4 and O3
Exposure
Technique
Natural 24 hour
exposure in urban
and rural areas
Natural 24 h
exposure in urban
and rural areas
Natural 24 h
exposure in 4
urban areas of
Mexico City and
1 rural area
Inhalation, whole-
body
Inhalation, flow-
past nose-only
Inhalation, nose-
only
Mass
Concentration
(ppm) or (,ug/m3)
0.8 ppm O3 and
5,000 or
50,000 /ig/m3 PM
10,000 //g/m3
carbon with or
without 5 to
20 ppm SO2 at 10%
or 85% RH
500 Aig/m3 H2SO4
aerosol (2 different
particles sizes) with
or without 0.6 ppm
03
Particle
Characteristics Exposure
Size (pm); ag Duration
3 months
Continuous
ambient
exposure
Continuous
ambient
exposure
Single 4 h
exposure
0.3 /an MMAD Single 4 h
withGSDof2.7 exposure
Fine (0.3 /an 4 h/day for
MMD with GSD 2 days
of 1.7) and
ultrafine
(0.06 /an with
GSD of 1.4)
Respiratory Effects of Inhaled Particles on Markers
in Lavage Fluid
Irritation characterized by mucus hypersecretion
and morphological changes in the epithelium in the
nasopharyngeal mucosa in lambs exposed in urban
area
Increased number of AM and decreased number of
lamellar bodies in type II epithelial cells in urban
pigeons
No significant differences in AM or total cell
counts in lavage from dogs studied among the
5 regions. A significant increase in lavage fluid
neutrophils and lymphocytes in the southwest
region, where the highest O3 levels were recorded,
compared to the 2 industrial regions with the
highest PM levels.
PM alone caused no change in cell proliferation in
bronchioles or parenchyma. Co-exposure with O3
greatly potentiated the proliferative changes
induced by O3 alone. These changes were greatest
in the epithelium of the terminal bronchioles and
alveolar ducts.
Macrophage phagocytosis was depressed only in
animals exposed to the combination of SO2 and
carbon at 85% humidity. This inhibition in
macrophage function lasted at least 7 days after
exposure.
The volume percentage of injured alveolar septae
was increased only in the combined ultrafine
acid/O3 animals. BrdU labelling in the periacinar
region was increased in a synergistic manner in the
combined fine acid/O3 animals.
Reference
Gulisano et al.
(1997)
Lorz and Lopez
(1997)
Vanda et al.
(1998)
Vincent et al.
(1997)
Jakab et al.
(1996)
Kimmel et al.
(1997)
O
HH
H
W
-------
0
<•}
(^ Species,
2> Gender, Strain
!_» Age, or Body
^O Weight
Rats, Fischer
NNia, male,
22 to 24
months old
Rats
TABLE 7-6 (cont'd). RESPIRATORY AND CARDIOVASCULAR EFFECTS OF MIXTURES
Gases and Exposure
Particles Technique
Carbon, Inhalation
ammonium
bisulfate, and
03
O3 and Ottawa Inhalation
urban dust
Mass Concentration
(ug/m3)
50 /ig/m3 carbon +
70 /cg/m3 ammonium
bisulfate + 0.2 ppm
O3 or 100 /ig/m3
carbon +140 /ig/m3
ammonium bisulfate
+ 0.2 ppm O3
40,000 /ig/m3 and
0.8 ppm O3
Particle
Characteristics
Size (pm); ag
0.4 /an MMAD
with GSD of
2.0
MMAD = 4.5 /j.
Exposure
Duration
4 h/day,
3 days/week for
4 weeks
Single 4 h
exposure followed
by 20 h clean air
Respiratory Effects of Inhaled Particles on Markers
in Lavage Fluid
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.
CO exposure 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
Reference
Bolarin et al.
(1997)
Bouthillier et al.
(1998)
O
o
2
o
H
O
C
o
H
W
O
O
HH
H
W
Rats
Healthy and
asthmatic
children
H2SO4 and O3 Inhalation,
whole body
H2SO4 Inhalation
SO2, and O3
20 to 150/ig/m3
H2SO4and0.12or
0.2 ppm O3
60 to 140 /ig/m3
H2SO4, 0.1ppmSO2,
and 0.1 ppm O3
0.4 to 0.8 /on
0.6 /an H2SO4
Intermittent
(12 h/day) or
continuous
exposure for up
to 90 days
Single 4 h
exposure with
intermittent
No interactive effect of H2SO4 and O3 on Last and Pinkerton
biochemical and morphometric endpoints. (1997)
A positive association between acid concentration Linn et al.
and symptoms, but not spirometry, in asthmatic (1997)
children. No changes in healthy children.
Sprague-
Dawley rats
(300 g)
O3 and H2SO4 Inhalation
coated carbon nose-only
0.2 ppm O3 0.26 pm
+ 50 /ig/m3 C GSD = 2.2
+ 100 /ig/m3 H2SO4
0.4 ppm O3
+250 /ig/m3 C
+500 pg/m3 H,SO4
4 h/day for 1 day
or 5 days
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 .
Kleinman et al.
(1999)
MMAD - Mass Median Aerodynamic Diameter
GSD - Geometric Standard Deviation
BrdU - See Table 3-C
O3 - Ozone
SO2 - Sulfur dioxide
H2SO4 - Sulfuric acid
-------
1 in studies of mixtures of carbon black and formaldehyde or of carbon black and acrolein (Jakab,
2 1992, 1993).
3 Chemical interactions between particles and gases can occur on particle surfaces, thus,
4 forming secondary products which may be more lexicologically active than the primary materials
5 and which can then be carried to a sensitive site. The hypothesis of such chemical interactions
6 has been examined in the gas and particle exposure studies by Amdur and colleagues (Amdur
7 and Chen, 1989; Chen et al., 1992) and Jakab and colleagues (Jakab and Hemenway, 1993; Jakab
8 et al., 1996). These investigators have demonstrated that synergism occurs as secondary
9 chemical species are produced, especially under conditions of increased temperature and relative
10 humidity. Thus, these studies suggest that air quality standards for individual air pollutants may
11 not be fully protective of human health for exposure to mixed ambient pollutants.
12 Another potential mechanism of gas-particle interaction may involve a pollutant-induced
13 change in the local microenvironment of the lung, enhancing the effects of the co-pollutant.
14 For example, Last et al. (1984) indicated that the observed synergism between ozone and acid
15 sulfates in rats was due to a decrease in the local microenvironmental pH of the lung following
16 deposition of acid, enhancing the effects of ozone by producing a change in the reactivity or
17 residence time of reactants, such as radicals, involved in ozone-induced tissue injury.
18 As noted in U.S. Environmental Protection Agency (1996a), the toxicology database for
19 mixtures containing PM other than acid sulfates is still quite sparse. Vincent et al. (1997)
20 exposed rats to 0.8 ppm ozone in combination with 5 or 50 mg/m3 of resuspended urban particles
21 for 4 h. While PM alone caused no change in cell proliferation (3H-thymidine labeling),
22 co-exposure to either concentration of resuspended PM with ozone greatly potentiated the
23 proliferative effects of exposure to ozone alone. These interactive changes occurred in epithelial
24 cells of both the terminal bronchioles and the alveolar ducts. The findings of these experiments
25 using resuspended dusts, although at high concentrations, are consistent with studies
26 demonstrating interaction between sulfuric acid (H2SO4) aerosols and ozone. Kirnmel and
27 colleagues (1997) examined the effect of acute co-exposure to ozone and fine or ultrafine H2SO4
28 aerosols on morphometry in rat lungs. They determined morphometrically that alveolar septal
29 volume was increased in animals co-exposed to ozone and ultrafine, but not fine, H2SO4.
30 Interestingly, cell labeling, an index of proliferative cell changes, was increased only in animals
31 co-exposed to fine H2SO4 and ozone as compared to animals exposed to ozone alone. Thus,
October 1999 7-94 DRAFT-DO NOT QUOTE OR CITE
-------
1 particle size of acid aerosols can influence the locale of the interactive effects of co-exposure
2 with ozone. Importantly, Last and Pinkerton (1997) have extended their previous work and
3 found that subchronic exposure to acid aerosols (20 to 150 //g/m3 H2SO4) had no interactive
4 effect on the biochemical and morphometric changes produced by either intermittent or
5 continuous ozone exposure (0.12 to 0.2 ppm). Thus, the interactive effects of ozone and acid
6 aerosol co-exposure in the lung disappeared during the long-term exposure.
7 Kleinman et al. (1999) examined the effects of ozone plus fine H2SO4 coated carbon
8 particles (MMAD = 0.26 //m) for 1 or 5 days. They found the inflammatory response with the
9 ozone-particle mixture was greater after 5 days (4 h/day) than after day 1. This contrasted with
10 ozone exposure alone (0.4 ppm) which caused marked inflammation on acute exposure, but no
11 inflammation after 5 consecutive days of exposure.
12 Two studies have examined interaction between carbon particles and gaseous co-pollutants.
13 Jakab et al. (1996) challenged mice with a single 4 h exposure to a high concentration of carbon,
14 10 mg/m3, in the presence of SO2 at low and high relative humidities. Macrophage phagocytosis
15 was significantly depressed only in mice exposed to the combined pollutants under high relative
16 humidity conditions. This study demonstrates that fine carbon particles can serve as an effective
17 carrier for acidic sulphates where chemical conversion of adsorbed SO2 to acid sulfate species
18 occurred. Interestingly, the depression in macrophage function was present as late as 7 days
19 post-exposure. Bolarin et al. (1997) exposed rats to only 50 or 100 //g/m3 carbon particles in
20 combination with ammonium bisulfate and ozone. Despite 4 weeks of exposure, they observed
21 no changes in protein concentration in lavage fluid or blood prolyl 4-hydroxylase, an enzyme
22 involved in collagen metabolism. Slight decreases in plasma fibronectin were present in animals
23 exposed to the combined pollutants versus ozone alone. Thus as, previously noted, the potential
24 for adverse effects in the lungs of animals challenged with a combined exposure to particles and
25 gaseous pollutants is dependent on numerous factors including the gaseous co-pollutant,
26 concentration, and time.
27 Linn and colleagues (1997) examined the effect of a single exposure to 60 to 140 //g/m3
28 H2SO4, 0.1 ppm SO2, and 0.1 ppm ozone in healthy and asthmatic children. The children
29 performed intermittent exercise during the 4 h exposure to increase the inhaled dose of the
30 pollutants. An overall effect on the combined group of healthy and asthmatic children was not
31 observed. A positive association between acid concentration and symptoms was seen, however,
October 1999 7-95 DRAFT-DO NOT QUOTE OR CITE
-------
1 in the subgroup of asthmatic children. The combined pollutant exposure had no effect on
2 spirometry in asthmatic children and no changes in symptoms or spirometry were observed in
3 healthy children. Thus, the effect of combined exposure to PM and gaseous co-pollutants
4 appeared to have less effect on asthmatic children exposed under controlled laboratory conditions
5 in comparison with field studies of children attending summer camp (Thurston et al., 1997).
6 However, prior exposure to H2SO4 aerosol may enhance the subsequent response to ozone
7 exposure (Linn et al., 1994; Frampton et al., 1995); the timing and sequence of the exposures
8 may be important.
9 Three unique animal field studies have examined the adverse respiratory effects of complex
10 mixtures in urban and rural environments. These studies have taken advantage of the differences
11 in pollutant makeup of urban and rural environments and studied animals under natural,
12 continuous exposure conditions. Gulisano et al. (1997) examined the morphologic changes
13 produced by continuous ambient exposure to air pollutants in lambs raised for 3 months in rural
14 (n=2) or urban (n=10) environments. Compared to the lungs of the rural lambs, irritation, as
15 characterized by mucus hypersecretion and morphological changes in the epithelial cells lining
16 the nasopharyngeal region, was present in the lambs exposed to urban air pollution. Lorz and
17 Lopez (1997) performed a similar study using pigeons as the test animal. They observed an
18 increase in the number of AM and a decrease in the number of lamellar bodies in Type II
19 epithelial cells in the lungs of urban pigeons. Extrapolation of these studies is hampered by an
20 incomplete characterization of the exposure atmospheres. A more thorough examination of the
21 ambient level of pollutants was performed in the study by Vanda et al. (1998) who studied the
22 effect of pollutant exposure in dogs raised in four urban regions of Mexico City and one nearby
23 rural area. They found no significant differences in AM number or total cell counts in lavage
24 fluid from the dogs among the 5 regions. A significant increase in lavage fluid neutrophils and
25 lymphocytes was found in dogs from the urban region with the highest ozone levels in
26 comparison to the regions with the highest PM levels. Thus, the effect of ozone on cellular
27 parameters in lavage fluid appeared to be greater than that for PM. In summary, each of these
28 3 animal field studies provides evidence that urban air pollutants can produce greater lung
29 changes than would occur from exposure to rural pollution. However, extrapolation of these
30 results is severely hampered by the uncontrolled exposure conditions, small sample size,
October 1999 7-96 DRAFT-DO NOT QUOTE OR CITE
-------
1 behavior patterns, and nutritional factors. Thus, in these field studies, it is difficult to assign a
2 role to PM in the observed adverse pulmonary effects.
3
4
5 7.6 MECHANISMS OF PM TOXICITY AND PATHOPHYSIOLOGY
6 7.6.1 Introduction
7 The characteristics of particles which may be responsible for lung injury after exposure to
8 ambient air pollution are not known. However injury has been postulated to be mediated by
9 ultrafine particles, biological agents (e.g., endotoxin), acid aerosols, organic fraction of PM and
10 oxidant generation catalyzed by transition metals associated with particles. Furthermore the
11 mechanisms which underlie injury are also unclear. This section discusses potential mechanisms
12 in relation to PM characteristics based upon available data.
13
14 7.6.2 Soluble Metals and Reactive Oxygen Intermediates
15 PM contains transition metals such as iron (most abundant), copper, nickel, vanadium, and
16 cobalt. These metals are capable of catalyzing the one-electron reductions of molecular oxygen
17 necessary to generate reactive oxygen species (ROS). These reactions can be demonstrated by
18 the iron-catalyzed Haber-Weiss reactions below:
19
20 Reductanf + Fe(III) - Reductantn+1 + Fe(II) (1)
21
22 Fe(II) + O2 ^ Fe(III) + O2'~ (2)
23
24 HO2' + O2- + H+ - O2 + H2O2 (3)
25
26 Fe(II) + H2O2 ^ Fe(III) +OH" + HO' (Fenton Reaction) (4)
27
28 Iron will continue to participate in the redox cycle in the above reactions as long as there is
29 sufficient O2 or H2O2 and reductants.
October 1999 7-97 DRAFT-DO NOT QUOTE OR CITE
-------
1 Soluble metals from inhaled PM dissolved into the milieu of the airway lumen can directly
2 react with biological molecules (acting as a reductant in the above reactions) to produce ROS.
3 For example, ascorbic acid in the human lung epithelial lining fluid can react with Fe(III) from
4 inhaled PM to cause single strand breaks in supercoiled plasmid DNA, (j)X174 RFI (Smith and
5 Aust, 1997). The DNA damage caused by a PM10 suspension can be inhibited by mannitol, an
6 hydroxyl radical scavenger, further confirming the involvement of free radicals in these reactions
7 (Gilmour et al., 1996; Donaldson et al., 1997; Li et al., 1997). Since the clear supernatant of the
8 centrifuged PM10 suspension contained all of the suspension activity, the free radical activity is
9 derived either from a fraction that is not centrifugable (on a bench centrifuge) or the radical
10 generating system is released into solution (Gilmour et al., 1996; Donaldson et al., 1997; Li et al.,
11 1997).
12 In addition to measuring the interactions of ROS and biomolecules directly, the role of
13 ROS in PM-induced lung injury can also be assessed by measuring the electron spin resonance
14 (ESR) spectrum of radical adducts or fluorescent intensity of dichlorofiuorescin (DCFH), an
15 intracellular dye that fluoresces upon oxidation by ROS. Alternatively, ROS can be inhibited
16 using free radical scavengers such as dimethylthiourea (DMTU); or antioxidants such as
17 glutathione or N-acetylcysteine (NAC); or antioxidant enzymes, such as superoxide dismutase
18 (SOD). The diminished response to PM after treatment with these antioxidants indicates the
19 involvement of ROS.
20 As described earlier, Kadiiska et al. (1997) used the ESR spectra of POBN adducts to
21 measure ROS in rats instilled with ROFA and demonstrated the association between ROS
22 production in lung lipid extracts and soluble metals in ROFA. Using DMTU to inhibit ROS
23 production, Dye et al. (1997) had shown that systemic administration of DMTU impeded
24 development of the cellular inflammatory response to ROFA, but did not ameliorate biochemical
25 alterations in BAL fluid. Goldsmith et al. (1998), as described earlier, showed that ROFA and
26 CAPs caused increases in ROS production in AMs. The water-soluble component of both CAPs
27 and ROFA significantly increased AM oxidant production over negative control values.
28 In addition, increased PM-induced cytokine production was inhibited by NAC. Li et al. (1996,
29 1997) instilled rats with PM10 particles (collected on filters from an Edinburgh, Scotland
30 monitoring station). Six hours after intratracheal instillation of PM10, they observed a decrease in
31 glutathione (GSH) levels in the BAL fluid. Although this study does not describe the
October 1999 7-98 DRAFT-DO NOT QUOTE OR CITE
-------
1 composition of the PM10, the authors suggest that changes in GSH, an important lung
2 antioxidant, support the contention that the free radical activity of PM10 is responsible for its
3 biological activity in vivo.
4 In addition to ROS generated directly by PM, resident or newly recruited AMs or PMNs are
5 also capable of producing these reactive species upon stimulation. The ROS produced during the
6 oxidative burst can be measured using a chemiluminescence (CL) assay. With this assay,
7 AM CL signals in vitro had been shown to be greatest with ROFA containing primarily soluble
8 V and were less with ROFA containing Ni plus V (Kodavanti et al., 1998a). As described
9 earlier, exposures to Dusseldorf and Duisburg PM increased the resting ROS production in
10 PMNs which could be inhibited by SOD, catalase and sodium azide (Hitzfeld et al., 1997).
11 Stringer and Kobzik (1998) showed that addition of NAC (1.0 mM) decreased ROFA-mediated
12 IL-8 production by approximately 50% in normal and TNF-cc-primed A549 cells. In addition,
13 exposures of A549 cells to ROFA caused a substantial (and NAC inhibitable) increase in oxidant
14 levels as measured by DCFH oxidation. In human AM, Becker et al. (1996) found a CL response
15 for ROFA but not urban air particles (Ottawa, Dusseldorf) or volcanic ash.
16 Metals are the most probable species capable of catalyzing ROS generation upon exposure
17 to PM. Soluble metals can be mobilized into the epithelial cells or AMs to produce ROS
18 intracellularly. Using ROFA and colloidal iron oxide, Ohio et al. (1997b, 1998a,b,c) have shown
19 that exposures to these particles disrupted iron homeostasis and induced the production of ROS
20 in vivo and in vitro. Treatment of animals or cells with metal-chelating agents such as DBF with
21 an associated decrease in response has been used to infer the involvement of metal in
22 PM-induced lung injury. Metal chelation by DBF (1 mM) caused significant inhibition of
23 particulate-induced AM oxidant production, as measured using DCFH (Goldsmith et al., 1998).
24 DBF treatment also reduced NF-KB activation and cytokine secretion in BEAS-2B cells exposed
25 to Provo, Utah PM (Kennedy et al., 1998). However, treatment of ROFA suspension with DBF
26 was not effective in blocking leachable metal induced acute lung injury (Dreher et al., 1997).
27 Dreher et al. (1997) indicated that DBF could chelate Fe(III) and V(II), but not Ni(II), suggesting that
28 nickel played a role in the observed lung injury.
29 Other than Fe, several vanadium compounds have been shown to increase mRNA levels for
30 selected cytokines in BAL cells and also to induce pulmonary inflammation (Pierce et al., 1996).
31 NaVO3 and VOSO4, highly soluble forms of vanadium, tended to induce pulmonary
October 1999 7-99 DRAFT-DO NOT QUOTE OR CITE
-------
1 inflammation and inflammatory cytokine mRNA expression more rapidly and more intensely
2 than the less soluble form, V2O5, in rats. Neutrophil influx was greatest following exposure to
3 VOSO4 and lowest following exposure to V2O5.
4
5 7.6.3 Intracellular Signaling Mechanisms
6 In has been shown that the intracellular redox state of the cell modulates the activity of
7 several transcription factors, including NF-KB, a critical step in the induction of a variety of
8 proinflammatory cytokine and adhesion-molecule genes. NF-KB is a heterodimeric protein
9 complex that in most cells resides in an inactive state in the cell cytoplasm by binding to
10 inhibitory kappa B alpha (iKBcc). Upon appropriate stimulation by cytokines or ROS, iKBcc is
11 phosphorylated and subsequently degraded by proteolysis. The dissociation of 1KB a from NF-KB
12 allows the latter to translocate into the nucleus and bind to appropriate sites in the DNA to
13 initiate transcription of various genes. Two studies in vitro have shown the involvement of
14 NF-KB in particulate-induced cytokine and intercellular adhesion molecule-1 (ICAM-1)
15 production in human airway epithelial cells (BEAS-2B) (Quay et al., 1998; Kennedy et al.,
16 1998). Cytokine secretion was preceded by activation of NF-KB and was reduced by treatment
17 with antioxidants or metal chelators. These results suggest that metal-induced oxidative stress
18 may play a significant role in the initiation phase of the inflammatory cascade following
19 particulate exposure.
20 A second well-characterized human transcription factor, AP-1, also responds to the
21 intracellular ROS concentration. AP-1 exists in two forms, either in a homodimer of c-jun
22 protein or a heterodimer consisting of c-jun and c-fos. Small amounts of AP-1 already exist in the
23 cytoplasm in an inactive form, mainly as phosphorylated c-jun homodimer. Many different
24 oxidative stress-inducing stimuli, such as UV light and IL-1, can activate AP-1. Exposure of rat
25 lung epithelial cells to ambient PM in vitro resulted in increases in c-jun kinase activity, levels of
26 phosphroylated c-jun immunoreactive protein, and transcriptional activation of AP-1-dependent
27 gene expression (Timblin et al., 1998). This study demonstrated that interaction of ambient
28 particles with lung epithelial cells initiates a cell signaling cascade related to aberrant cell
29 proliferation.
30 Early response gene transactivation has been linked to the development of apoptosis, a
31 unique type of programmed cell injury and a potential mechanism to account for PM induced
October 1999 7-100 DRAFT-DO NOT QUOTE OR CITE
-------
1 changes in cellular response. Apoptosis of human AMs exposed to ROFA (25 //g/ml) or urban
2 PM was observed by Holian et al. (1998). In addition, both ROFA and urban PM upregulated the
3 expression of the RFD1+ AM phenotype, while only ROFA decreased the RFDl+7+ phenotype.
4 It has been suggested that an increase in the AM phenotype ratio of RFDl+/RFDl+7+ may be
5 related to disease progression in patients with inflammatory diseases. These data showed that
6 ROFA and urban PM can induce apoptosis of human AMs and increase the ratio of AM
7 phenotypes toward a higher immune active state and may contribute to or exacerbate lung
8 inflammation.
9 Another intracellular signaling pathway that can lead to diverse cellular responses such as
10 cell growth, differentiation, proliferation, apoptosis, and stress responses to environmental
11 stimuli, is the phosphorylation-dependent mitogen-activated protein kinase (MAPK).
12 Noncytotoxic levels of ROFA have been shown to induce significant dose- and time-dependent
13 increases in protein tyrosine phosphate levels in BEAS cells (Samet et al., 1997). In a
14 subsequent study, the effects of As, Cr, Cu, Fe, Ni, V, and Zn on the MAPK, extracellular
15 receptor kinase (ERK), c-jun N-terminal kinase (JNK), and P38 in BEAS cells were investigated
16 (Samet et al., 1998). Noncytotoxic concentrations of As, V, and Zn induced a rapid
17 phosphorylation of MAPK in BEAS cells. Activity assays confirmed marked activation of ERK,
18 JNK, and P38 in BEAS cells exposed to As, V, and Zn. Cr and Cu exposure resulted in a
19 relatively small activation of MAPK, whereas Fe and Ni did not activate MAPK. Similarly, the
20 transcription factors c-Jun and ATF-2, substrates of JNK and P38, respectively, were markedly
21 phosphorylated in BEAS cells treated with As, Cr, Cu, V, and Zn. The same acute exposure to
22 As, V, or Zn that activated MAPK was sufficient to induce a subsequent increase in IL-8 protein
23 expression in BEAS cells. These data suggest that MAPK may mediate metal-induced
24 expression of inflammatory proteins in human bronchial epithelial cells.
25 To investigate the interaction between respiratory cells and PM, Kobzik (1995) showed that
26 scavenger receptors are responsible for AM binding of charged PM and that different
27 mechanisms mediate binding of carbonaceous dusts such as DPM. In addition, surfactant
28 components can increase AM phagocytosis of environmental particulates in vitro, but only
29 slightly relative to the already avid AM uptake of unopsonized particles (Stringer and Kobzik,
30 1996). Respiratory tract epithelial cells are also capable of binding with PM to secrete cytokine
31 IL-8. Using a respiratory epithelial cell line (A549), Stringer et al. (1996) found that binding of
October 1999 7-101 DRAFT-DO NOT QUOTE OR CITE
-------
1 particles to epithelial cells was calcium-dependent for TiO2 and Fe2O3, while cc-quartz binding
2 was not calcium-dependent. In addition, as observed in AM, PM binding by A549 cells was also
3 mediated by scavenger receptor(s), albeit those distinct from the heparin-insensitive acetylated-
4 LDL receptor. Furthermore, cc-quartz, but not TiO2 or CAPs, caused a dose-dependent
5 production of IL-8 (range 1-6 ng/mL), demonstrating a particle-specific spectrum of epithelial
6 cell cytokine (IL-8) response.
7
8 7.6.4 The Role of Particle Size and Surface Area
9 Most particles used in laboratory animal toxicology and occupational studies are greater
10 than 0.1 //m in size. However, the enormous numbers and huge surface area of the ultrafine
11 particles demonstrate the importance of considering the size of the particle in assessing response.
12 Ultrafine particles with a diameter of 20 nm when inhaled at the same mass concentration have a
13 number concentration that is approximately 6 orders of magnitude higher than for a 2.5 //m
14 diameter particle; particle surface area is also greatly increased (Table 7-7).
15
16
TABLE 7-7. NUMBERS AND SURFACE AREAS OF MONODISPERSE
PARTICLES OF UNIT DENSITY OF DIFFERENT SIZES AT A
MASS CONCENTRATION OF 10 //g/m3
Particle Diameter
(//m)
0.02
0.1
0.5
1.0
2.5
Particle Number
(per cm3 Air)
2,400,000
19,100
153
19
1.2
Particle Surface Area
(//m2 per cm3 Air)
3,016
600
120
60
24
Source: Oberdorster et al. (1995).
1 Many studies summarized in U.S. Environmental Protection Agency (1996a), as well as in
2 this document, suggest that the surface of particles, or substances that are released from the
October 1999 7-102 DRAFT-DO NOT QUOTE OR CITE
-------
1 surface (e.g., transition metals), interact with the biological system and that surface-associated
2 free radicals or free radical-generating systems may be responsible for toxicity. Thus, if ultrafine
3 particles were to cause toxicity by a transition metal-mediated mechanism, for example, then the
4 relatively large surface area for a given mass of ultrafine particles would mean high
5 concentrations of transition metals being available to cause oxidative stress to cells.
6 Several groups have examined the toxic differences between fine and ultrafine particles,
7 with the general finding that the ultrafine particles show a significantly greater response at
8 similar mass doses (Oberdorster et al, 1992; Yoo et al, 1995; Li et al, 1996, 1997). However,
9 only a few studies have investigated the ability of ultrafine particles to generate a greater
10 oxidative stress when compared to fine particles of the same material. Studies by Gilmour et al.
11 (1996) have shown that at equal mass, ultrafine TiO2 caused more plasmid DNA strand breaks
12 than fine TiO2. This effect could be inhibited with mannitol. Osier and Oberdorster (1997)
13 compared the response of rats (F344) exposed by intratracheal inhalation to "fine"
14 (approximately 250 nm) and "ultrafine" (approximately 21 nm) TiO2 particles with rats exposed
15 to similar doses by intratracheal instillation. Animals receiving particles through inhalation
16 showed a smaller pulmonary response, measured by BAL parameters, in both severity and
17 persistence, when compared with those animals receiving particles through instillation. These
18 results demonstrate a difference in pulmonary response to an inhaled vs an instilled dose, which
19 may be due to differences in dose rate, particle distribution, or altered clearance between the two
20 methods. Consistent with these in vivo studies, Finkelstein et al. (1997) has shown that exposing
21 primary cultures of rat Type II cells to 10 //g/ml ultrafine TiO2 (20 nm) causes increased TNF
22 and IL-1 release throughout the entire 48 h incubation period. In contrast, fine TiO2 (200 nm)
23 had no effect.
24 Oberdorster et al. (1999) recently completed a series of studies in rats and mice using
25 ultrafine particles of various chemical compositions (PTFE, TiO2, C, Fe, Fe2O3, Pt, V, V2O5).
26 In old rats sensitized with endotoxin, and exposed to ozone plus ultrafine carbon particles, they
27 found a nine-fold greater release of reactive oxygen species in the old than in similarly treated
28 young rats. Exposure to ultrafine PM alone in sensitized old rats also caused an inflammatory
29 response. These investigators also demonstrated translocation of UF Pt to the liver. Recovery of
30 ultrafine Pt after exposure was less than for fine particles, suggesting less uptake of ultrafine PM
31 by AM.
October 1999 7-103 DRAFT-DO NOT QUOTE OR CITE
-------
1 Only one study examined the influence of specific surface area on biological activity (Lison
2 et al., 1997). The biological responses to various MnO2 dusts with different specific surface area
3 (0.16, 0.5, 17 and 62 m2/g) were compared in vitro and in vivo. In both systems, the results show
4 that the amplitude of the response is dependent on the total surface area which is in contact with
5 the biological system, indicating that surface chemistry phenomena are involved in the biological
6 reactivity. Freshly ground particles with a specific surface area of 5 m2/g were also examined in
7 vitro. These particles exhibited an enhanced cytotoxic activity, which was almost equivalent to
8 that of particles with a specific surface area of 62 m2/g, indicating that undefined reactive sites
9 produced at the particle surface by mechanical cleavage may also contribute to the toxicity of
10 insoluble particles.
11
12 7.6.5 Summary
13 The mechanisms which underlie the biological responses to ambient PM are not clear.
14 Various studies using particulate matter having diverse physicochemical characteristics have
15 shown that these characteristics have a great impact upon the specific response which is
16 observed. Thus, there may, in fact, be multiple biological mechanisms that may be responsible
17 for observed morbidity/mortality due to exposure to ambient PM, and these mechanisms may be
18 highly dependent upon the type of particle in the exposure atmosphere. However, it should be
19 noted that many controlled exposure studies used concentrations of particulate matter which were
20 much higher than those occurring in ambient air. Thus, some of the mechanisms elicited may not
21 occur with exposure to lower levels. Clearly, controlled exposure studies have not as yet been
22 able to unequivocally determine the particle characteristics and the toxicological mechanisms by
23 which ambient PM may affect biological systems.
24
October 1999 7-104 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2
3 Adamson, I. Y. R.; Bowden, D. H. (1981) Dose response of the pulmonary macrophagic system to various
4 particulates and its relationship to transepithelial passage of free particles. Exp. Lung Res. 2: 165-175.
5 Adamson, I. Y. R.; Hedgecock, C. (1995) Patterns of particle deposition and retention after instillation to mouse
6 lung during acute injury and fibrotic repair. Exp. Lung Res. 21: 695-709.
7 Adamson, I. Y. R.; Prieditis, H. (1998) Silica deposition in the lung during epithelial injury potentiates fibrosis and
8 increases particle translocation to lymph nodes. Exp. Lung Res. 24: 293-306.
9 Amdur, M. O.; Chen, L. C. (1989) Furnace-generated acid aerosols: speciation and pulmonary effects.
10 In: Symposium on the health effects of acid aerosols; October 1987; Research Triangle Park, NC.
11 Environ. Health Perspect. 79: 147-150.
12 Amdur, M. O.; Dubriel, M.; Creasia, D. A. (1978) Respiratory response of guinea pigs to low levels of sulfuric acid.
13 Environ. Res. 15: 418-423.
14 American Conference of Governmental Industrial Hygienists (ACGIH). (1985) Particle size-selective sampling in
15 the workplace: report of the ACGIH technical committee on air sampling procedures. Cincinnati, OH:
16 American Conference of Governmental Industrial Hygienists.
17 Bayram, H.; Devalia, J. L.; Khair, O. A.; Abdelaziz, M. M.; Sapsford, R. J.; Sagai, M.; Davies, R. J. (1998a)
18 Comparison of ciliary activity and inflammatory mediator release from bronchial epithelial cells of nonatopic
19 nonasthmatic subjects and atopic asthmatic patients and the effect of diesel exhaust particles in vitro.
20 J. Allergy Clin. Immunol. 102: 771-782.
21 Bayram, H.; Devalia, J. L.; Sapsford, R. J.; Ohtoshi, T.; Miyabara, Y.; Sagai, M.; Davies, R. J. (1998b) The effect
22 of diesel exhaust particles on cell function and release of inflammatory mediators from human bronchial
23 epithelial cells in vitro. Am. J. Respir. Cell Mol. Biol. 18: 441-448.
24 Becker, S.; Soukup, J. M. (1998) Decreased CD1 Ib expression, phagocytosis, and oxidative burst in urban
25 particulate pollution-exposed human monocytes and alveolar macrophages. J. Toxicol. Environ. Health
26 55:455-477.
27 Becker, S.; Soukup, J. M.; Gilmour, M. I.; Devlin, R. B. (1996) Stimulation of human and rat alveolar macrophages
28 by urban air particulates: effects on oxidant radical generation and cytokine production. Toxicol. Appl.
29 Pharmacol. 141:637-648.
30 Bennett, W. D.; Ilowite, J. (1989) Dual pathway clearance of Tc99m-DTPA from the bronchial mucosa. Am. Rev.
31 Respir. Dis. 139: 1132-1138.
32 Bennett, W. D.; Zeman, K. L. (1998) Deposition of fine particles in children spontaneously breathing at rest.
33 Inhalation Toxicol. 10: 831-842.
34 Bennett, W. D.; Zeman, K. L.; Kim, C. (1996) Variability of fine particle deposition in healthy adults: effect of age
35 and gender. Am. J. Respir. Crit. Care Med. 153: 1641-1647.
36 Bennett, W. D.; Zeman, K. L.; Kang, C. W.; Schechter, M. S. (1997a) Extrathoracic deposition of inhaled, coarse
37 particles (4.5"mu"m) in children versus adults. Ann. Occup. Hyg. 41 (Suppl.): 497-502.
38 Bennett, W. D.; Zeman, K. L.; Kim, C.; Mascarella, J. (1997b) Enhanced deposition of fine particles in COPD
39 patients spontaneously breathing at rest. Inhalation Toxicol. 9: 1-14.
40 Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1998) Bronchial
41 airway deposition and retention of particles in inhaled boluses: effect of anatomic dead space. J. Appl.
42 Physiol. 85: 685-694.
43 Bennett, W. D.; Scheuch, G.; Zeman, K. L.; Brown, J. S.; Kim, C.; Heyder, J.; Stahlhofen, W. (1999) Regional
44 deposition and retention of particles in shallow, inhaled boluses: effect if lung volume. J. Appl. Physiol.
45 86: 168-173.
46 Blanc, P.; Wong, H.; Bernstein, M. S.; Boushey, H. A. (1991) An experimental human model of metal fume fever.
47 Ann. Intern. Med. 114: 930-936.
48 Blanc, P. D.; Boushey, H. A.; Wong, H.; Wintermeyer, S. F.; Bernstein, M. S. (1993) Cytokines in metal fume
49 fever. Am. Rev. Respir. Dis. 147: 134-138.
50 Blomberg, A.; Sainsbury, C.; Rudell, B.; Frew, A. J.; Holgate, S. T.; Sandstrom, T.; Kelly, F. J. (1998) Nasal cavity
51 lining fluid ascorbic acid concentration increases in healthy human volunteers following short term exposure
52 to diesel exhaust. Free Radical Res. 28: 59-67.
October 1999 7-105 DRAFT-DO NOT QUOTE OR CITE
-------
1 Bolarin, D. M.; Bhalla, D. K.; Kleinman, M. T. (1997) Effects of repeated exposures of geriatric rats to ozone and
2 particle-containing atmospheres: an analysis of bronchoalveolar lavage and plasma proteins. Inhalation
3 Toxicol. 9: 423-434.
4 Bonner, J. C.; Rice, A. B.; Lindroos, P. M.; O'Brien, P. O.; Dreher, K. L.; Rosas, I.; Alfaro-Moreno, E.;
5 Osornio-Vargas, A. R. (1998) Induction of the lung myofibroblast PDGF receptor system by urban ambient
6 particles from Mexico City. Am. J. Respir. Cell Mol. Biol. 19: 672-680.
7 Bouthillier, L.; Vincent, R.; Goegan, P.; Adamson, I. Y. R.; Bjarnason, S.; Stewart, M.; Guenette, J.; Potvin, M.;
8 Kumarathasan, P. (1998) Acute effects of inhaled urban particles and ozone: lung morphology, macrophage
9 activity, and plasma endothelin-1. Am. J. Pathol. 153: 1873-1884.
10 Brain, J. D.; Knudson, D. E.; Sorokin, S. P.; Davis, M. A. (1976) Pulmonary distribution of particles given by
11 intratracheal instillation or by aerosol inhalation. Environ. Res. 11: 13-33.
12 Brain, J. D.; Long, N. C.; Wolfthal, S. F.; Dumyahn, T.; Dockery, D. W. (1998) Pulmonary toxicity in hamsters of
13 smoke particles from Kuwaiti oil fires. Environ. Health Perspect. 106: 141-146.
14 Broeckaert, F.; Buchet, J. P.; Huaux, F.; Lardot, C.; Lison, D.; Yager, J. W. (1997) Reduction of the ex vivo
15 production of tumor necrosis factor alpha by alveolar phagocytes after administration of coal fly ash and
16 copper smelter dust. J. Toxicol. Environ. Health 51: 189-202.
17 Camner, P.; Anderson, M.; Philipson, K.; Bailey, A.; Hashish, A.; Jarvis, N.; Bailey, M.; Svartengren, M. (1997)
18 Human bronchiolar deposition and retention of 6-, 8-, and 10-(a particles. Exp. Lung Res. 23: 517-535.
19 Carter, J. D.; Ghio, A. J.; Samet, J. M.; Devlin, R. B. (1997) Cytokine production by human airway epithelial cells
20 after exposure to an air pollution particle is metal-dependent. Toxicol. Appl. Pharmacol. 146: 180-188.
21 Cave, A. C.; Apstein, C. S. (1996) Polymyxin B, aprotein Kinase C inhibitor, abolishes preconditioning-induced
22 protection against contractile dysfunction in the isolated blood perfused rat heart. J. Mol. Cell. Cardiol.
23 28:977-987.
24 Chen, L. C.; Fine, J. M.; Qu, Q.-S.; Amdur, M. O.; Gordon, T. (1992) Effects of fine and ultrafine sulfuric acid
25 aerosols in guinea pigs: alterations in alveolar macrophage function and intracellular pH. Toxicol. Appl.
26 Pharmacol. 113: 109-117.
27 Cheng, Y.-S.; Smith, S. M.; Yeh, H.-C.; Kim, D.-B.; Cheng, K.-H.; Swift, D. L. (1995) Deposition of ultrafine
28 aerosols and thoron progeny in replicas of nasal airways of young children. Aerosol. Sci. Technol.
29 23:541-552.
30 Cheng, K-H.; Cheng, Y.-S.; Yeh, H.-C.; Guilmette, R. A.; Simpson, S. Q.; Yang, Y.-H.; Swift, D. L. (1996) In vivo
31 measurements of nasal airway dimensions and ultrafine aerosol deposition in the human nasal and oral
32 airways. J. Aerosol Sci. 27: 785-801.
33 Cheng, K-H.; Cheng, Y.-S.; Yeh, H.-C.; Swift, D. L. (1997) An experimental method for measuring aerosol
34 deposition efficiency in the human oral airway. Am. Ind. Hyg. Assoc. J. 58: 207-213.
35 Chui, Y.-C.; Poon, G.; Law, F. (1988) Toxicokinetics and bioavailability of paraquat in rats following different
36 routes of administration. Toxicol. Ind. Health 4: 203-219.
37 Churg, A.; Brauer, M. (1997) Human lung parenchyma retains PM2.5. Am. J. Respir. Crit. Care Med.
38 155:2109-2111.
39 Churg, A.; Vedal, S. (1996) Carinal and tubular airway particle concentrations in the large airways of non-smokers
40 in the general population: evidence for high particle concentration at airway carinas. Occup. Environ. Med.
41 53:553-558.
42 Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.; Sioutas, C.; Paulauskis, J.; Coull, B.;
43 Ferguson, S.; Godleski, J. J. (1999) Urban air particulate inhalation alters pulmonary function and induces
44 pulmonary inflammation in a rodent model of chronic bronchitis. Inhalation Toxicol. 11: 637-656.
45 Cohen, B. S.; Harley, N. H.; Schlesinger, R. B.; Lippmann, M. (1988) Nonuniform particle deposition on
46 tracheobronchial airways: implications for lung dosimetry. In: Dodgson, J.; McCallum, R. L; Bailey, M. R.;
47 Fisher, D. R., eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung
48 dosimetry; September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32 (suppl. 1): 1045-1053.
49 Cohen, B. S.; Sussman, R. G.; Lippmann, M. (1990) Ultrafine particle deposition in a human tracheobronchial cast.
50 Aerosol Sci. Technol. 12: 1082-1091.
51 Cohen, M. D.; Yang, Z.; Zelikoff, J. T.; Schlesinger, R. B. (1996) Pulmonary immunotoxicity of inhaled ammonium
52 metavanadate in Fisher 344 rats. Fundam. Appl. Toxicol. 33: 254-263.
53 Cohen, M. D.; Becker, S.; Devlin, R.; Schlesinger, R. B.; Zelikoff, J. T. (1997) Effects of vanadium upon
54 polyl:C-induced responses in rat lung and alveolar macrophages. J. Toxicol. Environ. Health 51: 591-608.
October 1999 7-106 DRAFT-DO NOT QUOTE OR CITE
-------
1 Cohen, B. S.; Xiong, J. Q.; Fang, C.-P.; Li, W. (1998) Deposition of charged particles on lung airways. Health
2 Phys. 74: 554-560.
3 Cormier, Y.; Laviolette, M.; Bedard, G.; Dosman, J.; Israel-Assayag, E. (1998) Effect of route of breathing on
4 response to exposure in a swine confinement building. Am. J. Respir. Crit. Care Med. 157: 1512-1521.
5 Costa, D. L.; Dreher, K. L. (1997) Bioavailable transition metals in particulate matter mediate cardiopulmonary
6 injury in healthy and compromised animal models. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of
7 the sixth international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
8 September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1053-1060.
9 Crystal, R. G., et al., eds. (1997) The lung: scientific foundations. 2nd ed. Philadelphia, PA: Lippincott-Raven.
10 Cuddihy, R. G. (1984) Mathematical models for predicting clearance of inhaled radioactive materials. In: Smith, H.;
11 Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a meeting jointly
12 organized by the Commission of the European Communities and the National Radiological Protection Board;
13 March; Oxford, United Kingdom. Luxembourg: Commission of the European Communities; pp. 167-179;
14 report no. EUR 9384 EN.
15 Cuddihy, R. G.; Yeh, H. C. (1988) Respiratory tract clearance of particles and substances dissociated from particles.
16 In: Mohr, U.; Dungworth, D.; Kimmerle, G.; Lewkowski, J.; McClellan, R.; Stober, W., eds. Inhalation
17 toxicology: the design and interpretation of inhalation studies and their use in risk assessment. New York,
18 NY: Springer-Verlag; pp. 169-193.
19 Culp, D. J.; Latchney, L. R.; Frampton, M. W.; Jahnke, M. R.; Morrow, P. E.; Utell, M. J. (1995) Composition of
20 human airway mucins and effects after inhalation of acid aerosol. Am. J. Physiol. 269: L358-L370.
21 Dahl, A. R.; Felicetti, S. A.; Muggenburg, B. A. (1983) Clearance of sulfuric acid-introduced 35S from the
22 respiratory tracts of rats, guinea pigs and dogs following inhalation or instillation. Fundam. Appl. Toxicol.
23 3:293-297.
24 Diaz-Sanchez, D.; Tsien, A.; Casillas, A.; Dotson, A. R.; Saxon, A. (1996) Enhanced nasal cytokine production in
25 human beings after in vivo challenge with diesel exhaust particles. J. Allergy Clin. Immunol. 98: 114-123.
26 Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997) Combined diesel exhaust particulate and ragweed
27 allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
28 production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
29 Donaldson, K.; Brown, D. M.; Mitchell, C.; Dineva, M.; Beswick, P. H.; Gilmour, P.; MacNee, W. (1997) Free
30 radical activity of PM10: iron-mediated generation of hydroxyl radicals. In: Driscoll, K. E.; Oberdorster, G.,
31 eds. Proceedings of the sixth international meeting on the toxicology of natural and man-made fibrous and
32 non-fibrous particles; September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl.
33 105(5): 1285-1289.
34 Dong, W.; Lewtas, J.; Luster, M. I. (1996) Role of endotoxin in tumor necrosis factor a expression from alveolar
35 macrophages treated with urban air particles. Exp. Lung Res. 22: 577-592.
36 Dormans, J. A. M. A.; Steerenberg, P. A.; Arts, J. H. E.; Van Bree, L.; De Klerk, A.; Verlaan, A. P. J.; Bruijntjes,
37 J. P.; Beekhof, P.; Van Soolingen, D.; Van Loveren, H. (1999) Pathological and immunological effects of
38 respirable coal fly ash in male wistar rats. Inhalation Toxicol. 11: 51-69.
39 Dorries, A. M.; Valberg, P. A. (1992) Heterogeneity of phagocytosis for inhaled versus instilled material. Am. Rev.
40 Respir. Dis. 146: 831-837.
41 Dreher, K. L.; Jaskot, R. H.; Lehmann, J. R.; Richards, J. H.; McGee, J. K.; Ghio, A. J.; Costa, D. L. (1997)
42 Soluable transition metals mediate residual oil fly ash induced acute lung injury. J. Toxicol. Environ. Health
43 50:285-305.
44 Drew, R. T.; Kuschner, M.; Bernstein, D. M. (1987) The chronic effects of exposure of rats to sized glass fibres.
45 In: Walton, W. H.; Coppock, S. M., eds. Man-made mineral fibres in the working environment: [proceedings
46 of an international symposium]; October 1986; Copenhagen, Denmark. Ann. Occup. Hyg. 31: 711-729.
47 Driscoll, K. E.; Lindenschmidt, R. C.; Maurer, J. K.; Higgins, J. M.; Ridder, G. (1990) Pulmonary response to silica
48 or titanium dioxide: inflammatory cells, alveolar macrophage-derived cytokines, and histopathology. Am. J.
49 Respir. Cell Mol. Biol. 2: 381-390.
50 Driscoll, K. E.; Lindenschmidt, R. C.; Maurer, J. K.; Perkins, L.; Perkins, M.; Higgins, J. (1991) Pulmonary
51 response to inhaled silica or titanium dioxide. Toxicol. Appl. Pharmacol. 111: 201-210.
52 Dye, J. A.; Adler, K. B.; Richards, J. H.; Dreher, K. L. (1997) Epithelial injury induced by exposure to residual oil
53 fly-ash particles: role of reactive oxygen species? Am. J. Respir. Cell Mol. Biol. 17: 625-633.
54 Fabiani, R.; Pampanella, L.; Minelli, A.; Mezzasoma, L; Vecchiarelli, A.; Morozzi, G. (1997) Effect of airborne
55 particulate extracts onmonocyte oxidative metabolism. J. Environ. Pathol. Toxicol. Oncol. 16: 195-199.
October 1999 7-107 DRAFT-DO NOT QUOTE OR CITE
-------
1 Falk, R.; Philipson, K.; Svartengren, M.; Jarvis, N.; Bailey, M.; Camner, P. (1997) Clearance of particles from small
2 ciliated airways. Exp. Lung Res. 23: 495-515.
3 Ferin, J. (1977) Effect of particle content of lung on clearance pathways. In: Sanders, C. L.; Schneider, R. P.; Dagle,
4 G. E.; Ragan, H. A., eds. Pulmonary macrophages and epithelial cells: proceedings of the sixteenth annual
5 Hanford biology symposium; September 1976; Richland, WA. Oak Ridge, TN: Energy Research and
6 Development Administration; pp. 414-423. Available from: NTIS, Springfield, VA; CONF-760927. (ERDA
7 symposium series 43).
8 Ferin, J.; Feldstein, M. L. (1978) Pulmonary clearance and hilar lymph node content in rats after particle exposure.
9 Environ. Res. 16: 342-352.
10 Findeisen, W. (1935) Uber das Absetzen kleiner, in der Luft suspendierter Teilchen in der menschlichen Lunge bei
11 der Atmung [The deposition of small airborne particles in the human lung during respiration]. Pfluegers
12 Arch. Gesamte Physiol. Menschen Tiere 236: 367-379.
13 Fine, J. M.; Gordon, T.; Chen, L. C.; Kinney, P.; Falcone, G.; Beckett, W. S. (1997) Metal fume fever:
14 characterization of clinical and plasma IL-6 responses in controlled human exposures to zinc oxide fume at
15 and below the threshold limit value. J. Occup. Environ. Med. 39: 722-726.
16 Finkelstein, J. N.; Johnston, C.; Barrett, T.; Oberdorster, G. (1997) Particulate-cell interactions and pulmonary
17 cytokine expression. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
18 on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
19 NY. Environ. Health Perspect. Suppl. 105(5): 1179-1182.
20 Folinsbee, L. J.; Kim, C. S.; Kehrl, H. R.; Prah, J. D.; Devlin, R. B. (1997) Methods in human inhalation toxicology.
21 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press LLC; pp. 607-670.
22 Fortoul, T. I.; Osorio, L. S.; Tovar, A. T.; Salazar, D.; Castilla, M. E.; Olaiz-Fernandez, G. (1996) Metals in lung
23 tissue from autopsy cases in Mexico City residents: comparison of cases from the 1950s and the 1980s.
24 Environ. Health Perspect. 104: 630-632.
25 Foster, W. M.; Langenback, E.; Bergofsky, E. H. (1980) Measurement of tracheal and bronchial mucus velocities in
26 man: relation to lung clearance. J. Appl. Physiol.: Respir. Environ. Exercise Physiol. 48: 965-971.
27 Frampton, M. W.; Voter, K. Z.; Morrow, P. E.; Roberts, N. J., Jr.; Culp, D. J.; Cox, C.; Utell, M. J. (1992) Sulfuric
28 acid aerosol exposure in humans assessed by bronchoalveolar lavage. Am. Rev. Respir. Dis. 146: 626-632.
29 Frampton, M. W.; Morrow, P. E.; Cox, C.; Levy, P. C.; Condemi, J. J.; Speers, D.; Gibb, F. R.; Utell, M. J. (1995)
30 Sulfuric acid aerosol followed by ozone exposure in healthy and asthmatic subjects. Environ. Res. 69: 1-14.
31 Fry, F. A.; Black, A. (1973) Regional deposition and clearance of particles in the human nose. J. Aerosol Sci.
32 4: 113-124.
33 Fujimaki, H.; Katayama, N.; Wakamori, K. (1992) Enhanced histamine release from lung mast cells of guinea pigs
34 exposed to sulfuric acid aerosols. Environ. Res. 58: 117-123.
35 Gavett, S. H.; Madison, S. L.; Dreher, K. L.; Winsett, D. W.; McGee, J. K.; Costa, D. L. (1997) Metal and sulfate
36 composition of residual oil fly ash determines airway hyperreactivity and lung injury in rats. Environ. Res.
37 72: 162-172.
38 Gavett, S. H.; Madison, S. L.; Stevens, M. A.; Costa, D. L. (1999) Residual oil fly ash amplifies allergic cytokines,
39 airway responsiveness, and inflammation in mice. Am. J. Respir. Crit. Care Med.: submitted.
40 Gearhart, J. M.; Schlesinger, R. B. (1986) Sulfuric acid-induced airway hyperresponsiveness. Fundam. Appl.
41 Toxicol. 7: 681-689.
42 Ghio, A. J.; Meng, Z. H.; Hatch, G. E.; Costa, D. L. (1997a) Luminol-enhanced chemiluminescence after in vitro
43 exposures of rat alveolar macrophages to oil fly ash is metal dependent. Inhalation Toxicol. 9: 255-271.
44 Ghio, A. J.; Piantadosi, C. A.; Crumbliss, A. L. (1997b) Hypothesis: iron chelation plays a vital role in neutrophilic
45 inflammation. Biometals 10: 135-142.
46 Ghio, A. J.; Carter, J. D.; Richards, J. H.; Brighton, L. E.; Lay, J. C.; Devlin, R. B. (1998a) Disruption of normal
47 iron homeostasis after bronchial instillation of an iron-containing particle. Am. J. Physiol. 274: L396-L403.
48 Ghio, A. J.; Richards, J. H.; Dittrich, K. L.; Samet, J. M. (1998b) Metal storage and transport proteins increase after
49 exposure of the rat lung to an air pollution particle. Toxicol. Pathol. 26: 388-394.
50 Ghio, A. J.; Carter, J. D.; Samet, J. M.; Reed, W.; Quay, J.; Dailey, L. A.; Richards, J. H.; Devlin, R. B. (1998c)
51 Metal-dependent expression of ferritin and lactoferrin by respiratory epithelial cells. Am. J. Physiol.
52 274: L728-L736.
53 Ghio, A. J.; Carter, J. D.; Dailey, L. A.; Devlin, R. B.; Samet, J. M. (1999a) Respiratory epithelial cells demonstrate
54 lactoferrin receptors that increase after metal exposure. Am. J. Physiol. 276 (6 Pt 1): L933-L940.
October 1999 7-108 DRAFT-DO NOT QUOTE OR CITE
-------
1 Ohio, A. J.; Stonehuerner, J.; Dailey, L. A.; Carter, J. D. (1999b) Metals associated with both the water-soluble and
2 insoluble fractions of an ambient air pollution particle catalyze an oxidative stress. Inhalation Toxicol.
3 11:37-49.
4 Gift, J. S.; Faust, R. A. (1997) Noncancer inhalation toxicology of crystalline silica: exposure-response assessment.
5 J. Expo. Anal. Environ. Epidemiol. 7: 345-358.
6 Gilmour, M. L; Taylor, F. G. R.; Baskerville, A.; Wathes, C. M. (1989a) The effect of titanium dioxide inhalation
7 on the pulmonary clearance of Pasteurella haemolytica in the mouse. Environ. Res. 50: 157-172.
8 Gilmour, M. L; Taylor, F. G. R.; Wathes, C. M. (1989b) Pulmonary clearance of Pasteurella haemolytica and
9 immune responses in mice following exposure to titanium dioxide. Environ. Res. 50: 184-194.
10 Gilmour, P. S.; Brown, D. M.; Lindsay, T. G.; Beswick, P. H.; MacNee, W.; Donaldson, K. (1996) Adverse health
11 effects of PM10 particles: involvement of iron in generation of hydroxyl radical. Occup. Environ. Med.
12 53:817-822.
13 Goldsmith, C.-A.; Frevert, C.; Imrich, A.; Sioutas, C.; Kobzik, L. (1997) Alveolar macrophage interaction with air
14 pollution particulates. In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth international meeting
15 on the toxicology of natural and man-made fibrous and non-fibrous particles; September 1996; Lake Placid,
16 NY. Environ. Health Perspect. Suppl. 105(5): 1191 -1195.
17 Goldsmith, C.-A. W.; Imrich, A.; Danaee, H.; Ning, Y. Y.; Kobzik, L. (1998) Analysis of air pollution
18 particulate-mediated oxidant stress in alveolar macrophages. J. Toxicol. Environ. Health Part A 54: 529-545.
19 Goldsmith, C.-A. W.; Hamada, K.; Ning, Y.-Y.; Qin, G.; Catalano, P.; Murthy, G. G. K.; Lawrence, J.; Kobzik, L.
20 (1999) The effects of environmental aerosols on airway hyperresponsiveness in a murine model of asthma.
21 Inhalation Toxicol.: in press.
22 Gordon, T.; Nadziejko, C.; Schlesinger, R.; Chen, L. C. (1998) Pulmonary and cardiovascular effects of acute
23 exposure to concentrated ambient particulate matter in rats. Toxicol. Lett. 96-97: 285-288.
24 Grabowski, G. M.; Paulauskis, J. D.; Godleski, J. J. (1999) Mediating phosphorylation events in the
25 vanadium-induced respiratory burst of alveolar macrophages. Toxicol. Appl. Pharmacol. 156: 170-178.
26 Groth, M. L.; Macri, K.; Foster, W. M. (1997) Cough and mucociliary transport of airway particulate in chronic
27 obstructive lung disease. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an
28 international symposium on inhaled particles organised by the British Occupational Hygiene Society; August
29 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 515-521.
30 Guilmette, R. A.; Cheng, Y. S.; Griffith, W. C. (1997) Characterising the variability in adult human nasal airway
31 dimensions. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII: proceedings of an international
32 symposium on inhaled particles organised by the British Occupational Hygiene Society; August 1996;
33 Cambridge, UK. Ann. Occup. Hyg. 41(suppl. 1): 491-496.
34 Gulisano, M.; Marceddu, S.; Barbara, A.; Pacini, A.; Buiatti, E.; Martini, A.; Pacini, P. (1997) Damage to the
35 nasopharyngeal mucosa induced by current levels of urban air pollution: a field study in lambs. Eur. Respir.
36 J. 10: 567-572.
37 Hahon, N.; Booth, J. A.; Green, F.; Lewis, T. R. (1985) Influenza virus infection in mice after exposure to coal dust
38 and diesel engine emissions. Environ. Res. 37: 44-60.
39 Hamada, K; Goldsmith, C.-A.; Kobzik, L. (1999) Increased airway hyperresponsiveness and inflammation in a
40 juvenile mouse model of asthma exposed to air pollutant aerosol. J. Toxicol. Environ. Health: in press.
41 Harmsen, A. G.; Muggenburg, B. A.; Snipes, M. B.; Bice, D. E. (1985) The role of macrophages in particle
42 translocation from lungs to lymph nodes. Science (Washington, DC) 230: 1277-1280.
43 Hatch, G. E.; Boykin, E.; Graham, J. A.; Lewtas, J.; Pott, F.; Loud, K.; Mumford, J. L. (1985) Inhalable particles
44 and pulmonary host defense: in vivo and in vitro effects of ambient air and combustion particles. Environ.
45 Res. 36: 67-80.
46 Health Effects Institute. (1995) Diesel exhaust: a critical analysis of emissions, exposure, and health effects:
47 a special report of the Institute's Diesel Working Group. Cambridge, MA: Health Effects Institute.
48 Hegemann, L.; van Rooijen, L. A. A.; Traber, J.; Schmidt, B. H. (1991) Polymyxin B is a selective and potent
49 antagonist of calmodulin. Eur. J. Pharmacol. 207: 17-22.
50 Heistracher, T.; Hofmann, W. (1997) Flow and deposition patterns in successive airway bifurcations. In: Cherry, N.;
51 Ogden, T., eds. Inhaled Particles VIII: proceedings of an international symposium on inhaled particles
52 organised by the British Occupational Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg.
53 41(suppl.): 537-542.
October 1999 7-109 DRAFT-DO NOT QUOTE OR CITE
-------
1 Henderson, R. F.; Driscoll, K. E.; Harkema, J. R.; Lindenschmidt, R. C.; Chang, I.-Y.; Maples, K. R.; Barr, E. B.
2 (1995) A comparison of the inflammatory response of the lung to inhaled versus instilled particles in F344
3 rats. Fundam. Appl. Toxicol. 24: 183-197.
4 Henshaw, D. L.; Fews, A. P. (1984) The microdistribution of alpha emitting particles in the human lung. In: Smith,
5 H.; Gerber, G., eds. Lung modelling for inhalation of radioactive materials: proceedings of a meeting jointly
6 organized by the Commission of the European Communities and the National Radiological Protection Board;
7 March; Oxford, United Kingdom. Luxembourg: Commission of the European Communities; pp. 199-218;
8 report no. EUR 9384 EN.
9 Heyder, J.; Rudolf, G. (1977) Deposition of aerosol particles in the human nose. In: Walton, W. H.; McGovern, B.,
10 eds. Inhaled particles IV: proceedings of an international symposium, part 1; September 1975; Edinburgh,
11 United Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 107-126.
12 Hiller, F. C. (1991) Health implications of hygroscopic particle growth in the human respiratory tract. J. Aerosol
13 Med. 4:1-23.
14 Hitzfeld, B.; Friedrichs, K. H.; Ring, J.; Behrendt, H. (1997) Airborne particulate matter modulates the production
15 of reactive oxygen species in human polymorphonuclear granulocytes. Toxicology 120: 185-195.
16 Hofmann, W.; Martonen, T. B.; Graham, R. C. (1989a) Predicted deposition of nonhygroscopic aerosols in the
17 human lung as a function of subject age. J. Aerosol Med. 2: 49-68.
18 Hofmann, W.; Koblinger, L.; Martonen, T. B. (1989b) Structural differences between human and rat lungs:
19 implications for Monte Carlo modeling of aerosol deposition. In: Mahaffey, J. A., ed. 26th Hanford life
20 sciences symposium, modeling for scaling to man: biology, dosimetry, and response. Health Phys.
21 57(suppl. 1): 41-47.
22 Hofmann, W.; Balashazy, I.; Heistracher, T.; Koblinger, L. (1996) The significance of particle deposition patterns in
23 bronchial airway bifurcations for extrapolation modeling. Aerosol Sci. Technol. 25: 305-327.
24 Holian, A.; Hamilton, R. F., Jr.; Morandi, M. T.; Brown, S. D.; Li, L. (1998) Urban particle-induced apoptosis and
25 phenotype shifts in human alveolar macrophages. Environ. Health Perspect. 106: 127-132.
26 Hsieh, T. H.; Yu, C. P. (1998) Two-phase pulmonary clearance of insoluble particles in mammaliam species.
27 Inhalation Toxicol. 10: 121-130.
28 International Commission on Radiological Protection. (1960) Report of Committee II on permissable dose for
29 internal radiation (1959). Health Phys. 3.
30 International Commission on Radiological Protection. (1979) Limits for intakes of radionuclides by workers.
31 Oxford, United Kingdom: Pergamon Press; ICRP publication 30, part 1.
32 International Commission on Radiological Protection. (1994) Human respiratory tract model for radiological
33 protection: a report of a task group of the International Commission on Radiological Protection. Oxford,
34 United Kingdom: Elsevier Science Ltd. (ICRP publication 66; Annals of the ICRP: v. 24, nos. 1-3).
35 Irsigler, G. B.; Visser, P. J.; Spangenberg, P. A. (1999) Asthma and chemical bronchitis in vanadium plant workers.
36 Am. J. Ind. Med. 35: 366-374.
37 Jacobs, J.; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
38 1983-1992. Environ. Health Perspect. 105: 980-985.
39 Jakab, G. J. (1992) Relationship between carbon black particulate-bound formaldehyde, pulmonary antibacterial
40 defenses, and alveolar macrophage phagocytosis. Inhalation Toxicol. 4: 325-342.
41 Jakab, G. J. (1993) The toxicologic interactions resulting from inhalation of carbon black and acrolein on
42 pulmonary antibacterial and antiviral defenses. Toxicol. Appl. Pharmacol. 121: 167-175.
43 Jakab, G. J.; Hemenway, D. R. (1993) Inhalation coexposure to carbon black and acrolein suppresses alveolar
44 macrophage phagocytosis and TNF- a release and modulates peritoneal macrophage phagocytosis. Inhalation
45 Toxicol. 5: 275-289.
46 Jakab, G. J.; Clarke, R. W.; Hemenway, D. R.; Longphre, M. V.; Kleeberger, S. R.; Frank, R. (1996) Inhalation of
47 acid coated carbon black particles impairs alveolar macrophage phagocytosis. Toxicol. Lett. 88: 243-248.
48 John, J.; Wollmer, P.; Dahlback, M.; Luts, A.; Jonson, B. (1994) Tidal volume and alveolar clearance of insoluble
49 particles. J. Appl. Physiol. 76: 584-588.
50 Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Baggs, R.; Oberdorster, G. (1996) Characterization of the early
51 pulmonary inflammatory response associated with PTFE fume exposure. Toxicol. Appl. Pharmacol.
52 140: 154-163.
53 Johnston, C. J.; Finkelstein, J. N.; Gelein, R.; Oberdorster, G. (1998) Pulmonary inflammatory responses and
54 cytokine and antioxidant mRNA levels in the lungs of young and old C57BL/6 mice after exposure to Teflon
55 fumes. Inhalation Toxicol. 10: 931-953.
October 1999 7-110 DRAFT-DO NOT QUOTE OR CITE
-------
1 Kadiiska, M. B.; Mason, R. P.; Dreher, K. L.; Costa, D. L.; Ohio, A. J. (1997) In vivo evidence of free radical
2 formation in the rat lung after exposure to an emission source air pollution particle. Chem. Res. Toxicol.
3 10:1104-1108.
4 Katz, I. M.; Davis, B. M.; Martonen, T. B. (1999) A numerical study of particle motion within the human larynx
5 and trachea. J. Aerosol Sci. 30: 173-183.
6 Kennedy, T.; Ohio, A. J.; Reed, W.; Samet, J.; Zagorski, J.; Quay, J.; Carter, J.; Dailey, L.; Hoidal, J. R.; Devlin,
7 R. B. (1998) Copper-dependent inflammation and nuclear factor- KB activation by particulate air pollution.
8 Am. J. Respir. Cell Mol. Biol. 19: 366-378.
9 Kesavanathan, J.; Swift, D. L. (1998) Human nasal passage particle deposition: the effect of particle size, flow rate,
10 and anatomical factors. Aerosol Sci. Technol. 28: 457-463.
11 Killingsworth, C. R.; Alessandrini, F.; Krishna Murthy, G. G.; Catalano, P. J.; Paulauskis, J. D.; Godleski, J. J.
12 (1997) Inflammation, chemokine expression, and death in monocrotaline-treated rats following fuel oil fly
13 ash inhalation. Inhalation Toxicol. 9: 541-565.
14 Kim, C. S.; Hu, S. C. (1998) Regional deposition of inhaled particles in human lungs: comparison between men and
15 women. J. Appl. Physiol. 84: 1834-1844.
16 Kim, C. S.; Kang, T. C. (1997) Comparative measurement of lung deposition of inhaled fine particles in normal
17 subjects and patients with obstructive airway disease. Am. J. Respir. Crit. Care Med. 155: 899-905.
18 Kimmel, T. A.; Chen, L. C.; Bosland, M. C.; Nadziejko, C. (1997) Influence of acid aerosol droplet size on
19 structural changes in the rat lung caused by acute exposure to sulfuric acid and ozone. Toxicol. Appl.
20 Pharmacol. 144: 348-355.
21 Kitabatake, M.; Imai, M.; Kasama, K; Kobayashi, L; Tomita, Y.; Yoshida, K. (1979) Effects of air pollutants on the
22 experimental induction of asthma attacks in guinea pigs: sulfuric acid mist and mixture of the mist and sulfur
23 dioxide. Mie Med. J. 29: 29-36.
24 Kleinman, M. T.; Mautz, W. J.; Bjarnason, S. (1999) Adaptive and non-adaptive responses in rats exposed to ozone,
25 alone and in mixtures, with acidic aerosols. Inhal. Toxicol. 11: 249-264.
26 Knox, R. B.; Suphioglu, C.; Taylor, P.; Desai, R.; Watson, H. C.; Peng, J. L.; Bursill, L. A. (1997) Major grass
27 pollen allergen Lol p 1 binds to diesel exhaust particles: implications for asthma and air pollution. Clin. Exp.
28 Allergy 27: 246-251.
29 Koblinger, L.; Hofmann, W. (1985) Analysis of human lung morphometric data for stochastic aerosol deposition
30 calculations. Phys. Med. Biol. 30: 541-556.
31 Koblinger, L.; Hofmann, W. (1988) Stochastic morphological model of the rat lung. Anat. Rec. 221: 533-539.
32 Kobzik, L. (1995) Lung macrophage uptake of unopsonized environmental particulates: role of scavenger-type
33 receptors. J. Immunol. 155: 367-376.
34 Kodavanti, U.; Costa, D. L. (1999) Animal models to study for pollutant effects. In: Holgate, S. T.; Koren, H. S.;
35 Samet, J. M.; Maynard, R. L., eds. Air pollution and health. London, United Kingdom: Academic Press;
36 pp. 165-197.
37 Kodavanti, U. P.; Jaskot, R. H.; Bonner, J.; Badgett, A.; Dreher, K. L. (1996) Eosinophilic lung inflammation in
38 particulate-induced lung injury: technical consideration in isolating RNA for gene expression studies.
39 Exp. Lung Res. 22: 541-554.
40 Kodavanti, U. P.; Jaskot, R. H.; Su, W. Y.; Costa, D. L.; Ghio, A. J.; Dreher, K. L. (1997a) Genetic variability in
41 combustion particle-induced chronic lung injury. Am. J. Physiol. 272: L521-L532.
42 Kodavanti, U. P.; Jaskot, R. H.; Costa, D. L.; Dreher, K. L. (1997b) Pulmonary proinflammatory gene induction
43 following acute exposure to residual oil fly ash: roles of particle-associated metals. Inhalation Toxicol.
44 9:679-701.
45 Kodavanti, U. P.; Hauser, R.; Christiani, D. C.; Meng, Z. H.; McGee, J.; Ledbetter, A.; Richards, J.; Costa, D. L.
46 (1998a) Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals.
47 Toxicol. Sci. 43:204-212.
48 Kodavanti, U. P.; Costa, D. L.; Bromberg, P. A. (1998b) Rodent models of cardiopulmonary disease: their potential
49 applicability in studies of air pollutant susceptibility. Environ. Health Perspect. 106(suppl. 1): 111-130.
50 Kodavanti, U. P.; Jackson, M. C.; Ledbetter, A. D.; Richards, J. R.; Gardner, S. Y.; Watkinson, W. P.; Campen,
51 M. J.; Costa, D. L. (1999) Lung injury from intratracheal and inhalation exposures to residual oil fly ash in a
52 rat model of monocrotaline-induced pulmonary hypertension. J. Toxicol. Environ. Health A 57: 101-121.
53 Kohlhaufl, M.; Brand, P.; Scheuch, G.; Meyer, T. S.; Schulz, H.; Haussinger, K.; Heyder, J. (1999) Increased fine
54 particle deposition in women with asymptomatic nonspecific airway hyperresponsiveness. Am. J. Respir.
55 Crit. Care Med. 159: 902-906.
October 1999 7-111 DRAFT-DO NOT QUOTE OR CITE
-------
1 Kreyling, W. G. (1992) Intracellular particle dissolution in alveolar macrophages. Environ. Health Perspect.
2 97: 121-126.
3 Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
4 experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
5 Perspect. 105: 1234-1237.
6 LaBelle, C. W.; Brieger, H. (1961) Patterns and mechanisms in the elimination of dust from the lung. In: Davies,
7 C. N., ed. Inhaled particles and vapours: proceedings of an international symposium; March-April 1960;
8 Oxford, United Kingdom. New York, NY: Pergamon Press; pp. 356-368.
9 Lambert, A. L.; Dong, W.; Winsett, D. W.; Belgrade, M. K.; Gilmour, M. I. (1999) Residual oil fly ash exposure
10 enhances allergic sensitization to house dust mite. Toxicol. Appl. Pharmacol. 158: 269-277'.
11 Landahl, H. D. (1950) On the removal of air-borne droplets by the human respiratory tract: I. the lung. Bull. Math.
12 Biophys. 12: 43-56.
13 Last, J. A.; Pinkerton, K. E. (1997) Chronic exposure of rats to ozone and sulfuric acid aerosol: biochemical and
14 structural responses. Toxicology 116: 133-146.
15 Last, J. A.; Hyde, D. M.; Chang, D. P.Y. (1984) A mechanism of synergistic lung damage by ozone and a
16 respirable aerosol. Exp. Lung Res. 7: 223-235.
17 Lay, J. C.; Berry, C. R.; Chong, S. K.; Bennett, W. D. (1995) Retention of insoluble particles after local
18 intrabronchial deposition in dogs. J. Appl. Physiol. 79: 1921-1929.
19 Lay, J. C.; Bennett, W. D.; Kim, C. S.; Devlin, R. B.; Bromberg, P. A. (1998) Retention and intracellular
20 distribution of instilled iron oxide particles in human alveolar macrophages. Am. J. Respir. Cell and Mol.
21 Biol. 18:687-695.
22 Leduc, D.; Fally, S.; De Vuyst, P.; Wollast, R.; Yernault, J.-C. (1995) Acute exposure to realistic acid fog: effects
23 on respiratory function and airway responsiveness in asthmatics. Environ. Res. 71: 89-98.
24 Lehnert, B. E.; Morrow, P. E. (1985) Association of 59iron oxide with alveolar macrophages during alveolar
25 clearance. Exp. Lung Res. 9: 1-16.
26 Lehnert, B. E.; Valdez, Y. E.; Bomalaski, S. H. (1988) Analyses of particles in the lung free cell, tracheobronchial
27 lymph nodal, and pleural space compartments following their deposition in the lung as related to lung
28 clearance mechanisms. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fisher, D. R., eds. Inhaled particles
29 VI: proceedings of an international symposium and workshop on lung dosimetry; September 1985;
30 Cambridge, United Kingdom. Ann. Occup. Hyg. 32 (suppl. 1): 125-140.
31 Leikauf, G.; Yeates, D. B.; Wales, K. A.; Spektor, D.; Albert, R. E.; Lippmann, M. (1981) Effects of sulfuric acid
32 aerosol on respiratory mechanics and mucociliary particle clearance in healthy nonsmoking adults. Am. Ind.
33 Hyg. Assoc. J. 42: 273-282.
34 Leikauf, G. D.; Spektor, D. M.; Albert, R. E.; Lippmann, M. (1984) Dose-dependent effects of submicrometer
35 sulfuric acid aerosol on particle clearance from ciliated human lung airways. Am. Ind. Hyg. Assoc. J.
36 45:285-292.
37 Lennon, S.; Zhang, Z.; Lessmann, R.; Webster, S. (1998) Experiments on particle deposition in the human upper
38 respiratory system. Aerosol Sci. Technol. 28: 464-474.
39 Leong, B. J. K.; Coombs, J. K.; Imlay, M. M.; Sabaitis, C. P. (1986) Morphometric analysis of the pulmonary
40 distribution of particles given by nose-only aerosol inhalation (AI) or endotracheal nebulization (EN) or
41 instillation (El). Toxicologist 6: 35.
42 Leong, B. K. J.; Coombs, J. K.; Sabaitis, C. P.; Rop, D. A.; Aaron, C. S. (1998) Quantitative morphometric analysis
43 of pulmonary deposition of aerosol particles inhaled via intratracheal nebulization, intratracheal instillation or
44 nose-only inhalation in rats. J. Appl. Toxicol. 18: 149-160.
45 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1996) Free radical activity and pro-inflammatory effects of
46 particulate air pollution (PM10) in vivo and in vitro. Thorax 51: 1216-1222.
47 Li, X. Y.; Gilmour, P. S.; Donaldson, K.; MacNee, W. (1997) In vivo and in vitro proinflammatory effects of
48 particulate air pollution (PM10). In: Driscoll, K. E.; Oberdorster, G., eds. Proceedings of the sixth
49 international meeting on the toxicology of natural and man-made fibrous and non-fibrous particles;
50 September 1996; Lake Placid, NY. Environ. Health Perspect. Suppl. 105(5): 1279-1283.
51 Linn, W. S.; Shamoo, D. A.; Anderson, K. R.; Peng, R.-C.; Avol, E. L.; Hackney, J. D. (1994) Effects of prolonged,
52 repeated exposure to ozone, sulfuric acid, and their combination in healthy and asthmatic volunteers. Am. J.
53 Respir. Crit. Care Med. 150: 431-440.
54 Linn, W. S.; Gong, H., Jr.; Shamoo, D. A.; Anderson, K. R.; Avol, E. L. (1997) Chamber exposures of children to
55 mixed ozone, sulfur dioxide, and sulfuric acid. Arch. Environ. Health 52: 179-187.
October 1999 7-112 DRAFT-DO NOT QUOTE OR CITE
-------
1 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
2 California. Environ. Health Perspect. 105: 216-222.
3 Lison, D.; Lardot, C.; Huaux, F.; Zanetti, G.; Fubini, B. (1997) Influence of particle surface area on the toxicity of
4 insoluble manganese dioxide dusts. Arch. Toxicol. 71: 725-729.
5 Lorz, C.; Lopez, J. (1997) Incidence of air pollution in the pulmonary surfactant system of the pigeon ( Columbia
6 livid). Anat. Rec. 249: 206-212.
7 Lundborg, M.; Lind, B.; Camner, P. (1984) Ability of rabbit alveolar macrophages to dissolve metals. Exp. Lung
8 Res. 7:11-22.
9 Lundborg, M.; Eklund, A.; Lind, B.; Camner, P. (1985) Dissolution of metals by human and rabbit alveolar
10 macrophages. Br. J. Ind. Med. 42: 642-645.
11 Madden, M. C.; Thomas, M. J.; Ohio, A. J. (1999) Acetaldehyde (CH3CHO) production in rodent lung after
12 exposure to metal-rich particles. Free Radical Biol. Med. 26: 1569-1577.
13 Madl, A. K.; Wilson, D. W.; Segall, H. J.; Pinkerton, K. E. (1998) Alteration in lung particle translocation,
14 macrophage function, and micro filament arrangement in monocrotaline-treated rats. Toxicol. Appl.
15 Pharmacol. 153:28-38.
16 Matsui, H.; Randell, S. H.; Peretti, S. W.; Davis, C. W.; Boucher, R. C. (1998) Coordinated clearance of periciliary
17 liquid and mucus from airway surfaces. J. Clin. Invest. 102: 1125-1131.
18 Mauderly, J. L. (1993) Toxicological approaches to complex mixtures. Environ. Health Perspect.
19 101(suppl. 4): 155-165.
20 Medinsky, M. A.; Kampcik, S. J. (1985) Pulmonary retention of [ 14C]benzo/a/pyrene in rats as influenced by the
21 amount instilled. Toxicology 35: 327-336.
22 Mercer, T. T. (1967) On the role of particle size in the dissolution of lung burdens. Health Phys. 13: 1211-1221.
23 Michel, O.; Nagy, A.-M.; Schroeven, M.; Duchateau, J.; Neve, J.; Fondu, P.; Sergysels, R. (1997) Dose-response
24 relationship to inhaled endotoxin in normal subjects. Am. J. Respir. Crit. Care Med. 156: 1157-1164.
25 Morrison, D. C.; Jacobs, D. M. (1976) Binding of polymyxin B to the lipid A portion of bacterial
26 lipopolysaccharides. Immunochemistry 13: 813-818.
27 Morrow, P. E. (1973) Alveolar clearance of aerosols. Arch. Intern. Med. 131: 101-108.
28 Morrow, P. E. (1977) Clearance kinetics of inhaled particles. In: Brain, J. D.; Proctor, D. F.; Reid, L. M., eds.
29 Respiratory defense mechanisms (in two parts), part II. New York, NY: Marcel Dekker, Inc.; pp. 491-543.
30 (Lenfant, C., ed. Lung biology in health and disease: v. 5).
31 Morrow, P. E. (1986) Factors determining hygroscopic aerosol deposition in airways. Physiol. Rev. 66: 330-376.
32 Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
33 10:369-384.
34 Morrow, P. E. (1994) Mechanisms and significance of "particle overload." In: Mohr, U.; Dungworth, D. L.;
35 Mauderly, J. L.; Oberdorster, G., eds. Toxic and carcinogenic effects of solid particles in the respiratory tract:
36 [proceedings of the 4th international inhalation symposium]; March 1993; Hannover, Germany. Washington,
37 DC: International Life Sciences Institute Press; pp. 17-25.
38 Morrow, P. E.; Yu, C. P. (1993) Models of aerosol behavior in airways and alveoli. In: Moren, F.; Dolovich, M. B.;
39 Newhouse, M. T.; Newman, S. P., eds. Aerosols in medicine: principles, diagnosis and therapy. 2nd rev. ed.
40 Amsterdam, The Netherlands: Elsevier; pp. 157-193.
41 Morrow, P. E.; Gibb, F. R.; Gazioglu, K. M. (1967) A study of particulate clearance from the human lungs.
42 Am. Rev. Respir. Dis. 96: 1209-1221.
43 Mueller, E. J.; Seger, D. L. (1985) Metal fume fever-a review. J. Emerg. Med. 2: 271-274.
44 Miiller, H. L.; Drosselmeyer, E.; Hotz, G.; Seidel, A.; Thiele, H.; Pickering, S. (1989) Behaviour of spherical and
45 irregular (U,Pu)O2 particles after inhalation or intratracheal instillation in rat lung and during in vitro culture
46 with bovine alveolar macrophages. Int. J. Radiat. Biol. 55: 829-842.
47 Murphy, S. A.; BeruBe, K. A.; Pooley, F. D.; Richards, R. J. (1998) The response of lung epithelium to well
48 characterised fine particles. Life Sci. 62: 1789-1799.
49 Nadeau, D.; Vincent, R.; Kumarathasan, P.; Brook, J.; Dufresne, A. (1996) Cytotoxicity of ambient air particles to
50 rat lung macrophages: comparison of cellular and functional assays. Toxicol. In Vitro 10: 161-172.
51 National Council on Radiation Protection and Measurements. (1997) Deposition, retention and dosimetry of inhaled
52 radioactive substances. Bethesda, MD: National Council on Radiation Protection and Measurements; report
53 no. 125.
October 1999 7-113 DRAFT-DO NOT QUOTE OR CITE
-------
1 National Institutes of Health. (1997) Guidelines for the diagnosis and management of asthma: expert panel report 2.
2 Bethesda, MD: U.S. Department of Health and Human Services, National Heart, Lung, and Blood Institute;
3 publication no. 97-4051.
4 Naumann, B. D.; Schlesinger, R. B. (1986) Assessment of early alveolar particle clearance and macrophage
5 function following an acute inhalation of sulfuric acid mist. Exp. Lung Res. 11: 13-33.
6 Nel, A. E.; Diaz-Sanchez, D.; Ng, D.; Hiura, T.; Saxon, A. (1998) Enhancement of allergic inflammation by the
7 interaction between diesel exhaust particles and the immune system. J. Allergy Clin. Immunol. 102: 539-554.
8 Nikula, K. J.; Avila, K. J.; Griffith, W. C.; Mauderly, J. L. (1997) Lung tissue responses and sites of particle
9 retention differ between rats and cynomolgus monkeys exposed chronically to diesel exhaust and coal dust.
10 Fundam. Appl. Toxicol. 37: 37-53.
11 Noone, P. G.; Bennett, W. D.; Regnis, J. A.; Zeman, K. L.; Carson, J. L.; King, M.; Boucher, R. C.; Knowles, M. R.
12 (1999) Effect of aerosolized uridine-5'-triphosphate (UTP) on airway clearance with cough in patients with
13 primary ciliary dyskinesia. Am. J. Respir. Crit. Care Med.: 160: 144-149.
14 Oberdorster, G. (1993) Lung dosimetry: pulmonary clearance of inhaled particles. Aerosol Sci. Technol.
15 18:279-289.
16 Oberdorster, G.; Oldiges, H.; Zimmermann, B. (1980) Lung deposition and clearance of cadmium in rats exposed
17 by inhalation or by intratracheal instillation. Zentralbl. Bakteriol. Abt. 1 Orig. B 170: 35-43.
18 Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992) Role of the alveolar macrophage in
19 lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
20 Oberdorster, G.; Gelein, R. M.; Ferin, J.; Weiss, B. (1995) Association of particulate air pollution and acute
21 mortality: involvement of ultrafine particles? In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the
22 colloquium on particulate air pollution and human mortality and morbidity; January 1994; Irvine, CA.
23 Inhalation Toxicol. 7: 111-124.
24 Oberdorster, G.; Cox, C.; Gelein, R. (1997) Intratracheal instillation versus intratracheal inhalation of tracer
25 particles for measuring lung clearance function. Exp. Lung Res. 23: 17-34.
26 Oberdorster, G.; Finkelstein, J. N.; Johnston, C.; Gelein, R.; Cox, C.; Baggs, R.; Elder, A. (1999) Ultrafine particles
27 as inducers of acute lung injury: mechanisms and correlation with age and disease. Cambridge, MA: Health
28 Effects Institute: submitted.
29 Ohtoshi, T.; Takizawa, H.; Okazaki, H.; Kawasaki, S.; Takeuchi, N.; Ohta, K.; Ito, K. (1998) Diesel exhaust
30 particles stimulate human airway epithelial cells to produce cytokines relevant to airway inflammation in
31 vitro. J. Allergy Clin. Immunol. 101: 778-785.
32 Oldham, M. J.; Mannix, R. C.; Phalen, R. F. (1997) Deposition of monodisperse particles in hollow models
33 representing adult and child-size tracheobronchial airways. Health Phys. 72: 827-834.
34 Osier, M.; Oberdorster, G. (1997) Intratracheal inhalation vs intratracheal instillation: differences in particle effects.
35 Fundam. Appl. Toxicol. 40: 220-227.
36 Osier, M.; Baggs, R. B.; Oberdorster, G. (1997) Intratracheal instillation versus intratracheal inhalation: influence of
37 cytokines on inflammatory response. Environ. Health Perspect. Suppl. 105(5): 1265-1271.
38 Osman, K.; Zejda, J. E.; Schutz, A.; Mielzynska, D.; Elinder, C. G.; Vahter, M. (1998) Exposure to lead and other
39 metals in children from Katowice district, Poland. Int. Arch. Occup. Environ. Health71: 180-186.
40 Passali, D.; Bianchini Ciampoli, M. (1985) Normal values of mucociliary transport time in young subjects. Int. J.
41 Pediatr. Otorhinolaryngol. 9: 151-156.
42 Patten (1996).
43 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
44 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
45 Pierce, L. M.; Alessandrini, F.; Godleski, J. J.; Paulauskis, J. D. (1996) Vanadium-induced chemokine mRNA
46 expression and pulmonary inflammation. Toxicol. Appl. Pharmacol. 138: 1-11.
47 Pinto, M.; Birnbaum, S. C.; Kadar, T.; Goldberg, G. M. (1979) Lung injury in mice induced by factors acting
48 synergistically with inhaled particulate antigen. Clin. Immunol. Immunopathol. 13: 361-368.
49 Prahalad, A. K.; Soukup, J. M.; Inmon, J.; Willis, R.; Ghio, A. J.; Becker, S.; Gallagher, J. E. (1999) Ambient air
50 particles: effects on cellular oxidant radical generation in relation to particulate elemental chemistry.
51 Toxicol. Appl. Pharmacol. 158: 81-91.
52 Pritchard, J. N.; Holmes, A.; Evans, J. C.; Evans, N.; Evans, R. J.; Morgan, A. (1985) The distribution of dust in the
53 rat lung following administration by inhalation and by single intratracheal instillation. Environ. Res.
54 36:268-297.
October 1999 7-114 DRAFT-DO NOT QUOTE OR CITE
-------
1 Pritchard, J. N.; Jefferies, S. J.; Black, A. (1986) Sex differences in the regional deposition of inhaled particles in
2 the 2.5-7.5 urn size range. J. Aerosol Sci. 17: 385-389.
3 Pritchard, R. J.; Ohio, A. J.; Lehmann, J. R.; Winsett, D. W.; Tepper, J. S.; Park, P.; Gilmour, M. I.; Dreher, K. L.;
4 Costa, D. L. (1996) Oxidant generation and lung injury after particulate air pollutant exposure increase with
5 the concentrations of associated metals. Inhalation Toxicol. 8: 457-477.
6 Quay, J. L.; Reed, W.; Samet, J.; Devlin, R. B. (1998) Air pollution particles induce IL-6 gene expression in human
7 airway epithelial cells viaNF-KB activation. Am. J. Respir. Cell Mol. Biol. 19: 98-106.
8 Radford, E. P.; Martell, E. A. (1977) Polonium-210: lead-210 ratios as an index of residence time of insoluble
9 particles from cigarette smoke in bronchial epithelium. In: Walton, W. H.; McGovern, B., eds. Inhaled
10 particles IV: proceedings of an international symposium, part 2; September 1975; Edinburgh, United
11 Kingdom. Oxford, United Kingdom: Pergamon Press, Ltd.; pp. 567-581.
12 Rose, C. S.; Martyny, J. W.; Newman, L. S.; Milton, D. K.; King, T. E.; Bebe, J. L.; McCammon, J. B.; Hoffman,
13 R. E.; Kreiss, K. (1998) "Lifeguard lung": endemic granulomatous pneumonitis in an indoor swimming pool.
14 Am. J. Public Health 88: 1795-1800.
15 Roy, M. (1989) Lung clearance modeling on the basis of physiological and biological parameters. Health Phys.
16 57(suppl. 1): 255-262.
17 Rudell, B.; Sandstrom, T.; Stjernberg, N.; Kolmodin-Hedman, B. (1990) Controlled diesel exhaust exposure in an
18 exposure chamber: pulmonary effects investigated with bronchoalveolar lavage. J. Aerosol Sci.
19 21(suppl. 1):S411-S414
20 Rutland, J.; Cole, P. J. (1981) Nasal mucociliary clearance and ciliary beat frequency in cystic fibrosis compared
21 with sinusitis and bronchiectasis. Thorax 36: 654-658.
22 Sagai, M.; Furuyama, A.; Ichinose, T. (1996) Biological effects of diesel exhaust particles (DEP). III. Pathogenesis
23 of asthma like symptoms in mice. Free Radic. Biol. Med. 21:199-209.
24 Samet, J. M.; Reed, W.; Ohio, A. J.; Devlin, R. B.; Carter, J. D.; Dailey, L. A.; Bromberg, P. A.; Madden, M. C.
25 (1996) Induction of prostaglandin H synthase 2 in human airway epithelial cells exposed to residual oil fly
26 ash. Toxicol. Appl. Pharmacol. 141: 159-168.
27 Samet, J. M.; Stonehuerner, J.; Reed, W.; Devlin, R. B.; Dailey, L. A.; Kennedy, T. P.; Bromberg, P. A.; Ohio, A. J.
28 (1997) Disruption of protein tyrosine phosphate homeostasis in bronchial epithelial cells exposed to oil fly
29 ash. Am. J. Physiol. 272: L426-L432.
30 Samet, J. M.; Graves, L. M.; Quay, J.; Dailey, L. A.; Devlin, R. B.; Ohio, A. J.; Wu, W.; Bromberg, P. A.; Reed, W.
31 (1998) Activation of MAPKs in human bronchial epithelial cells exposed to metals. Am. J. Physiol.
32 275: L551-L558.
33 Scheuch, G.; Stahlhofen, W. (1988) Particle deposition of inhaled aerosol boluses in the upper human airways.
34 J. Aerosol Med. 1:29-36.
35 Scheuch, G.; Philipson, K.; Falk, R.; Svartengren, M.; Stahlhofen, W.; Camner, P. (1995) Retention of particles
36 inhaled in boli with and without induced bronchoconstriction. Exp. Lung Res. 21: 901-916.
37 Schlesinger, R. B. (1985) Clearance from the respiratory tract. Fundam. Appl. Toxicol. 5: 435-450.
38 Schlesinger, R. B. (1988) Biological disposition of airborne particles: basic principles and application to vehicular
39 emissions. In: Watson, A. Y.; Bates, R. R.; Kennedy, D., eds. Air pollution, the automobile, and public
40 health. Washington, DC: National Academy Press; pp. 239-298.
41 Schlesinger, R. B. (1995) Deposition and clearance of inhaled particles. In: McClellan, R. O.; Henderson, R. F.,
42 eds. Concepts in inhalation toxicology. 2nd ed. Washington, DC: Taylor & Francis; pp. 191-224.
43 Schlesinger, R. B.; Ben-Jebria, A.; Dahl, A. R.; Snipes, M. B.; Ultman, J. (1997) Disposition of inhaled toxicants.
44 In: Massaro, E. J., ed. Handbook of human toxicology. Boca Raton, FL: CRC Press; pp. 493-550.
45 Schwartz, J. (1994) Air pollution and hospital admissions for the elderly in Birmingham, Alabama. Am. J.
46 Epidemiol. 139: 589-598.
47 Seaton, A.; MacNee, W.; Donaldson, K.; Godden, D. (1995) Particulate air pollution and acute health effects.
48 Lancet (8943): 176-178.
49 Sioutas, C.; Koutrakis, P.; Burton, R. M. (1995) A technique to expose animals to concentrated fine ambient
50 aerosols. Environ. Health Perspect. 103: 172-177.
51 Sjogren, B. (1997) Occupational exposure to dust: inflammation and ischaemic heart disease. Occup. Environ. Med.
52 54:466-469.
53 Smaldone, G. C.; Perry, R. J.; Bennett, W. D.; Messina, M. S.; Zwang, J.; Ilowite, J. (1988) Interpretation of
54 "24 hour lung retention" in studies of mucociliary clearance. J. Aerosol Med. 1: 11-20.
October 1999 7-115 DRAFT-DO NOT QUOTE OR CITE
-------
1 Smith, K. R.; Aust, A. E. (1997) Mobilization of iron from urban participates leads to generation of reactive oxygen
2 species in vitro and induction of ferritin synthesis in human lung epithelial cells. Chem. Res. Toxicol.
3 10: 828-834.
4 Snipes, M. B. (1989) Long-term retention and clearance of particles inhaled by mammalian species. CRC Crit. Rev.
5 Toxicol. 20: 175-211.
6 Snipes, M. B.; Clem, M. F. (1981) Retention of micro spheres in the rat lung after intratracheal instillation. Environ.
7 Res. 24:33-41.
8 Snipes, M. B.; McClellan, R. O.; Mauderly, J. L.; Wolff, R. K. (1989) Retention patterns for inhaled particles in the
9 lung: comparisons between laboratory animals and humans for chronic exposures. Health Phys.
10 57(suppl. 1): 69-78.
11 Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G.; Philipson, K. (1986a) Clearance from the human airways of
12 particles of different sizes deposited from inhaled aerosol boli. In: Aerosols: formation and reactivity,
13 proceedings of the second international aerosol conference; September; Berlin, Federal Republic of
14 Germany. Oxford, United Kingdom: Pergamon Press; pp. 192-196.
15 Stahlhofen, W.; Gebhart, J.; Rudolf, G.; Scheuch, G. (1986b) Measurement of lung clearance with pulses of
16 radioactively-labelled aerosols. J. Aerosol Sci. 17: 333-336.
17 Stanley, P. J.; Wilson, R.; Greenstone, M. A.; Mackay, I. S.; Cole, P. J. (1985) Abnormal nasal mucociliary
18 clearance in patients with rhinitis and its relationship to concomitant chest disease. Br. J. Dis. Chest
19 79:77-82.
20 Stober, W.; Morrow, P. E.; Koch, W.; Morawietz, G. (1994) Alveolar clearance and retention of inhaled insoluble
21 particles in rats simulated by a model inferring macrophage particle load distributions. J. Aerosol Sci.
22 25:975-1002.
23 Stringer, B.; Kobzik, L. (1996) Alveolar macrophage uptake of the environmental particulate titanium dioxide:
24 role of surfactant components. Am. J. Respir. Cell Mol. Biol. 14: 155-160.
25 Stringer, B.; Kobzik, L. (1998) Environmental particulate-mediated cytokine production in lung epithelial cells
26 (A549): role of preexisting inflammation and oxidant stress. J. Toxicol. Environ. Health Part A 55: 31-44.
27 Stringer, B.; Imrich, A.; Kobzik, L. (1996) Lung epithelial cell (A549) interaction with unopsonized environmental
28 particulates: quantitation of particle-specific binding and IL-8 production. Exp. Lung Res. 22: 495-508.
29 Strom, K. A.; Chan, T. L.; Johnson, J. T. (1988) Pulmonary retention of inhaled submicron particles in rats: diesel
30 exhaust exposures and lung retention model. In: Dodgson, J.; McCallum, R. I.; Bailey, M. R.; Fischer, D. R.,
31 eds. Inhaled particles VI: proceedings of an international symposium and workshop on lung dosimetry;
32 September 1985; Cambridge, United Kingdom. Ann. Occup. Hyg. 32(suppl. 1): 645-657.
33 Svartengren, K.; Lindestad, P.; Svartengren, M.; Philipson, K.; Bylin, G.; Camner, P. (1995) Added external
34 resistance reduces oropharyngeal deposition and increases lung deposition of aerosol particles in asthmatics.
35 Am. J. Respir. Crit. Care Med. 152: 32-37.
36 Svartengren, K.; Philipson, K.; Svartengren, M.; Anderson, M.; Camner, P. (1996a) Tracheobronchial deposition
37 and clearance in small airways in asthmatic subjects. Eur. Respir. J. 9: 1123-1129.
38 Svartengren, K.; Ericsson, C. H.; Svartengren, M.; Mossberg, B.; Philipson, K.; Camner, P. (1996b) Deposition and
39 clearance in large and small airways in chronic bronchitis. Exp. Lung Res. 22: 555-576.
40 Svartengren, K.; Philipson, K.; Svartengren, M.; Camner, P. (1998) Effect of adrenergic stimulation on clearance
41 from small ciliated airways in healthy subjects. Exp. Lung Res. 24: 149-158.
42 Swift, D. L.; Strong, J. C. (1996) Nasal deposition of ultrafine 218Po aerosols in human subjects. J. Aerosol Sci.
43 27:1125-1132.
44 Takahashi, S.; Asaho, S.; Kubota, Y.; Sato, H.; Matsuoka, O. (1987) Distribution of 198Au and 133Ba in thoracic and
45 cervical lymph nodes of the rat following the intratracheal instillation of 198Au-colloid and 133BaSO4.
46 J. Radiat. Res. 28:227-231.
47 Takahashi, S.; Kubota, Y.; Hatsuno, H. (1992) Effect of size on the movement of latex particles in the respiratory
48 tract following local administration. Inhalation Toxicol. 4: 113-123.
49 Takano, H.; Yoshikawa, T.; Ichinose, T.; Miyabara, Y.; Imaoka, K.; Sagai, M. (1997) Diesel exhaust particles
50 enhance antigen-induced airway inflammation and local cytokine expression in mice. Am. J. Respir. Crit.
51 Care Med. 156: 36-42.
52 Takano, H.; Ichinose, T.; Miyabara, Y.; Shibuya, T.; Lim, H.-B.; Yoshikawa, T. Sagai, M. (1998) Inhalation of
53 diesel exhaust enhances allergen-related eosinophil recruitment and airway hyperresponsiveness in mice.
54 Toxicol. Appl. Pharmacol. 150: 328-337.
October 1999 7-116 DRAFT-DO NOT QUOTE OR CITE
-------
1 Terashima, T.; Wiggs, B.; English, D.; Hogg, J. C.; Van Eeden, S. F. (1997) Phagocytosis of small carbon particles
2 (PM10) by alveolar macrophages stimulates the release of polymorphonuclear leukocytes from bone marrow.
3 Am. J. Respir. Crit. Care Med. 155: 1441-1447.
4 Thurston, G. D.; Lippmann, M.; Scott, M. B.; Fine, J. M. (1997) Summertime haze air pollution and children with
5 asthma. Am. J. Respir. Crit. Care Med. 155: 654-660.
6 Timblin, C.; BeruBe, K.; Chrug, A.; Driscoll, K.; Gordon, T.; Hemenway, D.; Walsh, E.; Cummins, A. B.;
7 Vacek, P.; Mossman, B. (1998) Ambient particulate matter causes activation of the c- jun
8 kinase/stress-activated protein kinase cascade and DNA synthesis in lung epithelial cells. Cancer Res.
9 58:4543-4547.
10 Toms, N.; Hasani, A.; Pavia, D.; Clarke, S. W.; Agnew, J. E. (1997) Effect of mucus hypersecretion on initial time
11 course of inert particle clearance from the lung. In: Cherry, N.; Ogden, T., eds. Inhaled Particles VIII:
12 proceedings of an international symposium on inhaled particles organised by the British Occupational
13 Hygiene Society; August 1996; Cambridge, UK. Ann. Occup. Hyg. 41(suppl.): 509-514.
14 Tsien, A.; Diaz-Sanchez, D.; Ma, J.; Saxon, A. (1997) The organic component of diesel exhaust particles and
15 phenanthrene, a major polyaromatic hydrocarbon constituent, enhances IgE production by IgE-secreting
16 EBV-transformed human B cells in vitro. Toxicol. Appl. Pharmacol. 142: 256-263.
17 Tsuchiyama, F.; Hisanaga, N.; Shibata, E.; Aoki, T.; Takagi, H.; Ando, T.; Takeuchi, Y. (1997) Pulmonary metal
18 distribution in urban dwellers. Int. Arch. Occup. Environ. Health 70: 77-84.
19 U.S. Environmental Protection Agency. (1982) Air quality criteria for particulate matter and sulfur oxides. Research
20 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and Assessment
21 Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA; PB84-156777.
22 U.S. Environmental Protection Agency. (1989) An acid aerosols issue paper: health effects and aerometrics.
23 Research Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
24 Assessment Office; EPA report no. EPA-600/8-88-005F. Available from: NTIS, Springfield, VA;
25 PB91-125864.
26 U.S. Environmental Protection Agency. (1996a) Air quality criteria for particulate matter. Research Triangle Park,
27 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
28 Available from: NTIS, Springfield, VA; PB96-168224.
29 U.S. Environmental Protection Agency. (1996b) Ambient levels and noncancer health effects of inhaled crystalline
30 and amorphous silica: health issue assessment [draft]. Research Triangle Park, NC: National Center for
31 Environmental Assessment; EPA report no. EPA/600/R-95/115.
32 U.S. Environmental Protection Agency. (1999) Health assessment document for diesel emissions, SAB review draft.
33 Washington, DC: Office of Research and Development; report no. EPA/600/8-90/057C.
34 Utell, M. J.; Morrow, P. E.; Speers, D. M.; Darling, J.; Hyde, R. W. (1983a) Airway responses to sulfate and
35 sulfuric acid aerosols in asthmatics: an exposure-response relationship. Am. Rev. Respir. Dis. 128: 444-450.
36 Utell, M. J.; Morrow, P. E.; Hyde, R. W. (1983b) Latent development of airway hyperreactivity in human subjects
37 after sulfuric acid aerosol exposure. J. Aerosol Sci. 14: 202-205.
38 Vanda, B.; de Buen, N.; Jasso, R.; Valero, G.; Vargas, M. H.; Olmos, R.; Arreola, J. L.; Santillan, P.; Alonso, P.
39 (1998) Inflammatory cells and ferruginous bodies in bronchoalveolar lavage in urban dogs. Acta Cytol.
40 42:939-944.
41 Veronesi, B.; Oortgiesen, M.; Carter, J. D.; Devlin, R. B. (1999) Particulate matter initiates inflammatory cytokine
42 release by activation of capsaicin and acid receptors in a human bronchial epithelial cell line. Toxicol. Appl.
43 Pharmacol. 154: 106-115.
44 Vincent, R.; Bjarnason, S. G.; Adamson, I. Y. R.; Hedgecock, C.; Kumarathasan, P.; Guenette, J.; Potvin, M.;
45 Goegan, P.; Bouthillier, L. (1997) Acute pulmonary toxicity of urban particulate matter and ozone. Am. J.
46 Pathol. 151: 1563-1570.
47 Vogelzang, P. F. J.; Van Der Gulden, J. W. J.; Folgering, H.; Kolk, J. J.; Heederik, D.; Preller, L.; Tielen, M. J. M.;
48 Van Schayck, C. P. (1998) Endotoxin exposure as a major determinant of lung function decline in pig
49 farmers. Am. J. Respir. Crit. Care Med. 157: 15-18.
50 Watkinson, W. P.; Campen, M. J.; Costa, D. L. (1998) Cardiac arrhythmia induction after exposure to residual oil
51 fly ash particles in a rodent model of pulmonary hypertension. Toxicol. Sci. 41: 209-216.
52 Wheat!ey, L. M.; Platts-Mills, T. A. E. (1996) Perennial allergens and the asthma epidemic. Sci. Med. 3: 6-13.
53 Wolff, R. K.; Henderson, R. F.; Snipes, M. B.; Griffith, W. C.; Mauderly, J. L.; Cuddihy, R. G.; McClellan, R. O.
54 (1987) Alterations in particle accumulation and clearance in lungs of rats chronically exposed to diesel
55 exhaust. Fundam. Appl. Toxicol. 9: 154-166.
October 1999 7-117 DRAFT-DO NOT QUOTE OR CITE
-------
1 Woodin, M. A.; Hauser, R.; Liu, Y.; Smith, T. J.; Siegel, P. D.; Lewis, D. M.; Tollerad, D. J.; Christiani, D. C.
2 (1998) Molecular markers of acute upper airway inflammation in workers exposed to fuel-oil ash. Am. J.
3 Respir. Crit. Care Med. 158: 182-187.
4 Xu, G. B.; Yu, C. P. (1986) Effects of age on deposition of inhaled aerosols in the human lung. Aerosol Sci.
5 Technol. 5: 349-357.
6 Yeates, D. B.; Aspin, M. (1978) A mathematical description of the airways of the human lungs. Respir. Physiol.
7 32:91-104.
8 Yeates, D. B.; Aspin, N.; Levison, H.; Jones, M. T.; Bryan, A. C. (1975) Mucociliary tracheal transport rates in
9 man. J. Appl. Physiol. 19: 487-495.
10 Yeates, D. B.; Pitt, B. R.; Spektor, D. M.; Karron, G. A.; Albert, R. E. (1981) Coordination of mucociliary transport
11 in human trachea and intrapulmonary airways. J. Appl. Physiol.: Respir. Environ. Exercise Physiol.
12 51: 1057-1064.
13 Yeh, H.-C.; Schum, G. M. (1980) Models of human lung airways and their application to inhaled particle
14 deposition. Bull. Math. Biol. 42: 461-480.
15 Yoo, H. S.; Maheswaran, S. K.; Lin, G.; Townsend, E. L.; Ames, T. R. (1995) Induction of inflammatory cytokines
16 in bovine alveolar macrophages following stimulation with Pasteurella haemolytica lipopolysaccharide.
17 Infect. Immun. 63:381-388.
18 Yuetal. (1998).
19 Zelikoff, J. T.; Sisco, M.; Cohen, M. D.; Frampton, M. W.; Utell, M. J; Schlesinger, R. B. (1994) Interspecies
20 comparison of immunotoxicity of inhaled sulfuric acid. II. New Zealand white rabbits. In: 1994 international
21 conference sponsored by the American Lung Association and the American Thoracic Society; May; Boston,
22 MA. Am. J. Respir. Crit. Care Med. 149: A621.
23 Zelikoff, J. T.; Frampton, M. W.; Cohen, M. D.; Morrow, P. E.; Sisco, M.; Tsai, Y.; Utell, M. J.; Schlesinger, R. B.
24 (1997) Effects of inhaled sulfuric acid aerosols on pulmonary immunocompetence: a comparative study in
25 humans and animals. Inhalation Toxicol. 9: 731-752.
26 Zhang, Z.; Martonen, T. (1997) Deposition of ultrafine aerosols in human tracheobronchial airways. Inhalation
27 Toxicol. 9:99-110.
28 Zock, J.-P.; Hollander, A.; Heederik, D.; Douwes, J. (1998) Acute lung function changes and low endotoxin
29 exposures in the potato processing industry. Am. J. Ind. Med. 33: 384-391.
October 1999 7-118 DRAFT-DO NOT QUOTE OR CITE
-------
i 8. INTEGRATIVE SYNTHESIS OF KEY POINTS:
2 PM EXPOSURE, DOSIMETRY, AND HEALTH RISKS
3
4
5 8.1 INTRODUCTION
6 This chapter integrates key information on exposure-dose-response risk assessment
7 components drawn from the preceding detailed chapters (Chapters 3 to 7), in order to provide a
8 coherent framework for assessment of human health risks posed by ambient particulate matter
9 (PM) in the United States. Given that substantial additional important new information is
10 expected to become available during the next 6 to 9 mo for incorporation into a second external
11 review draft of this document, only very provisional conclusions can now be drawn from the
12 information discussed in earlier chapters and a preliminary attempt at an integrative synthesis
13 framework set forth below. The key elements of the integrative synthesis framework include
14 discussion of: (a) atmospheric science and exposure research evidence substantiating
15 distinctions between fine and coarse particles as separate subclasses of ambient PM air
16 pollutants; (b) factors affecting PM dosimetry, especially in the human respiratory tract; (c) the
17 expanding epidemiologic evidence regarding effects of ambient PM on human health; and
18 (d) new toxicological evaluations of pathophysiologic effects of PM and potential mechanisms
19 of action.
20
21
22 8.2 AIRBORNE PARTICLES: DISTINCTIONS BETWEEN FINE AND
23 COARSE PARTICLES AS SEPARATE POLLUTANT SUBCLASSES
24 As discussed in detail in Chapter 3 of this document, airborne PM is not a single pollutant
25 but several classes of pollutants, each consisting of a number of chemical species. One
26 classification is based on the natural division of the atmospheric aerosol into fine-mode and
27 coarse-mode particles. Fine-mode particles, in addition to being smaller than coarse-mode
28 particles, differ in many aspects such as formation mechanisms, chemical composition, sources,
29 physical behavior, human exposure relationships, and control approaches required for risk
30 reduction. Such differences alone are sufficient to justify consideration of fine-mode and
31 coarse-mode particles as separate pollutants. Table 8-1 compares several key points that
October 1999 8-1 DRAFT-DO NOT QUOTE OR CITE
-------
o
a
o
VO
VO
VO
TABLE 8-1. COMPARISON OF PHYSICAL AND CHEMICAL PROPERTIES OF AMBIENT PARTICLES:
FINE MODE (Nuclei Mode Plus Accumulation Mode) AND COARSE MODE
Fine
Coarse
Nuclei
Accumulation
Formed from:
Formed by:
Combustion, high temperature processes and atmospheric reactions Break-up of large solids/droplets
oo
to
H
6
o
*
o
H
O
o
&
o
Composed of:
Solubility:
Atmospheric
half-life:
Removal
Processes:
Travel distance:
Nucleation
Condensation
Coagulation
Sulfates
Elemental carbon
Metal compounds
Organic compounds with
very low saturation vapor
pressure at ambient
temperature
Probably less soluble than
accumulation mode
Minutes to hours
Grows into accumulation
mode
-------
1 differentiate fine-mode from coarse-mode particles. Various physical and chemical differences
2 between fine-mode particles and coarse-mode particles, their sources, factors affecting human
3 exposure, and their respiratory tract deposition are also summarized below as a prelude to more
4 in-depth discussion of key health effects associated with ambient PM exposures and other
5 information useful in assessing PM-related public health risks in the United States.
6
7 8.2.1 Size Distinctions
8 Four approaches are used to classify particles by size: (1) modes, based on formation
9 mechanisms and the modal structure observed in the atmosphere; (2) size cut point, based on the
10 50% cut point of the specific sampling device; (3) occupational classification based on
11 dosimetry, the ability of particles to enter certain regions of the respiratory tract, and
12 (4) regulatory size cuts. The modal structure is shown in Figure 8-1. In the ambient atmosphere
13 the fine particle mode (< 1.0 //m diameter) is composed of the nuclei mode and the accumulation
14 mode. The nuclei mode is clearly observable only near sources of condensible gases. Particles
15 in the nuclei mode rapidly grow into the accumulation mode but the accumulation mode does not
16 grow further into the coarse particle mode. The lognormal distribution (in units of particle
17 diameter) is frequently used to approximate the distribution of particle number, surface area,
18 volume, or mass. The nuclei mode (<0.1 //m) includes ultrafme particles (toxicology
19 terminology) and nanoparticles (aerosol physics terminology). Particle diameters are usually
20 given as aerodynamic equivalent diameter, dae, defined as the diameter of a particle with a
21 settling velocity equal to that of a sphere with unit density (1 g/cm3). This is the most appropriate
22 diameter for discussion of lung deposition and particle collection. The accumulation mode
23 typically has a mass median aerodynamic diameter (MMAD) of 0.3 to 0.7//m and a geometric
24 standard deviation, og (a measure of the size dispersion), of 1.5 to 1.8. The coarse particle mode
25 may also contain multiple modes but they are not readily distinguished. Therefore, the coarse
26 particle mode tends to have a broader size distribution, with og = 2.2 to 2.4. Measured MMADs
27 typically range from 6 to 20 //m diameter in the ambient atmosphere, but these values may be
28 low because of the difficulty of collecting particles in the upper tail of the coarse-mode
29 distribution.
30 Agreement has been reached between the International Standards Organization (ISO) and
31 American Council of Government Industrial Hygienists (ACGIH) who have promulgated
October 1999 8-3 DRAFT-DO NOT QUOTE OR CITE
-------
o
CO
Q.
a
O)
o
8
7
6
5
4
3
2
1
0
0.002
Mechanically
Generated
0.01 0.1 1
Geometric Diameter, Dp, |jm
Nuclei Mode Accumulation Mode
10
100
Fine-Mode Particles
Coarse Mode
• ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^H • I
Coarse-Mode Particles
Figure 8-1. Volume size distribution, measured in traffic, showing fine-mode and coarse-
mode particles and the nuclei and accumulation modes within the fine-particle
mode. DGV (geometric mean diameter by volume, equivalent to volume
median diameter) and og (geometric standard deviation) are shown for each
mode. Also shown are transformation and growth mechanisms
(e.g., nucleation, condensation, and coagulation).
Source: Wilson etal. (1977).
1 definitions of particle size fractions that are based on the ability of particles to penetrate to
2 various depths within the respiratory tract (Vincent, 1995). Inhalable refers to particles which
3 can enter beyond the external airway openings and, as discussed in Chapter 10, has a practical
4 upper limit of 40 to 60 //m. Thoracic refers to particles which can penetrate beyond the larynx;
5 about 50% of particles of 10 //m aerodynamic diameter will penetrate beyond the larynx.
6 Respirable refers to particles which can reach the air exchange portion of the lung.
October 1999
8-4
DRAFT-DO NOT QUOTE OR CITE
-------
1 The appropriate division between the fine and coarse fractions is not sharply defined, but
2 falls in the range between 1.0 and 3.0 //m dae, where fine-mode and coarse-mode particles
3 overlap but where particle mass is at a minimum. Thus, in general, particles less than 1.0 //m dae
4 are fine-mode particles and particles greater than 2.5//m dae are coarse-mode particles. However,
5 as the relative humidity approaches 100%, fine particles may grow beyond 1.0 //m and even
6 beyond 2.5 //m dae; and, in very dry environments, it may also be possible to find particles less
7 than 1.0 //m dae in the small size tail of the coarse particle mode. It is important to note that
8 PM25 may sometimes contain an appreciable quantity of coarse-mode particles in the 1 to 2.5 //m
9 dae size range.
10 PM25 particles are frequently referred to as fine particles, while the difference between
11 PM2 5 and PM10 (PM(10_2 5)), is sometimes referred to as coarse particles or as the coarse fraction of
12 PM10. In the present discussion, fine-mode particles and coarse-mode particles are used to
13 emphasize that important distinctions include not just size but also other additional fundamental
14 differences in sources, formation mechanisms, and chemical composition.
15 The indicators for the current PM standards are PM10 and PM2 5. Since neither the
16 respiratory tract nor particle samplers can separate particles with a sharp cut, PM10 is defined as
17 having a 50% cutpoint at 10 //m dae. PM10 samplers collect all fine-mode particles. They also
18 collect a decreasing fraction of coarse particles as the diameter increases above 10 //m dae and an
19 increasing fraction of particles as the diameter decreases below 10 //m dae. The mass of the
20 coarse fraction ranges from 20% of PM10 in some eastern urban areas to 80% of PM10 in some
21 dry western areas. The penetration curve for PM10 is very close to that of thoracic particles
22 (Figure 3-6). PM10 is a design standard where the specified design provides a 50% cut point at
23 10 //m dae. PM25 is a combination of design and performance specifications which provide a
24 50% cut at 2.5 //m dae. This cut insures that virtually all fine particles will be collected, even
25 under high relative humidity conditions, but it also includes a fraction of the small-size tail of the
26 coarse particle distribution.
27
28 8.2.2 Formation Mechanisms
29 Fine particles may be formed during combustion processes. Fine particles are also formed
30 from condensible gases by nucleation (gas molecules coming together to form a new particle),
31 and grow by condensation (gas molecules condensing onto a pre-existing particle). Condensible
October 1999 8-5 DRAFT-DO NOT QUOTE OR CITE
-------
1 gases (low saturation vapor pressure at ambient temperature) may be formed by volatilization of
2 material during combustion or other high temperature processes or by atmospheric reactions that
3 generate condensible gases. Gases may dissolve in a liquid droplet (either a solution particle or a
4 cloud or fog droplet), react with another dissolved gas, and form a low vapor pressure product.
5 When fog and cloud droplets evaporate, particulate matter remains, usually in the fine particle
6 mode. Gases may also react directly with solid particles or in water films on solid particles.
7 Coarse particles are formed by mechanical processes which produce small particles from
8 large ones. Energy considerations normally limit coarse mode particles to sizes greater than
9 about 1.0 //m dae.
10 Particles are designated as primary if they are emitted directly into the air as particles or as
11 vapors which condense to form particles without chemical reaction. Examples of primary
12 particles are (a) elemental carbon chain agglomerates formed during combustion and
13 (b) chemical species such as lead, cadmium, selenium, or sulfuric acid which are volatile at
14 combustion temperature but form PM rapidly as the combustion gases cool.
15 Particles are designated as secondary if they form following a chemical reaction in the
16 atmosphere which converts a gaseous precursor to a product which either has a low enough
17 saturation vapor pressure to form a particle or reacts further to form a low saturation vapor
18 pressure product. Examples are the conversion of sulfur dioxide (SO2) to sulfuric acid (H2SO4)
19 which nucleates or condenses on existing particles, or the conversion of nitrogen dioxide (NO2)
20 to nitric acid (HNO3) which may react further with ammonia (NH3) to form particulate
21 ammonium nitrate (NH4NO3).
22 Coarse particles are normally primary since they are formed by mechanical rather than by
23 chemical processes. An exception is the reaction of acid gases with carbonate (CO3) containing
24 particles in which the CO3 may be replaced by sulfate (SO42), nitrate (NO3), or chloride (Cl").
25 Other exceptions are the reaction of HNO3 with NaCl to form NaNO3 and HC1 gas and the
26 reaction of SO2 with wet NaCl to form Na2SO4 and HC1 gas. Similar reactions may occur
27 between other basic and acidic species.
28
29
October 1999 8-6 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.3 Chemical Composition
2 8.2.3.1 Fine-Mode Particulate Matter
3 In the ambient atmosphere, fine-mode particulate matter is mainly composed of varying
4 proportions of sulfate, nitrate, hydrogen, and ammonium ions; elemental and organic carbon;
5 trace elements such as metals; and particle-bound water.
6 Sulfates/Acid. Sulfur dioxide (SO2), emitted mainly from combustion of fossil fuel, is
7 oxidized in the atmosphere to form sulfuric acid (H2SO4) particles. The H2SO4 may be partially
8 or completely neutralized by reaction with ammonia (NH3). Since the particles usually contain
9 water, the actual species present are H+, HSO4, SO42, and NH4, in varying proportions depending
10 on the amount of NH3 available to react with the H2SO4. Particle strong acidity is due to free H+
11 or H+ available from HSO4 or H2SO4.
12 Nitrates. Nitrogen oxides (NOX= NO + NO2) are formed during combustion and other high
13 temperature processes involving air. The NO is converted to NO2 by ozone (O3) or other
14 atmospheric oxidants. During the daytime, NO2 reacts with the hydroxyl radical (OH) to form
15 nitric acid (HNO3). During nighttime, it forms nitric acid through a sequence of reactions
16 involving ozone and the nitrate radical (NO3). Ammonia reacts preferentially with sulfuric acid
17 to form sulfate particles, but, if sufficient NH3 is available, parti culate ammonium nitrate
18 (NH4NO3) will al so form.
19 Elemental Carbon. Chain agglomerates of very small elemental carbon (EC) particles are
20 formed during combustion, such as in open hearth fireplaces, wood stoves and diesel engines.
21 Organic Carbon. Several heterogenous categories of organic carbon (OC) compounds are
22 also often found in ambient air, as follows:
23 • Primary-anthropogenic. Incomplete combustion also leads to hundreds of organic
24 compounds with low enough vapor pressure to be present in the atmosphere as particles,
25 including mutagenic species such as some polyaromatic hydrocarbons (PAHs).
26 • Secondary-anthropogenic. Some organic compounds, including aromatics (larger
27 than benzene), cyclic olefms and diolefms, and other C7 or higher hydrocarbons, react
28 with O3 or OH to form polar, oxygenated compounds with ambient temperature,
29 saturation vapor pressures low enough to form particles.
30 • Primary biogenic. Viruses, some bacteria, and plant and/or animal cell fragments may
31 be found in the fine mode.
October 1999 8-7 DRAFT-DO NOT QUOTE OR CITE
-------
1 • Secondary biogenic. Monoterpenes, C10 cyclic olefms released by plants, and other
2 organic compounds from plants, also react in the atmosphere to yield organic particulate
3 matter.
4 Trace Elements. A variety of transition metals and non-metals are volatilized during the
5 combustion of fossil fuels, smelting of ores, and incineration of wastes and are emitted as fine
6 particles (or vapors which rapidly form fine particles).
7 Water. Sulfates, nitrates, and some organic compounds are hygroscopic, i.e., they absorb
8 water and form solution droplets. A variety of atmospheric pollutant gases can dissolve in the
9 water component of the particle. This provides a possible mechanism for carrying into the lung
10 soluble species such as SO2, H2O2, HCHO, etc., which, when in the gas phase, would normally be
11 removed in the nose, throat, or upper airways. The amount of particle-bound water increases as
12 the relative humidity increases; relative humidity in the respiratory tract approaches 100%.
13
14 8.2.3.2 Coarse-Mode Particulate Matter
15 Coarse-mode PM sources are primarily crustal, biological, or industrial in nature.
16 Crustal Crustal material, from soil or rock, primarily consists of compounds that contain
17 Si, Al, Fe, Mg, and K (small amounts of Fe and K are also found among fine-mode particles but
18 come from different sources). In urban areas, much crustal material arises from soil which is
19 tracked onto roads during wet periods and is suspended in the air by vehicular traffic. In rural
20 areas, tilling, wind blowing over disturbed soil, or vehicles traveling on unpaved roads can
21 generate coarse particles. Where farms have been treated with persistent pesticides or herbicides,
22 these materials may also be present in suspended soil particles.
23 Biological. Biological materials such as bacteria, pollen, spores, and other plant and
24 animal fragments are mostly found in the coarse size range (i.e., 2.0 to 10 //m dae for most,
25 >20 //m dae for some).
26 Industrial. A variety of industrial operations generate coarse particles. Examples are
27 construction and demolition, open pit mining, grain handling, coal handling, etc. Also, coal and
28 oil combustion generate fly ash which is similar in chemical composition to soil and crustal
29 material but can be differentiated by microscopic examination. In the U.S. almost all fly ash
30 from large scale coal and oil combustion is removed by effective particle control technologies.
31
October 1999 8-8 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.2.4 Atmospheric Behavior
2 Coarse-mode particles are large enough so that the force of gravity exceeds the buoyancy
3 forces of the air. Therefore, large particles tend to rapidly fall out of the air. Coarse-mode
4 particles are also too large to follow air streams, so they tend to be easily removed by impaction
5 on surfaces. The atmospheric half-life of coarse particles depends on their size, but is usually
6 only minutes to hours. However, vigorous mixing and convection, such as occurs during dust
7 storms, can lead to longer lifetimes for the smaller size range of coarse-mode particles.
8 In contrast, fine-mode particles are small enough that gravitational forces are largely
9 overcome by the random forces from collisions with gas molecules. Thus fine particles tend to
10 follow air streams and are typically not removed by impaction (unless the air stream is
11 accelerated in a particle sampler). Accumulation-mode particles are sufficiently larger than gas
12 molecules that their diffusion velocity is low. Removal by dry deposition is inefficient since they
13 do not readily diffuse through the boundary layer of still air next to surfaces. Therefore,
14 accumulation-mode particles have very long half-lives in the atmosphere, travel long distances,
15 and tend to be more uniformly distributed over large geographic areas than coarse-mode
16 particles. The atmospheric half-life of accumulation-mode particles with respect to dry
17 deposition is on the order of weeks. Removal of accumulation-mode particles occurs when the
18 particles absorb water, grow into cloud droplets, grow further to rain drops, and fall out as rain.
19 This process reduces the atmospheric half-life of accumulation-mode particles to a few days.
20 Nuclei-mode particles, formed by nucleation of low saturation-vapor-pressure substances,
21 tend to exist as disaggregated individual particles for very short periods of time (
-------
o
a
o
VO
VO
VO
TABLE 8-2. CONSTITUENTS AND MAJOR SOURCES OF ATMOSPHERIC PARTICLES
oo
I
o
H
6
o
2
o
H
O
o
?d
o
Sources
Aerosol species
so;2
Sulfate
NO3-
Minerals
NH4+
Ammonium
Organic carbon
(OC)
Elemental carbon
(EC)
Metals
Bioaerosols
Primary (PM < 2.5 ,um) Primary (PM > 2.5 ,um)
Natural Anthropogenic Natural Anthropogenic
Sea Spray Fossil fuel combustion Sea Spray —
— Motor vehicle exhaust2 — —
Erosion, Fugitive dust; paved, Erosion, re-entrainment Fugitive dust; paved,
re-entrainment unpaved roads; unpaved road dust,
agriculture and forestry agriculture and forestry
— Motor vehicle exhaust2 — —
Wild fires Open burning, wood — Tire and asphalt wear,
burning, motor paved road dust
vehicle exhaust,
cooking
Wild fires Motor vehicle — —
exhaust, wood
burning, cooking
Volcanic activity Fossil fuel Erosion, re-entrainment, —
combustion, smelting, organic debris
brake wear
Viruses, bacteria — Plant, insect fragments,
pollen, fungal spores,
bacterial agglomerates
Secondary PM Precursors (PM < 2.5 ,um)
Natural
Oxidation of reduced sulfur
gases emitted by the oceans
and wetlands; and SO2 and
H2S emitted by volcanism
and forest fires
Oxidation of NOX produced
by soils, forest fires, and
lightning
—
Emissions of NH3 from wild
animals, undisturbed soil
Oxidation of hydrocarbons
emitted by vegetation,
(terpenes, waxes); wild fires
—
Anthropogenic
Oxidation of SO2
emitted from fossil fuel
combustion1
Oxidation of NOX
emitted from fossil fuel
combustion; and in
motor exhaust
—
Emissions of NH3 from
animal husbandry,
sewage, fertilized land
Oxidation of
hydrocarbons emitted by
motor vehicles, open
burning, wood burning
—
'Major source of each component shown in boldface type.
2Relatively minor primary source of substance, included only for the sake of completeness.
-------
1 reactions in particles; and the condensation of low vapor pressure photochemical reaction
2 products. Coarse particles, on the other hand, are produced mainly by the mechanical processes
3 (e.g., wind erosion, tire friction).
4 For a variety of reasons, concentrations of aerosol constituents measured at specific
5 monitoring sites do not reflect the composition that would be obtained from a straightforward
6 application of the emissions shown in Chapter 4. Although mineral dust, from wind erosion,
7 agricultural activities and fugitive dust, represents the largest single category of PM25 emissions
8 by mass (accounting for 62% of the total), it rarely accounts for more than half of the mass of
9 ambient samples.
10 In contrast, the results of a number of monitoring studies show that in the eastern and
11 central United States, mineral dust contributes less than 10% of PM2 5 and about 15% of PM25 in
12 the West. During dust storms in arid regions of the West, the fractional contribution of mineral
13 dust to PM2 5 can be much higher. Sulfate and associated water of hydration constitute a larger
14 fraction of PM2 5 in the East than in the West.
15 There has been a marked improvement in recent years in the ability of receptor models to
16 apportion PM to its sources. This improvement has come about because of the use of organic
17 species (i.e., organic compounds in the OC fraction) as tracers to distinguish among different
18 forms of combustion (e.g., gasoline vs diesel fueled vehicles, biomass burning, and meat
19 cooking) and even to identify vegetative detritus. The removal of Pb from gasoline had limited
20 the ability of receptor models to apportion PM to mobile sources in the past. However, the use of
21 organic compounds as tracers has been employed in a few studies, mainly in the western United
22 States.
23
24 8.2.6 Community and Personal Ambient PM Concentrations Exposure
25 Relationships
26 As discussed in Chapter 5, atmospheric behavior differences between fine-mode and
27 coarse-mode particles lead to important differences in relationships between personal exposure to
28 these ambient PM constituents and their ambient concentrations measured at a central fixed-site
29 monitor. Fine particles tend to have long atmospheric half-lives, can travel long distances, and
30 therefore can result from distant or widely distributed sources. Evidence from some cities (e.g.,
31 Philadelphia), suggest that the concentrations of fine particles may be uniform over large urban
October 1999 8-11 DRAFT-DO NOT QUOTE OR CITE
-------
1 areas. Thus, a PM25 measurement at one site may give a reasonable estimate of the fine particle
2 concentration across a city or even wider regional areas, assuming the site is not unduly
3 influenced by a local source of fine particles. Ambient coarse particles, however, have more
4 localized and variable sources and, because such particles are rapidly removed, their
5 concentration decreases with distance from the source and the distribution of PM(10_25) may not be
6 uniform across a city or region. Thus, people in one part of a city may experience high
7 concentrations of coarse fraction particles on one day while people in a different part of the city
8 may experience high concentrations on another day, even though the city-wide average
9 concentration may be the same on both days. This unevenness of coarse mode particle
10 distribution across a city may need to be taken into account when assessing health impacts in
11 community epidemiological studies.
12 A further consideration arises with regard to relationships between ambient (outdoor) PM
13 concentrations and personal or indoor exposures to PM of the same AD size. Because people
14 spend most of their time indoors, the particle concentrations indoors tend to dominate personal
15 exposures. However, indoor exposure is due both to particles generated indoors and to ambient
16 particles generated outdoors but which have infiltrated indoors. Major indoor sources of fine
17 particles are smoking and cooking. The major indoor sources of coarse particles are indoor
18 activities that resuspend previously settled PM and that stir up and suspend other materials,
19 including a variety of biological materials such as mold spores and insect debris. Household
20 cleaning, especially dusting and vacuuming, can dramatically increase indoor coarse particle
21 concentrations. When doors and windows are open, both fine-mode and coarse-mode particles
22 will penetrate from outdoors to indoors with little loss in passage. When doors and windows are
23 closed, particle penetration depends on AD size and air exchange rate, with penetration of
24 ambient particles to indoor microenvironments decreasing with increasing AD size. Once
25 indoors, particle sizes influence their half-lives in that microenvironment. Coarse-mode particles
26 are rapidly removed by deposition, whereas ultrafine and accumulation-mode particles have
27 longer half-lives. The production of indoor-generated particles is controlled by daily indoor
28 activities, which Mage et al. (1999) have shown to be independent of ambient PM
29 concentrations. The exposure to indoor-generated particles will often not be significantly
30 correlated with the concentration of ambient (outdoor-generated) particles. Therefore,
October 1999 8-12 DRAFT-DO NOT QUOTE OR CITE
-------
1 time-series epidemiology based on ambient PM measurements is not capable of identifying
2 health effects related to indoor-generated particles.
3 The various penetration and removal processes for PM can be modeled, and the equilibrium
4 ratio of the concentration of ambient particles which have penetrated indoors and remained
5 suspended to the concentration of ambient particles outdoors (called the infiltration ratio) can be
6 calculated as a function of the air exchange rate, the penetration factor, and the deposition
7 removal rates which are a function of particle AD size. Infiltration ratio calculations, based on
8 data from the Particle Total Exposure Assessment Methodology Study (PTEAM), were depicted
9 in Figure 13-2 of U.S. Environmental Protection Agency (1996). As is evident in Figure 5-13
10 (Chapter 5, this document) the infiltration ratio of sulfate, which is almost completely of outdoor
11 origin and expected to be in the fine-mode, is greater than that of PM25. Figure 5-4 (Chapter 5,
12 this document) shows that PM25, in turn, has a greater infiltration ratio than PM(10.25).
13 The more uniform distribution of ambient fine-mode particles across a city and the higher
14 infiltration ratio for fine particles means that an ambient measure of fine particles at a central site
15 may provide a useful estimate of the average exposure of people in the community to fine-mode
16 particles of ambient origin. For example, experimental data on personal exposure to sulfates,
17 which are predominantly of outdoor origin and in the fine-mode particle size range, show
18 consistently high correlation of total human exposure to sulfate with outdoor central-site
19 measurements of ambient sulfates (0.78 < R2 < 0.92) (Figure 5-13). However, because of the
20 non-uniform regional concentrations and lower infiltration ratios, an ambient measure of coarse
21 particles, such as PM(10_2 5) at a central site, may not provide nearly as good an indication of
22 exposure of people in the community to coarse particles of ambient origin. Much of the
23 time-series epidemiology currently available is based on ambient TSP or PM10 measurements,
24 which represent the sum of both fine and coarse (in the case of TSP) or the sum of fine particles
25 and the coarse-mode fraction of TSP less than 10 //m AD (in the case of PM10). In Philadelphia,
26 and in some other cities to a lesser extent (where PM10 is not dominated by coarse wind-blown
27 dust), it has been shown that TSP and PM10 concentrations correlate better with PM2 5
28 concentrations than with the coarse fraction of PM10 (PM(10.25)). It is thus possible that the
29 observed statistical relationships between various ambient particle indicators and health
30 outcomes are largely due to an underlying relationship between fine-mode particles and health
31 outcomes, or an undiscovered pollutant highly correlated with ambient PM2 5 concentrations.
October 1999 8-13 DRAFT-DO NOT QUOTE OR CITE
-------
1 This hypothesis is supported by recent epidemiological analyses (e.g., Schwartz et al., 1996) for
2 cities where both PM2 5 and PM(10_2 5) data are available.
3
4
5 8.3 FACTORS AFFECTING PM DOSIMETRY
6 A full characterization of the exposure-dose-response continuum is needed to reduce
7 uncertainty in extrapolations from laboratory animals or healthy humans to susceptible humans.
8 In the case of PM, this characterization requires understanding of particle deposition and
9 clearance, toxicant-target interactions, and tissue responses.
10
11 8.3.1 Factors Determining Deposition and Clearance
12 Particles are deposited in the respiratory tract by mechanisms of impaction, sedimentation,
13 interception, diffusion, and electrostatic precipitation. Differences in ventilation rates, in the
14 upper respiratory tract structure, and in the size and branching pattern of the respiratory tract may
15 alter particle deposition. Air flow in the extrathoracic (ET) region is characterized by high
16 velocity and abrupt directional changes and, thus, deposition in this region, especially for PM
17 > 1 //m, is mainly by inertial impaction. However, for ultrafine particles, deposition in the
18 ET region is mainly by diffusion although electrostatic precipitation may also play a role. In the
19 alveolar (A) region, deposition by diffusion is important.
20 Disposition and retention of deposited particles depends on clearance and translocation
21 mechanisms that vary across respiratory tract regions. Sneezing and nose wiping or blowing and
22 mucociliary transport to the gastrointestinal tract via the pharynx are important clearance
23 processes for particles deposited in the ET region, whereas coughing, mucociliary transport,
24 endocytosis by macrophages or epithelial cells and dissolution and absorption into the blood or
25 lymph are important in the tracheobronchial (TB) region. Endocytosis by macrophages or
26 epithelial cells and dissolution and absorption into blood or lymph are important mechanisms in
27 the A region. The fate of a deposited particle depends on the deposition site, physicochemical
28 properties of the particle (e.g., solubility), and time.
29 Variation in respiratory tract architecture, especially in the smaller conducting airways and
30 gas exchange regions, can be critical to the dosimetry of inhaled particles. Deposition of ambient
31 particles in the lung depends in part upon their aerodynamic and physicochemical properties.
October 1999 8-14 DRAFT-DO NOT QUOTE OR CITE
-------
1 Structural changes in the respiratory tract with chronic obstructive pulmonary disease (COPD)
2 affect airflow and the aerodynamic behavior of inhaled particles. In severe COPD, the healthy
3 portion of the lung receives more of the tidal volume resulting in some ventilatory units receiving
4 a larger particle burden than others. Kim and Kang (1997) demonstrated greater deposition of
5 1 //m particles in people with varying degrees of airway obstruction than in healthy subjects. The
6 increase in deposition was greatest for COPD patients and asthmatics but was also increased for
7 smokers. Svartengren et al. (1994) showed enhanced deposition in asthmatics. Bennett et al.
8 (1997) reported a greater deposition rate (particles/time) in COPD patients and an increased
9 ventilation which resulted in a total deposition approximately 2.5 fold greater than in healthy
10 adults. Model simulations also predict that dose expressed in particle numbers per anatomical
11 unit would be increased in people with compromised lungs (Miller et al., 1995). Not only may
12 patients with preexisting COPD be susceptible because of an enhanced or altered deposited dose
13 pattern, but their disease may also predispose these patients to altered responses to the toxic
14 effects of ambient PM.
15 Physicochemical characteristics of particles (e.g., particle diameter, distribution,
16 hygroscopicity) interact with the anatomic (e.g., branching pattern) and physiologic (e.g.,
17 ventilation rate, clearance processes) factors to influence deposition and retention of inhaled
18 aerosols. Two key parameters which characterize size distribution are the MMAD and the og of
19 the particles. The relative contribution of these anatomic, physiologic, and physicochemical
20 properties as well as ambient concentration and exposure duration must be integrated to assess
21 overall deposition.
22 The influence of the particle size distribution on the fraction of particles deposited in the
23 respiratory tract is illustrated in Figure 8-2. This figure depicts the predicted deposition fractions
24 for an adult male, using a typical ventilation pattern, in the alveolar (A), tracheobronchial (TB),
25 and thoracic (A + TB) regions. The difference between total respiratory tract and total thoracic
26 deposition fractions represents the extrathoracic (ET) or upper airway deposition fraction. The
27 deposition fraction in the respiratory tract, relative to unit mass concentration in air, is shown for
28 particles of different MMAD, in the range of 0.1 to 100 //m, for a geometric standard deviation
29 of 1.8. A recent model simulation considered regional deposition in the respiratory tract regions
30 (nasopharyngeal, tracheobronchial, and pulmonary) for the fine and coarse fraction of PM10 for
31 both mass and number deposition (Table 8-3). This simulation is for a healthy person breathing
October 1999 8-15 DRAFT-DO NOT QUOTE OR CITE
-------
0.01
Total Respiratory
Tract
0.1 1 10
MMAD ([jm) withog = 1.8
100
• Alveolar (Normal) A TB (Normal) • Total Thoracic (Normal)
0 Alveolar (Mouth) A TB (Mouth) D Total Thoracic (Mouth)
Figure 8-2. Human respiratory tract PM deposition fraction and PM10 or PM2 5 sampler
collection versus mass median aerodynamic diameter (MMAD) with geometric
standard deviation, og = 1.8. Alveolar, tracheobronchial, or total thoracic
deposition fractions predicted for normal augmenter versus mouth breather
adult male using a general population (ICRP66) minute volume activity
pattern and the 1994 ICRP66 model.
1
2
3
4
5
6
150 //g/m3 PM10 (57% PM25) for 24-h at varying activity and ventilation patterns. The results
indicate the clear dominance of both mass and number of fine mode particles in the pulmonary
region and the dominance of coarse mode mass in the extrathoracic region.
Simulations, performed for U.S. Environmental Protection Agency (1996), showed that
alveolar deposition fraction is fairly uniform for aerosols between 0.5 and 4.0 //m MMAD.
Alveolar region deposition increases for particles less than 0.5 //m. Particles in the 0.3 to 0.5 //m
October 1999
8-16
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-3. MODEL SIMULATION CONSIDERING REGIONAL DEPOSITION
FOR FINE AND COARSE PM10 FOR MASS AND NUMBER
Region
Mass
Dose
NPL
TBL
PUL
Number
Dose
NPL
TBL
PUL
Fine/
Coarse
Fine
Coarse
Fine
Coarse
Fine
Coarse
Fine/
Coarse
Fine
Coarse
Fine
Coarse
Fine
Coarse
For PM10 = 150 //g/m3 (57% = PM25)
Fractional Mass Deposited,
//g/day
25-51
413-687
29-38
50-52
108-194
44-55
Fractional Number of
Particles Deposited/day
5-15 xlO8
6-10 xlO6
2.2-3. IxlO10
10.7-11. IxlO5
9.3-16.7 xlO10
13.6-17 xlO5
24 h exposure
% of Inhaled Fraction
Mass Deposited
1.5%
30%
1.5%
4%
6%
2%
% of Inhaled Fraction
Number Deposited
0.06%
13%
2%
2%
9%
3%
NPL = Nasopharyngeal
TBL = Tracheobronchial
PUL = Alveolar - Pulmonary region
Fine = PM2 5
Coarse = PM(i0.2.5)
Source: Venkataraman and Kao (1999)
1
2
3
4
5
6
7
range undergo the lowest deposition rate in the respiratory tract. In the aerodynamic range of
particles (> 1.0 //m MMAD), deposition fraction increases as particle size increases and
sedimentation and impaction become important deposition mechanisms, especially for larger
particles (> 5 //m MMAD) in the TB region. This pattern is altered slightly for mouth breathing
versus normal breathing, in that mouth breathers have a greater TB deposition of particles larger
than 2.5 //m (i.e., the coarse fraction of PM10) than they would if breathing PM only via the nose.
October 1999
8-17
DRAFT-DO NOT QUOTE OR CITE
-------
1 8.3.2 Factors Determining Toxicant-Target Interactions and Response
2 Differences in susceptibility can be due to factors influencing deposited and retained
3 particle mass or number, toxicant-target interaction, or tissue response. Deposited particle dose
4 may be characterized in terms of particle mass, particle number, or particle surface area.
5 Furthermore each of these parameters can be expressed per gram of tissue, per cm2 of respiratory
6 tract surface area, or as dose/unit time. The most suitable metric to characterize a response will
7 likely depend upon the mechanism of action of the particle or particle constituent. The
8 biologically-effective dose may be described by particle mass or number deposition alone if the
9 particles exert their primary action on the surface contacted (Dahl et al., 1991). For longer-term
10 effects, the deposited dose may not be a useful metric, since particles clear at varying rates from
11 the different respiratory tract regions. When considering the epidemiologic data, dose metrics
12 could be separated into two categories, pattern and quantity of acute deposition and the pattern
13 and quantity of retained dose. The deposited dose may be more important for daily mortality,
14 hospital admissions, work loss days, etc. On the other hand the retained dose may be more
15 important for chronic responses.
16 Many analyses have relied upon the particle mass concentration (//g/m3) breathed by
17 exposed individuals. If relative risk (RR) estimates were calculated based on various internal
18 dose metrics (e.g., deposited dose [mass] normalized per unit tracheobronchial or alveolar
19 surface area or normalized per critical cell type such as the alveolar macrophage), relationships
20 could change. The fine fraction contains by far the largest number of particles, and those
21 particles have a larger aggregate surface area than coarse-mode particles. Retardation of alveolar
22 macrophage phagocytosis due to particle overload appears to be better correlated with particle
23 surface area than particle mass (Morrow, 1988; Oberdorster et al., 1995a,b). Also, ultrafme
24 particles have been shown to be less effectively phagocytosed by macrophages than larger
25 particles (Oberdorster et al., 1992a,b).
26 Figure 8-3 presents an example which illustrates the complexities of considering PM
27 "dose" using different metrics (e.g., such as mass, surface area, and number of particles). For the
28 accumulation mode, which constitutes about 40% of the total mass in the illustrated sample, the
29 geometric mean for the volume distribution, DGV, equivalent to the volume median diameter, is
30 0.31 //m. When the median diameter is expressed in terms of surface area, or count, the
31 respective median diameters of the accumulation mode are 0.19 //m and 0.07 //m. By far the
October 1999 8-18 DRAFT-DO NOT QUOTE OR CITE
-------
CO
Nn = 7.7x 10
DGNn = 0.013
Nc = 4.2
DGNC = 0.97
30 H
(D
F -a 20 H
5 Q
10H
Vn = 0.33
DGVn = 0.031
0.001
0.01
0.1
1.0
i
10
100
Figure 8-3. Distribution of coarse (c), accumulation (a), and nuclei or ultrafine (n) mode
particles by three characteristics: volume (V), surface area (S), and number
(N) for the grand average continental size distribution. DGV = geometric
mean diameter by volume; DGS = geometric mean diameter by surface area;
DGN = geometric mean diameter by number; Dp = geometric diameter.
Source: Whitby (1978).
October 1999
8-19
DRAFT-DO NOT QUOTE OR CITE
-------
1 largest number of particles are contained in the nuclei mode, which is inconsequential in terms of
2 mass. The composition of the particles in each mode is different, as are their hygroscopicity,
3 solubility, translocation pathways, and toxicity.
4 How could particle size be important in biological activity? The mass of the particle may
5 be important if the mechanism of action of the particle is related to its persistence. Larger
6 particles will typically take longer to dissolve or to be degraded enzymatically. If presentation of
7 active groups to cell surfaces by less soluble particles is important in the mechanisms of action,
8 then the total surface area of the particles should be important. The largest aggregate surface area
9 is in the accumulation mode. Biological effects on epithelial cells or macrophages may depend
10 on the number of cell surface receptors that are contacted or the number of particles ingested by a
11 phagocytic cell.
12
13 8.3.3 Construction of Exposure-Dose-Response Continuum for PM
14 Toxicological data in laboratory animals typically can aid the interpretation of human
15 clinical and epidemiological data because these studies provide concentration- and duration-
16 response information on a more complex array of effects and exposures than can typically be
17 evaluated in humans. However the use of laboratory animal toxicological data has typically been
18 limited because of difficulties in quantitative extrapolation to humans. The various species used
19 in inhalation toxicological studies do not receive identical doses in comparable respiratory tract
20 regions (ET, TB, A) when exposed to the same aerosol (same composition, mass, concentration,
21 and size characteristics). Furthermore, a number of recent toxicological studies of in vivo PM
22 effects use intratracheal instillation, which is problematic in terms of extrapolation to inhalation
23 in humans. Such interspecies and methodological differences are important because the adverse
24 toxic effects are likely related more to the quantitative pattern of deposition within the respiratory
25 tract than to the exposure alone; this pattern determines not only the initial respiratory tract tissue
26 dose, but also the specific pathways by which the inhaled particles are cleared and redistributed.
27 Another difficulty in elucidating the exposure-dose-response continuum using laboratory
28 animal data is that different endpoints are typically assayed in the laboratory animals and the
29 relationship of these endpoints to the human health outcomes have not been established.
30 Epidemiological studies evaluate endpoints such as illness, hospital admissions, and emergency
31 room/doctor visits; homologous biochemical or pathological endpoints in laboratory animal
October 1999 8-20 DRAFT-DO NOT QUOTE OR CITE
-------
1 models are unknown. However, a growing number of newer PM laboratory animal studies have
2 been performed on compromised animal models and, increasingly, human clinical studies are
3 examining the responses of older adults and those with cardiac or pulmonary disease.
4 In summary, until the mechanism(s) of action for effects induced by ambient PM or its
5 important constituents can be more definitively characterized, the linkage between exposure and
6 response provided by dosimetry cannot be considered quantitative. Thus, any insights to be
7 derived from dosimetry will be limited until the dose metrics that correlate well with PM
8 mechanism(s) of action are determined.
9
10
11 8.4 EXPANDING EPIDEMIOLOGIC INFORMATION ON HEALTH
12 EFFECTS OF PARTICULATE MATTER
13 8.4.1 Introduction
14 The purpose of this section is to update the most recent previous assessment (U.S.
15 Environmental Protection Agency 1996) of the information on human health epidemiology
16 studies of effects of exposure to ambient PM. This includes assessment of new scientific
17 evidence in the following areas by: (1) reporting the strengths and limitations of the available
18 epidemiologic findings, particularly those that tend to support or to refute the findings previously
19 reported; (2) assessing the biomedical significance and the coherence of the new studies,
20 including aspects of human exposure to ambient PM and biological consequences of such
21 exposure; (3) evaluating the plausibility of inferences about the relationship(s) between human
22 health and ambient PM exposure based on aerometric studies, exposure assessments, studies of
23 deposited PM dose, and studies on mechanisms of toxicity; (4) assessing the extent to which
24 adverse health effects may be attributable to PM and to other environmental factors, with
25 particular attention to PM size fractions, sources, or specific chemical components;
26 (5) quantifying the relationships between ambient PM concentrations and adverse health effects
27 in susceptible sub-populations at different time scales, where epidemiologic and toxicologic
28 evidence supports such quantification. Table 8-4 summarizes some of this information for fine
29 and coarse mode particles.
30 In recent years, many epidemiologic studies have shown associations of short-term ambient
31 air pollution exposure with mortality, exacerbation of pre-existing illness, and physiologic
October 1999 8-21 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-4. FINE AND COARSE MODE: EXPOSURE, DEPOSITION,
EPIDEMIOLOGY, AND BIOLOGICAL EFFECTS
FINE MODE PARTICLES
(FMP)
COARSE MODE PARTICLES
(CMP)
EXPOSURE
LUNG DEPOSITION1
Mass:
Number:
Particles/macrophage:
Ambient monitor representative
of personal exposure to FMP
mass
AL deposition exceeds combined
NP and TB deposition by two
fold. Approximately 9% of
inhaled mass is deposited
Responsible for
> 99% alveolar,
> 99% TB
> 99%
Ambient monitor not as
representative of personal exposure
to CMP mass
AL deposition is about half of TB
deposition. NP deposition about 5x
(AL + TB). Approximately 35% of
inhaled particles are deposited
Responsible for
< 1% alveolar,
< 1% TB
EPIDEMIOLOGY
Acute: Increase in daily mortality
2% per increase of 10 ,ug/m3.
Statistically significant
0.5% or less per increase of
10,ug/m3. Often not statistically
significant
Chronic: Decrease in life span
Relationship statistically
significant
Relationship not statistically
significant
INITIATORS OF
BIOLOGICAL EFFECTS
(Hypotheses, Speculation)
Strong acidity, transition metals,
ultrafmes, dissolved gases
Silica, biological substances: spores,
insect and plant fragments
NP - Nasopharyngeal
TB - Tracheobronchial
AL - Alveolar
'Venkataraman and Kao (1999)
1 changes. These studies have increased concern as to whether ambient air pollution exposure
2 can promote, and perhaps even produce, harmful health outcomes, even when pollutant
3 concentrations are at or below current U.S. air quality standards.
October 1999 8-22 DRAFT-DO NOT QUOTE OR CITE
-------
1 The epidemiologic database regarding short-term ambient air pollution exposure is growing
2 rapidly, and its interpretation is evolving with insights from the new data. As recently as the
3 mid-1990's, many epidemiologic studies had reported associations of mortality and exacerbation
4 of pre-existing disease with ambient levels of PM, some of which had not only demonstrated
5 significant PM associations in single (PM) or multipollutant models but also had investigated or
6 reported such associations with gaseous pollutants (including CO and O3). Since then, a growing
7 number of epidemiologic studies have given more thorough consideration to both PM and
8 gaseous pollutants, and many have frequently observed positive, statistically significant
9 associations of harmful effects with both. Thus, although associations of PM with harmful
10 effects continue to be observed consistently across most of the new studies, the newer findings
11 appear to complicate further the task of trying to sort out relative contributions to the observed
12 epidemiologic associations of (a) PM acting alone; (b) PM acting in combination with gaseous
13 co-pollutants; (c) the gaseous pollutants per se; or (d) the overall ambient pollutant mix. With
14 considerable new experimental evidence also in hand, and after more analysis of this issue, it is
15 possible to hypothesize ways in which ambient exposure to multiple air pollutants (including not
16 only PM acting alone but also in combination with others) could plausibly be involved in the
17 complex chain of biological events leading to harmful health effects in the human population.
18 In epidemiologic studies of ambient air pollution, small positive estimates of air pollutant
19 health effects have been observed quite consistently. These estimates have frequently been
20 statistically significant at cc<0.05. If ambient air pollution actually promotes or produces harmful
21 health effects, relatively small effect estimates from current PM concentrations in the U.S. and
22 many other countries would generally be expected on biological and epidemiologic grounds.
23 Also, the magnitudes and significance levels of observed air pollution-related effects estimates
24 have varied somewhat from place to place. This would also be expected if the observed
25 epidemiologic associations denote actual effects, because, not only would the complex mixture
26 of PM vary from place to place but also the affected populations may also differ in characteristics
27 that could affect susceptibility to air pollution health effects. These characteristics include
28 demographic and socioeconomic factors, underlying health status, indoor-outdoor activities, diet,
29 medical care systems and access to them, and exposure to risk factors other than ambient air
30 pollution, such as extreme weather conditions.
October 1999 8-23 DRAFT-DO NOT QUOTE OR CITE
-------
1 Thus, although it has been argued that the observed effects estimates for ambient air
2 pollution are not sufficiently constant across epidemiologic studies and that epidemiologic
3 studies are trustworthy only if they show relatively large effects estimates (e.g., large relative
4 risks), these arguments have only limited weight in relation to ambient air pollution studies.
5 Also, in any large population exposed to ambient air pollution, even a small relative risk for a
6 widely prevalent health disorder could calculate to a substantial public health burden attributable
7 to air pollution exposure.
8 The ambient atmosphere contains numerous air pollutants, and it is important to continue to
9 recognize that health effects associated statistically with any single pollutant may actually be
10 mediated by multiple components of the complex ambient mix. Specific attribution of effects to
11 any single pollutant may therefore be overly simplistic. PM is one of many air pollutants derived
12 from combustion sources, including mobile sources. These pollutants include PM, CO, sulfur
13 oxides, nitrogen oxides, and ozone, all of which have been considered in various epidemiologic
14 studies to date. Numerous volatile or semivolatile organic compounds are also emitted by
15 combustion sources or formed in the atmosphere, which have not yet been systematically
16 considered in relation to the non-cancer health outcomes usually associated with exposure to
17 criteria air pollutants. In many of the newly available epidemiologic studies, harmful health
18 outcomes are often associated with multiple combustion-related or mobile source-related air
19 pollutants, and some investigators have raised the possibility that PM may be a surrogate or
20 marker for a larger subset of the overall ambient air pollution mix. However, others have
21 reserved judgment on this issue, and many, including the National Research Council (National
22 Research Council, 1998, 1999), have emphasized the need for further research on PM in order to
23 better address this issue.
24 As discussed above, small health effects estimates have generally been observed for
25 ambient air pollutants, and small effects would indeed be expected on biological and
26 epidemiologic grounds. In contrast to effects estimates derived for the 1952 London smog
27 episode with RR > 4 for extremely high (>2 mg/m3) ambient PM concentrations, effects
28 estimates in most current epidemiology studies at distinctly lower PM concentrations (often
29 <200 to 500 //g/m3) are small and the statistical estimates (a) are more often subject to relatively
30 small (but proportionately large) differences in estimated effects of PM and other pollutants,
31 (b) may be sensitive to a variety of methodological choices; and (c) may sometimes not be
October 1999 8-24 DRAFT-DO NOT QUOTE OR CITE
-------
1 statistically significant, reflecting low statistical power of the study design to detect a small but
2 real effect.
3 It seems likely that pollutant effects estimates in multi-pollutant models would be more
4 biologically and epidemiologically sound than those in single-pollutant models, although it is
5 conceivable that single pollutant models might also be credible if independent biological
6 plausibility evidence supported designation of PM or some other single pollutant as likely being
7 the key toxicant in the ambient pollutant mix being evaluated. However, neither of these
8 possibilities have been convincingly demonstrated, and scientific consensus as to optimal
9 modeling strategies for time series air pollution studies has not yet been achieved. Therefore, the
10 choice of effects estimates to employ in risk assessments for short-term ambient air pollution
11 effects remains open to question.
12 In available studies, statistical uncertainty has generally been assessed rather superficially,
13 without formal consideration of the model tuning performed by the investigators. For example,
14 lag times and averaging times for air pollutant metrics have usually been selected to maximize
15 statistical effects estimates for pollutants. This technique may have led not only to unrealistically
16 large reported effects estimates, but also to inappropriately narrow confidence intervals. In future
17 studies, uncertainty arising from model tuning should be more carefully assessed. In this effort,
18 resampling or simulation procedures, which would recreate the entire model estimation process,
19 should be considered.
20 Exacerbation of heart disease has been epidemiologically associated not only with ambient
21 PM, but also with other combustion-related ambient pollutants such as NO2 and CO. Thus, the
22 quantitation of the proportion of risk for such exacerbation specifically attributable to ambient
23 PM exposure is unclear. Recent studies, e.g. concentrated ambient particle studies (CAPS), have
24 demonstrated cardiovascular effects in response to ambient particle exposures and studies
25 utilizing other techniques have also produced various results suggesting some plausible
26 mechanisms for cardiovascular effects. However, much remains to be resolved with regard to
27 delineation of dose-response relationships for the induction of such effects and the extrapolation
28 of such to estimate effective human equivalent exposures to ambient PM (or specific constituent)
29 concentrations.
30 If observed associations of ambient PM with heart disease exacerbation prove to be causal
31 and specific to PM, they would be of genuine public health concern. In the U.S. in 1997, there
October 1999 8-25 DRAFT-DO NOT QUOTE OR CITE
-------
1 were about 4,188,000 hospital discharges with heart disease as the first-listed diagnosis
2 (Lawrence and Hall, 1999). Among these, about 2,090,000 (50%) were for ischemic heart
3 disease, 756,000 (18%) for myocardial infarction or heart attack (a subcategory of ischemic heart
4 disease), 957,000 (23%) for congestive heart failure, and 635,000 (15%) for cardiac
5 dysrhythmias. Also, there were 726,974 deaths due to heart disease (Hoyert et al., 1999). Even a
6 small percentage reduction in admissions or deaths due to heart disease would predict a large
7 number of avoided cases.
8 Many investigators have also observed associations of short-term fluctuations in ambient
9 PM with daily frequency of respiratory illness. In most cases, exacerbation of pre-existing
10 respiratory illness has been assessed, though some cases of acute respiratory infection may be
11 considered as occurrence of new illness, especially in young people. Symptoms of acute
12 respiratory distress in children have been linked to elevated PM concentrations in studies in the
13 U.S. and other countries, with asthmatics apparently more susceptible than non-asthmatics.
14 However, some studies have also found associations between child respiratory symptoms or
15 reduced lung function and other pollutants (such as O3) in addition to PM, or no significant
16 relationship with air pollution. The credibility of ambient PM plausibly being linked to
17 exacerbation of pre-existing respiratory disease (e.g., asthma) is enhanced by newly reported
18 dosimetry study results noted earlier, which show greater lung deposition of 1 //m particles in
19 people with varying degrees of airway obstruction than in healthy subjects. The increased
20 deposition was greatest for COPD patients and asthmatics, but smokers also showed increased
21 deposition as well.
22 In the United States in 1997, there were 3,475,000 hospital discharges for respiratory
23 diseases; 38% for pneumonia, 14% for asthma, 13% for chronic bronchitis, 8% for acute
24 bronchitis, and the remainder not specified (Lawrence and Hall, 1999). Of the 195,943 deaths
25 due to respiratory diseases, 44% were due to acute infections, 10% for emphysema and
26 bronchitis, 2.8% for asthma, and 42% for unspecified COPD (Hoyert et al., 1999).
27 A small number of recent studies have identified young infants as an additional subgroup
28 potentially at risk from PM exposures. Effects may include intrauterine growth reduction or low
29 birth weight, known to be infant health risk factors, as well as excess infant mortality. Some
30 studies have found that PM is not as good a predictor of these endpoints as other pollutants, such
31 as CO for example.
October 1999 8-26 DRAFT-DO NOT QUOTE OR CITE
-------
1 The epidemiologic evidence has expanded greatly in quantity, increasing the number and
2 quality of both reported positive findings and some null or negative findings. The most
3 important additions to the database assessed in the 1996 PM AQCD (as evaluated in Chapter 6 of
4 this document) are:
5 - More studies of health endpoints using ambient PM10 and closely related mass
6 concentration indices (e.g., PM13 and PM7), which lessen the need to rely on non-gravimetric
7 indices (e.g., BS or CoH);
8 - New studies on a variety of endpoints for which information on the ambient coarse PM
9 fraction (PM(10_2 5)), the ambient fine particle fraction (PM2 5), and even ambient ultrafme particle
10 mass concentrations (PM0 x and smaller) were observed or estimated from site-specific
11 calibrations, with somewhat mixed results. Also, a few new studies in which the relationship of
12 some health endpoints to ambient particle number concentrations were evaluated;
13 - Many new studies which evaluated the sensitivity of estimated PM effects to the
14 inclusion of gaseous co-pollutants in the model;
15 - Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures
16 including PM components, based on empirical combinations (e.g., factor analysis) or source
17 profiles;
18 - New studies of infants and children as a potentially susceptible population;
19 - New studies of cardiovascular endpoints with particular emphasis on assessment of
20 cardiovascular risk factors as well as symptoms;
21 - Studies on asthma and other respiratory conditions exacerbated by PM exposure.
22 The new studies are discussed below from the point of view of the aspects (1-5) listed in the first
23 paragraph of this section.
24
25 8.4.2 Strengths and Limitations of Newly Available Daily Time-Series Studies
26 The new studies show a somewhat greater diversity of findings than in U.S. Environmental
27 Protection Agency (1996), shifting attention to the possibility of a multiplicity of PM health
28 effects (and perhaps some non-effects) being associated with PM of different size ranges or
29 chemical composition in a variety of different environmental contexts. While a number of daily
30 time-series studies continue to show significant excess mortality or hospital admissions
31 associated with PM10 or PM2 5 exposure, some other credible new studies show smaller
October 1999 8-27 DRAFT-DO NOT QUOTE OR CITE
-------
1 non-significant (but positive) associations with ambient PM, whereas a few others report some
2 negative associations.
3 These differences are often found in multiple-city studies in which the investigators used
4 the same analytical strategies, and models adjusted for the same or similar co-pollutants and
5 meteorological conditions, raising the possibility of different findings being attributable to
6 statistical variability. Examples of newly reported multiple-city studies include the APHEA
7 mortality studies in several European cities (Katsouyanni et al., 1997; Zmirou et al., 1998),
8 mortality in the three Wasatch Front SMSAs in Utah (Pope et al., 1999), mortality in Seoul and
9 Ulsan, Korea (Lee et al., 1999), and hospital admissions in eight U.S. counties (Schwartz, 1999).
10 Findings of some apparent different effects between the adjacent "Twin Cities" of Minneapolis
11 and St. Paul, and between the nearby cities of Seattle and Tacoma (Schwartz et al., 1999) suggest
12 that some components of variability may not be adequately explained by the statistical models, or
13 that differences in adverse health effects attributable to PM may exist at a sub-regional scale.
14 Differences between findings for Western European cities (with relatively high and statistically
15 significant mortality), and for Central-Eastern European cities (with little indication of excess
16 mortality) are most likely attributable to differences in the PM index used and/or differences in
17 implementation of the analytical strategy, although statistical variability cannot be entirely ruled
18 out. The large 20-city and 100-city NMMAPS studies of mortality in U.S. cities now in progress
19 are expected to provide further useful information about (a) the relative importance of PM in the
20 presence of varying ambient co-pollutants and (b) between-region and within-region differences.
21
22 8.4.3 Combining Results from Hospital Admissions Studies
23 Questions about the proper way to combine information from diverse studies need to be
24 considered. It is clear that combined analyses of independent studies at different locations are the
25 most appropriate for meta-analyses, especially when carried out by the same team of researchers,
26 using common methods for data collection and analysis. Independent studies from different
27 investigators, reporting the same endpoints for the same exposure metrics, are next most useful.
28 Many studies for hospital admissions data are discussed in Section 6.2.3. Table 6-19 shows
29 21 studies of hospital admissions at various locations related to PM10. Some studies were
30 adjusted for two or more co-pollutants, with results shown for PM10 RR with O3 as a
31 co-pollutant, whereas other studies were not adjusted for O3. Some studies in the same city
October 1999 8-28 DRAFT-DO NOT QUOTE OR CITE
-------
1 report two or more different endpoints, some studies report results for two or more cities using
2 the same or similar methods, and sometimes the same city is analyzed by different investigators,
3 e.g., all-ages respiratory disease in Toronto in Thurston et al. (1994) and Burnett et al. (1997a,b).
4 Consequently, the combined analyses in Table 6-21 involve a smaller number of cities within the
5 same respiratory hospital admissions category and age sub-group: all-age respiratory admissions
6 and asthma admissions, elderly admissions for all respiratory admissions, pneumonia, and
7 COPD, with smaller numbers yet for the multi-pollutant models.
8 Both the single-pollutant models and two-pollutant combined analyses show positive RR
9 associated with PM10, statistically significant in four of the five endpoints. In two-pollutant
10 models with PM10 and O3, PM10 RR ranged from 1.02 (not significant) for all-ages asthma
11 admissions to 1.12 for all respiratory admissions and 1.14 for elderly COPD. Tables 6-20 and
12 6-22 show similar findings for SO42 as the PM metric, with fewer endpoints. The SO42 results
13 are positive and significant for all-ages respiratory admissions in single-pollutant and
14 two-pollutant models.
15 New PM10 studies emphasizing cardiovascular outcomes are also described, but are not
16 integrated with the findings for respiratory hospital admissions. A quantitative synthesis across
17 eight U.S. counties reported by Schwartz (1999) shows a statistically significant relative risk of
18 2.48 per 25 //g/m3 (6.15 //g/m3 per 50 //g/m3 PM10) for hospital admissions for heart disease, in a
19 model that also includes CO as a co-pollutant; the CO effect was also positive and significant.
20 Schwartz's U.S. meta-analyses for cardiovascular disease admissions do not include results from
21 Canada (Burnett et al., 1997a,b; 1999) or Morris and Naumova's (1998) Chicago findings. The
22 results are quantitatively consistent with elevated risk of hospital admission for heart disease,
23 especially in elderly populations exposed to PM10.
24 Meta-analyses for mortality studies have been reported for the APHEA studies in Europe
25 (Katsouyanni et al. 1997; Zmirou et al., 1998). No recent meta-analyses for mortality have been
26 reported for U.S. cities. The Zmirou et al. (1998) study tends to support findings of significant
27 excess cardiovascular and respiratory mortality associated with PM (as BS) exposure in western
28 European cities, but not in Central-Eastern European cities. These results, and earlier single-city
29 results of daily-time series studies, also find some evidence for effects on total mortality and
30 cause-specific mortality in many (but not all) U.S. cities, and are consistent with the findings of
31 the combined hospital admissions studies.
October 1999 8-29 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.4 Strengths and Limitations of Prospective Cohort Studies
2 The estimation of effects associated with long-term exposure to PM was discussed in some
3 detail in the previous PM AQCD (U.S. Environmental Protection Agency, 1996). The Harvard
4 Six Cities Study (Dockery et al., 1993), the American Cancer Society data base (ACS) Study
5 (Pope et al., 1995), and the Adventist Health Study of Smog in California (AHSMOG (Abbey
6 et al., 1991, 1995) all used the prospective cohort study design. In this design, a cohort of
7 individuals is recruited at the beginning of the study and followed for a long period. Individual
8 subject data are collected at the beginning of the study, and may or may not be updated during
9 followups. Whenever the health outcome of interest occurs, its time of occurrence is recorded.
10 Data analysis compares time to occurrence of the outcome in subjects with different levels of air
11 pollution exposure. These studies are not always designed to be representative of a certain
12 population; they may depend on recruiting volunteers, some of whom may live in the same
13 household or may choose to participate because their friends and relatives participate. Greater
14 randomization may be possible within focused sub-populations, as in AHSMOG.
15 The shortcomings of this study design in available air pollution studies were discussed in
16 U.S. Environmental Protection Agency (1996). The most important are: (1) community-level
17 concentrations measured at SAM; (2) exposure metrics in the Six Cities and ACS Studies were
18 limited to long-term averages; (3) adjustments for co-pollutants were often not made;
19 (4) important personal covariates may have been omitted; and (5) important personal covariates
20 may have changed during the course of the study. The Six Cities and ACS Studies found rather
21 similar effects of PM2 5 or SO42 between the most and least polluted cities in the study, for both
22 sexes, including substantially larger effects of PM on total and cardiopulmonary mortality than in
23 the daily mortality studies, and suggested an elevated lung cancer risk. New analyses of the
24 Harvard Six Cities and ACS Studies, sponsored by the Health Effects Institute, are expected to
25 evaluate the sensitivity of the findings to alternate model specifications, inclusion of other
26 personal risk factors, and inclusion of co-pollutants when feasible.
27 New results from the AHSMOG study (Beeson et al., 1998; Abbey et al., 1999) are
28 somewhat different, as discussed in section 6.3.3. No excesses were found for long-term PM
29 exposure (nor almost any other pollutant) for females for almost all mortality endpoints. There
30 was a very strong relationship between PM10 and mortality in males for two causes of death:
31 (1) death from any nonmalignant contributing respiratory cause in the death certificate (as
October 1999 8-30 DRAFT-DO NOT QUOTE OR CITE
-------
1
2
3
4
5
6
7
8
9
10
opposed to the first-listed cause in the other
prospective cohort studies);
Males also had a significantly increased incidence of diagnosed lung
The results are summarized in the following Tables, 8-5 to 8-7.
relative risks of total mortality,
status, for the three prospective
never-smokers, although some
cardiopulmonary mortality, and lung
cohort studies.
past smokers
The
maybe
subject to mis-classification in the other studies
and
cancer,
These
cancer,
AHSMOG subjects
are
(2) lung cancer.
but females did
not.
data show the
by sex and smoking
included with
misclassified. Smoking status is also
. The statistically most significant
the ACS Study, due to the much larger sample sizes
than in the other
results are in
two studies.
TABLE 8-5. RELATIVE RISK (RR) OF TOTAL MORTALITY IN THREE
PROSPECTIVE COHORT STUDIES, BY SEX AND SMOKING STATUS
SEX SMOKING STATUS
F NON-SMOKER
STUDY
Six Cities
ACS
PM INDEX PMINC.1
PM
PM
10
2.5
S04
PAST
PAST + CURRENT
AHSMOG
Six Cities
ACS
PM
PM
PM
10
10
2.5
S04
CURRENT
M NON-SMOKER
Six Cities
Six Cities
ACS
PM
PM
PM
10
10
2.5
S04
PAST
PAST + CURRENT
AHSMOG
Six Cities
ACS
PM
PM
PM
10
10
2.5
S04
CURRENT
Six Cities
PM
10
50
25
15
50
50
25
15
50
50
25
15
50
50
25
15
50
1
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
RR
.280
.215
.147
.879
.999
.102
.104
.442
.568
.245
.104
.242
.611
.164
.104
.858
RR
0
1
1
LCL RR UCL
.704
.020
.045
0.713
0
0
.704
.898
0.977
0.719
0
1
.674
.000
0.977
0.955
0
1
1
1
.930
.051
.037
.090
2.
1.
1.
1.
5.
1.
1.
o
J.
3.
1.
1.
.345
.440
.261
.085
.632
.338
.240
.166
.678
.554
.247
1.616
2.
1.
1.
o
J.
.825
.297
.176
.166
Sources: Dockery et al. (1993); Pope et al
'PM increment in //g/m3
(1995); Abbey etal. (1999).
October 1999
8-31
DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-6. RELATIVE RISK (RR) OF CARDIOPULMONARY MORTALITY IN
THREE PROSPECTIVE COHORT STUDIES, BY SEX AND SMOKING STATUS
SEX SMOKING STATUS STUDY
F NON-SMOKERS ACS
AHSMOG
AHSMOG - CRC
PAST + CURRENT ACS
M NON-SMOKERS ACS
AHSMOG
AHSMOG - CRC
PAST + CURRENT ACS
F+M ALL Six Cities
PM INDEX
PM25
S04
PM10
PM10
PM25
S04
PM25
S04
PM10
PM10
PM25
S04
PM10
PM INC.
25
15
50
50
25
15
25
15
50
50
25
15
50
RR
1.585
1.316
0.841
1.219
1.276
1.219
1.245
1.205
1.219
1.537
1.235
1.126
1.744
RRLCL
1.235
1.147
0.639
0.739
0.918
1.008
0.929
1.023
0.862
0.879
1.061
1.037
1.202
RRUCL
2.039
1.518
1.107
2.011
1.760
1.465
1.668
1.412
1.616
2.688
1.440
1.233
2.501
Sources: Dockery et al. (1993); Pope et al. (1995); Abbey et al. (1999).
1 The only significant Six Cities finding in Table 8-5 is for male current smokers; the
2 AHSMOG subjects have similar RR for total mortality as the ACS and Six Cities Studies.
3 A similar finding in shown in Table 8-6 for cardiopulmonary mortality (no Harvard Six Cities
4 data available by sex), with ACS results significant for non-smokers and smokers in both sexes,
5 and AHSMOG males similar to ACS non-smoking males (though not significant). Lung cancer
6 mortality is not significant for females in any study in Table 8-7, but male lung cancer is highly
7 significant for AHSMOG non-smokers and for ACS smokers in the 151-city study (SO4).
8 All of these long-term studies report many statistically significant findings associated with
9 long-term mean PM concentrations. The AHSMOG study also reports more significant and
10 larger relationships with a different PM metric, the mean number of days in the a year in which
11 PM10 levels exceed 100 //g/m3 (or some other level). Due to the small number of prospective
12 cohort studies, the differences in findings by sex and smoking status, and the "patchy" pattern of
October 1999 8-32 DRAFT-DO NOT QUOTE OR CITE
-------
TABLE 8-7. RELATIVE RISK (RR) OF LUNG CANCER MORTALITY IN THREE
PROSPECTIVE COHORT STUDIES, BY SEX AND SMOKING STATUS
SEX SMOKING STATUS STUDY
F NON-SMOKERS ACS
AHSMOG
PAST + CURRENT ACS
M NON-SMOKERS ACS
AHSMOG
PAST + CURRENT ACS
F+M ALL Six Cities
ACS
PM INDEX
PM25
SO4
PM10
PM25
S04
PM25
S04
PM10
PM25
SO4
PM10
PM25
S04
PM INC.
25
15
50
25
15
25
15
50
25
15
50
25
15
RR
0.644
1.432
1.808
0.949
1.074
0.483
1.261
12.385
1.123
1.316
1.744
1.031
1.261
RRLCL
0.203
0.731
0.343
0.563
0.781
0.086
0.501
2.552
0.827
1.104
0.689
0.796
1.082
RRUCL
2.091
2.800
9.519
1.595
1.479
2.714
3.190
60.107
1.533
1.577
4.390
1.338
1.465
Sources: Dockery et al. (1993); Pope et al. (1995); Abbey et al. (1999).
1
2
3
4
5
9
10
11
12
comparisons shown in Tables 8-5 to 8-7, a quantitative meta-analysis does not seem to be
feasible at this time. Reanalyses of the Harvard Six Cities and ACS Studies by HEI may report
the results in a form that facilitates comparison with the AHSMOG findings.
8.4.5 Evaluating the Coherence of the New Studies
Aspects of coherence (Bates, 1992) were discussed in detail by U.S. Environment
Protection Agency (1996). Determination of coherence involves assessment of the entire body of
epidemiology studies, as well as supporting medical and toxicological data, for consistency
across a variety of health outcomes by repeated observation in different populations of
individuals, under different circumstances of duration and level of ambient PM concentration,
and in different places. The adverse health effects associated with PM are: (1) lung function
decrements; (2) respiratory symptoms, or exacerbation of symptoms requiring bronchodilator
October 1999
8-33
DRAFT-DO NOT QUOTE OR CITE
-------
1 therapy; (3) hospital admissions for respiratory and cardiovascular causes; (4) emergency medical
2 visits; and (5) death largely from cardiopulmonary causes in the elderly. None of the currently
3 available time series studies are based on a temporal sequence of these outcomes in single
4 individuals. Panel studies of respiratory symptoms have assessed the repeated occurrence of
5 symptoms in individuals, but not at the progression of, for example, repeated respiratory
6 symptoms into hospital admissions, or repeated hospital admissions into cardiopulmonary
7 mortality. Indeed, the extent to which this progression occurs in the population is uncertain.
8 While some studies are currently underway that will examine large public health data bases, no
9 preliminary results have been published. It is therefore necessary to look at indirect indicators of
10 the quantitative consistency at a group level.
11 For example, from asthmatic panel studies, it is apparent that mild asthmatics selected for
12 study are seldom hospitalized; their asthma symptoms can largely be controlled by medication.
13 They may be sufficiently mild that changes in pulmonary function or occurrence of symptoms
14 can be detected and self-treated. Although less likely to be selected to participate in panel
15 studies, moderate to severe asthmatics are more likely to go to a hospital or emergency
16 department for asthmatic episodes. There is no clear evidence that ambient PM concentrations
17 are associated with a progression from mild to severe asthmatic symptoms as studied; mild and
18 severe asthmatics appear to be distinct sub-groups.
19 It is by no means self-evident that the numbers of events on some appropriate baseline of
20 time and reference group will follow a sequence of: lung function decrements > respiratory
21 symptoms > ER visits > hospital admissions > death. There are many causes other than PM for
22 each of these endpoints. For example, lung function decreases with age and chronic respiratory
23 illness and is affected by cigarette smoking and exposure to occupational air pollution.
24 A number of studies have found an association of mortality with reduced lung function
25 (Strachan, 1992; Higgins and Keller, 1970). Strachan (1992) notes that little is known about the
26 constitutional or environmental determinants of lung function decline. Longitudinal studies such
27 as the Harvard Six Cities Study might possibly be used to evaluate a hypothetical causal
28 pathway. Ambient PM concentration —> PM exposure —> FEVj decrease —> death in
29 individuals. The decline of FEVj may be either a precursor of ambient-PM-induced health
30 effects, or an independent factor in susceptibility to air pollution effects leading to hospital
31 respiratory admissions or mortality. The relationship of declines of lung function with death is
October 1999 8-34 DRAFT-DO NOT QUOTE OR CITE
-------
1 that a decline in FVC or FEVl3 is prognostic mainly in the later stages of CFIF or COPD. This
2 does not necessarily imply that modest declines in healthy people will be prognostic of life
3 shortening.
4 The relationship of hospital admissions and mortality in independent studies has been
5 studied. Hospital admissions may be affected by health status as well as by environmental
6 factors. For an individual, the relationship of prior hospital admissions to mortality is uncertain.
7 In general non-environmental studies (Seneff et al., 1995), hospitalization in the preceding few
8 months or year is a good predictor of subsequent hospital admissions or death. The major risk
9 factor for subsequent death was the development and severity of non-respiratory organ system
10 dysfunction. Medical intervention is typically provided to the most seriously ill individuals, but
11 if these interventions reduce the likelihood of death associated with elevated ambient PM
12 concentration and exposure, then it is possible that many of the deaths attributed to ambient PM
13 may occur in a less frequently hospitalized population. Consistency would be suggested if there
14 were more cause-specific hospital admissions than deaths from respiratory or cardiovascular
15 causes as described in U.S. Environmental Protection Agency (1996).
16 The results of peak flow analyses in acute asthma studies consistently show small
17 decrements for both PM10 and PM25. This was observed for both morning (AM) and afternoon
18 (PM) peak flow. Most studies showed increases in cough, phlegm, difficulty breathing, and
19 bronchodilator use, although these increases were generally not statistically significant. The
20 results for nonasthmatic groups of the acute peak flow analyses consistently show small
21 decrements for increases in PM10, similar to those found for asthmatics. The results of the
22 chronic morbidity studies are not consistent. Some studies show effects for some endpoints, but
23 other studies fail to find the same effects. It is generally more difficult to find a gradient in long
24 term exposures than for short term exposure studies. For this reason, it is not surprising that the
25 long-term studies show less consistency than the acute studies.
26 The results of recent studies of the association of PM mass with hospital admission are
27 generally consistent with and supportive of the studies presented in U.S. Environmental
28 Protection Agency (1996). Moreover, mathematical syntheses of multiple hospital admissions
29 studies for the various age and disease categories were conducted as part of this current
30 assessment. Statistically significant and reasonably consistent RR effect sizes (i.e., within their
October 1999 8-35 DRAFT-DO NOT QUOTE OR CITE
-------
1 respective confidence intervals) were generally found across admissions categories for both PM10
2 and SO42.
3 A review of the studies summarized in U.S. Environmental Protection Agency (1996)
4 indicated that previously-available epidemiologic studies of chronic PM exposures collectively
5 indicated increases in mortality to be associated with long-term exposure to airborne particles of
6 ambient origin. The PM effect size estimates for total mortality from these studies also
7 suggested that a substantial portion of these deaths reflected cumulative PM impacts above and
8 beyond those exerted by acute exposure events.
9 The new AHSMOG study (Abbey et al., 1999) provides all-cause mortality RR estimates
10 for adult males that are quantitatively and qualitatively consistent with prior semi-individual
11 studies, especially the Six Cities Study. Extensive new by-gender, by-cause, and multiple
12 pollutant sensitivity analyses, as well as a more comprehensive analyses of numerous potentially
13 uncontrolled factors in this study (such as of the effects of variations in the time spent outdoors)
14 provide important new evidence that is largely supportive of the association of mortality with PM
15 of ambient origin previously reported by the Six Cities and ACS Studies.
16 Published cross-sectional studies collectively indicate that older adults and infants are the
17 age groups most affected by ambient PM, while both the cross-sectional and semi-individual
18 studies indicate that deaths involving respiratory disease (either malignant or not) are most
19 associated with exposure to PM air pollution. These results are biologically plausible and
20 consistent with a causal relationship between mortality and exposure to PM of ambient origins.
21 With regard to the role of various PM constituents in the PM-mortality association,
22 cross-sectional studies have generally found that the fine particle component, as indicated either
23 by PM2 5 or sulfates, was the PM constituent most consistently associated with mortality.
24 In addition, the Six-cities prospective semi-individual study also indicates that the fine mass
25 components of PM are more strongly associated with the mortality effects of PM than the coarse
26 PM components.
27 Recent investigations of the public health implications of these PM mortality effect
28 estimates were also reviewed. Life table calculations by Brunekreef (1997) found that relatively
29 small differences in long-term exposure to airborne particulate matter of ambient origin can have
30 substantial effects on life expectancy. For example, a calculation from the 1969-71 life table for
October 1999 8-36 DRAFT-DO NOT QUOTE OR CITE
-------
1 U.S. white males indicated that a chronic exposure increase of 10 //g/m3 PM was associated with
2 a reduction of 1.31 years for the entire population's life expectancy at age 25.
3 Deaths due to respiratory causes may not be consistently observed over different time
4 scales for ambient PM exposures. For example, PM exposures on a short time scale are reported
5 as being associated with excess cardiopulmonary deaths and respiratory hospital admissions.
6 However, it is not clear that such deaths or hospital admissions in association with high ambient
7 PM concentrations represent a long-term difference in the mortality rate, or a displacement of
8 events that would have occurred a few days later without the high PM levels. The prospective
9 cohort mortality studies do not yet allow assessment as to whether the occurrence and frequency
10 of deaths from short-term exposures are consistent with a higher rate of occurrence of deaths,
11 associated with longer-term exposures, on a scale of months or years.
12 There is as yet little basis (conceptual, experimental, or mathematical) that would allow a
13 quantitative linkage of endpoints between mortality attributable to short ambient PM exposures
14 ("acute") and mortality attributable to longer ambient PM exposures ("chronic" or
15 "sub-chronic"). Exposure indices are more easily compared at different time scales when
16 endpoints are identified (Evans et al., 1984). One hypothesis for discussion is that an individual
17 with high susceptibility to ambient PM at a given moment (e.g., elderly, with an acute respiratory
18 infection as well as COPD or other serious pre-existing conditions) may succumb to a moderately
19 elevated ambient PM exposure or to some coincident cause, although that individual may have
20 survived the same PM exposure if it had occurred during a time of lesser susceptibility to
21 ambient PM.
22 U.S. Environmental Protection Agency (1996) reported that the relative risk for excess
23 daily mortality associated with a large PM increment (50 //g/m3 or 25 //g/m3) was only about
24 1.05, whereas the excess risk associated with 20 //g/m3 PM2 5 between Portage and Steubenville
25 was 1.26 in the Six Cities Study, and a comparable PM2 5 increment found a mortality RR of
26 1.17 in (Pope et al., 1995) for 50 U.S. cities. This is consistent with the hypothesis that there is a
27 risk from long-term exposure to PM that is substantially larger than just the accumulation of
28 mortality due to PM episodes involving short-term PM exposures.
29
30
31
October 1999 8-37 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.6 Evaluating the Plausibility of Inferences about the Relationships
2 Between Human Health and Ambient PM Concentrations
3 Chapter 5 reviewed the PM exposure literature and the latest advances in our understanding
4 of the relationships between human exposure to PM of ambient origin and ambient PM
5 concentrations (Mage et al., 1999). Until recently, there was some confusion in the air pollution
6 community between the correlation of personal exposure to total PM (PM of ambient origin plus
7 PM of non-ambient origin) and ambient PM concentrations, and correlations of personal
8 exposure to PM of ambient origin and ambient PM concentrations (e.g., Gamble, 1998). PM
9 generated from sources that influence ambient PM concentrations (such as traffic, industry and
10 photochemical reactions) is distinctly different in chemical and toxicological character from PM
11 generated from indoor sources, such as smoking and human-generated detritus (Siegmann et al.,
12 1999). Consequently, these categories of PM need to be treated separately, as they represent
13 distinctly different classes of pollutants.
14 Chapter 5 shows that the cross-sectional correlation of a group of people's personal total
15 PM exposures and ambient PM concentrations of same AD size is often of order zero because of
16 the highly variable large amounts of non-ambient PM generated by personal activities and indoor
17 sources independently of the ambient PM concentration. However, the longitudinal correlation
18 between a given person's total-exposure to PM and concentration of ambient PM of the same
19 AD size is often of order 0.7 to 0.9. Because people tend to visit the same residential and
20 occupational environments from day to day, where sources of PM and air exchange rates usually
21 have a small variability, the higher individual longitudinal correlations with total PM exposure
22 reflect the correlation of personal exposure to PM of ambient origin with ambient PM
23 concentrations.
24 Therefore the correlation of human health effects with ambient PM concentrations is a
25 reasonable basis for inferring (a) that there is either an actual relationship between ambient PM
26 concentrations and health effects or (b) that the statistical relationship is confounded by some
27 other gaseous pollutants or ambient PM constituents highly correlated with the total mass of
28 ambient PM.
29
30
31
October 1999 8-38 DRAFT-DO NOT QUOTE OR CITE
-------
1 8.4.7 Assessing the Extent To Which Adverse Health Effects Are Attributable
2 to PM Size Fractions or Components, or Other Environmental Factors
3 8.4.7.1 Introduction
4 Recent epidemiology studies have greatly extended the data base from which inferences
5 may be drawn about the health effects of PM size fractions, chemical components, or source
6 categories. The size categories of most current interest (PM10, PM25, PM(10.25)) are now included
7 in new epidemiologic studies of ultrafine particles (Peters et al., 1997; Norris et al., 1999) and in
8 new studies of morbidity and mortality specifically addressing PM25 and PM(10.25), discussed in
9 Section 6.4.6. Effects associated with PM2 5 have been found in some new studies, but not all.
10 The effects of aerosol acidity and sulfate concentration were extensively reviewed in U.S.
11 Environmental Protection Agency (1996) and are assessed again here in Chapter 6 of this
12 document. In general, H+ and SO;2 are less consistently reported to be significantly associated
13 with adverse health effects than are PM25 and PM10, but positive findings occur sufficiently
14 frequently that a contributing role of these species in causing or promoting health effects
15 (possibly in the presence of other pollutants or factors) cannot be eliminated. While
16 toxicological studies of the effects of acids, transition metals, and other specific chemical
17 components of PM (alone or in combination) on animals suggest that adverse effects in humans
18 may also occur, there is as yet little epidemiologic evidence to confirm the predicted effects.
19 A number of toxicological studies have been carried out with complex materials that
20 simulate specific sources or components of airborne particles, including residual oil fly ash
21 (ROFA) and diesel exhaust emissions (DE), as well as studies with concentrated particles
22 sampled directly from urban air. These studies show numerous physiological and biochemical
23 responses in laboratory animals, particularly in animals with natural or artificially induced
24 pre-existing disease. Only very limited data on concentrated ambient particles and some
25 constituents are available for humans.
26 Finally, much recent interest has focused on aspects of urban air pollution that can be
27 constructed or derived from multiple-element or multiple-component data. One approach is to
28 construct "source profiles" analogous to those used in various source apportionment studies, and
29 then to estimate the contribution of each source (identified by elemental tracers, for example)
30 to the airborne aerosol mass. Laden et al. (1999) has used this approach to conclude that particle
31 components of "crustal" origin (presumably rich in Al and Si) appear to have no effect on excess
October 1999 8-39 DRAFT-DO NOT QUOTE OR CITE
-------
1 mortality attributable to PM2 5 in the Harvard Six Cities Study. A second approach is to construct
2 factors that are combinations of air pollutants and meteorological variables, and to use these
3 factors as predictors of adverse health effects. In a study in Toronto, Ozkaynak et al. (1996)
4 found that some factors predictive of excess mortality involved combinations of air pollutants
5 (including TSP and gaseous co-pollutants), whereas others were almost "pure" factors involving
6 only a single component such as ozone. In particular, weather effects on mortality were largely
7 separable from air pollution effects in Toronto. This tends to confirm the findings of other
8 investigators that effects of weather on health may be large, but are also largely separable from
9 those of air pollution.
10 It is likely that U.S. cities have important similarities and differences in their source
11 profiles for PM. The "crustal" particles are likely relatively higher in the western states than in
12 the east, and in areas where agriculture or construction produce significant soil turnover.
13 Particles from mobile source internal combustion engines are widespread, although related PM
14 levels may vary substantially from city to city and season to season. In some cities, a measurable
15 contribution to airborne particles may also arise from use of wood or oil for home heating, which
16 may be nearly absent in other cities. In the absence of specific information, it would reasonable
17 to assume that the human health hazards associated with these components may vary from time
18 to time and place to place, so that a single PM mass concentration may imply different levels of
19 potential hazard to human health. Risks associated with a specific PM component within a
20 specific exposure scenario, for example, "heating oil combustion particles during winter air
21 stagnation with temperatures less than 0°C" may be comparable. However, it is unlikely that the
22 sum of the PM components in different exposure scenarios would all consistently add up to an
23 equivalent health risk at equivalent PM mass concentrations within a specific size range. This
24 argument is sufficiently important to warrant evaluation, in some detail, of the evidence for
25 differential toxicity for so-called "crustal" particles.
26
27 8.4.7.2 Epidemiology Evidence Suggesting that Crustal Particles Are Less Clearly Harmful
28 to Human Health
29 Three informative assessments published since 1996 deal explicitly with health effects of
30 PM10 in the western U.S. states. PM10 is sometimes dominated by particles of crustal or
31 geological origin that are believed to have been transported by wind from rural areas into the
October 1999 8-40 DRAFT-DO NOT QUOTE OR CITE
-------
1 metropolitan areas of Spokane, WA, (Schwartz, 1999) and the Utah MS As of Ogden, Salt Lake
2 City, and Provo/Orem (Pope et al., 1999). PM10 in the Coachella Valley of California is also
3 dominated by coarse particles. The very high PM10 concentrations on certain days were
4 presumably not dominated by the typical urban sources of PM2 5 and coarse particles.
5 Concentrations of urban source particle mixtures tend to be highest on days with low wind speed,
6 not high wind speed. None of these epidemiology studies used PM(10_25) as an index.
7 The identification of these particles as "crustal" is an indirect inference. Studies reviewed
8 in the 1996 PM AQCD (Gordian et al., 1996; Hefflin et al., 1994) also did not rely on specific
9 identification of coarse crustal particles. The assumption is that most PM25 particles are
10 produced by combustion sources, and that high concentrations of PM25 particles occur when
11 locally-produced particles are 'trapped' in an urban airshed by stagnant air (see Chapter 4).
12 During periods of high wind, elevated concentrations of PM10 will often be associated with
13 non-local particle sources, such as dusts from agriculture or mineral extraction processes,
14 construction, forest fires, and transport of sand or natural dusts. This can be indirectly assessed
15 by determining whether elevated PM10 concentrations occur at the same time as below-average
16 concentrations of PMl 0 or of gaseous pollutants that are typically produced by urban combustion
17 sources.
18
19 Spokane Study (Schwartz et al, 1999)
20 There was only limited information about PM2 5 particle (FP) concentrations in the Spokane
21 study (Schwartz et al., 1999). Measured PMX 0 concentrations were typically less than 10 //g/m3
22 even when PM10 concentrations were > 150 //g/m3 during dust storm episodes. Fluctuations in
23 PM25 tended to be proportional to PM10 from 12:00 to 24:00 during the high-wind episode.
24 There were also large and proportional fluctuations in PM(10_2 5) and PM(2 54 0) during the episode
25 while PMX 0 concentrations were reduced and stable. It is likely then that much of the PM25 was
26 also of crustal origin, especially in the "large fine particle" or "inter-modal" (PM^^ 0)) fraction
27 in Spokane.
28 Conversely, dust episodes may lead primarily to increases in the PM(10_2 5) fraction, with
29 little change in PM2 5 or PMX 0 based on size-fractionated PM measurements near Fort Ord,
30 California. The large increase in PM10 occurs almost completely in the PM(10_2 5) fraction, leaving
31 PM2 5 and PMX 0 fractions little changed. As shown in Figure 8-4, the lower "hump" represents
October 1999 8-41 DRAFT-DO NOT QUOTE OR CITE
-------
CO
,0
CO
E
60
50-
40-
^30
Q
O)
o
20-
10-
.01
HUNTER-LIGGETT
9-14-72
Q13:10
A 13:20
'I I I I I | I I I I I I II |
.1 1
Particle Diameter, |jm
iii i i i ii
\ i r
10
Figure 8-4. Fort Ord CA dust episode: fine and coarse particles.
1 fine "accumulation mode" particles, typically less than 1 //m. The upper hump represents
2 "coarse-mode" particles. The area between 1.0 and 2.5 //m includes a mixture of "accumulation
3 mode" and "coarse mode" particles. In these figures, PM(2 54 0) is dominated by small
4 coarse-mode particles.
5 In typical urban environments, most of the mass in PM25 is in the lower accumulation
6 mode hump, but this may be reversed during high wind episodes. The 24-hour particle size
7 distribution for high-wind episodes may be better represented by that shown in Figure 8-5.
8 In Figure 8-5, the coarse mode particles are the predominant component of PM2 5 and PM(2 54 0),
9 but a very minor component of PMj 0. Under these conditions,!! is possible that a high
10 concentration of PM25 of largely crustal origin would not be predictive of toxicity on the
11 assumption that the PMl 0 fraction is more toxic. If the PMl 0 concentration (primarily from
October 1999
8-42
DRAFT-DO NOT QUOTE OR CITE
-------
400
350 -
300 -
250 -
=g 200 H
CO
CO
o
O
150 -
100 -
50 -
High Wind Episode
-Coarse Fraction, PM10.25
-PMi
-Intermodal Fraction, PM25-i
50
- 45
- 40 "E
- 35 -
Q.
- 30 %
• 25 °~.
o
20 ro
12:00 AM 3:00 AM 6:00 AM 9:00 AM 12:00 PM 3:00 PM 6:00 PM 9:00 PM
Local Time
- 15
• 10
5
0
E
cu
Figure 8-5. Spokane dust storm episodes.
1 urban particles) is also very low under these circumstances, then there may be little excess
2 mortality associated with PM2 5. The findings of Schwartz (1999) support this hypothesis in that
3 they observed no excess mortality during summer dust storms.
4 Thus, there are at least some conditions under which high concentrations of crustal PM2 5
5 may not be associated with excess mortality in Spokane, and the original nature of the PM25
6 particles (both physical and chemical properties) likely determines its toxicity. The findings of
7 Laden et al. (1999) for the very low toxicity of the 'crustal component' of PM25 in the Six Cities
8 particles tends to support this interpretation.
9 The above interpretation may not apply to locations where there are substantial
10 contributions of non-crustal PM25 from non-local sources during high wind episodes. However,
11 it is unlikely that high wind episodes in the Wasatch Front region or in Spokane are associated
12 with greatly elevated PM2 5 or PMj 0 concentrations transported from other urban areas.
13
October 1999
8-43
DRAFT-DO NOT QUOTE OR CITE
-------
1 Utah Study (Pope et al, 1999)
2 Similarly, the Utah studies by Pope et al. (1999) strongly suggest that mortality may be
3 much greater in three Wasatch Front communities during periods when PM10 pollution is not
4 attributable to wind-blown dust, as characterized by elevated levels of the clearing index during
5 periods of elevated PM10. The 'clearing index' is an indicator for air movement calculated by the
6 National Weather Service for the Wasatch Front region. It characterizes vertical and horizontal
7 motion of the atmosphere, using wind speed and direction, mixing height, moisture, and
8 temperature, on a scale of 0 (no movement, stagnant air) to 1000. Under stagnant atmospheric
9 conditions, Salt Lake City PM10 was dominated by small particles, 55 to 70% in the fraction
10 <1 //m, and 70 to 90% in the fraction <2.5 //m during the winter of 1995-1996. Occurrences of
11 high PM10 in Salt Lake City were common during days of high clearing index in Salt Lake City,
12 probably due to wind-blown transport of dust from mine tailings piles, gravel pits, cement plants,
13 salt flats, agricultural fields etc (see Figure 8-6). Such episodes were rare in Ogden and Provo.
14 It should be noted that elevated PM10 concentrations were significantly associated with
15 excess mortality in all three Wasatch Front cities during periods in which crustal particles were
16 unlikely to have dominated the PM concentrations. This suggests that the usual combination of
17 PM from mobile and stationary sources and from secondary aerosols in these SMS As constituted
18 the major contributions to adverse health effects, but that even higher concentrations of PM10
19 from crustal particles added little additional hazard.
20
21 Coachella Valley, California, Study (Ostro et al, 1999)
22 Recent published studies by Ostro et al. (1999, in press) in the Coachella Valley of
23 California (Palm Springs and Indio) suggest that adverse health effects may be associated with
24 the coarse fraction of PM10. These studies have not been evaluated with respect to the crustal
25 versus non-crustal components of PM10 in the coarse and PM25 fractions. The mortality analyses
26 covered the years 1989-1992, when no PM25 particle monitoring was carried out. PM25 particle
27 monitoring during April 1996 - Feb. 1997 showed that PM10 in the Coachella Valley was
28 dominated by the coarse fraction (PM(10.25)) comprising 59% of PM10 mass at the Palm Springs
29 site and 64% of PM10 mass at the Indio site. The crustal fraction can exceed 90% of PM10 during
30 wind events.
31
October 1999 8-44 DRAFT-DO NOT QUOTE OR CITE
-------
>
RO -
en
3 Diameter
4i. C
0 C
1
.2 30-
<
U)
w
03
^ 20 -
10-
.0
k C<
Accumulation Mode
1
High Wind ^^~ — - x^^
i i i i i i 1 1 1 I l l 1 l l 1 1 1 I
1 .1 1
Darse Mode
fV
\\
\
\
\
\
\
\
\
i i i 1 1 1 1
1
£
Lj;_-i_ \ A /I •-» *J
High Wind
^ Normal vvinu
\
\\
— x.
i iii
0
Particle Diameter, urn
Figure 8-6. During normal wind conditions, accumulation mode particles dominate PM2 5.
During high-wind conditions, coarse mode particles may dominate PM2 5 in
some western sites, where coarse particle are predominantly of crustal origin.
1
2
3
4
5
6
1
Statistically significant associations of total mortality, cardiovascular and respiratory
mortality, and mortality in age 50+ populations were found for PM10 at lag 2 days, and with
3-day and 5-day moving averages, and were similar in magnitude (RR 1.04 to 1.05 per 50 //g/m3
PM10 for total mortality) to other U.S. urban areas. The PM10 effects were hardly changed when
gaseous co-pollutants were included in the models, whereas inclusion of PM10 rendered effects of
O3, NO2, and CO non-significant.
PM10 source apportionment studies in the Coachella Valley during 1989 showed that crustal
particles comprised 50 to 60 percent of the annual average PM10 mass, and much more during
'exceptional' (likely wind) events. During 'exceptional event' days, the percentage of "marine"
October 1999
8-45
DRAFT-DO NOT QUOTE OR CITE
-------
1 particles was also elevated (South Coast Air Quality Management District [SCAQMD], 1990),
2 likely reflecting contributions from the Salton Sea (especially at Indio). Windblown PM(10_2 5) and
3 PM2 5 from the Los Angeles region and the Pacific Ocean are probably not highly correlated, due
4 to distance > 160 km and intervening mountains. Thus, crustal particles in the Coachella Valley
5 may have components specific to this airshed. Coarse particles of biogenic origin contribute
6 13 to 15% of annual average PM10 in the Coachella Valley.
7 Days with highly elevated PM10 concentrations probably included days with high wind
8 speed. When days with extremely elevated PM10 concentrations (> 91 //g/m3) were removed
9 from the data set, the relative risk increased from 1.044 to 1.053 and was more significant. It is
10 therefore likely that the excess mortality was not attributable to elevated concentrations of crustal
11 material on days with elevated wind speed. This finding by Ostro et al. (1999) is thus consistent
12 with Schwartz et al. (1999) and Pope et al. (1999).
13 A complicating factor is that behavior affecting PM exposure may change substantially on
14 windy days. It is likely that many people spend more time indoors on days with blowing dust and
15 high winds, where coarse particle penetration is less than for fine PM, and their concentrations
16 may be further reduced by the use of air conditioning (especially in the Coachella Valley). The
17 resulting reduction in personal PM exposure during such episodes could contribute to the finding
18 of little ambient PM10 effect during such episodes. The extent to which the reduction in personal
19 PM exposure might offset the higher ambient PM10 concentrations is unknown.
20
21 Six Cities Study (Laden et al, 1999)
22 The recent abstract by Laden et al. (1999) also describes a method for identification of a
23 crustal component in PM2 5 using elemental composition for source apportionment. The "crustal"
24 component shows little relation to excess mortality.
25
26 Summary
27 Several hypotheses may be advanced as interpretations of these findings: (1) Crustal
28 materials in some western PM10 appear to have little influence on morbidity/mortality during
29 high wind episodes; (2) On days with high wind speed in which PM10 is dominated by crustal
30 materials, the atmospheric conditions also reduce the concentrations of anthropogenic PM25 and
31 other gaseous co-pollutants presumably associated with PM10 toxicity; (3) during non-wind
October 1999 8-46 DRAFT-DO NOT QUOTE OR CITE
-------
1 events with high concentrations of coarse particles, effects may be associated with coarse-mode
2 PM. Exposure to crustal materials in PM2 5 is not likely to be greatly reduced on windy days,
3 since a fraction of these particles readily penetrate indoors. The most plausible interpretation of
4 these findings is that either the composition of crustal PM2 5 particles or their environmental
5 co-factors contribute to the reduced mortality.
6
7 8.4.7.3 Toxicology Evidence Suggesting That Crustal Particles Are Less Harmful to
8 Human Health
9 Animal toxicology studies of "crustal" particles are mostly limited to inhalation or
10 instillation of volcanic ash particle or amorphous silica. Very high concentrations
11 (> 10,000 //g/m3) can produce transient inflammation. There are no comparable controlled
12 human exposures. Lifetime (human) occupational exposure to respirable quartz is, however,
13 associated with silicosis and occurs at average concentrations of 50 to 100 //g/m3 with a
14 prevalence of up to 20%. Nevertheless, exposure to ambient crustal particles is unlikely to have
15 adverse effects in healthy humans. It is likely that the composition of crustal particles in the
16 ambient environment is not the key factor in their relatively low apparent toxicity, but rather the
17 fact that the particles are large and therefore well filtered out in the upper respiratory tract (URT)
18 and easily cleared after their deposition, mainly in the conducting airways. Nevertheless, it
19 should be noted that 75% of PM2 5 particles retained in lung parenchyma in one autopsy study
20 were crustal in nature (Churg et al., 1997). Key components of crustal materials are silicates;
21 both ultrafme crystalline and amorphous silicates are reported to be more toxic than diesel UF
22 particles (Murphy et al., 1998) in laboratory animal inhalation studies.
23
24 8.4.8 Quantifying Relationships Between Ambient PM Concentrations and
25 Health Effects in Susceptible Subpopulations at Different Time Scales
26 Children
27 Children are often at greater risk of adverse health effects from environmental toxicants
28 than are adults, usually attributable to their greater exposure for body size, greater deposition or
29 uptake, reduced clearance and elimination, or greater susceptibility. On the other hand,
30 sometimes they may be at lower current risk, for example, from toxicants requiring a very long
31 duration of exposure to exert their effects. Some recent new studies appear to suggest that
October 1999 8-47 DRAFT-DO NOT QUOTE OR CITE
-------
1 prenatal or postnatal exposure to ambient PM may be associated with adverse effects at several
2 time scales. Other effects are associated with very short term PM or O3 exposures, 1 h or 8 h
3 (Delfino et al. 1998), as well as with 24 h exposures. Also, deposition studies tend to show
4 increased extrathoracic deposition of respirable particles in children. Deposition rate may or may
5 not be increased in children, but the higher relative ventilation of children coupled with a smaller
6 lung surface area suggests that greater deposition per unit surface area may occur in children
7 (Bennett et al., 1997b; Bennett and Zeman 1998).
8 Significant short term effects of PM on lung function and respiratory symptoms in
9 asthmatics, discussed in Sections 6.2.1.1 and 6.4.2, are frequently found in children aged 5 to
10 17 years. Peak expiratory flow reductions (PEFR) are not generally significantly associated with
11 PM10. Evening PEFR reductions were significant in one study and marginally significant in two
12 of six pediatric studies (lags of 2 to 5 days), with an average reduction of about 1 L/min per
13 50 //g/m3 PM10. Similarly, coughing, phlegm production, and difficulty breathing were elevated
14 with odds ratios on the order of 1.1-1.2 per 50 //g/m3 PM10.
15 Smaller effects have been found in non-asthmatic children (Section 6.2.1.2) with PEFR
16 reduction of about 0.4 L/min per 50 //g/m3 PM10. There was less indication of increased
17 incidence of coughing, lower respiratory illness (LRI), or URI in children, except for increased
18 coughing in the Utah Valley. Bronchodilator use and respiratory symptoms associated with
19 sulfates, ozone, or PM10 were also found in some studies, but not in others (Tables 6-17, 6-18).
20 The inclusion of co-pollutants in the models slightly reduced the magnitude of the estimated
21 effects, which remained statistically significant.
22 These studies suggest that children aged 5 to 17 years are susceptible to a variety of
23 respiratory symptoms and pulmonary function decrements associated with PM (Table 6.4.4.2).
24 Various PM indexes such as PM10, BS, or FT were significantly associated with increased contact
25 with the medical care system emergency department visits, hospital admissions, or doctor's visits
26 in about half of the studies, particularly for asthma.
27 Intrauterine or child mortality were reported to be significantly associated with short term
28 exposure to PM in four recent studies. PM2 5, PM10, or TSP were significantly related to
29 mortality in three of the five outcomes reported (Table 6.4.4.2.3). Additionally, five of seven
30 studies show long-term ambient PM indices were associated with neonatal and infant health
31 effects (Table 6.4.4.2.5), including total infant mortality, SIDS, infant mortality from respiratory
October 1999 8-48 DRAFT-DO NOT QUOTE OR CITE
-------
1 causes, low birth weight, or premature birth. At this time, however, no biological mechanisms to
2 explain how PM may cause these effects are evident.
3
4 The Elderly
5 Few of the recent time-series studies of mortality have explicitly examined age subsets,
6 particularly ages > 64-65 vs. ages < 64-65 years. Earlier studies evaluated in the 1996 PM
7 AQCD found higher mortality RR for older subjects.
8 One new study in Mexico City (Borja-Aburto et al., 1997) found RR for TSP of 1.059 for
9 age > 65 compared to 1.058 for all ages. Using PM25 as the PM indicator, Borja-Abuto et al.
10 (1999) found RR of 1.058 for age > 65, compared to RR of 1.043 for all ages, in a three-pollutant
11 model with O3 and NO2 (Table 6.4.6.11), without presenting comparisons for other age groups.
12 Loomis et al. (1999), in another study, found a RR of infant mortality from PM25 of
13 1.16 (marginally significant) in a similar three-pollutant model (Table 6.4.6.12). This suggests
14 that both the very young and the elderly may be elevated risk from ambient PM2 5 in Mexico City,
15 compared to those at intermediate ages. As for some new Western European studies, Bremner
16 et al. (1999) report effect sizes for mortality in London UK for age groups 0-64 yr, 65-74 yr,
17 > 74 yr, > 64 yr, and all ages. The respiratory and cardiovascular effects appeared to be much
18 larger for age groups 0-64 yr to 65-74 yr than 75+, using BS as an indicator, whereas only
19 respiratory deaths showed a higher elderly rate using PM10. As Bremner et al. (1999) note, "...
20 the number of deaths was three to four times greater in the older elderly group [> 74 years]; this
21 means that the attributable deaths were considerably greater in the older group. ... Young elderly
22 people, however, seem to be more susceptible to the effects of air pollution as judged by the
23 relative risks than their older counterparts who could perhaps be seen as healthy survivors."
24 Also, Prescott et al. (1998) showed BS to be related to significant excess mortality in people of
25 aged > 64 years and smaller positive excess mortality in people of age < 65 years, based on a
26 14.5 year time series study in Edinburgh UK. In the elderly group, excess mortality was
27 significant for respiratory causes and positive but not significant for cardiovascular causes.
28 Two Australian studies also found age effects. Morgan et al. (1998) found excess heart
29 disease hospital admissions in Sydney of 2.82% (significant) for ages > 64 years, but only 1.02%
30 (NS) in younger people associated with PM. Simpson et al. (1997) found slight increases in
October 1999 8-49 DRAFT-DO NOT QUOTE OR CITE
-------
1 elderly versus non-elderly mortality RR in Brisbane: 1.010 (signif.) versus 1.001 (NS) for total
2 mortality. Both studies used nephelometer measurements as the PM index.
3 Two studies of Asian populations (Ostro et al., 1998; Cropper et al., 1997) tended to see a
4 decline in the RR with increasing age, with the largest risk in young children. The relevance of
5 these studies, from a very different socioeconomic situation, to U.S. populations is dubious.
6 These findings may reflect a healthy survivor effect.
7 Combined analyses for respiratory hospital admissions, presented in Section 6.2, suggest
8 little basis for inferring different age-specific hospital respiratory admissions rates.
9 Overall the new studies in Mexico, western Europe, and Australia tend to support the
10 findings of somewhat larger associations of mortality and cardiopulmonary hospital admissions
11 with ambient PM in the elderly than in the non-elderly, excepting infants. The findings are not
12 wholly consistent, and suggest that differences between countries in socio-demographic factors
13 such as age structure, distribution of wealth, access to medical care and public health
14 interventions, as well as climate, may be responsible for differences in the susceptibility of the
15 elderly to ambient PM concentrations.
16 New findings on cardiovascular effects related to ambient PM are particularly important in
17 assessing PM effects on the elderly. The studies described in Section 6.2.4 and discussed in
18 Section 6.4.2.3 (e.g., Peters et al., 1997; Pope et al., 1999; Dockery et al., 1999) include studies
19 in elderly subjects exposed to ambient air, who may or may not have pre-existing respiratory or
20 cardiovascular conditions. These studies suggest adverse biological changes that could have
21 serious consequences in less healthy or more susceptible subjects and add to the plausibility of
22 cardiovascular hospital admissions or deaths associated with current ambient PM levels.
23
24
25 8.5 TOXICOLOGIC EVALUATION OF PATHOPHYSIOLOGIC
26 EFFECTS OF PM CONSTITUENTS AND MECHANISMS OF ACTION
27 8.5.1 Possible Mechanisms of PM-Induced Injury
28 Several potential pathophysiologic mechanisms can be proposed by which ambient
29 particles could conceivably contribute to morbidity and mortality. As discussed in Chapter 7,
30 PM has been identified as causing a variety of health effects including respiratory symptoms,
31 mechanical changes in lung function, alteration of mucociliary clearance, pulmonary
October 1999 8-50 DRAFT-DO NOT QUOTE OR CITE
-------
1 inflammatory responses and morphological alterations in the lung. In addition, PM has been
2 associated with respiratory illness, hospital admissions, and increased daily mortality.
3 In this section, attention is directed at pulmonary and cardiovascular mechanisms which
4 could hypothetically contribute to increased morbidity and mortality. The phenomenon of
5 particle related mortality may include: (1) "premature" death (or mortality displacement), that is
6 the hastening of death for individuals already near death (i.e., hastening of certain death by hours
7 or days); (2) increased susceptibility to infectious disease; and (3) exacerbation of chronic
8 underlying cardiac or pulmonary disease (Utell and Frampton, 1995). The distribution of
9 deposition of particles inhaled into the respiratory tract depends on their size, shape, chemical
10 composition, and the airway geometry and pulmonary ventilation characteristics of the organism.
11 The mechanisms responsible for the broad range of particle-related health affects will vary
12 depending on the site of deposition. Once deposited, the particles may be cleared from the lung,
13 translocated into the interstitium, sequestered in the lymph nodes, metabolized or otherwise
14 transformed.
15 Deposition of particulate matter in the human respiratory tract could initiate events leading
16 to increased airflow obstruction, impaired clearance, impaired host defenses, or increased
17 epithelial permeability. Airflow obstruction could result from laryngeal constriction or
18 bronchoconstriction secondary to stimulation of receptors in extrathoracic or intrathoracic
19 airways. In individuals with airways disease and localized airway narrowing or obstruction, PM
20 accumulates more rapidly.
21 Acid aerosols are known to cause slowing of mucociliary clearance. Since this mechanism
22 is important in clearing particles from the lung, including biologically active particles such as
23 spores, fungi, and bacteria, impairment of mucociliary clearance could lead to increased PM
24 burdens, inflammation, and infection. Alveolar clearance may also be impaired through
25 alterations in macrophage function including decreased phagocytosis, depression of mobility, and
26 decreased adherence to surfaces. Retention of PM may be associated with the initiation and/or
27 progression of COPD. Macrophages play an important role in removing and digesting particles
28 and may be involved in facilitating translocation of PM to either other parts of the lung or into
29 the vascular system.
30 Soluble transition metals such as iron, copper, nickel, vanadium and cobalt, are capable of
31 catalyzing the production of reactive oxygen species, such as hydroxyl radical. These reactive
October 1999 8-51 DRAFT-DO NOT QUOTE OR CITE
-------
1 oxygen species can cause cellular injury and DNA damage which can be inhibited by
2 antioxidants or free radical scavengers. The responses to residual oil fly ash are due in large part
3 to the presence of soluble transition metals; these responses (e.g., inflammation) are much
4 reduced when free radical scavengers are added to the milieu. Concentrated air particles and PM
5 from various sources also show evidence of generating reactive oxygen species. Furthermore,
6 metal chelating agents are capable of reducing responses to PM and ROFA. Both vanadium and
7 nickel appear to be important toxic constituents of ROFA; their role in ambient PM is likely to be
8 much less. The ability of "ambient PM" to induce production of ROS is variable based on the
9 geographic, seasonal, and chemical characteristics of the PM.
10 Both ROFA and ambient PM can induce or enhance inflammation in the lung; such an
11 effect may depend on particle size and deposition site as well as on chemical or biological
12 composition of the particles. Inflammatory responses can lead to increased permeability and
13 possibly diffusion abnormality. Metal compounds (e.g., NaVO3; Fe2O3; NiSO4), either
14 individually or as part of a complex mixture, can initiate an inflammatory cascade using
15 intracellular signaling mechanisms. In addition, ROS can initiate a signaling cascade in
16 epithelial cells that is related to cell proliferation.
17 Pulmonary changes that contribute to cardiovascular responses include mechanisms which
18 can lead to hypoxemia, including bronchoconstriction, apnea, impaired diffusion, and
19 inflammation. Hypoxic episodes can lead to cardiac arrhythmias and other cardiac
20 electrophysiologic responses. Additionally, many respiratory receptors have direct
21 cardiovascular effects. Stimulation of C-fibers leads to bradycardia and hypertension, while
22 stimulation of laryngeal receptors can result in hypertension, cardiac arrhythmia, bradycardia,
23 apnea, and even cardiac arrest. Nasal receptor or pulmonary J-receptor stimulation can lead to
24 vagally mediated bradycardia and hypertension (Widdicombe et al., 1988). Little is known about
25 age-related changes in airway receptor reflexes and their cardiac effects. Exposure to high
26 concentrations of particles (mainly by instillation) can hasten death in animals with compromised
27 cardiopulmonary systems. These responses are associated with increased inflammation and
28 increased incidence of serious arrhythmias. It is not clear that either inflammation or hypoxia is
29 related to increased mortality; few systemic effects have been examined. Among the systemic
30 effects being investigated are changes in blood coagulation parameters, peripheral blood
31 neutrophil levels, and levels of inflammatory mediators in heart tissue.
October 1999 8-52 DRAFT-DO NOT QUOTE OR CITE
-------
1 Particles that deposit in the lung can also damage respiratory tract cells. The response of
2 the respiratory tract to such particles includes the release of numerous cytokines from alveolar
3 macrophages and epithelial lining cells that promote healing and repair. Repeated cycles of acute
4 lung injury and repair can lead to fibrotic changes in the lung. Such responses are well known in
5 animal models and they typically occur after several weeks of exposure to particle concentrations
6 much higher than those in ambient air.
7
8 8.5.2 Ultrafine Particles
9 In U.S. Environmental Protection Agency (1996), concern was expressed over the potential
10 role of ultrafme particles (UF; defined as particles <100 nm) in human health effects of PM.
11 Ultrafine particles have a high deposition efficiency in the extrathoracic airways (naso- and
12 oropharyngeal). The abundant vascularity of the nasal region and the ability of UF to rapidly
13 enter the interstitium suggests that uptake of UF into the vascular system could occur relatively
14 quickly. Although UF also have a high deposition efficiency in the alveolar region, they
15 constitute a relatively small fraction of the deposited PM mass. However, because of their small
16 size, UF tend to evade endocytosis by macrophages.
17 Human exposures to ultrafme particles result in varying responses depending upon the
18 composition and concentration of the particles. Zinc oxide UF particles induce metal fume fever
19 and airway inflammation whereas magnesium oxide UF particles do not (Kuschner et al., 1997).
20 Polymer fumes (UF polymer particles) can also lead to fever, diffusion impairment, and
21 respiratory symptoms in humans; similar responses are seen in animals. Ultrafine metal particles
22 cause lung injury and inflammation in rats that are dependent on their composition; nickel is
23 highly toxic whereas TiO2 is less so (Zhang et al., 1998). However, UF acid aerosol was found
24 to be no more effective than fine acid particles in causing morphologic or ventilatory effects in
25 rats. Nevertheless, UF acid did increase responses of rats to ozone, whereas fine acid PM did not
26 (Kimmel et al., 1997). Ultrafine amorphous silica had more effect on rat respiratory epithelium
27 than either carbon black or diesel exhaust UF particles (Murphy et al., 1998). Thus particle size
28 per se does not appear to be a crucial single index of particle toxicity; rather, the chemical
29 composition of the particle appears to play a key role in its effects once it has been deposited.
30 However, as discussed elsewhere, particle size is the primary determinant of the deposition
31 pattern of inhaled PM. When an UF particle is deposited at a specific site within the respiratory
October 1999 8-53 DRAFT-DO NOT QUOTE OR CITE
-------
1 tract, the composition of the particle and the sensitivity of the tissues to those particle
2 components may determine its toxicity. Physical particle characteristics (size, solubility) which
3 influence the translocation properties of the particle (e.g., to the interstitium or vascular system),
4 may also play a key role in its toxicity. Oberdorster et al. (1994) compared fine and UF TiO2
5 using inhalation exposure and observed that the UF TiO2 elicited greater inflammation,
6 prolonged retention of UF particles with increased translocation to the epithelium, more
7 epithelial effects, and impairment of macrophage function. Thus, in this case of identical particle
8 composition, the physical properties associated with UF TiO2 particles leading to their greater
9 persistence within pulmonary tissues, especially in the alveoli and the interstitium, appear to be
10 responsible for their greater toxicity.
11
12 8.5.3 Bioaerosols
13 Ambient bioaerosols include fungal spores, pollen, bacteria, viruses, endotoxin, and animal
14 and plant debris. Bacteria, viruses and endotoxin are mainly found attached to aerosol particles,
15 while entities in the other categories are found as separate particles. Data for characterizing
16 ambient concentrations and size distributions of bioaerosols remain sparse. The proportion of
17 particles that are bioaerosols is variable and they are present in both the fine mode and the coarse
18 mode. The cytoplasmic content of spores and pollen may to be adhered to particles emitted by
19 motor vehicles or particles of crustal origin.
20 Fungal spores form the largest and most consistently present component of biological
21 aerosols in ambient air. Levels vary seasonally and spatially reaching higher levels during the
22 summer and near some anthropogenic sources (agricultural activities, compost, etc.).
23 Bioaerosols can contribute to increased mortality and morbidity. Association of fungal and
24 pollen spores and fragments with exacerbations of asthma or allergic rhinits is well known.
25 In addition to fungal spores and pollen, other bioaerosol material can exacerbate asthma
26 and can also induce responses in nonasthmatics. Occupational and experimental exposures of
27 humans to endotoxin (lipopolysaccharide) is associated with increased airway responsiveness,
28 changes in lung function, and airway inflammation. However, these levels vastly exceed those
29 typically present in U.S. ambient air. A classic series of studies (Anto and Sunyer, 1990) proved
30 that airborne dust from soybean husks was responsible for asthma epidemics and increased
31 emergency room visits in Barcelona, Spain. These studies indicate that airborne fragments of
October 1999 8-54 DRAFT-DO NOT QUOTE OR CITE
-------
1 biological substances can produce severe health effects when present in high concentrations.
2 A series of new studies suggest that diesel particulate matter can act as an adjuvant to increase
3 the response to certain antigenic substances, including bioaerosols. Particles may act as carriers
4 as well as promote the response to antigens, leading to the exacerbation of allergic rhinitis and
5 possibly allergic asthma (Diaz-Sanchez et al., 1997).
6 Because of the extremely limited knowledge of ambient levels of bioaerosols and their
7 composition and relative potency of various components, the small number of well conducted
8 epidemiologic studies of bioaerosols, and the absence of controlled studies of ambient
9 bioaerosols, the relative contribution of bioaerosols to the observed PM-associated morbidity and
10 mortality effects cannot be determined with confidence at the present time. However, it seems
11 unlikely that bioaerosols play more than a minor role in such effects.
12
13 8.5.4 Metals
14 Soluble transition metals play an important role in the effects induced by instillation
15 (suspensions or supernatant) and inhalation of residual oil fly ash (Kodavanti et al., 1998). The
16 predominant metals in ROFA particles are typically vanadium, nickel and iron. Other metals,
17 including zinc and copper, are also present depending on the source of the ROFA. The
18 characteristic responses to the individual metals vary as does their behavior in complex mixtures.
19 Vanadium appears to be associated with airway inflammation and production of reactive oxygen
20 species by alveolar macrophages. Occupational studies have identified vanadium as being
21 associated with "boiler maker's bronchitis." Nickel is associated with pulmonary injury caused
22 by ROFA, and thus increased levels of LDH and protein in bronchoalveolar lavage, and is also
23 classified as a human carcinogen. Vanadium activates macrophages whereas nickel does not.
24 There is some evidence that the effects of nickel are antagonized by the presence of vanadium.
25 On the other hand, vanadium has a greater effect on alveolar macrophage ROS production when
26 it is not in the presence of nickel and/or iron. Iron tends to produce less severe effects than either
27 nickel or vanadium; it is associated with inflammatory responses in both humans and laboratory
28 animals. It is apparent that different metal components of ROFA cause different effects and that
29 the individual effects of these metals may not always be expressed to the same extent in
30 mixtures; antagonism has been identified for some endpoints. These findings point to the
October 1999 8-55 DRAFT-DO NOT QUOTE OR CITE
-------
1 complexities of determining a mechanism of response to "PM" and further clouds the prospect of
2 identifying a "magic bullet" that relates increased PM levels to increased human mortality.
3
4 8.5.5 Concentrated Ambient Particles (CAPs)
5 A number of new studies have examined the effects of concentrated ambient particles
6 (CAPs) in animals and humans. Petrovic et al. (1999) have studied a small number of healthy
7 volunteers exposed to CAPs (max concentration of 124 //g/m3) from downtown Toronto,
8 Canada. Cardiovascular and respiratory functions showed no adverse effects and there was no
9 indication of pulmonary inflammation or respiratory symptom responses. However, there were
10 trends for plasma fibrinogen and nasal neutrophils to increase.
11 In laboratory animals exposed to CAPs in Boston, MA (207, 733 and 607 //g/m3) for
12 3 days, lavage fluid protein and neutrophils were elevated, suggestive of an inflammatory
13 response (Clarke et al., 1999). In vitro studies using CAPs demonstrate an increased oxidative
14 burst in exposed alveolar macrophages (Goldsmith et al., 1998). Gordon et al. (1998) found
15 increased peripheral blood neutrophils and decreased lymphocytes after exposure to
16 110-350 //g/m3 CAPs from New York City. Additionally they found a 10-20 beat/min increase
17 in heart rate following the nose-only inhalation exposure. This study shows that ambient PM can
18 cause systemic effects that influence the cardiovascular system.
19 Studies using diesel exhaust particles demonstrate that diesel particles can act as an
20 adjuvant for inhaled antigens in the nose. In addition to nasal inflammation, diesel exhaust
21 (particles and gases standardized to 300 //g/m3 DPM) cause increased neutrophils, ICAM-1, and
22 IL-8 in bronchoalveolar lavage fluid from exposed volunteers. It has also been shown that diesel
23 exhaust causes transcription of IL-8 and other cytokines which are responsible for attracting and
24 activating leukocytes. In support of the findings of Gordon et al. (1998), these studies also found
25 increased neutrophils and platelets in peripheral blood following exposure (Salvi et al., 1999,
26 2000).
27
28 8.5.6 Summary
29 Particle impacts on the lung depend on particle size, chemical composition, deposition
30 pattern, and particle retention. The effects range from pulmonary responses such as respiratory
31 symptoms, airways inflammation, impacts on particle clearance, and epithelial cellular injury, to
October 1999 8-56 DRAFT-DO NOT QUOTE OR CITE
-------
1 systemic effects such as changes in blood clotting factors, increased peripheral blood neutrophils
2 and changes in myocardial inflammatory mediators.
3 Smaller particles (PM < 2.5 //m) have greater deposition and long term retention in the
4 respiratory (alveolar) region of the lung because there are fewer large particles in ambient air,
5 their deposition efficiency in the respiratory region is low, and they are likely to be cleared rather
6 than translocated to the parenchyma. However, specific particle components remain to be
7 lexicologically identified with confidence as likely being responsible for the epidemiologically
8 observed effects of ambient PM exposures. Rather, the influence of a given particle constituent
9 varies with the biological function being examined as well as the presence (or absence) of other
10 PM constituents. For example, metals can act additively or antagonistically with each other, and
11 other PM chemical constituents may enhance the effect of biologic components. The number of
12 combinations and interactions of various physical, chemical, and biological components of PM is
13 just beginning to be investigated. Newer data expected to become available during the next
14 several months will likely provide valuable additional information on lexicologically important
15 PM constituents and mechanisms.
16
October 1999 8-57 DRAFT-DO NOT QUOTE OR CITE
-------
1 REFERENCES
2 Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
3 suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
4 Adventists. Environ. Health Perspect. 94: 43-50.
5 Abbey, D. E.; Hwang, B. L.; Burchette, R. J.; Vancuren, T.; Mills, P. K. (1995) Estimated long-term ambient
6 concentrations of PM10 and development of respiratory symptoms in a nonsmoking population.
7 Arch. Environ. Health 50: 139-152.
8 Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, W. L.; Yang, J. X. (1999)
9 Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
10 Care Med. 159: 373-382.
11 Anto, J. M.; Sunyer, J. (1990) Epidemiologic studies of asthma epidemics in Barcelona. Chest
12 90(suppl.): 185S-190S.
13 Bates, D. V. (1992) Health indices of the adverse effects of air pollution: the question of coherence. Environ. Res.
14 59: 336-349.
15 Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
16 lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
17 Bennett, W. D.; Zeman, K. L. (1998) Deposition of fine particles in children spontaneously breathing at rest.
18 Inhalation Toxicol. 10: 831-842.
19 Bennett, W. D.; Zeman, K. L.; Kim, C.; Mascarella, J. (1997a) Enhanced deposition of fine particles in COPD
20 patients spontaneously breathing at rest. Inhalation Toxicol. 9: 1-14.
21 Bennett, W. D.; Zeman, K. L.; Kang, C. W.; Schechter, M. S. (1997b) Extrathoracic deposition of inhaled, coarse
22 particles (4.5"mu"m) in children versus adults. Ann. Occup. Hyg. 41 (Suppl.): 497-502.
23 Borja-Aburto, V. H.; Loomis, D. P.; Bangdiwala, S. I.; Shy, C. M.; Rascon-Pacheco, R. A. (1997) Ozone,
24 suspended particulates, and daily mortality in Mexico City. Am. J. Epidemiol. 145: 258-268.
25 Borja-Aburto et al. (1999).
26 Bremner, S. A.; Anderson, H. R.; Atkinson, R. W.; McMichael, A. J.; Strachan, D. P.; Bland, J. M.; Bower, J. S.
27 (1999) Short term associations between outdoor air pollution and mortality in London 1992-4. Occup.
28 Environ. Med. 56: 237-244.
29 Brunekreef, B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
30 Burnett, R. T.; Dales, R. E.; Brook, J. R.; Raizenne, M. E.; Krewski, D. (1997a) Association between ambient
31 carbon monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities.
32 Epidemiology 8: 162-167.
33 Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997b) The role of paniculate size and chemistry in the
34 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
35 Environ. Health Perspect. 105: 614-620.
36 Burnett, R. T.; Smith-Doiron, M.; Stieb, D.; Cakmak, S.; Brook, J. R. (1999) Effects of paniculate and gaseous air
37 pollution on cardiorespiratory hospitalizations. Arch. Environ. Health 54: 130-139.
3 8 Churg, A.; Brauer, M. (1997) Human lung parenchyma retains PM2 5. Am. J. Respir. Crit. Care Med.
39 155:2109-2111.
40 Clarke, R. W.; Catalano, P. J.; Koutrakis, P.; Krishna Murthy, G. G.; Sioutas, C.; Paulauskis, J.; Coull, B.;
41 Ferguson, S.; Godleski, J. J. (1999) Urban air paniculate inhalation alters pulmonary function and induces
42 pulmonary inflammation in a rodent model of chronic bronchitis. Inhalation Toxicol. 11: 637-656.
43 Cropper, M. L.; Simon, N. B.; Alberini, A.; Arora, S.; Sharma, P. K. (1997) The health benefits of air pollution
44 control in Delhi. Am. J. Agric. Econ. 79: 1625-1629.
45 Dahl, A. R.; Schlesinger, R. B.; Heck, H. D' A.; Medinsky, M. A.; Lucier, G. W. (1991) Comparative dosimetry of
46 inhaled matenals: differences among animal species and extrapolation to man. Fundam. Appl. Toxicol.
47 16: 1-13.
48 Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H. (1998) Symptoms in pediatric asthmatics and air pollution:
49 differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging time.
5 0 Environ. Health Perspect. 106:751-761.
51 Diaz-Sanchez, D.; Tsien, A.; Fleming, J.; Saxon, A. (1997) Combined diesel exhaust paniculate and ragweed
52 allergen challenge markedly enhances human in vivo nasal ragweed-specific IgE and skews cytokine
53 production to a T helper cell 2-type pattern. J. Immunol. 158: 2406-2413.
October 1999 8-58 DRAFT-DO NOT QUOTE OR CITE
-------
1 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.
2 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
3 329: 1753-1759.
4 Dockery, D. W.; Pope, C. A., Ill; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999) Daily changes in oxygen
5 saturation and pulse rate associated with paniculate air pollution and barometric pressure. Cambridge, MA:
6 Health Effects Institute; research report no. 83.
7 Evans, J. S.; Kinney, P. L.; Koehler, J. L.; Cooper, D. W. (1984) The relationship between cross-sectional and time
8 series studies. J. AirPollut. Control Assoc. 34: 551-553.
9 Gamble, J. F. (1998) PM2 5 and mortality in long-term prospective cohort studies: cause-effect or statistical
10 associations? Environ. Health Perspect. 106: 535-549.
11 Goldsmith, C.-A. W.; Imrich, A.; Danaee, H.; Ning, Y. Y.; Kobzik, L. (1998) Analysis of air pollution
12 particulate-mediated oxidant stress in alveolar macrophages. J. Toxicol. Environ. Health Part A 54: 529-545.
13 Gordian, M. E.; Ozkaynak, H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Paniculate air pollution and respiratory
14 disease in Anchorage, Alaska. Environ. Health Perspect. 104: 290-297.
15 Gordon, T.; Nadziejko, C.; Schlesinger, R.; Chen, L. C. (1998) Pulmonary and cardiovascular effects of acute
16 exposure to concentrated ambient paniculate matter in rats. Toxicol. Lett. 96-97: 285-288.
17 Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
18 dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
19 Higgins, M. W.; Keller, J. B. (1970) Predictors of mortality in the adult population of Tecumseh. Arch. Environ.
20 Health 21: 418-424.
21 Hoyert, D. L.; Kochanek, K. D.; Murphy, S. L. (1999) Deaths: final data for 1997. Atlanta, GA: U.S. Centers for
22 Disease Control and Prevention, National Center for Health Statistics. (National vital statistics reports: v. 47,
23 no. 19).
24 Katsouyanni, K.; Touloumi, G.; Spix, C.; Schwartz, J.; Balducci, F.; Medina, S.; Rossi, G.; Wojtyniak, B.; Sunyer,
25 J.; Bacharova, L.; Schouten, J. P.; Ponka, A.; Anderson, H. R. (1997) Short term effects of ambient sulphur
26 dioxide and paniculate matter on mortality in 12 European cities: results from time series data from the
27 APHEA project. Br. Med. J. 314: 1658-1663.
28 Kim, C. S.; Kang, T. C. (1997) Comparative measurement of lung deposition of inhaled fine particles in normal
29 subjects and patients with obstructive airway disease. Am. J. Respir. Crit. Care Med. 155: 899-905.
30 Kimmel, T. A.; Chen, L. C.; Bosland, M. C.; Nadziejko, C. (1997) Influence of acid aerosol droplet size on
31 structural changes in the rat lung caused by acute exposure to sulfuric acid and ozone. Toxicol. Appl.
32 Pharmacol. 144: 348-355.
33 Kodavanti, U. P.; Hauser, R.; Christiani, D. C.; Meng, Z. H.; McGee, J.; Ledbetter, A.; Richards, J.; Costa, D. L.
34 (1998) Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals.
35 Toxicol. Sci. 43: 204-212.
36 Kuschner, W. G.; Wong, H.; D'Alessandro, A.; Quinlan, P.; Blanc, P. D. (1997) Human pulmonary responses to
3 7 experimental inhalation of high concentration fine and ultrafine magnesium oxide particles. Environ. Health
38 Perspect. 105: 1234-1237.
39 Laden, F.; Neas, L. M.; Schwartz, J. (1999) The association of crystal particles in the fine paniculate fraction with
40 daily mortality in six cities. Am. J. Respir. Crit. Care Med. 159: A332.
41 Lawrence, L.; Hall, M. J. (1999) 1997 summary: national hospital discharge survey. Hyattsville, MD: U.S.
42 Department of Health & Human Resources, Centers for Disease Control and Prevention, National Center for
43 Health Statistics. (Advance data from vital and health statistics: no. 308).
44 Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
45 Perspect. 107: 149-154.
46 Loomis, D.; Castilejos, M.; Gold, D. R.; McDonnell, W.; Borja-Aburto, V. H. (1999) Air pollution and infant
47 mortality in Mexico City. Epidemiology 10: 118-123.
48 Mage, D. T.; Wilson, W.; Hasselblad, V.; Grant, L. (1999) Assessment of human exposure to ambient paniculate
49 matter. J. Air Waste Manage. Assoc.: in press.
50 Miller, F. J.; Angilvel, S.; Menache, M. G.; Asgharian, B.; Gerrity, T. R. (1995) Dosimetric issues relating to
51 paniculate toxicity. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on paniculate air
52 pollution and human mortality and morbidity, part II; January 1994; Irvine, CA. Inhalation Toxicol.
53 7:615-632.
54 Morgan, G.; Corbett, S.; Wlodarczyk, J. (1998) Air pollution and hospital admissions in Sydney, Australia, 1990 to
55 1994. Am. J. Public Health 88: 1761-1766.
October 1999 8-59 DRAFT-DO NOT QUOTE OR CITE
-------
1 Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
2 evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
3 Morrow, P. E. (1988) Possible mechanisms to explain dust overloading of the lungs. Fundam. Appl. Toxicol.
4 10:369-384.
5 Murphy, S. A.; BeruBe, K. A.; Pooley, F. D.; Richards, R. J. (1998) The response of lung epithelium to well
6 characterised fine particles. Life Sci. 62: 1789-1799.
7 National Research Council. (1998) Research priorities for airborne paniculate matter. I. Immediate priorities and a
8 long-range research portfolio. Washington, DC: National Academy Press.
9 National Research Council. (1999) Research priorities for airborne paniculate matter. II. Evaluating research
10 progress and updating the portfolio. Washington, DC: National Academy Press.
11 Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Shappard, L.; Stout, J. W. (1999) An association
12 between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
13 Perspect. 107: 489-493.
14 Oberdorster, G.; Ferin, J.; Gelein, R.; Soderholm, S. C.; Finkelstein, J. (1992a) Role of the alveolar macrophage in
15 lung injury: studies with ultrafine particles. Environ. Health Perspect. 97: 193-199.
16 Oberdorster, G.; Ferin, J.; Morrow, P. E. (1992b) Volumetric loading of alveolar macrophages (AM): a possible
17 basis for diminished AM-mediated particle clearance. Exp. Lung Res. 18: 87-104.
18 Oberdorster, G.; Gelein, R. M.; Ferin, J.; Weiss, B. (1995a) Association of paniculate air pollution and acute
19 mortality: involvement of ultrafine particles? In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the
20 colloquium on paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA.
21 Inhalation Toxicol. 7: 111-124.
22 Oberdorster, G.; Ferin, J.; Gelein, R.; Mercer, P.; Corson, N.; Godleski, J. (1995b) Low-level ambient air paniculate
23 levels and acute mortality/morbidity: studies with ultrafine Teflon particles. Am. J. Respir. Crit. Care Med.
24 151(suppl.): A66.
25 Ostro, B.; Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1998) The impact of fine paniculate matter in Bangkok,
26 Thailand. In: Chow, J.; Koutrakis, P., eds. PM2.5: a fine particle standard. Volume II: proceedings of an
27 international specialty conference.; January; Long Beach, CA. Pittsburgh, PA: Air & Waste Management
28 Association; pp. 939-949. (AWMA specialty conference VTP-81).
29 Ostro et al. (1999, in press).
30 Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
31 pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
32 Environmental Health; March 25.
33 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
34 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
35 Petrovic, S.; Urch, B.; Brook, J.; Datema, J.; Purdham, J.; Liu, L.; Lukic, Z.; Zimmerman, B.; Tofler, G.;
36 Downar, E.; Corey, P.; Tarlo, S.; Broder, I.; Dales, R.; Silverman, F. (1999) Cardiorespiratory effects of
37 concentrated ambient PM2 5: a pilot study using controlled human exposures. Inhalation Toxicol.: accepted.
38 Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
39 (1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
40 Respir. Crit. Care Med. 151: 669-674.
41 Pope, C. A.; Hill, R. W.; Villegas, G. M. (1999) Paniculate air pollution and daily mortality on Utah's Wasatch
42 Front. Environ. Health Perspect.: submitted.
43 Prescott, G. J.; Cohen, G. R.; Elton, R. A.; Fowkes, F. G. R.; Agius, R. M. (1998) Urban air pollution and
44 cardiopulmonary ill health: a 14.5 year time series study. Occup. Environ. Med. 55: 697-704.
45 Salvi, S.; Blomberg, A.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (1999) Acute inflammatory
46 responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human
47 volunteers. Am. J. Respir. Crit. Care Med. 159: 702-709.
48 Salvi, S.; Blomberg, A.; Nordenhall, C.; Rudell, B.; Kelly, F.; Sandstrom, T.; Holgate, S. T.; Frew, A. (2000)
49 Acute short term exposure to diesel exhaust increases IL-8 and GRO-"alpha" production in healthy human
50 airways. Am. J. Resir. Crit. Care Med.: accepted.
51 Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
52 10: 17-22.
53 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996) Is daily mortality associated specifically with fine particles?
54 J. Air Waste Manage. Assoc. 46: 927-939.
October 1999 8-60 DRAFT-DO NOT QUOTE OR CITE
-------
1 Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Claiborne, C.; Koenig, J. (1999) Episodes of high coarse particle
2 concentrations are not associated with increased mortality. Environ. Health Perspect. 107: 339-342.
3 Seneff, M. G.; Wagner, D. P.; Wagner, R. P.; Zimmerman, J. E.; Knaus, W. A. (1995) Hospital and 1-year survival
4 of patients admitted to intensive care units with acute exacerbation of chronic obstructive pulmonary disease.
5 JAMA J. Am. Med. Assoc. 274: 1852-1857.
6 Siegmann, K.; Scherrer, L.; Siegmann, H. C. (1999) Physical and chemical properties of airborne nanoscale
7 particles and how to measure the impact on human health. J. Mol. Struct. (Theochem) 458: 191-201.
8 Simpson, R. W.; Williams, G.; Petroeschevsky, A.; Morgan, G.; Rutherford, S. (1997) Associations between
9 outdoor air pollution and daily mortality in Brisbane, Australia. Arch. Environ. Health 52: 442-454.
10 South Coast Air Quality Management District. (1990) Final state implementation plan for PM10 in the Coachella
11 Valley.
12 Strachan, D. P. (1992) Ventilatory function, height, and mortality among lifelong non-smokers. J. Epidemiol.
13 Community Health 46: 66-70.
14 Svartengren, K.; Lindestad, P.-A.; Svartengren, M.; Bylin, G.; Philipson, K.; Camner, P. (1994) Deposition of
15 inhaled particles in the mouth and throat of asthmatic subjects. Eur. Respir. J. 1467-1473.
16 Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
17 summertime haze air pollution in Toronto, Ontario: consideration of the role of acid aerosols. Environ. Res.
18 65:271-290.
19 U.S. Environmental Protection Agency. (1996) Air quality criteria for paniculate matter. Research Triangle Park,
20 NC: National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/00 laF-cF. 3v.
21 Available from: NTIS, Springfield, VA; PB96-168224.
22 Utell, M. J.; Frampton, M. W. (1995) Particles and mortality: a clinical perspective. In: Phalen, R. F.; Bates, D. V.,
23 eds. Proceedings of the colloquium on paniculate air pollution and human mortality and morbidity, part II;
24 January 1994; Irvine, CA. Inhalation Toxicol. 7: 645-655.
25 Venkataraman, C.; Kao, A. S. (1999) Comparison of particle lung doses from the fine and coarse fractions of urban
26 PM-10 aerosols. Inhalation Toxicol. 11: 151-169.
27 Vincent, J. H. (1995) Standards for health-related aerosol measurement and control. In: Aerosol science for
28 industrial hygienists. Oxford, United Kingdom: Elsevier Science Ltd.; pp. 204-237.
29 Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
30 Widdicombe, J.; Sant'Ambrogio, G.; Mathew, O. P. (1988) Nerve receptors of the upper airway. In: Mathew, O. P.;
31 Sant'Ambrogio, G., eds. Respiratory function of the upper airway. New York, NY: Marcel Dekker;
32 pp. 193-231. (Lung biology in health and disease: v. 35).
33 Wilson, W. E.; Suh, H. H. (1997) Fine particles and coarse particles: concentration relationships relevant to
34 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
35 Wilson, W. E.; Spiller, L. L.; Ellestad, T. G.; Lamothe, P. J.; Dzubay, T. G.; Stevens, R. K.; Macias, E. S.; Fletcher,
36 R. A.; Husar, J. D.; Husar, R. B.; Whitby, K. T.; Kittelson, D. B.; Cantrell, B. K. (1977) General Motors
37 sulfate dispersion experiment: summary of EPA measurements. J. Air Pollut. Control Assoc. 27: 46-51.
38 Zhang, Q.; Kusaka, Y.; Sato, K.; Nakakuki, K.; Kohyama, N.; Donaldson, K. (1998) Differences in the extent of
39 inflammation caused by intratracheal exposure to three ultrafine metals: role of free radicals. J. Toxicol.
40 Environ. Health. 53:423-438.
41 Zmirou, D.; Schwartz, J.; Saez, M.; Zanobetti, A.; Wojtyniak, B.; Touloumi, G.; Spix, C.; Ponce de Leon, A.;
42 Le Moullec, Y.; Bacharova, L.; Schouten, J.; Ponka, A.; Katsouyanni, K. (1998) Time-series analysis of air
43 pollution and cause-specific mortality. Epidemiology 9: 495-503.
44
October 1999 8-61 DRAFT-DO NOT QUOTE OR CITE
------- |